Cancer Therapy (2006) 13, 131–140 r 2006 Nature Publishing Group All rights reserved 0929-1903/06 $30.00 www.nature.com/cgt

ORIGINAL ARTICLE Small interfering RNA expression vector targeting hypoxia-inducible factor 1 alpha inhibits tumor growth in hepatobiliary and pancreatic cancers T Mizuno1, M Nagao1, Y Yamada1, M Narikiyo1, M Ueno1, M Miyagishi2, K Taira2,3 and Y Nakajima1 1Department of Surgery, Nara Medical University, Kashihara-city, Nara, Japan; 2Department of Chemistry and Biotechnology, School of Engineering, University of Tokyo, Bunkyo-ku, Tokyo, Japan and 3Gene Discovery Research Center, National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki, Japan

Hepatobiliary and pancreatic carcinomas are hypovascular tumors that can proliferate under hypoxic conditions. Recent reports have demonstrated that hypoxia-inducible factor 1 alpha (HIF1a) plays an important role in the survival of these cancers. Given these findings, the inhibition of the HIF1a pathway might prove to be a powerful tool in the treatment of these cancers. To inhibit HIF1a expression, we used small interference RNA (siRNA) expression vectors in this study. The transient transfection of siRNA expression vectors significantly reduced both HIF1a mRNA levels (13% of control) and levels (41% of control) and significantly inhibited the growth of cancer cell lines (Po0.05). VEGF, Glut1, and aldorase A expressions were also significantly reduced by transfection with these vectors (Po0.05), and we found that these vectors induced apoptosis but not cell cycle arrest. In a subcutaneous tumor model using nude mice, transfected MIA PaCa-2 cells, stably expressing siRNAs, barely formed tumors compared to control (Po0.05). This study thus demonstrates the usefulness of siRNA expression vector in targeting HIF1a and points to a potential clinical role in the treatment of pancreatic and hepatobiliary carcinomas. Cancer Gene Therapy (2006) 13, 131–140. doi:10.1038/sj.cgt.7700871; published online 12 August 2005 Keywords: siRNAs; HIF1alpha; pancreatic cancer; hepatobiliary cancer

Introduction modifications from prolyl hydroxylase and interaction with the Von Hippel–Lindau (VHL) protein.2 When the Solid tumors are highly dependent upon for oxygen content of the surroundings decreases, HIF1a their survival. However, in tumors such as pancreatic and degradation is suppressed by the inactivation of prolyl hepatobiliary carcinomas, the newly formed vasculature is hydroxylase, and HIF1a accelerates. Addi- often immature and unable to supply sufficient oxygen tionally, HIF1a can activate a large number of other and nutrients to rapidly dividing tumor cells. Nonetheless, , such as VEGF, glucose transporters, glycolytic these cancers are able not only to survive but also pathway , , and inducible nitric 3 proliferate, growing under more hypoxic conditions than oxide synthase. Recently, it has been reported that the corresponding normal tissues. increased expression of HIF1a protein correlates with Under these hypoxic conditions, one of the major tumor progression and prognosis in several types of 4–7 factors influencing tumor growth is the hypoxia-inducible tumors. It would appear, then, that inhibiting the factor 1 alpha (HIF1a). HIF1a is a basic helix–loop–helix/ HIF1a protein can block angiogenesis and glycolytic PAS that forms a dimer with the Ah in solid tumors, thus suppressing tumor nuclear translocator (ARNT) and binds to growth and metastasis. elements responsible for hypoxia.1 Under normal oxygen In order to inhibit HIF1a expression, various techno- 8–10 conditions, HIF1a is ubiquitinated and degraded by logies have been considered. One such technology, RNA interference (RNAi), is the process of sequence- specific post-transcriptional gene silencing that has been Correspondence: Dr Y Yamada, Department of Surgery, Nara Drosophila Medical University, 840 Shijo-cho, Kashihara, Nara 634-8522, performed in various organisms, including and Japan. Caenorhabditis elegans. This technique, however, has E-mail: [email protected] been limited in mammalian cells because of nonspecific Received 25 October 2004; revised 12 April 2005; accepted 27 April gene silencing that results from a physiological reaction 11 2005; published online 12 August 2005 involving an interferon pathway. However, it has Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 132 recently been reported that small interfering RNAs another 100 ml of Opti-MEM medium and incubated for (siRNAs), 21–23 nucleotide RNAs with two to three 5 min at room temperature. Then, both mixtures, one overhangs, exhibit gene-specific RNAi without activating containing plasmid in Opti-MEM and the other contain- the nonspecific interferon pathway in mammalian cells.12 ing Lipofectamine 2000 in Opti-MEM, were combined These siRNAs have been demonstrated to have stronger and incubated for 20 min at room temperature. This suppressive effects than antisense oligonucleotides.13 complex was added to cell cultures in 12-well dishes. After Furthermore, recent studies have demonstrated that incubation for 18 h, the medium was removed and siRNA expression vectors that transcribe siRNAs by the DMEM with 10% FBS was added. Incubation was U6 or H1 promoters more effectively suppress the continued in 20% O2 (normoxia) or 1% O2 (hypoxia) expression of endogenous genes and exhibit more conditions. persistent inhibition compared to synthetic siRNAs.14–21 Given these findings, siRNAs may have applicability in mammalian cells, as well as a potential role in gene Northern blot analysis targeting therapy. In this study, we focus on the siRNA Total RNA was extracted from cultured cells using an expression vector as a tool to inhibit HIF1a expression ISOGEN kit (Nippon Gene Ltd, Toyama, Japan). and investigate its usefulness in in vitro and murine Aliquots (20 mg) were electrophoresed on 1.5% agarose- models of hepatobiliary and pancreatic carcinomas. formaldehyde gels, transferred to nylon membranes (Hybond N plus, Amersham, Buckinghamshire, UK), and then hybridized with 1 Â 106 c.p.m./ml of the 32P- radiolabeled HIF1a cDNA probe. The probe was Materials and methods prepared using the DNA labeling kit (d-CTP) (Pharmacia siRNA expression plasmids Biotech Inc., Uppsala, Sweden) in Rapid-hyb buffer Tandem-type and hairpin-type siRNA expression vectors (Amersham, Buckinghamshire, UK) and 100 mg/ml salmon using the pU6icassette vector,14 which contains a human testis DNA at 651C for 2 h. Membranes were washed U6 promoter and two BspMI sites, were prepared. For twice in 0.2 Â saline-sodium citrate buffer (SSC) and the construction of tandem-type siRNA expression 0.1% sodium dodecyl sulfate (SDS) at 651C for 30 min. plasmids, we used polymerase chain reaction (PCR) to Then, HIF1a expression levels were analyzed and amplify DNA fragments that included the HIF1a target quantified by densitometric methods with a BAS 1000 sequences of the sense and antisense regions and the U6 image analyzer (Fuji Photo Film Co. Ltd, Minato-ku, promoter using the pU6icassette vector as a template. Tokyo, Japan). After digestion of its products, each fragment was ligated Next, the Northern blots were reprobed with a into the BspMI sites of the pU6icassette vector to yield a glyceraldehyde 3-phosphate dehydrogenase (GAPDH) series of siRNA expression vectors. For the construction cDNA probe to confirm that similar amounts of RNA of hairpin-type siRNA expression plasmids, we synthe- were loaded and transferred from each sample. Within sized oligonucleotides with the hairpin sequence, the each sample, the density of the HIF1a mRNA band terminator sequence, and overhanging sequences. Then, was determined by comparing the densities of GAPDH we annealed the fragments and inserted them into the bands. BspMI sites of the pU6icassette vector. The target sequence selections were based on the characterization 22 of siRNA by Elbashir et al. PCEP4siHIFno5 vectors Western blot analysis were generated using PCR-based subcloning of the Cells (2 Â 107 cells) were collected by centrifugation and pU6iHIFno5 vector into the pCEP4 vector (Invitrogen resuspended in lysis buffer. Lysis buffer consisted of Corp., Carlsbad, CA). 10 mM Tris-HCl, 1 mM EDTA, 10 mM KCl, 10% glycerol, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, Cell culture conditions and transfection of plasmids 100 mM iodoacetamide, and 0.02% sodium azide. Samples Three pancreatic carcinoma cell lines (MIA PaCa-2, were homogenized and placed on ice for 15 min. All PK-8, and PANC-1), two cholangiocarcinoma cell lines lysates were centrifuged for 15 min at 41C, and then the (TFK-1 and HuCCT1 (Cell Resource Center for Bio- supernatant was collected and stored at 41C as whole cell medical Research, Tohoku University)), and one hepato- protein lysates. The protein concentration was measured ma cell line (HepG2) were used. The cell lines were spectrophotometrically using protein standards. An maintained in Dulbecco’s modified Eagle’s Medium equivalent amount of Laemmli buffer was added to (DMEM)/F12 medium with 10% heat-inactivated FBS 20 mg of protein and then boiled for 3 min prior to at 371C in a humidified atmosphere of 5% CO2. loading. SDS-polyacrylamide gel electrophoresis was When cells reached 80–90% confluence, they were performed using a 7.5% resolving gel. were transfected with plasmids in the following manner using electrophoretically transferred to a nitrocellulose mem- Lipofectamine 2000 reagent (Invitrogen Corp., Carlsbad, brane in Tris-glycine buffer. Immunodetection was CA) according to the manufacturer’s instructions. A carried out using Immuno Star reagents (Wako, Osaka, 1.5 mg portion of plasmid was mixed in 100 ml of Opti- Japan). The first antibody was a mouse monoclonal MEM medium (Invitrogen Corp., Carlsbad, CA), and anti-HIF1a antibody (H1a67, Novous Biologicals Inc., 4.0 ml of the Lipofectamine 2000 reagent was diluted in Littleton) at 1/750 dilution.

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 133 Semiquantitative reverse transcription-polymerase In vivo tumor treatment chain reaction analysis We transfected pCEP4siHIFno5 into MIA PaCa-2 cell Total RNA was extracted as above. After DNaseI lines using Lipofectamine 2000, and these cells were treatment, first-strand cDNA was prepared using Omnis- cultured in DMEM with 10% FBS and hygromycin. They cript Reverse Transcriptase (QIAGEN, Hilden, Ger- were harvested from continuous culture and resuspended many) and oligo dT primers. Adequate PCR cycle in DMEM without FBS for inoculation into nude mice. numbers were chosen in the middle of an exponential Male BALB/C-nu/nu mice (4-week-old) (Oriental Bio- range of amplification. The following oligonucleotides Service Inc., Kyoto, Japan) were injected subcutaneously were used as primers: HIF1a,50-CCTGCACTCAAT with 1  107 MIA PaCa-2 cells. Tumor volume was CAAGAAGTTGC-30 (forward) and 50-TTCTGCTCTG measured in a blind fashion using calipers. Tumor volume TTTGGTGAGGCT-30 (reverse) resulting in a 619 bp (TV) was estimated using the following formula: fragment; b-actin, 50-ATCAAGATCCTGACCGAGC TV ¼ (A2B)/2, where A is the width (mm) of the tumor G-30 (forward) and 50-TACTTGCGCTCAGGAGGAG and B is its length (mm). C-30 (reverse) resulting in a 400 bp fragment; VEGF, 50-GAAGTGGTGAAGTTCATGGATGTC-30 (forward) Immunohistochemical staining and counting of and 50-CGATCGTTCTGTATCAGTCTTTCC-30 (reverse) microvasculature resulting in a 365 bp fragment; Glut1, 50-ATGAAGGAA CD31 was detected using a monoclonal antibody directed GAGAGTCGGCA-30 (forward) and 50-TGAAGAGTT against CD31 (Becton Dickinson, San Jose, CA; 1:100 CAGCCACGATG-30 (reverse) resulting in a 457 bp dilutions). The resected specimens were snap-frozen in fragment; aldorase A, 50-CACTGGGATCACCTTCC liquid nitrogen and stored at À801C until analysis. TGT-30 (forward) and 50-AAGACACCACACACCACT Cryosections were air-dried and fixed for 10 min in cold GT-30 (reverse) resulting in a 412 bp fragment. The acetone and then washed in PBS. The endogenous density of each band with background subtraction was peroxidase activity was blocked with methanol containing measured using an NIH image. The levels of each kind 0.3% hydrogen peroxidase for 20 min. After another of mRNA were normalized according to the density of washing, the sections were preincubated in blocking b-actin bands. serum and then incubated for 60 min at room temperature with the primary monoclonal antibody. Immunostaining was performed using the Histofine Simple Stein kit Colony formation assay (Nichirei Co., Tokyo, Japan), a labeled polymer method. Plasmid-transfected cells (1  105 cells/100 mm dish) were The staining was visualized with diaminobenzidine-tetra- incubated in medium. Approximately 2 weeks later, chloride. The sections were finally counterstained with cell colonies were rinsed with phosphate-buffered saline hematoxylin. To assess microvasculature, the vessels in (PBS), fixed with 5% glutaraldehyde (Nacalai Tesque, the four most highly vascularized areas were counted at Kyoto, Japan) in PBS, and then stained with 0.2% crystal  200 magnification, and the average count was calcu- violet (Wako, Osaka, Japan) and 100 mM CAPS (Wako, lated. Osaka, Japan). Colonies containing at least 50 cells were counted. Apoptotic labeling index in vivo The TUNEL assay was performed twice, according to the manufacturer’s instructions, on each cryosection in Cell cycle and apoptosis analysis order to detect fragmented DNA in situ (Apoptag Plus Cell cycle distribution was assayed by determining DNA peroxidase kit, Oncor, Gaithersburg, MD). In each content. To determine DNA content, incubated cells were examination, at least 500 tumor cells were counted, and fixed using the dropwise addition of methanol at 41C. the apoptotic index, which was defined as the ratio of After fixation, cells were washed with PBS, and incubated apoptosis cells against counted cells, was calculated as the for 30 min at room temperature with RNase (1 mg/ml) mean of the two counts. and propidium iodide (50 mg/ml). Before performing flow cytometric analysis, samples were filtered through a Statistical analysis 35-mm nylon mesh. Stained cells were analyzed with a Statistical analysis was performed using the Student’s FACScan 83178 cytometer (Becton Dickinson, San Jose, t-test (two-tailed) throughout the present study. A CA). This assay was independently carried out three P-value of o0.05 was considered statistically significant. times. We performed Hoechst 33342 staining for analysis of apoptosis. To detect apoptotic bodies, cells were fixed with 1% glutaraldehyde in PBS at 41C, washed twice Results with PBS, stained with 0.2 mM Hoechst 33342 (Sigma Expression of HIF1a protein in hepatobiliary and Chemical Co., St Louis, MO), and then observed under pancreatic cancer cell lines a fluorescence microscope. Apoptotic bodies were As shown in Figure 1, hypoxia strongly induced HIF1a counted at three different fields of microscopic observa- protein expression in all cell lines. Under normoxic tion. A total of 100 cells were judged under one field. This conditions, the HIF1a protein was undetectable in HepG2 assay was independently carried out three times. cells. However, pancreatic cancer cells and cholangio-

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 134 Pancreatic cancer cell lines untreated control No.1 No.2 No.3 No.4 No.5 MIA PaCa-2 PK8 PANC-1 N H N H N H HIF1α HIF1 α

Actin GAPDH

Hepatoma cell lines Cholangiocarcinoma cell lines HepG2 TFK1 HuCC-T1 Figure 2 Effects of siRNA expression vectors targeting HIF1a on N H N H N H mRNA expression. HIF1a and GAPDH mRNA expression levels were analyzed using Northern blots of the MIA PaCa-2 cell lines HIF1 α performed 2 days after transfection with the pU6iHIFno1 through PU6iHIFno5 vectors. PU6iGL3 is a control vector that targets the Actin firefly luciferase GL3.

Figure 1 Immunoblottings of HIF1a protein. Expression of HIF1a protein in three pancreatic cancer cell lines (MIA PaCa-2, PK-8, and untreated control pU6iHIFno3 pU6iHIFno5 PANC-1), one hepatocellular carcinoma cell line (HepG2), and two cholangiocarcinoma cell lines (TFK-1 and HuCCT1) under normoxic HIF1α 120kDa conditions and hypoxic conditions. N: normoxia; H: hypoxia

(incubated under 1% O2 conditions for 18–24 h).

Actin 48kDa Table 1 AF304431 2509 bp mRNA linear Homo sapiens HIF1a subunit mRNA, complete cds Location Sequence Figure 3 Effect of siRNA expression vectors targeting HIF1a on protein expression. HIF1a and actin protein expression levels were pU6iHIFno1 (tandem type) AAGGCCTCTGTGATGAGGCTT investigated using immunoblotting on MIA PaCa-2 cell lines 48 h 165–186 after transfection with the pU6iHIFno3 and PU6iHIFno5 vectors

pU6iHIFno2 (stem loop AAGGACAAGTCACCACAGGACAG under 1% O2 hypoxic conditions. PU6iGL3 is used as a control. type) 890–912 pU6iHIFno3 (stem loop AAGACACAGAAGCAAAGAACCCA type) 1625–1647 pU6iHIFno5 suppressed HIF1a mRNA and protein pU6iHIFno4 (stem loop AAGAACTATGAACATAAAGTCTG type) 536–559 expression such that their expression was 13 and 41%, pU6iHIFno5 (stem loop AAGAAACTACTAGTGCCACATCA respectively, of that of control cultures. type) 1944–1967 Suppression of HIF1a, VEGF, Glut1, and aldorase A mRNA expression after transfection of HIF1a siRNA expression vectors carcinoma cells expressed HIF1a protein under normoxia, We investigated the expression of genes activated by especially the MIA PaCa-2 cells. Given this finding, MIA HIF1a, such as VEGF, Glut1, and aldorase A, using PaCa-2 was primarily used in our experiments. reverse transcription-polymerase chain reaction (RT- PCR). As shown in Figure 4, hypoxia for 5 days induced Effects of HIF1a siRNA expression vectors on the MIA expression of Glut1 and aldorase A. After transfection PaCa-2 cell line in culture with HIF1a siRNA expression vectors, the levels of siRNA expression vectors were designed to target five HIF1a mRNA and the expression of its related genes different HIF1a mRNA sequences (AF304431) (Table 1). were suppressed over 5 days under normoxic as well as Transfection efficiency in MIA PaCa-2 cells was about hypoxic conditions. After transfection with HIF1a siRNA 78%, calculated using the pEGFP-N1 vector (BD expression vectors, levels of expression of HIF1a and its Bioscience, Clontech, CA). The PU6iGL3 vector target- inducible genes were reduced under hypoxia to levels ing the firefly luciferase GL3 was used as a control. After equivalent to those expressed under normoxia. On the 48 h of culture, two of the five siRNA expression vectors, other hand, the suppressive effect of these vectors pU6iHIFno3 and pU6iHIFno5, both stem loop type disappeared in measurements of all related genes except vectors, effectively suppressed the expression of HIF1a for HIF1a after 2 days (data not shown). mRNA and protein in MIA PaCa-2 cells (Figures 2 and 3). PU6iHIFno4 also suppressed expression of both Cell cycle analysis and apoptosis HIF1a and GAPDH mRNA at any time, although we Analyses of cell cycle distribution 5 days after transfection performed this assay independently three times. Consid- are shown in Figure 5a. Hypoxia increased the proportion ered somewhat influence was reacted, pU6iHIFno4 was of the G1 peak in DNA content compared to norm- omitted. Densitometric analysis of the blots revealed that oxic conditions. Transfection with pU6iHIFno3 and

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 135 a 150 b 150

100 100

* * * 50 50 * The ratio of expression The ratio of expression (% of Untreated:normoxia.) (% of Untreated:normoxia.)

* * * *

0 0 Control Control Control Control Untreated Untreated Untreated Untreated pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 Normoxia Hypoxia Normoxia Hypoxia

c 500 d 200

400 150

300

100 * * 200 * * * The ratio of expression * The ratio of expression (% of Untreated:normoxia.) (% of Untreated:normoxia.) 50 100

0 0 Control Control Control Control Untreated Untreated Untreated Untreated pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 Normoxia Hypoxia Normoxia Hypoxia Figure 4 HIF1a, VEGF, Glut1, and aldorase A mRNA levels in MIA PaCa-2 cells 5 days after transfection with different vectors. mRNA levels were quantified using semiquantitative RT-PCR and comparing each band’s density using an NIH image. Data were normalized using b-actin as an internal control. Data are presented relative to untreated cells. A: HIF1a; B: VEGF; C: Glut1; D: aldorase A; bars: standard deviation (s.d.);

N: normoxia; H: incubated under 1% O2 hypoxic conditions; *Po0.01 versus control.

pU6iHIFno5 decreased the proportion of the G1 peak Colony formation assay under hypoxia. Furthermore, transfection with these The colony formation assay was used to investigate the vectors significantly increased the area of the less than inhibitory effects on growth of HIF1a siRNA expression G1 peak and the number of Hoechst 33342-stained cells vectors in a hypoxic environment. The effects on growth compared to control (Figure 5b). of these vectors in three pancreatic carcinoma cell lines,

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 136 a Untreated (normoxia) Untreated (hypoxia) Control (hypoxia) pU6iHIFno3 (hypoxia) pU6iHIFno5 (hypoxia) 200 200 200 200 200 Cell number Cell number Cell number Cell number Cell number

0 0 0 0 0 DNA contents DNA contents DNA contents DNA contents DNA contents

Less than G1 peak (%) G1 peak (%) Untreated (normoxia) 6.98 ±1.7 49.41 ± 1.6 Untreated (hypoxia) 9.94 ± 0.8 65.54 ± 2.3 Control (hypoxia) 13.21 ± 2.1 63.33 ± 2.4 pU6iHIFno3 (hypoxia) 26.15 ± 4.6* 51.36 ± 3.6 pU6iHIFno5 (hypoxia) 29.21 ± 8.9* 41.79 ± 2.4

b Untreated (hypoxia)

Apoptosis / 100 cells

Untreated (normoxia) 3.33 ± 0.6 pU6iHIFno5 (hypoxia) Untreated (hypoxia) 5.33 ± 1.5 Control (hypoxia) 4.67 ± 1.5 pU6iHIFno3 (hypoxia) 20.33 ± 3.5* pU6iHIFno5 (hypoxia) 22.33 ± 3.1*

Figure 5 (a) DNA content of MIA PaCa-2 cells 5 days after transfection with different vectors under normoxic or 1% O2 hypoxic conditions. After incubation, cells were collected and their DNA content was determined using propidium iodide staining and FACS analysis. The cell cycle distribution was analyzed with the DNA Modifit program. The results are expressed as the mean of three independent assays. *Po0.01 versus control (hypoxia). (b) Photographs of apoptotic bodies using Hoechst 33342 staining in MIA PaCa-2 cells 5 days after transfection. Apoptotic bodies were counted at fields of microscopic observation. The results are expressed as the mean of three independent assays. *Po0.03 versus control (hypoxia).

two cholangiocarcinoma cell lines, and a hepatoma cell lines from forming colonies (Figure 6). In particular, these line were investigated. Because transfection efficiency vectors inhibited the growth of MIA PaCa-2, PK-8, differed, siRNA expression vectors were cotransfected TFK-1, and HuCCT1 under both normoxic and hypoxic with the pEGFP-N1 vector that contains the neomycin conditions. Conversely, these vectors barely inhibited the resistance gene. Cells were cultured in medium with G418 growth of HepG2 under either condition. sulfate (Invitrogen Co., Carlsbad, CA), which is inacti- vated by neomycin resistance gene, to eliminate any Effect of siRNA expression vector on tumor growth nontransfected cells. After incubation for 2 weeks, almost in vivo all cell lines grew more efficiently under hypoxic condi- The inhibitory effects on growth were investigated in vivo. tions compared to normoxic ones. Interestingly, the In order to establish stable transfectants, the pGEP4si- HIF1a siRNA expression vectors prevented several cell HIFno5 vector was generated by PCR-based subcloning

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 137 a 1000 b 1250 c 1000

1000 750 750

750

500 500 500 Counts /dish Counts /dish Counts /dish 250 * 250 250 * * * * 0 0 0 Control Control Control Control Control Control untreated untreated untreated untreated untreated untreated pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 Normoxia Hypoxia Normoxia Hypoxia Normoxia Hypoxia

d 600 e 250 f 400

500 200 300 400 150 300 200 100 Counts /dish Counts /dish

200 Counts /dish 100 * 100 50 * * 0 0 * 0 Control Control Control Control Control Control untreated untreated untreated untreated untreated untreated pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 pU6iHIFno3 pU6iHIFno5 Normoxia Hypoxia Normoxia Hypoxia Normoxia Hypoxia Figure 6 Effect of siRNA expression vectors that target HIF1a on the growth of several cell lines in culture. pU6iHIFno3, PU6iHIFno5, and pU6iGL3 (control) vectors were used to transfect cells plated on 100 mm dishes at a density of 1 Â 105 cells/dish under normoxic or hypoxic conditions. Colonies were allowed to grow for 2 weeks and then counted. The results are expressed as the mean (n ¼ 5). A: MIA PaCa-2; B: PK-8; C: Panc-1; D: HepG2; E: TFK-1; F: HuCCT1; bars: s.d.; hypoxia: incubated under 1% O2 hypoxic conditions; *Po0.01 versus control (normoxia); : Po0.01 versus control (hypoxia). of the pU6iHIFno5 vector into the pCEP4 vector, using ing, a single vessel was defined as a row of brown immu- the Epstein–Barr virus nuclear antigen 1 (EBNA-1)/Ori-P nostained endothelial cells that was separate from system. As shown in Figure 7a, the expression of the adjacent vessels, tumor cells, or other connective tissue HIF1a protein decreased in pCEP4siHIFno5-transfected elements. The number of microvessels in pCEP4siHIFno5- cells compared with control vector (pCEP4)-transfected transfected cells was not different compared to the cells and compared with cells that were not transfected at numbers in controls (Figure 7d). all. Transfected cells and cells that had no treatment were inoculated subcutaneously into male nude mice. Although the inoculations of control vector-transfected cells devel- oped into massive tumors, those mice that received the Discussion pCEP4siHIFno5-transfected cells barely formed palpable tumors (Figure 7b). In this study, we demonstrate how siRNA technology can be used to decrease HIF1a expression, thus inhibiting Apoptosis index and microvasculature staining in vivo solid tumor growth. Although chemically synthesized As shown in Figure 7c, the pCEP4siHIFno5-transfected siRNAs are powerful and convenient tools of gene cells showed a significantly higher apoptotic index than silencing, they can also be problematic. One issue, cells that were not transfected or those transfected with for example, is the transient inhibitory effect against the pCEP4 vector (P ¼ 0.003). In microvasculature stain- . While vector-produced siRNAs are as

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 138 a normal pCEP4 pCEP4siHIFno5 effective as chemically synthesized siRNAs in inhibiting gene expression, the inhibitory effect of chemically synthesized siRNAs lasts only 1 week. The potential HIF1α clinical application of siRNA technology requires an inhibitory effect of long duration. For these reasons, we chose not to use chemically synthesized siRNAs and Actin instead created siRNA expression vectors to suppress HIF1a expression. We synthesized two different types of siRNA expres- sion vectors. One is a tandem-type vector that uses the U6 800 b promoter to direct synthesis of sense and antisense pCEP4siHIFno5 siRNAs. The other is a stem loop type vector that uses untreated the U6 promoter to direct the synthesis of short hairpin 600 pCEP4 RNA molecules (shRNAs). ShRNAs have a double- strand stem of 23 bp connected by a loop of nine bases 23 )

3 and can be efficiently re-formed into siRNAs. Although we did not compare two types of siRNA expression 400 vectors using the same sequences, the stem loop type vector is likely to be more effective than the tandem type

Volume(mm in gene silencing. Further studies on the different inhibitory effects of vector types should be carried out 200 in the future. The RNAi’s, including siRNA, are extremely useful * tools in gene targeting therapy and have revolutionized * genetic studies. However, attention must be paid to 0 * selection of the most appropriate siRNA sequence, as 7 14212835 RNAi’s occasionally inhibit not only the target gene but Days also other nontarget genes.24 Indeed, the PU6iHIFno4 vector suppressed GAPDH expression as well as HIF1a c 18 P=0.003 mRNA at any time in this study, although we performed 16 Northern blotting three times. Both the pU6iHIFno3 14 vector and the pU6iHIFno5 vector were confirmed not to suppress other genes of this study except for HIF1a. Prior 12 to the application of siRNA technology, these inhibitory 10 effects against nontarget genes must be investigated. 8 Several studies have demonstrated the relationship 6 between HIF1a expression and human cancer progres- sion.4–7,25,26 Some of the solid tumor cell lines used in this 4 study expressed the HIF1a protein under not only

Percent of apoptotic cells (%) 2 hypoxia but also normoxia. Therefore, the transfection 0 of HIF1a siRNA expression vectors exhibited similar normal pCEP4 pCEP4 siHIFno5 d 25 Figure 7 (a) Immunoblottings of the HIF1a protein on transfectants. 20 MIA PaCa-2 cells transfected with the pCEP4 or with the pCEP4siHIFno5 vectors were incubated under 1% O2 hypoxic conditions. The pCEP4siHIFno5 vector is described in Materials 15 and methods. Actin bands are performed as control. (b) Suppression of tumorigenesis by the pCEP4siHIFno5 vector-transfected cells 7 10 in vivo. A total of 1 Â 10 transfectants were inoculated into male nude mice. Tumor diameter was measured at regular intervals with calipers, and tumor volume was calculated. Each data point No. of vessels / fields 5 represents the mean (N ¼ 8) of each group of mice. Bars: 7s.e.; *Po0.05 versus pCEP4. (c) Apoptotic index in vivo. Percent of apoptotic cells in the specimen transfected with the pCEP4siHIFno5 0 normal pCEP4 pCEP4 vector was greater than that in the specimen transfected with siHIFno5 the pCEP4 vector. (d) Microvasculature staining and vessel count in vivo. There were no significant differences in the numbers of microvessels between the specimen transfected with the pCEP4si- HIFno5 vector and that transfected with the pCEP4 vector.

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 139 inhibitory effects to the solid tumor cell lines used in this angiogenesis. Figure 7d suggests that there was no study under normoxic as well as hypoxic conditions significant difference in angiogenesis between tumors (Figures 4 and 6). On the contrary, in the case of cell lines transfected with pCEP4siHIFno5 and those transfected expressing HIF1a under normoxic conditions, the with pCEP4. Recent reports have demonstrated that inhibitory effects of HIF1a siRNA expression vectors inhibition of HIF1a suppressed tumor formation with- against these cells are expected only under hypoxia. out suppressing angiogenesis.36 Baek et al.37 reported Hypoxia alone did not alter levels of HIF1a mRNA in that VEGF exhibited an antiapoptotic effect on tumors this study, reflecting the fact that its expression is not under hypoxic conditions. Given these findings, the induc- regulated by transcription alone. HIF1a protein expres- tion of apoptosis that occurred in our study after trans- sion in solid tumor cells is regulated during the translation fection with siRNA expression vectors may have resulted phase or by ubiquitination and degradation in normoxia. primarily from the inhibition of VEGF expression. Usually under normoxic conditions, HIF1a is rapidly In conclusion, we have demonstrated that siRNA ubiquitinated and degraded by pVHL.27 Some tumors expression vectors can inhibit the expression of HIF1a have been reported to have mutations of two domains in and of those genes induced by HIF1a, thus effectively pVHL, thus stabilizing HIF1a in normoxia.28 From the suppressing the in vitro and in vivo growth of selected result of Figures 2 and 3, pU6iHIFno5 suppressed HIF1a hepatobiliary tumors. mRNA expression to 13% of control but HIF1a protein expression to only 41%. This dissociation of suppressive effect may result from difference of half-value period References between HIF1a mRNA and protein, due to the reduction of degradation in HIF1a protein. In addition, mutations 1 Wang GL, Jiang BH, Rue EA, Semenza GL. Hypoxia- of the oxygen-dependent degradation domain, one that inducible factor 1 is a basic–helix–loop–helix–PAS hetero- confers oxygen-dependent instability, in the HIF1a gene dimer regulated by cellular O2 tension. Proc Natl Acad Sci have been reported in renal cell carcinomas.29 PU6iHIF- USA 1995; 92: 5510–5514. no5 prevented almost all cell lines besides HepG2 from 2 Masson N, Willam C, Maxwell PH, Pugh CW, Ratcliffe PJ. Independent function of two destruction domains in forming colonies in this study, but pU6iHIFno3 did not hypoxia-inducible factor-alpha chains activated by prolyl prevent Panc-1 cells from forming colonies. This result . EMBO J 2001; 20: 5197–5206. may imply the presence of mutation of HIF1a gene near 3 Pugh CW, Ratcliffe PJ. Regulation of angiogenesis by the target sequence of pU6iHIFno3. If different types of hypoxia: role of the HIF system. Nat Med 2003; 9: 677–684. solid tumors are found to have specific mutations in 4 Shibaji T, Nagao M, Ikeda N, Kanehiro H, Hisanaga M, HIF1a, gene therapy may be rendered tumor specific by Ko S et al. Prognostic significance of HIF-1 alpha over- constructing siRNA expression vectors that target these expression in human pancreatic cancer. Anticancer Res 2003; exact mutation sequences. 23: 4721–4727. In our study, gene transfer of HIF1a siRNA expression 5 Kurokawa T, Miyamoto M, Kato K, Cho Y, Kawarada Y, vectors prevented cancer cell lines from forming colonies Hida Y et al. Overexpression of hypoxia-inducible-factor 1alpha(HIF-1alpha) in oesophageal squamous cell carci- in vitro and from developing into solid tumors in vivo. noma correlates with lymph node metastasis and pathologic Recent studies have reported that hypoxia could change stage. Br J Cancer 2003; 89: 1042–1047. cell proliferation in two distinct ways. One way is through 6 Buchler P, Reber HA, Buchler M, Shrinkante S, Buchler apoptosis, and the other is through growth arrest in the MW, Friess H et al. Hypoxia-inducible factor 1 regulates G1 phase.30–32 HIF1a has been shown to have both vascular endothelial growth factor expression in human antiapoptotic and proapoptotic effects according to cell pancreatic cancer. 2003; 26: 56–64. type and experimental conditions.33–35 The analysis of cell 7 Schindl M, Schoppmann SF, Samonigg H, Hausmaninger H, cycle in this study showed that the transfection of Kwasny W, Gnant M et al. Overexpression of hypoxia- pU6siHIFno3 and pU6siHIFno5 increased the area of inducible factor 1alpha is associated with an unfavorable the less than G1 peak and of Hoechst 33342-stained cells, prognosis in lymph node-positive . Clin Cancer Res 2002; 8: 1831–1837. confirming the induction of cell death by these transfec- 8 Sun X, Kanwar JR, Leung E, Lehnert K, Wang D, tions. These results suggest that transfection with these Krissansen GW. Gene transfer of antisense hypoxia induci- vectors did not cause G1 arrest in hypoxia but instead ble factor-1 alpha enhances the therapeutic efficacy of cancer caused apoptosis to occur and that HIF1a had an immunotherapy. Gene Therapy 2001; 8: 638–645. antiapoptotic effect in pancreatic cancer cell lines. In 9 Maemura K, Hsieh CM, Jain MK, Fukumoto S, Layne MD, addition, there were no differences in the number of Ki67 Liu Y et al. Generation of a dominant-negative mutant of staining cells between those tumors transfected with endothelial PAS domain protein 1 by deletion of a potent siRNA expression vectors targeting HIF1a and those C-terminal transactivation domain. J Biol Chem 1999; 274: transfected with the control vector in vivo (data not 31565–31570. shown). Taken together, the inhibition of HIF1a protein 10 Rose F, Grimminger F, Appel J, Heller M, Pies V, Weissmann N et al. Hypoxic pulmonary artery fibroblasts product results predominantly in apoptosis and not in trigger proliferation of vascular smooth muscle cells: role of cell cycle arrest. The induction of apoptosis may result hypoxia-inducible transcription factors. FASEB J 2002; 16: from the suppression of genes activated by HIF1a, such as 1660–1661. VEGF, aldorase A, and Glut1. VEGF is one of the 11 Hunter T, Hunt T, Jackson RJ, Robertson HD. The most important genes induced by HIF1a and promotes characteristics of inhibition of protein synthesis by

Cancer Gene Therapy Cancer gene therapy using HIF1alpha siRNAs T Mizuno et al 140 double-stranded ribonucleic acid in reticulocyte lysates. hypoxia-inducible factors HIF-1alpha and HIF-2alpha in J Biol Chem 1975; 250: 409–417. normal human tissues, cancers, and tumor-associated macro- 12 Elbashir SM, Harborth J, Lendeckel W, Yalcin A, Weber K, phages. Am J Pathol 2000; 157: 411–421. Tuschl T. Duplexes of 21-nucleotide RNAs mediate RNA 26 Zhong H, De Marzo AM, Laughner E, Lim M, Holton DA, interference in cultured mammalian cells. Nature 2001; 411: Zagzag D et al. Overexpression of hypoxia-inducible factor 494–498. 1alpha in common human cancers and their metastases. 13 Miyagishi M, Hayashi M, Taira K. Comparison of the Cancer Res 1999; 59: 5830–5835. suppressive effects of antisense oligonucleotides and siRNAs 27 Stebbins CE, Kaelin Jr WG, Pavletich NP. Structure directed against the same targets in mammalian cells. of the VHL–ElonginC–ElonginB complex: implications Antisense Nucleic Acid Drug Dev 2003; 13: 1–7. for VHL tumor suppressor function. Science 1999; 284: 14 Miyagishi M, Taira K. U6 promoter-driven siRNAs with 455–461. four uridine 30 overhangs efficiently suppress targeted gene 28 Tanimoto K, Makino Y, Pereira T, Poellinger L. Mechanism expression in mammalian cells. Nat Biotechnol 2002; 20: of regulation of the hypoxia-inducible factor-1 alpha by the 497–500. von Hippel–Lindau tumor suppressor protein. EMBO J 15 Brummelkamp TR, Bernards R, Agami R. A system for 2000; 19: 4298–4309. stable expression of short interfering RNAs in mammalian 29 Clifford SC, Astuti D, Hooper L, Maxwell PH, Ratcliffe PJ, cells. Science 2002; 296: 550–553. Maher ER. The pVHL-associated SCF 16 Lee NS, Dohjima T, Bauer G, Li H, Li MJ, Ehsani A et al. complex: molecular genetic analysis of elongin B and C, Expression of small interfering RNAs targeted against Rbx1 and HIF-1alpha in renal cell carcinoma. Oncogene HIV-1 rev transcripts in human cells. Nat Biotechnol 2002; 2001; 20: 5067–5074. 20: 500–505. 30 Graeber TG, Osmanian C, Jacks T, Housman DE, Koch CJ, 17 Paul CP, Good PD, Winer I, Engelke DR. Effective Lowe SW et al. Hypoxia-mediated selection of cells with expression of small interfering RNA in human cells. Nat diminished apoptotic potential in solid tumours. Nature Biotechnol 2002; 20: 505–508. 1996; 379: 88–91. 18 Paddison PJ, Caudy AA, Bernstein E, Hannon GJ, Conklin 31 Amellem O, Sandvik JA, Stokke T, Pettersen EO. The DS. Short hairpin RNAs (shRNAs) induce sequence-specific -associated cell cycle arrest in S-phase silencing in mammalian cells. Genes Dev 2002; 16: 948–958. under moderate hypoxia is disrupted in cells expressing 19 Sui G, Soohoo C, Affarel B, Gay F, Shi Y, Forrester WC. A HPV18 E7 oncoprotein. Br J Cancer 1998; 77: 862–872. DNA vector-based RNAi technology to suppress gene 32 Schmaltz C, Hardenbergh PH, Wells A, Fisher DE. expression in mammalian cells. Proc Natl Acad Sci USA Regulation of proliferation–survival decisions during tumor 2002; 99: 5515–5520. cell hypoxia. Mol Cell Biol 1998; 18: 2845–2854. 20 Yu JY, DeRuiter SL, Turner DL. RNA interference by 33 Leuenroth SJ, Grutkoski PS, Ayala A, Simms HH. expression of short-interfering RNAs and hairpin RNAs Suppression of PMN apoptosis by hypoxia is dependent on in mammalian cells. Proc Natl Acad Sci USA 2002; 99: Mcl-1 and MAPK activity. Surgery 2000; 128: 171–177. 6047–6052. 34 Hruban RH, Lacobuzio-Donahue C, Wilentz RE, Goggins 21 McManus MT, Petersen CP, Haines BB, Chen J, Sharp PA. M, Kern SE. Molecular pathology of pancreatic cancer. Gene silencing using micro-RNA designed hairpins. RNA Cancer J 2001; 7: 251–258. 2002; 8: 842–850. 35 Butz J, Wickstrom E, Edwards J. Characterization of 22 Elbashir SM, Lendeckel W, Tuschl T. RNA interference is mutations and loss of heterozygosity of and K-ras2 in mediated by 21- and 22-nucleotide RNAs. Genes Dev 2001; pancreatic cancer cell lines by immobilized polymerase chain 15: 188–200. reaction. BMC Biotechnol 2003; 3: 11. 23 Bernstein E, Caudy AA, Hammond SM, Hannon GJ. Role 36 Ryan HE, Poloni M, McNulty W, Elson D, Gassmann M, for a bidentate ribonuclease in the initiation step of RNA Arbeit JM et al. Hypoxia-inducible factor-1alpha is a interference. Nature 2001; 409: 363–366. positive factor in solid tumor growth. Cancer Res 2000; 60: 24 Jackson AL, Bartz SR, Schelter J, Kobayashi SV, Burchard 4010–4015. J, Mao M et al. Expression profiling reveals off-target gene 37 Baek JH, Jang JE, Kang CM, Chung HY, Kim ND, Kim regulation by RNAi. Nat Biotechnol 2003; 21: 635–637. KW. Hypoxia-induced VEGF enhances tumor survivability 25 Talks KL, Turley H, Gatter KC, Maxwell PH, Pugh CW, via suppression of serum deprivation-induced apoptosis. Ratcliffe PJ et al. The expression and distribution of the Oncogene 2000; 19: 4621–4631.

Cancer Gene Therapy