Diffusible Signals, Not Autonomous Mechanisms, Determine the Main Proximodistal Limb Subdivision

See allHide authors and affiliations

Science  27 May 2011:
Vol. 332, Issue 6033, pp. 1086-1088
DOI: 10.1126/science.1199489


Vertebrate limbs develop three main proximodistal (PD) segments (upper arm, forearm, and hand) in a proximal-to-distal sequence. Despite extensive research into limb development, whether PD specification occurs through nonautonomous or autonomous mechanisms is not resolved. Heterotopic transplantation of intact and recombinant chicken limb buds identifies signals in the embryo trunk that proximalize distal limb cells to generate a complete PD axis. In these transplants, retinoic acid induces proximalization, which is counteracted by fibroblast growth factors from the distal limb bud; these related actions suggest that the first limb-bud PD regionalization results from the balance between proximal and distal signals. The plasticity of limb progenitor cell identity in response to diffusible signals provides a unifying view of PD patterning during vertebrate limb development and regeneration.

The vertebrate limb bud arises from the lateral plate as a bulge of mesenchymal cells encased within an ectodermal hull. Late limb buds of all tetrapods contain three proximodistal (PD) segments, each expressing specific homeobox genes. The stylopod (upper limb) expresses Meis1/2, the zeugopod (lower limb) expresses Hoxa11, and the autopod (hand/foot) Hoxa13 (1, 2), although none of these markers is sufficient to specify limb-segment identity. The transition between stylopod (proximal) and nonstylopod (distal) structures represents the main PD subdivision of tetrapod limbs (3, 4). In the distal limb bud, the pool of undifferentiated cells responsible for limb generation is maintained by fibroblast growth factor (FGF) and Wnt signals produced by a distal epithelial structure called the apical ectodermal ridge (AER) (5). However, the importance of these and other signals in PD patterning remains controversial. Whereas the progress zone model proposes autonomous progressive distalization of undifferentiated cells under permissive AER influence (6, 7), classical transplantation experiments provide evidence for nonautonomous signals (8, 9). More recently, a two-signal model was proposed, with retinoic acid (RA) as proximalizer and FGFs as distalizers (1013); however, endogenous proximal signals have not been identified, and the role of endogenous RA has been questioned by genetic analyses in the mouse (14).

To investigate limb-proximalizing signals in the chicken embryo, we transplanted distal leg tips [200 μm thick, Hamilton-Hamburger (15) stage 19 to 20 (HH19-20)] to two potentially proximalizing regions: the somites and proximal wing bud of HH20 embryos (16) (fig. S1, A and B). These transplants were compared with transplants to tissues not expected to contain limb-proximalizing signals: HH24 distal wing bud (prospective zeugopod) and anterior HH20 hindbrain (fig. S1, A and B).

Graft development in these experiments was not influenced by the grafting site (Fig. 1, A and B and fig. S1D). These results thus support previous reports indicating autonomy of distal limb grafts (6, 7, 17). An alternative explanation, however, is that proximalizing signals were suppressed by distalizing FGFs from the graft’s AER (10). We tested this by treating the grafts with the FGFR1 inhibitor SU5402. Whereas untreated grafts transplanted to the somites did not express the proximal marker Meis1 and maintained Hoxa11 22 hours post grafting (hpg) (Fig. 1, C and D), SU5402-treated grafts expressed Meis1 along the entire PD axis and lost Hoxa11 expression (Fig. 1, E and F). Presumably, SU5402 action is enhanced by its effect on AER degeneration, which further diminishes FGF signaling. Notably, neither control nor SU5402-treated grafts to prospective HH24 zeugopod or anterior HH20 hindbrain activated Meis1 or down-regulated Hoxa11 expression 22 hpg (Fig. 1, I and J, and fig. S2, B to I), which indicated that the changes observed require a specific signal from the somites and not just release from FGF signaling. The somite region thus specifically contains signals that proximalize the limb bud expression profile, but these signals are counteracted by strong FGF activity from the AER.

Fig. 1

Endogenous RA reprograms distal limb bud cells when FGF signaling is diminished. (A and B) Skeletal preparations from HH19-20 200-μm leg grafts, 6 days after implantation at the somites [(A), n = 19] and prospective zeugopod [(B), n = 19]. Z, zeugopod; A, autopod. (C to J) HH19-20 200-μm limb grafts were treated and transplanted as indicated. Meis1 and Hoxa11 expression is shown on adjacent sections 22 hpg [(C and D), n = 6; (E and F), n = 6; (G and H), n = 7; (I and J), n = 5]. Asterisks, beads.

The somite region expresses the RA-synthesizing enzyme RALDH2 and the RA target RARβ (18) and contains biologically active RA levels (13), whereas anterior hindbrain and distal limb bud do not (fig. S1, A and B). To test whether endogenous RA was required for the proximalizing activity of the somites, we treated grafts with beads soaked in SU5402 plus RA antagonist (RAA). In this case, Meis1 expression was not up-regulated, and Hoxa11 expression was maintained (Fig. 1, G and H), which indicated that specifically blocking RA signaling disrupts the somite region’s proximalizing activity. Changes in skeletal patterning after these treatments could not be determined, because AER inhibition in SU5402-treated grafts disrupted further limb bud growth. We thus looked for an alternative approach.

Recombinant limbs (RLs) are made by dissociating limb mesenchymal cells and then reaggregating and packing them into an ectodermal hull, producing a ~500-μm-long structure (19). RLs can be made exclusively from undifferentiated distal HH19-20 cells (17). In the resulting RL, most of the originally distal cells are located beyond the range of high FGF signaling from the AER, which allows investigation of the response of distal cells to proximal signals without the influence of distal signals. We grafted RLs made from the distal 100-μm region of HH19-20 limb buds to the somites of HH20 hosts (sRLs) or to the prospective zeugopod of HH24 hosts (zRLs). As previously described (17), sRLs formed the three main limb segments (Fig. 2Ab); in contrast, zRLs formed only the two distal segments (Fig. 2Aa). Because the skeletal elements formed from RLs are often dysmorphic, we established a new approach to determine their PD identity. We found that the perichondrial-periosteal area of skeletal elements in late differentiating limbs retains segment-specific Meis1 and Hox expression, which allows molecular identification of skeletal elements’ PD identity (fig. S3A). This approach confirmed that the three segments formed by sRLs correspond to the normal PD sequence of limb segments (fig. S3B). Moreover, the proximal-most element formed in sRLs invariably expressed Meis1 but not Hoxa11 (Fig. 2, Bb and Cb), whereas the proximal-most element in zRLs expressed Hoxa11 but not Meis1 (Fig. 2, Ba and Ca). These results demonstrate that formation of a complete PD axis depends on signals from the graft environment.

Fig. 2

Distal mesenchymal cells in recombinant limbs alter their PD identity in response to endogenous RA. (Aa to Ad) Skeletal preparations from RLs 6 days after grafting, obtained as indicated. n = 6, 8, 5, 2, respectively. (Ba to Dd) Meis1 and Hoxa11 mRNA detection and cartilage staining (Alcian blue) on adjacent sections of RLs analyzed 3 days after grafting. Filled and empty arrowheads mark the presence and absence, respectively, of expression surrounding differentiated skeletal elements. n = 4, 3, 4, 2, respectively. Asterisks, beads. S, stylopod; Z, zeugopod; A, autopod.

We next explored the mechanism that allows production of a complete axis from sRLs but not zRLs. We first confirmed that alterations in cell death, proliferation, or sorting patterns of RL cells do not contribute to the differences observed (figs. S4 and S5). We then determined that, in undisturbed limb buds, the fate of the cells used to make RLs is zeugopod and autopod (fig. S6), consistent with published fate maps (20). Thus, sRLs form a stylopod through the proximalization of distal limb bud cells, whereas, in zRLs, cells maintain their original potential.

Although the RA targets Meis1 and RARβ (10, 18) are not expressed in the cells used to make RLs (figs. S1 and S6), both were induced in early sRLs, but not in most zRLs (fig. S7). To test whether RA signaling in the somite area could be involved in RL proximalization, sRLs were treated with RAA. Instead of three PD elements, RAA-treated sRLs formed only zeugopod and autopod, as shown by morphology (Fig. 2Ac) and late expression of PD markers (Fig. 2, Bc and Cc). In addition, early RAA-treated sRLs also failed to express Meis1 and RARβ (fig. S7). Consistently, zRLs treated with RA-soaked beads expressed both markers and developed all three PD segments (Fig. 2Ad and fig. S7). These experiments show that RA signaling correlates with Meis1 expression and stylopod formation in RLs.

The question remains, however, as to why RA cannot induce PD axis duplications during limb development when it does so in distal blastemas during regeneration (2123). A possible explanation is that distal limb bud cells proximalized by exposure to RA might merge with the proximal limb bud cells, precluding production of a tandem duplication phenotype (10). To test this, we inserted RA beads into HH19-20 distal limb bud tips (200 μm thick) and transplanted them to the anterior hindbrain, far from the developing limb (Fig. 3A). Whereas control transplants developed an incomplete PD axis, usually beginning at the zeugopod level (Fig. 3Ba), RA-treated transplants developed a complete PD axis starting at the pelvic girdle or stylopod (Fig. 3Bb). Expression analysis confirmed that the proximal-most skeletal element of control grafts expressed Hoxa11 and not Meis1 (Fig. 3Ea), whereas it expressed Meis1 and not Hoxa11 in RA-treated grafts (Fig. 3Eb). Exogenously applied RA thus overcomes AER’s FGF distalizing activity in early developing limbs, producing distal-to-proximal transformations similar to those observed in regenerating limb blastemas. In transplants using HH22 donors, controls produced only autopod (Fig. 3Ca), whereas RA promoted zeugopod formation and Hoxa11, but not Meis1, expression in the proximal-most skeletal element (Fig. 3, Cb and F). Finally, RA had no effect on HH24 grafts, which only formed autopod and expressed Hoxa13 (Fig. 3, Da, Db, and G). These results indicate that RA can proximalize cells with different PD fates, but limb bud responsiveness to RA is progressively lost.

Fig. 3

Exogenous RA overcomes the distalizing effect of FGF signaling and leads to generation of the whole PD axis. (A) Grafting procedure. (Ba to Db) Skeletal preparations from 6 days after grafting transplants. Untreated HH19-20 grafts (Ba) started at various zeugopod levels (n = 10 out of 14) or proximal autopod (n = 4 out of 14), whereas the RA-treated grafts (Bb) started at the pelvic girdle (PG, inset) (n = 4 out of 5) or proximal stylopod (n = 1 out of 5). Whereas control HH21-22 grafts developed autopod [(Ca), n = 7], RA-treated ones formed zeugopod [(Cb), n = 6 out of 7]. HH24 donors developed autopod in both conditions [(Da) and (Db), n = 5 and 6, respectively]. (Ea to G) Adjacent sections showing gene expression and histological analysis of skeletal elements formed by control HH19-20 grafts (Ea) and RA-treated HH19-20 (Eb), HH22 (F), and HH24 (G) grafts. Asterisks, beads. Arrowheads mark expression surrounding skeletal elements.

Our results, together with those in the accompanying study by Cooper et al. (24), indicate that signals, not autonomous mechanisms, establish the main PD subdivision of vertebrate limbs. We also demonstrate the presence of a proximalizing signal in the embryo trunk and show its ability to induce proximal limb structures in RL transplantation experiments. Our results suggest that RA signaling is the trunk proximalizing signal; however, we cannot exclude that it might be activating or mimicking alternative endogenous signals. Rather than absolute signal levels, our study suggests that the balance between FGF and the candidate proximal signal RA instructs PD identity, a finding with potential relevance to other patterning processes (25, 26). In addition, the PD molecular code found in differentiating skeletal elements will be a very useful tool in patterning and evolutionary studies.

Supporting Online Material

Materials and Methods

Figs. S1 to S7


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. K. L. Cooper et al., Initiation of proximal-distal patterning in the vertebrate limb by signals and growth.Science 332, 1083 (2011).
  3. Acknowledgments: We thank K. Storey and R. Diez del Corral for plasmids and C. Tabin for sharing unpublished results. Supported by grants RD06/0010/0008, BFU2009-08331/BMC (M.T.) and BFU2008-00397 (M.A.R.), from the Spanish Ministry of Science and Innovation (MICINN) and fellowship CPI/0051/2007 (A. R.-D.) from Madrid’s Regional Government (CAM). The CNIC is supported by the MICINN and the Pro-CNIC Foundation.
View Abstract

Stay Connected to Science

Navigate This Article