Key words
C–H functionalization - photoredox catalysis - carbohydrates - flow chemistry - hydrogen
atom transfer - radicals
The natural abundance and diversity in stereochemistry of carbohydrates make them
interesting starting materials for the development of novel pharmaceutical compounds
and organic materials. The lion’s share of the literature on the modification of carbohydrates
in this context is focused on reactions involving the hydroxyl groups, such as ether
or ester formation.[1]
[2]
As discrimination between these multiple hydroxy groups is difficult, protecting group
strategies are commonly used. An alternative approach is direct carbon–carbon bond
formation, but apart from reactions with the carbonyl group in carbohydrates (such
as aldol and related reactions, and Strecker reactions) this approach is not often
pursued. With the advent of synthetically attractive photoredox chemistry,[3]
[4] site-selective photoalkylation has come into play as a powerful strategy to extend
the carbon framework of carbohydrates. Carbohydrates are intrinsically attractive
substrates for radical-based reactions involving the carbon skeleton, because the
hydroxy groups themselves are not very reactive in such a setting.[5] Nevertheless, the challenge of site-selectivity remains.
In 2015, the group of MacMillan made use of the fact that hydrogen bond formation
between an acceptor and a hydroxy group weakens the α-C–H bond strength of that hydroxy
group and this leads to preferential hydrogen atom transfer (HAT).[6] The produced carbon-centered radical can subsequently react with an electron-deficient
alkene, a ‘somophile’ (Scheme [1]).
Scheme 1 The addition of a hydrogen bond acceptor allows for α-C–H activation[6]
This activation is sufficiently large to discriminate between α-C–H bonds in ethers
or acetals and a free hydroxyl group, and this was illustrated in the selective alkylation
of a protected galactose derivative, possessing one hydroxy group (Scheme [2]).
Scheme 2 Previous literature
We subsequently showed that this photoredox alkylation reaction based on hydrogen
bond based activation can also be applied on unprotected carbohydrates.[7] The site-selectivity in these cases is substrate inherent and the hydrogen bond
formation functions to increase the reactivity (or makes the reaction possible at
all). In gluco- and allopyranosides it was observed that alkylation occurs selectively
at the C3 carbon (Scheme [3]). Furthermore, when the α-anomers of the substrates were used, alkylation took place
from the top face, resulting in an allo-configured product. This selectivity likely originates from the steric shielding
by the anomeric substituent. With β-methyl glucosides, only partial inversion was
observed.
Scheme 3 Isomerization of the C3-OH for α-methyl glucoside alkylation
Next to hydrogen bond formation, several other methods have been reported, with the
common theme that the electron density of the α-carbon is increased. Boron, tin and
silicon are less electronegative than hydrogen, so diarylborinic acids,[8] borinates,[9] spirosilanes[10] and organostannanes[11] have all been used successfully, and are instrumental to induce regioselectivity
in the alkylation, with the latter being applied to rhamnopyranosides. Alkylation
via 1,6-HAT in a fructopyranoside has been reported as well.[12]
Also, the nature of the hydrogen atom transfer agent has been studied, with quinuclidine
being favored. Recently however, quinuclidine-[13] and DABCO-derived[14] catalysts have also proven useful for facilitating HAT. This type of photocatalysis
is not limited to alkylations: in more recent reports, Wendlandt and co-workers have
used the HAT reaction for isomerization, providing access to rare monosaccharides
from commonly available monosaccharides,[15]
[16] whilst the group of Taylor employed the system for oxidations.[17] Redox isomerization, a strategy that combines the two, gives rise to ketodeoxysugars,
with examples of both furanosides[18] and pyranosides[19]
[20] having been investigated. The photocatalytic system has also seen success in chiral
resolutions: with a chiral phosphate, racemic ureas could be enantiomerically enriched.[21]
It is therefore clear that the versatility of these photoredox systems makes them
very interesting candidates to explore otherwise challenging chemistry. Despite this
progress however, the photoalkylation of carbohydrates has not yet been used to prepare
building blocks or scaffolds on a preparative scale. In order to be synthetically
useful, the procedure needs to be scalable and cost-effective. In addition, the variety
in somophiles is still rather limited, restricting structural diversity. To improve
upon this situation, we envisioned applying a continuous flow chemistry approach.
One of the key advantages of continuous flow chemistry is the much larger surface
area offered by the tubing. This allows for more efficient irradiation compared to
a batch setup. In addition, the amount of product does not depend on the scale of
the reaction but on the overall runtime.[22]
Since our primary goal was the upscaling of these known alkylation reactions, use
of the well-studied iridium catalyst Ir[dF(CF3)ppy]2(dtbbpy)PF6 (Table [1]) was not cost-effective. Iridium prices have increased substantially in recent times
and producing multigram quantities of product without being able to recycle the catalyst
would be far too costly. In 2020, Wendlandt and co-workers reported on the selective
epimerization of carbohydrates[15] by photoredox chemistry and in this study, the iridium catalyst was replaced by
the organocatalyst 4-CzIPN. Given the similarity of our reaction, as well as the range
of redox potentials, we hypothesized that 4-CzIPN might also be effective in our alkylation
reactions. In addition to being a metal-free alternative, 4-CzIPN is prepared in a
single step from readily available materials[23] and has greater absorption in the visible light region, which better matches the
used light source.
Table 1 Photocatalysts Used for the Reactions in this Study
 
|
|
4-CzIPN
|
Iridium catalyst
|
|
Redox potentials:[24]
|
|
E
1/2(P*/P–)
|
+1.35 V
|
+1.21 V
|
|
E
1/2 (P/P–)
|
–1.21 V
|
–1.37 V
|
|
λabs (in MeCN):[25]
|
435 nm
|
380 nm
|
To facilitate workup and purification of the photoalkylated products, we chose methyl
4,6-O-benzylidene-α-d-glucopyranoside as the substrate. Although the photoalkylation reaction does not
explicitly require the presence of the 4,6-O-benzylidene moiety, as we showed previously,[7] the starting material is readily prepared on large scale and soluble in acetonitrile.[26] In addition, the product is readily modified in subsequent transformations, a necessity
for a building block to act as a scaffold. Acrylonitrile was chosen as the initial
somophile due to its potential for subsequent derivatization.
To perform the flow reactions, we used an in-house constructed flow reactor,[27]
[28] equipped with 500 W input power of blue (460 nm) LEDs (Figure [1] and Figure [2]). Perfluoroalkoxy (PFA) tubing is wound around a water-cooled copper core and provides
an internal reactor volume of about 20 mL (ø 0.8 mm) along which the reaction mixture
could be irradiated.
To prevent overheating and thermal side reactions, also the LEDs were water-cooled,
maintaining a consistent temperature within the reactor. A continuous syringe pump
offered precise flow control and pumps the reaction mixture through the PFA tubing
at a constant rate. Behind the pump, a sample loop has been installed, which was used
to inject small volume samples for conditions screening, bypassing the internal volume
of the pump to prevent sample dilution.
Figure 1 Schematic of the flow reactor
Figure 2 Reactor setup, water-cooled internals
The study commenced by translating the reaction system from our established batch
conditions[7] to the continuous flow setup, after which several parameters were modified to optimize
product yield.
It is crucial that over the course of the reaction the mixture stays homogeneous,
as in-line crystallization causes the flow system to fail due to channel blockage.[29] Crystallization can take place both in the tubing and in the pump, imposing restrictions
on the maximum concentration at which the reaction can be carried out.
Experiments to maximize the conversion were carried out varying several parameters.
We observed that when the light output was reduced, the conversion dropped significantly,
in an almost linear fashion.[30] Subsequently, the reaction was performed at different substrate concentrations.
It was found that there was no significant change in conversion when the concentration
of the substrate was either increased or decreased. We settled on a substrate concentration
of 0.12 M, well below the maximum solubility in order to prevent precipitation.
Upon changing the iridium catalyst for 4-CzIPN in our batch flow system, similar conversions
were observed albeit at a somewhat higher catalyst loading (Scheme [4]). This is analogous to Wendlandt’s observations in the epimerization reaction.[15] Although no significant change in conversion was observed when the catalyst loading
was varied between 2.5 mol% and 4 mol%, the latter was chosen for larger scale reactions
in order to have flexibility in the residence time. The apolar nature of 4-CzIPN allows
it to be removed easily from the reaction mixture by chromatography.
Scheme 4 Photoalkylation of methyl 4,6-O-benzylidene-α-d-glucopyranoside in flow
Although we initially started with a 25 mol% loading of tetrabutylammonium dihydrogen
phosphate (TBAP), reducing the loading of the TBAP cocatalyst to 15 mol% did not have
a pronounced effect on the overall conversion and reduced the risk of precipitation.
Increasing the equivalency of the somophile acrylonitrile above 1.5 did not have a
positive result. Instead, unreacted or polymerized somophile was recovered after the
reaction.
Purification of the reaction mixtures is rather straightforward. Dependent on the
somophile used, the excess was evaporated under reduced atmosphere, together with
the solvent. The TBAP and the quinuclidine were removed by aqueous wash, whilst the
remaining 4-CzIPN and its decomposition products were removed by separation by column
chromatography. With commercial equipment, it would be possible to carry out liquid–liquid
extraction directly in-line, further simplifying purification.[31]
When subjected to these optimized conditions, with a residence time of Tr = 20 minutes, 1 could be alkylated on 1-gram scale in 55% isolated yield (3a). Increasing the residence time did not result in a significantly higher conversion.
The impurity profile of the crude suggested that some decomposition of the benzylidene
acetal took place, although a clear side product could not be isolated. It is not
unlikely that the C–H bond of the acetal can also be activated in this reaction. The
use of an alternative protecting group without such a hydrogen, for instance a di-tert-butylsilyl group, may avoid this problem, at the expense of increased cost.
In addition to acrylonitrile, several other somophiles were used in the alkylation
reaction, with mixed results (Figure [3]). Notably, the required residence time for these somophiles was significantly larger
than that of acrylonitrile (0.25 mL/min vs 1 mL/min). Reactions with alkenes such
as phenyl vinyl sulfone and diethyl vinyl phosphonate performed well, providing 3b and 3c, respectively. In an attempt to increase the scope, we also studied styrene derivatives
and vinylpyridine, although these were expected to be very weak somophiles. We were
pleased to see that m-chlorostyrene and in particular 2-vinylpyridine do act as somophiles in this reaction,
giving 3e and 3f, respectively. Although the yields are somewhat unsatisfactory still, in particular
for the preparation of pharmaceutical scaffolds this broadening of the scope is significant.
Bromostyrenes were studied as well, but decomposed under these conditions.
Figure 3 Somophile scope of gram-scale flow reactions
To confirm the value of performing these reactions in flow, the alkylation of 1 with acrylonitrile was carried out on 40-gram scale. As increasing the residence
time with this somophile had only negligible effect on product yield, a residence
time Tr = 20 minutes was chosen. After purification, 45% (21.4 g) of 3a was isolated.
In summary, by performing the established photoalkylation in flow, it has been demonstrated
that this reaction can be carried out on a multigram scale. The replacement of the
iridium photocatalyst with an organo-photocatalyst allows the reaction to be carried
out cost-effectively, offering large-scale production potential. Reactions with several
somophiles were carried out on gram scale and a representative example was scaled
to 21 g of product. The set of applicable somophiles has been expanded with 3-chlorostyrene
and 2-vinylpyridine, which makes the products more interesting from a pharmaceutical
point of view. We are currently focused on the follow-up chemistry of these compounds.
In the future, it would be interesting to investigate the reaction with the recently
reported HAT agents, as well as different carbohydrate substrates.
All reagents were purchased from commercial sources. Somophiles were distilled prior
to use, with exception of those that are heat-sensitive. Reaction mixtures were pumped
through the flow reactor by means of a Syrris Asia syringe pump. NMR spectra were
recorded on a Bruker Avance NEO (400 MHz) spectrometer at room temperature. High-resolution
mass spectrometry (HRMS) was carried out on a Thermo Fisher Scientific LTQ Orbitrap
XL (FTMS) instrument. IR spectra were recorded on a Thermo Fisher Scientific Nicolet
380 FT-IR spectrometer.
Alkylation of Methyl 4,6-O-Benzylidene-α-d-glucopyranoside (1) in Flow; General Procedure
Alkylation of Methyl 4,6-O-Benzylidene-α-d-glucopyranoside (1) in Flow; General Procedure
To a 50-mL round-bottom flask with stir bar was added methyl 4,6-O-benzylidene-α-d-glucopyranoside (1; 1.00 g, 3.54 mmol), 4-CzIPN (112 mg, 142 μmol, 4 mol%), quinuclidine (40 mg, 0.36
mmol, 0.1 equiv) and tetrabutylammonium dihydrogen phosphate (TBAP; 180 mg, 530 μmol,
0.15 equiv). The flask was subsequently capped with a septum and purged with nitrogen.
Freshly degassed acetonitrile (28 mL) was added and the mixture was stirred until
dissolution. Subsequently, the somophile (1.5 equiv) was added by injection via the
septum, after which the solution was stirred for an additional 5 min. The mixture
was subsequently pumped through a 20-mL flow system (Tr = 80 min) under irradiation of blue light by automated syringe and collected in amber
glassware. After solvent evaporation, the remaining material was dissolved in DCM
(25 mL) and washed with water and brine (10 mL each, to remove the TBAP). After solvent
evaporation, the residue was purified by column chromatography, providing the desired
product.
3a by Large-Scale Alkylation of Methyl 4,6-O-Benzylidene-α-d-glucopyranoside (1) with Acrylonitrile in Flow
3a by Large-Scale Alkylation of Methyl 4,6-O-Benzylidene-α-d-glucopyranoside (1) with Acrylonitrile in Flow
To a 2-L round-bottom flask with stir bar was added methyl 4,6-O-benzylidene-α-d-glucopyranoside (1; 40.0 g, 142 mmol), 4-CzIPN (4.45 g, 5.6 mmol, 4 mol%), quinuclidine (1.58 g, 14.2
mmol, 0.1 equiv) and TBAP (7.21 g, 21.3 mmol, 0.15 equiv). The flask was capped with
a septum and purged with nitrogen. Freshly degassed acetonitrile (1.4 L) was added
by injection via the septum, after which the solution was stirred until homogeneous.
Acrylonitrile (14.1 mL, 213 mmol, 1.5 equiv) was then added through the septum, after
which the mixture was pumped through a 20-mL flow system (Tr = 20 min) under irradiation of blue light by automated syringe and collected in amber
glassware. After solvent evaporation, the remaining material was dissolved in EtOAc
and washed with water and brine (200 mL each, TBAP removal). After solvent evaporation,
the residue was purified by column chromatography, affording 3a (21.4 g) in 45% yield.
3a by Reaction of 1 with Acrylonitrile
3a by Reaction of 1 with Acrylonitrile
White solid; yield: 711 mg (55%); Rf
= 0.2 (EtOAc/heptane, 50:50).
IR: 3390, 2998, 2941, 2867, 2248, 1070, 1038, 1014, 997 cm–1.
1H NMR (400 MHz, DMSO-d
6): δ = 7.45 (dq, J = 6.5, 2.3 Hz, 2 H), 7.42–7.35 (m, 3 H), 5.61 (s, 1 H), 4.86 (d, J = 9.4 Hz, 1 H), 4.62 (d, J = 4.0 Hz, 1 H), 4.25 (dd, J = 10.1, 5.2 Hz, 1 H), 4.17 (s, 1 H), 3.94 (td, J = 10.0, 5.1 Hz, 1 H), 3.67 (t, J = 10.3 Hz, 1 H), 3.50 (dt, J = 9.3, 2.1 Hz, 2 H), 3.33 (s, 3 H), 2.71–2.55 (m, 2 H), 1.98 (t, J = 8.0 Hz, 2 H).
13C NMR (101 MHz, DMSO-d
6): δ = 138.26, 129.20, 128.50, 126.63, 121.75, 100.91, 100.60, 80.39, 72.70, 70.26,
68.80, 58.73, 55.86, 31.00, 12.29.
HRMS (ESI): m/z [M + Na]+ calcd for C17H21NO6: 358.126; found: 358.125.
3b by Reaction of 1 with Phenyl Vinyl Sulfone
3b by Reaction of 1 with Phenyl Vinyl Sulfone
Yellow solid; yield: 965 mg (61%); Rf
= 0.35 (EtOAc/heptane, 60:40).
IR: 3478, 2933, 1466, 1286, 1143, 1067, 1044, 998, 745, 688 cm–1.
1H NMR (400 MHz, CDCl3): δ = 7.85–7.79 (m, 2 H), 7.63–7.56 (m, 1 H), 7.50–7.44 (m, 2 H), 7.39 (dd, J = 6.9, 3.0 Hz, 2 H), 7.36–7.30 (m, 3 H), 5.50 (s, 1 H), 4.70 (d, 1 H), 4.32 (dd,
J = 10.3, 5.0 Hz, 1 H), 3.93 (td, J = 9.9, 5.0 Hz, 1 H), 3.70 (t, J = 10.3 Hz, 1 H), 3.55–3.45 (m, 3 H), 3.44 (s, 3 H), 3.38 (d, J = 9.6 Hz, 1 H), 3.00 (s, 1 H), 2.56 (d, J = 11.9 Hz, 1 H), 2.31–2.13 (m, 2 H).
13C NMR (101 MHz, CDCl3): δ = 139.15, 136.84, 133.54, 129.21, 129.17, 128.30, 127.99, 126.07, 101.73, 100.39,
81.65, 72.87, 71.32, 69.02, 58.72, 56.32, 51.84, 28.40.
HRMS (ESI): m/z [M + Na]+ calcd for C22H26O8S: 473.124; found: 473.122.
3c by Reaction of 1 with Diethyl Vinyl Phosphonate
3c by Reaction of 1 with Diethyl Vinyl Phosphonate
Off-white solid; yield: 872 mg (55%); Rf
= 0.15 (EtOAc/heptane, 50:50).
IR: 3395, 2934, 2863, 1984, 1067, 1039, 1022, 996, 958, 697 cm–1.
1H NMR (400 MHz, CDCl3): δ = 7.54–7.40 (m, 2 H), 7.37–7.27 (m, 3 H), 5.49 (s, 1 H), 4.70 (dd, J = 3.8, 1.6 Hz, 1 H), 4.31 (ddd, J = 10.4, 5.1, 1.5 Hz, 1 H), 4.10–3.94 (m, 5 H), 3.70 (td, J = 10.3, 1.5 Hz, 1 H), 3.52 (d, J = 3.9 Hz, 1 H), 3.43 (s, 3 H), 3.36 (dd, J = 9.5, 1.6 Hz, 1 H), 2.16–1.96 (m, 4 H), 1.23 (apparent t, J = 7.0, 1.3 Hz, 6 H).
13C NMR (101 MHz, CDCl3): δ = 137.17, 129.04, 128.18, 126.19, 101.76, 100.66, 80.75, 73.48, 73.34, 70.54,
69.18, 61.81, 61.75, 61.72, 61.66, 58.71, 56.24, 27.35, 27.31, 20.86, 19.46, 16.41,
16.35.
31P NMR (162 MHz, CDCl3): δ = 34.41–32.42 (m).
HRMS (ESI): m/z [M-2H+Na]– calcd for C20H31O9P: 467.144; found: 467.143.
3d by Reaction of 1 with tert-Butyl Acrylate
3d by Reaction of 1 with tert-Butyl Acrylate
White solid; yield: 885 mg (61%); Rf
= 0.4 (EtOAc/heptane, 50:50).
IR: 3512, 3340, 2970, 2932, 2860, 1731, 1367, 1137, 1081, 1001 cm–1.
1H NMR (400 MHz, CDCl3): δ = 7.51–7.42 (m, 2 H), 7.37–7.29 (m, 3 H), 5.52 (s, 1 H), 4.71 (d, J = 4.1 Hz, 1 H), 4.32 (dd, J = 10.3, 5.1 Hz, 1 H), 4.03 (td, J = 9.9, 5.0 Hz, 1 H), 3.71 (t, J = 10.3 Hz, 1 H), 3.50 (dd, J = 11.3, 4.1 Hz, 1 H), 3.45 (s, 3 H), 3.38 (d, J = 9.5 Hz, 1 H), 3.20 (s, 1 H), 2.84 (d, J = 11.3 Hz, 1 H), 2.60–2.40 (m, 2 H), 2.18–1.98 (m, 2 H), 1.38 (s, 9 H).
13C NMR (101 MHz, CDCl3): δ = 173.71, 137.24, 129.05, 128.24, 126.23, 101.72, 100.52, 80.72, 80.53, 73.32,
70.83, 69.19, 58.65, 56.27, 30.49, 29.84, 28.07.
HRMS (ESI): m/z [M + Na]+ calcd for C21H30O8: 433.183; found: 433.182.
3e by Reaction of 1 with 3-Chlorostyrene
3e by Reaction of 1 with 3-Chlorostyrene
Yellow solid; yield: 245 mg (16%); Rf
= 0.4 (EtOAc/heptane, 50:50).
IR: 3478, 2942, 2844, 1071, 1016, 987, 697 cm–1.
1H NMR (400 MHz, CDCl3): δ = 7.55–7.50 (m, 2 H), 7.46–7.38 (m, 3 H), 7.26–7.17 (m, 3 H), 7.12 (dt, J = 7.2, 1.7 Hz, 1 H), 5.53 (s, 1 H), 4.83 (s, 1 H), 4.41 (dd, J = 10.3, 5.1 Hz, 1 H), 4.10 (td, J = 9.9, 5.0 Hz, 1 H), 3.77 (t, J = 10.3 Hz, 1 H), 3.67 (dd, J = 12.0, 4.1 Hz, 1 H), 3.54 (s, 3 H), 3.49 (d, J = 9.6 Hz, 1 H), 2.96 (s, 1 H), 2.86–2.77 (m, 2 H), 2.68 (d, J = 12.1 Hz, 1 H), 2.20–2.08 (m, 2 H).
13C NMR (101 MHz, CDCl3): δ = 144.06, 137.14, 134.15, 129.66, 129.13, 128.52, 128.31, 126.58, 126.16, 126.11,
101.78, 100.75, 79.86, 74.21, 69.68, 69.22, 58.71, 56.38, 35.21, 30.22.
HRMS (ESI): m/z [M + Na]+ calcd for C22H25ClO6: 443.123; found: 443.122.
3f by Reaction of 1 with 2-Vinylpyridine
3f by Reaction of 1 with 2-Vinylpyridine
White solid; yield: 412 mg (30%); Rf
= 0.2 (DCM/MeOH, 98:2).
IR: 2931, 1068, 1044, 995, 748, 698 cm–1.
1H NMR (400 MHz, CDCl3): δ = 8.55–8.24 (m, 1 H), 7.61–7.46 (m, 1 H), 7.42–7.24 (m, 5 H), 7.12–7.01 (m, 2
H), 5.45 (s, 1 H), 4.73 (d, J = 4.1 Hz, 1 H), 4.31 (dd, J = 10.2, 5.1 Hz, 1 H), 4.14 (td, J = 9.9, 5.1 Hz, 1 H), 3.69 (t, J = 10.3 Hz, 1 H), 3.62 (d, J = 4.2 Hz, 1 H), 3.44 (d, J = 1.0 Hz, 3 H), 3.39 (d, J = 9.5 Hz, 1 H), 3.21–2.95 (m, 2 H), 2.38–2.10 (m, 2 H).
13C NMR (101 MHz, CDCl3): δ = 161.87, 148.39, 137.51, 136.87, 128.86, 128.11, 126.23, 123.19, 121.17, 101.57,
100.78, 80.84, 73.50, 70.78, 69.31, 58.74, 56.25, 33.62, 32.30.
HRMS (ESI): m/z [M + Na]+ calcd for C21H25NO6: 410.157; found: 410.157.
