Practical synthesis of chiral 2-morpholine: (4-Benzylmorpholin-2-(s)-yl)- (tetrahydropyran-4-yl) Methanone mesylate, a useful pharmaceutical intermediate

Organic Process Research & Development 2009, 13, 209–224

Practical Synthesis of Chiral 2-Morpholine: (4-Benzylmorpholin-2-(S)-yl)-

(tetrahydropyran-4-yl)methanone Mesylate, a Useful Pharmaceutical Intermediate

Michael E. Kopach,* Utpal K. Singh, Michael E. Kobierski, William G. Trankle, Michael M. Murray, Mark A. Pietz,

Mindy B. Forst, and Gregory A. Stephenson

Eli Lilly and Company, Indianapolis, Indiana 46285, U.S.A.

Vincent Mancuso

GlaxoSmithkline Biologicals, B-1330 Rixensart, Belgium

Thierry Giard

Albemarle, B-1348 LouVain-la-NeuVe, Belgium

Michel Vanmarsenille and Thierry DeFrance

UCB Pharma SA Belgium, 60B-1070, Brussels, Belgium

Scheme 1. Biologically active 2-substitututed morpholines

Abstract:

A commercial synthesis was developed for the production of (4-

benzylmorpholin-2-(S)-yl)-(tetrahydropyran-4-yl)methanone me-

sylate, 1a, a key starting material for a phase 2, new investigational

drug candidate at Eli Lilly and Company. The target compound

was produced in the clinical pilot plant by the combination of two

key steps: resolution of a morpholine amide intermediate to install

the S-morpholino stereocenter in 35% yield and a high-yielding

(89%) Grignard reaction to generate the title compound 1a,

isolated as a mesylate salt. The Grignard reaction was found to

proceed optimally when using a combination of I₂and DIBAL-H

for the initiation. In addition, the Grignard reagent forma-

tion was monitored by ReactMax calorimetry, and proof-

of-concept studies were completed, demonstrating that the

Grignard step could potentially be run as a continuous

process with magnesium recycling.

afﬁnity and selectivity toward the norepinephrine transporter.³

In addition, the 2-morpholine structural skeleton is found in

other pharmaceutically active compounds such as NK1-receptor

antagonist aprepitant⁴(4), antifungal agent amoroﬁne⁵(5), and

selective GABA ꢀ-antagonist SCH-50911⁶(6). Despite the

numerous examples of pharmaceutical compounds substituted

at the 2-position of the morpholine ring, general synthetic

methods are limited, and as in the case of S,S-reboxetine, the

morpholine ring is typically formed in the late stages of the

Introduction

Selective norepinephrine reuptake inhibitors such as rebox-

etine¹(2) and viloxazine²(3) which contain a common

2-morpholine backbone have been used for the treatment of

depression (Scheme 1). While reboxetine is marketed as a

racemic drug, the 2S,3S-enantiomer (S,S-reboxetine) has greater

* Author to whom correspondence may be sent. E-mail: Kopach_Michael@

lilly.com.

(3) Wong, E. H. F.; Ahmed, S.; Marshall, R. C.; McArthur, R.; Taylor,

D. P.; Birgerson, L.; Cetera, P. WO 2001001973, 2001.

(4) Zhao, M. M.; McNamara, J. M.; Ho, G.; Emerson, J. M.; Song, Z. J.;

Tschaen, D. M.; Brands, K. M. J.; Dolling, U.; Grabowski, E. J. J.;

Reider, P. J. J. Org. Chem. 2002, 67, 6743.

(1) (a) Berzewski, H.; Van Moffaert, M.; Gagiano, C. A. Eur. Neurop-

sychopharmacol. 1997, 7, S37. (b) Hajos, M.; Fleishaker, J. C.; Filipak-

Resiner, J. K.; Brown, M. T.; Wong, E. H. F. CNS Drug ReV. 2004,

10, 23. (c) Greenwood, D. T.; Mallion, K. B.; Todd, A. H.; Turner,

W. J. Med. Chem. 1975, 18, 573.

(5) Tatsumi, Y.; Yokoo, M.; Senda, H.; Kakehi, K. Antimicrob. Agents

Chemother. 2002, 46, 3797.

(2) (a) Greenwood, D. T.; Mallion, K. B.; Todd, A. H.; Turner, W. J. Med.

Chem. 1975, 18, 573. (b) Mallion, K. B.; Todd, A. H.; Turner, R. W.;

Bainbridge, J. T.; Greenwood, D. T.; Madinaveitia, J.; Somerville,

A. R.; Whittle, B. A. Nature (London). 1972, 238, 157.

(6) Blythin, D. J.; Kuo, S.; Shue, H.; McPhail, A. T.; Chapman, R. W.;

Kreutner, W.; Rizzo, C.; She, H.; West, R. Biorg. Med. Chem. Lett.

1996, 6, 1529.

10.1021/op800247w CCC: $40.75

Published on Web 01/23/2009

2009 American Chemical Society

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

209

Scheme 2. Early routes to 1b⁸

synthesis.⁷For these reasons, a potentially efﬁcient entry into

the reboxetine class of compounds is direct synthesis of the

chiral 2-morpholine backbone. Herein we report a practical

synthesis of (4-benzylmorpholin-2-(S)-yl)-(tetrahydropyran-4-

yl)methanone mesylate (1a) which is structurally related to S,S-

reboxetine (2) and is a key starting material for a phase 2 new

investigational drug candidate at Eli Lilly and Company.

Scheme 3. Compound 1a retrosynthetic analysis

Results and Discussion

First-Generation Process (Scheme 2). The synthesis of the

free base form of 1a (4-benzylmorpholin-2-(S)-yl)-(tetrahydro-

pyran-4-yl)methanone (1b), has previously been reported by

Eli Lilly and Company.⁸The (1S)-morpholino chiral center was

initially installed via chiral chromatography of a Weinreb amide

precursor 7, to produce S- 7a, which was then converted to the

1b S-isomer via a chemoselective reaction with the 4-chlorotet-

rahydropyran Grignard reagent, 8b (Scheme 2). Alternatively,

the racemic Weinreb amide intermediate 7 has been converted

to racemic 1 via Grignard chemistry, where it was resolved as

the R-mandelic acid salt, 1c which is readily converted to

optically pure S-isomer 1a (>99% ee, Scheme 2). Key

disadvantages to the aforementioned approaches include an

overall low yield (∼5%), the use of chiral chromatography, late-

stage resolution, and lack of crystalline intermediates, a neces-

sary feature for purity upgrade. In addition, little was known

surrounding the Grignard reaction kinetics which was required

prior to scale-up. Thus, we envisioned that development of more

attractive routes to rapidly produce 1a in multikilogram quanti-

ties could be realized using Grignard chemistry with potential

precursors 9,10, 11 or 12 (Scheme 3).

ethanolamine with 2-chloro acrylonitrile (2-CAN) fol-

lowed by ring closure with 2M Potassium tert-butoxide/

THF.⁹The synthesis of 4-benzylmorpholine-2-carboxylic

acid, 10 was then accomplished via the direct hydrolysis

of nitrile 9. However, the high aqueous solubility of the

acid (>200 mg/mL) made separation of the product from

the resulting aqueous layer exceedingly difﬁcult. Hydroly-

sis of nitrile 9 with 6 N HCl was found to be optimal and

the hydrochloride salt, 10, precipitated directly form the

reaction mixture. Use of acetone as a wet cake rinse

solvent was efﬁcient at removing residual HCl and under

these conditions the hydrochloride salt of 10 was rapidly

prepared via a telescoped process on a multikilogram scale

4-Benzyl-2-cyanomorpholine, 9 was prepared in 80%

yield by reaction of commercially available N-benzyl-

(9) (a) King, F. D.; Martin, R. T. Tetrahedron Lett. 1991, 32, 2281. (b)

Cases-Thomas, M. J.; Masters, J. J.; Walter, M. W.; Campbell, G.;

Haughton, L.; Gallagher, P. T.; Dobson, D. R.; Mancuso, V.; Bonnier,

B.; Giard, T.; Defrance, T.; Vanmarsenille, M.; Ledgard, A.; White,

C.; Ouwerkerk-Mahadevan, S.; Brunelle, F. J.; Dezutter, N. A.;

Herbots, C. A.; Lienard, J. Y.; Findlay, J.; Hayhurst, L.; Boot, J.;

Thompson, L. K.; Hemrick-Luecke, S. Bioorg. Med. Chem. Lett. 2006,

16, 2022. (c) Clark, B. P.; Gallagher, P. T.; Haughton, H. C.

WO2004018440 A1, 2004.

(7) (a) Henegar, K. E.; Ball, C. T.; Horvath, C. M.; Maisto, K. D.; Mancini,

S. E. Org. Process Res. DeV. 2007, 11, 346. (b) Henegar, K. E.; Cebula,

M. Org. Process Res. DeV. 2007, 11, 354. (c) Cossy, J.; Pardo, D. G.;

Metro, X. J. Org. Chem. 2008, 73, 707. (d) Melloni, P.; Della Torre,

A.; Lazzari, E.; Mazzini, G.; Meroni, M. Tetrahedron 1985, 41, 1393.

(8) Campbell, G. I.; Cases-Thomas, M. J.; Man, T.; Masters, J. J. ; Eugenie

Rudyk, H. C.; Walter, M. W. U.S. Patent Appl. 2007/0083046 A1,

2007.

210

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Scheme 4. Synthesis of 1a from substrates 9, 10, and 11

in 60-65% yield with >99% purity. In addition, treatment

of nitrile 9 with concentrated sulfuric acid in ethanol

results in an 87% yield of 4-benzylmorpholine-2-carboxy-

lic acid ethyl ester, 11.⁸With sufﬁcient quantities of 9,

10 and 11 on hand we sought to evaluate the direct

preparation of 1 by the Grignard reaction of the 4-chlo-

rotetrahydropyran Grignard reagent, 8b, with each sub-

strate.¹⁰The initial substrate evaluated was nitrile 9, the

Grignard reactions of which with aryl Grignard reagents

have been shown to proceed with high chemoselectivity

to produce aryl-2-morpholinones.^9bUnfortunately, under

the optimal conditions for aryl Grignard chemistry, Thorpe

addition was the dominant pathway for the reaction of

8b and nitrile 9 (Scheme 4).

In general, the main drawback of the Grignard reaction with

esters is the formation of tertiary alcohol byproducts because

intermediates of type I are not stable.¹¹Nevertheless, in our

case we wished to examine the impact of the morpholino

oxygen on the stability of intermediate I and assess the extent

of deprotonation alpha to the ester. It was found that under the

best conditions (-20 °C and 1.5 equiv of Grignard reagent) an

incomplete conversion resulted with 20-25% residual 11 and

signiﬁcant amounts (10-15%) of the tertiary alcohol byproduct.

In order to test conﬁgurational stability of the R morpholino

hydrogen, an aliquot of the reaction mixture was quenched with

acetic-d₃-acid-d, and by mass spectroscopy and H NMR we

observed signiﬁcant deuterium incorporation (45%). Based on

these results, optimization of the ethyl ester approach was

abandoned.

1

After observing that carboxylic acid 10 was very soluble in

aqueous media, we were obliged to convert the acid to a salt

using a hydride reagent or a base whose corresponding acid

could easily be removed by distillation. Thus, we evaluated

performance of the magnesium and lithium salts of 10 toward

the 8b Grignard reaction. Conversion of 10 to its magnesium

salt with i-PrMgCl in THF is facile between -10 and 0 °C.

However, we observed partial competitive nucleophilic addition

to form the isopropyl ketone byproduct (5-12%). Initial

attempts to prepare the lithium salt of 10 with butyllithium at

-20 °C resulted in signiﬁcant quantities of butyl ketone (30%)

and dibutyl alcohol impurities (5%). Formation of the lithium

salt was improved by the use of lithium hydroxide monohydrate

in THF, but in this case it was difﬁcult to remove water by

using azeotropic distillation after salt formation, because the

lithium carboxylate salt forms a gummy solid which entrains

residual water. The best conditions for the formation of the

lithium salt of 10 were achieved using 3.62 M MeOLi/THF,

which proceeded rapidly without any detectable methyl ester

byproduct. In this case the lithium carboxylate quickly precipi-

tated to produce a thick mixture. For this reason the lithium

salt was formed at 55 °C and the methanol byproduct was

removed by atmospheric distillation. The lithium salt slurry was

then cooled to room temperature prior to use in the Grignard

reaction.

(10) (a) 4-Chlorotetrahydropyran is available in small quantities from

Aldrich (CAS no. 1768-64-5) at $96.60/5 g. Larger quantities can be

readily prepared via the following methods: Nikolic, N. A.; Gonda,

E.; Longford, C. P.; Lane, N. T.; Thompson, D. W. J. Org. Chem.

1989, 54, 2748. (b) Bunnelle, W. H.; Seamon, D. W.; Mohler, D. L.;

Ball, T. F.; Thompson, D. W. Tetrahedron Lett. 1984, 25, 2653.

(11) (a) Bachman, W. E., III; Hetzner, H. P. Organic Syntheses; John Wiley

& Sons: New York, 1955; Collect. Vol. III, p 839. (b) Colonge, J.;

Marey, R. Organic Syntheses; John Wiley & Sons: New York, 1963;

Collect. Vol. IV, p 601.

The magnesium and lithium salts of acid 10 were both

evaluated toward the Grignard reaction with 8b, and in side by

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

211

Scheme 5. Grignard reaction with acid 10

Scheme 6. First-generation process summary

side experiments, the kinetics of the Grignard reaction strongly

favored the Li over the Mg salt by nearly an order of magnitude.

In addition, a key disadvantage for the magnesium salt was

generation of the tertiary alcohol byproduct which could not

be reduced. In the case where the lithium carboxylate was

generated, we also formed 1 equiv of lithium chloride which

could potentially have an impact on the Grignard reaction since

the more nucleophilic ate complex species could have been

formed.¹²However, spiking studies with lithium chloride

resulted in no difference in the kinetic proﬁle. The Grignard

reaction with the lithium carboxylate was found to run best with

2 equiv of Grignard reagent between 20 and 25 °C; performing

the Grignard chemistry with one equivalent of 8b results in

less than a 50% conversion. The kinetics were quite fast at early

reaction times (90% conversion after 3 h) then slowed signiﬁ-

cantly (95% conversion after 21 h). The observation that full

conversion was not achieved can be partially explained by

deprotonation R to the carboxylate function. In fact, 30-40%

of deuterated acid was observed by mass spectroscopy after

quenching an aliquot of the reaction mixture in acetic-d₃-acid-

d. In addition, the resulting magnesium intermediate II has the

potential to be sufﬁciently basic to deprotonate the resulting

product (Scheme 5). Thus, quenching conditions would need

to be carefully considered for performing the analogous

chemistry on a chiral substrate.

2-propanol and was isolated in 27% yield with >99% ee

(Scheme 6). The mandelate salt, 1c, was then readily trans-

formed to a methanesufonic acid salt 1a in 90% yield.¹³The

overall yield of the ﬁrst-generation approach from acid 10 was

24% and was sufﬁciently enabling to produce the initial clinical

quantities of 1a.

Second-Generation Process. During the course of develop-

ment we recognized that signiﬁcant technical and environmental

improvements were needed for the production of 1a in order

to make commercial-scale quantities.¹⁴Key potential improve-

ments identiﬁed were: (1) resolve or introduce chirality at an

earlier stage of the synthesis; (2) develop a strategy that employs

an intermediate not prone to racemization; (3) reduce the usage

of 4-chlorotetrahydropyran, 8a, which is the most expensive

reagent used in the synthesis; (4) develop a robust Grignard

process that would operate safely with a variety of grades of

magnesium. With these considerations in mind, we embarked

on the synthesis of morpholine amide 12, which we thought

would offer advantages with regard to stability, economy, and

puriﬁcation.

Racemic Morpholine Amide 12 Synthesis. The initial

strategy evaluated was conversion of acid 10 to amide 12 via

an intermediate acid chloride using oxalyl chloride as the

chlorinating agent. However, this approach was abandoned due

to issues with the phosgene byproduct.¹⁵Rather, the acid

With an enabling route to produce racemic 1 in hand, a

systematic screen of resolving agents was performed with the

objective of obtaining the optically pure 1b S-isomer. The best

results were achieved with (R)-mandelic acid, and an efﬁcient

crystallization was developed where 1c crystallized out of

(13) The mesylate salt 1a is deﬁned as a regulatory starting material. For

2-substituted morpholine systems, mesylate salts are common and

generally deliver excellent impurity control. See ref 7a and Berg, S.;

Larsson, L.-G.; Renyi, L.; Ross, S. B.; Thorbery, S.-O.; Thorell-

Swantessen, G. J. Med. Chem. 1998, 41, 1934.

(14) The Weinreb amide 7, described in reference 8, was chemically

unresolvable, and the high cost of N,O-dimethylhydroxyamine hy-

drochloride would be untenable.

(12) (a) Knochel, P.; Krasovsky, A. Angew. Chem., Int. Ed. 2004, 43, 3333.

(b) Murso, A.; Lang, S.; Hawk, D. Org. Process Res. DeV. 2006, 10,

733.

(15) Aldrich [79-37-8]; up to 1.0% phosgene content is listed for reagent

grade material.

212

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Scheme 7. IBCF approach to amide 12 synthesis

chloride intermediate was generated using thionyl chloride, but

due to the low solubility of the resulting intermediate salts it

was necessary to perform the chemistry in dichloromethane with

1 mol% DMF to assist with solubilization. The initial condition

used CH₂Cl₂(12 vol [L/ kg 10]) as solvent, Hu¨nig’s base (2

equiv) to sequester HCl, DMF (0.08 equiv), and morpholine

(2 equiv) as base and reactant. We quickly moved to evaluate

replacement of suspected carcinogen dichloromethane with

THF. When the THF conditions were run at 4-g scale it was

discovered that the addition of SOCl₂to a mixture of 10,

Hu¨nig’s base, and THF was very exothermic, but the reaction

mixture did not become homogeneous at any time. Quenching

a sample of the intermediate acid chloride mixture in MeOH

for HPLC analysis showed that the conversion to acid chloride

was incomplete, with 8% of acid 10 starting material remaining.

Optimization of the tandem chlorination/morpholine amide

formation reaction was attempted within the following general

design space: 10-12 volumes THF, 1.5-3.0 equiv of Hu¨nig’s

base (DIEA), 1.5-3.0 equiv of morpholine, 1.05-1.15 equiv

of SOCl₂, 0-0.5 volumes of DMF. Unfortunately, within these

ranges improvements were not made to the conversion, and

typically 8-16% residual starting material 10 remained. To

better control the reaction conditions, a European-style jacketed

ﬂask was employed along with a heated-ﬂuid circulator to

maintain a constant temperature of 30 °C throughout the

reaction. The amount of Hu¨nig’s base was increased slightly

from 2 to 2.5 equiv, but the SOCl₂used was decreased from

1.15 to 1.10 equiv. While the acid chloride formation left 14%

residual acid 10, by the end of the morpholine addition there

was <1% of acid 10 remaining. The fact that the acid was

eventually consumed intimated that the presence of the mor-

pholine might help the conversion, either by changing the

system’s solubility characteristics or by providing an irreversible

sink for the acid chloride. It was decided to test how the reaction

would behave if the SOCl₂was added to a mixture of all the

other materials, again using the jacketed ﬂask and heated ﬂuid

circulator to maintain temperature control. However, this

reaction performed poorly, leaving 17% of the acid unreacted,

and even after an additional 0.5 equiv of both morpholine and

SOCl₂were added in sequence the amount of remaining acid

was 10%. One notable observation made during this experiment

was that when the morpholine was added to the slurry of 10,

Hu¨nig’s base, DMF, and THF the slurry thickened, indicating

the morpholine hydrochloride salt was less soluble than the

DIEA·HCl salt. This was unfortunate because it indicated that

the Hu¨nig’s base might not be a very efﬁcient proton sponge

in the presence of morpholine. Perhaps having a larger amount

of DIEA present would increase the solubility of the morpholine

hydrochloride thereby setting up a stronger equilibrium between

the protonated morpholine and Hu¨nig’s base. To test this

hypothesis an experiment was conducted where the DIEA was

increased from 2.5 to 3 equiv, and the result was very surprising;

the conversion to acid chloride was much poorer in the presence

of increased DIEA, with 40% of 10 remaining. The deleterious

effect of larger amounts of Hu¨nig’s base on the reaction was

seen in later experiments as well when the reaction was

performed at twice the dilution, and again when inverse addition

was employed. These reactions never became homogeneous at

any point, so there was a possibility that mass transfer mixing

issues might be important. An inverse-addition experiment was

set up employing addition of a slurry of Hu¨nig’s base, 10, DMF,

and THF to a solution of SOCl₂and THF. The addition was

performed in small aliquots at 24 °C, and a stir time was allowed

between aliquots. At the end of the addition there was 8% acid

10 remaining, the same as the original conditions.

While the necessity of DMF in the reaction had not been

ﬁrmly determined, there was some question as to whether a

much larger amount could facilitate the reaction by increasing

the solubility of intermediate salts. To this end, the inverse-

addition reaction was repeated but with 0.5 vol of DMF rather

than the 0.1 equiv previously used. The amount of acid left

over rose to 10%, but that was still fairly close to the 8%

unconverted when much smaller amounts of DMF were

employed. Thus, it did not appear that DMF made much of an

impact at low levels, and higher levels were not desired due to

the difﬁculty of removing all of the DMF prior to resolution.

Thus, due mainly to the incomplete conversions we decided

that alternatives to SOCl₂would be evaluated.

A rapid screen of peptide coupling reagents revealed that a

mixed anhydride approach utilizing isobutyl chloroformate

(IBCF) as a possible SOCl₂replacement was feasible (Scheme

7).¹⁶Execution of the best prior conditions utilizing IBCF as a

replacement for SOCl₂using THF (10 vol), Hu¨nig’s base (2

equiv), IBCF (1.1 equiv), and morpholine (1.5 equiv) resulted

in a stalled conversion with 9% 10 remaining, consistent with

the thionyl chloride chemistry. The heterogeneous nature of the

acid-to-mixed anhydride conversion was still a concern, and

several attempts were made to overcome this issue. First, an

increase from 2 equiv of Hu¨nig’s base to 3.1 equiv was carried

out in order to complex all the acid possible and to see if the

increased base content would impart more solubility to the

intermediate salts. However, this reaction resulted in 10% of

unconsumed 10; not only was there no improvement but the

extra DIEA would have to be removed via distillations so this

avenue was not further pursued. Previous experience had shown

(16) (a) Prashad, M.; Har, D.; Ho, B.; Kim, H.; Girgis, M. J.; Chaudhary,

A.; Repic, O.; Blacklock, T. J. Org. Process Res. DeV. 2004, 8, 330.

(b) Prashad, M.; Prashad, K.; Repic, O.; Blacklock, T. J. Org. Process

Res. DeV. 1999, 3, 409. (c) Shie, H. W.; Carlson, J. A.; Shore, M. G.

Tetrahedron Lett. 1999, 40, 7167.

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

213

Scheme 8. (-)-DTTA resolution

Scheme 9. T3P coupling/resolution process

that the morpholine hydrochloride salt precipitated preferentially

to that of Hu¨nig’s base, but it was unknown what effect it would

have with the new coupling agent. Thus, a reaction was run in

which IBCF was added to a mixture of all the other reagents at

22-35 °C, which resulted in the formation of 20% of a

byproduct that was identiﬁed by LCMS as isobutylmorpholine

carbamate. Due to timeline constraints the mixed anhydride

IBCF chemistry was run in the pilot plant which replicated that

laboratory model where reaction progress stalled at ∼10%

residual acid 10 starting material. Ultimately, this was acceptable

since the procedure contained an aqueous workup which

rejected acid 10 to <1% of the crude amide product.

Morpholine Amide Resolution. A screening of pure

racemic amide 12 with resolving agents revealed enrichment

with the (S)-mandelate, quinate and (-)ditoluoyltartrate (-)-

DTTA) salts. The initial crystallization with (-)-DTTA gave a

very thick mixture, but in this case the mother liquor composi-

tion was 76% ee. This favorable eutectic prompted us to

evaluate crystallization of the (-)-DTTA salt in a variety of

solvents with varying stoichiometries. The optimal solvent was

2-propanol (20 volumes) with 0.5 equiv of resolving agent

which allowed the (-)-DTTA salt to be formed in 40% yield

with 100% ee from analytically pure 12. With successful proof

of concept established for the resolution, the next logical step

was to telescope the resolution and morpholine amide synthesis

steps. This was accomplished as several resolutions were

successfully run which involved addition of the resolving agent

either as a slurry or as a solution in 2-propanol. As discussed

vide supra, the elimination of Hu¨nig’s base from the racemic

12 solution prior to resolution was critical in order to obtain a

good yield. Many laboratory experiments were performed where

one or more vacuum distillations removed the majority of the

amine, and in those runs, samples were taken after each

distillation and solvent add-back in order to evaluate the

effectiveness of the distillations. A maximum level of residual

DIEA was determined to be 3 mg/mL in order to achieve at

least a 35% yield for the resolution step. Prior to this determi-

nation a pilot-plant resolution was performed with 10.7 mg/

mL of residual DIEA in the crude mixture which resulted in

only a 24% isolated yield. However, a second distillation was

installed for the next two pilot-plant lots, which reduced the

amount of residual DIEA to 1.5 and 1.4 mg/mL, respectively.

For these lots, which were performed on a 45-kg scale, 33%

and 38.4% yields of (-)-DTTA salt 13 were achieved with 98.8

and 97.5% ee, respectively (Scheme 8).

The use of 1-propanephosphonic acid cyclic anhydride

(T3P) as a peptide coupling agent has signiﬁcant prece-

dent.¹⁷During the course of development, a screen

revealed that T3P used under standard conditions (THF

[10 vol], Hu¨nig’s base [2 equiv], T3P [1.2 equiv], and

morpholine [1.5 equiv]) resulted in complete consumption

of the starting acid after 3.5 h; it was the ﬁrst experiment

in which total consumption of intermediate 10 was

observed. As a follow-up the T3P-mediated reaction was

scaled up to 50 g and was telescoped with the resolution

step. These experiments showed that the use of T3P

provided resolved salt 13 in 40-41% yield, a 5%

improvement relative to the IBCF process (Scheme 9).¹⁸

While the T3P process was not adopted for the pilot-plant

campaign due to timing constraints, it will potentially be

implemented for future campaigns.

In order to perform the Grignard chemistry, it was

necessary to develop an efﬁcient free base procedure that

would be amenable to facile water and (-)-DTTA

counterion removal. In this regard, toluene would be the

ideal extraction solvent because water could be removed

azeotropically. However, an attempted free basing in a

biphasic toluene/water system (9 vol of each) with 2.5

equiv of sodium carbonate failed to solubilize (-)-DTTA

salt 13, and the conversion failed. Fortunately, addition

of 9 vol of THF to this mixture produced a complete

solution after 15 min of stirring. HPLC analysis of both

layers showed that the di-p-toluoyltartaric acid had

partitioned almost exclusively into the aqueous layer

(0.52% in the organic layer), although 3.4% of 13

(17) (a) Boggs, S. D.; Cobb, J. D.; Gudmundsson, K. S.; Jones, L. A.;

Matsuoka, R. T.; Millar, A.; Patterson, D. E.; Samano, V.; Trone,

M. D.; Xie, S.; Zhou, X.-m. Org. Process Res. DeV. 2007, 11, 539.

(b) Rzepecki, P.; Geib, N.; Cernovska, K.; Konig, B.; Schrader, J. J.

Org. Chem. 2004, 69, 5169. (c) Liu, J. O.; Su, Z.; Dai, X. Tetrahedron

Lett. 2000, 44. (d) Wissman, H.; Hoescht, A. G. Phosphorous Sulfur

1987, 30, 643. (e) Wissman, H.; Kleiner, H. J.; Hoescht, A. G. Angew.

Chem. 1980, 92, 129.

(18) T3P is purchased as a 50 wt % solution in ethyl acetate. While the

cost is signiﬁcantly higher than IBCF ($200/kg vs $20/kg), the 5%

yield improvement would more than offset the extra reagent cost.

214

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Figure 1. Heat ﬂow for Grignard formation and coupling reaction.

Figure 2. Grignard initiator and dilution evaluation at 40 °C.

remained in the aqueous as well.¹⁹There was some

concern with possible racemization of the amide during

the free base step, but the optical purity decreased only

slightly from 97.0% in the starting material to 96.5% in

the free base. In order to further streamline the process,

an attempt was made to perform the free basing in THF/

water (with no toluene), but the layers would not separate.

Ultimately, the free base procedure was optimized with

6.5 vol (L/ kg 13) of toluene, 6 vol of water, 3 vol of

THF, and 3 equiv of sodium carbonate. These conditions

were effectively implemented, and two pilot-plant lots

were run with a 97% yield for the free base step.²⁰Overall,

the improved efﬁciency of this step strongly contributed

to the reduced E-factor for the synthesis of 1a.²¹

Grignard Chemistry. It is well-known that Grignard

reactions are extremely moisture-sensitive. Consequently

the organic solution of 12a needed to be dried to a low

level of water. After the free base procedure was

implemented, solvent exchange under conventional condi-

tions was performed to produce a 20-25 wt % solution

of chiral amide 12a with <300 ppm water. A method was

developed to monitor the Grignard reaction initiation by

¹H NMR since the active Grignard reagent itself was not

stable enough to analyze directly. An aliquot of the

Grignard reaction mixture was quenched into CD₃OD,

which would convert any active Grignard to THP. The

extent of reaction could then be determined by comparing

the integration of 4-chlorotetrahydropyran, 8a, to tetrahy-

1

dropyran (THP) H NMR resonances. The 4-chlorotet-

rahydropyran multiplet at 4.2 ppm was a good indicator

because this signal is far removed from other 8a or any

¹H NMR resonances in THP. In addition, the multiplets

in THP at 1.6 and 1.5 ppm were sufﬁciently resolved from

the Cl-THP aliphatic resonances. Therefore, the extent of

Grignard formation was calculated by comparison of the

4.2 ppm signal of Cl-THP with that of the 1.6 and 1.5

ppm signals of THP.

The Grignard reagent 8b was initiated by addition of

a small charge of 8a to a THF suspension of magnesium

(1.5 equiv) and iodine (cat.) and warmed to 60 °C; the

initiation was generally accompanied by a 3-4 °C

temperature rise. Once initiation was veriﬁed by ¹H NMR,

then the remaining 8a was fed at a rate such that a light

reﬂux was not exceeded. After full conversion to active

Grignard reagent was observed, the mixture was cooled

to 23 °C, and then an amide 12a toluene solution was fed

over at least an hour. In a pure THF solvent system the

Grignard reaction rate was dosing controlled. However,

use of amide 12a as a dilute toluene solution (<10 wt %)

slows the reaction, and it can take up to 5 h to achieve

reaction completion (<1% morpholine amide 12a). Ulti-

mately, the Grignard chemistry was found to be excep-

tionally sensitive to residual water, activator choice, and

grades of magnesium. For these reasons, additional

(19) Extraction of the aqueous layer reduced this amount to 0.27%.

(20) In the aqueous process streams ∼2-3% product was lost.

(21) (a) E-factor ) total kg of all materials/per kg of active pharmaceutical

ingredient (API). See: Sheldon, R. D. Chem. Ind. (London). 1992,

93. (b) Sheldon, R. D. CHEMTECH. 1994, 24, 38.

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

215

Figure 3. Grignard reaction sensitivity to water at 40 and 60 °C.

development was performed with reaction calorimetry

using the ReactMax calorimeter on 5-10-mL-scale ex-

periments. Such systems have been used industrially and

academically for reaction kinetics and thermal character-

ization.²²The measured heat ﬂow was proportional to the

measured reaction rate by the following relationship:

reaction, respectively (Figure 1). The resulting adiabatic tem-

perature rise can be calculated as follows:

N_subs∆H_rxn

Heat generated by reaction

Heat sink from reaction mixture

∆T_ad)

)

M_rxnC_p

(2)

N_subs, ∆H_rxn, M_rxn, and C_pare the moles of reactant, heat of

reaction, mass of the reaction mixture contents, and reaction

mixture heat capacity, respectively. These measurements played

a central role in determining safe pilot-plant operational

procedure. For example, conducting the Grignard reaction with

0.4 L/kg THF results in an adiabatic heat rise of approximately

160 °C.²⁵However, diluting the reaction mixture to 2.5 L/kg

of 8b results in a reduction of the adiabatic heat rise to 100 °C

for the Grignard formation and coupling reactions, respectively.

On pilot-plant scale, 8a was charged to a mixture of THF, Mg,

DIBAL-H,²⁶and iodine in ﬁve equal portions. The resulting

adiabatic heat rise for each injection would then be <20 °C

which would result in a maximum pressure rise of <1 psig in

a closed reactor.²⁷In addition, samples were taken after each

Q_meas) R_rxn∆H_rxn

(1)

where Q_meas, R_rxn, and ∆H_rxnare the measured heat ﬂow

(mW ) mJ/s), instantaneous reaction rate (mol/s), and heat

of reaction, respectively.²³The heat ﬂow proﬁles for the

Grignard formation at 40 °C and subsequent reaction with

morpholine amide 12a at 22 °C were measured using

standard conditions (Figure 1). The slow dissipation of

heat during the Grignard formation is indicative of the

reaction not being feed limited at 40 °C, with potential

accumulation.²⁴However, the heat load was proportional

to the charge amount, indicating that the heat effects are

uniformly distributed throughout the course of the Grig-

nard formation. The subsequent reaction of the Grignard

with amide 12a administered as a 20 wt % solution in

toluene was clearly feed limited; and the heat load was

distributed uniformly throughout the course of the reaction.

Integration of the heat ﬂow proﬁles yield a value of 54 kcal/

mol and 37 kcal/mol for the Grignard formation and coupling

1

20% addition of 8a and analyzed by H NMR to ensure

consumption before proceeding to the coupling reaction.

As with any Grignard, one of the key drivers during

development was to understand the factors that affected the

initiation kinetics. There were early indications that the use of

DIBAL-H along with iodine facilitated initiation kinetics

compared to either one alone. The initiation kinetics (in the form

of heat ﬂow) using 0.4 L THF/kg of 12a were measured in a

parallel reaction calorimeter (Figure 2). In this experiment, 8a

was injected to all three reactors simultaneously, and while use

of the DIBAL/iodine system initiated immediately, longer lag

times were observed in cases where DIBAL-H and iodine were

used separately. The apparent synergistic role of DIBAL-H and

iodine is not completely understood; however, decreasing the

(22) (a) Dale, D. J. Org. Process Res. DeV. 2002, 6, 933. (b) Delhaye, L.;

Stevens, C.; Merschaert, A.; Delbeke, P.; Brioˆne, W.; Tilstam, U.;

Borghese, A.; Geldoff, G.; Diker, K.; Dubois, A.; Barberis, M.;

Casarubios, L. Org. Process Res. DeV. 2007, 11, 1104. (c) Stefanick,

S.; Grim, J.; Liu, F.; Sorgi, K. L.; Maryanoff, C. A. Org. Process

Res. DeV. 2003, 7, 1067. (d) Cajaiba da Silva, J. F.; Machado e Silva,

C. F. P. Org. Process Res. DeV. 2002, 6, 829.

(23) Measurement of the instantaneous heat ﬂow is a direct measurement

of reaction rate which in turn allows calculation of conversion and

reaction rate constants. Effective use requires correlation of the

observed heat effect with other analytical techniques such as NMR or

LC to ensure that other thermal events, i.e., side reactions, mixing,

and crystallization are not confounding the results.

(25) Typical heat capacity value of 2 J/g/°C was used for calculation of

the adiabatic heat rise.

(26) Tilstam, U.; Weinman, H. Org Process Res. DeV. 2002, 6, 905.

(27) Based on vapor pressure of THF at 80 °C since the reaction temperature

in the pilot plant was 60 °C.

(24) Reactions near reﬂux were faster, but the kinetics were not feed-limited.

216

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

iodine charge from 4% to 0.8% suppressed the initiation

kinetics. The synergistic effect of the DIBAL-H/iodine system

was only present for Grignard initiation, and there was no such

synergistic effect of reaction kinetics post initiation.

As with most Grignard reactions, the impact of water needs

to be closely monitored and controlled. The initiation kinetics

at 40 °C revealed that water levels needed to be controlled to

around 100 ppm as higher levels (300 ppm) resulted in

suppressed initiation at 40 °C (Figure 3). Upon completion of

initiation at 40 °C, the reactor temperature was increased to 60

°C, and there was no impact of water on reaction kinetics post

initiation. Signiﬁcantly, increasing the Grignard formation

temperature to 60 °C resulted in reduced sensitivity to the impact

of water on initiation, as evidenced by an experiment at 465

ppm water that had a rapid initiation. However, water content

beyond 465 ppm resulted in failure of the Grignard to initiate.

In order to implement a safety margin, the water level was

controlled to below 300 ppm for the initiation as well as for

the coupling reaction with amide 12a.²⁸Most signiﬁcantly, the

initiation kinetics at 60 °C with water as high as 300 ppm

showed dosing-controlled initiation. Thus, the decision was

made to conduct the Grignard reaction in the pilot plant at 60

°C where the process was more robust and could be safely

executed with dosing control.

Recently, there have been several reports of IR technology

used to evaluate Grignard initiation and reaction chemistry.²⁹

Mid-IR and Raman spectroscopy were evaluated as online tools

for monitoring the disappearance of 8a, and early laboratory

scale evaluation indicated that a weak band at 1150 cm^-1could

be monitored for reaction progress. This peak was correlated

with heat ﬂow measurements, however, larger-scale experiments

resulted in overlapping features that clouded this region. A

weaker stretch at 830 cm^-1was measured, tracked, and

correlated with heat ﬂow and NMR data on laboratory scale.

The mid-IR bands for the pyran stretches had weak intensity;

and while they could be used to monitor initiation in the pilot

plant, they could not be used to monitor the reaction progress

after approximately 40-60% of 8a had been converted.

Grignard studies in the literature indicate that a band around

400-700 cm^-1might be present that could be ascribed to the

C-Mg stretch; however, this stretch was not detected by mid-

IR. In addition, Raman spectroscopy was also explored as a

means to track the disappearance of 8b, and future efforts for

monitoring this reaction via Raman might be feasible with

further development and calibration focused on the 337 cm^-1

peak (Figure 4). Unfortunately, solids formed during the

Grignard initiation that fouled monitoring by all IR approaches,

thus we relied on temperature monitoring and ¹H NMR to verify

that initiation had occurred.

Figure 4. Comparison of NIR and Raman spectra for initiation

monitoring.

Table 1. Impact of reverse and normal addition mode for

Grignard reaction

entry 1

entry 2

entry 3

addition mode

reverse

addition

30 min.

15 °C

normal

addition

1-2 min.

15 °C

normal

addition

1-2 min.

-10 °C

79%

feed rate

reaction temperature

conversion

40%

68%

was observed that the Grignard formation appeared to have

proceeded normally, but only 20% of amide 12a was converted

to product. The only notable deviation from the standard

procedure was that amide 12a had been added over 10 min

instead of the usual 1-2 h. In general, mixing effects are

observed when the intrinsic reaction rate is faster than or equal

to the mixing rate. To assess mixing effects, three experiments

were conducted in parallel examining the effect of addition rate

and mode at 15 °C: (1) Grignard addition to amide 12a (inverse

addition); (2) rapid addition of amide 12a to the Grignard

reagent at 23 °C, and; (3) rapid addition of amide 12a to the

Grignard reagent at -10 °C. The results show a trend consistent

with adverse effect of poor mixing on reaction performance

(Table 1). The reverse addition effectively mimics the extreme

case of a poor mixing condition that would result from

segregation of amide 12a in the reaction mixture as a result of

incomplete blending (Table 1, entry 1). The lower conversion

with the reverse addition highlights the need to maintain low

concentrations of amide 12a during coupling with Grignard

reagent 8b in the toluene/THF solvent system. The apparent

mixing sensitivity was partially reduced by lowering the reaction

temperature, which in turn lowers the reaction rate while not

affecting the mixing time scale. Nevertheless, these results

underscored the need for a slow reaction feed and it is

hypothesized that the active Grignard species is sufﬁciently basic

to deprotonate amide 12a when the latter is locally an excess

reagent. Consequently, for pilot plant operations, the addition

time of morpholine amide 12a was increased to 5 h to suppress

the mixing sensitivity and provide maximum performance. In

addition, the temperature during the addition of amide 12a was

kept between 15 and 22 °C for two reasons: to control the

exotherm and to minimize racemization of the product. Moni-

toring the coupling reaction is less complicated because

introducing the amide 12a moiety imparts a good chromophore

that can be easily detected by HPLC analysis. The reaction is

Grignard Reaction Mixing Effects. The effect of mixing

on coupling of the Grignard of 8a with morpholine amide 12a

was conducted by examining the effect of charge time and

addition mode on reaction performance. In one experiment, it

(28) Since 8a is added neat, a much higher water toleration is allowed for

this substrate (1500 ppm).

(29) (a) Wiss, J.; Lanzlinger, M.; Wermuth, M. Org. Process Res. DeV.

2005, 9, 365. (b) Wiss, J.; Ermini, G. Org. Process Res. DeV. 2006,

10, 1282. (c) am Ende, D. J.; Clifford, P. J.; Deantonis, D. M.;

Santamaria, C.; Brenek, S. J. Org. Process Res. DeV. 1999, 3, 319.

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

217

Scheme 10. Synthesis of mesylate salt 1a

typically done within two hours of completing the amide 12a

addition, though longer stir times are not harmful.

3 h and then whatever solid was still out of solution was ﬁltered

off and dried. For the 22 °C experiment, the mixtures were

heated to 60 °C for 2 h, cooled to 22 °C, and then allowed to

stir at that temperature overnight before ﬁltering and drying.

The ﬁltrates were all concentrated and dried to constant weight

and the ﬁltrate mass was the number that was used to calculate

solubility. The data are displayed as a function of solvent

composition and temperature (Figure 5).

Grignard Reaction Workup. During initial development

there was anecdotal evidence suggesting that performing the

reaction quench by addition of the reaction mixture to acetic

acid/water (inverse quench) rather than the typical addition of

acetic acid/water directly to the reaction mixture (normal

quench) was necessary to minimize stereochemical scrambling.

We reasoned that the highly basic Grignard reagent, 8b, could

deprotonate product 1b. However, because the inverse quench

was less convenient, and would require the use of an additional

tank when run in the pilot plant, we performed an experiment

to assess the direct quench protocol. Whereas the typical crude

1b enantiomeric excess is 96-98% by feeding the reaction

mixture directly into acetic acid/water, the stereochemical purity

was only 61.4% ee. Conversion of the crude to mesylate salt

1a and crystallization/isolation in 7:3 2-propanol/toluene resulted

in an ee upgrade only to 69%. In addition, the heat of reaction

for the quench into acetic acid was measured at 67 kcal/mol

which resulted in an adiabatic heat rise of 34 °C, of which 33%

was attributed to the sensible heat effects associated with adding

the reaction mixture at room temperature. For these reasons

pilot plant operations for the quenching step were conducted

by precooling the reaction mixture prior to direct addition to

the cold acetic acid/water solution. This lowered the adiabatic

heat rise to 23 °C and protected stereochemical integrity during

the quenching process.

Mesylate Salt Formation and Crystallization. The initial

procedure developed for crystallization of mesylate salt 1a was

successfully performed by heating a crude solution of free base

1b in 10 vol. of 7:3 (v/v) 2-propanol/toluene, charging meth-

anesulfonic acid at 60 °C, then the resulting slurry was heated

to 70-75 °C in order to effect dissolution. The mesylate salt

1a was then crystallized by slow cooling, and under these

conditions thick mixtures result along with enrichment of

undesired (R)-enantiomer. In addition, the resulting mesylate

salt was found to retain high levels of residual toluene and

2-propanol (3-5%). In order to optimize the salt formation/

crystallization sequence, three sets of experiments were run at

22 °C, 40 °C, and 60 °C. The primary rationale for evaluating

multiple temperatures was to ﬁnd a solvent ratio that would

give good solubility of 1a at higher temperatures and minimal

solubility at room temperature where the isolation would take

place. A secondary goal was to identify which solvent ratio

had a solubility curve that was not as steep in the range where

seeding (or initiation of crystallization) would take place. Each

reactor well was charged with 2 g of 1a and 20 mL of solvent

were added at 2-propanol percentages of 0, 20, 50, 80, and

100% with respect to toluene. For the 40 and 60 °C experiments,

the suspensions were heated to the prescribed temperature for

Figure 5. 1a solubility data.

An analysis of Figure 5 demonstrates that using 20% or less

of 2-propanol is not a practical option because 1a is not soluble

enough at 60 °C to prevent the salt from precipitating as soon

as methanesulfonic acid is added. The best choices based upon

solubility data at 22 °C are either 80% or 100% 2-propanol;

both of these compositions also provide good solubility at 60

°C. Ultimately, the choice was made to switch the crystallization

to 100% 2-propanol which was a simpler system and maximized

yield.³⁰Finally, eliminating toluene from the crystallization

would alleviate the problem of high levels of toluene being

retained in the isolated mesylate salt 1a. The following protocol

was devised for crystallization of the mesylate salt: crude 1b

was dissolved in 10 volumes (based on the theoretical amount

of 1b from the Grignard reaction) of 2-propanol, then heated

to 70 °C. Methanesulfonic acid (1.0 equiv) was added in one

portion and the solution was allowed to cool to 60 °C, which

caused the salt to precipitate. After aging at 60 °C for an hour,

the slurry was allowed to cool to 20-25 °C and was stirred for

at least an hour. Filtration, washing with 2.5 volumes of

(30) Residual toluene <1% was readily achieved via solvent exchange due

to the favorable azeotrope with 2-propanol/toluene.

218

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Scheme 12. Process impurities

2-propanol, and drying gave white crystalline 1a in >98%

enantiomeric purity in 90% yield (Scheme 10). In addition, the

absolute stereochemistry of 1a was conﬁrmed by single X-Ray

crystallography (Scheme 11).

Scheme 11. ORTEP for 1a acetone solvate

crystallization process was exposed to g60 °C temperatures

during the pilot plant campaign.

During the course of development a key impurity (S)-1-(4-

benzylmorpholin-2-yl)-3-methylbutan-1-one, 14, was identiﬁed

by LCMS as a contaminant in some of the key 1a lots (Scheme

12). The identity of this impurity was conﬁrmed via independent

synthesis by reaction of isobutylmagnesium chloride with amide

12a followed by crystallization as its methane sulfonic acid salt.

Initially, impurity 14 was thought to arise from small amounts

of isobutyl chloride contaminating the 8a feedstock, but GC

analysis did not reveal any isobutyl chloride or potential

precursors. An analysis of the historical lots of 1a revealed a

range of 0.00 -0.40% and the only tangible differences between

these lots appeared to be the method of Grignard activation. It

was noted that all lots produced with I₂activation only resulted

in no detectable 14 in the isolated product, whereas the lots

that utilized DIBAL-H alone or in combination with iodine

contained various levels of the impurity up to 0.40%. In order

to test the idea that DIBAL-H was the source of the isobutyl

ketone 14 impurity, a series of side by side Grignard reactions

were performed with different activators, and the experimental

data clearly supported this hypothesis (Table 3). During the

Impurity Removal. Efﬁcient rejection of the undesired

enantiomer in 1b was of utmost importance to the

crystallization. In addition, the crystallization would also

be required to remove amide 12a left unreacted from the

Grignard reaction. Typically, amide 12a taken into the

three step process had a chiral purity ranging from

96-99%. Minor erosion of stereochemistry occurred

during the Grignard reaction and quench so the crystal-

lization of 1a has been effected on crude 1b lots ranging

from 90-98% ee. Unless there were extenuating circum-

stances, the Grignard reaction generally proceeded to a

low level of residual 12 (<0.5%). In order to test the

robustness of the crystallization, a series of crude 1b

solutions with both typical and atypical levels of amide

12a and (R)-enantiomer were evaluated under standard

crystallization conditions (Table 2).

Table 3. Grignard initiator screen^a

water initiation 12a

(ppm) time (h) area % area % area %

14

1a

entry

activator

1

2

3

4

5

6

7

8

4% DIBAL-H 3% I₂

11% TMSCl^b

14% TMSCl

10% I₂

10% DIBAL-H 10% I₂176

4% TMSCl

120

520

176

2.5

3.0

18

2.5

18

1.28

0.35

2.20

0.23

0.63

0.25

1.01

0.00

3.0

0.00

0.31

97.7

96.1

93.8

94.4

93.1

94.6

57.1

59.2

Table 2. 1a crystallization data^a

75

crude

% ee

1a

% ee

isolated

yield %

crude 12a

area %

ﬁnal 12a

area %

10% Vitride

20% DIBAL-H

2.5

34.6

34.5

entry

1

2

3

4

5

6

7

94.7

95.9

96.2

93.4

90.4

98.0

97.3

99.9

98.1

99.9

90.7

91.8

91.6

92.7

83.5

89.2

84.6

0.11

0.16

0.10

5.7

ND

0.41

1.9

ND

2.0

ND

0.19

1.3

^aAll data reported is in situ HPLC area % collected at 250 nm. ^bGrignard

formation at reﬂux (67 °C); all others were completed at 60 °C.

course of this work, TMSCl appeared to be a promising

surrogate for DIBAL-H, but initiation was very slow at 60 °C

and only became reliable at reﬂux, which was outside the safe

operating temperature. Unlike the (R)-enantiomer, which can

be completely purged at levels 6-8% in the crude mixture,

impurity 14 was only reduced by ∼50% from levels present in

the crude during the mesylate salt crystallization process. Under

normal activation conditions, (Table 3, entry 1) up to 1% of

impurity 14 could be produced in the crude, which translated

to ∼0.5% in the isolated product. However, since the next step

in the synthetic sequence was found to effectively reduce the

level of this impurity it was not necessary to alter the initiation

conditions for long-term production.³¹

^aAll data reported for isolated solid 1a.

The data clearly show that 2-propanol was effective at

upgrading the chiral purity to acceptable levels from as low as

90% ee in the crude mixture. However, analysis of the ﬁltrate

revealed that the level of (R)-enantiomer was typically 2-3%

higher than predicted by the crystallization so it was suspected

that minor amounts of erosion was occurring during the

mesylate salt formation and crystallization process. In order to

test this hypothesis enantiomerically pure 1a was heated to 60

°C in 2-propanol and the stereochemical purity was monitored.

After 8 h the purity had dropped to 97.5% ee and after 24 h

the stereochemical purity had dropped further to 80% ee. This

set of data led us to keep careful control of the time the

(31) Speciﬁcation of residual 14 contained in 1a isolated solid for production

lots was established at 0.50%. If necessary, recrystallization with

2-propanol reduces the level by ∼50%.

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

219

Scheme 13. Resolution recycle

The optimized Grignard conditions were run in the clinical

pilot plant to produce 55 kg of 1a in two lots in high yield

(89%) and purity (99.4% ee) (Table 4). Critical process

recycle, the enantiomeric excess had dropped to 97% and

impurity 14 had enriched to double the amountobserved in the

parent lot so further recycles were not pursued.

From acid 10, an overall yield of 50% can be achieved for

the 1c mandelate salt epimerization/recycle process with three

separate ﬁltrate recycles which is a 14% yield improvement

from the T3P process (Table 6, entries 3 and 5). From a process

Table 4. 1a pilot-plant campaign summary^a

entry 13 (kg) 13 ee (%) 1a (kg) 1a yield (%) 1a ee (%)

1

56

112

98.9

97.8

98.3

27.6

27.4

55.0

88.5

88.9

88.7

99.4

2

Table 6. 1a process comparison

avg/total

overall

yield (%),

^aData reported is HPLC area % at 250 nm for isolated solid, 1a.

entry Scheme(s)

process

10 to 1a

parameters identiﬁed for the Grignard reaction included (1)

<300 ppm water for initiation; (2) 60 °C operational temper-

ature for Grignard; (3) amide 12a feed rate of at least 1-2 h to

active Grignard reagent on <50-g scale (5 h on pilot-plant

scale); (4) quench by feeding reaction mixture to 2 equiv of

acetic acid/water solution to preserve chiral integrity; (5)

minimize time at g60 °C during mesylate salt crystallization.

Recycling Resolution Studies. Based on the propensity of

1a and 1b to lose chiral integrity during the Grignard reaction,

quench, and methanesulfonic acid salt crystallization processes,

it was anticipated that development of a recycling process for

the 1b (R)-isomer should be possible (Scheme 13). In fact, the

ﬁrst-generation route to produce the desired 1a (S)-isomer was

accomplished via resolution of racemic 1b with (R)-mandelic

acid. We found that after isolation of (R)-mandelate 1c (40.9

g, 31.4% yield) under typical process conditions, concentration

of the ﬁltrate (80% (R)-isomer), followed by dissolution in THF

and treatment with 5 N NaOH at 55 °C resulted in complete

racemization within 2 h. After aqueous workup and solvent

exchange to 2-propanol, resolution with (R)-mandelic acid

resulted in isolation of 1c in 18.6% yield with 98.2% ee (Table

5). With successful proof of concept achieved, two additional

1

6

ﬁrst-generation (Grignard reac-

tion with acid 10/mandelate salt

resolution.

second-generation (ICBF/DTTA

resolution/1a direct)

second-generation (T3P/DTTA

resolution/1a direct)

second-generation (mandelate

salt/1 recycle per two lots)

second-generation (mandelate

salt/3 recycles per single lot)

24

2

8-10

9-10

14

31

36

38

50

3

4^a

5^b

14

^a10 f {79%} 12 HCl f {54%} 1c f {90%} 1a. ^b10 f {79%} 12 HCl f

{69%} 1c f {90%} 1a.

efﬁciency standpoint, the best processing sequence for recycling

would be to run two separate resolutions, combine the ﬁltrates,

then execute the recycling process and this scenario was

successfully developed to deliver a 54% yield of 1c from

racemic amide 12. With a single ﬁltrate recycle for every two

lots produced of 1c, efﬁciency is gained because roughly the

same size 1c lots are produced for both the main and recycled

streams. However, in this case, the overall yield from acid 10

is only slightly improved relative to the T3P process (38 vs

36%) (Table 6, entries 3 and 4). A disadvantage to the

mandelate recycle process relative to the DTTA resolution

processes is a less favorable crystallization eutectic. Another

disadvantage is that it is not possible to use amide 12 directly

in the Grignard chemistry, rather it needs to be puriﬁed by

conversion to an achiral salt form prior to use in the Grignard

reaction, which reduces throughput. While a slight economic

advantage could potentially be gained at present from the

mandelate salt resolution with three recycles per lot, additional

optimization is possible with the (-)-DTTA resolution. For

example, a single epimerization/recycle of the amide 12

R-enantiomer in a modest 20% yield recovery would match

the highest yielding mandelic acid recycle process. Thus, future

development efforts will focus in this area.

Table 5. Batch recycle resolution data^a

entry

wt (g)

yield (%)

ee (%)

14 (%)

main lot

recycle 1

recycle 2

recycle 3

overall

40.9

24.2

15.5

9.5

31.3

18.6

11.9

7.3

99.7

98.2

99.4

97.0

98.6

0.34

0.50

0.45

0.71

0.50

90.1

69.1

^aData reported is HPLC area % at 250 nm for isolated solid, 1c.

recycles were performed to produce 1c in 11.9% and 7.3%

yields, respectively (Table 5). With three recycles included we

were able to achieve a 69.2% yield of 1c from racemic amide

12 hydrochloride salt.⁸It should be noted that for the third

220

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Scheme 14. Magnesium recycle

Table 7. Magnesium recycling study^a

Magnesium Recycle. On small scale it is practical to quench

the Grignard reagent and excess magnesium on a lot by lot

basis. However, on pilot plant or commercial scale handling,

of magnesium can be particularly hazardous. For high-volume

processes it is often convenient to run Grignard reactions in

continuous or near continuous fashion.³²While the 8b Grignard

reagent is soluble in THF (∼2 M), small amounts of ﬁne metal

particulates are difﬁcult to avoid and may be particularly

hazardous. We envisioned that a continuous process could be

developed for production of 1a as shown in Scheme 14. Key

features include the use of dedicated vessels for the Grignard

reaction, aqueous workup and crystallization. We anticipated

that a large magnesium heel could be generated via a single

activation process in the Grignard reactor that could be reused

several times without addition of further magnesium or activa-

tion.³³After the reaction was complete the crude reaction would

be transferred through a ﬁlter directly into the quench tank

containing excess acetic acid/water, which could also be

potentially recycled. The crude 1b solution would then be

transferred to a crystallization vessel for production of mesylate

salt 1a.

In order to establish proof of concept for a continuous

process, a Grignard reaction was run with 15 equiv of

magnesium (based on amide 12a) under otherwise standard

conditions. After the Grignard reaction was veriﬁed to be

complete (<1% amide 12a), magnesium was removed by

ﬁltration, and the resulting ﬁltrate was quenched by addition

into acetic acid/water in the normal manner to produce crude

1b. The process was then repeated an additional ﬁve times with

the recycled magnesium (Table 7). A key observation was that

recycle rxn ee (%) quench ee (%) 1b (%) 12a (%) 14 (%)

1

2

3

4

5

6

7

95.6

96.1

96.2

95.0

96.3

97.2

94.6

79.4

85.2

87.6

78.6

85.8

1.4

97.0

98.1

97.7

97.9

98.1

98.3

98.0

0.0

1.56

0.54

0.47

0.34

0.26

0.24

0.29

0.0

0.42

0.27

0.0

90.8

^aIn situ HPLC area % reported at 250 nm.

the reactions in each case performed as expected, with in situ

ee of 95-97%. However, the stereochemical purities of the

quenched mixtures were signiﬁcantly lower than those observed

in the reaction mixture. Since the isolation of excess magnesium

was not accomplished in an airtight environment it was

hypothesized that the decrease in ee was due to unwanted

moisture. In fact, an extremely slow ﬁltration was observed for

the sixth recycle of magnesium which resulted in a complete

scrambling of the stereochemistry where water condensation

was possible (Table 7, entry 6). A separate recycle experiment

where the magnesium and process stream were handled in an

environment with reduced moisture resulted in an improved

product ee (Table 7, entry 7).³⁴In addition, performing the

quench by addition of the reaction mixture into a large excess

of acetic acid/water resulted in a high ee (98%). However, in

this case higher than typical levels of 1b were partitioned into

the aqueous phase and required two toluene extractions to

recover. Taken together, these data suggest that a continuous

process with magnesium recycle is potentially feasible with

further development.

(32) (a) Van Alsten, J. G.; Reeder, L. M.; Stanchina, C. L.; Knoechel, D. J.

Org. Process Res. DeV. 2008, 12, 989. (b) Smith, A. A. Org. Process

Res. DeV. 1997, 1, 165. (c) Harrington, P. J.; Lodewijk, E. Org. Process

Res. DeV. 1997, 1, 75.

(34) The results underscore the necessity to avoid moisture. This is less of

a concern on a pilot plant or commercial scale where process streams

can easily be fully inerted and dried.

(33) Klokov, B. A. Org. Process Res. DeV. 2001, 5, 234.

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

221

Conclusions

A second-generation process to produce 1a on pilot-plant

(S)-(4-Benzylmorpholin-2-yl)(morpholino)methanone Di-

p-toluoyl-^L-(-)-tartrate (13). Preparation I. A 500-mL ﬂask

was charged with the acid salt 10 (19.99 g, 98% potency, 76.01

mmol, 1 equiv) and THF (180 mL). Diisopropylethylamine

(27.0 mL, 153 mmol, 2.0 equiv) was added, and the mixture

was heated to 30 °C. Isobutyl chloroformate (12.0 mL, 91.4

mmol, 1.2 equiv) was added via addition funnel at a rate that

kept the temperature below 40 °C; the resultant slurry was

stirred at 30 °C for 35 min. A solution of morpholine (10.0

mL, 115 mmol, 1.5 equiv) and THF (20 mL) was added via

addition funnel at a rate that kept the temperature e40 °C. The

solution was stirred for 1.5 h, and then PhMe and water were

added (40 mL each). The pH of the mixture was adjusted to

>8 with 5 M NaOH (29 mL); the layers were separated, and

the aqueous layer was extracted with PhMe (40 mL). The

combined organic layers were concentrated on a rotary evapora-

tor to a volume of 100 mL, 2-propanol (200 mL) was added,

and the solution was again concentrated to a volume of 150

mL. The mixture was polish ﬁltered through a 0.45 µm PTFE

membrane (with a 30-mL 2-propanol rinse) to remove the

suspended salts from the workup. The ﬁltrate was transferred

to a 500-mL ﬂask with a 2-propanol rinse (20 mL), and the

solution was heated to 70 °C. Solid di-p-toluoyl-^L-(-)-tartaric

acid (13.81 g, 35.74 mmol, 0.5 equiv) was added, followed by

a 2-propanol rinse (40 mL). The solution was seeded with the

(S)-(4-benzylmorpholin-2-yl)(morpholino)methanone di-p-tolu-

oyl-^L-(-)-tartrate product (18.4 mg), and a thick slurry resulted.

This slurry was heated to reﬂux and was maintained at that

temperature for 2 h, after which the heat was shut off, and the

mixture was allowed to cool slowly as it stirred overnight. The

mixture was ﬁltered through polypropylene backed with paper

on a Bu¨chner funnel, and the cake was rinsed with 2-propanol

(50 mL) prior to drying in a 55 °C vacuum oven; 7.83 g (35%)

of 13 was obtained as a white solid that was >99% ee by chiral

HPLC analysis; ¹H NMR (400 MHz, DMSO-d₆) δ 7.86 (d, J

) 8.4 Hz, 2H), 7.35 (d, J ) 8.4 Hz, 2H), 7.32-7.22 (m, 5H),

5.76 (s, 2H), 4.26 (dd, J ) 10.1, 2.2 Hz, 1H), 3.78 (m, 1H),

3.67-3.57 (m, 3H), 3.49-3.36 (m, 8H), 2.77 (d, J ) 11.9 Hz,

1H), 2.70 (d, J ) 11.5 Hz, 1H), 2.36-2.24 (m, 8H); ¹³C NMR

(DMSO-d₆, 100 MHz) δ 167.9, 166.8, 165.1, 144.8, 136.4,

129.9, 129.9, 129.8, 128.7, 127.9, 126.4, 72.3, 72.0, 66.7, 66.4,

66.0, 62.0, 53.8, 52.4, 46.0, 42.1, 21.7; IR (KBr) 1129, 1225,

1243, 1648, 1722 cm^-1; MS (TOF) m/z 291.1701 (291.1703

calcd for C₁₆H₂₃N₂O₃, MH); Anal. Calcd for C₃₆H₄₀N₂O₁₁: C,

63.90; H, 5.96; N, 4.14. Found: C, 63.83; H, 5.95; N, 4.29;

scale in >98% ee and 31% overall yield from acid 10 was

developed, an improvement of 30% from the ﬁrst-generation

approach. Key features include resolution of (-)-DTTA salt

13 and highly chemo- and stereoselective Grignard chemistry.

We also demonstrated that the coupling/resolution step yield

could be improved by 5% using T3P as a replacement for IBCF.

Proof of concept studies were completed, demonstrating that it

might be possible to run the Grignard process in continuous

mode. In addition, feasibility of an alternate process to 1a

through a resolution/epimerization/recycle process with (R)-

mandelic acid was demonstrated.

Experimental Section

Reaction solvents and reagents were used as purchased, and

no special precautions were used to further dry them unless

otherwise noted; reactions were carried out under a dry nitrogen

atmosphere. Column chromatography, where necessary, was

carried out with silica gel (230-400 mesh). ¹H NMR (300 or

400 MHz) and ¹³C NMR (100 MHz) were recorded in CDCl₃

unless otherwise noted.

4-Benzylmorpholine-2-carboxylic Acid Hydrochloride

(10). A 1-L ﬂask was charged with 2-chloroacrylonitrile (2-

CAN, 98% potency, 36.8 mL, 453 mmol, 1.01 equiv) and PhMe

(105 mL). A solution of N-benzylethanolamine (66.0 mL, 95%

potency, 440 mmol, 1 equiv) and PhMe (35 mL) was added to

the ﬂask, and the resulting solution was stirred overnight.

Toluene (208 mL) was added, and the solution was cooled to

<-5 °C with a brine/dry ice bath. In a separate 500-mL ﬂask,

KO-t-Bu (50.84 g, 98% potency, 444.0 mmol, 1.01 equiv) and

THF (220 mL) were combined, and the resulting mixture was

added to the cooled solution in the 1-L ﬂask at a rate that kept

the temperature <-1 °C. After 50 min, the mixture was

quenched by adding water (175 mL), which raised the tem-

perature to 3 °C. The organic layer was ﬁrst rinsed with 10%

(w/v) NaCl solution (75 mL), and then was concentrated to

235 g of solution. The solution was extracted with 6 M HCl,

and the aqueous layer was transferred to a round-bottom ﬂask

where it was heated to reﬂux for 2.5 h. The heat was then turned

off, and, when cooled, the solution was seeded with 4-benzyl-

morpholine-2-carboxylic acid hydrochloride (65.6 mg). Pre-

cipitation of the product created a thick slurry that was cooled

to <10 °C for 45 min prior to ﬁltration through polypropylene

and paper on a Bu¨chner funnel. The ﬁlter cake was rinsed with

copious acetone (310 mL total), and was then dried in a 50 °C

vacuum oven to provide 78.71 g of product 10 (98% potency,

[R]²⁰-62.2 (c 0.1, EtOH); mp (DSC): 175.2 °C.

589

Preparation II. A 1-L ﬂask was charged with the acid salt

10 (49.99 g, 190.1 mmol, 98% potency, 1 equiv) and THF (450

mL). Diisopropylethylamine (68 mL, 380 mmol, 2.0 equiv) and

1-propanephosphonic acid cyclic anhydride (T3P, 50 wt % soln

in EtOAc, 137 mL, 230 mmol, 1.2 equiv) were added

sequentially, with the T3P addition being followed by a THF

rinse (50 mL). Morpholine (25.0 mL, 287 mmol, 1.5 equiv)

was added, and the temperature rose to 48 °C. After 1.25 h at

22 °C, the mixture was partitioned between water (200 mL)

and PhMe (50 mL), and the pH was adjusted from 4 to 8 with

5 N NaOH (73 mL). The layers were separated, and the aqueous

layer was rinsed with PhMe (2 × 100 mL); the combined

1

68% yield) of a tan solid: H NMR (400 MHz, DMSO-d₆) δ

7.61 (m, 2H), 7.42 (m, 3 H), 4.56 (d, J ) 10.1 Hz, 1H), 4.33

(dd, J ) 17.6, 13.2 Hz, 2H), 3.96 (m, 2H), 3.38 (d, J ) 11.5

Hz, 1H), 3.16 (d, J ) 11.9 Hz, 1H), 3.05 (m, 2H); ¹³C NMR

(DMSO-d₆, 100 MHz) δ 168.8, 131.9, 130.0, 129.6, 129.2, 71.6,

63.2, 59.4, 51.4, 50.1; IR (KBr) 696, 750, 1101, 1147, 1284,

1413, 1444, 1753, 2604, 2867 cm^-1; MS (TOF) m/z 222.1122

(222.1125 calcd for C₁₂H₁₆NO₃, MH); Anal. Calcd for

C₁₂H₁₆ClNO₃: C, 55.93; H, 6.26; Cl, 13.76; N, 5.43; O, 18.62.

Found: C, 55.91; H, 6.16; N, 5.46; mp (DSC): 252 °C (dec).

222

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

organic layers were concentrated to 175 mL of volume on a

rotary evaporator. The concentrate was diluted with 2-propanol

(500 mL) and reconcentrated to 175 mL of volume. This was

diluted a second time with 2-propanol (500 mL), and the

resulting mixture was ﬁltered through a Celite pad on a glass-

fritted ﬁlter; the cake was rinsed with 2-propanol (100 mL).

The ﬁltrate was charged to a1-L ﬂask with a 2-propanol rinse

(75 mL), and the solution was heated to reﬂux. In a 500-mL

Erlenmeyer ﬂask a mixture of di-p-toluoyl-^L-(-)-tartaric acid

(DTTA, 37.66 g, 98% potency, 95.52 mmol, 0.5 equiv) and

2-propanol (200 mL) was heated to 70 °C on a steam bath to

obtain a solution. The DTTA solution was poured into the

reﬂuxing solution of racemic amide followed by a 2-propanol

rinse (50 mL). After stirring the resulting solution at reﬂux for

30 min, the solution was seeded with the (S)-(4-benzylmor-

pholin-2-yl)(morpholino)methanone di-p-toluoyl-^L-(-)-tartrate

product (3.2 mg), and a thick slurry resulted. After 2 h at reﬂux

the heat was turned off, and the mixture was allowed to cool

as it was stirred overnight. The mixture was ﬁltered through

Polypro backed with paper on a Bu¨chner funnel, and the ﬁlter

cake was rinsed with 2-propanol (150 mL) prior to drying in a

55 °C vacuum oven; 51.31 g (40%) of 13 was obtained as a

white solid that was >99% ee by chiral HPLC analysis.

(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-

methanone Methanesulfonate (1a). (S)-(4-Benzylmorpholin-

2-yl)(morpholino)methanone (12a). A 2-L ﬂask was charged

with (S)-(4-benzylmorpholin-2-yl)(morpholino)methanone di-

p-toluoyl-^L-(-)-tartrate (13, 60.03 g, 88.71 mmol, 1 equiv),

PhMe (390 mL), water (360 mL), and THF (180 mL). Solid

Na₂CO₃(28.2 g, 266 mmol, 3 equiv) was added in one portion,

and the mixture was stirred at 23 °C for 1 h, during which time

all the solids dissolved. The layers were separated, and the

aqueous layer was extracted with PhMe (250 mL). The

combined organic layers were washed with water (300 mL),

after which they were concentrated to a volume of 90 mL. The

concentrated solution was diluted with PhMe (300 mL) and

was again concentrated to 90 mL of volume; this dilution/

concentration was repeated one additional time, and the

concentrated solution was ﬁnally diluted with PhMe to a mass

of 103 g with a water content of 173 ppm. This solution was

used as-is in the Grignard reaction, with an assumed quantitative

yield of S-free base 12a (25.74 g).

°C overnight. The resulting solution was added dropwise to a

solution of HOAc (13.8 mL, 241 mmol, 2.7 equiv) and water

(75 mL) at 2 °C. The addition took 40 min, and the temperature

was kept below 10 °C. When the addition was completed, the

mixture was allowed to warm to 22 °C, at which point the layers

were separated. The organic layer was rinsed sequentially with

water (120 mL), a solution of Na₂CO₃(9.44 g, 89.1 mmol, 1

equiv) in water (90 mL), and water (120 mL). The organic layer

was concentrated to a volume of 60 mL on a rotary evaporator,

and this concentrated solution was diluted with 2-propanol (250

mL) before being concentrated back to 60 mL of volume. The

concentrated solution was diluted to a volume of 235 mL and,

after being transferred to a 500-mL ﬂask, was heated to 70 °C.

Methanesulfonic acid (8.51 g, 88.6 mmol, 1 equiv) was added

in one portion. The solution was allowed to cool to 60 °C,

during which time the salt precipitated. The slurry was

maintained at 60 °C for 1 h, after which it was allowed to cool

to 22 °C. After a 4-h stir, the slurry was ﬁltered, and the cake

was rinsed with 2-propanol (80 mL) before being dried in a 45

°C vacuum oven to provide the product 1a as a ﬂuffy, white,

crystalline solid (30.47 g, 89% from 13). The solid was

1

determined to be >99% ee by chiral HPLC; H NMR (400

MHz, DMSO-d₆) δ 7.52 (m, 2H), 7.44 (m, 3H), 4.45 (m, 2H),

4.37 (s, 2H), 4.12 (dd, J ) 12.8, 2.6 Hz, 1H), 3.82 (m, 3H),

3.44 (m, 2H), 3.28 (m, 3H), 3.05 (m, 3H); 2.38 (s, 3H); 1.66

(m, 2H), 1.41 (m, 2H); ¹³C NMR (DMSO-d₆, 100 MHz) δ

207.6, 131.8, 130.2, 129.4, 76.1, 66.6, 66.5, 64.0, 60.0, 51.2,

50.6, 43.1, 27.9, 27.6; IR (KBr) 528, 553, 697, 1126, 1164,

1210, 1222, 1715 cm^-1; MS (TOF) m/z 290.1748 (290.1751

calcd for C₁₇H₂₃NO₃, MH); Anal. Calcd for C₁₈H₂₇NO₆S: C,

56.08; H, 7.06; N, 3.63. Found: C, 55.77; H, 7.09; N, 3.73;

[R]²⁰2.0 (c 0.1, EtOH); mp (DSC): 178.9 °C.

589

(S)-(4-Benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)-

methanone (R)-Mandelate (1c). Free-Basing of 13. A 3-L ﬂask

was charged with (4-benzylmorpholin-2-yl)(morpholino)metha-

none hydrochloride (200.01 g, 611.98 mmol, 1 equiv), EtOAc

(1 L), and water (750 mL), and the mixture was stirred

vigorously while Na₂CO₃(87 g, 820 mmol, 1.3 equiv) was

added in several portions. The resulting mixture was stirred for

1 h, and then the layers were separated. The aqueous layer was

extracted with EtOAc (500 mL), the combined organic layers

were rinsed with water (750 mL), and the resulting organic

solution was concentrated to an oil on a rotary evaporator. The

oil was dissolved in PhMe (600 mL), and this solution was

concentrated to an oil that was ﬁnally diluted with PhMe to a

total mass of 891.94 g; the concentration of amide 12 was 19.95

wt %.

Grignard Reaction with 4-Chlorotetrahydropyran (8a). The

Grignard reaction was the same as that employing (S)-(4-

benzylmorpholin-2-yl)(morpholino)methanone 12a described

above; this experiment used 430.04 g (295.47 mmol) of the

solution prepared above and resulted in a solution of racemic

(4-benzylmorpholin-2-yl)(tetrahydro-2H-pyran-4-yl)metha-

none, 1b, and 2-propanol with a total mass of 246 g, theoreti-

cally containing 85.49 g of free-base (34.7 wt %).

Grignard Reaction and Mesylate Salt Formation. A 500-

mL ﬂask was charged with Mg (3.31 g, 136 mmol, 1.54 equiv)

and THF (180 mL). Iodine (118.2 mg, 0.465 mmol, 0.005

equiv) and diisobutylaluminum hydride (3.6 mL, 1 M soln in

PhMe, 3.6 mmol, 0.04 equiv) were added, and the mixture was

heated to 60 °C. After 45 min, 4-chlorotetrahydropyran (8a,

16.52 g, 137.0 mmol, 1.54 equiv) was added in portions; the

ﬁrst portion was 20% of the total, and there was a 2-h stir at 60

°C after that addition. The remaining 80% of the material was

added in four equal portions with 45 min of stirring between

each addition. When all the 8b was added, the mixture was

stirred at 60 °C for 1.5 h. The mixture was allowed to cool to

22 °C and was then further cooled to 15 °C. The PhMe solution

of amide 12a (25 wt %) from above was added dropwise over

1 h, keeping the temperature between 15 and 22 °C. When the

addition was complete, the mixture was allowed to stir at 22

Resolution with (R)-Mandelic Acid. The 1b crude solution

(34.7 wt %) from the Grignard reaction was transferred to a

1-L ﬂask with a 2-propanol rinse (72 mL), and the solution

Vol. 13, No. 2, 2009 / Organic Process Research & Development

•

223

was heated to 50 °C. A solution of (R)-mandelic acid (44.99 g,

295.7 mmol, 0.48 equiv) and 2-propanol (232 mL) was added

dropwise over 50 min, maintaining a pot temperature of 50-55

°C. When the addition was completed, the heating mantle was

removed, and the solution was allowed to cool to 22 °C, during

which time the salt precipitated. The mixture stirred at 22 °C

overnight. The slurry was ﬁltered through paper on a Bu¨chner

funnel, and the cake was rinsed with 2-propanol (50 mL) prior

to being dried in a 45 °C vacuum oven. Obtained was 40.9 g

(92.6 mmol, 31% from the free-base) of 1c as a crystalline white

solid. Chiral HPLC analysis on the solid showed it to be >99%

ee; ¹H NMR (400 MHz, DMSO-d₆) δ 7.42 (m 2H), 7.28 (m,

8H), 5.01 (s, 1H), 4.10 (dd, J ) 9.7, 2.6 Hz, 1H), 3.88 (dt, J )

11.5, 2.6 Hz, 1H), 3.81 (m, 2H), 3.58 (dt, J ) 11.0, 2.6 Hz,

1H), 3.50 (d, J ) 1.8 Hz, 2H), 3.30 (dt, J ) 11.5, 2.2 Hz, 2H),

3.01 (tt, J ) 11.5, 3.5 Hz, 1H), 2.83 (m, 1H), 2.60 (dd, J )

11.5, 1.3 Hz, 1H), 2.13 (dt, J ) 11.5, 3.5 Hz, 1H), 2.00 (dd, J

) 10.1, 1.3 Hz, 1H), 1.62 (m, 2H), 1.41 (m, 2H); ¹³C NMR

(DMSO-d₆, 100 MHz) δ 210.4, 174.5, 140.7, 137.8, 129.4,

128.6, 128.5, 128.0, 127.5, 127.0, 79.3, 72.8, 66.7, 66.6, 66.4,

62.5, 54.1, 52.8, 42.7, 28.1, 27.9; IR (KBr) 504, 697, 735, 1088,

1707 cm^-1; MS (TOF) m/z 290.1748 (290.1751 calcd for

C₁₇H₂₃NO₃, MH); Anal. Calcd for C₂₅H₃₁NO₆: C, 68.01; H,

7.08; N, 3.17. Found: C, 68.10; H, 7.11; N, 3.33; [R]²⁰₅₈₉-42.3

(c 0.1, EtOH); mp (DSC): 101.5 °C.

acid was added in one portion causing an exotherm to 75 °C.

The solution was allowed to slowly cool to between 40 and 45

°C where precipitation of the salt occurred. The temperature

was held in this range for 1 h, and then the mixture was allowed

to cool to 22 °C as it stirred overnight. The slurry was ﬁltered

through paper on a Bu¨chner funnel, and the cake was rinsed

with 2-propanol (20 mL). A large quantity of precipitate form-

ed in the ﬁltrate as it cooled, so the ﬁltrate was poured back

onto the ﬁlter and was reﬁltered followed by a 2-propanol rinse

(10 mL). The solid was dried in a 45 °C vacuum oven to provide

14 as a white solid (8.35 g, 87%) that was shown to be 98% ee

by chiral HPLC; ¹H NMR (400 MHz, DMSO-d₆) δ 7.52 (m,

2H), 7.44 (m, 3H), 4.37 (s, 2H), 4.28 (m, 1H), 4.10 (dd, J )

12.8, 3.1 Hz, 1H), 3.80 (t, J ) 12.3 Hz, 1H), 3.42 (d, J ) 12.3

Hz, 1H), 3.25 (d, J ) 12.3 Hz, 1H), 3.03 (m, 2H), 2.45 (d, J )

6.6 Hz, 2H), 2.38 (s, 3H), 1.98 (hep, J ) 6.6 Hz, 1H), 0.82 (d,

J ) 6.6 Hz, 3H), 0.81 (d, J ) 6.6 Hz, 3H); ¹³C NMR (DMSO-

d₆, 100 MHz) δ 206.3, 131.8, 130.2, 129.3, 77.3, 63.9, 60.0,

51.1, 50.5, 47.1, 23.4, 22.7; IR (KBr) 528, 552, 697, 1153, 1221,

1461, 1715 cm^-1; MS (TOF) m/z 262.1800 (262.1802 calcd

for C₁₆H₂₄NO₂, MH); Anal. Calcd for C₁₇H₂₇NO₅S: C, 57.12;

H, 7.61; N, 3.92. Found: C, 57.44; H, 7.68; N, 4.08; [R]²⁰

589

60.0 (c 0.1, EtOH); mp (DSC): 163.0 °C.

Acknowledgment

(S)-1-(4-Benzylmorpholin-2-yl)-3-methylbutan-1-one Meth-

anesulfonate (14). A 250-mL round-bottom ﬂask (RBF) was

charged with THF (47 mL) and isobutylmagnesium chloride

(18.0 mL of a 2.0 M solution in THF, 36.0 mmol, 1.22 equiv).

The solution was cooled to 3 °C and a solution of (S)-(4-

benzylmorpholin-2-yl)(morpholino)methanone (12a, 8.58 g,

29.6 mmol, 1 equiv) and PhMe (40 mL) was added dropwise

over 1 h, maintaining a pot temperature between 0 and 10 °C.

When the addition was completed, the brown solution was

allowed to warm to 22 °C. After 2 h the solution was cooled

back down to <10 °C, and a solution of HOAc (4.6 mL, 80

mmol, 2.7 equiv) and water (60 mL) was added dropwise,

keeping the temperature below 10 °C. The layers were allowed

to separate, and the organic layer was washed with a solution

of Na₂CO₃(3.14 g, 29.6 mmol, 1 equiv) and water (60 mL),

followed by an additional wash with water (60 mL). The organic

layer was then concentrated to an oil that was chromatographed

(20% EtOAc/hexanes) to provide the pure product as a colorless

oil (7.01 g, 26.8 mmol, 91%).

We dedicate this article to Prof. A. I. Meyers who inspired

excellence in all those he worked with. In addition, we dedicate

this article to the memory Dr. Christopher Schmid, a former

colleague and friend. We also acknowledge analytical support

provided by Huijun Tian, Lin Wang, Brian Scherer, and Mark

Argentine; engineering support was provided by Ed Conder,

Guy Hansen, and Jim Ciula during the course of the project.

Finally, we thank all the brave men and women who worked

day and night to successfully execute this project in the Lilly

clinical pilot plant.

Supporting Information Available

¹H NMR, ¹³C NMR, DSC, and mass spectrometry data for

compounds 1a, 1c, 10, 13 and 14; X-ray crystallographic data

for 1a. This material is available free of charge via the Internet

at http://pubs.acs.org.

Received for review September 30, 2008.

OP800247W

The oil was dissolved in 2-propanol (70 mL) in a 250-mL

RBF. The solution was heated to 73 °C, and methanesulfonic

224

•

Vol. 13, No. 2, 2009 / Organic Process Research & Development

Article Doi

Practical synthesis of chiral 2-morpholine: (4-Benzylmorpholin-2-(s)-yl)- (tetrahydropyran-4-yl) Methanone mesylate, a useful pharmaceutical intermediate

DOI: 10.1021/op800247w

Source and publish data:

Authors:

Article abstract of DOI:10.1021/op800247w

Full text of DOI:10.1021/op800247w

Products guided by the article

R&D Labs maybe for 1120335-12-7

Relevant to this article

Hot Product