β-Selective C-arylation of silyl protected 1,6-anhydroglucose with arylalanes: The synthesis of SGLT2 inhibitors

Article abstract of DOI:10.1021/jo502839e

The stereoselective arylation of hydroxy protected 1,6-anhydro-β-d-glucose with arylalanes to provide β-C-arylglucosides is reported. Modification of triarylalanes, Ar₃Al, with strong Br?nsted acids (HX) or AlCl₃ produced more reactive arylating agents, Ar₂AlX, while the incorporation of alkyl dummy ligands into the arylating agents was also viable. Me₃Al and i-Bu₂AlH were found useful in the in situ blocking of the C3-hydroxyl group of 2,4-di-O-TBDPS protected 1,6-anhydroglucose. The utility of the method was demonstrated by the synthesis of the SGLT2 inhibitor, canagliflozin.

Full text of DOI:10.1021/jo502839e

Subscriber access provided by SELCUK UNIV
Article
beta-Selective C-Arylation of Silyl Protected 1,6-Anhydroglucose
with Arylalanes: the Synthesis of SGLT2 Inhibitors
Julian Paul Henschke, Ping-Yu Wu, Chen-Wei Lin, Shi-Feng Chen, Pei-Chen Chiang, and Chi-Nung Hsiao
J. Org. Chem., Just Accepted Manuscript • DOI: 10.1021/jo502839e • Publication Date (Web): 28 Jan 2015
Downloaded from http://pubs.acs.org on February 9, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
The Journal of Organic Chemistry is published by the American Chemical Society.
1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
βꢀSelective
CꢀArylation
of
Silyl
Protected
1,6ꢀ
Anhydroglucose with Arylalanes: the Synthesis of SGLT2
Inhibitors
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Julian P. Henschke,* PingꢀYu Wu, ChenꢀWei Lin, ShiꢀFeng Chen, PeiꢀChen Chiang
and ChiꢀNung Hsiao
Department of Exploratory Research, Research and Development, ScinoPharm Taiwan, Tainan,
Taiwan
email address: julian.henschke@scinopharm.com
KEYWORDS: Cꢀglycoside, Cꢀglucoside, SGLT2 inhibitor, diabetes, organoaluminium,
arylalane
TOC:
ABSTRACT: The stereoselective arylation of hydroxy protected 1,6ꢀanhydroꢀβꢀ
Dꢀglucose with arylalanes
to provide βꢀCꢀarylglucosides is reported. Modification of triarylalanes, Ar Al, with strong Brønsted acids
3
(
HX) or AlCl produced more reactive arylating agents, Ar AlX, while the incorporation of alkyl dummy
3
2
ligands into the arylating agents was also viable. Me
3
2
Al and iꢀBu AlH were found useful in the in situ
blocking of the C3ꢀhydroxyl group of 2,4ꢀdiꢀOꢀTBDPS protected 1,6ꢀanhydroglucose. The utility of the
method was demonstrated by the synthesis of the SGLT2 inhibitor, canagliflozin.
1
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 2 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
INTRODUCTION
1
While many approaches exist for the formation of Cꢀglycosidic bonds, the
coupling of electrophilic glycosyl donors with carbon nucleophiles is probably most
2
used in Cꢀglucoside synthesis. Since 2012 four members (Figure 1) of a new class of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
antiꢀdiabetes active pharmaceutical ingredient (API), known as Sodiumꢀcoupled
3
GLucose coꢀTransporter 2 (SGLT2) inhibitors (1), have received marketing approval.
These compounds, canagliflozin (1a), dapagliflozin (1b), ipragliflozin (1c) and
4
empagliflozin (1d), are structurally archetypical of a plethora of synthetic,
biologically active βꢀCꢀarylglucosides that have been reported. Varying by only the
diaryl methylene side chain the close structural similarity of these compounds render
them good targets for a new synthetic platform technology.
Figure 1. βꢀCꢀarylglucoside SGLT2 inhibitors approved for the treatment of
diabetes.
Conventionally, βꢀCꢀarylglucosides are prepared by variants of the Kraus et al. and
Czernecki et al. approach that comprises the low temperature addition of aryllithium
2
ACS Paragon Plus Environment

Page 3 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
or arylmagnesium compounds to perꢀhydroxyl protected gluconolactones followed by
reduction with silane reagents in the presence of BF
3
•OEt
2
, separation of the resulting
5
anomers and deprotection (see the top route in Scheme 5). The only moderate βꢀ
anomeric selectivity (4:1 β:α when using Et SiH/BF •OEt ) witnessed in the silane
3
3
2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
reduction step of the traditionally used perꢀbenzylꢀprotected gluconolactone was later
addressed at BristolꢀMyers Squibb by the use of bulky silane reagents (up to 45:1
6
a
β:α) or by conversion of the lactol 1,2ꢀaddition product to a perꢀacetyl protected
methyl Cꢀarylglucoside derivative followed by reduction with Et SiH/BF •OEt in the
3
3
2
6b
presence of water (19:1 up to >65:1 β:α). This and other improvements to the
gluconolactone approach have allowed for the multikilogram scale synthesis of
SGLT2 inhibitors as exemplified for a dapagliflozin analogue from commercially
7
available gluconolactone. Perhaps the best advance in βꢀCꢀarylglucosides synthesis
since publication of the original gluconolactone method was reported by Lemaire et
8
a
al. In their transitionꢀmetalꢀfree approach a perꢀOꢀpivaloyl protected glucosyl
bromide substrate was rapidly arylated with diaryl zinc reagents that were prepared by
the lithiation of aryl halides followed by transmetalation with a zinc bromideꢀlithium
bromide complex. Good yields and high stereoselectivity was achieved over a range
of aryl nucleophiles, and the usefulness of the method was exemplified by the
8b
synthesis of canagliflozin (1a) and dapagliflozin (1b). Sakamaki et al. demonstrated
the preparation of a selection of βꢀCꢀarylglucosides by the boronation of a triꢀOꢀsilyl
protected Dꢀglucal substrate, palladiumꢀcatalyzed Suzuki crossꢀcoupling of the thus
generated glucal pinacol boronate with aryl bromides followed by a hydroborationꢀ
oxidation sequence with boraneꢀTHF complex and basic hydrogen peroxide, and
finally TBAF promoted desilylation. Cossy et al. produced αꢀCꢀarylglucosides from
perꢀOꢀacetyl protected glucosyl bromide with moderate selectivity using Grignard
3
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 4 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
reagents in the presence of a Co(III) catalyst, however, these possessed the opposite
8c
C1ꢀconfiguration of SGLT2 inhibitors.
Inspired by experience in oligosaccharide synthesis, we were intrigued by the
possibility of utilising a protected derivative 2 of 1,6ꢀanhydroꢀβꢀ ꢀglucose (3) as the
D
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
electrophile in a nucleophilic coupling with aryl nucleophiles (Scheme 1). With 1,6ꢀ
anhydroglucose possessing the same oxidation state as glucose, no reduction step
would be required, in contrast to the gluconolactone approach. While the relative
inertness of the 1,3ꢀdioxolane ring of 2 rendered nonꢀLewis acid assisted nucleophilic
9
attack unviable, literature precedent for the allylation of 2,3,4ꢀtriꢀOꢀbenzylꢀ1,6ꢀ
3 2
anhydroglucose (2: PG = Bn) in the presence of BF •OEt or TMSOTf suggested that
Lewis acidꢀassisted nucleophilic attack would be possible but would favour the
undesired αꢀanomer. To mitigate this the requisite aryl nucleophile and a Lewis acid
could be combined into a single agent able to i) coordinate to and open the 1,3ꢀ
dioxolane ring of compound 2 to furnish an oxocarbenium ion, and ii) whilst
covalently bound to the C6 oxygen as an ate complex of the opened 1,3ꢀdioxolane
ring, deliver the aryl anion directly to the requisite βꢀface of the oxocarbenium ion,
thereby furnishing the desired βꢀCꢀarylglucosides 4. This approach would be
synthetically succinct and potentially highly stereoselective owing to substrate control
and finds precedent in the alkynylation and arylation of anhydrofuranose and
anhydropyranose derivatives (see below).
Having investigated this, we report herein a method that arylates with high
stereoselectivity the protected 1,6ꢀanhydroglucose derivative 2a with arylalanes,
delivering hydroxyꢀprotected βꢀCꢀarylglucosides 4 in one step; transition metals and
10
cryogenic temperatures are not required.
4
ACS Paragon Plus Environment

Page 5 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 1. Convention (top) and conceptual Cꢀarylation strategy (bottom) for the
synthesis of βꢀCꢀarylglucosides 4 and 1
RESULTS AND DISCUSSION
Preliminary results. Whilst tertꢀbutyldimethylsilylation of commercially available
11
1
,6ꢀanhydroglucose (3) furnished triꢀOꢀtertꢀbutyldimethylsilyl (TBS) protected
sugar 2b, exhaustive silylation with tertꢀbutyldiphenylsilyl (TBDPS) chloride gave
the 2,4ꢀdiꢀOꢀTBDPS protected derivative 2a (Scheme 2). In an initial screening study,
2
a was treated with mixtures of PhLi or PhMgBr in the presence of B, Al, Ga, Ti, Zn,
In, La or Hfꢀbased Lewis acids. Although various Ti(IV) salts or GaCl provided
3
3
promise, the combination of PhLi or PhMgBr with AlCl , that was presumed to
generate arylaluminium compounds and metal halide salts, led to efficient, albeit slow
12
conversion of 2a to the desired protected product βꢀCꢀphenylglucoside 4a. By
contrast, when triꢀOꢀTBS protected derivative 2b was subjected to the same
5
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 6 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
conditions, considerable decomposition occurred and subsequent studies were
therefore focused on the use of 2a.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 2. βꢀCꢀArylation of protected 1,6ꢀanhydroglucose
Organoaluminiums can donate alkyl or aryl carbanions to other metals including
Rh(I), Ti(IV) or Pd(II) making them useful in asymmetric addition reactions and
13
crossꢀcouplings. The relatively high Lewis acidity of the aluminium centres
combined with nucleophilic organic ligands renders organoaluminium reagents also
14
useful as synthetic reagents on their own. For example, in related work Yamamoto et
15
al. demonstrated the stereoretentive cleavage and alkylation of 1,3ꢀdioxolane and
16
1
,3ꢀdioxane rings using alkylaluminium derivatives and Vasella et al. used silylꢀ
protected propargylaluminium dichloride in the βꢀselective alkynylation of protected
17
1
,6ꢀanhydrohexoses. Rainier et al. have shown that Ph
3
Al and (2ꢀfuryl)
3
Al are able
to arylate an oxirane derivative of triꢀOꢀbenzylꢀDꢀglucal to provide αꢀCꢀ
arylglucosides, whilst βꢀCꢀarylarabinofuranoses were prepared by Sietz et al. using
an excess of triarylaluminium reagents and 3,5ꢀdiꢀOꢀTBSꢀ or TBDPSꢀprotected 1,2ꢀ
1
8
anhydroarabinose. Both of these arylation reactions, however, utilise cryogenic
6
ACS Paragon Plus Environment

Page 7 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
temperatures (≤−65 °C) due to the high reactivity of both the oxiranes and
arylaluminium compounds. The relatively low toxicity and affordability of aluminium
derivatives was attractive from a scaleꢀup perspective and detailed investigations
ensued in our laboratories.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Examination of arylalanes as arylating agents of 1,6ꢀanhydroglucose 2a. The
conversion of 2a to 4a was investigated in PhMe using commercially available Ph
3
Al
1
9
(1
equiv of Ph
M solution in nꢀBu O) (see Figure S1 in the supplementary section). Whereas 2
2
3
Al provided a 77% HPLC yield in 118 h, yields of <50% were observed
when using 0.25 up to 1.5 equiv, despite only 0.33 equiv theoretically being required
to effect full conversion.
Next, as per the preliminary screening study, phenylalane reagents were prepared
20
in situ from PhMgBr in Et
2 3
O and AlCl in THF. Varying the relative molar ratio of
these reagents allowed precise control over the composition of the arylating agent and
2
n
1
provided access to the more reactive aryl(halo)alanes, Ph AlX (where X is Cl and
m
2
Br). Arylation rates in PhMe increased when i) THF and Et O were removed by
evaporation before arylation, and ii) as the ratio of the aryl anion (Ar) to halide (X)
decreased, consistent with the increasing Lewis acidity of the aluminium centre.
m n
Arylating reagents Ph AlX (2 equiv) having m/n ratios 2.5:0.5 and 3.0:0 gave HPLC
yields of 4a of about 65–70% in about 25–55 h, whilst a 2:1 ratio led to around a 50%
HPLC yield (see Figure S2 in the supplementary section). Predictably, the nonꢀLewis
4
acidic tetraarylalane ate complex, [Ph Al]MgX, provided no significant reaction
22
within 24 h. Paralleling the 2 equiv threshold seen when using commercially
3
available Ph Al, while 1 equiv Ph2.5AlX0.5 in PhMe only gave a 40% HPLC yield, 2,
3, and 4 equiv all provided about a 70% yield of 4a (see Figure S3 in the
supplementary section). The necessity for at least 2 equiv of arylating reagent to effect
7
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 8 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
good conversion of 2a to 4a was explained by the presence of the unprotected C3
hydroxyl of 2a and is addressed below.
Solvent screening (see Figure S4 in the supplementary section) showed that while
23
arylation was ineffective in THF using 2 equiv of Ph2.5AlX0.5, yields of around 70–
0% were witnessed in PhMe, PhCl, PhOMe or nꢀBu O as solvent. Moderate yields
were seen in Ph O (65%), dioxane (51%) and benzonitrile (30%) whilst arylation in
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
2
2
the highly polar solvents diglyme, pyridine and NMP failed to provide good yields of
glucoside 4a. Although PhCl proved a particularly good solvent in terms of reaction
24
rate and yield (80%), it was undesirable from a toxicity perspective. Instead
2
5
PhOMe, that provided superior solubilisation, reaction rate or stirrability as
compared to PhMe, PhCl or nꢀBu O, was selected for the subsequent studies. Reꢀ
2
examination (see Figure S5 in the supplementary section) of the aryl anion/halide
ratios m/n using PhOMe as solvent revealed a similar trend to that in PhMe, however,
the best yields were witnessed using the more Lewis acidic m/n ratios of 2.5:0.5
(
83%; 28 h) and 2:1 (83%; 21 h) while severe decomposition occurred when m/n was
1:2.
In addition to the desired product 4a (t
13.7 min), benzene (t 10.1 min) and phenol (t
R
24.0 min) and small amounts of biphenyl
6.6 min), HPLC analysis of the
(t
R
R
R
reaction mixture revealed the formation of a closely eluting coꢀproduct (t 23.5 min).
R
Desilylation of the unpurified, crude product mixture using TBAF in THF followed
26
by HPLC comparison to a mixture of authentic αꢀ1e and βꢀ1e showed, to our
surprise, no detectable amounts of αꢀanomer αꢀ1e revealing that the arylation reaction
was highly βꢀselective. LCMS analysis of the coꢀproduct indicated that it was instead
a dehydrated derivative of TBDPSꢀprotected 1,6ꢀanhydroglucose 2a. When reaction
aliquots were treated with 5% TFA in MeCN before HPLC analysis, however, the
8
ACS Paragon Plus Environment

Page 9 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
major coꢀproduct (t
R
23.5 min) was no longer detectable and instead significant
amounts of tertꢀbutyldiphenylsilanol (TBDPSOH; t
R
14.2 min) were produced along
3.9 min). Under the
with a new, relatively weakly UVꢀactive, fast eluting impurity (t
R
same HPLC sample preparation conditions, however, neither TBDPSꢀprotected 2a nor
the βꢀCꢀarylglucoside product 4a produced TBDPSOH indicating that it was
produced by acidic catalysed/promoted decomposition of the coꢀproduct. Following
isolation by column chromatography, 1D and 2D NMR spectroscopy revealed that the
coꢀproduct was 1,6ꢀanhydrosilylenol ether 5. Consistent with this, treatment of 5 with
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5
% TFA in MeCN produced (−)ꢀlevoglucosenone (6), the fast eluting impurity, as
confirmed by reference to a commercial sample (Figure 2). The level of TBDPSOH,
produced when reaction aliquots were treated with 5% TFA in MeCN before HPLC
27
analysis, was then used as a proxy for 1,6ꢀanhydrosilylenol ether 5 formed in the
arylations and in fact its level was a good measure of reaction performance.
O
O
O
O
TBDPSO
OTBDPS
O
5
6
Figure 2. Silylenol ether 5 and (−)ꢀlevoglucosenone (6)
Scope of the arylation reaction and stereoselectivity. Having determined
stoichiometry, solvent and Ph AlX m/n ratio (Table 1, entry 1), the scope was
m
n
extended to other Cꢀarylglucosides by varying the aryl Grignard or lithium reagent.
Simple aromatics (entries 2 and 3) provided good HPLC yields, whilst the two 4ꢀhalo
substituted aromatics tested (entries 4 and 5) provided lower yields and required a
higher reaction temperature. While the heteroaromatic furyl group (entry 6) was
readily transferred to the substrate, thienyl (entry 7) and 4ꢀmethoxyphenyl (entry 8)
9
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 10 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
groups were more challenging due to competitive, albeit manageable, degradation of
the Cꢀarylglucoside products. Satisfyingly, this methodology was also applicable to
the synthesis of the 2,4ꢀdiꢀOꢀTBDPS protected derivatives 4i and 4j of canagliflozin
(1a; entry 9) and dapagliflozin (1b; entry 10), respectively. These were subsequently
desilylated with TBAF in THF to provide the respective SGLT2 inhibitors 1a and 1b
for proof of concept.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In the absence of reference samples of the αꢀanomers of the Cꢀarylglucosides, high
resolution LCMS analyses of the crude product mixtures were conducted for 4c, 4d,
4
f, and 4h and extractedꢀion chromatograms (XICs) using the mass to charge ratios of
the corresponding ammonium adducts were created from the resultant datasets. No
isomer peaks were detectable in the XICs indicating that the arylation reactions
proceeded with high stereoselectivity, consistent with that already shown for the
phenyl analogue 4a. This was further confirmed for canagliflozin (1a) following
deprotection of the crude mixture of 4i with TBAF in THF and reꢀanalysis without
1
purification; again, no isomer peak was detected. In addition to H NMR
spectroscopic identification of the desilylated products of 4a, 4i and 4j by comparison
26
to spectra of authentic samples of βꢀ1e, canagliflozin (1a) and dapagliflozin (1b),
respectively, the coupling constants J1,2 of 4aꢀ4j (9.2ꢀ10.0 Hz) were consistent with
8b
C1ꢀβꢀconfigured Cꢀarylglucosides.
1
0
ACS Paragon Plus Environment

Page 11 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
Table 1. Arylation of 2a to produce βꢀCꢀarylglucosides 4
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
β/α
Selectivity
a
Temp
(°C)
Yield 4
(%)
d
Entry
Ar
b
Time (h) Product
c
1
2
Ph
120
120
140
140
140
120
120
120
140
120
28
26
16
22
4
4a
4b
4c
4d
4e
4f
83
59
67
47
56
78
>99:1
ND
4ꢀTolyl
Mesityl
4ꢀClꢀPh
4ꢀFꢀPh
3
>99:1
>99:1
ND
4
5
e
6
2ꢀFuryl
16
4
>99:1
ND
f
7
2ꢀThienyl
4ꢀMeOꢀPh
4g
4h
4i
60
f,g,h
8
8
54
>99:1
>99:1
ND
1
i
9
Ar
Ar
5
68
51
2
j
10
8
4j
a
The arylating agent Ar AlX0.5 (2 equiv) was prepared in PhOMe using AlCl in
2.5
3
THF and the corresponding Grignard reagent at rt. 2a in PhOMe was deprotonated
with PhMgBr in Et O. The mixture was concentrated under reduced pressure to
remove THF and Et
2
b
c
2
O. Internal solution temperature. HPLC yields with
d
reference to 1,2,4,5ꢀtetramethylbenzene (internal standard). See explanation in
e
f
the main text. 2ꢀFuryllithium was used as the aryl source. Competing
g
decomposition of the arylation product was observed. Some demethylated
h
product formed from 4h as the reaction proceeded. The arylating reagent was (4ꢀ
i
1
MeOC
6
H
4
)
2
AlCl.
Ar
is
3ꢀ[[5ꢀ(4ꢀfluorophenyl)ꢀ2ꢀthienyl]methyl]ꢀ4ꢀ
j
2
methylphenyl. 2a was deprotonated with nꢀBuLi. Ar is 4ꢀchloroꢀ3ꢀ[(4ꢀ
ethoxyphenyl)methyl]phenyl.
1
1
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 12 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Examining the effect of metal halides. Although sometimes having a positive
2
8
influence, magnesium salts have also been reported to interfere in some reactions
13b,29
involving organoaluminium reagents.
Given this, a better understanding of the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
influence of the magnesium and lithium halide byꢀproducts, formed during arylalane
synthesis, was sought. Two equiv of the arylating agent were used to account for the
unprotected C3ꢀhydroxyl of 2a that results in the formation of alkoxy(diphenyl)alane
7
a (see below). Whereas 2a converted to 4a in a 81% HPLC yield using
commercially available Ph Al (Table 2, entry 1), the addition of 1 equiv of MgCl
entry 2) resulted in a decreased 61% HPLC yield, 17% unreacted 2a and a relatively
high level of TBDPSOH (51%), indicating increased elimination to produce 5.
Although the reaction rate increased when spiked with 1 equiv MgBr (entry 3), an
3
2
(
2
even higher level of TBDPSOH (66%) was observed. Notably, a low 26% yield of 4a
was witnessed along with 30% of unreacted 2a and 51% of TBDPSOH when the
reaction was conducted in the presence of 6 equiv of LiCl (entry 4). By contrast,
addition of 1 equiv or 3 equiv of Mg(OTf) (entries 5 and 6) to the reaction resulted in
2
about a 70% HPLC yield of 4a with similar conversion rates to that of the
unadulterated reaction, indicating that the halide ion(s) rather than the metal itself
were predominantly responsible for the observed deleterious effect. To confirm this,
reactions (entries 7–13) were spiked with [Ph
from metal ions (see Figure S6 in the supplementary section). Whereas 1 equiv of
Ph P]Cl (equivalent to 0.5 equiv of chloride per aluminium; entry 7) reduced the
4 4
P]Cl or [Ph P]Br to avoid interference
[
4
yield of 4a to 35% (19 h) leaving 34% unreacted 2a and 37% TBDPSOH, 2 and 4
equiv (entries 8 and 9) of the chloride salt inhibited arylation altogether. Although a
4
similar phenomenon occurred when reactions were spiked with [Ph P]Br (entries 10–
1
2
ACS Paragon Plus Environment

Page 13 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
1
3), even an excess failed to completely inhibit the reaction. It was therefore evident
that free halides were detrimental to this reaction whilst the presence of metal counter
ions (Li and Mg) appeared to moderate the phenomenon.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 2. Influence of additives on the arylation of 2a with Ph Al
3
Yield 4a
b
Entry
Additive
Time (h)
a
Unreacted 2a
TBDPSOH
(
%)
81
61
64
26
67
70
35
0
1
2
3
4
5
6
7
8
9
None
31
23
6
9
33
51
66
51
37
31
37
28
18
33
32
53
57
1 equiv MgCl
2
17
4
1 equiv MgBr
6 equiv LiCl
2
12
33
23
19
48
48
8
30
13
16
34
82
86
28
66
45
40
1 equiv Mg(OTf)
3 equiv Mg(OTf)
2
2
1 equiv Ph
2 equiv Ph
4 equiv Ph
1 equiv Ph
2 equiv Ph
3 equiv Ph
4
PCl
4
PCl
4
PCl
4
PBr
4
PBr
4
PBr
0
1
0
1
39
7
1
12
12
12
12
9
1
3
4 equiv Ph PBr
8
4
a
HPLC yields determined by reference to 1,2,4,5ꢀtetramethylbenzene (internal
b
standard). Reaction aliquots were treated with 5% TFA in MeCN before HPLC
analysis resulting in the conversion of 5 into TBDPSOH (the values presented may be
approximate).
Metal halideꢀfree, Brønstedꢀacid activated arylating agents. Attention was
turned to arylating agents free of magnesium or lithium halide salts (Table 3).
Additionally, the relationship between Lewis acidity of the alane and reactivity was
1
3
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 14 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
further explored. To this end, commercial Ph
3
Al was modified by premixing with
TfOH, C
Ph
with decreasing pKa of the Brønsted acids used (Figure 3). For the OTf (entry 1), Cl
6
F
5
OH, PhOH or iꢀPrOH, while saltꢀfree Ph
2
AlCl was prepared by mixing
30
3
Al and AlCl
3
.
As expected, the rate of arylation using these reagents increased
3
1
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
5
(
entry 2), or OC
yields of ≥80% much more rapidly than Ph
, entry 4) and OiꢀPr (entry 5) ligands, on the other hand, deactivated the arylating
reagent and reduced yields significantly, with respect to Ph Al, and more TBDPSOH
6
F
5
(entry 3) modified alanes, the arylation reactions reached HPLC
3
Al (entry 1, Table 2). Both the OPh (Table
3
3
was produced when HPLC samples were prepared under acidic conditions. That the
alkoxy(diphenyl)alane was not a good arylating agent is consistent with the low yields
3
generated when arylating 2a with 1 equiv of Ph Al or Ph2.5AlX0.5 that form 7a. That
is, the alkoxy(diphenyl)alane moiety of 7a is not thought to be a good intraꢀ or
intermolecular arylating agent. From a scaleꢀup perspective, the high reaction rate and
yield, the ease of preparation (i.e., mixing of Ar Al and AlCl ) and with cost
3
3
considerations in mind, aryl(chloro)alanes were favoured as arylating reagents and
were used in subsequent studies.
1
4
ACS Paragon Plus Environment

Page 15 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
Table 3. Arylation of 2a using activated phenylalanes
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a
Yield 4a Unreacted
c
Entry
Base
Ph
2
AlX
Time (h)
b
TBDPSOH
(
%)
81
85
82
69
44
67
90
94
2a
1
2
3
4
5
6
7
8
Ph
Ph
Ph
Ph
Ph
Ph
Ph
Ph
3
3
3
3
3
3
3
3
Al
Al
Al
Al
Al
Al
Al
Al
1 equiv Ph
1 equiv Ph
1 equiv Ph AlOC
1 equiv Ph AlOPh
1 equiv Ph AlOiꢀPr
0.5 equiv Ph AlCl
2
AlOTf
2.5
8
3
31
36
24
45
53
62
21
15
2
AlCl
1
2
6
F
5
12
24
21
5
3
2
10
14
8
2
2
2 equiv Ph
3 equiv Ph
2
2
AlCl
AlCl
3
0
4
0
9
nꢀBuLi
nꢀBuLi
1 equiv Ph
2 equiv Ph
3
3
Al
Al
31
30
3
0
0
68
89
2
9
1
0
1
24
35
54
1
nꢀBuLi
2 equiv Ph AlCl
86
72
2
d
1
a
2
Ph
3
Al 1 equiv Ph
2
AlCl–LiCl
2
0.8
b
1
equiv of base use. HPLC yields as determined by reference to 1,2,4,5ꢀ
tetramethylbenzene (internal standard). Reaction aliquots were treated with 5% TFA
in MeCN before HPLC analysis (the TBDPSOH values presented may be
approximate). 2a was reacted with 2 equiv of Ph2.5AlCl0.5 which comprises a
3 2
mixture of Ph Al and Ph AlCl. 1 equiv of LiCl was added.
c
d
1
5
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 16 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Arylation of 2a using activated arylalanes
3
Arylations of alkoxy(diphenyl)alane 7a, formed from 2a in the presence of Ph Al
(see later), using 0.5 equiv through to 3 equiv of Ph AlCl (Table 3, entries 6–8) were
2
typically complete within several hours, however, at least 1 equiv of Ph
2
AlCl (entry
2
) was required to achieve HPLC yields of over 80%. Yields increased as the amounts
of Ph
2
AlCl were increased. Notably, arylation of lithium alkoxide 7d (2a was
Al (entries 9
deprotonated using nꢀBuLi) instead of 7a with 1 equiv or 2 equiv of Ph
3
and 10) failed to produce 4a. This was explained by the formation of the nonꢀLewis
acidic, and therefore unreactive, arylaluminate complexes [Ph(7a)]Li or [Ph Al]Li
4
2
2
(
along with 7a), respectively. Under the same conditions, however, arylation of
lithium alkoxide 7d with 2 equiv of Ph AlCl provided efficient (3 h; entry 11)
conversion to 4a (86% HPLC yield), despite halide salts (LiCl: entry 4 and
Ph P]Cl/Br: entries 7–13, Table 2) retarding arylations using Ph Al. This contrast
2
[
4
3
indicated that triarylalanes and aryl(halo)alanes possessed different tolerances to
halides. Additionally, competitive scavenging of the halide by the halide metal
1
6
ACS Paragon Plus Environment

Page 17 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
counter ion and the aluminium arylating agent might explain why the arylation
reactions using aryl(halo)alanes prepared in situ from Grignard or organolithium
reagents described earlier were able to proceed effectively in the presence of these
salts. To test this, 2a was converted to 7a and was reacted with 1 equiv Ph AlCl in the
2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
presence of 1 equiv LiCl (entry 12). The arylation was very rapid and produced a
superior yield (72%) of 4a as compared to all halide spiking experiments in Table 2
3
that used the triarylalane, Ph Al, as arylating agent.
32
With the intent of increasing aryl anion efficiency, methyl, ethyl or isobutyl
18,33
dummy ligands
were incorporated into the arylalanes by preꢀmixing the
34
appropriate alkylalanes, Ph
3
Al and AlCl
3
.
Reasonably rapid conversion of lithium
alkoxide 7d and good yields of arylated product 4a (Table 4, entries 1–3) were
witnessed, but no advantage was seen.
Aluminiumꢀbased hydroxy blocking groups. In view of the unproductive
consumption of the first molar equiv of the arylalane upon reaction with 2a to form
alkoxy(diphenyl)alane 7a, blocking of the C3 hydroxyl group of 2a with alkyl
aluminium groups prior to the arylation was examined next. When sugar 2a was
treated with Me
gas was observed confirming that deprotonation had occurred. The corresponding
phenyl analogue 7a, formed during the arylation of 2a with Ph Al, was prepared for
3 2
Al in PhMe or iꢀBu AlH (DIBAL) in PhMe at rt rapid evolution of
3
reference by reaction of 2a with Ph
3
Al in nꢀBu
2
O. All three sugar aluminium adducts
for at least 4 days. In the case
7
a, 7b and 7c (Figure 4) were stable at rt in benzeneꢀd
6
of 7b, along with the disappearance of the C3ꢀOH signal from 2a, a sixꢀproton singlet
1
corresponding to a Al,Alꢀdimethyl group was seen at −0.92 ppm in the H NMR
35,36
spectrum, consistent with that reported for alkoxy(dimethyl)alanes.
While
successive treatment of 7b with another 2 equiv of Me Al produced significant
3
1
7
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 18 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
changes in the chemical shift of the ring protons, a fourth equivalent resulted in
essentially no change suggesting that 7b possesses two Lewis basic sites able to
coordinate organoalanes (see spectra in the supplementary section). A 1Dꢀselective
NOESY (DPFGSE) experiment indicated strong throughꢀspace interactions between
the six proton singlet of the dimethylaluminium moiety and H2, H3, H4 and H6a of
the sugar ring and a weaker interaction with the aryl groups of the TBDPS protecting
groups. Subsequent treatment of the alkoxyalanes 7a, 7b or 7c with aq. NaOH
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
returned sugar 2a unaltered, as confirmed by H NMR spectroscopy and LCMS
analysis, showing that the alumination was reversible.
1
8
ACS Paragon Plus Environment

Page 19 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
6
O
1
1
O
7a: R = Ph
1
(R )3Al or i-Bu2AlH
5
1
4
2
7b: R = Me
2a
3
TBDPSO
OTBDPS
1
Aq. NaOH
7c: R = i-Bu
OAl(R1)
2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
1
a) H NMR spectrum of 2a in benzeneꢀd
6
. b) Partial H NMR spectrum of 2a in
1
benzeneꢀd after treatment with 1 equiv of Me Al (in PhMe). c) Partial H NMR
6
3
spectrum of 2a in benzeneꢀd
6
after treatment with 1 equiv of iꢀBu
2
AlH (in PhMe).
Al (in nꢀBu O;
1
d) Partial H NMR spectrum of 2a after treatment with 1 equiv of Ph
3
2
6
seen at 3.1–3.5 ppm) in benzeneꢀd .
Figure 4. Disubstituted(alkoxy)alanes 7a, 7b and 7c
Arylation of dialkyl(alkoxy)alanes. The preformed dimethyl(alkoxy)alane 7b
described above was subjected to arylation in PhOMe with 1 equiv of Ph Al under the
3
standard conditions. Pleasingly, an HPLC yield of 78% (entry 4, Table 4) was
obtained along with only 3% of Cꢀmethylglucoside 8a while a slightly lower yield
(
72%) was observed when the preformed diisobutyl(alkoxy)alane 7c was arylated
under the same conditions (entry 5). This confirmed that dialkylalanyl groups could
1
9
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 20 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
serve as hydroxyl protecting groups. By contrast, and with reference to the high
yielding arylation of diphenyl(alkoxy)alane 7a (Table 3, entry 2), however, it was
2
surprising that arylation of dimethylaluminium derivative 7b with 1 equiv of Ph AlCl,
chosen for its greater reactivity and superior aryl anion economy to Ph Al, only gave
3
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a 33% yield (Table 4, entry 6) accompanied by relatively high amounts of TBDPSOH
(
68%) and Cꢀmethylglucoside 8a (10%). Increasing the amount of Ph
entry 7) or using 1 equiv of Ph AlCl with diisobutyl(alkoxy)alane 7c (entry 8) proved
Al
preꢀtreated with small amounts of AlCl was used instead (entries 9 and 10), arylation
2
AlCl to 2 equiv
(
2
better, but still fell short of expectation (>70%). Pleasingly, when 1 equiv of Ph
3
3
3
of 7b gave good yields (77–78%) and more rapid reaction than when Ph Al was used.
2
0
ACS Paragon Plus Environment

Page 21 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
Table 4. Arylation of 7 using arylalanes and modified arylalanes
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a
b
Time
(h)
Yield 4
(%)
TBDPSOH
(%)
Entry
Base
7
Arylalane
c
8 (%)
1
2
nꢀBuLi
nꢀBuLi
7d
7d
3 equiv Ph(Me)AlCl
3 equiv Ph(Et)AlCl
6
6
75
72
10 (8a)
4 (8b)
24
32
0.5 (8c)
3
4
5
nꢀBuLi
7d
3 equiv Ph(iꢀBu)AlCl
6
70
34
33
29
7 (8d)
3 (8a)
d
Me
3
Al
7b
1 equiv Ph
1 equiv Ph
3
3
Al
Al
17
19
78 (1)
72 (0.7)
1
2
(8c)
(8d)
e
iꢀBu
2
AlH 7c
d
f
6
7
Me
Me
3
3
Al
Al
7b
1 equiv Ph
2 equiv Ph
2
2
AlCl
AlCl
1
1
33 (14) 10 (8a)
68
d
7b
66 (2)
2 (8a)
50
2
8
(8c)
(8d)
e
8
9
iꢀBu
2
AlH 7c
1 equiv Ph
2
AlCl
9
53 (3)
52
d
Me
Me
Me
Me
Me
Me
Me
Me
3
3
3
3
3
3
3
3
Al
Al
Al
Al
Al
Al
Al
Al
7b
1 equiv Ph
1 equiv Ph
3
3
Al/10 mol% AlCl
3
8
3.5
19
2
78
77
7 (8a)
8 (8a)
7 (8a)
7 (8a)
2 (8a)
1 (8a)
23
21
52
30
29
26
6.9
34
29
53
d
d
d
d
d
d
d
d
1
0
1
7b
7b
7b
7b
7b
7b
7b
7b
Al/20 mol% AlCl
3
g
1
1 equiv Ph1.5AlMe1.5
2 equiv Ph1.5AlMe1.5
53 (8)
67 (3)
33 (40)
g
12
h
13
1 equiv Ph
3
Al–(LiCl)
3
18
61
26
13
5
i
j
14
1 equiv Ph
3
Al
83 (1)
1
k
l
15
1.5 equiv Ar
3
Al
16 (80)
62 (8)
ND
1
k
l
16
17
18
2 equiv Ar 2.5AlCl0.5
ND
1
k
m
l
Me Al
2 equiv Ar AlCl
62 (2)
ND
3
2
1
2
k
l
nꢀBuLi
7d
2 equiv Ar
AlCl
4
55 (5)
ND
a
b
c
1 Equiv of base was used. The internal reaction temperature was 140 °C. HPLC yields as
determined by reference to 1,2,4,5ꢀtetramethylbenzene (internal standard); amounts in the
parentheses are unreacted 2a. Reaction aliquots were treated with 5% TFA in MeCN prior to
HPLC analysis producing TBDPSOH (the values presented may be approximate). 7b was
Al at rt for 5 min. 7c was prepared by mixing 2a with iꢀBu
at rt for 5 min. Increased to 85% within another 1 h. Equimolar amounts of Ph Al and Me Al
were mixed. Prepared by mixing AlCl and PhLi in 1:3 ratio at 140 °C for 2 h. As per footnote h,
d
e
prepared by mixing 2a with Me
3
2
AlH
f
g
3
3
h
i
3
j
then filtered to remove solids. For comparison to entry 4, at 34 h 78% HPLC yield of 4 and 8%
k
1
unreacted 2a, 1.3% 8a and 20% TBDPSOH were detected. Ar is 3ꢀ[[5ꢀ(4ꢀfluorophenyl)ꢀ2ꢀ
2
1
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 22 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
l
thienyl]methyl]ꢀ4ꢀmethylphenyl. Due to impurity overlap, the level of 8a was not determined. In
1
the preparative scale run, H NMR spectroscopy indicated a >10:1 ratio of 4i/8a before
purification. A preparative scale furnished 75% isolated yield of protected canagliflozin 4i after
m
column chromatography.
Taking all results together, our original conclusion that effective arylation was
dependent upon there being at least 2 equiv of arylating reagent with respect to the
sugar substrate was revised; instead, only 1 equiv was needed, however, at least 3
equiv of aryl anion with respect to the sugar substrate were required for good
conversion (i.e., for >70%). Significant ligand mobility is believed to occur between
the sugar C3ꢀalkoxyalane and arylating agent, consistent with that reported for
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
3h,i
aryl(alkyl)alanes
Aryl(alkyl)alanes prepared by premixing of Ph
arylate 7b (entries 11 and 12) but did not prove better than Ph
and explaining the formation of side product 8.
18
13h
3
Al and Me
3
Al were able to
Al or the
3
phenyl(chloro)alane systems. As indicated during the preliminary screening studies,
arylation of triꢀOꢀTBS protected substrate 2b did not prove effective. This was reꢀ
examined using 1 or 2 equiv of Ph
Consistent with the original result, however, no better than a 12% yield of 2,3,4ꢀtriꢀOꢀ
tertꢀbutyldimethylsilylꢀ1ꢀCꢀphenylꢀβꢀ ꢀglucopyranoside could be isolated following
3
Al in PhOMe under the optimised conditions.
D
column chromatography. By contrast, the dimethylaluminium moiety serves as an
effective in situꢀgenerated C3ꢀhydroxyl blocking group.
Realꢀcase arylation: the synthesis of canagliflozin. Although some
1
3,20
triarylaluminium compounds are crystalline solids,
sensitive and require handling under strictly anhydrous conditions. With attempts to
synthesise trisꢀ(3ꢀ[[5ꢀ(4ꢀfluorophenyl)ꢀ2ꢀthienyl]methyl]ꢀ4ꢀmethylphenyl)alane,
required for canagliflozin synthesis, from the Grignard reagent or aryl lithium and
AlCl providing gumꢀlike, waxy semiꢀsolids, it was reasoned that it would be better to
they are nevertheless moisture
3
2
2
ACS Paragon Plus Environment

Page 23 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
generate the triarylalane in solution, filter off the metal halide salts and use the filtrate
directly in the arylation reaction. This was modelled by arylating
dimethyl(alkoxy)alane 7b in PhOMe with 1 equiv of a 2
2 3
M nꢀBu O solution of Ph Al
37
that was prepared inꢀhouse from PhLi and AlCl . When filtration to remove the LiCl
3
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
was omitted the arylation failed to reach completion; a 33% HPLC yield was
observed after 18 h (Table 4, entry 13) mirroring the arylation of 2a with commercial
3 3
Ph Al spiked with LiCl (Table 2, entry 4). By contrast, when the Ph Al solution was
filtered (Table 4, entry 14) to remove the precipitated LiCl the subsequent arylation of
dimethyl(alkoxy)alane 7b provided a comparable yield to the commercial reagent
(entry 4) reconfirming earlier observations that LiCl poisoned triarylalanes.
To further demonstrate the applicability of our methodology to the synthesis of
SGLT2 inhibitors, aryl bromide 9a in PhMe–iꢀPr
hexane at 0 °C, transmetalated with AlCl in nꢀBu
The iꢀPr
2
O was lithiated with nꢀBuLi in nꢀ
3
2
O at 90 °C and diluted with PhMe.
2
O was removed by evaporation, the solution was filtered to remove the LiCl
and was concentrated further. Disappointingly though, arylation of
dimethyl(alkoxy)alane 7b in PhOMe using 1.5 equiv of the PhMe solution of trisꢀ(3ꢀ
1
[
[5ꢀ(4ꢀfluorophenyl)ꢀ2ꢀthienyl]methyl]ꢀ4ꢀmethylphenyl)alane (10; Ar
3
Al) provided
only a 16% HPLC yield of 2,4ꢀdiꢀOꢀprotected canagliflozin 4i (entry 15) along with
0% unreacted starting material. Evidently, as compared to the corresponding Ph Al
8
3
model system (e.g., see entry 4) the more bulky aryl moiety reduced the reactivity of
the arylalane. To solve this the triarylalane 10 was preꢀactivated by treatment with
3 2
AlCl in nꢀBu O. Following the above conclusion that 3 equiv or more of aryl anion
were necessary for good reaction performance, 2 equiv of the aryl(halo)alanes were
used. This resulted in improved reaction times and a yield of greater than 60% (entries
16 and 17). For comparative purposes the sugar was deprotonated with nꢀBuLi instead
2
3
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 24 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
of Me
3
Al, giving 7d, and was then arylated under the same conditions (entry 18). A
slightly lower HPLC yield (55%) of 4i was obtained and more TBDPSOH (53%) was
seen than in the test (29%; entry 17) using Me Al. Comparison of this result to the
3
corresponding model reaction using Ph AlCl and nꢀBuLi (Table 3, entry 11) again
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
showed that the bulkier aryl moiety reduced reaction efficiency. Finally, entry 17 was
repeated on a preparative scale providing an improved 75% yield of
chromatographically purified 2,4ꢀdiꢀOꢀprotected canagliflozin (1a; Scheme 3) that
was converted into canagliflozin by desilylation using TBAF in THF.
AlCl , n-Bu O
3
2
F
F
F
X
S
90 °C
Al
S
3
10
9a: X = Br
9b: X = Li
n-BuLi in n-hexane
PhMe/i-Pr O, 0 °C
AlCl , n-Bu O
3
2
2
S
ClAl
2
11
11
Me Al
n-Bu2O, PhOMe,
40 °C, o/n
O
3
F
HO
S
1
PhMe
PhOMe
rt
2a
7b
1
OR1
R O
7
5% isolated yield
OH
1
4
i: R = TBDPS
TBAF, THF, rt
8%
1
1
a: R = H (canagliflozin)
6
Scheme 3. Synthesis of canagliflozin via the arylation of 2a with an
aryl(halo)alane
Over the course of these investigations it was noted that the formation of βꢀCꢀ
arylglucosides 4 and the enol ether sideꢀproduct 5 were inextricably linked. Although
the proportion of the two differed under different reaction conditions, the relative
rates of their formation implied a common intermediate. We propose that arylation
2
4
ACS Paragon Plus Environment

Page 25 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
first involves the formation of a complex between the arylalanes and the C6ꢀO ether
linkage of the sugar and ring opening of the 1,6ꢀanhydro ring then produces the
requisite oxocarbenium ion A that then undergoes either i) the desired arylation (path
3
8
a) to furnish Cꢀarylglucoside 4, possibly by ipso substitution, or ii) the undesired
elimination pathway (path b) via C2 deprotonation to ultimately produce enol ether 5
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(
Scheme 4). While path b is seemingly less favoured, examination of a Dreiding
model of the oxocarbenium ion suggests that selection of either path a or path b might
require only a small reorientation of the C6ꢀO arylaluminate moiety, making
39
16
exclusion of unwanted path b difficult. Finally, in related work by Vasella et al. on
the stereoretentive alkynylation of 1,6ꢀanhydrohexoses using trimethylsilylꢀprotected
propargylaluminium dichloride to yield βꢀCꢀalkynylglycosides, it was proposed that a
bidentate alkynylaluminium chloride ate complex bridging the C3ꢀO and C6ꢀO
positions delivered the alkyne to the oxocarbenium ion. Basic modelling of the
analogous complex B in our system showed it to be very rigid and the resulting
conformational restriction might hinder aryl group delivery to the oxocarbenium ion
whilst forcing the bulky C2ꢀOTBDPS and C4ꢀOTBDPS groups into an unfavourable
1,3ꢀdiaxial relationship. By contrast, a model of complex A shows that the C6ꢀOꢀ
aluminium ate complex possesses greater freedom allowing for the requisite
alignment of the aryl group with the oxocarbenium ion whilst relieving the 1,3ꢀdiaxial
interaction. Delivery of the aryl group from complex A to the αꢀface of the
oxocarbenium ion on the other hand appears highly unfavourable, requiring an
implausible contortion of the sugar ring, and would be further encumbered by the C2ꢀ
10
OꢀTBDPS group. Moreover, we have shown that 1,6ꢀanhydroꢀ2,3,4ꢀtriꢀOꢀbenzylꢀβꢀ
ꢀglucopyranose can be arylated in good yield with high stereoselectivity (>99:1 β:α
ratio) using Ph Al, confirming that a covalent bond between C3ꢀO and the arylating
agent is not necessarily required for arylation to occur nor for providing the observed
D
3
2
5
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 26 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
high stereoselectivity. Given these reasons it is believed that complex A sufficiently
explains the high stereoselectivity witnessed in this reaction.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 4. Proposed mechanism to account for the coꢀformation of βꢀ
arylglucoside 4 and silylenol ether 5
Cꢀ
CONCLUSION
A new approach useful for the preparation of βꢀCꢀarylglucosides that utilises
arylalanes that are readily prepared from AlCl and Grignard reagents or aryl lithium
3
compounds has been established. Modification of triarylalanes by replacing one or
more aryl groups with the conjugate bases of strong Brønsted acids produced more
reactive arylating agents, while incorporation of alkyl dummy ligands into the
arylating agents produces viable arylating agents too. Of the modified arylalanes,
2
6
ACS Paragon Plus Environment

Page 27 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
aryl(chloro)alanes were most practical and fine tuning of the stoichiometric ratio of
the aryl anion and halide ligands, which was required on a caseꢀbyꢀcase basis, resulted
in significantly different reactivity and product yields. In fact, triarylalanes,
aryl(halo)alanes, aryl(alkyl)haloalanes and aryl(alkyl)alanes were all shown to be
effective arylating agents. The reaction was tolerant of a range of aryl groups but was
sensitive to the presence of halide salts, particularly when using triarylalanes. This
problem was mitigated by removal of the salts by filtration before arylation. Despite
the presence of a free hydroxyl group on the sugar substrate, blocking was
accomplished by in situ treatment of the sugar with common alkyl aluminium
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3 2
reagents, including Me Al or iꢀBu AlH. This approach avoids a dedicated protection
and deprotection step and it has not escaped our attention that the blocking of
hydroxyl groups with alkyl aluminium reagents might find other synthetic
applications where a conventional protecting group would otherwise be required.
3
Over the course of our investigations using the model arylalane, Ph Al, reaction
times were reduced from 118 h in the early studies in PhMe to just 3.5 h in PhOMe by
activation of the arylalanes with AlCl whilst maintaining good yields. The loading of
3
Ph
3
Al was decreased from 2 equiv down to 1 equiv by blocking of the C3 hydroxyl
10
3
group of 2a by preꢀtreatment with Me Al. Finally, this methodology has been shown
workable using other hydroxyl protecting groups and will be reported in due course.
EXPERIMENTAL SECTION
Materials and Methods. Experiments were conducted under anhydrous conditions in
a nitrogen atmosphere using Schlenk techniques. Solvents were dried over 3Å
molecular sieves and ovenꢀdried glassware and gasꢀtight syringes were used.
Commercially obtained or inꢀhouse prepared solutions of organometallic reagents
2
7
ACS Paragon Plus Environment

The Journal of Organic Chemistry
Page 28 of 54
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
were regularly titrated prior to use using standard titration methods, including
40
Knochel’s method. AlCl
3
was titrated by Eriochrome cyanine R spectrophotometric
ꢀglucopyranose (3), Grignard reagents, PhLi and Ph Al
in nꢀBu O) were purchased and used as supplied. Reference samples of
41
method. 1,6ꢀAnhydroꢀβꢀ
(1.0
D
3
M
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
canagliflozin (1a), dapagliflozin (1b) and levoglucosenone (6) were aquired from
commercial sources. Authentic samples of αꢀ1e and βꢀ1e were prepared as per ref 26.
Analyses of crude reaction mixtures or isolated products were performed by reversedꢀ
phase HPLC on a XDBꢀC18 column (3.5 µm, 4.6×150 mm) at 30 °C monitoring at
2
10 nm eluting with a linear gradient from 10% to 100% 0.1% aq TFA–MeCN for 15
min followed by isocratic elution with MeCN for 12 min at a 1 mL/min flow rate.
Samples for HPLC analysis were preꢀtreated with 5% TFA in MeCN (1000 ꢁL per 10
ꢁ
L of reaction sample). When HPLC yields were required the following HPLC assay
method was used: 1,2,4,5ꢀtetramethylbenzene (ca. 0.4 equiv with respect to 2a) was
added into the reaction prior to heating and the peak area% of the peak of interest
(e.g., 2a, 4a, or TBDPSOH) was then corrected using its predetermined relative
response factor with respect to 1,2,4,5ꢀtetramethylbenzene at 210 nm (2a: 4.36, 4a:
4.78, TBDPSOH: 1.90; compound 4b–h were assumed to have similar response
factors as 4a while compound 8a–d were assumed to have the same response factor as
2
4
1
a) and compared to the peak area% of 1,2,4,5ꢀtetramethylbenzene. βꢀCꢀarylglucoside
a, biphenyl, benzene, phenol, silyl enol ether 5 were seen at t 24.0 min, 13.7 min,
0.1 min, 6.6 min and 23.5 min, respectively. βꢀCꢀarylglucosides 4b, 4c, 4d, 4e, 4f,
25.3 min, 28.3 min, 24.6 min, 23.0 min, 21.9 min, 22.9 min,
3.0 min respectively. βꢀCꢀarylglucosides 4i and 4j were analysed using the following
HPLC conditions: XBridge C8 column (3.5 µm, 4.6×150 mm) at 40 °C monitoring at
10 nm eluting with a linear gradient from 50:50 to 0:100 (v/v) of 0.02% TFA–H
and 0.02% TFA–MeCN for 18 min followed by isocratic elution with 100% 0.02%
R
4g and 4h were seen at t
R
2
2
2
O
2
8
ACS Paragon Plus Environment

Page 29 of 54
The Journal of Organic Chemistry
1
2
3
4
5
6
7
8
9
R
TFA–MeCN for 7 min at a 1.2 mL/min flow rate). 4i and 4j were seen at t 20.5 min
and 20.4 min, respectively, while the relative response factors in this system of 2a, 4i
and 4j were 3.45, 4.20 and 5.35, respectively. High resolution electrospray ionization
(ESI) mass spectrometry was performed using a QToF tandem mass analyzer.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
,6ꢀAnhydroꢀ2,4ꢀdiꢀOꢀtertꢀbutyldiphenylsilylꢀβꢀ
D
ꢀglucopyranose
(
2a). To
a
suspension solution of 1,6ꢀanhydroꢀβꢀ ꢀglucopyranose (3, levoglucosan, 1.83 g, 11.3
D
mmol) and imidazole (3.07 g, 45.2 mmol) in THF (10 mL) at 0 °C was added
dropwise a solution of TBDPSCl (11.6 mL, 45.2 mmol) in THF (10 mL). After the
mixture was stirred at rt until 3 was consumed (TLC), water (10 mL) was added and
the mixture was extracted with EtOAc (20 mL ×2), washed with brine (10 mL), dried
over Na
2
SO
4
and concentrated. Column chromatography (eluting with 1:20 EtOAc–nꢀ
1
heptane) afforded the title compound as a white powder (5.89 g, 81%). H NMR (400
MHz, CDCl
3
) δ 7.82–7.76 (m, 4H), 7.73–7.70 (m, 4H), 7.49–7.37 (m, 12H), 5.17 (s,
1
2
H), 4.22 (d, J= 4.8 Hz, 1H), 3.88–3.86 (m, 1H), 3.589–3.586 (m, 1H), 3.49–3.46 (m,
13
H), 3.30 (dd, J= 7.4, 5.4 Hz, 1H), 1.146 (s, 9H), 1.142 (s, 9H); C NMR (100 MHz,
) δ 135.87 (CH ×2), 135.84 (CH ×2), 135.82 (CH ×2), 135.80 (CH ×2), 133.7
(C), 133.5 (C), 133.27 (C), 133.21 (C), 129.92 (CH), 129.90 (CH), 129.87 (CH),
29.86 (CH), 127.81 (CH ×2), 127.80 (CH ×2), 127.75 (CH ×4), 102.3 (CH), 76.9
CH), 75.3 (CH), 73.9 (CH), 73.4 (CH), 65.4 (CH ), 27.0 (CH ×6), 19.3 (C ×2); FTꢀ
CDCl
3
1
2
2
2
(
2
3
IR (neat) 3482, 3071, 3049, 2958, 2931, 2894, 2857, 1472, 1427, 1391, 1362, 1024,
ꢀ
1
20
9
D
99, 900, 825, 841, 702 cm ; [α] = −26.3 (c 1.0, MeOH); ESI QTof calculated for
+
[C38
H
46
O
5
Si
2
+NH
4
] = 656.3222, found 656.3213; mp 126.7 °C.
2
,4ꢀDiꢀOꢀtertꢀbutyldiphenylsilylꢀ1ꢀCꢀphenylꢀβꢀ
in PhMe and 2.0 equiv Ph2.5AlCl0.5. AlCl
solution in THF) and phenylmagnesium bromide (1.9 mL, 5.0 mmol,
D
ꢀglucopyranoside
(
4a
)
Method
A
—using PhMgBr–AlCl
3
3
(4.0 mL, 2.0
mmol, 0.5
M
2
9
ACS Paragon Plus Environment