Characterization of a unique human single-chain antibody isolated

by phage-display selection on membrane-bound

mosquito midgut antigens

Brian D. Foy a,b,*, Gerry F. Killeen b,c, Ross H. Frohn b, Daniel Impoinvil b,Andrew Williams d, John C. Beier b

a Interdisciplinary Program of Molecular and Cellular Biology, Tulane University, USAb Department of Tropical Medicine, Tulane School of Public Health and Tropical Medicine, Tulane University Medical Center,

1430 Tulane Avenue SL29, New Orleans, LA 70112, USAc International Centre for Insect Physiology and Ecology, PO Box 30772, Nairobi, Kenya

d Cambridge Antibody Technology Ltd., The Science Park, Melbourn, Cambridgeshire, SG8 6JJ, UK

Received 27 March 2001; received in revised form 6 November 2001; accepted 6 November 2001

Abstract

The insect midgut is the primary site for food digestion, as well as for vector-borne pathogen infection into the invertebrate host.

Accordingly, antigens of this critical insect organ are targets for anti-vector vaccines, insecticidal toxins, and transmission-

blocking vaccines. We used midgut proteins of the African malaria vector mosquito Anopheles gambiae to select single-chain

human antibody fragments (scFv) from a high-diversity, phage-displayed library. Using a phage-display selection method on

western-blotted antigens, we selected an unusual truncated scFv clone, consisting of a heavy-chain only, which binds to An.

gambiaemidgut tissue. This clone binds a spectrum of mosquito antigens from the midgut and other mosquito tissues, as well as

various mammalian glycoproteins, but binding was reduced when these glycoproteins were enzymatically deglycosylated. We

also observed that this clone preferentially binds the lumenal midgut surface. Furthermore, antigen binding by our selected scFv

was limited by competition with increasing concentrations of certain soluble carbohydrates, most dramatically by galactose andN-

acetyl glucosamine. Our results show that the cognate epitope of this scFv is a carbohydrate moiety. This paper describes a phage-

display selection of antibody fragments on mosquito midgut tissue and it also describes a method for phage-display selection on

membrane-immobilized heterogeneous antigens. These selection methods resulted in the isolation of a novel, truncated,

carbohydrate-binding human antibody fragment from a naive phage-display library. D 2002 Elsevier Science B.V. All rights

reserved.

Keywords: Phage display; scFv; Antibody; Carbohydrate; Mosquito; Insect; Midgut

0022-1759/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.

PII: S0022-1759 (01 )00554 -3

Abbreviations: scFv, single-chain variable fragment; PBS, phosphate-buffered saline; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide

gel electrophoresis; PVDF, polyvinylidene fluoride; RT, room temperature; PCR, polymerase chain reaction; FITC, fluorescein isothiocyanate;

CDR, complementary determining region; Bt, Bacillus thuringiensis.* Corresponding author. Department of Tropical Medicine, Tulane School of Public Health and Tropical Medicine, Tulane University

Medical Center, 1430 Tulane Avenue SL29, New Orleans, LA 70112, USA. Tel.: +1-504-988-6785; fax: +1-504-988-6686.

E-mail address: bfoy@tulane.edu (B.D. Foy).

www.elsevier.com/locate/jim

Journal of Immunological Methods 261 (2002) 73–83

1. Introduction

Little is known about the molecular makeup of

mosquito midgut cells and tissue, despite the fact that

the midgut is the primary organ for food digestion

as well as pathogen entry into the insect. Two gene-

ral cell types, absorptive/storage and endocrine, have

been described in adult mosquito midgut epithelium

(Billingsley and Lehane, 1996). Several components

of the peritrophic matrix (PM) have been identified

but many of their functions need to be further elu-

cidated (Moskalyk et al., 1996; LeHane, 1997; Shen

and Jacobs-Lorena, 1998). Much more research has

focused on the identification and function of secreted

enzymes of the midgut (Barillas-Mury et al., 1995;

Muller et al., 1995; Lemos et al., 1996). The molecu-

lar makeup of the glycocalyx, cell membrane, and ba-

sal lamina is still largely uncharacterized.

Mosquito-transmitted pathogens have important

interactions with midgut antigens but, with a few

exceptions, our current understanding of these interac-

tions is confined to the ultrastructural level. For exam-

ple, in the midgut of Anopheles gambiae Giles, the

human malaria parasite must cross the PM, invade and

traverse midgut cells, and then form an oocyst between

the basolateral membrane and the basal lamina. Plas-

modium ookinetes have been shown to have a complex

interaction with the PM, involving the secretion of a

chitinase, to penetrate this barrier (Shahabuddin and

Kaslow, 1994). They have also been shown to adhere

to the luminal surface of Aedes aegypti midgut, spe-

cifically to carbohydrates of the glycocalyx (Zieler et

al., 1999). Likewise, Plasmodium oocysts have been

shown to recognize collagens and laminins that are

abundant components of the basal lamina (Adini and

Warburg, 1999).

Antibodies that bind midgut antigens could be of

great value in determining their cellular location (as

probes) and function (by binding and disrupting nor-

mal function), as well as their interaction with invading

pathogens such as Plasmodium parasites. Many at-

tempts have been made to produce antibodies with ac-

tive anti-vector or transmission-blocking capabilities

against mosquito midgut tissue, but success has been

limited (Kay and Kemp, 1994; Willadsen and Bill-

ingsley, 1996; Shahabuddin et al., 1998). All of these

attempts have utilized traditional approaches for the

generation of polyclonal (immunization) or mono-

clonal (immunization and hybridoma production)

antibodies. In this paper, we utilize a naı̈ve phage-

display library of human antibody fragments, cloned

from the variable regions of human antibody heavy

and light chains, which contains a very high diversity

of clones (>1010) (Vaughan et al., 1996). Phage-

display technology, in conjunction with recombinant

antibody technology, has enabled the in vitro selection

of monoclonal antibodies and can circumvent many of

the problems associated with generating monoclonal

antibodies through immunization (Chiswell and Mc-

Cafferty, 1992). Selected single-chain variable frag-

ments (scFvs) from this library can have binding

affinity to antigen equivalent to that of normal human

antibodies (Kd < 10 nM, off-rates 10 � 3 to 10 � 4

s� 1). With the use of this library, it is possible to ra-

pidly select monoclonal scFv to a large array of indi-

vidual antigens, including those antigens, which are

not normally recognized in vivo.

We have selected the aforementioned library on

western-blotted An. gambiae midgut antigens. We

characterize this selection process and identify a u-

nique scFv that binds An. gambiae midgut antigens.

These data represent the selection of a human anti-

body fragment, through phage-display technology,

which has affinity for the midgut antigens of a mos-

quito and which was selected on western-blotted tis-

sues.

2. Material and methods

2.1. Insects

Anopheles larvae (strain G3) were reared on ground

TetraTabMinR fish food tablets until pupation; Aedes

larvae were reared on ground liver powder. Adults

were maintained on cotton balls soaked in a 10% di-

luted Karok syrup. Mosquitoes were reared and main-

tained in an insectary on a 14 h:10 h light–dark cycle

at 27 �C andf 80% relative humidity. Coptotermes

formosansus termite soldiers were collected from lo-

cal habitat.

2.2. Selection antigen preparation

Adult mosquitoes (3–8 days old) were blown into

alcohol and then quickly transferred to cold phos-

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–8374

phate-buffered saline (PBS) before dissection. Mos-

quito midguts or other tissue were dissected in PBS

and then transferred to sodium dodecyl sulfate (SDS)

solubilization buffer [1% SDS (w/v), 2% CHES (w/v),

1% DTT (w/v), 10% glycerol (v/v)]. Midguts were

solubilized, rocking, in 1 ml of buffer for 10 min, then

frozen at � 80 �C until later use. The antigen solution

was defrosted and the remaining particulates were

pelleted out and was then either subjected to sodium

dodecyl sulfate-polyacrylamide gel electrophoresis

(SDS-PAGE) and western blotting onto polyvinyli-

dene fluoride (PVDF) membrane (Millipore), or was

directly dot blotted onto PVDF. Antigen-bearing mem-

brane was cut into strips and then immersed in 9 ml of

MPBS-T blocking solution (2% non-fat dry milk, 0.1%

Tween-20 in PBS, pH 7.4) in a sterile tube and stored

at 4 �C overnight in preparation for phage panning the

following day.

2.3. Antibody fragment library and phage panning

The high-diversity phage-displayed scFv library

was constructed by Cambridge Antibody Technology,

and is described in detail elsewhere (Vaughan et al.,

1996). The library was constructed from naı̈ve human

donors. We performed this selection with both dot-

blotted and western-blotted antigens; however, the

selection results were essentially identical and each

resulted in substantial enrichment of the same clone.

Therefore, we have presented only the results of the

western blot selections for expediency. Approximately

1012 transforming units (t.u.) of phage was preblocked

in 1 ml MPBS-T at room temperature (RT) for 30 min.

These phages were then incubated rocking, in a 15-ml

sterile tube, with the previously stored antigen blot

strip, typically five guts/lane, for 1 h at RT. After

phage incubation, the solution was poured out and the

membrane was rigorously washed 3� in PBS-T and

then 3� in PBS to remove non-specifically bound

phage. The membrane was then put in a new container

and incubated in 100 mM triethylamine at RT for 10

min to remove bound phage. After incubation, the

eluted phage was neutralized with 500 ml of 1 M Tris

buffer, pH 7.4. Eluted phage was infected into log

phase Escherichia coli of either the suppressor TG1

strain (K12, D(lac-pro) supE thi hsdD5 [FV traD36

proA +B + lacIq lacZDM15]) or the non-suppressor

HB2151 strain (K12, D(lac-pro) ara NaIr thi [FV

proA +B + lacIq lacZDM15]). Infected cells were titer-

ed, propagated, harvested and phage particles were

recovered from the TG1-infected bacteria as describ-

ed elsewhere (Kay et al., 1996). The process was re-

peated six times. Selection rounds that did not indicate

a positive enrichment factor were discarded and re-

peated.

2.4. Phage binding to blotted antigens

Midgut antigens were blotted as previously de-

scribed and western strips were incubated in blocking

solution supplemented with 25% glycerol (GMPBS-

T) at 4 �C overnight. 1011 t.u. of phage was preblock-

ed in GMPBS-T for 30 min at RT. The phage and

membranes were then incubated together for 1 h at

RT. After incubation, the blots were washed vigo-

rously, seven times at 30 min each using GMPBS-T.

Anti-M13–HRP conjugate (Amersham) was diluted

1:2000 in MPBS-T and incubated with the blots for 1

h at RT. The blots were then washed five times at 15

min apiece in PBS-T and developed using the ECL

system (Amersham).

2.5. scFv surface-enrichment western blots

Pooled phages from each selection round were

solubilized by mixing 1:1 with 2� SDS-PAGE buffer

and then boiling for 5 min. This mixture was subjected

to SDS-PAGE and western blotted. The membrane

was blocked with MPBS-T. The primary developing

antibody was mAb 9E10 (Sigma) diluted 1:500. The

secondary developing antibody was anti-mouse IgG

(g-chain specific)–HRP conjugate (Sigma) diluted

1:1000. Blots were developed using the ECL system

(Amersham).

2.6. Clone analysis and sequencing

Colonies of bacteria harboring single clones were

picked with a sterile pipette tip and put directly into

polymerase chain reaction (PCR) buffer for scFv insert

amplification. The insert expression region was ampli-

fied by the following vector primers: forward 5V-AGCGGATAACAATTT CACACAGG, reverse 5V-GTGGTGTTTCCAGACGTTAGT: with 30 cycles of

1 min at 94 �C, 1 min at 55 �C, and 1 min at 72 �C,followed by 10 min extension cycle at 72 �C. For clone

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–83 75

sequencing, these PCR amplicons were purified with

the Concert PCR purification system (Life Technolo-

gies) and sequenced using an automated sequencer

with the following inner primers: heavy chain 5V-ACCGCCAGAGCCACCTCCGCC, light chain 5V-CTCTTCTGAGATGAGTTTTTG.

2.7. Soluble scFv analysis and binding to blotted

antigens

Soluble scFv was recovered by picking individual

colonies of infected E. coli HB2151 strain. Each

colony was grown to saturation overnight in a sepa-

rate sterile tube or a separate well of a 24-well culture

plate in 2TYAG liquid medium (20 ml or 400 ml,respectively) at 30 �C. The following day, cultures

were diluted 2.5:1 in prewarmed 2TYAG and allowed

to grow for 1 h at 30 �C. The cells were then cen-

trifuged and resuspended in 2.5� their original (o-

vernight) volume with prewarmed 2TYA+ 100 mg/ml IPTG (2TYAI) and further incubated for 4 h at 30

�C. After incubation, the cells were centrifuged again

and resuspended in 5% of their original volume with

hypotonic extraction buffer [20% sucrose (w/v), 1

mM EDTA, 50 mM Tris–HCl, pH 8.0] and incubated

on ice for 10 min. Cells were removed by centrifu-

gation and the supernatant (periplasmic extract con-

taining soluble scFv) was harvested and frozen at

� 80 �C until further use. Alternatively, scFv was pu-

rified from periplasmic extract using a NTA-nickel

column according to the manufacturer’s protocol (Qia-

gen). For size analysis, scFv was denatured in SDS

sample, subjected to SDS-PAGE and western blot-

ting, and the membrane was probed with mAb 9E10

followed by anti-mouse IgG/HRP as previously de-

scribed. The soluble scFv binding assays were devel-

oped essentially as described for a normal western

blot.

2.8. Phage clone binding to whole midgut tissue

Midgut tissue was fixed and incubated with anti-

bodies according to a previously described protocol

(Shahabuddin and Pimenta, 1998). 1011 t.u. of primary

Table 1

Selection titers

Selection Input Output Antigen control Phage control Enrichment

Round 1 2� 1011 1.5� 105 N.D. 9.1�104 Max. 1.6�Round 2 6� 1012 3.5� 105 1.4� 105 2� 102 2.1�Round 3 4.1�1012 6.2� 105 3.6� 105 3.4� 103 1.7�Round 4 1�1012 1.7� 105 1.1�105 4� 102 1.5�Round 5 1�1012 6.2� 103 1.4� 103 < 2� 102 4.4�Round 6 8.7� 1012 4.9� 106 1.4� 106 1�103 3.5�The input, output, and control phage titers were recorded as the number of transforming units (t.u.) of phage. The enrichment factor was

calculated by dividing the output by the greater of the two controls.

Fig. 1. Pooled phage binding to western-blotted An. gambiae

midguts (a) or surface expression of scFv (b). (a) 1011 t.u. phage

was incubated with each lane of antigen: anti-FITC phage on BSA–

FITC (con+), anti-FITC phage on BSA (con� ), pCANTAB6 phage

on midguts (� ), or selection round total output phage on midguts

(1–6). Ten micrograms/lane of BSA and BSA-FITC and five

midguts/lane were blotted. (b) 1011 t.u. of output phage from each

round (1–6) or control phage anti-FITC (+), and pCANTAB6 (� ),

was solubilized and western blotted.

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–8376

phage antibody was incubated with the fixed midgut

tissue (pCANTAB6 phage or 6wAgB6 phage). Then

midguts were washed and then incubated in secondary

antibody, anti-M13-HRP at 1:2000 dilution. Phage

binding was detected with DAB substrate (Sigma).

2.9. Enzymatic deglycosylation of model antigens

The glycoproteins bovine ovalbumin and human

transferrin were deglycosylated by O-glycosidase DS,

NANase II, and PNGase F using BioRad’s Enzymatic

Deglycosylation Kit according to the manufacturer’s

instructions. Antigens were western blotted and de-

tected with phage as previously described. Protein

was detected on replicate western blots using Colloi-

dal Gold Total Protein Stain (BioRad).

2.10. Midgut antigen ELISA

Unfed adult An. gambiae midguts were dissected,

frozen, ground in PBS containing a cocktail of pro-

tease inhibitors (Sigma), and the solution was coated

onto 96-well plates at a concentration of five guts/

well overnight, 4 �C. Some positive control wells were

coated with BSA–fluorescein isothiocyanate (FITC)

conjugate (10 mg/well). The plates were washed with

PBS and blocked in MPBS for 2 h at 37 �C. ScFv-bearing phage (1�1011 t.u./well) was diluted inMPBS

or MPBS containing carbohydrates (Sigma) and pre-

blocked for 1 h at RT. Then these mixtures were al-

lowed to incubate in the wells for 1 h at 37 �C. Thewells were then washed with 3� PBS-T, and 3� PBS,

incubated with anti-M13–HRP conjugate (1:5000) for

1 h at 37 �C, washed again, developed with ABTS sub-

strate (Pierce), and read using 405 nm wavelength

light.

Fig. 2. scFv extract from four random sixth-round clones and

pCANTAB6 (pC) was tested for binding to midgut antigens An.

gambiae (A.g.), An. freeborni (A.f.), Ae. aegypti (A.a.) (five guts/

dot), and to BSA, avidin, (10 mg/dot) or buffer alone (none).

Fig. 3. The 6wAgB6 was identified to be a truncated clone both by

PCR fragment size (a) and scFv expression product (b). (a) The

insert region of various clones was PCR amplified. Phagemid clones

with normal scFv inserts have 1 kb amplicons (ns). Amplification of

the 6wAgB6 insert generated a f 650-bp amplicon (B6), while the

empty vector pCANTAB6 (pC) shows no scFv insertion. (b) These

same clones were induced to express soluble scFv, which was then

extracted from bacterial periplasms and detected on western blots.

While normal scFv is a f 30-kDa expression product, the 6wAgB6

clone is about half that size (f 17 kDa).

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–83 77

Fig. 4. Nucleotide and predicted amino acid sequences of 6wAgB6. Heavy-chain amino acid residues are grouped and numbered according to

Kabat (1991). FR1-3 and CDR1-3 identify the framework and CDR regions, respectively. Other scFv modifications are underlined and labeled.

Fig. 5. 6wAgB6 binds a variety of western-blotted insect tissues as well as various glycoproteins, but it does not bind bovine fetuin, or BSA.

Comparison with coomassie-stained midguts (c-Ag) reveals a pattern of binding to many, but not all, proteins of midgut tissue. Midguts of An.

gambiae (Ag), An. freeborni (Af ), Ae. aegypti (Aa), and C. formosansus (Cf ). An. gambiae ovaries (ov), and malphigian tubules (mp).

Glycoproteins: bovine carbonic anhydrase II (ca), ovalbumin (ob), human transferrin (tf ), bovine fetuin (fe), and bovine serum albumin (bsa).

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–8378

3. Results

3.1. Selections

The selection process was repeated through six

selection rounds for this study. The resulting phage

titers of each selection round, in terms of the number

of recovered (output) phage after each round, are

presented in Table 1. Dividing the selection round

output titer from the higher of the two control se-

lection titers gave us an enrichment factor by which

to determine if that particular selection round en-

riched for antigen-specific clones (versus clones that

had non-specific affinity toward the PVDF or anti-

gens).

3.2. Phage binding to blotted antigens

To gauge the number of selections required to en-

rich for a phage population with detectable affinity to

the midgut antigens, we tested the phage pools from

each round for reactivity to midgut antigens by wes-

tern blot. Phage from rounds one to four, as well as the

control phagemid vector pCANTAB6, showed no de-

tectable affinity toward An. gambiae antigens (Fig.

1a). Phage from the fifth round of selection exhibit-

ed minimal affinity. Sixth-round phage readily bound

blotted midgut antigens. Antigen binding is apparent

across the spectrum of protein molecular weights.

3.3. scFv surface-display on phage

By solubilizing the phage pools, themselves, in

SDS and western blotting these phage antigens, we

determined that enrichment of selection rounds for

binding ability could also be assessed by detection of

scFv expression on the surface of phage (Fig. 1b).

Surface display of scFv was assessed by detection of

the c-myc epitope, which is linked with the expressed

scFv, using an anti-c-myc antibody. As compared

with the positive control, monoclonal anti-FITC pha-

ge, efficient display of scFv by the selection pools

was not apparent until the sixth round. These results

imply that enrichment of surface-displayed scFv

corresponds with the antigen-binding nature of each

phage pool. The anti-FITC phage control was ob-

served as a single band around 97 kDa indicating a

fusion protein of the scFv (f 30 kDa) and geneIII

(f 52 kDa). Interestingly, the sixth-round phage,

scFv–geneIII fusion proteins appeared as a doublet,

indicating that a truncated fusion protein of f 65

kDa was enriched in the panning.

3.4. Isolation of the midgut binding clone

We selected individual clones from the sixth-round

selection and extracted scFv from phagemid-infected

E. coli to test for scFv affinity to midgut antigens as

well as to non-specific antigens. After an initial round

of primary screening on An. gambiae midgut antigens

alone (data not shown), reactive scFv clones were

stockpiled and tested again against a variety of anti-

gens (Fig. 2). All showed a similar pattern of reac-

tivity, which consisted of a general affinity towards

midguts of several mosquito species, but no compa-

Fig. 6. 6wAgB6 preferentially binds to the lumenal side of An.

gambiae midgut tissue. (a) A reddish-brown precipitate indicates

6wAgB6 phage binding to the lumenal midgut surface (L) but not to

the outer midgut that is covered by basal lamina (O). (b) pCANTAB6

phage does not bind midgut tissue.

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–83 79

rable binding of the unrelated antigens BSA or avidin,

nor buffer alone spotted on the membrane.

We amplified the DNA of all confirmed reactive

clones through PCR, and checked for clonal diffe-

rences by a BstN1 restriction pattern fingerprint (data

not shown). Our analysis indicated that all positive

clones were identical and this was later confirmed by

sequence analysis. Hence, we concluded that we had

selected for a single, midgut antigen binding clone

that we designated 6wAgB6. Upon PCR amplification

of our clone, it was clear that this scFv was actually

a truncated clone and subsequent expression of so-

luble scFv resulted in an antibody that was half of

the expected molecular weight (Fig. 3a and b). Se-

quencing of the 6wAgB6 verified these unusual

results and indicated that the clone consisted of a

heavy chain properly fused to the GS linker, but it

lacked a light chain (Fig. 4). The heavy chain is

derived from the human VH subgroup 3 germline and

most closely resembles VH germline sequenceDP48 as

determined by alignment using V-BASE (Tomlinson et

al., MRC Center for Protein Engineering, Cambridge,

UK).

3.5. Characterization of 6wAgB6 binding specificity

To identify the cognate epitope of 6wAgB6, we

tested this clone for binding to a variety of western-

blotted antigens (Fig. 5). 6wAgB6 exhibits a broad

pattern of reactivity with the full spectrum of midgut

protein molecular weights. Similar broad reactivity

patterns were seen on midguts of the mosquitoes An.

Fig 7. 6wAgB6 loses affinity for glycoproteins after enzymatic deglycosylation. Ovalbumin (top) and human transferrin (bottom) were

enzymatically treated with O-glycosidase, NANase II, and PNGase F and western blotted. The left blots were stained for total protein with

colloidal gold and the right blots were developed using 6wAgB6 phage as the primary antibody probe. (1) Enzymes/no antigen, (2) no enzymes/

antigen, (3) enzymes/antigen.

Table 2

Certain carbohydrates can compete against 6wAgB6 midgut binding. The numbers are mean adsorbance values from three separate ELISA

experiments

Control Glu Gal Man Fuc GlcNAc GalNAc NANA NANA-L

6wAgB6 1.13 1 mM 1.18 1.31 1.02 1.12 0.70* 0.74 0.94 0.87

10 mM 0.94 0.82 0.78 0.82 0.87 0.93 0.83 1.06

100 mM 0.81 0.71* 0.75 0.82 0.68* 1.10 1.03 1.00

1 mM 0.74* 0.69* 0.57* 0.56* 0.57* 0.79 0.73* 0.79

pCAN (� ) 0.49 1 mM 0.48 0.39 0.53 0.54 0.47 0.49 0.53 0.53

GF2 (+) 1.56 1 mM 1.11 1.10 1.04 1.18 1.40 1.64 1.43 1.35

Anti-FITC (+) 2.05 1 mM 2.14 2.11 2.09 1.98 2.14 2.03 1.87 1.86

Mean adsorbance of phage clones to midguts [6wAgB6, pCAN (� ), GF2 (+)] or to BSA–FITC [anti-FITC (+)] without carbohydrates is

reported as control. 6wAgB6 binding was competed with increasing concentrations of carbohydrates. Control phage clones were only competed

with the highest carbohydrate concentration (1 mM). Adsorbance values that significantly differed from their controls by Student’s t-test

( p< 0.05) are marked with an asterisk (*). Glu, glucose; Gal, galactose; Man, mannose; Fuc, fucose; GlcNAc, N-acetyl glucosamine; GalNAc,

N-acetyl galactosamine; NANA, N-acetylneuraminic acid; NANA-L, N-acetylneuraminyl-lactose.

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–8380

freeborni and Ae. aegypti, as well as the termite C.

formosansus. Furthermore, 6wAgB6 bound western

blots of An. gambiae larval midguts, ovary and mal-

phigian tubule tissue, and several mammalian glyco-

proteins. Negative control phagemid pCANTAB6 did

not bind any of these tissues (data not shown). Despite

the broad reactivity of 6wAgB6, it did not bind equiv-

alent amounts of fetuin or BSA, nor did it bind every

antigen on western-blotted midguts, as is shown by

comparison with the coomassie-stained western blot.

We also incubated 6wAgB6 phage with fixed midguts

to identify the binding epitope’s tissue location (Fig. 6).

From this figure, it is apparent that this clone prefer-

entially binds to the lumenal side of An. gambiae

midguts.

The binding pattern on western blots and the bind-

ing location on fixed midgut suggested that 6wAgB6

binds a carbohydrate moiety. Consequently, we were

able to reduce the apparent affinity of 6wAgB6 toward

the glycoproteins upon enzymatic deglycosylation

(Fig. 7). Furthermore, we were able to use certain so-

luble carbohydrates to compete against 6wAgB6 bind-

ing in a midgut-antigen ELISA (Table 2). None of the

carbohydrates altered the midgut binding of a full-

length scFv clone we isolated from the same library

(GF2), nor did they reduce binding of an anti-FITC

scFv clone toward BSA–FITC conjugate.N-acetylglu-

cosamine had the most dramatic effect on 6wAgB6

binding, even at small concentrations. Galactose see-

med to limit 6wAgB6 binding in a dose-dependant

manner.

4. Discussion

We describe the first phage-display selection of

human scFv on mosquito midgut antigens using a

membrane-selection technique, and we characterize

the resultant unique scFv clone as well as define its

tissue and antigen specificity. The 6wAgB6 clone i-

solated in this study binds to a spectrum of mosquito

and other antigens on western blots. Its binding ubiq-

uity, midgut localization, lack of affinity for degly-

cosylated model antigens, and failure to bind midgut in

competition with certain carbohydrates, support our

conclusions that the epitope is, at least in part, a

common carbohydrate shared by many different insect

proteins. The phage-bound clone shows preferential

affinity for the lumenal cell surface of dissected midgut

as opposed to the basal lamina surface. Affinity to the

former further suggests that this carbohydrate epitope

is primarily located within the glycocaylx, and not to

carbohydrates that may exist on the basal lamina.

Antibodies that bind the glycocalyx may be useful

because there is evidence that Plasmodium ookinetes

use brush border carbohydrate residues for recognition

and binding of midgut cells before cell invasion (Zieler

et al., 1999). Arboviruses from the family Flaviviridae

have also been shown to specifically recognize carbo-

hydrates to invade target cells (Chen et al., 1997).

The truncated nature of the 6wAgB6 clone is very

interesting and was quite unexpected due to the fact

that the library was specifically made as fusion pro-

teins between recombined heavy and light chains

(Vaughan et al., 1996). The size of this clone may be

directly related to its cognate epitope and explain its

selection. ScFv heavy and light chains are normally

interspersed with a flexible polypeptide linker to allow

them to fold back onto one another and form a

complementary determining region (CDR) binding

groove that molecularly mimics natural antibodies

(Huston et al., 1988). It may be possible that at least

some heavy-chain sequences were not linked in the

ligation step and slipped past the cloning and trans-

fection steps, or that the bacterial host excised the

light-chain DNA of this clone. Recently, there have

been studies aimed at making phage-displayed binding

units smaller to enhance their function as probes and

therapeutic molecules, including the display and selec-

tion heavy chain-only antibodies (Yokota et al., 1992).

The model for this work are the antibodies of drom-

edaries (camels, llama, etc.) which naturally produce

heavy chain-only antibodies that are homologous to

the VH3 germline family of human heavy chains,

differing by only three amino acids of the second

framework region (Riechmann and Muyldermans,

1999). Hence, it is interesting that 6wAgB6 is also

derived from the VH3 family. Furthermore, heavy

chain-only antibodies offer a unique binding site for

antigens due to their long third hypervariable loop,

which can protrude from the remaining paratope (Des-

myter et al., 1996). This feature can give them a

convex paratope architecture (as opposed to the con-

cave paratope of a normal human CDR) and allows

them to bind into antigen pockets, such at the active

site of enzymes. Notably, 6wAgB6 has the longest

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–83 81

CDR3 loop possible, according to the Kabat residue

numbering system, which might explain its affinity to

carbohydrates.

The selection technique we employed is novel and

it may be broadly applicable as it was performed on

membrane-blotted heterogeneous antigens. Although

western-blotted antigens are not generally in their

native conformation, the technique allows for selec-

tion of scFv that bind protein in a format which is

readily amenable to manipulation and analysis using

standard laboratory methods. Thus, this technique

may allow selection on groups of proteins or protein

bands which are minimally separated from each other

by SDS-PAGE gels, isoelectric focusing, or two di-

mensional gels. It has been suggested that the ra-

pidly growing field of proteomics is still limited by

the need to make antibodies against separated pro-

teins using traditional purification and immunization

methods (Service, 2000). Phage-display selection on

western-blotted 2-D gel spots could greatly enhance

the efficiency and speed of this research.

Despite possible advantages of these selection me-

thods, we still were not able to overcome certain bar-

riers that others have described when selecting on

heterogeneous antigens (Mutuberria et al., 1999),

primarily, the lack of diversity of resultant clones.

Our repeated rounds of selection, propagation, and

reselection may have led to a lack of clonal diversi-

ty and a ‘bottlenecking’ of the final selection to se-

veral or fewer clones. The ubiquity of this clone’s

midgut epitope may also help explain the dominan-

ce of a single scFv. Since this epitope seems to be

common among many, if not most of the midgut

proteins, it will most likely outnumber each unique

epitope on individual proteins. Thus, a clone that re-

gnizes such a common epitope, will outnumber all

others upon phage elution at the end of the selection

round. Many rounds of clonal expansion and reselec-

tion will then only exacerbate the matter and the ‘bot-

tlenecking’ would occur. However, the ability to se-

lect antibody panels with specificity to many epitopes

in complex antigen mixtures has been demonstrated

using whole cell and tissue mixtures (Edwards et al.,

2000). Such antibody panels would be ideal for high

throughput screening methods that test for in vivo

biological activity in insects (Killeen et al., 2000).

Better deselection and blocking techniques may be

able to circumvent the problems that we encountered

for the generation of either more diverse scFv panels,

or deliberately targeted antibody fragments.

Understanding the composition and function of the

insect midgut is essential to the fields of vector biology,

pest biology, and vector-borne disease transmission;

phage-displayed antibodies can help to define this

research. From an applied research perspective, the

small size of scFv, especially the truncated clone we

describe, may lend themmore ame-nable to fusion with

a toxic peptide, were they could serve as a functional

binding domain and thus enable localization and sub-

sequent increase of toxin activity in insects. There is

currently great interest in the creation of second

generation bioinsecticides, including toxic antibodies

which could be cloned into plants, thus decreasing

the reliance on transgenic Bacillus thuringiensis (Bt)

cultivars (Boulter, 1993). Finally, with continued re-

finement, phage-display selection on membrane-

bound complex antigen mixtures should enable faster

and easier generation of monoclonal antibody frag-

ments against both individual, hard-to-purify anti-

gens as well as panels of solubilized tissue antigens.

Acknowledgements

We thank our collaborators at Cambridge An-

tibody Technology for the generous use of their

phage-displayed scFv library, the bacterial strains,

the anti-FITC E2 clone, and their technical support.

We also thank Scott Michael for donation of the

termites. This work was supported by NIH grant A1

29000 and B.D.F was supported by the Louisiana

Educational Quality Scholarship Fund grant (1996-

01)-GF-23.

References

Adini, A., Warburg, A., 1999. Interaction of Plasmodium gallina-

ceum ookinetes and oocysts with extracellular matrix proteins.

Parasitology 119, 331.

Barillas-Mury, C.V., Noriega, F.G., Wells, M.A., 1995. Early trypsin

activity is part of the signal transduction system that activates

transcription of the late trypsin gene in the midgut of the mos-

quito, Aedes aegypti. Insect Biochem. Mol. Biol. 25, 241.

Billingsley, P.F., Lehane, M.J., 1996. Structure and ultrastructure of

the insect midgut. In: Billingsley, P.F., Lehane, M.J. (Eds.),

Biology of the Insect Midgut. Chapman & Hall, London, p. 3.

Boulter, D., 1993. Insect pest control by copying nature using ge-

netically engineered crops. Phytochemistry 34, 1453.

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–8382

Chen, Y., Maguire, T., Hileman, R.E., Fromm, J.R., Esko, J.D.,

Linhardt, R.J., Marks, R.M., 1997. Dengue virus infectivity

depends on envelope protein binding to target cell heparan sul-

fate [see comments]. Nat. Med. 3, 866.

Chiswell, D.J., McCafferty, J., 1992. Phage antibodies: will new

‘coliclonal’ antibodies replace monoclonal antibodies? Trends.

Biotechnol. 10, 80.

Desmyter, A., Transue, T.R., Ghahroudi, M.A., Thi, M.H., Poort-

mans, F., Hamers, R., Muyldermans, S., Wyns, L., 1996. Crystal

structure of a camel single-domain VH antibody fragment in

complex with lysozyme [see comments]. Nat. Struct. Biol. 3, 803.

Edwards, B.M., Main, S.H., Cantone, K.L., Smith, S.D., Warford,

A., Vaughan, T.J., 2000. Isolation and tissue profiles of a large

panel of phage antibodies binding to the human adipocyte cell

surface [In Process Citation]. J. Immunol. Methods 245, 67.

Huston, J.S., Levinson, D., Mudgett-Hunter, M., Tai, M.S., No-

votny, J., Margolies, M.N., Ridge, R.J., Bruccoleri, R.E., Haber,

E., Crea, R., et al., 1988. Protein engineering of antibody bind-

ing sites: recovery of specific activity in an anti-digoxin single-

chain Fv analogue produced in Escherichia coli. Proc. Natl.

Acad. Sci. U. S. A. 85, 5879.

Kabat, E.A., 1991. Sequences of proteins of immunological interest.

N.I.H. Publication, 91–3242.

Kay, B.H., Kemp, D.H., 1994. Vaccines against arthropods. Am. J.

Trop. Med. Hyg. 50, 87.

Kay, B.K., Winter, J., McCafferty, J., 1996. Phage Display of Pep-

tides and Proteins. A Laboratory Manual. Academic Press, San

Diego.

Killeen, G.F., Foy, B.D., Shahabuddin, M., Roake, W., Williams,

A., Vaughan, T.J., Beier, J.C., 2000. Tagging bloodmeals with

phagemids allows feeding of multiple-sample arrays to single

cages of mosquitoes (Diptera: Culicidae) and the recovery of

single recombinant antibody fragment genes from individual

insects. J. Med. Entomol. 37, 528.

LeHane, M.J., 1997. Peritrophic matrix structure and function. An-

nu. Rev. Entomol. 42, 525.

Lemos, F.J., Cornel, A.J., Jacobs-Lorena, M., 1996. Trypsin and

aminopeptidase gene expression is affected by age and food

composition in Anopheles gambiae. Insect Biochem. Mol. Biol.

26, 651.

Moskalyk, L.A., Oo, M.M., Jacobs-Lorena, M., 1996. Peritrophic

matrix proteins of Anopheles gambiae and Aedes aegypti. Insect

Mol. Biol. 5, 261.

Muller, H.M., Catteruccia, F., Vizioli, J., della Torre, A., Crisanti,

A., 1995. Constitutive and blood meal-induced trypsin genes in

Anopheles gambiae. Exp. Parasitol. 81, 371.

Mutuberria, R., Hoogenboom, H.R., van der Linden, E., de Bruine,

A.P., Roovers, R.C., 1999. Model systems to study the param-

eters determining the success of phage antibody selections on

complex antigens. J. Immunol. Methods 231, 65.

Riechmann, L., Muyldermans, S., 1999. Single domain antibodies:

comparison of camel VH and camelised human VH domains. J.

Immunol. Methods 231, 25.

Service, R.F., 2000. Proteomics. Can Celera do it again? Science

287, 2136.

Shahabuddin, M., Kaslow, D.C., 1994. Plasmodium: parasite chiti-

nase and its role in malaria transmission. Exp. Parasitol. 79, 85.

Shahabuddin, M., Pimenta, P.F., 1998. Plasmodium gallinaceum

preferentially invades vesicular ATPase-expressing cells in

Aedes aegypti midgut. Proc. Natl. Acad. Sci. U. S. A. 95, 3385.

Shahabuddin, M., Cociancich, S., Zieler, H., 1998. The search form

novel malaria transmission-blocking targets in the mosquito

midgut. Parasitol. Today 14, 493.

Shen, Z., Jacobs-Lorena, M., 1998. A type I peritrophic matrix

protein from the malaria vector Anopheles gambiae binds to

chitin. Cloning, expression, and characterization. J. Biol. Chem.

273, 17665.

Vaughan, T.J., Williams, A.J., Pritchard, K., Osbourn, J.K., Pope,

A.R., Earnshaw, J.C., McCafferty, J., Hodits, R.A., Wilton, J.,

Johnson, K.S., 1996. Human antibodies with sub nanomolar

affinities isolated from a large nonimmunized phage display

library. Nat. Biotechnol. 14, 309.

Willadsen, P., Billingsley, P.F., 1996. Immune intervention against

blood-feeding insects. In: LeHane, M.J., Billingsley, P.F. (Eds.),

Biology of the Insect Midgut. Chapman & Hall, London, p. 323.

Yokota, T., Milenic, D.E., Whitlow, M., Schlom, J., 1992. Rapid

tumor penetration of a single-chain Fv and comparison with

other immunoglobulin forms. Cancer Res. 52, 3402.

Zieler, H., Nawrocki, J.P., Shahabuddin, M., 1999. Plasmodium

gallinaceum ookinetes adhere specifically to the midgut epithe-

lium of Aedes aegypti by interaction with a carbohydrate ligand.

J. Exp. Biol. 202, 485.

B.D. Foy et al. / Journal of Immunological Methods 261 (2002) 73–83 83