Proximodistal identity during vertebrate limb regeneration is regulated by Meis homeodomain proteins

How cells interpret their position along the three axes of a morphogenetic field is fundamental to the correct patterning of developing and regenerating organs. This question has been studied extensively during vertebrate limb development (Tickle, 2003) and regeneration (Nye et al.,2003). Despite intensive effort in this field, how cells acquire positional information along the limb proximodistal (PD) axis remains poorly understood during both limb development(Duboule, 2002) and regeneration (Brockes,1997).

Urodeles and anuran tadpoles are the only vertebrates able to regenerate amputated limbs completely. Following Spallanzani's description of this phenomenon in the 18th century(Spallanzani, 1768),scientists have tried to identify the mechanisms underlying limb regeneration and patterning of the growing blastema. As is the case during limb development, undifferentiated cells in the regenerating blastema acquire positional information de novo to reproduce the original pattern. In contrast to limb development, however, not every positional value needs to be recreated during regeneration, as the limb amputation can occur at any PD level and the blastema will reproducibly regenerate only the missing portion. A system of positional memory recorded in the mature limb is thought to instruct blastema cells of the exact PD level from which regeneration should take place(Stocum, 1984).

Classical experiments showed the role of RA as a major instructor of positional identity along the regenerating limb PD axis. When anuran tadpole or urodele limbs are amputated at the wrist and allowed to regenerate in an excess of RA, a whole PD axis develops rather than just the missing part,generating a tandem duplication (Maden,1982; Niazi and Saxena,1978). Duplications induced by decreasing amounts of RA result in increasing distalization of the proximal-most identity of the duplicated regenerate, showing the ability of RA to change positional values in a dose-dependent manner (Thoms and Stocum,1984).

Cell affinity properties constitute a hallmark of PD positional identity of blastema cells (Nardi and Stocum,1983). Consistent with its ability to proximalize regenerating limbs, RA reprograms blastema cell affinity properties towards a proximal identity (Crawford and Stocum,1988) through the RARγ receptor isoform(Pecorino et al., 1996),downregulates the distal marker Hoxa13(Gardiner et al., 1995) and upregulates the proximal marker Hoxd10(Simon and Tabin, 1993). Recent results shed the first light on the molecular pathways controlling cell affinity that are activated by RA during limb regeneration. The GPI-anchored cell surface molecule Prod1 is more abundant in proximal than in distal blastemas, and is upregulated by RA (da Silva et al., 2002). Blocking Prod1 results in loss of the specific affinity properties of proximal blastemas, whereas overexpression of Prod1 in limb blastema cells causes them to translocate to more proximal positions, showing it contributes to the PD affinity code(Echeverri and Tanaka, 2005)Prod1 is also differentially expressed in the mature limb PD axis, suggesting that it forms part of the positional memory system(da Silva et al., 2002).

Some advances in the field of PD specification during limb development in amniotes came again from analyzing the RA pathway. During embryogenesis, RA is synthesized throughout the lateral plate mesoderm, but is excluded from the limb field as limb buds emerge from the trunk(Berggren et al., 1999; Niederreither et al., 1997). As determined by RA reporter transgene expression(Reynolds et al., 1991; Rossant et al., 1991), RA appears to be distributed in a proximal high-distal low gradient in the limb bud. RA is restricted proximally by limited diffusion from the trunk RA-producing region and by specific activation of the RA-degrading enzymes CYP26A1 and CYP26B1 in the distal limb(MacLean et al., 2001; Swindell et al., 1999). Study of Raldh2- and CYP26B-deficient mice revealed a role for RA in promoting limb outgrowth, as well as in limb PD patterning(Mic et al., 2004; Niederreither et al., 2002; Yashiro et al., 2004). Concurring with the view that RA instructs PD limb cell identities, distal limb bud region transplants generate proximal structures following RA treatment, and RA-exposed distal cells contribute to the formation of proximal skeletal elements (Ide et al.,1994; Mercader et al.,2000; Tamura et al.,1997). As during regeneration, cells of the developing limb bud display specific affinity according to their position in the PD axis(Ide et al., 1994), and RA can proximalize limb cell affinity in several experimental contexts(Ide et al., 1994; Mercader et al., 2000; Tamura et al., 1997).

Two closely related homeobox genes, Meis1 and Meis2, are crucial targets activated by RA during PD limb axis patterning(Mercader et al., 2000; Yashiro et al., 2004). Early subdivision of the limb bud into proximal Meis-positive and distal Meis-negative domains is necessary for correct PD limb development; ectopic Meis1 or Meis2 overexpression abolishes distal limb structures (Capdevila et al.,1999), leads to a proximal shift of limb identities along the PD axis (Mercader et al., 1999),and proximalizes distal limb cell affinity properties(Mercader et al., 2000).

Meis genes encode transcription factors of the TALE class of homeodomain proteins. Members of this family share a three-amino-acid loop extension between homeodomain helices 1 and 2(Bürglin, 1997). Meis proteins have an additional conserved N-terminal domain, through which they interact with members of a second TALE family, the PBC class, encompassing vertebrate proteins Pbx1 to Pbx4 (Chang et al., 1997; Knoepfler et al.,1997; Wagner et al.,2001). Meis-Pbx dimerization, as well as that of their insect counterparts, is required for nuclear localization of the proteins(Abu-Shaar et al., 1999; Berthelsen et al., 1999; Capdevila et al., 1999; Mercader et al., 2000; Rieckhof et al., 1997), and Meis-Pbx interaction is thus considered to be essential for their function.

During vertebrate limb regeneration, however, with the exception of Prod1,the elements in the molecular pathway activated by RA remain unknown. Here, we use recently developed electroporation methods(Echeverri and Tanaka, 2003; Echeverri and Tanaka, 2005; Schnapp and Tanaka, 2005) in the limb blastema to analyze the role of Meis genes in PD patterning during regeneration. We provide evidence of a role for Meis genes as crucial targets of RA proximalizing activity during limb regeneration. Our results show that the RA-Meis pathway, active during limb development, is essential to specify RA-induced proximal fates during limb regeneration.

We established an axolotl colony (Ambystoma mexicanum) from 40 founder animals imported from Indiana University (USA). For electroporation,blastemas were generated on animals with a snout-to-tail length of 3-5 cm. For RA treatment and RNA or protein extraction, 4-7 cm-long animals were used. Experimental animals were anesthetized in 0.01% ethyl-p-amino benzoate (Sigma)prior to amputation and limbs were allowed to regenerate at room temperature. For RA treatment, all-trans RA (Sigma) dissolved in DMSO was injected intraperitoneally (100 μg/g body weight) on day 4 post-amputation(Thoms and Stocum, 1984). To study the effect of RA during limb development, tadpoles were immersed (10-24 hours) in 10 mM all-trans RA in swimming water.

Primers used for axolotl Meis1 and Meis2 amplification from proximal medium bud blastema cDNA (Meis1) and whole tadpole cDNA(Meis2) were: Meis1, 5′-ACTAGTGATTCTGCACTC, and 5′-CATGTAGTGCCATTGCCCCTC; Meis2, 5′-CTTCAACGAGGACATCGCGGTC and 5′-TTGGGGAATATGCCTCTTTTCTTC. Full-length Meis sequences were completed by 5′ and 3′RACE using the following primers.

For 5′RACE: Meis1, 5′-TGCGATTGGTTAACTAGATGGT and 5′-GATTGGTTAACTAGATGGTGTTAC; Meis2, 5′-AGCGTAGGAACAGGGCGAAGTC and 5′-GCATCACGACAGGCCGCACATC.

For 3′RACE: Meis1, 5′-GTAAGTCAAGGTACACCA and 5′-CCATGGGAGGATTTGTGA; Meis2, 5′-GTGAACCAAGGGGACGGGC and 5′-ACAGCGTGGCCTCTCCTG.

The axolotl Pbx1 full-length sequence was identified in the axolotl EST sequencing project database (Habermann et al., 2004). Axolotl Meis1, Meis2 and Pbx1nucleotide sequences have been submitted to GenBank with accession numbers DQ100364, DQ100365 and DQ100366.

The following probes were used: axoMeis2 (+337 to +828) and axoHoxa13 (+552 to +834). Whole-mount in situ hybridization was performed as described (Wilkinson and Nieto, 1993), with proteinase K treatment (10 μg/ml) for 10 minutes for the youngest embryos and 15 minutes for stage (St) 36 to St43 tadpoles.

For in situ hybridization on sections, 12 μm cryosections were collected on Superfrost slides, air-dried (2 hours), and hybridized as described(Myat et al., 1996).

Tissue samples were collected and fixed immediately in 4% paraformaldehyde(PFA; overnight, 4°C), rinsed in PBS and dehydrated in methanol. Specimens were rehydrated to PBS, immersed in 30% sucrose in PBS (overnight, 4°C),OCT-embedded, sectioned (12 μm) on a cryostat and collected on Superfrost slides. Prior to immunohistochemistry, slides were air-dried (30-60 minutes,RT), refixed in ice-cold acetone (2 minutes, 4°C), rinsed in PBS, treated with 100 mM glycine in PBS (30 minutes) and washed in PBT (PBS 0.1% Tween-20). Sections were permeabilized with 0.05% Triton X-100 in PBS (10 minutes),followed by a PBT wash (5 minutes). A second permeabilization step was performed with PBS:DMSO (1:1; 8 min), and a PBT wash. Slides were blocked in 10% goat serum (overnight at 4°C), incubated sequentially with the primary antibody, with biotin-conjugated secondary antibody and with streptavidin-Cy3(Jackson) (1 hour each, room temperature).

Anti-Meis-a antibody was generated in rabbits with a synthetic peptide corresponding to the conserved C-terminal domain of Meis1a and Meis2a/Meis2b isoforms (GMNMGMDGQWHYM). Anti-Meis1 and -Meis2 antibodies were raised in rabbits using the peptide sequences HAARSMQPVHHLNHGPP and HAPRPIPPVHHLNHGPP,respectively. Anti-Pbx1/Pbx2/Pbx3/Pbx4 was purchased from Santa Cruz Biotechnology and used at 20 μg/ml. Overexpressed tagged Meis1, Meis2 and Pbx1 proteins were detected using anti-FLAG (Sigma) and anti-Myc monoclonal antibodies.

For western blot, blastemas were lysed in RIPA and 15 μg protein lysate was transferred to nylon membranes and counterstained with Ponceau Red to visualize total protein.

Quantitative PCR was performed on an ABIPrism 7700 thermocycler (Applied Biosystems) applying 1 cycle at 50°C for 2 minutes, 1 cycle at 95°C,10 minutes, 40 cycles at 95°C, 15 seconds, and at 67°C, 90 seconds,followed by 1 cycle at 95°C, 15 seconds, and at 60°C, 15 seconds, and a final step at 95°C, 15 seconds. For each cDNA sample, two dilutions(1/50 and 1/500) were amplified in duplicate. The following primers were used:

• Meis1, 5′-CCATCTACGGACACCCCCT and 5′-GGAAGAACACACGTCCCCG

• Meis2: 5′-AGTGGAGGGCACGCTTCT and 5′-GCTTCTTGTCTTTGTCCGGGT

• Hoxa13: 5′-TCTGGAAGTCCTCTCTGCCG and 5′-TCAGCTGGACCTTGGTGTACG

• Ef1α: 5′-CGGGCACAGGGATTTCATC and 5′-TGCCGGCTTCAAACTCTCC. Amplification reactions were carried out simultaneously using SYBR Green PCR Master Mix (Applied Biosystems). c_t values obtained for experimental primer sets were normalized with c_t values for EF1α control primers (Livak and Schmittgen, 2001)

Axolotl Meis1 and Meis2 cDNA cloned into pCMVTag2b and axolotl Pbx1 cloned into pCMVTag3b vectors (Stratagene) were diluted in PBS (0.25 μg/μl), together with a CMV-nuclear EGFP reporter plasmid(0.05 μg/μl). Using a microinjector (Inject+Matic, Geneva), DNA solution(∼7 nl), with Fast Green added to improve monitoring, was injected at three different blastema sites to ensure homogeneous distribution. Immediately after injection, five pulses were given (5 mseconds, 150 V/cm each) using flat circular electrodes (3 mm diameter) with an electroporator (BTX ECM-830, San Diego). To count GFP-positive cells, limbs were fixed in 4% PFA (overnight,4°C), soaked in 30% sucrose in PBS (2 hours) and OCT-embedded. Longitudinal cryosections (30 μm) were collected on slides, air-dried (2-6 hours), fixed in ice-cold acetone (2 minutes), rinsed in PBS and mounted in anti-fading mounting medium.

Fluorescein-coupled antisense morpholino oligonucleotides (MOs) were designed against 5′-coding regions of axoMeis1(5′-AAGTCGTCGTACCTTTGCGCCATCG) and axoMeis2(5′-GGAGTTCATCGTACAGAAACATGAT) (Gene Tools, Corvallis, OR). A fluorescein-coupled standard MO was used as control. MO were diluted in water and stored at –20°C as a 5 mM stock solution. For injection, MO were diluted to 1.5 mM in PBS and heated (68°C) for 10 minutes prior to use. Injection and electroporation were as indicated above for DNA constructs.

Using primers against conserved sequences of mouse Meis cDNA at low stringency, we amplified axolotl cDNA fragments corresponding to the three vertebrate Meis genes. We subsequently obtained full cDNA sequences of axolotl Meis1 and Meis2, and a partial Meis3 sequence, by 5′ and 3′ RACE. Axolotl Meis2, Meis1 and Meis3 proteins show 99%,97% and 90% identity to their respective mouse counterparts. Whole-mount mRNA in situ hybridization showed a Meis2 expression pattern in axolotl embryonic central nervous system (CNS) that is similar to that described in mouse embryos (Fig. 1A-D). Two major Meis isoforms differing in the C-terminal region are generated by alternative splicing in vertebrates, a short C terminus form (corresponding to mouse Meis1a, Meis2a and Meis2b) and a long C terminus form (corresponding to mouse Meis1b, Meis2c and Meis2d). Although our strategy was designed to clone either isoform, from several cDNA sources, we amplified mainly the short isoform-encoding cDNA of Meis1 and Meis2, suggesting that this is the predominant form in axolotls.

An antibody that recognizes an epitope conserved in all short isoforms of mouse and axolotl Meis proteins (anti-Meis-a) specifically detects a nuclear signal in the embryonic CNS, in a pattern that correlates with Meis2mRNA distribution (Fig. 1E-H). Using this antibody, we examined Meis protein distribution during limb outgrowth. At stage H36, when the limb bud has just formed, the Meis protein signal is strong, nuclear and uniformly distributed throughout the limb bud(Fig. 1I). From stage 40, we found that the limb bud was subdivided into two regions defined by Meis protein distribution, a proximal region retaining a strong nuclear Meis-a signal and a distal region showing a low, diffuse Meis-a signal with cytoplasmic enhancement (Fig. 1J). During subsequent limb bud stages, we found progressive growth of the distal, non-nuclear Meis region and maintenance of the proximal confinement of nuclear Meis expression(Fig. 1K). In early limb buds,Meis-a signal was detected in a nuclear pattern in the ectoderm(Fig. 1I). Ectodermal expression has been described in the mouse for Meis2(Oulad-Abdelghani et al.,1997) and Meis1 (N.M. and M.T., unpublished). Meis2 mRNA distribution during limb development was uniform with no proximal restriction at any stage in the mesenchyme, and was weak or absent in the ectoderm (Fig. 1N-P). These observations suggest that Meis protein expression could be subject to post-transcriptional, and/or subcellular distribution, control during limb development. We observed no expression by in situ hybridization using different Meis1 probes in the embryo, whereas Meis3 was detected in the CNS, but was not detected in developing limbs (not shown). We cannot disregard, however, the fact that technical problems might underlie our inability to detect Meis1 transcripts in situ, as RT-PCR analysis indicated the presence of Meis1 mRNA in embryos (data not shown). Meis1 expression may therefore also contribute to the proximal enrichment in Meis-a signal, as well as to the ectodermal Meis expression. We also analysed the expression of Pbx proteins during limb development using an antibody that recognizes a conserved epitope present in all Pbx proteins(Pbx1-4). Nuclear Pbx expression was detected ubiquitously along the limb bud PD axis, with higher expression in the proximal versus distal regions of the bud (Fig. 1M).

We next tested whether Meis expression was RA-regulated during axolotl limb development. Exposure of stage 43 axolotl larvae to 0.12 mM RA in the swimming water for 22 hours produced an extension of the Meis protein expression domain and an enhancement of its nuclear signal(Fig. 1L), as well as an overall increase in Meis2 mRNA expression(Fig. 1Q).

These results indicate that the proximal restriction of Meis activity during limb development and its positive regulation by the RA pathway are conserved between the axolotl and amniotes.

To analyze Meis pathway involvement in PD patterning during regeneration,we studied the Meis gene expression pattern in limb blastemas amputated at different PD positions. In situ hybridization showed no differences in Meis1 and Meis2 mRNA expression levels between proximal and distal blastemas (data not shown). Nonetheless, real-time PCR (RT-PCR) showed that Meis1 and Meis2 transcripts are more abundant in proximal than in distal medium bud blastemas, by a factor of 3.4 and 2.5,respectively (Fig. 2A). In these experiments, Hoxa13 was more abundant in distal than in proximal blastemas (Fig. 2A),as described previously (Gardiner et al.,1995). Meis3 was not detected by RT-PCR in blastemas from any axial amputation level or stage (not shown). RA treatment of distal blastemas increased Meis1 and Meis2 transcripts by 2.3- and 2.2-fold, respectively (Fig. 2A). As reported elsewhere(Gardiner et al., 1995), Hoxa13 expression was repressed after RA treatment of distal blastemas (Fig. 2A).

Similar samples were analyzed by western blot to determine protein variations. Meis proteins detected using anti-Meis-a were 4-fold more abundant, and Meis2 was 3-fold more abundant, in proximal versus distal blastemas (Fig. 2B). RA treatment of distal blastemas elevated Meis protein levels over those in proximal blastemas; Meis proteins detected with anti-Meis-a increased 6-fold,whereas Meis2 increased 4.5-fold compared with proximal blastemas(Fig. 2B). In immunohistochemical analysis, an endogenous Meis-a signal was not detected in distal blastemas from untreated animals at any regeneration stage(Fig. 2C,F and not shown);however, a nuclear Meis-a signal was detected in the majority of distal blastema cells three days after RA treatment(Fig. 2D,G). The Meis-a nuclear signal was stronger and more widespread in the blastema 6 days after RA treatment. At this stage, however, a narrow distal Meis-a low-expression area was observed (Fig. 2E,H). The strong Meis-a signal persisted 9 days post-injection in proximal blastema regions (not shown). The Hoxa13 expression pattern was complementary,with broad expression in blastemas from untreated animals, disappearing 3 days after RA treatment and re-appearing at 6 days post-treatment in the Meis-a low-expression area (Fig. 2I-K).

These results showed that Meis expression is regulated in limb blastemas depending on the PD amputation position, and that RA activates Meis mRNA and protein expression levels during regeneration. By contrast, in early to medium bud stage blastemas, a homogeneous nuclear Pbx expression pattern was observed, which did not increase significantly after proximalization of distal blastemas with RA (Fig. 2L-Nand data not shown).

Analysis of Meis mRNA or protein levels in tissues from several PD positions of the mature limb showed no differences for Meis1 or Meis2 (not shown). We thus found no direct evidence that Meis activity takes part in positional memory, although we cannot exclude that regulation of Meis activity in the mature limb takes place at the subcellular localization level.

We used overexpression experiments for functional testing of Meis protein involvement in PD patterning during limb regeneration. Because Meis function may require the presence of a Pbx partner, we included co-expression of Meis together with Pbx1 in our analysis. For these experiments, we used a full-length Pbx1 cDNA identified in the axolotl cDNA sequencing project(Habermann et al., 2004). Expression plasmids driving different Meis or Pbx1 proteins fused to the FLAG epitope or driving EGFP (Fig. 3A) were electroporated into the limb blastema 6 days after amputation. When overexpressed without Pbx1, Meis proteins were detected only until 48 hours after electroporation. By contrast, when introduced together,both Pbx1 and Meis proteins were detected until 5 days after electroporation(Fig. 3B). Meis or Pbx1 proteins were no longer detected from day 6, but EGFP persisted up to 20 days post-electroporation (Fig. 3Cand not shown). The proportion of electroporated cells ranged from 5 to 10% of blastema cells, precluding the use of this approach to study the consequences of Meis/Pbx1 overexpression in pattern formation(Fig. 3B,C). By contrast, the persistence of EGFP detection allowed us to determine the contribution of electroporated cell descendants to the regenerated limbs. Proximal blastemas were electroporated with different combinations of Meis/Pbx1 and EGFP expression vectors, and their contribution was compared with that of GFP-only-electroporated control cells. Control cells were widely dispersed throughout the PD extension of the regenerate(Fig. 3D-F). By contrast, any combination containing Meis1a or Meis2a promoted aggregation of electroporated cells and their preferential location in proximal regenerate regions, both evident from 3 days after electroporation(Fig. 3G-I and data not shown).

To quantify the results, regenerated limbs were sectioned horizontally and positive cells scored in the different PD regions. Any combination containing Meis1a or Meis2a produced a higher frequency of GFP-positive cells in proximal regions compared with controls (Fig. 3J). Whereas positive cell frequency in distal stylopod plus proximal zeugopod was 25% in controls, it was between 53 and 70% in Meis1a- or Meis2a-electroporated blastemas (Fig. 3J). Meis1b isoform activity in this assay was tested by electroporating the murine protein. In contrast to the results for Meis-a, we observed no effect of Meis1b on PD cell distribution(Fig. 3J). Pbx1 alone neither altered the relative PD distribution of electroporated cells nor contributed appreciably to the Meis effect on PD cell distribution(Fig. 3J). This result is consistent with its expression during limb regeneration, where no differential Pbx expression could be detected between proximal and distal blastemas. On the contrary, a homogeneous nuclear Pbx expression pattern in early to medium bud stage blastemas was observed, which did not increase significantly after proximalization of distal blastemas with RA(Fig. 2L-N and data not shown).

In some experimental conditions studied, positive cell counts in total limb were markedly reduced when compared with controls(Fig. 3K). This phenomenon was most evident in Meis-only electroporated limbs, such that absolute cell numbers in proximal regions increased only marginally compared with controls(Fig. 3K). Co-electroporation with Pbx1 rescued total numbers of GFP-positive cells, resulting in a net increase in absolute cell number in proximal regions(Fig. 3K).

The proximal-preferential distribution of Meis-expressing cells may result from a number of possible cellular effects induced by Meis overexpression,including distal-specific underproliferation, increased death of Meis-electroporated cells, relocation of Meis-expressing cells to proximal limb regions, or combinations of these. Proliferation and cell death rates of electroporated limbs were analyzed at different regeneration stages, with no notable differences between control and experimental conditions, or between proximal and distal blastema regions (data not shown). Undetectable differences in proliferation or cell death rates could nonetheless cause a significant difference in cell number if they accumulated during the 9-day experimental period.

To determine directly whether proximal relocation was a principal cause of the cell redistribution, we set up focal electroporation to target specific regenerate subregions (Fig. 4A,B). Focal electroporation of control plasmid in the distal third of the medium bud limb blastema (closer than 350 μm to the apical ectodermal cap) resulted in GFP-positive cells that contributed almost exclusively to distal regions, with very few regenerates showing GFP-positive cells in zeugopod (2/12) or stylopod (1/12)(Fig. 4I). By contrast, similar electroporation with Meis2a, alone or with Pbx1, resulted in a large proportion of limbs containing GFP-positive cells in zeugopod (20/23) or stylopod (14/23) (Fig. 4E,F,I). Total cell counts in each PD segment in horizontal sections of regenerates showed that Meis2a, alone or with Pbx1, produced proximal relocation of approximately 40% of the electroporated cells(Fig. 4J).

Similar experiments using mass electroporation in blastemas from wrist-level amputations showed that Meis1a/Meis2a, alone or with Pbx1, induced cell aggregation and confinement to the proximal-most region of the regenerate(Fig. 4C,D,G,H and data not shown). Control focal electroporation in distal regions of late bud proximal amputation blastemas produced cells contributing mainly to digits, less to the carpal-metacarpal regions, and not at all to more proximal regions(Fig. 4K). By contrast, under the same conditions, Meis2a±Pbx1-electroporated cells contributed mainly to the carpal-metacarpal and less to the digit region(Fig. 4K). The ability of Meis overexpression to modify the contribution of regenerating limb cells is thus not confined to sorting between stylopod/zeugopod versus more distal regions,but is also detected within distal limb regions.

The results show that PD cell sorting or migration is the major cause of the proximal-preferential distribution of Meis-overexpressing cells. The ability of Meis to modify limb cell PD affinity during regeneration is active throughout the PD axis of the regenerating limb. Our results suggest a role for Pbx1 in stabilizing Meis proteins and show that Pbx1 expression is not sufficient to promote proximal relocation.

To study the consequences of Meis knockdown during RA-induced proximalization of the limb blastema, morpholino oligonucleotides (MO) that inhibit Meis1 and Meis2 protein expression in vivo(Fig. 5) were used in regeneration experiments. Fluorescently labeled MOs electroporated much more efficiently into blastema cells than did plasmid DNA, with up to 80% of cells showing a strong signal after electroporation(Fig. 6A,B). Using this approach, we determined the consequences of Meis1 and Meis2 knockdown during limb regeneration. When Meis1 MO (n=20), Meis2 MO (n=20), or both MOs (n=38), were electroporated into early-bud proximal blastemas, no alterations in the regeneration pattern were observed (not shown).

We next studied the consequences of Meis knockdown in the context of RA-induced PD tandem duplications. To generate PD duplications, axolotl larvae forelimbs were amputated at mid-carpal level and RA (100 μg/g body weight)was administered intraperitoneally 4 days after amputation. These conditions gave rise to tandem PD duplications, ranging from partial zeugopod to the complete PD axis, including shoulder elements. To score duplication intensity,we used a duplication index (DI) modified from that of Maden(Maden, 1983): partial zeugopod, 1; total zeugopod, 2; partial stylopod, 3; total stylopod, 4;shoulder element present, 5.

Bilateral amputations were performed, and left and right DI scored after regeneration following RA injection. Differences in degree between left and right duplication were determined by calculating the asymmetry index (AI; left DI minus right DI). Positive AI values indicate stronger duplications in the left limb; negative AI values indicate the opposite. When both left and right limbs were electroporated with control MO, deviations from symmetry were minor(never more than one AI unit in a single animal), and were balanced between left and right sides, so that the average AI was 0(Fig. 6D, Fig. 7A,B). By contrast, when blastemas on one side were electroporated with control MO and those on the opposite with different combinations of Meis MO, less intense duplications were observed in the Meis MO-electroporated limbs in over half of the cases(Fig. 6C,E). This reduction in the DI affected animals showing any degree of duplication on the control side(Fig. 6E and not shown). The average AI was –0.8 for Meis1 and –1.0 for Meis2, when Meis MOs were electroporated on the left side (Fig. 7C,D). When Meis1 and Meis2 MOs were coelectroporated, the average AI variation was –0.8 when introduced on the left and +1 when introduced on the right. In all cases, DI reductions induced by Meis1 and/or Meis2 knockdown in single animals ranged from 1 to 3 duplication units, and affected 26 out of 46 treated animals(Fig. 7).

These results show that Meis knockdown does not result in an altered pattern of normal regeneration but indicate that Meis genes act downstream of RA to induce the proximalization of the regenerating limb.

During limb regeneration, positional information must be re-specified to achieve a correctly patterned regenerated structure. A plausible scenario is that adult tissue re-uses the molecular machinery employed during development to pattern the re-growing body part (Bryant et al., 2002; Gardiner et al.,2002). Indeed, grafting experiments suggest that early limb bud and blastema cells share similar differentiation programs(Muneoka and Bryant, 1982),and developmentally active genes, such as the signaling molecules Fgf8 and Shh, as well as the Hox, Tbx and Dlx transcription factors, are re-expressed during regeneration (Beauchemin and Savard,1992; Brown and Brockes,1991; Gardiner et al.,1995; Han et al.,2001; Simon et al.,1997; Simon and Tabin,1993; Torok et al.,1999). Here, we show that two closely related homeodomain proteins, Meis1 and Meis2, act as proximalizing factors during regeneration. When overexpressed, Meis proteins induce relocation to proximal regions of otherwise distal cells, a phenomenon also promoted by RA pathway activation(Crawford and Stocum, 1988; Pecorino et al., 1996). Meis knockdown inhibits RA proximalizing activity, showing that these genes represent key targets in the RA proximalizing pathway. By contrast, Meis knockdown induced no alterations in normal regeneration after a proximal amputation. One possible explanation for this result is that the RA-Meis pathway does not play any role during normal regeneration and thus its ability to reprogram distal blastemas would only correspond to the re-activation of an embryonic proximalizing pathway. However, our data do not exclude a role for Meis genes in establishing the proximal identity within the normal blastema. Indeed, when proximal stumps are confronted with distal blastemas in recombination experiments, instead of an interstitial deletion, normal regeneration takes place by intercalation of the missing identities from the stump (Iten and Bryant, 1975; Pescitelli and Stocum, 1981; Stocum, 1975). By contrast,when distal stumps are confronted with proximal blastemas, PD tandem duplications are obtained. In the light of these recombination experiments, in the case of an experimental distalization of the blastema, normal limb regeneration is still expected to occur, whereas in the case of an attenuation of an RA-induced proximalization, a change in the regeneration pattern is predicted. Further research is therefore necessary to establish the involvement of the RA-Meis pathway in normal regeneration.

During vertebrate limb development, Meis1 and Meis2 act downstream of RA in a proximalizing pathway(Mercader et al., 2000). Our results here show that the same pathway used during limb development is re-used in regeneration to specify proximal identity. In contrast to its effect on urodele limb regeneration, RA does not induce obvious PD duplications in the developing chick limb. A possible explanation is that during normal limb development, reprogrammed cells have the chance to migrate and integrate into proximal compartments, whereas during regeneration, they are `trapped' in a distal region and are thus forced to generate an ectopic proximal compartment.

During amniote limb development, the limit of Meis gene expression in the patterned chicken limb corresponds to the boundary between the stylopod and the zeugopod. Localized inhibition of RA signaling in the chick disrupts skeletal elements proximal to the elbow/knee, but does not affect regions distal to this point. Conversely, localized RA excess disrupts elements distal to the elbow/knee but spares elements proximal to this point(Mercader et al., 2000). These results suggested a major subdivision of the limb bud into an RA-Meis domain that generates regions proximal to the stylopod/zeugopod boundary, and an RA-Meis-negative area that generates regions beyond this boundary. We found here that Meis gene expression and regulation through RA is conserved between amniote and amphibian limb development. Nonetheless, in the chick, RA exposure or Meis overexpression does not impose a stylopod character, but proximalizes cells at any distal position in a graded manner(Mercader et al., 1999). Similar results are obtained after RA exposure during limb regeneration;duplications induced by decreasing amounts of RA result in increasing distalization of the proximal-most identity of the regenerate(Thoms and Stocum, 1984). The alterations in PD identity promoted by changes in Meis activity seem to follow similar rules; Meis overexpression proximalizes cell affinity at any position of the regenerating PD axis, and Meis knockdown distalizes the proximal-most identity of the regenerate, irrespective of the initial specification status of the blastema. These observations suggest that positional information is encoded uniformly, and is recognized continuously along the PD axis during both limb development and regeneration. In addition, the components of the positional code appear to remain sensitive to the RA-Meis pathway at any PD position.

Meis regulation is mainly transcriptional in amniotes, whereas in the axolotl it appears to be at least partially by translational and/or by subcellular protein localization. Mouse Meis proteins and their Drosophila ortholog Hth are translated and imported into the nucleus constitutively, whereas mouse Pbx1 and its Drosophila ortholog Exd require Meis/Hth expression to be imported into the nucleus and become functional (Abu-Shaar et al.,1999; Berthelsen et al.,1999; Capdevila et al.,1999; Mercader et al.,1999; Rieckhof et al.,1997). Data from Xenopus, zebrafish and spiders provide examples of the opposite situation; Pbx/Exd proteins are expressed constitutively in the nucleus whereas Meis/Hth proteins remain in the cytoplasm by default, and become nuclear only after co-expression with Pbx/Exd(Maeda et al., 2002; Prpic et al., 2003; Vlachakis et al., 2001). These observations suggest that the mechanisms for the regulation of Meis and Pbx activity have remained labile during evolution. This flexibility would only be possible if, as suggested by most Meis/Pbx functional experiments, Meis and Pbx proteins always work together, with no functional independence for each partner. Any change in the ability of a partner to enter the nucleus on its own would have no functional consequence, and would thus be unrestricted during evolution. In our experiments, however, Pbx neither induced alterations in cell behavior nor substantially changed the phenotype induced by Meis protein overexpression. These results might indicate that Meis alone can reprogram limb cell PD identity, although it is most likely to interact with Pbx proteins constitutively expressed in limb blastemas.

Another interesting aspect of limb PD specification mechanisms is the regulation of cell affinity. PD axial level-specific affinity is a hallmark of PD identity, and is modulated by RA during development and regeneration. The fact that Meis rapidly induces cell relocation in the early blastema suggests that PD-specific affinity is already established at this stage. Cell lineage tracing and transplantation experiments have also suggested an early PD sub-division of the limb blastema(Echeverri and Tanaka, 2005). Alterations of Meis activity at later stages similarly induced cell relocation, but within a restricted subregion of the regenerate, suggesting progressive PD compartmentalization of the blastema, or progressive limitation in cell migratory ability.

The GPI-anchored Prod1 molecule is downstream of RA in the proximalizing pathway, at least as part of the PD affinity code(da Silva et al., 2002; Echeverri and Tanaka, 2005). Prod1 appears to form part of the positional memory system that enables blastema cells to `know' which limb parts are missing. By contrast, we found no evidence of Meis genes as part of the positional memory system, suggesting that the Meis pathway belongs exclusively to the patterning network re-activated after amputation. In this case, after proximal amputation, the Meis pathway would be activated by the positional memory system, so that the initial Prod1 status may determine the level of Meis activation. The answers to these questions must await isolation of the axolotl Prod1 counterpart.

We thank R. Mueller for building the electrodes, K. Echeverri for advice on plasmid electroporation, E. Schnapp for advice on MO electroporation, E. Leonardo for technical help, the animal house personnel for support in setting up the axolotl colony, and C. Mark for editorial assistance. N.M. was supported by EU Marie Curie long-term and EMBO short-term fellowships. M.T. and N.M. are supported by grant SAF2003-04317 from the Spanish Ministerio de Educación y Ciencia. The Department of Immunology and Oncology was founded and is supported by the Spanish Council for Scientific Research (CSIC)and by Pfizer.