Selective chemical oxidation of risperidone: A straightforward and cost-effective synthesis of paliperidone

Article abstract of DOI:10.1002/ejoc.201001618

A very short and cost-effective synthesis of the commercial drug paliperidone has been achieved starting from its parent compound risperidone, through a thoroughly optimized oxidation with air under basic conditions. A very short and cost-effective synthesis of the commercial drug paliperidone (1) has been achieved starting from its parent compound risperidone, through a thoroughly optimized oxidation with air under basic conditions. Copyright

Full text of DOI:10.1002/ejoc.201001618

FULL PAPER
DOI: 10.1002/ejoc.201001618
Selective Chemical Oxidation of Risperidone: A Straightforward and
Cost-Effective Synthesis of Paliperidone
Renata Riva,*[a] Luca Banfi,[a] Graziano Castaldi,[b] Diego Ghislieri,[b] Luciana Malpezzi,[c]
Francesca Musumeci,[a] Roberto Tufaro,[b] and Marcello Rasparini*[b]
Keywords: Medicinal chemistry / Nitrogen heterocycles / Synthesis design / Antipsychotics / Hydroxylation / Oxidation
A very short and cost-effective synthesis of the commercial
drug paliperidone has been achieved starting from its parent
compound risperidone, through a thoroughly optimized oxi-
dation with air under basic conditions.
dal symptoms during risperidone treatment depends on the
proportion of 2 to 1 in the patient.[2] Actually, paliperidone
(1) is characterized by a very fast 50% dissociation from D2
(only 60 s), and thus it definitely is an atypical antipsy-
chotic. This knowledge has prompted the development of 1
as a separate antipsychotic drug. Paliperidone (1) is gen-
erally regarded as an improved drug compared to its parent
not only because of its reduced side effects, but also because
of its better pharmacokinetic properties, that is, having a
longer elimination half-life in the brain. For these reasons,
notwithstanding its higher price, it has recently been suc-
cessfully commercialized on the drug market as the race-
mate[6] and is sold as Invega®.
Introduction
Atypical antipsychotics are gradually replacing typical
antipsychotics in the treatment of schizophrenia because of
the decreased propensity to cause extrapyramidal signs and
symptoms (EPS) and the absence of sustained prolactin ele-
vation. These drugs act as dual dopamine-serotonin antago-
nists. The reasons for reduced extrapyramidal activity in
these drugs are not yet fully understood. The reduction may
be linked to the dual activity against D2 and serotonin re-
ceptors (mainly 5-HT2 and 5-HT7). However, it was shown
that the kinetics of binding to dopamine D2 receptors differ
among antipsychotic compounds, and therefore it was pro-
posed that fast-off kinetics (high koff) may be associated
with lower liability for extrapyramidal side effects (EPS).[1,2]
Thus, a loose binding with D2 receptors may be the distinc-
tive mark of atypical antipsychotics.
In continuation of our research focused on the efficient
synthesis of pharmaceutically active principles, we decided
to search for an original and scalable synthetic route for
this important drug.
Risperidone (2) is an antipsychotic widely used in clinical
practice for the treatment of schizophrenia, autism, and bi-
polar disorders. Though generally considered an atypical
antipsychotic, there is still some debate about that classifi-
cation. In fact, in vitro experiments showed a dissociation
time of 27 min, more similar to that of older “typical” anti-
psychotics.
The major route for elimination of 2 is hepatic oxidation
to both enantiomers of 9-hydroxyrisperidone (paliperidone,
1).[3–5] It has been suggested that the extent of extrapyrami-
The first synthesis of paliperidone was described by its
discoverers at Janssen,[7] and the route is very similar to the
one previously employed for risperidone (Scheme 1). The
key intermediates are compounds 4, which in turn are syn-
thesized from protected pyridopyrimidinone 5 by hydrogen-
ation of the pyridine ring (R = benzyl) or by simultaneous
hydrogenation and hydrogenolysis (R = H). This procedure
is not fully satisfactory. Compared to the industrial risperi-
done synthesis, the need to include an extra oxygen leads
to diminished yields for both the formation of 5 from ami-
nopyridine 6, as well as for the formation 4 from the hydro-
genation of 5. In our hands these steps were found to be
troublesome, and the purification steps were rather difficult.
A possible alternative is represented in the synthesis of 4
by nitrosation of 9, a key intermediate in the risperidone
synthesis. This approach was recently reported in a
patent,[8] after the completion of the present work, and in-
volves oxidation of 9 with isoamyl nitrite. Unfortunately,
conversion of the resulting oxime 8 into the key intermedi-
ate 4 required a multi-step sequence, using a stoichiometric
amount of TiCl3, making this approach not very attractive.
[a] Department of Chemistry and Industrial Chemistry, University
of Genova,
Via Dodecaneso 31, 16146 Genova, Italy
Fax: +39-010-3536118
E-mail: riva@chimica.unige.it
[b] Chemessentia Srl,
Via Bovio 6, 28100 Novara, Italy
Fax: +39-0321-479200
E-mail: Marcello.Rasparini@chemessentia.it
[c] Politechnic of Milan, Department of Chemistry, Materials and
Industrial Chemistry,
Via Mancinelli 7, 20131 Milan, Italy
Supporting information for this article is available on the
WWW under http://dx.doi.org/10.1002/ejoc.201001618.
Eur. J. Org. Chem. 2011, 2319–2325
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2319

R. Riva, M. Rasparini et al.
FULL PAPER
Scheme 1.
Due to these problems, the market price of paliperidone
(1) was expected to be more than four times higher than
that of risperidone (2). Thus, we reasoned that the direct
oxidation of 2 to 1, in one step, would lead to a synthesis of
1 which was more convenient than those previously known.
Though simple in theory, this approach was not expected
to be trivial in practice, because risperidone (2) possesses
several oxidizable sites. Indeed, it can form N-oxides at N-
1 or at the piperidine nitrogen, and it can be oxidized at
five different positions activated by being α to heterarom-
atic systems (benzylic positions). For example, it is well
known that in the liver the oxidation at C-9 is accompanied
by oxidation at C-6.
Although biological oxidation of 2 to 1 seems the most
obvious approach, as paliperidone is a natural metabolite
of risperidone, the use of cytochromes to convert 2 into 1
on a preparative scale was found to be inefficient by other
researchers in the field.[9] Furthermore, we could find in
the literature very few examples of biological oxidations of
benzylic carbons in nitrogen heterocycles.[10] For these
reasons we decided to concentrate our efforts on chemical
methodologies.
Scheme 2.
Other reagents, such as the Dess–Martin periodinane,[13]
directly afforded 9-oxo-risperidone (12), and no intermedi-
Results and Discussion
We first tried some neutral oxidizing agents. The oxi- ate alcohol 1 was observed. At first sight the formation of
dations at benzylic positions of electron-rich arenes by a 12 may be considered useful in view of possible reduction
variety of oxidants are well known, but in our case, the to 1. However, as already anticipated by a study on a model
electron-poor nature of the pyrimidinone ring made all compound,[14] reduction of 12 to 1 with NaBH4 or other
attempts unsuccessful. In general the reaction was very dif- reducing agents (for example, diisobutylaluminum hydride)
ficult, and in all cases, the nitrogens tended to be oxidized proceeded poorly, because of concomitant reduction of the
first. For example no reaction occurred with meta-chloro- pyrimidinone ring.
perbenzoic acid, whereas slow formation of the N-oxide at
Therefore we realized that the best strategy would be to
N-1 to give 10 was observed (Scheme 2) with peracetic acid exploit the electron-poor nature of the pyrimidone and to
(generated in situ from acetic acid and hydrogen peroxide). convert the benzylic proton at C-9 into an anion to facili-
Unfortunately, the reaction was not clean, because of con- tate oxidation. As already stated above, 2 has five benzylic
current formation of the N-oxide on the piperidine ring, positions, and at least two of them may form resonance
and the yield of 10 was poor. Nevertheless we tried to con- stabilized anions, that is, at C-9 and the methyl group. Thus,
vert this N-oxide into 1 by way of the Boekelheide re- to have an idea of possible selectivity and to check the com-
arrangement,[11,12] but this was unsuccessful.
patibility of the whole molecule with strong bases, we first
2320
www.eurjoc.org
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2011, 2319–2325

Synthesis of Paliperidone
studied the allylation of 2. Although 2 was found to be We were delighted to find that treatment of 2 with LDA
unstable in the presence of NaH, we succeeded in obtaining followed by bubbling O2 at 0 °C followed by a reductive
the 9-allyl derivative 11 in 43% yield by deprotonation of 2 work-up (Table 1, Entry 1) gave the desired product 1.
with LDA (lithium diisopropylamide) followed by reaction However, the reaction was largely incomplete, the isolated
with allyl bromide. Although a diallylated side product was yield was poor (10%), and various side products were de-
detected, NMR clearly showed that it was diallylated at C- tected. The most important one was ketone 12, which was
9. Although this reaction was not optimized, this encourag- formed in equal amounts with 1. In CDCl3, this compound
ing result demonstrated that kinetic deprotonation at C-9 is a mixture of ketone 12 and enol 13 (82:18) and is not
was favored with LDA and prompted us to treat the same very stable, slowly decomposing at –20 °C in the dry state.
carbanion with oxidizing agents. Probably the most versa- Interestingly ketone 12 was completely converted into the
tile and convenient reagent for the hydroxylation of enolates corresponding stable enamine 14 when chromatography of
derived from ketones and esters is Davis’ 3-phenyl-2-(phen- the crude product was carried out in the presence of ammo-
ylsulfonyl)oxaziridine,[15,16] which has also been used on nia. Apart from this ketone, other more polar impurities
large scale syntheses of active principles.[17] Although to the were present. Two of them were identified by HPLC-MS.
best of our knowledge it was not previously employed in The first one, with [M + H]+ = 819, was proposed to be
benzyl anions oxidations, we tried to trap the LDA derived dimer 15. The structure of the second one, with [M + H]+
carbanion of 2 with this reagent, but this was unsuccess- = 443, was more difficult to assign. It is clearly a product
ful.
of dioxygenation, but the second oxidation could in prin-
Among the rare examples of benzylic oxidations of nitro- ciple have occurred at N-1, at the piperidine nitrogen, or at
gen heterocycles, the oxidation of deoxyquinine with di- one of the other four benzylic positions. As described be-
oxygen under basic conditions performed by Uskokovic[18] low, the structure 16 was later unambiguously assigned to
and Stork[19] during their total synthesis of quinine caught this impurity by crystallographic methods. Other unidenti-
our attention. Indeed, it is a benzylic oxidation of a quino- fied polar impurities were also present.
line derivative within a quite complex structure containing,
Identification of the impurities allowed us to draw a pos-
as in our case, a tertiary cyclic aliphatic amine. Thus we sible mechanism for this reaction (Scheme 3).[21] Among the
decided to treat the LDA derived anion with dioxygen.[20] five possible benzylic positions, it is likely that two are more
Table 1. Optimization of conversion of 2 into 1.
Entry
Base[a]
Solvent
Conc. [m] Additive[b] Oxidant[c]
Temp. Time [h] Work-up[d]
Yield [%]
Main products[e]
1
2
12
1
2
3
4
5
6
7
8
LDA
tBuOK
KHMDS THF
LiHMDS THF
THF
THF
0.0244
0.0244
0.0244
0.0244
0.0244
0.0244
0.0244
0.0244
0.0244
0.122
none
none
none
none
none
none
none
none
none
O2
O2
O2
O2
O2
O2
O2
O2
O2
O2
O2
O2
O2
air
0 °C
–10 °C
0 °C
3.25
3.75
2
0.75
3.5
6
3
3.5
4
0.6
0.75
3
1.25
2.5
A
–
–
B
B
–
B
B
B
B
B
B
B
B
10
39
10
no reaction
no reaction
–10 °C
–30 °C
–10 °C
–10 °C
–10 °C
–10 °C
–10 °C
–10 °C
–40 °C
–40 °C
–40 °C
25
20
3
46
9
4
LiHMDS THF
LiHMDS DMF/THF[e]
LiHMDS CH2Cl2/Et2O[e]
LiHMDS toluene
no reaction
18
11
40
34
40
36
60
61
16
18
28
0
0
18
0
45
34
14
19
6
25
10
7
9
LiHMDS CH2Cl2/toluene[f]
LiHMDS CH2Cl2/toluene[f]
LiHMDS CH2Cl2/toluene[f]
LiHMDS CH2Cl2/toluene[f]
LiHMDS CH2Cl2/toluene[f]
LiHMDS CH2Cl2/toluene[f]
10
11
12
13
14
none
0.122
0.122
0.122
0.122
P(OEt)3
P(OEt)3
P(OMe)3
P(OMe)3
0
15
LiHMDS CH2Cl2/toluene[f]
0.122
P(OMe)3
air
–78 °C
7
B
66
56
0
8
16
17
18
19
20
LiHMDS toluene
0.122
0.122
0.122
0.122
0.122
P(OMe)3
P(OMe)3
P(OEt)3
P(OEt)3
P(OMe)3
O2
O2
O2
O2
O2
–30 °C
r.t.
r.t.
r.t.
r.t.
3
4.5
3
5
4.3
B
–
B
B
B
0
17
LiHMDS Diglyme/toluene[e]
no reaction
14
tBuONa
tBuOK
NaH
tBuOH/DMF
tBuOH/DMF
DMF
0
5
5
0
18
no reaction
21[g]
LiHMDS CH2Cl2/THF
0.244
P(OMe)3
air
–78 °C
5
C
70
0
8
[a] 1.5 equiv. of base were always used. HDMS is bis(trimethylsilyl)amide. [b] 1.15 equiv. of the additive were always used. [c] In all cases
bubbling of O2 or air into the reaction mixture was started 5 min after addition of the appropriate base. [d] Work-up conditions:
(A) quenching the reaction mixture with MeOH/H2O, aqueous KI, and AcOH, and then extraction with CH2Cl2; (B) quenching with
aqueous NH4Cl and extraction with CH2Cl2 after adjusting the pH to 10 with K2CO3; and (C) quenching with H2O followed by overnight
stirring, and then extraction with CH2Cl2. [e] The crude product was fractionated by chromatography affording a fraction containing
unseparated 1, 2, and 12. Then NMR was used to establish the relative ratio of these three compounds. [f] In these cases, a solution of
LiHMDS in THF, Et2O, or toluene was added to the solution of 1 in the first solvent. [g] Reaction was carried out on 10 g scale. For
details, see Exp. Sect.
Eur. J. Org. Chem. 2011, 2319–2325
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
2321

R. Riva, M. Rasparini et al.
FULL PAPER
easily deprotonated thanks to possible resonance stabiliza- both 1 and 12. The partial conversion was a problem too,
tion, that is, at C-9 and the methyl group. The preferential because separation of 1 from unreacted 2 was rather diffi-
hydroxylation at C-9 indicates that deprotonation at C-9 to cult.
give 17 is favored. This enolate can undergo a Single Elec-
tron Transfer (SET) process with triplet oxygen, giving the
radical 18, which accounts for the formation of dimer 15,[22]
and the superoxide anion. Combination of these two radi-
cals generates 19, which is probably in equilibrium with 20.
Anion 19 can in turn react with a reducing agent to give
alkoxide 21. The nature of the reducing agent is unclear
according to the literature precedents. A likely possibility is
that one molecule of 19 interacts in a redox process with
one molecule of anion 17 to give two molecules of alkoxide
21.[23]
Therefore we tried to optimize the reaction (see Table 1)
to reach a full conversion, to minimize formation of ketone
12, and to maximize the yield of 1 (Table 1). We first no-
ticed that the reductive work-up was not necessary, as we
obtained the same yields without it. Thus the hydroperoxide
species seemed to be completely converted under the reac-
tion conditions to afford either 1 or 12. With LDA in THF
(tetrahydrofuran), the conversions were always low and
increasing the reaction times was not beneficial. Thus we
screened various combinations of bases and solvents. The
solvent was particularly critical, because of the poor solu-
On the other hand, elimination of a hydroxide anion bility of risperidone (2) in most of them, and was reflected
from 20 gives ketone 12. This ketone could in principle be in a low active concentration of the substrate. Risperidone
derived from 1 (or 21) by a second oxygenation process. (2) was sufficiently soluble at low temperatures only in
However, a control experiment showed that paliperidone (1) THF, CH2Cl2, and alcohols. As shown by Entries 1–9, the
is not converted into 12 under the same reaction conditions. best results (40% yield) were obtained by addition of lith-
Finally anion 19 can also undergo an intramolecular pro- ium bis(trimethylsilyl)amide (LiHMDS) in toluene to a
ton transfer to form 22 which can then react with a second solution of 2 in CH2Cl2 (Entry 9). However, there was still
oxygen molecule to eventually lead to 16, analogous to the partial conversion, and substantial amounts of ketone 12
conversion of 17 to 1. Diol 16 could also be derived from were present. A real breakthrough occurred with an in-
an initial hydroxylation of the methyl group followed by a crease of substrate concentration, which led to full conver-
second hydroxylation at C-9. In this case, the product of sion (Entry 10). Unfortunately, the ratio of 1/12 was still
the single hydroxylation at the methyl site should have been unacceptably low.
formed as well, but it was never detected it in the crude
reaction mixtures.
To increase this ratio, we should accelerate the conver-
sion of hydroperoxide anion 19 into 21, compared to the
In this scenario, the reduction of 19 to 21 seems to be disproportionation leading to ketone 12. We reasoned that
the most critical step. This reaction is probably slow, as the introduction of a reducing agent in the reaction media
demonstrated in the literature that dioxygen oxidation of could be the ideal solution. For example, this reducing
enolates may be stopped, under appropriate circumstances, agent could be a sulfide, a phosphane, or a phosphite.[27,28]
at the hydroperoxide level.[23–26] The low reduction rate of For practical reasons we chose a phosphite, because of its
19 allows, through an intramolecular process, the conver- lower cost and the possibility of easily eliminating the re-
sion to ketone 12. As already stated, 12 is not a useful inter- sulting phosphates by extraction. This hypothesis proved to
mediate. When the crude mixtures obtained from the oxi- be correct. A comparison of Entry 10 with Entry 11 shows
dation reaction were treated with NaBH4 prior to purifica- that the ratio 1/12 increases from 1.8:1 to 6.7:1. A further
tion, the yields of 1 turned out to be lower, due to decom- improvement in the isolated yield of 1 and in the 1/12 ratio
position processes involving the two heterocyclic rings of was achieved by using P(OMe)3, instead of P(OEt)3, and by
Scheme 3.
2322
www.eurjoc.org
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2011, 2319–2325

Synthesis of Paliperidone
lowering the reaction temperature (Entries 13–15). More-
over we noticed that the reaction worked similarly by bub-
bling air instead of oxygen, with obvious advantages from
the procedural safety point of view. Under the best condi-
tions (Entry 15), an isolated yield of 66% of 1 was obtained,
with a 1/12 ratio higher than 8:1. This high ratio also makes
purification of 1 by chromatography easier, because 1 and
12 have similar polarities.
Having found benefits in the use of phosphites, we also
checked if these additives could improve the reactions with
other base/solvent combinations, but this was unsuccessful
(Entries 16–20).
Experimental Section
NMR spectra were recorded at 300 MHz (1H), and 75 MHz (13C)
with tetramethylsilane (1H NMR in CDCl3: δ = 0.000 ppm) or
CDCl3 (13C in CDCl3: δ = 77.16 ppm) as the internal standard.
Chemical shifts are reported in ppm (δ scale). Peak assignments
were made with the aid of gCOSY (gradient-selected COSY) and
gHSQC (gradient-selected heteronuclear single quantum corre-
lation) experiments. LRMS data were recorded with an Agilent
6310 instrument. Flash column chromatography was done using
220–400 mesh silica.
General Procedure for the Small Scale Oxidation of 2 (Table 1, En-
tries 10–15): A solution of risperidone (1, 500 mg, 1.22 mmol) in
The conditions for Entry 15 were used substantially for dry CH2Cl2 (10 mL) was cooled under an Ar atmosphere to the
appropriate temperature. Then (in the case of Entries 11–15), tri-
methyl or triethyl phosphite (1.58 mmol) was added. LiHDMS
(0.68 m in toluene, 2.69 mL, 1.83 mmol), freshly prepared from
bis(trimethylsilyl)amine and nBuLi (1.6 m in hexanes), was added
during 2 min. After 5 min, dioxygen or air (chromatographic grade)
was bubbled into the solution (about 1 bubble per second) for the
appropriate time always maintaining the solution at the same tem-
perature. At the end, after stopping the gas flow, the reaction was
quenched with saturated aqueous NH4Cl (5 mL) and saturated
aqueous K2CO3 (5 mL). After checking that the pH was between
scaling-up the methodology. We were able to carry out the
reaction on a 10 g scale (Entry 21), increasing even further
the concentration, and attain an isolated yield of pure 1
(after chromatography) of 70%. However, working in a
more concentrated solution required the substitution of tol-
uene (solvent for the base) with THF to avoid the partial
precipitation of 2 at low temperature. In this procedure we
also isolated ketone 12 (8%) and detected, by HPLC analy-
sis of the crude product, the presence of 3–4% of diol 16.
Preliminary results have shown that pure 1 can also be ob- 10 and 11, and adjusting it if needed, the mixture was extracted
with CH2Cl2 (3ϫ), and the organic extracts were washed with brine
and dried with Na2SO4. The solvents were evaporated to dryness,
and the crude product was purified by chromatography (CH2Cl2/
MeOH, from 90:10 to 85:15) to give a single fraction containing,
in the order of elution, 2, 12, and 1. After evaporation to dryness,
tained, with only a slight decrease in yield, by crystalli-
zation of the crude product without the need for
chromatography.
Although it was very difficult to obtain a pure sample of
16, after several chromatographic purifications and prepar-
ative TLC plates, we finally succeeded in isolating a few
milligrams of crystals. This made the X-ray diffraction
analysis feasible and allowed us to unambiguously establish
the molecular structure of 16 (Figure 1). The OH group
linked to C-9 was found to be disordered over the two pos-
sible positions, with occupation factors of 0.8 and 0.2 for
O-3 and O-3Ј atoms, respectively.
1
this fraction was weighed and examined by using H NMR to de-
termine the molar ratio of the three compounds. The yields re-
ported in Table 1 were determined from this ratio and from the
weight of the mixture.
Preparative Scale Synthesis of Paliperidone (1) from Risperidone (2):
A solution of risperidone (1, 10.0 g, 24.4 mmol) in dry CH2Cl2
(100 mL) was cooled to –78 °C under a N2 atmosphere and treated
with trimethyl phosphite (3.31 mL, 28.06 mmol, 1.15 equiv.). Then,
LiHDMS (1.0 m in THF, 25.6 mL, 1.05 equiv., commercially avail-
able from Chemetall) was slowly added so that the temperature did
not exceed –70 °C. After 5 min from the end of the base addition,
air (chromatographic grade) was slowly bubbled into the cooled
solution through a sintered steel plug until the disappearance of
the risperidone (2), as monitored by TLC (CH2Cl2/MeOH/satd.
NH4OH, 95:5:1, visualized with UV and ninhydrin) for 5 h typi-
cally. The mixture was quenched with water (50 mL) and stirred
overnight to destroy the excess phosphite. The phases were sepa-
rated, and the aqueous layer was extracted with CH2Cl2 (2ϫ).
Evaporation to dryness gave a crude foam which was purified by
chromatography (CH2Cl2/MeOH, from 90:10 to 80:10) to give first
ketone 12 (820 mg, 8%) and then 1 (7.29 g, 70%) as a foam, which
was pure according to TLC and NMR. The tail fractions weighed
1.56 g. HPLC analysis of the crude product showed that it con-
tained about 3–4% of the diol 16. An analytically pure sample of
1 was obtained by crystallization from ethyl acetate/n-hexane (m.p.
162.2–162.4 °C) or by crystallization from dimethylacetamide [m.p.
186.2 °C, differential scanning calorimetry (DSC)]. 1H NMR
(300 MHz, CDCl3, 297 K): δ = 1.76 (ddt, Jd = 4.3, 12.9 Hz, Jt =
9.5 Hz, 1 H, 8-H), 1.89–2.04 (m, 1 H), 2.07–2.21 (m, 6 H), 2.25–
2.42 (m, 2 H, CHHN of piperidine), 2.36 (s, 3 H, CH3), 2.55 (mc,
2 H, CH2CH2-N), 2.79 (mc, 2 H, CH2CH2-N), 3.10 (mc, 1 H, CH-
isoxazolyl), 3.18 (br. d, J = 11.4 Hz, CHHN of piperidine), 3.87–
4.03 (m, 2 H, 6-H), 4.42 (s, 1 H, OH), 4.51 (dd, J = 6.3, 10.2 Hz,
Figure 1. X-ray structure of 16 and selected atom labelling. ORTEP
view at the 30% probability level.
In conclusion, careful optimization of the hydroxylation
of 2 with dioxygen (or air) under basic conditions led to an
efficient synthesis of highly valuable paliperidone from its
cheaper parent risperidone in one step. Moreover, we feel
that the optimization study described in this paper will also
be helpful in future synthetic approaches involving autooxi-
dations of heterocyclic benzylic positions under basic con-
ditions. Indeed, this methodology has been used very sel-
dom so far, notwithstanding its great potential for the syn-
thesis of biologically active compounds.[18–20,22,29–31]
Eur. J. Org. Chem. 2011, 2319–2325
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
2323

R. Riva, M. Rasparini et al.
FULL PAPER
1 H, 9-H), 7.07 (dt, Jd = 2.2 Hz, Jt = 8.8 Hz, 5-H of benzoisox-
azole), 7.25 (dd, J = 2.2, 9.0 Hz, 7-H of benzoisoxazole), 7.72 (dd,
J = 5.4, 9.0 Hz, 4-H of benzoisoxazole) ppm. 13C NMR (75 MHz,
CDCl3, 297 K): δ = 18.5 (CH2), 21.2 (CH3), 24.0 (CH2), 27.1
(CH2), 30.7 (2 CH2), 34.7 (CH), 42.5 (CH2), 53.5 (2 CH2), 56.7
(CH2), 67.1 (CH), 97.5 (d, J = 27.0 Hz, CH), 112.4 (d, J = 25.3 Hz,
CH), 117.4 (Cquat), 120.6 (Cquat), 122.7 (d, J = 11.5 Hz, CH), 157.5
(Cquat), 157.9 (Cquat), 161.2 (Cquat), 162.1 (Cquat), 163.9 (d, J =
= 425.3 [M + H]+. C23H25FN4O3 (424.47): calcd. C 65.08, H 5.94,
F 4.48, N 13.20; found C 65.21, H 6.02, F 4.48, N 13.18.
Enamine 14: Ketone 12 was also characterized by converting it to
enamine 14. When the crude product, derived from risperidone oxi-
dation under the conditions of Entry 10, was purified by
chromatography with CH2Cl2/MeOH/saturated NH4OH, ketone 12
was quantitatively converted into enamine 14. (With high quanti-
ties of 12, it was difficult to separate 12 from 1 using only CH2Cl2/
MeOH). Enamine 14 was easier to characterize by NMR, because
of the lack of the keto-enol equilibrium, that is, there was no evi-
dence of the imine. 1H NMR (300 MHz, CDCl3, 297 K): δ = 1.68
(br. s, 2 H, NH2), 2.02–2.17 (m, 3 H), 2.20–2.36 (m, 3 H), 2.34 (s,
3 H, CH3), 2.43 (dt, Jd = 4.8 Hz, Jt = 7.2 Hz, 2 H, 7-H), 2.49–2.62
(m, 2 H), 2.72–2.90 (m, 2 H), 3.01–3.23 (m, 3 H), 4.11 (t, J =
7.2 Hz, 2 H, 6-H), 5.49 (t, J = 4.8 Hz, 8-H), 7.05 (dt, Jd = 2.0 Hz,
Jt = 9.0 Hz, 5-H benzoisoxazole), 7.24 (dd, J = 1.5, 8.4 Hz, 7-H of
benzoisoxazole), 7.72 (dd, J = 5.1, 8.4 Hz, 4-H of benzoisox-
azole) ppm. 13C NMR (75 MHz, CDCl3, 297 K): δ = 21.0 (CH2),
21.6 (CH3), 24.2 (CH2), 30.6 (2 CH2), 34.7 (CH), 39.4 (CH2), 53.5
(2 CH2), 56.7 (CH2), 97.5 (d, J = 26.8 Hz, CH), 104.3 (Cquat), 112.4
(d, J = 25.1 Hz, CH), 117.4 (Cquat), 121.5 (Cquat), 122.8 (d, J =
11.1 Hz, CH), 135.0 (Cquat), 147.9 (Cquat), 158.1 (Cquat), 161.2
(Cquat), 161.8 (Cquat), 164.0 (d, J = 13.7 Hz, Cquat), 164.2 (d, J =
249.0 Hz, Cquat) ppm. C23H26FN5O2 (423.48): calcd. C 65.23, H
6.19, N 16.54; found C 65.05, H 6.11, N 16.39.
13.2 Hz, Cquat), 164.0 (d, J = 248.5 Hz, Cquat) ppm. FT-IR: ν
=
˜
max
3287, 2934, 2783, 2754, 1616, 1535, 1413, 1338, 1269, 1183, 1143,
1130, 996, 955, 868, 853, 817, 792, 758, 698 cm–1. LRMS (ESI, ion
trap, 70 V): m/z (%) = 427 (9) [M + H]+, 209 (100). C23H27FN4O3
(426.48): calcd. C 64.77, H 6.38, F 4.45, N 13.14; found (sample
crystallized from dimethylacetamide) C 64.84, H 6.48, F 4.49, N
13.21. A small amount (few milligrams) of compound 16 was ob-
tained after several chromatographic purifications and preparative
TLCs of the tail fractions. A crystal suitable for single-crystal X-
ray diffraction was selected and used to determine the molecular
and crystal structure of the compound.
Crystal Data and Structure Refinement for 16: Colorless monoclinic
crystal of C23H27FN4O4 (0.78ϫ0.40ϫ0.04 mm in size), molecular
weight 442.5 g/mol, space group P21/n, a = 6.9000(2) Å, b =
21.9314(9) Å, c = 13.8711(6) Å, β = 92.212(1), V = 2097.5(2) Å3, Z
= 4, Dc = 1.401 gcm–3, room temperature. A total of 19923 reflec-
tions {independent 3284 [R(int) = 0.0248]} was collected by using
graphite monochromated Mo-Kα radiation (λ = 0.71069 Å) with a
SMART-APEX CCD area detector diffractometer. The structure
was solved by direct methods and refined by a full-matrix least-
squares procedure with anisotropic temperature factors for non-
hydrogen atoms. The hydrogen atoms were located in calculated
positions and refined in the riding model. Final R values were: R1
= 0.0380 and wR2 = 0.0958 for 308 refined parameters and 2932
observed reflections.
Allylated Risperidone 11: A solution of LDA was freshly prepared
from nBuLi (1.6 m in n-hexane, 4 mL, 6.40 mmol) and diisopropyl-
amine (995 μL, 7.04 mmol) in THF (8 mL). A solution of risperi-
done (2, 101.8 mg, 248 μmol) in dry THF (5 mL) was cooled under
argon to 0 °C and treated with the solution of lithium diispropyl-
amide (0.49 m in THF/n-hexanes, 755 μL, 372 μmol). At the end of
the addition, the cooling bath was removed. After 15 min, allyl
bromide (27.5 μL, 322 μmol) was added. After stirring for 15 min
at room temp., the reaction was quenched with saturated aqueous
NH4Cl, and the pH was adjusted to 11 with saturated aqueous
K2CO3. The mixture was extracted with Et2O, and the organic ex-
tracts were washed with brine and dried with Na2SO4. Evaporation
of the solvent gave a crude product which was purified by
chromatography (CH2Cl2/MeOH, from 97:3 to 92:8) to give pure
11 (34.9 mg, 31%). The 9,9Ј-diallylated product (9.9 mg, 8%) and
recovered 2 (28.2 mg) were also isolated. The yield of 11 based on
recovered starting material was 43%. 1H NMR (300 MHz, CDCl3,
297 K): δ = 1.60–1.74 (m, 1 H), 1.80–2.20 (m, 7 H), 2.24–2.46 (m,
3 H), 2.33 (s, 3 H, CH3), 2.55 (m, 2 H, CH2CH2-N), 2.72–2.93
(m, 4 H, CH2CH2-N and CH-CHH-CH=CH2), 3.09 (m, 1 H, CH-
isoxazolyl), 3.19 (br. d, J = 11.1 Hz, CHHN of piperidine), 3.85–
4.01 (m, 2 H, 6-H), 5.02–5.20 (m, 2 H, CH2=CH), 5.80 (dddd, J =
6.0, 7.8, 10.2, 17.7 Hz, CH=CH2), 7.06 (dt, Jd = 2.0 Hz , Jt =
8.8 Hz, 5-H benzoisoxazole), 7.24 (dd, J = 2.0, 8.7 Hz, 7-H of
benzoisoxazole), 7.72 (dd, J = 5.1, 8.7 Hz, 4-H of benzoisox-
azole) ppm. 13C NMR (75 MHz, CDCl3, 297 K): δ = 20.0 (CH2),
21.5 (CH3), 23.9 (CH2), 24.0 (CH2), 30.6 (2 CH2), 34.7 (CH), 38.3
(CH2), 39.9 (CH), 42.9 (CH2), 53.5 (2 CH2), 56.8 (CH2), 97.5 (d, J
= 26.5 Hz, CH), 112.4 (d, J = 25.4 Hz, CH), 117.4 (Cquat), 117.6
(CH2), 119.1 (Cquat), 122.8 (d, J = 11.1 Hz, CH), 135.8 (CH), 158.4
(Cquat), 158.6 (Cquat), 161.2 (Cquat), 162.8 (Cquat), 164.0 (d, J =
13.3 Hz, Cquat), 164.2 (d, J = 249.0 Hz, Cquat) ppm. LRMS (ESI,
ion trap): m/z = 451.1 [M + H]+. C26H31FN4O2 (450.55): calcd. C
69.31, H 6.94, N 12.44; found C 69.57, H 6.95, N 12.27.
CCDC-801682 contains the supplementary crystallographic data
for this paper. These data can be obtained free of charge from The
Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/
data_request/cif.
Independent Synthesis of Ketone 12 from Paliperidone (1): A solu-
tion of paliperidone (1, 1.00 g, 2.44 mmol) in glacial acetic acid
(12 mL) and acetonitrile (3 mL) was cooled to 0 °C and treated
with aqueous NaOCl (15%, 2 mL). After stirring for 1 h at 0 °C,
the mixture was diluted with water (150 mL) and neutralized with
NaHCO3 (pH = 7). The reaction mixture was extracted with
CH2Cl2 (3ϫ) and dried with Na2SO4. Evaporation of solvent fol-
lowed by purification of the crude product by chromatography
(CH2Cl2/MeOH, between 100:0 and 90:10) gave pure 12 as a foam
(453 mg, 45%). An 82:18 mixture of ketone and enol was present
1
1
in the H NMR sample. H NMR (300 MHz, CDCl3, 297 K): δ =
2.00–2.15 (m, 4 H), 2.18–2.40 (m, 4 H), 2.47 (s, 3 H, CH3), 2.48–
2.63 (m, 2 H), 2.78–2.90 (m, 4 H), 3.02–3.22 (m, 3 H), 4.11 (dt, Jd
= 5.1 Hz, Jt = 7.5 Hz, 0.18 H, 6-H of enol), 4.19 (t, J = 5.8 Hz,
0.82 H, 6-H of ketone), 5.68 (t, J = 4.6 Hz, 0.18 H, CH=C of enol),
7.04 (dt, Jd = 2.1 Hz, Jt = 8.8 Hz, 5-H benzoisoxazole), 7.22 (dd,
J = 2.0, 8.6 Hz, 7-H of benzoisoxazole), 7.68 (dd, J = 5.1, 8.6 Hz,
4-H of benzoisoxazole) ppm. Only the signals of the ketone are
reported in the 13C NMR data. 13C NMR (75 MHz, CDCl3,
297 K): δ = 20.5 (CH2), 21.6 (CH3), 24.5 (CH2), 30.6 (2 CH2), 34.5
(CH), 37.1 (CH2), 42.4 (CH2), 53.4 (2 CH2), 56.2 (CH2), 97.4 (d, J
= 26.5 Hz, CH), 112.3 (d, J = 25.1 Hz, CH), 117.3 (Cquat), 122.6
(d, J = 11.1 Hz, CH), 126.6 (Cquat) 145.6 (Cquat), 158.7 (Cquat), Supporting Information (see footnote on the first page of this arti-
161.1 (Cquat), 161.2 (Cquat), 163.8 (d, J = 13.3 Hz, Cquat), 164.1 (d,
J = 249.0 Hz, Cquat), 189.5 (Cquat) ppm. LRMS (ESI, ion trap): m/z
cle): 1H NMR, 13C NMR, IR, LRMS, HPLC and DSC of com-
1
pound 1; H NMR and 13C NMR of compounds 11 and 12.
2324
www.eurjoc.org
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2011, 2319–2325

Synthesis of Paliperidone
[16] S. G. Davies, D. R. Fenwick, O. Ichihara, Tetrahedron: Asym-
metry 1997, 8, 3387–3391.
Acknowledgments
[17] F. Stappers, R. Broeckx, S. Leurs, L. Van Den Bergh, J. Agten,
A. Lambrechts, D. Van den Heuvel, D. De Smaele, Org. Pro-
cess Res. Dev. 2002, 6, 911–914.
[18] J. Gutzwiller, M. R. Uskokovic, J. Am. Chem. Soc. 1978, 100,
576–581.
We thank Dr. Katharine Powles for her precious help in improving
the manuscript, and Fondazione San Paolo for financial support
(purchase of an NMR spectrometer).
[19] G. Stork, D. Niu, A. Fujimoto, E. R. Koft, J. M. Balkovec,
J. R. Tata, G. R. Dake, J. Am. Chem. Soc. 2001, 123, 3239–
3242.
[20] Y. Houminer, E. W. Southwick, D. L. Williams, J. Org. Chem.
1989, 54, 640–643.
[21] G. A. Russell, E. G. Janzen, A. G. Bemis, E. J. Geels, A. J.
Moye, S. Mak, E. T. Strom, Oxidation of Hydrocarbons in Basic
Solution, in: Advances in Chemistry, vol. 51 (Selective Oxidation
Processes) (Ed.: E. K. Fields), American Chemical Society,
Washington, DC, 1965, pp. 112–172.
[1] P. Seeman, Can. J. Psychiatry 2002, 47, 27–38.
[2] P. Seeman, Am. J. Psychiatry 2005, 162, 1984–1985.
[3] I. Locatelli, M. Kastelic, J. Koprivsek, B. Kores-Plesnicar, A.
Mrhar, V. Dolzan, I. Grabnar, Eur. J. Pharm. Sci. 2010, 41,
289–298.
[4] J. Fang, M. Bourin, G. B. Baker, Naunyn-Schmiedeberg’s Arch.
Pharmacol. 1999, 359, 147–151.
[5] J. A. Bork, T. Rogers, P. J. Wedlund, J. de Leon, J. Clin. Psychi-
atry 1999, 60, 469–476.
[6] L. P. H. Yang, G. L. Plosker, CNS Drugs 2007, 21, 417–425.
[7] C. G. M. Janssen, A. G. Knaeps, L. E. J. Kennis, J. Vanden-
berk, US Patent 5158952, 1989; Chem. Abstr. 1993, 118,
240939.
[22] A. Ohta, M. Ohta, Heterocycles 1982, 17, 151–154.
[23] D. A. Konen, L. S. Silbert, P. E. Pfeffer, J. Org. Chem. 1975,
40, 3253–3258.
[24] H. H. Wasserman, B. H. Lipshutz, Tetrahedron Lett. 1975, 16,
[8] J. Bartl, P. Benovsky, PCT WO 2010003702A1, 2009; Chem.
1731–1734.
Abstr. 2010, 152, 144708.
[25] N. J. Lawrence, O. Lamarche, N. Thurrab, Chem. Commun.
1999, 689–690.
[26] H. R. Gersmann, A. F. Bickel, J. Chem. Soc. B 1971, 2230–
2237.
[9] This oxidation was indeed reported in the Teva patent applica-
tion EP2035571A1, “deemed to be withdrawn” on 03/03/2010
(EPO database). In this patent the isolated yield of 1 was only
8%.
[27] J. N. Gardner, F. E. Carlon, O. Gnoj, J. Org. Chem. 1968, 33,
[10] M. D. Garrett, R. Scott, G. N. Sheldrake, Tetrahedron: Asym-
metry 2002, 13, 2201–2204.
[11] S. P. Tanis, T. T. Parker, J. R. Colca, R. M. Fisher, R. F. Kletz-
ein, J. Med. Chem. 1996, 39, 5053–5063.
[12] V. Boekelheide, W. J. Linn, J. Am. Chem. Soc. 1954, 76, 1286–
1291.
3294–3297.
[28] P. M. Burden, H. T. A. Cheung, T. R. Watson, J. Chem. Soc.
Perkin Trans. 1 1987, 169–172.
[29] M. Berlin, R. Aslanian, M. de Lera Ruiz, K. D. McCormick,
Synthesis 2007, 2529–2533.
[30] Y. Houminer, E. B. Sanders, J. Heterocycl. Chem. 1980, 17,
647–649.
[31] K. Teranishi, T. Goto, Bull. Chem. Soc. Jpn. 1990, 63, 3132–
[13] J. Krajcovic, J. Bartl, P. Benovsky, US Patent 20100069633A1,
2009; Chem. Abstr. 2010, 152, 144702.
[14] R. Riva, L. Banfi, A. Basso, P. Zito, Org. Biomol. Chem. 2011,
3140.
DOI: 10.1039/C0OB00978D, in press.
Received: December 1, 2010
[15] F. A. Davis, A. C. Sheppard, Tetrahedron 1989, 45, 5703–5742.
Published Online: March 4, 2011
Eur. J. Org. Chem. 2011, 2319–2325
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
2325