Analysis of origin and growth of the thyroid gland in zebrafish



The zebrafish thyroid gland shows a unique pattern of growth as a differentiated endocrine gland. Here, we analyze the onset of differentiation, the contribution of lineages, and the mode of growth of this gland. The expression of genes involved in hormone production and the establishment of epithelial polarity show that differentiation into a first thyroid follicle takes place early during embryonic development. Thyroid follicular tissue then grows along the pharyngeal midline, initially independently of thyroid stimulating hormone. Lineage analysis reveals that thyroid follicle cells are exclusively recruited from the pharyngeal endoderm. The ultimobranchial bodies that merge with the thyroid in mammals form separate glands in zebrafish as visualized by calcitonin precursor gene expression. Mosaic analysis suggests that the first thyroid follicle differentiating at 55 hours postfertilization corresponds later to the most anterior follicle and that new follicles are added caudally. Developmental Dynamics 235:1872–1883, 2006. © 2006 Wiley-Liss, Inc.


During vertebrate development, the endoderm gives rise to the intestinal tube with its appendages and derivatives such as pancreas, liver, and thyroid. During final differentiation, however, all these “endodermal” organs also contain tissues derived from other germ layers as well. Such contributions include parts of the peripheral nervous system and blood vessels, originating from the ectoderm and the mesoderm, respectively. Moreover, in particular in the vertebrate head, it has become increasingly clear that separating structures according to their origin from a single germ layer is questionable. The vertebrate skull for instance is composed of both neural crest and mesodermal derivatives, with both lineages mixing in bones of the neurocranium (Creuzet et al.,2005; Kuratani,2005).

The thyroid gland in humans and mice is composed of two cell types showing endocrine activity, follicle cells, and C-cells (De Felice and Di Lauro,2004). Follicle cells produce the thyroid hormones T4 (thyroxin) and T3, both generated by iodination of tyrosin residues in the common precursor protein thyroglobulin. T4 and T3 are required for multiple aspects in development and metabolism. C-cells produce the peptide hormone calcitonin-related polypeptide alpha (Calca), also known as calcitonin. Calca is thought to be involved in calcium regulation, however, its detailed role in humans is unclear (Inzerillo et al.,2002).

In many mammals such as humans and mice, the thyroid gland develops from two types of primordia (De Felice and Di Lauro,2004). One primordium, the thyroid diverticulum, evaginates from the midline of the pharyngeal floor, on the level between the first and second pharyngeal pouch. This diverticulum relocates and reaches a position deep in the cervical mesenchyme. During relocation, it fuses with a second type of primordia, the ultimobranchial bodies. These bodies occur as a bilateral pair of epithelial bodies that can be found earlier associated with the most posterior pair of branchial pouches (Pearse and Carvalheira,1967). After fusion of ultimobranchial bodies and diverticulum, the gland's cells differentiate into follicles and C-cells (Kusakabe et al.,2005; Fagman et al.,2006). It is widely believed that the ultimobranchial bodies give rise to C-cells and the diverticulum to thyroid follicle cells. However, due to the late time point of differentiation into both cell types after fusion of the primordia and the lack of lineage analyses, it remains unclear if such a strict separation into lineages exists (Manley and Capecchi,1998; Kusakabe et al.,2005 and references herein).

In “lower” vertebrates such as fish, amphibia, and birds, thyroid follicle cells and C-cells are found in separate organs. Thyroid follicle cells form the thyroid gland and do not fuse with the calcitonin producing ultimobranchial bodies that can be found as distinct glands elsewhere in the body (Le Douarin et al.,1974; Le Lievre and Le Douarin,1975). Based on morphological observations, the midline diverticulum gives rise to the follicular thyroid gland, so that the thyroid is traditionally perceived as an “endodermal” organ. Lineage analysis in chick–quail chimeras has shown that the ultimobranchial bodies derive from the neural crest and, therefore, should be regarded as of ectodermal origin (Le Douarin et al.,1974; Le Lievre and Le Douarin,1975).

In teleost fish, the thyroid gland differs from that in other organisms. Here, follicle tissue is loosely dispersed along the ventral midline of the pharyngeal mesenchyme, thus, is not encapsulated by connective tissue (Raine and Leatherland,2000; Raine et al.,2001; Wendl et al.,2002). In zebrafish development, immunostaining of thyroid hormone reveals a first follicle at around 60 hours postfertilization (hpf), when the embryo hatches from the egg shell. An increasing number of further follicles is generated throughout larval life (Wendl et al.,2002; Elsalini et al.,2003). Initial molecular studies suggest that the mechanisms of zebrafish thyroid development are generally comparable to human thyroid development (Wendl et al.,2002), so that the zebrafish will be useful as a model organism in further studies that aim to understand the molecular basis of vertebrate thyroid development.

In the present study, we address open questions about the origin and the mode of differentiation and growth of the zebrafish thyroid. Our expression analysis of genes involved in physiological function shows which cells differentiate into hormone producing cells. Integrating gene expression, morphological aspects, and antibody stainings provides information how the gland differentiates in zebrafish. Based on this knowledge, we then investigate to which degree the zebrafish thyroid is derived from the endoderm. We further cloned a zebrafish calcitonin precursor gene that allows us to determine the presence of C-cells. Finally, mosaic analysis provides first insights into the way follicles grow in zebrafish larvae. Taken together, these approaches present an anatomical picture of teleost thyroid development that provides a base for further molecular and genetic analysis.


Expression of Differentiation Markers Reveals Growth of the Differentiated Thyroid in Zebrafish Larvae Along the Ventral Pharyngeal Midline

To date, expression of differentiation markers for the thyroid have not been analyzed in zebrafish. To identify the cells that possess thyroid hormone producing properties, we analyzed expression of the genes encoding the sodium/iodide symporter (solute carrier family 5 member 5, slc5a5), thyroglobulin (tg), and cathepsin b (ctsb). The sodium/iodide symporter, usually abbreviated as Na(+)/I(−) symporter, NIS, is required for iodide uptake from the blood into the thyroid follicle cells, where iodination of thyroglobulin takes place. This reaction is dependent on the pendrin anion transporter at the apical membrane and the enzyme thyroperoxidase. Iodinated thyroglobulin (here called TG-T4) can either be stored in the lumen of thyroid follicles or is mobilized by the follicle cells. Mobilization involves solubilization and proteolytic cleavage of TG-T4 by proteases, including cathepsin b (Brix et al.,1996; Friedrichs et al.,2003). The resulting T4 is released into the blood stream as thyroid hormone (Fig. 1A; for an overview of thyroid hormone synthesis, see Nussey and Whitehead,2001).

Figure 1.

Differentiation markers start to be expressed in the thyroid primordium during embryonic development in zebrafish. A: Simplified sketch depicting the physiological activity of a vertebrate thyroid follicle cell (blue, nucleus light blue). The red arrow indicates that Ctsb is also required for human thyroglobulin (TG) degradation after endocytosis. Note that the gene slc5a5 encodes the sodium/iodide symporter NIS. For further details, see text. B–K: Expression of thyroid markers in zebrafish embryos. Whole-mounts (B–G) and sections (H–K); lateral views (B,C,F,G) and ventral views (D,E), anterior always to the left. The stage is shown bottom right, the probe/staining bottom left. B–I: Wild-type embryos; J,K: cross-sections of noitu29 homozygous embryos. Arrows point to the thyroid primordium, asterisks mark pharyngeal mesenchyme, and arrowheads expression of ctsb in the yolk syncytial layer. hy, hypothalamus; J, iodide; TPO, thyroperoxidase.

The tg expression in zebrafish is observed first exclusively in the thyroid primordium at approximately 32 hpf (Fig. 1B), with slc5a5 expression initiating slightly later at around 40 hpf (Fig. 1C). As the thyroid primordium subsequently grows, expression of both genes becomes visible as an irregular domain that extends along the midline of the ventral pharyngeal area (Fig. 1D,E). ctsb is widely expressed in embryonic and extraembryonic tissues throughout the stages tested (two-cell stage to 5 dpf; starting with maternal expression that was confirmed by reverse transcriptase-polymerase chain reaction [RT-PCR], Fig. 1F and data not shown). From 35 to 40 hpf, a domain of strong ctsb expression is detectable in the area where the thyroid primordium is expected to develop (Fig. 1G), in addition to widespread low-level expression throughout the pharyngeal area. This domain of ctsb expression appears more diffuse than that of other thyroid markers such as nk2.1a (Fig. 1H,I). In no isthmus (noitu29) mutant embryos lacking a thyroid primordium at 55 hpf (Wendl et al.,2002), the ctsb domain is completely absent, whereas low-level expression in pharyngeal tissue is unaffected (Fig. 1J,K). Therefore, the domain of strong ctsb expression predicts the location of primordial differentiation. Taken together, tg and slc5a5 are valuable markers for studying thyroid differentiation, whereas ctsb is less suitable due to its additional expression in other tissues.

To visualize differentiation of thyroid follicles, we use here an antibody against human thyroglobulin (TG). A first signal can be observed as a single small follicle at around 55 hpf. (Fig. 2A). During larval development, more follicles arise, aligning at the pharyngeal midline (Fig. 2B–D). This pattern of immunostaining during follicular development is identical to that observed using an antibody against the thyroid hormone T4 (thyroxin; Elsalini and Rohr2003). Combining cytoplasmic mRNA detection of thyroid developmental or differentiation genes and TG immunostaining shows that TG localization is confined to apical and some lateral membranes of the follicle cells and to the follicle lumen (Fig. 2E–G). This is also the case for T4 immunostaining and can be seen even in whole-mounts using Nomarski optics (Fig. 2H). Thus, TG and T4 immunostaining allows the visualization of the generation of epithelial polarity and follicular differentiation.

Figure 2.

Thyroglobulin (TG) immunostaining labels the apical and some lateral membranes of thyroid follicle cells. Ventral views (A–D,H–J), anterior always to the left, and cross-sections (E–G). Stage and probe/staining are indicated as in Figure 1. A–D: TG immunostaining labels thyroid follicles from approximately 55 hours postfertilization (hpf). The ceratohyal cartilage (the ventral portion of the hyoid arch) and the heart (out of focus) are highlighted by dotted lines. E–G: In situ hybridization combined with TG immunostaining. At 55 hpf, immunostaining labels clearly only the lumen of a first follicle when combined with nk2.1a (E). Later, the follicle has grown (E and F show same magnification). F: ctsb is not exclusively expressed in thyroid follicle cells, but also in subepidermal cells (arrowheads) and in pharyngeal mesenchyme. G: slc5a5 and tg are exclusively expressed in the cytoplasm of follicle cells (tg staining is identical to slc5a5, not shown). H: Close-up of T4 immunostaining, showing follicle cells (arrows) around the follicle lumen in Nomarski optics, same magnification as in G. I,J: In lia mutants, a row of distinct follicles is visualized with T4 immunostaining. In the strongly affected pharynx, most cartilaginous structures are missing. The dotted line in I demarcates the ceratohyal cartilage that is still present in this mutant (Herzog et al.,2004a). Note that the overall length of the pharyngeal area is reduced. In a wild-type sibling, the ceratohyal cartilage is marked for comparison (J, dotted line). h, heart; ch, ceratohyal.

As the TG immunostaining pattern resembles that of T4 immunostaining, we asked whether both antibodies detect the same epitope. This dual identification could be possible, because the epitopes recognized by these antibodies in zebrafish are unknown. We have shown previously that T4 immunostaining is sensitive to treatment of zebrafish larvae with goitrogens such as phenylthiourea (PTU), methimazole, or potassium perchlorate (KClO4; Elsalini and Rohr,2003). In contrast, TG immunostaining is not affected by treatment of larvae with any of these chemicals even at high concentrations (Table 1). Thiocarbamides such as PTU and methimazole act by inhibiting thyroperoxidase, the enzyme required for iodination of thyroglobulin (Nussey and Whitehead,2001). KClO4 acts by competing with iodine in the process of its uptake from the blood at the sodium iodine symporter (Nussey and Whitehead,2001). That both variants of goitrogens affect T4 immunostaining, but not TG immunostaining, shows that different epitopes are recognized in each case and that TG immunostaining occurs independent of thyroglobulin iodination.

Table 1. Treatment With Goitrogens Does Not Affect TG Immunostaininga
GoitrogenT4 immunostainingTG immunostaining
  • a

    T4 and TG immunostaining in thyroid follicles of zebrafish larvae (120 hpf) after continuous goitrogen treatment. TG, thyroglobulin; PTU, phenylthiourea.

PTU 0.001%++
PTU 0.01%+
PTU 0.6%+
KClO4 0.025%+
KClO4 0.25%+
Methimazole 0.3 mM++
Methimazole 3 mM+

Early Larval Growth of Thyroid Follicular Tissue and Initial Thyroid Hormone Production Is Tsh Independent

Thyroid stimulating hormone (Tsh) is regarded as a major growth factor for the differentiated thyroid in adult humans (Nussey and Whitehead,2001). Being part of the endocrine hypothalamus–hypophysis axis, Tsh is produced by thyrotropes, a subset of cells of the adenohypophysis. To investigate whether the growth of the differentiated thyroid tissue in zebrafish is dependent on Tsh, we investigated the thyroid in a zebrafish mutant that lacks thyrotropes.

In liat24149 mutants, the FGF3 signal required for early specification of the adenohypophysis is disrupted, leading to early loss of thyrotrope progenitors (Herzog et al.,2004a). In this mutant, expression of the tsh precursor gene is completely lost due to the lack of thyrotropes (Herzog et al.,2004a,b). At 120 hpf, a row of follicles is detectable with T4 immunostaining in homozygous liat24149 mutant larvae. In contrast to in wild-type siblings, the row of follicles is shorter, but nevertheless distinct and nearly normally sized follicles form a row along the anterior–posterior axis at the pharyngeal midline (Fig. 2I,J). Craniofacial defects leading to a shorter pharynx in liat24149 mutants (Herzog et al.,2004a) might be responsible for the reduced length of the strand of thyroid tissue. Nevertheless, absence of thyrotropes does not affect thyroid hormone production or growth of follicles and both processes, therefore, are independent of Tsh, at least at early larval stages.

Taken together, differentiation markers show that the thyroid gland in zebrafish grows as a strand of polarized, hormone producing follicular tissue during larval life. This observation was the basis for investigating whether this tissue is completely derived from the small endodermal midline primordium and establishing the directionality of follicle tissue growth.

Follicular Thyroid Is Completely Derived From the Endodermal Midline Primordium in Zebrafish

Continuous expression of the developmental marker nk2.1a suggests that the follicular thyroid originates from the midline primordium (Rohr and Concha,2000). The question remains whether additional cells contribute later to the growing thyroid. We addressed this question using a molecular approach that forces blastomeres to become endoderm and combined this approach with lineage analysis.

Endoderm development depends on the Nodal signalling pathway (Feldman et al.,1998). Correspondingly, the endoderm is affected in zebrafish mutants with compromised Nodal signalling. Concomitant with an absent or reduced endoderm in one-eyed pinhead (oep) and cyclops (cyc) mutants, respectively, the thyroid primordium is absent or reduced in size (Elsalini et al.,2003). This finding supports the notion that the midline primordium is derived from the endoderm, but structures apart from the endoderm are also affected in oep and cyc mutants (Schier et al.,1996,1997; Sampath et al.,1998). Abnormalities in ectodermal and mesodermal structures could indirectly account for defects in thyroid development. Furthermore, a smaller primordium in cycm294 mutant embryos at 26 hpf appears to recover during later growth, and at 5 dpf, a normal number of thyroid follicles forms (Elsalini et al.,2003). This compensation could be achieved by increased growth of the primordium or by increased recruitment of cells from surrounding tissues. Thus, from mutant analysis alone, we can neither conclude that the thyroid is completely derived from the endoderm nor that it is specifically in the endoderm where Nodal signalling is required for thyroid development.

To test whether the thyroid derives from Nodal-dependent endoderm, we used ectopic Tar* expression. Tar* is an activated form of the type I transforming growth factorβ related receptor TARAM-A (Renucci et al.,1996) that is normally expressed in a region encompassing endodermal progenitors and that activates the Nodal/activin pathway required for endoderm development. Injection of Tar* mRNA causes zebrafish blastomeres to become endoderm in a cell-autonomous manner (Peyrieras et al.,1998; David and Rosa,2001). Injection of Tar* mRNA into one of the outer blastomeres at the eight-cell stage causes this cell to take over a central role in endoderm formation, such that the progeny of the injected blastomere will form the endodermal derivatives. Co-injection of Tar* mRNA together with lineage tracers such as gfp mRNA, biotin–dextran, or rhodamine–dextran allows the development of the intestinal tube with its appendages to be monitored (Peyrieras et al.,1998).

In embryos where Tar* and gfp mRNA have been injected into one of the outer blastomeres, green fluorescent protein (GFP) visualizes pharyngeal endoderm development in vivo. At 24 hpf, the pharyngeal epithelium is the only structure stained in the head, together with staining in the hatching gland (Fig. 3A,B). From around 35 hpf, the thyroid diverticulum can be seen to evaginate from the pharyngeal endoderm (Fig. 3C). Later, GFP expression becomes increasingly weaker, but biotin–dextran detection results in strong staining of the pharyngeal epithelium and the thyroid in the visceral mesenchyme (Fig. 3D,E). These data confirm that the midline primordium of the thyroid derives from Nodal-dependent endoderm.

Figure 3.

Tar* expressing cells reveal the origin of the thyroid from the pharyngeal endoderm. Stage and probe/staining are indicated as in Figure 1. A–C: Green fluorescent protein (GFP) expression in embryos where Tar* and gfp mRNA was co-injected into one of the outer blastomeres at the eight-cell stage. In A and B, arrowheads indicate GFP expression in the hatching gland, arrows in the pharyngeal endoderm. In C, the arrow points to the thyroid diverticulum that has just lost contact to the pharyngeal endoderm, the arrowhead points to the mouth. A dorsal view, anterior to the top, B; C lateral views, anterior to the left. D: Biotin–dextran detection of a 7-day-old larva after Tar*+biotin–dextran+gfp injection into one of eight outer blastomeres. Biotin–dextran staining is visible in the pharyngeal epithelium (arrowheads) and in a long row of thyroid follicles (arrows). E: Double staining combining tg expresssion and biotin detection in a larva derived from an embryo injected with Tar* as in A–D. F–I: Consecutive staining of tg (F,H) and biotin–dextran (G,I) in an embryo (F,G) and a larva (H, I), both derived from embryos injected with Tar* as in A–E. tg labels the thyroid (blue) as tissue that stains completely overlappingly brown with biotin–dextran detection and, therefore, shows that the whole thyroid is endoderm-derived. lj, lower jaw; m, mouth; p, pharynx.

To address the question whether cells are incorporated into the differentiated thyroid from other tissues, we combined thyroid marker expression with lineage tracer detection. We found that all thyroid cells expressing the differentiation marker tg also contain the lineage tracer (Fig. 3F–I) and, hence, derive from the endoderm. We, therefore, conclude that in zebrafish, the midline primordium from the pharyngeal epithelium gives rise to the complete follicular thyroid at larval stages.

Grafted Tar* Expressing Cells Frequently End Up in the Thyroid Primordium

Fate mapping of the zebrafish blastula has shown that the pharyngeal endoderm originates from the dorsal part of the blastula, from the marginal blastomeres close to the shield (Warga and Nusslein-Volhard,1999). According to this study, further posterior endodermal derivatives originate from more ventral parts of the blastula. To test whether the thyroid originates from the area of pharyngeal precursor cells at blastula stages, we attempted to specifically fate map the thyroid using cell labeling techniques. Labeled cells were found in the pharyngeal epithelium (in 11 of 47 embryos) but in no case in the thyroid gland, presumably because only a very small portion of the pharynx forms the thyroid primordium.

To increase the likelihood of cells ending up in the thyroid, we grafted cells at dome and sphere stages from Tar* mRNA and rhodamine–dextran/biotin–dextran injected donor embryos into the marginal zone of uninjected host embryos. Such grafted cells always contribute to the endoderm (David and Rosa,2001). Shortly after grafting at shield stage, rhodamine–dextran labeled cells were still found at the margin, the part of the late blastula that is known to contribute to the endoderm (Warga and Nusslein-Volhard,1999). The position of the grafted cells shortly after transplantation was measured in vivo in relation to the shield (Fig. 4A). This was possible, as in nearly all cases the grafted cells still formed a coherent group at this stage (n = 63). Two cases where the group of labeled cells has fallen apart were omitted from analysis. Later at 55 hpf, we monitored cells in the thyroid primordium by biotin–dextran detection. In contrast to our first cell labeling approach without Tar* mRNA, cells were now repeatedly found in the thyroid primordium (in 16 of 63 embryos, Fig. 4B,C). A comparison of the thyroid hits with the position of the grafted cells at shield stage reveals that cells very close to the shield (zone I) have the highest likelihood to contribute to the thyroid. Other cells in the dorsal hemisphere (zones II and III) could also contribute to the thyroid, whereas cells in the ventral hemisphere (zones IV to VI) did not (Table 2). Statistically, the position of cells at shield stage is not random with respect to their contribution to the thyroid (χ2 test, 18.577; P = 0.001; Fisher's exact test, 16.458; P = 0.002). Although we cannot completely exclude an influence of Tar* on the behavior of the cells with respect to their distribution along the dorsoventral axis at shield stage (see the Discussion section), the similarity of the distribution of thyroid precursors and pharynx precursors (Warga and Nusslein-Volhard,1999) suggests that both precursor cell types share the lineage at blastula stage.

Figure 4.

Mosaic nature of Tar* expressing, labeled cells allows fate mapping and gives insight into the mode of thyroid follicle growth. Stage and probe/staining are indicated as in Figure 1. A: Schematic drawing showing the experimental procedure. See text for details. At shield stage, the embryo is subdivided into six zones (here shown for the right side only), and the position of labeled cells is noted. An example of a resulting embryo at 55 hours postfertilization (hfp) is shown in B and C. B,C: Consecutive staining of nk2.1a (B) and then biotin–dextran (C). In this embryo, at least two cells (arrows) can be found in the thyroid primordium as visualized by nk2.1a. In addition, part of the pharyngeal epithelium is stained (arrowheads). D–G: Selection of specimens where a mosaic of grafted cells (arrows) can be found in thyroid follicles. One-color staining that combines TG immunoreaction in the lumen of the thyroid follicles with cytoplasmic biotin–dextran detection in grafted cells. It is typical that cells around the anterior follicle show stronger biotin–dextran labeling than cells around more posterior follicles. D and E show specimens with many grafted cells. F and G show specimens with few grafted cells. In G, one follicle can be found that lacks grafted cells completely (asterisk), and a single labeled cell can be seen expanding posteriorly (arrowhead). Note that in the depicted specimens, as in most cases, the anterior follicle is larger than more posterior ones.

Table 2. Thyroid Precursor Cells are Found Close to the Shield at Blastula Stagesa
Position of grafted cells at shield stage (zone according to distance from shield, see Fig. 4A)Number of chimeric embryos analyzed with cells in respective areaCorresponding number of thyroid hits at 55 hpf (total 16 hits of 63 grafting experiments)
  • a

    Distribution of the progeny of Tar* mRNA injected grafted cells. hpf, hours postfertilization.


Mosaic Nature of Lineage Tracer in the Thyroid Suggests Anterior-to-Posterior Growth

We used the transplantation approach of Tar* mRNA injected cells to gain further insight into the way the thyroid follicles grow during larval life. For instance, it would be interesting to find out whether the first forming follicle can be predicted to correspond later to the most anterior, the most posterior, or an intermediate follicle. All embryos where biotin–dextran labeled cells ended up in the thyroid primordium at 55 hpf showed a mosaic of donor and host cells in the primordium. We allowed host embryos to develop further and analyzed the distribution of donor cells in the row of follicles (Fig. 4D–G).

The mosaic nature of grafted cells varied from follicles that were nearly completely surrounded by biotin–dextran labeled cells (Fig. 4D,E), to only three or four biotin–dextran labeled cells in the whole thyroid (Fig. 4F,G). Due to variable size and shape of follicle cells and their three-dimensional distribution around the lumina, it was not possible to clearly distinguish and count single cells. However, biotin–dextran appears to accumulate in vesicles at the apical membrane toward the lumen of the follicles, so that biotin–dextran labeled cells are clearly visible in their extent surrounding the follicle.

Comparing 30 larvae where grafted cells ended up in the thyroid, the following observations could be made. First, biotin–dextran labeling is usually strongest in the most anterior follicle (Fig. 4D–G). Second, the most anterior follicle contains labeled cells in all 30 larvae, whereas more posterior follicles are occasionally without grafted cells (Fig. 4G, Table 3). Third, even if clear counting of biotin–dextran labeled cells is not possible, it is apparent that the first follicle always contains more, or a similar number, but not less labeled cells than the more posterior follicles. Therefore, there is a bias of cells contributing to the most anterior follicle.

Table 3. In Larvae with a Contribution of Biotin–Dextran Labelled Cells to the Thyroid, Such Cells Are Always Found in the Most Anterior Follicle, but Not Necessarily in More Posterior Onesa
Follicle in anterior to posterior orderNumber of larvae with corresponding follicleNumber of larvae where the corresponding follicle was without labeled cells
  • a

    Distribution of follicles that lack labeled cells in Tar*+biotin–dextran mosaic larvae. Data taken from 30 larvae where biotin–dextran labeled cells ended up in the thyroid. Larvae had two to five follicles at fixation (for example, the embryo in Fig. 4G lacks biotin–dextran labeled cells in the second of three follicles).

1. Follicle300
2. Follicle302
3. Follicle295
4. Follicle252
5. Follicle213

It is clear from our Tar* mRNA experiments that all cells giving rise to the whole strand of thyroid tissue at larval stages derive from the small embryonic primordium visible at 36 to 40 hpf (Fig. 3C,F,G), or from the tissue surrounding the first differentiating follicle at 55 hpf (Fig. 4C), respectively. The bias of cells contributing to the most anterior follicle at 120 hpf suggests that this part of the thyroid corresponds to the remnant of the first differentiating follicle. In such a model, a proportion of cells would remain at the position of the first follicle during thyroid growth, with grafted cells maintaining a relatively high biotin content. The other cells would proliferate and add tissue caudally, due to their mosaicism, occasionally giving rise to follicles devoid of labeled cells. This model is further supported by the observation that in some specimens, grafted cells extend from the most posterior follicle further caudally (Fig. 4G, arrowhead). However, the notion of a strict anterior-to-posterior growth of the thyroid in zebrafish is preliminary and awaits further analysis (see the Discussion section).

calca Expression Visualizes the Ultimobranchial Bodies in Zebrafish Larvae

Ultimobranchial bodies can adopt various shapes in different teleost species (Sasayama et al.,2001), so we analyzed where the ultimobranchial bodies are located in zebrafish. To obtain a differentiation marker for the ultimobranchial bodies, we cloned cDNAs of the zebrafish calcitonin gene starting from predicted open reading frame information in the incomplete zebrafish genome (ENSEMBL: In mammals as well as the teleost species Fugu rubripes, alternative splicing results in two mRNAs: calca and cgrp. The transcript calca is translated into the preprotein for the calcitonin-related polypeptide alpha (Calca). cgrp is translated into the preprotein for the calcitonin gene-related peptide (Cgrp; Clark et al.,2002). We amplified both splice variants, calca and cgrp, using different primers (Fig. 5). Sequence comparison confirmed that they are orthologuous to the calca and cgrp transcripts in Fugu rubripes (data not shown). The 5′ part of both transcripts, comprising exon I and exon II according to ENSEMBL prediction, is identical in calca and cgrp, whereas an alternative third exon is recruited (Fig. 5) as in Fugu (Clark et al.,2002). The probes generated from the different splice variants show an identical expression pattern (data not shown). In this study, we used the calca probe for expression studies.

Figure 5.

Cloning of the cDNAs encoding Calca and Cgrp preproteins. The middle boxes represent exons that were predicted by the ENSEMBL database. Arrows indicate approximate positions of the primers “Ak2” (forward, used for both amplification of both transcripts), and “calca” and “cgrp” (both reverse). The two amplified transcripts calca and cgrp are indicated according to their composition of predicted exons.

calca is expressed from around 60 hpf in a bilateral set of cells close to the atrial side of the heart (Fig. 6A), a position that is characteristic for the ultimobranchial bodies in other vertebrate species (Sasayama,1995). The calca expression can also be detected in some parts of the brain. Combination with TG immunostaining shows that the ultimobranchial bodies are far away from the thyroid originating from the midline primordium (Fig. 6B). In a few wild-type embryos, four groups of calca expressing cells can be found instead of two groups (Fig. 6C). Sectioning reveals that the ultimobranchial bodies are positioned adjacent to the muscles surrounding the gut (Fig. 6D). Based on this information, we were able to locate the ultimobranchial bodies in the adult fish, where they lie as two groups of follicles at the transverse septum close to the sinus venosus (Fig. 6E). Although the morphology of the ultimobranchial glands varies in teleosts (Sasayama,1995, Sasayama et al., 2001), the glandular shape and the position clearly identify this structure. Thus, the ultimobranchial bodies adopt a position far away from the follicular thyroid in zebrafish and can be regarded as a separate organ.

Figure 6.

calca is expressed in the ultimobranchial bodies of zebrafish. A–D: calca expression (blue) in zebrafish larvae. Stages, bottom right. A: Expression in the brain and in the ultimobranchial bodies (arrow), lateral view, anterior to the left. B: Double staining with thyroglobulin (TG) immunostaining (arrowheads). C: A specimen with four ultimobranchial bodies instead of the usual two as in B. B and C are ventral views, anterior to the top. D: Cross-section on the level of the ultimobranchial bodies, note the muscle layer (arrows) around the gut that is still filled with yolk. E: Position of the ultimobranchial bodies in the adult zebrafish. Hematoxylin and eosin staining, section on the level of the sinus venosus of the heart. Inset: A close up of the ultimobranchial bodies visualizes their glandular appearance and organization as follicular epithelia. g, gut; h, heart; m, muscle layer of the gut.


The starting point of this study was the analysis of thyroid differentiation markers. To date, only developmental genes have been studied in zebrafish, and differentiation of follicles has been shown using T4 antibody staining. However, T4 is the hormone itself that is released into circulation, and only by means of presumed cross-reaction with intermediate stages of hormone synthesis serves as an indirect marker for thyroid follicles in fish (Raine and Leatherland,2000; Raine et al.,2001; Wendl et al.,2002). Our analysis reveals that thyroid differentiation occurs even earlier than assumed in previous studies. tg expression is observed even before the thyroid primordium has evaginated from the pharyngeal epithelium. Double stainings of cytoplasmic mRNA and membrane associated protein reveals that epithelial polarity is already established at 55 hpf, similarly very early in development. Due to the localized TG immunostaining, it can be assumed that, at this stage, the thyroid is fully differentiated (Fig. 7).

Figure 7.

Overview of the timing of differentiation and mode of growth of the thyroid in zebrafish. The upper sketches (I–VI) show different stages of thyroid development along the time scale of zebrafish embryonic (up to approximately 70 hpf) and larval development. Gray corresponds to the thyroid primordium and follicular tissue. The arrows indicate the onset of developmental gene expression (nk2.1, hhex, pax2.1) and differentiation markers used in this study. I: Expression of markers in a field of the pharyngeal endoderm before a pharynx has formed. II, III: Evagination of the midline diverticulum. IV: At around 55 hours postfertilization (hpf), the first follicle differentiates. V, VI: During growth of follicles, further tissue is added caudally. “1” indicates the most anterior and first follicle. e, endoderm; h, heart; ch, ceratohyal; p, pharynx.

It is striking how much earlier tg mRNA is detectable than TG immunostaining. The reason for this finding might be a slow accumulation of protein up to a detectable threshold. However, because tg expression is very strong and immunostaining becomes detectable only after 20 hr, it is more likely that differentiation of unpolarized primordial cells into polarized follicle cells is required for proper protein production or for the transport of thyroglobulin to the apical membrane. It should also be noted that the antibody was generated against human TG and, therefore, that the immunostaining relies on cross-reactivity against an unknown epitope in zebrafish. Nevertheless, a comparison of TG and T4 immunostaining in our goitrogen experiments suggests that zebrafish TG is recognized by this antibody in its noniodinated form. The sole immunostaining at membranes can be explained either by very low levels of cytoplasmic TG in thyroid follicle cells or by the failure of the antibody to detect cytoplasmic TG. Taken together, the early expression of all thyroid differentiation markers in zebrafish highlights the difference from mice, where these genes are expressed much later in development after substantial growth of the primordial gland has occurred (Elsalini et al.,2003).

To our knowledge, this study represents the first fate mapping for the thyroid in any species. Our data show that the thyroid derives completely from the endoderm. Mosaic analysis in Tar* grafted host embryos further suggests that the thyroid precursors are part of the pharyngeal precursors. This suggestion would be predicted from the evagination of the midline diverticulum from the pharyngeal epithelium. This evagination is not only visible using developmental marker gene expression but can also be followed in vivo using Tar* mRNA coinjected with a lineage tracer. Therefore, in zebrafish, the thyroid can be considered as an endoderm-derived organ in the strict sense. The concept of an organ derived from a germ layer-specific cell lineage proves correct in this case.

That the follicular thyroid is completely derived from its endodermal diverticulum suggests that, in cycm294 mutants, the initially smaller primordium reaches normal size at later stages by increased growth (Elsalini et al.,2003). Thyroid growth in adult vertebrates is regulated by Tsh from the pituitary (Nussey and Whitehead,2001), and it is possible that such an endocrine feedback mechanism might account for the growth compensation in this mutant. Alternatively, a nearly normal thyroid in lia mutant larvae reveals that, under wild-type conditions, Tsh does not play a major role in thyroid growth at these early larval stages. As judged from Tsh-receptor deficient mice that have a normal thyroid gland at birth, neither maternal nor zygotic Tsh is required for fetal thyroid growth in this species (Postiglione et al.,2002). Of interest, both early thyroid function (T4 immunostaining) and growth (a normal row of follicles) in zebrafish larvae is Tsh independent and, in this respect, similar to fetal thyroid growth in mice. However, as no zebrafish mutant is available that lacks the Tsh-receptor, it remains unclear if it is basal activity of this receptor that is responsible for normal thyroid follicle growth and positive T4 immunostaining in thyrotrope deficient zebrafish mutants.

Our mosaic analysis further suggests that endodermal precursor cells at shield stage can only contribute to the thyroid if they involute in the dorsal hemisphere. In this experiment, we cannot rule out an influence of Tar* mRNA on the position of the precursor cells. However, the thyroid itself always appears normal in Tar* injected embryos; thus, this manipulation does not interfere with thyroid development and, therefore, is unlikely to change precursor positions. Instead, the variable position of the precursors at the margin rather suggests that the exact position of the thyroid is not yet defined at shield stage. It will be interesting to investigate whether signals from the shield influence pharyngeal endoderm patterning and/or thyroid development.

The differentiated thyroid grows concomitantly with jaw elongation. The bias of grafted cells contributing to the most anterior follicle in our mosaic experiment suggests this tissue being the first to form. It is tempting to assume that further tissue is added caudally. However, the distribution of grafted cells in our mosaic experiment does not exclude the possibility that posterior follicles subsequently also grow cranially, for instance filling gaps that arise in the row of follicles during further increase in size of the head of the fish. In the adult fish, a very loose row of follicles exists and most follicles lack any direct cell–cell contact (Wendl et al.,2002). The exact mode of growth of the strand of thyroid tissue remains to be elucidated by time lapse movies. It will be interesting to investigate how growth is regulated and synchronized in such loosely distributed teleost thyroid tissue and whether specific thyroid stem cells are involved in the growth process.

In comparative morphological studies throughout the vertebrate clade, the thyroid has always been associated with the anterior pharynx on the level around the hyoid arch. Whereas in other vertebrates the thyroid has adopted a globular or lobed shape, in teleost fish it extends along the midline of the pharynx. We here conclude that it is the anterior part of the zebrafish thyroid that represents the ancestral position of the organ.

Due to a lack of developmental markers for the ultimobranchial bodies, the early development of these glands in zebrafish remains obscure; however, there is at present no indication that they interact with or contribute to thyroid development in zebrafish. In embryos co-injected with Tar* mRNA and lineage tracer, we never observed labeled cells in the region where calca expressing cells are found (data not shown), indicating that the Tar*-dependent endodermal lineage does not contribute to the ultimobranchial bodies. That the ultimobranchial bodies arise from the neural crest has been shown in lineage experiments in chick–quail chimeras (Le Douarin et al.,1974; Le Lievre and Le Douarin,1975) and it can be assumed that the same is true for zebrafish. For mammals, it has been discussed that neural crest-derived cells might also contribute to thyroid follicles. For instance, in Hoxa3 deficient mice lacking the neural crest contribution to the thyroid gland, the remaining follicular thyroid is smaller than expected if only C-cells are missing (Manley and Capecchi,1995,1998). In dogs, the ultimobranchial bodies form C-cell clusters without complete incorporation into the thyroid lobes. Nevertheless, TG expressing follicles also differentiate within the C-cell clusters (Kameda and Ikeda,1980). Murine ultimobranchial bodies contain a second poorly characterized cell type in addition to C-cells that might act as a stem cell pool capable of differentiating into both C-cells and thyroid follicle cells (Kusakabe et al.,2005). Whereas in mammals the contribution of cell lineages to thyroid cell types remains controversial, we can exclude the possibility that neural crest cells for zebrafish contribute to follicle cells at least during the early larval stages we analyzed. In the light of the massive growth of endoderm-derived cells in the early differentiated gland, it appears unlikely that such contributions occur later.

In mice, both ultimobranchial bodies and midline diverticulum express the transcription factor Nkx2.1 (Mansouri et al.,1998; Meunier et al.,2003; Fagman et al.,2006). The role of Nkx2.1 in the development of the diverticulum and physiological activity of thyroid follicle cells is well understood (reviewed in De Felice and Di Lauro,2004). In the ultimobranchial bodies, Nkx2.1 is required for C-cell differentiation and for fusion with the thyroid lobes (Kusakabe et al.,2005). The zebrafish orthologue nk2.1a is expressed in the thyroid diverticulum and has a conserved function with respect to thyroid follicle development (Elsalini et al.,2003). However, we were neither able to detect nk2.1a nor its zebrafish paralogue nk2.1b in any structure that might correspond to developing ultimobranchial bodies (data not shown). We cannot rule out low or transient expression in zebrafish but our data currently suggest that Nkx2.1 genes assumed a role in ultimobranchial body development later in evolution, probably concomitant with the fusion of, or other interactions between, both glands.



Zebrafish work was carried out according to standard procedures and staging in hours post fertilization refers to development at 28.5 to 29°C. The term “larva” is used in the text for fry that have hatched from the chorion and generally refers to an age older than 72 hr. Homozygous noitu29 mutant embryos (Brand et al.,1996) were identified by the missing midbrain hindbrain boundary. Homozygous liat24149 mutants were identified by their craniofacial defects at 120 hpf (Herzog et al.,2004a).

Preparation of Specimens

In situ hybridization on zebrafish was carried out according to standard procedures. Whole-mount immunohistochemistry with antibodies against the thyroid hormone T4 (polyclonal rabbit anti T4, ICN Biochemicals; Wendl et al.,2002) or thyroglobulin (polyclonal rabbit anti human thyroglobulin, Dako) in zebrafish larvae was performed as described elsewhere (Elsalini and Rohr,2003). H&E staining on paraffin sections was carried out as described (Wendl et al.,2002). Biotin–dextran was detected by using the ABC kit (Vectastain) as described for T4 immunostainings (Elsalini and Rohr,2003). For one-color TG immunostaining plus detection of biotin–dextran injected cells, both follicles and grafted cells stain positive due to the use of a biotinylated secondary antibody against the primary rabbit anti thyroglobulin antibody.

Cloning of Differentiation Markers

Both the ctsb and the tg gene were deposited in the ZFIN Expression database (Sprague et al.,2001) as clones zgc:65809 and cb717, respectively. Clones were obtained from RZPD (ctsb; Berlin, Germany) and from the University of Oregon (tg; Eugene, OR). BLAST comparison to other species and analysis of the alignment of the deduced amino acid sequences with related proteins from other species (data not shown) revealed the clone zgc:65809 (GenBank accession no. BC056688) to be a predicted full length ctsb gene and confirmed the identity of clone cb717 (GenBank accession nos. CB891044 and CB891045) as a fragment of the zebrafish thyroglobulin gene.

Sequences of the putative open reading frames of slc5a5 and the calcitonin transcripts calca and cgrp were identified in the incomplete zebrafish genome available at Transcripts were cloned from larval first strand cDNA preparations. Total RNA was prepared using Trizol; for reverse transcription, we used the SuperScript first strand synthesis system (both Invitrogen). Primer sequences for the slc5a5 gene are as follows: slc5a5-1 (forward), ATGGCTATGGACTCTGACAGACCAC; slc5a5-2 (reverse), CCTCCATAGTGACAGCAGCC. For amplification of calca and cgrp, primers were designed based on ENSEMBL predicted open reading frame and intron–exon structure information. Primer sequences are as follows: A2 (forward, used for both transcript variants), ATGGTTATGTTGAAGATCTCCGCTT; calca (reverse), AACGCCTGGGAGCCCACATTAGTGG; cgrp (reverse), TATCCATCACTCCCTTCCAGGCTGC (for primer positions, see Fig. 5). The identity of the obtained fragments as zebrafish orthologues of slc5a5, calca, and cgrp was confirmed by BLAST analysis and comparison of the deduced amino acid sequence with other species (data not shown). Sequences are deposited in GenBank under the accession numbers DQ402039 (slc5a5), DQ406589 (calca), and DQ406590 (cgrp).

Embryonic Manipulation

Tar* mRNA was injected into outer blastomeres at the eight-cell stage as described (Peyrieras et al.,1998). For grafting of Tar* mRNA injected cells, we followed essentially the procedures described (David and Rosa,2001), with the difference that Tar* mRNA was injected into donor embryos at the one-cell stage, not at the two- to four-cell stage. As lineage tracers, we injected gfp mRNA (100 ng/μl), tetramethylrhodamine–dextran (10,000 molecular weight, 5 mg/ml), and biotin–dextran (10,000 molecular weight, 5 mg/ml; both Invitrogen). Treatment with goitrogens was carried out as described in (Elsalini and Rohr,2003).


We thank our colleagues from the zebrafish groups in Köln for discussion and support, in particular Julia von Gartzen for excellent technical assistance. We also thank Gabriela Nica and Matthias Hammerschmidt (MPI, Freiburg) for kindly providing mutant larvae. Additionally, we thank Alexandra Franzmann for statistical analysis and John Chandler for corrections on the manuscript. This work is funded by the SFB572 of the Deutsche Forschungsgemeinschaft. O.A.E. is supported by the Libyan Ministry of Education/Garyounis University.