Streamlined Synthesis of a Bicyclic Amine Moiety Using an Enzymatic Amidation and Identification of a Novel Solid Form

Enzymatic Amidation and Identi ﬁ cation of a Novel Solid Form

Article

Streamlined Synthesis of a Bicyclic Amine Moiety Using an

Maria S. Brown, Michaella A. Caporello, Adam E. Goetz,* Amber M. Johnson, Kris N. Jones,

Kevin M. Knopf, Samir A. Kulkarni, Taegyo Lee,* Bryan Li, Cuong V. Lu, Javier Magano,

Angela L. A. Puchlopek-Dermenci, Giselle P. Reyes, Sally Gut Ruggeri, Lulin Wei,

Gerald A. Weisenburger, Richard A. Wisdom, and Mengtan Zhang

Cite This: Org. Process Res. Dev. 2021, 25, 1419 −1430

Read Online

ACCESS

Metrics & More

Article Recommendations

sı Supporting Information

ABSTRACT: We describe a series of improvements to the synthesis of a 3,8-diazabicyclo[3.2.1]octane derivative that result in a

reduced step count and higher overall eﬃciency compared to previously published syntheses. Our method includes optimization and

mechanistic understanding of a key diastereoselective cyclization to achieve a >95:5 diastereomeric ratio, as well as demonstration of

a unique enzyme-catalyzed amidation reaction using hexamethyldisilazane as both an ammonia source and scavenger. Finally, we

identify a novel cocrystal solid form of the target compound that provides improved purity and material properties. Demonstration

of the new chemistry to prepare >100 kg of the target compound serves to illustrate the robustness of the new process.

KEYWORDS: heterocyclic amine, enzymatic amidation, cocrystal, pyrrolidine, hexamethyldisilazane, bicyclic amine

INTRODUCTION

Scheme 1. Several Examples of Cyclic Motifs That Can Be

Accessed from meso-Dibromide 1

■

Pharmaceutical chemistry is well-known to rely heavily on

heterocyclic aromatic building blocks to construct target

molecules. More recently, heterocyclic fragments that contain

more sp content are becoming increasingly popular. These

fully or partially saturated analogues often have better

solubility than aromatic counterparts and oﬀer an increased

ability to achieve critical interactions with targeted biological

receptors due to their 3D shape. As a result, the development

of eﬃcient syntheses of these “sp -rich” heterocyclic fragments

is growing in relevance.

We were faced with the need to prepare large quantities of a

building block containing the 3,8-diazabicyclo[3.2.1]octane

system (3) to support a program with ongoing clinical trials.

Although multiple strategies exist for preparing this ring

system, the most commonly utilized precursor is meso-diethyl-

,5-dibromohexanedioate (1). In addition to the 3,8-

diazabicyclo[3.2.1]octane system, dibromide 1 has been

formation of the pyrrolidine ring via double S 2 displacement

demonstrated as a precursor to a variety of unique ring

followed by saponiﬁcation/acidiﬁcation to give di-acid 5.

Unfortunately, previous reports show that even when

dibromide containing ≥98% of the meso-diastereomer is

employed, the product d.r. is typically only ∼70−85% of the

desired syn-diastereomer, which was consistent with our

observations. This epimerization represents a loss in yield

systems (Scheme 1, 1→2, 1→4). The dibromide is produced

via nonstereoselective bromination of adipic acid, but the

desired meso-diastereomer (1) can be selectively crystallized

with a >95% diastereomeric ratio (d.r.) leaving a racemic

mixture of (R,R) and (S,S) enantiomers in the mother

,4c

liquor.

For our speciﬁc needs, we targeted the 8-benzyl derivative of

Received: April 6, 2021

Published: May 27, 2021

the diazabicyclo[3.2.1]octane ring system (3, R = Bn, R =

H). Published syntheses of benzylated analogue 3 exist in the

literature, and we utilized a modiﬁed version of these

published routes to access the desired bicyclic amine, which

was isolated as the bis-HCl salt (Scheme 2, top route).

Reaction of dibromide 1 with benzylamine resulted in

2021 American Chemical Society

Org. Process Res. Dev. 2021, 25, 1419−1430

419

Organic Process Research & Development

Article

Scheme 2. Original Large-Scale Synthesis of Bis-HCl Salt 8 and a Proposed Alternative Route

Figure 1. Inﬂuence of various bases and solvents for the conversion of dibromide 1 to syn-diester 9 using either inorganic bases (left) or organic

bases (right). The results from inorganic base screening show only reactions performed using an aqueous solution of the bases. The performance of

the original conditions (excess BnNH as both a nucleophile and a base) is indicated by a horizontal line.

since the syn geometry is required to ultimately form the

bicyclic scaﬀold. Nevertheless, conversion of syn-di-acid 5 to

imide 7 was accomplished using a straightforward albeit

somewhat lengthy sequence involving cyclic anhydride

formation using diisopropylcarbodiimide (DIC), ring opening

with methanolic ammonia to give amide 6, and formation of

imide 7 using 1,1′-carbonyldiimidazole (CDI). Reduction of

imide 7 using sodium bis(2-methoxyethoxy)aluminum hydride

attractive solid form due to its simplicity, it possessed several

distinct disadvantages that we wanted to address by exploring

alternative solid forms. This manuscript describes these

improvements to the process as well as an initial multikilo-

gram-scale demonstration of the new chemistry.

RESULTS AND DISCUSSION

■

Improving Diastereoselectivity in the Cyclization of

. As mentioned above, the cyclization of meso-diastereomer 1

(Red-Al) gave the desired amine as a freebase, which was

isolated as the corresponding bis-HCl salt 8. During our initial

large-scale campaign, 8 was prepared in seven steps with an

with various amines aﬀords primarily the syn-diastereomer of

the pyrrolidine diester, however, with a decrease in the

stereochemical purity relative to the starting dibromide. During

early campaigns, we utilized an excess of benzylamine (to serve

as both the nucleophile and base) in toluene, which gave an

∼85:15 mixture of syn:anti diastereomers despite starting with

1 that was ≥98% of the meso-diastereomer. We found that the

ﬁnal diastereomeric ratio did not further change with

prolonged reaction times, and subjecting puriﬁed syn isomer

1 to the reaction conditions did not result in any stereo-

∼

30% overall yield.

While subsequent campaigns resulted in slight enhance-

ments, we believed that we could streamline this route by

targeting three key areas for further development (Scheme 2,

bottom route). We sought to improve the modest diaster-

eoselectivity during formation of 9 to maximize yield as well as

reduce the overall step count via a direct conversion of bis-

ester 9 to amide 10. Finally, though the bis-HCl salt was an

420

Organic Process Research & Development

Org. Process Res. Dev. 2021, 25, 1419−1430

Article

Figure 2. Epimerization of bromide stereoisomers as reported in refs 5d and 13.

chemical scrambling. Epimerization of pyrrolidine diester

derivatives such as 9 has been demonstrated using stronger

During our attempts to understand the origin of the higher

diastereoselectivity with inorganic bases, we relied on a key

observation from published accounts focused on improving the

isolated yield of meso-dibromide 1 during crystallization. It has

been previously demonstrated that the dibromide stereo-

metal-alkoxide bases that results in a near 1:1 ratio of the two

diastereomers. This is also in line with quantum mechanical

calculations that suggest that the two diastereomers are very

similar in energy. However, since we did not observe any

measurable amount of epimerization of diester 9 using bases

such as benzylamine, triethylamine, or even DBU, we reasoned

that epimerization occurred prior to formation of the carbon−

nitrogen bonds and that we could potentially suppress this by

changing the reaction conditions.

isomers can be epimerized via S 2 displacement by bromide

−

anions (Br ), which can allow for recycling of the more soluble

racemic isomers to increase the recovery of the more highly

crystalline meso-diastereomer (Figure 2).

We have also observed that the same epimerization as

described above occurs during the reaction of dibromide 1

with benzylamine. By monitoring the diastereomeric purity of

the unreacted dibromide 1 (>95% meso at t = 0) as well as the

pyrrolidine diester 9 being formed during the reaction, we

found that stereochemical purity continually decreases

throughout the reaction. Especially in the case of the

dibromide, this erosion in the meso content occurs to a

much greater extent than could be expected solely from

reaction rate diﬀerences between the diﬀerent isomers (Figure

3, solid orange line). The extent of erosion seen in the

remaining dibromide 1 was less when the biphasic conditions

with an inorganic base were utilized (Figure 3, dashed orange

line). The lower selectivity during cyclization is therefore a

result of epimerization of meso-dibromide 1 caused by the

bromide anion byproduct. This results in an increasingly

unselective formation of esters 9 and 12 as the reaction

progresses. We hypothesize that the use of biphasic conditions

helps sequester the bromide anion in the aqueous layer (as the

alkali salt) due to the low solubility of bromide salts in the

We embarked on screening various solvents and bases to

explore the impact on the diastereomeric ratio during

formation of ester 9. We found that nonpolar solvents

(

toluene, MTBE, and cyclohexane) tended to give higher

diastereoselectivity than polar solvents (MeCN and 2-PrOH)

and that inorganic bases gave superior selectivity compared to

organic bases. A portion of the screening data is shown in

Figure 1, which highlights the signiﬁcantly higher selectivity for

when using inorganic bases vs organic bases. More detailed

screening data is available in the Supporting Information. Of

the >20 organic bases evaluated under various conditions, the

selectivity rarely exceeded the ∼85:15 ratio previously seen. In

contrast, many of the inorganic bases screened showed a >95:5

preference for the syn diastereomer. These inorganic bases

were screened using “anhydrous” conditions that gave a

heterogeneous reaction mixture, as well as conditions where

the inorganic base was dissolved in water, which resulted in a

biphasic mixture with the organic solvents. On vial scale,

similar results were seen between these two modes, but during

larger-scale experiments, the “anhydrous” conditions showed

more variability in terms of selectivity and reaction rates, so we

nonpolar organic solvents. Consistent with this hypothesis,

we have observed that spiking in a more organic-soluble

bromide anion such as tetrabutylammonium bromide (TBAB)

into the biphasic conditions results in lower diastereoselectiv-

ity.

elected to proceed with the biphasic solvent mixture. The use

of inorganic bases under biphasic conditions (K CO in

toluene/water) for cyclization of dibromide 1 has been

previously reported; however, in these cases, the reported

selectivity is only 75−80% syn, which is signiﬁcantly lower than

what we observed under our optimized conditions (vide

Despite the higher levels of selectivity seen using inorganic

bases, we found that many of the reactions were quite slow,

often taking ≥48 h to achieve full conversion. While some

reactions in 2-MeTHF were found to proceed with higher

selectivity in ∼24 h, the cost of this solvent relative to other

options from the initial screening (toluene, MTBE, and

cyclohexane) as well as the deleterious impact that it had on

the downstream enzymatic transformation prompted us to

pursue further optimization with alternative solvents. We

therefore performed a second round of screening that focused

c12

infra).

In addition to diastereomers 9 and 12, amide 13

was formed in small amounts (typically <5%) under many of

the conditions screened. We were pleased to see that despite

the presence of aqueous basic conditions, saponiﬁcation of the

diesters to the corresponding acid was not a signiﬁcant

problem under these conditions.

421

Organic Process Research & Development

Org. Process Res. Dev. 2021, 25, 1419−1430

0 −50 °C (Figure 4 ). Unfortunately, all attempts to utilize

Article

aqueous citric acid wash is performed to remove any residual

benzylamine while minimizing the losses of pyrrolidine 9.

Enzymatic Amidation of Ester 9. Despite the simplicity

of the original four-step sequence used to convert bis-ester 9 to

imide 7, we speculated that we could reduce the overall step

count by targeting primary amide 10 as a precursor to the

desired imide. For this transformation, we explored the

feasibility of an enzymatic amidation to form the amide

selectively. Speciﬁcally, we were pleased to ﬁnd that diﬀerent

preparations of immobilized Candida antarctica lipase B

(

CalB) showed the most promising for this transformation

using either solutions of ammonia in alcoholic solvents or

ammonium carbamate with MTBE as the organic solvent at

Figure 3. Monitoring the chiral purity of syn-diester 9 and unreacted

dibromide isomers during pyrrolidine formation. The solid lines are

from a reaction using the original conditions of excess BnNH in

toluene. The meso-dibromide is seen to epimerize signiﬁcantly over

the course of the reaction (orange solid line) resulting in the

selectivity for syn-product formation decreasing throughout the

reaction (yellow solid line). When using aqueous NaHCO /2-

MeTHF (orange dotted line), the meso-dibromide undergoes

signiﬁcantly less epimerization over the course of the reaction.

on identifying solvent/base pairs that could maintain the high

diastereoselectivity while reducing the overall reaction time

(

Table 1). Ultimately, we identiﬁed the use of aqueous

Table 1. Impact of Reaction Conditions on the Rate of Ester

Figure 4. Initial screening of amidation using various immobilized

CalB sources. Conditions: 2.5 mg of ester 9, 25 mg of an enzyme, 25

mg of 3 Å M.S., 2 M ammonia in i-PrOH (0.5 mL), 50 °C, and 22 h.

Formation

base

solvent

temp./time

ester (9) (% area)

NaHCO₃

2-MeTHF (5 V)

75 °C/24 h

55 °C/24 h

55 °C/120 h

55 °C/78 h

75 °C/48 h

75 °C/24 h

liquid enzyme preparations showed essentially no conversion

during this initial evaluation, highlighting the importance of the

solid support for stabilizing the enzyme under the reaction

conditions. After the initial screenings, further optimization

was performed with several of the top enzyme candidates to

understand diﬀerences in performance. We ultimately selected

CalB Immo Plus from Purolite as the preferred option due to a

good balance of activity and cost.

H O (5 V)

K CO₃

MTBE (2.5 V)

H O (2.5 V)

K CO₃

MTBE (5 V)

H O (5 V)

K PO₄

MTBE (5 V)

H O (5 V)

K PO₄

PhMe (5 V)

Unfortunately, both ammonia sources from the initial

screening work had aspects that made them unattractive.

The presence of ethanol is deleterious to enzyme activity,

requiring the use of ammonia solutions in either methanol or

isopropanol. Unfortunately, with both solvents, we observed

H O (5 V)

Na PO₄

PhMe (5 V)

H O (5 V)

Na PO₄

PhMe (2.5 V)

H O (5 V)

varying amounts of transesteriﬁcation byproducts. For

Reactions were run using 3.0 equiv of the base. Reactions were

monitored by UPLC until >97% conversion was achieved.

ammonium carbamate, the continual release of CO₂was

unattractive as well as the large amount of solids that were

present in the reaction. Molecular sieves (or CaCl ) were also

required in both cases to minimize carboxylic acid byproducts

that formed from the small amount of water introduced with

the ammonia sources as well as to adsorb the EtOH byproduct

that is generated from the desired transformation and prevent

enzyme poisoning.

To address these drawbacks, we explored the use of

hexamethyldisilazane (HMDS) as an alternative ammonia

trisodium phosphate (Na PO ) as the preferred base, using

toluene as the organic phase, which also allowed for a higher

reaction temperature than MTBE. By reducing the overall

volumes of the reaction, we could further accelerate the

reaction to give an ∼90% assay yield of 9 (97:3 syn:anti) with a

typical reaction time of 24−28 h. Following the reaction, an

422

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

Article

source. We hypothesized that reaction with any water or

alcohol present in the reaction mixture would gradually liberate

ammonia while also having the beneﬁt of removing these

problematic species by converting them to (presumably inert)

silylated species (Scheme 3, 16→17 and 16→18). To our

Scheme 3. Reaction of HMDS with Various Protic Species

Present during the Enzymatic Amidation

Figure 5. Recycling of a single batch of enzymes during conversion of

to amide 10. Conditions: 2.0 g of ester 9, 1 g/g CalB Immo Plus, 40

mL of MTBE, 1.5 equiv of HMDS, and 40 °C. Following the reaction,

the enzyme was washed with MTBE then soaked with EtOH in

MTBE (∼1 equiv of EtOH) for several hours before a ﬁnal MTBE

wash. See the Supporting Information for full details on the washing

protocol used after each run.

completely understood although numerous experiments show

that it provided a consistent and reproducible beneﬁt. We

hypothesize that a critical amount of water is needed to

maintain a hydration sphere around the enzyme for optimal

performance. Indeed, with too little water, enzymatic activity

drops dramatically (and the appearance of silylated analogues

of 10 is sometimes seen), but with too much water, the

undesired carboxylic acid byproduct is formed preferentially,

knowledge, the use of HMDS has not been demonstrated

previously in enzymatic transformations, but we were delighted

to ﬁnd that it gave the desired amide product in good yield.

Initially, the high enzyme load that was required for good

activity (∼1 g immobilized enzyme/g substrate) prompted us

to explore the possibility of recycling the enzyme through

multiple runs. We observed that at the end of the reaction, the

enzyme showed signiﬁcantly decreased activity, and we

speculated that HMDS could be reacting with various sites

on the enzyme leading to diminished activity (Scheme 3, 16→

which does not undergo conversion to the desired amide. We

have observed that maintaining the solution water content

around 0.02−0.1% (as measured by coulometric KF titration)

gives good performance although this most likely does not

represent the true water content since a portion of the water

is adsorbed onto the enzyme and not detected by solution

measurements. After signiﬁcant optimization, we settled on a

charging strategy where ∼1.3 equiv of HMDS and ∼0.5−0.6

equiv of water were utilized. At the start of the reaction,

approximately half of each reagent was charged, and then, the

remaining amount was charged in periodic doses to achieve full

conversion. Reactions were typically complete within 48−72 h

and showed >90% (a/a) amide 10, with <5% (a/a) of

19). Indeed, we found that soaking the enzyme beads in

MTBE containing a stoichiometric amount of EtOH (relative

to HMDS charged) regenerated the activity. Further

exploration of this enzyme washing protocol suggested that

the speciﬁc conditions employed had some impact on the

subsequent run (see the Supporting Information ); however, we

were able to successfully demonstrate that the same sample of

an enzyme could be reused for at least 8 cycles with minor

impact to the overall reaction proﬁle as shown in Figure 5.

Furthermore, the fact that there is no continual decline in

enzyme performance over sequential runs (e.g., cycle 5

outperforms cycle 2) is consistent with our belief that the

enzyme deactivation by HMDS is largely reversible (16→19→

carboxylic acid byproduct.

We had originally hoped to perform the pyrrolidine

formation, enzymatic amidation, and imide formation (9→

10→7) without isolation to help streamline the overall process.

While it was possible to telescope the solution of pyrrolidine 9

into the enzymatic reaction, we did observe variability across

diﬀerent ingoing batches of 9 with some reactions stalling out

before achieving full conversion. We performed numerous

spiking studies to understand the tolerability of diﬀerent

impurities/byproducts that were plausible in ester 9 as well as

studies to identify suitable washes or treatments that might

provide a consistent material for the enzymatic reaction.

Ultimately, we opted to isolate the HCl salt of bis-ester 9 and

then perform a salt-break prior to the enzymatic step, which

helped normalize the performance of the ingoing material.

Fortunately, we were able to telescope amide 10 into the imide

formation after a simple ﬁltration to remove the enzyme.

Attempts to cyclize amide 10 by heating in reﬂuxing toluene

showed no conversion to the desired imide. Indeed, previous

reports demonstrate that ring closure of amides similar to 10

can be accomplished thermally although temperatures in

8, Scheme 3) and that the washing protocol is critical to

restoring enzyme performance.

Ultimately, we sought to reduce the overall enzyme loading,

and through these eﬀorts, we discovered that the presence of

small amounts of water had a dramatic impact on reaction

performance. We were excited to ﬁnd that addition of a

substoichiometric amount of water (typically 0.5−0.8 equiv)

allowed for a 6−8-fold reduction in the immobilized enzyme

from 1 to 0.12−0.15 g/g while maintaining good performance.

In the case of immobilized enzymes, the actual mass of the

enzyme is quite low, with the bulk of the mass coming from the

solid support. For CalB Immo Plus, the protein loading is

reported to be approximately 24 mg/g , which corresponds

wet

e26

to an enzyme loading of 0.3−0.4% (w/w) relative to diester

excess of 200 °C are required.

Alternatively, we opted to

The exact role of water in this transformation is not

treat crude amide 10 with a solution of potassium tert-butoxide

423

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

KOtBu) in THF to eﬀect rapid ring closure at room

temperature and intercept imide 7, which served as the

penultimate intermediate in our initial synthetic route.

Numerous examples of imide reduction on substrates similar

Article

(

88-6) as the most attractive candidate. 4,4′-Dihydroxybiphen-

yl is produced on a >1000 MT/year scale, is used widely as a

key building block in the liquid crystal and high-performance

polymer ﬁelds, and with a cost of ∼$30/kg contributes little

to the economics of the overall process. We were further

pleased to ﬁnd that when this coformer was mixed with amine

11, it resulted in a crystalline material that was composed of a

2:1 ratio of amine 11 to biphenol 20 (Figure 6 ). A beneﬁt of

to 7 have also been reported, typically using LiAlH to produce

the bicyclic diamine. However, during work on our initial

route, we identiﬁed sodium bis(2-methoxyethoxy)aluminum

hydride (Red-Al) as an attractive alternative based on ease of

handling as well as the convenience of using a concentrated

toluene solution, which simpliﬁed the workup and allowed for

direct extraction of the desired freebase amine 11 into toluene

after the reduction.

Identiﬁcation of a Novel Cocrystal of 11. The desired

bicyclic amine 11 is a low melting solid with a melting point of

0−35 °C and is therefore more easily handled as a salt.

While the bis-HCl salt 8 is an obvious choice, we observed that

this material tends to have low bulk density (typically 0.25−

.30 g/cm ), poor ﬂowability, variable levels of certain

impurities and in a few cases caused minor corrosion to steel

containers used for storage due to excess HCl. As a result, we

explored other solid forms that might address some of these

drawbacks. Rather than limiting ourselves only to various acid

salts, we also explored the possibility of forming a cocrystal

given the propensity of certain amines to form these ordered

structures by serving as hydrogen bond acceptors. It was also

hoped that by employing a cocrystal, we could avoid

performing a salt-break prior to using the amine in downstream

reactions.

We utilized a combination of computational and exper-

imental screening by ﬁrst ranking a library of potential

coformers based on their calculated ability to form strong

hydrogen bonds with the desired amine 11 using commercially

available COSMOquick software (Table 2, see the Supporting

Figure 6. (a) Formation of cocrystal 21. (b) Microscopy comparison

of bis-HCl salt 8 (left) with cocrystal 21 (right). (c) X-ray crystal

structure of 21 showing the unit cell with key hydrogen bonding

interaction between O1 and N1 (distance is 1.73 Å).

Table 2. Top Computational Hits for Potential Coformers

for Amine 11 Using COSMOquick

coformer

phenol

,4-biphenol (20)

coformer ranking

H_m_i_x(kcal/mol)

f_ﬁt

−3.64

−4.33

−1.47

−1.76

−2.70

−2.12

−0.48

−1.88

−0.32

−0.21

−2.11

−1.78

0.06

0.79

0.87

0.94

1.05

1.18

1.21

1.32

this 2:1 complexation is that the active potency of amine 11 is

only slightly lower than the original bis-HCl salt (68.5% for the

cocrystal vs 73.5% for the bis-HCl salt). Furthermore, the

cocrystal formed well-behaved crystalline particles (typical bulk

density of 0.5−0.6 g/cm ), had much better ﬂowability, is

nonhygroscopic, and to date has given only a single polymorph

across various conditions.

saccharin

p-hydroxybenzamide

butyl acetate

fucose

thiabendazole

α-D-ribopyranose

imidazole

Further optimization of the crystallization procedure

ultimately produced a cocrystal in an 85−90% yield based

on the assay concentration of amine 11 present. Following

reduction of imide 7 and aqueous workup, a solution of amine

11 in toluene was obtained. This turned out to be an optimal

solvent for isolation of the desired cocrystal due to the low

-aminopyrimidine

Information for full results). For each potential coformer, the

software calculates the mixing enthalpy (H_m_i_x), which

corresponds to the enthalpy diﬀerence between a supercooled

mixture of the two components and the enthalpies of the two

individual components. This enthalpy value relates to the

tendency of the two species to cocrystallize. Application of a

correction to H_m_i_xbased on the number of rotatable bonds in

the two molecules gives the “f_ﬁt” parameter, which tends to

solubility of this species in toluene. The biphenol also has

poor solubility in toluene, but it can be readily dissolved to give

a ∼20% (w/w) solution in THF. We found that adding 0.55−

0.60 equiv of biphenol (as a THF solution) to a heated toluene

solution of amine 11 followed by gradual cooling resulted in

recovery of the desired cocrystal (Figure 6a). The addition of

the biphenol as a THF solution as opposed to as a solid charge

minimizes the risk of the ﬁnal solids containing free 4,4′-

biphenol due to the relatively low solubility of this coformer.

Crystallization of 21 was also found to provide a signiﬁcant

purity upgrade. Although prior batches of the bis-HCl salt 8

showed generally high purity (>98% by HPLC), various low-

give better predictive power.

The top computational hits were then evaluated exper-

imentally to determine whether mixing of amine 11 with each

coformer resulted in a new solid form, as determined by PXRD

analysis. Several new cocrystals were identiﬁed, and we

ultimately selected 4,4′-dihydroxybiphenyl (20, CAS no. 92-

424

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

Article

level impurities were observed across diﬀerent batches of this

material (Figure 7 ). When isolating cocrystal 21, we

layer, whereas the undesired 4,4′-dihydroxybiphenyl was

extracted into the aqueous layer as the bis-potassium salt. In

some cases, cocrystal 21 can also be used directly (without

complication from the biphenol) due to the higher

nucleophilicity of the amine fragment. Both amidation using

carboxylic acid 27 with propylphosphonic anhydride (T3P)

and S Ar reaction with pyrrolopyrimidine 29 proceeded

smoothly to give the respective products 28 and 30 resulting

from a reaction with the amine. Following each reaction, the

unreacted biphenol was readily removed via extraction with

aqueous KOH before isolation of the desired product.

Kilogram-Scale Manufacturing of 21. Having identiﬁed

a number of improvements to the overall process for producing

amine 11, we successfully demonstrated this technology on

multi-kilogram scale, as shown in Scheme 5. During the

formation of bis-ester 9, the biphasic conditions performed as

expected based on lab trials and resulted in high selectivity for

the syn diastereomer (>96:4 dr for both batches). Isolation as

the HCl salt 31 provided a further upgrade to yield a material

that had ≥99.5% purity by HPLC. During the initial batch, the

isolated yield was only 70%, which was much lower than

expected, and the mother liquors were found to contain an

additional ∼14% yield of the product. For the second batch,

the amounts of HCl and EtOH used during salt isolation were

decreased, which resulted in a much improved 83% isolated

yield without any change in purity.

Figure 7. Impurities observed in amine 11 used for spiking studies.

consistently observed HPLC purities of >99.8% (a/a). Even

when numerous impurities (22−26, Figure 7) were intention-

ally introduced to provide a solution of amine 11 with an

ingoing purity of ∼90% by HPLC, the ﬁnal cocrystal was still

isolated with >99.9% HPLC purity.

In addition to providing a robust puriﬁcation point for the

desired amine, cocrystal 21 can also serve as a useful precursor,

analogous to the bis-HCl salt (Scheme 4). Cocrystal 21 can be

easily broken using a single partition between aqueous

potassium hydroxide (KOH) solution and an organic solvent

such as MTBE, toluene, 2-MeTHF, or DCM. In all cases,

The initial batch of the amidation reaction utilized a 15%

98% of the ingoing amine 11 was recovered in the organic

(

w/w) enzyme loading and charged HMDS and water in only

two portions (t = 0 and 28 h). The reaction reached

completion after 42 h; however, it was found that the level of

the monoacid impurity was higher than desired (∼9% by area),

which is likely due to having too much water present during

the initial stages of the reaction. For the second batch, the

enzyme loading was decreased to 12.5% (w/w), and the

charging strategy was adjusted to utilize less water upfront

followed by an increased number of smaller charges during

later portions of the reaction. This had the desired eﬀect of

reducing the amount of the monoacid impurity formed,

although it resulted in a slower reaction that required ∼85 h to

reach full conversion. The proﬁle for this reaction is shown in

Figure 8 (solid lines), along with the levels of the various silyl

species that can be monitored by GC (Figure 8, dashed lines).

Further optimization of the charging strategy will likely allow

for shorter reaction times while still maintaining the high

selectivity for the desired amidation reaction. Given the

signiﬁcant reduction in enzyme loading that had been achieved

and the fact that only two batches using the enzymatic

transformation were planned, we opted not to recycle the

enzyme during this campaign. Ultimately, a further under-

standing of the economic impact of recycling the enzyme vs

purchasing fresh supplies may be beneﬁcial for larger

campaigns or processes where the enzyme cost is a larger

contributor.

Scheme 4. Synthetic Utility of Cocrystal 21

Following amidation, the product solution is ﬁltered to

remove the enzyme and then concentrated before potassium

tert-butoxide is added as a THF solution to eﬀect ring closure

and form imide 7. After pH adjustment using aqueous citric

acid and brine, the THF solution is again concentrated before

water is added to promote crystallization of imide 7. Overall,

we found that this 2-step sequence produced imide 7 in an

average of 81% yield from diester 9, despite the lower yield for

the initial batch due to elevated levels of the monoacid

impurity.

425

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

Article

Scheme 5. Pilot Plant-Scale Synthesis to Produce 110 kg of Cocrystal 21

Figure 8. Reaction proﬁle showing conversion of various pyrrolidine species by HPLC (solid lines) and separate tracking of various silyl species by

GC (dashed lines).

Reduction of imide 7 was performed using a nearly identical

process to earlier campaigns that targeted the bis-HCl salt. The

initial reduction is facile, giving amide intermediate 25 (see

Figure 7), which is then gradually heated to 65 °C to promote

the subsequent reduction. After the reaction is complete, it is

quenched via slow addition into a large excess of aqueous

sodium hydroxide, which gives a solution of amine freebase 11

in toluene. Past experience with the bis-HCl salt had shown

that the ﬁnal material sometimes possessed a light brown/tan

color that was diﬃcult to remove during downstream

synthesis, we have identiﬁed a shorter sequence to construct

the bicyclic moiety using an enzymatic amidation that employs

HMDS as both the ammonia source and the reagent to help

sequester byproducts that would otherwise reduce enzyme

performance. We have demonstrated the robustness and

scalability of this chemistry by preparing >100 kg of the target

compound with high purity in similar performance to lab runs.

When compared to the initial process to prepare bis-HCl salt

8, we have decreased both the number of steps and isolated

intermediates while also increasing the overall yield from 29.8

to 49.6%. This results in an ∼40% reduction in the amount of

the starting dibromide 1 needed to prepare the desired bicyclic

amine 21, even when accounting for potency diﬀerences of the

diﬀerent solid forms. Based on observations during this initial

large-scale campaign, we believe that further improvement to

the route presented here would oﬀer even more beneﬁt

compared to existing technology.

chemistry. Therefore, we opted to circulate this toluene

mixture through a carbon cartridge to help improve the color.

Formation of the cocrystal as described above resulted in the

isolation of 110 kg of cocrystal 21, which corresponds to a 79%

yield over the reduction and cocrystal formation sequence.

SUMMARY/CONCLUSIONS

■

In summary, we have demonstrated a streamlined process to

prepare bicyclic amine 11 as well as identiﬁed a novel cocrystal

that serves as a convenient solid form for isolation. In addition

to improving the performance of key reactions used in the

EXPERIMENTAL SECTION

■

Preparation of Pyrrolidine 31. To a 2000 L vessel were

added water (840 kg) and tribasic sodium phosphate (160 kg,

426

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

25 mol, and 2.12 equiv). To this cloudy suspension were

Article

The MTBE solution containing 10 was concentrated at 55−

60 °C to remove ∼1000 kg of the solvent. THF (862 kg) was

added, and the mixture was cooled to 15 °C. To this solution

was added potassium tert-butoxide (20% in THF, 180 kg, 321

mol, and 0.992 equiv) over 3.5 h. After stirring for another 40

min, the reaction was cooled to 10 °C and quenched via

addition of an aqueous solution of 10% citric acid and 10%

sodium chloride (676 kg). After allowing the layers to settle,

the aqueous layer was discarded. The organic layer was

concentrated to remove ∼1050 kg of the solvent, and then,

water (750 kg) was added at 35−40 °C over 45 min to

promote crystallization. The slurry was cooled to 20 °C and

ﬁltered, and the solids were washed with water (100 kg). Imide

7 was dried under vacuum to give 61.9 kg (83% yield from

added dibromide 1 (140 kg, 389 mol, and 1.00 equiv), toluene

315 kg, 2.25 g/g), and benzylamine (52 kg, 490 mol, and 1.2

(

equiv), and the vessel was made inert with nitrogen. The

contents were heated to 75 °C and held until <1% of

dibromide 1 remained (∼26 h). After cooling to 40 °C,

additional water (280 kg) and toluene (285 kg) were added,

and the layers were separated. The organic phase was washed

twice with an aqueous solution containing 10% (w/w) citric

acid and 10% (w/w) NaCl (2 × 530 kg) followed by a ﬁnal

wash with 10% aqueous NaCl (130 kg). The solution was

concentrated under vacuum to reduce the amount of toluene

to ∼400 kg, and then, distillation continued (with toluene

replacement) until a water content of 0.03% (as measured by

KF) was obtained. The reaction was cooled to 10 °C, and

ethanol (6 kg) and 1,4-dioxane (85 kg) were added. Into this

solution was slowly bubbled gaseous HCl (13 kg, 356.6 mol,

and 0.92 equiv), and the slurry was cooled to 5 °C. The solids

were ﬁltered and washed with toluene (110 kg), and the

pyrrolidine salt 31 was dried until 138.2 kg (110.6 kg

corrected, 83% yield, and ∼20% toluene by assay) of an easily

pyrrolidine 31) of the ﬁnal product. HPLC purity: >99.5% (a/

a); H NMR (400 MHz, DMSO-d ): δ 10.90 (s, 1H), 7.38−

7.31 (m, 2H), 7.31−7.25 (m, 3H), 3.70 (s, 2H), 3.62−3.57

(m, 2H), 2.30−2.20 (m, 2H), 1.85−1.76 (m, 2H); C NMR

(100 MHz, DMSO-d ): δ 174.5, 138.0, 129.0, 128.9, 127.8,

63.6, 53.1, 26.3; HRMS (ESI/Orbitrap) m/z: [M + H] calcd

for C H O N , 231.1128; found, 231.1130. Analytical data is

handled solid was obtained. HPLC purity: >99.5% (a/a,

consistent with previous reports.

excluding toluene); H NMR (400 MHz, DMSO-d ): δ 11.47

Preparation of Cocrystal 21. To a 1000 L vessel were

added imide 7 (53.9 kg, 234 mol, and 1.00 equiv) and toluene

(479 kg). The contents were cooled to 5 °C, Red-Al (70% in

toluene, 312 kg, 1080 mol, and 4.62 equiv) was added over ∼5

h to maintain a temperature of 0−10 °C, and then, the transfer

line was washed with toluene (20 kg). The reaction was heated

to 63 °C over 5 h and held for an additional 2 h until imide 7

and amide 25 had been consumed. The contents were then

cooled to 25 °C and transferred gradually over 2.5 h to a vessel

containing sodium hydroxide (33% in water, 550 kg, 4540 mol,

and 19.4 equiv) and water (175 kg) at 5 °C. The mixture was

stirred for an additional 30 min, and then, the layers were

separated. The aqueous layer was further extracted with

toluene (100 kg), and the organic layers were combined. The

combined organic layers were washed with two portions of

aqueous sodium chloride (5% in water, 66 kg). The toluene

solution was dried by distillation under vacuum at 50 °C until

a water content of <0.1% was obtained by KF measurement,

and then, toluene was added to give an ∼8% (w/w) solution of

amine 11 in toluene. Two identical batches produced using the

above procedure were combined and circulated through a plug

containing activated carbon (2.7 kg). The ﬁnal solution

showed an HPLC purity of 98.2% (a/a, toluene excluded).

To a vessel containing half of the toluene solution from

above (517 kg, ∼42.5 kg of amine 11, 210 mol, and 1.00

equiv) held at 60−65 °C was added 4,4′-dihydroxybiphenyl

(19.5% w/w in THF, 119 kg, 125 mol, and 0.595 equiv) over

2−3 h followed by THF (7 kg). The contents were cooled to 0

°C slowly, ﬁltered, and washed with 4:1 toluene:THF (55 kg).

The resulting solids were slurried with water (230 kg), ﬁltered,

washed with water (70 kg), and dried under vacuum at 45 °C.

Two batches were performed using the above procedure to

produce a total of 110.2 kg of cocrystal 21 (79.4% yield from

(

br s, 1H), 7.60−7.50 (m, 2H), 7.40−7.31 (m, 3H), 4.39 (s,

H), 4.20 (br s, 2H), 4.01 (app. qd, J = 7.2, 1.7 Hz, 4H), 2.4−

.26 (m, 2H), 2.10−1.98 (m, 2H); C NMR (100 MHz,

DMSO-d ): δ 170.0, 133.4, 131.0, 129.2, 128.8, 66.7, 61.8,

for C H O N, 306.1700; found, 306.1701. Anal. calcd for

C H ClNO : C, 59.73; H, 7.08; N, 4.10. Found: C, 59.76; H,

8.7, 28.0, 14.3; HRMS (ESI/Orbitrap) m/z: [M + H] calcd

.78; N, 4.13.

Preparation of Imide 7. To a 2000 L vessel were added

sodium phosphate (7.5 kg, 39 mol, and 0.12 equiv), water (140

kg), pyrrolidine 31 (138.2 kg, 110.56 kg corrected, 323.4 mol,

and 1.000 equiv), and MTBE (260 kg).While maintaining a

temperature of 10−15 °C, 33% NaOH (33.1 kg, 273.1 mol,

and 0.844 equiv) was added, resulting in a solution with pH =

.2. The layers were separated, and the organic phase was

washed with a 10% sodium chloride solution (30 kg) followed

by water (30 kg). The organic phase was concentrated at 40−

and then further distilled (with MTBE replacement) until a

water content of 0.1% (as measured by KF) was obtained.

To this MTBE solution were added the CalB Immo Plus

immobilized enzyme (12.0 kg), MTBE (540 kg), HMDS

5 °C under vacuum until ∼1/2 of MTBE had been removed

(

42.00 kg, 260.2 mol, and 0.8 equiv), and water (1.50 kg, 83.3

mol, and 0.258 equiv), and the reaction was warmed to 41−43

C. To the reaction was added additional water (1.0 kg, 0.48

kg, 0.17 kg, and 0.284 equiv total) after 24.5, 43.5, and 64 h,

and additional HMDS (25.00 kg, 154.9 mol, and 0.47920

equiv) was added after 25 h. After 86 h, the reaction was

ﬁltered to remove the immobilized enzyme, which was washed

with MTBE (130 kg) to give 1267 kg of an MTBE solution

containing amide 10. HPLC purity: 89−93% (a/a); H NMR

(

400 MHz, CDCl ): δ 7.87 (s, 1H), 7.22−7.06 (m, 5H), 5.83

s, 1H), 3.87 (q, J = 7.0 Hz, 2H), 3.77 (d, J = 13.2 Hz, 1H),

imide 7). HPLC purity: >99.9% (a/a); H NMR (400 MHz,

.67 (d, J = 13.2 Hz, 1H), 3.47−3.36 (m, 2H), 2.09−1.98 (m,

DMSO-d ): δ 7.40−7.33 (m, 8H), 7.29 (t, J = 7.3 Hz, 4H),

H), 1.95−1.86 (m, 1H), 1.80−1.69 (m, 1H), 1.02 (t, J = 7.2

7.21 (t, J = 7.3 Hz, 2H), 6.80 (d, J = 8.4 Hz, 4H), 3.39 (s, 4H),

2.90 (s, 4H), 2.75 (d, J = 12.0 Hz, 4H), 2.43 (dd, J = 12.0, 1.7

Hz, 3H); C NMR (100 MHz, CDCl ): δ 178.3, 175.3,

37.39, 129.4, 128.5, 127.6, 67.9, 66.4, 61.0, 59.1, 30.7, 30.5,

Hz, 4H), 1.92−1.84 (m, 4H), 1.71−1.64 (m, 4H); C NMR

4.1; HRMS (ESI/Orbitrap) m/z: [M + H]⁺calcd for

(100 MHz, DMSO-d ): δ 156.8, 140.4, 131.6, 128.9, 128.5,

C H O N , 277.1547; found, 277.1548.

127.45, 127.0, 116.1, 60.5, 57.7, 52.2, 25.4. Anal. calcd for

427

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

https://pubs.acs.org/doi/10.1021/acs.oprd.1c00120.

Article

C H N O : C, 77.25; H, 7.85; N, 9.48. Found: C, 77.06; H,

UPLC purity: 98.3% (a/a); H NMR (400 MHz, CDCl ): δ

.01; N, 9.48.

8.12 (d, J = 8.5 Hz, 2H), 7.45−7.25 (m, 8H), 6.53 (d, J = 4.1

Hz, 1H), 4.20 (br s, 2H), 3.60 (s, 2H), 3.42 (d, J = 10.1 Hz,

2H), 3.33 (s, 2H), 2.43 (s, 3H), 2.11−2.00 (m, 2H), 1.70−

Representative Breaking of Cocrystal 21. To a solution

of KOH in water (45%, 6.0 mL, and 8.2 equiv) were added

additional water (15 mL, 3 mL/g), an organic solvent (25 mL,

1.62 (m, 2H); C NMR (100 MHz, CDCl ): δ 158.6, 154.7,

mL/g), and cocrystal 21 (5.0 g, 8.5 mmol, and 1.0 equiv).

152.3, 145.7, 139.0, 134.6, 129.6, 128.8, 128.7, 128.4, 127.2,

121.3, 105.1, 103.4, 58.2, 56.9, 52.3, 25.5, 22.1; HRMS (ESI/

The mixture was stirred until all solids were dissolved, and the

solution was transferred to a separatory funnel. In all solvents

tested, the phase split was rapid and gave a clean interface. The

aqueous layer was separated and discarded. The remaining

organic layer was analyzed by qNMR (1,3,5-trimethoxyben-

zene as an internal standard) and found to contain >98% of the

theoretical amount of amine 11 based on the calculated

potency. An analytical sample was obtained by concentrating

Orbitrap) m/z: [M + H] calcd for C H ClN O S, 508.1568;

found, 508.1573.

ASSOCIATED CONTENT

sı Supporting Information

■

The Supporting Information is available free of charge at

the organic layer to give 11 as an oil. H NMR (400 MHz,

CD CN): δ 7.41 (d, J = 6.9 Hz, 2H), 7.34 (t, J = 7.4 Hz, 2H),

Spectral data for isolated compounds, solubility data for

isolated solids, computational data and additional

screening results for formation of 9, cocrystal screening

data, and X-ray data for cocrystal 21 (PDF)

.26 (t, J = 7.2 Hz, 1H), 3.47 (s, 2H), 2.99 (s, 2H), 2.86 (d, J =

2.0 Hz, 2H), 2.62 (s, 1H), 2.55 (dd, J = 12.1, 1.8 Hz, 2H),

.05−1.96 (m, 2H), 1.78−1.71 (m, 2H); C NMR (100

MHz, CD CN): δ 141.3, 129.6, 129.1, 127.6, 61.3, 58.5, 52.9,

5.8; HRMS (ESI/Orbitrap) m/z: [M + H] calcd for

AUTHOR INFORMATION

C H N , 203.1543; found, 203.1543.

■

Amide Coupling to Produce 28. To a slurry of acid 27

2.29 g, 18.6 mmol, and 2.2 equiv), cocrystal 21 (5.0 g, 8.45

Corresponding Authors

(

Adam E. Goetz − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States; orcid.org/

mmol, and 1.0 equiv), and N-methylimidazole (1.6 mL, 20

mmol, and 2.4 equiv) in acetonitrile (50 mL) was added

propylphosphonic anhydride (50% w/w in acetonitrile, 12.5

mL, 17.9 mmol, and 2.1 equiv) dropwise over 5 min, during

which the reaction became fully homogeneous. After 2 h, the

reaction was quenched by addition of water (3.0 mL, 20 equiv)

and concentrated on the rotavap to remove any volatiles. The

oil was rediluted with 2-MeTHF (75 mL) and washed twice

with a mixture of 6 N KOH (18 mL, 12.5 equiv) and water (15

mL). The resulting organic solution was dried over magnesium

sulfate, ﬁltered, and concentrated to give a crude oil containing

000-0003-3561-6502 ; Email: adam.goetz@p ﬁ zer.com

Taegyo Lee − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States;

Email: taegyo.lee@p ﬁ zer.com

Authors

Maria S. Brown − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States

Michaella A. Caporello − Chemical Research and

Development, Pﬁzer Worldwide Research and Development,

Groton Laboratories, Groton, Connecticut 06340, United

States; Present Address: Snapdragon Chemistry, Inc., 300

2nd Ave., Waltham, Massachusetts 02451, United States

(M.A.C.).

8 (4.57 g). Analysis of this oil gave a potency of 82.30% for

desired amide 28, which corresponds to a 72% yield. An

analytical sample was obtained by column chromatography

(

CH Cl /MeOH) for characterization purposes. UPLC purity:

2 2

5.9% (a/a); H NMR (400 MHz, DMSO-d ): δ 8.63 (dd, J =

.8, 1.5 Hz, 1H), 8.59 (d, J = 1.5 Hz, 1H), 7.82 (dt, J = 7.8, 1.7

Hz, 1H), 7.46 (dd, J = 7.8, 4.9 Hz, 1H), 7.38 (d, J = 7.5, 2H),

.32 (t, J = 7.5 Hz, 2H), 7.24 (t, J = 7.2 Hz, 1H), 4.22 (d, J =

2.5, 1H), 3.50 (s, 2H), 3.41 (d, J = 12.5 Hz, 1H), 3.22 (br s,

H), 3.17 (d, J = 12.5 Hz, 1H), 3.06 (br s, 1H), 2.97 (d, J =

2.5 Hz, 1H), 2.04 (m, 2H), 1.65−1.55 (m, 1H), 1.49−1.37

Amber M. Johnson − Chemical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

Kris N. Jones − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States

Kevin M. Knopf − Chemical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

Samir A. Kulkarni − Chemical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

Bryan Li − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States; orcid.org/

0000-0002-7972-9966

S_NAr Reaction to Produce 30. To a slurry of

Cuong V. Lu − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States

Javier Magano − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

428

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

A. A.; GWathmey, J. K.; Mullin, S.; van Duzer, J.; Murphy, C. K.

Article

Groton, Connecticut 06340, United States; orcid.org/

Agents Chemother. 1992, 36, 1671−1676. (c) Rothstein, D. M.;

Farquhar, R. S.; Sirokman, K.; Sondergaard, K. L.; Hazlett, C.; Doye,

Efficacy of Novel Rifamycin Derivatives against Rifamycin-Sensitive

and -Resistant Staphylococcus aureus Isolates in Murine Models of

Infection. Antimicrob. Agents Chemother. 2006, 50, 3658−3664.

000-0001-7151-8032

Angela L. A. Puchlopek-Dermenci − Chemical Research and

Development, Pﬁzer Worldwide Research and Development,

Groton Laboratories, Groton, Connecticut 06340, United

States

(d) Lemm, J. A.; Liu, M.; Gentles, R. G.; Ding, M.; Voss, S.;

Giselle P. Reyes − Chemical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

Sally Gut Ruggeri − Chemical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

Lulin Wei − Chemical Research and Development, Pﬁzer

Worldwide Research and Development, Groton Laboratories,

Groton, Connecticut 06340, United States

Gerald A. Weisenburger − Chemical Research and

Development, Pﬁzer Worldwide Research and Development,

Groton Laboratories, Groton, Connecticut 06340, United

States

Pelosi, L. A.; Wang, Y.-K.; Rigat, K. L.; Mosure, K. W.; Bender, J. A.;

Knipe, J. O.; Colonno, R.; Meanwell, N. A.; Kadow, J. F.; Santone, K.

S.; Roberts, S. B.; Gao, M. Preclinical Characterization of BMS-

791325, an Allosteric Inhibitor of Hepatitis C Virus NS5B

Polymerase. Antimicrob. Agents Chemother. 2014, 58, 3485−3495.

(e) Delbary-Gossart, S.; Lee, S.; Baroni, M.; Lamarche, I.; Arnone,

M.; Canolle, B.; Lin, A.; Sacramento, J.; Salegio, E. A.; Castel, M.-N.;

Delesque-Touchard, N.; Alam, A.; Laboudie, P.; Ferzaz, B.; Savi, P.;

Herbert, J.-M.; Manley, G. T.; Ferguson, A. R.; Bresnahan, J. C.;

Bono, F.; Beattie, M. S. A Novel Inhibitor of p75-neurotrophin

Receptor Improves Functional Outcomes in two Models of Traumatic

Brain Injury. Brain 2016, 139, 1762−1782. (f) Fensome, A.; Ambler,

C. M.; Arnold, E.; Banker, M. E.; Brown, M. F.; Chrencik, J.; Clark, J.

D.; Dowty, M. E.; Efremov, I. V.; Flick, A.; et al. Dual Inhibition of

TYK2 and JAK1 for the Treatment of Autoimune Diseases: Discovery

of ((S )-2,2-Difluorocyclopropyl)((1 R ,5 S )-3-(2-((1-methyl-1 H -pyra-

zol-4-yl)amino)pyrimidin-4-yl)-3,8-diazabicyclo[3.2.1]octan-8-yl)-

methanone (PF-06700841). J. Med. Chem. 2018, 61, 8597−8612.

Richard A. Wisdom − Euticals GmbH, 65926 Frankfurt am

Main, Germany

Mengtan Zhang − Analytical Research and Development,

Pﬁzer Worldwide Research and Development, Groton

Laboratories, Groton, Connecticut 06340, United States

(3) Singh, R. K.; Jain, S.; Sinha, N.; Mehta, A.; Naqvi, F.; Agarwal, A.

K.; Anand, N. A Simple and Efficient Synthesis of 8-methyl-3,8-

diazabicyclo[3.2.1]octane (azatropane) and 3-Substituted Azatro-

panes therefrom using Pyroglutamic Acid. Tetrahedron Lett. 2007,

Complete contact information is available at:

https://pubs.acs.org/10.1021/acs.oprd.1c00120

48, 545−548.

Author Contributions

(4) (a) Cignarella, G.; Nathansohn, G. Bicyclic Homologs of

Piperazine. Synthesis of 8-Methyl-3,8-diazabycyclooctanes. J. Org.

Chem. 1961, 26, 1500−1504. (b) Cignarella, G.; Nathansohn, G.;

Occelli, E. J Org. Chem. 1961 , 26 , 2747 −2750. (c) Blackman, S. W.;

Baltzly, R. The Synthesis of 3,8-Diazabicyclo[3.2.1]octane and Some

of Its N-Substituted Derivatives. J. Org. Chem. 1961 , 26 , 2750 −2755.

The manuscript was written by A.E.G. and T.L. All authors

have given approval to the ﬁnal version of the manuscript.

Notes

The authors declare no competing ﬁnancial interest.

d) Schipper, E.; Boehme, W. A Novel Ring System: 3,8-

ACKNOWLEDGMENTS

Diazabicyclo[3.2.1]octane. J. Org. Chem. 1961, 26, 3599−3602.

■

us, A. Org. Prep. Proced.

(

e) Paliulis, O.; Peters, D.; Miknius, L.; Sack

Int. 2007, 39, 86−90. (f) Huang, L. J.; Teng, D. W. Chin. Chem. Lett.

011, 22, 523−526.

5) (a) McDonald, R. N.; Reitz, R. R. Strained Ring Systems. XII.

We are grateful to Chris Morris for HRMS data, Hugh Clarke

for assistance with elemental analysis, and Ivan Samardjiev and

Brian Samas (Pﬁzer, Inc.) for help with X-ray crystallography.

We also acknowledge Matt Weekly, Dave Place, and Jade

Nelson (Pﬁzer, Inc.) for helpful scientiﬁc discussions.

The Synthesis of Several Dimethyl Δ -Cycloalkene-1,2-dicarboxylates

and Certain 4-Substituted Bicyclo[2.1.0]pentane-1-carboxylic acids. J.

Org. Chem. 1972, 37, 2418−2423. (b) Sturm, P. A.; Henry, D. W.;

Thompson, P. E.; Zeigler, J. B.; McCall, J. W. Antifilarial Agents.

Diazabicyclooctanes and Diazabicycloheptanes as Bridged Analogs of

Diethylcarbamazine. J. Med. Chem. 1974 , 17 , 481 −487. (c) Butina,

D.; Sondhemier, F. Syntheses of 1-Cyclopentene-1,2-dicarboxalde-

hyde and 1-Cyclobutene-1,2-dicarboxaldehyde. Synthesis 1980, 543−

545. (d) Yamamoto, Y.; Hoshino, J.; Fujimoto, Y.; Ohmoto, J.;

Sawada, S. A Convenient Synthesis of Enantiomeric Pairs of 2,5-

REFERENCES

■

1) (a) Lovering, F.; Bikker, J.; Humblet, C. Escape from Flatland:

Increasing Saturation as an Approach to Improving Clinical Success. J.

Med. Chem. 2009, 52, 6752−6756. (b) Lovering, F. Escape from

Flatland 2: Complexity and Promiscuity. Med. Chem. Commun. 2012 ,

, 515 −519. (c) Stockdale, T. P.; Williams, C. M. Pharmaceuticals

that Contain Polycyclic Hydrocarbon Scaffolds. Chem. Soc. Rev. 2015 ,

4 , 7737 −7763. (d) Mykhailiuk, P. K. Saturated Bioisosteres of

Benzene: Where to go Next? Org. Biomol. Chem. 2019, 17, 2839−

849. (e) Hanby, A. R.; Troelsen, N. S.; Osberger, T. J.; Kidd, S. L.;

Disubstituted Pyrrolidines of C -Symmetry. Synthesis 1993, 298−302.

(e) Guenoun, F.; Zair, T.; Lamaty, F.; Pierrot, M.; Lazaro, R.;

Viallefont, P. Pyrrolidinopiperazinedione as Chiral Auxiliary and its

use in Asymmetric Mannich Synthesis. Tetrahedron Lett. 1997, 38,

1563−1566. (f) Yu, W.; Vibulbhan, B.; Rosenblum, S. B.; Martin, G.

S.; Vellekoop, A. S.; Holst, C. L.; Coburn, C. A.; Wong, M.; Selyutin,

O.; Ji, T.; Zhong, B.; Hu, B.; Chen, L.; Dwyer, M. P.; Jiang, Y.; Nair,

A. G.; Tong, L.; Zeng, Q.; Agrawal, S.; Carr, D.; Rokosz, L.; Liu, R.;

Curry, S.; McMonagle, P.; Ingravallo, P.; Lahser, F.; Ansante-Appiah,

E.; Fells, J.; Kozlowski, J. A. Discovery of Potent Macrocyclic HCV

NS5A Inhibitors. Bioorg. Med. Chem. Lett. 2016, 26, 3793−3799.

(6) Guha, P. C.; Sankaran, D. K. Muconic Acid. Org. Synth. 1946,

22, 57−1541.

Mortensen, K. T.; Spring, D. R. Fsp -rich and Diverse Fragments

Inspired by Natural Products as a Collection to Enhance Fragment-

based Drug Discovery. Chem. Commun. 2020, 56, 2280−2283.

(f) Bauer, M. R.; Di Fruscia, P.; Lucas, S. C. C.; Michaelides, I. N.;

Nelson, J. E.; Storer, R. I.; Whitehurst, B. C. Put a Ring on it:

Application of Small Aliphatic Rings in Medicinal Chemistry. RSC

Med. Chem. 2021, 448.

(2) The diazabicyclo[3.2.1]octane system is a recurring motif in

bioactive compounds. For selected examples, see (a) Cignarella, G.;

Barlocco, D.; Tranquillini, M. E.; Volterra, A.; Brunello, N.; Racagni,

G. Interaction of 3,8-diazabicyclo(3.2.1)octanes with mu and delta

Opioid Receptors. Pharmacol. Res. Commun. 1988 , 20 , 383 −394.

(7) Willstatter, R.; Lessing, R. Ueber N -Methylpyyolidin-α -α -

dicarbonsaure. Chem. Ber. 1902, 35, 2065−2073.

(8) (a) Cignarella, G.; Testa, E.; Pasqualucci, C. R. Thermal and

(b) Girard, D.; Gootz, T. D.; McGuirk, P. R. Pharmacokinetic Studies

of CP-74,667, a New Quinolone, in Laboratory Animals. Antimicrob.

Chemical Intramolecular N →N Acyl Migration in 8-Acyl-3,8-

429

Org. Process Res. Dev. 2021, 25, 1419−1430

Organic Process Research & Development

Colombo, D.; Toma, L. Mono- and Disubstituted-3,8-

Article

Diazabicyclo[3.2.1]octanes. Tetrahedron 1963, 19, 143−148. (b) Bar-

locco, D.; Cignarella, G.; Tondi, D.; Vianello, P.; Villa, S.; Bartolini,

A.; Ghelardini, C.; Galeotti, N.; Anderson, D. J.; Kuntzweiler, T. A.;

diazabicyclo[3.2.1]octane Derivatives as Analgesics Structurally

Chem. 1998 , 41 , 674 −681. (c) Liu, H.; Cheng, T.-M.; Zhang, H.-M.;

Li, R.-T. Synthesis and Structure-Activity Relationship of Di-(3,8-

diazabicyclo[3.2.1]octane) Diquaternary Ammonium Salts as Unique

Analgesics. Arch . Pharm. 2003, 336 , 510 −513.

(21) (a) Gorman, L. A. S.; Dordick, J. S. Organic Solvents Strip

Water off Enzymes. Biotechnol. Bioeng. 1992 , 39 , 392 −397. (b) Yang,

L.; Dordick, J. S.; Garde, S. Hydration of Enzyme in Nonaqueous

Media is Consistent with Solvent Dependence of its Activity. Biophys.

J. 2004 , 87 , 812 −821. (c) Serdakowski, A. L.; Munir, I. Z.; Dordick, J.

S. Dramatic Solvent and Hydration Effects on the Transition State of

Soybean Peroxidase. J. Am. Chem. Soc. 2006 , 128 , 14272 −14273.

(d) Serdakowski, A. L.; Dordick, J. S. Enzyme Activation for Organic

Solvents Made Easy. Trends Biotechnol. 2008, 26, 48−54.

(22) The addition of a comparable amount of EtOH to the reaction

does not provide the same beneﬁt as water.

9) (a) Lowe, G.; Ridley, D. D. Synthesis of β -Lactams by Photolytic

Wolff Rearrangement. J. Chem. Soc., Perkin Trans. 1 1973 , 2024 −

029. (b) Kemp, D. S.; Curran, T. P. Base-Catalyzed Epimerization

(23) We have explored various strategies for trying to accurately

quantify the amount of free vs bound water in the system, which we

believe is important for further optimization; however, to date, this

has been unsuccessful.

Behavior and Unusual Reactivity of N-Substituted Derivatives of 2,5-

Dicarbalkoxypyrrolidine. Preparation of a Novel Mixed Carbamic

Carbonic Anhydride by a 4-(Dimethylamino)pyridine-Catalyzed

Acylation. J. Org. Chem. 1988, 53, 5729−5731.

(

24) The enzyme itself also contributes water to the system. The

manufacturer speciﬁes a water content of NMT of 5.0% (w/w).

25) Enzyme recycling studies were performed on runs using

(

10) See the Supporting Information for more details.

11) We attribute these inconsistencies to the poor solubility of

signiﬁcantly higher enzyme loadings (∼1 g/g) prior to identifying the

beneﬁcial impact of small amounts of water. When using the

optimized conditions, we observed that the enzyme that had been

subjected to the washing protocol showed decreased performance by

many inorganic bases in the solvents being utilized. As a result,

diﬀerences in the surface area and surface activity of the insoluble base

can translate to diﬀerences in reaction performance. On a small scale,

the grinding impact of magnetic stirrers can obviate these diﬀerences,

which become more apparent at a larger scale.

∼

40−60% in subsequent runs; however, we did not attempt to further

optimize the enzyme recycling for these conditions.

26) Smith, S. C.; Bentley, P. D. Tandem Cyclisation and [2,3]-

(12) (a) Kubyshkin, V. S.; Mikhailiuk, P. K.; Komarov, I. V.

Stevens Rearrangement to 2-Substituted Pyrrolidines. Tetrahedron

Lett. 2002, 43, 899−902.

Synthesis of 7-azabicyclo[2.2.1]heptane-1,4-dicarboxylic acid, a Rigid

Non-Chiral Analogue of 2-Aminoadipic Acid. Tetrahedron Lett. 2007,

(27) Like many amines, the freebase can discolor over time when

8, 4061 −4063. (b) Wu, G.; Wang, D.; Zhu, W.; Shen, H.; Liu, H.;

exposed to oxygen; however, signiﬁcant impurity growth is not

observed with this discoloration.

Fu, L. Synthesis of Bridged Macrocyclic Amino Alcohols as Compact

Modules for Medicinal Chemistry. Synth. Commun. 2019 , 49 , 12 −21.

(28) For recent reviews on cocrystals, see (a) Karimi-Jafari, M.;

(13) Watson, H. A., Jr.; O ’Neill, B. T. A Reinvestigation and

Padrela, L.; Walker, G. M.; Croker, D. M. Creating Cocrystals: A

Review of Pharmaceutical Cocrystal Preparation Routes and

Applications. Cryst. Growth Des. 2018 , 18 , 6370 −6387. (b) Karagianni,

A.; Malamatari, M.; Kachrimanis, K. Pharmaceutical Cocrystals: New

Solid Phase Modification Approaches for the Formulation of APIs.

Pharmaceutics 2018 , 10 , 18. (c) Childs, S. L.; Zaworotko, M. J. The

Remergence of Cocrystals: The Crystal Clear Writing is on the Wall.

Cryst. Growth Des. 2009, 9, 4208−4211.

Improvement in the Synthesis of meso -2,5-Dibromoadipates by

Application of Le Chatelier ’s Principle. J. Org. Chem. 1990, 55,

(

950−2952.

14) Given the biphasic conditions, adequate mixing is needed to

obtain high selectivity and conversion, although the reaction does not

appear to be highly sensitive to mixing eﬀ ects.

(15) (a) Gotor-Fernandez, V.; Gotor, V. Enzymatic Aminolysis and

Ammonolysis Processes in the Preparation of Chiral Nitrogenated

Compounds. Curr. Org. Chem. 2006, 10, 1125−1143. (b) Lima, R. N.;

dos Anjos, C. S.; Orozco, E. V. M.; Porto, A. L. M. Versatility of

Candida Antarctica Lipase in the Amide Bond Formation Applied in

Organic Synthesis and Biotechnological Processes. Mol. Catal. 2019,

(29) Abramov, Y. A.; Loschen, C.; Klamt, A. Rational Coformer or

Solvent Selection for Pharmaceutical Cocrystallization or Desolvation.

J. Pharm. Sci. 2012, 101, 3687−3697.

(30) Loschen, C.; Klamt, A. Solubility Prediction, Solvate and

Cocrystal Screening and Tools for Rational Crystal Engineering. J.

Pharm. Pharmacol. 2015, 67, 803−811.

(

66, 75−105.

16) Starting ester 9 and the ﬁnal amine 11 are both meso

(

31) 4,4′-Dihydroxybiphenyl has been reported as a potential skin

compounds and therefore achiral. Enzymatic amidation results in

desymmetrization to give amide 10, which is a chiral molecule. Since

this chirality is inconsequential to the ﬁnal product, we did not

monitor the enantioselectivity of the enzymatic transformation.

sensitizer and should be handled with appropriate safety precautions.

32) (a) Asrar, J.; Toriumi, H.; Watanabe, J.; Krigbaum, W. R.;

(

Ciferri, A.; Preston, J. Thermotropic Homopolyesters 1. The

Preparation and Properties of Polymers Based on 4,4’dihydroxybi-

phenyl. J. Polym. Sci., Part B: Polym. Phys. 1983, 21 , 1119 −1131.

(17) (a) De Zoete, M. C.; Kock-Van Dalen, A. C.; van Rantwijk, F.;

Sheldon, R. A. A New Enzymatic Reaction: Enzyme Catalyzed

b) Krigbaum, W. R.; Watanabe, J.; Ishikawa, T. Thermotropic

Ammonolysis of Carboxylic Esters. Biocatalysis 1994, 10, 307−316.

Polyesters 2. Investigation of the Mesophase Properties of Polymers

(b) Slotema, W. F.; Sandoval, G.; Guieysse, D.; Straathof, A. J. J.;

Based on 4,4 ′-Dihydroxybiphenyl. Macromolecules 1983, 16, 1271−

Marty, A. Economically Pertinent Continuous Amide Formation by

Direct Lipase-Cataltyzed Amidation with Ammonia. Biotechnol.

Bioeng. 2003 , 82 , 664 −669. (c) Gill, I.; Patel, R. Biocatalytic

Ammonolysis of (5 S )-4,5-dihydro-1 H -pyrrole-1,5-dicarboxylic acid,

(

279.

33) 4,4′-Dihydroxybiphenyl is a key monomer used to synthesize

polymers such as polyphenylsulfone (PPSU, PPSF, Radel) as well as

an antioxidant used in the processing of rubber.

-(1,1-dimethyl)-5-ethyl ester: Preparation of an Intermediate to the

Dipeptidyl Peptidase IV Inhibitor Saxagliptin. Bioorg. Med. Chem. Lett.

006, 16, 705−709.

18) In the case of methanol, the higher reactivity of methyl esters

(

34) See the Supporting Information for additional solubility data.

35) We believe that this color is at least partially related to the

(

ingoing quality of dibromide 1; however, the exact factors controlling

its appearance are not known.

resulted in formation of the corresponding bis-amide product,

whereas the isopropyl ester was inert to the desired amidation

reaction.

(36) In some cases, small amounts of inorganic salts have been

detected in the ﬁnal material. In this case, a water reslurry can easily

remove these, as the targeted cocrystal is essentially insoluble in water.

(19) Monitoring the reaction by GC shows the presence of both

hexamethyldisiloxane (HMDSO) and ethoxytrimethylsilane (EtO-

SiMe ) in addition to HMDS added.

(20) For more details, see Basso, A.; Froment, L.; Hesseler, M.;

Serban, S. Eur. J. Lipid Sci. Technol. 2013, 115, 468−472.

430