The development of a practical and scalable process for the asymmetric synthesis of sitagliptin is reported. Density functional theory calculations reveal that two noncovalent interactions are responsible for the high diastereoselection. The first is an intramolecular hydrogen bond between the enamide NH and the boryl mesylate S═O, consistent with MsOH being crucial for high selectivity. The second is a novel C-H···F interaction between the aryl C5-fluoride and the methyl of the mesylate ligand.
The development of a practical and scalable process for the asymmetric synthesis of sitagliptin is reported. Density functional theory calculations reveal that two noncovalent interactions are responsible for the high diastereoselection. The first is an intramolecular hydrogen bond between the enamide NH and the boryl mesylate S═O, consistent with MsOH being crucial for high selectivity. The second is a novel C-H···F interaction between the aryl C5-fluoride and the methyl of the mesylate ligand.
Stereoselective enamine reduction is one
of the most useful methods for the synthesis of enantiomerically pure
β-amino carbonyls. In order to attain stereoinduction in these
transformations, a sacrificial chiral auxiliary[1−6] or asymmetric metal-catalyzed method is used.[7−9] A noteworthy
example is found in the commercial synthesis of sitagliptin (2) from reduction of enamine 1 using an expensive
rhodium/chiral phosphine combination (Scheme 1, A).[10] The phosphate salt of sitagliptin,
discovered by Merck, has been approved by the US FDA for the management
of type 2 diabetes mellitus.[11] Sitagliptin
inhibits the proteolytic activity of dipeptidyl peptidase-4,[12] an enzyme that breaks down the incretins,[13] which play a key role in glucoregulation by
increasing insulin secretion and suppressing glucagon release. More
recently, an elegant biocatalytic route to sitagliptin utilizing an
engineered transaminase enzyme to covert dicarbonyl 3 to 2 has been developed and commercialized (Scheme 1, B).[14] Although numerous
reports on the synthesis of sitaliptin have appeared,[10,17,15] to the best of our knowledge,
an inexpensive reagent-based diastereoselective enamine reduction
toward sitagliptin remains unprecedented.
Scheme 1
Asymmetric Routes to Sitagliptin
Herein, we report
such an alternate approach exploiting the combination of an inexpensive
reducing agent (NaBH4/MsOH) and an inexpensive chiral auxiliary
(phenethylamine)[16] to accomplish the diastereoselective
reduction of β-enamino amide 4 to provide advanced
intermediate 5 (Scheme 1, C).
This approach also led to the diasteroselective reduction of several
analogues of 4, demonstrating the potential to provide
a general route for highly efficient and diastereoselective reductions
of enamines. Further, density functional theory was used to assess
the role of the chiral auxiliary and substrate on the observed diastereoselectivity.
Intramolecular hydrogen bond and CH···F interactions
were shown to play key roles in controlling the stereochemical outcome.
Subsequent experiments with additional substrates provided support
for the computed models.We initiated our studies by
surveying the entire synthetic landscape[17] to 2 to assess the relative merits of all the hitherto
known strategies. Although a few chiral auxiliary enamine reduction
approaches are reported toward 2, they utilize costly
metal catalysts.[18] Therefore, we attempted
to develop a homogeneous reduction system using inexpensive hydride
reagents. At the outset of our endeavor, we considered the application
of widely used chiral amine (R)-(+)-phenyl ethyl
amine as an auxiliary. Notably, this material is generated on large
scale[19] and is readily available at low
cost. Inspection of the key proposed substrate 4 reveals
an α,β-unsaturated amide which permits a very different
approach relative to metal-catalyzed hydrogenation. It was envisioned
that the carbonyl group of the enone would engage in hydrogen bonding
with the amino group of the auxiliary giving rise to conformationally
well-defined system 4 where the amino group is predominantly
sp2.[20] Consequently, allylic
strain[21] would cause the auxiliary to impose
a facial bias during external delivery of hydride to the β-site.
Specifically, approach to the less-hindered bottom face of the enamide
as shown in 4-HBR of Scheme 2 would generate the targeted stereocenter. Intermediate 4 was synthesized by combining ketoamide 3 and
(R)-(+)-phenylethylamine in toluene with 1.6 equiv
of acetic acid at 110 °C.[22] In this
experiment, only the Z-isomer was observed in the
quantitative conversion to 4 due to the highly stabilizing
hydrogen bond described above (see also below).[20]
Scheme 2
Synthetic Route to Sitagliptin
Having achieved an expedient synthesis of 4, the next task was to devise a suitable reagent system for
reduction. In this context, several borohydride species precedented
in the this reduction [NaBH4, NaB(OAc)3H, and
NaBH3CN][20] were studied but
resulted in no conversion with the β-enamino amide being recovered
(Scheme 2). Undeterred, we hypothesized that
a Lewis acidic reductant would allow internal hydride delivery, which
is not possible with the aforementioned reductants. Addition of 5–6
equiv of MsOH to 2 equiv of NaBH4 generated the requisite
B(MsO)2H in situ.[23] The internal
hydride delivery mode provides stereocontrol because interactions
from the (R)-(+)-phenethylamine would dictate the
geometry in the resultant 6-membered transition state (Scheme 3). To our delight, this combination afforded 5 in 71% yield with the target diastereomer predominating
(dr 92:8).[24] Furthermore, reaction of ent-4 with the NaBH4/MsOH reagent
system afforded ent-5 with same selectivity
and efficiency. Hydrogenolysis of the desired diastereomer 4 by means of 10% Pd/C (50% wet) in i-PrOH/water
at 65–70 °C and 105 psi hydrogen pressure for 10 h readily
afforded sitagliptin (1) in 88% yield. The diastereoselectivity
in the previous step was retained in this transformation and the Pd
catalyst could be recovered and regenerated.[25] Significantly, the previously reported enantioselective catalytic
hydrogenation[10] or diastereoselective catalytic
hydrogenations[18] all utilized high cost
transition metals (Rh, Ru, Pt). As such, this strategy represents
a cost competitive approach.
To gain insight into the mechanism
and diastereoselectivity, DFT (B3LYP and M062X) calculations[26] were undertaken on the model system using BH3 as the hydride source (Scheme 3).
In accord with experiment, the Z-6 isomer
is lower in energy (by ∼8 kcal/mol) than the E-6 enamine isomer due to an intramolecular hydrogen bond.
Further, the barrier for interconversion between these two isomers
is high (∼42 kcal/mol). Complexation with BH3 is
energetically favored (by ∼3–8 kcal/mol) and significantly
reduces the E/Z interconversion barrier (∼30
kcal/mol); however, this barrier is still high relative to the experimental
conditions (−90 °C) and isomerization was therefore not
considered. Finally, calculations showed that the hydride transfer
process occurs via a 6-membered transition state with overall barrier
of 23.5 kcal/mol (from Z-6-int) and
is slightly downhill in energy. Subsequent decomplexation/enol–keto
tautomerization delivers the amino amide product.
Free Energies of
the Model System Reduction
Conditions: 298
K, kcal/mol, B3LYP/6-31G(d) [toluene, CPCM].Next, the origin of the observed diastereoselectivity was investigated
by performing extensive conformational modeling of the transition
states with the leading to (R,R)-
and (R,S)-diastereomers using model
substrates Z-7-int (R= phenyl and 2,4,5-F3C6H2) and BH3 as hydride
source (see the Supporting Information for
structures and energetics). However, these calculations failed to
account for the observed diastereoselectivity predicting a low diastereomeric
ratio favoring the wrong stereoisomeric product (see Supporting Information).Experimentally it was shown
that the nature of the hydride source had a significant effect on
the overall diastereoselectivity (see Scheme 2). As depicted in Figure 1, modeling B(OMs)2H as hydride source correctly predicts the major diastereomer
(R,R)-7 albeit only
slightly [a Boltzmann distribution of the transition states gives
a 35:65 ratio favoring (R,R)-7]. This result highlights the subtle but important interactions
the OMs groups have on the diastereoselectivity (see the Supporting Information for other conformers).
In addition, these results show a different conformation than that
expected based on allylic strain considerations; namely, the methyl
group of the α-methylbenzyl is oriented anti-periplanar to the enamide C–N bond due to favorable π-stacking
interactions between the chiral auxiliary arene and the substrate
arene. As a consequence, hydride approach occurs over the α-methylbenzyl
hydrogen rather than the phenyl in the lowest energy approaches (Figure 1, Z-7-TS-eq-cis and Z-7-TS-ax-trans); other subtle, nonbonded interactions account for
the energetic differences. Notably, modeling with the substrate containing
the trifluorinated aryl group (values in parentheses) increased the
overall diastereoselectivity in excellent agreement with experiment;
the Boltzmann distribution predicts 5:95 ratio in favor of the (R,R)-diastereomer. The fluorinated
aryl ring caused the OMs groups to adopt the conformation shown in Z-7-F-TS-eq-cis (Scheme 4).
Figure 1
Effect of boron substitution on the diastereoselectivity.
Relative enthalpies (kcal/mol) at the B3LYP/6-31G(d) in toluene (CPCM).
The first value for each compound is the result for Ar = Ph, the value
in parentheses is for Ar = 2,4,5-F3C6H2. Structures shown with Ar = Ph.
Scheme 4
. Effect of Aryl Group Substitution on Diastereoselectivity
Relative enthalpies, (kcal/mol),
B3LYP/6-31G(d) toluene (CPCM); values in brackets and selected distances
(Å) are for M062X/6-31G(d) toluene (CPCM) optimized geometries.
Effect of boron substitution on the diastereoselectivity.
Relative enthalpies (kcal/mol) at the B3LYP/6-31G(d) in toluene (CPCM).
The first value for each compound is the result for Ar = Ph, the value
in parentheses is for Ar = 2,4,5-F3C6H2. Structures shown with Ar = Ph.
. Effect of Aryl Group Substitution on Diastereoselectivity
Relative enthalpies, (kcal/mol),
B3LYP/6-31G(d) toluene (CPCM); values in brackets and selected distances
(Å) are for M062X/6-31G(d) toluene (CPCM) optimized geometries.Closer inspection of the transition-state structures
with the fully fluorinated aryl (Scheme 4)
revealed a key intramolecular hydrogen bond between the NH of the
amine auxiliary and one of the mesylate S=O groups that stabilizes
the Z-7-F-TS-eq-cis transition state structure. Further, this transition state benefits
from an intramolecular CH···F interaction “hydrogen
bridge”[27,28] between the one of the C–F
bonds of the aryl group and the CH3 of the methanesulfonyl.
The 2.27 Å intermolecular distance is shorter than the sum of
the van der Waals radii which is 2.54 Å.[29] This interaction accounts for the higher selectivity with the fluoro
analog. Although CH···F interactions have been observed
in crystal structures (CH-F distance 2.26 Å), they are rarely
invoked in influencing selectivities and reactivities.[30]To explore the effects of the fluorines
on the diastereoselectivity, several analogues with varying fluorine
patterns around the ring were synthesized and subjected to the same
reduction (Scheme 5). In agreement with computations,
the fluorines do play a role in on the experimentally observerd diasteroselectivity.
Overall, the 5-fluoro substituent has the greatest impact on the stereoselectivity,
consistent with the calculations. However, the effects of the fluoro
groups do not appear additive indicating that additional factors such
as their influence on the conformational space are likely to play
a role in the diastereoselective reduction (see the Supporting Information). Due to the computational cost and
minor changes on diastereoselectivity (within the limits of computational
errors) the role of each individual fluorine atom was not explored
further.
Scheme 5
Effect of Aryl Group Substitution on Experimental Diastereoselectivity
In comparison to previously
reported synthesis of sitagliptin, this strategy offers a very efficient
and workable protocol with at least four salient features: (1) facile
stereoselective access to a single Z-isomer starting
material, (2) a protocol that does not require high hydrogen pressures,
(3) high conversion, and (4) very good diastereoselectivity (92:08).
Such a diastereoselective strategy permits ready separation of the
minor diastereomer from the mixure to obtain the major diastereomer
in very pure form (>99.9:0.1), thereby increasing the isolated
yield and overall efficiency of the procedure. Further, this diastereoselective
approach does not require repeated purification to achieve stereochemical
enrichment, as is the case in alternative enantioselective approaches.
In conclusion, a scalable and economical process has been developed
to access advanced sitagliptin intermediate 5 via homogeneous
reagent-based reduction with excellent diastereoselectivity. Density
functional theory was used to gain insight into the observed mechanism
of asymmetric induction. Analysis of the relevant tranisiton states
revealed two key noncovalent interactions, an intramolecular hydrogen
bond and a novel C–H···F interaction, between
the chiral auxiliary and the hydride source that serve as control
elements.
Authors: Barbara Weiner; Wiktor Szymański; Dick B Janssen; Adriaan J Minnaard; Ben L Feringa Journal: Chem Soc Rev Date: 2010-02-24 Impact factor: 54.564
Authors: Norihiro Ikemoto; David M Tellers; Spencer D Dreher; Jinchu Liu; Angie Huang; Nelo R Rivera; Eugenia Njolito; Yi Hsiao; J Christopher McWilliams; J Michael Williams; Joseph D Armstrong; Yongkui Sun; David J Mathre; Edward J J Grabowski; Richard D Tillyer Journal: J Am Chem Soc Date: 2004-03-17 Impact factor: 15.419
Authors: Karl B Hansen; Yi Hsiao; Feng Xu; Nelo Rivera; Andrew Clausen; Michele Kubryk; Shane Krska; Thorsten Rosner; Bryon Simmons; Jaume Balsells; Nori Ikemoto; Yongkui Sun; Felix Spindler; Christophe Malan; Edward J J Grabowski; Joseph D Armstrong Journal: J Am Chem Soc Date: 2009-07-01 Impact factor: 15.419
Authors: Debora Williams-Herman; Elizabeth Round; Arlene S Swern; Bret Musser; Michael J Davies; Peter P Stein; Keith D Kaufman; John M Amatruda Journal: BMC Endocr Disord Date: 2008-10-27 Impact factor: 2.763