Second-generation process for the hcv protease inhibitor biln 2061: A greener approach to ru-catalyzed ring-closing metathesis

Article abstract of DOI:10.1021/op800225f

The ring-closing metathesis (RCM) step, a key reaction in our process to BILN 2061, was dramatically improved from the first-generation process by the selection of a more appropriate substrate as well as the use of a more effective catalyst. The two RCM r

Full text of DOI:10.1021/op800225f

Organic Process Research & Development 2009, 13, 250–254
Second-Generation Process for the HCV Protease Inhibitor BILN 2061: A Greener
Approach to Ru-Catalyzed Ring-Closing Metathesis†
Vittorio Farina,*,§ Chutian Shu,* Xingzhong Zeng, Xudong Wei, Zhengxu Han, Nathan K. Yee, and Chris H. Senanayake*
Department of Chemical DeVelopment, Boehringer Ingelheim Pharmaceuticals, Inc., 900 Ridgebury Road,
Ridgefield, Connecticut 06877, U.S.A.
Abstract:
The ring-closing metathesis (RCM) step, a key reaction in our
process to BILN 2061, was dramatically improved from the first-
generation process by the selection of a more appropriate substrate
as well as the use of a more effective catalyst. The two RCM
reactions are compared in detail using criteria that are of high
significance to the process chemist.
Introduction
The ring-closing metathesis (RCM) reaction has recently
emerged as an important tool in organic synthesis.1 We have
described an application of this reaction to the synthesis of HCV
protease inhibitor BILN 2061, a 15-membered macrocyclic
compound containing a (Z)-olefin (Figure 1).2
This first-generation RCM (Scheme 1) was scaled to produce
>100 kg of the active pharmaceutical ingredient (API).3 In this
contribution we describe the optimization of the initial RCM,
leading to a second-generation process which dramatically
improves throughput and reaction efficiency, as measured by a
variety of parameters.
Figure 1. Structure of BILN 2061.
titatively using the concept of effective molarity (EM), which
is convenient in order to formulate the above competition with
one quantifiable parameter.5
In order to place the problem in our initial reaction on
quantitative grounds, we carried out a rough calculation of EM
() kintram/kinterm) for the process employing the first-generation
Hoveyda catalyst 46 (Figure 2), which operates under kinetic
conditions.7
In an effort to design a more effective second-generation
process to this and related chemical targets, the high dilution
of the RCM reaction (10 mM) had to be overcome; this problem
is typical of macrocyclizations, including all RCM macrocy-
clizations known to date. In fact, a high-concentration RCM
macrocyclization was unprecedented in the literature at the
outset of our work, and our first-generation process was actually
already at the high end of the concentration range used in the
published RCM macrocyclizations.4
Figure 2. Catalysts discussed in this study.
The reasons for this problem are easily understood: at higher
concentration, intermolecular processes start competing with
ring closure, thus lowering the yield of the desired cyclization
product. In practice, one is then forced to compromise between
high yield and high throughput. This can be expressed quan-
We made several simplifications. We assumed that the
several cyclic dimers (identified by LC/MS analysis)8 accounted
for the missing mass balance of the reaction and that all dimers
were formed at the same rate (an average rate), given that the
intermolecular step of the dimerization is likely to be the same
or similar. We then carried out the RCM at different diene
concentrations, and estimated the starting material and product
yield by a quantitative HPLC assay. We used the “product ratio”
approach described by Percy et al. to estimate the kinetic EM
† Dedicated to the memory of Chris Schmidt, a friend and esteemed colleague.
* To whom correspondence should be addressed. Telephone: 203-778-7876.
Fax: 203-791-6130. E-mail: chutian.shu@boehringer-ingelheim.com.
§ Current address: Johnson & Johnson Pharmaceutical Research and Develop-
ment, Turnhoutseweg 30, B-2340 Beerse, Belgium.
(1) (a) Grubbs, R. H. Handbook of Metathesis; Wiley-VCH: Weinheim,
2003. (b) Astruc, D. New J. Chem. 2005, 29, 42.
(5) (a) Illuminati, G.; Mandolini, L. Acc. Chem. Res. 1981, 14, 95. (b)
Galli, C.; Mandolini, L. Eur. J. Org. Chem. 2000, 3117.
(6) Kingsbury, J. S.; Harrity, J. P.; Bonitatebus, P. J.; Hoveyda, A. H.
J. Am. Chem. Soc. 1999, 121, 791.
(2) Yee, N. K.; Farina, V.; Houpis, I. N.; Haddad, N.; Frutos, R. P.; Gallou,
F. R.; Wang, X.-j.; Wei, X.; Simpson, R. D.; Feng, X.; Fuchs, V.;
Xu, Y.; Tan, J.; Zhang, L.; Xu, J.; Smith-Keenan, L. L.; Vitous, J.;
Ridges, M. D.; Spinelli, E. M.; Donsbach, K.; Nicola, T.; Brenner,
M.; Winter, E.; Kreye, P.; Samstag, W. J. Org. Chem. 2006, 71, 7133.
(3) Nicola, T.; Brenner, M.; Donsbach, K.; Kreye, P. Org. Process Res.
DeV. 2005, 9, 513.
(7) We have shown that 3 does not re-open to 2 under the reaction
conditions nor does it produce dimer, even under an ethylene
atmosphere or in the presence of an analogous RCM reaction in the
same pot; the main cyclic dimer does not equilibrate to 3 when treated
with catalytic 4.
(4) Gradillas, A.; Pe´rez-Castells, J. Angew. Chem., Int. Ed. 2006, 45, 6086.
250
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development
10.1021/op800225f CCC: $40.75
2009 American Chemical Society
Published on Web 01/14/2009

Scheme 1. First-generation RCM
highly streamlined commercial process, where all downstream
manipulations have been optimized, this step will be the
bottleneck, and therefore further reduce the already meager
throughput.
A fourth problem we have encountered from time to time
is the epimerization reaction at the vinyl-bearing cyclopropane
(P1) carbon atom, due to trace impurities of amines or
phosphines in the reaction medium. Although we have clarified
some of the mechanistic details of the epimerization,10 this
troublesome side reaction requires the most scrupulous quality
control in the RCM solvent as well as in the starting diene,
thus introducing extensive analytical work. Although this can
be done, ideally we would prefer an RCM reaction that could
be made robust under a more permissive range of conditions,
thus improving ease of operation.
Figure 3. Calculation of EM(kin) for reaction in Scheme 1.
Although second-generation catalysts like Grela’s 5 (Figure
2) address rather successfully the second and third problems,
due to their better turnover numbers (TON) and turnover
frequencies (TOF), they operate under thermodynamic condi-
tions instead of kinetic ones,2 producing more products of
intermolecular metathesis.11 This makes it difficult to compare
the efficiency of catalysts 4 and 5 quantitatively. On the other
hand, we wanted to express the concentration sensitivity of the
RCM reaction with catalyst 5 using a single parameter, which
would make comparison with other substrates easy. The
problem of expressing the thermodynamic effective molarity
under conditions where both cyclic and acyclic oligomers can
be formed has been tackled, for the most general case, by
Ercolani et al.12 The thermodynamic EM for cyclization is
expressed as the ratio of two equilibrium constants: one for the
cyclization reaction, the other for a corresponding intermolecular
reaction (e.g., the dimerization event). In order to simplify the
problem, we have only taken into consideration the cyclizations
of the monomer (starting diene) and its main dimer (eqs 2 and
3):
(EM(kin)).9 Expressing the rate equations and replacing the ratio
of the kinetic constants of the two competing processes with
EM, one obtains eq 1. Use of this equations leads to the plot in
Figure 3, and to a EM(kin) value of 0.046 M.
(% RCM) ⁄ (% dimers) ) EM ⁄ [diene]
(1)
This EM represents the initial concentration of diene that
will yield a ca.1:1 mixture of the RCM product and dimers.
This, in practical terms, means that the RCM reaction must be
run well below the EM value in order to obtain quantitative
yields (>90%) of the product resulting from the intramolecular
process. These concentrations are extremely impractical in a
plant setting, especially if the production target for the new API
is in the metric ton (MT) range. In addition, they generate
enormous volumes of spent solvents which must be eventually
incinerated. A “greener” process was therefore needed.
A second problem with the first-generation process was the
high catalyst load. In the early runs, 5-7 mol % of the
expensive catalyst 4 was routinely used. On a pilot-plant scale,
once the starting diene 2 was obtained in high purity, it was
possible to reduce this load to about 3 mol % and still obtain
high yields of 3. This limitation is due to the intrinsically low
stability and low efficiency of catalyst 4. On a commercial scale,
this is impractical, due not only to the high cost but also to the
extreme difficulty of removing the high levels of Ru metal from
the API. Therefore, a better catalyst was absolutely required.
A third problem was constituted by the long reaction times,
often in excess of 24 h and reaching, in some cases, 40 h. In a
diene ) RCM product + ethylene (EM1)
(2)
linear dimer ) cyclic dimer + ethylene (EM2) (3)
This, following Ercolani, leads to eq 4:
[RCM product]2⁄[dimer] ) (EM1)2⁄EM2
(4)
Thus, the ratio [RCM product]2/[dimer] should be constant
for any concentration of starting diene. This ratio is basically
(8) For the isolation and identification of the most abundant dimer, see
the Supporting Information of Shu, C.; Zeng, X.; Hao, M.-H.; Wei,
X.; Yee, N. K.; Busacca, C. A.; Han, Z.; Farina, V.; Senanayake, C. H.
Org. Lett. 2008, 10, 1303.
(10) Zeng, X.; Wei, X.; Farina, V.; Napolitano, E.; Xu, Y.; Zhang, L.;
Haddad, N.; Yee, N. K.; Grinberg, N.; Shen, S.; Senanayake, C. H. J.
Org. Chem. 2006, 71, 8864.
(9) Mitchell, L.; Parkinson, J. A.; Percy, J. M.; Singh, K. J. Org. Chem.
2008, 73, 2389.
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
251

Scheme 2. Second-generation RCM process
the equilibrium constant for the disproportionation of the RCM
product to a cyclic dimer (eq 5).
catalyst 5, but at the cost of product yield, and this was not
acceptable, especially because removal of the dimers required
multiple recrystallizations, with the associated losses.
The table also lists our goals for a second-generation
metathesis process. The robust and readily initiating Grela
catalyst,15 which in preliminary experiments could successfully
operate at 0.3-0.4 mol % load, seemed ideally suited. Although
we desired to maintain the higher performance of the Grela
catalyst, we needed to increase the EM(therm). We felt that a
10-fold improvement would allow us to operate in normal
process equipment and lead to a much greener process.
After considering various derivatization schemes for diene
2, we settled on diene 6 as the new substrate (Scheme 2). The
new diene 6 features a Boc group bound to the P1-P2 amide
nitrogen. The selection process and an interpretation of the data
have been communicated separately.16
We ran the RCM reaction of Scheme 2 under several initial
diene concentrations. Calculation of (EM1)2/EM2 for this process
(Figure 5) yielded 1.85 M (R2 ) 0.990), which represents a
19-fold improvement over the analogous modified EM(therm)
for diene 2 under the same conditions. A further improvement
could be realized by running the RCM reaction at 110 °C: the
(EM1)2/EM2 in this case was 2.56 M (R2 ) 0.955; plot not
shown).
2 RCM product ) dimer
(5)
We can estimate the latter concentration, as we have done
above, assuming [eq 6]:
[diene]0 ) [RCM product] + 2[dimer]
(6)
This leads to the plot in Figure 4; the resulting EM12/EM2
(a Keq) represents a Modified EffectiVe Molarity and has a value
0.096 M (R2 ) 0.998). Note that this modified thermodynamic
EM was obtained under slightly different conditions than the
kinetic one (PhMe at 60 °C instead of CH2Cl2 at 40 °C), and
the two parameters cannot, at any rate, be meaningfully
compared.
Table 1 summarizes our results with the two representative
first- and second-generation catalysts. We had also briefly
considered, as potential solutions, polymer-supported catalysts13
The effect of the temperature is qualitatively understood by
taking into account the entropic change in the two processes;
whereas the RCM process produces two olefinic products for
Figure 4. Calculation of modified EM(therm) for the reaction
in Scheme 1.
or pseudo-high dilution conditions,14 but these initial attempts
did not hold promise and are not discussed in detail here. As
Table 1 shows, a higher TON and TOF were achieved with
(11) (a) Miller, S. J.; Kim, S.-H.; Chen, Z.-R.; Grubbs, R. H. J. Am. Chem.
Soc. 1995, 117, 2108. (b) Conrad, J. C.; Ellman, M. D.; Duarte Silva,
J.A.; Monfette, S.; Parnas, H. H.; Snelgrove, J. L.; Fogg, D. E. J. Am.
Chem. Soc. 2007, 129, 1024.
Figure 5. Modified EM(therm) calculation for N-Boc sub-
strate 6.
(12) Ercolani, G.; Mandolini, L.; Mencarelli, P.; Roelens, S. J. Am. Chem.
Soc. 1993, 115, 3901.
each diene molecule, the dimerization process leads to one and
a half. Higher temperatures will tend to favor the RCM process
(13) (a) Deshmukh, P. H.; Blechert, S. Dalton Trans. 2007, 2479. (b)
Clavier, H.; Nolan, S. P. Chem. Eur. J. 2007, 13, 8029. (c) Clavier,
H.; Grela, K.; Kirschning, A.; Mauduit, M.; Nolan, S. P. Angew.
Chem,. Int. Ed. 2007, 46, 6786. Polymer-supported catalysts are
problematic for several reasons. First, many supported catalysts are
used in large molar amounts and are not sufficiently challenged with
a high enough number of recycles. Indeed, in publications of supported
catalysts there is seldom (if ever) a direct comparison between the
maximum TON obtained under homogeneous conditions and the
maximum TON obtained after all possible recycles of the supported
catalyst. This, on the other hand, is the decisive parameter that would
establish the utility of a supported catalyst. .
(14) The problem of low EM can sometimes be alleviated by running the
reaction by adding the substrate slowly to the catalyst. In practice,
this was unsuccessful using catalyst 4 due to its poor thermal stability,
whereas the Grela catalyst 5 operates in the thermodynamic regime,
and this approach cannot be used.
(15) Michrowska, A.; Bujok, R.; Harutyunyan, S.; Sashuk, V.; Dolgonos,
G.; Grela, K. J. Am. Chem. Soc. 2004, 126, 9318.
(16) Shu, C.; Zeng, X.; Hao, M.-H.; Wei, X.; Yee, N. K.; Busacca, C. A.;
Han, Z.; Farina, V.; Senanayake, C. H. Org. Lett. 2008, 10, 1303.
252
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development

Table 1. Current RCM process vs target
process with
Hoveyda catalyst 4
process with
Grela catalyst 5
desired
process
catalyst load
3-5 mol %
24-40 h
0.01 M
0.3-0.4 mol %
e1 h
e0.3 mol %
e1 h
g 0.1 M
90%
reaction time
initial diene concentration
RCM yield (in situ, HPLC quant.)
0.01 M
92%
85%
Table 2. Current RCM process vs results of second-generation process
process with
Hoveyda catalyst 4
process with
Grela catalyst 5
desired
process
second-generation
process
catalyst load
3-5 mol %
24-40 h
0.01 M
0.3-0.4 mol %
e1 h
e0.3 mol %
e1 h
g0.1 M
g90%
0.05-0.1 mol %
30 min
reaction time
initial diene concentration
RCM yield (HPLC assay)
0.01 M
0.2 M
92%
85%
93%
under thermodynamic conditions. OVerall, a 27-fold improVe-
ment in “modified” EM(therm) was recorded in our new RCM
Vs the old one, resulting in a much more concentrated reaction.
Given that the second-generation process is also a batch process,
like the first, the relative efficiencies of the two approaches can
be compared directly. Of course, the need to add and remove
an N-Boc group needs to be factored in when comparing the
two processes in terms of efficiency.
this sense. The reasons were discussed in our communication16
and can be summarized as follows: the resting state of the Ru
catalyst is not at the vinylcyclopropane moiety, but at the
nonenoic acid moiety, and there is no chance of epimerization
even in the presence of amines or phosphines, which cause
extensive epimerization in the reaction of Scheme 1. This feature
allows us to simplify our manipulations, our QC of solvents
and intermediates, and makes the RCM reaction much more
robust and easier to execute.
Table 2 compares the efficiencies of the first-generation
process (using either catalyst) vs the new process after reopti-
mization (vide infra). Although the Grela catalyst (5) improves
TON and TOF, it leads to large amounts of dimers, and this
results in a modest isolated yield (ca. 75% after two recrystal-
lizations, representing a further 10% yield loss), and therefore,
its use per se does not result in a practical process improvement.
The use of catalyst 5 in conjunction with modified substrate 6,
on the other hand, leads to a major improvement in all areas
and exceeds our original goals when we listed the attributes of
an ideally efficient, green process. The following considerations
hold regarding the new RCM process shown in Scheme 2, in
comparison with our first-generation manufacturing process.
Catalyst efficiency (TON and TOF). Second-generation
catalysts react, in the process shown in Scheme 1, about 3 orders
of magnitude more rapidly than first-generation ones like 4.17
In addition, metathesis of 6 is about 3 times faster than that of
2 under typical conditions.18 This allows reaction completion
in a manner of minutes instead of days (original process). The
RCM step of the new process is no longer a bottleneck
operation. The much higher TOF is complemented by a TON
for 5 which is 50-100 times higher than that of 4 under
identical conditions. This introduces major savings in catalyst
use. We have also observed that the reaction in Scheme 1
requires scrupulous degassing for optimum results, whereas the
one in Scheme 2 requires only a brief boil-out before adding
the catalyst. The much higher robustness of the Grela catalyst
5 vs 4 has allowed us to dispense with oxygen sensors and
oxygen specifications in the system, rendering the RCM much
easier to execute on scale.
Solvent Consumption. Whereas our old process utilized as
much as 150,000 L solvent to process 1 MT of diene 2, the
same amount of diene 6 can now be processed with only 7500
L using the new process. This has important repercussions on
the type of equipment used to carry out the reaction (simple
2000 L or 4000 L vessels are now adequate for production).
Also importantly, the new process does not require the instal-
lation of equipment for rapid evaporation of organic solvents
(such as wiped-film evaporators).
Ruthenium Removal. In the original process,2 quenching
of the catalyst and extractions required excess 2-mercaptoni-
cotinic acid (2 kg for 1 kg diene) and large volumes of
bicarbonate to remove some of the solubilized Ru. One silica
filtration was then employed in order to further reduce the Ru
level, right after the RCM reaction, to 100-200 ppm. Subse-
quent chemical steps had to be coupled with charcoal filtrations
in order to reduce the Ru levels in the API to below 10 ppm.
These heroic efforts could be spared using the new RCM
reaction: given the much lower catalyst load, the amount of
MNA used was 50-fold lower. No silica pad or charcoal
filtrations were necessary, and the Ru content of the RCM
product was typically <50 ppm, yielding an API with <5 ppm
Ru without further specific removal steps.
Overall Reaction Efficiency. Of all parameters that have
been proposed to evaluate the efficiency of a reaction, the
E-factor, introduced by Sheldon, is the easiest to use.19 This
parameter refers to the amount of waste (e.g. in kilograms) for
each unit (e.g. 1 kg) of useful product obtained, and includes
all reagents and solvents. We have calculated the E-factors for
both processes in Schemes 1 and 2. The one in Scheme 2
includes the introduction and the removal of the Boc group.
The first-generation RCM is accompanied by an E-factor of
370, whereas the one in Scheme 2 has an E-factor of 52. No
solvent recycle has been considered. The latter would, of course,
Epimerization Reaction at P1. Although the epimerization
reaction in the RCM shown in Scheme 1 can be avoided by
ensuring that all solvents used in the RCM are acid-washed,
the new RCM (Scheme 2) requires no special precautions in
(17) Simpson, R. D., unpublished observations.
(18) Shu, C., unpublished observations.
(19) Sheldon, R. Green Chem. 2007, 9, 1273.
Vol. 13, No. 2, 2009 / Organic Process Research & Development
•
253

improve the picture for both processes. We have not yet
evaluated the suitability and efficiency of such recycles. In
conclusion, the second-generation RCM process improves the
greenness of the process by about 1 order of magnitude.
mmol, 1.4 equiv) slowly over 30 min. The reaction was allowed
to reach rt over 1 h and stirred at rt for another 2 h. HPLC
showed complete reaction (>99% conversion). The reaction
mixture was washed with 0.1 M HCl (200 mL) and H2O (200
mL), concentrated to minimum volume, diluted with toluene,
and distilled under reduced pressure (total reaction volume )
1.42 L). An HPLC quantitative assay showed formation of
110 g of 6 (yield of 95%, concentration of 0.10 M).
Metathesis of 6. The toluene solution of diene 6 from the
previous step was refluxed at 110 °C under air for 30 min to
remove the oxygen, and then treated with a toluene (5 mL)
solution of 5 (100 mg, 0.15 mmol, 0.001 equiv) slowly over
30 min at 110 °C. The conversion and yield were determined
by quantitative HPLC with predetermined absorption/concentra-
tion curves vs a reference standard. After the reaction was
complete (>99.9%; typically <30 min), the vessel was cooled
to 60 °C, and 2-mercaptonicotinic acid (1.16 g, 7.5 mmol, 0.05
equiv) was added. The suspension was stirred at 60 °C for 2 h.
The mixture was cooled to rt and extracted with 5% NaHCO3
(200 mL) and water (200 mL) to afford a toluene solution of
7. HPLC quantitative assay of the organic phase showed 100 g
of 7 (95%).
Conclusions
We have demonstrated elsewhere16 that the RCM macro-
cyclization reaction can be made more practical when the
EM(therm) is optimized with respect to diene perimeter sub-
stitution, which affects the ring strain of the product. In this
contribution, we have presented the practical outcomes of such
optimization. The main benefits are increased throughput and
decreased solvent consumption, thus making the RCM reaction
much greener and operative under normal plant conditions in a
convenient batch mode at low catalyst loads. We further suggest
that parameters such as kinetic and thermodynamic effectiVe
molarities (EM(kin) and EM(therm), respectively) are extremely
useful concepts in analyzing intramolecular processes and
should be calculated and reported more systematically, in order
for the reader to evaluate the efficiency of a given cyclization.
We hope that these concepts can be extended to other RCM
macrocyclizations.
RCM Product 3. To the toluene solution of 7 obtained
above (1.4 L, 100 g by quantitative HPLC assay, 0.135 mol,
1.0 equiv) was added benzenesulfonic acid (42.8 g, 0.27 mol,
2.0 equiv). The mixture was heated to 75 °C. After 2 h, HPLC
showed complete reaction (>99.5%). The solution was washed
with saturated NaHCO3 (300 mL) followed by H2O (300 mL)
to afford a toluene solution of 3.
Experimental Section
Materials and Methods. All reactions were performed in
vacuum-dried jacketed reactors, unless otherwise indicated. All
reagents and solvents were purchased from commercial sources
and used as received, unless otherwise indicated. Compound 2
was synthesized according to a literature procedure. Conversion
was determined by HPLC using UV area % values. All
quantifications were performed by HPLC assay using a
predetermined calibration line of absorption/concentration vs
an external reference standard. All compounds disclosed here
have been previously described.20
HPLC quantitative assay showed 82 g of 3 (yield of 95%).
This material was further processed as previously described.2
Acknowledgment
We thank Scott Denmark, Erick Carreira, and Peter Wipf,
for useful discussions on the problem of low EM, and Donna
Blackmond, for valuable mechanistic suggestions.
Synthesis of Diene 6. To a reactor containing 2 (100 g, 150
mmol, 1.0 equiv), DMAP (5.50 g, 45 mmol, 0.3 equiv), and
ethyl acetate (600 mL) at 0 °C was added Boc2O (45.6 g, 210
Received for review September 11, 2008.
OP800225F
(20) See Experimental Section in ref 2 and Supporting Information in ref
16.
254
•
Vol. 13, No. 2, 2009 / Organic Process Research & Development