The study was approved by the Ethics Committee of Copenhagen and Frederiksberg. Skin biopsies were collected from 21 patients with MF after written informed consent.

Three CTCL cell lines have been used: MyLa2000 derived from a plaque biopsy of a patient with MF ^[25], SeAx ^[26] and Hut-78 ^[27] derived from peripheral blood of patients with Sézary syndrome. MyLa2000 and SeAx cells were cultured in DMEM containing 4.5 g/l glucose, 10% fetal bovine serum (FBS) and at 37°C under 5% CO₂. Hut-78 cells were cultured in RPMI 1640 containing 2 mM L-glutamine and 10% FBS. Human primary T-cells (CD4⁺ and CD8⁺, quiescent T-cells) from three healthy donors were purified using density gradient centrifugation (Ficoll-PlaqueTM PLUS, Amersham Biosciences, Uppsala, Sweden) as previously described ^[15]. Sézary cells (CD4⁺CD7⁻) were isolated from the peripheral blood of two patients (2 males; age of 56 and 65 years; CD4⁺/CD8⁺ ratio of 19 and 29) using anti-CD7 biotin conjugated antibody (eBioscence, San Diego, CA) and anti-CD4 micro beads (Milteny Biotec) together with the anti-Biotin Multisort Kit (Milteny Biotec, Bergisch Gladbach, Germany), as described by Marzec et al.^[28]. Lesional skin comprising 10 human punch biopsies were collected from patients with MF (6 males and 4 females; mean age 75 years; range 65 to 90 years; 7 plaque/patch (T2) and 3 tumour (T3) stadium) and stored in RNAlater (Ambion, Austin TX) before proceeding to tissue homogenization. In situ hybridization (ISH) was performed on paraffin sections of 11 patients with MF (7 males and 4 females; mean age 70 years; range 49 to 88 years; 6 plaque/patch (T2) and 5 tumour (T3) stadium). Patient diagnoses were confirmed by an expert pathologist in accordance with the WHO-EORTC classification ^[1], ^[29].

GSI-1 (Z-Leu-Leu-Nle-CHO), Ly294002 (2-(4-morpholino)-8-phenyl-4H-1-benzopyran-4-one), Akt inhibitor X (10-(4′-(N-diethylamino) butyl)-2-chlorophenoxazine, HCl) were from Merck Calbiochem (Darmstadt, Germany). Nutlin-3a was from Cayman Chemical (Ann Arbor, MI). Bortezomib (MG341) and MG132 (Z-Leu-Leu-Leu-CHO) were purchased from Selleck (West Paterson, NJ) and doxorubicin hydrochloride from Sigma (Aldrich, St. Louis, MO). Drugs were dissolved in dimethyl sulphoxide (DMSO). Mock-treated cells were cultured with DMSO at final concentrations of 0.04–0.2%.

Tissue samples were homogenized using a TissueLyzer II (Qiagen, Valencia, CA, USA) and processed as previously described ^[30]. Briefly, the cells were centrifuged at 600 rpm for 6 minutes and the pellet was washed twice in PBS. Total RNA (TRNA) and small RNAs were purified using the Rneasy® MiniKit (Qiagen, Hilden, Germany) and mirVana MicroRNA Isolation Kit (Ambion, Foster City, CA) according to the manufacturer's instructions. The concentration of TRNA was measured spectrophotometrically using NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE), and RNA integrity was confirmed with Agilent 2100 Bioanalyzer using Agilent Nano RNA kit (Agilent Technologies, Santa Clara, CA). High quality total RNA from human liver was from Ambion.

Mature miRNA expression was measured in the samples containing the same amount of TRNA and small RNAs with quantitative real-time qRT-PCR assay (TaqMan™ microRNA Reverse Transcription kit-4366596, and TaqMan™ Universal PCR Master mix-4324018, Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. miR-191 was used as a reference for the normalization of RT-PCR-data ^[31].

For detection of pri-miR-122, cDNA was prepared from 1 µg of TRNA per reaction using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and amplified using TaqMan® Pri-miRNA Assay and TaqMan® Gene Expression Master Mix mix-4324018 (Applied Biosystems) according to the manufacturer's instructions. 18S rRNA was used as an internal control gene.

The RT-PCR was performed in triplicates using a 7900HT Fast Real-Time PCR System (Applied Biosystems).

ISH was done on skin lesions of patients with MF as previously described ^[32]. Paraffin sections were mounted on SuperFrost®Plus slides (Dako, Glostrup, Denmark), air-dried for 1–2 h at RT, melted in the oven at 60°C for 45 min, deparaffinized in xylene and hydrated through decreasing ethanol concentrations into PBS. Sections were then treated by proteinase-K (15 µg/ml in PK-buffer, 5 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM NaCl) at 37°C for 8 min in a volume of 300 µL in a Dako hybridizer (Dako), washed, and dehydrated through an increasing gradient of ethanol solutions. Two different mercury locked nucleic acid (LNA) miRNA Detection Probes (Exiqon, Vedbaek, Denmark) were used; hsa-miR-122 (40 nM in a formamide-free ISH buffer) and scrambled probe (40 nM). Probes were denatured by heating to 90°C for 4 min and 50 µL probe mixture was hybridized with the tissue sections in the hybridizer at 55°C for 60 min. The slides were then placed at RT in 5× saline-sodium citrate (SSC) (Invitrogen) and washed for 5 min at 55°C in 5× SSC (one wash), 1× SSC (two washes) and 0.2× SSC (two washes). After a further washing in PBS, sections were blocked with DIG Wash and Block Buffer Set (Roche, Mannheim, Germany) and incubated for 60 min at RT with alkaline phosphatase (AP)-conjugated anti-DIG (diluted 1∶800 in blocking solution, Roche). Slides were then washed twice with PBS containing 0.1% Tween-20. Ready to use tablets (Roche) of 4-nitro-blue tetrazolium (NBT) and 5-brom-4-chloro-3′-indolylphosphate (BCIP) substrate were dissolved in aqueous 0.2 mM levamisole. Slides were incubated for 120 min at 30°C to develop the dark-blue NBT-formazan precipitate, washed twice for 5 min in KTBT buffer (50 mM Tris–HCl, 150 mM NaCl, 10 mM KCl), and then twice in water, dehydrated in the ethanol gradient and mounted.

Skin sections were classified into three categories depending on the percentages (<25%, 25–75%, >75%) of malignant T-cells positive for miR-122 staining.

In the case of SeAx cells, transfection was carried out using Lipofectamine RNAiMax (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol as previously described by ^[15]. 20 nM miRIDIAN miRNA Mimics and Hairpin Inhibitors from Dharmacon (Thermo Scientific, Chicago, Il) were used for specific inhibition and overexpression of miR-122. Small Interfering RNA was used at 50 nM for the specific knockdown of PTEN (6251, Cell Signalling, Beverly, MA) and p53 (6231, Cell Signalling).

In the case of MyLa2000 cells, transfection was carried out using 4 µl TurboFect™ in vitro Transfection Reagent (Fermentas, Burlington, Canada). 50 nM miRIDIAN miRNA Mimics and Hairpin Inhibitors from Dharmacon (Thermo Scientific, Chicago, Il) were mixed to 200 µl GIBCO™ Opti-MEM I (Invitrogen, Carlsbad, CA) and incubated for 20 min before application to 500.000 cells/mL.

Scrambled siRNA (Dharmacon) was used as a negative control. Efficiency of transfection was evaluated by SiGLO (Dharmacon) and by RT-PCR (Figure S1).

Apoptosis was determined by flow cytometry using FITC-annexin V/propidium iodide (PI) protocol of the manufacturer (Beckman Coulter), as previously described ^[15]. In some experiments the proportion of viable cells was determined by staining with only PI (5 µg/ml).

Whole cell extracts were prepared as described previously ^[15] and equal amounts of protein were separated by a 12% Bis-Tris gel electrophoresis at 200 V followed by electrophoretic transfer to a nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA). Membranes were then blocked for 1 hour at 4°C with Li-Cor blocking agent (Lincoln, NE), incubated with primary mouse or rabbit antibody overnight at 4°C followed by incubation for 1 hour with the appropriate secondary antibodies labeled with 800IR dye (anti-rabbit) (Li-Cor) or Alexa Fluor 680 (anti-mouse) (both from Molecular Probes, Invitrogen Cooperation, Carlsbad, CA). The following primary antibodies were used: Akt total, phosphorylated Akt (Ser⁴⁷³), total FoxO3a, phosphorylated FoxO3a (Ser^318/321), p53, phosphorylated MDM2 (Ser¹⁶⁶), P70 S6K total and phosphorylated P70 S6K (Thr^421/Ser424) from Cell Signalling (Beverly, MA), anti–β-actin antibody from Sigma (Aldrich, St. Louis, MO). Immunoreactivity was detected and quantified with the infrared Odyssey imaging System (Li-Cor).

Continuous data are reported as means with standard deviation (SD) and the differences were evaluated by the Student's t-test. RT-PCR data were analyzed using the ΔΔCT method by paired Student's t-test. Normality of the data was assessed by Kolmogoroff-Smirnov test. Correlation between miR-122 expression level in the MF biopsies and the stage was analyzed with χ² test. p-value<0.05 was considered significant. Statistical analysis was performed by the SPSS Version 17.0 (SPSS Inc., Chicago, Il), GraphPad Prism Version 4.03 (GraphPad Software Inc., San Diego, CA) or Excel (Microsoft Corp., Redmond, WA).

In order to identify the miRNA species involved in the regulation of apoptosis in CTCL, we investigated the influence of GSIs on miRNA expression pattern in CTCL cell lines. The rationale for choosing GSIs was our previous studies showing that these agents potently induce apoptosis via the inhibition of the proteolytic cleavage of Notch ^[8], ^[14], ^[15]. Preliminary results obtained from miRNA arrays were described by Manfé et al.^[22] and revealed miRNA species, differentially expressed between GSI-1 treated and control cells (miRNAs down-regulated: miR-27a, miR-92b, miR-181a, miR-18a, miR-425, miR-193b, miR-30d, miR-199a-5p, miR-19b, miR-222, miR-221, miR-9*, miR-125a-5p, miR-23b; miRNAs up-regulated: miR-122, miR-214, miR-185, miR-574-5p, miR-638, miR-149*, miR-766, miR-595). Subsequent functional analysis using the corresponding antagomiRs and mimic oligonucleotides showed a functional significance of one of the deregulated miRNAs: miR-122. miR-122 has been considered to be liver specific but recent evidence showed that this miRNA is more widely distributed and is also detected in human skin ^[30]. Here, we analyzed miR-122 expression by LNA-ISH in 11 sections of lesional skin biopsies of patients with MF. Figure 1A shows miR-122 presence in the epidermis and in the infiltrate of malignant T-cells in plaque/patch (T2) and tumour (T3) stage MF. LNA-ISH with a control scrambled RNA did not give any signal (Figure 1A). miR-122 positive signal was significantly increased in the advanced stage MF (χ² test, p-value = 0.046); 3 of 5 cases of tumour, but none of the 6 cases of plaque/patch, displayed more than 75% of malignant T-cells positive for miR-122 (Figure 1B). We concluded that miR-122 expression correlated with the T-stage in MF.

We also measured miR-122 expression level in vitro and in vivo in CTCL by TaqMan real-time RT-PCR. As shown in Figure 1C, miR-122 was not detectable in quiescent-T cells, but it was expressed in three CTCL cell lines (SeAx, MyLa2000, HuT-78), in leukaemic cells purified from the patients with Sézary syndrome and in the lesional skin from patients with MF. However, the miR-122 expression in CTCL was lower than the one in human tissue liver. Exposure to GSI-1 increased the expression of miR-122 in SeAx and MyLa2000 cells as previously reported ^[22], but not in Hut-78 cells. This effect was not specific to GSI since proteasome inhibitors (another class of chemotherapeutic agents which induce apoptosis in CTCL (15)) and doxorubicin increased miR-122 expression in SeAx and MyLa2000 cells (Figure 1D). The treatment of purified leukaemic Sézary cells with GSI-1 also increased the expression of miR-122 (Figure 1D).

We further examined the effect of miR-122 up-regulation on apoptosis in CTCL cell lines. We attenuated miR-122 using miR-122 hairpin inhibitor (antagomiR-122) in GSI-1-treated or untreated SeAx cells (Figure 2A, B). SeAx treatment with antagomiR-122, but not with the scrambled oligonucleotide, led to a substantial increase in apoptosis induced by GSI-1 but did not affect apoptosis in the untreated cells. mir-122 overexpression had an opposite effect and protected against GSI-1 induced apoptosis. We observed similar results with MG132 and bortezomib (Figure 2C). We also examined the impact of attenuation of miR-122 expression on apoptosis in another CTCL cell line, MyLa2000. Similar to SeAx cells, we observed an increase in the number of apoptotic cells after GSI-1, bortezomib and MG132 treatment in miR-122 depleted cells relative to the scrambled oligonucleotide control (Figure 2D). Taken together, our data suggested that miR-122 conferred a protective response of malignant T-cells to chemotherapy-induced apoptosis in CTCL.

Our next aim was to determine which signalling pathways involved in apoptosis are modulated by miR-122. It is known that apoptosis in SeAx cells can be caused by the inhibition of Stat3 or Akt signalling ^[8], ^[33], ^[34]. Overexpression of miR-122 did not inhibit either Stat3 or NF-kB (data not shown). However, we found that miR-122 overexpression induced Akt phosphorylation at Ser⁴⁷³ after 4 hours (∼2 fold increase compared with control siRNA) (Figure 3A). We confirmed the functional activation of Akt in miR-122 overexpressing cells by showing that the downstream targets of Akt, p70 S6K and Forkhead FOXO3a transcription factor (FXHRL1), were phosphorylated (Figure 3A). The pro-apoptotic drugs (GSI-1, bortezomib and MG132) decreased the Akt phosphorylation (Figure 3A, B) and the simultaneous treatment with miR-122 mimic and GSI-1 recovered Akt activity (Figure 3A). These data indicated that miR-122 inhibited apoptosis via activation of Akt. We further confirmed that Akt activation was functional in the regulation of apoptosis by showing that inhibition of PTEN, the major negative regulator of Akt, protected the SeAx cells from GSI-1-induced cytotoxicity (Figure 3C).

Interestingly, we observed a reciprocal regulation between miR-122 and Akt. In addition to the above described activation of Akt by miR-122, the inhibition of Akt activity by 2 different blockers, Ly294002 (a potent inhibitor of phosphoinositide 3-kinases, PI3Ks) and Akt Inhibitor X (inhibitor of the phosphorylation of Akt and its in vitro kinase activity) induced expression of miR-122 in SeAx cells (Figure 3D). Notably, no miR-122 increase following the Akt inhibition was observed in Hut-78 cells (Figure 3D).

To elucidate whether the observed increase in miR-122 is a direct consequence of Akt inhibition or is mediated by another protein we focused on a possible role of p53. The idea that p53 might be involved arose from the observations in another T-cell neoplasm, T-ALL, where hyperactivation of Notch-1 inhibits PTEN, which in turn leads to activation of Akt and p53 degradation via induction of human homolog of the mouse double minute 2 p53-binding protein (MDM2) (for a review see Dotto ^[35]). We observed that GSI-1, bortezomib and MG132 induced p53 accumulation in SeAx cells (Figure 4A, B) and that miR-122 mimic oligonucleotide inhibited this response (Figure 4B). Third, Akt inhibition by Ly294002 and Akt Inhibitor X increased p53 (Figure 4C) and the activation of Akt by siRNA against PTEN increased the phosphorylation of Ser¹⁶⁶ on MDM2 and caused a degradation of p53 (Figure 4D).

To further investigate the role of p53 in apoptosis in CTCL we used nutlin-3a that is a known inducer of p53 acting via disruption of p53-MDM2 interaction ^[36]. Nutlin-3a amplified the apoptosis induced by Akt inhibition (Figure 4E) and was also able to sensitize SeAx cells to GSI-1, MG132 and bortezomib (Figure 4F). However, the up-regulation of p53 by nutlin-3a did not change the phosphorylation status of Akt (Figure S2A) and did not cause apoptosis (Figure S2B). The down-regulation of p53 by a specific siRNA (Figure S2C) did not affect Akt activation either (Figure S2D). Thus, p53 modulated the sensitivity of SeAx cells to apoptosis, but did not cause cell death on its own. This conclusion was further reinforced by analysis of another CTCL cell line, Hut-78. These cells have a mutated P53 (P53^mut) ^[37] and did not accumulate p53 after nutlin-3a treatment (Figure 5A) and were more resistant than SeAx to the apoptosis after treatment with GSI-1, bortezomib and MG132 (Figure 5B).

The induction of miR-122 during apoptosis seemed to occur via p53. In SeAx cells nutlin-3a led to p53 accumulation (Figure 5A) and potently induced miR-122 (Figure 5C). In contrast, in Hut-78 cells, harbouring a truncated form of p53 ^[37], we observed neither a stabilization of p53 nor up-regulation of miR-122 after exposure to nutlin-3a (Figure 5A, C). Similarly, GSI-1, bortezomib and MG132 induced p53 (Figure 4A, B) and miR-122 in SeAx cells, but not in Hut-78 cells (Figure 1D). Therefore, we determined whether the miR-122 up-regulation by chemotherapies was mediated by p53. p53 in SeAx cells was down-regulated with siRNA and the cells were exposed for 24 h to GSI-1, bortezomib, MG132 and doxorubicin. The expression of miR-122 was significantly up-regulated after the treatment in the control cells transfected with scrambled oligonucleotide but not in the cells in which p53 was silenced by siRNA (Figure 5D). In order to examine whether the miR-122 modulation by p53 in SeAx cells was mediated transcriptionally or post-transcriptionally, we used qRT-PCR for examining changes in miR-122 primary transcript (pri-miR-122) and mature form (mature miR-122) levels following nutlin-3a exposure (Figure 5E). Time course analysis revealed no significant change in pri-miR-122 (less than two-fold), but showed an increased mature miR-122 expression upon nutlin-treatment. Thus, we suggest that p53 modulation of miR-122 occurs post-transcriptionally.

The results of our experiments led to a model of miR-122 action outlined in Figure 6. Akt inactivation is a major effector pathway in drug-induced apoptosis and cell death is additionally amplified by p53. On the other hand, the p53 signalling leads to the induction of miR-122, which is antiapoptotic via activation of Akt.