Genetic evidence that FGFs have an instructive role in limb proximal-distal patterning

Abstract

Half a century ago, the apical ectodermal ridge (AER) at the distal tip of the tetrapod limb bud was shown to produce signals necessary for development along the proximal-distal (P-D) axis, but how these signals influence limb patterning is still much debated. Fibroblast growth factor (FGF) gene family members are key AER-derived signals, with Fgf4, Fgf8, Fgf9 and Fgf17 expressed specifically in the mouse AER. Here we demonstrate that mouse limbs lacking Fgf4, Fgf9 and Fgf17 have normal skeletal pattern, indicating that Fgf8 is sufficient among AER-FGFs to sustain normal limb formation. Inactivation of Fgf8 alone causes a mild skeletal phenotype; however, when we also removed different combinations of the other AER-FGF genes, we obtained unexpected skeletal phenotypes of increasing severity, reflecting the contribution that each FGF can make to the total AER-FGF signal. Analysis of the compound mutant limb buds revealed that, in addition to sustaining cell survival, AER-FGFs regulate P-D-patterning gene expression during early limb bud development, providing genetic evidence that AER-FGFs function to specify a distal domain and challenging the long-standing hypothesis that AER-FGF signalling is permissive rather than instructive for limb patterning. We discuss how a two-signal model for P-D patterning can be integrated with the concept of early specification to explain the genetic data presented here.

Figure 1. Fgf8 is sufficient for normal limb development

(a) Schematic diagram illustrating temporal aspects of AER-FGF gene expression and the stages at which the Msx2-cre transgene functions to inactivate Fgf8 and Fgf4 floxed alleles in the AER. Note that development of forelimb buds, as marked by Fgf8 expression, commences before that of hindlimb buds, that Fgf8 expression precedes that of Fgf4, Fgf9, and Fgf17, and that Msx2-cre functions earlier in hindlimb than in forelimb buds,. (b) Schematic diagram of mouse chromosome 14 showing the map positions of Fgf9, Fgf17, and the Msx2-cre transgene, which lies within 1 centimorgan (cM) of Bmpr1a (not illustrated), 12.8 cM from the centromere (circle). Because of this linkage, once the parental animals (male on left; female on right) were generated we could produce progeny of the genotype illustrated, in which the Fgf4 conditional null allele (Fgf4fl) is inactivated by Msx2-cre function in the AER (Fgf4;Fgf9,Fgf17 TKO mutants) at a frequency of 12.5% (n=6/48), close to the expected frequency of 15.5%. (c) Fgf4;Fgf9,Fgf17 TKO forelimb (FL) and hindlimb (HL) skeletons at E17.5, which are indistinguishable from wild-type (not shown) except for an enlarged deltoid tuberosity caused by loss of Fgf9 function after condensation. (d) qRT-PCR assays for Fgf8 and Shh expression. A representative experiment on forelimb buds from embryos at ∼E11.0 (39−40 somites; n=4 limb buds for each genotype) is shown. Values are normalized to cyclophilin expression and are shown as means ± the standard deviation. The difference between control and mutant limb buds with respect to Fgf8 and Shh expression was not significant (Fgf8, p=0.61; Shh, p=0.80). A two-tailed Student's t-Test was employed, using the average of triplicate cycle count values for each limb bud. Similar results were obtained for limb buds at 33, 34, 35, and 37 somites. Abbreviations: A, autopod; DT, deltoid tuberosity; Fe, femur; Fib, fibula; Hu, humerus; Ra, radius; Sc, scapula; S, stylopod; som, somite number; Ti, tibia; Ul, ulna; Z, zeugopod; I-V, digit numbers from anterior to posterior.

Figure 2. Effects of inactivating AER-FGF genes on skeletal development

(a-g) Comparison of skeletal preparations of forelimbs from E17.5 embryos of the genotypes indicated. Asterisk (panels b, c and i) indicates that the mutant autopod has only 4 digits. Open triangle (panel c) indicates the lack of the anterior element (radius). The differences in humerus thickness/shape and deltoid tuberosity size among mutants of the various genotypes illustrated in panels b-d were also observed among mutants of the individual genotypes, suggesting they are due to background genes. (e and f) Two examples of Fgf8;Fgf4-DKO;Fgf9^−/+ forelimb skeletons, illustrating the more and less common phenotypes, respectively. The insets show the distal element of the limbs at higher magnification. The bracket (panel f) indicates the gap between the digit-like element and the distal end of the humerus. (h-o) Expression of Sox9 at E12.5 and of Dusp6 at E10.5 (37 somites), as detected by RNA in situ hybridization in whole mount in forelimb buds from embryos of the genotypes indicated. Sox9 expression marks the condensations that prefigure the skeletal elements. The dotted white lines (panels j and k) outline the condensations that will develop into the scapula. Note the absence of the developing radius (panel j). The bracket (panel k) demarcates the region devoid of Sox9-positive cells between the developing condylar portion of the humerus and a distal condensation that presumably represents the distal part of a digit. Abbreviations, as in legend to Fig. 1, and pRa, prospective radius; pUl, prospective ulna.

Figure 3. Effects of inactivating AER-FGF genes on cell survival, limb bud size, and Meis1 expression

(a-c) Confocal images of sections through E10.5 forelimb limb buds (35−36 somites) of the genotypes indicated stained with LysoTracker Red, which labels apoptotic cells as well as healthy cells engulfing apoptotic debris, and DAPI (blue). (d) Graph showing the percent of the limb bud volume that is LysoTracker-positive for each genotype. The difference between the two mutants was statistically significant (p=0.014, two-tailed Student's t-Test). (e-g) Dorsal views of E10.5 forelimb buds (36−37 somites) from embryos of the genotypes indicated. (h) Graph showing the total volume of E10.5 limb buds (35−37 somites) of the genotypes indicated. No statistically significant difference between mutants was detected (p=0.79, two-tailed Student's t-Test). (i-k) E10.5 forelimb buds (37 somites) were assayed by RNA in situ hybridization in whole mount for Meis1 expression. The dashed tan and solid yellow brackets in each panel demarcate the Meis1-positive proximal and Meis1-negative distal domains in the limb bud mesenchyme, respectively. The length of each bracket reflects the average of the four measurements made on each sample. (l) The percent of total limb bud proximal-distal (P-D) length (i.e. sum of the lengths of the proximal and distal domains) that is Meis1-negative is shown for each limb bud. The difference between mutants was statistically significant (p=0.0003, two-tailed Student's t-Test using the average measurements obtained for each limb bud). Abbreviations: A<->P, anterior-posterior; AER, apical ectodermal ridge; D<->V, dorsalventral; P<->D, proximal-distal. Error bars in graphs show the standard deviation.

Figure 4. AER-FGF mutant phenotypes can be explained by a two-signal dynamic specification model for limb proximal-distal patterning

(a) Diagrams illustrating: (left) the concept that a proximal (P) signal(s) from the mesoderm flanking the limb bud and opposing distal (D) signals from the AER (FGFs) specify limb bud P and D domains, depicted as tan and purple boxes, respectively, by E10.5; (center) the hypothesis, based on recent fate mapping studies in chicken limb buds suggesting a common progenitor for zeugopod and autopod cells, that a middle domain (purple and gold cross-hatching) forms within the D domain adjacent to the P domain, possibly as a consequence of cell-cell interactions; (right). Three domains develop into the three classically defined segments: stylopod (S, tan), zeugopod (Z, gold), autopod (A, purple). (b) Diagrams representing wild-type and AER-FGF mutant forelimb buds in dorsal view at E10.5 and E11.5. Note that at E10.5, all mutant limb buds are equal in size but smaller than wild-type (quantified in this study for all genotypes except F8;4-DKO); by E11.5 mutant limb bud size is reduced in proportion to the reduction in AER-FGF signal. At E10.5, dying cells (Fig. 3a-d ; data not shown; ref. 14) are represented by red dots; cell death in the AER is not depicted. By E11.5, dying cells are no longer observed in the mutant limb buds. (c) Meis1-positive P and Meis1-negative D domains, represented by tan and purple boxes, respectively, can be compared between mutants because the limb buds are the same size (see Fig. 3 e-h and text). Tan and purple circles represent the opposing P- and D-signals, respectively. In the mutants, the P-signal remains the same as in wild-type limb buds, while the D-signal decreases in proportion to the AER-FGF signal. Arrows indicate the extent of limb bud P-D length that is under the influence of these signals. Note that in mutants the P-signal extends over a region that, in wild-type limb buds, would normally be exposed to D-signal. This occurs both because the limb bud is reduced in size due to loss of cells specifically in the P domain and because the D-signal is much weaker than normal. Consequently, cells that would have been fated to form autopod or zeugopod if the AER-FGF signal were normal, are specified as proximal and develop into stylopod. (d) Schematic diagrams representing the size and number of elements present in forelimbs of each genotype at birth. In F8;9-DKO limb buds, enough cells were specified as distal to form most Z and A elements, and S was only slightly reduced in length due to proximal cell death before and after E10.5. In F8;4-DKO;F9^−/+ limb buds, the distal domain was reduced in size such that only the distal-most A and no Z elements formed, and the S was substantially reduced in length due to proximal cell death before and after E10.5.