Hematopoietic, angiogenic and eye defects in Meis1 mutant animals

Tomoyuki Hisa, Sally E Spence, Rivka A Rachel, Masami Fujita, Takuro Nakamura, Jerrold M Ward, Deborah E Devor‐Henneman, Yuriko Saiki, Haruo Kutsuna, Lino Tessarollo, Nancy A Jenkins, Neal G Copeland

Author Affiliations

  1. Tomoyuki Hisa1,,
  2. Sally E Spence1,,
  3. Rivka A Rachel1,
  4. Masami Fujita1,,
  5. Takuro Nakamura2,
  6. Jerrold M Ward3,
  7. Deborah E Devor‐Henneman3,
  8. Yuriko Saiki2,
  9. Haruo Kutsuna4,
  10. Lino Tessarollo1,
  11. Nancy A Jenkins1 and
  12. Neal G Copeland*,1
  1. 1 Center for Cancer Research, Mouse Cancer Genetics Program, National Cancer Institute, Frederick, MD, USA
  2. 2 The Cancer Institute, Japanese Foundation for Cancer Research, Tokyo, Japan
  3. 3 Center for Cancer Research, Veterinary and Tumor Pathology Section, National Cancer Institute, Frederick, MD, USA
  4. 4 Department of Hematology, Osaka City University Medical School, Osaka, Japan
  1. *Corresponding author. Center for Cancer Research, Mouse Cancer Genetics Program, National Cancer Institute, Frederick, MD 21702‐1201, USA. Tel.: +1 301 846 1260; Fax: +1 301 846 6666; E-mail: copeland{at}
  1. These authors contributed equally to this work

  • Present address: Department of Obstetrics and Gynecology, Osaka University Medical School, Osaka, Japan

View Full Text


Meis1 and Hoxa9 expression is upregulated by retroviral integration in murine myeloid leukemias and in human leukemias carrying MLL translocations. Both genes also cooperate to induce leukemia in a mouse leukemia acceleration assay, which can be explained, in part, by their physical interaction with each other as well as the PBX family of homeodomain proteins. Here we show that Meis1‐deficient embryos have partially duplicated retinas and smaller lenses than normal. They also fail to produce megakaryocytes, display extensive hemorrhaging, and die by embryonic day 14.5. In addition, Meis1‐deficient embryos lack well‐formed capillaries, although larger blood vessels are normal. Definitive myeloerythroid lineages are present in the mutant embryos, but the total numbers of colony‐forming cells are dramatically reduced. Mutant fetal liver cells also fail to radioprotect lethally irradiated animals and they compete poorly in repopulation assays even though they can repopulate all hematopoietic lineages. These and other studies showing that Meis1 is expressed at high levels in hematopoietic stem cells (HSCs) suggest that Meis1 may also be required for the proliferation/self‐renewal of the HSC.


Meis1 expression is upregulated by retroviral integration in ∼15% of BXH2 myeloid leukemias (Moskow et al, 1995; Nakamura et al, 1996c). Meis1 encodes a member of the three amino acid loop extension (TALE) family of homeodomain‐containing proteins. Within the homeodomain, MEIS1 is most closely related to the PBX family of homeodomain proteins. The founding member of this family, PBX1, was identified as the fusion partner of TCF3 (E2A) in human pre‐B‐cell leukemias carrying a t(1;19)(q23;p13.3) translocation (Kamps et al, 1990; Nourse et al, 1990). Three additional PBX1‐related genes have since been identified (Monica et al, 1991; Wagner et al, 2001), as have three other Meis1‐related genes (Nakamura et al, 1996a; Berthelsen et al, 1998). Most leukemias with retroviral integrations at Meis1 also have retroviral integrations upstream of Hoxa7 or Hoxa9 (Nakamura et al, 1996c), with concomitant activation of Hoxa7 or Hoxa9 expression. This striking selection for integration and activation of Meis1 in concert with Hoxa7/Hoxa9 strongly suggests that these genes cooperate to induce leukemia. Additional evidence for cooperation has been provided by Sauvageau and colleagues (Kroon et al, 1998), who showed that mouse bone marrow cells engineered to overexpress Meis1 and Hoxa9 induce growth factor‐dependent oligoclonal acute myeloid leukemia (AML) at <3 months when transplanted into syngenic mice. In contrast, Meis1 overexpression failed to transform these cells alone, while Hoxa9 overexpression induced AML but with a longer latency period (Thorsteinsdottir et al, 2001).

In humans, HOXA9 is fused to NUP98 in AMLs carrying a t(7;11)(p15;p15) translocation (Borrow et al, 1996; Nakamura et al, 1996b). MEIS1 and HOXA9 are also frequently coexpressed in human AMLs (Lawrence et al, 1999; Afonja et al, 2000) and are upregulated in acute lymphocytic leukemias carrying MLL translocations (Rozovskaia et al, 2001). MEIS1 may therefore also be a human leukemia disease gene.

The affinity and specificity of DNA binding by HOX proteins is augmented by their interaction with PBX. In vitro studies have shown that MEIS1 can also physically interact with HOX proteins (HOXA9) by forming heterodimeric‐binding complexes on a DNA target containing a MEIS1 site and an AbdB‐like HOX site (Shen et al, 1997). MEIS proteins are also major in vivo DNA‐binding partners for PBX proteins (Chang et al, 1997) and can form heterotrimeric complexes with PBX and HOX (Jacobs et al, 1999; Shanmugam et al, 1999; Shen et al, 1999; Schnabel et al, 2000). The basis for cooperation between MEIS1 and HOX is therefore likely to be complicated and possibly mediated by heterotrimers, heterodimers, and/or homodimers involving these different proteins.

The ortholog of Meis1 in Drosophila is encoded at the homothorax (hth) locus (Rieckhof et al, 1997; Pai et al, 1998). HTH acts upstream of extradenticle (EXD), which is the Drosophila ortholog of PBX, and is required for the nuclear localization of EXD. In Drosophila, many functions have been ascribed to HTH and EXD, including the regulation of eye development (Pai et al, 1998), patterning the embryonic peripheral nervous system (Kurant et al, 1998), and proximal–distal limb development (Mercader et al, 1999).

While homeobox genes were first recognized for their role in pattern formation, it has become increasingly clear in recent years that homeobox genes also have important functions in hematopoiesis (reviewed by, Chiba, 1998). Given MEIS1's role as a HOX and PBX DNA‐binding cofactor and its involvement in leukemogenesis, it would not be surprising if MEIS1 were also required for normal hematopoiesis. Here we confirm this prediction and show that mice carrying a germline mutation in Meis1 have multiple hematopoietic defects. We also show that Meis1 is required for the development of the microvasculature and the mammalian eye, a role consistent with a recent report suggesting that MEIS1 regulates Pax6 expression during vertebrate lens development (Zhang et al, 2002).


Targeted disruption of Meis1

The Meis1 knockout mutation was made by replacing part of the homeodomain (exon 8) with a pSAβgeo gene trap cassette (Figure 1A). pSAβgeo contains a splice acceptor site followed by stop codons in all three reading frames, a β‐galactosidase‐neomycin fusion gene initiated from an internal AUG, and a bovine growth hormone polyadenylation signal sequence (Friedrich and Soriano, 1991). βgeo will therefore be transcribed from the Meis1 promoter in mutant animals. Male chimeras carrying this mutation were mated to C57BL6/NCr (B6) females to generate a mixed B6;129 line and to 129S3/SvImJ females to produce a 129 line. Subsequent intercrosses of heterozygous mutant animals failed to produce any homozygous mutant pups in 297 B6;129 heterozygous intercross progeny or in 239 heterozygous 129 intercross progeny, indicating that Meis1 is required for embryonic development. Viable Meis1−/− B6;129 embryos were observed as late as 14.5 days post‐coitus (dpc) (Table I, Figure 1B), although after 13.5 dpc many of the embryos were pale and exhibited hemorrhaging (see below). Meis1−/− 129 embryos were not identified after 12.5 dpc, although they had similar hemorrhages. B6 and 129 animals therefore differ in one or more genetic modifiers of Meis1. In subsequent experiments only animals from the B6;129 background were used due to the early embryonic death of Meis1−/− 129 embryos.

Figure 1.

Meis1‐targeting cassette. (A) Meis1 exon 8 was replaced with a pSAβgeo gene trap targeting cassette. The targeting cassette also contains a thymidine kinase gene driven by a PGK promoter for the negative selection of incorrectly targeted ES cells. Homeodomains located in exons 8 and 9 are represented by black boxes. The 5′ and 3′ probes used for Southern blot analysis are indicated as striped boxes; the sizes of SacI‐hybridizing fragments are also indicated. Restriction sites in the introns flanking the targeted exons and the diagnostic restriction fragments for homologous recombinants are indicated (A, ApaI, C, ClaI, K, KpnI, S, SacI, X, XhoI). (B) Southern blot analysis of yolk sac DNA prepared from E13.5 embryos generated from a B6;129 Meis1+/− intercross and digested with SacI. The wild‐type Meis1 allele is 6.5 kb while the mutated allele is 2.2 kb in size.

View this table:
Table 1. Meis1 mutation is embryonic lethal

As expected, RT–PCR analysis showed that Meis1 exons 1–7 are fused to pSAβgeo sequences in the mutant embryos (data not shown). In contrast, no amplification products were detected when primers from exons 7 and 9 were used for amplification, indicating that all Meis1 transcripts terminate in pSAβgeo sequences. Western blot analysis of wild‐type E13.5 embryos using a MEIS1‐specific antibody identified a wild‐type 46 kDa MEIS1 protein that was virtually undetectable in mutant embryos. Heterozygous and homozygous mutant embryos also expressed a smaller, fainter 33 kDa MEIS1 protein (Supplementary Figure 1). This protein is approximately the size predicted for a truncated protein produced from exons 1–7. This protein would contain the MEIS‐homology (MH) domain, which interacts with PBX. The MH domain alone causes a phenotype when ectopically expressed in flies (Jaw et al, 2000), and mutational analysis of HTH has demonstrated that truncated forms of the protein lacking the homeodomain retain part of their activity (Kurant et al, 2001). It is therefore possible that this truncated MEIS1 protein can act as a dominant negative, or that the null phenotype would be slightly different from what is described here for our mutant animals. No differences, however, between heterozygous and wild‐type littermates were detected in any of our studies (see below), suggesting that this protein does not act as a strong dominant negative.

Meis1 expression

By β‐galactosidase staining, lacZ expression from the Meis1 promoter (hereafter Meis1‐lacZ) was detected in specific and discrete locations throughout the embryo in a pattern that changed from embryonic day 11 (E11) to E13 (Figure 2A–D). Meis1‐lacZ staining was comparable to the pattern observed by in situ hybridization using a Meis1 cDNA probe (not shown), but revealed distinct patterns of expression in small regions of the central nervous system and sensory structures of the head, including the ears, eyes, and nose (Figure 2A and C). In the head, strong staining in the eye (white arrowheads in Figure 2A and C) and external ear primordia (ea in Figure 2C) was visible at E11, as well as a faint band of expression in the cochlear neuroepithelium (arrow in Figure 2E). Other regions of intense staining were the soft tissues of the mediastinum (m) and midgut (mg) (Figure 2A, B and D) and the spinal cord, which displayed several intense longitudinal stripes (black arrows; Figure 2B and D). Internal organs with strong expression included all four cardiac chambers and the lungs. This pattern was recapitulated at E13 with some expansion of labeling in the brain. At E13, a thin band of staining in the neuroepithelium of the cochlear canal was also observed (arrow in Figure 2E). By this age, it is apparent that the labeling in the olfactory epithelium occurs transiently in a dorsal to ventral and medial to lateral progression, following the maturational gradient of the olfactory epithelium, with Meis1‐lacZ expressed in the newest part of the epithelium only (arrows in Figure 2F).

Figure 2.

Meis1 expression revealed by β‐galactosidase staining. Whole‐mount (A, C, G, H), sagittal (B, D, I, J), and transverse (E, F, K, L) sections of heterozygous Meis1+/− (A, C, G, I, K) and homozygous Meis1−/− (B, D, E, F, H, J, L) mutant embryos at E11.5 (A, B) and E13.5 (C–L) days of embryonic development stained for Meis1‐lacZ expression. Anteroposterior (AP) and dorsoventral (DV) axes are marked.

In general, staining in Meis1−/− animals was more widespread and intense than in Meis1+/− animals, which likely reflects a higher dosage of the reporter gene. However, several striking differences were apparent. At E13, a strong anteroposterior stripe of expression running on the surface of the face from the eye to the nose was found in homozygotes (white asterisks in Figure 2C, F and H) and this expression was completely absent in heterozygotes (Figure 2G). In heterozygotes, Meis1‐lacZ was also expressed in scattered cells in the liver at both E11 and E13; in homozygotes, expression was observed in only a few cells in the liver at E11 and in no cells at E13. These cells are megakaryocytes, which die in the absence of Meis1 activity (see below). Strong circumferential expression around the eye field was also observed only in homozygotes (black arrows in Figure 2H and L). By contrast, in heterozygotes, but not in homozygotes, Meis1‐lacZ was expressed in the posterior lens (arrows in Figure 2K) and surface ectoderm (arrowheads in Figure 2K).

Some structures that reacted positive for β‐galactosidase in the Meis1−/− embryos retained normal morphology and structure at E13.5. A striking exception was the eye. In homozygotes, the lens was smaller (asterisks in Figure 2J and I), but contact between the lens and surface ectoderm was maintained (white arrowheads in Figure 2L). The retina was also abnormal (compare Figure 2G, I and K to Figure 2H, J and L) and may be partially duplicated, although this possibility will need to be confirmed by additional marker analysis. These abnormalities are not visible at E11 (A) and could only be seen at E13 (Figure 2F, H, J and L). In addition to the eye, abnormalities were also observed in the brain, heart, lungs, and kidneys, among others, suggesting pleiotropic effects of this gene in organogenesis. While the exact nature of these abnormalities remains to be elucidated, the defects in the brain are consistent with the known role of Meis genes in patterning of the hindbrain (Dibner et al, 2001; Choe et al, 2002). By contrast, at this early stage we did not detect any obvious proximal or distal limb defects as has been reported in Drosophila hth embryos (Mercader et al, 1999). This was surprising, given that MEIS1‐related proteins play critical roles in proximodistal axis formation in the limbs of both flies (Wu and Cohen, 1999) and chicken (Mercader et al, 1999). The gene is also expressed in the proximal limb buds of mice (Saleh et al, 2000) and a Pbx1 knockout has demonstrated an obligatory role for this gene in patterning of the limbs (Selleri et al, 2001).

Megakaryocytes are absent in Meis1−/− embryos

Microscopic evaluation revealed extensive hemorrhaging in the brain and trunk of E13.5 Meis1−/− embryos, although the extent of hemorrhaging varied among the mutant embryos (Figure 3A). Detailed microscopic analysis of the yolk sac did not reveal significant morphologic differences between mutant and control littermates except for loss of blood within yolk sac blood vessels, and the bleeding seen in Meis1−/− embryos appeared to coincide with the time of embryonic death. Because bleeding could result in part (although not exclusively, see below) from megakaryocytes/platelet defects, we stained livers from E13.5 heterozygous and homozygous mutant embryos with hematoxylin and eosin and examined them for megakaryocytes. Megakaryocytes were clearly visible in Meis1+/− livers but not in Meis1−/− livers (Figure 4A and B). Likewise, β‐galactosidase‐positive megakaryocytes were seen in Meis+/− livers but not in Meis1−/− livers (Figure 4C and D). Cells expressing CD41, a megakaryocytic marker (Shattil et al, 1985; Gewirtz, 1995), were also absent in the Meis1−/− livers (Figure 4E and F). These results show that Meis1 is expressed in megakaryocytes and is essential for their development.

Figure 3.

Phenotype of Meis1 mutant embryos and transplant recipients. (A) A wild‐type E13.5 and two Meis1−/− mutant B6;129 littermates. At this stage, most Meis1−/− embryos exhibited hemorrhaging in the brain and trunk. The level of hemorrhaging varied between mutant embryos, with most embryos showing the very extensive hemorrhaging seen in the embryo on the right. (B) FACS analysis of bone marrow cells 130 days or more after transplantation of 2 × 106 Ly5.1+ fetal liver cells derived from the embryos shown in (A) mixed with 5 × 105 wild‐type C57BL/6‐Ly5.2+ adult bone marrow cells transplanted to lethally irradiated C57BL/6‐Ly5.2+ hosts. Transplant recipients were stained with FITC‐labeled antibodies directed against Ly5.1+ or Ly5.2+ cells to discriminate between donor and host hematopoietic cells.

Figure 4.

Megakaryocyte defects in Meis1 mutant embryos. Hematoxylin and eosin staining of paraffin sections through the liver of E13.5 Meis1+/− or Meis1+/+ embryos (A, C, E) compared to Meis1−/− (B, D, F) embryos show large cells consistent with megakaryocytes in Meis1+/− and Meis1+/+ embryos (arrows in (A)). These cells are completely missing in Meis1−/− embryos by E11 (B). Likewise, β‐galactosidase staining was found in a subpopulation of large liver cells (i.e., megakaryocytes) in Meis1+/− embryos (arrows in (C)), but this cell population was absent in Meis1−/− embryos (D). Immunoreactivity for CD41, a megakaryocyte marker (Gewirtz, 1995; Shattil et al, 1985), normally present on a subset of liver cells at E13.5 (brown spots in (E)), was also lacking in Meis1−/− embryos (F).

To confirm these observations, we cultured E13.5 fetal liver cells in vitro in the presence of thrombopoietin, interleukin‐3 (IL‐3), and IL‐6. These culture conditions stimulate hematopoietic megakaryocytic progenitors, colony‐forming units‐megakaryocytes (CFU‐Meg), to grow in vitro (Kaushansky, 1995; Broudy and Kaushansky, 1998). After 6–8 days, cultures were stained for acetylcholinesterase activity and the number of acetylcholinesterase‐positive CFU‐Meg scored. In one experiment, an average of 7.2±2.4 CFU‐Meg colonies was scored in cultures derived from 5 × 104 wild‐type fetal liver cells (n=4), while a similar number of CFU‐Meg colonies (6.3±3.4) were scored in Meis1+/− cultures (n=4). In contrast, no CFU‐Meg colonies were detected in cultures derived from Meis1−/− embryos (n=4, data not shown). Similar results were observed in two other experiments.

Recent studies have shown that MEIS1/PBX complexes can synergistically transactivate the platelet factor 4 (Pf4) gene in combination with GATA1 and ETS1 (Okada et al, 2003). Pf4 is a lineage specific marker that appears during late stages of megakaryocytic differentiation (Breton‐Gorius and Vainchenker, 1986). Retinoic acid, an upstream activator of Meis1 in vertebrate limb development, also stimulates megakaryocytopoiesis in the presence of thrombopoietin or IL‐3 and GM‐CSF (Visani et al, 1999). These results raise the possibility that MEIS1 may be a master regulator of megakaryocytic gene expression.

Myeloerythroid colony‐forming cells are reduced in Meis1−/− fetal livers

To determine whether other hematopoietic lineages are affected in Meis1−/− mutant embryos, 2 × 104 fetal liver cells from E13.5 embryos were grown in methylcellulose for 6–8 days in SCF, GM‐CSF, G‐CSF, and EPO and the colonies were scored by morphological criteria for BFU‐E, CFU‐E, GM‐CFU, and CFU‐mix colonies. Two‐tailed t‐tests of the total cellularity of the Meis1−/− fetal livers (4.3 × 106±3.1 × 105, n=40 livers) (mean and standard error) showed a decrease (P<0.001, t=6.361, 68 degrees of freedom) compared to wild‐type (7.5 × 106±4.1 × 105, n=30 livers) or Meis1+/− (5.9 × 106±3.3 × 105, n=61) (P=0008, t=3.472, 99 degrees of freedom) fetal livers. The total number of colony‐forming units per liver in Meis1−/− embryos also showed a two‐ to five‐fold reduction compared to wild‐type livers (Figure 5). CFU‐mix colonies were most affected (five‐fold), while the least affected were GM‐CFU colonies (two‐fold). The hematopoietic defects in Meis1−/− embryos are therefore not limited to megakaryocytes.

Figure 5.

Myeloerythroid colony‐forming cells are reduced in Meis1−/− fetal livers. Fetal liver cells were isolated from E13.5 wild‐type and Meis1 mutant embryos and scored for the presence of myeloerythroid colony‐forming cells. The data shown represent the means and standard deviations from duplicate platings of wild‐type (n=3, crosshatched bars), Meis1+/− (n=5, horizontally hatched bars) and Meis1−/− (n=2, solid bars) fetal livers, and are representative of two separate experiments.

Meis1−/− fetal liver cells compete poorly in reconstitution assays

To determine whether Meis1−/− fetal liver cells can reconstitute the hematopoietic system of lethally irradiated mice, 2 × 106 fetal liver cells from E13.5 embryos were transferred to lethally irradiated hosts. All mice receiving Meis1−/− cells (n=5) died within 10 days post‐transplantation, while mice receiving wild‐type (n=4) or Meis1+/− (n=7) cells survived for 40 days (data not shown). Meis1−/− fetal liver cells are therefore unable to radioprotect lethally irradiated mice. These data are consistent with the decrease in myeloerythroid colony‐forming cells observed in vitro in Meis1−/− fetal livers.

To determine whether definitive pluripotent hematopoietic progenitors are functionally impaired in Meis1−/− embryos, 2 × 106 fetal liver cells from E13.5 mutant embryos were mixed with 5 × 105 wild‐type C57BL/6NCr‐Ly5.2+ adult bone marrow cells and injected into lethally irradiated C57BL/6NCr‐Ly5.2+ hosts. Meis1 mutant embryos are Ly5.1+, and this cell surface marker can therefore be used to follow the presence of Meis1 mutant cells in the transplanted hosts. All mice receiving a mixture of mutant and wild‐type cells survived 130 days post‐transplant, with the exception of one animal that died from causes unrelated to the transplant (data not shown). Mutant Ly5.1+ hematopoietic cells in long‐term reconstituted recipients were then scored by FACS analysis. The bone marrow of mice transplanted with Meis1−/− cells contained 18.1±7.7% Ly5.1+ cells compared to 85.3±5.2 or 85.9±2.2% Ly5.1+ cells in recipients receiving Meis1+/+ or Meis1+/− cells, respectively (Figure 6). There was some variability in engraftment, however, as three of the recipients receiving Meis1−/− cells showed 30, 48, and 72% engraftment of Ly5.1+ cells. These transplant recipients were invariably engrafted with cells taken from embryos that had reduced hemorrhaging (Figure 3B). These embryos may carry one or more Meis1 modifier genes that protect them from the effects of Meis1 loss or, alternatively, are slightly delayed in their development compared to Meis1−/− embryos with more extensive bleeding, and thus have more Ly5.1+ hematopoietic stem cells (HSCs) that can repopulate the hematopoietic system of irradiated hosts.

Figure 6.

Meis1 mutant fetal liver cells compete poorly in hematopoietic reconstitution assays. Two million fetal liver cells isolated from E13.5 embryos produced from an N7 C57BL/6‐Ly5.1+ Meis1+/− intercross plus 4 × 105 adult bone marrow cells from C57BL/6‐Ly5.2+ mice were transplanted to irradiated C57BL/6‐Ly5.2+ hosts. The genotype of each embryo was determined by Southern blot analysis of yolk sac DNA. At 130 days or greater after transplantation, recipients of fetal liver cells from wild‐type (closed squares, n=3), Meis1+/− (closed circles, n=11), or Meis1−/− embryos (closed triangles, n=10) were analyzed for donor engraftment (Ly5.1+ cells) in the bone marrow, spleen, and thymus by flow cytometry. A horizontal line indicates the mean level of engraftment.

The level of engraftment in the spleen was similar to that observed in bone marrow. In the spleen, 18.2±3.3% of the cells in mice receiving Meis1−/− cells were Ly5.1+ compared to 86.0±3.2 or 82.9±2.7% in recipients receiving Meis1+/+ or Meis1+/− cells, respectively (Figure 6). In the thymus, 4.1±1.9% of the cells in mice receiving Meis1−/− cells were Ly5.1+ compared to 93.6±0.3 or 79.9±5.2% in recipients receiving Meis1+/+ or Meis1+/− cells, respectively (Figure 6). In the three mice showing >30% engraftment of Meis1−/− cells in the bone marrow, increased engraftment of Meis1−/− cells was also seen in the spleen of all mice (25, 25, 29%) and the thymus of one mouse (21%).

The bone marrow, spleen, and thymus of mice showing >30% engraftment in the bone marrow were further analyzed for the presence of Ly5.1+ cells expressing cell surface markers characteristic of primitive progenitors (CD117, CD34) or more committed lineage‐restricted cells (Mac‐1, Gr‐1, CD61, Ter119, B220, CD3, DX5) (data not shown). These studies showed that Meis1−/− cells are capable of differentiating into granulocytes, neutrophils, erythrocytes, B‐ and T‐cells, and NK cells in transplanted hosts, although the overall level of engraftment of each cell type was severely reduced compared to animals transplanted with Meis1+/+ or Meis1+/− cells.

Defective angiogenesis in Meis1−/− embryos

Although Meis1−/− embryos fail to make megakaryocytes/platelets, it is unlikely that this alone causes the hemorrhaging seen in Meis1−/− embryos since Nfe2−/− mutant embryos, which have platelet defects due to a block in megakaryocyte maturation (Shivdasani et al, 1995), are born in normal numbers and do not hemorrhage in utero. To determine whether hemorrhaging in Meis1−/− mutant embryos could result from defects in vasculogenesis (the differentiation of mesodermally derived endothelial cell precursors into vascular channels) or angiogenesis, (the formation of smaller vessels and the microvasculature), we evaluated the large blood vessels and microvasculature in Meis1−/− mutant embryos and control littermates at E13.5 (Figure 7). In Meis1−/− embryos, expressions of α‐smooth muscle actin (SMA) and PECAM (CD31), a marker for endothelial cells, were abnormal. Unlike normal embryos, where antibodies to these molecules outline branching capillaries with a clear lumen and consistent size (Figure 7A and C), in mutant embryos these markers were found together in close association but failed to delineate a clear lumen (Figure 7B and D). Moreover, the branching pattern and size of the vessels was abnormal. In normal embryos the capillaries were just large enough to enclose red blood cells (RBCs) (Figure 7C), while in Meis1−/− embryos the small vessels were either too narrow to admit RBCs (Figure 7B and D) or formed larger sinuses where the RBCs pooled. Numerous extravagated RBCs were also found outside the microvasculature in surrounding tissues in Meis1−/− embryos. In contrast, the major blood vessels were normal in Meis1−/− embryos (Figure 7E and F).

Figure 7.

Meis1 mutant embryos have angiogenesis defects. Confocal images (AF) were acquired of blood vessels identified by PECAM/CD31 (red) and SAM (green) from normal (A, C, E) and Meis1−/− mutant (B, D, F) E13.5 embryos. Nuclei are labeled with DAPI (blue). A 15–20 μm stack of images was combined to generate a three‐dimensional projection (A, B), showing a branching capillary with clear lumen in the normal (A) but lack of this structure in the mutant (B). Normal capillaries displayed consistent size, while those in the Meis1−/− tissue did not (arrows in A, B). In 1 μm confocal images, RBCs normally filled the lumen of capillaries (arrows in C). This feature was lacking in Meis1−/− embryos, which formed irregular and disorganized microvasculature (D). The major blood vessels showed exact colocalization of PECAM (red) and SMA (green), resulting in a yellow signal (E, F). Large blood vessels (e.g. major arteries leaving the thorax) were present in Meis1−/− embryos and displayed normal lumen formation (F). Confocal images (GI) of wild‐type E13.5 embryo sections stained with antibodies to MEIS1 (red) and either PECAM/CD31 (green, G, H) or SMA (green, I). Nuclei are labeled with DAPI (blue). Various regions of the embryo show MEIS1 and PECAM in close but nonoverlapping proximity (G). Other regions of angiogenesis express PECAM without MEIS1 (H). Yet other areas are positive for both an angiogenic marker (SAM) and MEIS1. Original magnification, 100 ×.

To determine whether the capillary defects observed in Meis1−/− embryos are cell autonomous or cell non‐autonomous, we stained tissue sections of E13.5 wild‐type embryos with antibodies to MEIS1 and PECAM or SMA, markers of vascular endothelium and vascular smooth muscle, respectively. While we sometimes observed MEIS1 expression in the vicinity of PECAM and SMA, we did not see a direct correlation between the sites of MEIS1 and PECAM and SMA expression (Figure 7G–I). Moreover, PECAM‐1 and SMA were often found in regions that lacked MEIS1 expression and vice versa. We also observed individual Meis1‐lacZ‐positive cells in regions of hemorrhage in the dorsal flanks of Meis1−/− embryos, but not in a pattern suggestive of capillaries (data not shown). In addition, in no region could we detect Meis1‐lacZ expression in a blood vessel‐specific pattern. Taken together, these results suggest that the angiogenic defects in Meis1−/− embryos are cell nonautonomous with respect to MEIS1 expression, although conditional gene knockouts will be required to confirm these results.


Here we show that Meis1, a member of the TALE family of homeobox genes, is required for embryonic viability, eye development, and maintenance of definitive hematopoiesis. Meis1‐deficient embryos also lack megakaryocytes, display extensive hemorrhaging in the brain and trunk, and have defects in angiogenesis. The eye defects observed in Meis1−/− embryos are consistent with recent studies suggesting that Meis1 is a regulator of Pax6 expression in prospective lens ectoderm (Zhang et al, 2002). Pax6 is a pivotal regulator of eye development throughout Metazoans, but the upstream regulators of Pax6 have remained elusive. A strong anteroposterior stripe of Meis1 expression running on the surface of the face from the eye to the nose was also seen in Meis1−/− embryos. This expression was absent in Meis1+/− embryos. The cause of this ectopic expression remains to be determined, but one possibility is that it represents the presence of additional eye primordia. Evidence for this hypothesis comes from flies where HTH and EXD function as negative regulators of eye develop and delimit the eye fields (Pai et al, 1998).

Defects in primitive hematopoiesis were not observed in Meis1−/− embryos. Definitive hematopoiesis is, however, initiated in Meis1−/− mutant embryos as evidenced by the colony formation of fetal liver committed progenitors and multilineage reconstitution of irradiated hosts following transplantation of Meis1−/−mutant fetal liver cells (Palis et al, 1999). Compared to wild‐type and heterozygous mutant cells, E13.5 Meis1−/− cells competed poorly in reconstitution assays, suggesting a possible decrease in the number of HSCs and/or a defect in the proliferation or differentiation of more committed progenitors. Meis1 appears to act later in embryonic development than other transcription factors such as GATA3 (Pandolfi et al, 1995), RUNX1 (Okuda et al, 1996; Wang et al, 1996a), or CBFB (Wang et al, 1996b), which produce more severe defects in definitive hematopoiesis. Hemorrhaging and death in these mutant embryos occurs between 11.5 and 12.5 dpc.

Pbx1 and Meis1 mutant animals share several phenotypes in common, including defects in definitive hematopoiesis (DiMartino et al, 2001). This is interesting since PBX1 is a major in vivo DNA‐binding partner of MEIS1. Definitive myeloerythroid lineages are present in Pbx1 and Meis1 mutant animals, and the number of colony‐forming cells is substantially reduced in both (DiMartino et al, 2001). Pbx1−/− fetal liver cells also perform poorly in competitive repopulation assays, as do Meis1−/− fetal liver cells. Both mutants also display defects in kidney development (Schnabel et al, 2003 and our unpublished observations). There are some differences, however, in the two mutant phenotypes. Meis1 mutant animals do not develop the skeletal abnormalities observed in Pbx1−/− mice (Selleri et al, 2001). They also fail to make megakaryocytes/platelets and they show extensive hemorrhaging, unlike Pbx1−/− mutant animals, which make platelets and do not hemorrhage. Pbx1−/− embryos die later in development than Meis1−/− embryos (15.5–16.5 dpc versus 12.5–14.5 dpc). Pbx1−/− fetal liver cells also radioprotect lethally irradiated mice, although they are only half as effective at radioprotection as Pbx1+/+ cells (Chang et al, 1997), in contrast to Meis1−/− cells, which are unable to radioprotect lethally irradiated mice. Perhaps other PBX proteins that can interact with MEIS1 provide partial compensation for the loss of Pbx1 in knockout mice.

Meis1 is expressed at the highest levels in the Sca‐1+Lin fraction of the adult bone marrow and E14.5 fetal liver (Pineault et al, 2002). This fraction is highly enriched in HSCs (Rebel et al, 1996). In contrast, the expression of Meis1 in more committed progenitors is barely detectable or undetectable. A close correlation between the onset of Meis1 expression in embryoid body development and the appearance of hematopoietic cells has also been seen (Pineault et al, 2002). Many Hox genes are expressed at their highest levels in the Sca1+Lin fraction and are rapidly downregulated during hematopoietic differentiation (Pineault et al, 2002), findings consistent with a major role of MEIS1 in normal hematopoiesis as a HOX cofactor. Meis1 might therefore be required for the proliferation/self‐renewal of the HSC, a hypothesis that is supported by studies presented here and the failure of Meis1−/− fetal liver cells to contribute efficiently to multiple hematopoietic lineages in transplanted lethally irradiated recipients.

HSCs produce angiopoeitin (AGPT), a molecule that is critical for angiogenesis (Takakura et al, 2000). Embryos deficient in HSCs (Takakura et al, 2000), or that have defects in Agpt (Suri et al, 1996) or the AGPT receptors Tie1 and Tie2 (Puri et al, 1995; Sato et al, 1995), exhibit microvascularization defects and show extensive hemorrhaging and death between E9.5 and E13.5. These results, and those suggesting that the vascular defects in Meis1−/− embryos are cell nonautonomous with respect to MEIS1 expression, provide further evidence for an HSC defect in Meis1−/− embryos and suggest that hemorrhaging in Meis1−/− animals is due to an HSC defect resulting from a decrease in the secretion of AGPT and the failure to form normal microvascular channels.

Materials and methods

Meis1 mutant animals

Meis1 exon 8 was replaced with a pSAβgeo cassette (Figure 1A). Homologous targeting events were identified by Southern blotting using 0.5 kb SacI‐KpnI (5′) or 1.9 kb SacI‐XhoI (3′) probes (Figure 1A), labeled using a Megaprime DNA kit (Amersham Biosciences, Piscataway, NJ). Several independent ES cell clones carrying the knockout allele were isolated and the mutation was introduced into the germline by standard techniques. Mice were maintained under limited access conditions at the National Cancer Institute (Frederick) and animal care was provided according to the procedures outlined in the Guide for the Care and Use of Laboratory Animals, under an approved animal care and use committee protocol.

Immunohistochemical and confocal analysis

Embryos were collected at 12.5–13.5 dpc with the morning of vaginal plug detection considered as 0.5 dpc. Embryos were dissected from the uterus and fixed in 4% paraformaldehyde in PBS. For some studies, embryos were embedded in paraffin. After dewaxing, sections of embedded embryos were incubated with a rat anti‐mouse CD41 antibody (MWReg30, BD Pharmingen, San Diego, CA) followed by a biotinylated antibody directed against rat IgG antibody labeled with the avidin/biotinylated enzyme complex (Vector Laboratories, Burlingame, CA). A signal was detected with 3,3′‐diaminobenzidine (Vector Laboratories) as chromogen. For other studies, embryos were embedded in 3% agarose, vibratome‐sectioned, and immunostained with DAPI (Molecular Probes, Eugene, OR) and antibodies to PECAM/CD31 (BD Pharmingen, San Diego, CA) and SMA (Dako, Glustrup, Denmark). Secondary antibodies were Cy2‐ and Cy3‐conjugated goat anti‐mouse or goat anti‐rat IgG (Jackson Immunoresearch, West Grove, PA). Images were acquired on a Zeiss 510 confocal laser‐scanning microscope (Carl Zeiss, Germany). For three‐dimensional analysis of blood vessels, stacks of images were acquired and converted to a maximum intensity projection.

X‐gal staining

Paraformaldehyde‐fixed embryos at E11.5 and E13.5 were stained overnight at 37°C in the dark with 1 mg/ml X‐gal (US Biological, Swampscott, MA), 4 mM K4Fe(CN)63H2O, 4 mM K3Fe(CN)6, 2 mM MgCl2 in PBS to detect beta‐galactosidase (beta‐gal) activity.

Myeloerythroid colony assays

Megakaryocyte colony‐forming units were assayed by plating 5 × 104 fetal liver cells on chamber culture slides in collagen‐based medium (Stem Cell Technologies, Vancouver, BC) containing 50 ng/ml human thrombopoietin (R and D Systems, Minneapolis, MN), 10 ng/ml murine interleukin‐3 (IL‐3), and 20 ng/ml murine IL‐6 (PeproTech, Rocky Hill, NJ). After 6–8 days of incubation, the slides were stained for acetylcholinesterase activity according to the manufacturer's directions and scored for the presence of red‐brown to brown‐black granules in cells. BFU‐e, CFU‐e, CFU‐GM, and CFU‐mix were assayed by plating 2 × 104 fetal liver cells in 35 mm dishes in methylcellulose medium (M3334, Stem Cell Technologies), which contains 3 units/ml of EPO and supplemented with 20 ng/ml SCF, 5 ng/ml murine GM‐CSF, and 5 ng/ml murine G‐CSF (PeproTech). Dishes were incubated in a humidified 5% CO2 atmosphere at 37°C and colonies were counted 3–8 days after plating.

Fetal liver transplantation

Eight to twelve week old female C57BL/6‐Ly5.2. mice were irradiated with 1100 rads from a 137Cs source at a rate of 167 rad per minute. After 4 h, 2 × 106 E13.5 fetal liver cells produced from B6.129 Meis1+/− intercrosses were transplanted to irradiated hosts by tail vein injection. In some experiments, 4–5 × 105 bone marrow cells from 8‐ to 12‐week‐old female C57BL/6Ncr‐Ly5.2 mice were injected along with the fetal liver cells. Single‐cell suspensions were made from hematopoietic organs of transplant recipients and red cells lysed by incubation in ice‐cold ACK lysis buffer (Quality Biological, Gaithersburg, MD) for 10 min. Cells were washed twice in DPBS, 1% fetal calf serum (FCS, Atlanta Biologicals, Norcross, GA) and incubated with 2 μg/106 cells rat anti‐mouse CD16/CD32 antibody (2.4G2, BD Pharmingen, San Diego, CA) for 30 min on ice, washed twice in DPBS, 1% FCS and incubated for 30 min on ice with fluorescin isothiocyanate (FITC) or phycoerythrin (PE)‐conjugated antibodies (BD Pharmingen): 0.5 μg/106 cells of rat anti‐mouse CD3 (17A2), CD4 (GK1.5), CD8a (53‐6.7), CD19 (1D3), CD23 (B3B4), CD25 (PC61), CD34 (RAM34), CD45/B220 (RA3‐6B2), CD62P (RB40.34), CD117 (ACK45), Gr‐1 (RB6‐8C5), Ly‐76(Ter119), and isotype controls (R35‐95 and MWReg30), 0.125 μg/106 cells of rat anti‐mouse CD11b (M1/70), and 0.25 μg/106 cells mouse anti‐mouse CD45.2 (Ly5.1, 104) and CD45.1 (Ly5.2, A20). Cells were then washed twice in DPBS, 1% FCS, fixed in 1% paraformaldehyde/PBS, and analyzed on a FACS Caliber flow cytometer (Becton Dickinson, San Jose, CA). Percentage engraftment was determined in comparison to isotype control antibodies.

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplemental Fig. 1. Western blot analysis of wild type, Meis1+/− and Meis1−/− embryos at E13.5 using an MEIS1‐specific antibody. The marker lane contains proteins of 30, 40, 50, and 60 kDa, respectively. In the other lanes, 100 μg of protein were loaded per well. The MEIS1 antibody recognizes a protein of approximately 46 kDa in wild type (WT) mice. Heterozygous mutant embryos (ko/+) express the 46 kDa protein along with a fainter 33 kDa protein. Homozygous mutant embryos (ko/ko) express almost exclusively the 33 kDa protein. [emboj7600038-sup-0001.jpg]


We thank Deborah Swing, Holly Morris, and Sandra Grimm for animal handling, Debra Gilbert, Norene O'Sullivan, Deborah Householder, Anna Trivett, and Tukari Yamazaki for technical help, Linda Cleveland for sequencing, Eileen Southon for ES manipulation, Mary Ellen Palko for providing primers, Ed Cho for confocal microscopy assistance, Keith Rogers for pathological advice, Linda Brubaker and Madeline Knoebel for secretarial assistance and Michael Cleary for a critical reading of this manuscript. This research was supported in part by a Grant in Aid for Scientific Research on Priority Area C from the Ministry of Education, Science, Sport and Culture, Japan and the National Cancer Institute, DHHS, USA.


View Abstract