Response to the comments by Elati et al. in response to our article examining one of their previous articles

Article abstract of DOI:10.1021/op800252w

This reply highlights and discusses what we observe as internal inconsistencies in the data and analysis presented by Elati and coauthors in conjunction with their resolution protocols, as well as inconsistencies between their original manuscript, the ass

Full text of DOI:10.1021/op800252w

Organic Process Research & Development 2009, 13, 38–43
Response to the Comments by Elati et al. in Response to Our Article Examining One
of Their Previous Articles
Robert James Dancer*,† and Heidi Lopez De Diego‡
Department for Process Research and Department for Preformulation, H. Lundbeck A/S, OttiliaVej 7-9,
DK-2500 Valby, Denmark
Abstract:
with (-)-DPTTA as the chiral resolving agent is not feasible
in the manner cited in our article”4 (Note: in the article it was
clearly stated in a number of places that the (-)-isomer of DTT
is used). However, they then describe a different resolution
protocol, derived from (but not identical to) an example in their
patent application. This procedure involves crystallisation of
racemic citalopram with (+)-DTT, leading to mother liquors
slightly enriched with the salt of (S)-citalopram. Elati et al. then
reported that successive repetitive crystallisations with (+)-DTT,
taking the mother liquors each time, led to essentially pure (S)-
citalopram. Elati et al. then stated in their letter that it is this
latter procedure (“back-to-front” resolution) that should have
been in the article, and that a minor omission in the text led to
this confusion. We will now address the matters raised here in
the following order: (a) the level of changes to the original
manuscript necessary in order to incorporate this new experi-
ment, (b) a discussion of the procedure referred to in the patent
application, (c) a discussion of the currently described procedure
(letter), along with mention of some of its internal inconsisten-
cies, and (d) our efforts to repeat this last procedure.
This reply highlights and discusses what we observe as internal
inconsistencies in the data and analysis presented by Elati and
coauthors in conjunction with their resolution protocols, as well
as inconsistencies between their original manuscript, the associated
patent, and the response to our disputing manuscript. We address
also their comments concerning their alkylation procedures.
Introduction
In our article “Attempted Resolution of Citalopram using
(-)-O,O′-Di-p-toluoyl-(R,R)-Tartaric Acid, and Reflections on
an Alkylation Reaction; Comment on an Article by Elati et al.”1
we drew attention to our inability to repeat the resolution of
citalopram reported by Elati et al.2 using (-)-O,O′-di-p-toluoyl-
(R,R)-tartaric acid (DTT; Elati et al. use the abbreviation
“DPTTA”). In addition, we expressed concern over their
procedure to produce the most important intermediate in their
article, didesmethylcitalopram. It should be noted here that there
was no experimental detail given for this reaction in the article,
but only a reference to a patent application3 by essentially the
same authors.
(a) Rewording. We revisit the wording of the original article
by Elati et al.2 in order to discuss the level of changes required
to the original manuscript in order to accommodate their revised
procedure. Although Elati et al. imply that only a few small
changes in the wording of their original article are required to
this end, our interpretation is that a much deeper and thorough
rewriting of certain sections would be required. We begin with
the first mention of the resolution of citalopram (we have placed
their footnote in brackets for clarity): “Of the many resoluting
agents screened, use of (-)-DPTTA, though found to be useful
(Klaus, P.B. reported that the attempts to resolve citalopram
by diasteromeric salt crystallization have not been successful
[Klaus, P. B.; Jens P. U.S. Patent 4,943,590, 1990]), in our hands
proved to be unsatisfactory for an industrial-scale application
due to low yields and multiple crystallisations.” In the Experi-
mental Section a yield of 36% is reported (relative to theoretical
maximum, i.e. 50% of starting material; in other words a yield
of 18% relative to starting racemic citalopram). We believe that
the text and experimental detail are in good agreement with
each other, as they both describe processes which can be used
to obtain a useful quantity of escitalopram (18% yield overall,
98.4% “chiral purity”) but which are unsuited to production
(by means of comparison, the patent example gives no yield
for the final purified compound, and the current version of this
procedure described in the letter of Elati et al. gives an overall
yield of 5.5%, “chiral purity” 96.8%). Note also that both the
Elati et al. have been invited to comment on our article,4
and we thank them for their clarifications and expansions.
However, we would like to point out what we believe are a
number of inconsistencies in their article, their patent applica-
tion, and their reply letter, in addition to what we believe are
some errors in matters of fact. Some are inconsistencies between
documents, and some are internal inconsistencies. This discus-
sion has been separated into two sections: (a) Resolutions
(including the revised procedure for the resolution of citalopram
by Elati et al. and also of their resolution of didesmethylcit-
alopram) and (b) Comments regarding their alkylation reactions
and the stability of 3-chloropropyl amine (CPA).
Resolutions
Resolution of Citalopram Using DTT. First, we appreciate
the clear comment by Elati et al. that “resolution of citalopram
* Author for correspondence. Telephone: (+45) 36 43 33 31. Fax: (+45) 36
43 89 04. E-mail: rda@lundbeck.com.
† Department for Process Research.
‡ Department for Preformulation.
(1) Dancer, R. J.; De Diego, H. L. Org. Process Res. DeV. 2009, 13, 23.
(2) Elati, C. R.; Kolla, N.; Vankawala, P. J.; Gangula, S.; Chalamala, S.;
Sundaram, V.; Bhattacharya, A.; Vurimidi, H.; Mathad, V. T. Org.
Process Res. DeV. 2007, 11, 289–292.
(3) Sundaram, V.; Mathad, V. T.; Venkavala, P. J.; Elati, C. R.; Kolla,
N.; Govindan, S.; Chalamala, S. R.; Gangula, S. (Dr. Reddy’s
Laboratories, Inc.). Preparation of Escitalopram. WO2005/047274 A1,
2005;CAN 2005:451372
(4) Elati, C. R.; Kolla, N.; H.; Mathad, V. T. Org. Process Res. DeV.
2009, 13, 34.
38
•
Vol. 13, No. 1, 2009 / Organic Process Research & Development
10.1021/op800252w CCC: $40.75
2009 American Chemical Society
Published on Web 11/21/2008

text and the footnote specify crystallisations, with no mention
of the use of mother liquors.
reported by Elati et al. runs as follows: “normal” (priority
application, Nov 2003), “back-to-front” (final patent application,
Nov 2004),3 “normal” (article, Jan 2007),2 “normal” (correction,
Jun 2007),5 “back-to-front” (letter, Sep 2008).4
To quote from the Experimental Section: “...the resulting
solid was filtered. The recrystallisation with acetonitrile/
methanol was repeated for two more times, and the resulting
solid was filtered. The filter cake was washed with acetonitrile
(20 mL) and dried... to afford 9.8 g of 1·(-)-DPTTA.”2 All of
the manipulations here refer clearly to the solids isolated by
filtration, and not the mother liquors, and the yields and purities
quoted differ significantly from those in the example provided
in their letter. On the basis of this information, it is our belief
that in order to accommodate the changes proposed in the letter
of Elati et al., no simple incorporation of additional text to the
existing article would be sufficient. Certain sections would need
to be more deeply rewritten, and the experimental for this
procedure would need to be completely replaced. It is also
worthy to note that the authors published a correction to
precisely the same Experimental Section approximately 6
months after publication of the original article.5 Precisely the
same wording of the Experimental Section (as quoted above)
was also used in the correction.
(b) Patent Application Example. The “back-to-front” resolu-
tion procedure given in the patent example is identical to that
described in the letter by Elati et al. (i.e., the same quantities of
materials, the same ratio between citalopram and DTT, and the
same ratios between citalopram, acetonitrile, and methanol (10
volumes acetonitrile and 0.8 volumes methanol)). However, the
results reported differ. The patent example reports that two
crystallisations were required in order to reach a “chiral purity”
of 98.89%, whereas the letter example reports that three
crystallisations were required to obtain a “chiral purity” of
96.8% (using the same ratios between reactants, reagents, and
solvents for each individual crystallisation. In addition, whilst
the patent example reports a yield of 12.0 g (free base) for the
first step, the letter example reports a yield of 15.0 g (free base)
from the first step. Unfortunately, other comparisons are not
possible, as the patent example only reports a yield for the first
step (no purity), and only purity for the second step (no yield).
We would like to point out a curiosity in the workup of the
patent example. The mother liquor after filtration contains
citalopram isomers and DTT (a diacid) in a 1:1 ratio. After the
first crystallisation, 12 g (37 mmol) citalopram is isolated after
freeing the base with sodium hydroxide (1.6 g in 100 mL; 40
mmol). This means that 1.08 equiv of sodium hydroxide was
used for the neutralisation of a monoamine salt of a diacid. It
would be expected that over 2 equiv would be required for such
a neutralisation, and indeed in our hands it was found that, after
the addition of the specified amount of sodium hydroxide,
essentially no citalopram was extracted into the toluene phase.
In our hands, a small excess over 2 equiv was required
(typically, 2.2-2.5 equiv).
(c) Letter Example. As previously stated, this procedure is
essentially the same as that used in the patent application, but
the results were not (patent required 2 crystallisations to exceed
S:R 96:4; letter required 3; after 2 crystallisations the ratio S:R
was 83.33:16.67, despite the fact that the procedures described
are identical). Therefore, although the letter procedure can be
derived from the patent procedure, it is not in agreement with
the patent procedure, contrary to the statement of Elati et al.4
As also mentioned previously, a comparison of yields for pure
escitalopram between the letter and the patent application is
not possible, since there are no yields for the purified escitalo-
pram in the patent application.
However, we would like to focus on a portion of a mass
balance analysis of the reported products of the letter procedure.
For each cycle of the crystallisation process, the starting mass
and the S:R ratio of the free base is reported, along with the
S:R ratios of the filtered solids and the residues after evaporation.
From those values it was possible to back-calculate some of
the reported values in order to perform a cross-check. We
present here two inconsistencies found in this manner. The first
was found at the end of the First Isolation. It was stated that
the mass of the DTT salt isolated from the mother liquor was
23 g (32 mmol). However, after the salt was neutralised, they
obtained a yield of 15 g (46 mmol) of the free base. In other
words, from 32 mmol salt, they obtained 46 mmol free base.
Another more serious inconsistency can be found in the crucial
Third Isolation. The starting material is quoted as having a mass
(free base) of 5 g, and an S:R ratio of 83.33:16.67, giving masses
of S-citalopram and R-citalopram (as the free base) of 4.17 and
0.83 g respectively. In a similar manner, it can be calculated
from the quoted mass of salt isolated from the mother liquor
(6 g, equivalent to 2.74 g free base) with an S:R ratio of 96.38:
3.62, that the masses of S-citalopram and R-citalopram (as the
free base) in the final product were 2.64 and 0.099 g,
respectively. Therefore, it follows that the masses of S-
citalopram and R-citalopram (as the free base) in the precipitated
solid were 1.53 and 0.73 g, respectively. This gives an S:R ratio
of 67.7:32.3, which is significantly different from the reported
ratio 57.56:42.44.
(d) Our Results. We repeated the resolution procedure
described in the letter of Elati et al. using the same quantities
of starting materials, the same solvent mixtures, and the same
ratios between citalopram mixtures and DTT, acetonitrile, and
methanol. Our yields from the mother liquors were somewhat
lower (10.9 g vs 15.0, and 3.9 g vs 5 g) and purities somewhat
higher (S%: 64.4 vs 59.3, and 88.3 vs 83.3) than those reported
by Elati et al. for the first and second isolations. However, when
we attempted to repeat the third isolation, we obtained no
precipitate whatsoever (in contrast to Elati et al., who obtained
6 g of a DDT salt from the mother liquor, and therefore must
have obtained approximately 5 g precipitated salt). One possible
explanation for this discrepancy could be that, whilst Elati et
al. began with a sample with an S:R ratio of 83.3:16.7, our
sample had an S:R ratio of 88.3:11.7, and subsequent differences
Finally, it may be of interest to observe that in the first
priority application for this patent application, only the “normal”
resolution of citalopram was claimed (using a variety of different
chiral acids). It therefore follows that the chronology of
resolution procedures for citalopram claimed or otherwise
(5) Elati, C. R.; Kolla, N.; Vankawala, P. J.; Gangula, S.; Chalamala, S.;
Sundaram, V.; Bhattacharya, A.; Vurimidi, H.; Mathad, V. T. Org.
Process Res. DeV. 2007, 11, 780.
Vol. 13, No. 1, 2009 / Organic Process Research & Development
•
39

in solubilities could have led to the observed differences in
results. We therefore prepared a 5 g sample containing citalo-
pram isomers in the ratio S:R 83.9:16.1, and then used precisely
the same procedure as reported by Elati et al. under “Third
Isolation”. After the prescribed 2 h at 25-30 °C, virtually no
precipitate was observed (only cloudiness). The mixture was
then seeded with a citalopram ·DTT salt from a previous step
(S:R of seed material 45:55), and allowed to stir for a further
1 h. The mixture was filtered, and the mother liquor was
evaporated to give the final salt (10.13 g, S:R 87.6:12.4; cf.
Elati et al. 6.0 g, S:R 96.4:3.6).
Alkylations. Alkylation reactions on heteroatoms where
CPA is added as the HCl salt, with in situ neutralisation have
been widely reported in the literature,7 although reported
C-alkylation is somewhat rarer.8 Furthermore, it is conceivable
that with an appropriate procedure for the neutralisation/isolation
and use of the free base, alkylation with the free base could be
effective. However, our concerns in our original article1 were
in part based upon the fact that Elati et al. in their article2 did
not give any experimental details for this reaction, but instead
refer to their patent application. Furthermore, in this patent
application, two of the examples require the isolation of the
free base with no discussion of how the base is to be isolated
(and in addition, one of these two examples is unworkable).
Finally, the third example reported neither yield nor purity.
In this context, we respond to the claim in the letter4 of Elati
et al. that (use of italics is ours) “the findings of Dancer and
Lopez De Diego with respect to C-alkylation are based solely
on the presumption that the free base of chloropropylamine is
unstable which is further based on the disclosure in the literature9
that the ‘separation of 3-chloropropylamine and 2-chloropro-
pylamine by distillation appeared hopeless, because of the
instability of the chloropropylamines.’ ” We believe that this
assertion is simply incorrect. Our concerns are based upon the
following considerations:
We consider it interesting and instructive that the greatest
deviation in our results from those of Elati et al., are for the
critical final crystallisation, where the data of Elati et al. are
internally inconsistent.
Resolutions of Didesmethylcitalopram using DTT. In their
reply letter,4 Elati et al. state that their intention was to focus
on the resolution of didesmethylcitalopram. Since the submis-
sion of our article, we have also begun examine this resolution.
In the article by Elati et al.2 a “chiral purity” after a single
crystallisation of 99.0% is claimed. In our hands, such a high
purity has never been obtainable. We have found that typical
S:R ratios obtained are around 84:16 (equivalent to ee 68%).
On closer inspection of the article by Elati et al. we observed
an oddity in their Figure 3.2 This figure is a graph of “Chiral
purity (%)” vs quantity of water in the acetonitrile used for the
crystallisation. It shows that the starting purity (no water) of
85% rises to approximately 100% when water is approximately
2, thereafter steadily declining. We believe that the casual reader
would assume that these experiments were performed on
racemic material. However, this is not the case. There is a note
in the article to say that the material used for this set of
experiments was didesmethylcitalopram with “85% chiral
purity”. This note is not in Figure 3, and nor in the body of the
text describing the figure. Instead, this information is only found
in a footnote to the body of the text.
We found the use of such enantiomerically enriched material
for a screening experiment for the resolution of a racemic
mixture to be unusual and worrying. The thermodynamic
properties (solubilities, S/R compositions of precipitates) of such
an enantiomerically enriched material will necessarily be
different to that of the racemate, and therefore results (such as
purity or yield) from such a screen will be meaningless when
applied to the racemate. Our physicochemical characterisation
of this system is not yet complete, but it is already clear from
our existing results that this is a case of a partial solid solution.
For a detailed description of what this entails we recommend
to the reader an excellent article by Coquerel’s group (in
particular his discussion of Figure 4).6 However, in summary,
it means that despite a reasonable difference in solubilities of
the diasteromeric salts, it is nonetheless impossible to obtain
pure material from the first crystallisation. Multiple crystalli-
sations are necessary, and to quote from Coquerel’s group,
“Unfortunately, in this case, as the excess in salt A tends to
100%, the yield tends to 0%.”
1. Despite the fact that this alkylation reaction was a
critical step in their synthesis, an Experimental Section
for this procedure has never before been described in
the nonpatent chemical literature.
2. We were surprised that the authors did not decide to
publish this information in Org. Process Res. DeV.
(OPRD), given OPRD’s emphasis on solid, robust
chemistry.
3. The authors referred instead to a patent application
written by essentially the same authors. In the patent
application there are three examples of this alkylation.
Two of the procedures employed solutions of CPA
derived from neat CPA, and the third employed a
solution of CPA in toluene.
4. CPA as the free base is unstable. This is strongly
implied by the fact that a Scifinder search failed to
find a preparation of the free base, nor any physical
properties of the free base. Furthermore, the reference
we cited9 gives evidence that CPA is much less stable
than N,N-dialkylated derivatives.
5. Despite the fact that Elati et al.2 employed solutions
derived from CPA free base, we were surprised that
they had not included details of this important isolation.
6. The second example of the alkylation in the patent
application employed acetone as solvent, and a yield
of 64% was claimed. Bretherick’s Handbook of Reac-
tiVe Chemicals, 6th ed.10 cites the reaction of potassium
tert-butoxide with acetone as hazardous, and in our
(7) For an early example, see: Turk, S. D.; Louthan, R. P.; Cobb, R. L.;
Bresson, C. R. J. Org. Chem. 1962, 2846–2853.
(8) See, for example: (a) Adger, B. M.; Mastrocola, A. R. (Smith Kline
& French Laboratories Limited). Synthesis of 2-Pyridylalkylamines.
U.S. 4,526,974, 1985,CAN 100:68182. (b) Cooper, D. G.; Durant,
G. J.; Ganellin, C. R.; Ife, R. J.; Meeson, M. L.; Sach, G. S. Farmaco
1991, 46, 3–19.
(9) Kharasch, M. S.; Fuchs, C. F. J. Org. Chem. 1945, 10, 159–169
.
(6) Marchand, P.; Lefe`bvre, L.; Querniard, F.; Cardinae¨l, P.; Perez, G.;
Counioux, J.-J.; Coquerel, G. Tetrahedron Asymmetry 2004, 15, 2455–
2465.
(10) Urben, P. G., Ed.; Pitt, M. J., Compiler; Bretherick’s Handbook of
ReactiVe Chemical Hazards, 6th ed.; Butterworth-Heinemann (Reed
Elsevier plc group): Oxford, 1999; Vol. 1, pp 430, 551.
40
•
Vol. 13, No. 1, 2009 / Organic Process Research & Development

hands the procedure gave no yield of the required
compound, nor the characteristic deep red colour of
the phthalane anion. Instead, a quantity of white solid
(a mixture of aldol products) was observed after a
vigorous/violent reaction.
quoted].2 Much discussion was not devoted to this
particular example as our aim was to isolate the desired
product instead of the aldol product.” (a) As noted
previously, the mixture of potassium tert-butoxide and
acetone is known to be hazardous. (b) We have
performed two alkylation reactions using methyl iodide
as the alkylating agent instead of CPA, one using
DMSO as solvent, and the other using acetone as
solvent, both on the same scales as reported by
Sundaram et al. in their patent application.3 The only
difference in the procedure was that the methyl iodide
was added neat instead of a very concentrated solution.
Where DMSO was used as solvent, addition of methyl
iodide to the reaction mixture at 25 °C gave an
immediate transformation of the deep red/purple anion
colour to a transparent yellow colour, with an immedi-
ate rise in temperature to around 65 °C. The reaction
was finished almost instantaneously, and workup gave
an almost quantitative yield of the methylated ph-
thalane. Where acetone was used as solvent, there was
none of the characteristic anion colour present. On
addition of methyl iodide, there was essentially no
temperature increase, and even after 24 h there was
no sign of methylation products by HPLC (principally
starting phthalane, plus a number of small impurities).
This indicates that alkylation is impossible in this
system, regardless of the alkylating agent used.
7. The final example of the alkylation in the patent
application gave no yields or purities and so was
therefore of very little use.
We will now examine how these points were affected by
the reply of Elati et al.4
1. No change.
2. Elati et al.4 stated that their intention was to focus on
the resolution step. Nonetheless, we find it surprising
that Elati et al.2,4 chose not to publish their optimum
procedure in OPRD, especially since (a) no useful
procedure had been published by them elsewhere, (b)
the following dimethylation step was included in the
article, and (c) they chose to publish their procedure
for the resolution of citalopram (which they have now
in essence withdrawn).
3. No change.
4. (a) Elati et al. confirmed that CPA is unstable as the
free base4 and did not give any physical properties for
the free base (apart from some quoted stabilities; we
refer again to these figures below). (b) They claimed
in their reply letter that the article we quoted suggested
simply that distillation of CPA is difficult, not the
isolation or handling. They then later state that, in their
hands, “handling and storing 3-CPA free base was a
major challenge as it is unstable”. The article we
quoted is a clear comparison of the stabilities of CPA
and the N,N-diethyl derivatives (the latter has a quoted
boiling point and published preparations, the former
does not). (c) Elati et al. state that (our use of italics)
“While 3-CPA freebase polymerized at 25-45 °C in
the absence of solvent, 3-CPA extracted into DCM,
MTBE, or toluene were relatiVely stable.” However,
in their table of stability of CPA free base under
different conditions, the relative stabilities of different
solutions (10% w/v) at 25-30 °C are given in the
following order (numbers in brackets denote time
required for decomposition): toluene (72 h), DMSO
(12 h), DCM (12 h), neat (8 h), MTBE (2 h), 1,4-
dioxane (2 h). In other words, the text states that an
MTBE solution is more stable than the isolated free
base, yet the reported data show the opposite.
5. Elati et al. in their letter state4 that neat CPA was
obtained by the complete distillation of DCM from a
DCM solution. However, they have still not provided
a procedure for this distillation.
7. In their letter of reply, Elati et al.4 gave yields and
purities for their third alkylation procedure. However,
they also make the statement “hence we can com-
pletely rule out Dancer’s perception that we utilized
the hydrochloride salt of 3-chloropropyl amine, as
under these conditions the hydrochloride salt would
be highly insoluble in toluene.” The first step of their
alkylation procedure as stated in their patent applica-
tion is the neutralisation of CPA· HCl using a mixture
of toluene and sodium hydroxide solution. However,
in the patent application it is clearly stated that the
free base of CPA is treated with toluene and sodium
hydroxide, not the salt. In our article we noted that
this was likely to be an error in the patent application,
and indeed Elati et al. have incorporated our suggested
correction into the Experimental Section in their letter
of reply. So in that context it is therefore surprising
that they “completely rule out” an alteration that they
have themselves made later on in their own letter.
Summary
We summarise our comments in point form, divided up into
(a) crystallisation issues and (b) alkylation issues:
(a) Crystallisation Issues.
6. We were somewhat suprised that Elati et al.4 continue
to claim that the alkylation reaction in acetone using
potassium tert-butoxide can be performed successfully,
and in their quoted yield of 64%. In their letter they
state that (our comments in square brackets; the
reference numbers refer to their references, not ours)
“In example 22 the alkylation was performed in
acetone. It is known that carbonyl compounds under
basic conditions often undergo condensation reaction.
In comparison to example 1 [performed in DMSO,
quoted yield 80 %], this transformation was not as
smooth and did result in a low yield [yield of 64 % is
In the letter by Elati et al.4 they state that all mention
of a “normal” resolution in their article was a mistake;
a simple error arising from the fact that they had “not
incorporated a few words in the text” of their original
article and that a different, “back-to-front” resolution
from their patent application should have been pub-
lished instead. We find it difficult to see how a simple
addition of words to the text could support such a
change.
Although Elati et al.4 claim that the procedures for the
“back-to-front” resolutions in their letter and patent
Vol. 13, No. 1, 2009 / Organic Process Research & Development
•
41

application are in agreement, they differ significantly
in terms of purity after the second crystallisation.
Furthermore, the data that Elati et al.4 report in their
letter for the final recrystallisation are internally
inconsistent in a mass-balance analysis.
We have been unable to repeat the “back-to-front”
resolution procedure described by Elati et al. in their
letter.4 Furthermore, we consider it noteworthy that the
point where our results most strongly disagreed with
those of Elati et al. was at the final critical crystalli-
sation, where the data of Elati et al. were internally
inconsistent.
Preliminary data from our physicochemical charac-
terisation of the resolution of didesmethylcitalopram
are in clear disagreement with the results reported by
Elati et al.2 This doubt is reinforced by an oddity in
the manner they have reported the results of some of
their optimisation reactions in their original article.
solution of (+)-DTT·H2O (62.8 g, 155 mmol) in acetonitrile
(250 mL) and the mixture was stirred at 27 °C for 60 min.The
resultant slurry was heated to 75 °C and methanol (40 mL)
was added to give a clear solution. The solution was cooled to
27 °C over 1 h and kept at this temperature for 1.75 h. The
resulting precipitate was filtered under vacuum and dried
overnight at 60 °C under vacuum to give a yield of 84.60 g
(119.0 mmol; S:R 45.0:55.0). The mother liquor was evaporated
to dryness (26.0 g, 36.6 mmol). This residue was basified with
NaOH (10% w/v 260 mL) at 27 °C for 10 min and extracted
with toluene (2 × 260 mL). The combined organic phases were
dried over anhydrous sodium sulfate (11 g) and evaporated to
dryness under vacuum to give an oil (10.92 g, 33.7 mmol; S:R
64.4:35.6).
This procedure was repeated using this oil (10.92 g, 33.7
mmol; S:R 64.4:35.6) in acetonitrile (55 mL) and (+)-DTT·H2O
(13.6 g, 33.6 mmol) in acetonitrile (55 mL). This gave a
precipitate (14.00 g, 19.7 mmol; S:R 52.4:47.6) and a residue
from the mother liquors (9.82 g, 13.8 mmol) This residue was
basified to give the free base (3.92 g, 12.1 mmol; S:R 88.3:
11.7).
(b) Alkylation Issues.
Despite the data on the stability of CPA in their letter,4
Elati et al. have still not produced details as to how to
isolate CPA.
This procedure was repeated using this oil (3.92 g, 12.1
mmol; S:R 88.3:11.7) in acetonitrile (20 mL) and (+)-DTT·H2O
(4.9 g, 12.1 mmol) in acetonitrile (20 mL). However, on this
occasion no precipitate was observed.
The article quoted in our article compares the relatiVe
stability of CPA and a tertiary CPA analogue. It is
therefore of relevance when discussing the isolability
of CPA. Again, ScifinderTM reports no preparation of
CPA as the free base, but reports numerous prepara-
tions of N,N-diethyl-CPA. Elati et al.4 agree that CPA
is unstable as the free base, but quote data supporting
the stability of certain solutions of CPA. However,
their statements and data contain a number of internal
inconsistencies and contradictions.
A solution was prepared containing citalopram enananti-
omers (5.0 g, 15.4 mmol; S:R 83.9:16.1) in acetonitrile (25 mL),
and the above procedure was repeated using (+)-DTT·1H2O
(6.06 g, 15.4 mmol) in acetonitrile (25 mL). After the usual
cooling/stirring was completed, only cloudiness was observed
without any significant precipitation. The solution was then
seeded with some of the precipitate from the first part of this
experimental (approximately 30 mg; S:R 45.0:55.0). After
stirring for a further 1 h a precipitate had formed. The precipitate
was removed by filtration and dried under vacuum to give a
solid (1.0 g, 1.41 mmol; S:R 56.3:43.7), and evaporation of the
mother liquors gave a residue as a salt (10.13 g, 14.25 mmol;
S:R 87.6:12.4).
Despite the continuing claims of Elati et al.2,4 to the
contrary, Example 2 from their patent application3
(alkylation in acetone) is completely unworkable.
Experimental Section
Resolution of Didesmethylcitalopram Using (-)-DTT
after Elati et al.2 A mixture of (-)-DTT (6.5 g, 17.0 mmol)
and acetonitrile (25 mL) was stirred for 5 min. To this mixture
was added a solution of didesmethylcitalopram (5.0 g, 17.0
mmol) in acetonitrile (25 mL) over 15 min at 27 °C. The slurry
was warmed to 55-60 °C and water (15 mL) was added
dropwise. After a further 50 min the solution was cooled to 0
°C for 1 h, followed by stirring at 27 °C for a further 45 min.
This warming/cooling cycle was repeated two more times, and
the resulting solid was filtered at 0-5 °C, washed with cold
acetonitrile (10 mL), and dried under vacuum (60 °C) to give
didesmethylcitalopram.(-)-DTT as a solid (4.15 g, 5.85 mmol;
S:R 84:16)
Alkylation of Cyanophthalane with Methyl Iodide in
DMSO Using the Method of Sundaram et al.3 A solution of
potassium tert-butoxide (7.5 g, 67 mmol) and DMSO (40 mL)
was warmed to 60-65 °C for 10 min. The solution was allowed
to cool to 25-30 °C, and a solution of cyanophthalane (10.0
g, 41.8 mmol) in DMSO (35 mL) was added dropwise over 10
min, and the resulting intense deep red/purple solution was
General Methods. Acetonitrile (HPLC grade) and methanol
(Anhydroscan) were purchased from LAB-Scan and were used
without purification. DTT (both monohydrate and anhydrous)
were purchased and were used without further purification.
Chiral HPLC analyses were performed using a Chiralcel OD
column (4.6 mm i.d., 250 mm) using a mixture of heptane/
ethanol/diethylamine (98.4:1.5:0.1) as eluent. A flow rate of 1
mL/min was used at 30 °C with UV detection at 240 nm.
Nonchiral HPLC analyses of reaction mixtures were performed
using a Lichrosphere 100 RP-8e (5 µm) column (4 mm i.d.,
250 mm) using a 50:50 (v/v) mixture of acetonitrile/water
buffered to pH 3 with a triethylammonium phosphate buffer
as mobile phase. A flow rate of 1 mL/min was used, with UV
detection at 220 nm. NMR spectra were acquired on a Bruker
Avance AV-500 spectrometer operating at 500.13 MHz for 1H
spectra and 125.77 MHz for 13C spectra. Selected NMR spectra
are available in Supporting Information.
Attempted Resolution of Citalopram Using (+)-DTT
after Elati et al.4 To a solution/suspension of citalopram (50.0
g, 154 mmol) in acetonitrile (250 mL) at 27 °C was added a
42
•
Vol. 13, No. 1, 2009 / Organic Process Research & Development

stirred for a further 20 min at that temperature. A water bath
was then added for cooling (25 °C), neat methyl iodide (17.8
g, 125 mmol) was added in one portion, and the temperature
rose rapidly to approximately 65 °C; the solution almost
immediately lost the previous deep colour and instead gained
an orange colour. After 10 min, cold water (200 mL) and
toluene (100 mL) were added. The organic phase was separated,
and the aqueous phase was washed further with toluene (2 ×
200 mL). The combined toluene phases were evaporated to give
the methylated phthalane as an oil (10.9 g, purity (HPLC)
94.3%).
Alkylation of Cyanophthalane with Methyl Iodide in
Acetone Using the Method of Sundaram et al.3 To acetone
(40 mL) at reflux (56 °C) was added potassium tert-butoxide
(7.5 g, 67 mmol). A vigorous/violent reaction ensued, and the
temperature rose to approximately 63 °C, with concomitant
formation of a large amount of white solids. The solution/slurry
was allowed to cool to 25-30 °C, and a solution of cyanoph-
thalane (10.0 g, 41.8 mmol) in acetone (35 mL) was added
dropwise over 10 min, and the resulting solution was stirred
for a further 20 min at that temperature. Apart from a large
amount of a white precipitate, no colour was observed in the
solution/slurry. A water bath was then added for cooling (25
°C), neat methyl iodide (17.8 g, 125 mmol) was added in one
portion, and the temperature rose slowly to approximately 30
°C. After 10 min, analysis by HPLC indicated that the mixture
contained mostly starting cyanophthalane, plus a number of
small impurities. The mixture was stirred overnight at 25 °C,
and thereafter at 40-45 °C for 1 h. HPLC indicated that the
composition of the reaction mixture was essentially unchanged.
The reaction mixture was then discarded.
Acknowledgment
One of us (R.J.D.) thanks Sir John Cornforth for inspiration
derived from a series of his articles in a similar case some years
ago.
Supporting Information Available
1H NMR spectrum of the crude product from the alkylation
of cyanophthalide with methyl iodide in DMSO; HPLC
chromatograms comparing reaction mixtures in the alkylation
reaction using DMSO or acetone as solvent. This material is
available free of charge via the Internet at http://pubs.acs.org.
Received for review October 6, 2008.
OP800252W
Vol. 13, No. 1, 2009 / Organic Process Research & Development
•
43