A non-chromatographic method for the purification of a bivalently active monoclonal IgG antibody from biological fluids

Article abstract of DOI:10.1021/ja9023836

This paper describes a method for the purification of monoclonal antibodies (rat anti-2,4-dinitrophenyl IgG: IgG^DNP; and mouse antidigoxin IgG: IgG^Dgn) from ascites fluid. This procedure (for IgG^DNP) has three steps: (i) p

Full text of DOI:10.1021/ja9023836

Published on Web 06/17/2009
A Non-Chromatographic Method for the Purification of a
Bivalently Active Monoclonal IgG Antibody from
Biological Fluids
Bas¸ar Bilgic¸er, Samuel W. Thomas III, Bryan F. Shaw, George K. Kaufman,
Vijay M. Krishnamurthy, Lara A. Estroff, Jerry Yang, and George M. Whitesides*
Department of Chemistry and Chemical Biology, HarVard UniVersity, 12 Oxford Street,
Cambridge, Massachusetts 02138
Received March 25, 2009; E-mail: gwhitesides@gmwgroup.harvard.edu
Abstract: This paper describes a method for the purification of monoclonal antibodies (rat anti-2,4-
dinitrophenyl IgG: IgGDNP; and mouse antidigoxin IgG: IgGDgn) from ascites fluid. This procedure (for IgGDNP
)
has three steps: (i) precipitation of proteins heavier than immunoglobulins with ammonium sulfate; (ii)
formation of cyclic complexes of IgGDNP by causing it to bind to synthetic multivalent haptens containing
multiple DNP groups; (iii) selective precipitation of these dimers, trimers, and higher oligomers of the target
antibody, followed by regeneration of the free antibody. This procedure separates the targeted antibody
from a mixture of antibodies, as well as from other proteins and globulins in a biological fluid. This method
is applicable to antibodies with a wide range of monovalent binding constants (0.1 µM to 0.1 nM). The
multivalent ligands we used (derivatives of DNP and digoxin) isolated IgGDNP and IgGDgn from ascites fluid
in yields of >80% and with >95% purity. This technique has two advantages over conventional
chromatographic methods for purifying antibodies: (i) it is selective for antibodies with two active Fab binding
sites (both sites are required to form the cyclic complexes) over antibodies with one or zero active Fab
binding sites; (ii) it does not require chromatographic separation. It has the disadvantage that the structure
of the hapten must be compatible with the synthesis of bi- and/or trivalent analogues.
Introduction
Monoclonal antibodies are important in biomedical research,
and antibody-based therapies have become increasingly impor-
tant in the past decade:1-7 55% of the drugs that are under
development currently are mAbs.8 The pharmaceutical industry
is considering a target for production of mAbs of 10 tons per
year.9 Bioreactors with a capacity from 15 000 to 25 000 L are
becoming more common, and an expression rate of 5 g of
antibody/L is standard.9 Purification of mAbssboth at process
scales, and for researchsremains an expense and inconvenience.
The current processes for purifying mAbs use multiple steps
that may have detrimental effects on the specific activity of
isolated mAbs. Methods for purification that are faster, less
expensive, more convenient at research scales, and result in
greater yields (of specifically active product) would be useful.10-12
Immunoglobulin (e.g., IgD, IgE, and IgG antibodies) consists
of two antigen recognition sites and is bivalent (Figure 1a). The
bivalency of IgG increases its avidity for antigens displayed on
This paper describes a non-chromatographic procedure for
purifying monoclonal IgG antibodies (mAbs) from a biological
fluid. This procedure is based on selective precipitation of cyclic
complexes of the targeted antibody and multivalent haptens with
ammonium sulfate (AMS) from a biological fluid (e.g., acites
fluid or a cell lysate). Because the cyclic oligomers of [IgG]n
(n ) 2, 3) have molecular weights that are two or three times
that of the monomeric IgG (150 kDa), the complexes precipitate
at lower concentrations of AMS than does monomeric IgG. The
key step in this procedure is the precipitation that separates the
oligomeric [IgG]n complexes from other monomeric antibodies
that do not form complexes (including IgGs that are not
bivalently active), and from other proteins in the ascites fluid.
We used two commercial IgGs (rat anti-2,4-dinitrophenyl,
IgGDNP and mouse antidigoxin, IgGDgn) as model systems in
developing this method. To the best of our knowledge, this
procedure is the first for purifying monoclonal IgGs that selects
for biValently active antibodies. We have successfully precipi-
(1) Brekke, O. H.; Sandlie, I. Nat. ReV. Drug DiscoV. 2003, 2, 52–62.
(2) Drews, J. Science 2000, 287, 1960–1964.
(3) Safarik, I.; Safarikova, M. J. Chromatogr. B 1999, 722, 33–53.
(4) Hoogenboom, H. R.; de Bruine, A. P.; Hufton, S. E.; Hoet, R. M.;
Arends, J. W.; Roovers, R. C. Immunotechnology 1998, 4, 1–20.
(5) Perico, N.; Remuzzi, G. Drugs 1997, 54, 533–570.
(6) Casadevall, A.; Scharff, M. D. Clin. Infect. Dis. 1995, 21, 150–161.
(7) Minunni, M.; Mascini, M. Anal. Lett. 1993, 26, 1441–146.
(8) Lerner, R. A. Angew. Chem., Int. Ed. 2006, 45, 8106–8125.
(9) Kelley, B. Biotechnol. Prog. 2007, 23, 995–1008.
tated complexes of IgGDNP using bi- and trivalent DNP (2,4-
affinity
dinitrophenyl) haptens (Kd
4-NP (4-nitrophenyl, Kd
≈ 0.8 nM), and bi- and trivalent
affinity
≈ 0.5 µM), and complexes of
IgGDgn using bivalent digoxin (Kd
≈ 0.1 nM). We believe,
affinity
based on these results, it should be possible to apply this
technique to antibody-ligand systems that have monovalent
dissociation constants ranging from micro- to nanomolar
(provided that the bi- and/or trivalent analogues of these haptens
are synthetically accessible).
(10) Farid, S. S. J. Chromatogr. B 2007, 848, 8–18.
(11) Farid, S. S. Biopharm. Int. 2008, 37–42.
(12) Shukla, A. A.; Hubbard, B.; Tressel, T.; Guhan, S.; Low, D.
J. Chromatogr. B 2007, 848, 28–39.
9
10.1021/ja9023836 CCC: $40.75  2009 American Chemical Society
J. AM. CHEM. SOC. 2009, 131, 9361–9367 9361

A R T I C L E S
Bilgiçer et al.
Figure 1. (a) Crystal structure of an antibody with the dimensions labeled (PDB: 1HZH). The flexibility of the hinge region gives rise to a range of values
for Θ and x (distance between binding sites). The complexes of antibodies (IgG) that can be formed by incubation with (b) trivalent hapten include bicyclic
antibody trimer, tricyclic antibody hexamer, and branched polymer; (c) bivalent hapten include cyclic dimer, cyclic trimer, and linear polymer.
cell surfaces and also allows it to form high-molecular-weight
complexes with soluble, multivalent antigens and allergens.13
During purification, several processes may yield monoclonal
antibodies with only one active Fab binding site; examples
include (i) protein unfolding, misfolding, and aggregation; (ii)
covalent modification (e.g., oxidation); (iii) enzymatic proteoly-
sis; and (iv) the “scrambling” of light chains.12 The biological
significance of bivalently active (as opposed to monovalently
active) immunoglobulins is presently unclear because there has
been no consistent route for preparing either.14-16
Procedures for purifying antibodies must remove a number
of contaminants that are associated with their expression, such
as host cell proteins, DNA, endotoxins, and cell culture media
additives. In addition, antibody-derived impurities, such as high-
molecular-weight aggregates and proteolytic fragments of
immunoglobulin, can also contaminate the desired product.
Current procedures for purifying therapeutic antibodies typically
rely on protein A (proA) chromatography (proA binds to the
Fc domain of IgGs), where the elution of the antibody is
achieved by decreasing the pH to 2-3.9 Although proA affinity
chromatography can yield products with 95% purity after a
single chromatographic step,17 it introduces a number of
additional challenges and new routes for contamination: hy-
drolysis and proteolysis can contaminate the isolated product
with cleaved proA and its truncated derivatives, and leached
proA can adhere to the eluting product.18-20 Furthermore, the
acidic pH of the mobile phase can cause the mAb to unfold
and lose activity, and/or aggregate nonspecifically and precipitate.
Because chromatographic procedures are labor-intensive,
expensive, and operationally demanding at large scales, there
is substantial effort directed toward developing new methods
of purification of mAbs. The Cohn fractionation process, which
currently provides yields of 80 tons per year for IgIV purifica-
tion, is a potential candidate for the ultimate goal for producing
10 tons per year of mAbs. The Cohn fractionation process
employs selective precipitation steps through control of pH,
temperature, concentration of ethanol, and ionic strength but
does not employ any chromatographic steps;21,22 additional unit
operations include microfiltration, ultrafiltration, and centrifuga-
tion. Nevertheless, the Cohn process is not yet a viable method
for the production of mAbs, and neither the Cohn process nor
the chromatographic techniques under development explicitly
select for bivalently active IgG.
The technique we introduce here is based on the formation
of discrete, cyclic complexes of antibodies, and therefore can
avoid many of these disadvantages when the appropriate
oligovalent ligand can be prepared. This procedure is an affinity-
based method, with the strengths and weaknesses of such
methods. In particular, it is specific for a hapten but requires
that that hapten be accessible and amenable to synthetic
manipulation. Importantly, this procedure differs from conven-
tional techniques in that the formation of the complexes requires
both Fab sites of an antibody to have binding activity.
(13) Bilgic¸er, B.; Moustakas, D. T.; Whitesides, G. M. J. Am. Chem. Soc.
2007, 129, 3722–3728.
Early in the development of molecular immunology, Pecht,
Baird, Posner, and others described the formation of discrete,
(14) Hlavacek, W. S.; Posner, R. G.; Perelson, A. S. Biophys. J. 1999, 76,
3031–3043.
(15) Posner, R. G.; Erickson, J. W.; Holowka, D.; Baird, B.; Goldstein, B.
Biochemistry 1991, 30, 2348–2356.
(19) Fahrner, R. L.; Whitney, D. H.; Vanderlaan, M.; Blank, G. S.
Biotechnol. Appl. Biochem. 1999, 30, 121–128.
(16) Erickson, J. W.; Posner, R. G.; Goldstein, B.; Holowka, D.; Baird, B.
Biochemistry 1991, 30, 2357–2363.
(20) O’Leary, R. M.; Feuerhelm, D.; Peers, D.; Xu, Y.; Blank, G. S.
Biopharm., Appl. Technol. Biopharm. DeV. 2001, 14, 10.
(21) Martin, T. D. Int. Immunopharmacol. 2006, 6, 517–522.
(22) Johnston, A.; Adcock, W. Biotechnol. Genet. Eng. ReV. 2000, 17, 37–
70.
(17) Gagnon, B. Purification Tools for Monoclonal Antibodies; Validated
Biosystems: Tucson, AZ, 2007.
(18) Fahrner, R. L.; Iyer, H. V.; Blank, G. S. Bioprocess Eng. 1999, 21,
287–292.
9
9362 J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009

Purification of a Monoclonal IgG Antibody
A R T I C L E S
cyclic dimers and trimers resulting from the interaction of IgEs
and IgGs with bivalent haptens (Figure 1c).23-28 On the basis
of analytical modeling of the assembly of antibody complexes,
Dembo and Goldstein predicted that the concentration of
bivalent hapten (Ctotal) at which maximum conversion (CTmax
)
took place would depend on the monovalent dissociation
constant (Kdaffinity) and the total concentration of antibody
([IgG]total) according to eq 1.28 Hence, in order to achieve
maximum conversion to cyclic complexes, the dissociation
constant (Kdaffinity) should be lower than the concentration of
antibody, and the ratio of the bivalent ligand to antibody should
be 1.
CTmax ) Kdaffinity/2 + [IgG]total
(1)
Results
Summary of Purification. Using this purification procedure,
we isolated pure, bivalently active anti-2,4-DNP and antidigoxin
from rat and mouse ascites fluids; ascites fluid and the
supernatant from hybridoma bioreactors are the two most
common biological sources for monoclonal antibodies for both
small and large scales.10,26 Ascites fluid contains 1-10 mg/mL
of globulins, and 10-30 mg/mL of other serum proteins
including albumin (MW ≈ 66 kDa) and transferrin (MW ≈ 80
kDa).
The procedure consists of three steps: (i) Addition of
ammonium sulfate (AMS) to a final concentration of 35% of
the saturated concentration precipitated of all proteins and
complexes heavier than an IgG (150 kDa). (ii) After removing
the precipitate, addition of bi- or trivalent hapten formed cyclic,
higher-molecular-weight complexes (described in Figure 1) of
the IgG of interest (here, anti-2,4-DNP or antidigoxin) (Figure
2).13,27-29 These complexesswith molecular weights of 300,
450, or 600 kDasprecipitated from the 35% AMS solution
immediately. (iii) Centrifugation separated the precipitated
complexes from the supernatant, which retained IgG molecules
incapable of forming complexes with the multivalent haptens
(i.e., those with different specificity, or with one or no active
Fab sites), as well as other proteins with MW e 150 kDa. We
finally solubilized and dissociated the complexes by incubation
with a large excess of monovalent hapten and removed the
haptens by dialysis.
Figure 2. Structures of the bi- and trivalent DNP and 4-NP haptens (1, 2,
3, and 4), bivalent digoxin hapten (5), monovalent 2,4-dinitrophenyl lysine
(6), and 4-nitrophenylglucose (7).
is applicable. (iv) The syntheses of bi- and trivalent haptens
are straightforward. (v) The lifetime of the IgG-DNP complex
is sufficiently long to allow the use of chromatography to
separate the aggregates relevant to this work for analysis and
to understand the mechanisms underlying the process.
Analytical Methods. We determined the efficiency of the
purification with size-exclusion chromatography (SE-HPLC).30,31
This analytical technique can resolve antibody complexes of
different molecular weights, provided that these complexes do
not dissociate over the time required to carry out a separation
(∼20 min).13 Using this technique, we separated the complexes
formed by mixing commercially available IgGDNP with bi- and
trivalent derivatives of DNP, as well as the complexes that
formed by mixing commercially available IgGDgn with bivalent
digoxin (Figure 3). The procedure developed with these
compounds also works with ligands that bind less tightly than
Anti-2,4-DNP rat monoclonal IgG1κ (from clone LO-DNP-
2) and antidigoxin mouse monoclonal IgG1 (from clone DI-
22) antibodies are appropriate for proof-of-principle demon-
strations for five reasons: (i) The purified antibodies and the
ascites fluids are both commercially available. (ii) Both antibod-
DNP
Dgn
ies have high affinity (Kd
≈ 0.8 nM and Kd ≈ 0.1 nM)
DNP and dissociate more rapidly (as we show using 4-
affinity
for their monovalent haptens (a requirement to observe and
isolate the complexes by SE-HPLC). (iii) IgGDNP has a
substantially lower affinity (Kd ≈ 0.5 µM) for monovalent 4-NP
than for 2,4-DNP; we used this low-affinity interaction to
demonstrate the range of values of Kd for which this procedure
nitrophenyl hapten, Kd
≈ 0.5 µM), but the ability to resolve
aggregates was very important in understanding the mechanism
of the purification. SE-HPLC was an effective tool for measuring
the amount of IgGDNP at each step of the purification procedure
(except in the starting ascites fluid), as well as determining the
mole fraction of each complex.
(23) Schweitzerstenner, R.; Licht, A.; Luscher, I.; Pecht, I. Biochemistry
1987, 26, 3602–3612.
By contrast, although this procedure was applicable to the
purification of IgG using the much more weakly binding 4-NP
as a hapten, SE-HPLC could not resolve complexes formed by
bi- and trivalent 4-NP ligands (the chromatograms showed only
(24) Hlavacek, W. S. Biophys. J. 1999, 76, 2421–2431.
(25) Schweitzerstenner, R.; Licht, A.; Pecht, I. Biophys. J. 1992, 63, 551–
562.
(26) Aldington, S.; Bonnerjea, J. J. Chromatogr. B 2007, 848, 64–78.
(27) Baird, B.; Posner, R.; Goldstein, B.; Holowka, D. FASEB J. 1991, 5,
A652–A652.
(30) Subramanian, K.; Holowka, D.; Baird, B.; Goldstein, B. Biochemistry
1996, 35, 5518–5527.
(28) Dembo, M.; Goldstein, B. J. Immunol. 1978, 121, 345–353.
(29) Posner, R.; Goldstein, B.; Holowka, D.; Baird, B. Biophys. J. 1990,
57, A295–A295.
(31) Xu, K. L.; Goldstein, B.; Holowka, D.; Baird, B. FASEB J. 1996, 10,
1254–1254.
9
J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009 9363

A R T I C L E S
Bilgiçer et al.
Figure 4. Size-exclusion chromatograms of antidigoxin ascites fluid. Top
SE-HPLC trace labeled “ascites IgGDgn” is untreated ascites fluid. Anti-
digoxin ascites fluid after incubation with bis-Dgn (5) (2 µM) is labeled
“ascites IgGDgn + bis-Dgn (5)”. The bottom trace is purified IgGDgn mixed
with bis-Dgn (this sample was prepared to match the original IgGDgn
concentration in the ascites). The peaks at 300, 450, and 600 kDa correspond
to cyclic dimer, trimer, and tetramer (Figure 3). The differences in the
absorbance illustrate that the concentration of the IgG in the ascites was
low relative to the other proteins.
ascites fluid. We did not carry out the ELISA procedure for
IgGDgn. The chromatograms in Figure 4 are of IgGDgn-containing
ascites with and without the bivalent Dgn ligand, and after
purification. The concentration of IgGDgn was the same in all
three injections and the peak intensities demonstrate the relative
concentration of IgGDgn to the rest of the proteins in the ascites
fluid.
Step 1. Removal of High-Molecular-Weight Impurities. The
first step consisted of filtering the ascites fluid (0.5 mL) through
glass wool to remove the majority of the liposaccharides. We
rinsed the glass wool with an additional 0.5 mL of phosphate-
buffered saline (PBS, pH 7.4, 10 mM phosphate, 150 mM NaCl)
for a final volume of 1 mL (diluting the ascites fluid 2-fold).
Addition of saturated AMS solution (540 µL; to a final
concentration of 1.4 M (35%)) to the filtered ascites fluid,
followed by centrifugation, separated the proteins having
molecular weights >150 kDa (Figure 5). The proteins that
precipitated have retention times similar to that of the cyclic
complexes that form upon addition of multivalent ligands. Using
an ELISA, we determined that this initial precipitation step
Figure 3. (a) SE-HPLC chromatograms of IgGDgn and IgGDNP complexes
formed upon binding to multivalent ligands. Each antibody-ligand complex
is labeled by the chromatogram. The schematic structures show the
aggregates expected to be formed from IgG; other proteins with the same
MW also seem to be present in small quantities. The products in a mixture
of bis-DNP ligand 1 and IgGDNP showed the cyclic antibody dimer (IgG)212
(MW ) 300 kDa and trimer (IgG)313 (MW ) 450 kDa) (Figure 3a). The
product of a mixture of tris-DNP ligand 2 to IgGDNP, as expected from
previous work, was a bicyclic trimer complex.13 Mixing bis-Dgn ligand 5
with IgGDgn yielded multiple peaks that corresponded to cyclic dimer, trimer,
and tetramer, as well as monomeric antibody. (b) Calibration of the size-
exclusion column using proteins with known molecular masses.
resulted in the loss of only ∼4% of the IgGDNP
.
Step 2. Isolation of Bivalently Active IgGDNP as Complexes.
The supernatant from step 1 contained active IgG, inactive IgG,
and serum proteins with molecular weights equal to or lower
than that of IgG. Adding multivalent haptens 1-4 (to a final
concentration of 5 µM) to the supernatant induced the aggrega-
tion of bivalently active IgG; these complexes immediately
formed a precipitate. To ensure maximum recovery, we
incubated the sample overnight at 4 °C. We then centrifuged
(16 000 g, 30 min) these samples, isolated the pellets, and
redissolved them in PBS for analysis by SE-HPLC (Supporting
Information, Figure S.1).
a monomeric IgG peak) as a result of the kinetic instability of
these complexes. All IgGDNP complexes dissociated rapidly (<1
min) in the presence of excess monovalent hapten (6).
The presence of many other proteins precluded the use of
SE-HPLC for determining the amount of IgGDNP in the ascites,
as many proteins eluted together with the antibody. Therefore,
we used an enzyme-linked immunosorbent assay (ELISA) to
quantify the concentration of IgGDNP in the ascites fluid. We
calculated the percentage of IgGDNP separated after each step
of the protocol by comparing the intensity of fluorescence from
the sample to the intensity of fluorescence from the original
SE-HPLC chromatograms of the pellets isolated using bis-
and tris-DNP haptens 1 and 2 had peaks that corresponded to
the IgG monomer and the cyclic antibody complexes (dimer
and trimer). The presence of antibody monomers in the
chromatogram suggests two possibilities: (i) a fraction of the
9
9364 J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009

Purification of a Monoclonal IgG Antibody
A R T I C L E S
multivalent ligand did not precipitate with the antibody or (ii)
an excess of ligand precipitated and, as a result, a fraction of
the complexes dissociated. Regarding the first possibility: the
addition of more ligand would have driven the formation of
more dimeric and trimeric complexes (but this was not observed;
data not shown). This distribution of antibody complexes
suggests, therefore, that an excess of multivalent ligand pre-
cipitated with the complexes as a result of nonspecific binding.
This type of precipitation, if it did occur, did not interfere with
the rest of the purification.
Analysis of pellets from AMS precipitation with bi- and
trivalent 4-NP haptens 3 and 4 with SE-HPLC showed a single
peak at a retention time that corresponded to monomeric IgGDNP
;
neither chromatogram showed peaks corresponding to higher
aggregates. This observation indicated that, although the
complexes that formed upon the interaction of antibody with
these haptens were sufficiently stable to facilitate precipitation
with AMS, the complexes were not sufficiently stable kinetically
to survive a 20-min SE-HPLC separation. The absence of peak
broadening indicates that the rate constant for dissociation must
be >2.5 × 103 s-1
.
According to the results of the ELISA experiments, the yields
of purified IgG (after step 2) with the multivalent DNP haptens
appear to be much lower than the yields with the multivalent
4-NP haptens: 16% ( 2% from bis-DNP, 11% ( 4% from tris-
DNP, 78% ( 12% from bis-4NP, and 82% ( 20% from tris-
4NP. SE-HPLC experiments with the purified antibodies,
however, indicated that the actual yield obtained by precipitation
with bis-DNP 1 was 66% (the peak area was ∼80% of the peak
area of the pellet generated using tris-4-NP). We hypothesized
that the reason for the low apparent yield of antibody purified
by multivalent DNP ligands, when determined with ELISA, was
that those aggregates remained stable when redissolved in PBS,
and were not accessible (either kinetically or thermodynamically,
we have not determined which) to the DNP on the surface of
the ELISA plate. To test this hypothesis, we reacted com-
mercially available, affinity purified IgGDNP with tris-DNP
(ligand 2) to form the bicyclic complex (IgG322). An ELISA of
this sample detected a concentration of IgGDNP that was only
∼6% of its actual concentration. Therefore, the IgGDNP purified
using the multivalent DNP haptens required purification step
3: dissociation of the complexes by the addition of monovalent
DNP (ligand 6), and removal of the monovalent and multivalent
ligands from the antibody by dialysis.
Step 3. Dissociation of the Cyclic Complexes with Monova-
lent DNP. We solubilized the pellets from step 2 in PBS buffer
and added excess (∼1 mM) DNP-Lysine (6); this monovalent
ligand completely dissociated the cyclic IgG complexes (as
confirmed by SE-HPLC). We dialyzed (10 kDa MWCO
membrane, 4 °C) the sample against monovalent 4-NP 7 (in
order to prevent the reformation of IgG complexes) and
eliminated all the multivalent ligands from the dialysis chamber.
Upon the completion of dialysis, the chamber contained
monovalent ligand 7 together with IgGDNP. A second dialysis
step against PBS at 4 °C removed the low molecular weight
monovalent ligand 7 (as monitored by UV absorbance, λ ) 360
nm). The final product was ∼ 80% of the total amount of the
IgGDNP in the starting ascites fluid (as estimated by ELISA) and
yielded a clean single peak corresponding to 150 kDa on the
SE-HPLC (Figure 6).
Figure 5. Schematic representation of the three steps (1-3) used to purify
bivalently active monoclonal IgGDNP from ascites fluid using AMS
precipitation. The starting material was ascites fluid, which contained a
mixture of IgGDNP with two active Fab binding sites that recognize 2,4-
DNP (active IgG), improperly folded or denatured IgGDNP (damaged IgG),
and IgG fragments (heavy or light chain), as well as other proteins present
in ascites with a range of molecular weights. (1) A low concentration of
AMS (35%) precipitated high molecular weight (g300 kDa) proteins. These
proteins (as a precipitate) were separated by centrifugation as a pellet (a).
(We carried the supernatant, which contained all IgG and low molecular
weight serum proteins, to the next step.) (2) The addition of bi or trivalent
hapten molecules to the supernatant formed complexes of IgGDNP (repre-
sented here as the cyclic dimer and bicyclic trimer), which immediately
precipitated from the solution. This precipitate (b) was isolated by
centrifugation. The supernatant, which now contained any damaged IgGDNP
,
immunoglobulins against other antigens, and other serum proteins, was
discarded. (3) The pellet (b) was dissolved in PBS, and the IgGDNP
complexes were dissociated by the addition of excess monovalent 4-NP 7
(∼1 mM). Dialysis of this solution against phosphate buffered saline (pH
7.4) removed both the monovalent and multivalent haptens, and gave the
This procedure yielded a final product that was >95% pure
by HPLC and, we believe, has two active Fab binding sites.
The most significant experiment to establish the bivalency of
final product as monomeric, bivalently active, IgGDNP
.
9
J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009 9365

A R T I C L E S
Bilgiçer et al.
to their purification. We believe, however, that the bivalent IgE
class of antibodies will be amenable to the purification approach
described here because the only difference between IgG and
IgE antibodies is in the Fc region.
AMS precipitation of cyclic complexes of antibodies provides
a convenient, nonchromatographic method to purify antibodies
from complex solutions and to purify bivalently active antibody
from inactive antibody and/or monovalently active antibody.
This method is, to our knowledge, the only purification
procedure for monoclonal antibodies that selectively isolates
monoclonal IgGs with two active Fab binding sites and is able
to start from a crude biological source of antibodies.
The logic of the method is straightforward, and the procedures
are easy to execute experimentally. They can be applied to small
quantities of solutions and antibodies; although we have not worked
with large volumes or quantities, this procedure should be scalable
to large quantities. We believe that the antibodies isolated using
this procedure will exclusively have two fully active Fab binding
sites, since both sites are required to form the cyclic complexes.
The primary limitation of this technique is its requirement
that appropriate bi- and trivalent haptens be synthetically (or
naturally) accessible. The antibodies that we used in this study
were directed against small-molecule haptens. The requirement
for a synthetically accessible bivalent derivative of the hapten
may limit the application of this technique to purify antibodies
directed toward a recognition site created by the tertiary structure
of a protein, although oligopeptides sometimes can be developed
that bind such proteins (by random combinatorial methods, if
necessary).34-36 For some antibodies directed against proteins,
the bivalent hapten could, in principle, be a dimer of the
antigenic protein. For antibodies raised against large or mem-
brane bound proteins, mimotopes (short peptide sequences that
mimic the binding site), or peptidomimetics (organic molecules
that mimic the function of mimotopes) could serve as multi-
valent molecules.37-39 Discovery of mimotopes usually requires
high throughput screening of peptide libraries, or phage
display.40-42
Figure 6. (Bottom) SE-HPLC trace of the IgGDgn purified using the
described procedure in Figure 5. IgGDgn was present as a mixture of cyclic
complexes (monomer, dimer, trimer, and tetramer), as we did not have a
weaker-binding monomeric ligand to dissociate its complexes. (Middle) SE-
HPLC trace of the IgGDNP purified using our protocol and (Top) trace after
mixing purified IgGDNP with trisDNP (ligand 2).
the purified IgGDNP was to affirm its capacity to form the
kinetically stable bicyclic complex ((IgGDNP)322) upon addition
of tris-DNP 2 (Figure 6b). Conversion of the monomeric IgGDNP
purified with this method to the complex proceeded with >95%
yield. This value was slightly greater than that observed for the
commercially available, affinity-purified IgGDNP (>90%).
This procedure can selectively isolate one IgG from a mixture
of ascites fluidssone containing IgGDNP, and the second IgGDgn
.
The procedure described above separated IgGDNP from this
mixture and gave results similar to those we have described in
detail in the preceding section. Bivalent digoxin ligand 5 could
also purify IgGDgn from this mixture. We conclude that the
procedure is capable of selective precipitation of a target IgG
from a mixture containing multiple IgG molecules with different
specificities. (The experiment summarized in Figure 4 implies
the same conclusion).
We believe that this technique has the potential to be useful
in many applications that require purifying substantial quantities
of antibodies for common biological and clinical analyses, and
perhaps for human therapeutics.43 This technique may also be
useful for fractionating mixtures of polyclonal antibodies from
serum on the basis of their affinity for a given hapten and/or
their specificity.
Discussion
The thermodynamic stability of the complexes, rather than
their kinetic stability, is critical to the effectiveness of this
protocol. Theoretical studies predict that this stability is directly
related to the monovalent affinity of the antibody for the hapten
and the concentration of antibody.28,32,33 SE-HPLC established
that isolation of IgGDNP using multivalent ligands of 4-NP
Experimental Section
Synthesis and Purification of Multivalent Ligands. We used
straightforward synthetic strategies (see Supporting Information)
affinity
affinity
(34) Meloen, R. H.; Puijk, W. C.; Slootstra, J. W. J. Mol. Recogn. 2000,
13, 352–359.
(Kd
≈ 0.5 µM) or of DNP (Kd
≈ 0.8 nM) gave
comparable yields. We believe, based on the results, that this
procedure is applicable for the purification of monoclonal
antibodies with affinities in the range from µM to nM for their
haptens/antigens, provided that bi- and/or trivalent analogues
of these haptens/antigens are synthetically available.
(35) Olson, G. L.; et al. J. Med. Chem. 1993, 36, 3039–3049.
(36) Giannis, A. Angew. Chem., Int. Ed. Engl. 1993, 32, 1244–1267.
(37) Scala, G.; Chen, X. N.; Liu, W. M.; Telles, J. N.; Cohen, O. J.;
Vaccarezza, M.; Igarashi, T.; Fauci, A. S. J. Immunol. 1999, 162,
6155–6161.
(38) Hanessian, S.; McNaughtonSmith, G.; Lombart, H. G.; Lubell, W. D.
Tetrahedron 1997, 53, 12789–12854.
We have not explored the application of this technique to
other antibody isotypes. The majority of previous studies of
cyclic complexes have used bivalent (IgG or IgE) anti-
bodies.13,23,25,29 Further study of the aggregating behavior of
IgAs and IgMs is required before this procedure can be applied
(39) Meola, A.; Delmastro, P.; Monaci, P.; Luzzago, A.; Nicosia, A.; Felici,
F.; Cortese, R.; Galfre, G. J. Immunol. 1995, 154, 3162–3172.
(40) Steward, M. W.; Stanley, C. M.; Obeid, O. E. J. Virol. 1995, 69, 7668–
7673.
(41) Motti, C.; Nuzzo, M.; Meola, A.; Galre, G.; Felici, F.; Cortese, R.;
Nicosia, A.; Monaci, P. Gene 1994, 146, 191–198.
(42) Folgori, A.; Tafi, R.; Meola, A.; Felici, F.; Galfre, G.; Cortese, R.;
Monaci, P.; Nicosia, A. EMBO J. 1994, 13, 2236–2243.
(43) FDA Guidelines for Monoclonal Antibodies for Human Use: http://
www.fda.gov/cber/gdlns/ptc_mab.txt.
(32) Whitesides, G. M.; Krishnamurthy, V. M. Q. ReV. Biophys. 2005, 38,
385–395.
(33) Hornick, C. L.; Karush, F. Immunochemistry 1972, 9, 325–340.
9
9366 J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009

Purification of a Monoclonal IgG Antibody
A R T I C L E S
based on previously reported syntheses to prepare and purify
multivalent ligands 1-5.13
the fluorescence results from the wells that contain samples to that
of a well that contains known concentration of pure IgGDNP. We
averaged four independent sets of ELISA results obtained from
purified antibodies using our purification procedure. We also
measured the enzymatic activity of a known concentration of
commercially available IgGDNP in parallel to quantify the amount
of antibody obtained from precipitation using each multivalent
ligand listed in Figure 2.
SE-HPLC. SE- HPLC measurements were carried out on Tosoh
TSK-GEL G3000SWXL and Tosoh TSK-GEL G4000SWXL size-
exclusion columns using a Varian ProStar 400 HPLC system with
autosampler. HPLC runs were performed with an isocratic solvent
system that was 50 mM phosphate buffer and 370 mM NaCl (to
adjust the ionic strength to 0.475 M) at pH 6.8, with a 0.5 mL/min
flow rate. The sample peaks were analyzed with a UV-vis detector,
as monitored at λ ) 214 nm. The chromatograms of the cyclic
complexes of commercial IgG were obtained from injections where
we kept the concentration of antibody constant while carrying out
serial dilutions of the bi- and trivalent DNP haptens 1 and 2. We
determined the concentrations of our samples using the reported
extinction coefficients for IgGs and DNP. We incubated all samples
for 12 h at 4 °C prior to injection onto the SE-HPLC column.
Samples from purified IgGs from ascites fluid were run after 1/3
dilution into running buffer.
Acknowledgment. This research was supported by the National
Institutes of Health (GM 30367). The American Cancer Society
(S.W.T.) and the NIH (B.F.S. and L.A.E.) provided postdoctoral
fellowships. The authors also thank Dr. David Wood of the
Department of Chemical Engineering at Princeton University for
stimulating discussions and Dr. Charles Mace for assistance with
characterizing the multivalent ligands.
Supporting Information Available: Additional experimental
details, including the synthesis of the multivalent ligands, and
SE-HPLC chromatograms of redissolved pellets after step 2 of
the purification. This material is available free of charge via
the Internet at http://pubs.acs.org.
ELISA. The wells of a 96-well ELISA plate were incubated with
DNP-BSA conjugate to adsorb it on the well surface. After treating
the plate with the sample to be assayed and washing, we incubated
the plate with a secondary antibody (antirat IgG from goat)
conjugated to horseradish peroxidase (HRP). We treated the wells
with Amplex Red, and monitored the HRP-catalyzed hydrolysis
of this substrate using fluorescence (ex, 545 nm; em, 590 nm). To
determine the concentration of IgGDNP in each well, we compared
JA9023836
9
J. AM. CHEM. SOC. VOL. 131, NO. 26, 2009 9367