Localised inhibition of FGF signalling in the third pharyngeal pouch is required for normal thymus and parathyroid organogenesis

The thymus is a bi-lobed organ that is located just above the heart and is the site of thymocyte selection and T-cell differentiation. The parathyroid glands are located next to the thyroid gland and regulate calcium homeostasis through the production of parathyroid hormone. Despite their unrelated functions, the thymus and parathyroid share a common developmental origin in that they are both derived from the third pharyngeal pouch (reviewed by Blackburn and Manley, 2004; Gordon and Manley, 2011). Four pairs of pharyngeal pouches arise sequentially in the mouse embryo from anterior to posterior by the bilateral outgrowth of the pharyngeal endoderm. As the evaginating pouches make contact with the pharyngeal ectoderm, the pharyngeal region is segmented into pharyngeal arches (Graham, 2003; Cordier and Haumont, 1980).

After pouch formation, the third pouch becomes subdivided into two molecularly distinct domains that can be visualised by embryonic day (E) 10.5 of mouse development (Gordon et al., 2001). An anterior-dorsal domain, which is fated to differentiate into the parathyroid, can be distinguished by the expression of glial cells missing 2 (Gcm2), which is required for parathyroid differentiation and survival (Günther et al., 2000; Liu et al., 2007). A prospective thymus domain is marked by Bmp4 expression at the posterior-ventral side of the pouch at E10.5 (Patel et al., 2006). The expression of the first definitive thymus marker, Foxn1, appears in the posterior, Bmp4-positive pouch by E11.25. Foxn1 expression spreads further along the posterior side of the pouch towards the Gcm2+ region and this Foxn1+ domain will form the primordium of the thymus (Gordon et al., 2001). After the establishment of two molecularly distinct organ primordia, they physically separate from each other and from the pharynx and migrate to positions at the mediastinum anterior-ventral to the heart (thymus) and adjacent to the thyroid (parathyroid). Little is known about the mechanisms that allow the separation of organs from the pharynx. Apoptotic epithelial cells have been detected in the region where the thymus separates from the pharynx at E11.75 just prior to separation, suggesting that apoptosis might be required for the separation process (Gordon et al., 2004). Although Pax1, Pax9 and Hoxa3 have been shown to be necessary for the normal separation of the thymus primordium from the pharynx, these studies did not investigate whether the lack of separation is associated with the loss of apoptosis in these mutants (Su et al., 2001; Hetzer-Egger et al., 2002).

Our understanding of the identity and actions of signalling molecules that coordinate pouch formation and subsequent steps in thymus and parathyroid organogenesis is still incomplete. The signalling cascades that control Gcm2 and Foxn1 expression are complex and several pathways and transcription factors have been implicated. These include a Hox-Eya-Six-Pax gene network required for initiation and maintenance of organ-specific gene expression (Chisaka and Capecchi, 1991; Manley and Capecchi, 1995; Su et al., 2001; Zou et al., 2006). Wnt signalling, presumably as a result of Wnt4 and Wnt5b expression in the pharyngeal region, has been implicated as a regulator of Foxn1 expression (Balciunaite et al., 2002). The transcription factor Gata3, which is mutated in HDR (hypoparathyroidism, sensorineural deafness and renal disease) syndrome, directly regulates the expression of human homologue of Gcm2, GCMB (GCM2 – Human Gene Nomenclature Database) (Grigorieva et al., 2010). In addition, the bone morphogenetic protein (BMP) and sonic hedgehog (Shh) signalling pathways are required for normal thymus and parathyroid organogenesis, respectively (Ohnemus et al., 2002; Storm et al., 2003; Bleul and Boehm, 2005; Moore-Scott and Manley, 2005; Soza-Ried et al., 2008; Gordon et al., 2010; Neves et al., 2012).

Fibroblast growth factor (FGF) signalling regulates the development of several tissues in the pharyngeal region. Mutations that affect FGF signalling have been shown to affect two developmental processes that impact on thymus and/or parathyroid organogenesis: (1) Fgf8 hypomorphic mutants fail to form normal third pouches resulting in thymus and parathyroid aplasia or hypoplasia (Abu-Issa et al., 2002; Frank et al., 2002); and (2) thymic epithelial cell (TEC) proliferation after E12.5 is reduced in Fgfr2(IIIb) and Fgf10 mutants (Ohuchi et al., 2000; Revest et al., 2001). Thus, although FGF signalling has been implicated in third pouch formation (∼E9.25) and TEC proliferation (after E12.5), the role of this pathway between E10.5 and E12.5 has not been investigated directly. Research over the last few years has shown that FGF feedback antagonists of the sprouty gene family are important regulators of organogenesis (Mason et al., 2006). In this study, we report that the localised inhibition of FGF signalling by sprouty proteins is essential for several key processes during thymus/parathyroid organogenesis: (1) the normal initiation of Gcm2, Bmp4 and Foxn1 expression in the third pouch, (2) apoptosis in the third pouch that is associated with organ detachment from the pharynx, and (3) Fgf10-mediated thymus growth after E12.5.

Mutant mouse lines were maintained on a mixed (129SvEv,C57BL/6J,FVB/N) genetic background, and genotyped by PCR as described in original publications. Mutant lines harbouring conditional (floxed) Spry1 and Spry2 alleles have been described (Basson et al., 2005; Shim et al., 2005). These mice were crossed to βactinCre mice to generate Spry1 and Spry2 null alleles (Lewandoski et al., 1997). Embryos lacking both genes (Spry1;2dko) were produced by crossing βactinCre/βactinCre;Spry1^+/–;Spry2^+/– males with Spry1^flox/flox;Spry2^flox/flox females, as described (Simrick et al., 2011). For rescue experiments, βactinCre²;Spry1^+/–;Spry2^+/–;Fgf8^+/– males were crossed with Spry1^flox/flox;Spry2^flox/flox females or βactinCre²;Spry1^+/–;Spry2^+/– males with Spry1^flox/flox;Spry2^flox/flox;Fgf3^+/– females. The Fgf3 and Fgf8 null alleles have been described (Meyers et al., 1998; Alvarez et al., 2003). Endoderm-specific mutants were produced using Sox17-2A-iCre (Engert, S. et al., 2009) by crossing Sox17-2A-iCre;Spry1^flox/+;Spry2^flox/+ males to Spry1^flox/flox;Spry2^flox/flox females.

Noon on the day a vaginal plug was detected was designated as E0.5. Embryos were accurately staged by determining the number of somite pairs, indicated as somite stage (ss). All experiments involving mice were approved by the UK Home Office. At least three separate embryos of each genotype were analysed per experiment.

Embryos were dissected in PBS and fixed overnight in 4% paraformaldehyde in PBS, followed by dehydration through a series of ethanol washes and embedding in wax. Sections were cut at 7 μm and dried overnight at 42°C. In situ hybridisation (ISH) was performed according to standard methods (Yaguchi et al., 2009). The probes used for section ISH have all been previously described as follows: Fgf3 (Robinson et al., 1998), Fgf8 (Crossley and Martin, 1995), Fgf15 (Borello et al., 2008), Etv4 and Etv5 (Klein et al., 2006), Dusp6 (Dickinson et al., 2002), Foxn1 and Gcm2 (Gordon et al., 2001), Spry1 and Spry2 (Minowada et al., 1999). Immunohistochemistry (IHC) was performed according to standard methods (Yu et al., 2009). Double IHC was performed using primary antibodies raised in rabbit against activated caspase-6 (Abcam #ab52295) to analyse apoptosis, and raised in mouse against E-cadherin (Fitzgerald #02660) to label epithelial cell membranes. A rabbit antibody to phospho-histone H3 (Ser10) (Cell Signaling #9701) was used to detect proliferating cells. Secondary Alexa fluorophore-labelled antibodies were obtained from Invitrogen. All primary and secondary antibodies were diluted at 1/200. Images were acquired with a 5M pixel Nikon DS colour camera connected to a Nikon Eclipse 80i microscope and subsequently collated and annotated using Adobe Photoshop and Illustrator.

Embryos for three-dimensional (3D) reconstruction of the pharyngeal region were sectioned in the coronal plane, stained with Haematoxylin and Eosin (H&E) and photographed as detailed above. 3D reconstruction from alternate sections was carried out using Winsurf version 4.3, as described by Patel et al. (Patel et al., 2006). Relative volumes of each organ/structure were determined. The mean volume of control glands was calculated, and the volumes of littermate controls and knockouts calculated as a percentage of this mean control volume. When no stage-matched littermate controls were available, knockouts were compared with the mean volume of all stage-matched Spry1^flox/+;Spry2^flox/+ controls collected during the course of the study. Statistical analyses were carried out using Graphpad and Microsoft Excel.

At E10.5 of development in the mouse, distinct, molecularly defined domains can be observed in the third pharyngeal pouch. An anterior-dorsal, Gcm2-positive domain represents the primordium of the parathyroid (Fig. 1A,K). At this stage of development, the ventral pouch can be identified by Bmp4 expression (Fig. 1B,K). Cells in this ventral Bmp4+ domain turn on Foxn1 expression from E11 and Foxn1+ cells, destined to become the thymus, occupy a ventral-posterior region of the third pouch, complementary to the Gcm2+ region, by E11.5 (Gordon et al., 2001). As a first step towards understanding whether FGF signalling plays a role in thymus/parathyroid development between E10.5 and E11.5, we determined the expression of genes encoding FGF ligands and downstream targets in and around the third pouch at E10.5. We focused this analysis on FGF genes that had been implicated in pharyngeal or thymus development in previous studies (Revest et al., 2001; Abu-Issa et al., 2002; Frank et al., 2002; Trokovic et al., 2005; Aggarwal et al., 2006). In situ hybridisation studies on sagittal sections through the third pouch identified a small region in the posterior pouch that expresses Fgf3, Fgf8 and Fgf15 (Fig. 1C-E). Fgf10 is expressed in the neural crest-derived mesenchyme surrounding the pouch, and in a small domain in the most ventral part of the Bmp4+ pouch region (Fig. 1F). To identify regions of active FGF signalling in the pouch, we determined the expression of Spry1 and Spry2, as these genes are induced by FGF signalling in several developmental contexts (Minowada et al., 1999). Spry1 and Spry2 are expressed at low levels throughout most of the pouch, with highest levels of expression in and around the Fgf3/8/15-expressing region in the posterior pouch (Fig. 1G,H). These two sprouty genes are also expressed in the neural crest surrounding the pouch, with highest expression in the cells adjacent to the Fgf3/8/15/Spry expression domain in the posterior pouch (Fig. 1G,H). Taken together, these observations suggest that FGF ligands, produced in the posterior pouch endoderm and the adjacent neural crest, can induce Spry1 and Spry2 expression in both tissues.

Next, we visualised the expression of three additional transcriptional targets of the FGF pathway: Etv4 (Pea3), Etv5 (Erm) (Roehl and Nüsslein-Volhard, 2001; Klein et al., 2008) and Dusp6 (Mkp3) (Li et al., 2007). High Etv4 (Fig. 2A) and Etv5 expression was evident in the surrounding neural crest, especially along the anterior and posterior sides of the pouch, confirming that these cells are responding to FGF signals (Fig. 1I). By contrast, the pouch endoderm appeared to be largely devoid of Etv4 and Etv5 expression, with the exception of a distinct expression domain in the posterior pouch where FGF ligands are expressed (Fig. 1I). Dusp6 is also expressed at high levels in the neural crest, with low expression in a small region in the ventral pouch where Bmp4 and Fgf10 are expressed, but no detectable expression in the posterior pouch (Fig. 1J). The observation that several genes that report FGF signalling are expressed in and around the third pouch suggests that FGF signalling is active during pouch patterning. Although these genes show slightly different expression patterns, which probably reflects subtle differences in gene regulatory mechanisms, our data suggest the presence of an FGF signalling centre in the posterior pouch endoderm and a potential negative feedback mechanism whereby sprouty inhibits intracellular FGF signal transduction in the pouch endoderm.

By comparing the expression patterns of FGF pathway genes with other region-specific markers, we found that the FGF signalling centre was localised in the posterior pouch between the anterior-dorsal Gcm2+ parathyroid domain and the ventral Bmp4+ thymus domain, with some overlap with the latter (summarised in Fig. 1K). Taken together, these observations suggest that strong negative feedback mechanisms have evolved to inhibit the activation of FGF signalling in the third pharyngeal pouch and that Spry1 and Spry2 might serve essential functions during third pouch patterning and subsequent events that affect thymus/parathyroid development.

To test the hypothesis that Spry1 and Spry2 regulate thymus/parathyroid development by preventing excessive FGF signalling in the pouch endoderm, we generated Spry1^–/–;Spry2^–/– (Spry1;2dko) embryos. Spry1^+/–;Spry2^+/– (Spry1;2^+/–) littermate embryos produced in these crosses were normal compared with wild-type embryos with respect to thymus and parathyroid development and were used as controls. We first compared the expression of three FGF reporter genes, Etv4, Etv5 and Dusp6, in Spry1;2^+/– control and Spry1;2dko embryos (Fig. 2A-F). As shown in Fig. 2, the regionalised Etv4, Etv5 and Dusp6 expression patterns were almost entirely disrupted in the Spry1;2dko embryos and these genes were ubiquitously expressed at high levels throughout the pouch (Fig. 2B,D,F). In addition, Etv5 and Dusp6 expression was notably upregulated in the surrounding neural crest (compare Fig. 2C,E with 2D,F).

These data are consistent with a model in which the characteristic pattern and level of FGF signalling in the third pouch endoderm is maintained by the localised inhibition by sprouty proteins.

Our data thus far suggested that Spry1 and Spry2 were essential for maintaining the pattern of FGF signalling in the third pouch endoderm at E10.5, i.e. at the time when the third pharyngeal pouch becomes subdivided into prospective Bmp4+ thymus and Gcm2+ parathyroid domains (Gordon and Manley, 2011). To determine whether the changes in FGF signalling in sprouty-deficient embryos were associated with altered patterns of thymus and parathyroid development, Gcm2 and Bmp4 gene expression was analysed at E10.5. In normal embryos, Gcm2 expression is initiated in a region of endoderm encompassing the second and third pouch at E9.5 and becomes restricted to a defined region in the anterior-dorsal third pouch by E10.5 (Fig. 2G) (Gordon et al., 2001). Only a few isolated Gcm2+ cells could be detected in the third pouch in E10.5 Spry1;2dko embryos (Fig. 2J). Bmp4 expression identifies the ventral pouch region that is fated to turn on Foxn1 expression and become the thymus (Fig. 2I). Bmp4 expression was downregulated in the ventral pouch in Spry1;2dko embryos (Fig. 2J). These observations suggest that the initiation of the patterning events that subdivide the third pouch into molecularly distinct regions is disrupted in Spry1;2dko embryos.

To understand the consequences of these early alterations in gene expression on the progression of thymus/parathyroid development, we followed Gcm2 expression and the initiation of Foxn1 expression further during development. At the 45 somite stage (ss), Gcm2 expression was still seen in only a few isolated cells in the mutant embryos compared with the robust expression observed in controls (Fig. 3A,B). By this stage, Foxn1 expression is visible in control embryos in the ventral pouch and has spread further along the posterior pouch towards the dorsal Gcm2+ domain (Fig. 3C), as described previously (Gordon et al., 2001). We find that Foxn1 expression is induced in mutant embryos, albeit in fewer cells, which seem to be localised primarily in the ventral pouch with little evidence of spreading towards the dorsal side (Fig. 3D). By the 50 ss, a distinct Gcm2+ region in the correct anterior-dorsal position of the common thymus/parathyroid primordium was present in the mutant embryos (Fig. 3F). However, the size of the Gcm2+ domain was reduced compared with control embryos (Fig. 3E,F). By contrast, the Foxn1+ domain in the ventral-posterior pouch of the mutant embryos appeared no different in terms of size and intensity of expression compared with controls (Fig. 3G,H). These data suggest that Foxn1 expression is delayed in Spry1;2dko embryos at early stages of development, but that this expression eventually recovers to form a normal thymus primordium by the 50 ss. We infer from these data that the reduced expression of Gcm2 and the formation of a small Gcm2+ domain in the third pouch by E11.75 is likely to result in parathyroid hypoplasia, whereas alterations in Foxn1 expression are unlikely to have functional consequences for thymus development.

A prediction from our gene expression analyses was that a normal thymus primordium would develop in Spry1;2dko embryos, whereas parathyroid primordia would be hypoplastic. To determine the consequences of sprouty gene deletion on thymus and parathyroid development, serial sections through E12.5 mutant and control embryos were reconstructed in three dimensions and the relative positions and sizes of the thymus and parathyroid glands were compared (Fig. 4A,B). In agreement with our prediction, a comparison of the relative sizes of the individual primordia (Fig. 4E) revealed that the parathyroid primordia were significantly hypoplastic in mutants compared with controls (31% of control, n=4, P=0.002), confirming that the smaller Gcm2+ domain at E11.5 (Fig. 3F) resulted in parathyroid hypoplasia. The thymus glands were of normal size (111% of control, n=4, P=0.2), indicating that the recovery of normal Foxn1 expression by the 50 ss (Fig. 3H) was sufficient to allow for normal thymus development up to E12.5.

A number of additional defects were evident in Spry1;2dko embryos upon 3D reconstruction (Fig. 4A,B) of histological sections (Fig. 4C,D). First, the organ primordia derived from the third and fourth pouches remained attached to the pharynx. In addition, the third pouch-derived endoderm also remained attached to the ectoderm (Fig. 4B, arrow). The pharynx itself was wider than in controls, and all embryos examined had tracheo-oesophageal fistula. These data indicate that Spry1 and Spry2 are required for normal parathyroid organogenesis as well as for separation of organ primordia from the pharynx.

A previous study reported that the apoptotic death of endodermal cells at the junction between the thymus primordium and the pharynx at E11.75 (47-50 ss) was associated with the detachment of the thymus from the pharynx (Gordon et al., 2004). We therefore hypothesised that the failure of organ primordia to detach from the pharynx in the Spry1;2dko mutants might be due to defects in the initiation of programmed cell death. To test this hypothesis, apoptotic cells were visualised in control embryos by staining third pharyngeal pouch sections with antibodies to cleaved caspase-6. No apoptosis was detected in the third pouch of E10.5 embryos (34 ss, data not shown). However, a population of apoptotic cells was located in the dorsal half of the posterior pouch, close to the junction between the pouch and pharynx in the third pouch endoderm a few hours later at the 38-40 ss (E10.75) (Fig. 5A,C). This region corresponded to the domain where high levels of Spry1 and Spry2 genes were expressed (compare Fig. 5C with similar sections in Fig. 1G,H). These observations are in agreement with previous data showing that the primary site of attachment of both primordia to the pharynx is at the site between the Gcm2+ and Foxn1+ domains (Gordon and Manley, 2011).

To test whether the loss of sprouty genes prevented apoptosis in the third pouch, we also stained sections from Spry1;2dko pouches with an antibody to cleaved caspase-6. Cell death was reduced or absent in the dorsal-posterior pouches of mutant embryos at E10.75 (Fig. 5B,D), suggesting that apoptosis was indeed sensitive to the level of FGF signalling. We also observed apoptotic cells in the junction between the thymus primordium and the pharynx shortly before separation in control embryos at E11.75 (47-50 ss) (Fig. 5E,E′) as previously reported (Gordon et al., 2004). Interestingly, although apoptotic cells could be detected in this region in Spry1;2dko embryos, these cells were mainly observed in the ventral, but not the dorsal, side of the epithelial bridge between the pharynx and thymus in Spry1;2dko embryos (Fig. 5F,F′). These observations suggest that the loss of sprouty genes prevents normal programmed cell death in the dorsal-posterior pouch endoderm. Furthermore, although cell death is evident in the region between the organ primordia and the pharynx at E11.75 in sprouty-deficient embryos, the distribution of apoptotic cells in these mutants indicates that an abnormal apoptotic programme is most likely to be the cause of persistent attachment to the pharynx.

Sprouty genes are not expressed in the pharyngeal endoderm alone and defects in early thymus and parathyroid organogenesis in Spry1;2dko mutants might be due to the combined affect of deleting these genes in several different cell types in the third pharyngeal arch (Fig. 1G,H). To determine whether the loss of sprouty genes in the pharyngeal endoderm alone could recapitulate some of the phenotypes in embryos that lack Spry1 and Spry2 in all tissues, mutants in which these genes were specifically removed from the pharyngeal endoderm using a Sox17-iCre line were produced (Engert, S. et al., 2009; Simrick et al., 2011). Cre activity in the pharyngeal pouch endoderm was confirmed by visualising the green fluorescent reporter in Sox17-iCre;RosaYFP mice (Srinivas et al., 2001) (Fig. 6A). Sox17-iCre; Spry1^flox/flox;Spry2^flox/flox conditional mutants exhibited similar phenotypes to Spry1;2dko mutants at E12.5. Parathyroids were hypoplastic (43% of controls, n=3, P=0.004), thymus volumes were not statistically different from controls, and these organs failed to separate from the pharynx and were localised in abnormal orientations relative to each other (Fig. 6B,C). Analysis of Gcm2 and Foxn1 expression at E11.5 also demonstrated a smaller Gcm2+ domain (Fig. 6D-G). Etv5 expression was again upregulated throughout the pouch endoderm at E10.5 (Fig. 6H,I). These data indicate that the function of sprouty genes as FGF inhibitors in the pouch endoderm is essential for establishing normal pouch patterning and organ separation from the pharynx.

The results presented so far suggest that sprouty genes function to restrict FGF signalling in the third pouch endoderm during thymus/parathyroid development. However, as sprouty proteins act intracellularly to antagonise conserved signalling pathways that are utilised downstream of many receptor tyrosine kinases (Mason et al., 2006), the possibility that the phenotypes in Spry1;2dko embryos were due to hyperactive signalling downstream of receptor tyrosine kinases (RTKs) other than FGF receptors remained. We therefore tested whether genetic reduction of Fgf3 or Fgf8, both expressed in the third pouch at E10.5 (Fig. 1) and reported to be involved in thymus organogenesis (Frank et al., 2002; Aggarwal et al., 2006), could rescue any of the observed phenotypes, which would indicate that hyperactive FGF signalling is responsible for that particular defect. Fgf3 heterozygosity had no effect on any of the phenotypes in Spry1;2dko embryos (Fig. 7A-F,I). By contrast, halving the Fgf8 gene dosage resulted in a partial rescue of the parathyroid and separation defects at E12.5. The parathyroids in Spry1;2dko,Fgf8^+/– embryos were consistently found in the correct position dorsal to the thymus (n=6/6), had separated from the thymus primordia and were significantly larger than parathyroids in littermate S1;2dko mutants, although still hypoplastic compared with controls (65% of controls, compared with 25% in Spry1;2dko littermates, n=3, P=0.03) (Fig. 7G-I). In addition, three out of six primordia examined had separated from the pharynx. Taken together, these observations indicate that reduced Fgf8 levels moderated the effects of sprouty mutants, suggesting that sprouty proteins inhibited signalling downstream of Fgf8 during third pouch patterning. We conclude that increased Fgf8 signalling is at least partially responsible for the separation defects and parathyroid hypoplasia in Spry1;2dko embryos.

Analysis of E14.5 embryos (Fig. 8A,B,E) indicated that the parathyroid hypoplasia observed in sprouty-deficient embryos at E12.5 failed to recover during development (mutant parathyroids were 23% the size of controls, n=4, P=0.003). Surprisingly, sprouty mutant thymi were also hypoplastic by E14.5 (56% of control, n=4, P=0.007), suggesting a role for sprouty genes in thymus growth between E12.5 and E14.5 (Fig. 8E). As a consequence of the physical attachment of the thymus to the pharynx (Fig. 8D), the hypoplastic thymus lobes were present in ectopic locations in the neck adjacent to the pharynx, compared with normal thymi, which by this stage had migrated to positions above the heart (Fig. 8A,B). Hypoplastic thymi located in ectopic positions in the neck were clearly visible in embryos just before birth (Fig. 8F,G).

As previous studies have suggested that FGFR2(IIIb) ligands such as Fgf10 are responsible for driving the rapid expansion of the thymus primordium after E12.5 (Ohuchi et al., 2000; Revest et al., 2001), we investigated whether sprouty genes were still expressed in the thymus primordium at this time of growth. In situ hybridisation experiments demonstrated that Spry1 and Spry2 genes were expressed specifically in cells of the thymic capsule that surrounds the thymus primordium at E13.5 (Fig. 8H,I) and not in the thymus itself (marked by Foxn1 expression in Fig. 8J). Comparison of Fgf10 expression in control and Spry1;2dko embryos at E13.5 revealed a clear reduction in Fgf10 expression in the thymus capsule of sprouty-deficient embryos at E13.5 (Fig. 8K,L). To determine whether the loss of Fgf10 expression was associated with reduced proliferation of thymic epithelial cells, cells in mitosis were quantified after staining with an antibody to phospho-histone H3. Spry1;2dko thymic primordia contained significantly fewer mitotic cells compared with control thymi at E13.5 (Fig. 8M-O). These observations suggest that thymus hypoplasia in Spry1;2dko embryos is due to reduced thymus growth as a result of the loss of expression of the thymic epithelial mitogen Fgf10 in the thymic capsule at E13.5 of development.

Our analysis of thymus and parathyroid development in sprouty-deficient embryos has revealed several key developmental processes that require tight regulation of FGF signalling. Inhibition of FGF signalling is required for the normal initiation of Gcm2 expression and the establishment of a normal-sized parathyroid primordium. Although deregulated FGF signalling is also associated with defects in the initiation of Bmp4 and Foxn1 expression, these abnormalities appear to recover such that a normal thymus primordium initially forms. However, we produce evidence for a continued requirement for sprouty gene function during later stages of thymus expansion, such that the loss of sprouty genes is associated with loss of Fgf10 expression in the thymus capsule, reduced TEC proliferation and thymus hypoplasia. Sprouty gene function is also required for the regulation of FGF signalling during organ detachment from the pharynx.

Our data indicate that increased FGF signalling in the absence of Spry1 and Spry2 results in diminished induction of Gcm2 expression resulting in the formation of a small Gcm2+ domain by E11.5 (Fig. 9). These observations suggest that defects in Gcm2 expression underlies the parathyroid hypoplasia observed in Spry1;2dko mutants. Currently, the exact mechanism by which Gcm2 expression is repressed in sprouty mutants is not known. Hyperactive FGF signalling might repress Gcm2 transcription either directly, or by inhibiting the expression or activity of upstream transcriptional regulators of Gcm2, such as Gata3 (Grigorieva et al., 2010). Previous studies have shown that Shh signalling induces Gcm2 expression (Moore-Scott and Manley, 2005), raising the possibility that the effect of sprouty gene deletion on Gcm2 expression might be indirect through, for example, the inhibition of Shh signalling, as recently reported in the developing cerebellum (Yu et al., 2011).

We found that Bmp4 expression, which marks the presumptive thymus domain, is reduced in Spry1;2dko embryos (Fig. 9). Changes in FGF signalling are documented to affect Bmp4 expression in other developmental contexts, such as forebrain patterning, raising the possibility that similar mechanisms might be responsible for these changes during third pouch patterning (Storm et al., 2006). Bmp4 has been implicated in the regulation of Foxn1 expression in mouse and chick and reduced BMP signalling results in thymus hypoplasia in the mouse (Ohnemus et al., 2002; Neves et al., 2012; Bleul and Boehm, 2005; Soza-Ried et al., 2008; Gordon et al., 2010). These observations suggested the possibility that the reduced Bmp4 expression in the sprouty-deficient pouch at E10.5 might result in thymus hypoplasia in Spry1;2dko embryos. However, despite these early alterations in Bmp4 expression, and a corresponding delay in the initiation of Foxn1 expression in sprouty-deficient embryos, Foxn1 expression appears largely normal by the 50 somite stage and thymus primordia are of normal size by E12.5 of development. These observations suggest that an early defect in Foxn1 expression can recover to allow for normal development to ensue. A similar example of a recovery of an early defect in tooth development due to deficient BMP signalling was recently reported, suggesting that this phenomenon of ‘developmental stalling’ might be a common feature of BMP-regulated developmental processes (Miletich et al., 2011).

Curiously, thymi were significantly smaller by E14.5, suggesting that the loss of sprouty genes affected thymus expansion between E12.5 and E14.5. As Fgf10/Fgfr2b signalling has been shown to be required for the growth of the thymus primordium after E12.5, one might have expected the opposite phenotype, i.e. thymus overgrowth when deleting FGF antagonists. However, we find that sprouty genes are not expressed in TECs and that their deletion did not result in FGF hyper-responsive, over-proliferating TECs owing to the loss of FGF inhibition in TECs. Instead, sprouty genes are expressed in the surrounding thymus capsule at E13.5 and the loss of sprouty genes was associated with downregulation of Fgf10 expression and reduced TEC proliferation. The downregulated expression of Fgf10 might be due to a direct repression of Fgf10 gene expression by FGF signalling. A previous study has shown that the Ets transcription factor Pea3, the product of the Etv4 gene, which is induced by FGF signalling, can bind directly to an Fgf10 promoter and repress Fgf10 transcription (Chioni and Grose, 2009). The identification of capsule-specific Fgf10 gene regulatory elements and their regulation by FGF signalling should provide further insights into this phenomenon. We also considered the possibility that the loss of Fgf10 expression might be due to general defects in the thymus capsule. Fgf10 expression in the neural crest surrounding the third pouch was unaffected at E10.5 (data not shown), ruling out severe defects in early neural crest development. Furthermore, the thymus capsule appears histologically normal in sprouty mutants at E13.5 (black arrows in Fig. 8M,N). Nevertheless, we cannot at this stage rule out the possibility that subtle changes in the differentiation state of these cells could account for a reduced ability to express Fgf10.

A striking consequence of deleting sprouty genes during development is the failure of organ primordia to detach from the pharynx and from each other. This phenotype was preceded by the loss of apoptosis normally present in the posterior-dorsal pouch from E10.75. This apoptotic domain is localised to a region between the prospective Gcm2+ parathyroid and Bmp4+ thymus domains in the dorsal half of the posterior pouch endoderm at E10.5 where the pouch is attached to the non-pouch pharyngeal endoderm (PE in Fig. 9). We have detected the expression of Fgf3, Fgf8 and Fgf15 in and adjacent to this region, suggesting that the regulation of FGF signalling might be intimately linked to the separation process. The observation that low levels of FGF signalling in the posterior-dorsal pouch are associated with apoptosis during normal development is in agreement with previous demonstrations that FGF signalling is required to maintain cell survival in several developmental contexts including the brain (Basson et al., 2008; Paek et al., 2009), limb (Sun et al., 2002), kidney (Grieshammer et al., 2005) and olfactory system (Kawauchi et al., 2005). In all these examples, abnormally reduced FGF signalling is associated with aberrant cell death and loss of tissues or structures. In the present study, we show that the localised inhibition of FGF signalling by sprouty genes is essential for correct morphogenesis by allowing the formation of a discrete domain of cell death in the pouch endoderm. Furthermore, our data suggest that the surviving Foxn1;Gcm2-negative cells persist as an epithelial bridge between the pharynx and the thymus/parathyroid primordia in Spry1;2dko embryos, easily observed a day later, at E11.75. We propose that these changes in morphology and apoptosis, arising due to increased FGF signalling, are collectively responsible for the failure of organ detachment.

A recent study showed that the deletion of an essential intracellular mediator of FGF signalling, Frs2α, is associated with a general failure of organ detachment from the pharynx (Kameda et al., 2009). Superficially, this observation appears to be at odds with our findings. However, previous studies focusing on nervous system development have found that increasing or decreasing FGF signalling could have similar effects on cell survival (Storm et al., 2003). It will be interesting to determine whether sprouty gene expression is sufficiently downregulated in the Frs2α mutants to prevent apoptosis in the posterior pouch endoderm in an analogous fashion to the findings in the forebrain.

The downstream, transcriptional effectors of FGF signalling that are responsible for mediating the various phenomena described in this manuscript are not known. Likely candidates are the Ets family transcription factors encoded by the Etv4 and Etv5 genes (Fig. 9), as recent studies have shown that these factors are the physiological effectors of FGF signalling in other developmental systems (Mao et al., 2009). As far as we are aware, thymus/parathyroid development has not been investigated in mutants that lack Etv4 and/or Etv5. Although it is difficult to predict the thymus/parathyroid phenotypes of these mutants at this stage, we predict that the inducible overexpression of Etv4 or Etv5 would result in similar effects on thymus and parathyroid development as we describe here for sprouty-deficient embryos.

In conclusion, we have identified several roles for FGF antagonists of the sprouty gene family during thymus and parathyroid organogenesis. The inhibition of FGF signalling by sprouty proteins is crucial for normal Gcm2 induction and parathyroid size, apoptosis required for organ separation from the thymus, and expansion of the thymus primordium by proliferation. This study suggests that a full understanding of these developmental processes will require studies aimed at elucidating the role of FGF signalling and the integration of FGF signalling with other developmental pathways.

We thank Gail Martin for the Spry2, βactincre, Fgf8 and R26R mouse lines, Frank Costantini for the RosaYFP line and Thomas Schimmang for the Fgf3 line. In situ probes were kind gifts from Silvia Arber (Etv4, Etv5), Stephen Keyse (Dusp6), Spry1, Spry2, Fgf8 (Gail Martin), and Fgf3 (Ivor Mason). We are grateful to Mohi Ahmed for assistance with the model diagrams and Hagen Schmidt and Samantha Martin for expert animal husbandry and technical assistance.