Mouse Polycomb M33 is required for splenic vascular and adrenal gland formation through regulating Ad4BP/SF1 expression

The spleen comprises 2 structurally distinct parts, the red pulp and white pulp, which predominantly contain a large number of erythrocytes and tightly packed lymphoid cells, respectively. In addition to these regions, the spleen possesses a unique vascular system composed of venous sinuses and blood vessels. The venous sinuses form a meshwork structure throughout the red pulp and the marginal zone of the white pulp. Blood from fine arteries ending in these regions flows into splenic cords and thereafter passes into the venous sinuses through narrow windows between endothelial cells that constitute the sinus wall. Finally, the blood drains directly into the pulp vein. These specialized endothelial cells are thought to be present only in the spleen; thus, the complicated meshwork structure of the venous sinuses is also thought to be unique to this organ, and it supports the spleen's special ability to eliminate abnormal or damaged blood cells. During their passage through the sinus wall, these blood cells are trapped by the wall and soon after are endocytosed by macrophages. Thus, the endothelial cells of the venous sinus are thought to be crucial for the clearance of abnormal and damaged blood cells. Nevertheless, the mechanisms underlying splenic development remain to be investigated. Progress on this topic includes the observation of Hox11,¹ Bapx1,^2,3 Wt1,⁴ Nkx2-3,^5,6 Ad4BP/SF1, and capsulin⁷  gene expression in the spleen and disruption of these genes resulted in various splenic defects. Of particular interest, it was noticed that an orphan nuclear receptor and transcription factor, adrenal 4 binding protein/steroidogenic factor-1 (Ad4BP/SF-1),⁸  was expressed specifically in these specialized endothelial cells, and subsequent knock-out of that gene revealed that it is essential for the development of functional splenic vasculature.

The Polycomb group (PcG) was identified as a functional group of proteins that repress Hox gene transcription at the chromatin level in Drosophila.^9,10  PcG, together with the counteracting trithorax (trxG) proteins,¹¹  have been found to establish a cellular memory by maintaining transcription states beyond cell division. Indeed, it was shown that PcG and trxG¹²  interact directly with DNA elements in the regulatory region of the Hox gene cluster to respectively silence and activate gene expression in flies.^13-19  These elements were designated Polycomb response elements (PREs) or trithorax response elements (TREs), and PcG and trxG are assumed to form multimeric complexes tethered near the target loci. Because of the binding patterns of PcG and trxG on polytene chromosomes of the larval salivary gland,²⁰  PREs have been predicted to be distributed among approximately 100 loci throughout the Drosophila genome.

In mammals, multiple orthologues of PcG component members have been identified.²¹ Polycomb (Pc) has 3 mouse counterparts, M33 (Mpc1), Mpc2, and Mpc3. Likewise, other PcG members also have multiple counterparts. Therefore, mammalian PcG complexes are thought to be composed of distinct sets of constituent group members. In fact, mammalian and Drosophila PcG complexes have been subdivided into 2 distinct types according to their biochemical and functional properties.²²  The first complex, containing embryonic ectoderm development (Eed), mouse enhancer of zeste (Enx) 1, and Enx2, appears to be required to initiate repression of target gene expression in early development, whereas a second complex containing M33, mouse prohormone convertase 2 (Mpc2), rae28/Mph1 (mouse polyhomeotic 1), B-lymphoma MO-MLV insertion region-1 (Bmi1), and mouse posterior sex comb (Mel18) appears to be required for stable maintenance of the repression state. Gene disruption of individual components of the second PcG complex invariably resulted in defects of skeletal transformation of anterior-posterior identities because of aberrant Hox gene expression.^23-27  However, defects specific to each individual knock-out mouse were observed in a variety of tissues. For example, M33–knock-out (KO) mice were unique in showing hypoplastic gonad formation in both sexes and male-to-female sex reversal.²⁷  Moreover, these mice also possessed abnormally few nucleated cells in the spleen and thymus, probably because of aberrant T-cell expansion.²⁶ 

In the present study, we report that M33-KO mice develop smaller spleens with irregularly distributed red blood cells and that this abnormality is probably caused by defects in the splenic vascular structure. Moreover, the adrenal glands of the KO mice were smaller than those of the wild type. Interestingly, these defects were similar to those seen in the corresponding tissues of the homozygous and heterozygous Ad4BP/SF1-KO mice.^8,28  Investigation of the phenotypes of the 2 mouse strains strongly suggested that M33 is essential for Ad4BP/SF1 gene expression. With chromatin prepared from Y-1 adrenocortical cells, this functional correlation of the 2 genes was substantiated by chromatin immunoprecipitation (ChIP) assay using an antibody for M33.

M33- and Ad4BP/SF1-KO mice were generated as described previously.^27,29 M33-disrupted mice were backcrossed with C57BL/6NJcl females (Japan Clea, Tokyo, Japan) for more than 13 generations. Ad4BP/SF1-KO mice were maintained as a closed colony (129/Sv). Severe combined immunodeficient (SCID; Fox CHASE SCID C.B-17/Icr-scid Jcl) mice were purchased from Japan Clea. All protocols were approved by the Institutional Animal Care and Use Committee of the National Institute for Basic Biology.

Abdominal regions, including the spleen, were excised from embryos and immersed in 3% glutaraldehyde in HEPES (N-2-hydroxylpiperazine-N'-2-ethanesulfonic acid) buffer (10 mM HEPES-NaOH [pH 7.4] and 145 mM NaCl) for at least 2 hours. After fixation with 1% osmium tetroxide for 1 hour, the tissues were dehydrated and embedded in epoxy resin. Ultrathin sections were prepared, stained with uranyl acetate and lead citrate, and thereafter subjected to electron microscopic observation. Electron micrographs were taken with a JEM 1200EX electron microscope (JEOL, Tokyo, Japan) at 80 kV acceleration voltage. The diameter of the objective aperture was 300 μm. The fine-grain positive film (Eastman Kodak, Rochester, NY) was developed with D-19 developer (Eastman Kodak) at 20°C for 9 minutes. Images were processed with Adobe Photoshop (Adobe Systems, San Jose, CA).

Using peripheral blood obtained from fetuses 18.5 days after coitus, blood smears were prepared on glass slides coated with a 1:9 solution of 10 mg/mL acridine orange aqueous solution:ethanol.³⁰  Howell-Jolly bodies were detected under a fluorescent microscope. At least 1000 erythrocytes were examined in each sample.

cDNA encoding the C-terminal half of M33 (from residues 255 to 519) was cloned into pET28A to synthesize His-tagged M33(255-519), and the recombinant plasmid was transformed into BL21 cells. Transformants were cultured at 37°C for an appropriate amount of time, followed by further incubation with 1 mM IPTG (isopropyl-1-thio-β-D-galactoside) to induce protein expression. After recovery by centrifugation, Escherichia coli cells were sonicated in 20 mM Tris-HCl (tris-hydroxymethyl aminomethane, pH 8.0) containing 0.1 M NaCl. After the expressed M33(255-519) protein was recovered as inclusion bodies, it was suspended in 50 mM Tris-HCl (pH 8.0) containing 6 M guanidine and 0.2 M NaCl, and then applied on a Ni-NTA Super Flow column to purify His-tagged M33(255-519) according to protocol recommended by the manufacturer (QIAGEN, Valencia, CA). The purified fraction was subjected to stepwise dialyses to obtain soluble His-tagged M33(255-519) in the absence of denaturing reagents.³¹  In brief, it was dialyzed first against TNE (50 mM Tris-HCl [pH 8.0], 0.2 M NaCl, 1 mM EDTA [ethylenediaminetetraacetic acid]) containing 6 M guanidine for 4 hours, and then successively dialyzed again against TNE containing 6 M guanidine for 6 hours and against TNE containing 3 M guanidine for another 6 hours. Dialysis continued for 6 hours against TNE containing 375 μM β-mercaptoethanol, 0.4 M arginine, 375 μM oxidized glutathione, and 2 M guanidine, and then for another 12 hours with the same buffer containing 0.5 M guanidine. The protein solution was finally dialyzed against this same buffer without guanidine for 12 hours, yielding soluble His-tagged M33(255-519) without any denaturing reagents present. The protein was used to immunize guinea pigs, and antisera were prepared from them. The antibody was purified using an antigen column as described previously³¹  using cyanogen bromide (CNBr)–activated Sepharose (Amersham Biosciences, Piscataway, NJ).

Spleens from fetuses 16.5 days after coitus were lysed with 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, and 0.1 mM EDTA containing 1% sodium dodecyl sulfate (SDS). Resultant lysed tissue sample (25 μg) was subjected to SDS–polyacrylamide gel electrophoresis. The conditions for detection of Ad4BP/SF-1 were described previously.²⁹  For detection of M33, purified polyclonal antibody was used as the primary antibody for overnight incubation at 4°C. Anti–rabbit immunoglobulin G (IgG; Amersham Biosciences) and anti–guinea pig IgG conjugated with horseradish peroxidase (Dako Cytomation A/S, Copenhagen, Denmark) were used as the secondary antibodies. As a control, α-tubulin expression was examined with mouse monoclonal IgG (Sigma-Aldrich, St Louis, MO). The Electrochemiluminescence Plus (ECL+ Plus) Western Blotting Detection System (Amersham Biosciences) was used to detect immunoreactive signals. The signals were analyzed quantitatively by a LAS-1000, Image Gauge system (version 3.1; Fujifilm, Kanagawa, Japan).

Spleens were fixed with filtered 4% paraformaldehyde in phosphate-buffered saline (PBS) for 12 hours, dehydrated, and embedded in paraffin (Merck, Darmstadt, Germany). Sections were stained with the antibody against Ad4BP/SF-1. Biotinylated antibodies against rabbit IgG (Jackson Immnoresearch Laboratories, West Grove, PA) and guinea pig IgG (Vector Laboratories, Burlingame, CA) were used as the secondary antibodies. Immunoreactive signals were detected by streptavidin-conjugated peroxidase using a Histofine SAB-PO (M) kit (Nichirei, Tokyo, Japan). For frozen sections, the spleens were fixed with 1:1 acetone/methanol at –20°C and frozen in Tissue-Tek OCT compound (Sakura Finetechnical, Tokyo, Japan). Frozen sections were subjected to incubation with a rabbit polyclonal antibody against Type-IV collagen (Chemicon International, Temecula, CA), mouse monoclonal antibodies against vimentin (Boehringer Mannheim, Mannheim, Germany) and laminin (Chemicon International), and rat antibody against mouse F4/80 antigen (MCA497R; Serotec, Oxford, United Kingdom). Immunoreactive signals were detected by streptavidin-conjugated alkaline phosphatase using Histofine SAB-AP (M) and SAB-PO (M) kits (Nichirei) and NBT/BCIP (5-bromo-4-chloro-3-indoylphosphate p-toluidine salt; Roche Applied Science, Indianapolis, IN). Histochemical images were photographed with an Axioskop (Zeiss, Oberkochen, Germany) equipped with Plan-NEOFLUAR 20 ×/0.50 NA (Figure 1E-H), 10 ×/0.30 NA (Figure 3A-I), or 40 ×/0.75 NA (Figures 1I-L, 4B-D, 5D-E) objectives (Zeiss) using Photograb imaging solution and a Fuji HC-2500 digital camera (Fujifilm). Images were processed with Adobe Photoshop.

Total RNA was prepared from the spleens of fetuses 16.5 days after coitus using an Absolutely RNA reverse transcriptase–polymerase chain reaction (RT-PCR) Miniprep Kit (Stratagene, Cedar Creek, TX). First-strand cDNA was synthesized from 1 μg total RNA, and subsequently 1/20 of the cDNA was used for PCR reactions (KOD-Plus-; TOYOBO, Osaka, Japan) with primers for Ad4BP/SF-1 (5′-GTACGGCAAGGAAGACAGCAT-3′ and 5′-CCACCAGGCACAATAGCAACT-3′) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 5′-GGCATGGCCTTCCGTGTTCCT-3′ and 5′-TCCTTGCTGGGGTGGGTGGTC-3′). PCR conditions were as follows: 40 cycles of 15 seconds at 95°C, 30 seconds at 65°C, and 60 seconds at 68°C. Quantitative RT-PCR were performed as described previously.³² 

Y-1 mouse adrenocortical cells (4 × 10⁷) were crosslinked with 1/10 volume of fix solution (50 mM HEPES [pH 8.0]), 1 mM EDTA, 0.5 mM EGTA [ethylene glycol bis {beta-aminoethyl ether}-N, N, N′, N′-tetraacetic acid], 100 mM NaCl, 11% formaldehyde) for 10 minutes at room temperature, and the crosslinking was terminated by the addition of 1/20 volume of 2.5 M glycine for 10 minutes at 4°C. After the cells were washed 3 times with PBS, they were suspended with 50 mM HEPES (pH 7.6) containing 1 mM EDTA, 140 mM NaCl, 0.5% NP-40 (Nonidet P-40, [Octylphenoxy]polyethoxyethanol), 0.25% Triton-X100, and 10% glycerol, and incubated subsequently for 10 minutes at 4°C. After 5 minutes of centrifugation, the precipitate was suspended with 10 mM Tris-HCl (pH 8.0) containing 1 mM EDTA, 0.5 mM EGTA, and 200 mM NaCl, and incubated for 10 minutes at room temperature. The fixed cells were finally suspended with 10 mM Tris-HCl (pH 8.0) containing 1 mM EDTA and 0.5 mM EGTA, and thereafter sonicated 3 times for 5 minutes each at 30% Out Put setting (Sonifier Cell Disruptor; Branson, Dunbary, CT). After centrifugation of the solution for 10 minutes at 4°C, the supernatant was precleared for 2 hours at 4°C with 20 μL protein A Sepharose beads (Amersham Biosciences) slurry that had been prepared by washing 3 times with RIPA buffer (50 mM Tris-HCl [pH 7.5], 5 mM EDTA, 150 mM NaCl, 0.1% SDS, 0.5% NP-40). After centrifugation for 10 minutes at 4°C, 500 μL supernatant was subjected to overnight incubation at 4°C with 0.4 μg anti-M33 or control guinea pig IgG (Santa Cruz Biotechnology, Santa Cruz, CA) in IP buffer (10 mM Tris-HCl [pH 8.0], 1 mM EDTA, 0.5 mM EGTA, 150 mM NaCl, 0.03% SDS, 0.5% NP-40). The immunocomplexes were recovered by 2 hours of incubation with 20 μL protein A Sepharose beads. The precipitates were washed 8 times with 1 mL RIPA buffer and subsequently twice with 1 mL of 10 mM Tris-HCl (pH 7.5) and 1 mM EDTA. The precipitated chromatin complexes were treated with 50 μg DNase-free RNase (Nippon Gene, Tokyo, Japan) at 55°C for 1 hour, and thereafter treated with 40 μg proteinase K (Merck, Darmstadt, Germany) at 55°C for 2 hours. After crosslinking was reversed by overnight incubation at 65°C, DNA was extracted with phenol:chloroform (1:1), precipitated with ethanol, and finally subjected to PCR. The primer sets used in the PCR analyses were as follows: primer 1, 5′-GAGCACAAATGGAGTTTGTAACAAA-3′, 5′-CATCCTGGCTGGGCTAAGTG-3′; primer 2, 5′-GGTCACATTCCAGAGCAGCTC-3′, 5′-GCCTGAGGGAAGTGAACCCT-3′; primer 3, 5′-TTGTTTTAGTCCCTGTGGCCA-3′, 5′-GGGCGTGAGGTGGGTGTA-3′; primer 4, 5′-GGGCTCTATTGGGCTTCATTC-3′, 5′-CCACTGCACCCCTACTCCA-3′; primer 5, 5′-GCCCCTTGGCCTTTAGTCC-3′, 5′-CAAAGCACACAGGCTCACTCC-3′; primer 6, 5′-CCTGTGTGCTTTGTTTGCCTT-3′, 5′-TGTGGACA TGGGAATAGCGTTATAT-3′; primer 7, 5′-AGTGGTGGAGCACCTCTATGAGTT-3′, 5′-GCTGAAAGCTTGGCTGCTCT-3′; primer 8, 5′-AGGACAGAGCAGCCAAGCTTT-3′, 5′-CCACTGCCCAGCTAGAACCT-3′; primer 9, 5′-TGGTGTTTTCTGGTCCATTTCTT-3′,5′-CTTCGGTCAAGCCTCTTTAACC-3′; primer 10, 5′-GTAGAAGGAAGGAATGCACAGG-3′, 5′-CTGCAGATATGGTCCATTGG-3′; primer 11, 5′-CAGTCCATCTTGACATCTGTC-3′, 5′-CACGAGAAACTGCGTGTGTAAG-3′; primer 12, 5′-GTTATCTCTCCTTGGGCTTG-3′, 5′-CCACAGTGTGCATATAGAGG-3′; primer 13, 5′-CAGGGAACAGATAACAGCCTTG-3′, 5′-CAACTGGAGCACTAACTCTTGG-3′; primer 14, 5′-GTCACTGCTATTGGGACAAC-3′, 5′-GAATCAGGTTCTCTGCAACC-3′; primer 15, 5′-CCATGTCTATCTGCAGCTTC-3′, 5′-GTGTATGCGGTTATGTCTGG-3′; primer 16, 5′-GAGAGTCAGATGTCTTCAGG-3′, 5′-CAGGTGTTGATCAAGCTCAG-3′; primer 17, 5′-CATTACACTAGCTGGTCCCTTG-3′,5′-GATCTGTTGCTCTGAGAGGTAC-3′.

Previously generated M33-KO mice²⁷  were backcrossed with C57BL/6 animals for more than 13 generations. Because splenic defects became obvious during this backcross, we started to examine defects associated with genetic background. Because the M33-KO mice died immediately after birth, spleens were analyzed from animals in the late-fetal stage in this study. At 18.5 days after coitus, the spleens from M33-KO mice were quite smaller than those from their wild-type littermates (Figure 1A-B). As a control, the spleens of SCID mice, which show lymphoid cell defect, were examined (Figure 1C). The wet weights of the wild-type, KO, and SCID mouse spleens were 1.5 ± 0.4 mg (n = 3), 0.5 ± 0.2 mg (n = 4), and 1.6 ± 0.6 mg (n = 7), respectively, whereas the body weights of the wild-type, KO, and SCID mouse fetuses were 1.15 ± 0.12 g (n = 5), 0.98 ± 0.11 g (n = 5), and 1.14 ± 0.09 g (n = 7), respectively. Upon macroscopic observation, the wild-type and SCID mouse spleens appeared uniformly red in color, indicating that immature hematopoietic cells had colonized and erythrocyte production had begun throughout the spleens (Figure 1A,C). In contrast, the M33-KO mice developed spleens with scattered red spots (Figure 1B, arrowheads). In more severe cases, the KO spleens failed to develop any red spots. Histologic examination of the wild-type spleen showed densely packed myeloid and lymphoid cells at various stages of maturation, as well as well-assembled vascular structures. As indicated in Figure 1E and I, both eosin-stained erythrocytes and hematoxylin-stained lymphocytes were distributed throughout the entire spleen of the wild-type. In particular, lymphocytes were obviously packed around the peripheral vascular structures. In contrast, KO spleens were only sparsely packed with these myeloid and lymphoid cells. Furthermore, the spleens failed to develop well-assembled vascular structures and showed a biased distribution of erythrocytes and disorganized assemblies of lymphocytes (Figure 1F,J). The few areas containing densely packed erythrocytes corresponded to the macroscopically observed red spots (data not shown). Although the spleens of SCID mice contained fewer lymphoid cells, these organs showed the development of well-assembled vascular structures (Figure 1G,K) similar to the wild-type mice, indicating that the defects seen in the M33-KO spleen are independent from lymphatic defect.

Because the splenic phenotype of these M33-KO mice seemed similar to that of Ad4BP/SF1 knock-outs,⁸  the spleens of these 2 strains were compared. Like the M33-KO mice, the Ad4BP/SF1 knock-outs developed significantly smaller spleens and exhibited either a few red spots (Figure 1D), or none at all, corresponding to a sparse distribution of erythrocytes and lymphocytes (Figure 1H,L). The well-assembled vascular structure was rarely observed. The apparent phenotypic similarity between the M33- and Ad4BP/SF1-KO mice prompted us to compare the splenic defects in light of their detailed architecture and function.

Previously,⁸  we showed that Ad4BP/SF-1 immunoreactive cells begin to assemble vascular structures during early stages of splenic development, and these cells eventually develop into the splenic venous sinuses and pulp veins. Moreover, it was strongly suggested that organization of vascular structures was affected in the Ad4BP/SF1-KO spleen. Therefore, in the present study we investigated the vascular structures of the M33- and Ad4BP/SF1-KO spleens by transmission electron microscopy.

In a wild-type spleen at 16.5 days after coitus, the vascular structures were observed to be organized by single or multiple endothelial cells (Figure 2A,D). These endothelial cells usually contained well-developed cytoplasm and were attached to each other by tight junctions (Figure 2G). Mature vascular structures were covered by a basement membrane and were tightly attached to surrounding cells (Figure 2D). Compared with the wild type, the number of vascular structures per unit area was unlikely decreased in the M33- and Ad4BP/SF1-KO spleens (data not shown). As is the case in wild-type animals, vascular structures comprised single or multiple endothelial cells in both M33-KO (Figure 2B,E) and Ad4BP/SF1-KO (Figure 2C,F) spleens. Although the structures appeared morphologically quite different from their wild-type counterparts, they were judged to be vascular structures based on the presence of erythrocytes. The amount of cytoplasm of the endothelial cells lining the vascular structures in the M33-KO spleen was quite reduced (Figure 2B,E), and in particular certain cytoplasmic regions were strikingly thin, as indicated in Figure 2B and E (enlarged in Figure 2I). This particular endothelial defect was also observed in the Ad4BP/SF1-KO spleens (Figure 2C). In Figure 2H, the endothelial cytoplasm can be seen clearly to be only approximately 15- to 30-nm thick. In the most severe cases, the vascular structure was completely ruptured (Figure 2F,J), with exposure of residual cytoplasmic parts and collagen fibers that are localized normally along the basement membrane. The vascular structure of the stomach, where Ad4BP/SF-1 is never expressed, appeared normal at 16.5 days after coitus in both mutants (data not shown). Because defects in endothelial cells are often due to or accompanied by defects in perivascular cells,^33-35  the structures of perivascular cells in the 2 KO spleens were compared with those in the wild-type spleens. As shown in Figure 2K, L, and M, the perivascular cells were present around the endothelial cells even in the M33- and Ad4BP/SF1-KO spleens. Careful observation indicated that the structure of the perivascular cells was unlikely affected in the M33- and Ad4BP/SF1-KO spleens.

In another effort to characterize these defects, the M33- and Ad4BP/SF1-KO spleens were stained with antibodies against type-IV collagen, a marker for extracellular matrix. In the wild-type spleens of fetuses 18.5 days after coitus, abundant fibrous and trabecular structures were observed throughout the whole spleen (Figure 3A). Although these fibrous structures were still evident in the M33- and Ad4BP/SF1-KO spleens, trabecular structures were rarely observed (Figure 3B,C). Striking differences were seen when the spleens were immunostained with an antibody to laminin, another extracellular-matrix protein. Large laminin-expressing vascular structures were detected in wild-type spleens (Figure 3D), but they were never seen in either of the knock-out spleens (Figure 3E,F). In addition, cells assembled around the vascular structure were labeled in the wild-type spleen, whereas no such cellular assemblies were seen in the knock-outs, although a small number of immunoreactive cells were still present sporadically. Similar to laminin staining, antibodies to vimentin, an intermediate filament, stained cells assembling around the vascular structure (Figure 3D,G), whereas unassembled and decreased numbers of the immunoreactive cells for vimentin were observed in the spleens of both mutants (Figure 3H,I). It was reported that vimentin is expressed in all developmental stages of T lymphocytes, monocytes, and granulocytes; in erythroid precursor cells; and in immature B lymphocytes.³⁶  In addition, endothelial and perivascular cells were characterized as immunoreactive cells for vimentin.^37-39  Considering our immunohistochemical observations together with histologic and morphologic studies, the decrease of vimentin-immunoreactive cells is primarily due to the decrease of the blood cells. However, these results do not exclude a possibility that loss of vimentin expression occurs in the splenic endothelial and perivascular cells.

Venous sinuses, which are lined by Ad4BP/SF-1 immunoreactive cells, are involved in clearance of abnormal blood cells. By counting the number of circulating erythrocytes containing Howell-Jolly bodies, a previous study found that Ad4BP/SF1-KO spleens exhibited dysfunctional clearance of abnormal erythrocytes. We used this technique to assess the function of the M33-KO spleen. When compared with the wild type, the number of the abnormal erythrocytes counted in peripheral blood smears was increased by 2.3-fold in M33-KO mice, suggesting that the M33-KO spleens were impaired in their ability to clear abnormal blood cells (Figure 4A). To provide a direct assessment of the phagocytotic activities of the KO spleens, we examined the distribution of phagocytes using F4/80 as the marker. In the wild-type spleen at 18.5 days after coitus, the F4/80-positive phagocytes assembled around the vessel-like structures (Figure 4B). The F4/80-positive phagocytes were present even in the M33- and Ad4BP/SF1-KO spleens, although the numbers of the cells were smaller than those in the wild type (Figure 4C,D). Moreover, when considering that the M33- and Ad4BP/SF1-KO spleens are quite smaller than the wild type, the total number of phagocytes should be quite smaller in the KO spleens. Therefore, the impaired ability to clear abnormal blood cells seemed to be caused by the decreased number of the phagocytes together with the small spleen.

To elucidate the molecular mechanisms underlying the phenotypic similarities between the spleens from M33- and Ad4BP/SF1-KO mice, we examined the expression levels of M33 and Ad4BP/SF-1 in the spleen. Western blot analysis using the antibody against M33 demonstrated the absence of M33 expression in the M33-KO mouse spleen (Figure 5A, top). Splenic Ad4BP/SF-1 expression was also clearly reduced in these mice compared with wild-type littermates (Figure 5A, middle), whereas α-tubulin was expressed invariably (Figure 5A, bottom). Quantification of the Western blot signals indicated that the level of Ad4BP/SF-1 was decreased significantly in homozygote M33 knock-outs and unchanged in heterozygote animals (Figure 5B). To determine whether the decrease in Ad4BP/SF-1 is due to a decrease in transcription, semiquantitative RT-PCR was performed. As with its protein product, we found that Ad4BP/SF1 mRNA level was significantly decreased in the M33-KO spleen, whereas this decrease was not detected in heterozygous littermates (Figure 5C). Expression of the positive control, GAPDH, was unaffected.

As reported previously⁸  and confirmed here, in wild-type animals, Ad4BP/SF-1 was expressed around 16.5 days after coitus in cells that were constructing vascular structures in the spleen, including primitive venous sinuses and pulp veins (Figure 5D). In contrast, even though the M33-KO spleen contained vascular structures, there were no Ad4BP/SF-1–immunoreactive cells present (Figure 5D, left panel). The expression of M33 in the Ad4BP/SF1-KO spleens was examined using the antibody against M33. Interestingly, as was seen in wild-type animals, Ad4BP/SF1-KO spleens showed M33 expression in cells constructing vascular structures, as well as in hematopoietic cells (Figure 5E). Moreover, Western blot analyses revealed that the amount of M33 expressed was unaffected in the Ad4BP/SF-1 KO spleen (data not shown).

Similar to the spleen, the adrenal gland was affected in both M33- and Ad4BP/SF1-KO mice. The adrenal gland of M33-KO mouse fetus was smaller than that of the wild type (Figure 6A), whereas, as previously shown by Bland et al,²⁸  the adrenal gland of Ad4BP/SF1 heterozygous KO mouse was also smaller than that of the wild type (Figure 6A). Therefore, the expression of Ad4BP/SF-1 in the adrenal gland of M33-KO mouse was examined by Western blot analysis. As shown in Figure 6B, the expression of Ad4BP/SF-1 was decreased in the M33-KO fetal adrenal gland but not in the heterozygote animals, whereas the expression of α-tubulin was not different. Quantification of the Western blot signals confirmed the low level of Ad4BP/SF-1 expression in the M33-KO fetal adrenal gland (Figure 6C). Showing a good correlation, quantitative RT-PCR indicated that the amount of mRNA for Ad4BP/SF-1 in the M33-KO fetal adrenal gland was decreased by approximately 50% compared with the wild-type or heterozygous M33-KO mice (Figure 6D). In contrast, the expression of M33 was never altered in the Ad4BP/SF1 heterozygous KO mouse (data not shown). Together with the splenic defects, the phenotypes of the M33- and Ad4BP/SF1-KO mice strongly suggested that M33 regulates Ad4BP/SF1 gene expression directly or indirectly.

On the basis of the observations in Figure 6, we next investigated whether M33-containing PcG complex binds directly to Ad4BP/SF1 gene locus, by using the ChIP assay. As a chromatin source, we examined several cell lines such as endothelial MSS31 and MSS62 cells and adrenocortical Y-1 cells to check whether M33 and Ad4BP/SF-1 are coexpressed in the cells. Although Ad4BP/SF-1 was not expressed in the endothelial cells, both genes were coexpressed in Y-1 cell (data not shown). Therefore, the chromatin was prepared from Y-1 cells and subjected to ChIP assays with anti-M33 antibody. With respect to the genomic structure, the mouse genome project revealed that the Ad4BP/SF1 gene is localized approximately 10 kb (kilobases) downstream of the last exon of Gcnf (germ cell nuclear factor; NR6a1) and 3 kb downstream of the last exon of the predicted gene, Gpr144 (Figure 7). Our recent investigation around the Ad4BP/SF1 gene locus showed that 3 DNase I hypersensitive sites (adHS1 to adHS3), a nuclear matrix attachment region (MAR), and Ctcf (an insulator binding protein) binding sites are localized at an intergenic region between Ad4BP/SF1 and Gcnf genes⁴⁰  (Figure 7). Moreover, a discontinuous pattern of histone H3 and H4 acetylation was observed in this region.

With special attention to these observations, we prepared 17 sets of primers for ChIP assays to cover the whole Ad4BP/SF1 gene locus. Interestingly, regions 4 and 5 placed at the adjacent upstream of adHS1 and placed to overlap with the Ctcf binding region, respectively, were recovered by anti-M33 antibody but not by control IgG, indicating that M33-containing PcG complex binds to these 2 regions. Region 3 localized at 600 bp (base pair) upstream of region 4, and regions 1 and 2 localized at 2.5 kb upstream were not recovered. Similarly, the M33-containing PcG complex did not bind to regions 6, 7, and 8 localized downstream of region 5. Region 9 at the 3′ region of the MAR was recovered even though less effectively, whereas region 10 localized at 500 bp upstream of the first exon of Ad4BP/SF1 gene was recovered. The M33-containing PcG complex failed to bind to regions 11 and 12 localized at the 4th intron, region 13 at the 5th intron, and regions 14 and 15 at the 6th intron. When examining regions downstream of the last exon of the Ad4BP/SF1 gene, M33-containing PcG complex binds to region 16 localized adjacent downstream of the last exon, whereas it did not bind to region 17 localized at the 19th intron of Gpr144 gene.

Knockout of the Hox11,¹ Bapx1,^2,3 Wt1,⁴ Nkx2-3,^5,6 Ad4BP/SF1, or capsulin⁷  gene gave rise to mice with splenic defects, and investigation of these mice revealed the roles of these genes in splenic development. Consistent with the early onset of expression of Hox11, Bapx1, Wt1, and capsulin, the corresponding gene disruptions resulted in crucial defects, beginning at the early stages of tissue development. Consequently, no spleen developed in these knock-out mice. In the cases of the Ad4BP/SF1 and Nkx2-3 genes, the homozygous KO mice developed spleens displaying structural abnormalities, although these defects were nonidentical. Considering the distinct types of developmental splenic defects, these genes were revealed to have crucial functions at certain steps of the splenic development. Thus, to understand the mechanisms of the splenic development, a gene cascade composed of these genes should be uncovered. In this regard, it was shown so far that Hox11 is the upstream gene for Wt1, based on the observation that the expression of Wt1 disappeared from the Hox11-KO splenic primordium.⁴¹  In addition, reporter gene analyses revealed that Wt1 gene transcription was activated by Hox11.⁴¹  However, the functional correlation between the other genes remains to be elucidated.

In the present study, we investigated splenic defects in M33-KO mice and found that they are strikingly similar to those seen in Ad4BP/SF1-KO spleens, especially those involving the structure of the vascular endothelial cells. The cytoplasm of the endothelial cells was thin, and in the most severe cases the vascular structures were ruptured, indicating that the closed circuit of the vascular systems had collapsed. Accordingly, the aberrant vascular system should not be able to deliver blood cells to the entire spleen, resulting in the development of totally white spleens. Moreover, hemorrhage probably occurred at the sites of vascular rupture, which resulted in the red spots seen in both knock-out spleens. We also found functional defects in both spleens. Clearance of abnormal blood cells is mediated by spleen-specific structures, the venous sinuses. In a previous study, we showed that the cells comprising the venous sinuses were immunoreactive for Ad4BP/SF-1, and thus, we predicted that Ad4BP/SF1-KO spleens would be impaired in their ability to clear abnormal blood cells. Indeed, analysis of peripheral blood smears revealed an increased number of abnormal blood cells containing Howell-Jolly bodies in both Ad4BP/SF1- and M33-KO mice. In addition to the spleen, we found that like Ad4BP/SF1 heterozygous mutant mice,²⁸ M33-KO mice developed smaller adrenal glands compared with wild-type animals. Moreover, previous studies indicated that both of these knock-out mice exhibited defects in the gonads, although unlike the spleen, the defects were different between the 2 strains. Taken together, the overlap in phenotypes in multiple tissues strongly suggests that Ad4BP/SF1 is a downstream gene of M33. This hypothesis was verified by Western blot, immunohistochemistry, and RT-PCR, which all showed that Ad4BP/SF-1 expression was largely down-regulated in the splenic vascular endothelial cells and adrenal glands of the M33-KO mice.

Upstream stimulatory factor 1 (USF-1) and USF-2,^42-46  Sry-type HMG box-9 (Sox9),⁴⁷  and Wilms tumor 1 (Wt1; –KTS isoforms)⁴⁸  have all been identified as direct-acting positive regulators of Ad4BP/SF1 gene transcription. These factors bind recognition sites in the 5′ upstream region of the Ad4BP/SF1 gene and thereby activate gene transcription. However, the present genetic analyses could not distinguish whether Ad4BP/SF1 is regulated directly or indirectly by M33. Because M33 has no DNA-binding activity by itself, the activity of PcG complex containing M33 as a component should be considered. In this regard, it is noteworthy that a PRE/TRE has been defined by transgenic and ChIP assays in Drosophila. Alternatively, a computer algorithm was developed recently to search for PRE/TREs in the Drosophila⁴⁹  genome. However, as described in the report, the program does not always find all PRE/TREs in Drosophila genome, and furthermore it is unclear whether the program is applicable to genomes of other animals.

Therefore, we tested whether M33 directly binds to Ad4BP/SF1 gene locus by using ChIP assay using an antibody to M33. Interestingly, assays using Y-1 adrenocortical cells revealed that the target sites of M33-containing PcG complex are present at the upstream region of the first exon and the immediately downstream region of the last exon of Ad4BP/SF1 gene. Importantly, our recent study demonstrated that the intergenic region between Ad4BP/SF1 and Gcnf genes contains functional architectures such as DNase I hypersensitive sites, a nuclear matrix attachment region (MAR), and Ctcf (an insulator binding protein) binding sites.⁴⁰  Moreover, the ChIP assays showed a discontinuous pattern of histone H3 and H4 acetylation over the intergenic region. These observations strongly suggested that this region forms a boundary, so-called insulator, between the 2 transcriptional units, Ad4BP/SF1 and Gcnf. This assumption was consistent with previous observation that Ctcf and nuclear matrix are required for the insulator activity.⁵⁰  In this regard, it seemed interesting to examine whether M33-containing PcG complex binds to this intergenic region because PcG complex is known to function in forming inactive chromatin arrays.^9,10  Confirming our expectation, the ChIP assays demonstrated that the M33-containing PcG complex binds to sites adjacent to and overlapped with the Ctcf binding region and MAR, respectively. Similarly, it is noteworthy that the M33-containing PcG complex bound to the region adjacent to the last exon of Ad4BP/SF1 gene. Although further examination is required, it is likely that M33-containing PcG complex is implicated in the formation of intergenic boundary in Y-1 cells.

This presumptive function of PcG complex is supported by previous studies in Drosophila. Mihaly et al¹³  showed that a cis-regulatory region, Fab-7, consists of a boundary element and PRE, both of which are implicated in the regulation of parasegmental-specific expression of bithorax genes. In the case of gypsy insulator, Gerasimova and Corces⁵¹  demonstrated that a gypsy insulator binding protein (mod(mdg4)) requires a PcG component (Polycomb) for the insulator function. Considering these observations, the present study strongly suggested that the PcG complex together with Ctcf and nuclear matrix form intergenic boundary between Ad4BP/SF1 and Gcnf genes, although the functional and physical correlations between the PcG complex and Ctcf and/or nuclear matrix remains to be elucidated.

There is a general agreement that PcG complex binds directly to target gene loci to keep silent chromatin conformation. Nevertheless, in the present study, multiple binding regions of the M33-containing PcG complex were found to lie at the 5′ boundary region of the transcriptionally active Ad4BP/SF1 gene. Although our findings seem contradictory to the accepted notion, the results of recent ChIP analyses using antibodies to PcG components are similar to those of the present study, showing that the PcG complex binds to not only inactive but also active gene loci.^52,53  Accordingly, it is reasonable to assume that PcG complex is involved in the formation of transcriptionally active chromatin unit.

Participation of PcG members in hematopoiesis has been elucidated by gene disruption studies. Mice lacking any one of the components of the PcG complex (M33, Rae28/Mph1, Mel18, and Bmi1) showed severe hypoplasia of the spleen and thymus because of a reduction in nucleated cell number in the weaning period.^23,26,54,55  The reduction of hematopoietic cells reflects, at least in part, an impaired mitotic response of lymphocyte precursor cells to stimuli such as interleukin 7 (IL-7).^23,55,56  As is the case for other PcG members, M33 transcripts were detected in hematopoietic cells purified from human bone marrow cells.⁵⁷  Furthermore, M33-KO mice during the weaning period also showed hypoplasia of the spleen and thymus resulting from a reduction of nucleated cell number. Hematopoietic progenitor cells from M33-KO fetal liver may have failed to induce expansion of mature T cells in recolonized alymphoid thymic lobes; indeed, splenic cells from 3-week-old M33-KO mice failed to respond to lipopolysaccharide (LPS) activation in culture.²⁶  Thus, in the fetal spleen of M33-knockouts, hematopoietic-cell proliferation could not be excluded. However, when considering the phenotypic similarity between the M33-and Ad4BP/SF1-KO spleens, it is highly likely that the altered framework due to the vascular endothelial defects, rather than the hematopoietic cell defects, underlay in the splenic phenotype in M33-KO animals.