A reevaluation of X-irradiation-induced phocomelia and proximodistal limb patterning.

Phocomelia is a devastating, rare congenital limb malformation in which the long bones are shorter than normal, with the upper portion of the limb being most severely affected. In extreme cases, the hands or fingers are attached directly to the shoulder and the most proximal elements (those closest to the shoulder) are entirely missing. This disorder, previously known in both autosomal recessive and sporadic forms, showed a marked increase in incidence in the early 1960s due to the tragic toxicological effects of the drug thalidomide, which had been prescribed as a mild sedative. This human birth defect is mimicked in developing chick limb buds exposed to X-irradiation. Both X-irradiation and thalidomide-induced phocomelia have been interpreted as patterning defects in the context of the progress zone model, which states that a cell's proximodistal identity is determined by the length of time spent in a distal limb region termed the 'progress zone'. Indeed, studies of X-irradiation-induced phocomelia have served as one of the two major experimental lines of evidence supporting the validity of the progress zone model. Here, using a combination of molecular analysis and lineage tracing in chick, we show that X-irradiation-induced phocomelia is fundamentally not a patterning defect, but rather results from a time-dependent loss of skeletal progenitors. Because skeletal condensation proceeds from the shoulder to fingers (in a proximal to distal direction), the proximal elements are differentially affected in limb buds exposed to radiation at early stages. This conclusion changes the framework for considering the effect of thalidomide and other forms of phocomelia, suggesting the possibility that the aetiology lies not in a defect in the patterning process, but rather in progenitor cell survival and differentiation. Moreover, molecular evidence that proximodistal patterning is unaffected after X-irradiation does not support the predictions of the progress zone model.

According to the Progress Zone Model7, PD structures develop under the influence of a continuous signal, now understood to be a fibroblast growth factor (FGF)8, produced by the overlying apical ectodermal ridge (AER). Initially, the entire limb mesenchyme has a proximal identity however, the mesenchymal cells at the tip of the limb bud under the AER—those in the “Progress Zone”—are exposed to FGF, allowing them to transition to more distal fates. As the progress zone cells divide and the limb grows, not all of these cells remain within range of the FGF signal. The fate of the cells exiting the progress zone becomes fixed, whereas those in this zone are once again respecified to more distal fates. The Progress Zone Model elegantly explained the effects of X-irradiation on developing limb buds5 by stating that the X-irradiation-induced cell death and resultantly smaller limb buds cause the proximal cells to remain within range of the AER-produced FGF signal for a longer time than normal, until the limb bud recovers and grows to a sufficient size. Consequently, proximal cells spend an increased amount of time in the progress zone, and ultimately are specified to distal fates5.

The X-irradiation experiments, however, were performed prior to the identification of molecular markers for each limb segment. The use of such markers could provide powerful additional data supporting the Progress Zone-based interpretation of irradiation-induced phocomelia or, alternatively, drawing into question the model’s validity. We therefore decided to reexamine this paradigm with modern molecular tools.

We irradiated embryos at embryonic day (E) 3.5 (stage 19-21) and grafted right limb buds onto host wings as described5. Skeletal elements were examined at E9.5 (6 days post irradiation, dpi). As in previous studies3, 4, the phenotype was dose-dependent. While unirradiated limbs developed normally on the host after grafting (Figure 1A(a,e), minor skeletal malformations such as elbow joint fusion were observed at low doses (Figure 1A(b,f)). Intermediate doses resulted in the loss of the most proximal element, the humerus, and anterior digits (Figure 1A(c,g)) and at the highest irradiation dose, only digits formed (Figure 1A(d,h)). To eliminate the possibility that the irradiation phenotype could be explained by reabsorption of proximal tissue following grafting, we performed a parallel series of experiments irradiating limbs while shielding the rest of the embryo. Using this method, we generated embryos with shorter limbs lacking proximal elements. Thus, at a dose of 17.5-20 Grays, 28% irradiated right forelimbs (RFLs) were missing the humerus, and 36% lacked humerus and radius/ulna (n=36; Figure 1B(a-c, lower limbs)).

Figure 1

X-irradiation of chick limbs causes phocomelia.

(A) Limb grafts (a - d, arrows; e-h) are shown next to host limbs. Controls (a,e) develop a humerus (h), radius(r), ulna (u), and digits (d). Low doses cause elbow fusion (asterisk) and reduced anterior digit 2 (f). Intermediate dosed-limbs have no humerus, but form a partial radius and ulna and normal posterior digits (c, g). At high doses, only digits form (d,h).

(B) Irradiated RFLs from shielded embryos (lower limb; a-c) lose proximal elements compared to LFLs (upper limb; a-c). Some irradiated specimens lack an anterior (a, b) or posterior digit (c, lower limb). Occasionally, LFLs had mild abnormalities (asterisks). a-c are ventral images at the same scale. Scale bars =1mm

For the Progress Zone Model to explain the phocomelia phenotype, the growth of the limb bud must be impaired by irradiation, while exposure to the permissive AER-FGF signal must not be disrupted. The irradiated limb buds are smaller than unirradiated and contralateral controls at 24, 48 and 72 hours post irradiation (hpi) (Supplementary Figure 2A, C, D). To determine whether FGF signaling continues unabated after X-irradiation, we performed whole mount in situ hybridization, examining Fgf8 expression in irradiated limbs from grafted or shielded embryos. In all irradiated limbs, at all stages examined, we found normal Fgf8 expression (100%, n=18; Supplementary Figure 2A, B). Maintenance of normal AER-FGF signaling levels requires the expression of a second signaling molecule, Sonic hedgehog (Shh), in the posterior limb bud mesenchyme9, 10. Consistent with this, we observed no loss in Shh expression in irradiated limb buds (100%, n=13; Supplementary Figure 2A, B). Moreover, AP-2, a transcription factor dependent on FGF signaling11, was expressed in irradiated limbs at all stages analyzed (100%, n=30; Supplementary Figure 2C, D), verifying that irradiation does not disrupt AER-FGF signaling to limb mesenchyme.

The Progress Zone Model predicts that continual FGF signaling within the geometry of a small limb bud would lead to respecification of proximal tissue to more distal fates, thereby explaining the phocomelia. To verify this hypothesis, we investigated the specification of the three major elements of the limb: the stylopod/humerus, zeugopod/radius-ulna, and autopod/digits. The best markers for these limb segments are, respectively, Meis1/2, HoxA11, and HoxA13 (Figure 2C). However, their expression patterns are dynamic at early stages and there is scant evidence that these transcription factors are required for PD specification12. Nonetheless, at later stages they uniquely delineate each segment of the limb along the PD axis, providing a molecular indication of their specification. The Progress Zone Model would predict that proximal cells, being respecified to distal fates, would no longer express Meis1,2 and instead express distal markers such as Hoxa11 or Hoxa13.

Figure 2

Expression analysis of irradiated limbs reveals defects in differentiation but not specification of the proximal segments.

(A) In irradiated limb grafts, Meis1, HoxA11, and HoxA13 expression is reduced at 24 hpi (i, j, k, l, arrowheads) compared with controls (a-d). At later timepoints, expression demarcates three limb segments (m-o; q-s, brackets) as in controls (e-g, brackets). Sox9 expression is found distally (p, t, arrows) compared with controls (d, h). (Grafted limbs are located in the anterior margin of host limbs.)

(B) At 24 hpi, irradiated RFLs displayed reduced expression of Meis1 (brown), HoxA13 (purple; b; 91%, n=11) and Sox9 (d; 100%, n=6) compared with LFLs (a, c). After 48 hpi, Meis1 and HoxA13 expression in irradiated RFLs (f, j, arrowheads; 100%, n=12) was similar to LFLs (e, i, arrowheads). Sox9 was only expressed distally in irradiated RFLs (100%, n=9) at 48 hpi (h, arrows). k and l were stained at 96hpi (u, ulna; d, digit).

(C) At stage 23 (left), Meis1 (red) is proximal; HoxA11 (grey) is distal and partially overlapping with HoxA13 (purple). After stage 28 (right), expression demarcates three segments (Pr, proximal; D, distal; A, anterior; P, posterior).

Surprisingly, whole mount in situ hybridization (WISH) shows no difference in the relative domains of expression of these markers in irradiated and unirradiated limbs (Figure 2A(q-s) and 2B(i,j)), although at an early time point, expression of Meis1, Meis2, and HoxA11 is reduced and HoxA13 expression was virtually undetectable in both grafted and shielded X-irradiated limbs (Figure 2A,B; data not shown). This was likely due to developmental delay as normal forelimb HoxA13 expression initiates after that of HoxA1113. We also found that hindlimb HoxA13 expression was not disrupted following irradiation, consistent with HoxA13 hindlimb expression preceding its expression in the forelimb (data not shown). After 48 hpi, Meis1, Meis2, HoxA11 and HoxA13 expression returned and demarcated the three limb segments. To confirm that there are three distinct PD domains in irradiated limbs, we performed double WISH with Meis1 and HoxA13. Similar to contralateral controls, three distinct domains were identifiable in irradiated limbs, a distal HoxA13 domain, an unstained middle region, and a proximal Meis1 domain (Figure 2B(e, f, i, j)).

Our data, thus suggesting that limb segment specification is not affected by X-irradiation, raises the question of why this treatment causes a preferential loss of proximal structures. We reasoned that this could be explained if X-irradiation led to changes in apoptosis, proliferation, or vasculature disproportionately in the proximal limb. X-irradiation causes both apoptosis and cell cycle arrest14. Consistent with this, we see an increase in apoptosis, using TUNEL staining at 3 and 24 hpi (Figure 3A, Supplementary Figure 3 and 4), as well as a 56% decrease in mitotic cells at 3 hpi by phosphorylated Histone H3 (pH3) staining (Figure 3B, Supplementary Figure 3 and 4). At later stages, neither apoptosis nor proliferation rates differ from contralateral and unirradiated controls or from control and irradiated grafts (Figure 3A(c, f, i); 3B(b, d, f), Supplementary Figure 3). Similar results were reported in prior analyses5. Importantly, however, neither the cell death nor the proliferative changes we observed were localized to or enriched in proximal domains of the limb. Similarly, there do not appear to be any differential changes in the vasculature along the PD axis (data not shown). Thus, these effects would not likely be responsible for the specific failure of the proximal structures to form.

Figure 3

Changes in cell death and proliferation do not explain the loss of proximal structures.

(A) An increase in TUNEL-positive cells in irradiated RFLs from shielded embryos (d, e) and irradiated grafts (h) compared with unirradiated controls (a, b, g). Levels are relatively normal at 48 hpi (c, f, i). Autofluorescence in (g, i) is due to blood cells.

(B) A decrease in pH3-positive cells at 3 hpi in irradiated RFLs (c) compared to controls (a). By 24 hpi, pH3-staining of irradiated RFLs (d) and grafts (f) is similar to controls (c, e).

Sections are longitudinal except for h and f, which are frontal.

We next examined whether X-irradiation differentially affects differentiation along the PD limb axis. In the limb, condensation of the skeletal elements occurs progressively, with proximal elements condensing before distal ones. Sox9 is an early marker of condensing mesenchyme and is necessary for cartilage and bone formation15. Sox9 expression is first detected in the limb bud at Stage 21, concurrent with or just after irradiation. We find that at 24 hpi, Sox9 expression is dramatically reduced or lost in irradiated limb buds compared with unirradiated and contralateral controls (Figure 2A(d, l) and 2B(c, d)). After 48 hpi, Sox9 expression is detected, but, strikingly, only in more distal regions of irradiated limb buds where condensation is initiated at later stages (Figure 2A(h, p ,t) and 2B(g, h, k, l)). Thus, this finding suggests a model in which X-irradiation depletes the number of cells throughout the mesenchyme. This depletion has dire consequences for the proximal tissue, which is in the early stages of chondrogenic condensation, and, as a result of this loss, cannot form skeletal structures. The failure of the proximal cells to condense is likely attributed to insufficient numbers of chondrocyte precursors. More distal tissue, in contrast, has time to recover before the proximal-to-distal wave of differentiation reaches it. A similar model has been proposed to explain segment loss in FGF mutants16, 17. Interestingly, the idea of a threshold number of cells required to form skeletal elements was proposed by Wolpert and colleagues as an explanation as to why structures that have already been specified were affected by increasing doses of irradiation5.

If this model is correct, irradiation at early stages affects the formation of proximal skeletal elements because cells in that domain initiate condensation first, and hence, undergo this process concurrently with the cellular response to irradiation damage. According to this view, we reasoned that irradiation at later stages should selectively affect the later condensing distal elements. We irradiated the limbs of shielded embryos at E4.5 (stage 24) and examined their skeletal elements at E8. The irradiated limbs had severely shortened zeugopods while other limb segments were minimally affected (Figure 4A(arrow in f)). As expected, there was an initial reduction in proliferation and an increase in apoptosis throughout the mesenchyme of the late irradiated limbs (Supplementary Figure 5A; data not shown). Corresponding to the reduction in zeugopod size, Sox9 expression was abnormal and reduced in the middle segment of irradiated limb specimens despite the presence of HoxA11 expression (Figure 4A(arrow b, e)).

Figure 4

X-irradiation disrupts chondrogenesis and does not convert proximal cells to distal fates.

(A) Expression of Sox9 (b; 100%, n=8) and HoxA11 (100%, n=6), and skeletal preparations of stage 24 irradiated RFLs (b, e, f; arrows) and LFLs (a, c, d).

(B) Expression of HoxA13 (100%, n=3) and Sox9 (100%, n=3) and Alcian Blue-staining of stage 26 irradiated RFLs (b, e, f; arrows) and LFLs (a, c, d).

(C) DiI (red fluorescence) labels the humerus in controls, (a, d, g, arrowheads; 100%, n=5). In irradiated RFLs, DiI did not expand distally (b, e, h, arrowheads; 100%, n=4 and c, f, i, arrowheads; 100%, n=2). Scale bar =1 mm; h, humerus; r, radius; u, ulna; d, digits.

We saw preferentially distal effects when we irradiated later E5.5 (stage 26) limb buds. Most strikingly, a complete loss of the autopod was observed in these irradiated limbs (Figure 4B(arrow in f)). The proximal limb cartilage had less TUNEL positive cells, consistent with our finding that the stylopod and zeugopod are little affected in these limbs and in agreement with the idea that differentiated cartilage is resistant to irradiation-induced cell death18 (Supplementary Figure 5B). Sox9 expression was virtually lost in the handplate at 24 and 48 hpi, in spite of continued HoxA13 expression (Figure 4B(b, arrow in e); not shown). Taken together, these results support the idea that X-irradiation at successively later stages sequentially affects more distal limb segments, where chondrogenic condensation is occurring, and as in the earlier stage irradiations, segment specification is not affected.

These conclusions depend on the use of markers which, although congruent with the limb segments, may not be involved in their specification. Therefore, we tested our conclusions in a second way. According to the Progress Zone Model where proximal tissue, and indeed, the entire limb bud is respecified as distal following irradiation, all surviving mesenchymal cells of the limb bud, including the proximal ones, should contribute to the distal skeletal structures that form. In contrast, if specification is unaffected, but skeletal elements fail to differentiate from proximal mesenchyme, then proximal tissues should not contribute to the distal elements forming after irradiation.

To follow the fates of proximal tissue after X-irradiation, we injected the lipophilic dye DiI approximately 300 μm from the AER in control and irradiated stage 19-20 limb buds. Consistent with the results of previous fate maps19, injection at this location in unirradiated limb buds labels the stylopod segment (Figure 4C(arrowheads in a, d, g)). In contrast, DiI injected proximally in irradiated limbs had limited expansion and did not contribute to the distal skeletal elements that formed (Figure 4C(arrowheads in b, c, e, f, h, i)). Another lipophilic dye, DiO, however, when injected distally within 50 μm of the AER in irradiated limb buds did expand into the distal regions (Supplementary Figure 6). This result supports our hypothesis that the loss of proximal skeletal elements is due to failure of those segment progenitors to condense and not because the cells occupy the progress zone for longer periods and are respecified to distal fates.

Our data, in conjunction with earlier studies reevaluating the effect of AER removal20, eliminate the major experimental support for the Progress Zone Model. While our study does not address the mechanism by which PD specification does occur, it is consistent with models in which specification is driven by the influence of traditional signaling centers12, 21 rather than by time spent within a progress zone. Most importantly, our data indicate that phocomelia caused by X-irradiation results from a loss of prechondrogenic progenitors and a consequent failure in differentiation, rather than from a defect in PD patterning.

Increased cell death has been thought to underlie thalidomide-induced limb truncations in chick embryos, but whether this is a result of direct activation of caspase pathways22, 23, or an indirect result of angiogenic inhibition, remains unclear 24. Although thalidomide treatment in chick causes distal truncations, thalidomide’s effects in humans on predominantly proximal segments likewise have been suggested to be due to cell death. Indeed, cell death has been linked to phocomelia in other contexts of teratogenic exposure, such as with nitrogen mustard25, 26 and the phocomelia observed in these experiments has been interpreted mainly as a patterning defect5,6. However, our results suggest that in these cases, as in the irradiation experiments, cell death may lead to phocomelia not by producing a smaller limb bud in the context of a Progress Zone, but by eliminating chondrogenic precursors during a time window when proximal condensation is compromised but distal differentiation has not yet commenced.

Supplementary Figure 1. Effect of irradiation on skeletogenesis of shielded embryos.

(A-C, top) Skeletons from shielded embryos were relatively normal. Enlarged images (lower panels) of the irradiated limbs prior to dissection for Figure 1B(a-c). Abnormalities in the vertebrae, clavicle, scapula and ribs were occasionally observed (asterisks). All are ventral views of the embryos.

(D) Unshielded embryos were irradiated at 1000 rads at stage 19-21 and harvested at E10. While LFLs (upper limb) were virtually normal, irradiated RFLs (lower limb) often exhibited elbow fusion. This result corresponds to the phenotype in the low irradiation dose grafts (Figure 1A(b, f).

Supplementary Figure 2. FGF signaling to the limb mesenchyme is not disrupted after irradiation.

(A) Expression of Shh (brown) and Fgf8 (purple) is similar between irradiated RFLs and LFLs. Irradiated RFLs are smaller than LFLs (both images are at the same magnification).

(B) Expression of Shh and Fgf8 (both purple stain) is unperturbed in irradiated grafts (upper limb, right panel) compared with unirradiated grafts (upper limb, left panel).

(C) AP-2 is expressed in the distal limb mesenchyme at all stages examined in irradiated RFLs and LFLs from shielded embryos and in irradiated (upper limb, right panels) and unirradiated grafts (upper limb, left panels).

Supplementary Figure 3. Analysis of cell death and proliferation following irradiation of shielded embryos at stage 19-21.

(A and B) Brightfield images (left panels) of all stained sections show the morphology of the limb tissue, corresponding to the sections shown in Figure 3. Images from Figure 3 were duplicated for direct comparision with the LFL and brightfield images.

(A) DAPI and TUNEL stained (right panels) embryos from unirradiated (control), irradiated RFLs, and LFLs at 3, 24 and 48 hpi.

(B) Staining for pH3 and DAPI (right panels) in unirradiated controls and irradiated RFL and LFL at 3 and 24 hpi.

Supplementary Figure 4. Quantitative Analysis of cell death and proliferation 3 hpi

(A) An increased number of TUNEL-positive cells were found in the stage 19-21 irradiated RFLs from shielded embryos compared with LFLs (p=7.5×10-6) and unirradiated limbs (p=1×10-6).

(B) The number of pH3-positive cells per total pixel unit area of the limb bud was 56% reduced in irradiated RFLs compared with unirradiated RFLs (Student’s t-test; p=3.4×10-6). There was a 52% decrease in irradiated compared to unirradiated limbs (p=6.6×10-12), and a slightly significant difference between irradiated RFLs and LFLs (approximately a 20% decrease; p=0.014). The standard deviations are shown and averages are shown at the base of each bar.

Supplementary Figure 5. Analysis of cell death and proliferation after irradiation at stage 24 or stage 26.

(A) An increase in cell death in the stage 24 irradiated RFLs compared with LFLs and unirradiated controls (not shown) at 24 hpi.

(B) An increase in cell death in the limb mesenchyme surrounding the condensed cartilage in the stage 26 irradiated RFLs compared with LFLs and unirradiated controls (not shown).

(C) and (D) No apparent changes in the number of pH3-positive cells cells in the stage 24 and stage 26 irradiated RFLs compared with LFLs and unirradiated controls (control) at 24 hpi.

Supplementary Figure 6. Proximal cells do not contribute to distal regions of the limb after irradiation.

(A, B) Unirradiated stage 19-20 limbs were injected with DiI (red) proximally (300 μm from the AER) and DiO (green) distally (within 50 μm of the AER). At 4 dpi, DiI labeled the stylopod (arrowheads) and DiO labeled the autopod (arrow).

(C, D) Irradiated stage 19-20 limbs were injected with DiI proximally and DiO distally as in controls (A, B). DiI cells remained in the most proximal regions (arrowhead indicating red fluorescence). DiO cells, similar to controls, expanded distally (arrow pointing to green fluorescence). Scale bars = 1mm.