Gene Therapy (2000) 7, 1906–1914  2000 Macmillan Publishers Ltd All rights reserved 0969-7128/00 $15.00 www.nature.com/gt ACQUIRED DISEASES RESEARCH ARTICLE TGF-␣ antisense therapy inhibits head and neck squamous cell carcinoma growth in vivo

S Endo1, Q Zeng1, NA Burke2,YHe3, MF Melhem4, SF Watkins2, MN Lango1, SD Drenning1, L Huang3 and J Rubin Grandis1,2 Departments of 1Otolaryngology, 2Cell Biology and Physiology, 3Pharmacology, 4Pathology, University of Pittsburgh School of Medicine, and University of Pittsburgh Cancer Institute, Pittsburgh, PA, USA

Unlike normal mucosal squamous epithelial cells, head and remained localized to the nucleus for up to 3 days. Direct neck squamous cell carcinomas (HNSCCs) overexpress inoculation of the TGF-␣ antisense (but not the correspond- TGF-␣ mRNA and which is required to sustain the ing sense) construct into established HNSCC tumors proliferation of HNSCC cells in vitro. To determine whether resulted in inhibition of tumor growth. Sustained antitumor TGF-␣ expression contributes to tumor growth in vivo, cat- effects were observed for up to 1 year after the treatments ionic liposome-mediated gene transfer was used to deliver were discontinued. Down-modulation of TGF-␣ was an antisense expression construct targeting the human TGF- accompanied by increased apoptosis in vivo. These experi- ␣ gene into human head and neck tumor cells, grown as ments indicate that interference with the TGF-␣/EGFR subcutaneous xenografts in nude mice. The TGF-␣ anti- autocrine signaling pathway may be an effective therapeutic sense gene was immediately detected in the cytoplasm of strategy for cancers which overexpress this /receptor the tumor cells, translocated to the nucleus by 12 h and pair. Gene Therapy (2000) 7, 1906–1914.

Keywords: transforming alpha; head and neck cancer

Introduction by the same tumor cell provides indirect evidence of an autocrine regulatory pathway. Reports suggest that the ␣ Transforming growth factor alpha (TGF- ) is a polypep- maintenance of an autocrine loop in HNSCC is depen- tide growth factor that binds exclusively to the epidermal dent on both elevated levels of EGFR and the presence 1,2 (EGFR). It is synthesized as a of TGF-␣.30,31 ␣ larger, membrane-bound glycoprotein (proTGF- ), which TGF-␣ overexpression has been implicated as a poor is cleaved to release a soluble 50 polypeptide prognostic factor in cancers of the kidney32 and eso- ␣ ␣ (TGF- ). TGF- plays an important role in regulating phagus.16 The biologic importance of this growth factor growth and development of a variety of cells including in head and neck cancer progression is supported by our 3,4 5 6 7 , mammary epithelium, GI tract mucosa, , findings demonstrating that survival of HNSCC patients 8 gynecologic and reproductive organs, kidney, and correlates significantly with TGF-␣ protein expression 9 bladder. levels in the primary tumor, independent of other clinical ␣ Overexpression of TGF- alone, or in combination with and pathological parameters including the presence of EGFR, leads to colony formation in soft agar and tumor regional metastases (N-stage).33 The observation that ␣ formation in nude mice, thus linking TGF- to transform- elevated levels of TGF-␣ are associated with the loss of 10 ␣ ation. TGF- transgenic mice display evidence of epi- growth control in many cancers has led to the develop- thelial hyperplasia, pancreatic metaplasia and carcinoma ment of strategies aimed at down-modulating this 11 ␣ of the breast. Co-expression of TGF- and EGFR has growth factor in an attempt to decrease tumor cell also been implicated in the pathophysiology of human proliferation. cancer. Overexpression of TGF-␣ alone, or in combination with EGFR, has been reported in primary malignancies and cell lines established from a wide variety of tumors, Results including glioblastoma,12,13 cancers of the breast,14,15 eso- 16 17,18 19 20,21 phagus, pancreas, liver, colorectal region, We previously reported that TGF-␣ antisense oligonucle- 20,22 9,23 24 25 26 lung, kidney, bladder, skin, ovary, vulva, cer- otides inhibited the growth of HNSCC but not normal 8 27–29 vix, endometrium, as well as the head and neck. mucosal cells in vitro, suggesting that the growth stimul- Increased production of a growth factor and its receptor ating effects of TGF-␣ are specific for transformed squam- ous epithelial cells.34 To determine the consequences of targeting TGF-␣ in vivo, we cloned a TGF-␣ antisense Correspondence: J Rubin Grandis, Dept of Otolaryngology, University of sequence designed against the translation initiation site Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA of the human TGF-␣ gene into an expression vector based Received 31 December 1999; accepted 24 July 2000 on the U6 small nuclear RNA (snRNA) promoter. TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1907

Figure 1 Schematic representation of the TGF-␣ antisense (or sense) oligonucleotide sequences contained in the U6 expression cassette.

Transduction of the TGF-␣ antisense gene into HNSCC Similar gene localization studies were performed in tumor cells in vitro HNSCC tumors in vivo. Tumors were injected and har- Experiments were performed to elucidate the conse- vested 1 week or 2 weeks following a single injection. quences of TGF-␣ down-modulation in HNSCC cells. To The antisense gene was detected in proportion of the determine the intracellular localization of the TGF-␣ anti- tumor cells 1 week after treatment. The labeling charac- sense gene in head and neck cancer cells, fluorecein- teristics in vivo were similar at both 1 week and 2 week labeled DNA (Figure 1) was used to tranfect the HNSCC time-points (Figure 3). cells in vitro followed by examination of the tumor cells using fluorescence microscopy at different time-points TGF-␣ antisense gene therapy inhibits tumor growth (Figure 2). By 12 h, labeling was detected within the To determine whether or not the injected gene could be nucleus of sectioned cells. Furthermore, there was no detected in the tumors, DNA was harvested from the quantitative difference in the amount of labeled DNA in treated xenografts followed by PCR using specific pri- the nucleus for up to 72 h (Figure 2b–e), demonstrating mers designed to detect the recombinant TGF-␣ sense or persistence of the antisense DNA in the cells. No label antisense gene. The recombinant DNA was detected in was detectable in cells which received fluorecein-labeled all tumors in contrast to tumors treated with liposomes liposomes alone or unlabeled DNA (data not shown). alone (Figure 4). To examine the consequences of the These results demonstrate that transduction of TGF-␣ transfected TGF-␣ antisense gene on tumor growth in antisense DNA complexed with DC-chol liposomes into vivo, established HNSCC xenografts were repeatedly head and neck tumor cells, results in nuclear localization inoculated with 50 ␮g TGF-␣ antisense DNA plus 50 of the transduced gene in tumor cells for at least 3 days nmol DC-chol liposomes. Previous studies from our lab- following transfection. oratory have demonstrated the necessity of a carrier mol-

Figure 2 Intracellular localization of the TGF-␣ antisense gene in vitro. HNSCC cells transfected with FITC-tagged TGF-␣ antisense (green) and counterstained with Hoechst nuclear stain (red) were imaged by fluorescence microscopy. Nuclear localization is indicated by yellow. Five hours after transfection (a) no nuclear localization of the DNA was observed. However, at the 12 h time-point (b) a small amount of anti-sense DNA is localized to the nucleus (white arrow). By 24 h (c), nuclear localization is clearly visible and persists 48 (d) and 72 (e) h after transfection.

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1908

Figure 3 Localization of antisense gene in vivo. Panel a shows a typical H&E stained image of the tumors, panel b shows the fluorescence microscope image captured before H&E staining at 1 week. Arrows in panels a and b indicate the same cells. Panel c is a representative sample showing intracellular gene 2 weeks after injection. Bar, 25 microns.

Figure 4 Detection of sense or anti-sense TGF-␣/U6 gene(s) in represen- tative HNSCC xenografts. Tumor-bearing mice were treated with intra- tumoral injections of TGF-␣ antisense (AS) or sense (S) DNA plus lipo- somes or with liposomes alone (control). PCR detection demonstrates chimeric DNA in the treated tumors as demonstrated on an ethidium bromide-stained 1% agarose gel.

ecule such as DC-chol to achieve an antitumor effect.35 Dose-response experiments using a similar approach to target EGFR showed equal antitumor efficacy using 25 ␮gor50␮g of DNA per injection with less antitumor Figure 5 In vivo growth inhibition of HNSCC xenografts. The growth effects seen at lower doses (2.5 ␮g, 0.25 ␮g, 0.025 ␮g).35 inhibitory effects of the TGF-␣ antisense construct in a representative In contrast to tumors treated with TGF-␣ sense DNA plus experiment is demonstrated. Groups of mice received intratumoral treat- ments (three times a week) with TGF-␣ antisense DNA plus liposomes liposomes or tumors treated with liposomes alone, ᭺ ␣ ȣ „ ␣ ( ), TGF- sense DNA plus liposomens ( ), or liposomes alone ( ) 14– HNSCC xenografts inoculated with TGF- antisense 21 days following tumor implantation. All cases received nine treatments. DNA plus liposomes were growth inhibited (Figure 5). Each point represents the mean value for eight to 10 tumors from an To determine the persistence of the antitumor effects, a individual experiment that was replicated three times. Fractional tumor subset of mice treated with the TGF-␣ antisense construct volume (tumor volume as a proportion of pretreatment volume) is plotted plus liposomes were maintained and followed for up to and the standard error of tumor volumes for all points was less than 10% 1 year after the treatments were discontinued. In these of the mean. Statistical analysis was performed comparing fractional tumor volumes in the TGF-␣ antisense-treated group with the sense- mice, the tumors completely disappeared after therapy treated group at each time-point and significant values (*) were obtained was discontinued and did not re-grow (data not shown). at nearly all time-points (two-sided; P Ͻ 0.05). Although the immune system may play a role in tumor clearance, use of an athymic nude mouse model in these studies, limits the ability to implicate immune Increased apoptosis and decreased proliferation are mechanisms. associated with TGF-␣ antisense gene therapy To investigate the mechanism of the antitumor effect Modulation of TGF-␣ protein expression by TGF-␣ induced by treatment with the TGF-␣ antisense DNA antisense gene therapy plus liposomes, we examined hematoxylin and eosin To determine whether the antitumor effects observed staining of the tumors and were unable to detect a differ- with TGF-␣ antisense gene therapy were associated with ence in necrosis between treatment groups (data not decreased TGF-␣ protein expression, tumors were har- shown). To determine whether the observed growth inhi- vested following treatment and immunohistochemistry bition was associated with an increase in the rate of was performed using a TGF-␣ monoclonal antibody. apoptosis, tumors were stained for DNA fragmentation Treatment with TGF-␣ antisense gene therapy resulted (ApopTag). Results demonstrated a two-fold elevation in in decreased TGF-␣ staining intensity in the tumor cells the number of apoptotic tumor cells in xenografts treated suggesting that the antitumor effects observed are a with the TGF-␣ antisense construct plus liposomes com- direct result of abrogation of TGF-␣ pared with tumors treated with the corresponding sense (Figure 6). construct (P Ͻ 0.001) or liposomes alone (P = 0.002);

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1909

Figure 7 (a) Elevated apoptosis rates in the TGF-␣ antisense-treated tumors. Mean rates of apoptosis (number of apoptotis cells per five high power fields) in eight representative TGF-␣ antisense-treated tumors com- pared with four representative control (liposome-treated) (P = 0.002) and eight representative TGF-␣ sense-treated tumors (P Ͻ 0.001) from an individual experiment that was replicated three times. Bar denotes 95% ␣ confidence intervals. (b) Decreased Bcl-xL expression following TGF- antisense gene therapy. Bcl-x and Bax immunoblot of representative Figure 6 Decreased TGF-␣ staining intensity following TGF-␣ antisense L HNSCC tumors treated with TGF-␣ antisense DNA plus liposomes com- gene therapy. TGF-␣ immunostaining in (a) a representative tumor pared with a representative control tumor treated with liposomes alone. treated with liposomes alone, (b) a representative tumor treated TGF-␣ sense DNA plus liposomes, or in (c) a representative TGF-␣ antisense plus liposome-treated tumor. , including Bcl-2 and Bax, were not affected by TGF-␣ antisense gene therapy. (Figure 7a). To investigate the mechanism of the apop- ␣ totic consequences of abrogating TGF- gene expression, Discussion we examined expression of apoptotic regulatory proteins in treated tumors. HNSCC xenografts injected with TGF- Overexpression of TGF-␣ is prevalent in a variety of ␣ antisense DNA plus liposomes demonstrated decreased malignant neoplasms as well as established tumor cell expression of the anti-apoptotic protein, Bcl-xL compared lines. Previous studies in our laboratory demonstrated ␣ with controls (Figure 7b). This down-modulation of Bc-xL increased expression of TGF- in HNSCCs compared appeared to be specific since other apoptotic regulatory with normal mucosa, primarily as a result of activated

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1910 gene transcription.36 Patients with HNSCC are at high localization to the tumor, head and neck cancers are risk of developing additional cancers of the aerodigestive amenable to intralesional therapy due to the accessibility tract. This predisposition to tumor formation is thought of the tumors and the relatively low incidence of distant to be due to toxin (tobacco and alcohol) exposure of the metastases. To deliver the antisense gene to the tumor entire mucosa (field cancerization) leading to cumulative cells, we chose a nonvirus-mediated gene transfer strat- genetic alterations. Our earlier findings of elevated TGF- egy. This approach offers several theoretical advantages ␣ in this mucosa from HNSCC patients as well as in pre- over virally mediated gene delivery including low tox- malignant, dysplastic lesions, suggests that up-regulation icity, lack of immunogenicity and inflammatory reac- of TGF-␣ represents an early event in HNSCC carcino- tions, and the relative ease of obtaining large quantities genesis.28,29,37 The observation that TGF-␣ gene of vector. DC-chol liposomes increase DNA uptake into expression may be regulated by chemopreventive agents, cells compared with naked DNA alone.47 Several clinical such as retinoic acid in HNSCC cells, provides further trials have used DC-chol liposomes with negligible tox- evidence that TGF-␣ may serve a target for prevention icity reported, including the delivery of the allogeneic strategies in aerodigestive tract cancers.36 The critical bio- MHC gene into tumor sites and CFTR gene logic role of TGF-␣ in HNSCC is supported by our find- transfer into the lungs of patients with cystic fibrosis.48,49 ing that TGF-␣ expression levels in the primary HNSCC The chief disadvantage of liposome-mediated gene trans- tumor correlates with decreased survival, independent of fer is the relatively low transfection efficiency compared other clinical and pathologic parameters, including with that of viral vectors. lymph node metastases.33 Several promoters are available for in vivo gene transfer Previous therapeutic approaches designed to abrogate strategies. We elected to use the U6 snRNA promoter, TGF-␣ in vitro have included the use of neutralizing anti- which was originally engineered to express large bodies or antisense oligonucleotides. TGF-␣ monoclonal amounts of short RNA sequences.50 U6 snRNA is one of antibody treatment has been shown to inhibit the growth the small nuclear RNAs which participates in RNA splic- of lung cancer cell lines,38 chemically induced mouse ing and is constitutively expressed (approximately half a intestine cancer cells,39 and human colon cancer cell lines million copies per cell) in almost all cells. In direct com- in vitro.40 DNA synthesis is reportedly decreased in both parison studies, the amount of U6 product is significantly HNSCC cells41 and ovarian cancer cells26 following TGF- higher than that of expressed under the control of ␣ antibody treatment. We and others have reported inhi- the CMV promoter.47 Localization and trafficing of vector bition of cancer cell line growth following treatment with expressing antisense RNA and/or antisense oligodeoxy- TGF-␣ antisense oligonucleotides or an expression vector nucleotides depends on their chemical modification as in vitro.34,42 In our earlier study, TGF-␣ antisense oligonu- well as the vector from which it is expressed. Our finding cleotides down-modulated TGF-␣ protein expression in of labeled DNA in the nucleus at 12, 24, 48 and 72 h after both HNSCC and normal mucosal epithelial cells. How- transduction suggests that liposome-mediated transfer of ever, the normal cells were not growth inhibited by TGF- the TGF-␣ antisense construct facilitates delivery of ␣ antisense oligonucleotide treatment suggesting that the sequence into the cell which is sustained for at least 3 proliferative effects of TGF-␣ are specific for trans- days. formed cells.34 In this study, antitumor efficacy of TGF-␣ antisense Several groups have reported the consequences of gene therapy was accompanied by increased apoptosis. TGF-␣ targeting in vivo. Implanted LE2 mouse cells trans- Growth factors, such as TGF-␣, often stimulate prolifer- fected with a TGF-␣ antisense expression vector showed ation of cells expressing (or overexpressing) receptors for slower growth compared with that of parental cells trans- that growth factor. The relationship between TGF-␣ fected with control vector alone.42 Using a prostate cancer function and apoptosis is less well understood. Studies xenograft model, Rubenstein et al43 demonstrated abro- examining nontransformed cells in vitro include the gation of tumor growth following intratumoral injection observation that rat enterochromaffin-like cells demon- of large doses (Ͼ400 ␮g) of TGF-␣ antisense oligonucleo- strate decreased apoptosis when exposed to TGF-␣,51 and tides. While the adverse effects of synthetic antisense oli- exogenous TGF-␣ reduces the incidence of apoptosis in gonucleotides appear to be minimal, there is little developing mouse blastocysts.52 In cancer cells, several evidence that they are efficacious against solid reports also suggest that TGF-␣ may serve as an anti- tumors.44,45 Administration of a TGF-␣ antisense oligonu- apoptotic factor. TGF-␣ inhibits c-myc-induced apoptosis cleotide directly into tumors implanted subcutaneously in a mouse mammary tumor model53 whereas apoptosis into nude mice resulted in rapid diffusion of the radiolab- is induced in liver cancer cells expressing TGF-␣ follow- eled oligonucleotide from the tumor to the digestive and ing treatment with TGF-␣ neutralizing antibodies.54,55 urinary tracts.46 Our finding of increased apoptosis in HNSCC cells fol- To determine the role of TGF-␣ in HNSCC in vivo,we lowing antisense targeting of TGF-␣ in vivo, accompanied

treated HNSCC-bearing mice with intratumoral adminis- by decreased Bcl-xL expression, suggests that autocrine tration of a TGF-␣ antisense gene therapy construct in production of TGF-␣ by HNSCC cells may abrogate combination with DC-chol liposomes. In these studies, apoptosis via selective modulation of apoptotic we elected to use an antisense expression construct regulatory proteins. designed to achieve high levels of antisense RNA TGF-␣ stimulates and exerts its biologic effects exclus- expression intracellularly. A corresponding sense con- ively through the EGFR. We have previously demon- struct was generated and utilized as a DNA control, strated that the autocrine production of TGF-␣ by although a mismatch or scramble sequence may have HNSCC cells which express EGF receptors results in a provided stronger evidence of a specific antisense mech- growth pathway that has prognostic implications. In a anism of action. Although the ideal cancer gene therapy survival analysis, levels of TGF-␣ in the tumor were would be administered systemically with preferential highly correlated with EGFR expression levels.33 Tar-

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1911 geting either ligand or receptor alone, in vitro, inhibits the bility. Intratumoral injection of plasmid DNA (50 ␮g) growth of HNSCC cells.34,56 Previous reports combining complexed with DC-chol liposomes (50 nmol) in a vol- TGF-␣ antisense oligonucleotides with an EGFR-specific ume of 50 ␮l (three times a week for 3 weeks or a total of tyrosine inhibitor or a protein kinase A inhibitor nine injections) was performed. Tumors were measured in vitro, suggested that increased growth inhibition could using calipers before each injection and tumor volumes be achieved by targeting the ligand simultaneously with were calculated (tumor volume = length × width2/2; frac- a downstream signaling .41,57 TGF-␣ antisense oli- tional tumor volume calculated as a proportion of the gonucleotides have also been used in conjunction with pretreatment tumor volume). Mice were killed when the one of several chemotherapy agents where combined tumors ulcerated or reached a maximum diameter of 2 treatment demonstrated increased growth inhibition of cm. Animal care was in strict compliance with colon cancer cells compared with TGF-␣ antisense oligo- institutional guidelines established by the University of nucleotide alone.58 Treatments designed to abrogate TGF- Pittsburgh, the Guide for the Care and Use of Laboratory ␣ production represent a potential means of inhibiting Animals, and the Association for Assessment and tumor growth in cancers which overexpress this Accreditation of Laboratory Animal Care International. growth factor. Polymerase chain reaction (PCR) To detect the U6/TGF-␣ chimeric DNA in the tumor fol- Materials and methods lowing intratumoral injection of the plasmid–liposome complex, nucleic acids were extracted from harvested Plasmid construct and cloning xenografts using Ultraspec (Biotecx Laboratories, Hous- We modified the original U6 expression plasmid ton, TX, USA). One microgram of the total RNA com- (pGEMmU6; a kind gift from S Noonberg, University of bined with plasmid DNA was digested with 10 ␮gof 50 California San Francisco Research Institute ), as RNase A (GIBCO BRL). The RNase was removed by phe- 35 described previously. Forty long sense and nol: chloroform: isoamyl alcohol (25:24:1) extraction. PCR antisense oligonucleotides corresponding to the ATG was performed using primers designed to detect U6 ␣ − + 59 start site of the human TGF- gene ( 20 to 20) were DNA and the U6/TGF-␣ chimeric DNA under the con- synthesized and cloned into the Xhol and Nsil sites of the ditions recommended by the manufacturer. For the TGF- ⌬ modified plasmid, p HU6. The sequences of the sense ␣ antisense gene, the forward primer was AAA CGC and antisense expression constructs were verified by ACC ACG TGA CGG and the reverse primer was TGC sequence analysis (Figure 1). ATT GCT GCC CGC CCG. For the TGF- ␣ sense gene, the forward primer was AAA CGC ACC ACG TGA CGG Cells and the reverse primer was CTG TCC AGC CGA GGG The cell line, 1483, is a well-described HNSCC cell line GAC. The PCR products were run on a 1% agarose gel. derived from a tumor of the retromolar trigone region of the oropharynx60 and was a generous gift of Dr R Lotan Intracellular localization of transfected DNA (The University of Texas MD Anderson Cancer Center, To determine the intracellular localization of the TGF-␣ Houston, TX, USA). We previously reported that 1483 antisense construct after delivery into HNSCC cells, we cells, like all HNSCC cell lines, produce TGF-␣ and are examined fluorecein-labeled DNA in transfected tumor significantly growth inhibited in vitro following treatment cells using fluorescence microscopy. The DNA was lab- with TGF-␣ antisense oligonucleotides.34 The cells were eled with Labet IT fluorecein nucleic acid labeling maintained in vitro in Dulbecco’s modified Eagle’s (Molecular Probes, Eugene, OR, USA). Five micrograms medium (DMEM) (Fisher Scientific, Pittsburgh, PA, USA) of DNA were mixed with 5 ␮lof10× labeling buffer A supplemented with 10% fetal calf serum (FCS) and anti- and 5 ␮l of reconstituted labeling reagent in a reaction biotics (Life Technologies (GIBCO BRL), Gaithersburg, performed at 37°C for 1 h followed by a purification with MD, USA). a G50 spin column. 1483 Cells were plated on 60-mm cell culture dished with a glass cover slip on the bottom at a In vivo tumor xenograft studies density of 2 × 106 cell per well. These cells were trans- We and others have previously reported that the 1483 cell fected 24 h later with a DNA–liposome mixture of 7 ␮g- line grows well as xenografts in nude mice.35,60 Cells in labeled DNA and 24 ␮l LipofectAMINE reagent (GIBCO log phase were harvested by trypsinization, resuspended BRL) for each dish according to the manufacturer’s in DMEM supplemented with 10% FCS, centrifuged at instructions. Cells alone, cells treated with labeled DNA 1000 r.p.m. for 10 min, and resuspended in culture media only, and cells treated with unlabeled DNA plus Lipofec- at a concentration of 1 × 107 cells/ml before subcutaneous tAMINE under the same transfection conditions were implantation into mice. Female athymic nude mice used as controls for substracting the false positive signals. nu/nu (4–6 weeks old; 20 ± 2 g (s.d.); Harlan Sprague Cells transfected with fluorescent DNA were coun- Dawley, Indianapolis, IN, USA) were implanted with 1 terstained with Hoechst nuclear strain (Molecular Probes) × 106 cells into the right flank with a 26-gauge needle/1 and imaged on a Zeiss 135 axivert scope (Zeiss, Thorn- ml tuberculin syringe. Approximately 14–21 days later wood, NY, USA) with an automated xyz stage. For each when the tumor nodules were palpable (approximately transfection time-point (5 h, 12 h, 24 h, 48 h, 72 h) five 2 × 2 mm in diameter), mice were randomly assigned to separate fields were imaged in the xy axis at 30 different the treatment groups. Mice were treated with liposomes z axis positions. Data stacks for each field were used to alone, TGF-␣ antisense DNA plus liposomes or TGF-␣ create 360° three-dimensional reconstructions to verify sense DNA plus liposomes. There were 10 mice in each nuclear localization. Standard fluorescent filter sets treatment group in an individual experiment. Experi- (Chroma Tech, Brattleboro, VT, USA) for FITC and DAPI ments were repeated three times to ensure reproduci- were used in these studies. Tissue samples were fixed in

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1912 2% parafomaldehyde in PBS for 1 h at 4°C, cryoprotected at 100°C. Proteins (50 ␮g per lane) were separated by in 30% sucrose overnight and chock frozen in liquid 12.5% SDS-polyacrylamide gel electrophoresis (SDS- nitrogen cooled isopentane. Sections (5 microns) were cut PAGE) and transferred on to a nitrocellulose membrane using a Microm (Walldorf, Germany) 505E cryostat (MSI, Westboro, MA, USA). Prestained molecular weight mounted on glass slides and counterstained with DAPI markers (Gibco, Gaithersburg, MD, USA) were included (1 mg/100 ml) for 2 min. The slides were then in each gel. Membranes were blocked for 30 min in Tris- coverslipped in glycerol and the sections examined using buffered saline (TBS: 10 mmol/l Tris-HCl, pH 7.5, and an Olympus (Melville, NY, USA) IX70 microscope with 150 mmol/l NaCl) with 0.5% Tween-20 (TBST) and 5% a40× planapochromat oil immersion objective and a bovine serum albumin (BSA). After blocking, membranes double pass (green/blue) cube. Images were captured were incubated for 60 min with Bcl-xL mouse anti-human using an Optronics Firewire camera (Galeta, CA, USA) monoclonal antibody or a Bax rabbit polyclonal antibody and stored. Subsequently, the coverslips were lifted from (Santa Cruz Biotechnology, Santa Cruz, CA, USA) in the slides by immersion in PBS and slides passed through TBST and 1% BSA. After washing the membranes three a standard hematoxylin and eosin series and remounted times with TBST (5 min each), they were incubated with in Permount. The exact areas imaged using fluorescence horseradish peroxidase-conjugated secondary antibody microscopy were then captured using brightfield illumi- in TBST and 1% BSA for 30 min. Subsequently, mem- nation. branes were washed three times with TBST and developed using the enhanced chemiluminescence (ECL) Immunohistochemistry detection system (Amersham Life Sciences, Arlington Tumor specimens (HNSCC xenografts) were fixed Heights, IL, USA). immediately following resection in 10% buffered neutral formalin and stained with hematoxylin and eosin for his- Statistical analysis topathologic analysis. Indirect immunohistochemical For in vivo experiments where tumor volumes of the staining for TGF-␣ (Antibody 2; Oncogene Science, same mice were measured over time, the statistical sig- Uniondale, NY, USA) was performed on paraffin-embed- nificance of differences between groups was examined by ded tissues using a murine monoclonal antibody from a use of repeated measurement analysis of variance (two- commercially available assay as described by us pre- sided). Comparisons were restricted to mice treated in viously.29,33 The labeled streptoavidin–biotin (LSAB) the same experiment. For apoptosis studies, the statistical method was used to visualize antibody positivity (DAKO significance of differences in apoptosis was assessed by LSAB + kits; DAKO Corporation, Carpinteria, CA, USA). use of Student’s t test (two-sided) that ensured unequal The primary antibody was a mouse antihuman IgG and variance. the secondary antibody was a horse antimouse biotinyl- ated IgG (Bio-Rad Laboratories, Hercules, CA, USA). Brown staining was considered positive. Positive and Acknowledgements negative controls were as described previously.33 We are grateful to Dr Reuben Lotan for the gift of the cell line 1483. This work was supported in part by Public Apoptosis determinations/DNA fragmentation Health Service grants CA01760, CA72526, and CA77038 The percentage of apoptotic cells in tumors treated with from The National Cancer Institute (to JRG), The Eye and the TGF-␣ antisense (versus sense) gene therapy was Ear Foundation, The Pittsburgh Foundation, The Mary determined by staining for DNA fragmentation with Hillman Jennings Foundation and the Veterans Research ApopTag (INTERGEN, Purchase, NY, USA). Tumors Foundation of Pittsburgh. were harvested, sectioned, fixed in formalin and paraffin- embedded, then incubated with proteinase K diluted in phosphate-buffered saline (PBS) for 20 min and washed References four times in water. Slides were then incubated in 3% 1 Todaro GJ, De Larco JE. Transformation by murine and feline H2O2 in PBS for 5 min and washed twice in PBS. Each sarcoma viruses specifically blocks binding of epidermal growth section was incubated with a terminal transferase factor to cells. Nature 1976; 264: 26–31. enzyme that catalyzes the addition of digoxigenin-lab- 2 Todaro GJ, Fryling C, De Larco JE. Transforming growth factors eled nucleotides to the 3Ј-OH ends of the fragmented produced by certain human tumor cells: polypeptides that inter- DNA for 15 min at 37°C. Slides were then placed in stop act with receptors. Proc Natl Acad Sci USA 1980; 77: 5258–5262. buffer for 30 min at 37°C, followed by washing three 3 Barrandon Y, Green H. Cell migration is essential for sustained times in PBS for 5 min. Negative controls were obtained growth of deratinocyte colonies: the roles of transforming by substituting dH2O for the terminal deoxynucleotidyl growth factor-alpha and epidermal growth factor. Cell 1987; 50: transferase mix. Slides were read and scored under 400 × 1131–1137. magnifications for the number of positive cells per five 4 Coffey RJ Jr et al. Production and auto-induction of transforming high power fields using computerized image analysis as growth factor-alpha in human keratinocytes. Nature 1987; 328: described previously61 (SAMBA 4000 Image Analysis 350–352. System; Image Products International; Chantilly, VA, 5 Liscia DS et al. Transforming growth factor-alpha messenger USA). RNA localization in the developing adult rat and human mam- mary gland by in situ hybridization. Dev Biol 1990; 140: 123–131. 6 Thomas DM, Nasim MM, Gullick WJ, Alison MR. Immunoreac- Immunoblotting × tivity of transforming growth factor alpha in the normal adult Whole-cell extracts from tumors were mixed with 2 gastrointestinal tract. Gut 1992; 33: 628–631. sodium dodecyl sulfate (SDS) sample buffer (125 mmol/l 7 Kudlow JE et al. Transforming growth factor-alpha in the mam- Tris-HCl, pH 6.8; 4% SDS; 20% glycerol; 10% 2- malian brain. Immunohistochemical detection in and mercaptoethanol) at 1:1 ratio and were heated for 5 min characterization if its mRNA. J Biol Chem 1989; 364: 3880–3883.

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1913 8 Xynos FP et al. Expression of transforming growth factor alpha factor-alpha and epidermal growth factor receptor in patients mRNA in benign and malignant tissues derived from gynecol- with squamous cell carcinoma of the head and necck. Cancer ogic patients with various proliferative conditions. Anticancer 1996; 78: 1284–1292. Res 1992; 12: 1115–1120. 30 Di Marco E et al. Autocrine interaction between TGF alpha and 9 Gomella LG et al. Expression of transforming growth factor the EGF-receptor: quantitative requirements for induction of the alpha in normal human adult kidney and enhanced expression malignant phenotype. Oncogene 1989; 4: 831–838. of transforming growth factors alpha and beta 1 in renal cell 31 Moroni MC, Willingham MC, Beguinot L. EGF-R antisense RNA carcinoma. Cancer Res 1989; 49: 6972–6975. blocks expression of the epidermal growth factor receptor and 10 Rosenthal A et al. Expression in rat fibroblasts of a human trans- suppresses the transforming phenotype of a human carcinoma forming growth factor-alpha cDNA results in transformation. cell line. J Biol Chem 1992; 267: 2714–2722. Cell 1986; 46: 301–309. 32 Atlas I et al. Growth regulation of human renal carcinoma cells: 11 Sandgren EP et al. Overexpression of TGF alpha in transgenic role of transforming growth factor alpha. Cancer Res 1992; 52: mice: induction of epithelial hyperplasia, pancreatic metaplasia, 335–339. and carcinoma of the breast. Cell 1990; 61: 1121–1135. 33 Rubin Grandis J et al. Tumor levels of TGF-␣ and EGFR protein 12 Gewirtz AM. Am Society of Clin Oncol-Educational Book. Bostrom predict survival in patients with head and neck squamous cell Corp: Chicago, 1995 pp 192–196. carcimona. J Natl Cancer Inst 1998; 90: 824–832. 13 Saxena A, Ali IU. Increased expression of genes from growth 34 Rubin Grandis J et al. Downmodulation of TGF-alpha protein factor signaling pathways in glioblastoma cell lines. Oncogene expression with antisense oligonucleotides inhibits proliferation 1992; 7: 243–247. of head and neck squamous carcinoma but not normal mucosal 14 Bates SE, Davidson NE, Valverius EM. Expression of trans- epithelial cells. J Cell Biochem 1998; 69: 55–62. forming growth factor alpha and its messenger ribonucleic acid 35 He Y et al. Inhibition of human squamous cell carcinoma growth in human breast cancer: its regulation by estrogen and its poss- in vivo by epidermal growth factor receptor antisense RNA tran- ible functional significance. Mol Endocrinol 1988; 2: 543–545. scribed from the U6 promoter. J Natl Cancer Inst 1998; 90: 15 Dickson RB et al. Growth factors in breast cancer: mitogenesis 1080–1087. to transformation. J Steroid Biochem Mol Biol 1992; 43: 69–78. 36 Rubin Grandis J, Zeng Q, Tweardy DJ. Retinoic acid normalizes 16 Iihara K et al. Prognostic significance of transforming growth the increased gene transcription rate of TGF-alpha and EGFR in factor-alpha in human esophageal carcinoma. Implication for head and neck cancer cell lines. Nature Med 1996; 2: 237–240. the autocrine proliferation. Cancer 1993; 71: 2902–2909. 37 Rubin Grandis J, Tweardy DJ, Melhem MF. Asynchronous 17 Smith JM, Derynck R, Korc M. Production of transforming modulation of transforming growth factor-alpha and epidermal growth factor alpha in human pancreatic cancer cells: evidence growth factor receptor protein expression in progression of pre- for a superagonist autocrine cycle. Proc Natl Acad Sci USA 1987; malignant lesions to head and neck squamous cell carcinoma. 84: 7567–7570. Clin Cancer Res 1998; 4: 13–20. 18 Barton CM et al. Transforming growth factor alpha and epider- 38 Imanishi KI et al. Inhibition of growth of human lung adenocar- mal growth factor in human pancreatic cancer. J Pathol 1991; cinoma cell lines by anti-transforming growth factor-alpha 163: 111–116. monoclonal antibody. J Natl Cancer Inst 1989; 81: 220–223. 19 Yeh YC et al. Elevation of transforming growth factor alpha and 39 Filmus J, Shi W, Spencer T. Role of transforming growth factor its relationship to the epidermal growth factor and alpha-feto- alpha (TGF-alpha) in the transformation of ras-transfected rat protein levels in patients with hepatocellular carcinoma. Cancer intestinal epithelial cells. Oncogene 1993; 8: 1017–1022. Res 1987; 47: 896–901. 40 Sizeland AW, Burgess AW. Antisense transforming growth fac- 20 Liu C, Woo A, Tsao MS. Expression of transforming growth fac- tor alpha oligonucleotides inhibit autocrine stimulated prolifer- tor-alpha in primary human colon and lung carcinomas. Br J ation of a colon carcinoma cell line. Mol Biol Cell 1992; 3: Cancer 1990; 62: 425–429. 1235–1243. 21 Tanaka S et al. Immunoreactive transforming growth factor 41 Solorzano CC et al. Inhibition of transforming growth factor alpha is commonly present in colorectal neoplasia. Am J Pathol alpha stimulation of human squamous cell carcinoma of the 1991; 139: 123–129. head and neck with anti-TGF-alpha antibodies and tyrphostin. 22 Tateishi M et al. Immunohistochemical evidence of autocrine Ann Surg Oncol 1997; 4: 670–678. growth factors in adenocarcinoma of the human lung. Cancer 42 Laird AD, Brown PI, Fausto N. Inhibition of tumor growth in Res 1990; 50: 7077–7080. liver epithelial cells transfected with a transforming growth fac- 23 Mydlo JH et al. Expression of transforming growth factor alpha tor alpha antisense gene. Cancer Res 1994; 54: 4224–4232. and epidermal growth factor receptor messenger RNA in neo- 43 Rubenstein M, Mirochnik Y, Chou P, Guinan P. Antisense oli- plastic and nonneoplastic human kidney tissue. Cancer Res 1989; gonucleotide intralesional therapy for human PC-3 prostate 49: 3407–3411. tumors carried in athymic nude mice. J Surg Oncol 1996; 62: 24 Mellon JK, Cook S, Chambers P, Neal DE. Transforming growth 194–200. factor alpha and epidermal growth factor levels in bladder can- 44 Gewirtz AM et al. Bone marrow purging with oligodeoxynucle- cer and their relationship to epidermal growth factor receptor. otides. Blood 1996; 88: 1517. Br J Cancer 1996; 73: 654–658. 45 Webb A et al. Bcl-2 antisense therapy in patients with non-Hodg- 25 Reiss M, Stash EB, Vellucci VF, Zhou Z. Activation of the kin lymphoma. Lancet 1997; 349: 1137–1141. autocrine transforming growth factor alpha pathway in human 46 Cammilleri S et al. Biodistribution of iodine-125 tyramine trans- squamous carcinoma cells. Cancer Res 1991; 51: 6254–6262. forming growth factor alpha antisense oligonucleotide in 26 Morishige K et al. Involvement of transforming growth factor athymic mice with a human mammary tumor xenograft follow- alpha/epidermal growth factor receptor autocrine growth ing intratumoral injection. Eur J Nucl Med 1996; 23: 448–452. mechanism in an ovarian cancer cell line in vitro. Cancer Res 47 He Y, Huang L. Growth inhibition of human papillomarvirus 1991; 51: 5951–5955. 16 DNA-positive mouse tumor by antisense RNA transcribed 27 Todd R et al. TGF-alpha and EGF-receptor mRNAs in human from U6 promoter. Cancer Res 1997; 57: 3993–3999. oral cancers. Carcinogenesis 1989; 10: 1553–1556. 48 Nabel GJ et al. Direct gene transfer with DNA–liposome com- 28 Rubin Grandis J, Tweardy DJ. Elevated levels of transforming plexes in melanoma: expression, biologic activity, and lack of growth factor alpha and epidermal growth factor receptor mess- toxicity in humans. Proc Natl Acad Sci USA 1993; 90: 11307– enger RNA are early markers of carcinogenesis in head and neck 11311. cancer. Cancer Res 1993; 53: 3579–3584. 49 Caplen NJ et al. Liposome-mediated CFTR gene transfer to the 29 Rubin Grandis J, Melhem MF, Barnes EL, Tweardy DJ. Quanti- nasal epithelium of patients with cystic fibrosis. Nature Med tative immunohistochemical analysis of transforming growth 1995; 1: 39–46.

Gene Therapy TGF-␣ antisense gene therapy for head and neck cancer S Endo et al 1914 50 Noonberg SB et al. In vivo generation of highly abundant Inhibition of epidermal growth factor receptor gene expression sequence-specific oligonucleotides for antisense and triplex gene and function decreases proliferation of head and neck squamous regulation. Nucleic Acids Res 1994; 22: 2830–2836. carcinoma but not normal mucosal epithelial cells. Oncogene 51 Mahr S et al. Growth factor effects on apoptosis of rat gastric 1997; 15: 409–416. enterochromaffin-like cells. Endocrin 1998; 139: 4380–4390. 57 Ciardiello F et al. Cooperative inhibition of renal cancer growth 52 Brison DR, Schultz RM. Increased incidence of apoptosis in by anti-epidermal growth factor receptor antibody and protein transforming growth factor alpha-deficient mouse blastocytes. kinase A antisense oligonucleotide. J Natl Cancer Inst 1998; 90: Biol Reprod 1998; 59: 136–144. 1087–1094. 53 Amundadottir LT et al. Cooperation of TGF alpha and c-myc 58 De Luca A et al. Antisense oligonucleotides directed against in mouse mammary tumorigenesis: coordinated stimulation of EGF-related growth factors enhance anti-proliferative effect of growth and suppression of apoptosis. Oncogene 1996; 13: 757– conventional anti-tumor drugs in human colon-cancer cells. Int 765. J Cancer 1997; 73: 277–282. 54 Seki S et al. Induction of apoptosis in a human hepatocellular 59 Qian JF, Lazar-Wesley E, Breugnot C, May E. Human trans- carcinoma cell line by a neutralizing antibody to transforming forming growth factor alpha: sequence analysis of the 4.5-kb and growth factor-alpha. Virchows Arch 1997; 430: 29–35. 1.6-kb mRNA species. Gene 1993; 132: 291–296. 55 Nakamura N, Shikoji Y, Moriwaki H, Muto Y. Apoptosis in 60 Sacks PG et al. Establishment and characterization of two new human hempatoma cell line induced by 4,5-didehydro geranyl- squamous cell carcinoma cell lines derived from tumors of the geranoic acid (acyclic acid) via down-regulation of transforming head and neck. Cancer Res 1988; 48: 2858–2866. growth factor-alpha. Biochem Biophys Res Commun 1996; 219: 61 Drenning SD et al. Bcl-2 but not Bax expression is associated 100–104. with apoptosis in normal and transformed squamous epi- 56 Rubin Grandis J, Chakraborty A, Melhem MF, Tweardy DJ. thelium. Clin Cancer Res 1998; 4: 2913–2921.

Gene Therapy