A Pbx1-dependent genetic and transcriptional network regulates spleen ontogeny

In vertebrates, the spleen is a lymphoid organ that serves important roles in hematopoiesis and the generation of primary immune responses, as well as acting as a filter that removes and processes aged and abnormal blood cells(VanRooijen et al., 1989; Zapata and Cooper, 1990). During development, the spleen originates from splanchnic mesoderm (mesenchyme that surrounds the gut endoderm), which arises from splitting of the lateral plate mesoderm into somatic mesoderm (body wall) and splanchnic mesoderm(Funayama et al., 1999). The latter undergoes gut-specific mesodermal differentiation(Apelqvist et al., 1997; Ramalho-Santos et al., 2000)in response to signals from the endoderm that gives rise to the epithelium of the gut and associated organs. Histological studies using frog, chick and mammalian embryos indicate that development of the splenic anlage is first detectable at approximately gestational day (E) 11, as progenitor cells derived from splanchnic mesoderm form a single condensation within the dorsal mesogastrium (Dm; the mesenchymal sheet that attaches the stomach to the dorsal body wall), adjacent to the stomach and dorsal pancreas(Thiel and Downey, 1921; Manning and Horton, 1969; Sty and Conway, 1985; Vellguth et al., 1985; Yassine et al., 1989).

In vertebrates, the mesodermally derived spleen normally displays left-handed asymmetry (Boorman and Shimeld, 2002) and has been considered to be a landmark organ for detecting laterality defects (Aylsworth,2001). But, interestingly, mice with left-right (LR) asymmetry defects, such as the Inv/Inv(Yokoyama et al., 1993) and ActRIIB (Oh et al., 2002) mutant mice exhibit either normally developed spleens or, infrequently, splenic hypoplasia. Furthermore, asplenic mouse models, such as the Dh spontaneous mutant(Green, 1967) and the Nkx3.2 (Bapx1 - Mouse Genome Informatics) mutant mouse(Lettice et al., 1999;Tribioli et al., 1999), appear to exhibit only regional perturbations of LR asymmetry in the primordial splenopancreatic mesoderm(Hecksher-Sorensen et al.,2004). Other asplenic mouse models, such as those mutant for Hox11 (Tlx1 - Mouse Genome Informatics)(Roberts et al., 1994; Dear et al., 1995), display asplenia as the sole organ abnormality. Likewise, in humans, asplenia may present as the sole organ anomaly, without perturbations of LR asymmetry(Rose et al., 1975; Waldman et al., 1977). Overall, these findings underscore the notion that, both in mice and humans,mechanisms other than the regulation of LR asymmetry must be responsible for the control of splenic cell fate specification and morphogenesis.

Recent advances in mouse genetics have led to the discovery of novel genes required for early spleen ontogeny. These include Hox11(Roberts et al., 1994; Dear et al., 1995; Kanzler and Dear, 2001),Nkx3.2 (Lettice et al.,1999; Tribioli et al., 1999), Pod1(Quaggin et al., 1999; Lu et al., 2000) and Wt1 (Herzer et al.,1999); however, the hierarchical relationships among these genes remain unknown. This limited collection of genes also includes Pbx1(Nourse et al., 1990; Kamps et al., 1990), which encodes a TALE class (Burglin,1997) homeodomain protein, the absence of which results in embryonic asplenia with 100% penetrance(Selleri et al., 2001). Although the role of Pbx1 in spleen development is undefined, its reported biochemical in vitro interaction with homeodomain protein Hox11 through the hexapeptide motif (Shen et al.,1996) raises the possibility that these two homeoproteins may cooperate in spleen ontogeny, as Hox11 is also required for spleen formation(Roberts et al., 1994; Dear et al., 1995).

Heterodimers of Pbx and other TALE proteins of the Meinox family, such as Meis (Bischof et al., 1998; Chang et al., 1997) and Pknox/Prep1 (Berthelsen et al.,1998; Knoepfler et al.,1997; Fognani et al.,2002), form stable nuclear complexes, and biochemical analyses suggest that these complexes regulate several genes(Swift et al., 1998). Indeed,we have found that loss of Pbx1 causes multiple organogenesis defects in the mouse and lethality in utero at E15.5(Selleri et al., 2001). These defects include abnormalities in patterning and development of the skeleton(Selleri et al., 2001), in pancreas morphogenesis and function (Kim and Selleri et al., 2002), in adrenal/urogenital development(Schnabel et al., 2003a; Schnabel et al., 2003b), and in caudal pharyngeal pouch-derived organ formation and patterning(Manley and Selleri et al.,2004), as well as impaired hematopoiesis(DiMartino and Selleri et al.,2001). These findings underscore the notion that Pbx1 serves as a key developmental regulator, although the crucial genetic and transcriptional pathways underlying its specific developmental roles have not been established.

In this report, we investigated spleen ontogeny by analyzing asplenic mouse models lacking Pbx1, Hox11, Nkx3.2 or Pod1 (capsulin). Our studies define a genetic hierarchy in which Pbx1 serves a central and crucial role as a common co-regulator in spleen ontogeny.

Asplenic knockout mice for Hox11(Roberts et al., 1994), Hox11^lacZ (lacZ insertion into the Hox11locus) (Dear et al., 1995), Pbx1 (Selleri et al.,2001) and Pod1(Quaggin et al., 1999) have been described and were genotyped according to previously published protocols. Nkx3.2 (Bapx1) asplenic knockout mice were generated and kindly provided by Drs W. Zimmer and R. Schwartz (unpublished). Pbx2knockout mice, which develop normally, were genotyped as described previously(Selleri et al., 2004).

Embryos from E10.5 to E15.5 were harvested and fixed overnight at 4°C in phosphate-buffered saline (PBS) containing 4% (w/v) paraformaldehyde (PFA). For histological analysis and immunohistochemistry with a mouse anti-Pbx1b primary antibody (Jacobs et al.,1999), protocols were followed as described(Selleri et al., 2001). Single-stranded sense and antisense riboprobes for in situ hybridization on frozen sections were specific for Hox11(Dear et al., 1995),Nkx3.2 (Tribioli et al.,1997), Nkx2.5 (Lyons et al., 1995), Pbx1 (3′ UTR), Pod1(Quaggin et al., 1998) and Wt1 (Herzer et al.,1999).

Embryos heterozygous for Hox11^lacZ were collected at E11.5 and E12.5 and stained for β-galactosidase as described previously(Dear et al., 1995).

Mice 6-8 weeks old were immunized by intravenous (i.v.) injection of 5×10⁷ sheep red blood cells (SRBC) in PBS. Mice were sacrificed and the spleens processed for immunohistochemistry with biotinylated lectin peanut agglutinin (PNA; Vector Laboratories, Burlingame,CA) as described (Inada et al.,1998).

Pregnant Pbx1^+/- and Hox11^+/-female mice, carrying embryos at E12.5 and E13.5, respectively, were injected intraperitoneally with 5-bromo-2-deoxy-uridine (BrdU) (50 μg/g of body weight) and BrdU incorporation was assayed as previously described(Selleri et al., 2001). The number of BrdU-positive cells (dark brown nuclei) within the developing spleen were counted in six to eight sagittal sections (10 μm thickness) for each genotype. Quantitative analysis of BrdU immunoperoxidase-stained sections was made on a Nikon microscope equipped with a video camera. TUNEL assays were performed as described by Gavrieli et al.(Gavrieli et al., 1992).

Embryonic spleens at day E16-17 were dissected and trypsinized (0.25% final concentration) for 10-15 minutes at 37°C. The cell suspension was washed twice and cultured in Dulbecco's Modified Eagle Medium (D-MEM), supplemented with 10% fetal calf serum (Celbio), 2 mM L-glutamine (Invitrogen), 100 U/ml penicillin and 100 mg/ml streptomycin (Invitrogen) in humidified 5%CO₂, and used as a primary cell culture for chromatin immunoprecipitation (ChIP) assays. Primary spleen cultures were immunophenotyped by using an α-smooth muscle actin Ab (ASMA; Santa Cruz Biotech), which stains mesoderm-derived cells. To generate immortalized spleen embryonic cell lines from Pbx2^-/- embryos, the NIH 3T3 protocol (Todaro et al., 1963) was used.

Western blot analysis was performed as described previously(Berthelsen et al., 1996; Jacobs et al., 1999). The following antibodies were used: anti-Hox11 (1:1,000) (Santa Cruz Biotech, CA),anti-Pbx1b (1:1,500) (Jacobs et al.,1999) and anti-Prep 1 (1:1,500; Upstate Biotechnology).

NIH 3T3 cells were cultured in D-MEM supplemented with 5% fetal calf serum and 5% delipidated fetal calf serum. Transient transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. For Hox11 promoter analysis, the following constructs were used: a luciferase reporter vector (pGL2); a pGL2 construct carrying an EcoRI-SalI 0.9 kb fragment, corresponding to the Hox11 promoter region (pGL2-540)(Arai et al., 1997); pcDNA3 constructs containing the cDNA of Pbx1a or Prep1(Berthelsen et al., 1998) and a pBlueScript-Hox11 construct containing the full-length cDNA of Hox11 (obtained from Dr N. Hoehler) (Koehler et al.,2000). Cells were lysed 40-45 hours after transfection and assayed for luciferase activity (Benasciutti et al., 2004). Values were normalized for β-gal activity. Data represent means of triplicate values from a representative experiment. All transfections were independently performed three times.

Electrophoretic mobility shift assays (EMSA) were performed as described(Berthelsen et al., 1996; Jacobs et al., 1999) using nuclear extracts from spleen cells. The following oligonucleotides, spanning Pbx1-binding sites within the Hox11 promoter (AB 000681)(Arai et al., 1997), were employed in EMSA reactions: 5′-CCAAAGGCTTGTGACTGCTTTTCAGG-3′ PX1 and 5′-CCAAAGGCTTGTCGACGCTTTTCAGG-3′ PX1 mutated.

The following antibodies were used: rabbit polyclonal antibodies specific for Pbx1 (P-20) and Hox11 (C-18; Santa Cruz Biotech, CA).

Formaldehyde crosslinking and chromatin immunoprecipitation were performed as described (Frank et al.,2001). Samples were immunoprecipitated overnight at 4°C with the following antibodies: mouse monoclonal specific for Pbx1b(Jacobs et al., 1999), rabbit polyclonal specific for Hox11 (Santa Cruz Biotech, CA), mouse monoclonal anti-Green Fluorescent Protein (anti-GFP; Santa Cruz Biotech, CA) and normal rabbit serum (Covenge Research). Immune complexes were recovered by adding 30μl of blocked protein G beads and incubated for 2 hours at 4°C. Beads were washed and eluted, and crosslinks were reversed as described (Aparicio et al., 1999). The eluted DNA was resuspended in 30-60 μl water. A region within the Hox11 promoter containing the Pbx1-binding sites (-359 s/-115 as and -258 s/-54 as) and a control region within the same promoter(-1170 s/-891 as) were amplified by PCR using specific primer pairs. One primer pair that amplifies a region of the Bmp4 promoter was also used as additional negative control: 5′-ACGCACTTCCCTGATTCTCGTC-3′(-359 s) and 5′-AGCAGTCACAAGCCTTTGGATTAC-3′ (-115 as) product size 244 bp; 5′-TCTCACAAACCCAGAGCCATTC-3′ (-258 s) and 5′-TAGCAGCCACTCCAACTCAGTCTC-3′ (-54 as) product size 204 bp;5′-TGAGAACAACTACCTGCTTCGTGC-3′ (-1170 s) and 5′-TGGAGACTTGACTTGCCCAACC-3′ (-891 as) product size 279 bp; and 5′-AATGAACAAACACCACTCTCCCTC (Bmp4 s) and 5′-AACACCAGACCGAAAAGATGACTG (Bmp4 as) product size 350 bp.

Analysis of Pbx1^-/- embryos at E15.5 demonstrated that they are asplenic (Fig. 1B,D),compared with wild-type controls, which exhibited a fully developed ribbon-shaped spleen (Fig. 1A,C). The defect in spleen development was evident at E13.5 because of the absence of the splenic anlage(Fig. 1F), which is easily recognizable in close proximity to the stomach on histological sections of wild-type embryos (Fig. 1E). Earlier in development, at E11-12.5, histological and in situ analyses(Fig. 2) indicated that the splenic primordium, which starts to be recognizable as a condensation of mesenchymal cells within the normal Dm, lateral to the stomach(Thiel and Downey, 1921), was hypoplastic in Pbx1^-/- embryos(Fig. 2). At even earlier stages of organogenesis (E10-10.5), however, the differentiation of the splanchnic mesoderm into the epithelial-like plate of cells that is well distinguished from the surrounding unorganized mesenchyme(Green, 1967), was completely preserved in Pbx1^-/- embryos(Fig. 1G-J). In fact, at this developmental stage, the general architecture of the splanchnic mesoderm that surrounds the stomach enlargement (Fig. 1G-J) and midgut primordia (not shown) within the coelomic cavity did not appear morphologically abnormal in Pbx1^-/-embryos. This contrasts with other cases of murine asplenia, which exhibit marked abnormalities of the splanchnic mesoderm, such as that documented in the Dh mutant, which lacks the epithelial-like plate of cells at E10-10.5 (Green, 1967; Hecksher-Sorensen et al.,2004).

Visualization of gene transcripts known to be present in the condensing splenic mesenchyme within the Dm was conducted to trace spleen ontogeny at early embryonic stages (E11-12.5) and to establish hierarchical requirements for specific transcription factors in splenic cell fate specification and morphogenesis. Using in situ hybridization, we assessed expression of Hox11 (Koehler et al.,2000; Kanzler and Dear 2001) (Fig. 2A-D)and Nkx2.5 (Patterson et al.,2000; Hecksher-Sorensen et al., 2004), which are first observed in Dm mesenchyme between E10.5 and E11 and are regarded as the earliest known markers for splenic cell fate.

By E11, Nkx2.5 is normally detectable in a domain of the murine Dm(Figs 2 and 3) that overlaps with the expression of Hox11, lateral to the stomach(Hecksher-Sorensen et al.,2004). In Pbx1^-/- embryos, however, Hox11 and Nkx2.5 were absent in the condensing splenic mesenchyme of the Dm from E11 to E12.5 (Hox11, Fig. 2A-D; Nkx2.5, Fig. 3K,L; analysis of Nkx2.5 at E11-11.5 is not shown). Hox11 expression was not detected up to E13.5 by in situ hybridization in Dm condensing splenic mesenchyme of Pbx1^-/- embryos (not shown), indicating a complete absence of Hox11, rather than a temporal delay of its expression in the spleen anlage. Thus, Hox11 expression is completely absent in splenic condensing mesenchyme and later in the spleen anlage. Conversely, Hox11 expression was unperturbed in other organ primordia, such as the developing pancreas at E12.5 and the branchial arches(not shown). By contrast, no alterations of Pbx1 expression were observed in the Dm splenic condensing mesenchyme of Hox11^-/- embryos (Fig. 3A,B). Taken together, these results indicate that the onset and continued expression of Hox11 in the splenic anlage, unlike in non-splenic domains, is Pbx1 dependent, whereas Pbx1expression is not dependent on the presence of Hox11 protein. Furthermore, our findings also suggest that, in the absence of Pbx1, splenic cell fate specification is, at least in part, compromised, as two of the earliest known markers of splenic cell fate, Nkx2.5 and Hox11, are absent in Dm condensing mesenchyme. Conversely, expression of other genes that mark splenic condensing mesenchyme, such as Pod1(Fig. 2I-L) and Nkx3.2(Fig. 2M-P) is maintained in Pbx1^-/- embryos, demonstrating that condensing mesenchymal cells are still present in Dm of Pbx1^-/- embryos and that splenic gene expression is not globally impaired by Pbx1 loss.

Wt1 is expressed during early spleen development in the mouse(Rackely et al., 1993) and is required for spleen development(Herzer et al., 1999). It is normally expressed in a domain of the Dm that, by E11, overlaps with that of Hox11 within the splenic condensing mesenchyme lateral to the stomach(Fig. 2E,G). Wt1 is regulated by Hox11 in spleen development, as its expression is diminished in Hox11^-/- embryos (Koehler, 2000) (data not shown). In addition, Wt1, like Pbx1, is highly expressed in the outer mesothelial lining of the splenic anlage(Fig. 2E,G) that will give rise to the splenic capsule (Sadler,1995). By contrast, Hox11 is not expressed in the mesothelial lining of the developing spleen(Fig. 4A, arrowhead).

Expression of Wt1 was severely down regulated in the condensing mesenchyme of Pbx1^-/- Dm, when compared with wild-type littermates. However, its expression was well maintained in the outer mesothelial lining of the Pbx1^-/- splenic primordium, both at E11-11.5 and E12-12.5 (Fig. 2E-H). As Hox11 is absent in Pbx1^-/-embryos, and Wt1 is regulated by Hox11 (Koehler, 2000) (data not shown), these findings suggest a genetic hierarchy whereby Pbx1regulates Hox11, which in turn regulates Wt1 in condensing splenic mesenchyme. Nevertheless, while Wt1 expression in Dm splenic mesenchyme is hierarchically dependent upon Pbx1, probably through Hox11, its expression is independent of Pbx1 in the developing splenic capsule.

Expression of Pod1 (Robb et al., 1998; Lu et al.,2000) (Fig. 2I-L)and Nkx3.2 (Fig. 2M-P)overlaps with Hox11 and Nkx2.5 in the condensing splenic mesenchyme. In addition, gene targeting studies have shown that Pod1(Quaggin et al., 1999; Lu et al., 2000) and Nkx3.2 (Lettice et al.,1999; Tribioli et al., 1999) are required for spleen development. Pod1 (Fig. 2I-L) and Nkx3.2 (Fig. 2M-P)expression is maintained in Pbx1^-/- embryos, and Pbx1 expression is unperturbed in the condensing mesenchyme of both Pod1^-/- (Fig. 3C,D) and Nkx3.2^-/-(Fig. 3E,F) embryos at E12-12.5. Thus, Pbx1 expression in Dm condensing mesenchyme is not dependent on the presence of either Pod1 or Nkx3.2. We also confirm that Nkx3.2 is expressed in the condensing mesenchyme of Pod1^-/- embryos at E12-12.5 (see Fig. S1A,B in the supplementary material), as already reported by others(Lu et al., 2000). Additionally, Pod1^-/- expression is unperturbed in Nkx3.2^-/- embryos at E12-12.5 (see Fig. S1C,D in the supplementary material). Thus, the Pod1 and Nkx3.2 pathways appear to be separate in early spleen development. Furthermore, our present studies (Fig. 2A-D) and work by Lettice et al. (Lettice et al.,1999) reveal that Pbx1 and Nkx3.2 independently regulate Hox11 in a hierarchical fashion. Therefore, Hox11expression is dependent on both Pbx1 (Fig. 2A-D) and Nkx3.2 (Lettice et al., 1999). Moreover, no expression of Nkx2.5 was observed in the condensing splenic mesenchyme of Pod1^-/-or Pbx1^-/- embryos(Fig. 3I,J; K,L). Conversely,we demonstrated that Nkx2.5 is still expressed in the condensing splenic mesenchyme of Nkx3.2^-/- embryos (see Fig. S1E,F in the supplementary material) at E12-12.5. Thus, Nkx2.5 expression is dependent on both Pbx1 (Fig. 3K,L) and Pod1 (Fig. 3I,J), although it is not dependent on Nkx3.2 (see Fig. S1E,F in the supplementary material). Taken together, these results suggest that Pbx1 impinges on the control of the separate Pod1 and Nkx3.2 pathways, both essential for spleen development, by functioning upstream of Hox11 and Nkx2.5, respectively.

Given the dependence of Hox11 expression on Pbx1, the potential colocalization of these two transcription factors was assessed in splenic mesenchymal cells. Sections from E11.5 and E12.5 Hox11^lacZ/+ embryos (in which the lacZ gene was inserted into the Hox11 locus)(Dear et al., 1995) were immunostained with an anti-Pbx1b monoclonal antibody(Jacobs et al., 1999) and simultaneously stained for β-galactosidase activity. In these mice, lacZ expression faithfully recapitulates expression of Hox11in the splenic anlage (Kanzler and Dear,2001). Pbx1 (visualized by brown staining) and Hox11 (visualized by blue staining) were present and colocalized(Fig. 4B,D; arrow) in cells of the condensing splenic mesenchyme. However, not all Pbx1-positive cells expressed Hox11 (Fig. 4B,D; arrowhead), which was notably absent in the thickened mesothelial lining that surrounds the mesenchyme of the splenic anlage(Fig. 4A,C; arrowhead).

Given the requirement for both Hox11(Roberts et al., 1994; Dear et al., 1995) and Pbx1 in spleen development and their cellular colocalization in the splenic anlage, their potential genetic interaction during spleen ontogeny was assessed. Pbx1^+/- and Hox11^+/- mice were intercrossed and offspring were examined at 6 to 8 weeks of age(Table 1). A high percentage(80%) of Pbx1^+/-;Hox11^+/- double heterozygous mice displayed hypoplastic and malformed spleens, compared with wild-type or single heterozygous littermates(Fig. 4E,F; Table 1), of which a very low percentage exhibited splenic morphological abnormalities such as minor indentations (Table 1). The spectrum of malformations of double heterozygous spleens(Fig. 4F) comprised sickle shapes, presence of indentations, tubercles and nodules, as well as fusions of two spleens (polysplenia). Thus, Pbx1 and Hox11 genetically interact in spleen development. Despite the observed morphological abnormalities, Pbx1^+/-;Hox11^+/- double heterozygous mice exhibited normal splenic architecture(Fig. 4G,H), germinal center(GC) (Dent et al., 1997)formation (Fig. 4I,J) and primary immune function (not shown).

Early expansion of the splenic anlage, prior to hematopoietic invasion at E14.5 (Sasaki and Matsumura,1988), is dependent primarily on proliferation of splenic mesenchymal progenitor cells. Evaluation of mesenchymal cell proliferation by BrdU in vivo labeling showed a marked reduction in the percentage of S-phase cells in Pbx1^-/- splenic mesenchyme compared with wild-type controls at E13.5 (Fig. 5A,B). No significant reduction of S-phase cells in Pbx1^-/- splenic mesenchyme was found at E12.5 (not shown). Conversely, no detectable increase of apoptosis was found in Pbx1^-/- splenic mesenchyme by TUNEL assay(Gavrieli et al., 1992),either at E12.5 or at E13.5 (not shown). Taken together, these results indicate that although loss of Pbx1 does not affect mesenchymal cell survival, by E13.5 it severely affects cell proliferation and, therefore,expansion of the splenic anlage.

Given the in vivo genetic interaction of Pbx1 and Hox11in spleen development and their cellular colocalization, BrdU in vivo labeling was also performed in E13.5 Hox11^-/- embryos. The splenic anlage of Hox11^-/- embryos(Fig. 5C,D) was remarkably similar to that of Pbx1^-/- embryos(Fig. 5A,B), with a reduction in the percentage of S-phase cells of ∼50% in both mutants. Interestingly,the splenic primordium of E13.5 Hox11^-/- embryos, unlike that of Pbx1^-/- embryos, also exhibited a modest increase in apoptosis, mostly localized to the mesothelium surrounding the mesenchyme of the splenic primordium (not shown). In sum, loss of either Pbx1 or Hox11 presents a comparable phenotype, affecting the proliferation of mesenchymal splenic progenitor cells and preventing normal expansion of the splenic anlage.

To investigate the possibility that Pbx1 may directly regulate Hox11 expression, sequences of the Hox11 promoter that are conserved between the mouse and human genes were examined for Pbx-binding sites. A potential Pbx-binding site (PX1) was identified within the 540 bp Hox11 region that displays promoter activity(Arai et al., 1997), as indicated in Fig. 6A. To determine if the PX1 element could support the formation of a Pbx1 DNA-binding complex, EMSA assays were performed using nuclear extracts from primary embryonic spleen cells. A slow-migrating band containing Pbx1, as demonstrated by its specific competition with an anti-Pbx1 antibody, was observed to form on an oligonucleotide containing the PX1 site(Fig. 6B: left panel, lanes 1 and 2).

Hox11 was also present in this DNA-binding complex, as indicated by the finding that an anti-Hox11 antibody inhibited complex formation(Fig. 6B: right panel, lane 2). The complex was not observed to form on a PX1 oligo mutated within the Pbx1 site (mPX1) (Fig. 6B: middle panel, lane 1), confirming the specificity of Pbx1-Hox11 binding. Additionally, complex formation was inhibited by excess unlabeled wild-type PX1 oligo, but not by its mutated form mPX1 (not shown), further confirming binding specificity. These results demonstrate that an endogenous Pbx1-Hox11 heterodimer assembles on the PX1 site of the Hox11 promoter in vitro.

Possible in vivo Hox11 promoter-specific binding by Pbx1 and Hox11 was examined by chromatin immunoprecipitation (ChIP) performed on primary cell cultures established from murine embryonic spleens at E16. These primary murine spleen cultures, which uniformly expressed the mesodermal marker alpha-smooth muscle actin (green fluorescence; Fig. 6C: top panel), exhibited high levels of Pbx1b, Hox11 and the Meinox co-factor protein Prep1, as detected by western blot analysis (Fig. 6C: bottom panel). The binding of Pbx1 and Hox11 was examined at the region of the Hox11 promoter(Arai et al., 1997; Fig. 6A) that contains the PX1 site, where assembly of the Pbx1-Hox11 heterodimer had been detected by EMSA. Two different pairs of primers within this region (spanning regions -258 to-54, depicted in red, and -359 to -115, depicted in teal) amplified sequences within the Hox11 promoter that had been immunoprecipitated by the anti-Pbx1b antibody (Fig. 6D:left and middle panels). As a control for Pbx1b antibody specificity, two different sets of primers (one on the Hox11 promoter: -1170 to -891;depicted in blue; and one on the Bmp4 promoter) did not amplify their respective intervening sequences after immunoprecipitation(Fig. 6D: right panel).

The potential binding of Hox11 at its own promoter in vivo in embryonic murine spleen cells was also examined by ChIP assay. Similar to Pbx1, Hox11 was present at the region of Hox11 promoter activity(Fig. 6E), as detected by both primer pairs that revealed Pbx1 on the Hox11 promoter(Fig. 6E: left panel and data not shown). Overall, these results indicate recruitment of both Pbx1 and Hox11 on the Hox11 promoter in vivo in spleen embryonic cells. Thus, Hox11 is a direct target of Pbx1. The simultaneous binding of Hox11 to its own promoter suggests that it may contribute to an auto-regulatory circuit in spleen development.

The functional consequences of potential interactions of Pbx1 and Hox11 for Hox11 expression were tested in transient transcription assays using NIH 3T3 fibroblasts. For these studies, we employed a luciferase reporter construct containing the promoter regulatory region of Hox11 (p540)(Arai et al., 1997), which spans the PX1 site. When the p540 reporter gene was co-transfected with the Pbx1 construct no activation above background was observed(Fig. 6F). Co-expression of Pbx1 and a representative Meinox family protein, Prep1 (highly expressed in spleen mesenchyme; Fig. 6C:bottom panel), resulted in transcriptional activation of two- to threefold above background levels (Fig. 6F). Significantly, co-transfection of Hox11 with Pbx1 and Prep1 resulted in an eight- to ninefold increase in transcription above the baseline(Fig. 6F). These results demonstrate that synergistic activation of the Hox11 promoter is achieved by the association of the three homeodomain proteins, consistent with a Pbx1-dependent autoregulatory role for Hox11 to enhance and/or maintain its own expression during spleen development.

In this report we investigated the genetic and transcriptional control of cell fate specification, morphogenesis and expansion of the spleen, using various asplenic mouse models. Our results establish a genetic network that regulates spleen ontogeny and is dependent upon Pbx1, which encodes a TALE class homeodomain protein that serves a reiterative role in early splenic morphogenesis as well as later anlage expansion. A crucial part of the Pbx1 role in spleen development is due to its genetic and transcriptional interaction with Hox11, which autoregulates its own promoter with Pbx1 in spleen progenitor cells. However, Pbx1 genetically regulates key genes downstream of both Nkx3.2 and Pod1, which we demonstrate to control spleen development via separate pathways. These results reveal a broad and crucial role for Pbx1 as a central hierarchical co-regulator in spleen genesis.

Regional perturbations of LR asymmetry have been associated with spleen agenesis in Dh spontaneous mutants and Nkx3.2^-/-embryos (Hecksher-Sorensen et al.,2004). However, our studies demonstrate that asplenia in Pbx1^-/- embryos is not the result of perturbations of LR asymmetry, providing further evidence that asplenia is not always associated with LR asymmetry defects. This notion is best supported by the presence of spleen agenesis as the sole organ abnormality, without perturbations of LR asymmetry, in mice that lack Pod1(Quaggin et al., 1999) or Hox11 (Roberts et al.,1994; Dear et al.,1995) (N. Dear, personal communication). Similarly, Pbx1^-/- embryos do not exhibit anomalies of LR asymmetry in the heart (C. P. Chang, unpublished), lungs (B. Hogan, personal communication), liver or stomach (data not shown). Although the Pbx1^-/- pancreas displays an anteroposterior (AP)patterning defect, it does not present LR asymmetry abnormalities(Kim and Selleri et al.,2002).

Furthermore, at E10-10.5, the architecture of the splanchnic mesoderm,which normally exhibits an epithelial-like cellular organization, remains well preserved in Pbx1^-/- embryos. This contrasts with Dh mutant embryos, in which the splanchnic mesoderm is replaced by unorganized mesenchyme, and Nkx3.2^-/- embryos, in which it is in part defective (Green,1967; Hecksher-Sorensen et al., 2004). Thus, loss of Pbx1 does not result in perturbations of very early developmental choices, such as LR position and differentiation of the splanchnic mesoderm.

In vertebrates Pbx1 is a key developmental regulator required for the ontogeny of most organ systems(Selleri et al., 2001). In this study, we demonstrate that Pbx1 is reiteratively required in spleen development. First, it affects, at least in part, the fate of condensing mesenchymal cells during early organogenesis, at E11-11.5, as demonstrated by the absence of Nkx2.5 and Hox11 in Pbx1^-/- Dm. Therefore, Pbx1 genetically regulates both Hox11 and Nkx2.5 in early spleen morphogenesis. Nkx2.5 is regarded as one of the earliest known markers for spleen progenitor cells (Patterson et al.,2000), and its expression overlaps with Hox11, although its requirement in spleen genesis has not yet been shown since Nkx2.5 loss causes lethality in utero at E9-10, before development of the splenic anlage (Lyons et al., 1995). Second, at later stages of organogenesis, by E13.5, Pbx1 is required again for splenic progenitor cell proliferation, as detected by BrdU incorporation experiments, underscoring the essential role of Pbx1 in organ expansion. Thus, Pbx1 exhibits temporally distinct roles in spleen ontogeny, reminiscent of its dual contributions to skeletal development(Selleri et al., 2001).

Regulation of another marker of early splenic anlage, Wt1, is also dependent on Pbx1, possibly through regulation of Hox11, in the splenic condensing mesenchyme. It is of interest that regulation of Wt1 expression is independent of Pbx1 in the outer mesothelial lining of the splenic anlage, which normally does not express Hox11, and will give rise to the splenic capsule. Taken together,these results support a scenario where Pbx1 regulates Hox11,which in turn regulates Wt1 in the splenic mesenchyme, whereas in the mesothelial lining of the developing spleen regulation of Wt1 is uncoupled from Pbx1. Thus, Pbx1 can be considered as the uppermost known genetic regulator within the Hox11-Wt1 pathway in the non-mesothelial splenic mesenchyme (Fig. 7A).

Expression of other genes that mark early condensing splenic mesenchyme,such as Pod1 and Nkx3.2, is well maintained in Pbx1^-/- embryos, demonstrating that splenic condensing mesenchymal cells are still present in Dm of Pbx1^-/-embryos and splenic gene expression is not globally impaired by Pbx1loss. Thus, requirement for Pod1 and Nkx3.2 in spleen ontogeny is not dependent upon Pbx1. Finally, the findings by Lu et al.(Lu et al., 2000), confirmed by our present studies, that Nkx3.2 is expressed in the condensing mesenchyme of Pod1^-/- embryos at E12-12.5, and our results that Pod1^-/- expression is unperturbed in Nkx3.2^-/- embryos, indicate that the Pod1 and Nkx3.2 transcription factors use separate pathways to regulate early spleen development.

Although previous work has demonstrated asplenia in Hox11 mutants(Roberts et al., 1994; Dear et al., 1995), until now the cellular basis of this defect was mostly unknown. Indeed, apoptosis was not detected in Hox11^-/- splenic primordium in a previous study (Roberts et al., 1995),while it was documented in another report that used a different Hox11-deficient model (Dear et al., 1995). In the present study, we observed a modest increase of apoptosis in Hox11^-/- splenic primordium. The apoptotic cells were mostly localized to the outer mesothelial lining of the splenic anlage, which gives rise to the spleen capsule and does not normally express Hox11. Thus, it appears that such a subtle increase of apoptosis cannot be responsible for the complete lack of spleen development in Hox11-deficient embryos. Conversely, our finding that, by E13.5, Hox11^-/- spleen progenitor cells exhibit a marked defect in cellular proliferation comparable with that in Pbx1^-/-embryos is consistent with the hypoplasia of Hox11^-/-splenic anlage (Roberts et al.,1994; Dear et al.,1995) and the demonstrated involvement of Hox11 in cellular proliferation and cell cycle control(Kawabe et al., 1997; Hough et al., 1998; Owens et al., 2003). Finally,the finding of a common cellular defect (i.e. impaired progenitor cell proliferation) in spleen development of Pbx1^-/- and Hox11^-/- embryos further corroborates the observation that Pbx1 genetically regulates Hox11.

In sum, the requirement for Pbx1 in spleen ontogeny appears to be reiterative. This reiterative role can account for the complete absence of the spleen, which would not otherwise be explained either by a partial impairment of splenic cell fate specification or by a 50% decrease in progenitor cellular proliferation in the splenic anlage, but probably results from the summation of these defects.

The finding of a high percentage (80%) of Pbx1^+/-;Hox11^+/- double heterozygous mice displaying severely hypoplastic and malformed spleens(Fig. 4E,F; Table 1), compared with single heterozygotes, demonstrates that Pbx1 and Hox11 genetically interact in vivo in spleen development. The wide spectrum of malformations of Pbx1^+/-;Hox11^+/- double heterozygous spleens, which includes fusions of two spleens, mimics polysplenia, a human congenital condition. In polysplenia two or more splenic masses, hypoplastic and irregularly shaped (splenules), are present lateral to the stomach(Lodewyk et al., 1972). Unlike human asplenia, which involves life-threatening infections in children(Waldman et al., 1977),polysplenia is associated with normal splenic function(Lodewyk et al., 1972). Despite their morphological abnormalities, Pbx1^+/-;Hox11^+/- double heterozygous mice exhibit normal splenic architecture, germinal center formation and primary immune function (not shown), thus closely modeling the human polysplenic condition.

Despite the growing understanding of Hox and TALE homeoprotein functions in development (Krumlauf, 1994; Mann and Affolter, 1998; Popperl et al., 2000; Selleri et al., 2001; Waskiewicz et al., 2002; Hisa et al., 2004; Selleri et al., 2004) and their functional interactions (Popperl et al., 1995; Maconochie et al.,1997; Jacobs et al.,1999; Ferretti et al.,2000; Manzanares et al.,2001; Samad et al.,2004), to date, only a few direct target genes have been reported(Rauskolb et al., 1993; Graba et al.,1997; Bromleigh and Freedman,2000; Theokli et al.,2003). In this study, we provide the first in vivo evidence that Pbx1 directly regulates Hox11 in embryonic spleen cells. Interestingly, at E9.5, Pbx1 is already expressed in the mid-gut mesenchyme (Schnabel et al.,2001), from which the spleen is derived, well before the onset of Hox11 expression (Kanzler and Dear, 2001). And indeed, Pbx1 controls the onset of Hox11 expression at E11 within the Dm, as demonstrated by our in situ hybridization experiments. Although Pbx1 expression starts to decrease in the splenic anlage after E13.5 (data not shown), Hox11persists in the spleen until birth(Kanzler and Dear, 2001),suggesting that Pbx1 is required for the onset of Hox11 expression and for its continued expression in early spleen development, until E13.5,although it is not necessary for Hox11 maintenance in later phases of organogenesis.

Hox11 is one of the earliest known markers for spleen cell progenitors (Dear et al.,1995). A useful tool to monitor Hox11 transcription in the developing spleen is provided by Hox11^lacZ mice(Dear et al., 1995), in which lacZ expression is dependent on Hox11 regulatory sequences and faithfully recapitulates Hox11 expression. Analysis of Hox11^lacZ/lacZ embryos previously demonstrated that lacZ expression is normally initiated in the absence of Hox11 (Dear et al.,1995), suggesting that the Hox11 protein is not required for initiation of its own transcription in the splenic mesenchyme. These findings indicate that other factors might be necessary for the onset of Hox11transcription. Here, we identify Pbx1 as one such factor that activates Hox11 transcriptional onset and early expression in the splenic anlage until E 13.5.

Pbx1 and Hox11 bind to a potential Pbx-binding site (PX1) within the Hox11 promoter, as shown by EMSA assays conducted on embryonic spleen primary cells. Additionally, Pbx1 and Hox11 bind the Hox11 promoter in vivo in embryonic spleen cells, as revealed by ChIP assays. Regulatory interactions of Hox genes, such as the induction of Hoxb1 segmental expression by Hoxb1 and Hoxa1 through auto- and cross-regulatory loops, have been documented in developmental processes(Popperl et al., 1995; Studer et al., 1996; Studer et al., 1998). Here, we reveal an autoregulatory loop for an orphan Hox gene, Hox11, which is non-clustered but bears a hexapeptide motif(Shen et al., 1996). Taken together, our findings establish that Hox11 is a direct target of Pbx1 and that, simultaneously, it regulates its own promoter. Significantly,co-transfection of Hox11 with Pbx1 and Prep1 resulted in a striking increase in transcription above baseline, demonstrating that synergistic activation of the Hox11 promoter is achieved by the association of the three homeoproteins, consistent with a Pbx1-dependent autoregulatory role for Hox11 to enhance and/or maintain, at least in part, its own expression during spleen development. Interestingly, additional potential Pbx-Meinox-binding sites were identified within the Hox11 promoter downstream of the PX1 site (not shown), suggesting that multiple binding sites within close proximity might be used, simultaneously or at different times, for a complex, multi-faceted transcriptional regulation of Hox11 in spleen development.

In addition to establishing that Pbx1 is the most upstream known direct regulator of Hox11 in spleen ontogeny(Fig. 7A), our studies demonstrate an even broader role for Pbx1 in spleen development. Pbx1 regulates key genes downstream of Nkx3.2 and Pod1, which we show to control spleen development through separate genetic pathways (Fig. 7A). Both Nkx3.2 (Lettice et al.,1999; Tribioli et al., 1999) and Pod1(Quaggin et al., 1999; Lu et al., 2000) are essential for spleen development, and their expression in condensing splenic mesenchyme overlaps with Hox11 and Nkx2.5. The specific mechanisms and cellular behaviors by which the Nkx3.2 transcription factor regulates spleen development are as yet unknown, while Pod1 has been proposed to control splenic cell survival (Lu et al.,2000). Interestingly, Lu et al. have reported that the spleen primordium of Pod1^-/- embryos does not further expand after E12.5 and starts to undergo apoptotic cell death. As a result, after E12.5, expression of all splenic markers, including Nkx3.2,disappears from the degenerating splenic primordium of Pod1^-/- embryos (Lu et al., 2000). Our studies demonstrate that Pbx1 expression is not dependent on the presence of either Nkx3.2 or Pod1 in the splenic mesenchyme (Fig. 7A). Likewise,the requirement for both of these transcription factors in splenic development is Pbx1 independent (Fig. 7A). In addition, our findings that Nkx3.2 is expressed in the condensing mesenchyme of Pod1^-/- embryos at E12-12.5, and that Pod1 expression is also unperturbed in Nkx3.2^-/- embryos, indicate that the Pod1 and Nkx3.2 transcription factors use separate pathways to regulate early spleen development.

Furthermore, our studies reveal that Pbx1 and Nkx3.2independently regulate Hox11 in a hierarchical fashion(Fig. 2M-P; Fig. 3E,F)(Lettice et al., 1999). And,in a similar scenario, Pbx1 and Pod1, but not Nkx3.2 (see Fig. S1E,F), independently control Nkx2.5 gene expression in a hierarchical fashion (Fig. 7A). Thus, Pbx1 impinges on the separate Nkx3.2 and Pod1 pathways by genetically regulating key players in both of these pathways, i.e. Hox11 and Nkx2.5(Fig. 7A). As a result, Pbx1 emerges as a central hierarchical co-regulator in spleen ontogeny (Fig. 7A). It will be of interest to determine the roles of additional transcription factors required for spleen development, such as Sox11(Sock et al., 2004) and Nkx2.3(Pabst et al., 1999; Wang et al., 2000; Tarlinton et al., 2003),within the genetic pathways established by our study.

In conclusion, we demonstrate here the essential role of the Pbx1-Hox11 transcriptional pathway in spleen ontogeny. We provide evidence that Pbx1 is reiteratively required during spleen development, as it is implicated, at least in part, in splenic cell fate specification and morphogenesis, and then is essential again, later in organogenesis, for anlage expansion through control of progenitor cell proliferation. Finally, we demonstrate that spleen ontogeny is dependent on the orchestration of a complex network of transcription factors, among which Pbx1 emerges as a central, master co-regulator. Overall, our study takes a significant first step towards understanding the genetic and transcriptional control of spleen development.

We thank W. Zimmer and R. Schwartz for generating and providing Nkx3.2 knockout mice (unpublished); S. Korsmeyer and T. H. Rabbitts for providing Hox11 and Hox11^lacZ knockout mice,respectively; C. Englert, R. Harvey, H. Popperl, T. H. Rabbitts, J. Rossant and R. Schwartz for in situ hybridization probes; N. Koehler and M. Hatano for the Hox11 constructs; N. Dear for sharing unpublished observations;K. Manova for immunohistochemistry support; C. Nicolas for technical assistance; L. Lacy, A. Koff, L. Niswander, K. Anderson and members of their laboratories for stimulating discussions; J. Giacalone for editing of the manuscript; K. Hadjantonakis, S. Kim, L. Lacy and V. Zappavigna for critical reading of the manuscript. L.S. personally thanks N. Copeland, B. Hogan and T. H. Rabbitts for invaluable input, and V. Zappavigna for sharing his expertise for the ChIP experiments. This work was supported by grants from Telethon and the Italian Ministry of Research (MIUR, PRIN) to F.B.; the National Institutes of Health to M.C. (CA42971 and CA90735) and L.S. (HD43997), and a grant from the March of Dimes and Birth Defects Foundation (6-FY03-071) to L.S. L.S. is an Irma T. Hirschl Scholar.