According to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (31), a systematic search was independently performed by 2 investigators (Chengxiang Shan and Lei Li) using electronic databases, including Medline, EBSCO, and Embase, to identify articles published before May 1st, 2022. The listing keywords were thyroid, regeneration, and/or organoid. The “related articles” function was used to broaden the search. All studies were checked for additional relevant material when appropriate. Only studies published in English were considered.

Abstracts or full-text articles were initially screened and then selected or rejected by the two reviewers (Huajin Zeng and Qixuan Sheng) based on the inclusion and exclusion criteria described below. The inclusion criteria were as follows: (1) Original experiments referring to regenerative medicine of the thyroid gland; (2) Research subjects could be derived from all species (human, mouse, rat, etc.) and different sources. The exclusion criteria were as follows: (1) Review articles; (2) Nonthyroid experiments; (3) Cultures or organoid models referring to thyroid cancer; (4) Books, posters, conference abstracts, or presentations; (5) Studies with the same research subjects published repeatedly by different journals. The two reviewers (Qiang Wang and Wei Zhang) independently extracted details from each eligible study, including first author, year of publication, subject species and source, research model, key experiment process, and significant findings.

Fourteen studies between 1959 and 1978 establishing 2D thyrocyte models in vitro were identified. These experiments were performed with a variety of species, such as human (32, 33), calf/bovine (12, 34–36), sheep/lamb (11, 13–15), rat (37), chick (38), and swine (39, 40). The majority were carried out using 2D thyroid models, which were generated from culturing isolated thyrocytes as a monolayer. All these earlier experiments demonstrated that in vitro isolated thyrocytes maintained the capacity of iodide uptake, organification, and deiodination (11, 40). However, Pulvertaft R J (32) observed that a proportion of thyrocytes transformed and dedifferentiated during long-term culture (>3 weeks) and lost the capacity for iodide concentration and incorporation, which was consistent with the findings by Kalderon A E (15). Moreover, the thyrocytes’ capacity for iodide metabolism could be enhanced if additional TSH was incorporated into the culture medium, which was partially mediated via adenyl cyclase activation (36). However, Tong W (12) demonstrated the opposite result that extra TSH did not affect the capacity of iodide uptake. In addition to TSH, prostaglandin E 1 (PGE1) and dibutyryl cyclic AMP (DBcAMP) also enhanced iodide uptake (36) and organification (40). 7-Oxa-13-prostynoic acid inhibited the effects of TSH and PGE1 upon iodide concentration (36), and high concentrations of iodine inhibited thyrocyte organic iodination (35), as did methimazole (13, 33). It was first demonstrated by Rousset B (40) that isolated thyrocytes maintained the capacity to secrete T4, and TSH (1-60 munits/ml) could stimulate the release of T4 in a dose-related manner in vitro.

As no follicles were observed in the suspension (40) or monolayer culture before, three special experiments at this stage were designed to show that isolated thyrocytes were capable of self-organization and forming functional follicles. It was described that adult or embryonic thyrocytes reaggregated and reconstituted follicles when cultured in a chick chorioallantoic membrane (CAM) or rat anterior eye chamber (AEC) (37, 38). Even in monolayer culture, highly concentrated thyrocytes were rearranged into follicular structures upon TSH stimulation (39).

Twenty-four studies between 1979 and 2003 focusing on thyrocyte-derived folliculogenesis were identified. The most common species involved in these studies were swine (16–18, 41–48), rats (17, 49–56) and humans (17, 19, 52, 57–60). On the one hand, a proportion of experiments revealed that both directly isolated thyroid follicles (17) and newly formed follicles retained the main follicular functional characteristics, such as active iodide transport, thyroglobulin (TG) and T3 production, and TSH responsiveness (57), compared with native thyroid follicles. On the other hand, the majority of experiments focused on exploring crucial factors affecting folliculogenesis based on isolated thyrocytes or thyrocyte cell strains.

First, a high cell density was proven to be a necessity for reconstructing functional follicles in an earlier experiment (39). Its importance was reemphasized here by Fayet G, who revealed that high cell density could be achieved if the plastic medium was treated with poly-L-lysine (41). Second, appropriate cell-matrix interactions in 3D systems have been observed to enhance thyrocyte organization and function (54). When culturing thyrocytes in untreated polystyrene dishes (16), similar to suspension culture, their aggregates formed hollow spheres encompassed by a single layer of cells, similar to follicles. However, such follicular-like spheres were not identical to physiological follicles, as they possessed the opposite polarity, with the apical microvilli pointing toward the culture medium and the basal membrane pointing toward the follicular lumen. These inverted follicles (so-called inside-out follicles), which were also demonstrated by Chambard M (42) and Kitajima K (45), could regain cellular polarity and reorganize normal follicle structures when the thyrocytes were embedded in gelatin (16), collagen gel (42, 46, 57), Matrigel (19), or an alginate bead system (53). Furthermore, if dispersed thyrocytes were cultured between two layers of collagen gel, instead of mixing with the collagen gel, they could form more follicle-like structures with distinct cellular polarity (58). Third, the importance of stimulation by TSH or other specific stimuli or inhibitors should also be highlighted during follicle formation. Upon stimulation with TSH, thyrocytes aggregate and rearrange in 3D structures resembling cross-sections of intact thyroid tissue (16, 41, 43). Adding insulin to the TSH-containing culture medium stimulated the construction of follicles and the recovery of iodine metabolism of thyrocytes. Epidermal growth factor (EGF) alone could induce a progressive migration of thyrocytes radiating from the preexisting (mother) follicles into the collagen lattice and might play a role in the process of multiplication and migration of thyrocytes and formation of new follicles in the thyroid (47). Abundant oxygen was also proven to protect thyroid follicles from death due to findings in the air exposure (AE) culture system (48). Adding methimazole to the alginate bead culture system significantly decreased follicular diameters, which indicated an inhibitory activity in modulating follicular morphology (54). Fourth, the role of the inter- or intracellular signals in promoting folliculogenesis was also established. Thyrocytes cultured in vitro formed widely different sizes of cell colonies, indicating that the intrinsic growth rates among the same cells were heterogeneous. However, these cells did not grow individually or independently from each other; instead, they interacted with each other with synchronization. The morphological change of one single cell might initiate the interaction among neighboring cells within the aggregate, thus leading to folliculogenesis (52). Additionally, when mixing FRTL thyroid cells with dermal or thyroid fibroblasts simultaneously in matrix medium, the integrated thyroid gland equivalent (TGE) forms functional thyroid follicles (49). Nevertheless, connexin 23 (Cx23) overexpression in FRTL cells increased cell-to-cell communication and induced 3D follicular-like structure formation (55). The low-shear culture environment present in simulated microgravity also assisted folliculogenesis. Compared with traditional culture, both human isolated thyrocytes and FRTL-5 thyroid cells aggregated and formed densely arranged multicellular follicles enclosing a central lumen in a rotary cell culture system (56, 60); however, additional supplements with basement membrane extract and keratinocyte growth factor needed to be added simultaneously. It is worth noting that newly formed descendant follicles could also be derived from preexisting mother follicles through three types, namely, the solid nest type, the budding type and the lumen-dividing type, and the frequencies of solid nest, budding and lumen-dividing types of follicle formation were approximately 20%, 40% and 40%, respectively (18).

Fifteen studies between 2003 and 2021 exploring thyrocyte differentiation from SCs were identified. The most common species involved in these studies were mice (9, 21, 28, 61–68) and humans (9, 21, 30, 69–71). The main purpose of these experiments was to elucidate the molecular mechanisms leading to the process of SC-derived thyrocyte specification, and these experiments were arbitrarily separated into two categories according to the source of differentiated thyrocytes, ESCs/iPSCs or ASCs/progenitor cells.

In mouse models, experiments have been conducted in which ESCs became committed to differentiating into thyrocyte-like cells in vitro ( 61, 63, 66), and the inactivation of one allele of the TSH receptor (TSHR) gene in ESCs did not alter the developmental program during differentiation (62). ESCs have the potential to differentiate into the thyrocyte lineage, and some additional prerequisites might still be required to promote follicular morphogenesis. Four thyroid transcription factors (NKX2-1/PAX8/FOXE1/HHEX) were shown to be essential regulators during thyroid organogenesis (72). Coexpression of NKX2-1 and PAX8 in ESCs initiated a change in the fate of cells toward a thyroid follicular cell lineage in mice, and knockdown of NKX2-1, PAX8 and FOXE1 decreased the responses to TSH of thyroid-specific genes (NIS, TSHR, TG, and TPO) to different degrees (71). However, in undifferentiated ESCs, merely coexpressing NKX2-1 and PAX8 appeared to be insufficient for full thyrocyte fate determination. With the removal of leukemia inhibitory factor (LIF) and supplementation with activin A (endoderm induction) plus TSH (stimulus) in the culture medium, NKX2-1+/PAX8+ ESCs were capable of forming embryoid bodies (EBs) first and then transforming to mature thyroid follicular cells with as high as normal expression of thyroid-specific genes (64). This identical differentiation process was also successfully demonstrated in a double-transfected NKX2-1+/PAX8+ human ESC line (Line H9) (70) and NKX2-1+/PAX8+ iPSCs (65). Intriguingly, if the endoderm-induction step (exclusion of activin A) in the above multistage differentiation protocol was omitted, NKX2-1+/PAX8+ ESCs differentiated into nonthyrocyte lineages and mature thyrocytes, suggesting that additional factors specifically restricted the progenitors to the thyroid lineage (68). In addition, bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) signaling pathways were found to be active in NKX2-1+ endodermal progenitors in vitro, and endogenous NKX2-1 expression could also be directly induced by culturing ESCs/iPSCs in medium supplemented with BMP4 and FGF2 (21). Furthermore, transforming growth factor-β (TGF-β) signaling was also demonstrated to be significant during the transition from the immature to mature thyrocyte state, increasing the capacity for iodide uptake and organification (68).

There was sufficient evidence that tissue-specific stem cells resided in certain adult tissues, as was demonstrated in the thyroid. Side population (SP) cells, recognized by the ability to efflux the vital dye Hoechst 33342, are generally considered to have characteristics consistent with stem cells. A cell population with the SP phenotype was proven to exist in both mouse thyroid and human thyroid. SP cells appeared in the interfollicular space but not in follicles. They represented 0.3% to 1.4% of the total population of cells in mouse thyroid (28) and 0.1% in human thyroid (30), which were eliminated by verapamil pretreatment. The positive expression of SCA1, CD34, Nestin, OCT4, and Nanog, as suggested stemness markers, was found in a proportion of putative thyroid SCs (69), while in some cases, emblematical stemness markers were not easily identified. It was known that isolated thyroid stem cells could maintain viability with phenotype during passaging in vitro, but they showed resistance to growth stimulations and were unable to differentiate directly, which was perhaps due to lack of interaction between stem cells and niche cells. One possible option to induce murine thyroid ASC differentiation into functional follicular cells is to generate thyrospheres or thyrospheroids containing ASCs in specific “spheroid medium” primarily supplemented with EGF and bFGF. For human thyrosphere formation, additional WNT and R-Spondin1 were required in the defined thyroid gland medium (9). Once thyrospheres are established and then cultured in the “differentiation medium”, they are specified into mature thyroid cells expressing thyroid-specific genes, producing thyroid hormones (69), and showing TSH-dependent I¹²⁵ uptake (30).

Twenty-two studies between 1984 and 2022 exploring thyroid reconstruction models and thyroid organoid models were identified. The most common species involved in these studies were mouse (9, 21, 25, 63, 65–68, 73, 74), human (9, 19, 21, 26, 52, 59, 60, 69, 74, 75), rat (49, 52, 54, 56, 75, 76) and swine (77). Detailed 3D matrices and medium components for different thyroid cell types were concluded in Table 1 .

As described above, differentiated thyrocyte lineages have been successfully induced from ESCs/iPSCs or ASCs/progenitor cells. Consequently, thyroid reconstruction could originate directly from these SCs (9). In a mouse model, NKX2-1+/PAX8+ ESC-derived thyroid cells organized into 3D follicular structures and maintained thyroid functionalities (63, 66, 68). The culture of NKX2-1+ ESCs/iPSCs in 3D Matrigel conditions triggered the formation of follicular-like clusters of cells (21, 65). Furthermore, in FGF10 EX1 ^mut/EX3 ^mut transgenic mice with severe thyroid hypoplasia, allogeneic normal ECSs could induce orthotopic thyroid tissue regeneration when microinjected into the embryos of FGF10 EX1 ^mut/EX3 ^mut mice via blastocyst complementation (25). Thyroid organoids can also be generated in vitro by culturing isolated adult or fetal thyroid cells in a specific medium supplemented with a necessary stimulus (such as EGF, FGF10, WNT, R-Spondin1, or Noggin) (26, 74). Human ASC-derived thyrospheres seeded in collagen gel in the presence of a “differentiation medium” started to generate functional thyroid follicles (9, 69). In a mouse model, isolated cells formed thyroid organoids when embedded into an organoid culture medium containing Matrigel and various growth factors, and each thyroid organoid appeared to be composed of clonally derived cells and did not simply represent cellular aggregates (67). Interestingly, some authors generated small or microthyroid tissues by directly culturing mature thyrocytes or thyroid cell lines in a 3D medium (19, 49, 52, 54, 56, 60) and proved their physiological functions with enriched critical genes (20, 59). However, their organoids’ capabilities of self-renewal and differentiation did not seem to be fully determined.

Some special protocols for constructing thyroid organoids have also been reported. Arauchi A (76) showed thyroid reconstruction using the cell sheet technique. After transplantation into thyroidectomy-inducible hypothyroid rats, these cell sheets were organized into honeycomb-like structures of thyroid follicles. Moreover, parafollicular cells and numerous functional microvessels containing red blood cells also existed among the follicles, telling the reliable capability of cell sheet technique to get new blood supply. Bulanova E A (73) described thyroid reconstruction from embryonic thyroid tissue-derived spheroids (TS) and embryonic allantoic tissue-derived spheroids (AS) using a multifunctional Fabion 3D bioprinter with the turnstile system. Fusion spheroids with three TS and six AS formed the rudiment of thyroid organoids, and during further culture in collagen gel, the endothelial progenitors in AS responded to the angiogenic factor secreted by TS and then initiated vascularization in bioprinted thyroid. Yang Y (77) accomplished thyroid organoid formation by microencapsulation of porcine thyroid cells by applying a designed 8-nozzle microfluidic device. Upon TSH stimulation, thyroid cells gathered and formed 3D follicles in the inner core of microbeads within 48 hours and produced more thyroxine than monolayer cultured thyrocytes. Pan J (75) also employed a decellularization-recellularization method to reconstruct multifunctional organoids consisting of both thyroid cells and parathyroid cells. A rat acellular thyroid scaffold with an intact follicular basement membrane and arterial elastic fiber network was first obtained by thyroid artery perfusion with sodium dodecyl sulfate. Alternative decellularization protocols were also provided by Alfieri M (78). Upon the subsequent coseeding of human thyroid cells and parathyroid cells into this rat decellularized matrix, all of these cells were capable of distributing throughout the scaffold and expressing critical proteins, such as TG, TPO, and parathyroid hormone (PTH).

Four studies between 2012 and 2021 focusing on in situ compensative thyroid regeneration after thyroid damage were identified (79–82), and all of them were performed in mice. It was demonstrated that the central part of the intact thyroid lobe where many small or micro follicles were present generally served as a center for proliferation, where newly formed follicles were generated and pushed to the peripheral zone during maturation. When partial thyroidectomy was performed, on the one hand, the primitive thyroid follicles in the center of proliferation underwent a regeneration process (the budding type of folliculogenesis) to produce new follicles (81). On the other hand, the proliferative center extended to the area near the cut edge, and some follicles in the area became irregular in shape, and this affected area was considered under repair after damage. In the proliferative area, some class of immature clear cells was found to participate in the repair or regeneration of the thyroid gland (79); however, the nature and origin of these clear cells and their relationship to follicular cells or parafollicular cells (C cells) were not fully elucidated. Several crucial stemness markers were monitored to demonstrate whether stem cells or progenitor cells played a part in the process of thyroid regeneration after thyroid damage. In a transgenic TPOCreER2/iDTR mouse model, after 4 weeks of combined administration of tamoxifen (TM) and diphtheria toxin (DT), more than 90% of TPO-expressing mature thyroid follicles were damaged in mice. A significant increase in stem cell markers (OCT4, Nanog, SOX and REX1) during and following TM/DT treatment with a transient surge at 2 weeks was intuitively displayed (82). Furthermore, stem cell antigen 1 (SCA1)-positive cells were initially found to have negative NKX2-1 expression in nonfollicular mesenchymal areas, suggesting that SCA1+ cells were not of thyroid follicular cell origin. However, at 120 days after surgery, these SCA1+ cells subsequently differentiated to SCA1+/NKX2-1+ cells, indicating that SCA1+ cells might exhibit stem-like or progenitor-like cell characteristics and be responsible for the regeneration of thyroid follicular cells after surgery (80). Keratin14 (KRT14), known as a liver progenitor marker, was also identified in a limited part of the proliferative area. However, KRT14+ cells were demonstrated to have no specificity, as they could be found in both follicular cells and nonfollicular areas (80).

Concerning thyroid function recovery after partial thyroidectomy or damage, in most conditions, serum FT4 or T4 levels could return to normal at 2 to 8 weeks after damage, but TSH levels were still elevated with statistical significance (81).

Eleven studies between 1984 and 2022 investigating the efficacy and potential tumorigenicity of thyroid grafts were identified. Eight of them elucidated the efficacy of artificial thyroid gland transplantation, mainly into hypothyroid animals, and 3 of them specifically displayed the potential tumorigenicity of the implanted grafts.

To assess regenerative efficiency and function, the explanted thyroid tissues were routinely examined by histology, immunodetection, protein, or gene expression assays (83). First, the morphological formation of thyroid-like tissue in host animals was shown after transplantation. It was found that no matter which kinds of thyroid organoids were implanted under the skin or beneath the kidney capsule, one or two months later, the transplanted grafts exhibited numerous follicles formed by monolayered epithelium at the grafting sites (9, 19, 21, 26, 49, 59, 63, 65, 73), verified by cytosolic TG expression and TG deposition within the follicular lumens. Second, the harmonic function of implanted thyroid organoids at different quantities was illustrated. In totally thyroidectomized rats, the experimental group with 1×10⁶ cells of TGE grafts showed a significant increase in body weight, although the final weight attained fell short of that of the control group with intact thyroid glands (49). Additionally, transplanting thyroid cell sheets with 2.5×10⁶ cells into rats following thyroidectomy significantly elevated both FT3 and FT4 at 2 weeks after surgery, while transplanting half the number only contributed to FT4 recovering significantly at 3 weeks and to less than the normal level when FT3 and FT4 levels were saturated at 4 weeks (76). It was noticeable that in nonthyroidectomized rats, the TGE graft was not able to form follicles or change their weights (49). Similarly, in euthyroid severe combined immunodeficient (SCID) mice, no matter how many thyrocytes of the human thyroid-derived organoid were transplanted, the total T4 level in recipient mice did not differ significantly from that in normal SCID mice (19, 59). An insufficient TSH feedback loop stimulation induced by normal TSH levels in euthyroid animals may account for these phenomena. In I¹³¹ inducible hypothyroid mouse models, 3 and 5 weeks after grafting three TS and six AS, recipient mice showed a gradual normalization of body temperature and a substantial elevation of serum T4 levels, while T4 levels were still far from the normal range (73). After transplanting 2.5-3×10⁶ cells from NKX2-1+/PAX8+ ESC-derived organoids beneath the unilateral kidney capsule, hypothyroid mice showed a full normalization of plasma thyroid hormone levels (21) and body temperature (63) 4 or 8 weeks after grafting. Reconstitution of T4 levels was accompanied by amelioration of elevated plasma TSH levels, further emphasizing that physiologic rescue was occurring (21). However, when injecting 6×10⁵ cells from adult thyroid-derived organoids underneath the kidney capsule, although the area of the follicular structure remained stable at the grafting site, the FT4 levels only increased very modestly (9).

Regarding the tumorigenicity of thyroid organoids, no reports of tumor events were noted after in vivo transplantation in all of the recipients during their research periods. No tumor development was observed even in xenografted SCID mice (69), nor was it observed when transplanted p53 knockout thyroid-derived organoids were subcutaneously injected into BALB/c nu/nu mice (67). In addition, no macroscopic tumors were detected at 1 year after transplantation within the thyroid organoid implants that were previously exposed to X-ray radiation before transplantation (9). Intriguingly, gene expression analysis showed that more than 70% of upregulated genes related to thyroid carcinogenesis were downregulated in early passage thyroid organoids and continued to be downregulated in later passage thyroid organoids. Correspondingly, more than 70% of downregulated genes related to thyroid carcinogenesis were upregulated and continued to be upregulated by prolonged culture. These data indicated that prolonged culturing did not result in the activation or inhibition of tumor-related markers.