Polyacrylamides bearing pendant α-sialoside groups strongly inhibit agglutination of erythrocytes by influenza virus: The strong inhibition reflects enhanced binding through cooperative polyvalent interactions

Article abstract of DOI:10.1021/ja953729u

An ELISA assay is described for measuring the binding of influenza virus A-X31 to α-sialoside groups that are linked to biotin-labeled polyacrylamides. The efficacy of these polymers in inhibiting the adhesion of influenza virus to erythrocytes (as measured by a hemagglutination assay) was shown to be directly related to the binding affinity of the polymers for the viral surface: the differences in inhibitory efficacy among the polymeric inhibitors and monomeric α-methyl sialoside, among fractions of a polymeric, polyvalent inhibitor with narrow molecular weight ranges, and among polymeric inhibitors prepared by copolymerization or modification of a preformed polymer chain, all correlated with differences in the affinity of the inhibitors for the surface of the virus. The polymeric inhibitors studied had affinities for the viral surface that ranged between 10³ and >10⁶ greater than α-methyl sialoside, on the basis of total sialic acid groups in solution. The role of steric stabilization in the mechanism by which these polymers inhibit hemagglutination was investigated. The ability of the polymeric, polyvalent inhibitors to inhibit the binding of a polyclonal antibody to the viral surface suggests that steric stabilization may also be an important effect in this system.

Full text of DOI:10.1021/ja953729u

VOLUME 118, NUMBER 16
APRIL 24, 1996
© Copyright 1996 by the
American Chemical Society
Polyacrylamides Bearing Pendant R-Sialoside Groups Strongly
Inhibit Agglutination of Erythrocytes by Influenza Virus: The
Strong Inhibition Reflects Enhanced Binding through
Cooperative Polyvalent Interactions
George B. Sigal, Mathai Mammen, Georg Dahmann, and George M. Whitesides*
Contribution from the Department of Chemistry, HarVard UniVersity, 12 Oxford Street,
Cambridge, Massachusetts 02138-2902
ReceiVed NoVember 6, 1995X
Abstract: An ELISA assay is described for measuring the binding of influenza virus A-X31 to R-sialoside groups
that are linked to biotin-labeled polyacrylamides. The efficacy of these polymers in inhibiting the adhesion of influenza
virus to erythrocytes (as measured by a hemagglutination assay) was shown to be directly related to the binding
affinity of the polymers for the viral surface: the differences in inhibitory efficacy among the polymeric inhibitors
and monomeric R-methyl sialoside, among fractions of a polymeric, polyvalent inhibitor with narrow molecular
weight ranges, and among polymeric inhibitors prepared by copolymerization or modification of a preformed polymer
chain, all correlated with differences in the affinity of the inhibitors for the surface of the virus. The polymeric
inhibitors studied had affinities for the viral surface that ranged between 103 and >106 greater than R-methyl sialoside,
on the basis of total sialic acid groups in solution. The role of steric stabilization in the mechanism by which these
polymers inhibit hemagglutination was investigated. The ability of the polymeric, polyvalent inhibitors to inhibit
the binding of a polyclonal antibody to the viral surface suggests that steric stabilization may also be an important
effect in this system.
Introduction
has been difficult because the binding pocket is small and
shallow.3,4 Solubilized HA binds only weakly (Kd ∼ 3 mM)
to R-methylsialoside (1).5 The best monomeric inhibitors to
date are SA derivatives linked to large hydrophobic groups at
the 2 and 4 positions.6,7 The tightest binding of these inhibitors,
2, has a dissociation constant in the micromolar range (Kd ) 3
µM).7
Although the interaction of a single HA binding site with a
single SA is weak, the binding of a particle of virus to the
surface of a cell is strong (i.e. stable complexes of viral particles
and red blood cells can form in suspensions containing total
The first step in the infection of a cell by the influenza A-X31
virus is the attachment of the virus to the cell membrane. The
attachment occurs through the interaction of hemagglutinin
(HA), a sialic acid (SA) receptor on the virus, with SA groups
linked to glycoproteins and glycolipids on the surface of the
cell.1,2 In theory, one approach to preventing infection by the
influenza virus might be to design analogs of SA that bind
tightly to HA and prevent attachment of the virus to cells. In
practice, however, the design of tight binding inhibitors of HA
X Abstract published in AdVance ACS Abstracts, April 1, 1996.
(1) Paulson, J. C. In Interactions of Animal Viruses with Cell Surface
Receptors, Conn, P. M., Ed.; Academic Press: Orlando, FL, 1985; Vol. 2,
pp 131-219.
(3) Sauter, N. K.; Hanson, J. E.; Glick, G. D.; Brown, J. H.; Crowther,
R. L.; Park, S. J.; Skehel, J. J.; Wiley, D. C. Biochemistry 1989, 28, 8388-
8396.
(4) Weis, W.; Brwon, J. H.; Cusack, S.; Paulson, J. C.; Skehel, J. J.;
Wiley, D. C. Biochemistry 1992, 31, 9609-9621.
(2) Wiley, D. C.; Skehel, J. J. Annu. ReV. Biochem. 1987, 56, 365-394.
0002-7863/96/1518-3789$12.00/0 © 1996 American Chemical Society

3790 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
Scheme 1. Cartoon Showing Volume Restriction and Loss of
Water (W) from a Sterically Stabilized Viral Particle on
Approach of a Red Blood Cella
concentrations in solution of HA and SA, on the surfaces of
the virus and cell, that are well under the Kd for the interaction
of a single HA with a single SA).8 We believe that this strong
binding reflects the interaction of multiple copies of HA on the
viral surface simultaneously with multiple SA groups on the
surface of the cell.9 Based on the idea that polyvalent interaction
is required for tight binding of the cell to the virus, we and
others have developed inhibitors that also present multiple copies
of SA to the virus; these polyvalent inhibitors are effective at
preventing the attachment of influenza virus to red blood cells
at very low concentrations of inhibitor.10-19
These studies determined the effectiveness of the inhibitors
by an assay measuring the inhibition of the agglutination of
red blood cells (hemagglutination inhibition, HAI). The inhibi-
tion constant (KHi AI) is the minimum concentration of inhibitor
required to inhibit hemagglutination by the virus in a 96-well
assay format.20 Here and throughout this paper, KHi AI for
polyvalent inhibitors is calculated on the basis of the total
concentration of SA groups in the system (whether as monomers
attached to polymers, or linked to the surface of liposomes or
cells) and not on the basis of the concentration of polymer
molecules. Liposomes incorporating lipids derivatized with SA
analogs prevent hemagglutination with inhibition constants as
low as KiHAI ) 10 nM .10,11 Polyacrylamide chains bearing
pendant R-sialoside groups are also effective inhibitors of
hemagglutination,12-19 and viral replication.12b Polymers formed
by the copolymerization of acrylamide with acrylamide
derivatives bearing SA analogs have inhibition constants as low
as KiHAI ) 200 nM.12-18 More recently, we have prepared
polyacrylamides presenting SA groupssusing the reaction of
a Both changes are associated with unfavorable energy terms due to
losses of both conformational entropy and favorable water-polymer
interactions.
preformed polymers of N-acryloyloxysuccinimide with an
amine-functionalized SA derivative and ammoniasthat are the
most effective known inhibitors of hemagglutination by influ-
enza virus (KHi AI ) 1 nM).19
A balance of three factorsstwo favorable and one un-
favorablesare believed to contribute to the effectiveness of these
polyvalent inhibitors in preventing viral attachment. First, there
is an increase in the affinity of the polyvalent inhibitor for the
surface of the virus, relative to a polymer having only one (or
a few) copies of the ligand, due to the binding of multiple SA
groups per inhibitor molecule. Second, the large effective
molecular volume of these high molecular weight inhibitors
prevents the virus from coming close enough to a cell for the
viral receptors and cell-surface ligands to interact. Third (and
unfavorable), a single ligand attached to a polymer binds less
tightly to the receptor than a ligand in a low molecular weight
form.18
While inhibition of viral attachment by a monomeric inhibitor
requires a concentration of monomer high enough to block at
least half of the SA binding sites on the virus, high molecular
weight inhibitors might only need to make enough attachments
to achieve a critical amount of excluded volume around the virus
and, in theory, could target receptors other than those involved
directly in viral attachment. For inhibitors having large
structures with a defined shape, such as liposomes whose
surfaces are decorated with ligands, the structures physically
block the approach of the red blood cell to the virus.21 For
inhibitors with no defined shape, such as the copolymers of
acrylamide, the blocking effect is believed to be more like the
well-known ability of adsorbed polymers to prevent the ag-
gregation (flocculation) of colloidal dispersions (steric sta-
bilization).22-25 Scheme 1 is a cartoon representing the approach
of a red blood cell to a viral particle coated with a well-solvated
layer of polymer. As illustrated in the cartoon, the approach
of the virus to the surface of the red blood cell results in
compression of the polymer chains. This compression results
in two, potentially large, repulsive energy terms. The firsts
volume restrictionsis due to loss of conformational entropy as
(5) Sauter, N. K.; Bednarski, M. D.; Wurzburg, B. A.; Hanson, J. E.;
Whitesides, G. M.; Skehel, J. J.; Wiley, D. C. Biochemistry 1989, 28, 8388-
8396.
(6) Toogood, P. L.; Galliker, P. K.; Glick, G. D.; Knowles, J. R. J. Med.
Chem. 1991, 34, 3138-3140.
(7) Weinhold, E. G.; Knowles, J. R. J. Am. Chem. Soc. 1992, 114, 9270-
9275.
(8) For detailed descriptions of the interaction of influenza virus with
cells see: Paulson, J. C. In The Receptors; Conn, P. N., Ed.; Academic
Press: New York, 1985; Vol. 2, Chapter 5. Wharton, S. A.; Weiss, W.;
Skehel, J. J.; Wiley, D. C. In The Influenza Viruses, Krug, R. M., Ed.;
Plenum: New York, 1989; Chapter 3.
(9) Matrosovich, M. N. FEBS Lett. 1989, 252, 1. Ellens, H.; Bentz, J.;
Mason, D.; Zhang, F.; White, J. M. Biochemistry 1990, 29, 2697.
(10) Kingery-Wood, J. E.; Williams, K. W.; Sigal, G. B.; Whitesides,
G. M. J. Am. Chem. Soc. 1992, 114, 7303-7305.
(11) Spevak, W.; Nagy, J. O.; Charych, D. H.; Schaefer, M. E.; Gilbert,
J. H.; Bednarski, M. D. J. Am. Chem. Soc. 1993, 115, 1146-1147.
(12) Spaltenstein, A.; Whitesides, G. M. J. Am. Chem. Soc. 1991, 113,
686-687.
(13) (a) Mastrovich, M. N.; Mochalova, L. V.; Marinina, V. P.;
Byramova, N. E.; Bovin, N. V. FEBS Lett. 1990, 272, 209-212. (b)
Mochalova, L. V.; Tuzikov, A. B.; Marinina, V. P. AntiViral. Res. 1994,
23, 179-190.
(14) Roy, R.; Laferriere, C. A. Carbohydr. Res. 1988, 177, C1-C4.
(15) Gamian, A.; Chomik, M.; Laferriere, G. A.; Roy, R. Can. J.
Microbiol. 1991, 37, 233-237.
(21) Transmission electron microscopy of liposomes presenting SA
groups bound to virus shows each virion is completely coated with several
liposomes (unpublished data).
(22) Napper, D. H. In Polymeric Stabilization; in Colloidal Dispersions;
Goodwin, J. W., Ed.; Royal Society of Chemistry: London, 1982; pp 99-
128.
(23) Fleer, G. J.; Scheutjens, J. M. H. M. In Modeling Polymeric
Adsorption, Steric Stabilization, and Flocculation; in Coagulation and
Flocculation; Dobia´sˇ, B., Ed.; Marcel Dekker Inc.: New York, 1993; pp
209-263.
(24) The Effect of Polymers on Dispersion Properties; Tadros, Th. F.,
Ed.; Academic Press: New York, 1982.
(16) Nagy, J. O.; Wang, P.; Gilbert, J. H.; Schaefer, M. E.; Hill, T. G.;
Callstrom, M. R.; Bednarski, M. D. J. Med. Chem. 1992, 35, 4501-4502.
(17) Sparks, M. A.; Williams, K. W.; Whitesides, G. M. J. Med. Chem.
1993, 36, 778-783.
(18) Lees, W. J.; Spaltenstein, A.; Kingery-Wood, J. E.; Whitesides, G.
M. J. Med. Chem. 1994, 37, 3419-3433.
(19) Mammen, M.; Dahmann, G.; Whitesides, G. M. J. Med. Chem. 1995,
38, 4179-4190.
(20) Rogers, G. N.; Pritchett, T. J.; Laue, J. L.; Paulson, J. C. Virology
1983, 131, 394.
(25) Sato, T.; Richard, R. Stabilization of Colloidal Dispersions by
Polymer Adsorption; Marcel Dekker: New York, 1980.

Inhibition of Agglutination of Erythrocytes
J. Am. Chem. Soc., Vol. 118, No. 16, 1996 3791
the polymer strands are restricted into a smaller volume. The
secondsosmotic repulsionsresults from the expulsion of solvent
(water) on compression; this second term may have both
entropic and enthalpic components.
Scheme 2. Polyacrylamides Bearing Pendant R-Sialoside
Groups (SA) Prepared (a) by Radical Copolymerization of
Acrylamide with an Acrylamide Derivative Containing an
SA Group or (b) by Modification of a Preformed, Activated
PolymersPoly(N-acryloyloxysuccinimide), pNASswith an
Amine Derivative Containing an SA Group and then with
Concentrated Ammonium Hydroxide
The relative importance of enhanced affinity and steric
stabilization in the inhibitory ability of SA-containing polymeric
inhibitors is unknown because the two effects cannot be
distinguished by the HAI assay. One way to separate these
two effects is to measure directly the affinity of SA-containing
polymers for the viral surface and to compare the binding ability
with the effectiveness in preventing hemagglutination. In this
paper we describe an ELISA assay for quantifying the binding
to the influenza virus of polymers containing both SA groups
and biotin. By comparison of the values of KHi AI with the
results of the ELISA assay, we show that the potency of the
polymeric inhibitors in inhibiting hemagglutination correlates
with (and, we presume, reflects) their enhanced affinity for the
viral surface. In particular, we describe experiments that
demonstrate the following: (i) the effective dissociation con-
stants of the polymeric inhibitors for the viral surface as
determined by the ELISA assay are roughly equal to the
measured values of KHi AI; (ii) the differences in the ability of
SA-containing polymers prepared by two different synthetic
routesscopolymerization and modification of a preformed
polymersto inhibit hemagglutination correlates with the relative
affinities of the two types of polymers for the viral surface;
and (iii) the ability of SA-containing polymers to inhibit
agglutination increases with increasing molecular weight, and
this dependence correlates with increases in the affinity of the
polymer for the virus with increasing molecular weight.
Steric stabilization is more difficult to measure directly than
enhancements in affinity. One approach for demonstrating a
role for steric stabilization is to show that the polymeric
inhibitors inhibit the binding to the virus of receptors directed
at sites other than the binding pocket of HA. We show that
the polymeric inhibitors inhibit the binding to the virus of a
polyclonal antibody generated against the viral surface and are,
therefore, capable of steric stabilization of the viral surface.
chain that present a particular group (i.e. øSA ) the number of
monomeric units containing SA divided by the total number of
monomeric units in the polymer chain).
The polymer pA(co-P: Am, SA) was prepared by the free-
radical copolymerization in water of a mixture of 10 equiv of
acrylamide 3 with 1 equiv of the SA-containing acrylamide
derivative 418 (an O-glycoside of SA) under UV light in the
presence of azobis(cyanovaleric acid) as a radical initiator.12,19
Under these conditions all of 4 was consumed over the course
of the polymerization; the proportion of monomer units linked
to SA in the resulting polymer was øSA ≈ 0.09 (as determined
by a colorimetric assay for O-glycoside of SA31). The polymer
pA(pre-P: Am, SA) was synthesized by the reaction of pNAS
(a polymer of N-acryloyloxysuccinimide 5 prepared by free-
radical polymerization in benzene) with 0.2 equiv of SA
derivative 6 (a C-glycoside of SA) per equiv of succinimide
ester groups in DMF, followed by aminolysis of the remaining
active ester groups in aqueous concentrated ammonium hy-
droxide.19 The SA content of this polymer was determined by
elemental analysis and 1H NMR spectroscopy to be øSA ≈ 0.17-
0.18.19 The proportion of monomers that were -CONH2 was
∼0.8 based on elemental analysis (i.e., the proportion of
activated esters that hydrolyzed to COOH was negligible).
Results
Preparation of Polyacrylamides Presenting SA Groups.
Polymers were prepared by two routes: copolymerization
(Scheme 2a) or modification of a preformed, activated, polymer
chain (Scheme 2b).12,18,19 The following nomenclature is used
in this paper to distinguish between the different polymer
preparations: pA(co-P: Am, SA) refers to a polyacrylamide
chain (pA) prepared by copolymerization (co-P) of acrylamide
and an acrylamide derivative bearing an SA group. In this
nomenclature, Am (for primary amide) and SA refer to the
groups presented on the polymer side chains. Similarly, pA-
(pre-P: Am, SA) refers to a polymer also presenting primary
amides and SA groups on the side chains, but prepared by
modification of a preformed polymer chain (pre-P). Inclusion
of biotin (BT) or fluorescein (F) derivatives in the polymer chain
by either of the two synthetic methods is indicated by including
the abbreviation for the label within the parentheses, for
example, pA(pre-P: Am, SA, F). We use the symbol ø to
represent the mole fraction of monomeric units in a polymer
Molecular Weight Distribution of Polyacrylamides Pre-
senting SA Groups. Polymers prepared by free-radical po-
lymerization usually have a broad molecular weight distribu-
tion.26 The molecular weight distribution of the SA-containing
polymers was determined by gel-filtration chromatography
(GFC). To ensure accurate calibration of the GFC column, we
hydrolyzed the polymers to polyacrylic acid under acidic
conditions,27 and compared the GFC elution profile of the
(26) Ishige, T.; Hamielec, A. E. J. Appl. Polym. Sci. 1974, 175, 3117.
(27) Moens, J.; Smets, G. J. Polym. Sci. 1957, 23, 931-948.
(28) The molecular weight distribution was calculated according to:
Meira, G. R. In Modern Methods of Polymer Characterization; Barth, H.
G., Mays, J. W, Eds.; Wiley-Interscience: New York, 1991; pp 67-101.
(29) For narrow molecular weight fractions, Mp ∼ (Mn × Mw)0.5, see
ref 28.
(31) Roy, R.; Tropper, F. D.; Romanowska, A. J. Chem. Soc., Chem.
Commun. 1992, 1611-1613.
(30) Skoza, L.; Mohos, S. Biochem. J. 1976, 159, 457-462.

3792 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
Table 1. Hemagglutination Inhibition Constant (KHi AI), Weight
Average Molecular Weight (Mw), Number Average Molecular
Weight (Mn), and Molecular Weight at the GFC Peak Retention
Time (Mp), for Unfractionated pA(co-P: Am, SA) and
pA(pre-P: Am, SA)
polymer
KHi AI (nM) Mp (kD) Mn (kD) Mw (kD)
pA(co-P: Am, SA)
pA(pre-P: Am, SA)
300
4
490
100
210
70
450
140
hydrolyzate to standards of polyacrylic acid of known molecular
weight. Figure 1a shows the GF chromatograms, after acid
hydrolysis, for pA(co-P: Am, SA) and pA(pre-P: Am, SA).
By calibrating the GFC column using the peak retention time
of the standards (Figure 1a inset), the molecular weight
distributions of the polymers were calculated (Figure 1b).28
Table 1 shows that although pA(pre-P: Am, SA) inhibits
hemagglutination more strongly than does pA(co-P: Am, SA),
it has a smaller chain length, as indicated by the weight and
number average molecular weights and the molecular weight
at the peak maximum (Mw, Mn, and Mp, respectively) for the
hydrolyzed polymer backbones.
Dependence of KHi AI on the Molecular Weight of the
Polymeric Inhibitors. To test the dependence of inhibitory
ability on molecular weight, we prepared narrow molecular
weight fractions of the polymers by preparatory GFC. Figures
2a and 2b show the GF chromatograms for the individual
fractions obtained from the two polymers. Figures 3a and 3b
show the chain length and the concentration of the polymers in
each fraction. We calculated the polymer chain length (N),
given as the number of monomeric units per chain, from the
value of Mp measured for each fraction after acid hydrolysis.29
Integration of the HPLC peak for the hydrolyzed polymers gave
the concentration of the polymer in each fraction. For the pA-
(co-P: Am, SA) fractions, a colorimetric assay gave directly
the concentration of the O-glycosides of SA.30 Figure 3a shows
that the concentration of SA was proportional to the concentra-
tion of polymer and indicates that polymers of all lengths had
similar ratios of the two monomeric components. This result
is evidence that the SA groups in pA(co-P: Am, SA) were
distributed evenly among polymers of different lengths (the
distribution of SA groups along the length of individual polymer
chains, however, remains unknown and could be nonrandom).
There is no convenient assay with sufficient sensitivity to
determine the concentration of C-glycoside SA groups in pA-
(pre-P: Am, SA) fractions, so the concentration of SA groups
was calculated directly from the polymer concentration by
assuming an even distribution of SA groups.
Figure 1. Characterization of acid-hydrolyzed pA(co-P: Am, SA) and
pA(pre-P: Am, SA) by gel filtration chromatography (GFC). (a) GFC
chromatograms for the two polymer samples. The complex peaks at
elution times greater than 1200 s are due to air and low molecular
weight salts in the sample; these peaks also occur after injection of
samples containing no polymer. The inset shows the molecular weight
calibration curve determined for the GFC column with polyacrylic acid
molecular weight standards. (b) Molecular weight distributions for the
two polymer samples. Distributions are given as a function of molecular
weight (MW) or chain length (N) calculated as the number of
monomeric units per chain.
inaccurate, in the sense that the assay underestimates the
effectiveness of the most effective inhibitors.
Due to the complexity of the interactions occurring in the
HAI assay, it is difficult to explain the molecular weight
dependence just on the basis of these data. The longer polymers,
because they contained more SA groups per polymer chain, may
have bound to the virus at a lower concentration of SA groups
due to increased polyvalent binding. Another possibility is that
the longer polymer chains, once adsorbed on the surface of the
virus, provided more steric stabilization. Some combination
of the two factors is also possible. To distinguish between these
possibilities, we developed a direct binding assay for the
interaction between virus and polymer.
We tested the narrow molecular weight fractions for their
ability to inhibit agglutination of chicken red blood cells by
the X-31 strain of influenza A virus (H3N2). Figure 4 gives
KHi AI as a function of molecular weight and chain length. The
log-log plot for pA(co-P: Am, SA) gives a linear least-squares
fit with a slope of -1.1 over a range of N covering close to
two orders of magnitude. The measured slope indicates that
for this range of molecular weights, KHi AI was roughly in-
versely proportional to chain length (i.e. the inhibitory potency
was roughly linearly proportional to chain length). For pA-
(pre-P: Am, SA), KHi AI was also inversely related to chain
length, but the dependence was weaker than linear (slope of
log-log plot ) -0.68). The slope calculated for the pre-P
polymers may be a lower bound; we will demonstrate later in
this paper that the values of KHi AI measured for the pre-P
polymers approach the limits of the HAI assay and that very
small values of KHi AI (that is, very tight binding) may be
Measurement of the Affinity of the Polymeric Inhibitors
for the Surface of Influenza Virus: Binding Studies Using
Biotin-Labeled Polymers. Roy et al. have reported an ELISA
method for measuring the interaction of polyacrylamide co-
polymers bearing sugar and biotin (BT) groups with lectins and
streptavidin.31 In their procedure, the polymer was adsorbed
on a polystyrene surface, labeled with either a lectin-enzyme
or a streptavidin-enzyme conjugate, and quantified using a
standard ELISA protocol. We used a similar approach in our
assay for measuring the binding of SA-containing polymers to
the influenza virus. For our binding assay we prepared
polyacrylamides containing both SA and BT groups both by
the copolymerization protocol (pA(co-P: Am, SA, BT)) and
by reaction with prepolymerized pNAS (pA(pre-P: Am, SA,
BT). The pA(co-P: Am, SA, BT) was prepared by copolymer-
izing acrylamide (3), the SA-containing acrylamide (4), and the
biotin-containing acrylamide (7) at a ratio of 100:10:1. The

Inhibition of Agglutination of Erythrocytes
J. Am. Chem. Soc., Vol. 118, No. 16, 1996 3793
Figure 2. Fractionation of SA-containing polyacrylamides by GFC.
GFC chromatograms from fractionation of (a) pA(co-P: Am, SA) using
a 4000 Å pore size column and (b) pA(pre-P: Am, SA) using a 300 Å
pore size column. Plots show chromatograms from unfractionated
polymer (dashed line) and narrow molecular weight fractions (solid
lines). The sharp peak at an elution time of 380 s in (a) and the complex
peaks at elution times greater than 400 s in (b) are due to air and low
molecular salts in the sample; these peaks also occur after injection of
samples containing no polymer.
Figure 3. Characterization of narrow molecular weight fractions of
(a) pA(co-P: Am, SA) and (b) pA(pre-P: Am, SA). The polymer chain
length (2), given as the number of monomeric units per chain (N),
and the concentration of polymer (O), given as the total concentration
of monomeric units [AAm], is provided for each fraction. For the pAA-
(co-P; Am, SA) fractions, the total concentration of polymer-bound
SA groups [SA] is also given (b).
pA(pre-P: Am, SA, BT) was prepared by sequentially reacting
pNAS with 0.2 equiv of the SA derivative (6) and then 0.01
equiv of the biotin derivative (8) in DMF, followed by
aminolysis of the remaining active esters in aqueous ammonium
hydroxide. Both preparations led to polymers containing a
proportion of BT-linked monomer units, øBT ) 0.01 (as
determined by a colorimetric assay for BT),32 and proportions
of SA-linked monomer units, øSA, roughly equal to the corre-
sponding polymers prepared without BT. The incorporation of
BT groups at low mole fractions into the polyacrylamides
presenting SA groups had no apparent effect on the properties
of the polymers prepared by either method: both the molecular
weight distribution and the KHi AI measured for the polymers
containing SA and BT were indistinguishable from those
prepared without BT groups.
Figure 4. Dependence of hemagglutination inhibition on molecular
weight and polymer chain length (N) for narrow molecular weight
fractions of pA(co-P: Am, SA) (b) and pA(pre-P: Am, SA) ([).
KHi AI is the minimum concentration of polymer (given as the concen-
tration in solution of polymer-linked sialic acid groups) required to
prevent hemagglutination of chicken red blood cells. The molecular
weight (Mp) of the polymer fractions was determined from the retention
time of the GFC peak for polymer that had been hydrolyzed in acid to
polyacrylic acid. The chain length (N) is defined as the number of
monomeric units per polymer chain. The dashed lines show the values
for the unfractionated polymers.
Third, the immobilized virus was treated with a solution of the
biotin labeled polymer, and incubated for 30 minutes (we have
shown previously that incubation for 30 min is sufficient for
polymer-virus mixture to reach equilibrium33a). Fourth, bound
The procedure we used in our ELISA binding assay is
outlined in Scheme 3. First, fetuin (a glycoprotein containing
SA groups in oligosaccharide side chains) was adsorbed non-
covalently on the hydrophobic surface of polystyrene 96-well
plates. Second, viral particles were immobilized on the plates
by treatment of the fetuin layer with a suspension of the virus.
(33) (a) Mammen, M.; Dahmann, G.; Whitesides, G. M. J. Med. Chem.
1995, 38, 4179-4190. (b) At the highest concentrations of polymers tested,
the ELISA signal decreases with increasing concentration. This behavior
could be caused by inaccessibility of biotin groups to the streptavidin-
enzyme complex at high surface coverage of polymer, or by an increase in
(32) McCormick, D. B.; Roth, J. A. Anal. Biochem. 1970, 34, 226-
236.

3794 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
Scheme 3. ELISA-Type Assay for Measuring the Binding of
a Biotin-Labeled Polymer-Bearing Sialic Acid Groups to the
Influenza Virusa
Figure 5. ELISA sandwich assay measuring the binding of unfrac-
tionated pA(co-P: Am, SA, BT) (b) and pA(pre-P: Am, SA, BT) ([)
to the immobilized influenza virus as a function of the concentration
of polymer-linked SA groups in solution during adsorption of the
polymer. The virus was immobilized on a fetuin-coated surface from
a suspension containing 50 µg of viral protein/mL. Adsorbed polymer
was quantified by binding a streptavidin-alkaline phosphate conjugate
and measured as the rate of enzymatic hydrolysis of p-nitrophenyl
phosphate (given as the rate of change of absorbance at 405 nm). PA-
(co-P: Am, BT), a polymer synthesized under the same copolymeri-
zation conditions as pA(co-P: Am, SA, BT) except for the omission
of the SA-labeled acrylamide monomer, was tested as a negative control
to verify the need for SA groups for binding (O). The concentrations
of the pA(co-P: Am, BT) solutions are given as the concentration of
polymer-linked SA groups present that would be present in equivalent
dilutions of pA(co-P: Am, SA, BT). The curves superimposed over
the data show nonlinear fits of the data to Langmuir isotherms as
described in the text (the fitted constants, KELISA and Vsat, are shown
with dashed lines). At the higher polymer concentrations there was a
reproducible decrease in rate with increasing [SA]; these data were
not included in the nonlinear fits.
a The cartoon shows schematically the important binding events
occurring during the assay.
polymer was determined by using a streptavidin-alkaline
phosphatase conjugate. Fifth, the rate of enzyme-catalyzed
hydrolysis of p-nitrophenyl phosphate was measured colori-
metrically with a kinetic microplate reader. Figure 5 shows
the binding curves as a function of the concentration in solution
of polymer-bound SA moieties. Although the interaction of
these polymers with the virus is probably complex mechanisti-
cally, the data can be adequately fit over most of the concentra-
tion range by assuming they follow a Langmuir isothermsa
one-step reaction with a single binding constant (eq 1).33b
[SA]
V ) Vsat
ELISA
(
)
K
+ [SA]
Where [SA] ) the concentration in solution of polymer-bound
SA groups during polymer binding; V ) the rate of p-nitrophenyl
phosphate hydrolysis due to bound streptavidin-enzyme; Vsat
) the rate at hydrolysis that results from saturation of the viral
surface with polymer; KELISA ) the effective dissociation
constant, calculated on a per SA basis (i.e. on the basis of the
total concentration of polymer-linked SA residues in solution).
The use of eq 1 to fit the ELISA binding isotherms requires
the assumption that the concentration of polymer in solution is
not significantly affected by the binding of polymer to im-
mobilized virus. As long as this assumption holds, the effective
dissociation constant that we measure, KELISA, should be
independent of the concentration of virus immobilized on the
surface of the 96-well plate. If this surface concentration of
virus is high enough to bind and remove a significant fraction
of the polymer from solution, the calculated value of KELISA
will be larger (weaker binding) than the actual value and will
increase with an increase in the surface concentration of virus.
To test the validity of the assumption, we measured the binding
isotherms of unfractionated pA(co-P: Am, SA, BT) and pA-
(pre-P: Am, SA, BT) on fetuin-coated plates that had been
treated with different concentrations of influenza virus. Figure
Figure 6. Dependence of binding constants, calculated for the binding
of SA-containing polymers to influenza virus by the ELISA assay, on
the amount of virus immobilized on the ELISA plates. The surface
concentration of immobilized virus was varied by treating fetuin-coated
ELISA plates with suspensions of virus containing different concentra-
tions of virus. For each concentration of virus used, the binding
constants KELISA and Vsat were determined for both pA(co-P: Am, SA,
BT) and pA(pre-P: Am, SA, BT) from binding isotherms as described
in the text and Figure 5.
6 shows the results of this study. The correlation of Vsat with
the concentration of virus shows that the number of viral
particles that adsorbed to the plates was dependent on the
concentration in solution of virus during the immobilization step.
Except for the highest surface concentration of virus, the KELISA
values determined for pA(co-P: Am, SA, BT) were independent
of the concentration of immobilized virus: for this polymer,
the effective binding constant KELISA ) 541 ( 112 nM. In
contrast, the values of KELISA determined for pA(pre-P: Am,
SA, BT) were strongly dependent on the concentration of
immobilized virus over the entire range of surface concentrations
that were tested. Thus, using this assay, we cannot accurately
the number of biotin groups bound per tetravalent streptavidin at high
concentrations of biotin groups on the surface. Another possibility is that
high concentrations of the SA-containing polymers are able to competitively
desorb immobilized virus from the fetuin-coated surface. Data from polymer
concentrations exhibiting this behavior were not included in the curve fits.
(34) The differences in inhibition constants are not due to the different
SA derivative and linker chains used in the pre-P and co-P polymers.
Polymers with the same linker and SA derivative as the pre-P polymer but
prepared by copolymerization behave similarly to the co-P polymer used
in this study (see ref 17).

Inhibition of Agglutination of Erythrocytes
J. Am. Chem. Soc., Vol. 118, No. 16, 1996 3795
Scheme 4. The Increase in the Mass of Adsorbed Polymer
with Polymer Concentration May Represent an Increase in
Occupancy of SA Binding Sites on the Virus (Top Arrow), a
Decrease in the Number of Attachment Sites per Polymer
Chain without a Change in Occupancy of SA Binding Sites
(Bottom Arrow), or a Combination of These Two
Mechanisms
instance, that the co-P class of polymers may not have a random
distribution of SA groups along the polymer backbone because
of differences in the relative rates of copolymerization of the
substituted and unsubstituted monomers.19 The uneven distribu-
tion of SA groups over the length of a polymer chain limits the
surface area of the virus (and therefore the number of SA
binding sites) that can interact with the polymer, and thereby
decreases the effective binding constant of the polymer. In
contrast, the preparation of the pre-P polymers is more likely
to give a random distribution of SA groups because the
distribution is not influenced by the relative rates with which
each functional group is incorporated into the polymer. The
distribution of SA groups in the pre-P polymers is controlled
by differences in the reactivities of the activated side chains
along the length of the polymer; these differences are probably
small because neutral polyacrylates in good solvents tend to
behave as random coils.35
No binding occurred, by the ELISA assay, if the polymers
did not present both SA and BT groups; this observation
demonstrated that no non-specific interactions were occurring.
The specificity of the binding interactions was confirmed by
transmission electron microscopy (TEM). A carbon film on a
copper grid was treated with fetuin, virus, and pA(pre-P: Am,
SA, BT) by the same procedure used to coat the ELISA plates.
The adsorbed polymer was then labeled with a streptavidin-
colloidal gold conjugate. Figure 7 is a TEM micrograph that
shows extensive labeling of the adsorbed polymer with the
colloidal gold tag. The colloidal gold is predominantly confined
to the surface of the viral particles; the polymer is therefore
also localized on the viral surface.
measure the effective binding constant, but we can place an
upper limit of KELISA < 3 nM. We were unable to decrease
the concentration of virus on the surface of the virus further
because the signal to noise of the assay became too low.
The effective dissociation constants (KELISA) that were
measured indicate that the polymers bound from 103 to >106
tighter, on a per SA group basis, than monomeric SA and its
simple analogs. This result is evidence for a strong enhancement
in the affinity of these polyvalent inhibitors through binding of
multiple binding elements. Lees et al. showed that ligands
attached to a polymer backbone may be less accessible sterically
to receptors than free ligands in solution.18 The enhancement
of binding due to polyvalency effect, therefore, may be even
greater than indicated. We note explicitly that the ELISA assay
gives no indication of the degree of occupancy of SA binding
sites on the virus. The mass of adsorbed polymer adsorbed on
virus may increase with polymer concentration for at least two
reasons. First, additional polymer chains may bind to unoc-
cupied sites on the virus. Second, it is possible to adsorb more
polymer chains per viral particle without changing the oc-
cupancy of the SA binding sites by making fewer attachments
per polymer chain. These two possibilities are illustrated in
Scheme 4. At very low concentrations of polymer on the
surface, the first mechanism will dominate. At higher concen-
trations on the surface, unfavorable interactions between polymer
chains may preclude the occupation of all SA binding sitessunder
these conditions, the adsorption of additional polymer chains
is likely to follow a combination of both mechanisms.
The values of KELISA that were measured were similar to the
corresponding values of KHi AI: inhibition of hemagglutination
requires concentrations of polymer on the viral surface that are
approximately half of the values at saturation. The 100-fold
difference in KHi AI for polymers prepared by copolymerization
compared to polymers prepared by modification of a preformed
polymer is completely accounted for by differences in the
effective dissociation constants of the polymers, as opposed to
differences in the abilities of the polymers, once bound to the
virus, to prevent the virus from binding the red blood cell
sterically. The reason for the large differences in dissociation
constants measured for polymers prepared by the two methods
is not clear but may reflect differences in the distribution of
SA groups along the polymer backbone.34 We believe, for
Narrow molecular weight fractions of the biotin-labeled
polymers were prepared exactly as described for the unlabeled
polymers, and the values of KHi AI showed the same dependence
on chain length. Figure 8a shows that for pA(co-P: Am, SA,
BT), KELISA was inversely dependent on chain length (slope of
the log-log plot ) -0.8), while Vsat did not vary with chain
length; fractions containing longer polymer chains (and, there-
fore, more SA groups per chain) had higher affinities, but the
total mass of polymer bound under conditions of saturation was
independent of chain length. Figure 8b shows that KHi AI and
KELISA had similar dependencies on chain length; the dependence
of inhibitory potency on chain length, therefore, is due to the
higher affinity of the longer chains. A similar analysis was
not possible for the pre-P inhibitor because the binding constants
were too tight to measure accurately by the ELISA assay (all
the fractions have a KELISA < 15 nM; data not shown).
One can develop a simple model for relating binding constants
to chain length by assuming that each additional SA on the
longer chains is capable of participating in an additional
interaction with HA on the viral surface, and that each additional
SA-HA interaction contributes roughly equally to the free
energy of binding. It follows from the exponential relationship
of equilibrium constants to free energy that the binding constant
should be an exponential function of the chain length. The
roughly linear (i.e. less than exponential) relationship that we
observed implies that, in the range of molecular weights we
analyzed, there must be a decreasing energetic contribution from
the binding of each additional SA group. The decrease in the
effectiveness of additional SA groups can be explained in two
ways: (i) a decrease in the accessibility of SA groups due to
the larger conformational entropy of the longer chains; and (ii)
geometric constraints that prevent some SA groups from
reaching unoccupied SA-binding sites.
(35) For a listing of the solution properties of various polyacrylic acid
derivatives in good solvents see: Kurata, M.; Stockmayer, W. H. AdV.
Polym. Sci. 1963, 3, 196-312.

3796 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
Figure 8. ELISA sandwich assay measuring the binding of narrow
molecular weight fractions of pA(co-P: Am, SA, BT) to immobilized
virus. (a) The constants KELISA and Vsat (obtained by fitting binding
isotherms to the Langmuir model described in the text) are plotted as
a function of molecular weight and chain length, N. (b) KELISA and
KHi AI are plotted for each polymer fraction. The molecular weight of
each fraction is listed next to the corresponding symbol on the plot.
Figure 7. Transmission electron microscope (TEM) pictures showing
the binding of pA(pre-P: Am, SA, BT) to virus under the conditions
used for the ELISA sandwich assay. Polymer bound to the virus was
labeled with streptavidin adsorbed on 5 nm diameter colloidal gold.
The adsorbed complexes were stained with a uranyl acetate negative
stain before TEM analysis. The picture shows a dense distribution of
collioidal gold label (seen as dark black dots) on and around the virus
particles. The arrow in the micrograph with the highest magnification
shows one group of colloidal gold labels. The scale bars indicate 100
nm.
Steric Stabilization of the Viral Surface by Adsorbed
Polymer: Inhibition of Antibody Binding by Adsorbed
Polymer. To determine whether adsorption of the polymeric
inhibitors causes steric stabilization of the viral surface, we
examined the effect of pA(pre-P: Am, SA, BT) on the binding
of a rabbit polyclonal antibody generated by immunizing a rabbit
with X-31 virus (R-X31). The polyclonal antibody should
include individual antibodies that bind to a variety of epitopes
over the entire surface of the virus. A significant inhibition of
the binding of R-X31 to the viral surface by the polymeric
inhibitor would indicate that the adsorbed polymer not only
blocks the SA binding site but also creates an excluded volume
around the virus that is large enough to interfere with binding
events removed from the SA binding site.
The format of the assay we used to measure the binding of
R-X31 to influenza virus was similar to the ELISA assay we
used for measuring the binding of the polymeric inhibitors. First,
we treated viral particles immobilized on fetuin-coated microtiter
plates with varying concentrations of a polymeric inhibitor. In
contrast to the assay for the binding of polymer (in which bound
polymer was quantified using a streptavidin-alkaline phosphate
conjugate), we instead treated the immobilized virus-polymer
complex sequentially with R-X31 followed by an alkaline
Figure 9. Inhibition of the binding of a polyclonal antibody (b) to
influenza virus by adsorbed pA(pre-P: Am, SA, BT) (O) as a function
of the concentration of polymer-linked SA groups in solution during
adsorption of the polymer. The virus was immobilized on a fetuin-
coated surface from a suspension containing 50 µg of viral protein/
mL. Polymer was adsorbed on the immobilized virus and quantified
by binding a streptavidin-alkaline phosphatase conjugate and measur-
ing the enzymatic activity as described in Figure 5. To measure the
binding of the polyclonal antibody, R-X31, the immobilized virus-
polymer complex was treated first with R-X31. Bound antibody was
quantified by binding a goat anti-rabbit Ig antibody linked to alkaline
phosphatase and measuring the bound alkaline phosphatase activity.
phophatase-labeled goat anti-rabbit Ig polyclonal antibody (to
quantify the amount of R-X31 that bound).
Figure 9 compares the relative mass of polymer adsorbed on
the virus (as determined by the polymer-binding assay) from
solutions containing different concentrations of pA(pre-P: Am,
SA, BT) with the relative mass of R-X31 that bound in the
presence of the adsorbed polymer (as determined by the
antibody-binding assay). The presence of adsorbed polymer
decreased the binding of R-X31 by as much as 64%. This
inhibition of R-X31 binding was not solely due to the blocking

Inhibition of Agglutination of Erythrocytes
J. Am. Chem. Soc., Vol. 118, No. 16, 1996 3797
of antibodies directed at the SA binding site; the presence of
the SA derivative 2 (Kd ) 3 µM) at a concentration of 300 µM
had no discernible effect on the binding of R-X31 (data not
shown). These results shows that pA(pre-P: Am, SA, BT) is
capable of steric stabilization of the viral surface. A threshold
concentration of polymer on the surface of the virus was
necessary for steric stabilization to occur; no inhibition was
observed until the concentration of polymer on the surface was
approximately half of the value at saturation. The similarity
of the threshold values of adsorbed polymer that were necessary
for both steric stabilization against antibody binding and
inhibition of hemagglutination (also approximately half of
saturation) suggests that steric stabilization may have a role in
the inhibition of hemagglutination.
We note that even in the presence of saturating concentrations
of polymer, the binding of R-X31 was not completely inhibited.
Although we have not investigated further this residual binding,
we propose two possible explanations: (i) incomplete coverage
of the virus by the polymer or (ii) penetration of the antibodies
through the polymer layer. In both cases, larger structures, such
as red blood cells, would be expected to be more strongly
influenced by the steric stabilization than would the relatively
small antibodies.
Measuring the Stochiometry of the Binding of the Poly-
meric Inhibitors to the Surface of Influenza Virus: Binding
Assays Using Fluorescein Labeled Polymers. To understand
better the requirements for both polyvalent binding of the
polymeric inhibitors to influenza virus and stabilization of the
viral surface, we needed a technique for measuring the amount
of polymer adsorbed onto the surface of the virus. A disad-
vantage of the ELISA assay for measuring the binding of
polymeric inhibitors to the virus is that the assay determines
only the relatiVe mass of bound polymer and, therefore, cannot
be used to calculate the stochiometry of binding. We developed
an alternative method for determining the absolute mass of
polymer bound to the virus. A fluorescein-labeled SA-contain-
ing polymer was bound to virus under conditions predicted to
lead to the maximum amount of polymer bound to the viral
surface (as determined by ELISA assay using a comparable
biotin-labeled polymer). Free and bound polymer were sepa-
rated by centrifugation of the virus. The amount of bound
polymer was determined by comparing the fluorescence of the
supernatant to a control solution that was not treated with virus.
The polymer used in these experimentsspA(pre-P: Am, SA,
F), where F stands for fluoresceinswas prepared by the same
procedure as pA(pre-P: Am, SA, BT) except the biotin-
containing amine 8 was replaced with the fluorescein-containing
amine 9 at a concentration calculated to give a polymer
containing a proportion of monomer units linked to fluorescein,
øF ) 0.005. Figure 10 shows that the fraction of a solution of
pA(pre-P; Am, SA, F) (present at a total concentration of
polymer-linked SA groups, [SA] ) 1 µM) that adsorbed on a
suspension of virus was linear with the concentration of virus
up to 16 µg of viral protein/mL.
Figure 10. Titration of a solution of the fluorescein-labeled polymer
pA(pre-P: Am, SA, F) with virus in suspension. The concentration
of polymer-linked SA groups in solution is [SA] ) 1.0 µM. The
polymer solutions were incubated with the virus for 30 min at 4 °C.
The bound polymer was then removed by centrifugation of the virus.
The percentage of polymer bound to virus was determined by
comparison of the fluorescence of the supernatant to a control run in
the absence of virus. The concentration of polymer chains adsorbed
on virus particles is given in terms of the concentration of polymer-
linked SA groups.
in one influenza virus has been determined (using a variety of
experimental techniques) to be roughly 2 × 108 g of protein/
mol of viral particles.36-38 Using this molecular weight, the
ratio of bound polymer to virus is given by the following
expression:
[polymer]bound/[virus] ) ([SA]bound (M)/[virus] (g))MWvirus
≈ (50 × 10-6 M/g)(2 × 108 g/mol)
≈ 10 000 polymer-linked SA groups per viral particle
The mole fraction of SA-substituted monomer units in the
polymer backbone is øSA ) 0.17. The average chain length
(Nav) of polymers derived from pNAS is equal to the average
molecular weight determined by GFC for acid-hydrolyzed pNAS
(Table 1) divided by the molecular weight of acrylic acid: Nav
) 70000 D/71 D ≈ 1000 monomer units per polymer chain.
Using these constants, the ratio of bound polymer to virus can
be rearranged to give the number of bound polymer chains per
viral particle:
[polymer]bound/[virus] )
(10 000 SA groups/virion)/(øSANav)
≈ 60 polymer chains per viral particle
The average radius of the viral particles is Rv ) 60 nm.37 The
average molecular weight of the polymer with the SA groups
attached is 140 kD. Using these values, a surface concentration
of polymer on the virus can be calculated:
[polymer]bound/areavirus
)
2
{60 chains × 140 kD/(6 × 1023)}/(4πRv ) ≈ 0.3 ng/mm2
These values are all semiquantitative but they provide a useful
picture of the interaction of the polymeric inhibitors with the
virus.
Discussion
From the stochiometry of binding, we can determine the
conditions required for inhibition of hemagglutination. Assum-
(36) Cusack, S.; Ruigrok, R. W. H.; Krygsman, P. C. J.; Mellema, J. E.
J. Mol. Biol. 1985, 186, 565-582.
(37) Ruigrok, R. W. H.; Andree, P. J.; Hooft Van Huysduynen, R. A.
M.; Mellema, J. E. J. Gen. Virol. 1984, 65, 799-802.
(38) Reimer, C. B.; Baker, R. S.; Newlin, J. E.; Havens, M. C. Science
1966, 152, 1379-1381.
The slope of this line (50 µmol of polymer-linked SA
groups/g of viral protein) is the ratio of bound polymer to virus.
The average molecular weight of the combined viral proteins

3798 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
ing that the inhibition of hemagglutination required a surface
concentration of polymer at least equal to half of the maximum
value (KiHAI ≈ KELISA), the minimum surface concentration on
the virus of pA(pre-P: Am, SA), øSA ) 0.17, required to prevent
hemagglutination was approximately 5000 polymer-linked SA
groups per viral particle or 0.15 ng/mm2. This value is
approximately the same concentration of polymer on the surface
that was required to cause significant inhibition of the binding
of a polyclonal antibody against the virus. The number of
polymer chains per viral particle that were required to inhibit
hemagglutination is remarkably small: For a polymer with a
chain length of 1000 monomer units (molecular weight ) 140
kD), only 30 polymer chains per viral particle were required
for inhibition. Since the weight of adsorbed polymer required
for inhibition was roughly independent of chain length, polymers
with molecular weights >1000 kD needed to bind <3 polymer
chains per viral particle to inhibit hemagglutination.
The measured stochiometry can be used to determine the
sensitivity limits of the HAI assay. The typical concentration
of virus used in the HAI assay is 50-100 µg of viral protein/L.
The required concentration of bound SA required to inhibit this
suspension (1-2 nM) sets a lower limit of 1-2 nM on the
values of KHi AI that can be measured using this assay. The
measured values of KHi AI for the pre-P class of polymeric
inhibitors approach this lower limit; the real values may be
lower.
The fact that adsorbed polymers inhibited the binding of a
polyclonal antibody to the viral surface is evidence that the
polymers were capable of steric stabilization of the viral
particles. Adsorbed or grafted polyacrylamide has been shown,
in other systems, to stabilize a variety of colloids in water.39
The required surface concentration of polyacrylamide required
for stabilization, however, has not been carefully measured in
a manner that allows for comparison between systems. Some
data are available for other polymers. The stabilization of AgI
sol in water by adsorbed poly(vinyl alcohol) requires a minimum
surface concentration of polymer of 0.4 ng/mm2.40 More
typically, stabilized colloids are prepared under conditions that
result in relatively high concentrations of polymer on the surface
(> 1 ng/mm2).41 The surface concentrations we observed for
our inhibitors (∼ 0.3 ng/mm2) were, therefore, slightly low
compared to these other systems, but close enough to previously
observed values, albeit in systems using different polymers, to
be consistent with the hypothesis that steric stabilization of the
viral surface was important in the phenomenon we were
examining (especially considering the semiquantitative nature
of our calculations of stochiometry).42 The relatively low
amounts of polymer that adsorbed on the viral surface may
explain why we observed only partial inhibition of the binding
of antibodies against influenza virus.
then steric stabilization of the virus by the polymer may be
irrelevant. Despite the large excess of polymer-linked SA
groups on the surface, however, the percentage of SA-binding
sites that are occupied is probably low due to geometric
constraints: the average distance between adjacent SA groups
on the polymer chain (∼30 Å for the pre-P polymers) is too
small to bridge the HA binding pockets (the intramolecular
distance between SA binding sites on an HA trimer is 46 Å,
the range is distances between HA trimers on the fluid viral
surface is 50-115 Å),18 and the binding of some SA groups
may be prevented by inter- or intra-chain interactions. A direct
measure of the occupancy of SA-binding sites would clarify
further the importance of steric stabilization; there is, however,
no good assay presently available for measuring this value.
Conclusions
SA-containing polyacrylamides are significantly better than
monovalent SA at preventing the agglutination of red blood cells
by influenza virus. Using an ELISA assay for the binding of
polymer to virus, we showed that the differences in inhibition
constants (KHi AI) between monomeric and polymeric inhibitors,
measured on the basis of total SA groups in solution, correlated
with differences in the binding affinities of the inhibitors for
the viral surface. Similarly, we showed that differences in the
affinity of binding also explained the differences in inhibitory
efficacy of polymeric inhibitors prepared by two different
methods (pre-P vs co-P) as well as the differences in inhibitory
efficacy of polymer fractions differing in molecular weight.
These results confirm the importance of enhanced binding
through polyvalent interactions in the efficacy of the polymeric
inhibitors. We expect that the synthesis of polymers bearing
other weak-binding ligands will provide a general route to high-
affinity polyvalent ligands for a variety of receptors on the
surfaces of viruses or cells.
We note that large enhancements in binding and inhibitory
efficacy were observed even for fractions of polymers with
relatively low molecular weights. For polymers prepared by
the pre-P method, a fraction containing just 150 monomer units
(corresponding to ∼30 SA groups) both bound and inhibited at
concentrations of polymer-bound SA groups <15 nM (>5
orders of magnitude better than the monomer). The fact that a
relatively small number of SA groups is adequate to provide
such large enhancements in binding is important in suggesting
the design of new tight-binding inhibitors. In principle, it may
be possible to construct polyvalent inhibitors with better
controlled structure than the polyacrylamide inhibitorssperhaps
based on dendrimers or modified cyclodextrinsswith similar
or better binding affinities. The detailed geometrical require-
ments imposed on the SA groups to achieve large binding
enhancements are not, however, currently understood.
The surface of each viral particle contains on average 500
HA trimers (i.e. 1500 SA-binding sites).37,43 At a surface
concentration of 5000 polymer-linked SA groups per viral
particle, there are sufficient SA groups available to block every
SA-binding site. If all SA binding sites are, indeed, occupied,
While enhanced affinity through polyvalent binding is clearly
important in the high potency of the polymeric inhibitors for
inhibiting hemagglutination by influenza virus, we as yet do
not have a complete picture of the mechanism of inhibition.
The fact that the polymers inhibit the binding of antibodies to
the surface of the virus is strong indirect evidence that steric
stabilization plays a role in the inhibition of virus-cell
interactions. Other recent studies have provided additional
indirect evidence for steric stabilization: (i) structure-activity
relationships determined for polymeric inhibitors with modified
side chains are consistent with a mechanism involving steric
stabilization,19 and (ii) high-affinity monomeric inhibitors of
neuraminidase (NA), a hydrolytic enzyme present on the surface
of influenza virus, enhance the efficacy of SA-containing
(39) Evans, R.; Davison, J. B.; Napper, D. H. Polym. Lett. 1972, 10,
449-453. Furusawa, K.; Tezuka, Y.; Watanabe, N. J. Colloid Interface
Sci. 1980, 73, 21-26. Nowicki, W.; Nowicka, G. Materials Sci. Forum
1988, 25-26, 497-500. Tsubokawa, N.; Maruyama, K.; Sone, Y.;
Shimomura, M. Polym. J. 1989, 21, 475-481.
(40) Fleer, G. J.; Koopal, L. K.; Lyklema, J. Kolloid-Z. Z. Polym. 1972,
250, 689-702.
(41) Klein, J.; Luckham, P. Nature 1982, 300, 429-431.
(42) Theoretical calculations predict that steric stabilization may be
possible at surface concentrations of polymer as low as 0.2 ng/mL, but
comparisons to experimental systems are not straightforward. See: Hes-
selink, F. Th.; Vrij, A.; Overbeek, J. Th. G. J. Phys. Chem. 1971, 75, 2094.
(43) Skehel, J. J.; Schild, G. C. Virology 1971, 44, 396-408.

Inhibition of Agglutination of Erythrocytes
J. Am. Chem. Soc., Vol. 118, No. 16, 1996 3799
polymers in inhibiting hemagglutination.44 This second obser-
vation, which is especially pronounced in polymers containing
high proportions of SA, suggests that SA groups on the polymers
bind the NA active site as well as the HA-binding pocket. The
enhancement of inhibition by the polymers in the presence of
NA inhibitors is probably due to expansion of the adsorbed
polymer layer after the competitive release of SA groups from
the NA active sites.44 Proof for steric stabilization as an
important factor would require showing that polymeric inhibitors
can prevent hemagglutination without occupying a large per-
centage of the SA-binding sites on the protein. We are currently
investigating this issue by testing the inhibitory efficacy of
polymeric inhibitors targeted specifically at NA, a protein that
does not play a role in virus-cell attachment.
7.83 (t, 1H), 6.77 (t, 1H), 6.42 (s, 1H), 6.36 (s, 1H), 4.29 (dd, 1H),
4.12 (t, 1H), 3.48 (s, 4H), 3.36 (q, 4H), 3.16 (q, 2H), 3.08 (m, 3H),
2.80 (dd, 1H), 2.55 (d, 1H), 2.05 (t, 2H), 1.59 (m, 1H), 1.46 (m, 3H),
1.36 (s, 9H), 1.28 (m, 2H). HRMS-FAB [M + H]+ calcd for
C21H38N4O6S 475.259, found 475.257.
N-Biotinyl-3,6-dioxaoctane-1,9-diamine (8). The Boc-protected
amine 10 was dissolved in 10 mL of trifluoracetic acid (TFA) per gram
of protected amine and stirred for 20 min at room temperature. The
TFA was evaporated under reduced pressure to give an oil. The oil
was dissolved in water and reconcentrated to remove residual TFA
leaving the product as the TFA salt. The oil was used without further
purification. 1H-NMR δ (D2O): 4.31 (dd, 1H), 4.11 (dd, 1H), 3.45 (t,
2H), 3.39 (s, 4H), 3.32 (t, 2H), 3.09 (t, 1H), 3.02 (m, 1H), 2.90 (t,
2H), 2.70 (dd, 1H), 2.47 (d, 1H), 1.97 (t, 2H), 1.3-1.5 (m, 4H), 1.11
(m, 2H).
N-Acryloyl-N′-biotinyl-3,6-dioxaoctane-1,9-diamine (7). The amine
8, prepared as described above from 0.71 g (1.5 mmol) of the BOC-
protected amine 10, was dissolved in 50 mL of 250 mM NaHCO3.
N-Acryloylsuccinimide46 was added dropwise with stirring in 10 mL
of dimethoxyethane. The solution was stirred for 3 h at 4 °C in the
dark. Unreacted starting materials and buffer salts were removed by
adding 40 g of Amberlite mixed bed ion-exchange resin. The
supernatant was lyophilized to give a crude product, which was then
applied to a silica column and eluted with 4:1 methylene chloride/
methanol to give 0.34 g of the product as a white solid, yield 52%, mp
123-125 °C. 1H-NMR δ (DMSO-d6): 8.20 (t, 1H), 7.85 (t, 1H), 6.43
(s, 1H), 6.37 (s, 1H), 6.23 (dd, 1H), 6.06 (d, 1H), 5.55 (d, 1H), 4.28
(dd, 1H), 4.11 (m, 1H), 3.49 (s, 4H), 3.43 (t, 2H), 3.37 (t, 2H), 3.27
(q, 2H), 3.16 (q, 2H), 3.07 (m, 1H), 2.80 (dd, 1H), 2.56 (d, 1H), 2.05
(t, 2H), 1.59 (m, 1H), 1.47 (m, 3H), 1.28 (m, 2H). HRMS-FAB [M +
H]+ calcd for C19H32N4O5S 429.217, found 429.215.
Synthesis of SA-Containing Polymers by Copolymerization. The
copolymer pA(co-P: Am, SA) was prepared as described previously.18
Stock solutions of acrylamide (2.2 M), SA-linked acrylamide 4 (0.22
M), and azobis(cyanovaleric acid) (100 mM adjusted to pH 7.2 with
10 N NaOH) in deoxygenated water were combined to give a solution
containing 1 M acrylamide, 0.1 M 4, and 10 mM azobis(cyanovaleric
acid). Irradiation with a long-wave UV lamp (Spectronics Corp.) gave
a polymer assumed to contain a 10:1 ratio of unsubstituted acrylamide
units to sialic acid linked units (øSA ) 0.09). The copolymer pA(co-
P: Am, SA, BT) was prepared by the same method except that the
polymerization solution also contained 10 mM biotin-linked acrylamide
7, and resulted in a polymer assumed to contain a ratio of 100:10:1 of
unsubstituted acrylamide units, sialic acid-linked acrylamide units, and
biotin-linked acrylamide units, respectively (øSA ) 0.09, øBT ) 0.01).
Synthesis of SA-Containing Polymers by Modification of a
Prepolymerized Chain.19 A solution of SA-containing amine 6 (243
µmol) in triethylamine (TEA, 0.5 mL) was added to a stirred solution
of pNAS (6.78 g, 1.22 mmol NHS ester) in dimethylformamide (DMF,
20 mL). The solution was stirred at room temperature for 20 h, heated
at 65 °C for 6 h, then stirred at room temperature for an additional 48
h. This procedure yielded a stock solution of preactivated polymer
with øSA ≈ 0.20 (the stock solution contained 2.0 µmol SA/mL and
8.0 µmol NHS/mL).
Experimental Section
Materials. The SA-linked acrylamide monomer 4 was prepared
according to Lees et al.18 The amine 6 and pNAS were prepared
according to Mammen et al.19 The amine-containing derivative of
fluorescein 9 was purchased from Molecular Probes, chicken red blood
cells from 2 week old chicks were purchased from Spafas, and fetuin
(from fetal calf serum) was purchased from Sigma. The alkaline
phophatase-linked streptavidin and goat anti-rabbit Ig conjugates used
in the ELISA sandwich assay were purchased as solutions from Fisher
Scientific. Streptavidin adsorbed on colloidal gold (particle diameter
) 5 nm) was purchased from Sigma. Influenza virus X-31 and the
rabbit polyclonal antibody against X-31 (R-X31) were obtained from
Prof. J. J. Skehel (National Institute for Medical Research, London).
Phosphate buffered saline (PBS) is 137 mM NaCl, 3 mM KCl, 8 mM
Na2HPO4, and 2 mM KH2PO4, adjusted to pH 7.2. BSA-PBS is 1.0
mg/mL bovine serum albumin in PBS. Polymer-linked BT groups and
O-glycosides of SA were assayed by colorimetric assays.30,32 Hemag-
glutination inhibition (HAI) assays were conducted as previously
described.18,20 Flash chromatography was performed using silica gel
60 (230-400 mesh, E. Merck). Melting points are reported without
correction.
N-Boc-N′-biotinyl-3,6-dioxaoctane-1,8-diamine (10). Tris(ethylene
glycol)-1,8-diamine (6.6 g, 45 mmol) and 12 N HCl (4.6 mL, 55 mmol)
were added to 60 mL of 1:1 water/dioxane and cooled to 0 °C. While
the solution was stirred, 2-(tert-butoxycarbonyloxyimino)-2-phenylac-
etonitrile (BOC-ON, 3.7 g, 15 mmol) was added and the solution was
allowed to warm to room temperature over 90 min. The solution was
concentrated to an oil, suspended in 30 mL of water, and acidified to
pH 4.0 with 6 N HCl. The aqueous solution was washed 3 times with
30-mL portions of ethyl acetate, adjusted to pH 11.0 with 10 N NaOH,
and then extracted 5 times with ethyl acetate. The combined ethyl
acetate fractions were dried over magnesium sulfate, then evaporated
under reduced pressure to give 2.0 g (55%) of the crude mono-Boc
protected diamine.
The crude mono-Boc protected diamine (1.0 g, 4.1 mmol) and 4.1
mL of 1 N HCl were combined in a mixture of 40 mL of 250 mM
NaHCO3 and 40 mL of dimethoxyethane. The N-hydroxysuccinimide
ester of biotin45 (1.32 g, 3.9 mmol) was added as a solid and the solution
was stirred overnight at room temperature. Water (100 mL) was added
to the solution and the product extracted six times into 100-mL portions
of ethyl acetate. The combined extracts were washed with a saturated
solution of NaCl, dried over magnesium sulfate, and concentrated to a
white solid. Purification by flash chromatography using 9:1 methylene
chloride/methanol as the eluent gave 1.3 g (71%) of the product as a
white solid, yield 71%, mp 101.5-102.5 °C. 1H-NMR δ (DMSO-d6):
(i) The polymer pA(pre-P: Am, SA) was prepared by adding the
stock solution (600 µL, 4.8 µmol NHS) dropwise to NH4OH (concen-
trated aqueous, 1.5 mL), then stirring at room temperature for 12 h
and dialyzing against PBS.
(ii) The polymers pA(pre-P: Am, SA, BT) (øBT ) 0.01) and pA-
(pre-P: AM, SA, F) (øF ) 0.005) were prepared by combining the
stock solution (115 µL, 1 µmol NHS) with the biotin-containing amine,
8 (0.01 µmol in 25 µL of DMF), or the fluorescein-containing amine,
9 (0.005 µmol in 25 µL of DMF) and heating for 6 h at 65 °C. The
solutions were added dropwise to NH4OH (concentrated aqueous, 1.5
mL), stirred at room temperature for 12 h, and dialyzed against PBS.
Gel-Filtration Chromatography. For determination of the mo-
lecular weight distribution of acrylamide polymers by GFC, the polymer
was first hydrolyzed in acid to acrylic acid. This procedure results in
(44) (a) Choi, S.-K.; Mammen, M.; Whitesides, G. M. In press (b) All
polymers made using C-glycosides gave similar values of KHi AI to those
made using O-glycosides when assayed at 4 °C. At room temperature,
however (19 °C), we see evidence that the hydrolysis of the O-glycosidic
linkages becomes more significant: for O-glycosides, the values KiHAI
measured at 19 °C were 10 times higher (worse) than that measured at 4
°C. We conclude that while the hydrolytic activity of NA may be significant
at room temperature, it is not significant at 4 °C.
the hydrolysis of >95% of the amide groups in polyacrylamide.27
A
(45) Becker, J. M.; Wilchek, M.; Katchalski, E. Proc. Natl. Acad. Sci.
U. S. A. 1971, 68, 2604-2607.
(46) Pollack, A.; Blumenfeld, H.; Wax, M.; Baughn, R. L.; Whitesides,
G. M. J. Am. Chem. Soc. 1980, 102, 6324-6336.

3800 J. Am. Chem. Soc., Vol. 118, No. 16, 1996
Sigal et al.
solution was prepared containing the polymer at a concentration of 1
mg/mL in 300 µL of 10% (w/v) thioglycolic acid/6 N HCl (in the
absence of thioglycolic acid, the hydrolysis of sugar-containing
polymers resulted in the formation of an insoluble tar). The solution
was sealed in a glass tube under nitrogen and heated to 105 °C for 24
h. Initially a precipitate forms due to formation of cyclic imides. All
the material, however, redissolves as the reaction goes to completion.
The hydrolysis mixture was quenched by the addition of 300 µL of 5
N NaOH, 200 µL of 1 M Na2HPO4, and 700 µL of water. The resulting
solution was dialyzed extensively against 10 mM sodium phosphate,
150 mM sodium sulfate, pH 7.2.
GFC analysis was conducted using a 7.8 × 300 mm column
containing a linear gradient of different pore size polyacrylamide beads
(Ultrahydrogel Linear, Waters Chromatography). The flow rate of
eluent (10 mM sodium phosphate, 150 mM sodium sulfate, pH 7.2)
was 0.5 mL/min. The elution of polymer samples (20 µL sample
injections) was monitored by the UV adsorption at 210 nm. The column
was calibrated using the peak retention times (Mp ∼ (MnMw)0.5) of
polyacrylamide molecular weight standards (Polysciences).28 Molecular
weight distributions were calculated from the GFC traces.28
The preparation of narrow molecular fractions of the substituted
polymers was carried out by GFC using similar conditions except that
smaller columns with narrower distributions of pore sizes (GPC4000
or GPC300, Alltech Chromatography) were used to maximize through-
put and minimize dilution. Unfractionated polymers (20 µL, ∼2 mg/
mL) were injected and 10 × 200 µL fractions were collected during
elution of the polymer. For each sample, the GFC separation was
repeated several times to obtain reasonable amounts of fractionated
polymer.
ELISA Sandwich Assay for Measuring the Relative Binding of
Biotin-Labeled, SA-Containing Polymers to Influenza Virus. The
wells of a 96-well ELISA plate (Pro-Bind, Falcon) were coated with
fetuin by treatment with 50 µL of a solution containing 100 µg/mL of
fetuin in PBS for 2 h at 4 °C. The wells were washed three times with
200-µL portions of BSA-PBS. The fetuin-coated wells were then
consecutively treated with 100 µL of a suspension of virus in BSA-
PBS for 1 h, 100 µL of the polymer sample in BSA-PBS for 30 min,
and 100 µL of a solution containing the streptavidin-alkaline phos-
phatase conjugate (prepared by a 1:1000 dilution of the stock solution
in BSA-PBS) for 30 min. Each well was washed 3 times with BSA-
PBS between each binding reaction and 3 times with PBS after binding
of the streptavidin-alkaline phosphatase conjugate. Alkaline phos-
phatase substrate (100 µL of a solution containing 1 mg/mL of
p-nitrophenyl phosphate, 5 mM MgCl2, 1.6 M diethanolamine, adjusted
to pH 9.8 with 6N HCl) was added to each well and the rate of
hydrolysis of p-nitrophenyl phosphate measured using a kinetic
microplate reader (Biorad) with a 405-nm filter.
drop containing a solution of fetuin in PBS on the surface of the film
for 2 h. The surface was washed with 2 drops of BSA-PBS then treated
sequentially with a drop of each of the following solutions: influenza
virus (50 µg of protein/mL in BSA-PBS, 1 h), pA(pre-P: Am, SA,
BT) ([SA] ) 1 µM in BSA-PBS, 30 min), and streptavidin-colloidal
gold ([streptavidin] ) 50 µg/mL, 30 min). The surface was washed
with 2 drops of BSA-PBS after each binding reaction and with 2 drops
of water prior to staining. The adsorbed complex was negatively stained
with uranyl acetate: a drop of an aqueous solution of uranyl acetate
(25 mM) was placed on the surface of the substrate; most of the drop
was removed by absorption onto a piece of filter paper; the remaining
thin film of liquid was allowed to evaporate. The sample was analyzed
on a Philips EM420 TEM using an accelerating voltage of 120 kV.
ELISA Sandwich Assay for Measuring the Inhibition by SA-
Containing Polymers of the Binding of Antibodies to Influenza
Virus. Solutions containing pA(pre-P: Am, SA, BT) were adsorbed
onto viral particles immobilized on 96-well microtiter plates as described
in the procedure for the ELISA assay for the binding of the polymeric
inhibitors to influenza virus. Without removing the solution of the
polymeric inhibitor, serum from a rabbit immunized against the X-31
virus (100 µL of a 1:10000 dilution in BSA-PBS) was added and
allowed to bind for 10 min. The wells were washed with 3 × 200 µL
portions of BSA-PBS and then treated for 30 min with a goat anti-
rabbit Ig antibody linked to alkaline phosphatase (100 µL of a 1:2000
dilution of the stock solution in BSA-PBS). The wells were washed
with PBS and the alkaline phosphatase measured as described for the
previous ELISA assay.
Fluorescence Assay for Measuring the Binding of Fluorescein-
Labeled, SA-Containing Polymers to Influenza Virus. Solutions
containing pAA(co-P; Am, SA, F) and virus in PBS were incubated
30 min at 4 °C. Adsorbed and free polymer were then seperated by
centrifugation of the virus at 100000 × g (Airfuge Centrifuge,
Beckman). For each polymer concentration tested, a control was also
run in the absence of virus. The concentration of free polymer was
determined by comparison of the fluorescence intensity (Ex ) 489 nm,
Em ) 515 nm) of the supernatant of the test sample with that of the
control.
Acknowledgment. This work was supported by NIH Grant
No. GM30367. We thank Prof. J. J. Skehel for kindly providing
us with Influenza virus X-31. G. D. was supported by
Forschungsstipendium of the Deutsche Forschungsgemeinschaft
(DFG) and M. M. by an Eli Lilly predoctoral fellowship (1994).
The NMR facilities at Harvard were supported by NIH Grant
No. 1-S10-RR04870-01 and NSF Grant No. CHE88-14019.
Transmission Electron Microscopy (TEM). A 50 Å thick carbon
film supported on a copper grid was coated with fetuin by placing a
JA953729U