POZ-, AT-hook-, and Zinc Finger-containing Protein (PATZ) Interacts with Human Oncogene B Cell Lymphoma 6 (BCL6) and Is Required for Its Negative Autoregulation*

Background: PATZ is a transcription factor, whose role in cancer is still under debate. Results: PATZ interacts with BCL6 and negatively modulates its expression. Consistently, Patz1 knockdown mice showed up-regulation of BCL6 expression and BCL6-dependent B cell neoplasias. Conclusion: PATZ is a tumor suppressor that acts by cooperating with BCL6 in its negative autoregulation. Significance: This work helps in understanding the pathology of BCL6-expressing lymphomas in which BCL6 is not mutated. The PATZ1 gene encoding a POZ/AT-hook/Kruppel zinc finger (PATZ) transcription factor, is considered a cancer-related gene because of its loss or misexpression in human neoplasias. As for other POZ/domain and Kruppel zinc finger (POK) family members, the transcriptional activity of PATZ is due to the POZ-mediated oligomer formation, suggesting that it might be not a typical transactivator but an architectural transcription factor, thus functioning either as activator or as repressor depending on the presence of proteins able to interact with it. Therefore, to better elucidate PATZ function, we searched for its molecular partners. By yeast two-hybrid screenings, we found a specific interaction between PATZ and BCL6, a human oncogene that plays a key role in germinal center (GC) derived neoplasias. We demonstrate that PATZ and BCL6 interact in germinal center-derived B lymphoma cells, through the POZ domain of PATZ. Moreover, we show that PATZ is able to bind the BCL6 regulatory region, where BCL6 itself acts as a negative regulator, and to contribute to negatively modulate its activity. Consistently, disruption of one or both Patz1 alleles in mice causes focal expansion of thymus B cells, in which BCL6 is up-regulated. This phenotype was almost completely rescued by crossing Patz1+/− with Bcl6+/− mice, indicating a key role for Bcl6 expression in its development. Finally, a significant number of Patz1 knock-out mice (both heterozygous and homozygous) also develop BCL6-expressing lymphomas. Therefore, the disruption of one or both Patz1 alleles may favor lymphomagenesis by activating the BCL6 pathway.

The PATZ1 gene encodes four main alternative proteins ranging from 537 to 687 amino acids that contain an N-terminal POZ domain, one or two AT-hooks in the central region, and four to six C2H2 zinc finger motifs at the C terminus (1)(2)(3). Both the AT-hooks and the POZ domain are characteristic of protein factors involved in gene transcription by interacting with a number of other protein factors. Indeed, PATZ protein, also known as MAZR, ZNF278, or ZSG, is a transcriptional regulatory factor that may function either as activator or as repressor depending upon the cellular context; it has been reported to either activate or repress c-Myc (1,2), to activate mast cell protease 6 (4), and to repress androgen receptor (5) and CD8 (6) genes.
Different functional and genetic evidences suggest that PATZ might be directly involved in human tumors. Indeed, PATZ1 is rearranged and deleted in small round cell sarcoma (3), and the chromosomal region where it is located (22q12) is in the human fragile site FRA22B, which suffers loss of heterozygosity in tumors (7). Increased expression of PATZ1 mRNA has been observed in human malignant neoplasias, including colorectal (8), breast (9), and testicular (10) tumors. Moreover, PATZ knockdown by siRNA either blocks the growth or induces apoptosis of cell lines derived from colorectal cancer or gliomas, respectively (8,11). However, in testicular tumors alone, PATZ protein expression has been analyzed, demonstrating that it was mislocalized to cytoplasm (10,12). Therefore, although PATZ1 is strongly suggested to be a cancer-related gene, its role as tumor suppressor or oncogene is still controversial.
In the present study, starting from a yeast two-hybrid screening using PATZ1 full-length cDNA as bait, we demonstrate that PATZ associates with BCL6, 3 a protein that shares with PATZ the N-terminal POZ domain, responsible for such association, and is involved in B and T cell development and lymphomagenesis (13)(14)(15)38). We show that PATZ participates in BCL6 function by enhancing its activity of transcriptional repressor on BCL6 promoter in GC-derived lymphoma B cells. We also show that the knock-down of PATZ in mice causes BCL6-expressing thymus B cell hyperplasias (that eventually lead to B cell lymphomas), in which BCL6 is critical for their onset. The development of BCL6-expressing lymphomas in Patz1 knockdown mice indicates a potential haploinsufficient tumor suppressor role for the PATZ1 gene, whose disruption may lead to the lymphomas by activating the BCL6 pathway.

EXPERIMENTAL PROCEDURES
Two-hybrid Analysis-Two-hybrid screens were performed in yeast using full-length PATZ1 (isoform 4) cDNA as a bait. Human heart and placenta cDNA libraries (Clontech) were simultaneously analyzed. A total of about 2 ϫ 10 6 clones were tested for each library, and the specificity of interaction was assessed as described previously (16).
Plasmids-Full-length and truncated (devoid of the BTB/ POZ domain) cDNAs for the human PATZ protein (isoform 4), Myc-tagged at their 3Ј-end, were subcloned into the XbaI-HindIII sites of the pcDNA3.1 plasmid (Invitrogen). The cDNA for the human BCL6 protein was subcloned into the EcoRI site of the pCEFL-HA vector (17), in-frame with the upstream HA tag. The BCL6i-luc reporter construct was obtained by cloning the Ϫ4.9 to ϩ2.0-kb fragment of the BCL6 promoter into the pGL3Basic vector as described previously (18).
Protein Extraction, Immunoprecipitation, and Immunoblot Analysis-Tissues and cells were lysed in buffer containing 1% Nonidet P-40, 1 mmol/liter EDTA, 50 mmol/liter Tris-HCl (pH 7.5), and 150 mmol/liter NaCl supplemented with Complete protease inhibitors (Roche Applied Science). Total proteins were immunoprecipitated, in the presence or absence of 100 ng/ml ethidium bromide, as described previously (19), or they were directly resolved in a 10% polyacrylamide gel under denaturing conditions and transferred to nitrocellulose filters for Western blot analyses. Membranes were blocked with 5% BSA in TBS and incubated with the primary antibodies. The antibodies used were: anti-HA (sc-805), anti-Myc (sc-40), anti-BCL6 (sc-858), anti-tubulin (sc-5546), anti-vinculin (sc-7649) (Santa Cruz Biotechnology, Santa Cruz, CA), and anti-PATZ (polyclonal antibody raised against a conserved peptide recognizing all PATZ isoforms of mouse and human origin).
Cell Cultures and Transcriptional Activity Assays-Raji cells, originally derived from a Burkitt lymphoma, were cultured in RPMI 1640 medium adjusted to contain 1.5 g/liter sodium bicarbonate, 4.5 g/liter glucose, 10 mM HEPES, 1.0 mM sodium pyruvate, penicillin/streptomycin (Invitrogen/Life Technologies Italia, Monza, Italy), and 10% FBS (JRH Biosciences, Lenexa, KS). They were transfected by using the Amaxa Nucleofector kit V (Lonza, Cologne, Germany) following the manufacturer's instructions. COS-1 cells were cultured in DMEM with 10% FBS (JRH Biosciences). They were transfected using Lipofectamine Plus reagents (Invitrogen/Life Technologies Italia) according to the manufacturer's instructions and harvested 42 h after transfection. Cell lysates were extracted as described below and analyzed for luciferase activity. For transcriptional activity assays, a total of 2 ϫ 10 6 cells (Raji) were seeded into each well of a 12-well plate and transiently transfected with 4 g of BCL6i-luc, 5 g of HA-BCL6, and 0.25-0.5 g of PATZ-Myc, together with 1 g of Renilla and various amounts of the backbone vectors to keep the total DNA concentration constant. Transfection efficiency, normalized for Renilla expression, was assayed with the Dual-Luciferase system (Promega Corp., Madison, WI). All transfection experiments were repeated at least three times. Aliquots of the same lysates were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with anti-BCL6, anti-PATZ, and antivinculin antibodies, as above described.
Generation of Patz1-Knock-out Mice-The Patz1 gene targeting vector was derived from a ⌽XII phage library of a 129SvJ mouse strain (Stratagene, La Jolla, CA). It was designed to delete a 2317-bp PstI-XhoI fragment, including the start codon, the coding regions for the POZ domain, the AT-hook, and the first four zinc fingers. It was constructed by subcloning the 5Ј-flanking region (the SpeI-PstI 3-kb fragment), the neo cassette, and the 3Ј-flanking region (the XhoI-XbaI 3.2-kb fragment) into the Bluescript plasmid (Stratagene) that contained a PacI digestion site inserted at a distance from the multicloning site. The targeting vector was linearized with PacI before electroporation into embryonic stem (ES) cells (Incyte Genomics, Palo Alto, CA). Among 700 G418-resistant ES clones examined, 10 (1.4%) underwent homologous recombination. Two correctly targeted ES cell lines were injected into C57Bl/6J blastocysts. Both ES cell lines gave rise to germ line chimeras that were backcrossed to C57Bl/6J females to obtain Patz1 heterozygous offspring. For Southern blot analysis, tail DNA samples were digested with StuI and probed with an external 5Ј genomic fragment that would detect 9.3-or 8-kb fragments, corresponding to the wild-type and mutant alleles, respectively. The mice were maintained under specific pathogen-free conditions, and all studies were conducted in accordance with Italian regulations for experimentations on animals.
Isolation of mRNA and Quantitative RT-PCR-Total RNA was extracted using TRI-reagent solution (Sigma) according to the manufacturer's protocol, treated with DNase I (Invitrogen/ Life Technologies Italia), and reverse-transcribed using random hexanucleotides as primers and MuLV reverse transcriptase (PerkinElmer Life Sciences) following the manufacturer's instructions. For quantitative RT-PCR, each reaction was performed three times in triplicate using the SYBR Green PCR master mix (Applied Biosystems, Foster City, CA) under the following conditions: 10 min at 95°C followed by 40 cycles (15 s at 95°C and 1 min at 60°C). Subsequently, a dissociation curve was run to verify amplification specificity. The 2 Ϫ⌬⌬CT method was used to calculate relative expression levels (20). Primers specific for the glucose-6-phosphate dehydrogenase (G6PD) gene were used for normalization of real-time quantitative PCR data. The following primers were used: Patz, forward, 5Ј-GAG-CTTCCCCGAGCTCAT-3Ј; Patz, reverse, 5Ј-CAGATCTCG-ATGACCGACCT-3Ј; G6pd, forward, 5Ј-CAGCGGCAACTA-AACTCAGA-3Ј; G6pd, reverse, 5Ј-TTCCCTCAGGA-TCCCACAC-3Ј.
IgH Gene Rearrangement Analysis-Genomic DNA was isolated from tumor masses or normal tissues from control mice, and Southern blotting was performed with the 32 P-labeled DNA probe PJ3 representing the J H 4 region of the IgH locus (21). A 1.3-kb Pst-1 fragment from the Gapdh gene was used for loading control.
Analysis of Lymphocyte Cell Surface Antigens-Spleens and thymi removed from mice were dissociated into single cells and stained for FACS analysis on a FACSCalibur flow cytometer (BD Biosciences, Buccinasco, Italy) as described previously (22). All the antibodies used were obtained from Pharmingen.
Statistical Analyses-Kaplan-Meyer survival curves were used to analyze the percentage of tumor-free mice. Differences were analyzed by the log rank test. The one-way analysis of variance followed by Tukey's multiple comparison test was used to compare groups of experiments. The statistical significant difference was considered when the p value was Ͻ0.05.

RESULTS
PATZ Interacts with BCL6-To identify PATZ interacting proteins, two-hybrid screenings of human heart and placenta pretransformed libraries were performed. A total of 87 positive clones (61 and 26 from heart and placenta, respectively) were isolated. Fourteen clones from heart and two from placenta libraries contained most of the coding sequence of the BCL6 gene (data not shown). To confirm the interaction between PATZ and BCL6 in mammalian cells, co-immunoprecipitation experiments were performed. To this aim, total cell extracts from COS-1 cells transiently transfected with constructs encoding Myc-tagged PATZ, Myc-tagged PATZ-⌬POZ (devoid of the BTB/POZ domain), and HA-tagged BCL6 (Fig.  1A) were subjected to immunoprecipitation using anti-Myc antibody. As shown in Fig. 1B, co-precipitation of the BCL6 protein was observed when constructs for BCL6 and full-length PATZ were co-transfected, and not when BCL6 was transfected together with PATZ-⌬POZ or with the backbone vector. Therefore, PATZ and BCL6 form a complex in mammalian cells, and the POZ domain of PATZ is necessary for such interaction. Next, to ensure that the interaction between PATZ and BCL6 takes place in a physiological context, endogenous PATZ and BCL6 were co-immunoprecipitated in Raji cells, deriving from a Burkitt lymphoma, which is a GC-derived lymphoma expressing both proteins. The result, shown in Fig. 1C, further demonstrates that PATZ and BCL6 interact in vivo.
PATZ Is Involved in BCL6 Negative Autoregulation-To study the functional consequences of the interaction between PATZ and BCL6, we investigated whether PATZ and BCL6 associate on a BCL6 endogenous target promoter. Previous experiments demonstrated that the BCL6 gene contains two non-STAT6 functional BCL6 binding sites located within exon 1 (23). Thus, we examined whether PATZ could bind exon 1 of the BCL6 gene in vivo. Endogenous PATZ and BCL6 bind to exon 1 of BCL6, as determined by ChIP experiments in Raji cells ( Fig. 2A). This result was further confirmed in vivo using mouse spleens, a tissue that harbors GC cells, where both PATZ and BCL6 proteins were immunoprecipitated on BCL6 exon 1 (Fig.  2B). Interestingly, BCL6 binding to its own promoter was enhanced (Fig. 2B) in spleens lacking PATZ (obtained from Patz1 knock-out mice described below). The absence of immunoprecipitated PATZ in Patz1-null tissues confirmed the binding of PATZ to BCL6 exon 1 observed in the wild-type mice.
It is known that BCL6, by binding to its exon 1 and recruiting the distal cis-acting factor ZEB1, represses its own transcription, thus establishing a circuit of negative autoregulation (23,24). To further investigate the role of PATZ in such a function of BCL6, we used a reporter construct driving luciferase gene expression under control of a BCL6 genomic sequence of 6.9 kb, including the ZEB1-responsive element, exon 1, and 1.5 kb of intron 1 (BCL6i-luc) (18). Raji cells were co-transfected with BCL6i-luc and plasmids expressing PATZ, BCL6, or both proteins (Fig. 2C). As expected, BCL6 expression resulted in downregulation of the BCL6 promoter activity. A similar, but less strong, activity was also achieved by PATZ, leading to a dosedependent repression of this BCL6 promoter region. Interestingly, PATZ and BCL6 co-transfection resulted in the enhancement of the BCL6-dependent repression (Fig. 2C). Western blot analysis of Raji cells transfected or not with increasing amounts of PATZ expression plasmid, as in Fig. 2C, confirmed the negative dose-dependent effect of PATZ on BCL6 expression (Fig. 2D). To finally assess the dependence of BCL6 expression from PATZ, we analyzed, by Western blot, the BCL6 protein levels in mouse embryonic fibroblasts from Patz1 ϩ/ϩ and Patz1 Ϫ/Ϫ mice (described below). As shown in Fig. 2E, Patz1null mouse embryonic fibroblasts showed a significant increase in BCL6 expression, which confirms the negative regulatory role of PATZ on it.
Disruption of Mouse Patz1 Gene Causes Thymus B Cell Expansion, Which Eventually Leads to B Cell Lymphomas-To gain insights into the physiological role of PATZ in vivo, a targeting vector was designed to knock out the mouse Patz1 gene The recovered DNA was used as a template for PCR with primers that specifically amplify exon 1 of BCL6 or the promoter region of the LPL gene (negative control). input indicates PCR products with chromosomal DNA without immunoprecipitation. Nonspecific IgG was used as a negative control. B, ChIP assay, revealed by quantitative PCR, on spleens from Patz1 ϩ/ϩ and Patz1 Ϫ/Ϫ mice to detect the binding of PATZ and BCL6 proteins to exon 1 of Bcl6 DNA. The percentage of immunoprecipitated DNA with respect to the input is reported. C, luciferase activity of the BCL6 promoter (Ϫ4.9 to ϩ2.0 kb) in Raji cells. Where indicated, 0.25 and 0.5 g of PATZ and/or 5 g of BCL6 expression vectors were co-transfected with the BCL6i-luc plasmid. Where required, backbone vectors were also transfected to reach the same total amount of DNA for each experimental point. Data express mean Ϯ S.D. of three independent experiments. Asterisks indicate the statistical results of a multiple comparison test versus promoter basic activity. *, p Ͻ 0.05; **, p Ͻ 0.01, ***, p Ͻ 0.001. D, representative Western blot analysis to detect BCL6 expression in Raji cells transfected or not with increasing amounts of PATZ expression plasmid, as in C. As a control for equal loading, the same blot was incubated with antibodies against vinculin. E, Western blot analysis of BCL6 expression in mouse embryonic fibroblasts from Patz1 ϩ/ϩ and Patz1 Ϫ/Ϫ mice. Vinculin has been analyzed as a loading control. by homologous recombination in ES cells (Fig. 3A). Inactivation of the Patz1 allele was verified by analyzing Patz1 expression in adult tissues using quantitative RT-PCR and Western blot (Fig.  3, B and C). As expected, Patz1 mRNA and protein were undetectable in homozygous mutants, whereas they were present at about half-levels in heterozygous as compared with wild-type controls. Heterozygous mice were viable and fertile, but heterozygous intercrosses produced viable homozygous mutants at non-Mendelian frequencies. Indeed, the large majority of homozygous mutants (about 75% on average) died prenatally due to developmental defects in the cardiac outflow tract. 4 Here we focused on some phenotypes, observed in adult mice (both heterozygous and homozygous for the Patz1-null mutation), which can be due to an impaired regulation of the BCL6 gene. By 2-3 months of age, with a similar frequency (75%), Patz1 ϩ/Ϫ and Patz1 Ϫ/Ϫ mice develop thymus hyperplasias characterized by focal expansion of an intramedullary B cell population (Fig. 4A). B cells are the minority cell type (0.82 Ϯ 0.50%) in the thymus of wild-type animals, as confirmed by FACS analysis in a cohort of 20 animals, but significantly increase to an average of 5.94 Ϯ 4.42 (p Ͻ 0.05) in an equal number of Patz1 ϩ/Ϫ mice (Fig. 4B). A similar increase of B cells was also observed in Patz1 Ϫ/Ϫ , where, for the limited number of available animals, we could only do a qualitative analysis (data not shown). To exclude that this B cell expansion could be due to formation of intrathymic reactive B cell follicles possibly induced by an altered CD4 helper versus cytotoxic subset ratio (25), we analyzed CD4ϩ and CD8ϩ lymphocyte populations by FACS in Patz1 ϩ/Ϫ thymi showing B cell hyperplasia, without finding any differences in comparison with wild-type controls (data not shown). Interestingly, in 4 out of 75 Patz1 ϩ/Ϫ mice sacrificed at an advanced age (17-22 months old), we also found thymus B cell lymphomas, characterized by large cells with marked cell-to-cell variation in size and shape and abundant pale cytoplasm, which were diagnosed as diffuse large B cell lymphomas by our pathologists (Fig. 4C and data not  shown).
Thymus B Cell Lesions in Patz1 Knock-out Mice Are Dependent on Bcl6 Expression-Pathological thymus B cells are considered to originate from the GC (26). Accordingly, focal expression of BCL6 was detected in Patz1-knock-out thymus lesions (either hyperplasia or lymphomas) but was absent in the wild-type controls (Fig. 5A). Western blot analyses on thymus tissues from Patz1 ϩ/Ϫ mice confirmed the expression of BCL6, which was absent in Patz1 ϩ/ϩ controls (Fig. 5B). Up-regulation of BCL6 in these cells is consistent with the role of PATZ in BCL6 autoregulation (see above), which leads us to suggest that decreased or null levels of PATZ causes up-regulation of BCL6 expression, which in turn could be responsible for the thymus pathological phenotype. To validate our hypothesis of a role for increased BCL6 expression in the development of certain phenotypes in Patz1 knock-out mice, we crossed Patz1 ϩ/Ϫ with Bcl6 ϩ/Ϫ mice (27) to generate double mutants. Patz1 Ϫ/Ϫ ; Bcl6 ϩ/ϩ , Patz1 ϩ/ϩ ;Bcl6 Ϫ/Ϫ , and Patz1 Ϫ/Ϫ ;Bcl6 Ϫ/Ϫ mice died during embryogenesis. However, because we observed the thymus phenotype in Patz1 heterozygous mice, we analyzed the alterations present in Patz1 ϩ/Ϫ ;Bcl6 ϩ/ϩ mice and compared them with the double heterozygous Patz1/Bcl6 mutants.
Cohorts of 10 Patz1 ϩ/Ϫ ;Bcl6 ϩ/ϩ and 10 Patz1 ϩ/Ϫ ;Bcl6 ϩ/Ϫ mice were equally distributed by gender and sacrificed at 12 months of age. Their thymus was analyzed by histological, immunohistochemical, and FACS assays. As shown in Fig. 5C, where representative FACS analyses are shown, 100% of the Patz1 ϩ/Ϫ ;Bcl6 ϩ/Ϫ mice analyzed did not show any significant thymus B cell expansion as compared with the Patz1 ϩ/Ϫ ; Bcl6 ϩ/ϩ controls. Immunohistochemical analyses confirmed this result in 80% of the cases, whereas in the remaining 20% of double heterozygous mice, an aberrant thymus B cell hyperplasia was observed, but it was strongly reduced as compared with that observed in Patz1 ϩ/Ϫ ;Bcl6 ϩ/ϩ mice (Fig. 5D). These results confirm a key role for the up-regulation of BCL6 in the pathological thymus phenotype of Patz1 knock-out mice.
Increased Tumorigenesis in Patz1 Knock-out Mice-Although initially healthy, many adult Patz1 ϩ/Ϫ and most of the few available Patz1 Ϫ/Ϫ mice developed signs of morbidity and displayed visible tumors as they aged. As evident in the survival curves in Fig. 6A, Patz1 Ϫ/Ϫ and Patz1 ϩ/Ϫ mice showed an increased incidence of neoplastic lesions at ϳ17 or 21 months of age, respectively. Only 12% of the Patz1 Ϫ/Ϫ and 67% of the Patz1 ϩ/Ϫ animals were tumor-free at 20 months, in contrast to the 95% of wild-type mice that were tumor-free at the same age.
The histological analyses revealed that 43 of 75 (57%) Patz1 ϩ/Ϫ and 9 of 11 (82%) Patz1 Ϫ/Ϫ mice developed malignant tumors, versus only 7 of 63 (11%) wild-type mice. Patz1 knock-out tumors were mainly lymphomas, but hepatocellular carcinomas and rare sarcomas and lung adenocarcinomas were also detected. Wild-type animals developed lymphomas and one hepatocellular carcinoma (Fig. 6B and data not shown). Lymphomas from wild-type mice were all of the B cell lineage but did not show BCL6 expression nor PATZ down-regulation as compared with normal controls (data not shown).
To determine whether the tumors in Patz1 ϩ/Ϫ mice occurred via loss of heterozygosity or haploinsufficiency, PATZ protein expression and Patz1 gene sequence were analyzed in the tumor tissues from Patz1 ϩ/Ϫ mice. Western blot (Fig. 6C) and immunohistochemical analysis (data not shown) revealed that PATZ was present in all of the tumors examined. Furthermore, sequence analysis of Patz1 cDNA isolated from eight Patz1 ϩ/Ϫ tumors showed no mutation or rearrangement (data  MAY 25, 2012 • VOLUME 287 • NUMBER 22 not shown). These results indicate that haploinsufficiency, rather than loss of heterozygosity, accounts for the tumor occurrence in Patz1 ϩ/Ϫ mice, thus suggesting Patz1 as a haploinsufficient tumor suppressor gene.

PATZ Is Crucial for BCL6 Negative Autoregulation
Spleen lymphomas were the most representative malignant diseases in Patz1 knock-out mice, and they were further characterized to determine the mechanism by which reduced Patz1 expression induces such neoplasias. Tumor sections from Patz1 ϩ/Ϫ mice were stained with antibodies raised against T cell-and B cell-specific markers to determine the cell type of origin of the lymphoma. Interestingly, we found both B type and T type lymphomas with a prevalence of B (28%) versus T lymphomas (17%). The B type lymphoma (BCL) subtypes, as defined by Morse et al. (28), were as follows: (a) diffuse large B cell (16.8%), (b) follicular B cell (5.6%), (c) pre-B cell (2.8%), and (d) small B cell lymphomas (2.8%) (Fig. 6D). Flow cytometry analysis confirmed the immunohistochemical data (data not shown). To further characterize lymphoid neoplasias, IgH gene configuration was analyzed by Southern blot of EcoRI-digested spleen DNA from Patz1 ϩ/ϩ and Patz1 ϩ/Ϫ mice using a probe representing the J H 4 region of the IgH locus (Fig. 6E). The 6.5-kb fragment represents the germ line configuration of the gene. All BCLs showed oligoclonal rearrangements of the germ line IgH joining region (Fig. 6E, asterisks). Interestingly, benign lymphoproliferative diseases often show aberrant V(D)J recombination at the IgH locus, suggesting that they may represent an FIGURE 6. Increased lymphomagenesis in Patz1 knock-out mice. A, Kaplan-Meyer tumor incidence analysis of Patz1 ϩ/ϩ , Patz1 ϩ/Ϫ , and Patz1 Ϫ/Ϫ mice. Cohorts of 63 wild-type, 75 heterozygous, and 11 homozygous Patz1 knock-out mice were monitored and harvested when they exhibited symptoms of disease. The curves were significantly different (p Ͻ 0.0001) as determined by log rank test. B, the lymphoid phenotype, as a function of the relative percentage, in mice homozygous and heterozygous for the Patz1-null mutation versus wild-type controls was plotted as pie charts. The number of mice analyzed for each genotype is the same as in A. C, Western blot analysis showing expression of BCL6 and PATZ in spleen samples from Patz1 ϩ/ϩ and Patz1 ϩ/Ϫ mice. ␣-Tubulin expression was evaluated as a loading control. N, normal; Hy, hyperplasia; Ly, lymphoma. D, immunohistochemical staining for the phenotypic characterization of lymphomas in Patz1 knock-out mice. Representative Patz1 ϩ/Ϫ spleen samples of all types of lymphomas observed are shown (for the percentage of each type, see under "Results"). Antibodies used for the staining are indicated on the top. H&E, hematoxylin and eosin staining; DLBCL, diffuse large cell B cell lymphoma; TCL, T cell lymphoma; pre-BCL, pre-B cell lymphoma; FBL, follicular B cell lymphoma; SBL, small B cell lymphoma. Scale bars, 100 m. E, IgH gene rearrangements were analyzed by Southern blot on EcoRI-digested spleen DNA. Control (ϩ/ϩ) is the wild-type mouse with the genomic 6.5-kb fragment representing the gene in its germ line configuration. All hyperplastic (Hy) and B cell lymphoma samples from Patz1 ϩ/Ϫ mice, and normal spleen from a Patz1 Ϫ/Ϫ mouse, show rearranged extra bands (asterisks). The histological diagnoses of the spleens are indicated above. early stage in the development of BCLs. The same Southern blot was also probed for gapdh, which gave rise to a unique band for all loaded samples (data not shown). Both BCLs and lymphoproliferative diseases expressed BCL6, whereas it was not detectable in wild-type controls or normal spleens from Patz1 knock-out mice (Fig. 6C), suggesting a key role for this oncoprotein in development of such neoplasias.

DISCUSSION
Previous studies suggest a cancer-related role for PATZ (3, 8 -11), but the mechanisms by which PATZ is involved in the process of carcinogenesis are still controversial. Our data indicate that PATZ acts as a tumor suppressor in lymphomagenesis by inhibiting BCL6 expression. We first showed that PATZ and BCL6 bind to each other. This finding is consistent with the notion that POK proteins, such as PATZ and BCL6, commonly aggregate in large nuclear complexes by self-interaction, as heterodimers with other POK proteins and in association with unrelated partners, such as transcriptional corepressors, as a way to extend the repertoire of their target genes and/or the ways they act on their expression (29). BCL6 is the most commonly altered proto-oncogene in non-Hodgkin lymphomas, the majority of which derive from normal GC B cells (30). In fact, its sustained expression causes malignant transformation of GC B cells (31). Nearly half of human diffuse large B cell lymphomas, the most common form of non-Hodgkin lymphomas, express BCL6 constitutively, mainly as a consequence of BCL6 gene rearrangements and activating point mutations that target the 5Ј regulatory region of this gene (23,24,32,33). However, different studies showed that BCL6 expression in lymphoma is largely independent from the corresponding chromosomal alterations, suggesting that mechanisms other than gene rearrangements or mutations can deregulate its expression in lymphomas (34,35). BCL6 expression is tightly regulated in a lineage-and developmental-stage-specific manner, and disruption of normal controls can contribute to lymphomagenesis (13,14,36). Transcription of the BCL6 gene is negatively selfregulated by means of the interaction of two BCL6 binding sites within exon 1 of the gene and the BCL6 protein itself, which is a potent transcription repressor, and both chromosomal translocations and activating mutations allow lymphoma cells to bypass this negative autoregulation mechanism (23,24). To date, a variety of corepressors have been described. Among them, the CtBP1 and ZEB1 transcriptional repressors are required for BCL6 autoregulation (18,37). In a recent model, a repressive complex at the BCL6 locus that involves binding of transcription factors to both exon 1 and a distant cis-acting element (HSS-4.4) has been suggested to regulate BCL6 transcription. In this complex, ZEB1 and BCL6, which bind to distant sites, are linked through binding of the common corepressor CtBP1 (18). In the present study, we show that PATZ can bind BCL6 exon 1 and negatively modulate BCL6 promoter activity. Therefore, we speculate that PATZ can be one of the transcription factors involved in the complex and that, because of its ability to bind the minor groove of DNA via the AT-hooks, could be crucial for the bending of DNA required to put together distant cis-acting transcription factors. Because PATZ can interact and cooperate with BCL6, we can suggest that it acts together with BCL6 in its autoregulation. However, we cannot exclude that PATZ can also act without BCL6, thus contributing to keep BCL6 expression off when BCL6 is not expressed at all. Consistent with a role of PATZ in BCL6 negative regulation, we also show that mice carrying a null mutation of the Patz1 gene develop an aberrant expansion of thymus B cells, in which BCL6 expression is up-regulated. We believe that this phenotype, which we demonstrated to be dependent on BCL6 expression, is also responsible for the development of BCLs that we observe in both thymi and spleens of Patz1 knockout mice at a later age.
In conclusion, our data indicate a haploinsufficient tumor suppressor role for PATZ that would act in lymphomagenesis by down-regulating BCL6 expression.