Human immunoglobulin A (IgA)-specific ligands from combinatorial
engineering of protein A
Jenny Ro¨ nnmark
1
, Hans Gro¨ nlund
2
, Mathias Uhle
´
n
1
and Per-A
˚
ke Nygren
1
1
Department of Biotechnology, Royal Institute of Technology (SCFAB), Stockholm, Sweden;
2
Department of Medicine,
Unit of Clinical Immunology and Allergy, Karolinska Institute, Sweden
Affinity reagents capable of selective recognition of the dif-
ferent human immunoglobulin isotypes are important
detection and purification tools in biotechnology. Here we
describe the development and characterization of affinity
proteins (affibodies) showing selective binding to human
IgA. From protein libraries constructed by combinatorial
mutagenesis of a 58-amino-acid, three-helix bundle domain
derived from the IgG-binding staphylococcal protein A,
variants showing IgA binding were selected by using phage
display technology and IgA monoclonal antibodies (mye-
loma) as target molecules. Characterization of selected
clones by biosensor technology showed that five out of eight

affinity ligand; phage display.
Efficient and selective methods for immunoglobulin (Ig)
detection and purification are of major importance for a
vast number of applications within areas such as immuno-
logy, diagnostics and biotechnology. For these purposes, an
often recruited class of reagents is derived from so-called
receptin proteins, produced as surface-anchored or soluble
proteins by some bacteria [1]. Many of these proteins show
binding to one or more mammalian serum or cell surface
proteins, including for example Ig, serum albumin and
fibrinogen [1]. One of the most well-known Ig-binding
receptins is staphylococcal protein A, widely used in many
formats for its capability to bind a wide spectrum of Igs via
Fc or V
H
region recognition [2–5]. Other examples of
frequently used Ig-binding receptins include streptococcal
protein G [6,7] and Peptostreptococcus magnus protein L
[8,9]. The binding specificities displayed by different
Ig-binding receptins differ significantly in terms of Ig
isotype, subclass, species origin and subfragment type (Fc,
Fab, scFv, etc.), which makes the choice of reagent for a
particular application important.
IgA is the most abundant Ig isotype in humans where it is
present in two subclasses, IgA
1
and IgA
2
, differing mainly in
their hinge region sequences [10]. IgA is predominantly

interface on the IgA Fc fragment.
The biotechnological use of this class of proteins has been
recognized and initially investigated for immunodetection
applications using the IgA binding protein B from group B
streptococci [20].
In this work, we describe an alternative approach to
obtain novel IgA binding ligands, using combinatorial
protein engineering to change the binding specificity of an
already existing receptin protein. Employing a 58-amino-
acid domain derived from one of the immunoglobulin
binding (IgG) domains of staphylococcal protein A as a
scaffold, we have previously described the construction of
Correspondence to P A
˚
. Nygren, Department Biotechnology, Royal
Institute of Technology, SCFAB, SE-106 91 Stockholm, Sweden.
Fax: + 46 855378481, Tel.: + 46 855378328,
E-mail:
Abbreviations: c.f.u., colony forming units; K
d
, dissociation constant;
ABD, albumin binding domain; HSA, human serum albumin; NHP,
normal human plasma; CIC, circulating immune complexes.
(Received 31 January 2002, revised 5 April 2002,
accepted 10 April 2002)
Eur. J. Biochem. 269, 2647–2655 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02926.x
combinatorial protein libraries from which novel affinity
proteins (denoted Z-affibodies) have been selected for
binding to desired target proteins using phage display
technology [21,22]. In common with their protein A

Phage stocks
The phage stocks were prepared as follows. Cells hosting the
respective phagemid library were cultivated in shake flasks
containing 50 mL tryptic soy broth supplemented with
yeast extract (TSB+YE) and ampicillin (100 lgÆmL
)1
) at
37 °CtoD
600
¼ 0.5. An aliquot (10 mL) was incubated
with 10
11
colony forming units (c.f.u.) of M13K07 helper
phage for 30 min at 37 °C followed by centrifugation at
2500 g for 10 min The cells were resuspended and trans-
ferred to 100 mL TSB+YE with ampicillin (100 lgÆmL
)1
),
kanamycin (25 lgÆmL
)1
) and isopropyl thio-b-
D
-galacto-
side (100 l
M
) and cultivated overnight ( 20 h) at 30 °C.
The supernatant of the overnight culture was subjected to
poly(ethylene glycol) precipitation: 25 mL poly(ethylene
glycol)/NaCl [20% poly(ethylene glycol) 6000, 2.5
M

targets during selections. The IgA antibodies were sepa-
rately biotinylated in vitro using a
D
-biotinoyl-e-amidocap-
roic acid N-hydroxysuccinimide ester (Boehringer
Mannheim GmbH, Germany), according to the supplier’s
recommendations. Biotinylated antibodies were bound to
streptavidin-coated paramagnetic DynabeadsÒ (SA-beads)
(Dynal AS, Norway) at conditions resulting in  30 lgIgA
per mg beads. In each selection cycle, approximately
10
10
)10
11
phages were mixed with 5 mg target-containing
beads in a total volume of 100 lL, resulting in a final IgA
concentration during selection of approximately 10 l
M
.
Prior to selection, IgA-containing beads were incubated
with 0.1% gelatine for 30–60 min Prior incubation with
IgA-coated beads phage stocks were incubated with clean
SA-beads (cycles two to five). The two IgA
1
antibodies
were used alternatively as targets during selections
(IgA2167: cycles 1, 2 and 4; IgA1116: cycles 3 and 5).
Phage stocks were incubated with target-containing beads
for 2–4 h, after which the beads were washed with portions
of 500 lLofNaCl/P

using BigDye Terminators (PerkinElmer Applied Biosys-
tems, USA).
Production and purification of proteins from phagemid
constructs
Sequenced clones were expressed as periplasmically secreted
affibody-ABD fusion proteins in the nonsuppressor E. coli
strain, RV308. From an overnight culture in the same
medium, 1 mL was inoculated to 100 mL TSB+YE
supplemented with ampicillin (100 lgÆmL
)1
) and cultivated
at 37 °CtoaD
600
of  1. This culture was induced by
2648 J. Ro
¨
nnmark et al. (Eur. J. Biochem. 269) Ó FEBS 2002
adding isopropyl thio-b-
D
-galactoside to a final concentra-
tion of 1 m
M
and cultivated at room temperature overnight.
The periplasmic fraction was collected by an osmotic shock
procedure and the clones were purified by affinity chroma-
tography on human serum albumin Sepharose column as
follows: the column was pulsed with 0.5
M
HAc, pH 2.8 and
TST (25 m

NaCl, 3.4 m
M
EDTA, 0.005% Surfactant
20, pH 7.4) was used at a flow rate of 5 lLÆmin
)1
in all
experiments. Between injections, the surfaces were regene-
rated with 10 m
M
HCl and/or 0.05% SDS. Proteins were
immobilized on CM5 chip surfaces (Biacore) using amine
coupling chemistry according to the manufacturer’s
recommendations. At one chip,  3400 RU (response
units) of HSA and 5300 RU of IgA2167 were immobilized
in different flow cells. Initial analyses of selected variants
were performed using the HSA surface by injection of
25 lL (25–70 l
M
) protein dissolved in HBS, directly
followed by injection of 25 lL IgA2167 (2 l
M
) in HBS.
The IgA2167 surface was used for binding affinity studies
through steady state response value analyses for different
concentrations of injected analyte employing
BIAEVALUA-
TION
3.0 software (Biacore). The program fits the data to
the formula K
a

IgA1
-ABD protein
(720 RU), Z
wt
-ABD protein (640 RU) and ABD protein
(reference, 380 RU), immobilized in different flow cells for
initial IgG/IgA binding specificity studies and the same chip
was also used for subsequent human isotype and subclass
specificity analyses. 35 lLsamples(0.2l
M
) of human
polyclonal IgA (Sigma–Aldrich Chemie, Germany, cat. no
I-1010), human polyclonal IgA
1
(Calbiochem, USA, cat.
no. 400105), human myeloma IgA
2
(Calbiochem, cat.
no. 400110), human myeloma IgM (Pharmacia Diagnos-
tics), human polyclonal IgG (Pharmacia, Sweden), human
myeloma IgD (Chemicon, USA, art. no AG740), secretory
IgA (Nordic Immunological Laboratories, Netherlands,
cat. no P020) and a monoclonal recombinant IgE
mouse::human chimera consisting of murine light chain
and VH domains and human epsilon 1–4 heavy chain
domains [28] were separately injected and the responses
recorded. The reference surface (ABD) was used to produce
subtractive sensorgrams.
Construction and production of dimeric (head-to-tail)
affibody constructs

ding a dimeric (head-to-tail) (Z
IgA1
)
2
-His
6
affibody con-
struct. The (Z
IgA1
)
2
-ABD affibody fusion protein encoded
by phagemid pKN1-dZ
IgA1
was produced in strain RV308
and purified as described above. The (Z
IgA1
)
2
-His
6
affibody
fusion protein was produced in strain RR1DM15 under the
same conditions as above and purified from the periplasm
by immobilized metal ion affinity chromatography (IMAC)
using 2 mL TALON
TM
(Co
2+
) media (Clontech Laborat-

TM
-10
column (Amersham Biosciences) with a simultanous
buffer change to 5 m
M
NH
4
Ac pH 5.5, lyophilized and
than analysed on a homogenous 20% SDS/PAGE
PhastGel
TM
(Amersham Biosciences) stained with Coomas-
sie Brilliant Blue.
Ligand immobilization and affinity chromatography
The (Z
IgA1
)
2
-ABD or (Z
IgA1
)
2
-His
6
affibody fusion proteins
(2.5 mg and 2 mg, respectively), were immobilized onto
1 mL HiTrap NHS-activated Sepharose
TM
High Perform-
ance columns (Amersham Biosciences), according to the

column at
25 °C and after washing with 10 mL phosphate buffer
Ó FEBS 2002 IgA-specific affibody ligands (Eur. J. Biochem. 269) 2649
pH 7.5 and 2 mL of 5 m
M
NH
4
Ac,pH 5.5,thecolumnwas
eluted with 0.5
M
HAc pH 3.0. Applied sample, flow-
through and eluted proteins were analysed by SDS/PAGE
on a 10–20% NovexÒ gel (Invitrogen life technologies, UK)
and stained by Coomassie Brilliant Blue. Sample, flow-
through, wash fractions and eluted proteins were further
analysed for IgA, IgG and IgM content by immunoprecip-
itation (nephlometry) at the routine laboratory at the
Karolinska Hospital, Stockholm, Sweden, using a Swedac
accredited IMMAGE instrument (Beckman Coulter, Stock-
holm, Sweden). The detection limits for IgG, IgM and IgA
assays were 0.33, 0.04 and 0.02 gÆL
)1
, respectively.
Western blotting
Sample, flow-through and eluate fractions from the affinity
chromatography experiment with unconditioned NHP were
applied on a 10–20% SDS/PAGE gel (NovexÒ) and
proteins were electroblotted to nitro-cellulose membranes
(NovexÒ). The blotted membranes were preblocked in
blocking solution (1% milk powder in TST) for 30 min at

human IgA. In order to direct the selection of IgA binding
affibody variants towards nonvariable domains, two differ-
ent human IgA
1
monoclonal antibodies (myelomas) were
used alternatively as microbead-immobilized panning tar-
gets during five cycles of affibody phage library selection. To
increase the selection stringency, increasing number of
washing steps were introduced in later cycles. DNA
sequence analyses were performed on 10 clones derived
from the fifth cycle eluate, which revealed two clones that
were represented twice (Z
IgA7
and Z
IgA4
) and an additional
six unique variants. An alignment of their amino-acid
sequences showed upon regions of similarities between some
variants (Fig. 1; see Discussion), although no obvious
overall consensus sequence could be identified. For subse-
quent binding analyses, all eight variants were produced
from their respective phagemid vectors as fusions to a
5-kDa albumin binding domain (ABD) affinity tag, facili-
tating their recovery from periplasmic fractions by HSA-
affinity chromatography. Expression levels for all eight
affibody-ABD fusions were in the 10–20 mgÆmL
)1
culture
range.
Biosensor analyses

binding responses were chosen for further binding studies in

Fig. 1. Schematic representation of the location of positions for rando-
mization and their amino-acid occupancies in selected variants. (A)
Ribbon diagram of the three-helix bundle wild type Z domain (PDB
file 1SPZ), with the 13 positions randomized during affibody library
constructions indicated (labeled green and numbers. (B) Deduced
amino-acid occupancies at the variegated 13 positions in the investi-
gated affibody variants selected for human IgA binding in this work.
+/–, confirmed IgA binding or not in biosensor studies (see text for
details).
2650 J. Ro
¨
nnmark et al. (Eur. J. Biochem. 269) Ó FEBS 2002
which their IgA binding affinities (K
d
) were determined to
0.5–3 l
M
, with the highest affinity observed for the Z
IgA1
variant which was chosen for further studies.
The binding specifity of the Z
IgA1
variant was further
analyzed in a series of biosensor binding studies. To
investigate if the IgA binding capability of the Z
IgA1
affibody could possibly be explained by any IgA binding
activity present already in the Z domain scaffold used for

polyclonal IgA
1
and human myeloma IgA
2
samples were
analyzed for binding. The results showed that the Z
IgA1
affibody was indeed capable of binding to both IgA
1
and
IgA
2
(Fig. 4). In addition, human polyclonal IgA, contain-
ing both subclasses, was also efficiently bound by the Z
IgA1
affibody (Fig. 4). Binding was also demonstrated to secre-
tory IgA, composed of dimeric IgA linked by the J-chain
and secretory component (Fig. 4). Taken together, the
binding specificity studies suggested that the Z
IgA1
affibody
recognized an IgA-specific epitope present in all known
forms of human IgA.
Affinity recovery of IgA from a spiked
E. coli
total lysate
To investigate the potential use of the Z
IgA1
affibody as an
IgA-specific ligand in affinity chromatography applications,

Z
IgA4
Z
IgA5
Z
IgA8
Z
IgA7
Z
IgA6
1
2
3
Fig. 2. Overlay sensorgram from IgA binding studies of eight affibody-
ABD fusion proteins. Samples were injected over a HSA-coated surface
for an initial directed immobilization via the ABD tag–HSA inter-
action (1), directly followed by injections of a 2-l
M
solution of human
IgA
1
(myeloma) (2), ending at (3). Five variants, denoted Z
IgA1
-Z
IgA5
were confirmed to bind human IgA under these conditions.
-100
100
300
500

Time [s]
Response units [RU]
Z
wt
surface
Z
IgA1
surface
IgG injected
B
Fig. 3. Initial binding specificity analysis of the Z
IgA1
variant. Resulting
overlay sensorgrams from injections of human IgA (A) or human
polyclonal IgG (B) over separate sensor chip surfaces containing either
Z
IgA1
–ABD or Z
wt
–ABD proteins, as indicated.
-100
400
900
1400
1900
2400
0 200 400 600 800 1000 1200 1400
Time [s]
Response units [RU]
pIgA

complex background of proteins for testing of the selectivity
of the ligand, monoclonal IgA protein was spiked
(0.1 mgÆmL
)1
) into an E. coli total lysate obtained by
sonication of a stationary phase bacterial culture. After
sample loading and subsequent washing, the column was
eluted with low pH. Analysis by SDS/PAGE (reducing
conditions) of the flow-through and eluate fractions showed
that the IgA material had been efficiently recovered with
high purity (Fig. 5). Analysis by nonreducing SDS/PAGE
showed that the eluate contained a band of high molecular
mass but no material migrating as separate heavy and light
chains, indicating that intact IgA had been recovered (data
not shown). This showed that the micromolar affinity of the
interaction was sufficient for affinity chromatography and
that the selectivity of the ligand was high.
Affinity recovery of IgA from human plasma
The potential use of the Z
IgA1
affibody as affinity ligand for
selective chromatographic recovery of IgA directly from
human blood fractions was also investigated. In this
experiment unconditioned normal human plasma was used
as the source of IgA, containing in addition to IgA also all
other antibody isotypes and a highly complex mixture of
other proteins [30]. For this experiment, due to the presence
of serum albumin in plasma, an affinity column was
prepared through coupling of an alternative dimeric ligand
devoid of the albumin binding ABD tag [denoted (Z

in the applied sample. For the flow-through fractions,
similar staining intensities as in the sample fractions were
observed for IgG, IgM and IgD, indicating that these
isotypes were not significantly retarted in the column.
Interestingly, IgA could not be detected in the flow-through
fraction, supporting the notion that this isotype had been
efficiently recovered from the plasma sample. The analyses
of the eluate fraction showed upon a major IgA content as
expected, but with small amounts of contaminating IgG and
IgM (see below).
DISCUSSION
In this work, a combinatorial protein engineering approach
was used to obtain novel IgA binding proteins. Using the
well-characterized protein A-derived IgG-binding, three-
helix bundle domain Z as scaffold for library constructions,
Fig. 5. SDS/PAGE analysis (reduced conditions, Coomassie Brilliant
Blue staining) of different samples from the recovery of IgA protein from
aspikedE. coli total lysate. Lane 1, applied sample; lane 2, flow-
through fraction and lane 3, eluate fraction. The reducing conditions
results in the separation of the IgA into heavy and light chain frag-
ments.
-200
0
200
400
600
800
1000
1200
1400

-His
6
ligand. Intervals for sample loading,
NaCl/P
i
buffer washing, ammonium actetate buffer washing and low
pH elution are indicated. The A
280
was monitored throughout the
experiment.
2652 J. Ro
¨
nnmark et al. (Eur. J. Biochem. 269) Ó FEBS 2002
five out of eight investigated variants selected using phage
display technology showed IgA binding. The reasons
leading to the selection of non-IgA binding variants were
not investigated further, but could be due to background
binding of such variants to components other than IgA
present during selection, such as streptavidin beads and
phages. In several of the confirmed binders, some substitu-
tions were identical with a preference to the second
variegated helix (Fig. 1). This could indicate that these
variants binds to a common site on the IgA protein.
One of the confirmed IgA-binders (clone Z
IgA1
) was
characterized further and showed to be selective for human
IgA in a series of biosensor binding studies, where the
specificity was challenged with all other human isotypes,
including polyclonal IgG. Further studies showed that both

–CH
3
region interface of the Fc fragment, overlapping with the
binding site for the human CD89 IgA receptor [19],
suggesting biological significance. It is worth noting that
in the interactions between IgG and protein A or protein G,
the corresponding domain interface is involved in the
binding [32]. However, the IgA binding proteins described
in this work were selected in vitro, in the absence of any
obvious biological selection pressure, suggesting that more
or less any site on the IgA heavy chain not protected by the
secretory component could contain the binding epitope. The
recognition of both the IgA
1
and IgA
2
subclasses suggests
that the IgA hinge region, which differs in length and
sequence between the two forms of IgA, is not significantly
involved in the binding.
In the affinity chromatography experiments, head-to-tail
dimeric constructs of the Z
IgA1
affibody were used, resulting
in divalent ligands. Although the primary reason for the
dimerization was to increase the likelihood that at least one
moiety in each dimeric ligand was biologically active after
the amine coupling procedure, a divalent capturing ligand
could potentially also result in adventageous avidity effects
in the binding between the ligand and the target as well as to

the isotype analysis different human isotype-specific horse radish-conjugated rabbit IgG sera were used as follows: anti-(human IgA) Ig (B), anti-
(human IgG) Ig (C), anti-(human IgM) (D) and anti-(human IgD) Ig (E).
Ó FEBS 2002 IgA-specific affibody ligands (Eur. J. Biochem. 269) 2653
antibodies of the IgG isotype are effectively purified by
protein A affinity chromatography, which is also imple-
mented at industrial scale processes [35]. However, the
IgA isotype has also been suggested for immunotherapy
applications, owing its strong association with neutrophils
and the possibility to direct antibodies to luminal surfaces
[17]. For example, Streptococcus mutans-specific IgA has
been investigated for administration to the oral cavity for
the prevention of caries [36]. In analogy to the efficient
protein A-based purification of IgG, the development of
corresponding affinity chromatography media containing
a robust affinity ligand selective for IgA should facilitate
the purification of recombinant antibodies of this isotype.
Based on the stable structure of a single protein A
domain, the ligands of the type described in this work
could constitute interesting candidates for such applica-
tions.
ACKNOWLEDGEMENTS
This work has been supported by support from the Swedish Research
Council for Engineering Sciences (TFR), the Swedish Natural Science
Research Council (NFR) and Amersham Biosciences.
REFERENCES
1. Kronvall, G. & Jo
¨
nsson, K. (1999) Receptins: a novel term for an
expanding spectrum of natural and engineered microbial proteins
with binding properties for mammalian proteins. J. Mol. Recognit.

¨
rck, L. & A
˚
kerstro
¨
m, B. (1992)
Protein L from Peptostreptococcus magnus binds to the kappa
light chain variable domain. J. Biol. Chem. 267, 2234–2239.
9. Nilson, B.H., Logdberg, L., Kastern, W., Bjo
¨
rck, L. & A
˚
kerstro
¨
m,
B. (1993) Purification of antibodies using protein
L
-binding
framework structures in the light chain variable domain.
J. Immunol. Methods 164, 33–40.
10. Torano, A., Tsuzukida, Y., Liu, Y.S. & Putnam, F.W. (1977)
Location and structural significance of the oligosaccharides in
human IgA1 and IgA2 immunoglobulins. Proc. Natl Acad. Sci.
USA 74, 2301–2305.
11. Mestecky,J.&McGhee,J.R.(1987)ImmunoglobulinA(IgA):
molecular and cellular interactions involved in IgA biosynthesis
and immune response. Adv. Immunol. 40, 153–245.
12. Lamm, M.E. (1997) Interaction of antigens and antibodies at
mucosal surfaces. Annu. Rev. Microbiol. 51, 311–340.
13. Feehally, J. & Allen, A.C. (1999) Pathogenesis of IgA nephro-

22. Nord, K., Gunneriusson, E., Ringdahl, J., Sta
˚
hl, S., Uhle
´
n, M. &
Nygren, P.A
˚
. (1997) Binding proteins selected from combinatorial
libraries of an a-helical bacterial receptor domain. Nat. Biotechnol.
15, 772–777.
23. Andersson, C., Liljestro
¨
m, P., Sta
˚
hl,S.&Power,U.F.(2000)
Protection against respiratory syncytial virus (RSV) elicited in
mice by plasmid DNA immunisation encoding a secreted RSV G
protein-derived antigen. FEMS Immunol. Med. Microbiol. 29,
247–253.
24. Nord, K., Gunneriusson, E., Uhle
´
n, M. & Nygren, P.A
˚
. (2000)
Ligands selected from combinatorial libraries of protein A for use
in affinity capture of apolipoprotein A-1M and Taq DNA poly-
merase. J. Biotechnol. 80, 45–54.
25. Nord, K., Nord, O., Uhle
´
n, M., Kelley, B., Ljungqvist, C. &

30. Burnouf, T. & Radosevich, M. (2001) Affinity chromatography in
the industrial purification of plasma proteins for therapeutic use.
J. Biochem. Biophys. Methods 49, 575–586.
31. Lindahl, G., A
˚
kerstro
¨
m, B., Vaerman, J.P. & Stenberg, L. (1990)
Characterization of an IgA receptor from group B streptococci:
specificity for serum IgA. Eur. J. Immunol. 20, 2241–2247.
2654 J. Ro
¨
nnmark et al. (Eur. J. Biochem. 269) Ó FEBS 2002
32. Sauer-Eriksson, A.E., Kleywegt, G.J., Uhle
´
n, M. & Jones, T.A.
(1995) Crystal structure of the C2 fragment of streptococcal pro-
tein G in complex with the Fc domain of human IgG. Structure 3,
265–278.
33. Lacroix-Desmazes, S., Kaveri, S.V., Mouthon, L., Ayouba, A.,
Malanche
`
re, E., Coutinho, A. & Kazatchkine, M.D. (1998) Self-
reactive antibodies (natural autoantibodies) in healthy individuals.
J. Immunol. Methods 216, 117–137.
34. Glennie, M.J. & Johnson, P.W. (2000) Clinical trials of antibody
therapy. Immunol. Today 21, 403–410.
35. Fahrner, R.L., Whitney, D.H., Vanderlaan, M. & Blank, G.S.
(1999) Performance comparison of protein A affinity-chromato-
graphy sorbents for purifying recombinant monoclonal anti-