Recombinant adenoviral delivery for in vivo expression of ... - Nature

Gene Therapy (1998) 5, 770–777  1998 Stockton Press All rights reserved 0969-7128/98 $12.00 http://www.stockton-press.co.uk/gt

Recombinant adenoviral delivery for in vivo expression of scFv antibody fusion proteins HA Whittington, LJ Ashworth and RE Hawkins University Department of Oncology, Bristol Oncology Centre, Horfield Road, Bristol, BS2 8ED, UK

Antibodies and their recombinant fragments have enormous potential for therapy of malignant and other diseases, but there can be problems associated with their production and purification in the quantities required for therapeutic use. We investigated the use of gene therapy for the production of such recombinant antibody fragments in vivo. We generated two recombinant adenoviruses expressing the single chain Fvs (scFvs) fused to murine GM-CSF (mGM-CSF). The scFvs used are MFE-23 which binds to a human tumour-associated marker carcinoembryonic antigen (CEA) and B1.8 which binds the hapten 4-hydroxy-3-nitro-5-iodo-phenylacetyl (NIP). Using scFvs to target GM-CSF to tumour cells should reduce the systemic toxicity of GM-CSF but retain its ability as a cytokine to induce systemic immune responses to tumour targets. Cell lines infected with the recombinant adenoviruses in

vitro express and secrete high levels of the scFv.mGMCSF fusion proteins. The scFv retains specificity while the mGM-CSF portion is fully bioactive and there is no detectable degradation of the fusion product. We also demonstrated effective in vivo expression of the scFv.mGM-CSF proteins. C57Bl/6 mice injected intravenously with the adenovirus encoding the MFE-23.mGM-CSF fusion produce high levels of the fusion protein by 2 days after infection. The scFv.mGM-CSF protein can be detected in the serum, at biologically active levels, for at least 20 days and the level of protein produced is related to the amount of adenovirus injected. This approach has the potential to streamline the testing of the many therapeutic strategies based on recombinant scFvs and we are currently testing these constructs in an animal model for antitumour activity.

Keywords: recombinant adenovirus; single-chain Fv; fusion; mGM-CSF

Introduction Antibodies have enormous potential for the therapy of malignant and other diseases. For the therapy of malignant disease the use of small fragments and fusion proteins derived from them has potential advantages.1,2 A particular format of antibody fragments which has been extensively studied in recent years is the scFv.3 These have advantages as they are small and thus penetrate tumours and solid tissues relatively well. They can be isolated directly using phage technology4 and can be subsequently engineered to enhance their affinity.5 Indeed, such fragments made from phage libraries6 appear superior to those derived from pre-existing monoclonal antibodies in tumour targeting models.7 Recently their activity in patient imaging studies has been established and this work also confirms the short half-life of such fragments in the human circulation.8 Fusion proteins with such scFv fragments are attractive for therapeutic use but the short half-life means that continuous infusion for a period of time may be required to obtain sufficient target tissue concentrations and that large quantities will need to be administered.1 Difficulties in production and purification of bulk quantities remain and an attractive

Correspondence: RE Hawkins, Medical Oncology, Paterson Institute for Cancer Research, Wilmslow Road, Manchester M20 9BX, UK Received 17 November 1997; accepted 27 February 1998

alternative is the use of gene therapy to produce such recombinant antibody fragments in vivo. There are a number of different fusion proteins which could be used1 but particularly interesting are antibody cytokine fusion molecules with their potential to recruit specific effector cells. Such fusion proteins have been described with interleukin-29 and GM-CSF.10 The latter may be particularly attractive since as well as having a direct effect on targeted cell killing10 GM-CSF appears to be a particularly effective cytokine for the induction of systemic immune responses to tumour cells.11 This is achieved by altering the local immunological environment to increase the immunogenicity of the tumour cells.11 Practically, however, it is preferable to treat tumours directly in vivo. Systemic toxicity means that it is necessary to devise methods that increase the concentration of cytokine in the immediate vicinity of the tumour. One approach has been to use gene transfer to introduce cytokine genes into tumour cells in vivo. Recombinant adenovirus has been used to introduce the GM-CSF gene into a murine lung tumour model system and suppress growth.12 However, tumours are often not readily accessible for the direct introduction of adenoviruses and in the adjuvant setting (where immunotherapy is most likely to be effective) all known tumour has been removed so it is generally preferable to target the cytokine itself if appropriate methods can be devised. To address the above issues we devised and tested an approach which avoids many of these problems. We present data describing the construction of scFv-murine

Adenoviral delivery of recombinant antibodies HA Whittington et al

GM-CSF (scFv.mGM-CSF) fusions which are efficiently expressed from eukaryotic cells and use recombinant adenoviruses to deliver this genetic construct for in vivo gene delivery. We demonstrate that cells infected with these adenoviruses both in vitro and in vivo produce bioactive fusion protein at potentially therapeutic levels. Such an approach may have many advantages for cancer immunotherapy or for the use of antibodies in other diseases.

Results Construction and in vitro expression of the scFv.mGMCSF gene fusion To determine the optimum cassette for eukaryotic expression of the scFv.mGM-CSF fusion proteins various vectors were constructed containing different combinations of promoter sequence and leader sequence with scFv.mGM-CSF DNA (Figure 1). The promoters studied were derived from the Rous sarcoma virus (RSV) and cytomegalovirus (CMV). The leader sequences were VH1 from the murine antibody heavy chain13 and the OM signal sequence from oncostatin-M.14 The expression of the B1.8.mGM-CSF fusion cloned into these vectors was studied in 791T (ATCC CRL-7798) and HeLa (ATCC CCL-2) cells. Two cell lines were used to allow for natural differences in the level of expression and protein secretion. The B1.8.mGM-CSF protein was produced from all the different expression vectors but the level of expression and secretion was higher using the CMV promoter in conjunction with the OM leader sequence in both 791T and HeLa cells (data not shown). The OM leader sequence produced higher levels of B1.8.mGM-CSF secretion than the VH1 leader with both the RSV and CMV promoters (data not shown). Hence, the combination of CMV promoter and OM leader sequence was used for subsequent experiments involving the expression of scFv.mGM-CSF proteins in eukaryotic cells. Overall this resulted in four-fold and nine-fold higher expression than our original vector (pVAC1) in HeLa and 791T cells, respectively. Generation of recombinant adenoviruses For these studies we selected the scFvs B1.85 and MFE23,6 which bind to 4-hydroxy-3-nitro-5-iodo-phenylacetyl (NIP) and carcinoembryonic antigen (CEA), respectively. The B1.8.mGM-CSF and MFE-23.mGM-CSF gene fusions were subcloned into the plasmid pAdlox to yield plasmids pAdlox.B1.8.mGM-CSF and pAdlox.MFE-23.mGMCSF. The plasmids were individually cotransfected into

Figure 1 Map of the scFv.mGM-CSF expression cassette. A restriction map showing the components of the scFv.mGM-CSF expression cassette. The HindIII–XbaI fragment containing the OM leader sequence, scFv and mGM-CSF DNA was cloned into the multiple cloning site of pAdlox for constructing the recombinant adenoviruses. The CMV promoter is already present within pAdlox. (L = the OM or VH1 leader sequence).

CRE8 cells with adenovirus ⌿5 DNA. The ⌿5 adenovirus is replication deficient and has deletions in the viral E1 and E3 regions. The packaging site is directly flanked by loxP sites so that the packaging site is deleted by recombination when the virus is cultured in CRE8 cells, a derivative of 293 cells containing the Cre recombinase gene. Hence there is negative selection for growth of the ⌿5 virus in CRE8 cells. Growth is restored by recombination with the loxP site in pAdlox, which contains a packaging site.15 The CRE8 cells promoted recombination between the loxP site in pAdlox.B1.8.mGM-CSF or pAdlox.MFE23.mGM-CSF and ⌿5 virus. Cytopathic effects on the CRE8 cells were observed at 10 days after transfection. Control CRE8 cells transfected with either of the pAdlox plasmids or ⌿5 DNA did not exhibit any cytopathic effects. The recombinant viruses were passaged twice in CRE8 cells to reduce contamination by ⌿5 virus. Digestion of viral DNA, isolated from cells at each passage, with BsaBI confirmed the presence of the recombinant viruses ⌿.B1.8.mGM-CSF and ⌿.MFE-23.mGM-CSF. The titre of virus present after the second passage was approximately 1010/ml cell lysate.

Expression of the scFv.mGM-CSF proteins in vitro The ⌿B1.8.mGM-CSF and ⌿MFE-23.mGM-CSF viruses were used to infect HeLa cells. Approximately 106 cells were infected in triplicate and medium samples were collected at time-points up to 48 h. The samples were assayed for mGM-CSF by ELISA and also examined by Western blot. At t = 0 there was no detectable scFv.mGMCSF protein in any of the medium samples, confirming that scFv.mGM-CSF present in the viral lysate had been removed by washing after infection of the HeLa cells. B1.8.mGM-CSF and MFE-23.mGM-CSF were detected in the culture supernatants from 4 h after infection (Figure 2). A Western blot of the samples from the infection confirmed the data obtained by ELISA. The ability of ⌿.MFE-23.mGM-CSF to direct the expression of the scFv fusion protein in a variety of cells was tested in two murine and five human cell lines. MFE-23.mGM-CSF was produced in all those tested. There was only a minor difference in the level of secretion of MFE-23.mGM-CSF in HeLa compared with HeLa-CEA (HeLa expressing surface CEA following transfection with full-length human CEA). The lower efficiency of adenoviral infection of murine cells resulted in lower levels of expression than obtained in human cell lines (data not shown). A 100 ng sample of each scFv.mGM-CSF protein in culture supernatant (as determined by ELISA) was examined by SDS PAGE and Western blot in comparison with 10 ng of recombinant mGM-CSF, expressed in E. coli (Sigma, St Louis, MO, USA), to check for any specific proteolytic cleavage in the fusion proteins (Figure 3). Protein bands of the correct size (46 kDa) were observed for the scFv.mGM-CSF fusion proteins. There were no other bands representing possible specific proteolytic cleavage products of the fusion protein. As the band of recombinant mGM-CSF is clearly visible, it suggests that any proteolytic products would represent a small percentage (Ⰶ10%) of the protein sample. Bioactivity of the scFv.mGM-CSF proteins Antigen binding activity of B1.8 and MFE-23 scFvs as fusion proteins with mGM-CSF proteins was tested by ELISA and FACS analysis, respectively. Samples of

771


772

culture supernatant from HeLa cells infected with ⌿.B1.8.mGM-CSF or ⌿.MFE-23.mGM-CSF were taken immediately after infection and after 24 h incubation. Samples were tested by ELISA against NIP antigen (Figure 4a). The B1.8.mGM-CSF protein binds to the NIP antigen while, as expected, the MFE-23.mGM-CSF protein does not. The B1.8.mGM-CSF and MFE-23.mGM-CSF proteins, in culture supernatants taken 24 h after infection, were tested by FACS analysis against LS 174T cells. LS 174T, a human colon carcinoma cell line (ATCC CL188), expresses high levels of CEA. The FACS analysis shows that MFE-23.mGM-CSF protein binds to LS 174T cells while B1.8.mGM-CSF does not (Figure 4b). The ability of mGM-CSF within the scFv.mGM-CSF proteins to promote cell proliferation was demonstrated by bioassay using FDC-P1 cells. FDC-P1 (ATCC CRL-

Figure 2 In vitro expression of the scFv.mGM-CSF fusion proteins. Approximately 106 HeLa cells at 70% confluence were infected at a MOI of 1000 with either ⌿MFE-23.mGM-CSF or ⌿B1.8.mGM-CSF as described in Materials and methods. The amount of mGM-CSF (ng/ml of medium), present after 4, 21 and 46 h, was determined by ELISA using the mGM-CSF ELISA kit (Endogen).

Figure 3 Western blot of MFE-23.mGM-CSF, B1.8.mGM-CSF and mGM-CSF. 100 ng MFE-23.mGM-CSF (lane A), 100 ng B1.8.mGMCSF (lane B) and 10 ng recombinant mGM-CSF, expressed in E. coli (lane C), were electrophoresed through a 7.5% SDS PAGE gel and transferred to 4.5 ␮ nitrocellulose. The proteins were detected using rabbit antimGM-CSF according to Materials and methods. The molecular weight markers are Rainbow coloured protein markers (Amersham).

Figure 4 Antigen binding activity of scFv.mGM-CSF fusion proteins. (a) ELISA of MFE-23.mGM-CSF and B1.8.mGM-CSF against NIP antigen. Wells of a 96-well microtitre assay plate were coated with 20 ␮g/ml NIP.BSA conjugate. Uncoated wells were used as a negative control. Wells were incubated with culture supernatants containing either MFE23.mGM-CSF or B1.8.mGM-CSF. Bound protein was detected using the biotinylated anti-mGM-CSF antibody from the mGM-CSF ELISA kit (Endogen) as described in Materials and methods. All samples were assayed in triplicate. (b) FACS analysis of MFE-23.mGM-CSF and B1.8.mGM-CSF against LS 174T. 2 × 105 LS 174T cells, a colon carcinoma line expressing high levels of CEA, incubated with culture supernatants of either MFE-23.mGM-CSF or B1.8.mGM-CSF. Bound fusion protein was detected with rabbit anti-mGM-CSF (Sigma) and FITC-conjugated goat anti-rabbit Ig (Sigma) according to Materials and methods. The data were acquired using a Becton Dickinson FACScan and LYSIS II software and analysed using WinMDI 2.4.


773

12103) is a murine myeloid cell line dependent on IL-3 or mGM-CSF for growth. The B1.8.mGM-CSF and MFE23.mGM-CSF proteins were compared to recombinant mGM-CSF (Sigma), expressed in E. coli. HeLa cell culture medium was used as a negative control. Samples between 0.001 ng and 50 ng were incubated with FDCP1 cells for 48 h and proliferation measured by 3H-thymidine incorporation (Figure 5). The B1.8.mGM-CSF and MFE-23.mGM-CSF proteins appear to have similar bioactivity to recombinant mGM-CSF, as determined by FDCP1 proliferation. The EC50 of the scFv.mGM-CSF proteins, which is the effective concentration of growth factor that elicits a 50% increase in cell growth in a cell-based bioassay, is approximately 0.5 ng/ml. Hence, both the scFv and mGM-CSF domains of the fusion proteins are bioactive in vitro.

Expression of MFE-23.mGM-CSF in vivo To assess the expression of MFE-23.mGM-CSF protein in vivo C57Bl/6 mice were injected intravenously with ⌿.MFE-23.mGM-CSF adenovirus. Groups of five mice (Groups A, B and C) were injected with 10 ␮l, 2 ␮l and 0.4 ␮l of adenovirus, respectively, at a titre of approximately 1012 plaque forming units (p.f.u.)/ml in a total volume of 100 ␮l. Serum samples, before and after infection, were tested by ELISA (Figure 6a). Mice receiving 10 ␮l or 2 ␮l of virus produced high levels of MFE-23.mGMCSF and were killed upon exhibiting signs of systemic toxicity caused by GM-CSF. The mice injected with 0.4 ␮l of virus produced lower levels of MFE-23.mGM-CSF (⬍20 ng/ml serum). Protein expression was detected 2 days after infection, with levels of expression falling to Figure 6 In vivo expression of MFE-23.mGM-CSF in C57Bl/6 mice. (a) ELISA of serum samples from C57B1/6 mice infected with ⌿.MFE23.mGM-CSF. Groups of five C57B1/6 mice (Groups A, B and C) were injected intravenously with 10 ␮l, 2 ␮l or 0.4 ␮l of ⌿MFE-23.mGMCSF adenovirus, respectively (titre approximately 1012 p.f.u./ml; purified by CsCl centrifugation) in a total volume of 100 ␮l. Serum samples taken up to 26 days after infection, were assayed for mGM-CSF using the ELISA kit (Endogen) and the average level of fusion protein calculated for each group of mice. Mice were killed upon exhibiting symptoms of systemic toxicity caused by the high levels of mGM-CSF. No mGM-CSF could be detected in the remaining mice at 26 days. (b) Western blot of C57B1/6 mouse serum containing the MFE-23.GM-CSF protein. C57B1/6 serum before adenoviral infection (lane A), C57B1/6 mouse serum at 5 days after adenoviral infection containing 10 ng of MFE23.GM-CSF (as determined by ELISA) (lane B) and 100 ng of MFE23.GM-CSF in culture medium from HeLa cells infected with ⌿.MFE23.GM-CSF (lane C) were electrophoresed through a 7.5% SDS PAGE gel and transferred to 4.5 ␮ nitrocellulose. The proteins were detected using rabbit anti-mGM-CSF according to Materials and methods. The molecular weight markers are Rainbow coloured protein markers (Amersham).

Figure 5 FDC-P1 cell proliferation assay for mGM-CSF bioactivity. Aliquots of 5000 FDC-P1 cells, a mGM-CSF-dependent cell line, were incubated with between 50 ng and 0.001 ng of each of the following: MFE23.mGM-CSF, B1.8.mGM-CSF and mGM-CSF. The mGM-CSF, which was prepared commercially from protein expressed from bacteria, was the positive control. Culture medium taken from HeLa cells was the negative control. All samples were set up in triplicate. After 48 h incubation, tritiated thymidine (1 ␮Ci per sample) was added for a further 8 h incubation. Proliferation was measured using a scintillation counter and is expressed as the number of counts/20 s of incorporated thymidine. There was no incorporation by the negative control. The EC50 is the concentration of mGM-CSF that elicits a 50% increase in cell growth and hence thymidine incorporation.

⬍1 ng/ml serum at 20 days after infection. No MFE23.mGM-CSF could be detected by 26 days after infection. A Western blot of mouse serum containing MFE23.mGM-CSF showed that the fusion protein was present and there was no free murine GM-CSF protein detectable in the serum (Figure 6b). Serum samples were tested for anti-adenoviral antibodies by ELISA using ⌿.MFE23.mGM-CSF-coated assay plates.16 Anti-adenoviral antibodies were not present before adenoviral infection but could be detected in serum samples from 4 days after infection (data not shown).


774

Discussion To address practical problems associated with the production and clinical use of antibodies and scFvs for immunotherapy of cancer, we produced recombinant adenoviruses encoding scFv fusion proteins as a model for the production of antibodies and their fragments in vivo. The recombinant adenoviruses, ⌿.B1.8.mGM-CSF and ⌿.MFE-23.mGM-CSF, have been generated by a method that utilises Cre recombinase15 rather than relying on the more inefficient process of overlap recombination.17 They contain a gene fusion encoding murine GM-CSF fused to the scFvs B1.8 or MFE-23, which bind to the antigens NIP or CEA, respectively. These adenoviruses are able to infect a variety of human and murine cell lines in vitro and promote expression and secretion of the scFv.mGM-CSF proteins. In the case of MFE23.mGM-CSF, the presence of CEA on the cell surface in a number of these lines does not appear to have a major effect on secretion of the fusion protein. This suggests that secretion of the fusion protein is not prevented by aggregation with or binding to the CEA intracellularly. Importantly, both parts of the scFv.mGM-CSF fusion proteins are bioactive. The scFvs are able to bind specifically to their respective antigens in vitro and the mGM-CSF domain promotes in vitro proliferation of the murine myeloid cell line FDC-P1, which is dependent on either GM-CSF or IL-3 for growth. Quantitative assessment of the mGM-CSF activity suggests full biological activity is retained and importantly no significant proteolysis can be detected. The ⌿.MFE-23.mGM-CSF virus when injected intravenously into a mouse infects murine cells efficiently enough to produce substantial amounts of the fusion protein. The MFE-23.mGM-CSF protein can be detected in mouse serum by ELISA for at least 20 days after infection, which is in accordance with previous data for the expression of proteins from adenoviral vectors in vivo.18 One of the potential advantages of using an antibodyGM-CSF fusion protein for therapy in comparison to systemic GM-CSF administered alone or in conjunction with a monoclonal antibody, is that it should be possible to achieve high concentrations of GM-CSF in the immediate vicinity of the targeted cells but maintain low levels of systemic GM-CSF, hence reducing toxicity. However, these proteins have a short half-life in the body. ScFvs and Fab⬘ fragments (which have a similar molecular weight to scFv.mGM-CSF protein) have a half-life in plasma of 2.8 and 7.5 h, respectively.19 An antibody-GMCSF fusion protein has a half-life of approximately 30 h.10 Hence, it is likely that frequent repeated administration would be required to maintain effective therapeutic levels. Using a recombinant adenovirus to direct expression of these proteins in vivo should allow low levels of expression to be maintained over a longer period of time. Levels of MFE-23.mGM-CSF exceeding 10 000 ng/ml serum have been produced in mice. These levels induce fatal toxicity, causing leukocytosis, hepatosplenomegaly and pulmonary haemorrhage.11 In this experiment mice were killed due to toxicity as indicated and the gross pathological findings were consistent with GM-CSF toxicity. Lower doses of recombinant virus produce lower potentially therapeutic but nontoxic levels (⬍20 ng/ml serum) of the protein. The production of antibodies against viral proteins and

an immune response to the scFv.mGM-CSF proteins is probably the major limitation on the duration of scFv.mGM-CSF expression. Previous in vivo studies using E1-deleted adenoviruses have shown that these antibodies can trigger a cell cytotoxic response leading to destruction of the adenoviral infected cells and hence loss of transgene expression.20 Transgene expression is substantially prolonged in studies using nu/nu mice or animals treated with immunosuppressing drugs.21 A memory-type immune response substantially diminishes the efficiency of gene transfer and hence levels of transgene expression on repeat administration of the virus.18,22 To some extent use of a different serotype of adenovirus delays and reduces the appearance of these antibodies and does not affect gene transfer.23 However, there is some evidence that overemphasis may have been placed on the anti-adenoviral immune response and that the use of non-murine and highly immunogenic transgene products is a major factor reducing the period of transgene expression.16 Overall, it is clear that there are a number of strategies which could allow longer expression from each administration and effective repeated use of recombinant adenoviruses if that is required. Certainly, the requirement for repeated administration of adenovirus in therapy protocols has prompted studies on ways of reducing the host immune response to the adenovirus and hence extending the period of transgene expression. Immunomodulatory agents that have an antagonistic effect on the immune response have been studied.24–26 Second generation adenoviral vectors containing further deletions of adenoviral genes encoding structural proteins have also been constructed. This markedly increases the period of transgene expression and highlights a role for adenoviral protein expression in eliciting a host immune response.21 The most extreme of these adenoviral vectors have all viral coding sequences deleted.27,28,15 They require a helper virus for the production of adenoviral particles. The absence of adenoviral protein expression in infected cells in vivo increases the persistence of the adenovirus. Studies in animals with induced tolerance to the transgene product have shown that expression from such vectors can last for ⬎80 days after gene delivery. In animals with no induced tolerance, expression was lost by 42 days.29 Thus, the use of these adenoviral vectors may allow us to achieve expression of scFv.GM-CSF proteins for a longer period in vivo. The level of protein produced may also be controlled by the amount of virus administered relative to body weight and also by the incorporation of gene regulatory sequences.30 The recently described positive feedback tetracycline control system may be particularly effective.31 Even with the optimal adenoviral delivery system the immune response to the expressed scFv fusion will become an issue. Its immunogenicity is low (murine scFv in mice or human scFv in patients) but for long-term expression this may be an important issue as antiidiotype responses can be generated.13 Additionally, the immunogenicity of our murine-derived scFv fusion protein could be enhanced by the presence of the immunomodulatory protein, mGM-CSF. Preliminary data suggest that there is some anti-scFv antibody response detectable 6 weeks after infection (data not shown). However, these problems could be overcome by changing the scFv used as well as the adenovirus serotype.32 Finally, although


adenoviruses are currently the most efficient method of in vivo gene delivery, in the future, other methods based on plasmid DNA may come to replace viral approaches.33 At present, in preliminary tests we could not detect any systemic expression using such approaches. These experiments show that it is possible to express a scFv with therapeutic potential at levels which are biologically active and are sustained for around 20 days. This approach has a number of theoretical and practical attractions and is being tested in an animal model for antitumour activity. In the future it may be possible to extend this system as indicated above to allow repeated administration and also to target the expression to organs of interest. For example, physical means could be used to direct expression to the liver (a common site of colorectal cancer metastasis) or the peritoneum (a common site of ovarian cancer metastasis) to obtain higher levels of scFv fusion proteins where they are most needed. It may also be possible to target expression of the scFv fusion to specific tissues either using tissue-specific promoters34 or by targeting gene delivery.35 Overall, it is hoped that this approach to immunotherapy will provide a useful adjuvant treatment for a variety of cancers.

Methods Cell culture conditions Cell lines were cultured in DMEM (Sigma) supplemented with 10% FCS, 100 U/ml streptomycin, 100 U/ml penicillin and 2 mm l-glutamine at 37°C/5% CO2. Antibodies and scFv.mGM-CSF expression cassette Two scFv antibody fragments were used. One scFv B1.8 is specific for the haptens 4-hydroxy-3-nitrophenyl acetyl/4-hydroxy-3-nitro-5-iodo-phenylacetyl (NP/NIP) and is derived from a hybridoma.36 The other, MFE-23, derived from a phage library binds to human carcinoembryonic antigen (CEA), which is highly expressed on many tumour types including lung, colorectal and ovarian cancers.6 To construct a scFv.mGM-CSF expression cassette the HindIII–PstI fragment containing the VH1 leader sequence in the eukaryotic expression vector pVAC113 was replaced with the oncostatin-M leader sequence (OM).14 The engineered signal sequence contains an optimal Kozak start sequence (ACCATGG)37 and a ClaI site introduced at 69 bp to allow for future manipulations. The mGM-CSF DNA was generated by PCR from mGMCSF cDNA using oligonucleotides to introduce NotI and XbaI restriction sites, to allow cloning into pVAC1, and replace the mGM-CSF signal sequence with a short linker peptide (Ala Ala Ala Gly) to allow fusion to the C-terminus of scFv proteins. The individual scFv DNAs were inserted into the vectors as PstI–NotI restriction fragments. Vectors to allow comparison of expression using the RSV promoter versus the CMV promoter were generated by excision of the RSV promoter on a BglII–HindIII fragment and addition of the CMV promoter from pCEP4 (Invitrogen, NV Leek, The Netherlands) as a AccI–HindIII fragment. An XbaI site present in this fragment was deleted to leave a unique XbaI site within pVAC1. Different constructs were tested to determine optimal expression by transient transfection into 791T or HeLa cells and assay of mGM-CSF in the media after 72 h using an ELISA kit (Endogen, Cambridge, MA, USA).

Production of recombinant adenoviruses The ⌿5 adenovirus , pAdlox plasmid and CRE8 cell line were generously provided by Dr S Hardy (Somatix, Alameda, CA, USA).15 The DNA encoding the OM leader sequence and scFv.mGM-CSF proteins, MFE-23.mGMCSF and B1.8.mGM-CSF, in the eukaryotic expression vector pVAC1 were excised by digestion with HindIII and XbaI. The fragments were subcloned into the HindIII and XbaI sites of plasmid vector pAdlox to yield plasmids pAdlox.MFE-23.mGM-CSF and pAdlox.B1.8.mGM-CSF. The recombinant viruses were then constructed by cotransfection of SfiI-digested pAdlox.MFE-23.mGM-CSF or pAdlox.B1.8.mGM-CSF with ⌿5 DNA into CRE8 cells, according to the method described, to yield adenoviruses ⌿MFE-23.mGM-CSF and ⌿B1.8.mGM-CSF.15 The recombinant viruses were passaged twice in CRE8 cells to reduce contamination with ⌿5 adenovirus. CRE8 cell lysates were prepared by three rounds of freeze–thawing. The viral titre was approximately 109–1010/ml cell lysate. Recombinant adenoviral particles were purified on CsCl gradients and titrated by comparison to a control virus in cytopathic effect assays, using standard methods. Adenoviral infection of human and murine cell lines Cells were cultured to 70% confluence and after removal of the culture medium, viral lysate at a multiplicity of infection (MOI) of 1000 was added to the cells in a low volume of culture medium. After incubation for 1 h at 37°C/5% CO2, the viral lysate was removed and the cells washed three times in an excess of culture medium before addition of an appropriate volume of fresh medium (5 ml/106 cells) and further incubation at 37°C/5% CO2. Culture supernatant was removed at appropriate time-points for analysis by ELISA and Western blot. Immunoassays To assay for NIP binding Falcon MicroTest III Flexible assay plates (Becton Dickinson, Oxnard, CA, USA) were coated overnight at 4°C with NIP.BSA conjugate (20 ␮g/ml) in PBS.36 After blocking for 1 h with PBS/2% BSA and washing with PBS, culture supernatants diluted in PBS/2% BSA were added to the appropriate wells and incubated for 1 h at room temperature. After washing three times, bound scFv.mGM-CSF protein was detected using the biotinylated anti-mGM-CSF antibody, streptavidin–POD conjugate and BM Blue POD substrate according to the protocol in the mGM-CSF ELISA mini-kit (Endogen, Cambridge, MA, USA). FACS analysis To assay for binding to CEA on the surface of cells aliquots of LS 174T cells (2 × 105), washed with PBS/2% BSA, were incubated with culture supernatants diluted 50:50 in PBS/2% BSA for 1 h at 4°C. Bound scFv.mGMCSF protein was detected using rabbit anti-mGM-CSF (Sigma) (1/100 in PBS/2% BSA for 30 min at 4°C) and FITC-conjugated goat anti-rabbit Ig (Sigma) (1/160 in PBS/2% BSA for 30 min at 4°C). Between incubations cells were washed three times in PBS/2% BSA. Data were acquired using a Becton Dickinson FACScan and LYSIS II software and analysed using WinMDI 2.4. SDS PAGE and Western blots To assess the size of the expressed proteins and to detect possible degradation Western blots were used.38 Samples

775


776

were electrophoresed through a 7.5% SDS polyacrylamide gel under denaturing conditions using a Mini-Protean II cell (Bio-Rad Laboratories, Hercules, CA, USA). Proteins were transferred on to nitrocellulose using the SemiPhor (Hoefer, San Francisco, CA, USA) semi-dry transfer system. Recombinant mGM-CSF and scFv.mGM-CSF proteins were detected using rabbit anti-mGM-CSF (Sigma) (1/1000 in PBS/0.1% Tween-100), HRP-conjugated goat anti-rabbit Ig (Sigma) (1/2000 in PBS/0.1% Tween-100) and then detected using either a chemiluminescent detection system according to the protocol supplied with the ECL Western blotting kit (Amersham, Little Chalfont, UK) or 3,3⬘-diaminobenzidine tetrahydrochloride (Sigma) according to the manufacturers’ instructions.

Cell proliferation assay mGM-CSF bioactivity was measured using the mGMCSF/IL-3-dependent haemopoietic precursor line FDCP1.39 FDC-P1 cells were cultured in DMEM/10% FCS/10% WEHI-3 CM at 37°C in 5% CO2. The cells were harvested, washed three times in PBS, counted and resuspended at a concentration of 105 cells/ml in DMEM/10% FCS. Recombinant mGM-CSF and culture supernatant containing MFE-23.mGM-CSF or B1.8.mGM-CSF were serially diluted in a final volume of 50 ␮l DMEM/10% FCS in 96-well flat-bottomed tissue culture microwell plates. A 50 ␮l aliquot (5000 cells) of FDC-P1 cells was then added to each well. All wells were set up in triplicate. Wells containing no growth factor (HeLa cell culture supernatant) were used as a negative control. Plates were incubated for 48 h at 37°C in 5% CO2. Proliferation was measured by incorporation of tritiated thymidine 3H-TdR (1 ␮Ci/well). After 8 h incubation, cells were harvested on to absorbent glass fibre paper and the radioactivity determined using a scintillation counter. Mouse studies Groups of five 8-week-old C57Bl/6 mice, bred under specific pathogen-free conditions and maintained according to Home Office approved institutional protocols, were injected in the tail vein with different amounts (total volume 100 ␮l) of pure ⌿C23.mGM-CSF diluted in 10 mm Tris pH 7.5, 1 mm MgCl2, 135 mm NaCl, 10% glycerol. Blood samples were collected at different time-points and the sera assayed for the presence of mGM-CSF using the mouse GM-CSF ELISA kit (Endogen, Cambridge, MA, USA).

4

5

6 7

8

9

10

11

12

13 14

15 16

17

18

19

20

Acknowledgements HAW was supported by the Friends of Bristol Oncology Centre. LJA is supported by the Cancer Research Campaign. We thank Stephen Hardy (Somatix, Alameda, CA, USA) for supply of the Cre-Lox adenoviral system and Nick Dibb (RPMS, London) for the FDC-P1 cell line.

References 1 Chester KA, Hawkins RE. Opportunities with phage technology and antibody engineering of fusion proteins. Adv Drug Del Rev 1996; 22: 303–314. 2 Chester KA, Hawkins RE. Clinical issues in antibody design. TIBTECH 1995; 13: 294–300. 3 Huston JS et al. Protein engineering of antibody binding sites:

21

22

23

24

25

recovery of specific activity in an anti-digoxin single-chain Fv analogue produced in Escherichia coli. Proc Natl Acad Sci USA 1988; 85: 5879–5883. Winter G, Griffiths AD, Hawkins RE, Hoogenboom HR. Making antibodies by phage display technology. Annu Rev Immunol 1994; 12: 433–455. Hawkins RE, Russell SJ, Winter G. Selection of phage antibodies by binding affinity: mimicking affinity maturation. J Mol Biol 1992; 226: 889–896. Chester KA et al. Phage libraries for generation of clinically useful antibodies. Lancet 1994; 343: 455–456. Verhaar MJ et al. A single chain Fv derived from a filamentous phage library has distinct tumour targeting advantages over one derived from a hybridoma. Int J Cancer 1995; 61: 497–501. Begent RHJ et al. Single-chain Fv antibody selected from a combinatorial library: clinical evidence of efficient tumour targeting. Nature Med 1996; 2: 979–984. Savage P, So A, Spooner RA, Epenetos AA. A recombinant single chain antibody interleukin-2 fusion protein. Br J Cancer 1993; 67: 304–310. Hornick JL et al. Chimeric CLL-1 antibody fusion proteins containing granulocyte–macrophage colony-stimulating factor or interleukin-2 with specificity for B cell malignancies exhibit enhanced effector functions while retaining tumor targeting properties. Blood 1997; 89: 4437–4447. Dranoff G et al. Vaccination with irradiated tumor cells engineered to secrete murine granulocyte–macrophage colony-stimulating factor stimulates potent, specific, and long-lasting antitumor immunity. Proc Natl Acad Sci USA 1993; 90: 3539–3543. Lee CT et al. Genetic immunotherapy of established tumors with adenovirus-murine granulocyte–macrophage colony-stimulating factor. Hum Gene Ther 1997; 8: 187–193. Stevenson FK et al. Idiotypic DNA vaccines against B cell lymphoma. Immunol Rev 1995; 145: 211–228. Malik N et al. Molecular cloning, sequence analysis, and functional expression of a novel growth regulator, oncostatin M. Mol Cell Biol 1989; 9: 2847–2853. Hardy S et al. Construction of adenovirus vectors through Cre– lox recombination. J Virol 1997; 71: 1842–1849. Michou AI et al. Adenovirus-mediated gene transfer: influence of transgene, mouse strain and type of immune response on persistence of transgene expression. Gene Therapy 1997; 4: 473–482. Bett AJ, Haddara W, Prevec L, Graham FL. An efficient and flexible system for construction of adenovirus vectors with insertions or deletions in early regions 1 and 3. Proc Natl Acad Sci USA 1994; 91: 8802–8806. Kozarsky KF et al. In vivo correction of low density lipoprotein receptor deficiency in the Watanabe heritable hyperlipidemic rabbit with recombinant adenoviruses. J Biol Chem 1994; 269: 13695–13702. Colcher D et al. In vivo tumor targeting of a recombinant singlechain antigen-binding protein. J Natl Cancer Inst 1990; 82: 1191–1197. Yang Y et al. Cellular immunity to viral antigens limits E1deleted adenoviruses for gene therapy. Proc Natl Acad Sci USA 1994; 91: 4407–4411. Engelhardt JF, Ye X, Doranz B, Wilson JM. Ablation of E2A in recombinant adenoviruses improves transgene persistence and decreases inflammatory response in mouse liver. Proc Natl Acad Sci USA 1994; 91: 6196–6200. Smith TA et al. Adenovirus mediated expression of therapeutic plasma levels of human factor IX in mice. Nat Genet 1993; 5: 397–402. Zabner J et al. Safety and efficacy of repetitive adenovirusmediated transfer of CFTR cDNA to airway epithelia of primates and cotton rats. Nat Genet 1994; 6: 75–83. Lei D et al. Nondepleting anti-CD4 antibody treatment prolongs lung-directed E1-deleted adenovirus-mediated gene expression in rats. Hum Gene Ther 1996; 7: 2273–2279. Jooss K, Yang Y, Wilson JM. Cyclophosphamide diminishes inflammation and prolongs transgene expression following


26

27

28 29 30

31

32

delivery of adenoviral vectors to mouse liver and lung. Hum Gene Ther 1996; 7: 1555–1566. Kay MA et al. Long-term hepatic adenovirus-mediated gene expression in mice following CTLA4Ig administration. Nat Genet 1995; 11: 191–197. Kochanek S et al. A new adenoviral vector: replacement of all viral coding sequences with 28 kb of DNA independently expressing both full-length dystrophin and ␤-galactosidase. Proc Natl Acad Sci USA 1996; 93: 5731–5736. Fisher KJ et al. Recombinant adenovirus deleted of all viral genes for gene therapy of cystic fibrosis. Virology 1996; 217: 11–22. Chen HH et al. Persistence in muscle of an adenoviral vector that lacks all viral genes. Proc Natl Acad Sci USA 1997; 94: 1645–1650. Hu SX et al. Development of an adenovirus vector with tetracycline-regulatable human tumor necrosis factor ␣ gene expression. Cancer Res 1997; 57: 3339–3343. Agha-Mohammedi S, Hawkins RE. Efficient transgene regulation from a single tetracycline-controlled positive feedback regulatory system. Gene Therapy 1998; 5: 76–84. Mack CA et al. Circumvention of anti-adenovirus neutralizing immunity by administration of an adenoviral vector of an alternate serotype. Hum Gene Ther 1997; 8: 99–109.

33 Liu Y et al. Factors influencing the efficiency of cationic liposome-mediated intravenous gene delivery. Nat Biotechnol 1997; 15: 167–173. 34 Rothmann T et al. Heart muscle-specific gene expression using replication defective recombinant adenovirus. Gene Therapy 1996; 3: 919–926. 35 Watkins SJ, Mesyanzhinov VV, Kurochkina LP, Hawkins RE. The adenobody approach to viral targeting: specific and enhanced adenoviral gene delivery. Gene Therapy 1997; 4: 1004–1012. 36 Hawkins RE, Winter G. Cell selection strategies for making antibodies from variable gene libraries: tapping the memory pool. Eur J Immunol 1992; 22: 867–870. 37 Kozak M. Point mutations define a sequence flanking the AUG initiator codon that modulates translation by eukaryotic ribosomes. Cell 1986; 44: 283–292. 38 Towbin H, Staehelin T, Gordon J. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications. Proc Natl Acad Sci USA 1979; 76: 4350–4354. 39 Dexter TM et al. Growth of factor-dependent hemopoietic precursor cell lines. J Exp Med 1980; 152: 1036–1047.

777