The T-Cell Oncogenic Protein HOX11 Activates Aldh1 Expression in NIH 3T3 Cells but Represses Its Expression in Mouse Spleen Development

Abstract

Hox11 is a homeobox gene essential for spleen formation in mice, since atrophy of the anlage of a developing spleen occurs in early embryonic development in Hox11 null mice. HOX11 is also expressed in a subset of T-cell acute leukemias after specific chromosomal translocations. Since the protein has a homeodomain and can activate transcription, it probably exerts at least some of its effects in vivo by regulation of target genes. Representational difference analysis has been used to isolate cDNA clones corresponding to mRNA species activated following stable expression of HOX11 in NIH 3T3 cells. The gene encoding the retinoic acid-synthesizing enzyme aldehyde dehydrogenase 1 (Aldh1), initially called Hdg-1, was found to be ectopically activated by HOX11 in this system. Study of Aldh1 gene expression during spleen development showed that the presence of Aldh1 mRNA inversely correlated with Hox11. Hox11 null mouse embryos have elevated Aldh1 mRNA in spleen primordia prior to atrophy, while Aldh1 seems to be repressed by Hox11 during organogenesis of the spleens of wild-type mice. This result suggests that expression of Aldh1 protein is negatively regulated by Hox11 and that abnormal expression of Aldh1 in Hox11 null mice may cause loss of splenic precursor cells by aberrant retinoic acid metabolism.

The HOX11 protein is a member of the family of proteins carrying a DNA-binding homeodomain. HOX11 was discovered by its activation in a subset of T-cell acute leukemias with t(10;14)(q24;q11) or t(7;10)(q35;q24) (6, 10, 18, 23). This ectopic expression is a contributary factor in the leukemogenesis of those T-cell tumors with translocations affecting chromosome 10, band q24. While activation of HOX11 is a feature of some T-cell tumors, the gene is normally not expressed in T cells. In developing mouse embryos, Hox11 expression is restricted to some parts of the brain, the developing branchial arches, and the developing spleen (3, 35, 36). The last site of expression seems particularly crucial in mouse development, as creation of null mutations in Hox11 causes mice to be born without spleens (36) as a result of atrophy of the spleen anlage around embryonic days 13 to 14 (E13 to E14) (3).

The primary structure of the HOX11 protein suggested that it may be a transcription factor, mainly because of the homeodomain (4). Molecular experiments to dissect functional domains of HOX11 confirmed its ability to bind to specific DNA sequences via the homeodomain (4, 43) and to activate transcription (26, 50). The latter ability was due to the presence of modular transcriptional activation domains (50), of which an NH₂-terminal module is needed for optimal transcription of a chromosomal target gene (26). In addition, the ability of HOX11 to interact with the catalytic subunits of protein phosphatases 1 and 2A has also been implicated in tumorigenesis (26), but the significance of this interaction for Hox11 function in normal mouse embryogenesis is unclear.

Thus, both the expression of Hox11 during embryogenesis (particularly in the developing spleen) and the ectopic expression of HOX11 following chromosomal translocation in T cells may have varying consequences for the cell. These consequences include activation and repression of gene expression mediated by the homeodomain binding to DNA in a sequence-specific manner and modulation of protein function via protein-protein interactions. The latter mode of action presumably also has the effect of modulating gene expression indirectly. In the light of these considerations, it was pertinent to assess the effect on gene expression after ectopic activation of HOX11 expression. As a model system, NIH 3T3 cells, which stably express HOX11, were made, and mRNAs expressed as a result of this were cloned as cDNA fragments. In this system, we have cloned and characterized two genes, class 1 aldehyde dehydrogenase (Aldh1) and Slim1. Aldh1 gene expression seems crucial for mouse spleen development since it shows an inverse relationship to Hox11 expression in this developing organ.

MATERIALS AND METHODS

Cell lines.

NIH 3T3 fibroblasts stably expressing HOX11 or ΔH3-HOX11 (the latter carries a deletion of the third helix of the HOX11 homeodomain) from the pEF-BOS vector (28) have been previously described (26) and were grown in Dulbecco modified Eagle medium supplemented with 10% fetal calf serum.

cDNA synthesis.

Total cytoplasmic RNA was isolated from NIH 3T3 cell lines by Nonidet P-40 lysis, phenol extraction, and ethanol precipitation. Polyadenylated [poly(A)⁺] RNA was purified from total RNA by oligo(dT) cellulose chromatography. In preparing cDNA, we followed a protocol kindly supplied by Hubank and Schatz (15). Poly(A)⁺ RNA (5 μg) and oligo(dT)_12–18 (2 μg) were heated to 70°C for 10 min in a total volume of 20 μl and then incubated with 1× first-strand buffer (Gibco BRL), 10 mM dithiothreitol, 0.5 mM each deoxynucleotide triphosphate, 20 U of RNasin (Promega), and 600 U of Superscript II reverse transcriptase (Gibco BRL) in a total volume of 40 μl at 37°C for 1 h. Second-strand synthesis was achieved by adding 40 μl of 5× second-strand synthesis buffer [100 mM Tris-HCl (pH 7.5), 500 mM KCl, 25 mM MgCl₂, 50 mM (NH₄)₂SO₄, 50 mM dithiothreitol, 0.25 mg of bovine serum albumin per ml], 2 μl of 15 mM β-NAD, 2 U of Escherichia coli DNA ligase (New England Biolabs [NEB]), 2.8 U of RNase H (Pharmacia), and 40 U of E. coli DNA polymerase I (Boehringer Mannheim) in a final volume of 200 μl and incubating the mixture at 15°C for 2 h and then at 22°C for 1 h. Double-stranded cDNA was phenol-chloroform extracted, ethanol precipitated in the presence of 2 μg of glycogen carrier, and resuspended in 30 μl of 10 mM Tris-HCl (pH 7.5)–1 mM EDTA (TE).

RDA oligonucleotides.

Oligonucleotides used for representational difference analysis (RDA) (15) were as follows: RBgl12, GATCTGCGGTGA; RBgl24, AGCACTCTCCAGCCTCTCACCGCA; JBgl12, GATCTGTTCATG; JBgl24, ACCGACGTCGACTATCCATGAACA; NBgl12, GATCTTCCCTCG; and NBgl24, AGGCAACTGTGCTATCCGAGGGAA.

RDA.

RDA was carried out as described previously (15). Double-stranded cDNA (approximately 2 μg) was digested with DpnII (NEB), extracted with phenol-chloroform, and ethanol precipitated. After resuspension, half of the restricted cDNA was ligated to the RBgl12 (4 μg) and RBgl24 (8 μg) oligonucleotides overnight at 16°C with 1,200 U of T4 DNA ligase (NEB) in a 60-μl reaction mixture. Prior to addition of DNA ligase, the partially complementary RBgl oligonucleotides were annealed by heating the ligation mixture to 50°C for 1 min and slowly cooling it to 10°C. The ligations were diluted to 200 μl with TE and amplified by PCR with the RBgl24 oligonucleotide as the primer and 5 U of Amplitaq (Perkin-Elmer Cetus) in a 200-μl volume. Twenty identical reaction mixtures were prepared, with each receiving 2 μl of the diluted ligation mixture. PCR samples were incubated at 72°C for 3 min to melt the RBgl12 oligomer before addition of the polymerase. After a further 5 min at 72°C to fill in the ends, the samples were subjected to 20 cycles of PCR (95°C for 1 min and 72°C for 3 min) and then combined, extracted with phenol-chloroform, precipitated with isopropanol, and resuspended in 600 μl of water.

Driver DNA was prepared by digesting the cDNA representations (600 μl) with DpnII, followed by phenol-chloroform extraction and isopropanol precipitation. Tester DNA was prepared by gel purifying 1/10 of the driver DNA with the Qiaex II system (Qiagen) and ligating it to the JBgl12 and JBgl24 adapters as described above. The first hybridization was performed by combining 0.4 μg of the NIH 3T3 HOX11 tester DNA with 40 μg of NIH 3T3 driver DNA (a ratio of 1:100). The mixture was ethanol precipitated and resuspended in 4 μl of EE × 3 buffer, i.e., 30 mM N-(2-hydroxyethyl)piperazine-N-(3-propanesulfonic acid) (pH 8.0)–3 mM EDTA, and covered with mineral oil. After heating the resuspension to 98°C for 5 min, 1 μl of 5 M NaCl was added and the solution was incubated at 67°C for 20 h. Following hybridization, the mixture was diluted to 400 μl with TE along with 40 μg of yeast RNA as the carrier.

Amplification of the reannealed tester was performed by adding 20 μl of the diluted hybridization mix to a 200-μl PCR mixture, with the JBgl24 oligonucleotide as the primer as described above. After 10 cycles of PCR, the reaction mixtures were phenol-chloroform extracted, isopropanol precipitated, and resuspended in 40 μl of 0.2× TE. To remove single-stranded amplification products, half of the initial PCR products were incubated with 20 U of mung bean nuclease in a 40-μl volume at 30°C for 35 min. The reaction was stopped by adding 160 μl of 50 mM Tris-HCl (pH 8.9) and incubating the mixture at 98°C for 5 min. PCR amplification was continued by adding 20 μl of the mung bean nuclease-digested products to a 200-μl PCR mixture as described above. After a further 18 cycles of PCR, the products were extracted with phenol-chloroform, precipitated with isopropanol, and resuspended in 100 μl of TE as the first difference product (DP1).

DP1s were digested with DpnII and ligated to new adapters (NBgl12 and -24), and the RDA procedure was repeated as described above with tester-driver ratios of approximately 1:800, 1:400,000, and 1:8,000,000 for the second, third, and fourth rounds, respectively. J adapters were ligated to the tester for the third round, and N adapters were ligated for the fourth round. At the end of the procedure, DPs were cut with DpnII, gel purified, and cloned into the BamHI site of pBluescript KS(+) (Stratagene). Transformed TG1 bacterial colonies were screened by colony hybridization with the NIH 3T3 and NIH 3T3 HOX11 cDNA representations as probes.

Filter hybridization analysis.

Total RNA was isolated from cultured NIH 3T3 cells by Nonidet P-40 lysis. For Northern hybridizations, RNA samples (10 μg) were denatured with glyoxal (50°C for 1 h), electrophoresed on 1.4% agarose gels, and transferred to nylon membranes (45). ³²P-labelled probes were prepared by random priming (7), and the filters were hybridized in hybridization buffer containing 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 10× Denhardt’s solution, 0.1% sodium dodecyl sulfate, and 250 mg of salmon sperm DNA per ml at 65°C for 16 h (5). The filters were washed in 0.1× SSC–0.1% sodium dodecyl sulfate at 65°C and subjected to autoradiography at −70°C.

DNA filter hybridization (5, 41) was carried out essentially as described previously (21). The HOX11 probe used was a 1,025-bp coding region cDNA fragment. The β-actin probe was a 500-bp cDNA fragment generated by reverse transcription (RT)-PCR. The Slim1 (483 bp), Aldh1 (517 bp), and ATP synthase (250 bp) probes were DpnII cDNA fragments obtained from the RDA procedure. The T-cell receptor β chain (TCRβ) probe was a 750-bp cDNA isolated from pM131B10BB1 (39). The Hox2.9 probe was an 876-bp cDNA fragment.

Western blotting.

Western blotting was carried out as described previously with rabbit polyclonal antiserum raised against a bacterially expressed HOX11 protein (26).

In situ RNA hybridization.

Cryosections (14 μm) of mouse embryos, either from Hox11 homozygous mutant (−/−) (3) or wild-type (C57BL.6) embryos at E12.5 and E13.5, were fixed in 4% paraformaldehyde and pretreated with 0.25% acetic anhydride in 0.1 M triethanolamine for 10 min. Hybridization was performed overnight with probes 3′ end labelled with ³⁵S-dATP as previously described (47, 48).

Antisense oligonucleotides used were as follows: for Hox11, GTGAACCTGTCCTTTGTGTATCTGCGGTTACTCTCCATCC and AGTGTAGGCGCCTCCAACCAAACAGCCAAGGCCATAGTCT, which correspond to mRNA positions 536 to 575 and 129 to 168, respectively; for Aldh1, ACAGCTAAGGGCAGGGCCTATCTTCCAAATGAACATGAGC and CTGAGGGACAGAGCTTAAGTCACAACTAGTCCTCCTCACC, which correspond to mRNA positions 519 to 558 and 1805 to 1844, respectively; for Aldh2, GGTTGCCAGGGCTGGGCCCAGTTTCCATGCTTGCATCAGG and GCACATCTGGACGCGTGATCGGTCTGCACTGAGCTTCATC, which correspond to mRNA positions 573 to 612 and 1742 to 1781, respectively; and for Tal1 (or Scl), CAACCCAGGGGCCTAAAGGCCCCTGCCTGCTGGCCCTGGGCAGCC, which corresponds to mRNA positions 1000 to 1044. All numbering is relative to the start site of the ATG initiation codon.

RESULTS

RDA cloning of HOX11-activated genes in NIH 3T3 cells.

Differential cDNA cloning methods, such as RDA (15, 22), provide the means to compare different mRNA expression profiles. We used cDNA RDA to compare levels of gene expression between NIH 3T3 cells and NIH 3T3 cells ectopically expressing HOX11 (3T3-HOX11) following stable transfection. RDA was performed with NIH 3T3 or 3T3-HOX11 cDNA representations as either the driver or the tester (15). Only when we used 3T3-HOX11 as the tester could consistent bands be detected (by DP2) after agarose gel electrophoresis (Fig. 1A, 3T3 HOX11 lanes DP2 to DP4), and one of these, a 267-bp band, hybridized strongly with a HOX11 probe (Fig. 1B). Thus, HOX11 was rapidly enriched by the RDA procedure, reaching a maximal level by the DP3 round (Fig. 1B). Conversely, hybridization with a β-actin probe showed depletion of this sequence, which is common to the driver and tester, from the DP3 and DP4 rounds of subtracted cDNA when RDA subtraction was carried out with cDNA of either polarity (Fig. 1B).

FIG. 1.

Enrichment of 3T3-HOX11 DPs by cDNA RDA. (A) Ethidium-stained agarose gel electrophoresis of starting cDNA DpnII fragment representations and various RDA enriched populations obtained by mutual subtraction of NIH 3T3 cells (3T3) with 3T3-HOX11 (left lanes show products obtained with NIH 3T3 cells as the tester and right lanes show products obtained with 3T3-HOX11 cells as the tester). Tester representation (R; unselected cDNA) and DP1, DP2, DP3, and DP4 are shown. The arrow indicates an enriched HOX11 DpnII fragment identified by Southern filter hybridization with a HOX11-specific probe. (B) Southern filter hybridizations (with Slim1, Aldh1, HOX11, and β-actin probes as indicated) of the agarose gel-fractionated RDA DPs shown in panel A. Fragment sizes are indicated. (C) Northern filter hybridization of 10 μg of total RNA prepared from untransfected NIH 3T3 cells and three independent 3T3-HOX11 clones and hybridized with HOX11, Slim1, Aldh1, and ATP synthase probes (as indicated). 3T3-HOX11 clones 5 and 18 were found to express HOX11 protein, while clone 11 did not (data not shown). Hybridization of the filter with ATP synthase was used to assess the quality of the RNA transferred.

cDNA sequences enriched when 3T3-HOX11 cDNA was used as the tester were identified by cloning gel-purified DPs, followed by sequential hybridization with NIH 3T3 and 3T3-HOX11 cDNA representations as probes. After we screened 264 recombinant colonies, we identified 126 differentially expressed clones, of which 77% represented HOX11. DNA sequencing revealed that the other clones fell into two main groups. One group, 10% of the clones, corresponded to the human and mouse SLIM1 and Slim1 genes, which encode a novel LIM-only protein with four and a half LIM domains (29), also called Kyo-T1 (44), and another group, 7% of the clones, corresponded to the mouse Ahd-2 gene, which encodes class 1 (cytosolic) aldehyde dehydrogenase (37). This gene was provisionally called Hdg-1 when it was first isolated (26).

The characteristics of enrichment of Slim1 and Aldh1 cDNAs were assayed by hybridization of appropriate probes to filters with aliquots of cDNA from each selected round (DP1 to DP4) and either untransfected NIH 3T3 or the 3T3-HOX11 clone as the tester mRNA population (Fig. 1B). This hybridization analysis revealed that both Slim1 and Aldh1 cDNAs were enriched in the 3T3-HOX11 tester in a manner which paralleled that of the HOX11 cDNA. The consistent ability of HOX11 to stimulate both Slim1 and Aldh1 mRNA production was assayed by Northern hybridization of RNAs from three different NIH 3T3 clones stably transfected with the HOX11 expression vector, two of which expressed HOX11 mRNA (Fig. 1C, clones 5 and 18) and one of which did not (Fig. 1C, clone 11). These assays demonstrated that HOX11-positive clones 5 and 18 express both Slim1 and Aldh1 mRNAs but that untransfected NIH 3T3 cells or transfected, nonexpressing clone 11 does not. However, in a wider analysis we observed that while Aldh1 was always activated in HOX11-expressing NIH 3T3 clones, Slim1 was activated only in a proportion (about 50%; data not shown). The reason for this difference is obscure at present but may reflect clonal variation in responsiveness in the NIH 3T3 population, perhaps with respect to expression of a cofactor. Nonetheless, the data indicate a clear relationship between Aldh1 activation and HOX11 protein in this system.

ALDH1 expression in human lymphoid cell lines.

The HOX11 gene was first discovered through its activation by the chromosomal translocation t(10;14) or t(7;10) in some T-cell acute leukemias. Possible effects of HOX11 expression on both ALDH1 and SLIM1 mRNA levels in human lymphoid cell lines were assayed by Northern filter hybridization (Fig. 2). Each lane of RNA was assessed for integrity and loading by hybridization with an ATP synthase probe (Fig. 2, bottom section) and with a TCRβ or HOX11 probe for the T-cell lines. ALDH1 mRNA expression was surprisingly restricted and was detectable only in HEL cells (a cell line derived from a human erythroleukemia) and a small cell lung carcinoma line (LUDLU). Low levels of ALDH1 mRNA were found in HepG2. It is noteworthy that none of the T-ALL lines (including two derived from T-ALL tumors with translocations of chromosome 10, band q24, involving the HOX11 gene) express sufficient ALDH1 mRNA to be detected in this Northern blot assay. Conversely, SLIM1 mRNA was found in most of the T-ALL lines (Fig. 2) as well as in one pre-B-ALL line (ACVA-1) and a neuroblastoma line (N417). High levels of SLIM1 mRNA were also found in the HepG2 liver cell line. The expression of SLIM1 in these cancer cell lines is noteworthy, as previous studies of normal tissues have shown that levels of expression of SLIM1, as judged by Northern blot analysis, were high in skeletal and heart muscles but were barely detectable in lymphoid tissues and completely absent in liver (29, 30, 44).

FIG. 2.

Expression of ALDH1 and SLIM1 mRNAs in T- and B-lymphoid cell lines. Results of Northern filter hybridization of total RNAs from the indicated cell lines with Slim1, Aldh1, HOX11, TCRβ, and ATP synthase probes are shown. RNAs were extracted from the T-cell lines (T) KOPT-K1, RPMI 8402, ALL-SIL, PER-255 (the last two carrying 10q24 translocations involving HOX11), SUP-T1, SUP-T13, CCRF-CEM (CEM), NALM, HBP-ALL, PER-117, and Jurkat or from the B-cell lines. (B) 38B9, WEHI-231, RAJI, and ACVA-1. N417 is a neuroblastoma line; LUDLU is a small cell lung carcinoma line; HEL, K562, and 707 are erythroid lines, and HepG2 is a hepatoma line.

Specificity of Aldh1 and Slim1 induction in 3T3-HOX11 cells.

The specificity of Aldh1 gene activation was analyzed by distinct transfection experiments involving NIH 3T3 cells expressing various HOX11 constructs as well as an unrelated Hox gene, Hox2.9 (8, 31). First, the effect of inactivating the HOX11 homeodomain was tested by stably transfecting NIH 3T3 cells with either a mutant of HOX11 lacking the third helix of the homeodomain, ΔH3-HOX11 (26), complete HOX11, or the vector only (Fig. 3A). RNAs were isolated from three independent cell clones from each transfection, and Northern blot analysis was carried out. While HOX11 mRNA was detected in HOX11 and ΔH3-HOX11 transfected clones (but not in the vector-only control clones) (Fig. 3A, BOS lanes), only the HOX11 clones showed transcriptional activation of the Aldh1 gene. Little or no Aldh1 mRNA was found in the mutant HOX11 clones despite both Hox11 mRNA (Fig. 3A) and protein being detectable in the clones (in Fig. 3A, the bottom section shows a Western blot developed with anti-HOX11 antiserum).

FIG. 3.

Specificity of Aldh1 induction by ectopic expression of HOX11 RNA, extracted from NIH 3T3 cell clones stably transfected with various HOX11 or Hox2.9 expression constructs, was analyzed by Northern filter hybridization with Aldh1, HOX11, Hox2.9, and ATP synthase probe as indicated. (A) Induction of Aldh1 mRNA by ectopic expression of HOX11 protein. RNAs were isolated from three independent NIH 3T3 clones stably transfected with either a HOX11 expression construct (HOX11 controlled by the pEF-BOS expression cassette [HOX11]), a HOX11 expression construct encoding a mutant HOX11 protein lacking the third helix of the homeodomain (ΔH3-HOX11), or the expression vector only (BOS). Aliquots were fractionated and transferred to nylon membranes prior to hybridization with an Aldh1, a HOX11, or an ATP synthase probe. Protein extracts were also prepared from these NIH 3T3 clones and analyzed by Western blotting. The presence of HOX11 protein was assayed (bottom section) with anti-HOX11 antiserum (26). (B) Stimulation of Aldh1 mRNA expression by HOX11 via the MT promoter. NIH 3T3 clones expressing HOX11 were incubated in the presence or absence of 50 μM zinc for 36 h. RNAs were prepared, fractionated, and transferred to nylon membranes. These were hybridized to Aldh1 and ATP synthase probes. Clone 1 is a control not expressing HOX11 protein (as shown by the Western blot developed with anti-HOX11 antiserum [bottom section]), and clone 2 is a HOX11-expressing line (see bottom section). (C) Aldh1 gene expression in response to HOX11 protein expression. We obtained NIH 3T3 clones which were stably transfected with pEF-BOS-HOX11 or pEF-BOS-Hox2.9. Two clones were selected for their expression of HOX11 or Hox2.9 mRNA. RNA was prepared, electrophoresed, and transferred to nylon filters. These filters were hybridized with Aldh1, HOX11, Hox2.9, and ATP synthase probes as indicated.

A similar dependence of Aldh1 induction on the ectopic expression of HOX11 in NIH 3T3 cells was found when HOX11 expression was controlled by the metallothionein (MT) promoter. Proteins extracted from NIH 3T3 clones isolated after transfection with the MT-HOX11 expression vector or the vector only were analyzed by Western blotting with anti-HOX11 antiserum following induction with zinc. A moderate induction of HOX11 protein production was found in this system (three- to fourfold) (Fig. 3B, lower section), and HOX11 protein was found only in the MT-HOX11 clones. In Northern (RNA) blot analysis, when the control ATP synthase mRNA levels were compared to those of Aldh1, it was found that Aldh1 gene expression is confined to the MT-HOX11 clones. However, only a slight difference was seen in the Aldh1 levels in uninduced and induced cells (Fig. 3B, top section), indicating that HOX11 protein levels in uninduced cells, due to leakiness of the MT promoter, are sufficient for its role in activating Aldh1 expression in the NIH 3T3 cells.

The effect of the Hox2.9 protein (an unrelated homeodomain protein) (8, 31) on Aldh1 mRNA synthesis was analyzed to assess the specificity of the HOX11 protein effect. NIH 3T3 clones were obtained following transfection with expression vectors bearing genes coding for either HOX11 or Hox2.9, and RNA was isolated. Figure 3C shows the results of Northern hybridization analysis with two independent clones from each transfection. Using an ATP synthase probe to control RNA integrity, we confirmed the presence of HOX11 or Hox2.9 mRNA in the appropriate clones. Aldh1 mRNA was found only in the NIH 3T3 clones expressing HOX11 (Fig. 3C), confirming the specificity of HOX11 for Aldh1 mRNA induction.

The Aldh1 gene is repressed in the spleen primordium in the presence of Hox11.

During mouse embryogenesis, while Hox11 is expressed in a variety of locations (3, 35, 36), the sole defect which can be observed in mice lacking functional Hox11 genes is the absence of spleens at birth (3, 36). During early development at around E12 to E13, mesoderm forms the developing-spleen anlage but the organ involutes and disappears by the E14 stage in Hox11 null mice. The Hox11 null mutation, however, does not affect viability or vigor in any other observable way, and true breeding Hox11^−/− mice can be obtained (3, 36). The genes which we have identified by ectopic expression of HOX11 in NIH 3T3 cells may therefore play a role in spleen development, and the absence of Hox11 protein in the null mice might result in the absence of Aldh1 and/or Slim1. This possibility was investigated by in situ RNA hybridization analysis of wild-type mouse embryos and Hox11 null mutant embryos at E12.5 and E13.5.

This analysis failed to demonstrate any difference between levels of Slim1 expression in the wild type and Hox11 null embryos (data not shown). However, a surprising inverse relationship was found between Hox11 and Aldh1 mRNA levels in the developing spleen (Fig. 4). At E12.5 and E13.5 Hox11 mRNA can be observed in wild-type embryos in the developing hindbrain and the spleen anlage (Fig. 4) whereas Aldh1 mRNA is not detectable. As a control, the related mRNA for Aldh2 was analyzed and, unlike mRNA for Aldh1, showed widespread expression. As the spleen anlage contains rather few cells at E12.5, it may be possible to miss this developing organ as the sagittal sections are made. Accordingly, we prepared serial parasagittal sections of an E13.5 wild-type embryo and hybridized one section with the Hox11 oligonucleotide, thereby detecting the spleen anlage, the next section with the Aldh1 oligonucleotide, and the next with the Hox11 oligonucleotide (Fig. 4, bottom section, left-hand embryo panels). Hox11 mRNA was detectable in both flanking sections of the spleen anlage, but no Aldh1 mRNA was detectable in the intermediate section.

FIG. 4.

Aldh1 expression in embryonic mouse spleen is inversely correlated with Hox11. Wild-type (+/+) or Hox11 homozygous mutant (−/−) mouse embryos were taken at E12.5 or E13.5, and parasagittal cryosections were hybridized in situ with oligonucleotide probes specific for Hox11, Aldh1, Aldh2, or Tal1/Scl mRNA. The upper section shows E12.5 embryos. Consecutive sections of +/+ or −/− embryos were hybridized with Hox11, Aldh1, and Aldh2 probes. The lower section shows E13.5 embryos. Consecutive sections of +/+ or −/− embryos were hybridized with Hox11 and Aldh1 probes in the order shown. For comparison, we hybridized Tal1/Scl, which showed its specific location within the hindbrain and fetal liver. Nissl-stained embryo sections are shown for comparison. SP, spleen; HB, hind brain; L, fetal liver.

The data with wild-type mouse embryos show that Aldh1 mRNA expression is undetectable in the spleen anlage of E13.5 embryos. We next examined in situ hybridization patterns in parasagittal sections of E12.5 and E13.5 null Hox11 embryos. As previously shown (3, 36), Hox11 mRNA is undetectable in Hox11 null embryos at stage E12.5, although a transient spleen anlage does form (3). Aldh2, related to Aldh1, is expressed in a broad spectrum of developing tissues of Hox11 null embryos with a profile which closely approximates that of wild-type mice. However, unlike wild-type embryos, Aldh1 mRNA expression was clearly detectable in the spleen anlage in the Hox11 null embryos, suggesting negative regulation of Aldh1 expression by Hox11 in embryogenesis, rather than positive regulation as determined by the ectopic HOX11 expression in NIH 3T3 cells (Fig. 4). To confirm the significance of this finding, an unrelated gene but one which is also activated by chromosomal translocation in some human T-cell acute leukemias, viz., Tal1/Scl (2), was examined in the Hox11 null embryos to assess whether altered Aldh1 mRNA levels reflected a generalized alteration of gene expression. As with Aldh2 expression, Tal1 mRNA levels at the E13.5 stage did not vary between wild-type and Hox11 null embryos (Fig. 4), the main sites of expression being the fetal liver and specific regions of the hindbrain. We conclude, therefore, that the effects observed in the Hox11 null embryos reflect a specific negative regulatory role of Hox11 protein on the Aldh1 gene in mouse spleen development.

DISCUSSION

HOX11 causes gene activation in NIH 3T3 cells.

HOX11 appears to function as an important regulator of cell growth and survival. Evidence for this comes from its normal role in the development of splenic precursor cells as well as from its tumorigenic role, implied by its association with the breakpoints of chromosomal translocations in T-ALL and its ability to immortalize murine hematopoietic progenitors in vitro (11, 12). HOX11 has been postulated to promote oncogenesis by disrupting a G₂/M checkpoint by binding to protein phosphatase 2A (17). However, it seems likely that HOX11 can also function by activation or repression of genes crucial for cell proliferation and/or differentiation in organogenesis of the spleen.

We have used cDNA RDA (15, 22) to search for potential target genes of HOX11, a T-cell oncogenic homeodomain transcription factor. Using this technique, we have identified Aldh1, a gene encoding an enzyme involved in retinoic acid synthesis, as a gene which can be transcriptionally upregulated by HOX11 in NIH 3T3 cells. Slim1, a novel LIM-only gene, may also be a target for regulation by HOX11 in NIH 3T3 cells, but factors other than simple intracellular expression of HOX11 appear to be involved. SLIM1 belongs to a new family of LIM proteins, which include SLIM2 and DRAL (9, 29, 44), that contain four and a half LIM domains in which the N-terminal domain consists only of the second half of the LIM consensus. Two other LIM-only proteins have been identified via their association with chromosomal translocations, viz., LMO1 and LMO2 (38). This finding suggests an interesting possible link to HOX11, also activated by chromosomal translocations in T-ALL, especially in view of the observed expression of SLIM1 mRNA in most T-ALL cell lines examined. However, the possible significance for development of T-ALL will need to be addressed by functional studies, particularly since not all the 3T3-HOX11 clones display Slim1 activation.

The activation of Aldh1 gene transcription following HOX11 expression in NIH 3T3 cells was found in all NIH 3T3 clones tested, and mutations of HOX11 which affect DNA binding and transcriptional activation (26) affect Aldh1 expression. At present it is not known whether this relationship is a direct effect of HOX11 on the Aldh1 promoter. Although the HOX11 homeodomain is required for Aldh1 expression, it is possible that HOX11 may bind to other intermediate genes whose protein products themselves bind and activate the Aldh1 gene. Alternatively, protein-protein interaction, which must be mediated through the homeodomain, may be responsible for the molecular effect on Aldh1 gene transcription.

A possible Hox11-Aldh1 control system for developing mouse spleen.

In the activation experiments described in this paper, we show that HOX11 ectopically expressed in NIH 3T3 cells results, directly or indirectly, in positive regulation of the Aldh1 gene. Conversely, we present evidence that Aldh1 is negatively regulated by Hox11 in the spleen anlagen of mouse embryos. This paradox is possibly explained by the physiological status of the Aldh1 gene in the two different situations. There is evidence of other homeodomain transcription factors that can act as both positive and negative regulators in differing situations. Two notable examples are the Drosophila homeodomain proteins Antennapaedia and Ultrabithorax, which have been shown to regulate the centrosomin gene. Antennapaedia negatively regulates centrosomin in the visceral mesoderm and positively regulates it in the central nervous system. For Ultrabithorax, the centrosomin regulatory role was found to be reversed in these tissues (13). A mechanism to account for these observations has emerged from a recent study suggesting that HOX protein interactions with cofactors such as PBX1 can determine whether target gene activation or repression will occur (34).

A major function of Aldh1 appears to be the biosynthesis of retinoic acid by conversion of vitamin A (retinol) to its biologically active form, retinoic acid, a local regulator of cellular proliferation, differentiation, and death. During retinoid signalling, retinol is converted to retinoic acid via two oxidative reactions in target cells (19). The irreversible second step in which retinoic acid is generated from retinal is catalyzed mainly in the adult and exclusively in the early embryonic mouse by NAD-dependent Aldh enzymes (27). The class 1 or cytosolic form (Aldh1) and the closely related V1 and V2 or Raldh-2 retinal dehydrogenases are the only Aldh isoenzymes known to have this catalytic capacity (49, 51). It is noteworthy that aldehyde oxidase, an enzyme which also has the capacity to synthesize retinoic acid in vertebrates (32, 46), has recently been proposed to be downstream of Drosophila homeoproteins (20). Retinoic acid acts as a ligand for two families of nuclear receptors, the retinoic acid and retinoid X receptors. This diffusable morphogen is crucial to vertebrate embryogenesis, where it is distributed in patterns that are highly regulated, both temporally and spatially, by the localized expression of retinoic acid-synthesizing enzymes, including Aldh1 (1, 25, 33). As a developmental signal, retinoic acid is involved in numerous processes ranging from anteroposterior patterning, both along the main body axis and in developing limb buds, to whole-organ formation. The need for vigilant control to be maintained over retinoic acid biosynthesis is emphasized by the highly teratogenic effect of a deficiency or excess of retinoic acid during development (14, 40). Our results are compatible with a role for Hox11 in regulating retinoic acid availability in an organ-specific manner by suppressing Aldh1 expression in the developing spleen. Aberrant expression of Aldh1 in spleen primordia of Hox11 null mice provides a plausible molecular mechanism to account for splenic involution, since it is known that retinoids can regulate apoptosis. This has been demonstrated in vitro as well as in whole organs in vivo (16, 24, 42). Very little is known about the endogenous levels of retinoic acid or its biosynthetic enzymes in the developing spleen; however, at least in the adult art, the spleen does not normally demonstrate detectable retinoic acid-synthesizing activity (32). Therefore, an aberrant increase in retinoic acid synthesis in the spleen anlage at that particular stage of development may lead to the observed asplenic phenotype of Hox11 null mutant mice.