Functional Analyses of Two Alternative Isoforms of the Transcription Factor Pax-5*

The Pax-5 gene plays a central role in B cell development, activation, and differentiation. At least four different isoforms have been identified, of which isoform Pax-5a has been extensively studied, while functions for alternative isoforms were previously unknown. Here, using a transient transfection system, we provide evidence that alternative isoform Pax-5d acts as a domi-nant-negative regulator by suppressing activity of Pax-5a in a dose-dependent manner. In contrast, co-ex-pression in the presence of alternative isoform Pax-5e causes an increase in Pax-5a activity. Protein studies on Pax-5e using Western blot analysis revealed that this 19-kDa isoform migrates as a 27-kDa species on SDS-polyacrylamide electrophoresis gels, while a mutant Pax-5e form in which a C-terminal cysteine residue has been mutated, runs at the expected 19 kDa. Using both Western blot and immunoprecipitation assays, we further provide evidence that this size discrepancy may be caused by a tight association between Pax-5e and a thi-oredoxin-like factor. Comparison of various B cell lines as well as resting and lipopolysaccharide-activated mature B lymphocytes shows that increased B cell proliferation correlates with increased levels of Pax-5e/thi-oredoxin, whereas increased Pax-5d amounts correlate with inhibition of cell growth. Together, our results suggest that during activation and differentiation of B lymphocytes, Pax-5a function

B lymphocytes are the major players of the humoral immune system and are essential to the detection and elimination of pathogens. Spatial and temporal gene expression of B cellspecific transcription factors largely determines the maturation and activation pathways in a B cell, and is a tightly regulated process. A number of transcription factors have now been identified as essential for B cell development and activation (reviewed in Refs. 1 and 2), including one of the products encoded by the Pax-5 gene, the B-cell specific activator protein (BSAP). Pax-5 expression is first detected in the developing central nervous system (3,4). After birth and throughout life, Pax-5 transcripts are found in cells of the B-lymphoid lineage and in testis of the adult mouse (3). Within the B cell lineage, Pax-5 is expressed during early stages of B cell development up to the mature B-cell, but is greatly down-regulated or absent in plasma cells (5). Inactivation of the Pax-5 gene in mouse results in a complete block of B cell development at the pro-B cell stage, revealing the essential role of this gene in early B cell lymphopoiesis (6).
Pax-5-binding sites have been identified on the promoters of a number of B cell-specific genes (reviewed in Ref. 7). Among the positively regulated Pax-5 targets are genes encoding the CD19 co-stimulatory receptor (8) and the protein-tyrosine kinase Blk (9). Pax-5 functions as a repressor for the immunoglobulin J chain and the Ig 3Ј␣ enhancer (10 -12). In addition to its role in B-lymphopoiesis, Pax-5 has been implicated in activation and proliferation of B lymphocytes since its decreased expression resulted in reduced numbers of cells post-activation (13).
Recent reports by Tell et al. (14,15) provide evidence that Pax-5a activity is regulated through a redox mechanism that involves Ref-1. The authors show that an oxidized form of Pax-5a is unable to interact with DNA, whereas the reduced form binds strongly, and that an intramolecular disulfide bond within the paired domain of Pax-5a causes interference with specific DNA binding (14). Furthermore, exposure of B cells to H 2 O 2 results in rapid transfer of the cytoplasmic redox factor Ref-1 into the nucleus and this correlates with an increase in Pax-5 binding activity (15).
The Pax-5 gene produces four isoforms as a result of alternative splicing: Pax-5a (full-length Pax-5 or BSAP), Pax-5b, Pax-5d, and Pax-5e (16). Pax-5a and Pax-5d isoforms, but not Pax-5b, are expressed at detectable levels in normal B cells, although the levels of Pax-5d transcripts are significantly lower than those of Pax-5a (16,17). As shown in Fig. 1, both Pax-5a and Pax-5d, but not Pax-5e, possess an intact DNAbinding domain, enabling them to interact with and compete for Pax-5-binding sites on DNA in vitro ( Fig. 1; Ref. 16). In contrast, neither Pax-5d nor Pax-5e possesses transactivation, repression, or partial homeodomain homology regions at the C terminus. Instead, in Pax-5d and -5e, the region encoded by exons 6 -10 is replaced with a 128 nt 1 novel sequence (16) with unknown function. Based on the DNA binding abilities and expression pattern of Pax-5a and -5d, we hypothesize that the two isoforms compete for DNA-binding sites and have opposite effects on transcription of target genes in vivo.
Although a number of studies have shown clear functional significance for isoform Pax-5a (8,18,19), no prior work had yet characterized the functions of isoforms Pax-5d and -5e. Thus, the first goal of our studies was to determine the transactivation properties of both isoforms. Here we show, using a transient transfection system, that Pax-5d has a transactivating function opposite to that of Pax-5a, whereas, unexpectedly, isoform Pax-5e increases the activity of Pax-5a. Upon further characterization we found evidence that isoform Pax-5e forms a strong complex in the nucleus with a thioredoxin-like molecule. Furthermore, using B cell lines as well as resting and LPSactivated B cells, we show that the ratio of Pax-5d to Pax-5e correlates with the proliferation state of the B cell. The observed changes in Pax-5d/5e ratio are regulated at the protein level, as we found no evidence for changes in ratios of Pax-5d to Pax-5e transcript levels. In summary, data presented here suggest that during B cell activation/proliferation, the activity of transcription factor Pax-5a may be regulated through changes in relative amounts of alternative isoforms Pax-5d and Pax-5e, and such changes likely affect expression of Pax-5 target genes.
DNA Constructs-The ␥42(3i)AS-CAT reporter construct was created using the p␥42CassI CAT reporter (20) (Fig. 2A). Expression of CAT is driven by the truncated rat ␥42-fibrinogen promoter (Ϫ54 to ϩ36), which includes a TATA box and a single Sp-1-binding site. Three copies of the high-affinity Pax-5-binding site from the CD19 promoter (5Ј-CAGACAC-CCATGGTTGAGTGCCCTCCAG-3Ј) were inserted into the polylinker upstream of the ␥42-fibrinogen promoter. Recombinant constructs were se-quenced to determine copy number and orientation of Pax-5-binding sites. The effector constructs pcDNA.5a and pcDNA.5d were made by cloning the cDNA sequences of Pax-5a, Pax-5d, or Pax-5e into NotI restriction sites of the expression vector pcDNA3 (Invitrogen). The pcDNA3 construct was used as a negative control effector construct, and the HBIICAT construct (9) was used as a control for transfection efficiency. The pcDNA.5e.mC5 construct was made using a PCR-directed site-specific mutagenesis approach. Primers m5b/5eATG2.S (5Ј-ccgtgcggccgc 323 CCATGTTTGCCTGGGAG 339 -3Ј, containing in small letters a linker and the NotI restriction site) and m5d/ 5e.mC5,AS, 5Ј-gcttctaga 202 CTAGGACCCTGGGAAGCCCGGTCCTCTG CTGCTA 735 -3Ј, including in small letters a linker XbaI restriction site) were used to amplify the region encoding the complete Pax-5e isoform, using pcDNA.5e as a template. A single replacement mutation of nt T (from codon TGC (Cys) to AGC (Ser), underlined as T on antisense primer) replaced the most 3Ј-terminal cysteine (Cys 5 ) codon with the structurally related Serine residue. The amplified sequence was cloned into the NotI and XbaI sites of pcDNA3 (Invitrogen). The cysteine replacement mutation was confirmed by dideoxy sequencing. Transient Transfections and Chloramphenicol Acetyltransferase Assay-Transient transfections of B cell lines A20/2J and Sp2/0 were performed by the DEAE-dextran (21) method as described previously (9). Nonlymphoid cell lines COS-1 and NIH 3T3 were transfected using LipofectAMINE Plus (Life Technologies, Inc.) according to the manufacturer's protocols. Data were quantified using an NIH Image software analysis program (rsb.ihfo.nih.gov/nih-image/). Relative CAT conversion was determined as described previously (9).
Isolation of Cell Fractions and Activation of Small Resting B Cells-Splenic B cells were obtained from 3-6-month-old BALB/c mice (bred at The College of William and Mary). Small resting B cells (SRBs) were isolated from a 70% Percoll gradient (Amersham Pharmacia Biotech) as described previously (18). SRB populations were activated by culturing in complete RPMI 1640 medium (supplemented as above) in the presence of 20 g/ml bacterial lipopolysaccharide (LPS) (Sigma) for the required period of time.
Nuclear Extract Preparation-Cells were collected at specified times and processed for nuclear extracts as described elsewhere (22). Procedures for nuclear extract preparation were carried out on ice in a cold room at 6°C.
Western Blot Analysis-Nuclear extracts and cytoplasmic fractions from SRBs, LPS-activated B cells, or cell lines were separated on 12-15% denaturing SDS-polyacrylamide gels and electrophoretically transferred onto nitrocellulose filters (Schleicher and Schuell) as described previously (18). Antibody probing was performed as described previously (18). The data were quantified using an NIH Image software analysis program (rsb.ihfo.nih.gov/nih-image/).
Anti-Pax Antibodies-Information about isotype-specific Pax-5 antibodies used in this study is summarized in Table I. Pax-5d/Pax-5especific mouse monoclonal antibody 6G11, recognizing the C-terminal "novel" sequence, was generated in our lab (17). 6G11 supernatants were used at a 1:60 to 1:100 dilution and detected with a horseradish peroxidase-conjugated goat anti-mouse IgG secondary antibody (Zymed Laboratories Inc.). ED-1 antiserum (18) was used at a 1:2000 dilution. Pax-5/N-19 and Pax-5/C-20 were used at a 1:400 dilution and detected with a horseradish peroxidase-conjugated rabbit anti-goat IgG (Zymed Laboratories Inc.). OC-1 was used at 1:1000. Rabbit polyclonal antiserum to the transcription factor TFIID (Santa Cruz Biotechnology) was used at a dilution 1:200. The ED-1, OC-1, and anti-TFIID antibodies were detected with a horseradish peroxidase-conjugated donkey anti-rabbit IgG secondary antibody (Amersham Pharmacia Biotech). Hybridoma supernatant containing anti-TRX/IgA mAbs developed against catfish TRX was a gift from Drs. Khayat and Clem (University of Mississippi Medical Center) (23) and was used at a 1:30 dilution, and detected using a horseradish peroxidase-conjugated goat anti-mouse IgA secondary antibody (Zymed Laboratories Inc.).
In Vivo Cell Labeling and Immunoprecipitation-SRBs were partially purified by Percoll gradients and grown in culture in the presence of 20 g/ml LPS for 2 or 6 days. Cells from both time points were always processed at the same time: first, cells were cultured in long term labeling mixture (1% fetal bovine serum, 90% RPMI minus methionine, and 10% complete RPMI) in the presence of 0.1 mCi/ml [ 35 S]methionine (Ͼ800 Ci/mmol; Amersham Pharmacia Biotech) for 8 h (24), and nuclear extracts prepared. Nuclear extracts for both time points were then incubated overnight at 4°C in the presence of ED-1 (1:250 dilution), ␣-TRX supernatant (1:30 dilution), or 6G11 supernatant (1:30 dilution), and mouse anti-Thy-1.2 mAb HO-13-4 (ATTC) hybridoma supernatant (1:30) as negative control. For ED-1 and 6G11, Protein G-Sepharose was then added for 2 h at 4°C, followed by immunoprecipitation, denaturing of the complexes, and finally, SDS-PAGE analysis. For ␣-TRX/IgA containing nuclear extracts, Protein G-Sepharose was preincubated with goat anti-IgA antiserum (1:50; Zymed Laboratories Inc.) for 2 h, followed by repeated washes of the beads before their addition to the nuclear extracts. Gels were dried, fixed, and incubated in fluorographic reagent using an Amplify kit (NAMP100; Amersham Pharmacia Biotech), followed by exposure to x-ray film for 2-6 days.

RESULTS
Pax-5a plays essential roles during B cell development and activation, and this isoform has been studied extensively. In contrast, no prior studies had yet investigated the regulatory properties of two alternative Pax-5 isoforms, Pax-5d and -5e. The transactivation function of isoform Pax-5a on the tyrosine kinase gene blk was previously determined in our laboratory using the chloramphenicol transferase (CAT) reporter gene: Pax-5a acts as an activator of blk expression (18). In the present studies, we initially used the same target promoter sequence as was used for Pax-5a to investigate Pax-5d and -5e function, but due to low transfection efficiency in B cell lines combined with low activity of the blk promoter (9), we were unable to determine reliable quantitative differences. As an alternative approach, we created an artificial promoter containing multiple Pax-5 DNA-binding sites, similar to an approach previously used by others (25). Three copies of a doublestranded oligonucleotide containing the high affinity Pax-5 DNA-binding site from the murine CD19 promoter were cloned in sense or antisense orientation upstream of the TATA element of the truncated rat ␥42-fibrinogen promoter driving expression of the CAT gene, as shown in Fig. 2A. Both sense and antisense reporters gave similar promoter activities and the antisense construct, named ␥42(3i)AS-CAT, was used in all subsequent experiments.
To verify the specificity of ␥42(3i)AS-CAT, namely, that reporter expression was expressed in the presence, but not absence, of endogenous Pax-5a protein, transient transfections using DEAE-dextran were performed in the mature B cell line A20/2J and the plasma cell line SP2/0. As expected, CAT expression was detected in the Pax-5 positive A20/2J line (Fig.  2B), but no activity was detectable in the Pax-5 negative SP2/0 line (Fig. 2C). As a control for transfection efficiency, the HBI-ICAT construct was used. This construct contains a portion of the RNA polymerase II promoter and drives high expression of CAT (9).
Can Isoform Pax-5d Affect the Activity of Pax-5a?-Based on earlier in vitro data (16) showing that Pax-5a and -5d have similar affinity for Pax-5 DNA-binding sites on the blk promoter, we first sought to investigate whether the potentially dominant negative isoform Pax-5d can inhibit Pax-5a activity in vivo. For this set of experiments, the NIH3T3 fibroblast cell line was used because this cell type does not express endogenous Pax proteins, allowing for transfection of highly controlled amounts of various Pax-5 isoforms.
To determine the activity of Pax-5a alone on the ␥42(3i)AS-CAT reporter in this cell line, co-transfections were performed with the expression vector pcDNA.5a. The pcDNA3 vector without insert was used in all subsequent transfections as a negative control, and added where necessary to maintain equal amounts of total transfected DNA. Results from repeated CAT assays showed that the reporter gene was expressed at high levels in the presence, but not absence, of Pax-5a (Fig. 3A) confirming a positive transactivating function for this isoform using the ␥42(3i)AS-CAT reporter. In contrast, isoform Pax-5d alone was unable to activate the reporter gene, yielding only basal levels of transcription that were similar to those produced in the presence of the control plasmid pcDNA3 (Fig. 3A).
Next, it was tested whether alternative isoform Pax-5d could affect the activity of Pax-5a. Reporter construct ␥42(3i)AS-CAT was transiently co-transfected into NIH3T3 cells using fixed amounts of pcDNA.5a and increasing amounts of pcDNA.5d (Fig. 3A). The data showed that Pax-5d was able to suppress Pax-5a dependent activity of the CAT reporter in a dose-dependent manner, although relatively high amounts of Pax-5d were needed for efficient suppression (discussed in later section). To verify that amounts of transfected DNA corresponded with the correct amounts of Pax-5a and Pax-5d protein, nuclear extracts from transfected cells were analyzed by Western blot analysis. Results using the ED-1 antibody (which detects Pax-5a and -5d protein with the same intensity, as both have a complete paired domain), showed the expected expression patterns of both proteins (Fig. 3B, upper panel), and Pax-5d expression was independently verified using the Pax-5d/5e specific monoclonal 6G11 (Fig. 3B, lower panel).
Function of Pax-5d in COS-1 Cells-Kidney cells contain endogenous Pax proteins which have been shown to interact with Pax-5 DNA-binding sites (27,28). The kidney cell line COS-1 does not express Pax-5 but expresses the closely related Pax-8 protein (26,27). Both Pax-5 and Pax-8 belong to the same subclass of Pax proteins and both recognize the Pax-5 DNA binding sequence from the human CD19 promoter with high affinity (27,28). Having demonstrated that Pax-5d is able to inhibit Pax-5a activity, we further hypothesized that Pax-5d might also be inhibitory to the activity of the related transcription factor Pax-8.
Before we could test this, the presence of Pax-8 protein in COS-1 cells needed confirmation, and this was done by EMSA using a 28-nt CD19/BSAP probe (18) and four different anti-Pax-5 antisera, ED-1, OC-1, N-19, and C-20 (see Table I). As shown in Fig. 4A (lane 2), a Pax-like DNA-protein complex was detected which migrated slower than in vitro translated (ivt) Pax-5a in agreement with the larger size of Pax-8 (Pax-5a, 392 aa, Pax-8, 450 aa). ED-1 antiserum recognizes the highly conserved paired domain of all Pax proteins, and was able to remove the complex as expected (Fig. 4A, lane 3). Furthermore, incomplete removal of the same complex was observed using the Pax-5-specific OC-1 antiserum, due to partial (52%) sequence homology between Pax-5 and Pax-8 in this region (Ref. 27; Fig. 4A, lane 4). In contrast, two Pax-5-specific antibodies, Pax-5/N-19 and Pax-5/C-20, were unable to remove the complex (Fig. 4A, lanes 5 and 6). Together, the data provided the necessary evidence that COS-1 cells indeed express Pax-8.
First the transactivating effect of endogenous Pax-8 on the ␥42(3i)AS-CAT reporter was tested using transient transfections in the COS-1 cell line. Endogenous Pax-8 was able to induce high CAT reporter expression in the absence of effector constructs as shown in Fig. 4B (pcDNA3 column), showing that Pax-8 recognizes the Pax-5 DNA-binding sites on the reporter and acts as an activator in this system. In contrast, co-transfections using reporter plus increasing amounts of co-transfected Pax-5d effector construct resulted in a dose-dependent decrease in CAT activity (Fig. 4B).
Addition of increasing amounts of pcDNA.5a did not significantly increase CAT activity of the reporter, possibly because endogenous Pax-8 already has a saturating transactivating effect on the reporter (result not shown). Together, these results indicate that alternative isoform Pax-5d suppresses endogenous Pax-8 activity in a dose-dependent manner in COS-1 cells, most likely through competition for Pax-5 DNA-binding sites. These data also suggest that Pax-5d may serve as a general inhibitor for this class of Pax proteins.
Transactivating Activity of Isoform Pax-5e-To study effects of isoform Pax-5e on Pax-5a activity, we next performed cotransfections in NIH3T3 cells using different ratios of pcDNA.5a and pcDNA.5e, similar to the approach described for Pax-5d. Unexpectedly, results showed that co-transfections of pcDNA.5a in the presence of an equal amount of pcDNA.5e (1:1) resulted in a significant increase of CAT activity over Pax-5a effector alone; thus, the presence of isoform Pax-5e increased the transactivating activity of Pax-5a (Fig. 5A). Because of this unexpected result, the experiment was repeated seven times and the activating effect was consistently observed. Interestingly, when excess Pax-5e was added (either 10-or 5-fold), we did not observe the increase in activity ( Fig.  5A and not shown).
Subsequent Western blot analyses of transfected samples using ED-1 confirmed that the amounts of transfected pcDNA.5a and pcDNA.5e DNA correlated with the amounts of expressed Pax-5a and -5e protein, respectively (Fig. 5B). Western blot signals for Pax-5a were stronger than for Pax-5e for the same amount of transfected DNA (e.g., Fig. 5B, lane 4), because the ED-1 antiserum recognizing the complete paired domain. In Pax-5e protein, the presence of an incomplete paired domain leads to reduced signal intensity.
Transfections were repeated in COS-1 cells to determine whether the Pax-5e activating effect represented a more general activating mechanism shared by other Pax members such as Pax-8. However, repeated co-transfections did not show any effect of isoform Pax-5e on the activity of Pax-8 (not shown), suggesting that the effect is specific to Pax-5 proteins.
In summary, transfection studies both in NIH3T3 and COS-1 cells demonstrate that: 1) neither Pax-5d nor Pax-5e have detectable transactivating activities in the absence of activators Pax-5a or Pax-8; 2) isoform Pax-5d can suppress activity of both Pax-5a and Pax-8 in a dose-dependent manner; and 3) isoform Pax-5e is able to specifically enhance the activity of Pax-5a but not of the related Pax-8 protein. Based on these data, we hypothesize that the relative amounts of isoforms Pax-5a, -5d, and -5e in the cell play important and specific regulatory roles during B cell development and/or activation.
Interestingly, we noted from our co-transfection studies in NIH3T3 and COS-1 cells that: 1) Pax-5d does not repress Pax-5a activity as efficiently as that of Pax-8; and 2) Pax-5e has an enhancing effect on the activity of Pax-5a, but not of Pax-8. Previously (16), we observed that small amounts of Pax-5e protein can be translated off Pax-5d transcripts through the use of a second, in-frame start codon. Expression of low levels of Pax-5e proteins during Pax-5a/5d transfections in NIH3T3 may increase Pax-5a activity (but not Pax-8 activity), resulting in higher promoter activities and consequently less efficient repression by Pax-5d. Thus the relative levels of nuclear Pax-5a, -5d, and -5e proteins in the cell are crucial for Pax-5a activity on target genes.
Identification of a 27-kDa Pax-5e Species-During the Western blot analysis of our transfection studies we observed a 27-29-kDa Pax-5-like protein of unknown origin. In previous studies, we had already observed a 27-kDa Pax-5-like species in normal, mature B cells (17). This Pax-5-like species reacted both with the paired domain-specific ED-1 antibody as well as with the Pax-5d/5e-specific monoclonal antibody 6G11, but not with a Pax-5a/5b-specific OC-1 antibody (17). Additionally, the protein was not detectable by EMSA using the CD19/Pax-5 probe (17). Given the characteristics of this 27-29-kDa protein  species, including the presence of the novel 5d/5e-specific sequence but not the homeodomain homology region (see Fig. 1), the inability to interact with DNA, and the significantly smaller size as compared with Pax-5d, we hypothesized that this Pax-5-like species may represent isoform Pax-5e, although the 137-aa Pax-5e protein has a predicted size of only 18 -19 kDa.
To further investigate whether the observed 27-29-kDa protein species corresponded with the 27-kDa Pax-5-like protein identified in normal B cells (17), pcDNA.5e DNA was transfected into NIH3T3 cells and nuclear extracts prepared. Samples were analyzed by SDS-PAGE in parallel with nuclear extracts prepared from normal, resting mature B cells (SRB cells), as well as the immature B cell line WEHI231. After probing the blots with mAb 6G11, it was found that NIH3T3transfected Pax-5e protein was present as a 27-kDa species only (Fig. 6A). Additionally, the 27-29-kDa Pax-5e-containing band ran in a position identical to the 27-kDa band observed previously in both normal B cells (17) and the WEHI231 B cell line (Fig. 6A). Use of additional blots and/or reprobing of the same blots using OC-1 and ED-1 (not shown), confirmed that the 27-kDa band contained Pax-paired domain sequence but not homeodomain homology sequence. Together, the results showed that the Pax-5e protein preferentially runs as a 27-kDa band on denaturing SDS-PAGE gels. Importantly, our results provide the first evidence that B cells and B cell lines express readily detectable levels of alternative isoform Pax-5e.
We hypothesized that the size discrepancy was either the result of extensive secondary structure formation of Pax-5e, or perhaps more likely, the association of Pax-5e with another (small) protein. In attempts to obtain the 19-kDa Pax-5e band, we took several approaches aimed to fully denature and reduce the 27-kDa Pax-5e-like protein. These included increasing the percent of ␤-mercaptoethanol in the gel-loading buffer (from 5 to 10%) together with increased boiling times (up to 15 min) prior to loading. Both treatments had only a minor effect on the pattern, resulting in a slight shift toward the 19-kDa Pax-5e species (not shown). Furthermore, nuclear extracts were incubated with dithiothreitol, followed by exposure to iodoacetamide, which irreversibly blocks reduced sulfhydryl residues on proteins. Such treatment has been shown to prevent oxidationdependent aggregation of some cysteine-containing proteins such as interferon-inducible protein-10, eliminating formation of homodimers during SDS-PAGE (29). However, iodoacetamide treatments did not result in a significant shift from the 27-to 19-kDa Pax-5e species (not shown). Isoform Pax-5e contains a total of three cysteine residues, one on its partial paired domain, and two on its C-terminal novel sequence (Fig. 1). Assuming that cysteine residues should be critical for the formation of either intra-or intermolecular disulfide bonds, we mutated one of the two cysteine residues present on the novel sequence of Pax-5e. If disulfide bonds were involved, such a mutation would be expected to cause a shift in migration of Pax-5e from the 27 to the 19-kDa position. Using a PCR-directed site-specific mutagenesis approach, we replaced the most C-terminal cysteine residue (Cys5) on construct pcDNA.5e with the structurally related aa serine. Transfections of the pcDNA.5e.mC5 construct into NIH3T3 followed by Western blot analysis showed that the Pax-5e.mC5 mutant indeed migrated in the expected 19-kDa position on the gel, as shown in Fig. 6B. Based on these results, two scenarios are possible: 1) 27-kDa Pax-5e protein represents the oxidized form of Pax-5e, and intramolecular disulfide bond formation is dependent upon the presence of residue Cys 5 . The oxidized form of Pax-5e severely slows down migration through the polyacrylamide mesh. 2) Alternatively, and perhaps more likely, Cys 5 may be necessary for the formation of a heterodimeric complex through the formation of an intermolecular disulfide bond. In this scenario, the 27-kDa species consists of Pax-5e as well as another, unidentified protein. Given the large size discrepancy (27 versus 19 kDa), the latter possibility seems more probable. Before the nature of the 27-kDa Pax-5e species was investigated more thoroughly, we first wished to establish the expression patterns of isoforms Pax-5d and Pax-5e in B cell lines and normal B cells thus supporting their functional significance in B cell maturation/activation.
Expression Patterns of Pax-5d and -5e in B Cell Lines-In previous experiments we observed that isoforms Pax-5d and -5e were expressed at detectable levels in B cell nuclear extracts from WEHI231. To determine potential functions for both isoforms during B cell development and/or activation, we analyzed their expression patterns in a number of B cell lines representing different stages of B cell development. Nuclear extracts were prepared from pro-B, pre-B, immature-B, mature-B, presecretors, and plasma cell lines and analyzed using Western blot analysis with the 5d/5e novel sequence-specific monoclonal Ab 6G11. Results from a representative experiment are shown in Fig. 7A and complete analysis from a total of 12 cell lines is summarized in Fig. 7B. Together, the data showed that pro-B cell lines possessed low levels of Pax-5d while Pax-5e was undetectable. Mature B cell lines expressed varying levels of both isoforms, and the presecretor (activated B) cell line CH12 expressed mostly Pax-5e.
As evident from Fig. 7, A and B,  tremendous variation in ratio of Pax-5d to 5e between experiments. In an attempt to identify the cause of variation, we used an earlier observation namely that "overgrown" mature B cell cultures (with cell densities higher than 1.5 ϫ 10 6 cells/ml) tended to display a shift toward more Pax-5d and less Pax-5e protein. Cell overgrowth is likely to correlate with decreased proliferation in such cultures. To test for a possible effect of cell density/starvation on the 5d/5e ratio, we used the mature B cell line B17.10 that has a doubling time of 12-14 h during log phase growth (at cell densities below 1 ϫ 10 6 cells/ml). Log phase B17.10 cultures were freshly seeded at 0.5 ϫ 10 6 cells/ml (time point 0), and cells remained in culture without feeding. Samples were then collected at different time points and nuclear extracts prepared, followed by Western blot analysis using mAb 6G11. As shown in Fig. 7C (right panel), the 5d/5e protein ratio had increased significantly after 1 day in culture, with a clear increase in Pax-5d, and a decrease in Pax-5e, and this trend continued in subsequent days.
To further explore the possibility that the change in 5d/5e ratio was related to cell proliferation and/or cell-cycle related events, a cell-cycle inhibition experiment was performed by growing B17.10 cultures in the absence of serum and essential amino acids followed by Western blot analysis of nuclear extracts (using mAb 6G11). The results, as shown in Fig. 7C (left  panel), were essentially the same: the Pax-5d/5e ratio had increased after 1 day of cell-cycle blocking and this trend continued on day 2. The changes in ratios of Pax-5d/5e where quantified using densitometric analysis as described under "Materials and Methods," and the ratios are shown below the panel in Fig. 7C. Together, the starvation and cell-cycle block experiments suggested that in mature B cell lines, low Pax-5d/5e levels correlated with actively proliferating cells, whereas high Pax-5d/5e ratios correlated with inhibited cell proliferation or growth.
To determine whether the changes in 5d/5x ratio observed in B17.10 cells were the result of changes in alternative splicing of Pax-5d and Pax-5e transcripts, we performed RT-PCR on RNA isolated from B17.10 cells collected in parallel with the cell density experiments described above: both day 0 (log phase growth) and day 4 (starvation) cell samples were collected. Two PCR-amplified DNA bands were expected, a 571-nt band that is Pax-5d specific, and a 413-nt band representing both Pax-5d and Pax-5e. If the ratio between the two PCR fragments were similar in both cell samples, this would indicate that similar amounts of both Pax-5d and Pax-5e transcripts were present, suggesting alternative splicing does not play a role in the observed changes in 5d/5e ratios. As a control, we performed parallel PCR amplifications with different ratios of plasmids pcDNA.5d and pcDNA.5e. Results (Fig. 7D) show that the ratios of the two Pax-5 PCR fragments are very similar between log phase B17/10 cells (day 0) as compared with starving B17.10 cells (day 4), while control PCR (plasmid) samples clearly showed the expected ratio-dependent changes in intensity of 413-nt versus 571-nt bands. We conclude that the observed increase in Pax-5d:5e protein ratio as seen during starvation is not the result of a change in alternative splicing of Pax-5 RNA. It is possible that the increase in Pax-5d protein levels is the result of a shift in start codon usage from the second ATG to the first ATG on Pax-5d transcripts (see Fig. 1). Alternatively, the transcript or protein stability of the two isoforms may change depending on the state of cell proliferation or growth of the B cells.
Does LPS Activation of Mature B Cells Affect the Ratio of Pax-5d to Pax-5e?-Our studies in B cell lines suggested that: 1) the Pax-5d/5e protein ratio may decrease as B cells mature; and 2) a decrease in cell growth correlates with a shift toward more Pax-5d and less 5e. Next, we sought to study a related question, namely, whether induction of cell proliferation during B cell activation leads to a decrease in 5d/5e protein ratio. This was investigated using normal SRBs that had been partially purified from mouse splenic cell suspensions by Percoll gradients (18). SRBs were activated in culture with bacterial LPS, several time points after activation were collected, and nuclear Pax-5d and -5e protein levels were assessed using Western blot analysis. We used in vitro translated Pax-5a and Pax-5d as controls and monitored total nuclear protein levels by reprobing the filters with an antibody that detects the basal transcription factor TFIID.
Using the anti-Pax-5d antibody 6G11 (Fig. 8A, upper panel), it was shown that levels of Pax-5d protein were highest in resting cells, and decreased gradually in activated B cells over the course of the LPS response. Pax-5e levels showed the opposite pattern: protein levels were low in nuclear extracts from resting B cells (day 0) and then increased gradually after LPS stimulation with a peak at day 6, and slightly decreasing at day 8 (Fig. 8A). Lastly, Pax-5a and TFIID levels were measured as controls (Fig. 8A). These results strongly support our observations in B cell lines: in the case of LPS activation, increased cell proliferation correlates with a reduction in nuclear Pax-5d protein levels with a concomitant increase in Pax-5e protein.
Does Isoform Pax-5e Interact with Another Protein?-Once we had established that Pax-5d and Pax-5e show specific expression patterns during B cell development and activation, thus suggesting functional significance for both isoforms, we addressed a final question, namely, what is the molecular nature of the Pax-5e protein? This issue was addressed by searching for a putative Pax-5e partner, as the presence of such a partner would explain why wild-type but not mutant Pax-5e migrates at 27 kDa on SDS-PAGE gels. Based on the predicted size of Pax-5e (18 -19 kDa, Ref. 16), a putative partner-protein would need to be in the range of 8 -12 kDa. Earlier reports by Tell et al. (15) had found a functional interaction between Pax-5a and the redox factor Ref-1, while our own recent data also suggested that Pax-5 activity may be regulated through redox mechanisms (17). One candidate as a potential Pax-5e partner was the redox factor TRX. This small (12 kDa) protein factor was recently reported to interact directly with the redox factor Ref-1 as well as with the transcription factor NF-B/p50 (30) and the DNA-binding domain of the glucocorticoid receptor (31). The highly conserved TRX molecule is expressed in all organisms and in many cell types, including proliferating B cells (32). Its molecular size is within the range of a potential Pax-5e partner, making it a good candidate for our study.
The presence of TRX as part of a Pax-5e-containing complex was tested using a catfish monoclonal anti-TRX (IgA) mAb that recognizes the highly conserved active site Trp-Cys-Gly-Pro-Cys (23). Using the same nuclear extracts prepared during the LPS activation experiments (Fig. 8A), we prepared independent Western blots and probed with anti-TRX mAbs. Significantly, a 27-kDa protein species was detectable, as shown in Fig. 8A (bottom  panel). Furthermore, the level of this 27-kDa TRX-like protein increased over time with a peak at day 6 and decreased slightly at day 8, a pattern identical to that seen for Pax-5e. Thus, TRX levels correlate with the level of the 27-kDa Pax-5e species. The anti-TRX mAb did not interact with in vitro translated Pax-5d or -5e, ruling out the possibility that this antibody cross-reacts with Pax-5d/5e proteins (not shown). These data support the idea that the 27-kDa Pax-5e-like species is composed of Pax-5e and TRX, and additionally, that this complex is involved in B cell proliferation.
Based on the Western blot patterns, we next investigated whether Pax-5e⅐TRX complexes could be immunoprecipitated using in vivo metabolic labeling of activated SRBs. Nuclear 27-kDa Pax-5e complexes should be readily detectable after 6 days of LPS activation, but not after 2 days of LPS activation. "Day 6" LPS-activated cells were isolated and cultured 4 days prior to "day 2" activated cells, so that both time points could be metabolically labeled and processed simultaneously. Nuclear extracts were immunoprecipitated with ED-1, TRX, or 6G11, or an unrelated (anti-Thy1.2) supernatant from the hybridoma line HO-13-4 as negative control. Complexes were separated on denaturing SDS-PAGE, and analyzed by fluorography. As shown in Fig. 8B, a 27-kDa band was specifically immunoprecipitated at high levels in day 6 samples, but only at very low levels in day 2 samples, using either anti-Pax-5e (6G11 or ED-1) or anti-TRX antibodies, while control anti-Thy1.2 mAb resulted in background bands only. The presence of the 27-kDa band specifically corresponded with the significant increase observed in Pax-5e at day 6 after LPS activation (Fig. 8A). This specific presence of the 27-kDa Pax-5e species at day 6 provides further evidence that the complex contains both Pax-5e as well as a TRX-like protein.

DISCUSSION
Alternative splicing of transcription factors increases the regulatory capacities for gene expression, and this mechanism is commonly used within the Pax family of transcription factors. In addition to Pax-5, alternatively spliced transcripts have been observed in Pax-2, -3, -6, and -8 genes (27,(33)(34)(35). Functional significance of alternative splicing within this family was first shown for the Pax-8 gene by Kozmik et al. (27). Six alternative isoforms were found to be developmentally expressed and to display differential transactivating potential based on the structure of their C termini. Similarly, Pax-5 isoforms may differ in transactivating activities as well.
In this study we analyzed possible functions for alternative isoforms Pax-5d and -5e, as both are potentially important regulators of Pax-5a activity. Based on structure alone, Pax-5d likely represents a dominant-negative isoform as it lacks a transactivation domain but binds to Pax-5 DNA-binding sites with similar affinity as isoform Pax-5a (16). Furthermore, Pax-5d possesses a 42-aa cysteine-rich novel C-terminal sequence with unknown function which may play a role in specific protein-protein interactions.
Isoform Pax-5e, like Pax-5d, has no transactivating domain and possesses the novel C-terminal sequence, but in contrast to Pax-5d, it has an incomplete DNA-binding domain. Thus, Pax-5e is unlikely to interact with Pax-5 DNA-binding sites directly, but may compete with Pax-5a for functional associations with other co-repressors or -activators. Based on our data, we also explore the possibility that this smallest of Pax-5 isoforms may have a recruitment, shuttling, and/or redox function in the B cell. Such functions may serve strictly to regulate the activity of Pax-5a, and/or (indirectly) to regulate B cell growth. These possibilities will be discussed further below.
Functions of Pax-5d-Our transfection data clearly demonstrate that isoform Pax-5d has the ability to act as a transcriptional suppressor. In addition to showing that Pax-5d can repress activity of Pax-5a in NIH 3T3 cells in a dose-dependent manner, we also show that it can similarly repress the activity of endogenous Pax-8 in the kidney cell line COS-1. Thus Pax-5a and Pax-5d can have an opposite regulatory function in vivo. Since Pax-5a and Pax-5d possess similar affinities for Pax-5 DNA-binding sites in vitro (16), the repressor function of Pax-5d may involve competition with Pax-5a for specific DNAbinding sites, which suggests a possible dominant-negative function for isoform Pax-5d. In the absence of both a transactivating domain and a homeodomain region, Pax-5d, although bound to DNA, is unable to interact with the basal initiation complex and/or other regulatory factors. By occupying Pax-5 DNA-binding sites, Pax-5d prevents transcription initiation of target genes by Pax-5a.
The Pax-5 gene has been suggested by others to play an important role in B cell proliferation although the exact mechanism for this is unclear (13). In this regard, it is of interest that Pax proteins have been shown to promote oncogenesis in tissue culture cells and in mice, and this depends on the presence of the paired domain (36). Pax genes are thus considered proto-oncogenes (36). Furthermore, overexpression of Pax-5a in splenic B cells stimulates cell proliferation (13), whereas Pax-5a activity is increased when B cells are activated by LPS (13). Last, the human PAX-5 gene is reportedly involved together with the Ig HC locus in a translocation associated with the onset of non-Hodgkins's lymphoma (37).
Given the growth-stimulating properties of Pax-5a together with our observations that Pax-5d inhibits Pax-5a (and Pax-8) activity in transfection assays, it is likely that Pax-5d has a reducing or inhibiting effect on cell proliferation. In agreement with this prediction, we observed that increased levels of Pax-5d protein correlated with B cell growth inhibition in B cell lines. When normal resting B cells are stimulated to undergo active cell proliferating, Pax-5d protein levels decrease. This should increase the likelihood that Pax-5a protein associates with Pax-5 DNA-binding sites on target genes.
The Structure of Pax-5e-Unexpectedly, we found that Pax-5e protein is expressed in B cells, which had not been anticipated (16,17). The reason for an apparent absence of Pax-5e expression observed in previous studies was a size discrepancy of the protein in SDS-PAGE analyses. Transfection studies using Pax-5e in NIH3T3 showed clearly that Pax-5e migrates as a 27-kDa species on SDS-PAGE. Subse-quent SDS-PAGE analysis of a mutated form of Pax-5e in which a cysteine residue in the novel sequence of Pax-5e was replaced with a serine residue, showed that mutated Pax-5e protein runs in the expected 19-kDa position. This suggests that Pax-5e protein can be present as a fully denatured monomer under conditions when cysteine residue Cys 5 is absent, although this experiment does not necessarily prove a (hetero)dimeric nature for the Pax-5e complex.
Further investigations on the nature of the Pax-5e-like species supported the possibility that Pax-5e is present in strong association with a TRX-like protein. The ubiquitous TRX protein has dithiol-disulfide oxidoreductase activity and has been shown to have a wide variety of functions (38 -42). In support of the existence of Pax-5e⅐TRX complexes in vivo, we have also detected the 27-kDa Pax-5e species in whole B cells analyzed by SDS-PAGE. 2 No apparent protein sequence homology was found to exist between TRX and Pax-5e, and no evidence for antibody cross-reactivity was found in our control samples. The interaction between Pax-5e and TRX appears to be extremely tight in vitro, and only very low levels of unbound Pax-5e are detectable in nuclear extracts. Both Western blot analysis and immunoprecipitation experiments using nuclear extracts from in vivo labeled B cells supported the existence of a Pax-5e⅐TRX complex, as both approaches suggest that the 27-kDa species contains both Pax-5e and TRX.
Pax-5e Function-Our Western blot studies in B cell lines indicate that the ratio of Pax-5d to Pax-5e decreases as B cell progenitors become (pre)mature B cells: no Pax-5e protein was detectable in progenitor B cell lines, but this protein species was readily detectable in late stage B cell lines. In contrast, the relative amount of Pax-5d protein decreased during B cell maturation of B cell progenitors and was undetectable in presecretor (activated) B cell lines. This trend of decreasing Pax-5d/5e ratios during B cell development was also seen in LPSactivation studies using normal mature B cells (SRBs). Over the time course of the LPS response, activated SRBs showed a decrease in Pax-5d amounts, while Pax-5e levels increased simultaneously. Starvation and cell cycle blocking experiments in B cell lines also indicated that active B cell proliferation correlates with high Pax-5e levels and low Pax-5d levels, while inhibited growth or resting B cell stages showed the opposite pattern.
Pax-5e appears to interact strongly with a TRX-like protein, and the presence of this complex in the nucleus positively correlates with B cell proliferation. This is particularly interesting considering that high TRX levels have previously been linked to B cell proliferation (47). The mechanism by which isoform Pax-5e accomplishes its potential role(s) in cell proliferation is unclear at this time. It is possible that Pax-5e function involves shuttling TRX into the nucleus, especially as it has been suggested by others that TRX itself lacks a NLS (32). The redox molecule TRX has been shown to have the ability to directly enhance the activity of other factors including Ref-1, NF-B and Pax-8 (43)(44)(45)(46). In this regard, the study by Kambe et al. (46) is especially significant as it reports that TRX is able to up-regulate Pax-8-mediated promoter activity of the thyroglobulin gene by reducing disulfide bonds of the Pax-8 protein.
2 P. Zwollo, unpublished data. Additionally, recent studies investigating the redox regulation of Pax-5 show that only reduced Pax-5 protein can interact with DNA, and that redox factor Ref-1 is involved in the reduction of Pax-5a disulfide bonds (14,15). Thus a functional relationship between TRX, REF-1, and Pax-5 isoforms likely exists, but this needs further investigation.
The C-terminal novel sequence in Pax-5e may play a crucial role in redox regulation. Comparison of this 42-aa novel sequence with the available protein data banks showed that it has significant homology with a number of oxidoreductase molecules, most notably with part of the bacterial hox F gene encoding the ␣ subunit of NAD-reducing hydrogenase, including the presence and position of two cysteine residues, as illustrated in Fig. 9 (16,48,49). Homology included 62% identical aa, 83% positive aa, over a stretch of 24 aa (Fig. 9). NADreducing hydrogenase contains structural features of a NADH oxidoreductase (49), thus supporting a role for the novel sequence of Pax-5e in redox regulation.
Functional studies showed that Pax-5e alone had no effect on the promoter activity of a Pax-5 DNA-binding site containing reporter construct. In contrast, co-transfections using equal amounts of Pax-5a and Pax-5e resulted in increased expression of the reporter gene, suggesting a role for Pax-5e as a coactivator. Although a 1:1 ratio had a significant positive effect on transactivating activity, 5-or 10-fold excess of Pax-5e did not. We do not know why this activating effect is dose-sensitive and observed only in the presence of similar amounts of both isoforms. It is possible that in the presence of high Pax-5e levels (excess Pax-5e), this protein will preferentially form homodimers which may lead to its own inactivation, but this needs further investigation. Alternatively, excess Pax-5e may have a quenching effect on proteins of the basal initiation complex or on other-co-activators.
Protein-protein interactions between isoform Pax-5e and other nuclear factors are likely to be dependent upon the presence of the centrally located octamer sequence. Recently, the octamer sequence of Pax-5a was shown to interact with corepressor Groucho4, and this highly conserved sequence (present on all four Pax-5 isoforms) is likely involved in repressing transcriptional activity of Pax-5a (50). Based on the presence of an octamer sequence, it is likely that Pax-5e, like Pax-5a, has the ability to interact with Groucho4. In such a model, increased levels of Pax-5e during cell proliferation would act as a quenching system by capturing the repressor Groucho4, increasing Pax-5a activity in the nucleus. It remains to be determined whether Pax-5e would be able to interact with regulatory proteins including Groucho4, Ref-1, and/or TRX, in vivo. Future studies involve investigation of protein-protein interactions in vivo using yeast two-hybrid systems.
In summary, our studies provide evidence that during activation and differentiation of B lymphocytes, Pax-5a function is modulated by the relative amounts of two alternative spliced isoforms. Dominant negative isoform Pax-5d may mediate inhibition of Pax-5a activity in resting B cells, while alternative isoform Pax-5e associated with a TRX-like molecule may increase Pax-5a activity in proliferating B cells.