A DNA vaccine encoding genetic fusions of carcinoembryonic antigen(CEA) and granulocyte/macrophage colony-stimulating factor (GM-CSF)
Jose Lima, Connie Jenkins, Antonio Guerrero, Pierre L. Triozzi,Denise R. Shaw, Theresa V. Strong∗
Division of Hematology and Oncology, Department of Medicine, and the Comprehensive Cancer Center,University of Alabama at Birmingham, WTI 558, 1824 6th Avenue South, Birmingham, AL 35294-3300, USA
Received 16 December 2003; received in revised form 3 August 2004; accepted 24 August 2004Available online 7 October 2004
locyte/macrophage colony-stimulating factor (GM-CSF) were examined. Immunization of C57BL/6 mice with the CEA–GMCSlasmids in a three injection, high-dose immunization schedule led to T cell and antibody responses specific for CEA. Mice injeEA–GMCSF fusion plasmids also developed IgG autoantibodies to GM-CSF. Tumor challenge with the CEA-expressing syngendenocarcinoma line, MC38-CEA-2, showed delayed tumor growth in mice immunized with the CEA–GMCSF fusion plasmids butrotection in mice immunized with plasmid encoding CEA alone. In contrast, a single low-dose immunization with CEA–GMCSlasmids provided better tumor protection than low-dose CEA plasmid alone and resulted in lower titers of GM-CSF antibodies.2004 Elsevier Ltd. All rights reserved.
Plasmid DNA vaccines encoding tumor-associated anti-ens have emerged as a potentially nontoxic adjuvant ther-py for cancer. DNA immunization offers several advantagesompared to other types of vaccines. Chief among these isntracellular production of the immunogen, leading to the in-uction of long-term cell mediated immunity[1]. DNA vac-ines are also relatively easy to prepare, stable, and compar-tively inexpensive. Importantly, they do not induce vector
mmunity, making repeat dosing feasible, and they have in-erent adjuvant effects due to the presence of unmethylatedpG dinucleotides[2]. Phase I trials have shown that DNAaccines are generally safe and well tolerated, and evidencef immune response has been demonstrated for several vac-ines directed against foreign antigens[3–6]. Recently, we
reported the results of a phase I study using plasmid Dencoded carcinoembryonic antigen (CEA), a human tuantigen which is overexpressed by many common cance[7].In this trial in patients with advanced colorectal carcinothere were no objective clinical responses, and in vitrodence of vaccine-induced anti-CEA immune responsesfound in only a minority of vaccinated patients. This suggthat additional immunostimulatory signals or adjuvantsbe needed to break tolerance to weakly immunogenic hutumor antigens so as to elicit effective antitumor immunesponses.
One potential strategy to enhance the activity of DNA vcines is to utilize cytokines to attract, differentiate and actiantigen-presenting cells (APCs). Granulocyte/macropcolony-stimulating factor (GM-CSF) may be a particulaeffective adjuvant for cancer vaccine approaches[8]. Themechanism of GM-CSF adjuvant activity appears to be mated in part by chemo-attraction and activation of APC win turn internalize, process and present tumor antigens to
phocytes[9,10]. GM-CSF has been shown in mice to pref-erentially expand myeloid dendritic cells (DCs) and to en-hance responses to vaccines by increasing both the numberand the maturation state of local DC[11,12]. GM-CSF alsoenhances macrophage phagocytic activity, major histocom-patibility class II molecule expression, antigen-processingactivity and tumor cell cytotoxicity[8].
It has been shown that immunization of C57BL/6 micewith a plasmid encoding CEA efficiently induces CEA im-mune responses that are protective against challenge withmouse tumor cells expressing human CEA[13], and thatco-delivery of mouse GM-CSF on a separate plasmid canenhance the immune response[14]. Physical linkage of theexpressed target antigen and GM-CSF in situ may more effec-tively target the uptake and presentation of antigen by GM-CSF-receptor expressing APC. Several chimeric fusion pro-teins containing GM-CSF have been reported which retainbiological activity and exhibit enhanced immunogenicity ofthe tumor antigen[15–17]. Here, we show that fusion of CEAwith GM-CSF in a plasmid vaccine elicits a more potent anti-CEA immune response in a mouse model of adenocarcinoma,but that the fusion construct can elicit potentially deleteriousGM-CSF autoantibodies.
2
2
c M.C useG int ctorL toa FP,w lto,C L3v ),w 38-CAa T-GS ncese ) ini n-t tived peatd am-p inalG 25a bes intoS chw ion
plasmids with CEA at the amino terminus. For fusionconstructs encoding CEA at the carboxy terminus, CEA wasamplified from MC38-CEA-2 cells using forward primer 5′-GGAAGGTACCAGCAGGTTCAGGTGGCGGAGGCTCT-AAGCTCACTATTGAATCCACGC-3′ and reverse primer5′-CCTTTCTAGATCAAGATGCA GAGACTGTGATGC-TCTTG-3′, with KpnI and XbaI sites underlined and theflexible linker in italics, again removing the GPI anchor site,and the cDNA was cloned intoKpnI/XbaI sites of pNGVL3.
For construction of plasmids encoding mouse GM-CSFfused to the carboxyl terminus of CEA(70), GM-CSF wasamplified from pGMCSF using forward primer 5′-GGA-AGGTACCAGCACCCACCCGCTCACCCATC-3′ and re-verse primer 5′-CCTTTCTAGATCATTTTTGGCTTGGTT-TTTTGCA-3′ (KpnI and XbaI sites underlined), and thefragment cloned into theKpnI/XbaI sites of pCEA(70)to generate the plasmid called pCEA(70)–GMCSF.For fusion of GM-CSF at the amino terminus ofCEA, GM-CSF was amplified using forward primer5′-GGAAGTCGACATGTGGCTGCAGAATTTACTTTTC-3′ and reverse primer 5′-GCCTGAATTCCTTTTTGGCTT-GGTTTTTTGCATTC-3′ (SalI andEcoRI sites underlined)and cloned into theSalI/EcoRI site of the plasmid con-taining CEA with flexible linker at the amino terminus.All constructs were verified by complete DNA sequencing( ility,U idD cyD )u tem( er’sd
2
EA-2 freyS stc ol-l nesw on,V an,U celll edig 8)c
2
ereo an-c d int nsiveC Alla rec-
. Materials and methods
.1. Plasmid DNA constructs
The mammalian expression plasmid pGT37[18] en-oding full length CEA was obtained from Dr. R.onry and is referred to as pCEA(full length). The moM-CSF plasmid (pNGVL1-mGMCSF) was obtained
he pNGVL1 backbone from the National Gene Veaboratories (NGVL, Ann Arbor, MI) and is referreds pGMCSF. The GFP encoding plasmid, pCMV-EGas obtained from Clontech Laboratories Inc. (Palo AA). Additional plasmids were constructed in the pNGVector (NGVL). A short, secretory form of CEA, CEA(70as obtained as a PCR product from the cell line MCEA-2 [19] using the following primers: forward 5′-GTA-GTCGACGCGACCATGGAGTCTCCCTCGGCCCC-3′nd reverse 5′-CCTTGAATTCCAGAGCCTCCGCCACCAACCTGCTGATGCAGAGA CTGTGATGCT-3′, withalI andEcoRI sites underlined and reverse primer sequencoding a flexible linker (amino acids AGSGGGGS
talics. The MC38-CEA-2 cell line contains a spoaneous internal deletion within the three repetiomains of CEA which creates a single chimeric reomain [19]. The reverse primer was designed tolify CEA sequences upstream of the carboxyl termPI membrane-anchoring region, omitting the finalmino acid residues of CEA, so the product wouldecreted. The amplified DNA fragment was clonedalI/EcoRI sites of pNGVL3 to generate pCEA(70) whias used for immunization and for construction of fus
Center for AIDS Research DNA Sequencing Core Facniversity of Alabama at Birmingham). Validated plasmNA constructs were prepared from Max EfficienH5� cells (Life Technologies Inc., Rockville, MDsing endotoxin-free qiagen plasmid purification sysQiagen Inc., Valencia, CA), according to manufacturirections.
.2. Cell lines
The murine colon adenocarcinoma cell line MC38-C[19], expressing human CEA, was provided by Dr. Jefchlom, National Cancer Institute. C2C12, a mouse myoblaell line, was obtained from the American Type Culture Cection (ATCC CRL-1772, Manassas, VA). These cell liere maintained in DMEM medium (Mediatech, HerndA) containing 10% heat inactivated FCS (Hyclone, LogT). The GM-CSF/IL-3 dependent murine bone marrow
ine, FDC-P1 (ATCC CRL-12103) was routinely culturn DMEM containing 10% heat inactivated FCS, 4 mMl-lutamine (Mediatech) and 25% WEHI-3 (ATCC TIB-6onditioned medium[20].
.3. Mice
Wild-type female C57BL/6 mice 4–8 weeks of age wbtained from the National Cancer Institute, Frederick Cer Research Facility (Frederick, MD) and were househe Pathogen-Free Rodent Shared Facility (Compreheancer Center, University of Alabama at Birmingham).nimal procedures were performed in accordance with
J. Lima et al. / Vaccine 23 (2005) 1273–1283 1275
ommendations for the proper care and use of laboratory ani-mals.
2.4. Immunoprecipitation, Western blot analysis, andquantitation of fusion proteins
To validate plasmid constructs, 1�g of plasmid DNA wastransfected into 60–80% confluent C2C12 cultures in 12-wellplates using 15�l of Lipofectamine (Life Technologies) asrecommended by the manufacturer. Forty-eight hours aftertransfection, 1 ml of media from transfected cultures wascollected, concentrated 10-fold, and subjected to immuno-precipitation with 5�g/ml of biotinylated single-chain CEAantibody ([21], kindly provided by Dr. M.B. Khazaeli) by in-cubation at 4◦C overnight with rotation, followed by additionof 5�g/ml of streptavidin-agarose beads (Sigma ChemicalCo.–Aldrich, St. Louis, MO) and further incubation at roomtemperature for 2 h with rotation. Tubes were centrifuged at1000× g for 2 min, supernates discarded, and pellets washedonce with PBS. Beads were resuspended in sample buffer(50 mM Tris–HCl, pH 6.8, 100 mM DTT, 2% SDS, 0.1%bromophenol blue, 10% glycerol) and proteins separated byelectrophoresis on a 10% polyacrylamide SDS gel. Proteinswere electroblotted to nitrocellulose membranes (BioRad,Richmond, CA) for 1 h at 4◦C and 75 V. Membranes wereb on-f bedw toC eG /2%n reds witha gG org Inc.,B lk.M evel-o b-o unto tf by ane form A).T d byc rmedi
2
entC 48 hl lter( n-t ,C in9 -s the
concentrated transfected cell supernates. Cultures were thenpulsed with 1�Ci/well 3H-thymidine (New England Nuclear,Boston, MA), and cultured for an additional 12 h before har-vest using a Micro 96 cell harvester (Skatron Instruments,Sterling, VA). Assay of incorporated radioactivity was per-formed with the Matrix 9600 Direct Beta Counter (Packard,Downers Grove, IL). To assay for neutralization of GM-CSFfunction by serum antibodies, FDC-P1 cells were culturedas above in the presence of 2.5 ng/ml of recombinant mouseGM-CSF (Peprotech, Rocky Hill, NJ) with the addition of0.22�m filtered immune mouse sera at a final dilution of1:12.
2.6. DNA vaccination and tumor challenge
Two immunization protocols were used. In the high-intensity protocol, mice received 50�g of plasmid intramus-cularly (i.m.) in the tongue three times, 3 weeks apart. Seven-teen days after the last immunization, mice were challengedwith 3× 105 MC38-CEA-2 cells s.c. in the right flank. In thelow-intensity protocol, a single dose of 5�g of plasmid wasinjected i.m., with tumor challenge 3 weeks later as above.Tumors were measured every 2–3 days and volume (cm3)calculated by the formulaL×W2/2.
2
tar3 In-ds andt ruma ses du-p ew atedw ora y)d Af-t ffer,p per-a aMaxm ularD latesw withot Fore mal-i oata pu-r wnc de-t es.T ove onE F( ated
locked with Tris-buffered saline (TBS) containing 2% nat milk for 2 h at room temperature. Membranes were proith either 2.5�g/ml COL-1 mouse monoclonal antibodyEA (NeoMarkers, Fremont, CA) or 5�g/ml rat anti-mousM-CSF antibody (Pharmingen, San Diego, CA) in TBSonfat milk for 2 h, washed three times with Tris-buffealine plus 0.5% Tween-20 (TBST), and then incubatedlkaline phosphatase (AP) conjugated goat anti-mouse Ioat anti-rat IgG2 (Southern Biotechnology Associatesirmingham, AL) diluted 1:2000 in TBS/2% nonfat miembranes were washed three times with TBST and dped using the AP Substrate Kit IV BCIP/NBT (Vector Laratories Inc., Burlingame, CA). To quantitate the amof GM-CSF produced by the plasmids, 50�l of supernatan
rom C2C12 cells, transfected as above, were measurednzyme linked immunosorbant assay (ELISA) specificouse GM-CSF (Biosource International, Camarillo, Che amount of GM-CSF produced per ml was calculateomparison with a standard curve. The assay was perfon quadruplicate.
.5. GM-CSF functional activity assay
One�g of plasmid DNA was transfected into conflu2C12 cells as above, and culture medium was collected
ater and concentrated 10-fold with a Centricon YM-10 fiMillipore Corporation, Bedford, MA), with the final concerate sterilized by filtration through 0.22�m (Spin-X, Costarambridge, MA). FDC-P1 cells were cultured for 48 h6-well flat bottom plates at 5× 104 cells/well, in the abence of WEHI-3 conditioned medium, with 50% (v/v) of
.7. ELISA for serum antibodies
For CEA antibody detection, 96-well EIA plates (Cos590) were coated with human CEA protein (Fitzgeraldustries International Inc., Concord, MA) at 1�g/ml in boratealine (BS) buffer, pH 8.4, for 4 h at room temperature,hen blocked with borate saline plus 1% (w/v) bovine selbumin (BS–BSA). Serial three-fold dilutions of mouerum in BS–BSA (1:50 to 1:109,350) were added tolicate wells and incubated overnight at 4◦C. Plates werashed with PBS + 0.05% (v/v) Tween-20 and incubith either AP conjugated goat anti-mouse IgG, anti-IgMnti-IgG isotypes�1,�2a,�2b,�3 (Southern Biotechnologiluted 1:2000 in BS–BSA for 4 h at room temperature.
er washing, AP substrate (Sigma) in diethanolamine buH 9.0, was added and incubated for 20 min at room temture. Absorbance was measured at 405 nm on a Versicroplate reader using SoftMax Pro software (Molecevices, Sunnyvale, CA). Absorbance on CEA-coated pas corrected for absorbance on parallel plates coatedvalbumin (Sigma). COL-1 mouse monoclonal�2a antibodyo CEA (NeoMarkers) was used as a positive control.stimation of antibody isotype content, data were nor
zed to artificial controls using EIA wells coated with gnti-mouse Ig (H + L) and subsequently incubated withified mouse IgM, IgG1, IgG2a, IgG2b or IgG3 at knooncentrations (Southern Biotechnology), followed byection with the� or � isotype-specific antibody conjugato detect GM-CSF antibodies, sera were assayed as abIA plates coated with 1�g/ml recombinant mouse GM-CS
Peprotech) and subsequently incubated with AP conjug
1276 J. Lima et al. / Vaccine 23 (2005) 1273–1283
goat anti-mouse IgG (Southern Biotechnology) followed byAP substrate and absorbance measurement.
2.8. Cytokine release assays
Single cell suspensions of splenocytes were preparedby mincing and forcing spleen tissue through a 100�msterile nylon strainer (Falcon 35-2360) in PBS. Erythro-cytes were removed by hypotonic lysis and cells culturedin RPMI-1640 + 10% FCS, 4 mMl-glutamine and 12.5�M�-mercaptoethanol at 1× 105 cells/well in round bottom 96-well plates (Linbro 75-042-05). Cells were cultured in thepresence of 25�g/ml purified human CEA protein (AspenBioincorporated, Littleton, CO), or as negative controls, me-dia alone or 50�g/ml ovalbumin (Sigma). After 3 days, cul-ture supernatants were collected and assayed for mouse IFN-� and IL-4 by ELISA kits (Biosource International, Camar-illo, CA) according to the manufacturer’s instructions.
3. Results
3.1. Construction and validation of fusion plasmids
andm SFp ino[ m-p r. Toa ns,w theM ris-i ainso vet pro-t tides SF)w
usionp et d byi owni , se-c ecificf SFp thes -d ithp ing8 -t rtiono asd SFd
Fig. 1. Validation of plasmid-encoded fusion protein secretion and function.(A) Supernates from C2C12 cells transfected with the indicated plasmidswere concentrated, subjected to immunoprecipitation with an anti-CEA an-tibody, separated by electrophoresis and transfered to nitrocellulose mem-branes. Proteins were detected with either rat anti-mouse GM-CSF (left)or mouse anti-human CEA (right). Supernates from cells transfected withplasmids encoding green fluorescent protein (pGFP) or mouse GM-CSF(pGMCSF) served as negative controls. (B) Cytokine activity of GM-CSFfusion proteins was assessed by proliferation of FDC-P1 cells, measured by3H-thymidine uptake, following incubation of cells with supernates fromC2C12 cells transfected with the indicated plasmids. Data are mean incor-porated cpm for quadruplicate samples, with error bars showing S.D. of themean.
3.2. High-intensity plasmid immunization
We first tested plasmid immunization using a high-intensity regimen. Mice were immunized three times with50�g plasmid DNA every 3 weeks. Sera from immunizedmice were assayed for CEA antibodies 2 weeks following thefirst and the third immunizations. None of the control groups(naıve, empty vector or pGM-CSF alone) had detectable an-tibodies against CEA. In contrast, all CEA containing plas-mids elicited IgG antibody responses to CEA by day 14 afterthe first immunization (data not shown), with higher titersobserved on day 56, two weeks after the third immuniza-tion (Fig. 2A). Isotype analysis of CEA antibodies at day 56(Fig. 3A) demonstrated a similar pattern of responses for all
Two plasmids encoding fusions between CEA(70)ouse GM-CSF were designed, with mouse GM-Claced either on the carboxyl [pCEA(70)–GMCSF] or am
pGMCSF–CEA(70)] terminus of CEA, and the fusion coonents separated by a glycine/serine-rich flexible linkellow flexibility in the construction of genetic fusion proteie cloned a short version of CEA, CEA(70), derived fromC38-CEA-2 cell line, which contains a spontaneously a
ng deletion within the homologous internal repeat domf CEA [19]. CEA(70) was additionally modified to remo
he GPI membrane anchor region so that the encodedein would be efficiently secreted. The natural signal pepequence of the amino-terminal moiety (CEA or GM-Cas preserved.To validate expression and secretion of the encoded f
roteins, supernatants from C2C12 cells transfected with thwo fusion plasmids or control plasmids were evaluatemmunoprecipitation and Western blot analysis. As shn Fig. 1A, proteins of the expected size were producedreted and appropriately recognized by antibodies spor CEA and mouse GM-CSF. Quantitation of the GM-Croduced by each plasmid was performed by ELISA ofupernatants from transfected C2C12 cells. The plasmids prouced similar amounts of GM-CSF by this analysis, wGMCSF, pCEA–GMCSF, and pGMCSF–CEA produc99± 52 pg, 866± 50 pg, and 837± 49 pg of GM-CSF pro
ein per ml of supernatant, respectively. The GM-CSF pof both plasmid fusion proteins was biologically activeetermined by the ability to support growth of the GM-Cependent cell line FDC-P1 in vitro (Fig. 1B).
J. Lima et al. / Vaccine 23 (2005) 1273–1283 1277
Fig. 2. CEA antibody titers in sera from DNA immunized mice. Pools ofsera (n= 12) from each group collected after immunization were assayed byELISA for reactivity with purified human CEA. Each dilution was tested induplicate, and mean absorbance on plates coated with CEA was correctedfor absorbance on parallel plates coated with ovalbumin. (A) High-intensityimmunization protocol sera collected on immunization day 56, 14 days afterthe third 50�g vaccine dose. (B) Low-intensity immunization protocol seracollected on day 20 following a single 5�g vaccine dose.
CEA-containing plasmids, with the predominant isotype be-ing IgG1, and detectable IgG2a and IgG2b, but no IgG3 orIgM antibodies to CEA.
To evaluate CEA-specific T cell activation, spleen cellsfrom day 56 of immunization were stimulated in vitro withhuman purified CEA protein and culture supernatants assayedfor IFN-� and IL-4 release by ELISA. Spleen cells frommice immunized with all CEA containing plasmids showedCEA-specific IFN-� release (Fig. 4), but no antigen-specificIL-4 release was detected (data not shown). Together withthe antibody isotype analysis, these data support inductionof a Th1 type immune response in mice immunized with
CEA-encoding plasmids, although the presence of high titerCEA-specific IgG1 antibodies suggests the response was notexclusively Th1.
Mice were challenged with syngeneic MC38-CEA-2 cells17 days following the third immunization, and tumor growthmonitored (Fig. 5A). All three control groups demonstrated100% progressive tumor growth by day 21 after challenge.In contrast, all mice immunized with CEA-expressing plas-mids (including fusion plasmids) were tumor free until day38 following challenge (p< 0.0001 compared to negativecontrols). After this time, late onset tumor growth was ob-served in 60–70% of the mice injected with either of theGMCSF–CEA fusion plasmids, whereas 90% of the miceimmunized with pCEA(70) and 100% of the animals vac-cinated with pCEA(full length) were tumor free at day 60.The difference in tumor free survival at day 60 was signifi-cant (p≤ 0.026) for the CEA–GMSCF fusion plasmids com-pared to the CEA alone plasmids, pCEA(70) and pCEA(fulllength).
GM-CSF recombinant proteins have been reported to in-duce autoantibodies, which in turn might decrease the effec-tiveness of vaccine immune response[16,17,22]. We there-fore evaluated serum antibody to recombinant mouse GM-CSF by ELISA. As shown inFig. 6A, sera from control im-munization groups showed no detectable GM-CSF reactivity,w GM-C M-C ionc (datan -s t as s toG tan-d sig-n ncer ed totv onw SF)dw lizec m-b acci-n -s withb hert sionp SFb
3
midi s im-m dyr
hereas sera from mice vaccinated with the pCEA(70)–SF fusion had significant titers of IgG reactive with GSF; mice immunized with the pGMCSF–CEA(70) fusonstruct showed similar levels of such autoantibodiesot shown). Sera from all mice (n= 24) immunized with fuion constructs were individually evaluated by ELISA aingle dilution (1:100) and all had detectable antibodieM-CSF, with a mean absorbance reading of 0.70 (sard deviation 0.31, range 0.24–1.4). There was noificant difference between the mean ELISA absorbaeadings from the mice that developed tumors comparhose that remained tumor free (mean± S.D. of 0.67± 0.26ersus 0.74± 0.44, respectively). Of note, immunizatiith plasmid encoding mouse GM-CSF alone (pGM-Cid not elicit detectable autoantibodies (Fig. 6A). To testhether the elicited GM-CSF antibodies could neutraytokine activity, FDC-P1 cells were cultured with recoinant mouse GM-CSF in the presence of sera from vated mice (Fig. 6B). Inhibition of FDC-P1 growth was oberved in the presence of sera from mice immunizedoth GMCSF–CEA fusion plasmids, but not in the ot
ested sera, indicating that autoantibodies induced by fulasmid immunization could specifically neutralize GM-Ciologic activity.
.3. Low-intensity plasmid immunization
We next tested the effectiveness of a low-intensity plasmmunization schedule, and also evaluated whether thi
unization regimen would elicit an anti-GM-CSF antiboesponse. Mice received a single injection of only 5�g of
1278 J. Lima et al. / Vaccine 23 (2005) 1273–1283
Fig. 3. CEA antibody isotypes in sera from plasmid immunized mice. Pools of sera (n= 12) from each immunization group as in Fig. 2 were assayed by ELISAon CEA-coated plates and developed with AP conjugated antibodies specific for the indicated mouse immunoglobulin heavy chain isotypes. Data representmean absorbance of duplicates at a single serum dilution relative to absorbance of the 5�g/ml mouse isotype artificial standard. (A) High-intensity protocolsera as in Fig. 2 tested at a 1:450 dilution. (B) Low-intensity protocol immunization sera as in Fig. 2 tested as a 1:50 dilution.
Fig. 4. CEA-specific IFN-� release by mouse spleen cells following high-intensity DNA immunization. Pooled splenocytes (n= 2) collected on day 56 of thehigh-dose immunization schedule were cultured with either CEA, ovalbumin or medium alone for 3 days. Culture supernates were collected and assayed forIFN-� by ELISA. Data represent mean IFN-� concentrations of duplicate samples, with error bars showing S.D. of the means.
J. Lima et al. / Vaccine 23 (2005) 1273–1283 1279
Fig. 5. Tumor-free survival of mice immunized with DNA vaccines. Immunized mice were challenged s.c. with MC38-CEA-2 cells and tumor developmentassessed every 2–3 days. Data are presented as percent tumor-free survival. (A) High-intensity immunization protocol mice (n= 10) were challenged with tumorcells 17 days following the third vaccination with 50�g of the indicated plasmids. (B) Low-intensity immunization protocol mice (n= 20 in each group, exceptfor the pCEA(70)–GMCSF group wheren= 19) were challenged with tumor cells 21 days following a single vaccination with 5�g of the indicated plasmids.Data presented are combined results from two independent experiments.
plasmid vaccines, a 10-fold lower dose than that used foreach of the three immunizations of the high-intensity proto-col. We had previously determined that this regimen protectsapproximately 50% of mice immunized with plasmid encod-ing full length CEA from MC38-CEA-2 tumor challenge (14,and data not shown). On day 10 following the single low-doseimmunization, no antibodies against CEA were detected inany of the injected mice (data not shown). By day 20, serumIgG antibodies to CEA were detected in all groups vacci-nated with CEA containing plasmids, but not in the controlimmunization groups (Fig. 2B). CEA antibody titers obtainedwith low-intensity immunization regimen were substantiallylower than those obtained with the high-intensity schedule(Fig. 2A). For low-intensity immunization with plasmids en-coding CEA alone, CEA-specific antibody IgG isotypes pro-files were similar to those observed following high-intensityimmunization (Fig. 3B). However, mice immunized with thelow-intensity regimen of CEA–GMCSF fusion plasmids ex-hibited lower proportions of IgG2a and IgG2b serum anti-bodies to CEA as compared to the respective high-intensityimmunization sera (Fig. 3). No IgM antibodies to CEA were
detected in any studied sera. Cytokine release assays usingspleen cell cultures as described above for the high-intensityimmunization failed to produce any detectable IFN-� or IL-4 in the low-intensity immunized mouse cells cultured withCEA protein (data not shown).
GM-CSF autoantibody titers were also substantially lowerfollowing low-intensity as compared to high-intensity im-munization. Again, IgG autoantibodies were detected onlyin mice vaccinated with the fusion constructs, and not inmice immunized with pGMCSF alone or with pGMCSF co-administered with pCEA(70) (Table 1). Neutralization assaysusing the GM-CSF-dependent FDC-P1 cell line and recom-binant mouse GM-CSF failed to demonstrate neutralizing ac-tivity in sera of mice immunized with low-intensity plasmidfusion proteins (data not shown).
Fig. 5B presents MC38-CEA-2 tumor challenge data frommice immunized with the low-intensity protocol, combiningdata from two independent experiments. Control mice im-munized with pGMCSF alone developed tumors by day 25following challenge. In contrast, 60% of mice immunizedwith pCEA(full length) and 50% of mice immunized with
1280 J. Lima et al. / Vaccine 23 (2005) 1273–1283
Fig. 6. GM-CSF autoantibodies following immunization with fusion plas-mids. (A) Pools of sera (n= 12) from high-intensity immunization groups asin Fig. 2 were assayed by ELISA on plates coated with recombinant mouseGM-CSF in duplicate, and the mean absorbance corrected for absorbance onparallel plates coated with ovalbumin. (B) FDC-P1 cells were cultured in thepresence of recombinant mouse GM-CSF, in the presence of day 56 pooledsera (1:12 final dilution) from the indicated high-intensity immunized mice.After 3 days, FDC-P1 cell proliferation was measured by3H-thymidine up-take. Data are mean incorporated cpm of quadruplicate samples, with errorbars indicating S.D. of the means.
Table 1Autoantibodies to mouse GM-CSF following DNA immunization
Pools of sera (n= 12) from each immunization group at a final dilution of1:100 were assayed for IgG and IgM antibodies to recombinant mouse GM-CSF by ELISA. High-dose sera were from day 56 following three immu-nizations with 50�g, and low-dose sera were from day 20 following a singleimmunization with 5�g. Data are mean absorbances of duplicate samples.
pCEA(70) were tumor free at the end of the experiment. BothGM-CSF fusion plasmids afforded a greater percent protec-tion than either CEA plasmid alone: 80% of mice immu-nized with pGMCSF–CEA(70) remained tumor free at day90 [p≤ 0.05 by comparison to pCEA(70) alone], and 65%of mice immunized with pCEA(70)–GMCSF were protected(not statistically significant compared to CEA plasmid alone).Co-injection of pGMCSF and pCEA(70) as separate plasmidsalso afforded 65% protection against tumor challenge. In con-trast to results following high-dose immunization (Fig. 5A),mice immunized with the CEA–GMCSF fusion plasmids bythe low-dose schedule did not demonstrate excess late-stagetumor development (Fig. 5B).
4. Discussion
DNA-based immunization may offer significant advan-tages compared to other forms of immunization for cancerimmunotherapy applications[1]. Importantly, the intracel-lular synthesis of the target antigen by host cells followinginjection of plasmid vaccines may promote the induction of aTh1-associated cellular immune response that is critical to thedevelopment of effective antitumor immunity. Further, DNAvaccines stimulate innate immunity via immunostimulatoryC e re-s igatera highd se tot ge fort smidD calt ciniaa le ofb mod-e c-t notp nsesi
pG dinucleotides, while escaping the specific immunponse (pre-existing or vaccine induced) that can mitepeat delivery of viral vaccine vectors[23]. Many tumor-ssociated antigens are self-antigens to which there is aegree of tolerance; and eliciting cellular immune respon
hese weak antigens represents a considerable challenumor vaccines. Despite the potential advantages of plaNA vaccines, it remains difficult to break immunologi
olerance to self antigens using this vector. Whereas vacnd/or fowlpox-based viral vectors have proven capabreaking tolerance to human CEA in transgenic mousels expressing CEA[24–27], to date, intramuscular inje
ion of naked plasmid DNAs encoding CEA alone hasroven effective in generating antitumor immune respo
n a transgenic mouse model or in the clinical setting[7].A number of strategies have been evaluated to augme
otency of immune response induction by DNA-basedunization. The co-injection of plasmids encoding sev
ytokines has been shown to enhance the cellular immesponse, humoral immune response, or both[28]. Amonghese, GM-CSF is attractive due to its ability to recruit Ao the site of antigen synthesis and its ability to stimulateration of DCs[8]. The co-injection of GM-CSF plasmids heen effective in enhancing the activity of DNA immunog
n several, but not all, preclinical studies[14,29–33]. Fusiononstructs consisting of the target antigen linked to GM-ave been tested by several groups, with the goal of en
ng immunogenicity as well as targeting the fusion protor uptake by APC expressing GM-CSF receptors[15–17].
Previous studies in a nontransgenic mouse modelEA-expressing tumors showed enhanced antitumo
J. Lima et al. / Vaccine 23 (2005) 1273–1283 1281
sponses with co-injection of plasmids encoding the humanCEA tumor antigen and GM-CSF[14]. In the current study,we examined the immunologic and anti-tumor activity of aDNA vaccine encoding CEA fused with GM-CSF in wildtype animals, as a prelude to their evaluation in the morestringent transgenic setting. The results indicate that fusionconstructs were capable of eliciting critical Th1-asssociatedcellular immune responses as evidenced by antibody iso-type induced (low intensity and high-intensity immunizationregimens) and the production of IFN-� in response to CEA(high-intensity regimen). More importantly, both immuniza-tion regimens produced in vivo tumor protection.
A potential drawback to the clinical use of GMCSF–CEAfusion construct is the induction of GM-CSF autoimmuneresponses that may interfere with antitumor immune re-sponse. Immunization with plasmids expressing cytokines orchemokines has resulted in autoantibodies capable of modu-lating immune activity in mouse studies[34,35]. More specif-ically, GM-CSF autoimmunity induction has been reportedin several clinical studies. Immunizations of cancer patientswith recombinant GM-CSF protein mixed with a tumor anti-gen[22], with an idiotype–GM-CSF fusion protein[16], orwith DC pulsed with a GM-CSF protein fusion to prostaticacid phosphatase[17] have been reported to elicit cellularand/or humoral immune response to GM-CSF. In addition,a 21h ents[ pli-c her,i SFr loodc
be-t re-s ei zinga up-p e andc gh-d rg CFa niza-t ,on cesb cinei thatd ctivev
tigeni im-m ere isn uatet e oft aveb une
system. In the 38C13 murine lymphoma, low-dose adminis-tration of GM-CSF was more effective in potentiating vaccineresponses than a high-dose[39]. Tumor cells transduced withGM-CSF were also more effective in stimulating immunity inother murine models if they were low-producers as comparedto high-producers of GM-CSF[40]. The data presented heresuggest that a lower dose of the GMCSF–CEA(70) plasmidwas better able to protect animals from challenge with CEA-expressing tumors, as directly compared to parallel immu-nization with CEA plasmids alone. Although autoantibodiesto GM-CSF were still detectable, they were of lower titer thanthose induced with the high-dose immunization schedule, andno cytokine neutralizing activity was detected. Of note, im-munization with plasmid DNA encoding mGM-CSF alonedid not induce detectable GM-CSF antibodies even when co-administered with plasmids encoding CEA alone (Table 1).
CEA is an attractive target for cancer vaccine approaches[41]. This 180 kDa glycosylated protein is displayed on thecell surface via a GPI membrane anchor and can be releasedinto the circulation via mechanisms apparently common tosuch GPI anchors[42]. In this study, we elected to constructthe GMCSF–CEA fusion plasmids using a shortened formof CEA, CEA(70), that spontaneously arose in the construc-tion of the MC38-CEA-2 cell line by deletion of internalrepeated domains of CEA[19,43]. pCEA(70) was furtherm teinw ws:F lll (70)w thes ientr ns-d F re-c liza-t pho-c endi y bel et ck-b (b)e ciatedC 70)i s arec
A-b maye t de-s e re-s uner e ands plas-m igensc thew s fort thed ol of
dministering the GM-CSF/IL-3 fusion protein PIXY 3as also elicited anti-GM-CSF antibody in cancer pati
36]. To date, there has been no evidence of clinical comations from such GM-CSF specific autoimmunity. Furt
n a mouse model of hyperimmunization with a GM-Cecombinant protein, no detrimental effects on white bells or immune responsiveness were noted[37].
In our study, only plasmids encoding protein fusionsween GM-CSF and CEA elicited detectable immuneponses to GM-CSF (Table 1, Fig. 6A). At least some of thnduced anti-GM-CSF antibodies had cytokine neutralictivity in vitro (Fig. 6B), suggesting that this may have sressed the vaccine-induced anti-CEA immune responsontributed to the late growth of tumors following the hiose immunization schedule (Fig. 5A). The kinetics of tumorowth warrant further investigation, because anti-GMSntibodies would be expected to be returning to preimmu
ion levels at the time the tumors appear[37]. Interestinglyur in vitro assays of immune response (e.g.,Figs. 2 and 3) didot reveal correlative quantitative or qualitative differenetween CEA immune responses induced by CEA vac
nducing GM-CSF autoantibodies as compared to thoseid not; however, these in vitro assays have limited predialue with respect to in vivo anti-tumor response[38].
It can be speculated that because CEA is a foreign ann these mice, CEA may have acted as a “carrier” to induce
une responses to the fused GM-CSF polypeptide. Tho homologue of human CEA in mice, which may accent
his hypothesized carrier effect. Alternatively, the structurhe GM-CSF in the context of the fusion protein may heen altered in a way to promote its activation of the imm
odified to remove the GPI anchor so that the fusion proould be efficiently secreted. Our rationale was as folloirst, pCEA(70) protein should be∼100 kDa smaller than fu
ength CEA, and thus plasmid-encoded fusions of CEAith GM-CSF may be more efficiently expressed due toignificant reduction in fusion protein size. Second, efficelease of the CEA(70)–GMCSF fusion proteins from trauced cells in vivo would be necessary to engage GM-CSeptors on large numbers of APC for subsequent internaion of the antigen for processing and presentation to lymytes. Although not statistically significant, there was a trn tumor protection studies suggesting that pCEA(70) maess effective pCEA(full length) (Fig. 5). This might be duo several factors including (a) the different plasmid baones of pCEA(full length) and the CEA(70) constructs,xpression of secreted CEA versus the membrane-assoEA and (c) loss of immunogenic domains in the pCEA(
nternal deletion. Studies to address these possibilitieurrently underway.
In summary, the incorporation of GM-CSF into DNased cancer immunotherapy protocols targeting CEAnhance efficacy. The GMCSF–CEA(70) fusion construccribed here can elicit Th1-associated cellular immunponses to CEA and elicit a protective antitumor immesponses in mice. Enhanced anti-tumor efficacy is doschedule dependent. However, as this study indicates,id encoded fusions between cytokines and tumor ant
an break immune tolerance to the cytokine moiety. Inorst case this may have long-term detrimental effect
he immunized host, and at minimum it could undermineesired cytokine immune adjuvant effects. Careful contr
1282 J. Lima et al. / Vaccine 23 (2005) 1273–1283
GM-CSF dose and appropriate monitoring will be necessaryto avoid a potentially detrimental GM-CSF autoimmune re-sponses. On the other hand, the immunogenicity of the fusionprotein may render it more effective in breaking immunolog-ical tolerance to CEA. Evaluation of the antitumor efficacy ofthis plasmid DNA vaccine in a CEA transgenic mouse modelshould allow optimization of this vaccine strategy.
Acknowledgements
This work was supported by the National Cancer Institutegrant 1 P50 CA89019 and the U.S. Army Medical ResearchAcquisition Activity award #DAMD17-00-1-0122.
References
[1] Gurunathan S, Klinman DM, Seder RA. DNA vaccines: immunology,application, and optimization. Annu Rev Immunol 2000;18:927–74.
[2] Haupt K, Roggendorf M, Mann K. The potential of DNA vaccinationagainst tumor-associated antigens for antitumor therapy. Exp BiolMed (Maywood) 2002;227:227–37.
[3] Calarota S, Bratt G, Nordlund S, et al. Cellular cytotoxic responseinduced by DNA vaccination in HIV-1-infected patients. Lancet1998;351:1320–5.
l ofency
Dis
cell-DNA
ty ofwith
nic-is B
Res
lony-atol
yte-anti-cells.
[ no-
[ nceating
[ col-notra-
[ nicene
[ s to4.
[ tionf an
anti-idiotype antibody mimicking a breast cancer-associated antigenand the cytokine GM-CSF. Hybridoma 1999;18:193–202.
[16] Chen TT, Levy R. Induction of autoantibody responses to GM-CSFby hyperimmunization with an Id-GM-CSF fusion protein. J Im-munol 1995;154:3105–17.
[17] Small EJ, Fratesi P, Reese DM, et al. Immunotherapy of hormone-refractory prostate cancer with antigen-loaded dendritic cells. J ClinOncol 2000;18:3894–903.
[18] Conry RM, LoBuglio AF, Loechel F, et al. A carcinoembryonicantigen polynucleotide vaccine for human clinical use. Cancer GeneTher 1995;2:33–8.
[19] Robbins PF, Kantor JA, Salgaller M, et al. Transduction and ex-pression of the human carcinoembryonic antigen (CEA) in a murinecolon carcinoma cell line. Cancer Res 1991;51:3657–62.
[20] Dexter TM, Garland J, Scott D, Scolnick E, Metcalf D. Growthof factor-dependent hemopoietic precursor cell lines. J Exp Med1980;152:1036–47.
[21] Rodenburg CM, Mernaugh R, Bilbao G, Khazaeli MB. Productionof a single chain anti-CEA antibody from the hybridoma cell lineT84.66 using a modified colony-lift selection procedure to detectantigen-positive scFv bacterial clones. Hybridoma 1998;17:1–8.
[22] McNeel DG, Schiffman K, Disis ML. Immunization with recom-binant human granulocyte-macrophage colony-stimulating factor asa vaccine adjuvant elicits both a cellular and humoral response torecombinant human granulocyte-macrophage colony-stimulating fac-tor. Blood 1999;93:2653–9.
[23] Conry RM, Khazaeli MB, Saleh MN, et al. Phase I trial of recombi-nant vaccinia virus encoding carcinoembryonic antigen in metastaticadenocarcinoma: comparison of intradermal versus subcutaneous ad-ministration. Clin Cancer Res 1999;5:2330–7.
[ r JW.gention676–
[ lomcell
[ gy ofs and
[ rapytumortoim-
[ ionBiol
[ es awingthol
[ ionlony-es T-
[ ngod-
toxicffectsLett
[ ssionigenVac-
[4] MacGregor RR, Boyer JD, Ugen KE, et al. First human triaa DNA–based vaccine for treatment of human immunodeficivirus type 1 infection: safety and host response. J Infect1998;178:92–100.
[5] Wang R, Epstein J, Baraceros FM, et al. Induction of CD4(+) Tdependent CD8(+) type 1 responses in humans by a malariavaccine. Proc Natl Acad Sci USA 2001;98:10817–22.
[6] Timmerman JM, Singh G, Hermanson G, et al. Immunogenecia plasmid DNA vaccine encoding chimeric idiotype in patientsB-cell lymphoma. Cancer Res 2002;62:5845–52.
[7] Conry RM, Curiel DT, Strong TV, et al. Safety and immunogeity of a DNA encoding carcinoembryonic antigen and hepatitsurface antigen in colorectal carcinoma patients. Clin Cancer2002;8:2782–7.
[8] Warren TL, Weiner GJ. Uses of granulocyte-macrophage costimulating factor in vaccine development. Curr Opin Hem2000;7:168–73.
[9] Morrissey PJ, Bressler L, Park LS, Alpert A, Gillis S. Granulocmacrophage colony-stimulating factor augments the primarybody response by enhancing the function of antigen-presentingJ Immunol 1987;139:1113–9.
10] Steinman RM. The dendritic cell system and its role in immugenicity. Annu Rev Immunol 1991;9:271–96.
11] Ishii KJ, Weiss WR, Klinman DM. Prevention of neonatal toleraby a plasmid encoding granulocyte-macrophage colony stimulfactor. Vaccine 1999;18:703–10.
12] Daro E, Pulendran B, Brasel K, et al. Polyethylene glymodified GM-CSF expands CD11b(high)CD11c(high) butCD11b(low)CD11c(high) murine dendritic cells in vivo: a compative analysis with Flt3 ligand. J Immunol 2000;165:49–58.
13] Conry RM, LoBuglio AF, Loechel F, et al. A carcinoembryoantigen polynucleotide vaccine has in vivo antitumor activity. GTher 1995;2:59–65.
14] Conry RM, Widera G, LoBuglio AF, et al. Selected strategieaugment polynucleotide immunization. Gene Ther 1996;3:67–7
15] Tripathi PK, Qin H, Bhattacharya-Chatterjee M, et al. Construcand characterization of a chimeric fusion protein consisting o
24] Kass E, Schlom J, Thompson J, Guadagni F, Graziano, GreineInduction of protective host immunity to carcinoembryonic anti(CEA), a self-antigen in CEA transgenic mice, by immunizawith a recombinant vaccinia-CEA virus. Cancer Res 1999;59:83.
25] Hodge JW, Sabzevari H, Yafal AG, Gritz L, Lorenz MGO, SchJ. A triad of costimulatory molecules synergize to amplify T-activation. Cancer Res 1999;59:5800–7.
26] Grosenbach DW, Barrientos JC, Schlom J, Hodge JW. Synervaccine strategies to amplify antigen-specific immune responseantitumor effects. Cancer Res 2001;61:4497–505.
27] Greiner JW, Zeytin H, Anver MR, Schlom J. Vaccine-based thedirected against carcinoembryonic antigen demonstrates antiactivity on spontaneous intestinal tumors in the absence of aumunity. Cancer Res 2002:6944–51.
28] Haupt K, Roggendorf M, Mann K. The potential of DNA vaccinatagainst tumor-associated antigens for antitumor therapy. ExpMed (Maywood) 2002;227:227–37.
29] De Rose R, McKenna RV, Cobon G, et al. Bm86 antigen inducprotective immune response against Boophilus microplus folloDNA and protein vaccination in sheep. Vet Immunol Immunopa1999;71:151–60.
30] Kamath AT, Hanke T, Briscoe H, Britton WJ. Co-immunizatwith DNA vaccines expressing granulocyte-macrophage costimulating factor and mycobacterial secreted proteins enhanccell immunity, but not protective efficacy againstMycobacteriumtuberculosis. Immunology 1999;96:511–6.
31] Kumar S, Villinger F, Oakley M, et al. A DNA vaccine encodithe 42 kDa C-terminus of merozoite surface protein 1 of Plasmium falciparum induces antibody, interferon-gamma and cytoT cell responses in rhesus monkeys: immuno-stimulatory eof granulocyte macrophage-colony stimulating factor. Immunol2002;81:13–24.
32] Sun X, Hodge LM, Jones HP, Tabor L, Simecka JW. Co-expreof granulocyte-macrophage colony-stimulating factor with antenhances humoral and tumor immunity after DNA vaccination.cine 2002;20:1466–74.
J. Lima et al. / Vaccine 23 (2005) 1273–1283 1283
[33] Weiss WR, Ishii KJ, Hedstrom RC, et al. A plasmid encod-ing murine granulocyte-macrophage colony-stimulating factor in-creases protection conferred by a malaria DNA vaccine. J Immunol1998;161:2325–32.
[34] Wildbaum G, Youssef S, Karin N. A targeted DNA vaccine augmentsthe natural immune response to self TNF-alpha and suppresses on-going adjuvant arthritis. J Immunol 2000;165:5860–6.
[35] Youssef S, Maor G, Wildbaum G, et al. C–C chemokine-encodingDNA vaccines enhance breakdown of tolerance to their gene productsand treat ongoing adjuvant arthritis. J Clin Invest 2000;106:361–71.
[36] Miller LL, Korn EL, Stevens DS, et al. Abrogation of the hema-tological and biological activities of the interleukin-3/granulocyte-macrophage colony-stimulating factor fusion protein PIXY321 byneutralizing anti-PIXY321 antibodies in cancer patients receivinghigh-dose carboplatin. Blood 1999;93:3250–8.
[37] Chen TT, Levy R. Induction of autoantibody responses to GM-CSFby hyperimmunization with an Id-GM-CSF fusion protein. J Im-munol 1995;154:3105–17.
[38] Lyerly HK. Quantitating cellular immune responses to cancer vac-cines. Semin Oncol 2003;30(Suppl 8):9–16.
[39] Kwak LW, Young HA, Pennington RW, Weeks SD. Vaccinationwith syngeneic, lymphoma-derived immunoglobulin idiotype com-bined with granulocyte/macrophage colony-stimulating factor primesmice for a protective T-cell response. Proc Natl Acad Sci USA1996;93:10972–7.
[40] Abe J, Wakimoto H, Yoshida Y, et al. Antitumor effect induced bygranulocyte/macrophage colony-stimulating factor gene-modified tu-mor vaccination: comparison of adenovirus- and retrovirus-mediatedgenetic transduction. J Cancer Res Clin Oncol 1995;121:587–92.
[42] Kinugasa T, Kuroki M, Yamanaka T, Matsuo Y, Oikawa S, NakazatoH, et al. Non-proteolytic release of carcinoembryonic antigen fromnormal human colonic epithelial cells cultured in collagen gel. Int JCancer 1994;58:102–7.
[43] Hand PH, Robbins PF, Salgaller ML, Poole DJ, Schlom J.Evaluation of human carcinoembryonic-antigen (CEA)-transducedand non-transduced murine tumors as potential targets foranti-CEA therapies. Cancer Immunol Immunother 1993;36:65–75.
