of April 12, 2018.This information is current as

Antigen-Binding RepertoireSingle-Domain Antibody Contributes to the A Novel Promiscuous Class of Camelid

Raes, Patrick De Baetselier and Serge MuyldermansNick Deschacht, Kurt De Groeve, Cécile Vincke, Geert

http://www.jimmunol.org/content/184/10/5696doi: 10.4049/jimmunol.0903722April 2010;

2010; 184:5696-5704; Prepublished online 19J Immunol 

MaterialSupplementary

2.DC1http://www.jimmunol.org/content/suppl/2010/04/19/jimmunol.090372

Referenceshttp://www.jimmunol.org/content/184/10/5696.full#ref-list-1

, 11 of which you can access for free at: cites 47 articlesThis article

        average*  

4 weeks from acceptance to publicationFast Publication! •    

Every submission reviewed by practicing scientistsNo Triage! •    

from submission to initial decisionRapid Reviews! 30 days* •    

Submit online. ?The JIWhy

Subscriptionhttp://jimmunol.org/subscription

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/About/Publications/JI/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/alertsReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists, Inc. All rights reserved.Copyright © 2010 by The American Association of1451 Rockville Pike, Suite 650, Rockville, MD 20852The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

The Journal of Immunology

A Novel Promiscuous Class of Camelid Single-DomainAntibody Contributes to the Antigen-Binding Repertoire

Nick Deschacht,*,†,1 Kurt De Groeve,*,†,1 Cecile Vincke,*,† Geert Raes,*,†

Patrick De Baetselier,*,† and Serge Muyldermans*,†

It is well established that, in addition to conventional Abs, camelids (such as Camelus dromedarius and Lama glama) possess unique

homodimericHchainAbs (HCAbs)devoidofL chains.TheAg-binding site of theseHCAbsconsists of a singlevariable domain, referred

to as VHH. It is widely accepted that these VHHs, with distinct framework-2 imprints evolved within the V(H) clan III-family 3, are

exclusively present onHCAbs. In this study, we report the finding of a distinct leader signal sequence linked to variable genes displaying

a high degree of homology to the clan II, human VH(4) family that contributes to the HCAb Ag-binding diversity. Although the VHH

framework-2 imprints are clearly absent, their VH(4)-D-JH recombination products can be rearranged to theH chains of both classical

andHCAbs.This suggests that for theseVdomains thepresenceof aLchain to constitute theAg-binding site is entirelyoptional.As such,

the capacity of this promiscuous VH(4) family to participate in two distinct Ab formats significantly contributes to the breadth of the

camelidAg-binding repertoire. Thiswas illustrated by the isolation of stable, dendritic cell-specific VH(4) single domains from aVH(4)-

HCAbphagedisplay library. The high degree of homologywithhumanVH(4) sequences is promising in that itmay circumvent the need

for “humanization” of such single-domain Abs in therapeutic applications. The Journal of Immunology, 2010, 184: 5696–5704.

Camelidae are known to possess a dichotomous immunesystem that enables them to generate 1) classical Abscomposed of twoH and twoL chains and 2) Abs composed

of two H chains and no L chains. The latter Abs are referred to as Hchain Abs (HCAbs) (1, 2), and their H chain also lacks the firstconstant domain (CH1), which in a classical Ab interacts with theconstant domain of the L chain. The Ag-binding fragment of theseHCAbs is reduced in size to a single variable domain (referred to asVHH) and is generated from a V-D-JH gene rearrangement forwhich a distinct set ofV genes, theVHHgermline genes, is availablein the camelid genome (3–5). Different subfamilies have been dis-tinguished for these VHH genes, but all are closely related to thehuman VH(3) family of clan III (4, 6). AVHH(3) domain of drom-edary or llama HCAbs differs from the VH(3) domain of classicalAbs only by a few crucial substitutions of amino acids normallyinvolved in VL pairing (7, 8). These framework region 2 (FR2)mutations, Val37Phe/Tyr, Gly44Glu, Leu45Arg, and Trp47Gly [Kabatnumbering system (9)], are encoded in a dedicated subset of Vgermline genes of camelids and are therefore not the result of so-

matic hypermutation (2). Obviously, such hallmarkmutations in theFR2 will reshape the VL interface and abrogate a possible VH–VLassociation. These VHH hallmark residues are also considered to beimportant for the solubility of the autonomous VHH(3) domain (10,11) and were demonstrated to affect its stability (12, 13). However,classical Ag-binding single-domain Ab (sdAb) fragments of familyVH(3) (i.e., lacking the VHH characteristic hallmark residues in FR2)have been isolated occasionally from immune V libraries, althoughsome controversy remains about the exact IgG isotype (HCAbs orclassical) at its origin (14–17). Apart from the absence of the VH(3)hallmark residues, these VH(3) sdAbs also lack the typical interloopdisulfide bridge and the long CDR3 loops, yet they remain highlysoluble, stable, and functional. So far, it was assumed that all of thesevariable domains of HCAbs are part of the same clan III VH(3) family.In this study, we report the presence of novel variable domains ofclan II anddescribe their ability tobepart of classical aswell asHCAbs.Indeed, although all of theVdomains of this clanwere classicalAbVHsequences, they are able to generate functional HCAbs, so-called VH-HCAbs. Analysis of these novel clan II VH-HCAbs and sequencecomparison with human V sequences indicate a strong resemblancewith the human VH(4) family. We also report that the same VH(4)-D-JH recombination products can be rearranged to both camelid classicalandHCAbs. Therefore, the variable domain obtained after a VH(4)-D-JH recombination is functional as an Ag-binding entity as a singledomain or after combiningwith aVL as aVH–VL pair. Hence, the useof a L chain is entirely superfluous to generate Abs with these VH(4)genes. That these V domains are functional in Ag binding was dem-onstrated by the retrieval of dendritic cell-specific VH(4) sdAbs froma llama sdAb library derived from the VH(4)-HCAbs. These llamasdAbs exhibit good solubility and are the first autonomous single-domain Ab fragments reported from the VH(4)-HCAbs repertoire.

Materials and MethodsAnimals

Mice. Female C57BL/6J mice (6–8 wk old) were used from the in housebreeding facility. Animal care and treatment were conducted in conformitywith institutional guidelines of the Belgian Council for Laboratory Animal

*Laboratory of Cellular and Molecular Immunology, Vrije Universiteit Brussel; and†Department of Molecular and Cellular Interactions, VIB, Brussels, Belgium

1N.D. and K.D.G. contributed equally to this work.

Received for publication November 19, 2009. Accepted for publication March 8,2010.

This work was supported by the Instituut Voor de Aanmoediging van Innovatie doorWetenschap en Technologie in Vlaanderen (to N.D.). K.D.G. is a Ph.D. fellow of theHorizontale Onderzoeks Actie of the Vrije Universiteit Brussel.

Address correspondence and reprint requests to Dr. Serge Muyldermans, Laboratoryof Cellular and Molecular Immunology, Vrije Universiteit Brussel, Building E, Plein-laan 2, 1050 Brussels, Belgium. E-mail address: svmuylde@vub.ac.be

The online version of this article contains supplemental material.

Abbreviations used in this paper: BM, bone marrow; BMDC, bone marrow-deriveddendritic cell; CD, circular dichroism; CH1, first constant domain; DC, dendritic cell;FR, framework region; HCAb, H chain Ab; HIS, histidine; IMAC, immobilized metalion affinity chromatography; LB, Luria-Bertani; LS, leader signal; rmGM-CSF, re-combinant murine granulocyte macrophage colony-stimulating factor; sdAb, single-domain Ab; SEC, size exclusion chromatography.

Copyright� 2010 by TheAmericanAssociation of Immunologists, Inc. 0022-1767/10/$16.00

www.jimmunol.org/cgi/doi/10.4049/jimmunol.0903722

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

Science as inspired by the Federation of European Laboratory AnimalScience Associations recommendations.

Llama. The Lama glama were kept at the premises of the veterinary schoolof the University of Ghent. All of the animal experiments were conductedwith the approval of the Ethical Committee of the Faculty of VeterinaryMedicine (University of Ghent, Ghent, Belgium) (18).

Dromedary. The Camelus dromedarius was housed at the Central Veteri-nary Research Laboratory (Dubai, United Arab Emirates).

Cells

Splenic CD11c+ dendritic cells. Four female C57BL/6J mice were infectedwith 5000 Trypanosoma brucei brucei parasites (administered i.p. to recruitdendritic cells [DCs] to the spleen) and sacrificed 7 d postinfection. Spleenswere taken and treated with 400 U/ml collagenase III (Roche, Basel,Switzerland) for 30 min at 37˚C in HBSS (Gibco, Invitrogen, Paisley, U.K.)and neutralized by adding 2% heat-inactivated FCS (Hyclone, Logan, UT)and 2 mM EDTA (Invitrogen). Cells were treated with Tris-buffered am-monium chloride to lyse erythrocytes and passed through a nylon mesh toremove undigested material. The splenic myeloid cell fraction was enrichedby centrifugation (2000 rpm for 25 min at room temperature) on a Nyco-denz gradient (OD360nm = 0.660–0.670). The cellular fraction of the in-terphase was purified by positive selection on magnetic separation columns(MACS) with CD11c+ beads (Miltenyi Biotec, Gladbach, Germany) ac-cording to manufacturers’ recommendations.

Murine bone marrow-derived CD11c+ DCs. Femursand tibiaswere removedand mechanically purified from surrounding tissues. Subsequently, epiphyseswere removed, and bone marrow (BM) was flushed using RPMI 1640 (Invi-trogen). Erythrocytes in the single-cell suspensions were lysed as mentionedabove. FreshBMcells were cultured in 10mlR10mediumconsisting ofRPMI1640 (Invitrogen) supplemented with penicillin (100 U/ml; Invitrogen),streptomycin (100 mg/ml; Invitrogen), 2-ME (50 mM; Sigma-Aldrich, St.Louis, MO), and 5% heat-inactivated FCS (Hyclone). At day 0, BM cells wereseededat 23106 cells per 100-mmdiameter bacteriological Petri dish (BectonDickinson) in 10mlR10mediumcontaining 200U/ml (20 ng/ml) recombinantmurine granulocyte macrophage colony-stimulating factor (rmGM-CSF) (giftfrom laboratory of Prof. K. Thielemans). At day 3, another 10ml R10mediumcontaining 200 U/ml rmGM-CSFwas added to the plates. At day 6, half of theculture supernatant was collected and centrifuged, and the cell pellet was re-suspended in 10 ml fresh R10 medium containing 100 U/ml rmGM-CSF andpoured into the original plate. Reducing the concentration of the rmGM-CSFduring culture diminishes the granulocyte contamination. Harvesting the cells1 d later yielded the immatureDCpopulation.For completematurationofDCs,the 7-d-old nonadherent cells were collected by gentle pipetting, centrifuged at1500 rpm for 5 min at room temperature, and resuspended in 6 ml fresh R10medium into a fresh six-well plate containing 100U/ml rmGM-CSF and 1mg/ml LPS (Sigma-Aldrich). Cells were then cultured for an additional 1 or 2 d toobtain the mature DC population (19).

RACE of dromedary IgG

PBLs were isolated from dromedary or llama using Lymphoprep (Axis-Shield,Oslo, Norway), and mRNA was extracted (20). Then, with a SMART RACEcDNA amplification kit (BD Clontech, Heidelberg, Germany), cDNAwas gen-erated, and 59RACEPCRproductswere amplified by using a FR4-specific prim-er (59-GGACTAGTGCGGCCGCTGGAGACGGTGACCTGGGT-39). These 59RACE products were cloned into the pCR2.1-TOPO vector using the TOPOTAcloning kit (Invitrogen), and insert-positive clones were identified for sequencingby blue–white selection. Subsequently the 59 ends encoding our dromedary andllama leader signal (LS) sequences and V elements were aligned to the corre-sponding sequences of the functional human IgV germline genes (21), as thehuman IgH locus is well established (22, 23).

cDNA analysis and phylogenetic study

On the basis of the 59 RACE-positive clones, a novel primer (CALL5) forthe new LS sequence was designed (59-TGGTGGCAGGTCCCCAAGGT-39). Standard PCR with CALL5 and a CH2-IgG–specific primer (59-GGTACGTGCTGTTGAACTGTTCC-39) was performed on llama cDNA(PBLs). The two amplification products (6700 and 1000 bp) were gel-purified, cloned separately into TOPO TA, and identified for sequencing asabove. Alignments were calculated with CLUSTAL X (version 1.8) usingthe GONNET series protein matrices. From this alignment, rooted phy-logenetic trees were constructed by using the neighbor-joining method ofMEGA 4.0 (24). The amino acid sequences of FR1–FR3 as defined by theKabat numbering system (9) were used to generate the tree. From theamino acid sequences of the VH(4)s, the isoelectric point (pI) andweighted hydrophobicity were calculated by Expasy proteomics (25).

Llama immunization and VH(4) library construction

A llama (L. glama) was injected s.c. with 108 immature mouse BM-derivedDCs (BMDCs) without adjuvant. The immunization was repeated six timesat weekly intervals. After the last boost, anticoagulated peripheral bloodwas taken from the jugular vein to prepare an immune library. Genefragments from the V region leader sequence up until the CH2 domain ofthe IgGs were first amplified as described above. The PCR fragmentscontaining the variable gene fragments of the IgG HCAbs were purifiedfrom agarose gels. Subsequently, the gene fragments encoding the entireV region were amplified with one additional PCR with nested primersannealing at FR1 (59-CCAGCCGGCCATGGCTCAGGTGCAGCTGGT-CCAGTCTGG-39) and FR4 (59-GGACTAGTGCGGCCGCTGGAGAC-GGTGACCTGGGT-39). The final PCR fragments were ligated into thephagemid vector pHEN4 (20), using the restriction sites NcoI and NotI(underlined). Ligated material was transformed in Escherichia coli cells(TG1) and plated on selective medium (16). The colonies were scrapedfrom the plates and washed, and this cell library was stored at 280˚C inLuria-Bertani (LB) medium supplemented with glycerol (50% final con-centration).

Selection of Ag-specific sdAb fragments

A representative aliquot of the library was cultured in LB medium withampicillin and infected with M13K07 helper phages (16) to express thecloned V region at the tip of phage particles as fusion products with thegene III bacteriophage coat protein. DC-specific V domains from this li-brary were enriched using whole-cell panning on an in vivo cell populationof CD11c+ monocytes from the spleens of mice (26). Freshly harvestedsplenic DCs (1 ml, 4 3 107 cells) were incubated with phage particles(1.5 3 1011) for 1 h at room temperature. Cells were then pelleted by low-speed centrifugation and resuspended in 2 ml MACS buffer, and anti-CD11c beads (Miltenyi Biotec) were added to the suspension. Once mixedwith the beads, cells were then isolated via CD11c+ selection by MACSaccording to the manufacturer’s protocol (Miltenyi Biotec). After the finalwash, the column was removed from the magnet, and the bead-coated/phage-coated/CD11c+ cells were eluted from the column with 5 ml sterilePBS. The CD11c+ cells were immediately centrifuged at 1800 rpm for 10min and resuspended in 200 ml triethylamine (pH 10) to elute cell-boundphages and lyse cells. After 10 min incubation at room temperature, thesolution of phage particle eluate and cellular debris was neutralized with200 ml Tris-HCl (1 M, pH 7.4) and used to infect exponentially growingE. coli TG1 cells that were further cultured for the second round of se-lection. An aliquot of the neutralized phage particle eluate was also takento infect bacteria and plated on LB agar plates supplemented with ampi-cillin. After each round of panning, individual colonies were picked ran-domly from the titration plate, and their sdAb gene was sequenced.

Purification and identification of DC-specific sdAb fragments

Purification of the sdAb. The variable regions of the clones that appearedmultiple times after panning in our sequence analysis were recloned into thepHEN6 expression vector (27) using the restriction enzymes NcoI andEco91I (Fermentas) and transformed into E. coli WK6 cells. The V gene,under control of the lac promoter, is preceded by a PelB sequence to directthe translation product to the periplasm and followed by a His6 tag tofacilitate the purification of the sdAb protein by immobilized metal ionaffinity chromatography (IMAC). Liter-scale production and purificationof these sdAbs by IMAC and gel filtration (Sephadex 75 column) in PBSwere performed as described previously (16).

Biophysical stability. To analyze the oligomeric state of the sdAbs, the VH(4)sdAbs and two (family 3) sdAbs (28)were expressed and subsequently purified(16). The purified monomeric fractions of the sdAbs in PBS at two differentconcentrations (35mM [0.5 mg/ml] and 175mM [2.5mg/ml]) were incubatedat 37 and 4˚C for up to 14 d. Samples (0.5 mg of each sdAb) were taken atdifferent time points (day 0, 1, 7, and 14), and their integrity was analyzed bysize exclusion chromatography (SEC) on an AKTA Explorer (GE Healthcare,Munich, Germany) using Sephadex 75 in PBS and by circular dichroism (CD)measurement. CD spectra were obtained with a Jasco J715 spectropolarimeterin the UV region (200–350 nm), using a protein concentration of 0.2 mg/mland a 0.1-cm cell path length. Data were acquired with a reading frequency of1/20 s21, a 1 s integration time, and a 2 nm bandwidth. Data analysis wasperformed according to Dumoulin et al. (13).

Temperature-induced denaturation. Thermal unfolding was followed by CDwith a Jasco J715 spectropolarimeter. A total volume of 300 ml of eachsample was heated in 50 mM sodium phosphate buffer (pH 7). The tem-perature was increased from 35 to 95˚C at a rate of 1˚C/min, and the signalintensity at 205 nm was recorded as a function of temperature.

The Journal of Immunology 5697

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

Flow cytometry. For flow cytometric analysis, the DC population was stainedwith Abs (from BD Biosciences [San Jose, CA], unless otherwise stated) for20 min at 4˚C using standard protocols. Cells (106 per 100 ml) were incubatedwith 1 mg anti-FcgR Ab (clone 2.4G2) before adding allophycocyanin-conjugated CD11c (HL3) Ab and 1 mg FITC-conjugated MHC II (clone M5/114.15.2), FITC-conjugatedCD86 (Gl-1), or FITC-conjugatedCD40Ab (clone3/23). Cell recognition of different sdAbs (at 1mg) wasmonitored using 0.1mgFITC-conjugated anti-histidine (HIS) (Serotec, Oxford, U.K.) as a secondaryAb. Multiparameter acquisition was performed on an FACSCanto II (BD Bio-sciences), followed by analysis with FlowJo (TreeStar, Ashland, OR).

Alexa Fluor 488 labeling of sdAb. The sdAbs were labeled using the AlexaFluor 488 protein labeling kit (Molecular Probes, Eugene,OR). Briefly, 100mlof 1 M sodium bicarbonate buffer was added to 1 ml sdAb solution at a con-centration of 1 mg/ml. The sdAb solution (either sdAb_31 or sdAb_32) wassubsequently transferred to a vial of reactive dye (Alexa Fluor 488 carboxylicacid, tetrafluorophenyl ester) and incubated overnight at 4˚C under mild agi-tation. Labeled sdAb_31 and sdAb_32 were separated from unreacted dye bySEC on Sephadex 75 in PBS. The protein concentration and the degree ofAlexa Fluor 488 labeling of the sdAb was determined via spectrophotometry(Nanodrop ND-1000; Isogen Life Sciences, De Meern, The Netherlands).

In vitro competition study. Each BMDC population (106 cells in 100 ml)was preincubated with 40 mg unlabeled sdAb (sdAb-BCII10, sdAb_31 orsdAb_32) for 20 min at 4˚C. The cells with bound sdAb were recovered ina pellet by low-speed centrifugation (supernatant containing the unboundsdAb was removed), resuspended in PBS, stained with 1 mg Alexa Fluor488-labeled sdAb_31 or sdAb_32, and incubated for 20 min at 4˚C.Multiparameter acquisition was performed on an FACSCanto II (BectonDickinson), followed by analysis with FlowJo (Tree Star).

ResultsIdentification in camelids of a distinct LS linked to clan IIvariable genes

A 59 RACE on cDNA of llama or dromedary PBLs with a JH-specific primer (i.e., FR4 of VH) and a 59 universal primer was used toinvestigate theV repertoire. This 59RACE resulted in a PCR fragmentof∼500 bp that was cloned into TOPOTA. Sequence analysis of thesecloned inserts revealed a few clones encoding a LS sequence thatdeviates from the established LS sequence of camelid VH(3) andVHH(3) genes (3) (Supplemental Fig. 1). The V region downstreamof the novel LS was aligned to the human IgV genes (23). Whereasprevious reports assigned the dromedary V genes to the VH(3) familyof clan III (3, 4), the International ImMunoGeneTics informationsystem amino acid numbering/V-QUEST (www.imgt.org) and Ig-Blast (www.ncbi.nlm.nih.gov/igblast) search allocated the novel LSand its V sequences to the human V genes of clan II.Note that onprevious occasionswealways amplified theVHHgenes

withLS-specificprimersoftheVHH(3)familythatwouldfailtoamplifythesenewVgenes,and therefore theseVgeneswereeliminatedfromallofour immuneVlibraries (15).Hence, anewprimer (CALL5)basedonthis novel LS (Supplemental Fig. 1) was designed. Interestingly, thePCR amplification with CALL5 and a CH2-IgG–specific primer gen-erated two distinct amplicons (Fig. 1). From their sizes, it can be in-ferred that the longest fragment (61000 bp) corresponds to theH chainfragment of classical IgGAbs,whereas the shorter fragment (6700bp)probably encodes a fragment of the H chain of the HCAbs. The dif-ference in size accounts for the absence or presence of the CH1 regionthat occurs only in H chains of classical IgGs (29). Fragments of both,the HCAbs and the classical Abs, were gel-purified, cloned in-dependently in TOPOTAvector, and sequenced. These particular clanII variable genes occurred in the HCAb of all camelid samples tested,although the band intensity varied and the fraction in the classical Hchain was always more pronounced (Fig. 1). Surprisingly, a 59 RACEIgG analysis of the classical H chains demonstrated that the VH(4)genes contributed up to 50% of the classical Ab repertoire.

Phylogenetic sequence analysis

A total of ∼150 individual clones, randomly chosen over bothclonings (i.e., classical and HCAb-derived), were sequenced. After

removing ambiguous sequence readings, nonfunctional sequences(containing a reading frame shift), or incomplete sequences, weassembled two sets of 58 V gene sequences for the H chain of theclassical and HCAb IgGs, respectively (Supplemental Figs. 2, 3). Se-quence alignment demonstrated that the V regions of both Ab classesbelong to the same V gene family and both groups of V regions en-coded classical VH elements (not VHH). Furthermore, a neighbor-joining analysis was performed with the V-encoding cDNA sequencesand the functional human IgV germlines. Analysis of these llama Velements (of classical Abs and HCAbs) showed that they all cluster inone branch that is most closely related to the branch of the human VH(4) family, which forms part of clan II (Fig. 2). The llama VH(4) andhuman Ig VH(4) domains display high sequence similarity, yet someframework residues (Thr30, Ile48, Thr66, Arg71, Gln81, Pro84, andGlu85)deviate regularly from the humanVH(4) sequences (23). The impact ofthese deviations is visualized by the hydrophobicity plot (Fig. 3). Thedecreased hydrophobicity profile at the outer framework-3 (asterisks inFig. 3) for the llama VH(4)s compared with that of human VH(4) issuggestive for a reduced aggregation-prone behavior for llama VH(4)relative to that reported for human VH(4) (30).

VH(4) enlarges both the classical and the HCAb diversity

Analysis of VH(4) sequences from both isotypes (i.e., classical Absand HCAbs) showed that sequence variation was the highest in theCDRsof theVregions (SupplementalFigs. 2, 3).Variabilitywasalsonoticed in theFR3 regions, although to a lesser extent comparedwiththat of the CDRs. No obvious differences in CDR1 sequence areobserved between VH(4)s from classical Abs and HCAbs (Sup-plemental Figs. 2, 3). With the notable exception of three clones(conv12,32,andHCAb28),bothVH(4)groupscontain, similar to thehuman VH(4), 7 aa in their CDR1. Such a CDR1 loop of 7 aa ispredicted to adopt canonical structure type-3 (31). The HCAb28clone possesses an enlarged CDR1 and forms potentially a novelloop structure, whereas the other two classical VH(4) clones (clones12 and 32) probably adopt canonical loop structure type-1 and type-2, respectively. Likewise, no major differences are detected forCDR2 between VH(4)s of both groups (Supplemental Figs. 2, 3).The vast majority of the CDR2 loops display a length of 16 aa,except clones conv56 and conv57 (Supplemental Fig. 2). Therefore,with the exception of these clones, theCDR2 regions of llamaVH(4)are predicted to fold into the known canonical loop structures of thehuman VH(4) domains (31). In summary, the CDR1 and CDR2Ag-binding loops of llama VH(4) apparently adopt the canonical loopstructures observed for human VH(4) domains. In sharp contrast,these loops are routinely deviating from the canonical confor-mations in the case of VHH of V family 3 (11, 32). Nonetheless it isobvious that the mere presence of these VH(4) domains within the

FIGURE 1. H chain amplification of IgG of two different L. glama.

Transcripts encoding the classical IgG and the H chain IgGs were am-

plified by PCR with primers specific to the clan II leader sequence and the

CH2 exon. A m.w. size marker was applied in the left lane.

5698 NOVEL CAMELID FAMILY 4 SINGLE-DOMAIN Abs

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

HCAb type enlarges the known structural repertoire of these uniquecamelid Abs (4). In contrast to CDR1 and CDR2 that is imprinted inthe V germline gene, CDR3 arises from an imprecise V-D-JH re-combination, which creates a much larger variability in loop lengthand sequence. Within the VH(4) dataset, CDR3 ranges from 7 to 22aa (Supplemental Figs. 2, 3), although the average length of CDR3from classical Abs (12.8 aa) does not deviate significantly from thatof HCAbs (13.5 aa). In contrast, for dromedary VH(3) and VHH(3),a net difference in CDR3 length is observed (9 and 15 aa, re-spectively) (33), although this difference is less pronounced in llama(9 and 13 aa) (8).From inspection of the CDR3 sequences, it appears that the VH

(4) sequences of classical Abs and HCAbs are recombined to acommon set of D and JH genes, although such analysis is com-plicated by imprecise recombination and somatic hypermutation. Aminor difference is that VH(4) of the classical Ab group lacks inter-or intraloop disulfide bridges (Supplemental Figs. 2, 3). In contrast,for the HCAb VH(4) sequences, clones 36 and 12 possess twocysteine residues in their CDR3 loop that probably form anintraloop disulfide bond.Inconjunction,nomajordifference insequenceor loopstructurewas

discerned between the VH(4) from classical Abs and HCAbs (Sup-plemental Figs. 2, 3). Interestingly, three different VH(4) sequences

were found twice, once in the classicalAbgroup andonce in theHCAbgroup (clones HCAb10/conv46, HCAb8/conv19, and HCAb52/conv7). When aligned pairwise (Fig. 4A), these VH(4)s containeda minimal amount of point mutations and a homologous CDR3 se-quence, indicating that each pair probably originates from an identicalV-D-JH recombination. One sequence of each pair was linkedwith (orproperly spliced to) an intact CH1 region and will therefore be part ofanAbwith anAg-binding fragment comprising aVH andVLdomain.The twin sequence partner definitely lacked theCH1 region, and theVregion was immediately followed by the hinge sequence, which dem-onstrates that this V domain was derived from a HCAb devoid of Lchains. Therefore, this latter VH(4) domain will function as a single-domain Ag-binding fragment in a HCAb. This indicates that a VH(4)gene involved in a V-D-JH recombination in a pre-B cell generatesa VH domain that might function either independent from a VL ina HCAb or paired with a VL domain in a classical Ab.

Generation of VH(4) library and selection of Ag-specificsingle-domain Abs

To verify that the isolated VH(4) domains from HCAbs (i.e., in theabsence of a VL partner) are indeed functional in Ag binding, wegenerated a VH(4)-HCAb–derived single-domain phage displaylibrary of a llama immunized with immature BMDCs. Two roundsof panning were performed on purified splenic CD11c+ cells (26).After the second round of panning, the number of phage particleseluted from CD11c+ cells increased 100-fold relative to the firstround using the same amount of input phage particles and iden-tical washing steps. Individual bacterial colonies (6100) from thesecond round were randomly picked and sequenced. Two se-quences (referred to as sdAb_31 and sdAb_32) were representedrepeatedly, although several point mutations throughout the frame-work or CDR occurred in the homologous clones of the domi-nating sdAb_31 sequence. The alignment of the deduced aminoacid sequences, against the human VH consensus sequence (Fig.4B) confirmed their VH(4) signature. The CDR3 lengths of 14 and15 aa for sdAb_31 and sdAb_32, respectively, contrast with theshorter average CDR3 length of llama VH(3) (8). In addition,cysteine residues were absent in the CDR, and therefore these longloops cannot be tethered by an inter- or intraloop disulfide bondthat might assist in the stabilization of the sdAb (34).

FIGURE 2. Neighbor-joining phylogenetic tree of human IgV germline

genes and llama VH sequences. The tree is based on the amino acid se-

quences of FR1–FR3 as defined by the Kabat annotation (9). Sequences of

llama are represented by gray branches; human sequences in black rep-

resent human IgV germline genes as specified by the gene family anno-

tations. The three designated clans are represented by Roman numbers.

FIGURE 3. Weighted Kyte-Doolittle hydrophobicity plot for VH(4)s.

Weighted hydrophobicities (y-axis) were calculated for the distribution of

amino acids observed at each position (x-axis) of the human VH(4)

germline genes (gray line) and the llama rearranged conventional and

HCAb VH(4)s (black line). The asterisks depict the areas in FR3 with the

most notable differences. In the Kyte-Doolittle scale, the higher positive

values correspond to the more hydrophobic amino acid stretches.

The Journal of Immunology 5699

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

Biochemical properties of Ag-specific VH(4) sdAb

The twoVH(4) sdAbswere recloned in the pHEN6expressionvector.After induction, the periplasmic proteins were extracted, and therecombinant protein was purified by IMAC and SEC (SupplementalFig. 4). Ayield of 5–7 mg protein per liter of culture was obtained forthese sdAbs, which is within the range (5–10 mg/l culture) normallyobtained for sdAbs of the VHH(3) family (15). The sdAb_31 elutedfromaSuperdex 75 gel filtration columnas a single symmetrical peakas expected for a soluble, monomeric protein, whereas the elutionprofile of the sdAb_32 showed an additional minor fraction of di-merizedmaterial. The thermal stability of the sdAbswasmeasured byCD, andTmvalues of 62.4 and 63.5˚Cwere recorded for sdAb_31 andsdAb_32, respectively (Supplemental Fig. 5). Although these Tmvalues are similar to thevaluesmeasured forVHH(3) sdAbs (13), theyare high for sdAbs belonging to theVH(4) family, becauseVdomainsof the human VH(4) family form consistently less stable domains(30). Earlier studies proposed for sdAbs a direct correlation betweentheir net acidic charge and aggregation resistance (35, 36). Therefore,we calculated the theoretical pI of VH(4)s of both Ab groups. Thedistribution of the individual pIs of VH(4)s from classical Abs orHCAbs as well as the pI ranges are indistinguishable (SupplementalFig. 6A). Remarkably, in this current dataset of VH(4)s of unknownspecificity, no pI values between 7.0 and 7.5were recorded. However,the DC-enriched sdAb_32 and sdAb_31 display a near neutral andbasic pI, respectively, and are soluble and monomeric entities, insharp contrast to predictions from previous work (35, 36).We next analyzed the integrity and aggregation tendency of our

twoDC-specificVH(4) sdAbs over time, relative to the properties ofVHH(3)-derived sdAbs and DC-specific VH(3) sdAbs (28). Thesefour sdAbs at two different concentrations (35 and 175 mM) wereincubated at 4 and 37˚C for different time intervals up to 14 d. Gelfiltration was used to profile the quaternary structure of the sdAbs,and the CD signal was taken to ascertain the maintenance of theirstructural integrity. The VH(4) sdAbs eluted as a single symmetricpeak at all incubation times, and their chromatographic profileswere indistinguishable from those of sdAbs of family 3 (Supple-mental Fig. 7A–F). The prolonged incubation of the sdAbs at 37˚Cdid not provoke oligomerization because their profiles were iden-tical to those of the sdAbs kept at 4˚C. The VH(4)s at differentconcentrations also yielded identical profiles, and even the VH(4)sdAbs at the highest concentration (10 mg/ml) remained mono-meric (data not shown). In addition and in agreement with the gelfiltration results, the incubation condition (concentration, time, and

temperature) had no effect on the CD profile of the sdAb (Sup-plemental Fig. 7G, 7H). We observe that under these conditions noprominent changes to the secondary structure are seen and thatcamelid VH(3), VHH(3), and VH(4) sdAbs appear robust.The Ag-binding profiles of sdAb_31 and sdAb_32 on mature and

immature BMDCs were analyzed by flow cytometry (Fig. 5A). Theresults of this in vitro binding study demonstrate that sdAb_31 andsdAb_32 are functional in Ag binding, and as expected from theresults above, even after 2 wk incubation at 37˚C, our sdAbs re-mained functional (data not shown). Their epitope is present onCD11c+ cells, including both mature and immature DCs (Fig. 5A).The observed signals are due to the sdAbs recognizing specificallytheir cognate cell surface markers, because control experiments ofBMDCswith sdAb_BCII10, a sdAbwith specificity forb-lactamase,gave background signals that were indistinguishable from thoseobtainedwith cellsmixedwith FITC-labeled anti-HIS secondaryAb.In addition, the experiments with other non-DC-specific VH(4)sdAbs all failed to give a fluorescence intensity shift (Fig. 5A). Thisprecludes that the signal originates from a nonspecific adsorption ofVH(4) sdAbs on the BM-DCs. Furthermore, the flow cytometrysignal intensity with the sdAbs directed against the immature DCs isslightly higher than that for mature DCs, and the relative shift of thefluorescence intensity peak is more pronounced for sdAb_32 thanthat for sdAb_31. Finally, experiments with Alexa Fluor 488-labeledsdAb tested in competition with an excess of unlabeled sdAb con-firmed the specific recognition of their cognate epitope and indicatedthat sdAb_31 and sdAb_32 recognize different (i.e., noncompeting)epitopes on their target cells (Fig. 5B).

DiscussionThe Ab generation process in Camelidae should deviate, at least insome details, from the general scheme in other vertebrates becausecamelids are actively producing unique, functional HCAbs. For theproduction of HCAbs, the camelids use dedicated Cg germline geneswith a truncated 59 splicing site at the “CH1 hinge” intron so thatsplicing of the RNA transcript leads to a direct connection of the Vexon to the hinge exon (Fig. 6) (3, 5, 37). In addition, the camelidgenome contains two sets of V germline genes (VH and VHH), pro-bably interspersed in their H locus (5, 33), and downstream clusters ofD genes and JHgenes (4, 5, 37). In our previous reports, wementionedonly the presence of VHHs belonging to the V clan III, family 3 (4).During Ab generation the rearranged VHH(3)-D-JH is eventu-

ally coexpressed with either the dedicated Cg2 or Cg3 and folds in

FIGURE 4. Amino acid sequence alignment VH(4) variable domains, numbered according to Kabat (above) and International ImMunoGeneTics in-

formation system amino acid numbering (below), with framework regions and CDR indicated (9, 21). A, Sequence alignment of three variable domains

[clan II, family VH(4)] present on llama classical IgG and H chain IgG. Differences between the sequences are underlined and in bold. B, sdAb_31 and

sdAb_32 are aligned with human IgHV4_30-4p01 (Z14238) and human IgHV4_39p03 (305715), respectively, because these sequences display the highest

nucleotide identity according to the IgBlast analysis. Differences in sequence from the reference human sequence are underlined and in bold.

5700 NOVEL CAMELID FAMILY 4 SINGLE-DOMAIN Abs

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

a homodimeric Ig. In contrast, VH(3)-D-JH is cotranslated withthe classical Cg1a or Cg1b, and the polypeptide associates with aL chain to form a classical Ig (Fig. 6). However, it has been re-ported that occasionally VH(3)D-JH products lacking the CH1exon can be cloned from cDNA, indicating that they probably arepart of the HCAb pool (14, 16). The majority of such classical Vdomains of a HCAb are (most often) generated after an unusual D-JH gene recombination whereby the codon for the conservedTrp103 (a residue that is essential for VL pairing in classical Abs)is substituted mostly by an arginine codon (6, 14, 16). In thisstudy, we demonstrate that the V repertoire of camelids alsocontains functional V genes that belong to clan II. These V genescluster with the Ig VH(4) family in the phylogenetic tree and theIgBlast search (Fig. 2). The occurrence in alpaca of such geneswas briefly mentioned recently (5). Interestingly, in llama, the VH(4) gene is transcribed in nearly 50% of the expressed classicalIgGs (possessing CH1 and by default also the L chain). Thispercentage comes from a 59 RACE cloning of the H chain ofclassical Abs and shotgun sequencing that was not biased for anyparticular V element. The abundant usage of VH(4) within this

pool suggests that these V genes fulfill a central role in formingthe Ag-binding repertoire of classical Abs in llama. The reasonwhy their presence was overlooked until now is dual: first, mostwork on camelid Abs focuses on the HCAbs and neglects theclassical Abs (14, 15). Secondly, the V genes are normally am-plified after PCR with primers designed to anneal at the LS se-quence or at the FR1, and these primers fail to hybridize to thecorresponding regions in the VH(4) family. In this study, wedemonstrate that this VH(4) family contributes significantly to theclassical Ig Ab repertoire and, more importantly, is numerouslypresent as part of HCAbs as well.Nonetheless, the frequency of the VH(4) occurrence within the

HCAbpool is lower than that in the classicalAb group. Interestingly,the very same VH(4)-D-JH rearrangement can occasionally befound within both classical Abs and HCAbs. In the former case, it iscoexpressed with a CH1 domain, and probably the VH domain willpair with a VL domain to constitute a functional Ag-binding site. Inthe latter case, the CH1 exon is definitely removed duringthe splicing reaction within the mRNA of the HCAb-specific Cg2or Cg3 genes and as such impeding a possible association with

FIGURE 5. Binding profile of VH(4) sdAb_31 and sdAb_32 on BMDCs. The activation state of the DC populations was evaluated using different cell

surface markers (CD40, CD86, andMHC II). A, DC populations were stained with 1 mg sdAb-BCII10 [VHH(3); negative control], a nonspecific VH(4) sdAb,

or VH(4) sdAb_31 or sdAb_32 visualized by 0.1 mg anti–HIS-FITC Ab gated on CD11c+ BMDCs (solid black line and open profiles). The profiles in dashed

lines and gray filling are from cells treated with 0.1 mg anti–HIS-FITC only. B, Immature DCs preincubated with 40 mg unlabeled sdAb-BCII10, sdAb_31, or

sdAb_32 (gray-filled, dashed-lined contoured profiles) followed by staining with 1 mg sdAb_32-Alexa Fluor 488 (open, solid-line contoured profiles).

The Journal of Immunology 5701

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

a L chain. Because the VH(4)-D-JH translation product seems to becompetent to function on both types of Abs, the role of the L chain(and the VL in particular) appears to be superfluous for the ex-pression of the BCR and for proper Ag recognition. Consequently,the VH(4) domain is forced to act as an autonomous, single-domainAg-binding fragment in a HCAbmolecule. Hereto, the stability andsolubility of the VH(4) domain and its secretion from the B cellshould be assured, and the Ag recognition of the autonomous do-main should be guaranteed.The dual usage of VH(4) raises two questions. First, is VH(4) of

the HCAbs a particular subset of the VH(4)-D-J rearrangementproducts or are all of the VH(4) domains capable to appear on eitherAb type? Second, what makes these camelid VH(4) different fromthe human VH(4) so that they can function as an isolated sdAbwhereas a humanVH(4) ismuch less adapted to fulfill this role (30)?With regard to the first question, VH(4) on HCAbs and classical

Abs seems to be indistinguishable. There is apparently no differ-ence in usage of the VH(4) or JH genes. In both cases, they re-arrange with the D gene to generate CDR3s of equal lengths. Also,the amino acid content and hydrophobicity of CDR3 within the VH(4) domains of both Abs reveal no significant bias (SupplementalFig. 6). Additionally, the distribution of the pI of the VH(4)domains from HCAbs or classical Abs is identical. Remarkably,few domains are formed with a pI between 7.0 and 7.5, which mightreflect a counterselection against such domains because thesewould be more prone to aggregation at a physiological pH. We alsoanalyzed areas on the VH(4) domain, such as FR2 and Trp103 (firstamino acid of FR4), where VH(3)s and VHH(3)s of HCAbs andclassical Abs are distinct. Yet again, FR2 of VH(4)s from HCAbsand classical Abs are identical. Although a few sequences fromthe HCAb pool show some amino acid substitutions in this con-served VL interface region, similar substitutions were noted forthe classical Abs as well (Supplemental Figs. 2, 3; HCAb23,HCAb24, and conv33, respectively). The Trp103 (encoded by the59 end of the JH gene) is another major component of the VL-contacting side of a VH domain. Indeed, for the VH(3)-derived

Ag-binding fragments of HCAbs, we observe regularly an argi-nine at this position, which will abrogate the VL pairing (7, 38).Remarkably, the Trp103Arg substitution is highly employed inVH(3)-HCAbs having a short CDR3, whereas a small number ofVH(3)-HCAbs have a long CDR3 loop but lack this Trp103Argsubstitution. This indicates that the CDR3 amino acid content inconjunction with the long sequence that collapses on the erstwhileVL interface mimics the Trp103Arg effect. The situation for VH(4) is different. The VH(4) of HCAbs and classical Abs possessesa CDR3 loop of equal average length, and in contrast with the VH(3)-HCAbs the majority of the VH(4)-HCAbs do normally not useArg103 substitution to impede the VH–VL association.Concerning the second question, we searched for possible

determinants that enhance the solubility and good biophysicalproperties of the isolated camelid VH(4) domain compared with thepoor behavior of the isolated humanVH(4) (30). These determinantsshould belong to theCDRs or to the framework, and in the latter casepreferably the VL binding interface (39–42). Although the sub-stitution of Trp103, Leu45, or both of FR2 by a charged residue (ar-ginine) may be favorable for the solubility of sdAbs, previousstudies clearly demonstrate that it is not a prerequisite (17, 36, 41,43). Despite the high sequence similarity between human and llamaVH(4), profound sequence analysis of the camelid VH(4)s high-lighted a few residues that deviate subtly but consistently from thehuman VH(4) framework. Remarkably, these residues are mainlypresent in FR3 (Supplemental Fig. 4) outside the VL binding face ofthe domain and render this area more hydrophilic than the humanVH(4)s. Because framework residues influence the biochemicalproperties of sdAbs (44–46), it is conceivable that the unfavorableproperties described for human single domains of familyVH(4) (30)are attenuated through these substitutions. However, it would besurprising that these differences are the sole determinant for theautonomous character of the camelid VH(4)-HCAbs. Additionalkey residues are probably located in the CDR loops because theseregions are reported to participate in the soluble behavior of au-tonomous human VH(3) domains (36, 41, 43, 47). Furthermore, the

FIGURE 6. Schematic representation of the H locus with V genes of VH(4) and VH(3) family followed by D and JH gene clusters and constant genes

(only Cg genes are shown). Note that the current organization of V genes (clustered or interspersed) is currently unknown, and therefore the order of VH(4),

VHH(3), and VH(3) is entirely tentative. The dedicated HCAb Cg2 and Cg3 have a deficient CH1 splicing site denoted by a star. A, (i) Generation of

classical IgG is obtained after rearrangement of a classical VH(3) gene and coexpression with classical Cg(1a,1b) genes, (ii) generation of HCAbs is

possible when a classical VH(3) gene recombines with a D-JH that encodes an Arg103 mutation (although sometimes Arg103 is absent) and coexpresses with

the dedicated HCAb Cg2 and Cg3 genes, and (iii) the majority of HCAbs are formed when a VHH(3) gene that differs from the VH(3) genes in its hallmark

FR2 amino acids is linked after a D-JH rearrangement to the dedicated Cg2, Cg3 constant genes. B, Generation of HCAbs with a VH(4) gene without the

Trp103 substitution can be joined to both sets of Cg genes to produce classical Abs or to produce HCAbs.

5702 NOVEL CAMELID FAMILY 4 SINGLE-DOMAIN Abs

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

nature of the CDR3 loop has been shown to contribute to the sta-bilization and solubility of sdAbs (30, 40). However, the inherenthypervariable character (in length and sequence) of the CDR re-gions precludes the identification of the critical residues to convertVH–VL domains into functional monomeric sdAbs.That the VH(4) sdAbs are indeed functional in Ag binding was

shown by immunizing a llama with DCs, cloning the VH(4) rep-ertoire from the HCAb-specific pool, and retrieving DC-specificsdAbs after phage display.Of note, in the absence of any informationon a possible Ag preference of the VH(4) of HCAbs, we preferred toimmunize with cells as opposed to individual proteins, becausetheoretically these harbor an infinite number and complex type ofepitopes, which maximizes our chances to retrieve binders after-ward. In addition, the retrieved VH(4) sdAbs produce readily inbacterial expression systems, display good solubility, thermal sta-bility, and shelf life, and are functional in the recognition of CD11c+

cells as measured by flow cytometry. Furthermore, the VH(4)s ofcamelids are closely related and share a high degree of identity withhuman VH(4)s. It is therefore conceivable that for human therapyonewould prefer to select specifically VH(4)-derived sdAbs insteadof VHH(3)-based sdAbs because the latter might require moredrastic humanization efforts to minimize immunogenicity (12).We also foresee two extra possible developments whereby our

insights on camelid autonomous VH(4)s could become practical inthe future. In one example, we propose to remediate an “unstable”human VH(4) that was identified (either as an autonomous sdAb orpaired with L domains in a scFv or Fab format) by substituting itsFR3 beginning and ending sequences with the more hydrophilicamino acids as those of the llama VH(4). For the second possibility,a strategy could be developed whereby a synthetic library would beconstructed by randomizing the CDR3 sequence on a llama VH(4)sequence that occurred in a HCAb and a classical Ab. After re-trieving Ag binders from this sdAb library, one could supply thisparticular sdAb with a library of (human) VLs and search for VH–VL Ag-binding molecules of enhanced affinity or specificity,whereby the VH–VL pair would function as a “clamp” like thetethered fibronectin 3 and PDZ domains (48).In summary, we demonstrated the presence within camelids of

a novel family of V elements, closely related to human VH(4). Themembers of this family contribute abundantly to the classical Abrepertoire but surprisingly alsoparticipate, albeit at a lower frequency,to the VH(4)-HCAb repertoire. No apparent sequence difference isnoted between the VH(4)-derived domains from HCAb or classicalAbrepertoires.Remarkably, it thus appears that the sameVH(4)-D-JHrecombination product from the pre-B cell can end up in both types ofAbs. Therefore, it looks as if the L chain is superfluous in llama Absharboring VH(4) elements. Hence, the VH(4)s enlarge the currentlyknown Ag-binding repertoire of camelid HCAbs (4) and can be ex-ploited in a sdAb format because they produce soluble, stable re-combinant Ag-specific recognition units with a high degree ofsequence identity to human VHs.

AcknowledgmentsWe thank C. Heirman and K. Thielemans of the Laboratory for Molecular

and Cellular Therapy for providing the GM-CSF and R. Van den Bergh of

the Laboratory for Molecular and Cellular Therapy for invaluable com-

ments on this paper.

DisclosuresThe authors have no financial conflicts of interest.

References1. Hamers-Casterman, C., T. Atarhouch, S. Muyldermans, G. Robinson, C. Hamers,

E. B. Songa, N. Bendahman, and R. Hamers. 1993. Naturally occurring antibodiesdevoid of light chains. Nature 363: 446–448.

2. Nguyen, V. K., A. Desmyter, and S. Muyldermans. 2001. Functional heavy-chainantibodies in Camelidae. Adv. Immunol. 79: 261–296.

3. Nguyen, V. K., S. Muyldermans, and R. Hamers. 1998. The specific variabledomain of camel heavy-chain antibodies is encoded in the germline. J. Mol. Biol.275: 413–418.

4. Nguyen, V. K., R. Hamers, L. Wyns, and S. Muyldermans. 2000. Camel heavy-chain antibodies: diverse germline VHH and specific mechanisms enlarge theantigen-binding repertoire. EMBO J. 19: 921–930.

5. Achour, I., P. Cavelier, M. Tichit, C. Bouchier, P. Lafaye, and F. Rougeon. 2008.Tetrameric and homodimeric camelid IgGs originate from the same IgH locus. J.Immunol. 181: 2001–2009.

6. Harmsen, M. M., R. C. Ruuls, I. J. Nijman, T. A. Niewold, L. G. Frenken, andB. de Geus. 2000. Llama heavy-chain V regions consist of at least four distinctsubfamilies revealing novel sequence features. Mol. Immunol. 37: 579–590.

7. Muyldermans, S., T. Atarhouch, J. Saldanha, J. A. Barbosa, and R. Hamers.1994. Sequence and structure of VH domain from naturally occurring camelheavy chain immunoglobulins lacking light chains. Protein Eng. 7: 1129–1135.

8. Vu, K. B., M. A. Ghahroudi, L. Wyns, and S. Muyldermans. 1997. Comparisonof llama VH sequences from conventional and heavy chain antibodies. Mol.Immunol. 34: 1121–1131.

9. Kabat, E. A., and T. T. Wu. 1991. Identical V region amino acid sequences andsegments of sequences in antibodies of different specificities. Relative con-tributions of VH and VL genes, minigenes, and complementarity-determiningregions to binding of antibody-combining sites. J. Immunol. 147: 1709–1719.

10. Davies, J., and L. Riechmann. 1994. ‘Camelising’ human antibody fragments: NMRstudies on VH domains. FEBS Lett. 339: 285–290.

11. Desmyter, A., T. R. Transue, M. A. Ghahroudi, M. H. Thi, F. Poortmans, R. Hamers,S. Muyldermans, and L. Wyns. 1996. Crystal structure of a camel single-domain VH

antibody fragment in complex with lysozyme. Nat. Struct. Biol. 3: 803–811.12. Vincke, C., R. Loris, D. Saerens, S. Martinez-Rodriguez, S. Muyldermans, and

K. Conrath. 2009. General strategy to humanize a camelid single-domain anti-body and identification of a universal humanized nanobody scaffold. J. Biol.Chem. 284: 3273–3284.

13. Dumoulin, M., K. Conrath, A. Van Meirhaeghe, F. Meersman, K. Heremans,L. G. Frenken, S. Muyldermans, L. Wyns, and A. Matagne. 2002. Single-domainantibody fragments with high conformational stability. Protein Sci. 11: 500–515.

14. Harmsen, M. M., and H. J. De Haard. 2007. Properties, production, and appli-cations of camelid single-domain antibody fragments. Appl. Microbiol. Bio-technol. 77: 13–22.

15. Muyldermans, S. 2001. Single domain camel antibodies: current status. J. Bio-technol. 74: 277–302.

16. Saerens, D., J. Kinne, E. Bosmans, U. Wernery, S. Muyldermans, and K. Conrath.2004. Single domain antibodies derived from dromedary lymph node and pe-ripheral blood lymphocytes sensing conformational variants of prostate-specificantigen. J. Biol. Chem. 279: 51965–51972.

17. Tanha, J., G. Dubuc, T. Hirama, S. A. Narang, and C. R. MacKenzie. 2002.Selection by phage display of llama conventional VH fragments with heavy chainantibody VHH properties. J. Immunol. Methods 263: 97–109.

18. Roovers, R. C., T. Laeremans, L. Huang, S. De Taeye, A. J. Verkleij, H. Revets,H. J. de Haard, and P. M. van Bergen en Henegouwen. 2007. Efficient inhibitionof EGFR signaling and of tumour growth by antagonistic anti-EFGR nanobodies.Cancer Immunol. Immunother. 56: 303–317.

19. Lutz, M. B., N. Kukutsch, A. L. Ogilvie, S. Rossner, F. Koch, N. Romani, andG. Schuler. 1999. An advanced culture method for generating large quantities ofhighly pure dendritic cells from mouse bone marrow. J. Immunol. Methods 223:77–92.

20. Arbabi Ghahroudi, M., A. Desmyter, L. Wyns, R. Hamers, and S. Muyldermans.1997. Selection and identification of single domain antibody fragments fromcamel heavy-chain antibodies. FEBS Lett. 414: 521–526.

21. Lefranc, M. P. 2003. IMGT, the international ImMunoGeneTics database. Nu-cleic Acids Res. 31: 307–310.

22. Tomlinson, I. M., G. Walter, J. D. Marks, M. B. Llewelyn, and G. Winter. 1992.The repertoire of human germline VH sequences reveals about fifty groups of VH

segments with different hypervariable loops. J. Mol. Biol. 227: 776–798.23. Matsuda, F., K. Ishii, P. Bourvagnet, K. Kuma, H. Hayashida, T. Miyata, and

T. Honjo. 1998. The complete nucleotide sequence of the human immuno-globulin heavy chain variable region locus. J. Exp. Med. 188: 2151–2162.

24. Tamura, K., J. Dudley, M. Nei, and S. Kumar. 2007. MEGA4: Molecular Evo-lutionary Genetics Analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24:1596–1599.

25. Gasteiger, E., A. Gattiker, C. Hoogland, I. Ivanyi, R. D. Appel, and A. Bairoch.2003. ExPASy: The proteomics server for in-depth protein knowledge andanalysis. Nucleic Acids Res. 31: 3784–3788.

26. Guilliams, M., K. Movahedi, T. Bosschaerts, T. VandenDriessche, M. K. Chuah,M. Herin, A. Acosta-Sanchez, L. Ma, M. Moser, J. A. Van Ginderachter, et al.2009. IL-10 dampens TNF/inducible nitric oxide synthase-producing dendriticcell-mediated pathogenicity during parasitic infection. J. Immunol. 182: 1107–1118.

27. Conrath, K. E., M. Lauwereys, M. Galleni, A. Matagne, J. M. Frere, J. Kinne,L. Wyns, and S. Muyldermans. 2001. Beta-lactamase inhibitors derived fromsingle-domain antibody fragments elicited in the camelidae. Antimicrob. AgentsChemother. 45: 2807–2812.

28. De Groeve, K., N. Deschacht, C. De Koninck, V. Caveliers, T. Lahoutte, N. Devoogdt,S. Muyldermans, P. De Baetselier, and G. Raes. 2010. Nanobodies as tools for in vivoimaging of specific immune cell types. J. Nucl. Med. In press.

The Journal of Immunology 5703

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

29. Nguyen, V. K., R. Hamers, L. Wyns, and S. Muyldermans. 1999. Loss of spliceconsensus signal is responsible for the removal of the entire CH1 domain of thefunctional camel IGG2A heavy-chain antibodies. Mol. Immunol. 36: 515–524.

30. Ewert, S., T. Huber, A. Honegger, and A. Pluckthun. 2003. Biophysical prop-erties of human antibody variable domains. J. Mol. Biol. 325: 531–553.

31. Chothia, C., A. M. Lesk, E. Gherardi, I. M. Tomlinson, G. Walter, J. D. Marks,M. B. Llewelyn, and G. Winter. 1992. Structural repertoire of the human VH

segments. J. Mol. Biol. 227: 799–817.32. Decanniere, K., S. Muyldermans, and L. Wyns. 2000. Canonical antigen-binding

loop structures in immunoglobulins: more structures, more canonical classes? J.Mol. Biol. 300: 83–91.

33. Conrath, K. E., U. Wernery, S. Muyldermans, and V. K. Nguyen. 2003. Emer-gence and evolution of functional heavy-chain antibodies in Camelidae. Dev.Comp. Immunol. 27: 87–103.

34. Saerens, D., K. Conrath, J. Govaert, and S. Muyldermans. 2008. Disulfide bondintroduction for general stabilization of immunoglobulin heavy-chain variabledomains. J. Mol. Biol. 377: 478–488.

35. Arbabi-Ghahroudi, M., R. To, N. Gaudette, T. Hirama, W. Ding, R. MacKenzie,and J. Tanha. 2009. Aggregation-resistant VHs selected by in vitro evolution tendto have disulfide-bonded loops and acidic isoelectric points. Protein Eng. Des.Sel. 22: 59–66.

36. Jespers, L., O. Schon, K. Famm, and G. Winter. 2004. Aggregation-resistantdomain antibodies selected on phage by heat denaturation. Nat. Biotechnol. 22:1161–1165.

37. Woolven, B. P., L. G. Frenken, P. van der Logt, and P. J. Nicholls. 1999. Thestructure of the llama heavy chain constant genes reveals a mechanism forheavy-chain antibody formation. Immunogenetics 50: 98–101.

38. Reiter, Y., P. Schuck, L. F. Boyd, and D. Plaksin. 1999. An antibody single-domain phage display library of a native heavy chain variable region: isolation offunctional single-domain VH molecules with a unique interface. J. Mol. Biol.290: 685–698.

39. Barthelemy, P. A., H. Raab, B. A. Appleton, C. J. Bond, P. Wu, C. Wiesmann,and S. S. Sidhu. 2008. Comprehensive analysis of the factors contributing to thestability and solubility of autonomous human VH domains. J. Biol. Chem. 283:3639–3654.

40. Bond, C. J., J. C. Marsters, and S. S. Sidhu. 2003. Contributions of CDR3 toVHH domain stability and the design of monobody scaffolds for naive antibodylibraries. J. Mol. Biol. 332: 643–655.

41. Jespers, L., O. Schon, L. C. James, D. Veprintsev, and G. Winter. 2004. Crystalstructure of HEL4, a soluble, refoldable human VH single domain with a germ-line scaffold. J. Mol. Biol. 337: 893–903.

42. Vranken, W., D. Tolkatchev, P. Xu, J. Tanha, Z. Chen, S. Narang, and F. Ni.2002. Solution structure of a llama single-domain antibody with hydrophobicresidues typical of the VH/VL interface. Biochemistry 41: 8570–8579.

43. To, R., T. Hirama, M. Arbabi-Ghahroudi, R. MacKenzie, P. Wang, P. Xu, F. Ni,and J. Tanha. 2005. Isolation of monomeric human VHs by a phage selection. J.Biol. Chem. 280: 41395–41403.

44. Ewert, S., C. Cambillau, K. Conrath, and A. Pluckthun. 2002. Biophysicalproperties of camelid VHH domains compared to those of human VH3 domains.Biochemistry 41: 3628–3636.

45. Ewert, S., A. Honegger, and A. Pluckthun. 2003. Structure-based improvementof the biophysical properties of immunoglobulin VH domains with a generaliz-able approach. Biochemistry 42: 1517–1528.

46. Honegger, A., A. D. Malebranche, D. Rothlisberger, and A. Pluckthun. 2009.The influence of the framework core residues on the biophysical properties ofimmunoglobulin heavy chain variable domains. Protein Eng. Des. Sel. 22: 121–134.

47. Dudgeon, K., K. Famm, and D. Christ. 2009. Sequence determinants of proteinaggregation in human VH domains. Protein Eng. Des. Sel. 22: 217–220.

48. Huang, J., A. Koide, K. Makabe, and S. Koide. 2008. Design of protein functionleaps by directed domain interface evolution. Proc. Natl. Acad. Sci. USA 105:6578–6583.

5704 NOVEL CAMELID FAMILY 4 SINGLE-DOMAIN Abs

by guest on April 12, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from