Differential splicing generates Tvl-1/RFXANK isoforms with different functions.

Earlier studies have shown that Tvl-1 gives rise to at least two differentially spliced mRNAs, one of which (Tvl-S) encodes a protein that lacks amino acids 91-112. DNA binding of RFX complexes assembled in the presence of Tvl-S is impaired. As a result, Tvl-S does not support the expression of Class II major histocompatibility complex (MHC) genes. Here, we show that the reason Tvl-S is inactive as a transcriptional regulator of Class II MHC genes is that the RFX complexes assembled in the presence of Tvl-S are unstable. Additionally, we show that interferon-gamma, which induces Class II MHC gene expression in 293 cells, promotes a shift in the splicing pattern of RFXANK/Tvl-1 toward the transcriptionally active Tvl-L isoform, suggesting that differential splicing of Tvl-1 is a signal-regulated process. Finally, we show that Tvl-1 regulates the expression of non-MHC genes. One such gene encodes the ephrin receptor EphA3. Since both Tvl-L and Tvl-S are identical in their ability to induce the expression of EphA3, we conclude that Tvl-1 regulates the expression of non-MHC genes by RFX-independent mechanisms.

Peptides derived from antigens processed by antigenpresenting cells are presented to lymphocytes in association with the components of the major histocompatibility complex (MHC). 1 CD4 ϩ helper T-cells recognize antigens presented in association with Class II MHC molecules. Such molecules, therefore, are essential for antigen presentation. Class II MHC genes are expressed in antigen presenting cells such as dendritic cells, macrophages, and B-cells (1)(2)(3). Failure of the antigen-presenting cells to express Class II MHC genes is a central feature of a severe immunodeficiency disorder, the bare lymphocyte syndrome (BLS) (4 -9). Mutations responsible for this syndrome belong to four complementation groups (4, 10 -13). Complementation group A is caused by mutations in CIITA (12% of all cases) (14 -17), whereas complementation groups B, C, and D are caused by mutations in RFXANK/ Tvl-1 (62% of all cases), RFX5 (12% of all cases), and RFXAP (14% of all cases), respectively (18 -23). Expression of Class II MHC depends on transcription factors encoded by the genes that define the four complementation groups of BLS.
CIITA, a transcriptional co-activator that binds the acetyltransferase p300/CBP and stimulates transcription by promoting histone acetylation (24 -27) is responsible for the restricted expression of Class II MHC genes. This is underscored by observations showing that the induction of CIITA by IFN-␥ in cells that do not normally express CIITA is sufficient to induce expression of Class II MHC genes (28 -31). RFX5, RFXAP, and RFXANK/Tvl-1 form a complex that binds a specific site in the Class II MHC promoter (21,(32)(33)(34). DNA binding is mediated by RFX5 (35)(36)(37). However, it has been suggested that RFXAP and RFXANK/Tvl-1 may also contact DNA (38). The RFX complex also interacts with CIITA and other transcription factors that bind the Class II MHC promoter (6,8).
Molecular interactions involved in the assembly of the RFX complex are somewhat controversial. Published studies suggest that RFXAP binds both RFX5 and RFXANK/Tvl-1 and that it serves as a bridge between these two proteins (39,40).
Our studies indicate that similar to RFXAP, Tvl-1 also interacts with the other two proteins in the complex, suggesting that complex formation depends on three way interactions between all the proteins in the complex. The finding that Tvl-1 contributes to the assembly of the RFX complex is in agreement with studies suggesting that RFXANK/Tvl-1 functions as a scaffold that controls the assembly of other multiprotein complexes (41).
Earlier studies have shown that RFXANK/Tvl-1 gives rise to two differentially spliced mRNAs (19), which, in this report, are referred to as Tvl-L (L for long) and Tvl-S (S for short). The protein encoded by Tvl-S lacks amino acids 91-112 and fails to induce expression of Class II MHC genes (39). In this report, we present evidence that this failure is caused by the instability of the RFX complex assembled in the presence of Tvl-S. We also show that IFN-␥, which induces CIITA and Class II MHC expression in 293 cells, promotes a shift in the splicing pattern of RFXANK/Tvl-1 toward the Tvl-L form. Finally, we show that Tvl-1 also regulates the expression of non-MHC genes and that Tvl-L and Tvl-S are identical in their ability to induce the expression of the ephrin receptor EphA3, which is encoded by one of these genes. The fact that both Tvl-L and Tvl-S induce the expression of EphA3 suggests that the role of Tvl-1 in the expression of this, and perhaps other non-MHC genes, is RFX-independent.

Plasmids
FLAG epitope-tagged Tvl-L (FLAG-Tvl-L) and Tvl-S (FLAG-Tvl-S) constructs were generated by inserting Tvl-L and Tvl-S in-frame with the FLAG epitope tag into the pcDNA3 or pREP4 expression vectors (Invitrogen). RFXAP-HA and GST-RFX5 were cloned in pcDNA3. GST-Tvl-L and GST-Tvl-S were cloned in pGEX5X3 (Amersham Biosciences).

Cell Culture
BLS-1 is a Class II MHC-deficient B-lymphoblastoid cell line derived from a patient with a mutation in RFXANK/Tvl-1. The BLS-1 mutation consists of a 58-bp deletion that removed the last 23 nucleotides of exon 6 and the adjacent splice donor site (18). BLS-1 cells transfected with pREP4-based constructs were selected with hygromycin (200 g/ml). BLS-1 and its derivative cell lines were maintained in RPMI 1640 medium supplemented with 15% heat-inactivated fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 g/ml). 293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin, and streptomycin. 293 cells were treated with recombinant human IFN-␥ (500 units/ml) (Endogen) for 24 h.

Preparation of RNA and RT-PCR
Total RNA was prepared using the RNeasy mini kit (Qiagen). Quantitative RT-PCR was carried out using the RETROscript kit (Ambion Inc.). To check the expression of Tvl-L and Tvl-S, we used the primer pair: 5Ј-CCG GCA GCG AGG GAA CGA CGT GT-3Ј and 5Ј-CGC TCG TCT GGC TTG TTG ACG AG-3Ј from exons 4 and 6 of Tvl-1, respectively. Primer pairs for HLA-DRA (5Ј-CGG GAT CCA TGG CCA TAA GTG GAG TC-3Ј and 5Ј-CGG AAT TCT TAC AGA GGC CCC CTG CGT T-3Ј) and for ␤-actin (5Ј-CGG GAT CCA TGG ATG ATG ATA TCG CC-3Ј and 5Ј-CGG AAT TCC TAG AAG CAT TTG CGG TG-3Ј) were used to amplify these genes as controls. To check the expression of EphA3, we used the primer pair: 5Ј-GCT GAG AAC AAA CTG GGT CCC CAG-3Ј and 5Ј-GAG GCG GGC ACT TAG CAC ACT TC-3Ј.

In Vitro Transcription and Translation
pcDNA3-based constructs were transcribed and translated in vitro using the TnT T7 quick coupled transcription/translation system (Promega).

In Vitro Binding Assays
Escherichia coli BL21(DE3) transformed with GST-Tvl-L and GST-Tvl-S constructs were induced with 1 mM IPTG for 4 h. Cells were sonicated in GST lysis and binding buffer (25 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 1% Triton X-100, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride). GST fusion proteins were bound to glutathione-Sepharose 4B beads (Amersham Biosciences) for 1 h at 4°C. Following washing with a buffer containing 150 mM NaCl, 50 mM Tris, pH 8.0, 1% Nonidet P-40, and 1 mM phenylmethylsulfonyl fluoride, 20 g of bead-bound GST-Tvl-L, GST-Tvl-S, or GST were incubated with 10 l of in vitro translated RFX5 and/or RFXAP at 30°C for 1 h. Bound proteins were eluted by boiling the beads in the electrophoresis sample buffer.
Immunofluorescence BLS-1 cells stably transfected with either the empty pREP4 vector, pREP4-FLAG-Tvl-L, or pREP4-FLAG-Tvl-S were mounted on glass slides using a cytospin centrifuge. Cells were fixed on the slides using ice-cold acetone. Following blocking with 5% goat serum for 30 min, slides were incubated with polyclonal anti-Tvl-1 primary antibody (1: 200 dilution in the blocking buffer) at 4°C overnight. Slides were washed with phosphate-buffered saline and incubated with FITC-conjugated anti-rabbit IgG (1:150 dilution) for 1 h. Slides were washed again with phosphate-buffered saline, and they were analyzed by fluorescent confocal microscopy. Non-transfected Raji cells stained with the same antibody were used as a control.

Binding of Tvl-L-and Tvl-S-containing Complexes to DNA
Electrophoretic Mobility Shift Assay (EMSA)-EMSA was carried out following standard procedures (40). The X1-box oligonucleotide used as the EMSA probe contained sequences that extend from Ϫ144 to Ϫ69 base pairs of the DRA promoter.
Binding of Tvl-L and Tvl-S to Biotinylated DNA-5Ј-Biotinylated X1-box oligonucleotides with 6C spacer arms were custom synthesized by Midland Laboratories, and they were attached to streptavidine-coated magnetic beads (streptavidine M280, Dynal laboratories) according to the manufacturer's protocol. Protein binding to bead-associated DNA was carried out following standard procedures (39,40).

Complex Assembly and Complex Stability Assays
10 l each of in vitro translated RFX5 and RFXAP were incubated on ice with bacterially produced GST-Tvl-L or GST-Tvl-S (20 g each) bound to glutathione-Sepharose beads for the indicated time points (Fig. 6). Alternatively, nearly equimolar concentrations of in vitro translated RFX5, RFXAP, FLAG-Tvl-L, or FLAG-Tvl-S were incubated at 4°C for 1 h. Protein complexes were pulled down by binding to either glutathione-Sepharose beads or anti-FLAG M2 agarose beads (Sigma). All the in vitro translated proteins were metabolically labeled with [ 35 S]methionine. Bead-bound proteins were analyzed by SDS-PAGE and autoradiography. To determine complex stability, complexes were assembled using 50 l each of in vitro translated GST-RFX5 and RFXAP and in vitro translated, [ 35 S]methionine-labeled Tvl-L or Tvl-S. Complexes were pulled down by binding to glutathione-Sepharose beads. The washed beads were incubated at 4°C in 50 l of binding buffer (40). Supernatants were analyzed by SDS-PAGE and autoradiography. In parallel experiments, the amounts of Tvl-L and Tvl-S that remained bound to the beads at the 0 h and the 6 h time points were also examined. To further analyze the stability of the Tvl-L-and Tvl-S-containing complexes, we examined whether Tvl-L can compete Tvl-S and whether Tvl-S can compete Tvl-L out of the assembled complexes. To this end, 50 l each of in vitro translated RFX5 and RFXAP were first incubated with 50 l of in vitro translated FLAG-Tvl-L or FLAG-Tvl-S at room temperature. One hour later, 25 g of bacterially produced GST-Tvl-S or GST-Tvl-L were added to the complexes and they were incubated at 4°C for an additional hour. The final protein complexes were pulled down with anti-FLAG M2 agarose beads, and they were analyzed by SDS-PAGE and autoradiography.

DNA Microarray
DNA microarray experiments were carried out using Affimatrix chips (U133A-Affimatrix Inc.). Specifically, cRNAs from BLS-1 cells stably transfected with pREP4, pREP4-Tvl-L, or pREP4-Tvl-S constructs were hybridized to a microarray of 12,000 clones, which includes both known genes and expressed sequence tags. Comparison of gene expression in the three cell lines allowed us to identify genes whose expression was altered in pREP4-Tvl-L and pREP4-Tvl-S, relative to the pREP4-transfected BLS-1 cells.

RESULTS
Earlier studies have shown that Tvl-1 gives rise to two differentially spliced mRNAs (19), Tvl-L and Tvl-S (Fig. 1A). Exon 5, which is spliced out in Tvl-S, encodes amino acids 91-112. Tvl-S, the protein encoded by Tvl-S, failed to induce expression of Class II MHC genes. In this report we addressed the molecular defect responsible for the inability of Tvl-S to support Class II MHC expression. In addition, we examined whether differential splicing of Tvl-1 in cells that do not normally express Class II MHC genes is regulated by Class II MHCinducing signals, and we addressed the function of Tvl-S.
Earlier studies have shown that, although Tvl-L restores Class II MHC expression in lymphoblastoid cell lines derived from bare lymphocyte syndrome patients, Tvl-S does not (39). To confirm these results, we stably expressed Tvl-L or Tvl-S in one of these cell lines (BLS1). The cells expressing the two Tvl-1 isoforms were stained with FITC-conjugated anti-HLA-DP antibody, and they were analyzed by flow cytometry (Fig. 1B).
Subcellular Localization of Tvl-L and Tvl-S-The inability of Tvl-S to support Class II MHC expression may result from its inability to translocate into the nucleus. To evaluate this possibility, we compared the subcellular localization of Tvl-L and Tvl-S by immunofluorescence staining and confocal microscopy. Previous studies have shown that Tvl-1 expressed in transiently transfected 293 cells can be detected in both the nucleus and the cytoplasm (42). The significance of this finding, however, was questioned because it is possible that the subcellular localization of a given protein may be artificially altered if Tvl-1⌬91-112 Promotes the Assembly of an Unstable RFX Complex the protein is expressed at high levels from a transiently transfected construct.
Here, we examined the subcellular localization of Tvl-L expressed at physiological or near physiological levels. In addition, we compared the subcellular localization of Tvl-L with that of similarly expressed Tvl-S. Since transcription of the endogenous gene gives rise to mRNAs that encode both proteins, we stably expressed Tvl-L and Tvl-S constructs separately in Tvl-1-null BLS-1 cells. The level of expression of these proteins in the stably transfected cells was similar to the level of expression of endogenous Tvl-1 in Raji cells. Immunofluorescence staining with a polyclonal anti-Tvl-1 antibody, followed by confocal microscopy, confirmed that Tvl-L can be detected in both the nucleus and the cytoplasm and showed that the highest concentration of the protein was in the perinuclear region (Fig. 2). The same analysis showed that Tvl-L and Tvl-S exhibit similar subcellular distributions. Therefore, the subcellular localization of Tvl-S does not explain its inability to induce Class II MHC expression.
Both Isoforms of Tvl-1 Interact with RFX5 and RFXAP in Vitro in the Presence as Well as the Absence of the X1-box Oligonucleotides-Previous studies have shown that complexes containing Tvl-1, RFX5, and RFXAP can be pulled down from nuclear extracts of Raji cells with an RFX5-specific antiserum (19). The inability of Tvl-S to support the expression of Class II MHC genes could be caused by its inability to bind the other components of the RFX complex. To address this question, we addressed its binding to RFX5 and RFXAP in vitro. Earlier studies had investigated the binding of Tvl-L but not Tvl-S to these proteins. Some of these earlier studies have shown that bacterially produced GST-Tvl-L interacts with in vitro translated RFX5 and RFXAP (39). However, other studies had found that Tvl-L interacts only with RFXAP and not with RFX5 (40). Here, we revisited the interactions between Tvl-L and RFX5 or RFXAP, and we examined whether Tvl-L and Tvl-S differ in their ability to interact with these molecules. To this end, purified GST-Tvl-L and GST-Tvl-S produced in E. coli were incubated with in vitro translated, [ 35 S]methionine-labeled RFX5 and RFXAP separately or in combination. The in vitro assembled complexes were pulled down using glutathione-Sepharose beads, and they were analyzed by SDS-PAGE. The results showed that both Tvl-L and Tvl-S bind RFX5 and RFXAP, either alone or in combination (Fig. 3A).
To determine whether X1-box oligonucleotides affect the assembly of the Tvl-L-or Tvl-S-containing complexes, we repeated the previous experiment in the presence or absence of 1 g of unlabeled oligonucleotides. The results confirmed that both Tvl-L and Tvl-S support the assembly of the RFX complex in vitro and that the X1-box oligonucleotides do not affect the assembly (Fig. 3B).
DNA Binding of Tvl-L-or Tvl-S-containing RFX Complexes-Earlier studies have shown that RFX complexes assembled in the presence of Tvl-L bind DNA, whereas complexes assembled in the presence of Tvl-S do not (39). This was demonstrated using EMSAs. Using a similar approach, we first sought to confirm these data. The results showed that Tvl-Scontaining complexes exhibit detectable but significantly weaker binding to X1-box oligonucleotides than Tvl-L-containing complexes (Fig. 4A). Given that both Tvl-L and Tvl-S support the assembly of RFX complexes (Fig. 3, A and B), it is not clear why the complexes assembled in the presence of Tvl-S do not efficiently bind DNA. One possibility is that amino acids 91-112, which are missing from Tvl-S, may define a DNAbinding motif. Alternatively, complexes assembled in the presence of Tvl-S may exhibit an altered conformation. In either case, their association with DNA may be weak and unstable. Finally, the protein complexes assembled in the presence of Tvl-S may themselves be unstable.
To confirm that Tvl-S-containing complexes bind DNA, we examined their association with DNA under conditions that do not favor the dissociation of protein-DNA complexes following binding. To this end, 5Ј-biotinylated X1-box oligonucleotides were incubated with RFX complexes assembled in the presence of different combinations of RFX5, RFXAP, Tvl-L, and Tvl-S. The assembled protein-DNA complexes were precipitated using streptavidine-coated magnetic beads. The beads were then stringently washed and the bead-bound proteins were analyzed by SDS-PAGE and autoradiography. The results showed that all the RFX5-containing complexes, including those assembled in the presence of Tvl-S, bind DNA (Fig. 4B). Moreover, RFX5 alone in the absence of any co-factors also binds DNA. The different sensitivities of this assay and EMSA suggest that protein-DNA complexes may dissociate in EMSA due to the shearing stress applied to the complexes as they traverse the gel during electrophoresis.
Given that EMSA assays clearly show that Tvl-S-containing complexes bind DNA with significantly reduced efficiency, we hypothesized that the binding of these complexes to DNA may be unstable. To test this hypothesis, we examined whether Tvl-S and Tvl-L compete with each other when incubated with Tvl-L-or Tvl-S-containing protein-DNA complexes. Our results showed that although excess Tvl-L displaces Tvl-S, excess Tvl-S does not displace Tvl-L (Fig. 4C). This finding suggests that the inability of the Tvl-S-containing RFX complexes to bind DNA efficiently may result from the instability of these complexes.
In Vitro Assembly of Tvl-S-or Tvl-L-containing RFX Complexes-The experiment in Fig. 3 revealed that both Tvl-L and Tvl-S support the assembly of RFX complexes. However, complex assembly in this experiment was carried out in the presence of a large excess of Tvl-1. The very high concentration of Tvl-L and Tvl-S may obscure the efficiency with which these two proteins promote complex assembly in vitro. To address this hypothesis, we examined the assembly of Tvl-L-and Tvl-S-containing RFX complexes using in vitro translated proteins at nearly equimolar amounts. The results showed that Tvl-L is more efficient in promoting the assembly of the complex than Tvl-S (Fig. 5).
Stability of Tvl-S-or Tvl-L-containing RFX Complexes-The lower efficiency of RFX complex assembly in the presence of Tvl-S could result from a lower rate of assembly or from a lower stability of the assembled complexes or both. To evaluate these possibilities, we incubated in vitro translated, [ 35 S]methioninelabeled RFX5 and RFXAP with excess of bacterially expressed GST-Tvl-L or GST-Tvl-S bound to glutathione-Sepharose beads. The complexes assembled in the presence of the two Tvl-1 isoforms were harvested at 5, 10, 15, 20, and 30 min from the start of the incubation, and they were analyzed by SDS-PAGE. The relative abundance of RFX5 and RFXAP in these complexes was determined by autoradiography and exposure to a phosphorimager screen followed by analysis with the Image Quant software. The results showed that the assembly of Tvl-Sand Tvl-L-containing complexes proceeds with similar efficiency in the first 5 min of incubation. However, although the assembly of Tvl-S-containing complexes reaches a plateau at 5 min, the assembly of Tvl-L-containing complexes continues up to, and perhaps beyond, the 30-min time point (Fig. 6). These data suggest that the assembly of Tvl-L-and Tvl-S-containing Tvl-1⌬91-112 Promotes the Assembly of an Unstable RFX Complex complexes proceeds at a similar rate. However, the assembly of Tvl-S-containing complexes reaches an equilibrium very rapidly. Given that the assembly is a reversible process, these data strongly support the hypothesis that the Tvl-S-containing complexes are unstable.
To evaluate the stability of Tvl-L-and Tvl-S-containing complexes, we incubated in vitro translated GST-RFX5 and RFXAP with in vitro translated, [ 35 S]methionine-labeled Tvl-L or Tvl-S. The assembled complexes were pulled down by binding to glutathione-Sepharose beads. Following extensive washing, the beads were incubated in the binding buffer at 4°C. Supernatants were collected at the indicated time points, and they were analyzed by SDS-PAGE and autoradiography. The results showed that complexes assembled in the presence of Tvl-S release their components more rapidly than complexes assembled in the presence of Tvl-L (Fig. 7, A1 and A2). In agreement with this data, a higher amount of Tvl-L than Tvl-S remained bound to the complex at 6 h from the start of the incubation (Fig. 7, B1 and B2).
To confirm the relative instability of the Tvl-S-containing complexes, we incubated complexes assembled in the presence of in vitro translated FLAG-Tvl-L or FLAG-Tvl-S with a 50-fold excess of bacterially expressed GST-Tvl-S or GST-Tvl-L, respectively. The composition of the complexes, before and after incubation with the bacterially expressed Tvl-1 isoforms, was determined by SDS-PAGE and autoradiography of the proteins immunoprecipitated with the anti-FLAG M2 antibody. The results showed that although excess GST-Tvl-L almost completely displaces Tvl-S from the complex, GST-Tvl-S does not displace Tvl-L (Fig. 7C). This confirmed that Tvl-S-containing complexes are indeed less stable than those containing Tvl-L. (28 -31). Here, we examined whether, in addition to its effect on CIITA, IFN-␥ treatment changes the expression of the two differentially spliced forms of Tvl-1. To measure the expression in untreated and IFN-␥-treated (500 units/ml for 24 h) 293 cells, we employed RT-PCR using primers from exons 4 and 6, which flank the differentially spliced exon 5. The results showed that while both isoforms were abundant in untreated 293 cells, treatment with IFN-␥ induced a shift toward Tvl-L, which encodes the transcriptionally active isoform of Tvl-1. Simultaneous RT-PCR amplification of HLA-DRA confirmed that IFN-␥ was effective in inducing Class II MHC expression (Fig. 8). We conclude that differential splicing of the Tvl-1 mRNA is a process that is regulated by cytokine signals. Moreover, since only Tvl-L participates in the assembly of stable transcriptionally active RFX complexes, we propose that the IFN-␥-induced increase in the level of Tvl-L may contribute to the induction of Class II MHC genes.

Interferon-␥ Enhances the Relative Expression of Tvl-L by Promoting a Shift in the Splicing Pattern of the Tvl-1 mRNA-IFN-␥ up-regulates the expression of CIITA in nonhematopoietic cells and induces Class II MHC expression
Tvl-L and Tvl-S Regulate Genes Other than Those of the Class II MHC Complex-The preceding results raised the question whether Tvl-S is biologically active or whether it is a non-functional by-product of an inefficient splicing process. To bead-associated DNA were analyzed by SDS-PAGE and autoradiography. Beads without tethered DNA were used as a control (B2). The input lanes contain one-fifth of the amount of the individual proteins used in the binding assay. C, competition between Tvl-L and Tvl-S in RFX complex assembly and DNA binding. Following assembly, Tvl-Lor Tvl-S-containing RFX complexes were incubated with labeled X1-box oligonucleotides. 5-Fold excess of in vitro translated competitor (Tvl-L in the case of Tvl-S-containing complexes and Tvl-S in the case of Tvl-L-containing complexes) was added (indicated by x), and the samples were incubated for 4 additional hours. The DNA bound protein complexes were resolved by EMSA. Tvl-1⌬91-112 Promotes the Assembly of an Unstable RFX Complex address this question, we first examined whether Tvl-S functions as a dominant negative variant that interferes with the function of Tvl-L. To this end, we carried out luciferase reporter assays in BLS-1 cells transfected with an HLA-DRA promoterluciferase reporter construct, in combination with Tvl-L and Tvl-S constructs. In parallel experiments, we transiently transfected a Tvl-S construct into BLS-1 cells stably expressing Tvl-L, and we stained the transfected cells with a phycoerythrin-conjugated anti-HLA-DRA monoclonal antibody. Both experiments showed that Tvl-S did not inhibit the activity of the Class II MHC promoter (data not shown). This result was not unexpected, however, given the fact that Tvl-S cannot replace Tvl-L in RFX complexes. Next, we examined whether Tvl-S, which differs from Tvl-L in that it does not support the expression of Class II MHC, regulates the expression of other cellular genes. To address this question, we carried out microarray experiments that compared gene expression between BLS-1 cells stably transfected with the mammalian expression vector pREP4 and BLS-1 cells stably transfected with pREP4-Tvl-L or pREP4-Tvl-S constructs. The results confirmed that Class II MHC genes are induced only by Tvl-L. In addition, they showed that several other genes are up-regulated by Tvl-L and/or Tvl-S. One of these genes encodes the ephrin receptor EphA3 (Fig. 9), a receptor tyrosine kinase that has been reported to have a role in early tissue patterning events during embryogenesis and that is overexpressed in some leukemias and solid tumors (43)(44)(45)(46). These results show that Tvl-S, an isoform of Tvl-1 that is encoded by a differentially spliced mRNA, is a functionally active transcription factor that regulates the expression of genes other than those of the MHC Class II complex.

DISCUSSION
It has been shown previously that the protein encoded by a differentially spliced RFXANK/Tvl-1 mRNA (Tvl-1⌬91-112) is inactive as a transcriptional regulator of Class II MHC genes (39). In this report, we systematically addressed the reasons behind the functional inactivity of this protein. In addition, we showed that differential splicing of Tvl-1 is a process regulated by cytokine signals known to regulate the expression of Class II MHC genes. Finally, we presented evidence showing that both Tvl-1 isoforms regulate the expression of non-MHC genes.
The inability of Tvl-S to induce expression of Class II MHC genes suggested that Tvl-S may not translocate into the nucleus. Earlier studies have shown that Tvl-1, transiently transfected into 293 cells, can be detected in both the nucleus and the cytoplasm (42). Given that the marked overexpression of the protein could have affected its subcellular distribution, we sought to confirm these data using cells that express near physiological levels of Tvl-1 from stably transfected constructs. The studies presented in this report confirmed that Tvl-1 is distributed in both subcellular compartments. In addition, they showed that this is not caused by the differential distribution of the two isoforms because they are both distributed similarly. Therefore, the inability of Tvl-S to induce Class II MHC expression is not due to its inability to enter the nucleus.
Previous studies have shown that RFX complexes can be assembled in the presence of Tvl-S, but that these complexes do not bind DNA (39). The data presented in this report confirmed that Tvl-S indeed supports the assembly of RFX complex in vitro. However, the same data showed that contrary to previous reports, the Tvl-S-containing complexes bind DNA, albeit very weakly, compared with the Tvl-L-containing complexes. Inter- Tvl-1⌬91-112 Promotes the Assembly of an Unstable RFX Complex estingly, excess Tvl-L displaced Tvl-S, whereas excess Tvl-S did not displace Tvl-L in the DNA-bound RFX complexes. These data combined suggested that the protein-DNA complexes assembled in the presence of Tvl-S may be unstable. One reason for the instability of protein-DNA complexes assembled in the presence of Tvl-S is that the protein complexes themselves may be unstable. It is self-evident that the binding of an unstable complex to DNA may give rise to a protein-DNA complex that is also unstable.
To address the stability of the RFX complexes assembled in the presence of Tvl-S or Tvl-L, we first examined whether Tvl-L and Tvl-S differ in their ability to promote complex assembly when they are used at concentrations similar to those of RFX5 and RFXAP. The results were consistent with the hypothesis that Tvl-S-containing complexes are less stable in that they showed that under these conditions, Tvl-L promotes complex assembly more efficiently than Tvl-S. The decreased stability of the Tvl-S-containing complexes was confirmed by three independent experiments. First, we examined the rate of assembly of complexes formed in the presence of excess GST-Tvl-L or GST-Tvl-S and the rate at which the assembly of these com-plexes reached the equilibrium. Following this, we monitored the dissociation of complexes assembled in the presence of Tvl-L or Tvl-S. Finally, we carried out competition experiments in which the competitor was bacterially expressed GST-Tvl-L or GST-Tvl-S. The results showed that although the initial rate of assembly of Tvl-L-and Tvl-S-containing complexes is similar, the assembly of Tvl-S-containing complexes reaches equilibrium much faster than the assembly of Tvl-L-containing complexes. Moreover, complexes assembled in the presence of Tvl-S were shown to dissociate at a much faster rate than complexes assembled in the presence of Tvl-L. Finally, the competition experiments showed that GST-Tvl-L competes Tvl-S out of assembled complexes, while GST-Tvl-S does not compete out Tvl-L. These data combined confirmed the hypothesis that the complexes assembled in the presence of Tvl-S are unstable. Therefore, the Tvl-1 peptide between amino acids 91 and 112 contributes to the stabilization of the RFX complex.
Previous studies on the inability of the Tvl-S-containing complexes to bind DNA had been interpreted to suggest that the domain defined by the sequences between amino acids 91 and 112 may be directly involved in DNA binding (39). Our data do FIG. 7. Stability of RFX complexes assembled in the presence of Tvl-L or Tvl-S. A1 and A2, in vitro translated GST-RFX5 and RFXAP were incubated with in vitro translated, [ 35 S]methioninelabeled Tvl-L or Tvl-S as described under "Experimental Procedures." Protein complexes were pulled down by binding to glutathione-Sepharose beads. Washed complexes were resuspended in 50 l of the binding buffer, and they were incubated at 4°C. Supernatants and pellets collected at the indicated time points were analyzed by SDS-PAGE and autoradiography. A1, autoradiography of the SDS-PAGE showing the Tvl-L and Tvl-S released from the pelleted complexes over time. A2, the radioactivity in the Tvl-L and Tvl-S bands in A1 was measured by phosphorimager. This experiment was repeated twice with similar results. B1, autoradiography of the SDS-PAGE showing the Tvl-L and Tvl-S bound to the complexes over time. B2, the radioactivity in the Tvl-L and Tvl-S bands in B1 was measured by a phosphorimager. To combine the results of three independent experiments, we calculated the mean intensity of the Tvl-L and Tvl-S, as measured in these experiments, and we assigned it the relative value of 1. This allowed us to calculate the relative decrease in intensity of the Tvl-L and Tvl-S bands over time. C, in vitro translated, [ 35 S]methionine-labeled RFX5, RFXAP, and FLAG-Tvl-L or FLAG-Tvl-S were incubated for 1 h at room temperature. Complexes were incubated at 4°C with 50-fold excess of bacterially purified GST-Tvl-L or GST-Tvl-S. Four hours from the start of the incubation, protein complexes were immunoprecipitated with anti-FLAG M2 antibody bound to agarose beads, and they were analyzed by SDS-PAGE.
Tvl-1⌬91-112 Promotes the Assembly of an Unstable RFX Complex not exclude the possibility that this domain contacts the DNA helix. However, given the fact that Tvl-S-containing protein complexes are unstable, we do not need to invoke a direct role for this domain in DNA binding to explain the weak binding of Tvl-S-containing complexes to DNA. Database searches revealed no similarity of the domain missing from Tvl-S to known DNA-binding domains. Based on these considerations, we propose that the region between amino acids 91 and 112 defines a domain that stabilizes the assembled RFX complexes (complex stabilization domain).
Our data showed that Tvl-L-containing RFX complexes are very stable in vitro and suggested that they may also be very stable in vivo. Stable RFX complexes may support the stable expression of Class II MHC molecules in mature dendritic cells during the primary immune response. This may be physiologically important, because an efficient primary immune response depends on sustained antigen presentation to CD4 ϩ helper T-cells (47)(48)(49)(50). Therefore, stable expression of Class II MHC may be a requirement for an efficient primary immune response.
IFN-␥ modulates the ratio of Tvl-L-and Tvl-S-expressing transcripts in 293 cells in favor of the Tvl-L transcripts. Since only Tvl-L contributes to the assembly of a stable RFX complex, this shift may play a role in the induction of Class II MHC genes by IFN-␥. Our data do not exclude the possibility that IFN-␥ may differentially modulate the stability of the two RNAs. However, the most likely mechanism for the observed changes in the abundance of the two messages is the modulation of the differential splicing of Tvl-1 by IFN-␥. Although the process of RNA splicing has been extensively studied, little is known about its regulation (51,52). The IFN-␥-induced changes in the abundance of the two differentially spliced mRNAs, which is described in this report, represents an im-portant form of regulation of gene expression that needs to be further explored.
The only genes known to be regulated by Tvl-1 to-date are the Class II MHC genes (8,18,19). Given the fact that Tvl-S has neither a positive nor a negative influence on Class II MHC gene expression, the question was raised whether it regulates the expression of non-MHC genes. The results showed that BLS-1 cells engineered to express Tvl-S or Tvl-L express non-MHC genes that are not expressed in vector transfected cells. One such gene, whose induction was documented in this report, is the ephrin receptor EphA3. The fact that both Tvl-L and Tvl-S induced the expression of EphA3 suggests that its induction does not depend on the formation of RFX-containing complexes.
The Eph (ephrin) receptor family is the largest subfamily of receptor tyrosine kinases (43,53). The receptors can be divided into two groups, EphA (A1 to A8) and EphB (B1 to B6). EphA receptors bind to glycosylphosphatidylinositol-linked ligands (ephrin A ligands-A1 to A5), while EphB receptors bind to transmembrane EphB ligands. Receptor-ligand interactions play an important role in early tissue patterning events during embryogenesis (44). Best described is their role in axon guidance and fasciculation during retinal development (54,55). Non-neuronal tissues expressing ephrin receptors and ligands include the endothelium where they appear to be involved in cell growth and migration (56). One mechanism by which eph-rinA signaling may regulate cell attachment and migration is the regulation of integrin function (57). The human EphA3 (HEK) receptor is expressed at low levels in normal human tissues like thymic lymphocytes, bone marrow, and brain (45,46). It is also expressed at high levels in lymphoid tumor cell lines including the pre-B-cell line LK63, the T-cell lines Jurkat, JM, HSB-2, and Molt-4 as well as in solid tumors, including melanoma, lung carcinoma, and sarcoma (45, 46) These findings suggest that both Tvl-L and Tvl-S may be involved in normal development and oncogenesis by regulating the expression of the ephrin receptor EphA3.
In summary, the data presented in this report showed that the inability of Tvl-S to promote Class II MHC expression may be caused by the instability of the RFX complexes assembled in the presence of Tvl-S. In addition, they provided evidence for the regulation of a differential splicing event by IFN-␥-generated signals. Finally, they showed that Tvl-1 regulates the expression of non-MHC genes and that its role in the regulation of these genes differs from its role in the regulation of Class II MHC.
FIG. 8. Interferon-␥ promotes a shift in the splicing pattern of Tvl-1 toward the transcriptionally active Tvl-L isoform. A, RT-PCR was carried out using total RNA isolated from untreated or IFN-␥-treated 293 cells. The cDNAs amplified via PCR, using oligonucleotides specific for Tvl-1, HLA-DRA, or ␤-actin, were electrophoresed in a 3% agarose gel and blotted onto nylon membranes. Blots were hybridized to a 32 P-labeled Tvl-1 probe that extends from nucleotide 248 to nucleotide 268 of the Tvl-1 mRNA (upper panel), a 32 P-labeled HLA-DRA cDNA probe (nucleotides 390 -413) (middle panel), or with a ␤-actin cDNA probe (lower panel). This experiment was repeated three times. B, quantitation of the data shown in A. The radioactivity counts associated with the individual RT-PCR bands were measured by a phosphorimager. The Tvl-L and Tvl-S levels were normalized based on the levels of ␤-actin.

FIG. 9. Both Tvl-L and Tvl-S induce the expression of EphA3.
RNAs were isolated from three mass cultures each of BLS-1 cells transfected with the empty vector pREP4 or pREP4-based expression constructs of Tvl-L or Tvl-S. The expression of EphA3 was measured by RT-PCR. Parallel RT-PCR amplification of ␤-actin from the same RNAs was used as a loading control. EphA3 was amplified for 25 cycles, while ␤-actin was amplified for 20 cycles.