Process development and scale-up of AZD7545, a PDK inhibitor

Article

pubs.acs.org/OPRD

Process Development and Scale-Up of AZD7545, a PDK Inhibitor

Bharti Patel,^†Catherine R. Firkin,^†Evan W. Snape,^†Shelley L. Jenkin,^†David Brown,^†

Julian G. K. Chaffey,^†Philip A. Hopes,^†Carl D. Reens,^‡Mike Butters,^†and Jonathan D. Moseley*^,†,§

^†Process Research and Development, Avlon Works, AstraZeneca, Severn Road, Hallen, Bristol BS10 7ZE, United Kingdom

^‡Development Manufacture, R&D Charnwood, AstraZeneca, Bakewell Road, Loughborough LE11 5RH, United Kingdom

ABSTRACT: A brief comparison of the early manufacturing routes to AZD7545 is given. Process development of the preferred

long-term manufacturing route is reported in detail, and changes from the initial kilogram-scale route are discussed. Scale-up

experience from the pilot-plant manufacture is included in the discussion of each stage. Noteworthy aspects throughout the

development of AZD7545 concerned chemical hazards, mechanisms, analysis, and impurities, upon which this case study will

focus.

A contractor was chosen to supply the next delivery of

1−5 kg which was soon increased to 10−12 kg when the low

potency of AZD7545 became apparent. This used a

significantly modified route that emerged from early collabo-

ration between AstraZeneca and the contractor and in which

chiral acid 2 was added at the penultimate stage (Scheme 2).

The functionalities on both aromatic rings required for the key

Ullmann coupling step were transposed in 7a−d and 10

(compared to 3 and 4 in Scheme 1), which also avoided the

relatively expensive starting materials of 3 and 2-chloro-4-

iodoaniline required in the original route. The dimethylamide

unit was conveniently incorporated into the proposed

benzamide portion (7a−d) early in this synthesis, rather than

after structure 5. Although this route was more convergent than

the original one, repeat manufactures were required to support

ongoing toxicity studies.

Due to manufacturing problems, which effectively indicated

that this route was not reliably scalable, concurrent develop-

ment of the proposed long-term manufacturing began at

AstraZeneca. The next deliveries were to be of 23 and 35 kg,

respectively, and the early stages were to be manufactured

jointly, with half of the material accelerated to the later stages to

meet clinical trial deadlines. Table 1 summarises this infor-

mation, from which it will be seen that there was considerable

overlap in the delivery schedule. The result was several small

deliveries at short intervals with little or no time between them

to incorporate improvements. On the other hand, it did increase

the learning that was possible on each stage for both routes

(campaigns 2 and 3) such that the chemistry for each stage was

thought to be well-understood.

INTRODUCTION

■

AZD7545 (1) is a pyruvate dehydrogenase kinase (PDK)

inhibitor¹designed to provide an oral treatment for type II

diabetes²and was the first of a number of compounds to enter

development from AstraZeneca’s PDK project.³Consideration

of the structure of AZD7545 reveals that it can be assembled in

numerous ways. To aid in the route selection process, we

applied a Kepner-Tregoe decision analysis (KTDA) tool as

described in an earlier article,⁴from which the preferred manu-

facturing route was identified, as discussed in a companion

article.⁵This article concludes the series by presenting the

scale-up issues involved with the manufacturing processes that

were developed and as experienced on the pilot plant. It will

focus on those aspects of the development and manufacture

most likely to be of interest to readers of this journal, namely

issues of chemical hazards, mechanism, analysis, and impurities.

Additional important points to note regarding AZD7545 are its

low activity (∼300 mg/day dosing) and the high cost of chiral

acid (2),⁵which immediately presented a cost of goods issue

for the project. This provided a strong incentive to introduce

the chiral acid unit late in the synthesis.

The data for choosing the new route over the old one has

been presented before,⁵but a brief review may be helpful.

Although the campaign 2 route was more convergent, a major

advantage of the linear campaign 3 route was that it avoided the

difficult Ullmann reaction (Scheme 2). This still had a mode-

rate yield with a difficult and dilute workup procedure, albeit

somewhat improved over the equivalent Discovery chemistry

stage (cf. Scheme 1). However, the contractor had to introduce

COMPARISON OF THE MANUFACTURING ROUTES

■

Our Discovery colleagues had produced 270 g of bulk drug 1

(purified by chromatography) prepared by a convergent

route^1,5from 3 and 4 to give 5 which contained much of the

structure of AZD7545 (Scheme 1). This introduced the

expensive chiral acid 2 before the key bond formation in the

moderately yielding Ullmann reaction. This was a major dis-

advantage although it did facilitate easy introduction of

analogues into the left-hand portion of 1 (as drawn).

Received: November 28, 2011

Published: February 3, 2012

447

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

Scheme 1

Scheme 2. Comparison of the manufacturing routes

a recrystallization to purify the sulfide aniline 13 and to remove

copper residues which were simply not present in the campaign

3 route. There was also considerable doubt that the Ullmann

reaction would be scalable at campaign 3 scale.

Despite the linear nature of the new route, much of the

chemistry was expected to be simple and robust, including the

hydrolysis of the nitrile from 11 to 12. Formation of thio-

cyanate 10 from 2-chloroaniline (9) used the same reaction

(but not process) as the campaign 2 route. Sulfide aniline 13

was the key common intermediate from which the subsequent

steps would be identical, although in fact chiral acid 2 was

added as its unprotected acid chloride (14), rather than as the

TMS-protected acid chloride (16), which avoided another

difficult step.

The remainder of this article will discuss the development

and scale-up of the campaign 3 processes, although learning

448

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

a

Table 1. Summary of planned and actual manufacturing campaigns

planned

actual

delivery

date

planned

actual

delivery

date

a

campaign

delivery (kg)

location

plant

1

n/a

10−12

12

0.25

4.5

n/a

Aug. 2000

Sept. 2001

Jul. 2002

AZ (Discovery)

CRO

kilo lab

pilot plant

kilo lab

pilot plant

n/a

2a

May 2001

Apr. 2002

Mar. 2002

June 2002

Dec. 2002

Nov. 2002

mid 2003

2b

4.5

CRO

2b′

1.5

2.3

Apr. 2002

AZ (PR&D)

not decided

b

3a

23

n/a

3b (plan)

4 (plan)

5 (plan)

35

n/a

55

250

n/a

a

b

n/a = not applicable; Halted in April 2002 after first batch of sulfide aniline.

from the analogous campaign 2 stages will be included where

relevant. To aid the reader, the development of each stage is

presented as a smooth and continuous process, rather than in

the order in which some developments actually occurred, which

would be less easy to follow. On a final point, note that in

AstraZeneca terminology, the unrecrystallised bulk drug is

termed “Crude” and the recrystallized but unformulated bulk

drug is termed “Pure”. The chemical structures for both are

identical in this case (1), since there was no salt formation.

There was a crystallization step to further purify crude 1 to pure

1, but this will not be discussed here.

Scheme 3. Mechanism of thiocyanogen formation

presented another problem due to the interaction of bromine

with methanol which can produce methyl hypobromite

(Scheme 4), a member of the generally unstable class of alkyl

CAMPAIGN 3 ROUTE DISCUSSION

■

Thiocyanate Stage (10). The chemistry for the thio-

cyanate stage was based on the use of thiocyanogen (SCN-

NCS) from an older ICI/Zeneca project (ZD5522).⁶In that

procedure, a full equivalent of thiocyanogen had been prepared

in acetonitrile solution before being added to its reaction partner.

This had required cooling and careful control on plant scale⁷due

to the instability of thiocyanogen as noted in Bretherick (poly-

merises explosively above its melting point of −2 °C).⁸The

AZD7545 process as developed for the campaign 2 route⁵

generated the thiocyanogen in situ in the presence of 9 which

could react immediately. The workup used solid NaHCO₃to

quench the excess acidic residues, generating much gas, and

involved a solvent exchange from methanol into toluene. A final

peculiarity of the process involved the addition of NaBr to the

reaction for which there was little obvious reason.

Scheme 4

hypohalite compounds.⁸NaBr had been used in both the

ZD5522 process and the contractor’s campaign 2 process,

presumably with the aim of increasing [Br⁻] to help deter the

forward reaction. Comparison of the chemical hazard assess-

ments on processes with and without added NaBr revealed that

its presence raised the self-heating onset temperature by 18 K

from 11 to 29 °C, but that the size of the eventual exotherm

was nearly doubled. Overall, this suggested that NaBr was

helpful in delaying a thermal event, but that the consequences

would be worse if it was allowed to develop.

Work at AstraZeneca for campaign 3 manufacture started

with the campaign 2 process. A drown-out with aqueous

NaHCO₃precipitated the product 10 in good quality, which

avoided the solvent exchange into toluene and subsequent

distillation to a low volume. Using NaOH instead did avoid

CO₂evolution but tended to produce a lot of the disulfide 19,

which had been a problem in the campaign 2 process. A reverse

drown-out into aqueous bicarbonate was found to work even

better for the physical form of the product and had the

advantage of keeping the main reaction vessel dry for the next

batch.

Before discussing the chemical hazard issues of this process

further, which were adequately controlled for both ZD5522 and

AZD7545 campaign 2 manufacture, some discussion of the mech-

anism is necessary, since this will explain some of the previous

points. The reaction mechanism requires 2 equiv of NaSCN for

the formation of thiocyanogen (NCS-SCN) which is generated

by reaction with bromine, as shown in Scheme 3. Two full

equivalents of NaBr are produced as the byproduct.

It is thiocyanogen, which can behave as a pseudohalide,

which then achieves a standard electrophilic aromatic substitu-

tion reaction directly with electron-rich aniline 9.⁹One full

equivalent of NaSCN is generated as the byproduct (cf. Scheme 2).

Care must be taken not to acidify the reaction liquors, since acidic

hydrolysis of thiocyanogen can produce HCN (the ZD5522 proce-

dure required base treatment in any case for the ongoing reaction).^6,7

The campaign 2 process avoided the potential chemical

hazard issue of generating stoichiometric thiocyanogen by

instead forming it in situ so that it could react with 9

immediately. However, conducting the reaction in methanol

449

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

Addition of the bromine/methanol solution over 1.5 h was

not only better for controlling the heat of reaction but also gave

lower levels of the mono- and dibrominated impurities 20 and

21; a dump charge of bromine (on small scale) resulted in a

notably higher level of 20. Both impurities are presumably

formed by direct electrophilic aromatic substitution reaction. It

was decided to use a precise 1.00 equivalent charge of bromine

to minimize formation of these impurities, especially as they

were crystalline solids, whereas residual 9 was a liquid and more

easily washed out of the product cake. An alternative approach

would have been to preform the thiocyanogen separately as

in the ZD5522 process, which probably would have further

reduced the direct bromination of 9. Since the levels of

brominated impurities 20 and 21 were not high, this idea was

not investigated to avoid stoichiometric thiocyanogen.

off since they were not needed, but the tungsten temperature

probe and bottom-runoff valve would need replacing. Engineer-

ing modifications were not easy to put in place, especially at a

relatively late date before manufacture. As an alternative, the

reaction was tried using NBS to generate the thiocyanogen,

which appeared to work well in principle.¹⁰Even so, the effect

on the quality of the product 10 was not fully assessed, and it

was felt to be too late to make such a major change to the

process, which would have required its own hazard assessment

and development of the workup and isolation stages. However,

it did establish the principle of using NBS as an alternative to

bromine, which would certainly have been reviewed in a follow-

ing campaign.

The MoC issues were resolved when a glass-lined mild steel

(GLMS) reactor module became available sooner than

expected (in place of the planned Hastelloy reactor), and the

tantalum temperature probes were found to be sufficiently

resistant to bromine on the basis of previous use. Other Hastelloy

components in the plant module were replaced or bypassed.

As a final minor improvement, the product was also isolated

after cooling to 5 °C, which gave a small increase in the yield

with no loss in quality. Analysis of aqueous reaction residues

indicated that thiocyanate 10 gradually decomposed to the

thiol/thiolate (18a/b), from which the disulfide 19 then

formed. Although this had been earlier thought to be a problem

in the campaign 2a/b manufactures, the knowledge that the

disulfide 19 could react in the cyano aniline (11) stage to give

the required acid aniline (12) meant that a more generous

specification level could be set with regard to this “impurity”. It

also meant that a long hold point was potentially available if

required during plant processing. Finally, the assay of the

product could be somewhat low in some batches (80−90%)

due to their high inorganic content. Although the following two

stages coped well with high levels of inorganic salts, slurry

washes were implemented in place of displacement washes, and

the wash sizes increased, which improved product assay without

loss of yield.

However, the chemical hazard assessment showed that the

bromine/methanol solution still presented a significant risk on

plant scale; there was an adiabatic temperature rise of 36 K

on addition of bromine to methanol, and a further 45 K rise on

adding this solution to the NaSCN solution. The bromine

could have been added neat to the methanolic solution of

NaSCN and 9, and this was a safer recommendation, but the

slow addition required would have been difficult to implement

on plant scale due to the small charge size of undiluted bro-

mine. Even on laboratory scale, slow addition of bromine

proved difficult to achieve, and any uneven addition resulted in

an uneven temperature profile and raised levels of the

brominated impurities 20 and 21.

The ZD5522 process had used acetonitrile as the reaction

solvent.⁶When this was tried instead of methanol (keeping all

other factors identical), conversion and quality were very good,

but the product failed to precipitate on addition of the aqueous

NaHCO₃solution due to a phase separation with acetonitrile.

The obvious solution was to dilute the bromine with

acetonitrile (thus avoiding the hazardous interaction with

methanol whilst providing a useful charge volume) and add this

to a methanolic solution of NaSCN and 9, thus keeping methanol

as the bulk reaction solvent. With this modification both the

reaction and the drown-out worked well with only a few per-

cent drop in yield and no loss in quality. The chemical hazard

assessment confirmed that the acetonitrile/bromine mixture

was stable at room temperature once formed (adiabatic

temperature rise of only 4 K, presumably due to mixing, with

no self-heating up to 125 °C in a sealed tube test over 6 h).

There was also no accumulation of bromine in the main

methanolic reaction mixture, so the chemical hazards were now

well controlled. Yields of 80−85% with assay of 90−95% were

reliably obtained.

The results for the campaign 2a batches prepared at the

contractor using the original method, their repeat campaign 2b

manufacture, and the results for the campaign 3a manufacture

are shown in Table 2. Campaign 3a batches had improved assay

Table 2. Summary of AZD7545 thiocyanate (10)

manufacturing campaigns

campaign 2a

campaign 2b

campaign 3a

no. of batches

input of 9 (kg)

batch size (L)

assay range (%)

quality range (%)

yield range (%)

overall yield (%)

total delivery (kg)

location

11

6

4

1.34

17

6.7

38−41

500−530

92−96

94−95

78−81

80

87

79−91

n/a

88−94

95−97

63−83

74

a

58−83

75

15.9

36.1

CRO

183

CRO

AZ (PR&D)

a

n/a = not available.

Although we had designed a chemically safe process using

bromine, accommodating it on the plant now proved proble-

matic due to an incompatibility with several materials of

construction (MoC). A coupon test showed that the bromine/

methanol solution could significantly corrode Hastelloy,

tantalum, and tantalum/tungsten components (∼20% loss of

mass over 1 week). The tantalum condensors could be blanked

and yield but had marginally reduced quality due almost

entirely to raised levels of the dibromide impurity 21 in each

batch (∼2−3%). This may have been due to uneven mixing on

larger scale since for mechanical reasons a smooth addition

proved difficult to achieve. Even so, this clearly shows that the

450

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

improved campaign 3 process could be safely scaled up to 500 L

with no loss in yield or quality.

At the same time, the assay and quality of the Na₂S was also

investigated. The nonahydrate (Na₂S·9H₂O) is in fact 33%

Na₂S by weight, whereas the technical grade is ∼60% Na₂S by

weight. It was found that the source of the Na₂S did not matter,

but the quantity of water did, with the nonahydrate performing

poorly or not at all, depending on the stoichiometry. Using

Na₂S·9H₂O at 2.0 equiv put 12% water into the reaction

mixture (v/v with DMF), compared to only 4% when using the

technical grade (water from other sources such as the

thiocyanate and DMF was insignificant). To confirm this,

technical grade Na₂S (>60%) was used and water added up to

the 12% level, whereupon the reaction failed. Other prepara-

tions at intermediate water content levels showed a general

decline in formation of 11, 22, and 23 as the water content

increased, with a significant drop above ∼9%, as shown in

Figure 1. Azeotropically drying the intermediate thiolate 18b

Cyano Aniline Stage (11). Although we planned to

generate the intermediate thiol/thiolate 18a/b in situ from

thiocyanate 10 for the formation of cyano aniline 11, to

simplify the process, the first attempt used isolated thiol 18a

with 4-fluorobenzonitrile (8) in NMP with K₂CO₃. This gave a

conversion of 92% after only 1.5 h at 80 °C and established the

principle of the benzonitrile route (campaign 3) as a viable

alternative to the benzamide route (campaign 2). Significantly,

it showed that the key coupling reaction could be achieved by a

simple uncatalysed S_NAr reaction when using a more electron-

deficient aromatic acceptor species (i.e., 8 instead of 7a−d).

This avoided the high-temperature copper-catalysed Ullmann

reaction with its moderate yield, inefficient isolation procedure,

and copper waste stream. This also avoided the need to test for

and remove copper in the final bulk drug. After initial develop-

ment, the S_NAr reaction was found not to require K₂CO_3,but

two significant and variable impurities were identified as the

primary amide 22 and the thioamide 23. The 4-chlorobenzo-

nitrile analogue of 8, which was cheaper and had a higher mp,

was also tried and worked in principle but gave rise to higher

levels of the amide 22 and an unknown impurity, so this

approach was not investigated further.

Having established the principle of reaction, however,

development for scale-up proved somewhat troublesome. A

solvent screen gave no product in the cases of acetonitrile, ethyl

acetate, methanol, or toluene, with the disulfide 19 being the

major product in the case of acetonitrile and ethyl acetate;

NMP proved to be the best solvent, with DMA and DMF also

good candidates. However, cyano aniline 11 was difficult to

isolate from NMP, did not have good physical form, and was

of low and variable assay despite being apparently pure by

¹H NMR spectroscopy. It was suspected that residual inorganic

compounds were present in the NMP solvent, since the mass

balance could not be corrected accounting for the known

masses of the amide, thioamide, and residual 10 which were

present.

Throughout this work the thioamide 23 level also varied

considerably, being generally higher when more Na₂S was used

in the thiocyanate hydrolysis and for longer reaction times.

When the product was isolated by extraction into ethyl acetate

instead of drowning out with water from NMP, the inorganic

content (ash assay) was low. A long reaction time (>18 h) had

been used, and 22 and 23 accounted for 30% of the product

mass in this case. With a short reaction time and extraction into

toluene, product of 87% assay in an 83% yield was obtained

with only 2−3% 22 and 23. Whilst this worked well in the lab,

the workup with toluene extraction had some other problems,

tending to oil out and give poor recovery. It was also a poor fit

with possible telescoping with the following aqueous hydrolysis

stage. Changing to DMA or DMF from NMP and using an

aqueous drown-out reliably gave a solid in the case of DMF but

not with DMA.

Figure 1. Formation of cyano aniline (11) and thioamide (23) against

water content of reaction.

before addition to the benzonitrile solution increased the yield

by a few percent but was judged not to be worth the additional

operational complexity at pilot scale.

The process now gave a solid by direct aqueous drown-out

from DMF with better physical form, good assay (87%), and

acceptable yield (72%) with less thioamide (6%). However,

both impurities 22 and 23 could, in practice, be hydrolysed in

the next stage to give the desired acid aniline 12; as a result, the

yield was nominally higher. Disulfide 19 was also tried instead

of thiocyanate 10 with 4-fluorobenzonitrile (8) and found to

give a good conversion to cyano aniline 11; thus, formation of

disulfide in the thiocyanate cleavage of 10 to 18a/b was no

longer a chemical concern. A factorial experimental design (FED)

of eight experiments was run, covering four factors (Na₂S equi-

valents, temperature, concentration, and reaction time) plus

four additional centre points (12 experiments in total). As a

result, minor adjustments were made by increasing the reaction

temperature to 60 °C, which also enabled better solution of

Na₂S in DMF, and increasing the charge of Na₂S to 1.8 equiv.

This led to higher levels of thioamide, as did longer reaction

times. However, the centre points gave reproducible results and

proved that the process was already effectively optimized. A

final modification was to use a reverse drown-out, which

improved the form of the isolated solid 11 still further.

There were no major chemical hazard concerns with any part

of the process. The adiabatic temperature rise on addition of 10

to the DMF/NaSCN solution was 38 K, and there was a further

451

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

30 K rise on addition of 8 to the thiolate solution. However,

both of these exothermic additions were well controlled by slow

additions, and there were no other concerns.

Cyano Aniline/Acid Aniline Telescope (11/12). Given

the initial difficulty experienced in designing a suitable isolation

procedure for cyano aniline 11 in good form, its low assay, and

the fact that the two major impurities (22 and 23) were partial

hydrolysis intermediates on the way to acid aniline 12, it had

been desirable from an early stage to try and telescope these

two reactions together. Once problems with the cyano aniline

stage were understood and the Hastelloy reactor became

available, the telescoped reaction could be investigated.

As a starting point, a standard cyano aniline preparation was

drowned-out with water and extracted with toluene, to which

48% aqueous NaOH in ethanol was added. The reaction

mixture was heated to 80 °C for 6 h, which converted all of

intermediate 11 into desired acid aniline 12. This batch was

then split, and one-half isolated as the sodium salt, the other

neutralized to pH 6 and isolated as the neutral species 12. The

assays of both products were low (67 and 76%, respectively),

and the overall yield of 51% was lower than the individual

steps. However, it did establish the principle of the telescoped

process. Further scrutiny showed that toluene extraction

resulted in high losses and poor form; changing to MTBE

was much better and allowed a solvent exchange into higher-

boiling ethanol for the hydrolysis step. Acidification with

concentrated HCl gave on isolation a product of improved

physical form with 93% assay in 74% overall yield from

thiocyanate 10. However, a strongly sulfurous-smelling grey

scum was noted from the reaction which was difficult to clean

out and would be a problem on plant scale.

Various solvents were screened to see if a single solvent

could be used for both processes, and whilst several worked

well in the cyano aniline stage, only MIBK worked in the hydro-

lysis step. Due to the imminent start of plant manufacture, this

result could only be noted for future study. NMP was retried

for the cyano aniline stage and found to give less grey scum and

bad odor, and also less thioamide 23. DMF had been the

preferred solvent for the single reaction and gave a better form

of 11, but now that extraction was replacing isolation, the need

for a good drown-out solvent was not required. Although 11

could crystallise from MTBE on storage, it remained stable in

solution when ethanol was added, so could be stored for several

weeks.

The charge of Na₂S was reduced to 1.5 equiv which also gave

less grey scum. When using less pure batches of thiocyanate 10,

the grey scum was again higher, suggesting that it was related to

impurities in 10 more than the charge of Na₂S. A large lab-scale

preparation (5-L scale) using plant-derived 10 worked well with

the solvent exchange from MTBE to ethanol after the

extraction procedure and gave expected yield and quality.

Typical laboratory batches now had only two major impurities,

residual amide 22 and dibromide 21 carried over from the

thiocyanate stage, both at ∼1% each and lost in the following

sulfide aniline stage. In this state, the process was expected to

give ∼75−80% yield of product with ∼90−95% assay. There were

no additional chemical hazard concerns beyond those of the

individual stages. The change from DMF back to NMP meant that

exothermic activity was slightly more noticeable due to the lower

Cp of NMP, but less of an issue due to its much higher bp (202 °C).

Manufacture proceeded with five batches as planned in fully

telescoped manner on 25-kg (225 L) scale. However, the first

two batches were found to have unusually high amide 22

content as measured by LC at the cyano aniline stage (19−

35%). In theory, this should not have been a problem since the

amide 22 would be hydrolyzed to acid aniline 12 in the second

Lastly, there was agreement from the analytical department

to include the amide 22 and thioamide 23 levels in the

specification for cyano aniline (based on aniline base assay)

with thiol 18a and disulfide 19 set at <0.5% each. This meant

the expected yield based on the sum of 11, 22, and 23 was now

in the range 84−94%. Ironically, just as the process was fully

optimized and a robust isolation procedure finally developed,

the telescoped reaction with the acid aniline stage became

viable. However, the understanding gained on this stage was

not wasted. The acid aniline stage is treated briefly below

before discussing the telescoped process for the joint stages.

Acid Aniline Stage (12). It was quickly established that

cyano aniline 11 could be hydrolysed to acid aniline 12 with

either strong acid or base, either from isolated 11 or from crude

reaction liquors. In both cases, HPLC analysis confirmed that

the amide and thioamide impurities converted to acid aniline

(12), which simplified the analysis and proved that both

“impurities” could contribute to the overall yield of this

hydrolysis stage.

Initially, acidic conditions were preferred by the pilot plant to

avoid concerns regarding the use of strong, hot alkaline

mixtures. Concentrated HCl at 100 °C looked promising and

gave good conversion and yield of the product as its HCl salt,

but the form was poor and difficult to isolate, possibly due to

hygroscopicity. Neutralisation and extraction into an organic

solvent at various pHs to obtain the neutral acid aniline avoided

these issues but gave disappointing yields for an apparently

trivial reaction (60−65%) with moderate assay (80−85%).

Hydrolysis of 11 by concentrated H₂SO₄gave a new product

initially thought to be an electrophilic ring sulfonation product

but, in fact, was later identified as amide 22. However,

concentrations of aqueous H₂SO₄at 5, 10, and 15 M all gave

good conversions to the sulfate salt of 12 in good form. Direct

use of the sulfate salt in the following stage (sulfide aniline, 13)

failed completely, however, even with the addition of extra base

to neutralize the acidic counterion in the coupling step. In

addition, both HCl- and H₂SO₄-mediated hydrolyses suffered

from incomplete reaction conversions and gave products with

low assays. Parallel lab work showed that hot alkaline processes

appeared to hydrolyse 11 both faster and more cleanly, leaving

less residual amide 22.

Fortunately, the Hastelloy reactor planned for the

thiocyanate stage became available in the pilot plant at this

time, and so work switched to developing the alkaline process

in earnest. Ethanolic NaOH worked well, as expected, giving

full conversion after 4 h with impurities at the same level

as those from the acidic processes, but methanol was used

instead because the form of the isolated solid was improved.

Other bases were also tried (LiOH and KOH) but offered no

advantage over the standard NaOH. The neutralization process

at the end of reaction was not straightforward, but yields of

80−85% could be obtained, corrected for a moderate assay of

80%. Overall, the base-mediated process was faster than the

acid-mediated ones, giving full conversion in less time with

equivalent quality and in better form as the neutral species of

12 rather than as the salts. There were no chemical hazard

issues of concern, and the process was ready to be scaled up as

a stand-alone process when the decision to telescope it with the

preceding cyano aniline stage was made.

452

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

step, but in practice it was found that much of 22 was lost to the

aqueous phase in the intermediate extractions, resulting in low

isolated yields of 12 (49−64%). An extended extraction procedure

was introduced to aid partition of 22 into the MTBE phase, but this

gave only marginal improvement. A laboratory user trial with plant-

supplied NMP and Na₂S indicated that this combination resulted in

the high level of 22. For the last two batches, the reaction

temperature was reduced to 45 °C for the cyano aniline stage and

the reaction time lengthened to 4 h which resulted in only 2−3%

amide before the extraction and an 86% yield of cyano aniline.

In total, 135 kg of acid aniline 12 were produced in five

batches in an average yield of 71%. Notably, however, the

isolated yield for the last two batches was 87% compared to

54−66% for the first three in which high levels of amide 22 had

resulted in high losses. Irrespective of the yield, all batches were of

Oxalyl and thionyl chlorides were briefly reinvestigated now

that higher-quality acid aniline 12 was available, but they still did

not work. Otherwise, only CDI was investigated, and this looked

promising from the first reaction in acetonitrile at 50 °C. The

CDI (1.05 equiv) was added as a solid in one portion which had

no exothermicity but did evolve some CO₂. Aqueous dimethyl-

amine was then added which reacted instantaneously (adiabatic

temperature change of 10 K). The product was isolated after an

aqueous drown-out (also moderately exothermic) and cooling to

0 °C, which gave good yield (>90%), quality (>95%), and

physical form. To avoid the minor inconvenience of solid

charging, CDI was dissolved in acetonitrile at 50 °C and added to

a slurry of 12 at 50 °C, whereupon it reacted instantly. This also

allowed the heat of reaction to be controlled better. This simple

process was prepared for manufacture.

1

excellent quality by H NMR assay (98−100%) and purity by

However, it was noted that CDI was prone to lose activity on

reaction with moisture, which meant the stoichiometry could

not be too tight. Furthermore, the CDI had to be assayed by

¹H NMR spectroscopy to determine its strength prior to charging

on the plant. A troublesome impurity had been observed, especially

when the CDI charge was higher, which was difficult to remove in

the following stages. We planned to scale the batch size to whole

units of the supplier’s pack size of CDI to minimize the possibility

of exposure to water. CDI could crystallise from acetonitrile at the

planned concentration but stayed in solution in warm aceto-

nitrile for a long period (24 h) which led to a successful reaction.

Although the reaction was heterogeneous at the start, it appeared to

perform well, and there were no special issues with chemical

hazards, both exothermic additions being well within plant

capability and CO₂generation being addition-rate controlled.

An FED indicated that the charge of CDI should be increased

from 1.1 to 1.2 equiv, and that of dimethylamine from 1.5 to 1.75

equiv. The acetonitrile volume for dissolving the CDI was increased

to 6 relative volumes, and this ensured the CDI stayed in solution

even at room temperature. A reverse drown-out had also been

investigated, but the normal drown-out was found to be better.

Lastly, just before the pilot-plant manufacture, a 5 L laboratory

preparation provided a final confirmation that the process was ready.

Only one batch was performed on plant scale (30 kg, 330 L)

before the project was put on hold. Filtration, washing, and

drying times were much slower than expected during isolation

on the pressure filter i.e. days instead of hours. Pilot-plant vessel

agitation was subsequently thought to be significantly better

(using several methods) than on laboratory scale, and analysis

of the solid indicated that some particle attrition had occurred.

In addition, the very poor form of the damp paste damaged the

agitator drive in the pressure filter during the prolonged drying

regime. The batch had to be discharged as a 12% water-wet

paste (∼34 kg), and the assay was low at 86%. It also had two

impurities at 3.6 and 1.2%, which were identified by LCMS as

di- and tri-adducts respectively, the diadduct being tentatively

identified as structure 24a. These impurities had been noted in

the process before, but well below 1%.

HPLC (97−99%), demonstrating a robust purification procedure.

Overall, the telescoped process was a worthwhile improve-

ment over the individual stages, and compensated in some

measure for the linear nature of this route by avoiding one

isolation step, often the lengthiest part of processing. The

telescope gave a higher yield with similar quality, reduced plant

time, and operations and avoided form and analytical issues at the

cyano aniline stage. The solvent exchange distillation worked well,

and cleaning was not an issue in NMP. The only remaining issue

to be resolved was the poor performance of the initial plant

batches, which had been attributed to the source and quality of the

plant NMP and Na₂S, but further investigations were not needed.

Sulfide Aniline Stage (13). As with the preceding two

stages, this one also had no counterpart in the campaign 2 route,

and looked simple on paper. Both reagents used in campaign 2

route (oxalyl chloride or thionyl chloride) were tried to activate

12 as its acid chloride to couple with dimethylamine, but

neither conditions would work. This did not appear to be due

to polymerization of the bifunctional acid aniline 12.

Changing to a standard peptide coupling reagent such as EDCI

gave the desired reaction. When the level of hydroxybenzotriazole

(HOBt) was low, however (10 mol %), the reaction was

incomplete, and an impurity formed, assumed to be one of the

N-acyl ureas.¹²At stoichiometric levels, the reaction conversion was

complete, but 20% w/w HOBt contaminated the product which

was not easily removed by recrystallization. Attempts to use the

sulfate salt of 12 directly in the reaction, with additional dim-

ethylamine or Et₃N to neutralize the acidic counterion, did not

work, even with a preneutralisation step. However, the neutralized

salt of 12 could be extracted and taken directly into the EDCI

reaction successfully. The sodium salt of 12 also coupled directly

with aqueous dimethylamine mediated by EDCI after some study.

Aqueous dimethylamine was next used instead of ethanolic

dimethylamine, as had been used in the campaign 2 bromo-

benzamide stage (6 to 7) which had given rise to an ethyl ester

impurity. Aqueous dimethylamine avoided this problem and

gave a faster reaction and higher yield, but the addition became

more exothermic and required a longer addition time. A

modest increase in the charge of EDCI to 1.2 equiv achieved

complete conversion after 2 h at 20 °C, and after further

addition of water, a 95% yield of product with 95% assay could

be obtained directly in fair form. The reaction mixture was

heterogeneous, but this did not appear to be a concern.

Residual HOBt could be washed out with a citrate wash. The

process was prepared for plant manufacture with EDCI, but at a

late stage, there was a desire to find an alternative coupling reagent,

given the relatively high cost of EDCI (a noncontributory reagent).

453

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

The proposed cause of their formation was from an

overcharge of CDI, which was traced back to an error in the

method used to determine its assay. This figure had been used

in the FED which had set the level of CDI required for plant

scale, but had effectively led to an overcharge. A laboratory

experiment with a significant overcharge (1.5 equiv compared

to planned 1.2) gave a 98% yield of product but containing 15%

of 24a and ∼1% of the triadduct.

Investigations to rescue the first batch by treating a portion

with a large excess of aqueous dimethylamine removed or

destroyed both impurities but also gave another significant

impurity. The solvent combination used by the contractor in

campaign 2 to remove copper residues from their sulfide aniline

also proved of no help in removing 24a. Postponing the prob-

lem and taking a sample through to the final stages did not

reduce these impurities sufficiently in the final crystallization.

Furthermore, LCMS analysis in the penultimate oxidation

process indicated they had transmuted into the analogous

sulfoxide (24b) and sulfone (24c) diadducts.

and GC could not monitor the heavier components. Use of the

unprotected chiral acid 2 would therefore avoid many of these

analytical, chemistry, and storage issues and remove one step

from the reaction sequence.

Direct coupling of chiral acid 2 with sulfide aniline 13 using

EDCI under standard conditions gave no reaction. Oxalyl

chloride appeared to work well, but any slight excess reacted

with 13 to generate an oxalate-bridged diadduct (25) which

could not be removed and which tracked through in subsequent

stages. This had also been identified as a troublesome impurity in

the campaign 2a/b manufactures when forming 17 from 15 and

16, and was the reason both manufactures had performed poorly.

The formation of 25 could be prevented by the use of only one

equivalent of oxalyl chloride, but under these conditions the acid

chloride formation (14) was slow to reach completion.

The project was canceled before a decision on the following

batch(es) was required. Clearly the charge of CDI could be set

correctly once the problems with the assay were understood.

The value of the FED study was that the results could be

readjusted by taking account of the known error, and no experi-

mental repetition was required. In the short term, batch 1 could

be quarantined whilst a suitable crystallization solvent was

identified, since the split campaigns 3a and 3b meant that the

full campaign delivery did not have to be made at the first pass.

Chiral Amide Stage (15). Throughout the previous

campaigns, there had been a desire to add the unprotected

chiral acid unit (2) via its acid chloride (14) instead of through

the TMS-protected derivative (16). The double TMS-

protection with bis-TMS urea had proven capricious at

Discovery chemistry scale, and the intermediate TMS-protected

acid chloride (16) was volatile and difficult to isolate without

high losses. Furthermore, the TMS-protected sulfide aniline

product (17) was also difficult to isolate and had to be

protected from moisture and heat; a sample held for 5 days in

dichloromethane, for example, lost most of the TMS group

over this period, although the subsequent oxidation worked

well and easily removed the remainder of the TMS group. This

would not be acceptable as a long-term manufacturing method

and in any case would require the solution assay of 16 to be

determined for reaction with sulfide aniline. Furthermore, no

single analytical method could monitor the reaction. HPLC

could not analyze the non-UV-active derivatives (14 and 16),

Use of standard thionyl chloride coupling conditions in

acetonitrile with triethylamine gave a 66% yield of 15 after an

aqueous drown-out on the first attempt. This proved the

viability of using the unprotected chiral acid 2 but the con-

version was incomplete. Use of more thionyl chloride or triethyl-

amine, a higher temperature or a longer reaction time, and

purging with nitrogen beforehand all gave no improvement.

An alternative approach to the formation of α-hydroxy

amides had been used on another Zeneca project (ZD6169)¹¹

taking advantage of neighbouring group participation of the

α-hydroxyl group. In such cases, a weakly nucleophilic amine

(aniline) can be activated using thionyl chloride to give the

N-sulfinyl amine (26)¹²which is then reacted directly with an

α-hydroxyacid (2) to give the desired α-hydroxy amide (i.e.,

chiral amide 15). The proposed mechanism is based on the

report of Shim and Kim (Scheme 5).¹³They proposed inter-

molecular protonation by the carboxylate group and then an

intramolecular attack by the α-hydroxyl group which, after

further protonation and regeneration of the amine (13),

effectively activates the cyclic intermediate (28) as a mixed

anhydride to attack by the free amine (13). They also proposed

reversible steps for all but the last, and claimed that addition of

either stronger acids or bases hindered the reaction by suppressing

Scheme 5. Mechanism of α-hydroxy amide formation

454

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

the forward reaction to zwitterionic intermediate 27. This last

point is contested by Chidambaram et al. who nevertheless

used this approach very successfully in nonracemic systems.¹⁴

Our experience was that the reaction proceeded well in the

presence of residual thionyl chloride used in excess for the

N-sulfinyl amine formation.

This had started with 16 equiv of H₂O₂in neat glacial acetic

acid at 90 °C, which might have been acceptable on small scale,

but would cause issues on the plant. The contractor wisely

reduced this to 2.0 equiv of H₂O₂heated at 60 or 80 °C, which

gave complete conversion at an acceptable rate. The inter-

mediate sulfoxides 29 were confirmed as intermediates by

LCMS (as co-eluting diastereomers). On 40-g scale however,

there was an exothermic temperature rise to 102 °C from 80 °C

and significant gassing, assumed to be from decomposition of

H₂O₂.

Following the ZD6169 project, we adopted the “all-in”

approach without isolation of the N-sulfinyl amine (26) to form

the chiral amide 15. To determine whether the reaction pro-

ceeded by N-sulfinyl amine 26 or the more obvious acid

chloride 14, both chiral acid 2 and sulfide aniline 13 were

prereacted with thionyl chloride for 1 h under the standard

reaction conditions before their reaction partners (13 or 2,

respectively) were added to each mixture. The acid chloride-

mediated process via 14 proceeded with a fast initial reaction

which faded away to an incomplete reaction. The N-sulfinyl

amine-mediated process via 26 progressed steadily to

completion over 18−24 h. It is proposed that the irreversible

loss of SO₂from the cyclic reaction intermediate 27 helps to

drive the reaction completion.¹²Since sulfide aniline 13 cannot

be reformed from intermediate 26, it continues to react with

the chiral acid 2 until it is all consumed. Attempts to confirm

the structure of the (in our case) nonisolated N-sulfinyl amine

26 using IR techniques were inconclusive, and the project was

canceled before a more thorough investigation of this intriguing

mechanism could take place.

Initial process development indicated a slight excess of

thionyl chloride (1.15 equiv) over triethylamine (1.05 equiv)

was better for the reaction, which could be conducted in

acetonitrile. After formation of the N-sulfinyl amine inter-

mediate over 2 h at 40 °C, an excess (1.15 equiv) of chiral acid

2 was added and the reaction stirred at 20 °C overnight. The

workup proved troublesome, in that the product 15 tended to

oil out. Normal and reverse drown-outs into water, brine and

bicarbonate solutions were all tried with mixed success; a warm

drown-out into a brine/bicarbonate solution was the preferred

option and gave the best physical form. Despite these problems,

reliable yields of 85−90% of product with 96−99% assay could

be obtained reproducibly.

A series of tube-scale experiments were performed to

investigate various parameters in both formic and acetic acids.

The reaction was found to be generally faster in formic acid

than acetic acid, not much affected by concentration, and not

requiring more than 3.0 equiv to go to completion. Slow

addition of H₂O₂also worked well and showed that the risk of

accumulation could be avoided. Laboratory-scale preparations

were performed in acetic acid at 80 °C and in formic acid at

60 °C (which was quicker and safer), and the results were

passed to the contractor for their campaign 2 studies.

Further studies within AstraZeneca in formic acid showed

that the first equivalent of H₂O₂gave complete conversion to

the sulfoxides 29 after 30 min; addition of a further equivalent

then gave the sulfone (i.e., crude 1) after 50 min, suggesting a

long reaction time was not needed. In practice, a small excess of

H₂O₂was used, which could be quenched with sodium meta-

bisulfite solution. A new impurity was detected and identified as

the benzoic acid derivative (30). Although the product some-

times crystallised on cooling and could be isolated and washed

with water in ∼90% crude yield, generally the workup proved

problematic because the product oiled out or had poor form.

Eventually, an 80% yield of product with 96% assay could be

isolated, which retained 1.0% of the sulfoxides 29 and 0.7% of

the benzoic acid 30.

A use-test of the damp sulfide aniline 13 from the first plant

batch failed as expected due to the high water content (12%).

When the sample of 13 was dried azeotropically, reaction

conversion was almost complete at 98% and an oven-dried

sample gave complete conversion. However, in both cases

the physical form of the product chiral amide 15 was notably

poorer than when prepared from laboratory samples of 13,

and the level of diadduct 24 was not reduced. This simply

confirmed that sulfide aniline batches would need to be

dry before further reaction and with a low level of di- and

triadducts.

Attempting an extraction procedure instead of direct iso-

lation was also problematic. The solubility of 1 in low polarity

solvents such as dichloromethane or toluene was low, resulting

in high volumes; more polar solvents were miscible with the

formic acid, and intermediate ones tended to extract some

formic acid, which resulted in higher levels of 30 if concen-

tration was attempted without neutralization of the acidic

residues. Attempts to distill off the formic acid before using any

of these extraction procedures had the same effect. Eventually,

MIBK was identified as a suitable midpolarity solvent, from

which most of the extracted formic acid could be removed by

an aqueous back-wash, followed by neutralisation of the

remainder with KHCO₃. Addition of a hydrocarbon solvent

There were no major chemical hazard concerns with the

process. The addition of thionyl chloride was exothermic and

SO₂was evolved when adding the chiral acid, but both

parameters were addition-rate controlled and well managed by

slow additions. There were some remaining minor concerns

with compatibility of thionyl chloride and MoC, and physical

form and isolation issues, but in all other respects, the chiral

amide stage was essentially ready for plant manufacture when

the project was halted. The batch size would have been on

25 kg (350 L).

Crude Stage (1). The oxidation process of chiral amide to

give “crude” bulk drug 1 was inherited from the campaign 2

process, itself adopted from the Discovery chemistry process.

455

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

such as octane helped to crystallise the product 1 in good form

and up to 90% yield with >98% purity. The benzoic acid 30 was

largely removed under these conditions, and the sulfoxides 29

were not detected in any case now that the H₂O₂charge had

been increased to 2.3 equiv, which was important because these

were not readily removed in the recrystallization stage.

only five for the linear route and seven for the convergent

route, which no doubt contributed to this result, in addition to

the improved yields.

The benzonitrile route was felt to be a viable long-term

manufacturing route and a significant improvement over the

more convergent benzamide route for many other reasons. The

improved thiocyanate (10) process was inherently safer and

worked better than before. Use of benzonitrile 8 allowed for a

facile S_NAr reaction, and avoided the problematic Ullmann

reaction between 7c and 10. The more linear nature of the

route was mitigated by the simple chemistry of the cyano

aniline/acid aniline telescope which proved reliable after initial

modification giving excellent quality product (12). The root

cause of the problem with the first batch of sulfide aniline (13)

was well understood and should have been controlled by

adjusting the CDI charge. The chiral amide (15) stage was

improved over the analogous sulfide anilide (17) stage by an

understanding of the N-sulfinyl amine mediated reaction with

an α-hydroxy acid, which avoided unnecessary protection of the

chiral acid 2 and the resulting losses. The remaining two

processes for the pilot-plant manufacture were also well

developed and had good scale-up precedent (chiral amide

type process on ZD6169 and crude-stage oxidation in campaign

2 and/or Sibenadet), and two options were available for the

final stage oxidation. Hazard issues for all stages were well

understood and controlled.

Although the process now worked well, the pilot plant had

some residual concerns about chemical hazards. The formic

acid/H₂O₂process was preferred due to giving the cleanest

reaction profile, especially when there were issues with

removing impurities in the final recrystallization process, but

the chemical hazards group were concerned that formic acid/

H₂O₂mixtures were known to be potentially explosive, albeit at

high concentrations.⁸The reagent controlled addition of dilute

H₂O₂avoided these conditions, but even so, the chemical

hazard assessment indicated an adiabatic temperature rise of

57 K. Therefore it was agreed to investigate the use of base-

mediated H₂O₂in acetonitrile¹⁵which had been developed for

plant scale on Sibenadet hydrochloride.¹⁶Using the standard

3.6 equiv of H₂O₂for this process with K₂CO₃at 45 °C in

acetonitrile gave an initial reaction with 63% yield of product

with 98% assay. The equivalent chemical hazard assessment on

the same scale indicated an adiabatic temperature rise of only

28 K. The sulfoxides 29 were present at slightly higher levels,

but not significantly so. However, as with the formic acid-

mediated process, product 1 oiled out; some improvements

were hastily made in subsequent preparations.

Both processes were then subjected to a Kepner−Tregoe

decision making process to compare them,¹⁷which confirmed

that there was little to choose between the two processes when

taking all factors into account. However, the Sibenadet-derived

process was judged to be marginally safer on scale (although

neither process was judged to be unsafe) and since this was the

preference of the plant personnel, this process was scheduled

for manufacture.

Use-tests of plant-derived sulfide aniline 13 were conducted

for both processes to determine if they could cope with the

high diadduct 24a impurity levels. In both cases, the diadduct

was reduced to <0.05%, but a new impurity was detected at

1.5%, unsurprisingly determined by LCMS as the fully oxidized

diadduct sulfone 24c; the intermediate diadduct sulfoxides 24b

were present at 0.2−0.3% in the final (pure) stage which was

acceptable. Neither process was trouble-free when performed

on the plant-derived materials, and simply confirmed that the

following processes could not cope with the diadduct 24a at

3.6%; the problems with the sulfide aniline stage would there-

fore have to be resolved at the that stage. Pilot-plant manu-

facture of AZD7545 was halted in April 2002 after the first

batch of 13 had been prepared, and development of AZD7545

was terminated in July 2002 with no further work conducted.

We are confident the benzonitrile route would have become

the long-term manufacturing route, probably with acid aniline

12 or sulfide aniline 13 as registered starting material, along

with chiral acid 2.¹⁸

EXPERIMENTAL SECTION

■

General Procedures. Reaction mixtures and products were

analysed by reverse phase HPLC on Hewlett-Packard 1050 or

1100 instruments according to the following conditions:

column, Genesis C18 100 mm × 4.6 mm i.d., particle size

3 μm; eluent A, 99.9% purified water with 0.1% v/v trifluoro-

acetic acid; eluent B, 90% acetonitrile and 9.9% purified water

with 0.1% v/v trifluoroacetic acid; flow rate 0.60 mL/min.;

wavelength 215 nm; temperature 40 °C; injection volume 5 μL;

at t = 0 min, 20% eluent B; at t = 15 min, 85% eluent B; at t =

16 min, 85% eluent B; 4 min post time. H and ¹³C NMR

1

spectra were recorded on a Varian Inova 400 spectrometer at

400 and 100.6 MHz, respectively, with chemical shifts given in

ppm relative to TMS at δ = 0. Electrospray (ES⁺) mass spectra

were determined on a Micromass LCT with time-of-flight.

Analytical TLC was carried out on commercially prepared

plates coated with 0.25 mm of self-indicating Merck Kieselgel

60 F₂₅₄and visualised by UV light at 254 nm. Preparative scale

silica gel flash chromatography was carried out by standard

procedures using Merck Kieselgel 60 (230−400 mesh). Product

1

CONCLUSIONS

purity was determined in most cases by both H NMR assay

■

The decision to choose the linear benzonitrile route (campaign 3)

over the convergent benzamide route (campaign 2) was readily

justified by their respective overall yields of 45.5% and 22.5%

for the longest linear sequence from thiocyanate 9. The benzo-

nitrile route may appear to have benefitted from more develop-

ment, but the benzamide route was effectively as good as could

be achieved after two or three iterations on each stage, and

further scale-up was thought to be impractical. Furthermore,

these figures do not account for the fact that all the steps in the

linear route are counted, whereas two steps in the convergent

route are not. Indeed, the step-count of actual isolations was

against an internal standard of either maleic acid or tetra-

chloronitrobenzene, and by HPLC area % standardized against

a sample of the product of known concentration.

In the preparations that follow, the largest scale plant

processes have been given for those stages manufactured at

campaign 3. For stages which did not reach manufacture, or

where laboratory scale alternatives have been discussed, details

are supplied for the finalized laboratory processes on which the

campaign 3 plant processes would have been scaled.

Plant-Scale Preparation of AZD7545 Thiocyanate (10).

[Safety note #1: unlike the initial process for this stage

456

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

reported previously,⁵this procedure has been safely scaled up

to 530 L (41 kg based on 9). Even so, there are still chemical

hazards of note, including a significant interaction between

bromine and alcohols as noted in Bretherick.⁸Although this

procedure largely mitigates against this issue, it remains

important that all chemical hazards are fully considered, especially

if any changes are made to this process. See the discussion in

the text for a full presentation of the potential concerns.

Safety note #2: late in the project, AZD7545 thiocyanate

was identified as a potential skin-sensitizing agent, which

seemed to be supported by experience during manufacture.

This was not further investigated before the project was

canceled but should be borne in mind if conducting future

preparations].

133.22, 132.61, 121.53, 116.71, 115.54; MS (ES⁺) 317/319/

321 (M + H⁺, 100%, 9:6:1).

Laboratory-Scale Preparation of AZD7545 Cyano

Aniline (11) (single process). Sodium sulfide (technical

grade reagent, 19.3 g @ 60%, 11.6 g @ 100%, 148 mmol,

1.80 equiv) and DMF (61 mL) were charged to a suitably

serviced reaction vessel A, and the mixture was stirred and

heated to 60 °C. Some sodium sulfide remained out of solution

as a yellow solid; the liquid phase of the mixture became dark

green/blue. AZD7545 thiocyanate (10) (15.2 g @ 100%, 82.2

mmol, 1.00 equiv) and DMF (53 mL) were charged to a

second vessel, B, and stirred to form a dark-yellow solution. (If

the thiocyanate contains a high level of inorganic impurities,

then some material remains out of solution at this point. These

fines can be removed by passing through a line filter in the

following transfer, but this is not necessary for standard quality

thiocyanate.) This solution was added to vessel A over 10 min

(caution: exothermic addition). The colour of the batch

changed from green/blue to orange. Vessel B was rinsed with a

line-wash of DMF (4.0 mL) and added into vessel A. The

reaction mixture was stirred at 60 °C for 30 min (but up to 4 h

is also acceptable and a longer hold period is unlikely to be

detrimental). 4-Fluorobenzonitrile (8) (10.5 g, 86.4 mmol,

1.05 equiv) and DMF (15 mL) were charged to a third vessel,

C, and stirred to form a colourless solution. This solution was

added to vessel A over 10 min (caution: exothermic addition;

an exotherm of 4 K is observed on this scale). Vessel C was

rinsed with a line-wash of DMF (4.0 mL) and added into vessel A.

The batch was stirred at 60 °C for 3 h, during which time the

batch turned paler in colour and a yellow solid formed. Water

(380 mL) was added to a fourth vessel D and heated to 60 °C.

The reaction mixture was added to vessel D over 15 min,

during which time the batch became cloudy (caution: mode-

rate exothermic addition). Vessel A was rinsed with a line-wash

of DMF (8.0 mL) and added into vessel D. The batch was

cooled evenly to 20 °C at a rate of 0.5 K/min and then stirred

at 20 °C for at least 1 h. A yellow solid precipitated during the

cooling period and hold time. The solid was isolated by vacuum

filtration, slurry-washed twice with water (76 mL each), and

dried in a vacuum oven at 50 °C to yield the title compound as

a yellow solid (19.4 g @ 87.1% assay, not including related

impurities 22 and 23; absolute yield of 11 79% yield corrected

for assay). HPLC (t_R12.8 min, assay 87.1%); ¹H NMR

(400 MHz, d₆-DMSO) δ 7.69 (2H, d, J = 8.7 Hz), 7.43 (1H, d,

J = 2.1 Hz), 7.23 (1H, dd, J = 8.5, 2.1 Hz), 7.13 (2H, d, J = 8.7 Hz),

6.91 (1H, d, J = 8.5 Hz), 5.96 (2H, s); ¹³C NMR (100.6 MHz,

d₆-DMSO) δ 147.2, 146.7, 136.0, 135.5, 132.6 (2C), 125.4

(2C), 118.8, 117.4, 116.3, 113.1, and 107.0; MS (ES⁺) 280/282

(M + H⁺). Safety note: the reaction liquors and aqueous washes

will contain NaCN and may also contain Na₂S. Both will present a

serious toxic hazard if the liquors become acidified. Na₂S also has a

strong smell. In the laboratory, the aqueous waste was treated with

bleach before being sent for disposal; on plant scale, more

substantial waste disposal measures will be required.

Sodium thiocyanate (49.1 kg, 607 mol, 2.0 equiv) and

methanol (270 L) were charged to a suitably serviced reaction

vessel, A, and cooled to 2 °C. 2-Chloroaniline (9) (38.7 kg,

303 mol) was added with stirring at 100 rpm over 10 min

(caution: moderate exothermic addition). The addition vessel

was washed with methanol (116 L) and added into vessel A.

Acetonitrile (97 L) was added to a second vessel, B, and cooled

to 2 °C. Bromine (48.5 kg 303 mol, 1.0 equiv) was added

slowly from a bomb to vessel B with stirring at 100 rpm over

30 min, maintaining a temperature of 0−5 °C (caution: moderate

exothermic addition). The line was washed with acetonitrile

(20 L) and added into vessel B. Both reaction mixtures were

recooled to 2 °C, and the bromine/acetonitrile solution from

vessel B was added slowly into the aniline/methanol solution

over 2.5 h with stirring at 100 rpm, maintaining the

temperature at 3 °C (caution: exothermic addition). Vessel B

was washed with acetonitrile (20 L) and added into vessel A.

The reaction mixture was stirred at 2 °C for 15−19 h. Sodium

bicarbonate (53.5 kg, 637 mol, 2.1 equiv) and water (830 L)

were added to a third vessel C with stirring at 10 rpm and

cooled to 10 °C. The reaction mixture from vessel A was added

into vessel C with stirring at a rate sufficient to maintain the

temperature below 20 °C and commensurate with the gas

disengagement capacity of vessel C (30 min on this scale)

(caution: a vigorous effervescence of CO₂and an exothermic

addition is observed). The product crystallises as a white or

pale-yellow precipitate during this addition. Vessel A was

washed with methanol (40 L) and added into vessel C. The

reaction mixture was cooled to 0−5 °C for 30 min before

isolation by filtration. (The reaction mixture can be left for 18 h

at 0−5 °C before being isolated but slow decomposition of

product 10 is observed with some loss of product quality). The

reaction mixture was transferred to a filter-dryer in several

portions using positive N₂pressure to deliquor the product

cake. The product was slurry-washed twice with water (155 L

each) on the filter-dryer, deliquored fully and dried using

positive N₂pressure at 40 °C to yield the title compound as an

off-white to pale yellow solid (46.7 kg @ 96% assay; overall

1

yield 80%). HPLC (t_R9.3 min, quality 94.6%); H NMR

(400 MHz, d₆-DMSO) δ 7.57 (1H, d, J = 2.3 Hz), 7.35 (1H,

Data on Related Impurities Amide 22 and Thioamide

23. Samples of AZD7545 cyano aniline may also contain signifi-

cant quantities of two impurities, the amide and thioamide,

both of which are hydrolysed in the following process to yield

the desired AZD7545 acid aniline. Data for AZD7545 Amide

dd, J = 8.5, 2.3 Hz), 6.86 (1H, d, J = 8.5 Hz), 6.04 (2H, s); ¹³

C

NMR (100.6 MHz, d₆-DMSO) δ 147.25, 133.81, 133.17,

117.29, 116.12, 112.68, 106.54; MS (ES⁺) 185/187 (M + H,

30%, 3:1), 226/228 (M + MeCNH⁺, 45%, 3:1), 283/285

(100%, 3:1). Data for AZD7545 Disulfide (19): HPLC (t_R

14.9 min); ¹H NMR (400 MHz, d₆-DMSO) δ 7.21 (2H, d, J =

2.0 Hz), 7.09 (2H, dd, J = 8.4, 2.0 Hz), 6.76 (2H, d, J = 8.4 Hz),

5.81 (4H, bs); ¹³C NMR (100.6 MHz, d₆-DMSO) δ 145.98,

1

(22): HPLC (t_R8.8 min); H NMR (400 MHz, d₆-DMSO) δ

7.89 (1H, br s), 7.76 (2H, d, J = 8.4 Hz), 7.39 (1H, d, J =

2.0 Hz), 7.30 (1H, br s), 7.21 (1H, dd, J = 8.4, 2.0 Hz), 7.07

(2H, d, J = 8.0 Hz), 6.88 (1H, d, J = 8.4 Hz), 5.86 (2H, br s);

457

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

¹³C NMR (100.6 MHz, d₆-DMSO) δ 167.27, 146.19, 143.33,

135.52, 135.07, 131.07, 128.19 (2C), 125.26 (2C), 117.24,

116.15 and 115.07; MS (ES⁺) 279/281 (M + H⁺). Data for

required. Vessel A was then rinsed with MTBE (250 L) which

was also charged into vessel D. The reaction mixture was stirred

vigorously for 10 min, and the resulting layers were allowed to

separate. The lower aqueous layer was removed and the upper

MTBE layer containing the intermediate crude cyano aniline

washed twice more with water (150 L each) before being

transferred back to vessel A. Vessel D was rinsed with absolute

ethanol (150 L) and added to the yellow MTBE solution in

vessel A. A second wash of absolute ethanol (150 L) was rinsed

through vessel D and added into vessel A. The reaction mixture

was distilled until a head temperature of 75−80 °C was

achieved to remove the MTBE solvent. The mass of cyano

aniline (11) extract retained was 175 kg with ∼275 L distillate

collected. The reaction mixture was cooled back to 50 °C, and

sodium hydroxide liquor (33.7 kg @ 32% w/w, 270 mol,

2.0 equiv) was added, followed by a line-wash of water (25 L).

The stirred reaction mixture was heated to 80 °C, held for 18−

22 h, and then cooled to 50 °C. (NB: It is important to achieve

reflux to remove gaseous byproduct.) Concentrated hydro-

chloric acid (37.0 kg @ 36% w/w, 365 mol, 2.7 equiv) was

added over 5 min with good agitation (caution: moderate

exothermic addition), followed by a line-wash of water (25 L).

During this addition, the product crystallized from solution.

The reaction mixture was cooled to 0 °C at 0.5 K/min and

transferred to an oyster filter for isolation. The product cake

was washed by displacement twice with water (60 L each) in

the oyster filter, deliquored fully, and dried at 50 °C to yield the

title compound as a pale-brown solid (33.4 kg @ 99% assay;

overall yield 88%). HPLC quality 98.8%. Other data for

AZD7545 acid aniline are as noted above.

1

AZD7545 Thioamide (23): HPLC (t_R11.4 min); H NMR

(400 MHz, d₆-DMSO) δ 9.77 (1H, br s), 9.40 (1H, br s), 7.81

(2H, d, J = 8.4 Hz), 7.39 (1H, d, J = 2.0 Hz), 7.21 (1H, dd, J =

8.4, 2.0 Hz), 7.03 (2H, d, J = 8.4 Hz), 6.88 (1H, d, J = 8.8 Hz),

5.87 (2H, s); ¹³C NMR (100.6 MHz, d₆-DMSO) δ 198.88,

146.22, 143.60, 136.09, 135.50, 135.05, 128.04, 124.88, 117.26,

116.16, and 114.97.

Laboratory Scale Preparation of AZD7545 Acid

Aniline (12) (single process). AZD7545 cyano aniline (11)

(5.0 g @ assumed 100% assay, 19.2 mmol) was dissolved in

absolute ethanol (50 mL) with stirring at 25 °C to form a

yellow solution. Sodium hydroxide liquor (2.6 mL @ 47% w/w,

45.1 mmol, 2.35 equiv) was added, followed by water (5.0 mL).

The stirred reaction mixture was heated to reflux at 80 °C for

22 h. (NB: It is important to achieve reflux to remove gaseous

byproduct.) The reaction mixture was cooled to 50 °C and

water (5.0 mL) added. Concentrated hydrochloric acid (4.3 mL

@ 32% w/w, 45.1 mmol) was added and the product crystal-

lized from solution (caution: moderate exothermic addition).

The reaction mixture was cooled to 0 °C and the solid product

isolated by vacuum filtration. The reactor vessel was rinsed with

water (12 mL) and the rinse wash passed through the product

cake by displacement. The water wash (12 mL) was repeated

and the damp solid dried in a vacuum oven at 50 °C to yield

the title compound as a pale yellow solid (5.3 g @ 92.3% assay;

1

overall yield 91%). HPLC (t_R9.8 min); H NMR (400 MHz,

d₆-DMSO) δ 12.81 (1H, br s), 7.79 (2H, d, J = 8.8 Hz), 7.38

(1H, d, J = 1.6 Hz), 7.20 (1H, dd, J = 8.4, 1.6 Hz), 7.06 (2H, d,

J = 8.4 Hz), 6.87 (1H, d, J = 8.0 Hz), 5.87 (2H, br s); ¹³C NMR

(100.6 MHz, d₆-DMSO) δ 168.87, 146.40, 145.82, 137.78,

135.32, 129.95 (2C), 127.30, 125.05 (2C), 117.30, 116.22, and

114.30; MS (ES⁺) 280/282 (M + H⁺).

Alternative Laboratory-Scale (EDCI) Preparation of

AZD7545 Sulfide Aniline (13). AZD7545 acid aniline (12)

(8.0 g, 28.6 mmol), hydroxybenzotriazole (4.1 g, 30.0 mmol,

1.05 equiv) and acetonitrile (50 mL) were charged to a suitably

serviced reaction vessel A and stirred at 20 °C to form a slurry.

EDCI.HCl (6.7 g, 34.3 mmol, 1.20 equiv) was dissolved in

acetonitrile/water (1:1, 19 mL total) in a second vessel, B, to

give a clear solution which was added to vessel A over 10 min at

20 °C. Vessel B was rinsed with a line-wash of acetonitrile

(1.0 mL) and added into vessel A. The reaction mixture was

stirred at 20 °C for 1.5 h, during which time a cream-coloured

precipitate formed. Aqueous dimethylamine solution (40% w/v,

9.7 mL, 76.4 mmol, 2.7 equiv) was added over 1 h (caution:

moderate exothermic addition) followed by a line-wash of

acetonitrile (1.0 mL) and the reaction mixture stirred for

15 min. Water (50 mL) was then added over 50 min (caution:

moderate exothermic addition) and stirred for a further 1 h,

during which time the product crystallised. The reaction

mixture was cooled to 2 °C and stirred at this temperature for

2 h. (Additional water may be added at this point if crystalliza-

tion is incomplete.) The product was isolated by vacuum

filtration, the slurry washed twice with water (40 mL each) and

dried in a vacuum oven at 50 °C to yield the title compound as

an off-white solid (7.9 g @ 96.7% assay; overall yield 93%).

Plant Scale Preparation of AZD7545 Acid Aniline (12)

(telescoped process from AZD7545 Thiocyanate 10).

Sodium sulfide (technical grade reagent, 15.8 kg @ 100%,

203 mol, 1.50 equiv) and NMP (100 L) were charged to a

suitably serviced reaction vessel, A, and the mixture was stirred

and heated to 45 °C. Some sodium sulfide remained out of

solution as a yellow solid; the liquid phase of the mixture

became dark coloured. AZD7545 thiocyanate (10) (25.0 kg @

100% 135 mol, 1.0 equiv) and NMP (87 L) were charged to a

second vessel, B, and stirred to form a dark-yellow solution with

a small quantity of fines. This solution was added to vessel A

through an in-line filter over 10 min (caution: exothermic

addition). Vessel B was rinsed with a line-wash of NMP (6 L)

and added into vessel A. The reaction mixture was stirred at

45 °C for 60 min (but up to 4 h is also acceptable, and a longer

hold period is unlikely to be detrimental), during which time

the colour became orange. 4-Fluorobenzonitrile (8) (18.1 kg,

149 mol, 1.1 equiv) and NMP (25 L) were charged to a third

vessel, C, and stirred to form a colourless solution. This solu-

tion was added to vessel A over 10 min (caution: an exothermic

addition is observed). Vessel C was rinsed with a line-wash of

NMP (6 L) and added into vessel A. The reaction mixture was

stirred at 60 °C for 4 h before being transferred to a fourth

vessel, D. During this time it became paler in colour, and a

yellow solid formed. Vessel A was rinsed with water (600 L)

which was charged into vessel D, and the resulting mixture

cooled to 20 °C for 60 min. The mixture can be held at both

temperatures for longer periods without detriment to quality if

1

HPLC (t_R10.5 min); H NMR (400 MHz, d₆-DMSO) δ 7.39

(1H, d, J = 1.8 Hz), 7.31 (2H, d, J = 8.4 Hz), 7.22 (1H, dd, J =

8.4, 1.8 Hz), 7.06 (2H, d, J = 8.4 Hz), 6.87 (1H, d, J = 8.4 Hz),

5.85 (2H, s), 2.93 (3H, br s), 2.90 (3H, br s); ¹³C NMR (100.6

MHz, d₆-DMSO) δ 169.54, 146.17, 141.01, 135.58, 135.12,

131.18, 127.91 (2C), 125.45 (2C), 117.23, 116.15, 115.23,

38.80 and 34.77; MS (ES⁺) 308/310 (M + H⁺).

Plant-Scale (CDI) Preparation of AZD7545 Sulfide

Aniline (13). AZD7545 acid aniline (12) (30.0 kg, 107.3 mol)

458

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

and acetonitrile (300 L) were charged to a suitably serviced

reaction vessel A and heated to 50 °C to form a mobile slurry.

CDI (21.3 kg @ assumed 100% assay, 131.3 mol, 1.22 equiv.)

and acetonitrile (150 L) were charged to a second vessel, B, and

heated to 50 °C to to form a clear solution. (NB: The assay of

CDI was incorrect; see the main discussion for explanation.)

The CDI solution was added to vessel A over 30 min at 50 °C

(caution: gas evolution), during which time a yellow solu-

tion formed. Vessel B was rinsed with a line-wash of acetonitrile

(60 L) and added to vessel A. The reaction mixture was stirred

at 50 °C for 75 min and then cooled to 25 °C. Aqueous

dimethylamine solution (14.6 kg @ 40% w/v, 129.5 mol, 1.20

equiv) was added smoothly over 15 min (caution: exothermic

addition). The line was washed with water (15 L) and added

into vessel A. The yellow colour disappeared, and a white

precipitate formed. The reaction mixture was stirred at 25 °C

for 30 min, then cooled to −2 °C. Water (480 L) was added

over 30 min (caution: moderate exothermic addition) and the

reaction mixture stirred for a further 30 min at 5 °C. The reaction

mixture was transferred to a filter-dryer in several portions using

positive N₂pressure to deliquor the product cake (this took 3 days

on plant scale). The product was washed by displacement twice

with water (160 L each) on the filter-dryer (this took 8−15 h on

plant scale), deliquored fully, and dried using positive N₂pressure

at 70 °C to yield the title compound as an off-white solid (34.4 kg

@ 86% assay; overall yield 89%). HPLC quality 94.4%. Other data

for AZD7545 sulfide aniline as noted above, allowing for lower

quality in this case.

Laboratory-Scale Preparation of AZD7545 Chiral

Amide (15). AZD7545 sulphide aniline (13) (20.0 g, 65.2

mmol) and acetonitrile (100 mL) were charged to a suitably

serviced reaction vessel A and stirred at 20 °C to give a mobile

slurry. Triethylamine (6.9 g, 9.6 mL, 68.4 mmol, 1.05 equiv)

was added directly to the vessel over 1 min, followed by a line-

wash of acetonitrile (40 mL). Thionyl chloride (9.0 g, 5.5 mL,

75.0 mmol, 1.15 equiv) was added over 1 h (on this scale by

syringe pump) keeping the temperature below 22 °C (caution:

exothermic addition), followed by a line-wash of acetonitrile

(40 mL). A dark solution resulted which was stirred at 20 °C

for 2 h (and not longer than 4 h). AZD7545 chiral acid (2)

(11.9 g, 75.0 mmol, 1.15 equiv) was dissolved in acetonitrile

(40 mL) in a second vessel, B. This solution was added to

vessel A over 1 h (to minimize gassing, the addition is not

exothermic), followed by a vessel rinse and line-wash of

acetonitrile (40 mL). The resulting dark reaction mixture was

stirred at 20 °C for ∼16 h (but up to 40 h has no detrimental

effect). Sodium chloride (57.6 g) and water (250 mL) were

charged to a third vessel, C, and heated to 50 °C to form a clear

solution. The reaction mixture from vessel A was added into

the solution in vessel C smoothly over 30 min to form two clear

phases. Vessel A was rinsed with acetonitrile (20 mL) which

was washed into vessel C, and the reaction mixture was stirred

at 50 °C for 30 min. The mixture was allowed to separate and

the lower aqueous phase removed. Fresh saturated brine

solution (120 mL) was added, followed by aqueous sodium

carbonate solution (1.0 M, typically ∼50 mL), added carefully

over 30 min until the pH was in the range 6.5−7.5. The mixture

was allowed to separate and the lower aqueous phase removed.

The temperature was adjusted to 42 °C and water (100 mL)

added over 1 h with stirring. The reaction mixture was stirred at

42 °C for 30 min, during which time the product started to

crystallise. Further water (100 mL) was added over 1 h and the

mixture stirred at 42 °C for an additional 30 min before cooling

to 20 °C over 1 h. The reaction mixture may be stirred over-

night at this temperature for convenience; a minimum of 8 h is

recommended for full product crystallization; longer periods

will not affect yields but may result in crystal attrition. The

product was isolated by filtration, washed by displacement with

water (100 mL), and dried in a vacuum oven at 60 °C to yield

the title compound as an off-white solid (26.8 g @ 98.2% assay;

overall yield 90%). HPLC (t_R12.1 min, quality 99%); ¹H NMR

(400 MHz, d₆-DMSO) δ 9.77 (1H, s), 8.04 (1H, dd, J = 8.8,

3.0 Hz), 7.89 (1H, s), 7.61 (1H, d, J = 2.2 Hz), 7.43 (1H, dd,

J = 8.8, 2.2 Hz), 7.40 (2H, d, J = 8.4 Hz), 7.32 (2H, d, J =

8.4 Hz), 2.96 (3H, s), 2.90 (3H, s), 1.61 (3H, s); ¹³C NMR

(100.6 MHz, d₆-DMSO) δ 169.31, 167.10, 136.53, 135.15,

133.61, 132.26, 131.69, 130.84, 129.22 (2C), 128.23 (2C),

125.78, 123.92, 74.80 (q, J = 27.6 Hz); MS (ES⁺) 447/449

(M + H⁺); (ES⁻) 446/448 (M − H⁻).

Laboratory-Scale Preparation of Crude AZD7545 (1)

(draft formic acid process finalized for pilot-plant

manufacture). AZD7545 chiral amide (15) (5.0 g, 10.6 mmol)

was dissolved in formic acid (25 mL) in a suitably serviced

reaction vessel and heated to 60 °C with stirring. Hydrogen

peroxide (12% w/w solution, 7.2 mL, 24.5 mmol, 2.3 equiv)

was added over 2 h (on this scale by syringe pump), main-

taining the temperature at or below 60 °C (caution: exo-

thermic addition). The reaction mixture was stirred at 60 °C

for a further 1 h, then cooled to 40 °C. Sodium metabisulfite

(0.90 g, 4.6 mmol, 0.43 equiv) as a solution in water (2.0 mL)

was added smoothly over 3 min and the mixture stirred for

10 min before confirming residual peroxide residues had been

quenched. MIBK (40 mL) was added over 3 min, maintaining

the temperature at 40 °C, followed by water (50 mL), also

added over 5 min, maintaining the temperature at 40 °C. The

reaction mixture was stirred vigorously for 10 min, and then the

phases were allowed to separate, and the lower aqueous 3was

phase removed. Fresh water (50 mL) was added over 10 min,

maintaining the temperature at 40 °C (so as not to precipitate

the product prematurely). The reaction mixture was stirred

vigorously for 10 min, and then the phases were allowed to

separate, and the lower aqueous phase was removed. Fresh

water (10 mL) was added over 5 min, maintaining the

temperature at 40 °C. Potassium carbonate (15.0 g, 14.9 mmol)

as a solution in water (60 mL) was added carefully with good

agitation (caution: gas evolution) to achieve pH 7. The phases

were allowed to separate, and the lower aqueous phase was

removed. Fresh water (30 mL) was added over 5 min,

maintaining the temperature at 40 °C (so as not to precipitate

the product prematurely). The reaction mixture was stirred

vigorously for 10 min, then the phases were allowed to separate,

and the lower aqueous phase was removed. The MIBK phase

was heated to 45 °C and octane (40 mL) added smoothly over

60 min. The reaction mixture was then cooled to 20 °C and

stirred overnight at this temperature, during which time the

product crystallised. The product was isolated by filtration,

washed by displacement with octane (10 mL) and then with

petroleum ether (10 mL), and dried in a vacuum oven at 50−

60 °C to yield the title compound as a white solid (4.4 g @

1

100% assay; overall yield 87%). HPLC (t_R9.8 min); H NMR

(400 MHz, d₆-DMSO) δ 9.94 (1H, s), 8.36 (1H, d, J = 8.7 Hz),

8.21 (1H, d, J = 2.1 Hz), 8.07 (2H, d, J = 8.2 Hz), 8.03 (1H, dd,

J = 8.7, 2.1 Hz), 7.65 (2H, d, J = 8.2 Hz), 3.00 (3H, s), 2.84

(3H, s), 1.62 (3H, s); ¹³C NMR (100.6 MHz, d₆-DMSO)

δ 168.3, 167.4, 141.7, 141.1, 138.3, 137.3, 128.6, 128.1

(2C), 127.7 (3C), 125.1, 124.3 (q, J = 286.2 Hz), 122.6, 74.9

459

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Organic Process Research & Development

Article

(q, J = 27.6 Hz), 38.7, 34.6, and 19.4. MS (ES⁺) 479/481 (M +

H⁺, 45%, 3:1), 520/522 (M + MeCNH⁺, 100%, 3:1); (ES⁻)

477/479 (M − H⁻, 100%, 3:1).

(17) (a) Kepner, C. H.; Tregoe, B. B. The New Rational Manager;

Kepner-Tregoe Inc.: Princeton, 1997. (b) http://www.kepner-tregoe.

com/.

(18) Shaw, N. M.; Naughton, A.; Robins, K.; Tinschert, A.; Schimid,

E.; Hischier, M.-L.; Venetz, V.; Werlan, J.; Zimmermann, T.; Brieden,

AUTHOR INFORMATION

■

W.; de Riedmatten, P.; Roduit, J.-P.; Zimmermann, B.; Neumuller, R.

̈

Org. Process Res. Dev. 2002, 6, 497−504.

Corresponding Author

*E-mail: jonathan.moseley@catsci.com

(19) Moseley, J. D. Route Selection and Process Development of

AZD7545, a PDK Inhibitor, presented at the 24th International

Conference and Exhibition: Organic Process Research and Develop-

ment, Budapest, Hungary, 19−21st September, 2011.

Present Address

^§CatScI Ltd, CBTC2, Capital Business Park, Wentloog, Cardiff

CF3 2PX, UK.

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS

■

We thank Paul Gillespie (AZ Macclesfield) and Paul Gadd and

Julian Wyer (both AZ Charnwood) for chemical hazard

assessments; Steve Baldwin and Matt Young for analytical

support; Paul Sadler (AZ Charnwood) and Maureen Young for

some additional laboratory preparations; and John Gilday and

Bill Moss for additional guidance. A shorter form of this case

study has been presented previously in oral form (see ref 19).

REFERENCES

■

(1) Butlin, R. J.; Nowak, T.; Burrows, J. N.; Block, M. H. PCT Int.

Appl. WO 1999/9962506 A1, 1999.

(2) Mayers, R. M.; Leighton, B.; Kilgour, E. Biochem. Soc. Trans.

2005, 33, 367−370.

(3) Morell, J. A.; Orme, J.; Butlin, R. J.; Roche, T. E.; Mayers, R. M.;

Kilgour, E. Biochem. Soc. Trans. 2003, 31, 1168−1170.

(4) Parker, J. S.; Moseley, J. D. Org. Process Res. Dev. 2008, 12, 1041−

1043.

(5) Moseley, J. D.; Brown, D.; Firkin, C. R.; Jenkin, S. L.; Patel, B.;

Snape, E. W Org. Process Res. Dev. 2008, 12, 1044−1059.

(6) Warren, K. E. H. ICI D5522 Stage 1. Procedure for the

Manufacture of Xylidine Disulphide on 100 L Scale in the LSL. ICI

Fine Chemicals Manufacturing Organisation (FCMO) Process Descrip-

tion, Internal Document, January 1992.

(7) Plumb, J. B. Final Chemical Hazard Assessment ZZ03845. D5522

Stage 1: 4-Amino-2,5-dimethylphenylthiocyanate and Its Conversion

to the Disulphide. ICI Fine Chemicals Manufacturing Organisation

(FCMO) Chemical Hazard Assessment, Internal Document, March 1992.

(8) See relevant volume and entry ofBretherick’s Handbook of Reactive

Chemical Hazards, 6th ed.; Urban, P. G., Ed.; Butterworth-Heinemann:

Oxford, 1999 (2 volumes).

(9) See relevant volume and entry of Encyclopedia of Reagents for

Organic Synthesis; Paquette, L. A., Ed.; John Wiley and Sons:

Chichester UK, 1995 (8 volumes).

(10) Toste, F. D.; De Stefano, V.; Still, I. W. J. Synth. Commun. 1995,

25, 1277−1286.

(11) Ohnmacht, C. J.; Russell, K.; Empfield, J. R.; Frank, C. A.;

Gibson, K. H.; Mayhugh, D. R.; McLaren, F. M.; Shapiro, H. S.;

Brown, F. J.; Trainor, D. A.; Ceccarelli, C.; Lin, M. M.; Masek, B. B.;

Forst, J. M.; Harris, R. J.; Hulsizer, J. M.; Lewis, J. J.; Silverman, S. M.;

Smith, R. W.; Warwick, P. J.; Kau, S. T.; Yochim, C.; Dryoff, M. C.;

Kirkland, M.; Neilson, K. L. J. Med. Chem. 1996, 39, 4592−4601.

(12) Shin, J. M.; Kim, Y. H. Tetrahedron Lett. 1986, 27, 1921−1924.

(13) Kim, Y. H.; Shin, J. M. Tetrahedron Lett. 1985, 26, 3821−3824.

(14) Chidambaram, R.; Zhu, J.; Penmetsa, K.; Kronenthal, D.; Kant,

J. Tetrahedron Lett. 2000, 41, 6017−6020.

(15) Ando, T.; Cork, D. G.; Kimura, T. Chem. Lett. 1986, 665−666.

(16) Giles, M. E.; Thomson, C.; Eyley, S. C.; Cole, A. J.; Goodwin,

C. J.; Hurved, P. A.; Morlin, A. J. G.; Tornos, J.; Atkinson, S.; Just, C.;

Dean, J. C.; Singleton, J. T.; Longton, A. J.; Woodland, I.; Teasdale, A.;

Gregertsen, B.; Else, H.; Athwal, M. S.; Tatterton, S.; Knott, J. M.;

Thompson, N.; Smith, S. J. Org. Process Res. Dev. 2004, 8, 628−642.

460

dx.doi.org/10.1021/op2003419 | Org. Process Res. Dev. 2012, 16, 447−460

Article Doi

Process development and scale-up of AZD7545, a PDK inhibitor

DOI: 10.1021/op2003419

Source and publish data:

Authors:

Article abstract of DOI:10.1021/op2003419

Full text of DOI:10.1021/op2003419

Products guided by the article

R&D Labs maybe for 1286573-32-7

Relevant to this article

Hot Product