Flow Chemistry at Pfizer

Paul Richardson Pfizer-La Jolla

10th October 2013

Expanding Chemical Space through Synthesis and Technology

• Making unique molecules with cutting edge synthesis… – Improved ADME properties – Unique IP space

• High-throughput Reaction Optimization Technology

– Enable challenging chemistry – Expedite synthesis – Limit material needs

• Utilize novel technologies to access forbidden chemistries

– Hazardous gas chemistry in batch and plate – Flow Chemistry

• Approach combinations offer a competitive advantage

120uL Reaction Volume

Pfizer-La Jolla – “Continuous” and “Segmented” Flow

Range of Complexity and Capabilities.

Classical Meerwein Chlorosulfonylation

• Extremely useful process due to low cost and ready availability of aniline starting materials. Also allows the chlorosulfonyl group to be installed with

precise regio-control.

• Reported by Meerwein in 1957 as a modification of the Sandmeyer Reaction. Original modification generates diazonium salts (often as a slurry) under

acidic conditions, and adds this to saturated solution of SO2 in acetic acid containing 0.2-0.4% CuCl2.

• Suggested that CuCl2 is reduced to CuCl by SO2, which enters the catalytic

cycle to effect the reaction.

NH2

RNaNO2 (aq)

conc. HCl/AcOH

N2+

R

Cl-S

R

OO

Cl

AcOH

CuCl2 , SO2

• Modification in 1960 to use CuCl directly. Careful control of temperature

required throughout the reaction.

• Significant safety concerns due to highly explosive diazonium intermediate, large associated exotherm, and generation of stoichiometric quantities of

nitrogen.

• Reaction has been carried out on scale in batch. Key to dilute diazonium salt generation with acetonitrile to keep this in solution.

Modified Meerwein Chlorosulfonylation

NH2

RNaNO2 (aq)

conc. HCl/AcOH

N2+

R

Cl-S

R

OO

Cl

AcOH

CuCl , SO2

• Need to avoid precipitates.

• Numerous components in the reaction system.

• Will the highly acidic reactions conditions present a problem ?

• How to deliver SO2 to the system ?

• How to effectively deliver the copper catalyst to the system ?

• STABILITY and ISOLATION of the Sulfonyl Chlorides.

Challenges to Developing a Flow Process

NH2

RNaNO2 (aq)

conc. HCl/AcOH

N2+

R

Cl-S

R

OO

Cl

AcOH

CuCl , SO2

Initial Development of Flow Process

• First Generation Process – Replace NaNO2 with t-BuONO for

solubility reasons. – Replace HCl as chloride source with

BTEAC. – Solubilize SO2 in acetonitrile (7.9M

soln). – Amberlyst 21 with CuCl2 (validated in

batch).

• Results – Clean conversion to product observed. – Visible leaching of copper/pressure

fluctuations. – Blocking of frit, slowing of reaction

rate.

Initial Flow – Variation of Cu Source

• First Generation Modifications – Switch to bidentate carboxylic

based ligand to hold copper more tightly.

• Results – Clean conversion to product

observed. – Reduces leaching. Still slight

pressure errors. – Copper is blocking frit. Switch to

glass wool. – Pressure is constant. However,

reaction slower, and by-products. Conclude active copper species is soluble, and is now being washed away.

NO

Cu(II)O

O

O

Evaluate Homogeneous Catalysis……

• CuCl2 is poorly soluble in organic solvents (particularly acetonitrile !!) • Nitrogen ligands evaluated. Either ineffective for solubilization or generated

precipitates. • tBuBOX with Cu(OTf)2 is homogeneous. Leads to low yield of impure

products (26%). • Ethylene glycol a good copper ligand. 0.25 equiv of CuCl2 dissolves in 2

equiv of EG in MeCN. Require twice as much water to dissolve this amount. EG easily removed by aqueous wash.

• How to introduce all components ??

?

Variation of the Chloride Source

i) 0.5mmol substrate, SO2-MeCN (0.6mL, 7.9M, 4.75 mmol), BTEAC, CuCl2, ethylene glycol (1 mmol), tBuONO (1.5 mmol), DCM-MeCN, 0 °C to rt, 30 min ; (ii) morpholine (1.5 mmol), DCM, rt, 30 min

NH2

S OO

N

CN

CN

S OO

Cl

CN

O

SO2, CuCl2, glycolBTEAC, tBuONO

DCM/MeCN 0 oC

Morpholine

DCM, rt

“Second Generation” Flow Process

• Keys to Success….. – Devised a three channel flow-

set as shown. – Order of mixing events critical

for success of reaction as well as avoiding precipitation.

– Experiments showed room temperature had a detrimental effect on yield

– Pre-combining tBuONO and CuCl2-ethylene glycol also lead to depressed yields.

– Mixing aniline with copper catalyst resulted in precipitate formation after 20 mins.

“Second Generation” Flow Process

• Keys to Success….. – Collect output of reactor. Key to

evaporate excess SO2 prior to work-up for good yield. Also avoids decomposition of sensitive sulfonyl chlorides.

– Work-up simple water wash to remove minor acidic impurities.

– System works extremely well for certain substrates (4-nitrophenylamine – 81% yield).

– HOWEVER, for other substrates, precipitation immediately occurs on mixing Solutions A and B (4-aminobenzonitrile).

– What causes this precipitation, and how can we avoid it ??

Accepted Mechanism for the Meerwein Reaction

Aryl Diazonium Chloride

Diazenyl Radical

Aryl Radical

Oxidation/Reduction of Copper Species drives

the Catalytic Cycle.

Proposed Alternative Mechanism……

Triazene formed by reaction of diazonium species with amine starting material. Can Isolated

as precipitate, and shown to be stable, by mixing solution A and B in batch for

4-aminobenzonitrile.

Neutral Conditions

Triazenes are known to generate aryl

radicals at elevated temperatures with tBuONO, but here

believe breakdown is promoted by HCl

Monitoring of Triazene formation using ReactIR Flow Cell

Aniline/BTEAC/SO2

CuCl2/Ethylene Glycol tBuONO

OPRD, 2010, 10, 393

“Optimized” Final Process

• Keys to Success – Triazene formation still

leads to desired product but necessary to avoid due to precipitation issues.

– Alter order of mixing of reagents (Solutions A and C now combined initially).

– Precipitates avoided, and reaction proceeds smoothly with a wide variety of substrates.

– Trap products as sulfonamides.

“Optimized” Final Process

• Keys to Success – Reactions run in MeCN.

Possible to use mixtures of DCE or DCM and MeCN.

– Strong acids are avoided. Reaction mixture does become mildly acidic, but only in the PTFE reactor. Easy work-up.

– Easy to scale up by scaling out. Increase flow rate, and number of coils (demonstrated 2g/hr).

NNC S

O

O

N

O

SO

OO2N

NS

O

ONO2

NS

O

OMeO

NS

O

OBr

OCF3

NO S

O

OCF3

Cl

Cl

83% 81% 90%

25% 85% 68%

Org. Biomol. Chem., 2010, 8, 5324

Chlorosulfonylation – Substrate Scope

Further Transformations of Sulfonyl Chlorides using Functionalized Monoliths.

“Continuous” v “Segmented” Flow • Material Sparing.

• Reaction Optimization and Library Synthesis.

• Discovery Chemistry.

• Material Intensive.

• Reaction Scale-up.

• Process Chemistry.

A DISCONNECT ??

CONJURE SEGMENTED FLOW REACTOR

• Segmented Flow Reactor, that can mix up to 4 components into a “segment” as small as 150 μL. Small size reduces risk of “clogging”

• Using a carrier solvent, a reaction “segment” is pushed through a 0.75 mm ID. tube at any temperature between –20°C to 300°C and up to 150 bar.

• Each segment can be analyzed using and inline sampler that takes an aliquot of the segment, dilutes and injects onto an Agilent HPLC.

• Designed and built in collaboration with Pfizer, Wyeth and Accendo.

Feed Working Solvent

Sample/Dilution Analysis/Collection

Reactor Temp. Residence T.

Material 4-36 source vials

Sample mixed. Segment prepared.

Injected.

Initial Chemistry Example

Sonagashira Reaction Screening – 4 Inputs

I+

1) 0.02 eq CuI 0.01 eq PdCl2(PPh3)2

THF, aq 1M NH3 RT, 87%

R2

R2

R1R1

BATCH REACTION

• If solvents degassed reaction is fast.

• Process scale reaction is typically aged for 8-12 hours to drive reaction to completion.

• Heating is difficult due to presence of ammonia.

• Copper-free conditions screened and found less effective.

ISSUES WITH TRANSLATION TO FLOW

• Catalyst and Copper insoluble in all solvents tried.

• At least 5 inputs and only 4 streams

• Volatile Ammonia

Sonagashira Reaction – Flow Screening

I+

Pd(OAc)2(0.02 eq) Triphenylphosphine(0.04 eq)

DMF

BaseAmount of Base

CopperTemperature

R2

R1

R1

R2

Flow Screening Example: • Stream 1: 1.1 eq of Acetylene + Iodide + Triphenyl Phosphine dissolved in DMF (acetylene)

• Stream 2: 2 different Copper amine complexes

• Stream 3: 3 different bases + 2 different base amounts. Bases Screened: DABCO, Piperidine, and Ammonium Hydroxide (DMF:Water solutions)

• Stream 4: Pd(OAc)2 dissolved readily in DMF, and was added separately.

Overall Screened: 3 coppers(no copper added) X 3 bases X 2 ratios of base X 2 Temperatures + 2 control replicates = 38 reactions

38 experiments 10 minute residence time 2 temperatures

Sonagashira – Flow Screening

I+

Pd(OAc)2(0.02 eq) Triphenylphosphine(0.04 eq)

DMF

BaseAmount of Base

CopperTemperature

R2

R1

R1

R2

Phenylacetylene Iodide

Product

PPh3

Reaction solvent

Sonagashira – Flow Screening

0 10 20 30 40 50 60 70 80 90 100

1

4

7

10

13

16

19

22

25

28

31

34

37

Expe

rimen

t #

Product Area %

Is external Copper source necessary with a Copper Coil Reactor ?

Copper Reactor-Mediated Sonagashira

• Issues

• Butyne b.p. 8°C • Pressurized system • Difficult in batch • Selectivity challenges

• Flow Experimental

• Accendo optimization

completed in 2hrs • Demonstrates 125°C

for 4 mins as optimal for conversion

NN

I

Br HN

NBr

DIPEAEtOH/Dioxane

"Cu"

+Pd(PPh3)4

Copper Reactor-Mediated Sonagashira

Conclusions Scalable process

(24g/day) Cu reactor removes

requirements for Cu additive

1.2g delivered to project team

NN

I

Br HN

NBr

DIPEAEtOH/Dioxane

"Cu"

+Pd(PPh3)4

Knife edge optimal conditions 275°C, 2000psi, 15 minutes >50% product

Conclusions Key template accessed IP free chemical space

Flow Decarboxylation – Accessing New Chemical Space

• Issues

• Extreme temperatures • Extreme pressures • Not possible in batch

p

21.9291

30.1321

38.3351

38.3351

46.5382

46.5382

54.7412

54.7412

4Sharp

Decrease in Yield

Sharp Decrease in Yield

Flow Experimental Flow DOE optimization

completed in 1hr 15 experiments

300°C 200°C Temp

5 min

15 min

Time

NN

OH

OH

O

NN

OH

IPA

-CO2

• Azide Formation – Can we prepare low MW azides

in-situ from halides and NaN3 in flow ?

• The click reaction is known to work

with extremely low concentrations of Cu.

1,4-Triazole Click Chemistry in Flow • Conditions

– DMF / Water used for NaN3 solubility and solubility of triazole products

• 2hr Accendo Optimization

5min 175°C

>75% y

Advanced Synthesis & Catalysis 2009, 351, 849

NaX+OHBr

N

NN

OH

0.25M NaN3

(1eq.)DMF / H2O 10:1150oC / 5min

Application to Medicinal Chemistry Projects

N

N NO

ONH2

Cl

Cl

O

N

N NO

ONH2

Cl

Cl

O

NN

NH

N

N NO

ONH2

Cl

Cl

O

NN

N

N

N NO

ONH2

Cl

Cl

O

NN

N

O

N

N NO

ONH2

Cl

Cl

O

NN

N

N

N

N NO

ONH2

Cl

Cl

O

NN

N

ONH2

N

N NO

ONH2

Cl

Cl

O

NN

N

O

NO S

Cl

N

0.5M NaN3 (1eq.)

DMF / H2O 10:1 150°C / 5min

Flow Cu Reactor

NO S

Cl

N

NN

N

NO S

Cl

N

NN

N

OH

O

NO S

Cl

N

NN

N

NO S

Cl

N

NN

N

O

NO S

Cl

N

NN

N

NH

O

NO S

Cl

N

NN

N NH2

O

R1 X

R2

X=Cl, Br, I

Click chemistry is extremely mild and functional group tolerant. Over 500 triazoles have been

prepared across 30 projects over the past 5 years WHAT ABOUT 1,5-TRIAZOLES ?

Ruthenium-Catalyzed Azide-Alkyne Cycloaddition

JACS, 2008, 130, 8928

New C-N formed in step B between

more electronegative and less sterically-

demanding alkyne carbon and terminal

nitrogen of the azide.

Regioselective Oxidative Coupling

Displacement of Spectator Ligands Product Release

Reductive Elimination

What about 1,5-Triazole Click Chemistry in Flow ?

• Using typical conditions for 1,4-Click reaction, no desired product is observed.

• NaN3 is found to poison Ru catalyst. • Ru catalyst is not stable to H2O or O2 (needs to be made up

fresh daily) • How to generate azide in situ ?

JACS, 2008, 130, 8923 ; JOC, 2011, 76, 2355

Attempting to adapt learnings from the 1,4-system

R

NH2N

R

NH2N

NN

N

R1R2Ruthenium Catalyst (0.1eq.)

R1 R2

N3 How to generate in

situ ?

What about 1,5-Triazole Click Chemistry?

• Due to catalyst poisoning, 2 steps are required. • Excess halide ensures sodium azide is consumed.

• Azide stock solution is isolated (albeit small scale and prepared on demand), and passed back through the flow reactor.

• Success rate of 19/22 halides/tosylates.

JOC, 2011, 76, 2355

DON’T USE THE COPPER REACTOR !!!!

R

NH2N

i) Flow Cycle 1Alkylating Agent (1eq.)

NaN3 (0.83eq.), 150C, 5mins R

NH2N

NN

N

R1R2ii) Flow Cycle 2

Ruthenium Catalyst (0.1eq.)100C / DMAc

R1 R2

X

X=, OTs, Cl, Br, I

Batch Screening to Enable Flow Processes

Initial reaction carried out in batch with

10mol% AuCl3 in batch – Isolated yield from reaction was 44-

57% (0.56mmol scale) Scaled to 0.3mol (x50)

– Significant deterioration in yield – Reluctance to heat such an energetic

mixture

• Can Flow Help (mixing/heating) ?

• Flow translation issues

– Reaction forms a thick reddish paste using Au conditions

– Reaction is sluggish and stalls after 20hrs

• How to find alternative conditions? – Reaction screening in

flow remains a challenge…

N+

R

NH2I-

N

N

SiR1

NN

N

N

R1

R

K2CO3, DMF25C, 18h

Batch 3+2 Cycloaddition Screen

• Batch high-throughput experimentation (HTE) screen examined 16 bases

x 6 solvents = 96 reactions • Strong base required to liberate acetylene (KHMDS, NaHMDS, KOtBu)

• Best conditions utilize KHMDS although several ‘flow applicable’ conditions identified

• Side note – Phosphazene bases have huge potential in flow

Mass Ion Count

%Area UV

PNN N

N

Vapourtec R4 Scale-Up

• Initially KHMDS / Dioxane chosen as reaction solvent.

• Abandoned as pyridine salt showed poor solubility and blocked system

• Switched to KHMDS / DMF • 15 mins at 50°C shows complete

disappearance of starting materials and >60% conversion to product by LC/MS

• 57% isolated yield as yellow solid. 1.5 days total work for study, scale-out and isolation

• Issues though demonstrate the problem with direct TRANSLATION of batch reaction screening to flow processes

Macrocycles in Drug Discovery Today…..

J.Med.Chem., 2011, 54, 1961

Nature Reviews Drug Discovery,

2008, 7, 608

The Challenge of Macrocyclization

Reactivity profile for Ring Formation

• Worst cases 8-11 membered rings (macrocycle defined as at least one large ring with > 12 atoms).

• Dependence on enthalpy and entropy.

• Reduction of entropy responsible for conformation restrictions in final molecules comes at a price during synthesis.

• For medium rings, entropy < enthalpy. For large rings, entropy > enthalpy.

• Classic techniques to increase yield. Slow addition, and high dilution.

• Conformation also plays a key role in efficiency of ring closure. • Other strategies to increase efficiency of reaction.

Acc. Chem. Res. 1981, 14, 95

Strategies for Macrocyclization

Br

N

Br

X X

N

Y

Ar

NBr

Y

BrN

Ar

Ar

Ar Ar

+ Ar

Ar

Ar

Aryl-Aryl bond formation

Aryl-Aryl bond formation

Condensationcoupling

Condensationcoupling

< 5%

20-40%

10-50%

60-70% • Successful route requires additional protection steps. • Stereospecific reactant functionalization is required.

•Condensation is amide bond formation. • Variable success of aryl-aryl bond formation.

• Formation of 12 membered rings.

WO 2013132376

Nitrile Directing C-H Activation

Org. Lett., 2011, 13, 1286

Can we take advantage of our nitrile in our lead series ?

Alternative Strategy for Macrocyclization

Br

N

Br

X X

N

Y

Ar

NH

Y

HN

Ar

Ar

Ar Ar

+ Ar

Ar

Ar

Aryl-Aryl bond formation

Aryl-Aryl bond formation

Condensationcoupling

Condensationcoupling

C-H Activation

< ????

20-40%

10-50% with bis-halide

60-70%

• Avoids unneccesary functionalization. • High yielding initial condensation.

• Potentially greater chemoselectivity. • Catalytic, Greener, Approach.

• More Efficient. • WILL IT WORK ??

Initial C-H Activation Ring Closure

• Initial conditions based on literature reactions. • Major by-product is deshalogenation – challenging

analysis. • Capricious reaction – microwave effect ??

• DMA as the solvent in final step is not ideal. • Can yield/robustness of the reaction be improved ??

Br

NH

NAr

Ar

Ar

ArPd(OAc)2, CataCXIUM A

KOAc/PivOHDMA/120 oC/uwave

1 hr / 0.07M 30-46%

C-H Activation Ring Closure Screen

Screen shows reaction

currently being run in the

worst solvent.

SOLVENTS

t-AmOH DMA

Further C-H Activation Screening…….

• Screen 2 (34 reactions) – Part 1 - DOE examining effect of T(100°, 120° and 140°C)

and H2O content (0, 1 and 10eq.) – Part 2 – Re-examination of solvent – Part 3 – Alternative bases (SOLUBILITY !!!)

• Results

– All DMAc reactions with no H2O failed (<10% by UV). With 1eq. H2O yield improves to 25%

– t-AmOH still preferred solvent – KOAc preferred base – 140°C > 120°C > 100°C – Is PivOH required ??

Potential Influence of Water

min0.1 0.2 0.3 0.4 0.5 0.6 0.7

mAU

0

500

1000

1500

DAD1 H, Sig=330,2 Ref=off (1097\1097 2012-02-24 11-34-41\1BJ-2201.D)

0.2

87

32

1 -

Pro

duct

MW

408+

ve

min0.1 0.2 0.3 0.4 0.5 0.6 0.7

0500000

10000001500000200000025000003000000

MSD1 TIC, MS File (C:\CHEM32\2\DATA\1097\1097 2012-02-24 11-34-41\1BJ-2201.D) ES-API, Pos, Fast Scan, Frag: 70, "+v

0.1

91

0.2

12

0.2

40

0.2

57

0.2

74

0.2

95

0.3

29

0.3

79

0.4

01

0.4

20

0.4

49 0

.462

0.4

81

0.5

16 0.5

41 0.5

71

0.6

09

0.6

27 0

.640

0.6

78

0.7

05

0.7

45 0

.755

0.7

95

min0.1 0.2 0.3 0.4 0.5 0.6 0.7

0

500000

1000000

1500000

2000000

MSD1 408, EIC=408:409 (C:\CHEM32\2\DATA\1097\1097 2012-02-24 11-34-41\1BJ-2201.D) ES-API, Pos, Fast Scan, Frag: 70,

33

0 -

Pro

duct

MW

408+

ve

min0.1 0.2 0.3 0.4 0.5 0.6 0.7

0

20000

40000

60000

80000100000

120000

MSD1 411, EIC=410:412 (C:\CHEM32\2\DATA\1097\1097 2012-02-24 11-34-41\1BJ-2201.D) ES-API, Pos, Fast Scan, Frag: 70,

33

0 -

Pro

duct

MW

408+

ve

99%, 20:1 selectivity

UV

TIC

Extract 410

Extract 408

Co-elution of peaks an issue –

Challenging Analysis

Why such a problem ?

1 equivalent of water

EQUIVALENTS OF WATER

Potentially explains the capricious nature of the reaction. Screen under glovebox conditions.

Modified C-H Activation Conditions

• Robust procedure (8:1 required/des-bromo). • Initial scale-up reaction overnight in a pressure vessel at 120 °C

(63%). • Further scale up – move to a higher boiling alcohol or investigate a

Flow process. • Challenges for Flow are (i) heterogeneous system, and (ii) reaction

kinetics/selectivity.

• Replace insoluble KOAc (initially attempted to use water as a co-solvent) with Cesium Pivalate (CsPiv) and perform proof of concept

(microwave) followed by iterative optimization in Flow.

Br

NH

NAr

Ar

Ar

ArPd(OAc)2, CataCXIUM A

KOAc/PivOHt-AmOH/140 oC

uwave or thermal1 hr / 0.67M

69 %

Flow Optimization – Selected Results

Pfizer Confidential │ 47

Pd (eq)

Ligand CsOPiv T° (C)

Time (min)

B (%) A (%) C (%) D (%) Comments

0.1 0.2 3 150 50 61 36.4 2.6 0 < 5 eq base gave lower yield

0.1 0.2 5 190 50 68.4 0.3 8.8 22.5 Temp > 150˚C gave more by-product D

0.1 0.2 5 140 50 50.6 47.3 2.1 0 Temp < 150˚C gave lower yield

0.1 0.2 5 140 100 57.4 39.7 2.7 0.2 Longer residence increased yield

0.1 0.2 5 150 50 72.9 20.7 3.9 2.5 150˚C optimal temp

0.15 0.3 5 150 50 91.6 3.7 2.8 1.9 Higher catalyst loading increased yield

(Optimized conditions)

Br

NH

Ar

Ar

N

ArAr

N

HAr

Ar

Ar

Ar

(A) (B) (C) (D)

Pd(OAc)2CataCXiumA

t-AmOHCsPiv

Desired

+ +

Key to use t-AmOH as carrier solvent

Flow Macrocyclization – Initial Scale-up

• Reaction demonstrated for two compounds from different lead series. Discrete optimization for each series. Both scaled up in

Flow. • Minor precipitation of Pd black observed in flow coils.

• Solid-supported Pd demonstrated to be effective in microwave, but experiment in flow using a tube reactor failed (Pd-Si).

• Key further optimization is to lower Pd/Ligand loadings in order to make this an economically viable approach.

D C

B

A

500mg Scale 65% Isolated Project Lead

N

ArAr

Acknowledgements

• Neal Sach • Judy Deal • Andrew Bogdan • Laia Malet-Sanz • Julia Madrzak • Steve Ley • Ian Baxendale • Heiko Lange • Catherine Carter • Jason Hein • Simon Bailey

• Joel Hawkins • Jan Hughes • Terry Long • Kristin Price • Larry Truesdale • Robert Tinder • Mike Collins • John Braganza • Bryan Li • Jen Lafontaine

• Martin Edwards