Development of an improved process for doxercalciferol via a continuous photochemical reaction

Article abstract of DOI:10.1021/op200346g

Doxercalciferol (1α-hydroxyvitamin D2) is a commercially approved vitamin D derivative used to treat chronic kidney disease (CKD) patients whose kidneys cannot metabolically introduce a hydroxyl group at C1. A new process for the production of doxercalcif

Full text of DOI:10.1021/op200346g

Article
pubs.acs.org/OPRD
Development of an Improved Process for Doxercalciferol via a
Continuous Photochemical Reaction
Bruce G. Anderson,† William E. Bauta,* and William R. Cantrell, Jr.‡
Genzyme Corporation, 14805 Omicron Drive, San Antonio, Texas 78245, United States
S
* Supporting Information
ABSTRACT: Doxercalciferol (1α-hydroxyvitamin D2) is a commercially approved vitamin D derivative used to treat chronic
kidney disease (CKD) patients whose kidneys cannot metabolically introduce a hydroxyl group at C1. A new process for the
production of doxercalciferol from ergocalciferol was developed using a continuous photoisomerization of a known vitamin D
intermediate as the key step, thus circumventing the limitations of batch photoisomerization processes. Doxercalciferol is
produced in an overall yield of about 10% from ergocalciferol.
the sulfur dioxide approach, in which the triene is isomerized
via a sulfur dioxide cycloadduct. Application to doxercalci-
ferol would thus entail synthesis of the cycloadduct 6 from
ergocalciferol (1) and SO2. Protection of the hydroxyl and
thermal retrocycloaddition would afford the isomeric triene (7),
which could be cleanly oxidized at C1 with selenium dioxide
and protected to triene (8). This then would be photo-
chemically isomerized and the alcohols deprotected to yield
doxercalciferol (3).8
In our experience with the cyclovitamin route, a number of
byproducts required removal through multiple and laborious
normal phase chromatography steps. These included the cis-1
β-hydroxyvitamin D2 (9) and the trans-1α-hydroxyvitamin D2
(10) (Figure 2). Byproduct 9 is derived from nonstereospecific
oxidation of 4, whereas 10 results from lack of selectivity in the
cyclopropane opening. Alternative purification approaches for
the cyclovitamin route have included the photoisomerization of
10 to 3,7,9 and the selective Diels−Alder cycloaddition of the
undesired trans-isomer with maleic anhydride10 in order to
facilitate chromatographic purification.
INTRODUCTION
■
In healthy humans, dietary vitamin D2 (ergocalciferol, 1)
undergoes two metabolic oxidations to afford the physiologi-
cally active form of vitamin D2 (1α,25-dihydroxyvitamin D2, 2)
(Figure 1).1 Oxidation at carbon 25 occurs in liver microsomes
by the action of vitamin D 25-hydroxylase, whereas introduc-
tion of the 1α-hydroxyl group is achieved in the kidney by
25-hydroxyvitamin D3 1α-hydroxylase.2 In patients with chronic
kidney disease (CKD), failure to undergo normal oxidation at
position 1α results in a variety of detrimental health effects.3
To circumvent this metabolic deficiency, 1α-hydroxyvitamin
D2, doxercalciferol (3), the active ingredient in the drug
Hectorol, which already contains the 1α-hydroxyl group that
would normally be introduced in the kidney, is used in the
treatment of patients with moderate to severe CKD, includ-
ing patients with end-stage renal disease receiving dialysis.4
Given that doxercalciferol is clinically active at microgram doses
and that the annual demand is less than 1 kg, it is a salient
example of a high-value, low-volume, active pharmaceutical
ingredient.
Owing to the difficulties we encountered with the cyclo-
vitamin route, we turned our attention to the sulfur dioxide
route and its inherent obstacle of batch photoisomerization,
which is typically conducted with an immersion lamp in a
cylindrical reactor. The key problem with batch photoreactors
from a process standpoint is that the reaction occurs very close
to the light source, with most of the batch being unaffected
at a given time. As volume increases this limitation decreases
efficiency, especially when mixing is not optimal. Longer reac-
tion times and increased levels of impurities may result under
such a scenario. In order to circumvent this obstacle, we in-
vestigated the application of a continuous flow isomerization
reactor for the conversion of 10 to doxercalciferol (3) using
a flow photoreactor apparatus adapted from that reported
by Hook et al.11 Flow photochemical processes have been
reported by several groups.7,12 The advantages of continuous
Various synthetic approaches to vitamin D structures are
documented in the chemical literature.5 Although the synthesis
of doxercalciferol (3) from commercially available ergocalciferol
(1) involves the introduction of a single allylic hydroxyl group,
this transformation is encumbered by multiple obstacles,
including competing reactivity at other allylic positions and
the inherent resistance of position 1 to direct oxidation by
selenium reagents, which is believed to result from unfavorable
steric and conformational factors.6 The problem of C1
oxidation has been addressed by different approaches involving
modification of the triene moiety in a manner that masks
competitive sites of reactivity and modifies the steric environ-
ment at C1. In one approach, the cyclovitamin route originally
developed by Paaren and DeLuca,6,7 the triene is modified via
internal displacement of the carbinol at C3 as its tosylate to
afford cyclovitamin (4) and subsequent oxidation at C1 to the
carbinol (5) with selenium dioxide (Scheme 1). Solvolytic
cycloreversion and purification then affords 3. This approach
was successfully applied to the synthesis of doxercalciferol. An
alternative that has been used in related vitamin D molecules is
Special Issue: Continuous Processes 2012
Received: December 1, 2011
Published: January 29, 2012
© 2012 American Chemical Society
967
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
Figure 1.
a
Scheme 1. Synthesis of doxercalciferol (3) using the cyclovitamin and sulfur dioxide routes.
a
Reaction conditions and reagents: a) TsCl, pyridine; b) NaHCO3, MeOH; c) SeO2, t-BuOOH, CH2Cl2; d) Ac2O, pyridine; e) H+, H2O; f) NaOH,
EtOH. g) SO2; h) protection of hydroxyl group; i); NaHCO3, EtOH, heat j) SeO2, NMO; k) irradiation; l) deprotection of hydroxyl groups.
RESULTS
■
Our objective was to use the sulfur dioxide route combined
with continuous photoisomerization of the trans-triene from
the sulfur dioxide route. We also anticipated that careful
selection of protecting groups would afford a crystalline
intermediate through which we might avoid chromatographic
purification. To this end we targeted the crystalline bis-tert-
butyldimethylsilyl (bis-TBS) ether (15) previously reported by
Calverley14 as the key intermediate (Scheme 2).15
Our synthesis of 15 employed a slight modification of
Calverley’s procedure. Starting from commercially available
ergocalciferol (1), the crude intermediates could be taken
Figure 2. Process impurities observed in the cyclovitamin route.
forward without purification to 15, which was purified by
crystallization. Thus, cheletropic cyclization of 1 with excess
flow processes from the standpoint of scalability, safety, and
intensification are well recognized.13 Flow photoreactors place
the reaction solution in close proximity to the light source, and
the duration of light exposure can be optimized through the
flow rate so as to reduce secondary reactions derived from
irradiation of the product. We hypothesized that the greater
efficiency of a flow photoreactor would lead to a cleaner
reaction that could be scaled beyond the limits of a batch
photolysis apparatus.
neat sulfur dioxide at reflux afforded crude sulfone 11 in
quantitative yield (Scheme 2). Some SO2 always remained
entrapped in the crude product, even after prolonged pumping
under high vacuum. An aqueous NaHCO3 workup was
attempted, but this introduced new colored impurities.
However, by calculating the residual SO2 in the crude product
on the basis of the excess crude weight, an adjustment was
made to the imidazole charge in the subsequent step, and the
crude product (11) was carried forward successfully.
968
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
Scheme 2. Synthesis of triene 15 from ergocalciferol
Protection of the carbinol of 11 with a TBS group proceeded
smoothly with TBSCl and imidazole in dichloromethane to
give a 95% yield of 12. Unacceptably low solubility in ethyl
acetate precluded its use in this reaction. The crude, oily
silylated sulfone intermediate (12) was taken to the next step
without purification.
Scheme 3. Photoisomerization apparatus
Retrocycloaddition of 12 to afford the (E,E)-triene isomer 13
was accomplished by treatment with sodium bicarbonate in
anhydrous ethanol at reflux, by analogy with reported
procedures.16 Interestingly, use of 95% ethanol led to a longer
reaction time, as did the use of ethyl acetate. Workup consisted
of solvent exchange for ethyl acetate, followed by aqueous
extraction of the ethyl acetate solution. The resultant oily triene
13 was taken to the next step.
Allylic oxidation of 13 to the alcohol 14 was accomplished
with catalytic SeO2 and N-methylmorpholine-N-oxide (NMO)
as the stoichiometric oxidant17 in dichloromethane. The
oxidation favored the desired 1α-isomer (14a) over the
1β-isomer (14b) with an 85/15 area ratio by HPLC. Attempts
to change the reaction solvent to EtOAc failed due to poor
solubility and slow reaction rate. The crude mixture was taken
to the next step. Following a solvent switch to EtOAc, TBS
protection of the crude isomeric alcohols mixture (14a/14b)
was accomplished by TBSCl with imidazole as base. The crude
mixture of bis-silyl ethers (15) was purified by silica gel
filtration followed by two crystallizations from MeOH/EtOAc
(2/1 v/v) to afford pure 15 in 20% overall yield from 1.
In addition to removing minor impurities that could complicate
the photoisomerization reaction, this purification sequence
increased the isomeric C1 ratio from 6:1 to 121:1 (α:β). As we
anticipated, triene 15 was an attractive intermediate for
purification owing to its ease of crystallization, stability, and
abundant characterization data from the literature, as well as
our desire to avoid introduction of photochemically active
impurities into the subsequent photoisomerization step.
For the continuous photoisomerization of 15, we employed
an apparatus analogous to that used by Hook et al. in a
photocycloaddition reaction (Scheme 3).11 The assemblage
consisted of a PTFE tubing coiled around a photolysis cell,
equipped with an immersion well with cooling jacket, a Pyrex
sleeve, and a 450 W mercury vapor lamp. The immersion well
was connected to a chiller to control the temperature at the
interface of the well and tubing. Temperature was monitored
with a thermocouple placed between the tubing and the
immersion well. The photoisomerization apparatus was
wrapped in aluminum foil to contain the light and was further
969
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
placed into a safety cabinet. A helium-sparged18 heptane
solution of 15 and a photosensitizer, 9-acetylanthracene (9-AA),
was continuously passed via a metering pump through the
apparatus into a collection chamber. Both anthracene and 9-AA
have been previously reported as required and effective triplet
photosensitizers in the photoisomerization of related vitamin D
triene systems.9,19 Starting material and collection flasks were
wrapped in foil to avoid fortuitous exposure to light.
Removal of the photosensitizer 9-AA presented a challenge
due to its lipophilicity and solubility, which made separation
from 16 through crystallization very difficult. However, 9-AA
could be effectively removed from the product solution with an
in-line, single-use 1:2 carbon/Celite column (3 g mixture per
gram of starting material 15) placed between the photoreactor
and the collection vessel.
product to be isolated in a few minutes. For this reason, a full
factorial design was selected as opposed to more abbreviated
models such as fractional factorial designs. Our initial favorable
result was obtained at 20 °C, and the temperature range for the
DoE experiment ( 10 °C) was set to test robustness about this
center point. Concentration and flow rate are both important
variables with implications for the throughput of the process on
a manufacturing scale, and it was anticipated that these variables
would be associated in their effects. For the purpose of the DoE
study, wider ranges about the center point settings were thus
selected for these variables. A principal concern with the
sensitizer range was the ease with which it could be removed
from the reaction mixture and a lower charge of this com-
ponent would be considered preferable; a range of 4 wt % was
selected.
Individual runs were generally very clean, with 16 and
residual 15 accounting for most of the chromatogram area.
Data analysis revealed that the variation in the data was mostly
attributable to variable response effects rather than exper-
imental error. Our three center point replicates afforded a
standard deviation of 1.50% with 95% confidence limits of
6.46. The regression analysis using our model had a standard
error of 0.598 (R2adj = 0.693) and F = 14.56 (p = 0.0001) (see
Supporting Information). Our model was thus appropriate. The
model terms are listed in Table 2. Flow rate and concentration
To our satisfaction, when a heptane solution of 15 (27.5 mg/mL)
and 9-AA (10 wt %) was passed through the photoreactor
apparatus at a flow rate of 5.5 mL/min at 20 °C, the resultant
product solution contained 16 with 96.0% area purity by HPLC,
as well as residual 15 (3.21%). Evaporation of the product
solution afforded crude 16 in 95% yield. Given this favorable
result, we studied the effects of concentration, flow rate, temper-
ature, and 9-AA charge on the HPLC area purity (271 nm
detection) of the crude reaction mixture using a full factorial
design of experiments (DoE) model including three replicates
(runs 4, 7, and 15) at the center point conditions (Table 1).
Table 2. DoE Model terms
Table 1. Four-parameter full factorial experiment in heptane
coefficient
model term
(constant)
2.65
0.042
0.014
−0.0032
flow rate
concentration
flow rate * concentration
occur as individual terms with positive coefficients and there is
also a mixture term with a negative coefficient containing the
product of these two variables. The effect of temperature was
relatively constant over the range of 10−30 °C (Figure 3). The
relationship between concentration and flow rate was more
complex and showed significant curvature (Figure 4). The data
suggest that operating at a higher concentration of 15 with
lower flow rate, or at lower concentration of 15 with higher
flow rate, results in high conversion to 16. This interaction was
anticipated since higher concentration would require longer
light exposure and hence a lower flow rate and vice versa.
Interestingly, an automated search for the optimal flow rate and
concentration revealed a flow rate of 21.8 mL/min and
concentration of 5.0 mg/mL, which corresponds to low
concentration and high flow rate within the ranges studied
(see Supporting Information). The 9-AA sensitizer charge had
little effect over the range studied (Figures 5 and 6). The lack
of response over the range of sensitizer charges could have
several explanations, and although this factor was not studied
extensively, further experiments verified that the reaction could
be conducted using 2.5 wt % 9-AA, indicating that the original
range chosen for the DoE study was probably too narrow.
The photoisomerization reaction worked equally well in
ethyl acetate and ethanol. The latter offered the additional
advantage of being the solvent in the subsequent deprotection
step.
temp.
flow rate
conc. (15)
9-AA
16
15
run
(°C)
(mL/min)
(mg/mL)
(wt%)
(% a/a)
(% a/a)
1
30
30
10
20
30
30
20
10
30
10
10
10
10
10
20
30
30
10
30
1.9
21.8
21.8
5.5
50.0
50.0
50.0
27.5
5.0
4
16
16
10
4
94.3
90.3
73.8
92.4
95.1
92.6
95.0
96.0
92.2
95.6
47.1
93.1
91.8
94.7
95.0
92.2
61.9
92.9
92.0
4.5
7.1
25.0
6.6
3.9
5.0
4.0
3.1
6.1
3.5
51.5
5.0
5.4
4.3
4.0
5.2
36.5
5.6
6.7
2
3
4
5
21.8
1.9
6
5.0
16
10
4
7
5.5
27.5
5.0
8
21.8
1.9
9
50
16
4
10
11
12
13
14
15
16
17
18
19
1.9
50.0
50.0
5.0
21.8
1.9
4
16
4
1.9
5.0
21.8
5.5
5.0
16
10
4
27.5
5.0
1.9
21.8
1.9
50.0
50.0
5.0
4
16
16
21.8
The design and data analysis were conducted using S-Matrix
Fusion Pro software. See Supporting Information for further details.
Using approximately 100 mg per run at the indicated flow
rates allowed for individual runs to be conducted and the crude
Deprotection of 16 with tetrabutylammonium fluoride in
THF, or a combination of THF and either ethyl acetate or
heptane, afforded crude doxercalciferol (3) that was accompanied
970
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
Figure 3. Percent area (16) as a function of temperature and flow rate.
Figure 4. Percent area (16) as a function of temperature and concentration of 15.
by dark-colored impurities which were not characterized. The
resultant mixture was not readily purified by crystallization, so
an alternative deprotection was explored. Treatment of crude
16 with HCl, generated in situ from acetyl chloride in the
presence of an alcohol at room temperature, cleanly afforded
crude 3 (Scheme 4).20
Purification of crude 3 from the acidic deprotection was
possible by crystallization from either methyl formate, methyl
formate and ethyl acetate, or isopropyl acetate (Table 3). Using
the streamlined photoisomerization in tandem with acidic
deprotection and crystallization allowed generation of high-
purity doxercalciferol (3) without purification of the crude
triene (16). Notably, the crystallization procedures were
efficient in the removal of the isomeric triene impurity (10)
derived from 15 in the deprotection step.
CONCLUSION
■
We developed an improved process for doxercalciferol
employing the continuous photoisomerization of a known
silylated triene (15), which was purified by crystallization. This
key triene intermediate was prepared from commercially
available ergocalciferol (1) in five steps, without purification
of intermediates, in an overall yield of about 20%. The
photoisomerization of 15 was accomplished using a continuous
photoreactor apparatus with inline removal of the photo-
sensitizer (9-AA). The photoisomerization reaction can be run
in several solvents at room temperature, but optimal results
require careful selection of the flow rate and concentration of
15. Desilylation of 16 to crude doxercalciferol (3) was
accomplished with HCl in the presence of alcohol solvent.
Purification of the crude product by crystallization afforded
971
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
Figure 5. Percent area (16) as a function of flow rate and concentration of 15.
Figure 6. Percent area (16) as a function of concentration and 9-AA charge.
Scheme 4. Deprotection of triene (16) to give doxercalciferol (3) under acidic conditions
high-purity doxercalciferol (3) in about 50% overall yield from
16. Thus, our process can produce doxercalciferol (3) from
ergocalciferol in about 10% overall yield without the need
for chromatography. Given the small annual demand for
doxercalciferol, the present photochemical isomerization
process could be applied using the same photochemical
972
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
a
evaporate under a stream of dry nitrogen. The residue was
pumped under high vacuum to further remove SO2. The
residual SO2 contained in the solid was calculated as follows:
The weight of crude 11 was 24.30 g, approximately 110% of the
theoretical yield (22.03 g), from which the estimated SO2
content of 2.27 g (35.4 mmol) was calculated. The solid was
taken directly to the next step.
SO2−Cycloadduct of 3(S)-[tert-Butyldimethylsilyl)-
oxy]-9,10-secoergosta-5,7(E),10(19),22(E)-tetraene (12).
To the carbinol (11) (24.30 g, 47.8 mmol) under nitrogen
was added a suspension of imidazole (7.81 g, 115 mmol) in
CH2Cl2 (132 mL), followed by TBSCl (8.65 g, 57.4 mmol).
The solution temperature increased from 22 to 27 °C. After
80 min, additional TBSCl (1.44 g, 9.6 mmol) was charged. After
stirring 165 min, the reaction was 97% complete by HPLC
analysis. The solvent was removed by rotary evaporation, and
the residue was dissolved in EtOAc (200 mL) and washed with
water (100 mL) followed by NaCl brine (100 mL). The EtOAc
phase was dried over MgSO4 and filtered. Solvent removal
afforded crude 12 (27.91 g). The product was initially an oil,
but became a solid at −20 °C.
5(E)-Vitamin D2 TBS Ether (13). A solution of crude 12
(27.49 g, 48 mmol), 95% EtOH (250 mL), and NaHCO3
(25 g, 298 mmol) was heated to reflux for 140 min, after which
additional NaHCO3 (5.00 g, 59 mmol) was charged, and reflux
continued for 2 h. Ethanol was removed by rotary evaporation,
then EtOAc was added to the residue to give a milky solution
that was filtered to afford a turbid filtrate. The EtOAc phase
was washed with water (2 × 100 mL) and NaCl brine (100 mL)
and then was dried with MgSO4 and filtered. Rotary evaporation
afforded crude 13 (21.01 g, 41 mmol).
Table 3. Crystallization of crude doxercalciferol
entry
solvent
3 (%)
10 (%) yield from 16 (%)
1
2
3
methyl formate
methyl formate-EtOAc
i-PrOAc
99.00
99.36
98.69
0.58
0.06
0.38
47
56
54
a
Samples were heated to reflux and then cooled to −20 °C (see
Experimental Section).
apparatus, thus avoiding scale-up-related complications in the
key isomerization step. This, combined with the avoidance of
chromatography, makes this process an attractive option for
manufacturing.
EXPERIMENTAL SECTION
■
HPLC data was collected on Waters Alliance 2690 System 600
instruments with PDA detection. HPLC methods were as
follows:
Method A for analysis of compounds 1, 11, 12, 13, 14,
15, 16: Column: Waters XBridge Phenyl 3.5 μm 4.6 mm ×
150 mm; Flow rate: 1.0 mL/min; Column temperature: 30 °C;
Detection: photodiode array (see individual chromatograms for
wavelength); Gradient elution: (initial) 10% H2O/90% MeOH
→ (27 min) 5% H2O/95% MeOH. Typical retention times
(min): 1 (4.86), 11 (3.22), 12 isomers (7.19, 7.55), 13 (12.83),
14a (6.83), 14b (7.26), 15 (17.24), 16 (19.78).
Method B for analysis of compound 3: Column:
Prodigy ODS (3), 100A, 3 μm, 4.6 mm × 150 mm; Flow rate:
1.4 mL/min; Column temperature: 30 °C; Detection: 265 nm;
Mobile phase: H2O/MeCN (1:4 v/v). Typical retention time
(min): 3 (19.85).
1(S),3(R)-3-(tert-Butyldimethylsilyl)-9,10-secoergosta-
5(E),7(E),10(19),22(E)-tetraene-1-ol (14). Powdered SeO2
(3.19 g, 28.8 mmol) was dissolved in MeOH (210 mL) with
sonication. Intermediate 13 (21.01 g, 41.1 mmol) was charged
and dissolved in DCM (210 mL). To this was added NMO
(19.3 g, 164 mmol), and the mixture was heated to reflux
(∼42 °C) to give a clear solution. To this refluxing solution was
added the methanolic SeO2 solution. After 2 h the reaction was
cooled to room temperature, and the solvent was rotary evapo-
rated. The residue was dissolved in EtOAc (300 mL), trans-
ferred to a separatory funnel, and washed with water (140 mL)
and then NaCl brine (130 mL). Both layers had an orange
color. The EtOAc phase was dried over MgSO4 and filtered.
Evaporation afforded the crude carbinol 14 (21.67 g, 41.1 mmol)
as an isomeric mixture (6:1 α/β at C1) which was taken to the
next step.
1(S),3(R)-Bis(tert-butyldimethylsiloxy)-9,10-secoer-
gosta-5(E),7(E),10(19),22(E)-tetraene (15). Crude carbinol
14 (21.60 g, 41.0 mmol) was dissolved in CH2Cl2 (110 g), and
imidazole (6.14 g, 90.2 mmol) was added. To this was charged
TBSCl (6.80 g, 45.1 mmol) whereupon a precipitate formed.
After 195 min, HPLC analysis revealed the reaction was com-
plete. The reaction mixture was dissolved in EtOAc (300 mL)
and washed with water (150 mL) and then NaCl brine (150 mL).
The EtOAc phase was dried over MgSO4 and filtered. Solvent
removal afforded 15 (25.39 g) as a crude foam. A portion of
this material was further purified as follows. The foam was
dissolved in hexanes (13 mL/g) and filtered through a bed
of silica gel eluting with 1% (v/v) EtOAc in hexanes. Sol-
vent was removed by rotary evaporation and high vacuum
to give crude 15 (1.13 g) as a white solid. This was mixed
with MeOH and EtOAc (2/1 v/v, 34 mL) heated at 61 °C,
The photoreactor was assembled as follows: An immersion
well apparatus (Ace cat. 7874−35) was connected with flexible
tubing to a chiller unit (Lauda-Brinkman RM-6). A 450 W
UV lamp (Ace cat. 7825-35) surrounded by a Pyrex sleeve
(Ace 7835-44) was inserted into the immersion well. The lamp
was attached to a power supply (Ace 7830-60). FEP tubing
(Saint-Germain Performance Plastics, AXI00002 Tube, 0.125″
(OD) × 0.062″ ID, 50 ft; VWR cat. 63014-692) was wrapped in
a tight, single layer around the immersion well, and aluminum
foil was wrapped around the tubing. A KNF STEPDOS liquid
pump was attached to the FEP tubing. The inline 1:2 w/w
carbon/Celite column (3 g/g starting material) (Ace glass filter
column cat. 5813) was placed between the tubing exiting the
photoreactor and the collection flask.
DoE analysis was conducted with S-Matrix Fusion Pro soft-
ware. An automated optimizer using a random walk algorithm
beginning from 10 different points on the response surface was
employed for optimization searches.
SO2−Cycloadduct of Ergocalciferol (11). A flask with a
dry ice condenser charged with dry ice and isopropanol, a
thermocouple, and a magnetic stir bar was assembled. Nitrogen
was introduced through a gas inlet adapter and the outlet
exhausted through a dry ice and isopropanol cold trap attached
to a 10% NaOH (aq) scrubber. Solid ergocalciferol (1) (18.97 g,
47.8 mmol) was charged. The tare weight of the flask and stir
bar was recorded prior to charging reagents. A dry ice−
isopropanol bath (∼−10 °C) was applied as SO2 was slowly
introduced with stirring. The solid 1 dissolved during the
course of the addition. After 23 min a clear orange solution
was achieved as the solid ergocalciferol gradually dissolved, at
which time SO2 addition was discontinued. The cold bath and
condenser were removed, and the residual SO2 was allowed to
973
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
whereupon no solid was present; the mixture was then
cooled to 0 °C. After cooling for 55 min, the crystalline solid
was suction filtered and washed with cold MeOH and EtOAc
(2/1 v/v, 10 mL) and dried overnight at high vacuum to afford 15
(0.61 g, 36:1 α/β at C1) as a white solid. This was recrystallized
by adding a mixture of MeOH and EtOAc (2/1 v/v, 18.3 mL)
and heating to reflux. The clear solution was cooled to room
temperature over 20 min and then further cooled in an ice bath
for 35 min. The solid was suction filtered, washed with the
cold recrystallization solvent (6.1 mL), and dried at room
temperature under high vacuum to afford 15 (0.46 g, 0.80 mmol,
121:1 α/β at C1). The overall yield from ergocalciferol (1) was
HPLC analysis of the product solution showed it to contain 16
in (97.1%, a/a). The solution was rotary evaporated, the residue
was dissolved in EtOH (10 mL) and a solution of ethanolic
HCl (0.5 mL of 9% v/v 12 N HCl in ethanol) at room
temperature. This mixture was stirred overnight at room
temperature, at which point HPLC analysis showed complete
conversion of 16 to doxercalciferol (3). Sodium bicarbonate
was added until the pH of the mixture was 7, then the mixture
stirred for 1.5 h and filtered. The filtrate was concentrated by
rotary evaporation and the residue mixed with methyl formate
(7.0 mL) and heated to reflux. The resultant solution was
cooled to −20 °C in a freezer for 2 h. The resultant solid was
filtered and washed with methyl formate to afford 3 (0.23 g,
0.56 mmol) in 38% yield. The solid was analyzed by HPLC
1
20%. H NMR (CDCl3) δ 6.46 (d, 1H, J = 11.4), 5.82 (d, 1H,
J = 11.4), 5.12−5.28 (m, 2H), 4.99 (s, 1H), 4.94 (s, 1H), 4.54
(dd, 1H, J = 8.9, 4.0), 4.22 (br s, 1H), 2.87 (d, 1H, J = 13.2),
2.55 (dd, 1H, J = 14.0, 5.2), 2.32 (d, 1H, J = 13.6), 1.81−2.08
(m, 6H), 1.62−1.80 (m, 4H), 1.42−1.53 (m, 3H), 1.23−1.39
(m, 3H), 1.02 (d, 3H, J = 6.6), 0.92 (d, 3H, J = 6.8), 0.88 (d,
18H, J = 16.8), 0.81−0.85 (m, 6H), 0.55 (s, 3H), 0.02−0.10
(m, 12H) ppm.
General Procedure for Photoisomerization of 15 to
1(S),3(R)-Bis(tert-butyldimethylsiloxy)-9,10-secoergosta-
5(E),7(E),10(19),22(E)-tetraene (16) for DoE Experiments.
The photoisomerization flow reactor with an inline absorption
column was assembled as described above and the lamp turned
on prior to adjusting the jacket temperature to the indicated
value using the chiller. Triene 15 (105−110 mg) and 9-AA
were dissolved in heptane at the indicated concentrations in an
amber-colored flask. The solution was sparged with He for at
least 5 min prior to being pumped into the reactor. The solu-
tion was pumped through the reactor into an amber-colored
collection flask. The product (16) solution was analyzed by
HPLC and concentrated via rotary evaporation to calculate
mass recovery.
Photoisomerization of 15 to 1(S),3(R)-Bis(tert-butyldi-
methylsiloxy)-9,10-secoergosta-5(E),7(E),10(19),22(E)-
tetraene (16) in Heptane. A solution of triene 15 (106.2 mg,
0.19 mmol) and 9-AA (14.1 mg, 0.06 mmol) in heptane
(20 mL) was sparged with He in an amber flask. The solution
was pumped through the flow reactor and the inline 9-AA
absorption column at a rate of 21.8 mL/min at 16 °C into an
amber collection flask. HPLC analysis of the heptane solution
revealed 16 (96.0%, a/a) and residual 15 (3.21%, a/a). Rotary
evaporation afforded crude 16 (103 mg, 0.18 mmol) in 95%
yield. Proton NMR data were consistent with those reported in
the literature.14 See Supporting Information.
Photoisomerization of 15 to 1(S),3(R)-Bis(tert-butyldi-
methylsiloxy)-9,10-secoergosta-5(E),7(E),10(19),22(E)-
tetraene (16) in Ethyl Acetate. A mixture of 15 (6.17 g, 9.62
mmol) and 9-AA (0.25 g, 1.15 mmol) was dissolved in ethyl
acetate (420 mL) to afford a 15 mg/mL solution, and the
mixture was sparged with He. The solution was pumped at a
rate of 7 mL/min through the photoreactor apparatus as
described above. Upon completion (∼1 h) the product solution
was analyzed by HPLC and found to contain 16 (97.23% a/a)
and 10 (2.77% a/a). Rotary evaporation afforded 16 (5.59 g,
8.71 mmol) as a foam in 93% yield.
1
and found to contain 3 (99.09%) and 10 (0.02%). H NMR
(CDCl3) δ 6.40 (d, 1H, J = 11.2), 6.04 (d, 1H, J = 11.2),
5.35 (s, 1H), 5.15−5.29 (m, 2H), 5.03 (s, 1H), 4.45 (dd, 1H,
J = 7.3, 4.0), 4.21−4.31 (m, 1H), 2.81−2.90 (m, 1H), 2.62
(d, 1H, J = 13.3), 2.34 (dd, 1H, J = 13.3, 6.5), 1.83−2.11
(m, 6H), 1.42−1.79 (m, 7H), 1.21−1.40 (m, 3H), 1.04 (d, 3H,
J = 6.6), 0.94 (d, 3H, J = 6.8), 0.86 (t, 6H, J = 7.3), 0.58
(s, 3H) ppm.
Alternative Crystallization of Doxercalciferol (3) from
Methyl Formate and Ethyl Acetate. Crude 3 (0.71 g, mmol)
was combined with methyl formate (14 mL) and heated to
reflux. EtOAc (7.0 mL) was gradually added to the refluxing
mixture until dissolution was achieved. The mixture was cooled
to −20 °C overnight and then was filtered. The resultant solid
was dried at room temperature at high vacuum overnight to 3
(0.38 g, 0.92 mmol) in 56% yield. HPLC analysis of this
material showed: 3 (99.36%) and 10 (0.06%). HPLC data were
consistent with a reference standard. Proton NMR data were
consistent with those reported in the literature.7 See Supporting
Information.
Alternative Crystallization of Doxercalciferol (3) from
Isopropyl Acetate. Crude 3 (356 mg) was combined with
i-PrOAc (10 mL) in a vial and heated to reflux to give a slightly
hazy solution. The mixture was cooled to room temperature
and stored at −20 °C overnight. The resultant solid was filtered
and washed with cold i-PrOAc. Drying at room temperature
under high vacuum afforded 3 (190 mg, mmol) in 54% yield.
HPLC analysis of the resultant solid was as follows: 3 (98.69%,
a/a), 10 (0.38% a/a).
ASSOCIATED CONTENT
■
S
* Supporting Information
DoE experimental details and analysis. NMR spectra of com-
pounds 15 and 3. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
william.bauta@genzyme.com
Present Addresses
Photoisomerization of 15 to 16, Deprotection, and
Crystallization from Methyl Formate to Doxercalciferol
(3). An amber vial was charged with triene 15 (0.844 g,
1.48 mmol) followed by 9-AA (31.3 mg, 0.14 mmol, 4 wt %)
and EtOAc (60 mL). This solution was passed through the photo-
isomerization apparatus at a flow rate of 14 mL/min at 20 °C.
†Incube Laboratories, 12500 Network Drive, Suite 112, San
Antonio, TX 78249.
‡307 Laurel Ridge, San Antonio, TX 78253.
Notes
The authors declare no competing financial interest.
974
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975

Organic Process Research & Development
Article
ACKNOWLEDGMENTS
■
We thank Dr. Antony Bigot for review of this manuscript and
NMR data. We also thank Dr. Leon LeVan and Dr. James
Goebel for review and useful discussions.
REFERENCES
■
(1) Holick, M. F. Vitamin D: Physiology, Molecular Biology, and
Clinical Applications; Springer Verlag: New York, 2009.
(2) Takeyama, K. I.; Kitanaka, S.; Sato, T.; Kobori, M.; Yanagisawa,
J.; Kato, S. Science 1997, 277, 1827−1830.
(3) (a) Gravellone, L.; Rizzo, M. A.; Martina, V.; Mezzina, N.;
Regalia, A.; Gallieni, M. Int. J. Nephrol. 2011, 1−13. (b) Holick, M. F. N.
Engl. J. Med. 2007, 357, 266−281. (c) Jones, G.; Strugnell, S. A.; DeLuca,
H. F. Physiol. Rev. 1998, 78, 1193−1231.
(4) (a) Upton, R. A.; Knutson, J. C.; Bishop, C. W.; LeVan, L. W.
Nephrol. Dial. Transplant. 2003, 18, 750−758. (b) Frazao, M.;
Chesney, R. W.; Coburn, J. W. Nephrol. Dial. Transplant. 1998, 13
(suppl 3), 68−72. (c) Coburn, J. W.; Tan, A. U.; Levine, B. S.; Mazess,
R. B.; Kyllo, D. M.; Knutson, J. C.; Bishop, C. W. Nephrol. Dial.
Transplant. 1996, 11, 153−157. (d) Knutson, J. C.; Hollis, B.; LeVan,
L. W.; Valliere, C.; Gould, K. G.; Bishop, C. W. Endocrinology 1995,
136, 4749−4753.
(5) Zhu, G.-D.; Okamura, W. H. Chem. Rev. 1995, 95, 1877−1952.
(6) Paaren, H. E.; Hamer, D. E.; Schnoes, H. K.; DeLuca, H. F. Proc.
Natl. Acad. Sci. U.S.A. 1978, 75, 2080.
(7) Paaren, H. E.; DeLuca, H. F.; Schnoes, H. K. J. Org. Chem. 1980,
45, 3253−3258.
(8) Triene modification through cycloaddition with phthalazine-1,4-
dione was employed in the synthesis of 1α,25-dihydroxyvitamin D3:
Andrews, D. R.; Barton, D. H. R.; Hesse, R. H.; Pechet, M. M. J. Org.
Chem. 1986, 51, 4819−4828.
(9) Gielen, J. W. J.; Koolstra, R. N.; Jacobs, H. J. C.; Havinga, E. Recl.
Trav. Chim. Pays-Bas 1980, 99, 306.
(10) (a) Bishop, C. W.; Horst, R. L.; Jones, G.; Kozewski, N. J.;
Knutson, J. C.; Moriarty, R. M.; Reinhardt, T. A.; Penmasta, R.;
Strugnell, S.; Guo, L.; Singhall. S. K.; Zhao, L. U.S. Patent 5,786,348,
1988. (b) Knutson, J. C.; Bishop, C. W.; Moriarty, R. M. U.S. Patent
5,488,120, 1996.
(11) Hook, B. D. A.; Dohle, W.; Hirst, P. R.; Pickworth, M.; Berry,
M. B.; Booker-Milburn, K. I. J. Org. Chem. 2005, 70, 7558−7564.
(12) (a) Coyle, E. E.; Oelgemoller, M. Photochem. Photobiol. Sci.
̈
2008, 7, 1313−1322. (b) Fukuyama, T.; Hino, Y.; Kamata, N.; Ryu, I.
Chem. Lett. 2004, 33, 1430.
(13) (a) Luis, S. V.; Garcia-Verdugo, E., Eds. Chemical Reactions and
Processes under Flow Conditions; Royal Society of Chemistry:
Cambridge, U.K., 2010. (b) Reay, D.; Ramshaw, C.; Harvey, A.
Process Intensification - Engineering for Efficiency, Sustainability and
Flexibility; Butterworth Heinemann (Elsevier): Burlington, MA, 2008.
(14) Calverley, M. J. Tetrahedron 1987, 43, 4609−4619.
(15) The corresponding bis-tert-butyldiphenylsilyl ether has been
reported: Shimizu, M.; Ohno, A.; Yamada, S. Chem. Pharm. Bull. 2001,
49, 312−317.
(16) (a) Xu, W.-M.; He, J.; Yu, M.-Q.; Shen, G.-X. Org. Lett. 2010,
12, 4431−4433. (b) Shimizu, M.; Ohno, A.; Yamada, S. Chem. Pharm.
Bull. 2001, 49, 312−317. (c) Manchand, P. S.; Yinnikouros, G. P.;
Belica, P. S.; Madan, P. J. Org. Chem. 1995, 60, 6574−6581.
(17) Hesse, R. H. U.S. Patent 4,554,105, 1985.
(18) Helium sparging was conducted as a precaution due to the
known photooxidation of vitamin D compounds in solution. See ref 9.
Vitamin D is generally unstable in the presence of oxygen. (a) Grady,
L. T.; Thakkes, K. D. J. Pharm. Sci. 1980, 69, 1099−1102. (b) Stewart,
B. A.; Midland, S. L.; Byrn, S. R. J. Pharm. Sci. 1984, 73, 1322−1323.
(19) (a) Calverley, M. Tetrahedron Lett. 1987, 28, 1337−1340.
(b) Choudhry, H. C.; Belica, P. S.; Coffen, D. L.; Focella, A.; Maehr,
H.; Manchand, P. S.; Serico, S.; Yang, R. T. J. Org. Chem. 1993, 58,
1496−1500. (c) Manchand, P. S.; Yiannikouros, G. P.; Belica, P. S.;
Madan, P. J. Org. Chem. 1995, 60, 6574−6581.
(20) Nelson, T. D.; Crouch, R. D. Synthesis 1996, 1031−1069.
975
dx.doi.org/10.1021/op200346g | Org. Process Res. Dev. 2012, 16, 967−975