One important manipulation that acts to respecify PD identity is to expose and early blastema to retinoic acid (RA), precursor retinoids such as Vitamin A, or synthetic retinoids. This converts distal to proximal in a concentration dependent manner. The concentration range for this effect is about 2.5 fold with both Vitamin A and synthetic retinoids, a surprisingly narrow range [8,9]; this encourages the view that the primary specification of blastemal cells depends on a graded level of expression, possibly of a critical surface component. In view of these considerations, we conducted a differential screen on newt distal limb blastemal cDNAs that was designed to identify (1) sequences that are up-regulated or down regulated by RA, (2) the subset of 1 that are co-ordinately regulated by the PD location of the blastema such that RA up, P>D or RA down, D>P, (3) the subset of 2 that are expressed at the cell surface. We identified approximately 300 different examples of (1), but only six of (2), one of which was a small protein expressed at the cell surface with a GPI glycolipid anchor [10]. This is referred to as Prod 1, although as discussed below it has now been definitively identified as the newt ortholog of the mammalian protein called CD59.

The protein sequence revealed significant similarity with members of the Ly6 family, in particular the conserved family motif between residues 77–84 as well as eight conserved cys residues. The presence of the GPI anchor was verified by expression of Prod 1 or Prod 1/GFP fusion proteins after transfection of mouse cells, followed by release of the protein following digestion with the specific phospholipase that cleaves the anchor [10]. The amino acid sequence of CD59 varies markedly between different species and analysis of protein structure has been critical in evaluating the newt sequence. The structure of human CD59 contains one a-helical region and five stretches of β-strand, and all are present in bacterially expressed Prod 1 as determined by solution NMR, including the critical helical region, which is only found in CD59 when compared to other Ly6 family members (A. Garza and P. Driscoll, personal communication).

If limb blastemas from two different PD levels are confronted in culture, they adhere and then the P engulfs the D over the next 3–5 days, whereas two P explants or two D explants remain with a stable boundary. This assay, originally established by Nardi and Stocum [11], is quite robust and provides a clear test for cellular recognition of PD identity. If the enzyme PIPLC is introduced into the culture medium after adhesion, then PD conjugates behave like PP or DD [10]. This result is consistent with a functional role for a GPI anchored protein, but does not identify which one. Two affinity purified rabbit anti-peptide antibodies to Prod 1 sequences were introduced in the same context, and both gave dose and time dependent inhibition of the engulfment response in PD conjugates, while a control rabbit antibody to the blastemal cell surface protein tomoregulin had no activity [10]. It could be argued that Prod 1 was concerned with the effector mechanism, for example, the engulfment and cell movement, rather than the recognition of PD disparity itself, but taken together with its graded expression and RA responsiveness, the antibody experiments are supportive of a direct role in PD identity.

This has received strong support from experiments of Echeverri and Tanaka [12], who have developed a system for focal electroporation of plasmids into the limb blastema of the larval axolotl. If the distal cells of the blastema are electroporated with a plasmid expressing GFP, then the labelled cells maintain a distal location during subsequent regeneration, and contribute to tissues in the hand. If the contralateral blastema is electroporated with plasmids expressing RFP and Prod 1, the labelled cells relocate and contribute to the more proximal structures. Some cells are even found in stump tissue proximal to the amputation plane. These experiments show that elevated expression of Prod 1 is sufficient to convert distal blastemal cells to proximal cells. These findings have been confirmed in our laboratory (J. Godwin et al., unpublished).

If Prod 1 is a direct determinant of PD identity, how does the confrontation of cells expressing different levels lead to activation (or inhibition) of activities such as movement, adhesion or division? We have suggested a simple model whereby molecules on opposing cell surfaces interact by homophilic adhesion, thereby allowing neighbouring cells to ‘titrate’ their expression [10]. In PD conjugates this would leave spare Prod 1 molecules on P cells at the cellular interface, and these could serve as receptors for either a cell bound or diffusible ligand. A similar model has recently been proposed for the regulation of cell division by morphogens in Drosophila imaginal discs. It may be possible to test the model in cell culture and this is considered later. An obvious requirement for the model is the existence of diffusible or cellular ligands, and this is one focus of our current efforts.

We have recently established conditions for reproducible detection of Prod 1 by immunohistochemical staining of sections of the limb. In order to preserve the immunogenicity of the Prod 1 epitopes we use a procedure originally developed for another GPI linked protein, the prion protein PrP [13]. This employs glutaraldehyde fixation and low temperature wax embedding, along with horseradish peroxidase based detection. This manuscript is concerned with some results obtained with this method, which are chosen because they contribute to some long-standing questions about PD identity.

In Fig. 2a, the expression of Prod 1 in an adult newt limb has been analysed by preparing RNA from five sections along the PD axis, followed by normalised RNAse protection analysis. The x axis is expressed as a percentage of the entire PD axis, as determined by measuring its extent in a group of adult newts. The most distal section of the limb, which has the lowest level of expression, has been assigned a normalised value of unity. Most genes show a distribution over the axis that is not significantly different from uniform, whereas a few show a gradient that is in the opposite direction, i.e., distal high. Prod 1 shows a proximal high gradient, and it should be noted that this was not part of the screen leading to its identification. The gradient appears not to be linear and the data points fall on a single exponential with a maximum proximal value of approximately 2.5 (Fig. 1b).

We also performed the RNAse protection analysis on the expression of the newt homeobox gene called Meis 2, which has been implicated in PD identity during limb regeneration [7]. In a comparison of expression in proximal and distal blastemas in parallel with the analysis of Prod 1, the difference was observed to be approximately two fold (Fig. 1c). The expression along the PD axis in normal limb was quite different from Prod 1, however, since it was a shallow D>P gradient (Fig. 1c).

It is difficult to investigate the expression of Prod 1 at the proximal limit by RNAse protection for technical reasons and we have used the immunohistochemical procedure instead. Rather than trying to examine the different tissues of the limb at each level, we focussed on the major branches of peripheral nerve that run from the proximal limit of the axis to more distal regions of the limb. Fig. 3 illustrates sections taken at different levels of a single nerve branch, which were cut and processed together on the same slide. The nerve axons do not express the antigen but it is expressed by the cells of the nerve sheath, principally Schwann cells. It can be seen that the most proximal section, taken at about 5% of the axis, is clearly stronger than the others are, and in general, the most distal section is the weakest. We have analysed five nerves from different animals in this way and although the staining of the middle three sections is somewhat variable, all examples are broadly consistent with the distributions outlined in Fig. 2a and b.

Prod 1 was isolated in part, because it is markedly up-regulated by a proximalising dose of RA. These analyses were performed at the RNA level, and in order to study its regulation at a tissue level, newts with distal limb blastemas were injected either with RA or with its vehicle, dimethylsulfoxide (DMSO), prior to analysis of Prod 1 in sections. The RA-treated distal blastemas showed a clear upregulation of Prod 1 expression throughout the mesenchymal tissues relative to the parallel samples from DMSO-treated animals (Fig. 4a and b). In observing these samples, our attention was drawn to the appearance of the RA-treated distal dermis of the regenerate. The effect of RA was particularly marked in up-regulating Prod 1 in this tissue as shown at higher power (Fig. 4c and d). The dermis is recognised as a particularly important source of cells for tissue patterning during regeneration as discussed below. These examples of antibody staining on tissue sections illustrate the utility for understanding pattern formation in a complex tissue architecture such as that of the limb.