Cancer Therapy (2013) 20, 94–100 & 2013 Nature America, Inc. All rights reserved 0929-1903/13 www.nature.com/cgt

ORIGINAL ARTICLE Transforming -b gene silencing using adenovirus expressing TGF-b1 or TGF-b2 shRNA

SOh1,2, E Kim1,2, D Kang1,3, M Kim1, J-H Kim4 and JJ Song1

Tumor cells secrete a variety of to outgrow and evade host immune surveillance. In this context, transforming growth factor-b1 (TGF-b1) is an extremely interesting because it has biphasic effects in cancer cells and normal cells. TGF-b1 acts as a growth inhibitor in normal cells, whereas it promotes tumor growth and progression in tumor cells. Overexpression of TGF-b1 in tumor cells also provides additional oncogenic activities by circumventing the host immune surveillance. Therefore, this study ultimately aimed to test the hypothesis that suppression of TGF-b1 in tumor cells by RNA interference can have antitumorigenic effects. However, we demonstrated here that the interrelation between TGF-b isotypes should be carefully considered for the antitumor effect in addition to the selection of target sequences with highest efficacy. The target sequences were proven to be highly specific and effective for suppressing both TGF-b1 mRNA and protein expression in cells after infection with an adenovirus expressing TGF-b1 short hairpin RNA (shRNA). A single change in the shRNA sequence completely abrogated the suppressive effect on TGF-b1. Surprisingly, the suppression of TGF-b1 induced TGF-b3 upregulation, and the suppression of TGF-b2 induced another unexpected downregulation of both TGF-b1 and TGF-b3. Taken together, this information may prove useful when considering the design for a novel cancer immunogene therapy.

Cancer Gene Therapy (2013) 20, 94–100; doi:10.1038/cgt.2012.90; published online 11 January 2013 Keywords: TGF-b isotype; shRNA; adenovirus; antitumor; immunosuppression

INTRODUCTION homology-dependent degradation of the target mRNA.15,16 In Transforming growth factor (TGF)-b is a homodimeric protein that mammalian cells, small interfering RNAs (siRNAs) are produced has three different isoforms (TGF-b1, TGF-b2 and TGF-b3), which through cleavage of longer double-stranded RNA precursors by 17,18 have both overlapping and distinct functions.1,2 These isoforms the RNaseIII endonuclease dicer. After dicer processing, the are secreted as latent 25 kDa active homodimer complexes that 21–23 nucleotide siRNA is incorporated into a protein complex 19 bind TGF-b receptors to initiate intracellular signaling.3 Because called RNA-induced silencing complex. Argonaute 2, a multi- various oncogenic pathways directly inactivate the TGF-b functional protein contained within RNA-induced silencing receptor-Smad pathway, which inhibits proliferation and induces complex, cleaves and releases the sense strand (passenger apoptosis, tumor cells can use TGF-b for their advantage to initiate strand), thereby activating RNA-induced silencing complex with 16,17 immune evasion, growth factor production and metastasis.4,5 All a single-stranded antisense strand (guide strand) that human tumors overexpress TGF-b and accordingly induce tumor selectively degrades complementary mRNA. cell invasion and metastasis.5 The TGF-b isoforms contribute to a Because of the efficacy and specificity of siRNA, it is emerging 20 number of cellular activities like inhibiting normal epithelial cell as a potential new basis for the treatment of human diseases. proliferation while inducing cancer cell proliferation, and However, one of the main drawbacks in the development of promoting an invasive phenotype that is characterized by siRNA therapy is the instability of siRNA in vivo. To overcome epithelial-to-mesenchymal transition.6 Another major role of this instability, DNA-based RNA interference drugs that utilize a TGF-b that is produced by tumors is to block the immune vector delivery system are being developed. For the expression response.7 The local immunosuppressive environment of the of siRNA in cells, short hairpin RNAs (shRNAs) are annealed tumor induced by TGF-b has proved to be the major obstacle to to the vector system and processed into siRNAs by dicer after 16 immunogene therapy with cytokines.8 Among the three isoforms introduction into the cell. Here, we used adenovirus as a delivery of TGF-b, TGF-b1 is the most abundant, universally expressed,7,9 vector for TGF-b shRNA. Adenoviral vectors are still frequently and when silenced, increases antitumor immunity.10,11 In addition used as a method of gene transfer because of efficient transgene to the effect on immunosuppression, TGF-b1 silencing also delivery, expression in both dividing and non-dividing cells and efficiently inhibits proliferation and migration.2 Recently, TGF-b1 ease of propagation to high titers in spite of shortcomings like has also been reported to suppress apoptosis.12–14 increased immunogenicity, the prevalence of pre-existing anti- RNA interference is a regulatory mechanism of most adenovirus immunity in the human population and the inability to 21,22 eukaryotic cells that uses small double-stranded RNA to direct specific targeting.

1Institute for Cancer Research, College of Medicine, Yonsei University, Seoul, South Korea; 2Brain Korea 21 Project for Medical Science, College of Medicine, Yonsei University, Seoul, South Korea; 3Department of Oncology, Affiliated Hospital of Yanbian University, Yanji, Jilin Province, People’s Republic of China and 4Department of Internal Medicine, College of Medicine, Yonsei University, Seoul, South Korea. Correspondence: Professor J-H Kim, Department of Internal Medicine, College of Medicine, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul, South Korea. or Professor JJ Song, Institute for Cancer Research, College of Medicine, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul, South Korea. E-mail: [email protected] or [email protected] Received 14 August 2012; revised 23 November 2012; accepted 30 November 2012; published online 11 January 2013 Target specificity and interrelation between TGF-b isotypes SOhet al 95 MATERIALS AND METHODS TCTCTTGAGAATCTGATATAGCTCAATCCTTTTA-30 and the bottom strand Cell culture sequence 50-AGCTTAAAAGGATTGAGCTATATCAGATTCTCAAGAGATTGAG AATCTGATATAGCTCAATCCG-30 was selected and annealed for subcloning The human cancer cell lines, DU145 (human prostate adenocarcinoma) into pSP72DE3-U6. and A375 (human skin melanoma), were cultured in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum (HyClone, Logan, UT, USA). B16BL6 and B16F10 (mouse melanoma) cells were also cultured in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum Recombinant adenovirus (HyClone). HeLa (human cervix adenocarcinoma) cells were cultured in First, an adenovirus containing the green fluorescence protein (GFP) or minimal essential medium with 10% fetal bovine serum. Cells were LacZ gene was constructed to examine the infection efficiency. In the case maintained in a 37 1C humidified atmosphere containing 5% CO2. of GFP, two different adenoviruses with the GFP gene in either the E1 or E3 region were produced. For the insertion of the GFP gene at the E1 region, pCA14 was used as a shuttle vector. After digestion of pEGFPN1 E1 Reagents (Clontech, Mountainview, CA, USA) with XhoI/XbaI, the insert was Trizol was purchased from Life Technologies (Carlsbad, CA, USA). All other subcloned into XhoI/XbaI-digested pCA14. For the insertion of the GFP chemicals were purchased from Sigma-Aldrich (St, Louis, MO, USA). gene at the E3 region, pSP72DE3 was used as a shuttle vector. After pSP72DE3 was digested with KpnI, the ends were blunted with a Quick blunting kit (New England Biolabs, Ipswich, MA, USA) and further digested Construction of TGF-b shRNAs with BamHI, generating a linearized vector with one cohesive end. For the To construct human TGF-b1 (hTGF-b1) shRNA, we screened three insertion of GFP, pEGFN1 was predigested with AflII, blunted, and then candidate sequences. The selected target sequence was 50-ACCAGAAATAC digested with BamHI to insert the linearized pSP72DE3. For the expression AGCAACAATTCCTG-30, and the loop sequence was 50-TCTCTC-30. Target of LacZ, the LacZ gene was removed from pcDNA3.1/hygro/LacZ selection was performed using an algorithm developed by Genolution (Invitrogen, Carlsbad, CA, USA) by digestion with BamH/XbaI. Thereafter, Pharmaceuticals Inc. (Seoul, South Korea). For expression of hTGF-b1 it was inserted into linearized pCA14 with BamH/XbaI. Finally, pCA14-LacZ shRNA in adenovirus, the top strand sequence was 50-GATCCGCCAGA and pCA14-GFP were recombined into the dl324 adenoviral vector and AATACAGCAACAATTCCTGTCTCTCCAGGAATTGTTGCTGTATTTCTGGTTTTTTT pSP72-GFP was recombined into the dl324-IX adenoviral vector by A-30 and the bottom strand sequence was 50-AGCTTAAAAAAACCAGAAA homologous recombination in BJ5183 bacterial-competent cells, generat- TACAGCAACAATTCCTGGAGAGACAGGAATTGTTGCTGTATTTCTGGTG-30 was ing three different replication-defective adenoviruses to be used as control selected and annealed for subcloning into pSP72DE3-U6. To construct viruses. mouse TGF-b1 (mTGF-b1) shRNA, we screened 11 candidate sequences. The adenoviral shuttle vector, pCA14 (Microbix, Mississauga, ON, The selected target sequence was 50-CCCTCTACAACCAACACAACCCGGG-30 Canada), containing the IX gene of adenovirus type 5, was linearized by and the loop sequence was 50-TCTCTC-30. For expression of mTGF-b1 XmnI digestion. The adenovirus vector dl324-BstBI containing the Ad5 shRNA in adenovirus, the top strand sequence was 50-GATCGCCTCT genome lacking the E1 region (340–4641 in nucleotide of Ad5) and E3 ACAACCAACACAACCCGGGTCTCCCCGGGTTGTGTTGGTTGTAGAGGGTTTT-30 region (28 592–30 470 in nucleotide of Ad5) was linearized by BstBI and the bottom strand sequence was 50-AGTCAAAACCCTCTACAACCAA- digestion. The linearized vectors were cotransformed into Escherichia coli CACAACCCGGGGAGACCCGGGTTGTGTTGGTTGTAGAGGGG-30 was selected BJ5183 cells for homologous recombination, generating the recombinant and annealed for subcloning into pSP72DE3-U6. In the case of hTGF-b2 adenoviral plasmid dl324-IX. shRNA, we screened five candidate sequences. The selected target The shRNA construct and U6 promoter were subcloned into the sequence was 50-GGATTGAGCTATATCAGATTCTCAA-30 and the loop pSP72DE3, E3 shuttle vector. The adenoviral shuttle vector, pSP72DE3-U6- sequence was 50-TCTC-30. For expression of hTGF-b2 shRNA in adenovirus, shRNA, was linearized by XmnI digestion. The adenoviral vector dl324-IX the top strand sequence 50-GATCCGGATTGAGCTATATCAGATTCTCAA was linearized by SpeI digestion and the linearized vectors were

Target 1 : 5’ CCAGATCCTGTCCAAACTAAGGCTC 3’ Target 6 : 5’ GGCAGTAGCTCCCCTATTTAAGAAC 3’ Target 1 : 5’ CTGCAAGACTATCGACATGGAGCTG 3’ Target 2 : 5’ GCCCGAAGCGGACTACTATGCTAAA 3’ Target 7 : 5’ GCAACAACGCCATCTATGAGAAAAC 3’ Target 2 : 5’ ACCAGAAATACAGCAACAATTCCTG 3’ Target 3 : 5’ CTTCAATACGTCAGACATTCGGGAA 3’ Target 8 : 5’ GGAATACAGGGCTTTCGATTCAGCG 3’ Target 3 : 5’ CACCAACTATTGCTTCAGCTCCACG 3’ Target 4 : 5’ GGGACCCTGCCCCTATATTTGGAGC 3’ Target 9 : 5’ GCCCTGGATACCAACTATTGCTTCA 3’ Target 5 : 5’ GCAGTTGTCCAACATGATTGTGCGC 3’ Target 10 : 5’ CCCTCTACAACCAACACAACCCGGG 3’ Target 11 : 5’ AAGACCATCGACATGGAGCTGGTGA 3’

120 120 120 100 100 100 80 80 80 1 (%) 1 (%)   60 1 (%) 60

 60 40 40 mTGF- 40 mTGF- hTGF- 20 20 20 0

0 of level Relative expression Relative expression level of level Relative expression

Relative expression level of level Relative expression 0 luciferase hTGF-β1 hTGF-β1 hTGF-β1 shRNA shRNA#1 shRNA#2 shRNA#3

Figure 1. Screening of human and mouse transforming growth factor (hTGF and mTGF)-b1 short hairpin RNAs (shRNAs). (a) Sequences of three shRNA oligomers targeting hTGF-b1. The selected target sequence was indicated in bold. (b) Relative expression level of hTGF-b1 mRNA. Three oligomers of the target and control shRNA (luciferase shRNA) were transfected into HeLa cells. The knockdown efficiency of these oligomers was measured by quantitative real-time polymerase chain reaction (PCR) amplifying hTGF-b1. The relative expression level of hTGF- b1 was plotted after normalization to the luciferase shRNA control. (c) Sequences of 11 shRNA oligomers targeting mTGF-b1. The selected target sequence was indicated in bold. (d) Relative expression level of mTGF-b1 mRNA. All candidate shRNAs and control shRNA (transformation-related protein 53 (Trp53) shRNA) were transfected into NIH3T3 cells. The knockdown efficiency of these oligomers was measured by quantitative real-time PCR amplifying mTGF-b1 mRNAs. The relative expression level of mTGF-b1 was plotted after normalization to the transfection control.

& 2013 Nature America, Inc. Cancer Gene Therapy (2013), 94 – 100 Target specificity and interrelation between TGF-b isotypes SOhet al 96 cotransformed into E. coli BJ5183 cells for homologous recombination. The TGF-b1 protein is present in fetal bovine serum, infected cells were homologously recombined adenoviral plasmids, dl324-IX-DE3-U6-NC and incubated for one more day after changing to serum-free medium and the dl324-IX-DE3-U6-shTGF-b1, were then digested with PacI and transfected ELISA was performed after harvesting the medium. After the harvested into 293 cells to generate the replication-incompetent adenovirus. The medium was activated, the activated sample or TGF-b1 standard protein infectious titer of the adenovirus was determined by a limiting dilution was incubated for 2 h at room temperature. Then, the samples were assay using the Adeno easy vector system (Qbiogene, Carlsbad, CA, USA). completely washed and horseradish peroxidase-conjugated TGF-b1 anti- body was added for 2 h at room temperature. Finally, after incubating with the substrate solution for 30 min and adding stop solution, the optical Quantitative real-time PCR density of each well was determined at 450 nm by using a microplate DU-145 or A375 cells were infected with a defective adenovirus expressing reader (Molecular Devices Corporation, Sunnyvale, CA, USA). For the hTGF-b1, hTGF-b2, mTGF-b1 or scrambled shRNA. After 2 days of infection, measurement of mTGF-b1, hTGF-b1 and hTGF-b2, ELISA was performed cells were lysed with Trizol reagent (Life Technologies) and the total RNA according to the manufacturer’s instructions for Quantikine mTGF-b1 (R&D was isolated via chloroform extraction. The RNA concentration was Systems, Minneapolis, MN, USA) or Quantikine hTGF-b1 or Quantikine determined by using the Nanodrop 2000 (Thermo Scientific, Fremont, hTGF-b2 (R&D Systems), respectively. In the case of TGF-b3, ELISA was CA, USA). The real-time polymerase chain reaction (PCR) was assayed with purchased from Koma Biotech Inc. (Seoul, South Korea) for the the Power SYBR Green RNA-to-CT 1-Step Kit (Life Technologies). The measurement of hTGF-b3, hTGF-b3 ELISA kit. reaction mixture was prepared with RT enzyme mix, reverse transcription (RT)-PCR mix, forward primer, reverse primer, RNA template and nuclease- free water. hTGF-b1 was amplified with the forward primer, 50-CAA Statistical analysis GGGCTACCATGCCAACT-30 and reverse primer, 50-AGGGCCAGGA Data were presented as mean±standard error of mean (s.e.m.), and the CCTTGCTG-30. Human b-actin was amplified with the forward primer, significant differences between groups were determined by unpaired two- 50-ACTCTTCCAGCCTTCCTTC-30 and reverse primer, 50-ATCTCCTTCTGC tailed t-test. P-values were calculated using GraphPad Prism version 6.0. ATCCTGTC-30. mTGF-b1 was amplified with the forward primer, 50-TTGC A value of Po0.05 or Po0.01 was considered as statistically significant. TTCAGCTCCACAGAGA-30 and reverse primer, 50-TGGTTGTAGAGGGCAAG- All experiments were performed three times independently. GAC-30. Mouse b-actin was amplified with the forward primer, 50- GGCTGTATTCCCCTCCATCG-30 and reverse primer, 50-CCAGTTGGTAA- CAATGCCATGT-30. RESULTS Suppression of TGF-b1 with an adenoviral vector expressing TGF- ELISA b1 shRNA Enzyme-linked immunosorbent assays (ELISAs) were performed to To identify the best target sequence for suppression of hTGF-b1, estimate the level of secreted TGF-b1 after infection with an adenovirus three oligomers of TGF-b1 shRNA plus control shRNA (luciferase expressing TGF-b1 shRNA. Adenovirus infection was carried out one day shRNA) were tested by Genolution Pharmaceuticals Inc. using real- after seeding 2 Â 105 A375 or DU-145 cells into a six-well plate. Because time RT-PCR. Of the three selected TGF-b1 shRNAs (Figure 1a),

900 120 sh-NC cell only 800 shTGFβ1 sh-NC 100 700 shTGFβ1 600 80 500 1 (%) 1(pg/ml)  60  400 40 300 hTGF- hTGF- 200 20 100

Relative expression level of 0 0 1 5 10 50 100 200 0151050100 MOI MOI

350 cell only sh-NC 120 sh-NC shTGFβ1 300 shTGFβ1 100 600 250 80 500 200 1(pg/ml) 1 (%) 400   60 cell only 150 1(pg/ml) 300 sh-NC  40 shTGFβ1 100 hTGF-

200 mTGF- 20 hTGF- 100 50

Relative expression level of 0 0 0 1 5 10 50 100 0 1 5 10 50 100 0500100 MOI MOI MOI Figure 2. Downregulation of human transforming growth factor (hTGF)-b1 mRNA by TGF-b1 short hairpin RNA (shRNA). Cancer cells (human A375 cells (a, b) or DU-145 cells (c, d)) were infected with adenovirus-expressing shRNA targeting human TGF-b1 (dl324-shTGF-b1) or scrambled DNA (dl324-sh-NC). TGF-b1 mRNA (a, c) and protein levels (b, d) were assayed by quantitative real-time polymerase chain reaction (qRT-PCR) and enzyme-linked immunosorbent assay (ELISA), respectively. (e) After infection of adenoviruses (100 and 500 MOI) expressing shRNA targeting mouse TGF-b1 (dl324-sh-mTGF-b1) or scrambled DNA (dl324-sh-NC) to B16F10, which were samples for ELISA. See Materials and methods for more details. MOI, multiplicity of infection; NC, negative control.

Cancer Gene Therapy (2013), 94 – 100 & 2013 Nature America, Inc. Target specificity and interrelation between TGF-b isotypes SOhet al 97 sh-NC sh-NC 400 sh-NC 140 140 shTGFβ1 shTGFβ1 shTGFβ1 350 120 120 300 100 100 250 1 (%) 3 (%) 80 2 (%) 80   β 200 60 60 150 hTGF- hTGF- hTGF- 40 40 100 20 20 50 Relative expression level of Relative expression level of Relative expression level of 0 0 0 1 5 10 50 100 1 5 10 50 100 1 5 10 50 100 MOI MOI MOI

sh-NC sh-NC sh-NC 120 120 120 β shTGFβ2 shTGFβ2 shTGF 2

100 100 100

80 80 80 3 (%) 1 (%)   60 2 (%) 60 60  hTGF- hTGF- 40 40 40 hTGF-

20 20 20 Relative expression level of Relative expression level of Relative expression level of

0 0 0 1 5 10 50 100 1 5 10 50 100 1 5 10 50 100 MOI MOI MOI

P<0.01 P<0.001 P<0.001

P<0.001 P<0.01 500 150 150 400

100 100 300 3 mRNA (%)  2 mRNA (%) 1 mRNA (%) 200   50 50

hTGF- 100 hTGF- hTGF- Relative expression level of Relative expression level of Relative expression level of 0 0 0

TGF-β1 TGF-β2 TGF-β3 TGF-β1 TGF-β2 TGF-β3

1000 250 150 800 800 250 600 100 600 80

150 25 (pg/ml) (pg/ml)  400 60

 20 400 100 40 50 15 10 hTGF- 200 hTGF- 50 200 20 5 0 0 0 0 0 0

Figure 3. Correlation of the level of transforming growth factor (TGF)-b isotypes with the decrease in TGF-b1orTGF-b2. A375 cells were infected with various multiplicity of infections (MOIs) of defective adenovirus carrying short hairpin RNA (shRNA) targeting human TGF-b1(hTGF-b1) (dl324-shTGF-b1) or scrambled DNA (dl324-sh-NC) (a)orTGF-b2 (dl324-shTGF-b2) or scrambled DNA (dl324-sh-NC) (b). Infected cells were cultivated for 2 days and harvested for RNA preparation. Quantitative real-time polymerase chain reaction (qRT-PCR) was used to determine the expression level of hTGF-b1, hTGF-b2and hTGF-b3 by using subtype-specific primer sets, with normalization to actin. (c) DU-145 cells were infected with 100 MOIs of defective adenovirus carrying shRNA targeting hTGF-b1 (dl324-shTGF-b1) or TGF-b2 (dl324-shTGF-b2), or scrambled DNA (dl324-sh-NC). Infected cells were cultivated for 2 days and harvested for RNA preparation. qRT-PCR was used to determine the expression level of hTGF-b1, hTGF-b2andhTGF-b3 by using subtype-specific primer sets, with normalization to actin (t-test, Po0.001 or Po0.01). (d) A375 (left) or DU-145 cells (right) were infected with 100 MOIs of defective adenovirus carrying shRNA targeting hTGF-b1 (dl324-shTGF-b1) or TGF-b2 (dl324-shTGF-b2), or scrambled DNA (dl324-sh-NC). Infected cells were cultivated for 2 days and harvested for TGF-b1orTGF-b2, or TGF-b3 by performing enzyme-linked immunosorbent assay (ELISA). NC, negative control.

& 2013 Nature America, Inc. Cancer Gene Therapy (2013), 94 – 100 Target specificity and interrelation between TGF-b isotypes SOhet al 98

TGF-1

TGF-1

TGF-1

target gene : 5'- ACCAGAAATACAGCAACAATTCCTG-3‘ one mismatch : 5'- ACCAGAAATATAGCAACAATTCCTG-3’ two mismatch : 5'- ACCAGAAATATAGCAACAGTTCCTG-3’

TGF-2

TGF-2

TGF-2

TGF-2

target gene : 5’- GGATTGAGCTATATCAGATTCTCAA -3’ one mismatch : 5’- GGATTGAACTATATCAGATTCTCAA -3’ two mismatch : 5’- GGATTGAACTATATCAGATCCTCAA -3’

P <0.01 P <0.01 250 150

200 ** 100 150 1 mRNA (%) 2 mRNA (%)  100  50 * 50 * hTGF- hTGF- 0

Relative expression level of 0 Relative expression level of

P<0.05 200 P<0.01 150

150 100 **

100 ** * 2 mRNA (%) 1 mRNA (%)   50 50 * hTGF- hTGF- 0 0 Relative expression level of Relative expression level of

Figure 4. Specificity of short hairpin RNAs (shRNAs) targeting human transforming growth factor (hTGF)-b1 or hTGF-b2. (a) Sequence alignment of human shTGF-b1 with the selected target sequence. In the case of TGF-b1 coding sequences (1–1173 base), the target sequences are 512–536 bases. The mismatched (mis) bases are indicated in bold (top). Sequence alignment of human shTGF-b2 with the selected target sequence. In the case of TGF-b2 coding sequences (1–1329 base), the target sequences are 578–602 bases. The mismatched bases are indicated in bold (bottom). (b) Human cell line (DU-145 (upper) or A375 (bottom)) were transfected with shuttle vectors expressing shRNAs targeting hTGF-b1 (pSP72DE3-shTGF-b1, pSP72DE3-shTGF-b1 with one mismatch, and pSP72DE3-shTGF-b1 with two mismatch) (left) or hTGF- b2 (pSP72DE3-shTGF-b2, pSP72DE3-shTGF-b2 with one mismatch, and pSP72DE3-shTGF-b2 with two mismatch) (right). After 2 days of transfection, total RNA was prepared from each sample, and quantitative real-time polymerase chain rection (PCR) was performed to investigate sequence specificity. Each shuttle vector expressing mismatched shRNA was constructed based on the ‘Construction of TGF-b shRNAs’ as in the Materials and methods section. NC indicates the shuttle vector expressing a scrambled shRNA. NC, negative control. Error bars represent the s.e. from three independent experiments. *Significant difference compared with NC (t-test, Po0.01 or Po0.05); **significant difference compared with NC (t-test, P40.1). TGF-b1 sh2 showed the greatest reduction, 89.7%, in TGF-b1 (luciferase shRNA) were tested (Figure 1c). The analysis indicated mRNA levels when expressed in HeLa cells (Figure 1b). In the case that the greatest reduction of TGF-b1 mRNA levels (76.5%) was of mouse, 11 TGF-b1 shRNA oligomers and control shRNA obtained with TGF-b1 sh10 (Figure 1d).

Cancer Gene Therapy (2013), 94 – 100 & 2013 Nature America, Inc. Target specificity and interrelation between TGF-b isotypes SOhet al 99 TGF-b1 expression in cells after infection with an adenovirus TGF-b1 and TFG-b3, which can further attenuate the expressing shTGF-b1 immunosuppressive effects of cancer cells induced by TGF-b.24 The infection efficiency of adenovirus type 5 was examined both Despite the high similarity between the active domains of the in human and mouse cells before confirming the repression of TGF-b isoforms, TGF-b2 differs from the other isoforms in that it TGF-b1 mRNA and protein. For this purpose, an adenovirus with binds TbRII through different residues and is dependent on the either the GFP gene inserted at the E3 region or the LacZ gene coreceptor b-glycan for function.25,26 Although TGF-b1 and TGF- inserted at the E1 region was constructed. We used A375 and DU- b3 are both capable of binding directly to the type II receptor, the 145 cells to determine infection efficiency in human cells and a3 helical region of TGF-b1 is structurally ordered, while that of B16F10 and B16BL6 cells to determine the infection efficiency in TGF-b3 is structurally disordered, resulting in a more flexible TGF- mouse cells. We observed that the adenovirus could infect most of b3 structure than TGF-b1.27,28 These results suggest that the the human cells with a multiplicity of infection (MOI) of 50, structure of the ligand/receptor complexes for TGF-b1 and TGF-b3 whereas it could infect only a few mouse cells with an MOI of may be very different, leading to different downstream signaling more than 500 (Supplementary Figure 1). Then, we examined the pathways. Furthermore, Akhurst et al.29 found that the different viral efficacy of adenoviruses expressing shTGF-b1 with various temporal–spatial expression of the TGF-b isoforms in embryogen- MOIs (1, 5, 10, 50 and 100) in human or mouse cells (100 and esis is a sign of uncompensated non-overlapping functions during 500 MOI). To determine whether these viruses decreased TGF-b1 development. However, our results (Figure 3) suggest that, expression at the mRNA and protein level, we used real-time PCR although they may have some unique biological roles, TGF-b1 and ELISA, respectively. TGF-b1 mRNA was decreased by 42% at and TGF-b3 likely also have some similarities as they are able to only 1 MOI and TGF-b1 mRNA was suppressed by 98% at 200 MOI compensate for one another in some situations. In cancer, TGF-b1 in DU-145 cells (Figure 2a), suggesting that the viral efficacy is is generally more abundant than TGF-b2 or TGF-b3.30 Interestingly, dependent on the viral MOI. TGF-b1 protein was decreased when TGF-b1 was selectively downregulated by the shTGF-b1 similarly to TGF-b1 mRNA (Figure 2b). Similar results were adenovirus, endogenous TGF-b3 dramatically increased (Figures observed in A375 cells (Figures 2c and d). Moreover, as expected, 3a (right) and c (right)). These results raise the possibility that at the viral efficacy of adenoviruses expressing shRNA no. 10 of least one of the function of TGF-b1, suppression of the antitumor mTGF-b1 in mouse cells could be confirmed only at 10-fold higher immune response, could still be intact when TGF-b1 is down- MOIs compared with human cells by determining the TGF-b1 regulated owing to the compensatory overexpression of TGF-b3.31 protein level (Figure 2e). However, this compensatory mechanism did not include TGF-b2 (Figures 3a (middle) and c (middle)). Moreover, TGF-b2 down- Level of TGF-b isotypes correlates with the decrease in TGF-b1or regulation by TGF-b2 shRNA strikingly correlated with down- TGF-b2 mRNA regulation of the other TGF-b isotypes. This was a surprising result and it is not yet fully understood how TGF-b2 could have an effect To examine the possibility that a compensatory mechanism was on the endogenous levels of the other TGF-b isotypes. One triggered during TGF-b1 suppression, the expression level of TGF- possibility is that a common transcription factor exists that b isotypes was examined by real-time PCR. Strikingly, when the regulates both TGF-b1 and TGF-b3, which is regulated by TGF-b2. expression of TGF-b1 was decreased by adenovirus-expressing This possibility is currently under investigation. For one reason or shTGF-b1 (at the highest MOI), the expression of TGF-b3 mRNA another, this study provides a novel therapeutic strategy to significantly increased, while the expression of TGF-b2 mRNA was overcome TGF-b-related immunosuppressive activities by only not changed too much (Figure 3a). Then, we examined shRNA TGF-b2 downregulation. corresponding to the target sequence of hTGF-b2, which has a similar immunosuppressive function in tumor cells as TGF-b1. Unexpectedly, when the expression of TGF-b2 was suppressed by ABBREVIATIONS adenovirus-expressing shTGF-b2, the expression level of both TGF- TGF-b, transforming growth factor-b; shRNA, short hairpin RNA; b3 and TGF-b1 was decreased (Figure 3b). We further validated GFP, green fluorescence protein; ELISA, enzyme-linked immuno- the interplay among TGF-b variants in response to treatment with sorbent assay shTGF-b1 RNA or shTGF-b2 RNA in DU-145 cells. We found that shTGF-b2-expressing adenovirus also suppressed the expression of TGF-b1 and TGF-b3, as well as TGF-b2, with the exception that CONFLICT OF INTEREST the expression of TGF-b2 was suppressed by adenovirus-expres- The authors declare no conflict of interest. sing shTGF-b1(Po0.01) (Figure 3c). Then, we confirmed these patterns by measuring the secreted level of TGF-b variants (Figure 3d). ACKNOWLEDGEMENTS This work was supported by the Industrial Strategic Technology Development Target specificity of TGF-b1 shRNA program (10035562: Development of nucleic acid-based anticancer drugs over- coming the immunotherapy resistance) funded by the Ministry of Knowledge Next, we investigated the requirement for sequence specificity in Economy (MKE, Korea). This work was also supported by the Basic Science Research shRNA experiments by transfection of shuttle vector expressing Program through the National Research Foundation of Korea (NRF) and funded by human shTGF-b1 or shTGF-b2 and their corresponding one- or the Ministry of Education. Science and Technology (2012-0002108). S Oh and E Kim two-mismatched shRNA (Figure 4a). This analysis showed that are funded by the Brain Korea 21 project for Medical Science, Yonsei University, even a single base pair difference in the shRNA completely College of Medicine, Seoul, South Korea. abrogated the ability of the shuttle vector expressing shTGF-b1or shTGF-b2 to suppress TGF-b1 or TGF-b2 at the mRNA level (Figure 4b). REFERENCES 1 Massague J. The TGF-beta family of growth and differentiation factors. Cell 1987; 49: 437–438. DISCUSSION 2 Moore LD, Isayeva T, Siegal GP, Ponnazhagan S. Silencing of transforming growth factor-beta1 in situ by RNA interference for breast cancer: implications for pro- In this study, we showed the overriding of TGF-b1 downregulation liferation and migration in vitro and metastasis in vivo. Clinical Cancer Res 2008; by other TGF-b isoform (TGF-b3) due to the partly overlapping 14: 4961–4970. 23 functions of three isoforms of TGF-b, whereas TGF-b2 3 Annes JP, Munger JS, Rifkin DB. Making sense of latent TGFbeta activation. JCell downregulation was accompanied by downregulation of both Sci 2003; 116(Part 2): 217–224.

& 2013 Nature America, Inc. Cancer Gene Therapy (2013), 94 – 100 Target specificity and interrelation between TGF-b isotypes SOhet al 100 4 Massague J. TGFbeta in Cancer. Cell 2008; 134: 215–230. 18 Zhang H, Kolb FA, Jaskiewicz L, Westhof E, Filipowicz W. Single processing center 5 Pardali K, Moustakas A. Actions of TGF-beta as tumor suppressor and pro-meta- models for human dicer and bacterial RNase III. Cell 2004; 118: 57–68. static factor in human cancer. Biochim Biophys Acta 2007; 1775: 21–62. 19 Rand TA, Ginalski K, Grishin NV, Wang X. Biochemical identification of Argonaute 2 6 Massague J, Blain SW, Lo RS. TGFbeta signaling in growth control, cancer, and as the sole protein required for RNA-induced silencing complex activity. Proc Natl heritable disorders. Cell 2000; 103: 295–309. Acad Sci USA 2004; 101: 14385–14389. 7 Elliott RL, Blobe GC. Role of transforming growth factor beta in human cancer. 20 Kumar LD, Clarke AR. Gene manipulation through the use of small interfering J Clin Oncol 2005; 23: 2078–2093. RNA (siRNA): from in vitro to in vivo applications. Adv Drug Deliv Rev 2007; 59: 8 Kim S, Buchlis G, Fridlender ZG, Sun J, Kapoor V, Cheng G et al. Systemic blockade 87–100. of transforming growth factor-beta signaling augments the efficacy of immuno- 21 Sharma A, Tandon M, Bangari DS, Mittal SK. Adenoviral vector-based strategies for gene therapy. Cancer Res 2008; 68: 10247–10256. cancer therapy. Curr Drug Ther 2009; 4: 117–138. 9 Ivanovic V, Demajo M, Krtolica K, Krajnovic M, Konstantinovic M, Baltic V et al. 22 Douglas JT. Adenoviral vectors for gene therapy. Mol Biotechnol 2007; 36: 71–80. Elevated plasma TGF-beta1 levels correlate with decreased survival of metastatic 23 Schlingensiepen KH, Jaschinski F, Lang SA, Moser C, Geissler EK, Schlitt HJ et al. breast cancer patients. Clin Chim Acta 2006; 371: 191–193. Transforming growth factor-beta 2 gene silencing with trabedersen (AP 12009) in 10 Conroy H, Galvin KC, Higgins SC, Mills KH. Gene silencing of TGF-beta1 pancreatic cancer. Cancer Sci 2011; 102: 1193–1200. enhances antitumor immunity induced with a dendritic cell vaccine by reduc- 24 Olivares J, Kumar P, Yu Y, Maples PB, Senzer N, Bedell C et al. Phase I trial of ing tumor-associated regulatory T cells. Cancer Immunol Immunother 2012; 61: TGF-beta 2 antisense GM-CSF gene-modified autologous tumor cell (TAG) 425–431. vaccine. Clin Cancer Res 2011; 17: 183–192. 11 Wei H, Liu P, Swisher E, Yip YY, Tse JH, Agnew K et al. Silencing of the TGF-beta1 25 De Crescenzo G, Hinck CS, Shu Z, Zuniga J, Yang J, Tang Y et al. Three key residues gene increases the immunogenicity of cells from human ovarian carcinoma. underlie the differential affinity of the TGFbeta isoforms for the TGFbeta type II J Immunother 2012; 35: 267–275. receptor. J Mol Biol 2006; 355: 47–62. 12 Ding Y, Kim JK, Kim SI, Na HJ, Jun SY, Lee SJ et al. TGF-{beta}1 protects against 26 Laverty HG, Wakefield LM, Occleston NL, O’Kane S, Ferguson MW. TGF-beta3 and mesangial cell apoptosis via induction of autophagy. J Biol Chem 2010; 285: cancer: a review. Cytokine Growth Factor Rev 2009; 20: 305–317. 37909–37919. 27 Hinck AP, Archer SJ, Qian SW, Roberts AB, Sporn MB, Weatherbee JA et al. 13 Dogar AM, Towbin H, Hall J. Suppression of latent transforming growth factor Transforming growth factor beta 1: three-dimensional structure in solution and (TGF)-beta1 restores growth inhibitory TGF-beta signaling through microRNAs. comparison with the X-ray structure of transforming growth factor beta 2. J Biol Chem 2011; 286: 16447–16458. Biochemistry 1996; 35: 8517–8534. 14 Schafer H, Struck B, Feldmann EM, Bergmann F, Grage-Griebenow E, Geismann C 28 Bocharov EV, Blommers MJ, Kuhla J, Arvinte T, Burgi R, Arseniev AS. Sequence- et al. TGF-beta1-dependent L1CAM expression has an essential role in macro- specific 1H and 15N assignment and secondary structure of transforming growth phage-induced apoptosis resistance and cell migration of human intestinal factor beta3. J Biomol NMR 2000; 16: 179–180. epithelial cells. Oncogene 2013. 29 Akhurst RJ, Lehnert SA, Faissner A, Duffie E. TGF beta in murine morphogenetic 15 Almeida R, Allshire RC. RNA silencing and genome regulation. Trends Cell Biol processes: the early embryo and cardiogenesis. Development 1990; 108: 645–656. 2005; 15: 251–258. 30 Bierie B, Moses HL. Transforming growth factor beta (TGF-beta) and inflammation 16 Aagaard L, Rossi JJ. RNAi therapeutics: principles, prospects and challenges. Adv in cancer. Cytokine Growth Factor Rev 2010; 21: 49–59. Drug Deliv Rev 2007; 59: 75–86. 31 Li C, Wang J, Wilson PB, Kumar P, Levine E, Hunter RD et al. Role of transforming 17 Whitehead KA, Langer R, Anderson DG. Knocking down barriers: advances in growth factor beta3 in lymphatic metastasis in breast cancer. Int J Cancer 1998; siRNA delivery. Nature reviews. Drug Discov 2009; 8: 129–138. 79: 455–459.

Supplementary Information accompanies the paper on Cancer Gene Therapy website (http://www.nature.com/cgt)

Cancer Gene Therapy (2013), 94 – 100 & 2013 Nature America, Inc.