Stable antibody expression at therapeutic levelsusing the 2A peptide

Jianmin Fang, Jing-Jing Qian, Saili Yi, Thomas C Harding, Guang Huan Tu, Melinda VanRoey & Karin Jooss

Therapeutic monoclonal antibodies (mAbs) are currently being developed for the treatment of cancer and other diseases. Despite

clinical success, widespread application of mAb therapies may be limited by manufacturing capabilities. In this paper, we

describe a mAb delivery system that allows continuous production of a full-length antibody at high-concentrations in vivo after

gene transfer. The mAb is expressed from a single open reading frame by linking the heavy and light chains with a 2A self-

processing peptide derived from the foot-and-mouth disease virus. Using this expression system, we generated a recombinant

adeno-associated virus vector encoding the VEGFR2-neutralizing mAb DC101 (rAAV8-DC101). A single dose of rAAV8-DC101

resulted in long-term expression of >1,000 lg/ml of DC101 in mice, demonstrating significant anti-tumor efficacy. This report

describes the first feasible gene therapy approach for stable delivery of mAbs at therapeutic levels, which may serve as an

attractive alternative to direct injection of mAbs.

mAbs have become important therapeutic agents for the treatment ofcancer, inflammation and infectious disease. With technical advance-ments in antibody engineering such as human antibody phage dis-play1, mice transgenic for human IgG2 and antibody humanizationtechniques1, highly specific human monoclonal antibodies can bereadily generated for various disease targets.

Chronic diseases such as cancer require long-term therapies, andmAbs are often infused into people with cancer, frequently at highdoses over a long period of time, which can induce adverse effects3. Analternative approach to long-term delivery of therapeutic antibodies isto express the antibodies in vivo after gene transfer. For mostantibodies, however, therapeutic serum concentrations range fromone to several hundred mg/ml, levels that have been impossible toachieve using gene transfer. Unfortunately, the recombinant adeno-associated virus (rAAV) vector, which is an attractive vector system forachieving long-term gene transfer in vivo4, cannot accommodateconventional antibody expression cassettes that drive the mAb heavyand light chains from two individual promoters, because the vectorcannot package more than B5 kb efficiently.

A potentially advantageous approach for in vivo delivery of anti-bodies is to express mAb heavy and light chains in a bicistronic vectorthat uses a single promoter. The conventional method for bicistronicexpression cassettes, however, uses internal ribosomal entry sites(IRES) that leads to substantially lower expression of the secondgene than the catabolite activator protein (CAP)-dependent firstgene5. In this study, we describe an antibody expression systemthat uses the foot-and-mouth-disease virus (FMDV)-derived 2Aself-processing sequence to express full-length antibodies from a singleopen reading frame (ORF). 2A sequences are oligopeptides located

between the P1 and P2 proteins in some members of the picornavirusfamily and can undergo self-cleavage to generate the mature viralproteins P1 and P2. Among various 2A or 2A-like sequences, FMDV2A is particularly short (minimum of 13 amino acids) and is able to‘cleave’ at its own C terminus between the last two amino acidsthrough an enzyme-independent but undefined mechanism, probablyby ribosomal skip, during protein translation6–11. Using a FMDV 2Asequence adjacent to a furin cleavage site to link the antibodyheavy and light chain sequences, we were able to engineer a mAbexpression cassette that, in the context of AAV-mediated gene transfer,results in high levels of full-length, functional monoclonal antibodiesin vitro and in vivo. Sustained mAb serum levels of 41,000 mg/ml wereachieved in mice with a single administration of an rAAV8 vectorexpressing DC101, an anti-angiogenic mAb targeting vascularendothelial cell growth factor receptor-2 (VEGFR2 or Flk-1)12.The rAAV8 mediated gene transfer of DC101 resulted in signifi-cant (P o 0.001) anti-tumor efficacy in two tumor models, demon-strating the generation of functional antibodies in vivo using thisexpression system.

RESULTS

2A-mediated mAb expression from a single ORF

DC101, a rat anti-mouse VEGFR2 (Flk-1) IgG1 mAb, which has beenwell characterized for its anti-angiogenic effects in mouse tumormodels12, was chosen as a model antibody to evaluate expression offull-length antibodies from a single ORF using the FMDV 2A self-processing peptide. An expression cassette termed H2AL, in which theheavy and light chain sequences of the DC101 mAb were linkedtogether by the FMDV 2A self-cleavage sequence, was generated and

Published online 17 April 2005; doi:10.1038/nbt1087

Department of Preclinical Oncology and Immunology, Cell Genesys, Inc., 500 Forbes Blvd., S. San Francisco, California 94080, USA. Correspondence should beaddressed to J.F. (jianminf@cellgenesys.com).

5 84 VOLUME 23 NUMBER 5 MAY 2005 NATURE BIOTECHNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

cloned into an expression plasmid driven by the CAG promoter(Fig. 1). The CAG promoter is comprised of the cytomegalovirus(CMV) immediate early enhancer region, chicken b-actin promoter/splice donor and rabbit b-globin enhancer13 and is constitutivelyactive. A control plasmid, which linked the DC101 heavy and lightchains by an IRES derived from encephalomyocarditis virus (EMCV)(Fig. 1b, HIRESL), was also generated. Both plasmids were transientlytransfected into human embryonic kidney (HEK) 293 cells, and theantibody concentrations in the supernatants were determined byenzyme-linked immunosorbent assay (ELISA) after 48 h. The H2ALconstruct resulted in about 1.6 mg/ml of DC101, whereas the HIRESLconstruct resulted in 0.1 mg/ml of the mAbs (Fig. 2a). Thus, the 2Asequence efficiently facilitates antibody heavy and light chain expres-sion from a single ORF.

To further characterize the DC101 heavy and light chains expressedfrom the 2A-containing construct, the proteins in the supernatants oftransiently transfected HEK 293 cells were separated on SDS-PAGEgels under both reducing and nonreducing conditions and subjectedto western blot analysis using a polyclonal goat anti-rat IgG antibodythat recognizes antibody heavy and light chains. Under reducingconditions, two protein bands at molecular weights of B55 and25 kDa were detected, corresponding to the IgG heavy and lightchain proteins of DC101, respectively (Fig. 2b). Bands of similarsize were detected in samples from the parental DC101 hybridoma

cells. The heavy chain protein from the H2AL plasmid migratesslightly more slowly than the native DC101 antibody heavychain (Fig. 2b) due to 23–amino acid residues that are derived fromthe 2A sequence and remain after cleavage. Under nonreducingconditions, a single band of approximately 160 kDa was detected inthe cell culture supernatant of H2AL-transfected HEK 293 cells(Fig. 2c), which is the expected size of a dimerized full-lengthantibody containing two heavy and two light chains. The molecularweight of the mAb expressed from the H2AL plasmid was slightlyhigher than that of the native antibody, because of the additionalamino acid residues at the C terminus of the heavy chain. Noadditional protein bands, which are expected when the ratio betweenheavy and light chains is imbalanced, were detected under nonredu-cing conditions (Fig. 2c), suggesting that the antibodies in the super-natant of H2AL transfected cells are properly dimerized and neitherheavy nor light chains are in excess.

Biological activity of mAbs expressed from 2A plasmids

The biological activity of the antibodies expressed from the2A-containing plasmid was evaluated using a binding assay thatmeasures mAb-binding activity to immobilized mVEGFR2 protein(see Methods). DC101 antibodies expressed from the H2AL constructin HEK 293 cells were able to recognize the mouse VEGFR2 withsimilar binding activity as the parental antibodies (Fig. 3a) and werecapable of blocking the VEGF and mVEGFR2 interaction in a dose-dependent manner in a VEGF-mVEGFR2 (ligand-receptor) bindingassay with the same potency as the native antibodies (Fig. 3b). Thus,antibodies expressed from the 2A-containing construct retain fullbiological activity.

Removal of 2A-derived amino acid residues by furin cleavage

Since the 2A self-processing cleavage occurs between the last twoamino acids at the C terminus of the 2A peptide, the first protein inthe cassette (that is, the heavy chain in the H2AL construct) has 23additional amino acid residues at its C terminus (Fig. 1a). Toeliminate possible adverse effects caused by the remaining 2A residues,a DC101 expression cassette was engineered that included a furincleavage site sequence (RAKR), located between the 2A sequence andthe mAb heavy chain (HF2AL) (Fig. 1b, lower construct).

mAb was expressed from the HF2AL construct in HEK 293cells after transient plasmid transfection (Fig. 2a). The heavychain proteins in the supernatants were separated by SDS-PAGEgel and analyzed by western blot using a goat anti-rat IgGantibody. The antibody heavy chains expressed from the HF2ALconstruct have a similar molecular weight as the native antibodyheavy chains, suggesting successful cleavage at the furin cleavagesite (Fig. 2b). Furthermore, the antibodies expressed from theHF2AL construct appeared as a single band at a molecular weightcorresponding to the antibody dimer (Fig. 2c) and demonstrated fullbiological activity in antibody binding (Fig. 3a) and neutralizationassays (Fig. 3b).

To confirm successful removal of the residual 2A amino acids, wegenerated a construct that contained six histidine residues (his-tag) atthe C terminus of the antibody light chain in the HF2AL cassette. Thehis-tagged antibody was expressed in vivo by hydrodynamic injectionof the HF2AL expression plasmid into mice, and mAb was purifiedfrom mouse serum using a nickel column purification system. Thepurified antibody appears as two protein bands in a reducing SDS-PAGE gel at B52 and 25 kDa (data not shown), corresponding to theantibody heavy and light chains, respectively. The heavy chain bandwas excised from the gel, digested with trypsin and analyzed by mass

Stop codon2AFurin

cleavagesite

Start codon

Promoter

Promoter

Promoter

HF2AL Ab heavy chain

Ab heavy chain

Ab heavy chain

Ab light chain

Ab light chain

Ab light chain

Poly A

Poly A

Poly A

Stop codon

Stop codonStart codon

Stop codonStart codon Startcodon

HIRESL IRES

2A

H2AL

2A residues

2A residue (1 aa)2A residues

Heavy and lightchains after signal-peptide removalduring secretion

Protein translationand self-processing

Cleavage site

2A

mRNA

Signal peptide Heavy chain Signal peptide

Light chaina

b

Figure 1 Full-length mAb expression cassette using the FMDV 2A sequence.

(a) Schematic illustration of the biosynthesis of mAb heavy and light chains

by the 2A, peptide-containing expression cassette. (b) Antibody expression

cassettes in which mAb heavy and light chain sequences are linked by the

2A sequence (H2AL), an IRES sequence (HIRESL) or a combination of furin

cleavage site and 2A sequence (HF2AL).

NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 5 MAY 2005 58 5

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

spectrometry. We detected no peaks that represent peptide frag-ments derived from the 2A sequence (APVK, QTLNFDLLK andLAGDVESNPG) in the mass spectrum, demonstrating that the 2Aresidues had been removed from the antibody heavy chain of theHF2AL construct. Moreover, the mass spectrum analysis revealed a1,039.53-Da fragment corresponding to the C-terminal fragmentsequence, SLSHSPGKRA, that contains the native C-terminalsequence of the antibody heavy chain plus two additional aminoacids (arginine and alanine) derived from the furin cleavage site. Theactual amino acid sequence of the 1039.53 Da fragment was furtherconfirmed by the post-source decay (PSD) analysis. In summary, thesedata demonstrate that the addition of a furin cleavage site to the 2Aself-processing cleavage site results in the removal of the 23–aminoacids that remain after 2A cleavage.

2A-mediated, high-level expression of mAbs in vivo

Using the 2A sequence, the expression cassette for DC101 could be fitinto a rAAV vector. We generated rAAV vectors pseudotyped with thecapsid proteins from AAV serotype-8 (ref. 14) that express the DC101antibody from an H2AL or HF2AL cassette driven by the CAGpromoter, termed rAAV8-DC101(H2AL) and rAAV8-DC101(HF2AL).Mice were injected through the hepatic portal vein with one of three

dose levels of rAAV8-DC101(H2AL) vector (1 � 1011, 2 � 1011 or4 � 1011 vector genomes (vg)/mouse), and DC101 serum levelswere evaluated over time. We detected, 28 d after administrationof the rAAV8-DC101(H2AL) vector, peak serum levels of 3,286and 1,877 mg/ml DC101 in mice that had received 4 � 1011 or2 � 1011 vg/animal, respectively (Fig. 4a). Antibody serum levelsdeclined somewhat thereafter but remained at about 600 mg/ml inthe animals treated with the highest dose of vector throughout the4-month study (Fig. 4a). Interestingly, two- to tenfold higher DC101serum levels were achieved when the antibody was expressed fromthe HF2AL expression cassette (Fig. 4b). Peak expression levelsin mice receiving 4 � 1011 and 2 � 1011 vg/mouse of therAAV-DC101(HF2AL) vector were 48,000 mg/ml and remainedabove 1,000 mg/ml up to 4 months after rAAV vector administration.Furthermore, DC101 antibody expressed from either cassette exhibitedfull biological activity in the antibody binding (Fig. 3a) and neutra-lization assays (Fig. 3b).

We monitored serum alanine aminotransferase (ALT) and aspartateaminotransferase (AST) levels in these animals to evaluate liverfunction after gene transfer and were not able to detect elevatedALT or AST serum levels in all vector-treated mice throughout theentire experiment (data not shown).

Figure 2 In vitro expression of DC101 mAb using

the 2A self-processing sequence–containing

expression plasmids. HEK 293 cells were

transiently transfected with DC101-expressing

plasmids and supernatants were harvested

for protein analyses. (a) ELISA analysis of

supernatants 48 h after transfection of HEK 293

cells with the H2AL, the HIRESL or the HF2ALplasmids (mean 7 s.d.). (b) Western blot

analysis of DC101-containing supernatants under

reducing conditions. Proteins in the supernatants

of the parental hybridoma cells (hybridoma),

the HEK 293 cells transfected with the H2AL

plasmid (H2AL) or the HF2AL plasmid (HF2AL),

or untransfected HEK 293 cells (mock) were

separated by SDS-PAGE under reducing conditions and probed with a goat anti-rat IgG (H+L) polyclonal antibody. (c) Western blot analysis of supernatants

from the parental DC101 hybridoma cells (hybridoma), from HEK 293 cells transfected with the H2AL plasmid (H2AL) or HF2AL plasmid (HF2AL), or from

untransfected HEK 293 cells (mock). Proteins in supernatants were separated in SDS-PAGE under nonreducing conditions and probed with a goat anti-rat

IgG (H+L) polyclonal antibody.

HF2ALHIRESLH2AL0

0.5

1

1.5

2

2.5

3

DC

101

mA

b (µ

g/m

l)

14

17

17

28

28

38

38

49

49

62

62

98

98

188

188

MW(kDa)

MW(kDa)M

ock

Hyb

ridom

a

H2A

L

HF

2AL

Moc

k

Hyb

ridom

a

H2A

L

HF

2AL

a b c

010301003001,0003,000

Antibody (ng/ml)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

OD

405

OD

405

Control

HF2AL in vivo

HF2AL in vitro

H2AL in vitro

Hybridoma

H2AL in vivoControl

HF2AL in vivo

HF2AL in vitro

H2AL in vitro

Hybridoma

H2AL in vivo

031.2562.51252505001000

Antibody (ng/ml)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6a b

Figure 3 Biological activity of DC101 mAbs generated from 2A-containing expression cassettes. (a) Antibody binding activity to immobilized Flk-1 protein.

DC101 antibody concentration was determined by ELISA. Binding of the mAb to Flk-1 was detected at an absorbance of 405nm after incubation with an

anti-rat IgG-HRP antibody, followed by the addition of the HRP substrate (mean 7 s.d.). (b) Neutralizing properties of DC101 evaluated in a VEGF-Flk-1

binding assay. VEGF-Flk-1 binding was detected at an absorbance of 405 nm with an anti-Flk-1 antibody conjugated to HRP.

5 86 VOLUME 23 NUMBER 5 MAY 2005 NATURE BIOTECHNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

Anti-tumor efficacy of 2A-mediated mAbs in vivo

After demonstrating high-level antibody expression in mice afterrAAV8-mediated gene transfer, we investigated whether DC101generated from the rAAV8 vector results in anti-tumor efficacyin vivo as demonstrated for recombinant DC101 antibodies12. Givenits higher levels of serum DC101 expression in vivo, only the furin-containing rAAV8-DC101(HF2AL) vector was tested.

rAAV8-DC101(HF2AL) or rAAV8-null control vector (2 � 1011 vg/mouse) was injected into nude mice through the tail vein, and DC101antibody serum levels were evaluated by ELISA. At day 24 after vectoradministration (Fig. 5), the mice were injected subcutaneously witheither 1 � 105 cells/mouse of murine B16F10 melanoma cells or 5 �106 cells/mouse of human U87 malignant glioblastoma cells. rAAV8-DC101 injection resulted in milligram levels of mAb in the serum ofmice (Fig. 5a,d). Significant anti-tumor activity was observed in bothmodels in mice treated with the rAAV8-DC101(HF2AL) vector(Fig. 5b,e; P o 0.05), which resulted in a significantly prolongedsurvival time (Fig. 5c,f; P o 0.001). In the B16F10 model, mediansurvival time (MST) increased from 30 d in the control group to41.5 d in the treated group. In the U87 MG model, DC101 antibodygene transfer resulted in tumor dormancy in 6 out of 11 mice formore than 3 months and three mice tumors that had reached avolume of 400–700 mm3 regressed completely (data not shown). Insummary, a single administration of a rAAV8 vector expressing DC101mAb results in stable and high mAb serum levels that are able tocontrol tumor burden.

DISCUSSION

This study demonstrates that a 2A self-processing peptide derivedfrom FMDV facilitates efficient and apparent equimolar expression offull-length antibody heavy and light chains in vitro and in vivo from asingle ORF and that the antibody chains self-assemble to form afunctional antibody. To our knowledge, no one else has shown thatfull-length mAb gene transfer can provide potent anti-tumor activityin vivo, which may be useful as an alternative therapy to directinjection of monoclonal antibodies.

Given the potential clinical benefits of antibody gene therapy, greateffort has been devoted to the expression of full-length antibodiesin vivo after gene transfer15. To achieve therapeutic effects, however,most antibodies require high and sustained serum levels, typicallyranging from one to several hundred mg/ml16–18. Such high mAbserum levels can be achieved only by repeated administration ofhigh doses of recombinant protein, typically ranging from 5 mg/kgto 20 mg/kg of body weight. The high mAb serum concentrationsrequired for clinical efficacy and the typically low expression levelsafter gene transfer have made the development of antibody genetransfer technologies challenging.

Electroporation of mAb plasmids in muscle achieved mAb serumlevels of 1.5 mg/ml in mice19, whereas implantation of ex vivo transducedcells with retroviral vectors, such as myoblasts20 and fibroblasts21,resulted in mAb serum levels of 1–3 mg/ml. rAAV vectors encodingthe heavy and light chains of a human anti-HIV mAb driven by aminimal CMV or EF1 alpha promoter yielded antibody serum levels of

140120100806040200Days

140120100806040200Days

1

10

100

1,000

10,000

100,000

DC

101

(µg/

ml)

1

10

100

1,000

10,000D

C10

1 m

Ab

(µg/

ml)

1 × 1011

2 × 1011

4 × 1011

1 × 1011

2 × 1011

4 × 1011

a b

10090807060504030

40

Days

Days

00

20

60

80

100

120

Per

cent

sur

viva

l

40

0

20

60

80

100

120

Per

cent

sur

viva

l

120100806040200

rAAV8 null

rAAV8 DC101

rAAV8 nullrAAV8 DC101

rAAV8 nullrAAV8 DC101

rAAV8 null

rAAV8 DC101

***

***

U87 MG

U87 MG

U87 MG

B16F10B16F10

B16F10

201713100

Days

3935322825211814117

Days

1,400

1,200

1,000

800

600

400

200

0

Tum

or v

olum

e (m

m3 )

Tum

or v

olum

e (m

m3 )

1,250

1,000

750

500

250

0

7550250

Days

50403020100

Days

1

10

100

1,000

10,000

1

10

100

1,000

10,000

DC

101

(µg/

ml)

DC

101

(µg/

ml)

2 × 1011

2 × 1011

a b c

d e f

*

*

Figure 5 Anti-tumor efficacy after rAAV8-mediated gene transfer of DC101. (a–f) rAAV8-CAG-HF2AL or rAAV8-null control vectors (2 � 1011 vg/mouse) were

administered through the tail vein into NCr nude mice on day 0 and serum DC101 concentrations were determined by rat IgG1 ELISA (mean 7 s.e.m.)

(a,d). On day 24, B16F10 (1 � 105 cells/mouse) (a–c) or U87 (5 � 106 cells/mouse) (d–f) tumor cells were injected subcutaneously into these mice and

tumor volume (mean 7 s.e.m.) (b,e) and survival (c,f) were evaluated. For tumor growth (b,e) and survival (c,f), the day of tumor challenge was regarded

as day 0. *, P o 0.05; ***, P o 0.001.

Figure 4 Expression of DC101 mAb in vivo by

rAAV8 vector-mediated gene transfer. (a,b) rAAV8

CAG H2AL (a) or rAAV8 CAG HF2AL (b) vectors

were administered to NCr nude mice at three

doses (1 � 1011, 2 � 1011 or 4 � 1011 vg/

mouse) through the portal vein. Mice were bled

weekly and DC101 serum levels determined by

ELISA (mean 7 s.e.m.).

NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 5 MAY 2005 58 7

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

4–5 mg/ml22. The mAb serum levels described in these studies are belowthe therapeutic levels. So far, only high doses of recombinant adenoviralvectors have occasionally achieved mAb serum levels of B200 mg/ml23.However, mAb expression by adenoviral vectors is transient23, due tothe instrinsic immunogenicity of the vector backbone24.

rAAV is a preferred vector system when long-term gene expressionis desired. rAAV vectors can stably transduce host cells and are capableof expressing therapeutic proteins at constant levels in vivo followinga single vector administration. rAAV vectors have been shown toefficiently transduce quiescent cells of the muscle, liver, brain andeye24–27 and are currently under clinical evaluation28–31. One of theproblems of using rAAV vectors as mAb expression vectors, however,is their limited packaging capacity, which makes it difficult to expressmAb heavy and light chains from individual promoters in one vector.We used the short FMDV 2A sequence to mediate antibody heavy andlight chain expression from a single ORF, which enabled us toconstruct a single rAAV vector for the production of a full-lengthmAb. In combining the 2A technology with the recently identifiedAAV8 serotype that efficiently transduces the liver, mAb serum levelsof 41,000 mg/ml were achieved in mice after a single administrationof the rAAV vector. Antibody expression remained at high levels forover 4 months. In summary, this expression system presents a feasiblegene therapy approach for long-term delivery of antibodies at highlevels in vivo. Furthermore, the high concentration of mAb achievedin vivo after rAAV8-mediated gene transfer will make vector manu-facturing possible for human use.

In our study, we were able to show that the 23 additional aminoacids derived from the 2A sequence at the C terminus of the heavychain can be efficiently removed by adding a furin cleavage site next tothe 2A sequence. This modification results in a significant increase ofmAb serum levels and generates an antibody that more closely re-sembles the native protein, thereby eliminating possible adverse effects.

There are currently contrasting results regarding whether theleading signal peptide is necessary for the second protein to enterinto the endoplasmic reticulum (ER) for protein secretion. In yeast,the lack of the signal peptide at the N terminus of the second proteinresulted in cytosolic localization of the protein10. In contrast, inmammalian cells, the second protein can enter the ER lumen withouta signal sequence11. It is currently not clear, how the signal sequence ofthe second protein affects overall antibody expression. Nevertheless,the mAb light chain in our H2AL and HF2AL constructs contains thenative signal peptide at its N terminus. Since the signal peptide iscleaved during protein secretion, inclusion of the leading sequenceremoves the amino acid proline from the N terminus of the antibodylight chain, that is derived from the 2A sequence. Therefore, thesecreted antibody light chain expressed from these cassettes is expectedto retain its native sequence.

In addition to providing an option for an alternative long-term anti-body therapy, we believe that this technology may be of great value forthe generation of mAb producer cell lines. Current technologies usedfor the generation of stable mAb producer cell lines are labor intensiveand time consuming. We are currently evaluating this technology forthe generation of stable mAb producer cell lines and have preliminarydata demonstrating that the 2A technology combined with viral vectorsenables the rapid identification of high mAb producer cell clones.

Furthermore, the 2A-furin/rAAV8 technology described within thispaper may be a useful tool to validate mAb targets in vivo for drugdevelopment, study protein function in vivo by blocking their biolo-gical activity or carry out cell-depletion research to study the functionof a particular cell type in an in vivo system. The 2A-furin/rAAV8technology for mAb gene expression may also have a great potential

for delivery of neutralizing mAbs to induce passive immunity in someinfectious diseases.

In summary, the 2A-furin technology may have broad applicationas an in vivo mAb therapy for the treatment of cancer or other chronicdiseases, and as a research tool for studies in which high and sustainedmAb serum levels are required or for the rapid generation of stablemAb producer cell lines.

METHODSPlasmid construction. Total RNA from DC101 hybridoma cells was purified

using RNeasy kit (Qiagen). First stream cDNA was synthesized from total RNA

using specific primers to heavy or light chain constant region sequences.

Variable regions of the antibody heavy and light chains, including their signal

peptide sequences, were amplified with the rapid amplification of cDNA ends

cloning kit (BD Biosciences Clontech). The VH and VL were cloned into the

pCR 2.1 plasmid using the TA cloning kit (Invitrogen) and sequenced. The

consensus VH and VL sequences were determined based on sequence data from

the clones derived from multiple independent PCRs. Constant regions of both

heavy and light chains were also cloned from cDNA. Variable and constant

regions of heavy and light chains were joined together by PCR reaction to

generate full-length heavy and light chains.

To generate the constructs containing the 2A self-processing sequence, the

cDNA oligo for a 24–amino acid FMDV 2A peptide was synthesized (Bio-

source) based on the sequence APVKQTLNFDLLKLAGDVESNPGP32. The

cDNA encoding antibody heavy chain, 2A and light chain was assembled by

PCR and was cloned into a plasmid downstream of a CAG promoter. The final

plasmid, pH2AL, contains a single ORF consisting of a full-length heavy chain,

the 2A sequence, and full-length light chain (Fig. 1). Both heavy and light

chains include their native signal peptide sequences at their N termini. The

plasmid also includes a bovine growth hormone poly A sequence at the 3¢ end.

For the construct that also contains a furin cleavage site, the sequence that

encodes the furin cleavage site, RAKR, was inserted by PCR between the

antibody heavy chain and the 2A sequence. The cDNA that encodes an antibody

heavy chain, a furin cleavage site (RAKR), the 2A sequence and an antibody light

chain was cloned into the plasmid downstream of the CAG promoter (pHF2AL).

To express full-length antibody heavy and light chains with IRES sequences,

the DC101 heavy chain, an IRES sequence derived from the EMCV33 and the

DC101 light chain were inserted into the plasmid downstream of the CAG

promoter. Both heavy and light chain cDNAs end with a stop codon (Fig. 1).

rAAV vector preparation. Subconfluent HEK 293 cells were cotransfected

using the calcium phosphate method with the pAAV-CAG-DC101 vector

plasmid in combination with the AAV8 serotype helper plasmid p5e18-VD2/8

(ref. 14) and pXX-6 (ref. 34). Forty-eight hours after transfection, cells were

harvested using PBS/EDTA (10 mM) and lysed by three freeze/thaw cycles in

cell lysis buffer (150 mM NaCl, 50 mM HEPES, pH 7.6). Lysates were treated

with 250 U/ml benzonase for 15 min at 37 1C and cellular debris was removed

by centrifugation. The cleared cell lysate was fractionated by ammonium sulfate

precipitation and the rAAV virions were isolated on two sequential CsCl

gradients. The gradient fractions containing rAAV were dialyzed against sterile

PBS containing CaCl2 and MgCl2, and stored at –80 1C. Viral titers were

determined by dot-blot analysis. Briefly, rAAV preparations were treated with

DNaseI followed by proteinase K in the presence of 0.5% SDS and 10 mM

EDTA to liberate the rAAV genomes, followed by phenol chloroform extraction

and ethanol precipitation. Viral DNA was denatured in alkali and applied to a

nylon membrane. Dilutions of the corresponding vector plasmid were used as

standards to determine the rAAV virion copy number. A radioactive probe

specific for the rAAV transgene was hybridized to DNA on the filter and the

filter was exposed to film followed by quantification of radioactivity by a

b-counter (Perkin Elmer). Biological activities of the purified rAAV were

determined by DC101 antibody expression in HEK 293 or HuH7 cells following

rAAV transduction in vitro.

Cell culture and transfection. HEK 293, B16F10, U87MG and DC101

hybridoma cell lines were obtained from ATCC. HuH 7 cells were from Jing-

Hsiung Ou, University of Southern California. HEK 293 cells were cultured in

Iscove’s Modified Dulbecco’s Medium (Invitrogen), supplemented with 3 mM

5 88 VOLUME 23 NUMBER 5 MAY 2005 NATURE BIOTECHNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

L-glutamine and 10% fetal bovine serum. HuH 7, B16F10, U87 and DC101

hybridoma cells were cultured in DMEM medium (Invitrogen), supplemented

with 3 mM L-glutamine and 10% fetal bovine serum. To produce DC101 mAb

from the hybridoma, exhausted cell culture supernatants were harvested. For

DC101 expression in vitro, plasmid DNA was purified using a plasmid DNA

mega purification kit (Qiagen) and cells were transfected in 6-well tissue

culture plates or 10-cm dishes with FuGene 6 transfection reagent (Roche).

Twenty-four hours after transfection, the cell culture medium was removed and

the cells were fed fresh medium with or without 10% FBS. The supernatants

were collected after 48 or 72 h.

ELISA and western blots. The DC101 antibody concentrations in mouse

serum or cell culture supernatants were determined using a commercial ELISA

assay kit for rat IgG1 (Bethyl Lab). For protein analysis in polyacrylamide gels,

protein samples were separated in precast Tris-glycine gels (Invitrogen) under

reducing or nonreducing conditions. For western blot analysis, proteins in

polyacrylamide gels were transferred to nitrocellulose membranes. The mem-

branes were blocked with 5% nonfat dry milk and incubated with a goat anti-

rat IgG antibody (Calbiochem) conjugated with horseradish peroxidase (HRP).

Protein bands were visualized by exposure on X-ray films (Kodak) after the

membranes were treated with enhanced chemiluminescence solution (Pierce).

Antibody binding and blocking assay. DC101 mAb antibodies were expressed

in the supernatants of hybridoma cells (hybridoma), from HEK 293 cells

transfected with the H2AL (H2AL in vitro) or the HF2AL plasmids (HF2AL

in vitro), or in the sera of mice injected with AAV8 H2AL (H2AL in vivo) or

HF2AL (HF2AL in vivo) vectors. To evaluate binding activity of the DC101

antibody to mVEGFR2 (Flk-1), 96-well ELISA plates were coated with 200 ng/

ml of recombinant Flk-1-Fc protein (R&D Systems). The plates were blocked

with 5% nonfat dry milk and incubated with various concentrations of DC101

antibody. The plates were incubated with goat anti-rat IgG antibody conjugated

to HRP and staining revealed by peroxidase substrate. Medium from naive

HEK 293 cells served as a negative control (control). The plates were read in a

microplate reader at absorbance of 405 nm.

To evaluate the neutralizing effect of DC101 mAb on VEGF-Flk-1 binding,

96-well plates were coated with 500 ng/ml of recombinant human VEGF165

(R&D Systems). Recombinant Flk-1-Fc protein was preincubated with

various concentrations of DC101 mAb expressed from hybridoma cells

(Hybridoma), HEK 293 cells transfected with the H2AL (H2AL in vitro) or

the HF2AL (HF2AL in vitro) plasmids, or sera of mice injected with AAV8

H2AL (H2AL in vivo) or AAV8 HF2AL (HF2AL in vivo) vectors. After blocking

with 5% nonfat dry milk, 50 ng/ml of recombinant Flk-1-Fc (R&D Systems),

which had been preincubated with various concentrations of DC101 antibody,

was added to each well and incubated at 37 1C for 1 h. The plates were washed

with Tris-buffered saline, incubated with biotin-conjugated goat anti-Flk-1

antibody (R&D Systems), washed again and the staining revealed with strepta-

vidin-HRP (DB Pharmingen) and peroxidase substrate. Medium from naive

HEK 293 cells was used as a negative control (control). VEGF-Flk-1 binding

was detected at an absorbance of 405 nm with an anti-Flk-1 antibody

conjugated to HRP.

His-tagged antibody expression, purification and mass spectrum analysis. A

hydrodynamic gene transfer method35 was used to express his-tagged DC101

antibody from plasmid in mice. Briefly, a pAAV-CAG-DC101 antibody HF2AL

construct with 6� histidine residues (his-tag) at the C terminus of the light

chain was constructed. Plasmid DNA (50 mg in 2 ml of PBS) was rapidly

injected into a NCr nu/nu mouse via the tail vein. His-tagged DC101 antibody

in mouse serum was purified using a nickel column (Qiagen). Purified proteins

were separated in a SDS-PAGE gel under reducing conditions and the antibody

heavy chain band was isolated, trypsin digested and analyzed in a mass

spectrometer. To determine the amino acid sequence, we isolated the peptide

fragment peaks from the mass spectrum and analyzed them by post source

decay (PSD) analysis.

Antibody expression in vivo by rAAV vector–mediated gene transfer. Female

NCr nu/nu mice (6–8 weeks old) were obtained from Taconic. All mice were

housed under specific-pathogen-free conditions and treated according to the

Institute for Laboratory Animal Research Guide for the Care and Use of

Laboratory Animals. rAAV vector at 1 � 1011, 2 � 1011 and 4 � 1011 vg/

mouse was injected into mice (n ¼ 5 in each group) via a surgically implanted

portal vein catheter. Mice were bled by alternate retro-orbital puncture at each

scheduled time point for up to 6 months for analysis of DC101 expression.

Blood samples may be collected from the orbital sinus of anesthetized mice at

scheduled intervals.

Mouse tumor models. Female NCr nu/nu mice (n ¼ 10–12 in each group)

were injected with rAAV8-CAG-HF2AL vector through intravenous adminis-

tration via tail veins at a dose of 2 � 1011 vg/mouse in 200 ml of PBS. Mice in

the control group were injected with the same dose of rAAV8-null vector. To

monitor serum DC101 levels, mice were bled weekly by alternate retro-orbital

puncture. At day 24 after rAAV administration, B16F10 melanoma (1 � 105

cells/mouse in 200 ml PBS) or human U87 MG glioma (5 � 106 cells/mouse in

200 ml PBS/Matrigel at 1:1 ratio) cells were implanted subcutaneously into the

flanks of mice. Tumor volume was measured twice a week with a caliper and

calculated by the formula of [(width � length � height)/2]. For survival

studies, end points were based on the pre-established criteria that include

tumor volume, body weight loss, degree of tumor necrosis and the general

health of animals.

ACKNOWLEDGMENTSThe authors would like to thank Mingxia Shi, Sandra Sanchez, Lei Xu, GailColbern and the animal service group of Cell Genesys for technical assistance, JohnLeszyk at the University of Massachusetts Medical School for carrying out massspectrometry analysis and Peter Working for critical reading of the manuscript.

COMPETING INTERESTS STATEMENTThe authors declare competing financial interests (see the Nature Biotechnologywebsite for details).

Received 20 January; accepted 10 March 2005

Published online at http://www.nature.com/naturebiotechnology/

1. Hudson, P.J. & Souriau, C. Engineered antibodies. Nat. Med. 9, 129–134 (2003).2. Green, L.L. Antibody engineering via genetic engineering of the mouse: XenoMouse

strains are a vehicle for the facile generation of therapeutic human monoclonalantibodies. J. Immunol. Methods 231, 11–23 (1999).

3. Maloney, D.G. et al. IDEC-C2B8 (Rituximab) anti-CD20 monoclonal antibody therapy inpatients with relapsed low-grade non-Hodgkin’s lymphoma. Blood 90, 2188–2195(1997).

4. Grimm, D. & Kay, M.A. From virus evolution to vector revolution: use of naturallyoccurring serotypes of adeno-associated virus (AAV) as novel vectors for human genetherapy. Curr. Gene Ther. 3, 281–304 (2003).

5. Mizuguchi, H., Xu, Z., Ishii-Watabe, A., Uchida, E. & Hayakawa, T. IRES-dependentsecond gene expression is significantly lower than cap-dependent first gene expressionin a bicistronic vector. Mol. Ther. 1, 376–382 (2000).

6. Ryan, M.D. & Drew, J. Foot-and-mouth disease virus 2A oligopeptide mediated cleavageof an artificial polyprotein. EMBO J. 13, 928–933 (1994).

7. Donnelly, M.L., Gani, D., Flint, M., Monaghan, S. & Ryan, M.D. The cleavage activitiesof aphthovirus and cardiovirus 2A proteins. J. Gen. Virol. 78, 13–21 (1997).

8. Donnelly, M.L. et al. Analysis of the aphthovirus 2A/2B polyprotein ‘cleavage’ mechan-ism indicates not a proteolytic reaction, but a novel translational effect: a putativeribosomal ‘skip’. J. Gen. Virol. 82, 1013–1025 (2001).

9. Szymczak, A.L. et al. Correction of multi-gene deficiency in vivo using a single ‘self-cleaving’ 2A peptide-based retroviral vector. Nat. Biotechnol. 22, 589–594 (2004).

10. de Felipe, P., Hughes, L.E., Ryan, M.D. & Brown, J.D. Co-translational, intraribosomalcleavage of polypeptides by the foot-and-mouth disease virus 2A peptide. J. Biol.Chem. 278, 11441–11448 (2003).

11. de Felipe, P. & Ryan, M.D. Targeting of proteins derived from self-processing poly-proteins containing multiple signal sequences. Traffic 5, 616–626 (2004).

12. Prewett, M. et al. Antivascular endothelial growth factor receptor (fetal liver kinase 1)monoclonal antibody inhibits tumor angiogenesis and growth of several mouse andhuman tumors. Cancer Res. 59, 5209–5218 (1999).

13. Niwa, H., Yamamura, K. & Miyazaki, J. Efficient selection for high-expression trans-fectants with a novel eukaryotic vector. Gene 108, 193–199 (1991).

14. Gao, G.P. et al. Novel adeno-associated viruses from rhesus monkeys as vectors forhuman gene therapy. Proc. Natl. Acad. Sci. USA 99, 11854–11859 (2002).

15. Bakker, J.M., Bleeker, W.K. & Parren, W.H.I. Therapeutic antibody gene transfer: anactive approach to passive immunity. Mol. Ther. 10, 411–416 (2004).

16. Davis, T.A. et al. Rituximab anti-CD20 monoclonal antibody therapy in non-Hodgkin’slymphoma: safety and efficacy of re-treatment. J. Clin. Oncol. 18, 3135–3143 (2000).

17. Lin, Y.S. et al. Preclinical pharmacokinetics, interspecies scaling, and tissue distribu-tion of a humanized monoclonal antibody against vascular endothelial growth factor.J. Pharmacol. Exp. Ther. 288, 371–378 (1999).

18. Armbruster, C. et al. A phase I trial with two human monoclonal antibodies (hMAb 2F5,2G12) against HIV-1. AIDS 16, 227–233 (2002).

NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 5 MAY 2005 58 9

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy

19. Perez, N. et al. Regulatable systemic production of monoclonal antibodies by in vivomuscle electroporation. Genet. Vaccines Ther. 2, 2 (2004).

20. Noel, D. et al. In vitro and in vivo secretion of cloned antibodies by genetically modifiedmyogenic cells. Hum. Gene Ther. 8, 1219–1229 (1997).

21. Noel, D., Pelegrin, M., Brockly, F., Lund, A.H. & Piechaczyk, M. Sustained systemicdelivery of monoclonal antibodies by genetically modified skin fibroblasts. J. Invest.Dermatol. 115, 740–745 (2000).

22. Lewis, A.D., Chen, R., Montefiori, D.C., Johnson, P.R. & Clark, K.R. Generation ofneutralizing activity against human immunodeficiency virus type 1 in serum by anti-body gene transfer. J. Virol. 76, 8769–8775 (2002).

23. Noel, D. et al. High in vivo production of a model monoclonal antibody on adenoviralgene transfer. Hum. Gene Ther. 13, 1483–1493 (2002).

24. Jooss, K. & Chirmule, N. Immunity to adenovirus and adeno-associated viral vectors:implications for gene therapy. Gene Ther. 10, 955–963 (2003).

25. Monahan, P.E., Jooss, K. & Sands, M.S. Safety of adeno-associated virus gene therapyvectors: a current evaluation. Expert Opin. Drug Saf. 1, 79–91 (2002).

26. Lu, Y. Recombinant adeno-associated virus as delivery vector for gene therapy–a review.Stem. Cells. Dev. 13, 133–145 (2004).

27. Flotte, T.R. et al. Phase I trial of intramuscular injection of a recombinant adeno-associated virus alpha 1-antitrypsin (rAAV2-CB-hAAT) gene vector to AAT-deficientadults. Hum. Gene Ther. 15, 93–128 (2004).

28. Moss, R.B. et al. Repeated adeno-associated virus serotype 2 aerosol-mediated cysticfibrosis transmembrane regulator gene transfer to the lungs of patients with cysticfibrosis: a multicenter, double-blind, placebo-controlled trial. Chest 125, 509–521(2004).

29. Manno, C.S. et al. AAV-mediated factor IX gene transfer to skeletal muscle in patientswith severe hemophilia B. Blood 101, 2963–2972 (2003).

30. Janson, C. et al. Clinical protocol. Gene therapy of Canavan disease: AAV-2 vector forneurosurgical delivery of aspartoacylase gene (ASPA) to the human brain. Hum. GeneTher. 13, 1391–1412 (2002).

31. Luo, J. et al. J. Subthalamic GAD gene therapy in a Parkinson’s disease rat model.Science 298, 425–429 (2002).

32. Ryan, M.D., King, A.M. & Thomas, G.P. Cleavage of foot-and-mouth disease viruspolyprotein is mediated by residues located within a 19 amino acid sequence. J. Gen.Virol. 72, 2727–2732 (1991).

33. Jang, S.K., Pestova, T.V., Hellen, C.U., Witherell, G.W. & Wimmer, E. Cap-independenttranslation of picornavirus RNAs: structure and function of the internal ribosomal entrysite. Enzyme 44, 292–309 (1990).

34. Xiao, X., Li, J. & Samulski, R.J. Production of high-titer recombinant adeno-associatedvirus vectors in the absence of helper adenovirus. J. Virol. 72, 2224–2232 (1998).

35. Liu, F., Song, Y. & Liu, D. Hydrodynamics-based transfection in animals by systemicadministration of plasmid DNA. Gene Ther. 6, 1258–1266 (1999).

5 90 VOLUME 23 NUMBER 5 MAY 2005 NATURE BIOTECHNOLOGY

A R T I C L E S©

2005

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

ebio

tech

nolo

gy