Rapid improvement of a reductive sulfonylation using design of experiment methods

Article abstract of DOI:10.1021/op034018g

Sulfone 2 was initially prepared using a zinc-mediated one-pot coupling procedure. A traditional sodium sulfite-mediated two-pot procedure was found to eliminate a major impurity, yet modest yields prevailed. Design of experiment methods with the aid of a basic parallel synthesizer rapidly led to a high-yielding one-pot process.

Full text of DOI:10.1021/op034018g

Organic Process Research & Development 2003, 7, 313−317
Rapid Improvement of a Reductive Sulfonylation Using Design of Experiment
Methods
Jiong Jack Chen,*,§ Thomas C. Nugent,*,| Cuong V. Lu,§ Sudharani Kondapally,| Peter Giannousis,| Ying Wang,§ and
Jeremy T. Wilmot§
Early Process Research and DeVelopment, Pharmacia, Kalamazoo, Michigan 49001, U.S.A., and Chemical DeVelopment,
SUGEN, Inc., South San Francisco, California 94080, U.S.A.
Scheme 1
Abstract:
Sulfone 2 was initially prepared using a zinc-mediated one-pot
coupling procedure. A traditional sodium sulfite-mediated two-
pot procedure was found to eliminate a major impurity, yet
modest yields prevailed. Design of experiment methods with
the aid of a basic parallel synthesizer rapidly led to a high-
yielding one-pot process.
Design of experiment (DoE) has increasingly been applied
to the process development of active pharmaceutical ingre-
dients. Process chemists engaged in the preparation of
preclinical or early clinical supplies have generally shied
away from the use of DoE, due to the perception that DoE
is for nominal optimizations and requires a large commitment
of time and effort. Furthermore, this view is exacerbated by
a general lack of knowledge concerning the fundamentals
of DoE and when it should be applied. In this note, we would
like to dispel some of these perceptions by disclosing an
application of DoE that has had immediate impact on our
early development efforts.
The conversion of sulfonyl chlorides to sulfones had been
studied extensively prior to the1960s. It was traditionally
accomplished via reduction of a sulfonyl chloride, with zinc
or sodium sulfite;1 the resulting sulfinic acid salt2 was then
alkylated to form the sulfone.1 The procedures usually
required two pots, with the sulfinic acid salt as the lone
isolated intermediate. Similar one-pot transformations em-
ploying various metals have been reported in recent years,
e.g., Zn,3 In,4 and Fe.5
ammonium chloride served well to activate the Zn metal in
situ. Simple addition of DMF6 and the sulfonyl chloride (1)
to the aqueous Zn medium allowed sulfinic acid salt
formation within 30 min; finally addition of DCBB allowed
the desired product to form in 6 h. The yields for the one-
pot transformation were consistent (40-45%), and a sulfide
byproduct (3) was always noted in a range of 1.6-2.5 area
% (HPLC). While the yield was low, this was offset by the
low cost of the starting materials and our immediate need
for compound 2. As a consequence the process was pursued
on the tens of grams scale.
However, even on modest scale-up (10-50 g) the yield
(10-30%) and purity (sulfide byproduct 5-11 area %) of
the isolated sulfone decreased significantly. With our process
in question we looked closely at a variety of reaction
parameters. Modification of the workup, namely cooling (5
°C) and addition of fewer equivalents of HCl (0.9 equiv,
2.0 N), reduced the sulfide byproduct to the manageable
range of 2-5 area % (HPLC).7 Encouraging as this was, a
closer evaluation of the process showed it lacked robustness.
For the preparation of an active pharmaceutical ingredient,
we needed quick access to tens of grams and later kilograms
of sulfone (2). After a literature survey we settled on a Zn-
mediated one-pot coupling of a sulfonyl chloride (1) and 2,6-
dichlorobenzylbromide (DCBB) (Scheme 1). During our
initial research, on gram scale, we found that aqueous
* Corresponding authors. (J.J.C) E-mail: jiong.j.chen@pharmacia.com.
Fax: 269-833-9282. (T.C.N.) E-mail: thomas-nugent@sugen.com. Fax: 650-
837-3326.
§ Early Process Research and Development, Pharmacia.
| Chemical Development, SUGEN, Inc.
(1) Drabowicz, J.; Kielbasinski, P.; Mikolajczyk, M. In The Chemistry of
Sulphinic Acids, Esters and Their DeriVatiVes; Patai, S., Ed.; John Wiley
& Sons: New York, 1990; p 351-430.
(2) Zoller, U. In The Chemistry of Sulphinic Acids, Esters and Their
DeriVatiVes; Patai, S., Ed.; John Wiley & Sons: New York, 1990; p 185-
215.
(3) Sun, P. P.; Wang, L.; Zhang, Y. M. Tetrahedron Lett. 1997, 38, 5549.
(4) Sun, X.-H.; Wang, L.; Zhang, Y.-M. Synth. Commun. 1998, 28, 1785.
(5) Saikia, P.; Laskar, D. D.; Prajapati, D.; Sandhu, J. S. Chem. Lett. 2001,
512.
(6) To our knowledge this is the first reported use of DMF for this type of
transformation. When the typical solvent THF was employed, g1 day was
required for complete reaction.
(7) The extreme insolubility of the product 2 and the sulfide 3, ruled out the
use of traditional crystallization scenarios. Ultimately we found that
enrichment could be effected by heating the sulfide-rich sulfone in an
EtOAc/DMF (2:1) mixture for 15 min at 60 °C, followed by hot filtration.
When repeated twice, this allowed a 70 wt % recovery with a 50% reduction
in the sulfide content.
10.1021/op034018g CCC: $25.00 © 2003 American Chemical Society
Published on Web 03/28/2003
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
313

Scheme 2
Table 1: Results of 23 full factorial design
factors
response
standard
order
A: sulfite amt
(equiv)
B: time
(h)
C: temp
(°C)
yielda,b
(%)
1
2
3
4
5
6
7
2
4
2
4
2
4
2
4
4
6
6
24
24
6
60
60
60
88
42
93
49
87
58
75
31
48
60
100
100
100
100
60
6
24
24
6
8
2*c
a Step 1: All the ingredients (1-mmol scale) were shaken in water (2.0 mL)
at the stated temperature and time. b Step 2: Mixture from step 1 was cooled
to 60 °C, and DCBB (240 mg) and acetone (2 mL) were added and stirred for
1 h. The mixture was then poured into water (14 mL), cooled to room
temperature, and filtered to give compound 2. Yields were corrected for potency.
c Duplicate of 2. The difference of 6% in yield is considered as the error for
these runs.
Seemingly subtle adjustments to a large number of reaction
parameters (mode of agitation, slight increases in alkylating
agent, concentration of aqueous NH4Cl, etc.) dramatically
lowered the yield and quality of the isolated sulfone. The
optimal conditions found at the 40-g scale provided a 39%
yield of the sulfone, contaminated with 1.9 area % of the
corresponding sulfide when magnetic stirring8 was employed.
At this juncture, we decided to pursue alternative methodol-
ogy for this coupling.
We turned our attention to the two-pot procedures,
particularly the ones using Na2SO3.1 Sodium sulfite is water-
soluble and has inherently lower reductive potential than Zn.
This was important to us, if the sulfide byproduct was to be
mitigated. Using the traditional reaction conditions,9 the
initial trial was quite promising. Treatment of compound 1
with Na2SO3 and sodium bicarbonate in water provided the
presumed sulfinic acid salt intermediate in 73% yield
(Scheme 2). When this intermediate was reacted with DCBB
in DMF, compound 2 was isolated in 78% yield. The 53%
overall yield was a modest improvement over the Zn
procedure, but importantly, no sulfide impurity was observed.
However, several problems did exist: (1) extreme foaming
was noted when the sulfonyl chloride and sodium bicarbonate
were mixed, (2) filtration of the intermediate was very slow,
due to its small particle size, and furthermore (3) the one-
pot sequence, without intermediate isolation, only resulted
in an overall 30% yield.
thereby avoiding gas evolution. When stronger bases were
employed, a side reaction became dominant and resulted in
much lower yields of the sulfone. Presumably, the side
reaction resulted from reaction at the acid labile 3-position
of the oxindole.
During our solvent screen for step 2, we also gained
further understanding for the alkylation step, noting it was
facile and unlikely to be problematic. Thus, step 1, the
reduction, would be the focal point for improving the yield.
Consequently, it would be possible to use a one-variable-
at-time (OVAT) approach. However, the time required to
develop a successful process was of critical importance to
our overall goals. Furthermore, OVAT may not identify the
interdependence of significant reaction factors, i.e., interac-
tions. Therefore, we decided to employ a DoE method with
the aid of a parallel synthesizer to accelerate identification
of significant factors and the possible interactions between
them. The ultimate goal was to define a simple and high-
yielding one-pot procedure.
On the basis of our initial data and intuition, we identified
three factors likely to have the greatest impact on step 1.
They were equivalents of sodium sulfite, reaction tempera-
ture, and reaction time. We did not consider the amount of
base as a factor, since in theory only 1.0 equiv is required.
A total of nine reactions were performed (reaction 9 was a
duplicate of reaction 2 to gauge reproducibility) in a parallel
synthesizer10 (Table 1). The resulting yields were entered
into the Design Expert software.11 The software analyzed
the data and yielded several key plots. The normal probability
plot (Figure 1) and ANOVA12 showed that factor A (amount
of sodium sulfite) and interaction BC were significant. Next,
examination of a one-factor plot (Figure 2) revealed that less
sodium sulfite (factor A) should allow greater yield. The
interaction graph of BC (Figure 3) showed that reactions at
We still preferred a one-pot procedure for scale-up.
However, before initiating a study of the one-pot process, a
few concerns needed to be addressed. The use of DMF was
undesirable considering the moderate solubility of product
in this solvent. A solvent screen for step 2 identified acetone
as a convenient and environmentally friendly replacement
for DMF. In addition, we replaced NaHCO3 with Na2HPO4,
(8) Zinc dust (Aldrich Chemical Co., catalog no. 20, 998-8) aggregation was
noted when overhead mechanical stirring was employed as the mode of
stirring, and low yields (10-30%) resulted. The agglomeration phenomena
observed with the zinc dust is relatively common. Magnetic stir bars have
the advantage of physically grinding agglomerates against the face of the
glass wall. This is not possible with an overhead stirrer. As a consequence,
any reaction requiring physical action to liberate occluded material will
not scale well unless something is used to provide this action. On scale-
up, depending on the strength of the agglomerate, the use of a recirculating
loop with a rotor/stator mill or a grinding sand mill generally corrects the
problem. We would like to thank John VanAlsten of Pharmacia for his
input regarding these insights.
(10) We used a parallel synthesizer, LabMate, that is made by Advanced
ChemTech Inc. It has four temperature zones, six positions for each zone,
and uses shaking for mixing.
(11) Design Expert, version 6; DoE software developed by Stat-Ease, Inc. The
software is intuitive, but some basic knowledge of DoE is desirable. Our
company statisticians provided an overview in a 1-h demonstration.
(9) Field, L.; Clark, R. D. Organic Syntheses; Wiley: New York, 1963; Collect.
Vol. IV, p 674.
314
•
Vol. 7, No. 3, 2003 / Organic Process Research & Development

Figure 1.
Figure 3.
Table 2: Solubility of inorganic salts
equiv at start equiv at end
mol/L
mol/L
salt
NaCl
Na2SO3
Na2SO4
Na2HPO4
NaH2PO4
KCl
K2SO3
K2SO4
K2HPO4
KH2PO4
of reaction of reaction (0 °C or 20 °C) (100 °C)
0
2.0
0
1.0
0
0
2.0
0
1.0
0
1.0
1.0
1.0
0
1.0
1.0
1.0
1.0
0
6.1
1.0
0.33
0.39
0.44
4.6
5.1
0.69
9.6
6.7
2.2
2.9
0.66
3.1
7.6
5.1
1.4
>9.6
6.1
1.0
2.4
DCBB. These facts led us to further speculate that the two
saturated phases could generate a biphasic reaction mixture
that may have a poor mixing profile. However, the corre-
sponding potassium salts are generally more soluble than
the sodium salts (Table 2), and we felt might give superior
results in this reaction.
Figure 2.
high temperature required less time than those at lower
temperature.
There are a number of factors in step 2 with potential
mass-transfer implications, such as reaction concentration,
counterion (Na or K), phase-transfer catalyst (PTC), solvent,
and reaction time. Time constraints would not allow these
parameters to be studied using the one-variable-at-a-time
approach. On the other hand, a fractional factorial design
could potentially identify at least some significant factors.
We decided on a 28-4 study of a one-pot procedure that
incorporated most mass-transfer factors plus the factors in
the earlier study. We opted to use this low-resolution design
due to limited resources. Our goal was to identify if any
mass-transfer factors were significant (Table 3).
From the normal probability plot (Figure 4) and ANOVA,
the AC interaction is significant (AC is confounded with the
BG, DF, and EH interactions). None of the main effects that
have mass transfer implications were actually significant,
such as concentration (A), counterion (E), presence of phase-
transfer catalyst (F), reaction time of step 2 (G), and solvent
type (H). The lower yield in one of the 10-mmol scale-ups
may be simply because of a poor magnetic stirring in an
isolated incident, as we had speculated. This analysis gave
us the confidence that mass transfer should not be an
Although an optimization using more advanced DoE
methods, such as the response surface method (RSM) may
further improve yields, the results from experiments 1 and
3 (Table 1) appeared to be a reasonable starting point from
a process chemist’s perspective. Thus, experiment 1 was
repeated on a 10-mmol scale using normal laboratory
glassware and magnetic stirring. It provided sulfone (2) in
85% yield. However, a similar 10-mmol scale-up of experi-
ment 3 resulted in a much lower yield (70%), probably as a
result of inefficient magnetic stirring. This led us to speculate
that mass transfer was an issue for this reaction. We then
took a closer look at the solubility of the inorganic salts
(Table 2) that were charged and generated during the
reaction. It revealed that the aqueous reaction mixture was
saturated with inorganic sodium salts. At the same time, the
organic layer was a very concentrated acetone solution of
(12) The definition of these terminologies, such as normal probability plot, factor,
interaction, etc. may be found in any statistics or experimental design books,
such as DOE Simplified: Practical Tools for EffectiVe Experimentation,
written by the developers of Design Expert. We found a contribution
published in this journal very informative: Owen, M. R.; Luscombe, C.;
Lai, L.-W.; Godbert, S.; Crooke, D. L.; Emiabata-Smith, D. Org. Process
Res. DeV. 2001 5, 308.
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
315

Table 3: Results of 28-4 fractional factorial design
factors
response 1
B: amount
of sulfite
(equiv)
D: time of
step 1
G: time of
step 2
A: concentration
(mmol/mL)
C: temperature
H: solvent
(min)
yield
(%)
run
(°C)
(h)
E: Na or K F: PTC
(h)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
9*
0.25
0.5
0.25
0.5
0.25
0.5
0.25
0.5
0.25
0.5
0.25
0.5
0.25
0.5
0.25
0.5
2
2
2.5
2.5
2
60
60
60
60
80
80
80
80
60
60
60
60
80
80
80
80
60
6
6
6
6
6
6
6
6
24
24
24
24
24
24
24
24
24
Na
Na
K
K
K
no
yes
no
yes
yes
no
yes
no
yes
no
yes
no
no
yes
no
1
2
2
1
2
1
1
2
1
2
2
1
2
1
1
2
1
Acet
DMF
DMF
Acet
Acet
DMF
DMF
Acet
DMF
Acet
Acet
DMF
DMF
Acet
Acet
DMF
DMF
74.2
82.6
60.1
84.6
73.6
79.8
75.8
78.1
72.5
77
71.3
75.8
71.9
68.8
79.8
65.2
73.6
2
K
2.5
2.5
2
Na
Na
K
2
K
2.5
2.5
2
Na
Na
Na
Na
K
K
K
2
2.5
2.5
2
yes
yes
0.25
solution. Chemical shifts are reported as δ (ppm) values from
Me4Si as an internal standard. HPLC was analyzed on an
Agilent 1100 series system. Elemental analysis was con-
ducted in Analytical Service within Pharmacia Corporation.
Preparation of 5-[(2,6-Dichlorobenzyl)sulfonyl]-1,3-
dihydro-2H-indol-2-one (2). Zn-Mediated Sulfonylation. Zn
dust (11.3 g, 17.3 mmol), was added to DMF (420 mL) and
aqueous NH4Cl (105 mL, 2.0 M) and stirred for 15 min;
2,3-dihydro-2-oxo-1H-indole-5-sulfonyl chloride (1) (40.0 g,
17.3 mmol) was then added over 30 min, allowing a
controlled exotherm to occur. The heterogeneous solution
was then stirred for 45 min followed by the addition of 2,6-
dichlorobenzyl bromide (41.4 g, 17.3 mmol) and then heated
at 40 °C for 6 h. The reaction mixture was then diluted with
200 mL of water, cooled to 5 °C, quenched with HCl (40
mL, 2.0 N), and stirred for 10 min. Finally the reaction
mixture was filtered and the cake washed with water (100
mL × 2), EtOAc (40 mL × 2), and MeOH (30 mL). After
air-drying in a large crystallizing dish, 23.50 g of compound
2 (39% yield) was obtained with 1.88 area % of sulfide
byproduct 3. Enrichment of the crude product is possible.7
Sulfite-Mediated Sulfonylation. Na2HPO4 (142 g, 1.0 mol)
and Na2SO3 (252 g, 2.0 mol) were dissolved in 2 L of water
and heated to 30 °C. This solution was then added to
compound 1 (232 g, 1.0 mol). The creamy suspension was
heated at 60 °C (became clear, then suspended) for 16 h.
2,6-Dichlorobenzyl bromide (240 g, 1.0 mol) in 1.8 L of
acetone was added to the above suspension over 60 min with
vigorous stirring at 60 °C followed with an acetone rinse
(200 mL). White solids formed immediately. The mixture
was stirred at 60 °C for 2 h and then quenched into 5 L of
water. The white slurry was stirred at room temperature for
1 h. It was then filtered and washed with water (1 L) and
acetone (1 L), and subsequently dried in vacuo at 60 °C
overnight to give 314 g of compound 2 as a white powder
(88%). 1H NMR (DMSO-d6) δ 7.55-7.50 (m, 4H), 7.40 (m,
1H), 6.96 (d, 1H, J ) 8 Hz), 4.80 (s, 2H), 3.57 (s, 2H).
Figure 4.
obstacle. Thus, by simply increasing the reaction time of step
2 and using vigorous mechanical stirring, we were able to
achieve a 90% isolated yield with >99% purity on a 30-g
scale, with no detectable amount of sulfide impurity 3 noted
(HPLC). Finally, a 300-g scale reaction incorporating all our
findings resulted in an 88% isolated yield.
In summary, with a basic parallel synthesizer and several
simple DoE designs, we were able to develop a reproducible
and high-yielding reductive sulfonylation in less than 3
weeks. Moreover, the final process only uses water and
acetone as solvents, in addition to a benign reducing agent,
sodium sulfite. This study demonstrated that a combination
of DoE and parallel synthesizers can be important tools for
early process chemists.
Experimental Section
General Procedures. Solvents and reagents were used
as received from vendor. 2,3-Dihydro-2-oxo-1H-indole-5-
sulfonyl chloride (1) (CAS 199328-31-9) was supplied by
Austin Chemical Company, Inc. H and 13C spectra were
1
obtained using a Bruker Avance 400 in dilute DMSO-d6
316
•
Vol. 7, No. 3, 2003 / Organic Process Research & Development

HPLC: rt ) 5.41 min (sampled prepared in ACN:DMF )
9:1, 0.5 mg/mL). HPLC conditions: Zorbax RX-C8, 250 mm
× 4.6 mm, UV ) 254 nm, flow rate ) 1.0 mL/min, inject
volume ) 5.0 µL, gradient (ACN:0.1 M NH4OAc): 0 min,
50:50; 7 min, 50:50; 10 min, 100:0; 13 min, 100:0; 15 min,
50:50; 18 min, 50:50. Anal. Calcd for C15H11Cl2NO3S: C,
50.58; H, 3.11; N, 3.93. Found: C, 50.21; H, 3.06; N, 3.86.
Acknowledgment
We thank M.J. Witt and G.L. Larner for their assistance
in statistical analysis.
Received for review January 28, 2003.
OP034018G
Vol. 7, No. 3, 2003 / Organic Process Research & Development
•
317