Regeneration of thyroid function by transplantation of differentiated pluripotent stem cells

Nkx2-1⁺ thyroid cells are specified from ESC-derived endodermal precursors upon stimulation with exogenous BMP4 and FGF2

We previously developed a directed differentiation protocol to produce ESC-derived definitive endodermal cells competent to form thyroid and lung epithelial lineages (Longmire et al., 2012). We employed an ESC line (Nkx2-1^GFP) carrying a knock-in GFP reporter cDNA targeted to the locus encoding the homeodomain-containing transcription factor Nkx2-1, an essential regulator of thyroid and lung development. Within endoderm, Nkx2-1 is only expressed in lung or thyroid epithelia and is the first known protein induced upon endodermal specification to lung or thyroid fates. Nkx2-1⁺ cells induced from ESC-derived endoderm using our published (Longmire et al., 2012) 6 factor cocktail (WFKBE+2; Figure. 1A) expressed thyroid epithelial-specific genes, such as thyroglobulin (Tg). However, these cells did not display full thyroid maturation as the genes required for iodine metabolism and thyroid hormone biosynthesis were not robustly expressed (ie the sodium iodine symporter (Nis) and thyroid peroxidase (Tpo)); (Figure S1A and data not shown).

Microarray analysis of the global mRNA expression profiles of these ESC-derived Nkx2-1^GFP+ endodermal progenitors (Figure 1A and Tables S1 and S2), revealed multiple active signaling pathways including BMP, Wnt, EGF, and FGF signaling. Hence, to identify the essential factors for thyroid cell development we performed sequential withdrawal of each individual factor from the WFKBE+2 cocktail. We found BMP4 and FGF2 to be indispensible for efficient Nkx2-1⁺ endodermal induction. An induction media composed of BMP4, FGF2 and Wnt3a led to the maximum efficiency of Nkx2-1⁺ cell specification by Day 14 and sorted Nkx2-1^GFP+ cells at this time point, either with or without Wnt3a, were competent to subsequently express thyroid differentiation markers, Tg and Tsh-r (thyroid stimulating hormone receptor), and maturation markers, Nis and Tpo (Figure 1B and Figure S1A). Therefore, BMP4 and FGF2 are sufficient to specify endodermal cells in vitro towards Nkx2-1⁺ endodermal progenitor cells with thyroid potential.

Nkx2-1 haploinsufficiency affects the differentiation of thyroid epithelial cells, but not the efficiency of Nkx2-1⁺ endodermal lineage specification

Because the Nkx2-1^GFP knock-in reporter in our engineered ESCs replaced one Nkx2-1 allele potentially creating haploinsufficiency (Figure 1C), we sought to test the effects of Nkx2-1 gene dosage on thyroid differentiation in vitro. We compared the Nkx2-1^GFP ESC line to a non-haploinsufficient Nkx2-1^mCherry ESC line where the mCherry fluorophore is knocked-in 3’ to the coding sequence (Bilodeau et al., 2014) (Figure 1C). The two lines were differentiated in parallel and Nkx2-1^GFP+ or Nkx2-1^mCherry+ cells were sorted to purity on Day 14 (Figure 1D), at which time the expression of Nkx2-1 mRNA was approximately 50% lower in the haploinsufficient Nkx2-1^GFP+ cells than in the non-haploinsufficient Nkx2-1^mCherry+ cells (Figure 1E). The expression of both early (Pax8) as well as all mature thyroid markers analyzed (Tg, Tsh-r, Nis, and Tpo) in the outgrowth of sorted Nkx2-1⁺ cells was diminished in Nkx2-1 haploinsufficient cells (Figure 1E). This was confirmed when the Nkx2-1^GFP line was compared to its parental syngeneic, non-haploinsufficient clone (Figure S1C). In contrast we did not detect any difference in the rates of lineage specification of Nkx2-1⁺ endodermal cells on Day 14, comparing haploinsufficient to normal ESC lines. The percentages and cell numbers of Nkx2-1⁺ cells specified by Day 14 was similar in the Nkx2-1^GFP line and its parental clone (Figure S1B) as well as in the Nkx2-1^mCherry line (data not shown). Thus, Nkx2-1 haploinsufficiency does not affect specification of Nkx2-1⁺ endodermal cells, but does affect subsequent maturation of thyroid epithelial cells. This is supported by observations in mice and humans who have Nkx2-1 haploinsufficiency and hypothyroidism (Krude et al., 2002).

The expression of Pax8 distinguishes a subpopulation within Nkx2-1⁺ endodermal cells with thyroid potential

Having established a protocol able to produce putative thyroid follicular cells, next we sought to determine whether thyroid and lung lineage specification occurred in distinct endodermal Nkx2-1⁺ precursors as is thought to occur in vivo. The Pax8 transcription factor is co-expressed with Nkx2-1 only in thyroid epithelial cells, beginning at mouse E8.5 and human E20 in the foregut endoderm within the region of the prospective thyroid primordium (Fagman and Nilsson, 2010; Trueba et al., 2005). In contrast Pax8 is not expressed at any developmental stage in lung epithelial cells. Hence, we developed a bifluorescent reporter system to track thyroid and non-thyroid fates in Nkx2-1⁺ cells by breeding mice with the GFP reporter targeted to the Nkx2-1 locus (Nkx2-1^GFP (Longmire et al., 2012), the Cre recombinase targeted to the Pax8 locus (Pax8^Cre (Bouchard et al., 2004)), and a conditional Rosa26 tdTomato reporter (lox-stop-lox-tdTomato). As expected, these mice (hereafter Nkx2-1^GFP;Pax8^{tdTomato trace}) co-expressed both GFP and tdTomato reporters only in the thyroid epithelium, whereas GFP without tdTomato was expressed throughout the developing respiratory epithelium (Figures 2A, 2B). As previously reported, GFP was also expressed within developing ectoderm in the ventral forebrain (Longmire et al., 2012), and tdTomato without GFP was expressed in tissues known to express Pax8 including the kidney, hindbrain, midbrain, and inner ear (Figure S2A and as published (Bouchard et al., 2004)). We derived iPSC clones carrying Nkx2-1^GFP;Pax8^{tdTomato trace} reporters by reprogramming mouse embryonic fibroblasts (MEFs) isolated from the bifluorescent mouse (Figures 2C, 2D and Figure S2B) and differentiated 6 lines in vitro using our Wnt3a, BMP4, and FGF2 Nkx2-1 induction protocol (Figures 2E, 2F). By Day 14 all clones exhibited induction of the Nkx2-1^GFP reporter with kinetics and efficiency similar to our previous report (Longmire et al., 2012) and Pax8^{tdTomato trace+} cells represented 4.9% (+/−4.7) of all Nkx2-1⁺ cells. Based on GFP and tdTomato expression, we sorted four distinguishable populations on Day 14 (Figure 2F) and further differentiated each population until Day 25 (Figure 2E and Figure S2C). Analysis of cells at the time of sorting (Day 14) as well as until the full 25 days of differentiation of 3 clones revealed that only the Day14 sorted GFP⁺ tdTomato⁺ population gave rise to cells expressing the constellation of 4 transcription factors (Parlato et al., 2004) that together uniquely identifies thyroid epithelia (Nkx2-1, Pax8, Foxe1, and Hhex) and the differentiation marker Tg (Figure 2E). In contrast, this population was depleted of cells competent to express differentiated lung marker genes, Sftpc and Scgb1a1 (data not shown). In separate experiments using ESC lines that lack Pax8 reporters, we used FACS analyses of intracellular protein to verify that Pax8 protein is expressed in approximately 5% of Nkx2-1⁺ ESC-derived cells (data not shown). We conclude that, as in the developing embryo, the thyroid lineage is specified in a distinct Nkx2-1⁺ endodermal population.

BMP and FGF signaling are necessary and sufficient for thyroid lineage specification from mouse and Xenopus anterior foregut endoderm or ESC-derived definitive endoderm

Having demonstrated that Pax8⁺/Nkx2-1⁺ thyroid progenitors were induced from PSC-derived endoderm by addition of Wnt3a, BMP4, and FGF2 (WB+2), we next sought to determine whether each pathway was necessary for thyroid lineage specification. In keeping with prior publications suggesting Wnt is not required for thyroid specification from endoderm in vivo (Goss et al., 2009), we found Nkx2-1^GFP+ cells (Figure S1A) or Nkx2-1^mCherry+ cells (Figure S3A) derived with BMP4 and FGF2 induction media (Day 6–14) in the absence of Wnt3a displayed undiminished subsequent expression of thyroid markers after outgrowth. While either BMP4 or FGF2 alone induced low numbers of Nkx2-1⁺ cells with little thyroid competence (Figures 3A, 3B), combinatorial use of BMP4 and FGF2 induced the highest percentage of thyroid competent Nkx2-1⁺ progenitors, as evidenced by subsequent robust expression of all early and mature thyroid markers including Pax8, Tg, Tsh-r, Tpo, or Nis (Figure 3B and Figure S3A). Consistent with these results, withdrawal of Wnt3a and/or addition of Wnt inhibitors Dkk1 or XAV-939 had no significant impact on the efficiency of lineage specification of Nkx2-1^GFP+;Pax8^{tdTomato trace+} iPSC-derived cells by Day 14, whereas there was a significant decrease in the percent yield of GFP+/tdTomato+ cells when either BMP or FGF2 was removed from the Nkx2-1 induction media (data not shown).

To determine which downstream BMP and FGF signaling cascades are required for induction of thyroid fate, we supplemented the BMP4 + FGF2 media with specific inhibitors of either SMAD-dependent BMP signaling (Dorsomorphin), MEK1/2-dependent FGF or BMP signaling (PD98059; hereafter PD), PI3 kinase (PI3K)-dependent FGF signaling (LY294002; hereafter LY), or p38-MAPK-dependent BMP signaling (SB203580). We found reduced numbers and percentages of Nkx2-1⁺ cells in conditions supplemented with Dorsomophin or LY (Figure 3C). Furthermore, when the rare Nkx2-1⁺ cells that were induced in the presence of these inhibitors were sorted to purity on Day 14 and cultured further to allow completion of the protocol (without inhibitors), the progeny of these Nkx2-1⁺ cells showed reduced capacity to express mature thyroid markers, such as Tg and Tpo (Figure 3D).

Taken together, our findings in the ESC/iPSC in vitro model system suggested that canonical Wnt signaling may be dispensable for thyroid lineage specification while combinatorial SMAD-dependent BMP signaling together with FGF signaling are required. To test this hypothesis in primary cells of the developing embryo, we employed both the murine and Xenopus models of embryonic foregut organogenesis (Figure 4 and S4). First, in developing Xenopus embryos we found evidence in support of active BMP and FGF signaling in the region of the thyroid primordium by immunostaining embryos for nuclear phospho-SMAD1/5/8, nuclear phospho-ERK1/2, and phospho-AKT ser473 proteins during early thyroid specification (Figure S4A). These phosphorylated signaling effectors were present in the foregut epithelium prior to and during the evagination of Nkx2-1⁺ cells of the developing thyroid anlage as well as in the surrounding mesenchyme (Figure S4A; stages NF20 and NF33).

Next, to assess whether FGF and BMP signaling are required for thyroid specification in vivo, we incubated developing mouse as well as Xenopus embryos in inhibitors of BMP or FGF signaling (Figure 4 and S4). Developing mouse foreguts were isolated by dissection at 6–8 ss (~E8.0) prior to detectable Nkx2-1 expression in the thyroid field and incubated for 2 or 3 days with the BMP inhibitor, DMH-1. DMH-1 caused a marked reduction in phosphorylation of SMAD1/5 (Figure 4A, right panel Western blot) and blocked induction of both Nkx2-1 and Pax8 in the region of the mouse endodermal thyroid primordium (Figure 4A, left panel). Similarly, we incubated developing Xenopus embryos in inhibitors of BMP signaling (DMH-1 or an injected dominant negative BMPR) or FGF signaling (SU5402, PD161570, or an injected dominant negative FGFR), starting just after gastrulation (stage NF13) until stage NF20 (6–7 somite stage; ss). The inhibitors were then removed and embryos allowed to develop until stage NF34 (36 ss); a time by which thyroid and lung lineages are normally specified (Shifley et al., 2012). In situ hybridization for markers of pharyngeal endoderm and thyroid lineage specification foxe1, nkx2-1, pax2, and hhex, revealed that inhibition of either BMP or FGF signaling abrogated induction of all four thyroid markers in the vast majority of Xenopus embryos (Figure 4B and Figure S4B). Furthermore, inhibition of PI3K-AKT with LY also prevented thyroid specification (Figure S4B). Immunostaining of pSmad1/5/8, pERK and pAKT confirmed the efficacy of the inhibitors (Figure S4C). In contrast, when Xenopus embryos were incubated from stage NF13 through NF35 in inhibitors of canonical Wnt signaling (XAV939), RA signaling (BMS493), or VEGF signaling (KRN633), we observed normal nkx2-1 induction in the thyroid primordium (Figure S4B), indicating that Wnt, RA, and VEGF signaling at these developmental stages are dispensable for thyroid specification.

To assess the stage-dependence of these signaling requirements we varied the timing of BMP and FGF loss of function during foregut endoderm development. We observed that early inhibition of BMP or FGF signaling beginning at stage NF13 (analogous to mouse E7.5) blocked induction of nkx2-1 (Figure S4B), whereas inhibition beginning later (at stage NF20; Figure S4D) did not, suggesting that the requirement for BMP and FGF signaling in thyroid lineage specification is restricted to a narrow developmental window between stages NF13-20.

Since our mouse ESC model had predicted that FGF2 and BMP4 were sufficient to induce thyroid lineage specification, we next asked whether exogenous FGF2 and BMP4 were sufficient to induce thyroid development in Xenopus foregut endoderm (Figure 4C). Foregut explants were micro-dissected at stage NF15, prior to thyroid specification and the mesoderm was removed. The foregut endoderm explants were then cultured until stage NF35 either without growth factors or with a combination of FGF2 and BMP4. In situ hybridization revealed that only explants incubated with FGF2 and BMP4 expressed nkx2-1, pax2, and hhex (Figure 4C). We did not detect expression of sftpc in explants from sibling embryos (data not shown) suggesting that the nkx2-1 expression was thyroid and not respiratory epithelium. Taken together these results from Xenopus and mouse embryo models extended our observations made in differentiating mouse ESCs and iPSCs, confirming that FGF and BMP signaling are evolutionarily conserved pathways required for the specification of thyroid fate from developing endoderm both in vitro and in vivo (Figure 4D).

Thyroid stimulating hormone and 3D culture promotes ESC-derived thyroid follicular maturation and organoid formation

Having interrogated the signals required for the induction of thyroid fate, next we focused on augmenting the maturation state of the thyroid epithelial progenitors generated from PSCs, using the Nkx2-1^mCherry ESCs. In contrast to lineage specification and early development, the expression of thyroid genes necessary for iodine metabolism, Nis and Tpo, is associated with later gland maturation (Figure 5A) and has been shown in vivo to require TSH receptor activation (Postiglione et al., 2002). Hence we tested the effect of TSH at various developmental stages of our in vitro protocol, either on induction of Nkx2-1⁺ progenitors (Day 9–12) or during outgrowth of sorted Nkx2-1⁺ cells (Day 14–22). We found addition of TSH prior to lineage specification (Day 9–12) had no detectable effect on the efficiency of thyroid lineage specification or the competence of thyroid cells to subsequently differentiate (data not shown). In contrast, addition of TSH after lineage specification (Day 14–22) resulted in increased Nis, Tsh-r, and Tpo expression (Figure S5A), but no significant change in the expression of lineage markers Nkx2-1 and Pax8 (data not shown). An additional 3 days of culture maturation in the presence of TSH was accompanied by further augmentation of Nis and Tsh-r expression (Figure S5A).

We previously published augmented thyroid gene expression in ESC-derived Nkx2-1⁺ cells when using a thyroid media containing IGF-1, insulin, transferrin, and selenium (Longmire et al., 2012). Therefore, we included these additives, together with TSH to our prior base media of FGF2 + FGF10. This “thyroid outgrowth media” was used for replating the sorted Nkx2-1⁺ cells on Day 12 (Figure S5B). For the final maturation stages after expansion of sorted cells in “thyroid outgrowth media” we sought to further optimize expression of Nis and Tpo. Since we had observed induction of these markers in response to either DCI+K (Figure 1E) or in response to 1–100 mU/ml of TSH (data not shown), we tested the effect of combinations of each of these factors (Figure 5C; Day 22–26). We found that withdrawal of either cAMP or dexamethasone from the DCI+K cocktail adversely affected maturation (Figure S5C), however substitution of TSH for cAMP in the DCI+K media retained maturation and resulted in the most robust expression of Nis (Figure S5C). Thus, the combination of dexamethasone and TSH was included in the final 4 days of maturation (Day 26–30), resulting in the thyroid directed differentiation protocol summarized in Figure S5B and achieving the derivation from ESCs of a population of cells expressing thyroid marker transcripts at levels equal to (Nkx2-1, Hhex, Foxe1, Tpo) or within 10-fold of (Pax8, Tg, Tshr, Nis) post-natal murine thyroid control tissue (Figure 5B).

To facilitate thyroid follicular structural formation during outgrowth of ESC-derived Nkx2-1⁺ cells, based on prior reports of Matrigel as a favorable thyroid culture substrate (Martin et al., 1993), we tested the effect of 3D culture in Matrigel (Days 12–30; Figures 5C, 5D). Compared to 2D culture, we found culture of purified Nkx2-1^mCherry+ cells in 3D Matrigel conditions resulted in higher expression of thyroid genes by qPCR (Nkx2-1, Pax8, Tg, Tsh-r, Tpo, and Nis; Figure 5C), increased epithelial gene expression (increased E-Cadherin and EpCam expression and reduced mesenchymal markers, Snail1, Twist, and Col1a1; Figure 5D), and triggered the formation of follicular-like clusters of cells (Figure 5E). By Day 30 we observed 130+/−17 follicular clusters emerging from each 65,500 Nkx2-1^mCherry+ cells plated on Day 12 in a well of a 24-well plate. By Day 30 in culture, 50% of the progeny of the Nkx2-1^mCherry+ sorted cells expressed Pax8 protein by FACS, 46% expressed thyroglobulin, and 96% continued to express Nkx2-1 by either microscopy or FACS for the mCherry reporter (data not shown). Most importantly, phenotyping of the resulting clusters or organoids, revealed them to have the 3D structure and molecular phenotype typical of thyroid follicles, including formation of organized monolayered Nkx2-1⁺ Pax8⁺ epithelia surrounding a lumen filled with thyroglobulin (Figures 5F, 5G). These organoids could be further passaged and maintained in these culture conditions with stable or increased expression of Nkx2-1, Pax8, Hhex, Foxe1, Tg, Nis, Tshr, and Tpo until at least Day 52 (Figure S5D). In addition, after incubation in iodinated media for 48 hours the organoids displayed the functional capacity to organify iodine and produce small amounts of T4 hormone bound to TG in vitro at different time points during the differentiation protocol, beginning on Day 30 (Figure S5E).

In vivo function of purified ESC-derived thyroid follicular organoids following transplantation into hypothyroid mouse recipients

The defining functional characteristic of thyroid follicular epithelial cells is their capacity to synthesize and secrete thyroid hormones in vivo. Hence, we sought to test whether ESC-derived Nkx2-1^mCherry+ thyroid progenitors sorted on Day 12 of differentiation and further matured into organoids until Day 30 could function in vivo after transplantation, and moreover whether these cells could rescue hypothyroid mice. Syngeneic mouse recipients underwent radioiodine ablation of their native thyroid tissue after 8 days of a low iodine diet in order to induce severe hypothyroidism, evident as significantly reduced circulating plasma T4 and T3 levels and elevated circulating TSH levels (Figures 6A, 6B). Four weeks later, we prepared 1 positive control mouse group (n=5) that had not undergone radioiodine ablation, 1 negative control group that received sham surgery (hereafter “sham”; n=4), and 2 experimental recipient groups that received either undifferentiated (Day 0 ESCs; hereafter “D0-ESCs” n=5) or differentiated ESCs that had been sorted on Day 12 based on Nkx2-1^mCherry+ expression and further differentiated until Day 30 in our complete 3D protocol (hereafter D30-ESCs; n=10). Following ESC transplantation with 2.5–3 × 10⁶ cells beneath the left kidney capsule of each mouse recipient we followed all recipient groups for at least 8 weeks post transplantation monitoring circulating T4, T3, and TSH levels, and we employed both MRI and SPECT/CT imaging to screen for in vivo graft growth and function (Figure 6 and Figure S6).

Starting 2–4 weeks post transplantation, recipients of D30-ESCs displayed augmented circulating T4 and T3 levels that continued to increase over time until they approached normal levels by 8 weeks after transplantation (Figure 6B and Figure S6A, S6C). Reconstitution of T4 levels was accompanied by amelioration of the elevated plasma TSH levels (Figures 6B, 6C and Figure S6B), further emphasizing that physiologic rescue was occurring. In a repeat experiment some recipients followed for longer periods were observed to display circulating T4 reconstitution at normal levels for at least 29 weeks after transplantation (data not shown). In contrast, we observed no reconstitution of circulating thyroid hormone levels in either sham transplanted controls or recipients of D0-ESCs, excluding the possibility that either residual undifferentiated cells or re-growth of native thyroid tissue was responsible for the return of circulating T4 or the dampening of the TSH response.

To further evaluate the function of the transplanted cells we incorporated imaging with sodium pertechtenate (Tc99^M) and SPECT/CT co-registered with MRI imaging. Importantly the uptake of Tc99^M into thyroid tissue is mediated by Nis allowing its visualization to be a surrogate for iodine uptake (Zuckier et al., 2004). In recipients of D30-ESCs we observed Tc99^M uptake occurring in the region of the kidney capsule grafts (Figures 6D, 6E and Figure S6E). We also observed low levels of Tc99^M uptake in the large masses that derived from kidney capsule transplantation of D0-ESCs (Figure S6D), but in no D0-ESC recipient was this signal accompanied by detectable return of circulating T4 levels. As expected, Tc99^M uptake was not detected in the anatomical region of the native thyroid tissue of any ablated recipient (Figure 6E and Figure S6E), further suggesting that return of thyroid function in D30-ESC recipients was not due to regrowth of native thyroid tissue.

To confirm the presence of engrafted D30-ESCs in vivo we evaluated the histology of the renal lesions that had been visualized by MRI and SPECT (Figures 6F-6I). As expected, recipients of D0-ESCs developed large teratomas (Figure S6D) without evidence of organized follicles and only rare, if any, cells expressed nuclear Nkx2-1 protein (data not shown). In contrast, no teratomas were observed in recipients of D30-ESCs; instead these recipients displayed persistent grafted cells beneath their kidney capsules organized into many follicular-like structures per high power field that were morphologically consistent with thyroid epithelial follicles (Figure 6F). These structures consisted of mCherry⁺ epithelial cells expressing nuclear Nkx2-1 and Pax8 proteins by immunostaining, surrounding central lumens filled with colloid-like material that contained Tg and T4 (Figures 6F-6H). A subset of the grafted cells expressed the proliferation marker, PCNA (Figure 6I). We did not detect expression of pro-Sftpc in any grafted cell suggesting that Nkx2-1 expression resulted from thyroid and not respiratory epithelium (Figure S6F).

To assess whether the engrafted D30-ESCs functioned in a regulated fashion we exposed additional D30-ESC transplanted mice, negative sham controls, as well as positive normal controls (n=5 per group) to a high dose of exogenous T3 expected to fully suppress endogenous TSH and hence T4 production. Indeed in all groups administration of T3 resulted in TSH suppression and abrogated T4 secretion (data not shown). To further interrogate TSH responsiveness (Figure S6G), we then injected each mouse with exogenous bovine TSH and three hours later observed an increase in circulating T4 in positive control mice as well as in recipients of D30-ESCs, but not in sham controls, consistent with the regulation of engrafted cells by TSH.

These results indicate that ESC-derived thyroid follicular epithelial cells prepared by directed differentiation in serum-free, feeder-free conditions exhibit in vivo functional potential, including the capacity to rescue hypothyroid mice who otherwise lack native functional thyroid tissue.

Derivation of human thyroid progenitors from ESCs and hypothyroid patient-specific iPSCs

Finally, we sought to adapt our protocol to direct the differentiation of human ESCs and iPSCs into thyroid progenitors. Having observed that thyroid specification was not affected by Nkx2-1 haploinsufficiency in mouse PSCs (Figure 1 and Figure S1), we hypothesized that BMP4 and FGF2 might also generate human thyroid progenitors in normal as well as in haploinsufficient lines. Hence we obtained dermal fibroblasts from three hypothyroid children previously diagnosed with the Brain-Lung-Thyroid syndrome arising from three different respective NKX2-1 coding sequence mutations (Figure S7A, S7B) that were predicted to cause NKX2-1 haploinsufficiency (Hamvas et al., 2013). Using the STEMCCA reprogramming system (Somers et al., 2010) we generated iPSC lines from each child’s fibroblasts (hereafter T1, T3, and T4; Figures 7A, 7B, supp. Figure 7C–7E). Next, we derived definitive endoderm from each iPSC line as well as from control human ESCs (RUES2) and control iPSCs (BU3 and iPS17; Figure 7C and supplemental methods). We employed dual inhibition of TGFβ and BMP signaling as originally described by Green et al. (Green et al., 2011) in order to mimic anterior foregut patterning of each ESCs/iPSCs-derived endodermal population and then tested the capacity of varying concentrations of BMP4 and FGF2 to specify putative thyroid cells. Using the control RUES2 line, we varied the doses of BMP4 and FGF2 to identify a dosing combination that resulted in induction of clusters of NKX2-1+/PAX8+ co-expressing cells (Figure 7D, and Figure S7F). This same dose induced NKX2-1+/PAX8+ cell clusters from human iPSC lines BU3 and iPS17, as well as from all three patient-specific iPSC lines (Figure 7E left panel and Figure S7G). BMP4 or FGF2 alone induced rare, if any, NKX2-1+/PAX8+ co-expressing cells in the T1, T3, T4, and RUES2 lines (data not shown). Furthermore, in BU3, iPS17, and T3 iPSC-derived endodermal cells subjected in parallel to published lung inducing media (Huang et al., 2014) that contains BMP4 (with Chir, FGF10, FGF7, and RA; “CFKBRA”) but lacks FGF2, we observed only induction of NKX2-1+ cells without co-expression of PAX8, suggesting these cells were respiratory rather than thyroid (Figure 7E right panel). As in our mouse ESC/iPSC experiments, NKX2-1+/PAX8+, putative thyroid progenitors induced after 10 days of exposure to BMP4 and FGF2 expressed low to undetectable levels of TPO and TSH-R transcripts (data not shown). However, after further culture in the same thyroid differentiation and maturation conditions used for mouse ESCs these thyroid differentiation markers were upregulated and NKX2-1, PAX8, and TG expression was maintained for at least 42 days in culture (Figure 7F). As expected, without the use of a human NKX2-1 reporter for purification of human thyroid cells from these heterogeneous endodermal cultures, these thyroid markers were detected at significantly lower levels than adult human thyroid control tissue biopsies (data not shown). Taken together these findings support the capacity of combinatorial BMP and FGF signaling to induce thyroid lineage specification from developing endoderm across species, from Xenopus to mice to humans, and enable the generation of patient-specific thyroid progenitors from individuals with genetic or congenital hypothyroidism.

Animal maintenance

All studies involving mice carrying GFP and tdTomato reporters were approved by the Institutional Animal Care and Use Committee of Boston University School of Medicine. All murine transplantation experiments were approved by Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee.

Mouse ESC and iPSC lines

Derivation of the Nkx2-1^GFP and Nkx2-1^mCherry knock-in reporter ESC lines by homologous recombination was previously published and is further detailed in the supplement (Bilodeau et al., 2014; Longmire et al., 2012). The Nkx2-1^GFP;Pax8^{tdTomato trace} iPSC reporter line was generated by transduction of MEFs with the dox-inducible Tet-STEMCCA reprogramming lentiviral vector. All ESCs and iPSCs were maintained on fibroblast feeder layers using serum-containing media and LIF (ESGRO Chemicon).

Mouse ESC/iPSC directed differentiation

Cell differentiation was performed in complete serum-free differentiation media (cSFDM, see Extended Experimental Procedures) supplemented with the indicated factors. Briefly, definitive endoderm was induced in embryoid body suspension cultures with 50 ng/ml Activin A (R&D) as previously published (Longmire et al., 2012). Next, 100 ng/ml mNoggin (R&D) and 10 µM SB431542 (Sigma) was applied for 24 hours to generate anterior foregut endoderm. Expression of Nkx2-1 was induced with specification media containing 100 ng/ml mWnt3a, 10 ng/ml mKGF, 10 ng/ml hFGF10, 10 ng/ml mBMP4, 20 ng/ml hEGF, and 250 ng/ml mFGF2 (all from R&D). After Nkx2-1 induction, cells were sorted on Day 12–14 by flow cytometry based on expression of GFP, mCherry, or tdTomato reporters and then replated for either 2D or 3D culture outgrowth as indicated. For 2D outgrowth plated cells were grown in gelatin-coated dishes for 8 more days in cSFDM with 250 ng/ml mFGF2, 100 ng/ml hFGF10 and 100 ng/ml Heparin Sodium Salt. On Day 22, the media was switched to DCI+K maturation media (see Extended Experimental Procedures). For 3D outgrowth cells were plated in pure growth factor reduced Matrigel drops (Corning) in cSFDM supplemented with 250 ng/ml mFGF2, 100 ng/ml hFGF10, 50 ng/ml mIGF-1, 25 ng/ml hEGF (all from R&D), 100 ng/ml Heparin Sodium Salt, 10 µg/ml insulin (Sigma), and 1 mU/ml bTSH (Los Angeles Biomedical Research Institute). On Day 26 media was switched to thyroid maturation media (Ham’s F12 supplemented with 15 mM HEPES, 0.8 mM CaCl₂, 100 ng/ml Heparin Sodium Salt, 0.25% BSA, 50 ng/ml mIGF-1, 5 µg/ml insulin, 5 µg/ml ITS, 25 ng/ml hEGF, 50 nM Dexamethasone, and 100 mU/ml bTSH until the harvest day indicated.

Human iPSC derivation

Procurement of all human specimens was approved by the Institutional Review Boards of University of Colorado and Boston University. Three hypothyroid individuals diagnosed with Brain-Lung-Thyroid Syndrome arising from mutations in NKX2-1 were identified and their dermal fibroblasts were reprogrammed with the STEMCCA reprogramming system to generate iPSC lines T1, T3, and T4 (Somers et al., 2010). Control human PSC lines included: BU3 (peripheral blood-derived normal iPSC line; from the CReM of Boston University and Boston Medical Center), iPS17 (fibroblast-derived iPSC) (Crane et al., 2015); (gift of Dr. Brian Davis, University of Texas Health Sciences), and RUES2 human ESCs (gift of Dr. Ali Brivanlou, Rockefeller University). Differentiation of human ESC/iPSC into thyroid cells is detailed in the supplement.

Statistical Analysis

Data are presented as sample means and standard deviations (SD) or standard errors of the mean (SEM), as indicated in the text and in each figure legend. Sample numbers are also indicated in each case. Differences between groups were analyzed by unpaired two-tailed Student’s t-test, 1-way analysis of variance (ANOVA), or 2-way ANOVA with Tukey's multiple-comparison post hoc test as stated in the text or figure legends; p<0.05 was used to indicate significant differences between groups.