Oncogene (1997) 15, 2349 ± 2359  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Retinoic acid enhances the expression of -induced : evidence for multiple mechanisms of action

Luis Pelicano1, Fengsheng Li2, Christian Schindler2 and Mounira K Chelbi-Alix1

1CNRS-UPR 9051; 1, avenue Claude Vellefaux, HoÃpital St Louis, 75010 Paris, France; 2Division of Molecular Medicine, College of Physicians and Surgeons of Columbia University, New York, New York 10032, USA

Retinoic acid (RA) and (IFNs) are negative Tyk2, phosphorylate Stat1, Stat2 and Stat3. The IFN- regulators of proliferation. In vitro and in vivo, their stimulated factor 3 (ISGF3) is formed between combination leads to a more potent growth inhibition. Stat1 and Stat2 (Stat1 : 2) in association with the DNA- However, the molecular mechanisms by which RA and binding , p48, a member of the interferon IFNs potentiate each other are not fully understood. As regulatory factor (IRF) family (Fu et al., 1992; some IFN-induced gene products regulate cell growth Schindler et al., 1992). This complex binds to the and/or antiviral activity, we analysed the e€ects of RA IFN-stimulated response element (ISRE) found in on their expressions. RA increases the level of promoters of IFNa/b-stimulated . In addition, 2'5'oligoadenylate synthetase, p68 , the promyelo- homo- and heterodimers of Stat1 and Stat3 (Stat1 : 1, cytic leukemia protein (PML) and Sp100 in both HL-60 Stat3 : 3 and Stat1 : 3) bind to a palindromic version of and WISH cells. Moreover, RA and IFN act coopera- the IFNg-activated site (GAS), regulating the expres- tively to increase the expression of these proteins. RA sion of distinct ISGs (Kanno et al., 1993; Khan et al., also inhibits vesicular stomatitis replication and 1993; Pine et al., 1994). In response to IFNg, Jak1 and induces a higher antiviral state and growth inhibition Jak2, phosphorylate Stat1 which forms a homodimer when combined with IFN. RA stimulates the IFN and binds to the GAS motif. regulatory factor 1 (IRF-1) directly The diverse biological e€ects of IFNs are believed to through the GAS motif and causes the induction and be mediated by a speci®c set of more than 40 proteins secretion of IFNa. Additional mechanisms could be directly involved in response to IFN treatment. involved as RA increases the level of signal transducing However, only a few of these proteins, namely the activators of (STAT) proteins, and en- p68 kinase (PKR) (Koromilas et al., 1992; Meurs et al., hances the IFN-induced STAT activation, suggesting 1992, 1993), the 2'5' oligoadenylate (2'5'A) synthetase that cooperative e€ects by RA and IFN are mediated (Chebath et al., 1987), certain forms of Mx proteins through multiple pathways. (Staeheli et al., 1993) and the promyelocytic leukemia (PML) protein (Mu et al., 1994; Chelbi-Alix et al., Keywords: interferon; retinoic acid; STAT; IRF-1; 1995, 1996; Koken et al., 1995; Liu et al., 1995; Stadler signalisation pathways et al., 1995) are implicated in the antiviral state and/or anti-tumor and anti-growth activities of IFNs. The enzymatic properties of PKR and 2'5'A synthetase have been well characterized (reviewed in Sen and Introduction Ransoho€, 1993). In short, PKR, once phosphorylated, acquires the capacity to phosphorylate the a subunit of Interferons (IFNs) are a family of secreted the translation initiation factor eIF-2, leading to the with antiviral, antiproliferative and immunomodula- inhibition of protein synthesis. The 2'5'A synthetase tory activities (Pestka et al., 1987; Romeo et al., 1989; catalyses the synthesis, from ATP, of unusual 2'5' Sen and Ransoho€, 1993). IFNs a and b (type I) are oligoadenylates, which in turn activate a latent produced by many types of cells in response to viral ribonuclease. The mechanism of action of Mx or infection, double stranded RNA treatment or other PML proteins remains to be elucidated. PML is a stimuli (Chelbi-Alix et al., 1994 and references within). nuclear matrix protein with growth suppressing IFNg (type II) is induced in T-lymphocytes and natural properties, whose expression is deregulated during killer cells in response to antigens or mitogens. These oncogenesis (Mu et al., 1994; Koken et al., 1995; Liu IFNs, after interacting with di€erent speci®c high- et al., 1995). PML is associated to nuclear compart- anity receptors on the plasma membrane, activate ments called nuclear bodies where it colocalizes with di€erent cascades of signal transduction reactions two IFN-induced proteins, Sp100 (an autoantigen leading to the transcriptional activation of IFN- detected in primary biliary cirrhosis, Guldner et al., stimulated genes (ISGs) (Schindler, 1995). IFN- 1992), and NDP52 (Korioth et al., 1995). induced signal transduction is mediated by the Retinoic acid (RA) is structurally related to vitamin -activation of the signal transducers A. RA exerts di€erent e€ects such as inhibition of and activators of transcription (STAT) proteins proliferation, induction of di€erentiation and immuno- (Schindler, 1995). In response to IFNa/b, Jak1 and modulation. RA modi®es gene expression by binding to speci®c nuclear receptors (Mangelsdorf et al., 1994). Both RA and IFNs exert antiproliferative e€ects on various cell types (Frey et al., 1991). Several groups Correspondence: MK Chelbi-Alix have reported a higher inhibition of cell proliferation Received 22 April 1997; revised 8 July 1997; accepted 9 July 1997 by a combination of RA with IFN, compared to Retinoic acid action on interferon system L Pelicano et al 2350 treatment by each agent alone (Lancillotti et al., 1995 the level of Stat1 in MCF-7, a breast tumor cell line and references within). However, the mechanism of this resistant to growth inhibition by IFNb (Kolla et al., collaborative e€ect is not yet fully elucidated. It is also 1996). not known whether these events are accompanied by In this report, we analysed the e€ect of RA, alone an enhancement of the IFN-induced antiviral state. It and in combination with IFNa, on the expression of has been recently shown that RA upregulates IRF-1 2'5'A synthetase, PKR, Sp100 and PML in di€erent gene expression in myeloid cells (Matikainen et al., human cell lines. The capacity to inhibit cell 1996; Gianni' et al., 1997) and that RA also increases proliferation and vesicular stomatitis virus (VSV)

a

b C IFN RA

PML

Sp100

Figure 1 Capacity of RA to increase IFN-induced proteins. (a) Kinetics of increase of 2'5'A synthetase activity in HL-60 cells after 76 RA, IFNa and poly(I)-poly(C) treatments. HL-60 cells were untreated (&) or treated by RA (10 M)(*), 100 mg/ml of poly(I)- poly(C) (^) or IFNa (500 units/ml) (*). At the indicated times, 2'5'A synthetase activity was determined in cell extracts as described in Materials and methods. Values are an average of three experiments. (b) Increase of PML and Sp100 proteins by RA. 76 HL-60 cells were treated by RA (10 M) for 3 days. Immuno¯uorescence was revealed by monoclonal anti-PML antibodies followed by Texas red, and polyclonal anti-Sp100 antibodies were visualized with FITC Retinoic acid action on interferon system L Pelicano et al 2351 replication was also tested. We demonstrate that the cells treated with RA plus IFNa vs IFNa alone. In e€ects of RA could be mediated through multiple HeLa cells, we failed to detect an increase of 2'5'A mechanisms involving increase of STAT expression, synthetase activity by RA alone or RA-IFNa IFN-induced STAT activation, direct IRF-1 gene compared to IFNa alone (Table 1). induction and IFNa secretion. We then investigated the e€ects of RA and IFN on PML, Sp100 and PKR proteins expression. Equal amounts of total protein extracts from control and treated HL-60 cells for 24 h were analysed by Western Results blotting. At that time, RA per se had no e€ect and IFNa at 100 units/ml only slightly increased the Increase of 2'5'A synthetase, PML and Sp100 proteins expression by RA

The 2'5'A synthetase activity induced by 1076 M RA Table 1 Induction of 2'5'A synthetase activity in HL-60, WISH and was studied in HL-60 cells and compared with that HeLa cells by RA, IFNa and their combination induced by 100 mg/ml of poly(I)-poly(C), a commonly Cells used IFN inducer. The basal level of 2'5'A synthetase HL-60 WISH HeLa activity remained constant in HL-60 cells for periods Treatment 2'5'A synthetase activity (units/mg) from 1 ± 7 days, whereas it increased in the presence of None 1+0.1 46+3 12+1 RA. The kinetics of enzyme activity for RA (Figure 1a) RA 8.1+0.3 100+5 13+1.5 shows a sixfold enhancement at day 1, reaching a IFNa 61+9 110+6 120+8 IFNa+RA 139+15 179+19 126+7 maximum at day 3 (50-fold), and a subsequent decline 76 75 Cells were treated for 24 h with 10 M RA (HL-60), 10 M RA of 80% at day 7. The maximal induction with poly(I)- (WISH and HeLa), IFNa (100 units/ml) or their combination. 2'5' A poly(C) was only fourfold after 3 days. Thus, clearly synthetase activity was measured in cell extracts as described in RA, compared to poly(I)-poly(C), is a better inducer of Materials and methods. Each value is the mean of three experiments 2'5'A synthetase activity in HL-60 cells. Note that, compared to IFN, RA induced enzyme activity with slower kinetics. In WISH cells, the maximum increase of RA-induced 2'5'A synthetase activity occurred at HL60 α 24 h at a concentration of 1075 M (Bourgeade and

Besancon, 1984 and Table 1). These results, considered MW α

in the of the recent observations that three (kDa) T RA IFN RA+IFN nuclear bodies-associated proteins, PML, Sp100 and 144— NDP52 are IFN-induced (Guldner et al., 1992; Chelbi- Alix et al., 1995; Korioth et al., 1995; Stadler et al., PML 1995), prompted us to ask whether the 2'5'A synthetase 87— inducing action of RA could be generalized to other IFN-inducible proteins, such as PML and Sp100. Both of these proteins localize to discrete speckles within the nucleus and their expression can be monitored with a 144— high degree of sensitivity by immuno¯uorescence (Koken et al., 1994). Figure 1b shows that a 3 day 87— Sp100 treatment of HL-60 cells with RA enhanced the levels of PML and Sp100 as revealed by an increased immuno¯uorescence staining. Western blot analysis of 44— extracts from these cells indicated that Sp100, PML and PKR (data not shown) are increased upon RA 87— treatment with kinetics similar to that of 2'5'A synthetase activity. PKR

E€ect of the combination of RA and IFNa on IFN- 44— inducible gene products To evaluate the molecular basis of the previously described antiproliferative e€ect caused by the combi- nation of RA and IFNa in di€erent cell lines (Frey et ACTIN al., 1991; Lancillotti et al., 1995), we ®rst analysed the expression of 2'5'A synthetase, PKR, PML and Sp100, using HL-60, WISH or HeLa cells. Cells were treated for 24 h with 1076 M RA for HL-60 or 1075 M for Figure 2 E€ect of RA, IFNa and their combination on PML, WISH and HeLa, in the presence or absence of 100 Sp100 and PKR proteins in HL-60 cells. HL-60 cells were treated 76 units/ml of IFNa. As shown in Table 1, 2'5'A by 10 M RA, IFNa (100 units/ml) or their combination for synthetase activity was increased by RA or IFNa in 24 h. 20 mg of extracts were analysed by Western blot and both HL-60 and WISH cells. However, when RA was revealed by rabbit anti-PML, anti-Sp100, anti-actin and mouse anti-PKR antibodies. Note that all bands visualized by anti-PML combined with IFNa,2'5'A synthetase expression was or anti-Sp100 antibodies are likely to isoforms derived from even more enhanced. The enzyme activities were 1.6- alternative splicing of unique gene. Molecular size markers are and 2.3-fold higher respectively in WISH and HL-60 indicated on the left Retinoic acid action on interferon system L Pelicano et al 2352 expression of these proteins. However, when RA was induce 2'5'A synthetase activity in MDBK cells, and combined with IFN, an enhancement in the expression that this was completely abolished by the addition of of PML, Sp100 and PKR was observed (Figure 2). anti-IFNa antibodies, demonstrating that IFNa is Note that, as previously described (Chelbi-Alix et al., induced and secreted by RA-treated HL-60 cells. To 1995), di€erent forms of PML and of Sp100 are estimate the amount of secreted IFN, the culture induced by IFNs. They are derived from alternative media, taken from RA-treated HL-60 cells, were RNA splicing of a single PML or Sp100 gene. A titrated on MDBK or HeLa cells challenged with similar enhanced level of expression of these proteins VSV (Figure 3 and data not shown). IFN titers were with combined treatment was observed in WISH cells, expressed in relation to the human IFNa reference. The but, again, not in HeLa cells (data not shown). results show that RA caused the induction and Treatment of HL-60 cells with 1000 units/ml of IFNa secretion of 32 international units/ml of IFNa (at day leads to a much higher increase in the level of PML, 3) in HL-60 cells. Untreated HL-60 cells did not secrete Sp100 and PKR proteins and the combination with detectable IFN activity (data not shown). It should be RA does not give a measurable enhancement of their noted that we failed to detect IFN activity in the expression (data not shown). Taken together, these culture medium from RA-treated WISH cells. How- data suggest that RA and IFNa act through similar ever, a very small amount of IFN would not have been intracellular pathways. detected by our assay (e.g. 54 units/ml).

Characterization of RA-induced IFN in the culture E€ect of RA, IFNa and their combination on medium antiproliferative and antiviral activities RA-treated HL-60 and WISH cells exhibited IFN-like Since the RA-treated HL-60 cells showed an increase in responses such as: increased expression of IFN-induced 2'5'A synthetase activity, PML, PKR and Sp100 proteins (Figure 1), higher antiproliferative and proteins expression (Figure 1), it was of interest to antiviral capacities (See below, Figure 4 and Table 3). determine if RA could modulate the biological However, since previous studies have only identi®ed responses to IFNs (e.g. antiproliferative e€ect and IFN as inducing these activities, we needed to antiviral state) in these cells. We studied the e€ect of determine whether these e€ects were due to IFN combined treatment with RA (1076 M for HL-60 and induction. To test this possibility, anti-human IFN 1075 M for WISH cells) and IFNa (100 units/ml). RA antibodies were added to RA-treated HL-60 and and IFNa inhibited the proliferation of HL-60 and WISH cells and 2'5'A synthetase activities were WISH cells as indicated by a decreased number of evaluated after 72 h for HL-60 and 24 h for WISH viable cells (Figure 4). Treatment with IFNa and RA cells. Table 2 shows that anti-human IFNb antibodies for 4 or 6 days increased the inhibitory e€ect of IFNa had no e€ect on RA-induced 2'5'A synthetase activities by 14 to 20% in both cell types (Figure 4). In order to in HL-60 and WISH cells, wherease anti-human IFNa see whether the increase in the growth inhibition by antibodies abolished in part RA-induced enzyme RA was also accompanied by the establishment of an activities, in both cell lines. This suggests that part of antiviral state, HL-60 cells were treated with RA or the RA e€ect could be due to the induction and IFNa as control. After 3 days, cells were counted and subsequent secretion of IFNa. However, as the amounts of antibodies used in the experiment were amply sucient to neutralize 500 units/ml of IFN, the remaining 2'5'A synthetase activity suggest that RA can act independently of IFN secretion. The secreted IFN was tested by the ability of the culture medium of RA-treated HL-60 cells to induce 2'5'A synthetase and antiviral activities in bovine MDBK cells, in which these activities were highly induced by IFNa (Chelbi- Alix and Thang, 1986) but not by RA (data not shown). It is clear from data shown in Figure 3 that culture medium from RA-treated HL-60 cells is able to

Table 2 E€ect of anti-IFN antisera on 2'5' A synthetase activity induced in RA-treated HL-60 and WISH cells Cells HL-60 WISH Treatment 2'5' A synthetase activity (units/mg) None 1.5+0.5 42+4 RA 64+3 116+11 Figure 3 Kinetics of appearance of RA-induced IFN in the RA+Anti-IFNa Antibodies 16+1 83+4 culture medium of RA-treated HL-60 cells. HL-60 cells (26105) RA+Anti-IFNb Antibodies 62+2 120+12 76 were treated with RA at ®nal concentration of 10 M. At the 76 HL-60 and WISH cells were treated with 10 M RA (72 h) or indicated times, IFN titer value was determined on a portion of 75 10 M RA (24 h) respectively in the presence or absence of anti-IFN the culture medium (*). The remaining portion was added to a or b antibodies, the amounts of anti-IFN antibodies used were untreated MDBK cells (46105) with (&) or without (&) anti- sucient to neutralize 500 units/ml of IFN and were added daily. 2'5' IFNa antibodies. After 18 h incubation at 378C, 2'5'A synthetase A synthetase activity was measured in cell extracts as described in activity (&,&) was determined as described in Materials and Materials and methods methods Retinoic acid action on interferon system L Pelicano et al 2353 an equal number of cells from each sample was ml of IFNa (Figure 3). The results, given in Table 3, infected with VSV at a multiplicity of 0.1. Immuno- show that the presence of anti-IFNa antibodies had no ¯uorescence analysis revealed that VSV antigens e€ect on the inhibition of VSV replication induced by expression was inhibited in RA- as well as in IFN- RA, whereas they completely abolished the action of treated HL-60 cells compared to control infected cells IFNa. Thus, RA can inhibit VSV growth independently (Figure 5). The titers of VSV in RA- or IFNa-treated of IFN secretion. cells were 150 and 1200 times lower respectively than in control infected cells. Next, we studied the e€ect of the E€ect of RA, IFNa and their combination on the PML combination of RA and IFNa on virus replication. HL-60 and WISH cells were treated, as described in Table 1, with RA and/or IFNa for 24 h, and infected The results presented above show that the addition of with VSV at a multiplicity of 0.1. Table 3 shows that RA enhanced the IFN-stimulated induction of several the treatment of HL-60 and WISH cells with RA-IFNa gene products. In part, this can be explained by IFN combination a€orded more antiviral protection against secretion upon RA treatment. In order to know VSV infection than that with IFNa alone. RA boosted whether this combination acts on promoters of the the antiviral activity by 40-fold in HL-60 cells and 11- target genes, HepG2 cells were transfected with fold in WISH cells. plasmids containing the luciferase reporter gene under The inhibition of VSV replication by RA in HL-60 the control of the entire PML promoter (pPromPML and WISH cells might have been due to the induction of 1.44), and were treated for 24 h with RA (1075 M) and IFN by this compound. In order to test this possibility, IFNa (100 units/ml). In accordance to our previous we examined the e€ect of anti-IFNa antibodies on the results (Stadler et al., 1995), the PML promoter antiviral state induced by RA in these cells. As a responded to IFNa by a 2.5-fold increase of luciferase control, cells were treated with IFNa in the presence of activity (Figure 6a) consistently with identi®cation of anti-IFNa antibodies. Note that, upon RA treatment ISRE and GAS motifs. Interestingly, in the same for 24 h, WISH cells did not secrete detectable IFN conditions, RA alone caused a signi®cant increase of (data not shown) and HL-60 cells secreted only 4 units/ 1.6-fold. Moreover, combined treatment led to a much

Table 3 Induction of antiviral activity in HL-60 and WISH cells by RA, IFNa or their combination VSV Growth PFU/ml Cells Treatment HL-60 WISH None 36107 56107 RA 106 4.56106 RA+Anti-IFNa Antibodies 1.26106 46106 IFNa 46104 105 IFNa+Anti-IFNa Antibodies 36107 5.56107 IFNa+RA 103 96103 HL-60 and WISH cells were treated for 24 h with RA, as described Figure 4 E€ects of RA, IFNa and their combination on the in Table 1, or IFNa (100 units/ml) in the presence or absence of anti- growth inhibition of HL-60 and WISH cells. HL-60 (206104) and IFNa antibodies, the amounts of anti-IFN antibodies used were WISH (356103) cells untreated (&) and treated with RA sucient to neutralize 500 units/ml of IFN. HL-60 and WISH cells 76 75 (10 M or 10 M respectively) (^), IFNa (100 units/ml) (*) were also treated for 24 h by RA-IFNa combination. Cells were used and with their respective combination (~), were counted at the for the determination of VSV growth as described in Materials and indicated periods. The results are presented as mean values of methods. Each value is the mean of three experiments, the di€erence viable cells counted in duplicate in the number of plaques counted at each dilution did not exceed 5%

C RA IFN

VSV

Figure 5 RA protects against VSV infection. Expression of VSV antigens in control (C), IFN- or RA-treated HL-60 cells. HL-60 76 cells were treated with IFNa (100 units/ml) or RA (10 M). After 3 days, cells were counted and infected with VSV at a multiplicity of 0.1. Immuno¯uorescence with anti-VSV antibodies was performed 18 h post-infection Retinoic acid action on interferon system L Pelicano et al 2354 a protein synthesis (Matikainen et al., 1996), suggesting a direct activation of the gene. Moreover, in addition to the GAS regulatory motif (7114 to 7124) (Pine et al., 1994), the IRF-1 gene promoter may also contain potential retinoic acid receptor element (from 7455 to 7440) (Matikainen et al., 1996). To determine whether RA could act directly on the IRF-1 gene promoter through this site or through the GAS element, three constructions previously described (Pine et al., 1994) were used to transfect HepG2 cells: the IRF-1 upstream sequence (EcoRI ± PstI) linked to a luciferase reporter, the GAS element derived from IRF1 promoter driven luciferase reporter (WTGAS), and an analogous construct with a conservative GAS mutation (3MGAS) (Pine et al., 1994). Transfected cells were treated for 24 h with RA (1075 M), IFNa or IFNg (1000 units/ml) which were used as positive controls. Figure 6b shows the induction in response to RA, IFNa or IFNg. As reported before (Pine et al., 1994), IFNg caused a tenfold induction with the entire (EcoRI ± PstI) or WTGAS constructions, while IFNa b induction was about 2.5-fold. In the same conditions, RA caused an induction of 2.2-fold. Furthermore, WTGAS oligonucleotide responded to RA by a 2.1- fold increase, while the response of 3MGAS was reduced by 84% (Figure 6b). Although, the e€ect of RA was modest, it was reproducible and observed in the presence of anti-IFNa antibodies (data not shown). Note that HepG2 cells did not secrete IFN upon RA treatment (data not shown). Thus, the GAS (and not the putative RARE) was sucient to account for RA inducibility of IRF-1 gene. Moreover, RA-IFNa combination led to a higher augmentation of reporter activity with the entire or WTGAS constructions (Figure 6b). These results indicated that RA could increase the expression of IFN-regulated genes through direct induction of IRF-1 Figure 6 E€ect of RA on target gene promoters. (a) E€ect of RA, IFNa and their combination on PML promoter. HepG2 cells independently of IFN secretion. were seeded on six dishes plate and transfected with 3 mgof pPromPML1.44 promoter plasmid (Stadler et al., 1995) with the calcium-phosphate method. Twelve hours later medium was RA enhances the activation of STAT by IFNa and IFNg changed. Then cells were divided and treated for 24 h with RA in HL-60 and WISH cells 75 (10 M), IFNa (100 units/ml) or both. (b) E€ect of RA and IFNs on IRF-1 promoter. HepG2 cells were seeded on six dishes STAT proteins are latent transcription factors acti- plate and transfected with 3 mg of pIRF-1 (EcoRI ± PstI), vated by tyrosine phosphorylation in response to WTGAS and 3MGAS plasmids (Pine et al., 1994) with the stimulation of several transmembrane spanning recep- calcium-phosphate protocol. Twelve hours later the medium was 75 tors (Schindler and Darnell, 1995). They were changed. Then cells were divided and treated with RA (10 M), IFNa or IFNg (1000 units/ml in each case). Luciferase activity originally identi®ed as part of the IFNs signaling was measured as described in Materials and methods. Each system. It has been recently shown that RA increases experiment included triplicate transfections for each construct. the level of Stat1 in MCF-7, a breast tumor cell line The data shown are the means and standard deviations of fold resistant to growth inhibition by IFNb (Kolla et al., inductions 1996). To determine whether RA could modify biological responses in HL-60 and WISH cells through the STAT pathway, we studied the e€ect of RA on higher augmentation of reporter activity (eightfold). STAT activation and expression. First, we examined Similar results were obtained by using RA 1076 M the ability of RA to a€ect the activation of these (data not shown). Thus, RA modulation of IFN- proteins by studying their DNA-binding activity to inducible gene expression occurs at the transcriptional oligonucleotides in gel shift experiments. Since Stat1 is level. involved in both type I and type II IFN responses, a GAS element DNA-binding assay is sensitive to stimulation with either ligand. To take advantage of E€ect of RA on IRF-1 promoter this assay, extracts were prepared from HL-60, WISH IRF-1 is a that regulates the and HeLa cells and evaluated with a GAS probe. expression of IFN (Fujita et al., 1989), and IFN- Nuclear extracts from IL6-treated HepG2 cells were inducible genes (Pine et al., 1990). It has been recently used as positive control for the formation of Stat1 : 1, shown that induction of IRF-1 gene expression in RA- Stat1 : 3 and Stat3 : 3 complexes (®rst lane on the left, treated NB4 cells was also observed in the absence of Figure 7a) as previously described (Rothman et al., Retinoic acid action on interferon system L Pelicano et al 2355 a HL60 WISH HeLa

STAT 3:3

STAT 1:3

STAT 1:1

RA / – + + + – – – + + + – – – + + + – – IFNα / – – + – + – – – + – + – – – + – + – IFNγ / – – – + – + – – – + – + – – – + – +

b HL-60 WISH HeLa MW days post- kDa 0 2 4 0 2 4 0 2 4 treatment 116— Stat 1 84—

116— Stat 2

Stat 3 84—

Figure 7 Action of RA on STATs proteins. (a) E€ects of RA, IFNa or their combination on STAT activation. HL-60, WISH or HeLa cells were treated with RA (at concentrations indicated in Table 1) for 24 h, and then incubated with IFNa (100 units/ml) or IFNg (0.2 ng/ml) for 15 min. Nuclear extracts were prepared and assayed by EMSA with an IRF-1 GAS probe as previously described (Rothman et al., 1994). The positions of Stat1 : 1, Stat1 : 3 and Stat3 : 3 are indicated on the left. The ®rst lane on the left: nuclear extract from IL-6-treated HepG2 cells were used as positive control. After autoradiography the shift bands were quanti®ed by a phosphorimager. (b) E€ect of RA on the STAT protein expressions. HL-60, WISH or HeLa cells were treated with RA as indicated in Table 1 and, at the indicated times, cell extracts were used for Western blot as described in Materials and methods. The positions of STATs are indicated on the right

1994). The cells used in these studies were RA- IFN responses in HeLa cells; it may even cause a small pretreated for 24 h prior to stimulation with IFNa or decrease (510%) in STAT DNA-binding activity IFNg for 15 min. As shown in Figure 7a, treatment (Figure 7a). To verify if RA a€ects also Stat1 : 2 with RA alone never led to the activation of STAT heterodimers, part of the ISGF3 complex, electro- DNA-binding activity. However, RA pretreatment did phoretic mobility shift assays (EMSA) were performed signi®cantly augment STAT activation in response to with an ISRE probe. RA pretreatment increased by IFNa and IFNg in HL-60 and WISH cells: for 30% the DNA-binding activity of ISGF3 in WISH example, in HL-60 cells there was a 35% increase in cells and had no e€ect on HeLa cells (data not shown). Stat1 : 1 homodimer activation compared to IFNa However, we failed to detect this same complex in HL- (45% increase in Stat1 : 3 heterodimer), as quanti®ed 60 cells, as well as the endogeneous p48 protein (data by phosphorimager analysis. In RA-pretreated WISH not shown). In HL-60 cells, RA as well as IFNa seems cells, there was a 40% increase in Stat1 : 1 and a 89% to stimulate gene expression through the GAS element in Stat3 : 3 homodimers compared to IFNa (61% in via activated STATs and/or the ISRE via IRF-1. Thus, Stat1 : 3 heterodimer). In addition, these cells exhibited in HL-60 and WISH cells, transcriptional activation by a 35% increase in Stat1 : 1 compared to IFNg.In IFNa or IFNg was enhanced by RA at the level of contrast, RA pretreatment had no e€ect in increasing trans-acting factor binding. Retinoic acid action on interferon system L Pelicano et al 2356 Concurrently to the increase of the IFN-inducible RA enhances STAT levels in HL-60 and WISH cells proteins expression, RA inhibited VSV growth and cell Next, we determined whether the ability of RA to proliferation. The antiviral and the growth inhibitory increase IFN-induced STAT activation might be e€ects are potentiated by addition of IFN to both HL- attributed to changes in the level of expression of 60 and WISH cells. However, in these cells treated for these proteins. Most cells have a constitutive level of 24 h by RA alone, inhibition of VSV replication is STAT proteins (Darnell et al., 1994). Extracts prepared observed in the presence of anti-IFNa antibodies. from unstimulated or RA-stimulated HL-60, WISH Thus, RA can establish an antiviral state against and HeLa cells were examined by a sensitive STAT VSV (this paper) and also, as previously shown, speci®c immunoblotting assay (Figure 7b). These against EMCV (Bourgeade and Besancon, 1984) studies indicate that in HL-60 cells, RA increased the independently of IFN secretion. Our studies suggest level of Stat1 dramatically after 4 days treatment, but that some of the cooperative e€ects could be explained only to a least extend increased Stat2 after 2 or 4 days in part by the synthesis and secretion of IFNa but that and did not alter Stat3 levels. Note that on a longer other mechanisms exist. exposure, endogeneous Stat1, Stat2 and Stat3 were The transcription factor IRF-1 plays an important clearly visible (data not shown). In WISH cells, Stat1 role in the expression of IFN and IFN-inducible genes and Stat3 levels increased at day 2 and diminished at (Lengyel, 1993). This is achieved by binding the ISRE day 4, whereas Stat2 level were maximal at day 4. In site (Pine et al., 1990), mediating responses. HeLa cells, RA did not increase levels of any of these Overexpression of IRF-1 leads to resistance to viral STATs, a small decrease could even be observed at day infection (Pine, 1992). A recent study indicates that 4 (Figure 7b). In addition, RA at 24 h did not increase IRF-1 mRNA levels are elevated 1 h after RA the level of expression of the 48 kDa component of the treatment in HL-60 and NB4 cells, even in the absence ISGF3 complex in any of these cell lines (data not of protein synthesis (Matikainen et al., 1996) suggest- shown). Taken together, these results suggest that the ing that IRF-1 is a primary target gene for RA. In this ability of RA to increase STAT protein expression in report, we demonstrate that RA stimulates the IRF-1 HL-60 and WISH cells could contribute to the gene expression with the same eciency as IFNa.In observed potentiation of gene activation. These studies addition, the GAS motif, which mediates the induction also suggest that the e€ect of RA on STAT expression of IRF-1 in response to IFNs, seems to be the major or may vary between cell lines. unique RA responsive element in the IRF-1 promoter. Thus, RA-dependent upregulation of IRF-1 is appar- ently not mediated by the classical pathway involving Discussion nuclear retinoic acid receptor. Stat1 activation and/or induction caused by RA seems to be responsive for the Several in vitro studies have demonstrated that a increase in the transcriptional rate of IRF-1. These combined administration of RA and IFNs leads to a results are consistent with a model in which RA modulation in the expression of some IFN-induced increased IRF-1 expression, which in turn promotes genes (Ferbus et al., 1985; Lancillotti et al., 1995) and the expression of ISRE-driven genes like PML (Stadler an increased inhibition of cell proliferation (Frey et al., et al., 1995), Sp100 (GroÈ tzinger et al., 1996), PKR 1991; Mangelsdorf et al., 1994). However, the (Tanaka and Samuel, 1994) and 2'5'A synthetase mechanisms by which RA and IFNs potentiate each (Rutherford et al., 1988). other in their antiproliferative action remain to be The synergistic e€ects between RA and IFN on elucidated, and are the focus of this study. We both the induction of genes and more general determined that several mechanisms are implicated: responses is likely to be achieved by several RA directly stimulates IRF-1 gene expression through complementary mechanisms. First, in some cells, RA its GAS motif and induces IFNa synthesis; RA also increased the levels of STAT protein expression after 2 augments STAT levels and enhances IFN-induced days. However, the nature of the STAT and the level STAT activation. of increase vary substantially between cells. Thus, RA We have shown that RA increased the expression of augmentation of STAT may be a crucial step in the IFN-inducible proteins, such as 2'5'A synthetase, PKR, enhancement by RA of biological responsiveness to Sp100 and PML in HL-60 and WISH cells. Anti- IFNs. The Stat1 gene seems to be a direct target gene human IFNa but not anti-IFNb antibodies inhibited for RA since in HL-60 cells, induction of Stat1 the induction of 2'5'A synthetase activity by RA. mRNA by RA was observed in the presence of However, this inhibition was not complete, suggesting cycloheximide (data not shown). Moreover, very that IFN-independent mechanisms may participate in recently, Weihua et al. show that the promoter of the increase of IFN-inducible gene products by RA. Stat1 directly respond to RA treatment. Second, The combination of RA and 100 units/ml of IFNa had STATs can be activated directly by RA in NB4 cells a cooperative e€ect on these IFN-inducible proteins (Gianni' et al., 1997). However, RA alone failed to expression, in HL-60 and WISH cells. RA modulation induce STATs activation in MCF-7, HL-60 or WISH of IFN-induced gene expression occurred at the cells (Kolla et al., 1996 and this paper). Our data transcriptional level as demonstrated by the capacity suggest that RA stimulates the production of IFN of RA to strongly enhance PML promoter stimulation which in turn activates STAT proteins in HL-60 cells. by IFNa. Third, the synergy observed between RA and IFN Interestingly, only HL-60 cells (Breitman et al., 1980), during short treatment protocols suggest a more rapid but not WISH cells, di€erentiate upon RA treatment; mechanism. Since RA does not alter the expression or thus, the increased expression of IFN-inducible gene the activation of Jak1, Jak2 or Tyk2 proteins in products by RA is not related to cell di€erentiation. NB4 cells (Gianni' et al., 1997), it may enhance Retinoic acid action on interferon system L Pelicano et al 2357 titers, determined as the amounts of IFN required to produce 50% inhibition of the cytopathic e€ect, were expressedinrelationtothehumanIFNareference (G-023- 902-527, National Institute of Health).

Assay of VSV yield VSV was grown in the L929 cells. Treated cells were counted and infected with VSV at a multiplicity of infection of 0.1. After 18 h, cultures were frozen, thawed three times and cell debris removed by centrifugation. The supernatants were serially diluted and virus titers were measured by a plaque assay in L929 cells, and expressed as plaque-forming units/ml (PFU/ml). VSV yield in control cultures was usually 108 PFU/ml. The antiviral activity of IFN or any other compound tested was measured by reduction in the virus yield.

Figure 8 A model for activation of IFN-stimulated genes by RA Assay of 2'5'A synthetase

After treatment, cells were washed with bu€er A (140 mM NaCl, 3 mM MgCl ,35mMHEPES pH 7.5) and then lysed STAT activity through, for instance, the activation of 2 for 5 min at 08C in bu€er B (0.5% NP-40, 5 mM MgAc, MAP , which have been shown to modulate 1mM DTT, 25 mM KCl, 10% glycerol, 20 mM Tris-HCl STAT activation (Wen et al., 1995). Studies are in pH 8). Cell lysates were pelleted for 6 min in an Eppendorf progress to elucidate whether phosphorylation of Stat1 centrifuge and the cell extracts were stored at 7808C until is implicated in this phenomenon. assayed. For the 2'5'A synthetase assay, 100 mlofeachcell Taken together, these results lead us to propose a extract were tested as described (Justesen et al., 1980). One model for activation of IFN-stimulated genes by RA unit of 2'5'A synthetase activity corresponds to 1 nmol (Figure 8). One of the primary events of RA could be ATP converted/h at 378C and its speci®c activity is the induction of IRF-1 gene expression through the expressed as units/mg of protein in the cell extract. GAS motif and activated STATs. De novo synthesised IRF-1 activates directly a variety of IFN-inducible Immuno¯uorescence and Western blot analysis genes and the IFNa gene itself, leading to a positive Cells were ®xed in 4% paraformaldehyde for 10 min at feedback on IRF-1 expression. On long-term, RA also 48C, followed by 5 min in 100% methanol at 48C. increases STAT levels, which could be activated by the Immuno¯uorescence was performed with monoclonal RA-induced IFNa. Moreover, exogeneous IFN could anti-PML (1/100), rabbit anti-Sp100 (1/400) or anti-VSV enhance, on short-term or long-term, the activation of (1/2000) antibodies and revealed with FITC or Texas red STAT. conjugated secondary antibodies. For Western blot analysis, cells were scraped in PBS, centrifuged and pellet was resuspended in lysis bu€er (125 mM Materials and methods Tris pH 7, 1% SDS, 10% glycerol, 0.75 mM PMSF). Each sample was sonicated to reduce viscosity. Western blot analysis was performed according to standard procedures Reagents (Harlow and Lane, 1988). Twenty micrograms of whole cell Adenosine 5'[a32P]triphosphate (410 Ci/mmol) was pur- extract were analysed using rabbit anti-PML (1/2000), anti- chased from Amersham, Polyriboinosinic-polyribocy- Sp100 (1/400) or monoclonal anti-PKR (1/5000) (from tidylic acid (Poly(I)-poly(C)) was from Pharmacia and RiboGene, Hayward, CA) antibodies and revealed by all-trans-retinoic acid (RA) from Sigma. DMEM, RPMI chemoluminescence (ECL, Amersham). The polyclonal anti- and fetal bovine serum were supplied by Gibco, France. carboxy terminal Stat1 and Stat2 antibodies were used as described (Gupta et al., 1996). Stat3 (C20) was purchased from Santa Cruz. Rabbit anti-actin antibodies were from Cell culture Sigma. Cells were grown at 378C: HL-60 human promyelocytic leukemia cells in RPMI medium, bovine MDBK, human DNA constructions WISH, HeLa and HepG2 cells and mouse L929 cells, in DMEM supplemented with 10% heat inactivated fetal calf The pPromPML1.44-Luc plasmid was described elsewhere serum. (Stadler et al., 1995). Brie¯y, a genomic 1.44 kb fragment, containing the 5' region upstream of the ®rst ATG of PML gene, was subcloned in the Bluescript II SK+ phagemid Human IFNs and anti-human IFN antibodies according to standard procedures (Ausubel et al., 1994). A Recombinant IFNa2(108 international units/mg) was from full-length, intronless luciferase construct, LD5' (de Wet et Schering (USA). Anti-human IFNa (G-030-501-553) and al., 1987) was then introduced at a downstream NcoI site IFNb (G-028-501-568) antibodies, having 7.5 105 and 1.2 to generate above mentioned construction. The IRF-1 104 neutralizing units/ml respectively, were obtained from promoter containing plasmid was previously described the National Institute of Health. (Pine et al., 1994). EcoRI ± PstI fragment which contains the entire IRF-1 promoter was subcloned into a HindIII site of the pZluc promoterless luciferase reporter vector. Determination of IFN titers WTGAS is an oligonucleotide spanning the IRF-1 GAS RA-induced IFN was titrated on MDBK and HeLa cells and 3MGAS is a IRF-1 GAS mutant obtained by ligation challenged with vesicular stomatitis virus (VSV). IFN of a synthetized oligonucleotide containing the AA to CC Retinoic acid action on interferon system L Pelicano et al 2358 mutation at positions 7115 and 7116 into the HindIII site prepared (Gupta et al., 1996). Each of the extracts was of ptkLuc as described in (Pine et al., 1994). examined by EMSA with either a GAS from the IRF-1 gene (gatcGATTTCCCCGAAAT) or an ISRE probe from the ISG15 gene promoter (gatcTTTCACTTTCTAGTTT- Transient transfection CACTTT CCCTTT) as described (Gupta et al., 1996). HepG2 cells (105/ml) were seeded in six dishes plates and After autoradiography the shift bands were quanti®ed with transfected with 3 mg of plasmid DNA using the calcium a Biorad GS250 phosphorimager. phosphate method as described previously (Stadler et al., 1995). To avoid di€erences in transfection eciency, the transfected cells were split. One half was treated as described in the Figure 6 legend, the other half served as control. Twenty hours later, cells were harvested and Acknowledgements relative luciferase units were measured with a luminometer. We are grateful to H de The for helpful discussions and encouragements. The help of B Boursin for the artwork were highly appreciated. We acknowledge MC Guillemin Electrophoretic mobility shift assays for polyclonal anti-Sp100 antibodies, D Blondel for anti- When HeLa or WISH cells reached 70% con¯uency or in VSV sera and M Stadler for PML promoter construct. We the case of HL-60 cells 56105 to 16106 cells/ml, RA was thank C Chopin for technical assistance, M Koken and A added for 24 h to a ®nal concentration of 1076 M (HL-60 Saib for critical reading of the manuscript. This work was cells) or 1075 M for WISH and HeLa cells. Then the RA supported by grants from Ligue contre le cancer and by treated or untreated cells were stimulated with IFNa (100 CNRS. L Pelicano was supported by Ligue contre le cancer units/ml) or IFNg (0.2 ng/ml) for 15 min. Cells were fellowship and grants of the PMP foundation and F Li by harvested and washed with cold PBS, and nuclear extracts Amgen.

References

AusubelFM,BrentR,KingstonRE,MooreDD,Seidman Justesen J, Ferbus D and Thang MN. (1980). Nucleic Acids JG, Smith JA and Struhl KI. (1994). Current Protocols in Res., 8, 3073 ± 3085. Molecular Biology. Wiley (Interscience), New York. Kanno Y, Kozak CA, Schindler C, Driggers PH, Ennist DL, Bourgeade MF and Besancon F. (1984). Cancer Res., 44, Gleason SL, Darnell Jr J and Ozato K. (1993). Mol. Cell. 5355 ± 5360. Biol., 13, 3951 ± 3963. Breitman TR, Selonick SE and Collins SJ. (1980). Proc. Natl. Khan KD, Shuai K, Lindwall G, Maher SE, Darnell Jr J and Acad. Sci. USA, 77, 2936 ± 2940. Bothwell AL. (1993). Proc. Natl. Acad. Sci. USA, 90, Chebath J, Benech P, Revel M and Vigneron M. (1987). 6806 ± 6810. Nature, 330, 587 ± 588. Koken MHM, Linares-Cruz G, Quignon F, Viron A, Chelbi- Chelbi-Alix MK, Brouard A, Boissard C, Pelaprat D, Alix MK, Sobczak-The pot J, Juhlin L, Degos L, Calvo F Rostene W and Thang MN. (1994). J. Cell. Physiol., and de The H. (1995). Oncogene, 10, 1315 ± 1324. 158, 47 ± 54. Koken MHM, Puvion-Dutilleul F, Guillemin MC, Viron A, Chelbi-Alix MK, Pelicano L, Quignon F, Koken MHM and Linares-Cruz G, Stuurman N, de Jong L, Szostecki C, de The H. (1996). Tumor Biology, Regulation of cell growth Calvo F, Chomienne C, Degos L, Puvion E and de The H. di€erentiation and genetics in cancer. Tsiftsoglou AS, (1994). EMBO J., 13, 1073 ± 1083. Sartorelli AC, Housman DE and Dexter MT (eds). Kolla V, Lindner DJ, Weihua X, Borden EC and Springer-Verlag: Berlin Heidelberg, pp. 17 ± 27. Kalvakolanu DV. (1996). J. Biol. Chem., 271, 10508 ± Chelbi-Alix MK, Pelicano L, Quignon F, Koken MHM, 10514. Venturini L, Stadler M, Pavlovic J, Degos L and de The H. Korioth F, Gie€ers C, Maul GG and Frey J. (1995). J. Cell (1995). Leukemia, 9, 2027 ± 2033. Biol., 130, 1±13. Chelbi-Alix MK and Thang MN. (1986). Biochem. Biophys. Koromilas AE, Roy S, Barber G, Katze MG and Sonenberg Res. Commun., 141, 1042 ± 1050. N. (1992). Science, 257, 1685 ± 1689. Darnell Jr J, Kerr IM and Stark GR. (1994). Science, 264, Lancillotti F, Giandomenico V, A€abris E, Fiorucci G, 1415 ± 1421. Romeo G and Rossi GB. (1995). Cancer Res., 55, 3158 ± de Wet JR, Wood KV, DeLuca M, Helinski DR and 3164. Subramani S. (1987). Mol. Cell Biol., 7, 725 ± 737. Lengyel P. (1993). Proc. Natl. Acad. Sci. USA, 90, 5893 ± Ferbus D, Testa U, Titeux M, Louache F and Thang MN. 5895. (1985). Mol. Cell. Biochem., 67, 125 ± 133. Liu J-H, Mu Z-M and Chang K-S. (1995). J. Exp. Med., 181, Frey JR, Peck R and Bollag W. (1991). Cancer Lett., 57, 1965 ± 1973. 223 ± 227. Mangelsdorf DJ, Umesomo K and Evans RM. (1994). The Fu XY, Schindler C, Improta T, Aebersold R and Darnell Jr Retinoids: Biology, Chemistry and Medicine, Sporn MB, J. (1992). Proc.Natl.Acad.Sci.USA,89, 7840 ± 7843. Roberts AB and Goodman DS (eds), Raven Press, Ltd: Fujita T, Kimura Y, Miyamoto M, Barsoumian EL and New York, pp. 319 ± 349. Taniguchi T. (1989). Nature, 337, 270 ± 272. Matikainen S, Ronni T, Hurme M, Pine R and Julkunen I. Gianni' M, Terao M, Fortino I, LiCalzi M, Viggiano V, (1996). Blood, 88, 114 ± 123. Barbui T, Rambaldi A and Garattini E. (1997). Blood, 3, Meurs E, Galabru J, Barber G, Katze MG and Hovanessian 1001 ± 1012. AG. (1993). Proc. Natl. Acad. Sci. USA, 90, 232 ± 236. GroÈ tzinger T, Jensen K and Will H. (1996). J. Biol. Chem., Meurs E, Watanabe Y, Kadereit S, Barber G, Katze M, 271, 25253 ± 25260. Chong K, Williams B and Hovanessian A. (1992). J. Guldner H, Szostecki C, GroÈ tzinger T and Will H. (1992). J. Virol., 66, 5805 ± 5814. Immunol., 149, 4067 ± 4073. Mu ZM, Chin KV, Liu JH, Lozano G and Chang KS. (1994). Gupta S, Yan H, Wong LH, Ralph S, Krolewski J and Mol. Cell. Biol., 14, 6858 ± 6867. Schindler C. (1996). EMBO J., 15, 1075 ± 1084. Pestka S, Langer JA, Zoon KC and Samuel CE. (1987). Harlow E and Lane D. (1988). (eds). Cold Spring Harbor Annu. Rev. Biochem., 56, 727 ± 777. Laboratory: NY. Pine R. (1992). J. Virol., 66, 4470 ± 4478. Retinoic acid action on interferon system L Pelicano et al 2359 Pine R, Canova A and Schindler C. (1994). EMBO J., 13, Sen GC and Ransoho€ RM. (1993). Adv. Virus Res., 42, 57 ± 158 ± 167. 102. Pine R, Decker T, Kessler DS, Levy DE and Darnell Jr J. Stadler M, Chelbi-Alix MK, Koken MHM, Venturini L, Lee (1990). Mol. Cell. Biol., 10, 2448 ± 2457. C, SaõÈ b A, Quignon F, Pelicano L, Guillemin M-C, Romeo G, Fiorucci G and Rossi GB. (1989). Trends Genet., Schindler C and de The H. (1995). Oncogene, 11, 2565 ± 5, 19 ± 24. 2573. Rothman P, Kreider B, Azam M, Levy D, Wegenka U, Eilers Staeheli P, Pitossi F and Pavlovic J. (1993). Trends Cell. A, Decker T, Horn F, Kashleva H, Ihle J et al. (1994). Biol., 3, 268 ± 272. Immunity, 1, 457 ± 468. Tanaka H and Samuel CE. (1994). Proc. Natl. Acad. Sci. Rutherford MN, Hannigan GE and Williams BR. (1988). USA, 91, 7995 ± 7999. EMBO J., 7, 751 ± 759. Weihua X, Kolla V and Kalvakolanu DV. (1997). J. Biol. Schindler C. (1995). Receptor, 5, 51 ± 62. Chem., 272, 9742 ± 9748. Schindler C and Darnell Jr J. (1995). Annu. Rev. Biochem., Wen Z, Zhong Z and Darnell Jr J. (1995). Cell, 82, 241 ± 250. 64, 621 ± 651. SchindlerC,FuXY,ImprotaT,AebersoldRandDarnellJr J. (1992). Proc.Natl.Acad.Sci.USA,89, 7836 ± 7839.