Retinoic acid is a key regulatory switch determining the difference between lung and thyroid fates in Xenopus laevis

Temporal expression of lung and thyroid markers

To establish a baseline of when the different markers of lung and thyroid are first detectable, their expression was compared by whole mount in situ hybridization (Figure 1). Nkx2.1 expression was first detectable in the thyroid at stage 30 and appeared as a small dot immediately caudal to the cement gland. There was only a slight delay before expression of pax2 was detected in the thyroid. At stage 34, expression of nkx2.1 was detectable in the forming lung. As expected, the expression of these key transcription factors preceded the expression of the lung differentiation markers sftpb and sftpc both of which were detectable at stage 38.

Retinoic acid and FGF are required for lung development in Xenopus

A requirement for both RA and Fgf signalling has been demonstrated in mouse lung development [26–29] and we wished to confirm that this requirement is conserved in Xenopus. Using a pan-retinoic acid receptor antagonist [30], we were able to show that there is a requirement for RA signalling. However, the ability of the antagonist to suppress lung differentiation, as assayed by sftpb expression is restricted to very early stages of lung development. If RA signalling was blocked at stage 26, well before the expression of nkx2.1, lung differentiation occurred normally (Figure 2). Expression of sftpb was detectable in some embryos when RA signalling was blocked as early as stage 20, although most had no expression and those that did had lower expression levels than control embryos. These results are mirrored by the expression of nkx2.1. If RA signalling is blocked prior to stage 20, nkx2.1 expression is lost in the lung although maintained in the thyroid. At stage 20 there is lung expression in some embryos and if treatment with the RA antagonist is delayed until stage 26, essentially all embryos express nkx2.1 in the lung.

If Fgf signalling was blocked using the inhibitor SU5402 after stage 12, 20, or even stage 26, expression of nkx2.1 was lost in the lung region whereas the expression in the forming thyroid was maintained (Figure 3). Again, as expected, blocking Fgf signalling also resulted in a loss of sftpb expression, although some embryos did show some very weak sftpb expression (Additional File 1, Figure S1). The loss of Fgf signalling was also confirmed by looking at the expression of sprouty2, a direct target of Fgf signalling (Additional File 1, Figure S1). In addition, expression of foxe4 was maintained in the forming thyroid in the presence of SU5402. This suggests that Fgf signalling was required for lung specification but not thyroid specification at the times tested. Expression of foxe4 in the forming lens was maintained when FGF signalling was blocked but the lens morphology was aberrant when compared to controls in Fgf inhibitor treatments that began at stage 20 (Figure 3). Given that Fgf signalling has been implicated in thyroid development in other organisms [7], we also looked at the expression of the other early thyroid marker, pax2. We found that blocking Fgf signalling immediately after gastrulation (stage12) resulted in no thyroid expression of pax2, although expression in other tissues remained unaffected (Figure 3). When Fgf signalling was blocked at stage 20, pax2 expression in the thyroid was apparently normal, although expression was not observed in all embryos. Treatment at stage 26 had no effect on pax2 expression in the thyroid.

Exogenous retinoic acid causes the thyroid to express the lung developmental program

In order to determine if this simply represented ectopic expression of sftpb in the thyroid or a more global change towards lung differentiation in the thyroid, we examined expression of sftpc, another highly specific lung specific marker and found that it was also expressed in the thyroid primordium when embryos were exposed to RA (Figure 4). The presence of sftpb and sftpc mRNA suggests either that RA causes lung differentiation markers to be expressed in the thyroid or that it causes the conversion of presumptive thyroid tissue into lung. To distinguish between these possibilities, we examined the expression of the key early transcription factors in lung and thyroid development in response to treatment with retinoic acid. As nkx2.1 is expressed in both the early thyroid and lung, it was perhaps not surprising that there was no difference in expression of nkx2.1 when embryos were exposed to RA or when RA signalling was blocked using the RA inhibitor at stage 26 (Figure 5). However, pax2 and foxe4, both markers of the early thyroid are not expressed in the thyroid in RA treated embryos, although expression at other sites is not lost in the same embryos. Foxe1 has been described as another early marker of the thyroid [25] but we find that it does not have the same strong midline expression seen with other thyroid markers, although we cannot rule out low levels of expression. Nevertheless, RA had no discernable effect on expression of foxe1 in that region although some patterning differences were noted in the pharyngeal region (Figure 5).

Hhex is a homeobox-containing transcription factor that is first expressed in a broad domain in the anterior endoderm and expression then becomes restricted to the forming thyroid and liver [1, 31]. Exposure to retinoic acid during late gastrulation is known to eliminate hhex expression in the thyroid [16] and we wanted to ensure that our later retinoic acid treatments would still result in loss of thyroid expression. Treatment with RA at stage 26 was also able to eliminate the expression of hhex in the thyroid without loss of expression in the liver (Figure 6), further suggesting that RA blocks the full repertoire of early thyroid transcription factors. We also examined whether blocking RA signalling would change hhex expression. If RA signalling was blocked between stage 14 and 26, there was no obvious change in hhex expression in either the liver or thyroid (Figure 6).

One simple model for explaining differentiation of the thyroid might be that gastrulation movements position the presumptive thyroid far enough towards the rostral end of the embryo that it is not exposed to retinoic acid. If so, we would predict that simply transplanting the presumptive thyroid to the flank of the embryo, where there is robust aldh1a2 expression, may be sufficient to cause those precursors to differentiate as lung. This would also directly test whether endogenous levels of retinoic acid would be sufficient to cause the thyroid to lung switch. When presumptive thyroid explants were removed and cultured to stage 38 they did not express sftpb but robust expression of pax2 could be seen in the cultured explants (Figure 7A,B). When explants were transplanted into the flank of a donor embryo and allowed to heal, the explants could be seen to express sftpb (Figure 7C, E) suggesting that endogenous levels of RA can cause the presumptive thyroid to express sftpb. However, the transplants also continued to express pax2 suggesting that, similar to when RA is added between stages 30 and 34, the tissue has not lost all thyroid characteristics (Figure 7D,F). The size of the expression domain also corresponded well with the normal size of the thyroid suggesting that it was only the presumptive thyroid domain that was expressing sftpb.

RA and FGF are required for lung development in Xenopus

Little is known about the signalling mechanisms required for lung development in Xenopus. A recent study has suggested that wnt and hedgehog signalling are needed for later lung development based on the expression patterns of shh, wnt7b, wnt5a and wif1 but their functional roles have yet to be elucidated. Here we show that both RA (Figure 2) and Fgf (Figure 3) signalling are necessary for differentiation of the lung in Xenopus. Both of these signalling systems are required for lung differentiation in the mouse and the two systems have been linked as loss of RA results in reduced levels of Fgf10 and subsequent lung agenesis [14]. It would appear that both RA and Fgf are very high in the hierarchy of lung specification as both are required well before the initial expression of nkx2.1.

The requirement for RA signalling is lost quite early in the development of the lung. If the antagonist is added as early as stage 20, well before the expression of nkx2.1, there is still detectable expression of sftpb. If the antagonist is added at stage 26, the expression of sftpb is essentially the same as control embryos. The sftpb promoter in humans has binding sites for both retinoic acid receptors and for nkx2.1 and the two transcription factors act synergistically to drive expression of sftpb [32]. Although we cannot make any conclusions regarding the levels of sftpb expression using whole mount in situ hybridization, our results demonstrate that the activation of the retinoic acid receptors by RA is not a requirement for sftpb expression.

The requirement for Fgf also occurs early in Xenopus as blocking Fgf signalling at stage 20 causes a loss of nkx2.1 expression, although unlike RA, the requirement is maintained later as blocking at stage 26 still showed a complete loss of nkx2.1 (Figure 3) and subsequent sftpb expression (Additional File 1, Figure S1) in the lung.

Requirements of RA and Fgf signalling for development of the thyroid gland

When RA signalling was blocked, expression of thyroid markers was not affected (Figure 5). In mouse, Fgf8, regulated by tbx1 and arising from the secondary heart field, is required for normal thyroid development although loss of fgf8 did not eliminate the thyroid [11]. In zebrafish, hand2 is upstream of fgf8 and this pathway is essential for thyroid development and again it is the cardiac mesoderm that is the source of the Fgf signal [7]. Thus it is surprising that the SU5402 treatments had little effect on expression of either nkx2.1 or foxe4, although it was clear that SU5402 was able to cause developmental defects in the treated embryo, including loss of nkx2.1 expression in the lung, severe loss of tail (Figure 3) and loss of sprouty2 expression (Additional File 1, Figure S1).

We did find that the expression of pax2 in the thyroid was lost with early SU5402 treatments. The thyroid itself does not appear to be a direct target of the Fgf signal in zebrafish [7]. If the potential intermediate tissue were already established before we treated with SU5402 in Xenopus, this could provide an explanation for the lack of effect on nkx2.1 and foxe4 expression. Identification of the intermediate tissue is needed in order to test this possibility. The lack of pax2 expression while nkx2.1 and foxe4 expression is maintained is perhaps expected. In pax8 knockout mice, the early thyroid primordium is maintained with normal nkx2.1 expression although the thyroid is eventually lost at later stages due to apoptosis [5, 33]. Thus, eventual loss of the thyroid might eventually be expected in Xenopus when Fgf signalling is blocked at very early stages.

RA causes presumptive thyroid tissue to express lung-specific markers

We have observed that addition of exogenous RA results in expression of sftpb and sftpc in the thyroid (Figure 2,4), both of which are clear markers of the differentiated lung. This does not seem to be simple ectopic expression as thyroid expression of hhex, pax2, and foxe4 are lost with the addition of RA (Figure 5,6). Hhex, pax8, foxe4, and nkx2.1 have been identified as forming a key transcription factor network for maintaining the thyroid [5] and it appears that a similar network exists in Xenopus. Pax2 substitutes for the expression of pax8 [6] and based on our expression data, it appears that foxe4 is more closely related to the central thyroid region than foxe1 (Figure 5). Nevertheless, the early thyroid markers that we have examined are lost with the addition of RA suggesting that there has been a loss of the thyroid developmental program. The one exception is that expression of nkx2.1 is maintained but it is also normally expressed in the lung. The suppression of hhex expression by exogenous RA has also been observed in the chick although in that system there was also loss of nkx2.1 expression [13] suggesting that RA is sufficient to block thyroid development but not able to cause the fate switch that we observe in Xenopus. In the same chick study it was concluded that the thyroid must form in the absence or at very low levels of RA [13]. Our results confirm this and suggest that in Xenopus, the absence of RA is actually required for thyroid development, as presence of RA would result in potential differentiation as lung tissue.

We were not able to look at any differentiation markers of thyroid. Although thyroid hormone has been extensively studied in terms of metamorphosis in Xenopus, little is known about the ontogeny of the enzymes required for thyroid hormone production. A study examining the expression of thyroid peroxidase, type II idothyronine deiodinase, and type III iodothyronine deiodinase showed that none of these were expressed in the thyroid before stage 41 [34]. Interestingly, although initially described as thyroid gland expression, thyroid peroxidase is expressed in the lung at stage 43 [34] further demonstrating the close relationship between these two organs. We also hoped to determine whether RA-treated thyroid progenitors would form structures reminiscent of a mature lung such as branching or alveoli. However, these are very late events in Xenopus development [35] and embryos treated with RA usually developed swelling around the heart that did not allow culturing to later stages.

The recent finding that sfrp5 and wnt11, both linked to the planar cell polarity pathway [36, 37], are necessary for foregut specification suggests that morphogenetic movements of the endoderm may be important for proper specification of the different organ systems [38]. The requirement for morphogenesis suggests a relatively simple model whereby the early movements of the endoderm simply result in the furthest anterior endoderm, the presumptive thyroid, being beyond the mesodermal domains where retinoic acid is being synthesized. By simply physically moving the presumptive thyroid endoderm back to regions that would be predicted to have high levels of RA, the transplanted endoderm was able to express lung differentiation markers (Figure 7). However, we cannot rule out the possibility that other factors, present in the posterior embryo, might also be responsible for the transition.

Embryo Collection

Female Xenopus laevis frogs were injected with 600-700 IU of human chorionic gonadotrophin (Sigma) to induce ovulation. In vitro fertilization of ovulated eggs was performed in 80% Steinberg's solution containing minced testis. Embryos were dejellied with 2.5% cysteine, pH 8.0, and cultured in 20% Steinberg's solution. Embryos were staged according to the Nieuwkoop and Faber staging table [35].

Cloning of sftpb and sftpc

Adult Xenopus laevis lung cDNA was used to amplify both sftpb and sftpc. The initial clones were amplified by polymerase chain reaction using the following sequences: sftpb forward 5'-cagtgggtccacaggatgac-3' and reverse 5'-gacggccccacactctttag-3' and sftpc forward 5'-ggctgcacatgagtcaaaaac-3" and reverse 5'-ctgttccggatccatttgtg-3'. The sequence obtained from the sftpb amplicon was used to generate new primers for a 5' rapid amplification of cDNA ends procedure that gave a longer cDNA clone that included the start codon. Sequencing of the clone showed that it matched previously identified sftpb sequence (Genbank Ref. NM_001096917.1). The sftpc sequence matched the previously identified sftpc sequence (Genbank Ref NM_001096721.1).

Embryo Treatments

Embryos were treated with 1μM all-trans RA (Sigma) or 1μM pan RAR antagonist (Allergan #193109) [30, 39] in 20% Steinberg's solution at various stages depending on the experiment. Stock solutions for both RA and RAA were 1mM dissolved in DMSO, and therefore a control treatment was performed with 0.1% DMSO in 20% Steinberg's solution. Fgf signaling was inhibited by the addition of 10μM SU5402 (EMB Biosciences) in the presence of 100 μM ATP. Control treatments were DMSO and ATP alone.

Explants were removed at stage 18/20 using the cement gland as a guide. Explants were cut and maintained in 1 × MBS. Explants were transplanted by creating a small wound in the recipient embryo at the site for transplantation and the transplantation was also conducted in 1× MBS. After the explant had healed, the embryos were transferred back to 20% Steinberg's solution.

In situHybridization

Whole-mount in situ hybridizations were performed according to [40] with modifications: embryos were blocked in Tris buffered saline, 20% heat treated sheep serum; and the Proteinase K, and RNase A steps were omitted. Antisense Dig-labelled riboprobes for sftpb and sftpc were generated by cutting the plasmids with Xba1 and transcribing with SP6 RNA polymerase according to established protocols [40]. Antisense Dig-labelled riboprobes for pax2a [6], foxe1, foxe4, [25], nkx2.1 [24], and hhex [41] were also generated using the same methods. BM Purple (Roche Diagnostics) was used as the alkaline phosphatase substrate throughout and embryos were fixed for twenty minutes in MEMPFA before removing endogenous pigment in 1% hydrogen peroxide, 5% formamide, and 0.5%SSC for several hours after the colour reaction. For some images, the embryos were cleared by immersion in 1 part benzyl alcohol: 2 parts benzyl benzoate after dehydration in methanol. Embryos were visualized using a Leica MZ12 dissecting microscope and images were taken using Northern Eclipse software (Empix Imaging; Mississauga, ON, Canada).