High-affinity disaccharide binding by tricyclic synthetic lectins

Article abstract of DOI:10.1002/anie.201200447

Stay flexible: Rigid preorganization is not always the best approach to molecular recognition. Unlike previous synthetic lectins, new receptors (see picture) were synthesized that possess conformational freedom which allows hydrophobically driven collapse

Full text of DOI:10.1002/anie.201200447

.
Angewandte
Communications
DOI: 10.1002/anie.201200447
Carbohydrate Recognition
High-Affinity Disaccharide Binding by Tricyclic Synthetic Lectins**
Bunyarithi Sookcharoenpinyo, Emmanuel Klein, Yann Ferrand, D. Barney Walker,
Peter R. Brotherhood, Chenfeng Ke, Matthew P. Crump, and Anthony P. Davis*
Carbohydrate recognition mediates a wide range of biological
processes,[1] including protein folding and trafficking,[2] cell–
cell recognition,[3] infection by pathogens,[4] and many aspects
of the immune response.[5] Molecules capable of selective
carbohydrate binding are, therefore, valuable as tools for
biological research, and potentially as medicinal agents.[6,7]
There are many saccharide-binding proteins, notably the
group known as lectins,[8] but they often show low affinities
and (from a researcherꢀs viewpoint) non-ideal selectivities.[8,9]
Moreover, they are generally too unstable and toxic for use in
medicine. Opportunities thus exist for synthetic systems,[10]
provided they can compete with lectins in terms of binding
strength and selectivity.
Over the past few years we have shown that macro-
polycyclic lactams such as 1 and 3 (Scheme 1) can bind
carbohydrates with all-equatorial substitution patterns in
water, by employing a combination of hydrogen bonding and
hydrophobic/CH-p interactions.[11] Selectivities compare well
with those of lectins; for example, both 1 and 3 bind their
respective targets 2 and 4 with preferences of > 20:1 versus
closely related substrates.[11b,c] However, affinities have been
less competitive. While lectins typically bind monosacchar-
Scheme 1. Previously reported macropolylactam synthetic lectins 1 and
ides and disaccharides with Ka = 103–105 mÀ1 [8,9]
the values for
,
3, with favored substrates d-GlcNAc-b-OMe (2) and d-cellobiose (4).
Binding is driven by hydrogen bonding to equatorial polar groups in
the carbohydrates, and hydrophobic/CH-p interactions to an axial CH
group of the substrate.
1 and 3, at approximately 600mÀ1, lie short of this range. We
now report two new synthetic lectins 5 and 6, which are
related to tetracycle 3 but with a less preorganized and more
accessible architecture. Remarkably, these systems yield
increased performance in key respects, with Ka values up to
4500mÀ1 and extreme selectivity for di- versus monosacchar-
ides. This study provides the best evidence yet that true lectin
mimicry may be achieved with simple, practically viable
receptor structures.
the central aromatic rings in each terphenyl unit (Figure 1). In
water, this collapsed conformation was predicted to be
ꢀ 25 kJmolÀ1 more stable than open structures, presumably
compromising disaccharide binding. We therefore turned to
the more highly connected 3 which, as previously descri-
bed,[11b] proved very effective.
The decision to undertake this study was made by an
indirect route, which nicely illustrates the limitations of
current molecular design. When initially we planned to target
all-equatorial disaccharides, tricyclic system 5 seemed an
obvious solution.[12] However, modeling studies[13] showed
that 5 can undergo a twisting motion which brings together
Despite the success of 3, we remained curious whether the
rigid tetracyclic architecture was necessary or, alternatively,
the tricyclic (and synthetically more accessible) 5 might show
at least moderate binding properties. Having recently found
that 4,4’-alkoxybiphenyl substituents could improve affinities
in monosaccharide receptors,[11d] we were also interested to
test the dimethoxy analogue 6. We therefore synthesized 5
and 6 from haloarenes 7 by sequential Suzuki–Miyaura cross-
coupling reactions, deprotections, and macrolactamizations
(Scheme 2).[14] The method was largely based on earlier work,
but a new development was the use of diazido aryl halides 8 as
coupling partners for boronates 9. While care was required to
prevent azide degradation during coupling, the Staudinger
reduction of tetraazido macrocycles 11 proceeded especially
smoothly. This protection strategy proved superior to alter-
natives such as benzyloxycarbonyl (Cbz), for which removal
could be problematic.
[*] B. Sookcharoenpinyo, Dr. E. Klein, Dr. Y. Ferrand, Dr. D. B. Walker,
Dr. P. R. Brotherhood, Dr. C. Ke, Dr. M. P. Crump, Prof. A. P. Davis
School of Chemistry, University of Bristol
Cantock’s Close, Bristol, BS8 1TS (UK)
E-mail: anthony.davis@bristol.ac.uk
[**] This work was supported by the Royal Thai Government (Fellowship
to B.S.), the Engineering and Physical Sciences Research Council
(grant number EP/D060192/1), and the Royal Society (Newton
International Fellowship to C.K.)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201200447.
4586
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4586 –4590

Angewandte
Chemie
Figure 1. a) Formulas for receptors 5 and 6. b) Ground-state confor-
mation of 5 as predicted by Monte Carlo Molecular Mechanics
(MCMM).[13,14] The two ends of the molecule have twisted relative to
each other, bringing the central terphenyl aromatic rings into proximity.
The central rings are shown in space-filling mode, and the terminal
terphenyl aromatic rings are highlighted in cyan. Water-solubilizing
groups are omitted for clarity.
Scheme 2. Synthesis of receptors 5 and 6. Yields refer to Y=H/
Y=OMe, respectively: a) N-bromosuccinimide, benzoyl peroxide
(cat.), MeOAc, 858C, 48 h, 40/99%; b) NaN3, DMF, 608C, 24 h,
ca. 99%; c) PPh3, THF/H2O, 608C, 24 h, 95/99%; d) Boc2O, DIPEA,
THF, RT, 24 h, 75/83%; e) pin2B2, [PdCl2(dppf)] (3 mol%), KOAc,
DMSO, 808C, 18 h, 80/83%; f) 4-bromo-iodobenzene, [PdCl2(dppf)]
(3 mol%), Na2CO3, DMSO, 508C, 24 h, 85%; g) pin2B2, [PdCl2(dppf)]
(3 mol%), KOAc, DMSO, 808C, 2 days, 85/94%; h) bisazide 8a or 8b,
[PdCl2(dppf)] (3 mol%), Na2CO3, DMF, 808C, 4 h, 72/74%; i) TFA,
CH2Cl2, 08C, 5 h, 78/89%; j) DIPEA, THF, 30 h injection, RT, 24 h,
20/52%; k) PPh3, THF/H2O, 608C, 24 h, 78/91%; l) 10, THF, 30 h
injection, RT, 24 h, 29/58%; m) TFA, CH2Cl2, 08C, 5 h, 98/99%;
n) aq NaOH, 95/98%. Boc=tert-butoxycarbonyl, DIPEA=N,N-diiso-
propylethylamine, pin2B2 =bispinacolato diboron, [PdCl2(dppf)]=
dichlorodiphenylphosphinoferrocene palladium(II) complex, TFA=tri-
fluoroacetic acid.
Macrocycles 5 and 6 were examined first by 1H NMR
spectroscopy in D2O. The parent system 5 gave broadened
signals which could not be used for binding studies, but
fortunately the tetramethoxy analogue 6 gave well-resolved
spectra. Dilution of an NMR sample from 0.5 mm to 0.15 mm
caused negligible changes to the spectra, thus implying that 6
does not aggregate in this concentration range. Carbohydrate
recognition by 6 was then studied by 1H NMR titrations, using
cellobiose (4) and saccharides 12–23 (Scheme 3) as substrates.
The addition of 4 as well as several other substrates caused
movements of signals in the aromatic region of the spectrum,
which implies binding with fast-medium exchange on the
1H NMR timescale (Figure 2). Broadening was observed for
some signals in some cases, but at least one proton could
always be followed easily throughout a titration. The move-
ments gave excellent fits to a 1:1 binding model, and were
analyzed to give the binding constants Ka listed in Table 1.
Interestingly, the signals for aromatic protons A and C were
split in some cases, while those for B and D remained singlets.
This is consistent with the D2h symmetry and conformational
properties of the host. A carbohydrate entering the host
renders all the protons different but, when fast exchange
occurs, movement between four substrate orientations results
in equivalence for groups of four protons. Thus the signal for
B (4 protons) remains a singlet, while those of A (8 protons)
and C (8 protons) are split into two groups of 4. The 8 protons
of D appear as a singlet, because the two environments
generated by the carbohydrate are exchanged by rotation of
the C6H4 unit.
To support the NMR studies on 6, we also employed
fluorescence titrations and isothermal titration calorimetry
(ITC). The addition of carbohydrates to a solution of 6 in
water caused increases in the fluorescence output which,
though modest, fit well to a 1:1 binding model (for example,
see Figure 3). The Ka values obtained by this method are
compared to the NMR-derived figures in Table 1. The ITC
traces (e.g. Figure 3) were also consistent with 1:1 complex-
ation,[14] and gave the thermodynamic parameters shown in
Table 2. The fluorescence and ITC techniques were also
applicable to parent tricycle 5. Receptor 5 behaved similarly
to 6 in these experiments, and gave the binding constants
Angew. Chem. Int. Ed. 2012, 51, 4586 –4590
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
4587

.
Angewandte
Communications
Figure 2. Assigned partial 1H NMR spectra (500 MHz, D2O, 298 K) of
6 alone (0.5 mm) and with added methyl b-d-cellobioside (12), N,N’-
diacetyl-d-chitobiose (14), d-cellobiose (4), d-lactose (15), and
d-xylobiose (13). Correlations between signals are indicated by dotted
lines.
Scheme 3. Carbohydrates used as substrates for binding studies.
Table 1: Association constants (Ka) for 1:1 complexes of receptors 5 and
6 with carbohydrate substrates in water, as determined by 1H NMR and
fluorescence titrations. Data for 1 and 3 are given for comparison.[a]
Carbohydrate
Ka [mÀ1
]
1
3
5[b]
6[c]
6[b]
methyl b-d-cellobioside (12)
d-cellobiose (4)
d-xylobiose (13)
850
580
270
4500 4470
3140 3340 3330
17
210
910
N,N’-diacetyl-d-chitobiose (14) ca. 0 120
d-lactose (15)
d-maltose (16)
ca. 0 14
ca. 0 11
5
230
67
270
72
35
ca. 0 ca. 0
ca. 0 ca. 0
87
2
9
ca. 0
240
80
36
d-gentiobiose (17)
d-trehalose (18)
d-sucrose (19)
d-cellotriose (20)
d-glucose (21)
ca. 0
ca. 0
9
56
2
11[d]
24
3
9
N-acetyl-d-glucosamine (22)
d-galactose (23)
[a] T=298 K unless otherwise stated. Values for 1 and 3 from
Refs. [11a,b], respectively. Values denoted ca. 0 were too small for
Figure 3. Data and analysis curves for (left) fluorescence and (right)
ITC binding studies on receptor 6 + cellobiose 4 in H2O at 298 K.[14]
Both analyses employ a simple 1:1 binding model. Ka =3330 and
3300mÀ1, respectively.
analysis. Values for reducing sugars are weighted averages of those for
the two anomers, as discussed in Ref. [11d]. [b] Fluorescence titration in
H2O. Errors estimated at ꢁ4%. [c] 1H NMR titration in D2O. Errors
estimated at ꢁ5%. [d] T=278 K.
listed in Tables 1 and 2, respectively. For both 5 and 6, the
agreement between different methods was excellent, espe-
cially for stronger-binding substrates. For example, in the case
of 6 +4, the NMR spectroscopic, fluorescence, and ITC
measurements gave Ka = 3340, 3330, and 3300mÀ1, respec-
tively.
Two features stand out when the binding constants in
Tables 1 and 2 are considered. Firstly, against our expect-
4588
www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4586 –4590

Angewandte
Chemie
Table 2: Association constants (Ka) and thermodynamic quantities for
1:1 host–guest complexes between receptors 3, 5 and 6 and carbohy-
drate substrates in water at 298 K, as measured by ITC.
Complex
Ka
DH
TDS
DG
[mÀ1
]
[kJmolÀ1
]
[kJmolÀ1
]
[kJmolÀ1
]
3 + cellobiose
(4)[a]
650
À13.46
2.59
À16.05
5 + cellobiose (4) 3110
5 + lactose (15) 220
5 + maltose (14) 61
À18.37
À8.41
À3.60
À18.86
À9.82
À4.45
1.52
4.90
6.82
1.23
3.82
6.39
À19.89
À13.30
À10.42
À20.08
À13.64
À10.83
6 + cellobiose (4) 3300
6 + lactose (15)
250
6 + maltose (14) 89
Figure 4. Possible structure for complex 6·12 derived from molecular
[a] From Ref. [11b].
modeling studies. Shown is the ground-state geometry from an
MCMM search in which both the receptor and substrate were allowed
conformational freedom.[14] Terphenyl aromatic groups are shown in
space-filling mode, isophthalamide aromatic rings are colored cyan.
Water-solubilizing groups are omitted for clarity.
ations, the affinities of both 5 and 6 for their target substrates
4 and 12 are considerably higher than achieved by the more
rigid system 3. Removing the fifth isophthalamide bridge,
straightening the terphenyl units, and allowing the framework
to flex has strengthened binding by factors of 5–6. Indeed, the
affinity of 6 for methyl cellobioside (12), at 4500mÀ1, is the
highest yet observed for a synthetic receptor binding an
uncharged carbohydrate substrate in water. Perhaps surpris-
ingly, given our previous experience with monosaccharide
receptors,[11d] the methoxy substituents made little difference;
the binding constants to 5 and 6 were almost identical.
Secondly, the selectivity between disaccharides has been
reduced (e.g. cellobiose/lactose = 40:1 for 3, 13:1 for 6), but
the preference for disaccharides versus monosaccharides has
been increased. Thus the cellobiose/glucose selectivity has
risen from 50:1 for 3 to about 1300:1 for 6. Selectivity
between di- and monosaccharides may not be especially
difficult to achieve, but again this system sets new records.
An advantage of the generally high affinities is that
accurate ITC measurements are feasible for 5 and 6 with both
target and nontarget substrates. As shown in Table 2, binding
is driven by both enthalpy and entropy changes, but with
enthalpy–entropy compensation such that enthalpy domi-
nates for strongly bound complexes. Comparison with liter-
ature data[9] shows that these numbers are within the bounds
observed for lectins, although with entropy contributions
which are above-average for the natural systems. An
enthalpy–entropy plot based on Table 2 is almost linear (see
Figure S69 in the Supporting Information), as observed for
other closely related host–guest pairings.[15]
of the cavity. MCMM calculations generated a number of
conformations consistent with these data (Figure 4). Typically,
these feature 4–6 intermolecular hydrogen bonds and approx-
imately 10 CH-p interactions. Molecular dynamics simula-
tions predicted that the complex should stay intact for at least
10 ns at 300 K.[14]
It is interesting to compare the structure in Figure 4 with
the NOESY-based model previously obtained for 3·4. Unsur-
prisingly, given the additional spacer unit in 3, the latter
complex features more intermolecular hydrogen bonds
(ca. 10). On the other hand, the linear p-terphenyl unit in 6
may be slightly more compatible with the cellobiose CH
group than the bent, m-substituted system in 3 (see Figur-
es S74/75 in the Supporting Information). Thus, the strength
of binding to 5 and 6 tends to reinforce our view that apolar
hydrophobic/CH-p interactions provide the major driving
force for carbohydrate recognition in water.[16] Whatever the
reason, it is remarkable that 5 and 6 can outperform 3 despite
the (at most) transient nature of their cavities (cf. Figure 1).
In conclusion, we have shown that tricyclic synthetic
lectins 5 and 6 are even more effective than tetracyclic 3 at
binding all-equatorial disaccharides under biomimetic con-
ditions. The success of these less-connected structures sug-
gests that an “induced fit” or “conformational selection”[17]
approach can be superior to rigid preorganization in carbo-
hydrate recognition, and may point the way to new, even
simpler systems with potential for applications. In particular,
the disaccharide substrates are representative of major
biopolymers (cellulose, xylan, chitin), and receptors which
bind the polymers themselves could have biological activity or
serve as aids to processing (e.g. by promoting solubility).
Future studies will focus on these possibilities.
Structural aspects of binding were investigated by NMR
spectroscopy and molecular modeling studies. We focused
especially on 6 +12, which form a strongly bound complex
and avoid the complications caused by substrate anomers.
Intermolecular cross-peaks in the NOESY spectrum provided
clear evidence that, as expected, 12 enters the cavity of 6 to
make hydrophobic/CH-p contacts with the terphenyl units.
Thus, strong interactions were observed between axial sub-
strate CH and receptor protons C and D (see Figure 2 for
labeling), while cross-peaks from C/D to the C(6) protons
were weaker. In particular, a correlation between the
cellobioside MeOCH proton and one receptor proton C
places the methyl glycosidic unit unambiguously in one corner
Received: January 17, 2012
Published online: March 29, 2012
Keywords: biomimetic hosts · carbohydrates ·
molecular recognition · receptors · supramolecular chemistry
.
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4586 –4590
www.angewandte.org 4589

.
Angewandte
Communications
Int. Ed. 2010, 49, 7340; S. Striegler, M. G. Gichinga, Chem.
Commun. 2008, 5930; T. Reenberg, N. Nyberg, J. O. Duus, J. L. J.
van Dongen, M. Meldal, Eur. J. Org. Chem. 2007, 5003; H. Goto,
Y. Furusho, E. Yashima, J. Am. Chem. Soc. 2007, 129, 9168; M.
Mazik, H. Cavga, J. Org. Chem. 2006, 71, 2957.
[1] The Sugar Code—Fundamentals of Glycoscience (Ed.: H.-J.
Gabius), Wiley-Blackwell, Weinheim, 2009; H. J. Gabius, H. C.
Siebert, S. Andre, J. Jimꢁnez-Barbero, H. Rudiger, ChemBio-
Chem 2004, 5, 740; R. A. Dwek, T. D. Butters, Chem. Rev. 2002,
102, 283; C. R. Bertozzi, L. L. Kiessling, Science 2001, 291, 2357.
[2] J. Roth, Chem. Rev. 2002, 102, 285.
[3] S.-i. Hakomori, K. Handa in Carbohydrates in Chemistry and
Biology (Eds.: B. Ernst, G. W. Hart, P. Sinay), Wiley-VCH,
Weinheim, 2000, p. II 771.
[4] T. Angata, A. Varki, Chem. Rev. 2002, 102, 439.
[5] R. Kannagi in Carbohydrate-Based Drug Discovery (Ed.: C.-H.
Wong), Wiley-VCH, Weinheim, 2003, p. 803.
[6] J. Balzarini, Nat. Rev. Microbiol. 2007, 5, 583.
[7] S. Jin, Y. F. Cheng, S. Reid, M. Y. Li, B. H. Wang, Med. Res. Rev.
2010, 30, 171.
[8] M. Ambrosi, N. R. Cameron, B. G. Davis, Org. Biomol. Chem.
2005, 3, 1593; H. Lis, N. Sharon, Chem. Rev. 1998, 98, 637.
[9] E. J. Toone, Curr. Opin. Struct. Biol. 1994, 4, 719.
[10] For reviews, see Ref. [7] and: D. B. Walker, G. Joshi, A. P. Davis,
Cell. Mol. Life Sci. 2009, 66, 3177; M. Mazik, Chem. Soc. Rev.
2009, 38, 935; T. D. James, M. D. Phillips, S. Shinkai, Boronic
Acids in Saccharide Recognition, RSC, Cambridge, 2006; A. P.
Davis, R. S. Wareham, Angew. Chem. 1999, 111, 3160; Angew.
Chem. Int. Ed. 1999, 38, 2978; For recent examples see: G.
Fukuhara, Y. Inoue, J. Am. Chem. Soc. 2011, 133, 768; A. Pal, M.
Berube, D. G. Hall, Angew. Chem. 2010, 122, 1534; Angew.
Chem. Int. Ed. 2010, 49, 1492; M. Rauschenberg, S. Bomke, U.
Karst, B. J. Ravoo, Angew. Chem. 2010, 122, 7498; Angew. Chem.
[11] a) E. Klein, M. P. Crump, A. P. Davis, Angew. Chem. 2005, 117,
302; Angew. Chem. Int. Ed. 2005, 44, 298; b) Y. Ferrand, M. P.
Crump, A. P. Davis, Science 2007, 318, 619; c) Y. Ferrand, E.
Klein, N. P. Barwell, M. P. Crump, J. Jimꢁnez-Barbꢁro, C. Vicent,
G. J. Boons, S. Ingale, A. P. Davis, Angew. Chem. 2009, 121, 1807;
Angew. Chem. Int. Ed. 2009, 48, 1775; d) N. P. Barwell, M. P.
Crump, A. P. Davis, Angew. Chem. 2009, 121, 7809; Angew.
Chem. Int. Ed. 2009, 48, 7673.
[12] We had previously found that a lipophilic analogue of 5 binds an
organic-soluble cellobiose derivative in chloroform/methanol
mixtures; see G. Lecollinet, A. P. Dominey, T. Velasco, A. P.
Davis, Angew. Chem. 2002, 114, 4267; Angew. Chem. Int. Ed.
2002, 41, 4093.
[13] MacroModel 9.9 (MMFFs force field, water GB/SA solvation)
accessed through the Maestro 9.2 interface.
[14] For further details, see the Supporting Information.
[15] D. H. Leung, R. G. Bergman, K. N. Raymond, J. Am. Chem. Soc.
2008, 130, 2798.
[16] E. Klein, Y. Ferrand, N. P. Barwell, A. P. Davis, Angew. Chem.
2008, 120, 2733; Angew. Chem. Int. Ed. 2008, 47, 2693.
[17] P. Csermely, R. Palotai, R. Nussinov, Trends Biochem. Sci. 2010,
35, 539; D. Bucher, B. J. Grant, J. A. McCammon, Biochemistry
2011, 50, 10530.
4590 www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4586 –4590