In humans, the organization of the endoderm and posterior foregut (PFG) occurs during the first 2–3 weeks (E7–E8.5 in mice). Subsequently, liver organogenesis starts at week 4, and the liver bud is created, where the hepatoblast migration and proliferation take place between weeks 5 and 7 (17, 18) (E9–E14 in mice; refs. 17, 19). Lastly, hepatoblasts differentiate into hepatocytes or cholangiocytes during weeks 7–30 (E13.5–E18.5 in mice). Here, we studied the developmental role played by DIO2 by using human iPSCs to generate hepatic organoids, which exhibit developmental stages that resemble those in human hepatic organogenesis.

Differentiation of iPSCs into hepatic organoids

To monitor the progression of the differentiating iPSCs into hepatic organoids, we tracked the expression of the stage-specific cell markers over time (Figure 1, A–F). As expected, the expression of the iPSC markers (20) POU5F1 (also known as OCT4) and SOX2 decreased after the differentiation into definitive endoderm (DE), at the same time that the DE markers OTX2, CER1, and FOXA2 increased (Figure 1B). Subsequently, the increase in the mRNA levels of HNF4A, TBX3, and CDX2 (21, 22) signaled the organization of the PFG (Figure 1B). The changes in the expression of these markers (OTX2, HNF4α, and TBX3; ref. 23) were also tracked through immunofluorescence (Figure 1C).

Differentiation of human iPSC into hepatic organoids. (A) Graphic representation of the differentiation phases. The first phase involves iPSC (day 0), DE (day 4), and PFG (day 10) in a 2-dimensional culture, and the second phase comprises IHO (day 14 to day 18), HBO (day 22 to day 26), and HO, which differentiated into a 3-dimensional structure since day 10. The maturation of the HO encompasses 2 phases, from day 27 to day 38 (HO1) and the second from day 39 to day 46 (HO2). C1–C6 are the indicated differentiation cocktails used (see Methods). (B) Relative mRNA levels of iPSC (POU5F1, SOX2), DE (OTX2, CER1, FOXA2), and PFG (TBX3, HNF4A, CDX2) markers; β-actin was used as the internal control. Entries are the mean of duplicates, represented as aligned scatter dot plots and the mean ± SD (iPSC, n = 8; DE and PFG n = 7; HNF4A in PFG, n = 6), and each differentiation stage is indicated at the bottom of the graphs. (C) Immunofluorescence of OTX2 (red), HNF4α (green), and TBX3 (red) in iPSC, DE, and PFG cells; nuclei are in blue. Scale bar: 150 μm. (D) Representative bright-field images of self-organized hepatic organoids from day 10 to day 38. (E) Immunofluorescence of HNF4α (green) and albumin (red) at day 18, day 26, and day 46. (F) Immunofluorescence of the proliferative markers MKI67 (green) and TBX3 (red) at day 18, day 26, and day 46. Nuclei are shown in blue; the inset is magnified on the bottom left of the first panel.

On day 10, the 2D cultures were disrupted by dissociation, and subsequently, the cells self-reestablished into a 3D structure that grew progressively under stimulation of key reagents in the differentiation cocktails — i.e., BMP4, BMP7, FGF7, and the WNT agonist CHIR99021 — to promote the hepatic specification (Figure 1D) (17). This occurred during the transition into 3D organoids and started with immature hepatic organoids, followed by hepatoblast organoids, and finally hepatic organoids (Figure 1A). These transitions were monitored by immunofluorescence of the developmental markers HNF4α (Figure 1E) and TBX3 (Figure 1F), together with the hepatocyte marker albumin (Figure 1E) and the proliferative marker MKI67 (Figure 1F).

The presence of maturing hepatocytes was detected through HNF4A and albumin (ALB) mRNA and by human albumin in the conditioned medium (Figure 2, A and B). Furthermore, HNF4α and albumin+ cells were visualized on day 46 organoids (Figure 1E). AFP is highly expressed by the embryonic hepatocytes in the fetal liver. Here, AFP mRNA levels were briefly detected on day 10 and increased progressively to reach an ~3-fold increase by day 42 (Figure 2A). In addition, we also measured mRNA for the hepatocyte genes CYP3A7, a major fetal hepatic enzyme, and CYP3A4, a cytochrome P450 isoform involved in the drug metabolism in adults (Figure 2C). CYP3A7 expression increased at day 14 by ~10-fold; it then diminished and remained steady (~2-fold) until day 46 (Figure 2C). A similar pattern was observed for CYP3A4 except that, at day 42, there was an ~7-fold increase in expression (Figure 2C). At the same time, KRT7 mRNA, a typical cholangiocyte marker, appeared on day 22 (Figure 2B), increased by ~3-fold, and reached a plateau from day 29 to day 35 (Figure 2B). These results indicate that the differentiation of iPSCs into hepatic organoids generates developing hepatocyte-like and cholangiocyte-like cells (Figure 2D).

Gene markers and deiodinase expression patterns during the process of differentiation into hepatic organoids. (A) Relative mRNA levels of ALB and AFP (left y axis) and albumin levels in the medium (right y axis). Albumin levels in the medium from day 14 to day 46; for day 14 and day 18, each sample was obtained from 10 individual wells and combined; for day 22 to day 46, there were 10 organoids per well (n = 2, day 14, day 22, day 46; n = 3, D18 and day 26; n = 4, day 29, day 42; n = 5, day 32, day 38; n = 6, day 35). (B) Relative mRNA levels of KRT7 and HNF4A. (C) Relative mRNA levels of cytochrome P450 isoforms, CYP3A7 (fetal liver marker) and CYP3A4 (adult liver marker). (D) Immunofluorescence at day 46 of albumin (hepatocyte marker; red; top left; scale bar: 150 μm) and KRT7 (cholangiocyte marker; green; middle left; scale bar: 150 μm); the inset is magnified (bottom left; scale bar: 40 μm), and the merge of both images (right). Nuclei are shown in blue. (E) Relative mRNA levels of DIO1, DIO2, and DIO3. (F) Relative mRNA levels of nuclear thyroid receptor, THRA and THRB. (G) Relative mRNA levels of nuclear coactivator NCOA1 and corepressors NCOR1 and NCOR2. β-Actin was used as the internal control; all entries are the mean of duplicates and shown as aligned scatter dot plots (mRNA samples: day 4, n = 3; day 14, n = 7; rest, n = 4). #Data not available for DIO1. The days of the differentiation are shown on the x axis (see Figure 1).

A transient surge in DIO2 expression during hepatoblast differentiation

We previously identified Dio2 expression in E13.5–E18.5 mouse liver embryos (4), a timing during which there is differentiation of hepatoblasts into hepatocytes (17). Here we also identified a peak of DIO2 during the differentiation of hepatoblasts into hepatocytes. DIO2 mRNA surged on day 14 to reach an ~6-fold expression peak by day 22 that was sustained until day 29 (Figure 2E) and subsequently subsided (Figure 2E). D2 catalytic activity reached ~0.35 nmol T3/h/mg protein on day 22 and dropped by half on day 38 (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.173780DS1). At the same time, DIO3 expression (the enzyme that inactivates TH) remained low throughout the differentiation process, except for day 4 and day 14, when a subtle but short-lived elevation was observed (Figure 2E and Supplemental Figure 1B). DIO1 mRNA emerged to ~2-fold at day 22 until day 32 (Figure 2E). It then increased again ~2-fold at day 38 (Figure 2E). THRA mRNA levels were low during the early stages of differentiation, but after day 10, THRA mRNA levels increased progressively to reach ~2-fold by day 46 (Figure 2F). THRB expression was induced after day 26 by about 2-fold and remained high through day 46 (Figure 2F), along with a steady presence of the TR coregulators NCOR1, NCOR2, and NCOA1 (also known as SRC1) (Figure 2G).

Local D2-mediated T3 production affects the differentiation of hepatoblasts

The D2 pathway is only active if T4 is available, whether availability is from plasma or supplied in the medium. The differentiation cocktails (C2–C6) do not contain T4, but B27 contains T3 at above the physiological levels (24). Thus, to eliminate the T3 signaling and the local D2-mediated T4-to-T3 activation during the development of liver organoids, we cultured DE cells (day 4) in B26 (without T3) medium containing no TH (vehicle-only [V-HOs]) or containing only ~10 pM free T3 (T3-HOs), which is equivalent to physiological plasma T3 concentration. The results were contrasted with cells grown with ~15 pM free T4 (T4-HOs), supporting physiological D2-mediated T3 production (Figure 3A).

Formation of hepatic organoids in the presence or absence of TH. (A) Graphic representation of the development of hepatic organoids in the presence or absence of TH. The yellow arrow indicates the addition of TH (from day 5 to day 50). The final concentration of T4-HOs (red) and T3-HOs (blue) was 1 nM T4 (free T4 = ~15 pM) and 200 pM T3 (free T3 = ~10 pM); V-HOs were grown in the absence of T4 or T3 (black). C1–C6 are the indicated differentiation cocktails used (see methods). (B) Bright-field images (3 conditions) of hepatoblast at day 22 and hepatic organoids at day 29. Scale bar: 200 μm. (C) Relative mRNA levels of HNF4A (hepatocyte marker) and CEBPA from day 26 to day 32 (n = 4). (D) Albumin levels in the medium from day 35 to day 50 (n = 4). (E) Relative mRNA levels of KRT7 at day 46 and day 50 (n = 4, except T4-HOs, n = 3). (F) ALB/KRT7 mRNA ratio during from day 38 to day 46 (n = 4 except T4-HOs and T3-HOs at day 42, n = 2; T4-HOs at day 46, n = 3). (G and H) Apolipoprotein B (APOB) and Apolipoprotein A1 (APOA1) levels from the medium from day 35 to day 50, comparing V-HOs (black) versus T4-HOs (red). Ten organoids per well (n = 4). Two-tailed Student’s t test for comparing V-HOs versus T4-HOs, and 1-way ANOVA and Tukey test were used for multiple comparisons. Data are the mean of duplicates, represented as aligned scatter dot plots and their mean. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. The days of the differentiation are shown on the x axis (see legend Figure 1).

Our first approach was to measure the volume of the developing organoids, which increased progressively from day 15 to day 38 (Supplemental Figure 1C). This is reflected in the ~2-fold–higher mRNA levels for the proliferative marker MKI67 in day 22 as compared with day 10 (Supplemental Figure 1D). The growth of the organoids was affected by the functionality of the D2-generated T3 pathway. At day 22, T4-HOs were 20%–25% smaller than V-HOs and ~15% smaller than T3-HOs (Figure 3B and Supplemental Figure 1E). No differences were observed between T3-HOs and V-HOs (Supplemental Figure 1E). On day 29, both T4-HOs and T3-HOs were ~20% smaller than V-HOs (Figure 3B and Supplemental Figure 1E). Accordingly, the absence of T4 in the medium increased MKI67 mRNA levels by ~20% and ~75% on day 14 and day 18, respectively (Supplemental Figure 1D). These differences confirmed our hypothesis that the local D2-mediated T3 production affects the differentiation process of hepatic organoids.

We next measured the expression of 2 TFs involved in hepatocyte differentiation, HNF4A and CEBPA, on day 26 to day 32 (Figure 3C), along with expression of 2 differentiation-specific markers, ALB and KRT7, during day 38 to day 50 (Figures 3, D–F, and Supplemental Figure 1F). First, we noticed that HNF4A and CEBPA mRNA levels began to increase at day 29, with HNF4A reaching higher levels in T4-HOs (Figure 3C). In addition, on day 42 and day 46, ALB mRNA levels remained substantially higher in the T4-HOs as compared with T3-HOs or V-HOs (Supplemental Figure 1E). These differences were also reflected in the albumin (Figure 3D), and apolipoprotein (APO) levels in the conditioned media of T4-HOs. Albumin levels were 3.0-fold and 2.5-fold higher than in V-HO on day 42 and day 46 (Figure 3D), respectively, whereas APOB levels were ~10-fold and 3-fold higher (Figure 3G) and APOA1 levels were ~3-fold and ~2-fold higher (Figure 3H) at day 42 and day 46, respectively. Conversely, KRT7 mRNA levels on day 46 and day 50 were lower in T4-HOs but not in T3-HOs (Figure 3E). The magnitude of differences between ALB and KRT7 expression during day 46 can be appreciated by plotting the ratio of ALB/KRT7 (Figure 3F). These results suggest that the local D2-mediated T3 production favors the differentiation of the hepatoblasts toward hepatocytes versus cholangiocytes.

The temporal expression of TFs is modified by local D2-mediated T3 production

The local D2-T3 signaling also affected other key TFs, including the pioneer factors GATA4 (role in liver bud expansion) and FOXA1 (Figure 4, A–C). GATA4 was highly expressed on day 10 to day 18 and decreased progressively during organoid maturation (Figure 4A). In cells capable of local D2-mediated T3 production, GATA4 mRNA levels were ~2.5-fold higher on day 18 and ~3-fold higher on day 22 (Figure 4A). FOXA2 mRNA levels were not modified by the presence of T4 (Figure 4B). In these cells, FOXA1 mRNA peaked on day 14, decreased during the hepatoblast, and slowly increased again until organoids (Figure 4C). A functional D2-T3 pathway had a generalized positive effect on FOXA1 expression, with ~130% and 70% higher mRNA levels on day 22 and day 26, respectively (Figure 4C); much higher fold inductions by D2-T3 were observed at day 29 and day 32 (respectively, ~3-fold and ~2-fold; Figure 4C).

Expression of TFs involved in liver development. (A–J) Relative mRNA levels of TFs (GATA4, FOXA2, FOXA1, PROX1, HHEX, HNF1A, HNF4A, CEBPA, HNF1B, ONECUT1) involved in the network regulation of the hepatic organoids from day 10 to day 46 in V-HOs (black) and T4-HOs (red) (n = 4 except T4-HOs at day 42, n = 2; day 46, n = 3; T4-HOs at day 18 in FOXA2, PROX1, and at day 10 in HHEX n = 3; and V-HOs at D29 in HHEX, n = 3). β-Actin was used as the internal control. The yellow square represents the hepatoblast expansion (HBO), and the purple squares indicate the period of hepatocyte and cholangiocyte maturation (HO1-2). Two-tailed Student’s t test was used to compare groups each day. Data are the mean of duplicates and represented as aligned scatter dot plots. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. The days of the differentiation are shown on the x axis (see Figure 1).

PROX1 promotes the migration and proliferation of the hepatoblast during the early liver bud formation. Its mRNA levels were low up until day 10, when it increased progressively to reach a plateau by day 18 and subsequently dropped over time (Figure 4D). In hepatoblasts and in early hepatic organoids, a functional D2-mediated T3 production increased the expression of PROX1 by ~2-fold (day 22) (Figure 4D). HHEX is another early TF that plays a role in the formation of the liver bud and, later on, the development of the biliary tree. The HHEX mRNA levels were high by day 10 and decreased sharply by day 14 (Figure 4E). A functional D2-mediated T3 production mechanism increased this early expression peak of HHEX by ~3-fold but, during the later phases of development, markedly reduced HHEX expression by ~10-fold (Figure 4E).

The 3 hepatocyte-promoting TFs, HNF1A, HNF4A (which, together with CEBPA promotes albumin production and with HNF1A maintains and directs the hepatocyte fate; refs. 25, 26), and CEBPA, exhibited low expression levels in the early hepatoblasts but increased over time; by organoids, their expression levels had increased 2- to 4-fold (Figure 4, F–H). A functional D2-mediated T3 production mechanism had an overall stimulatory effect on these 3 TFs. During the different developmental periods, D2-T3 caused HNF1A mRNA levels to increase by ~2-fold (Figure 4F), HNF4A mRNA levels by 2- to 5-fold (Figure 4G), and CEBPA mRNA by 40%–75% (Figure 4H).

HNF1B and ONECUT1 (also known as HNF6) are TFs that play a role in bile duct morphogenesis (Figure 4, I and J). HNF1B expression remained stable throughout development (Figure 4I), but ONECUT1 mRNA levels peaked at day 10, sharply diminished to low levels by day 14, and remained low (Figure 4J). Here, the local D2-mediated T3 production stimulated HNF1B expression by 25% to 2-fold in hepatoblasts and early organoids (Figure 4I), whereas ONECUT1 was potently induced by D2-T3 (~4.5- to 5-fold) exclusively at day 46 (Figure 4J).

Hepatic organoids are made up of hepatocyte-like and cholangiocyte-like cells

Single-cell RNA-Seq (scRNA-Seq) was used to analyze a total of 10,887 day 45 cells (Figure 5A), which were sorted into 21 clusters (Figure 5B) distributed in 4 groups based on (a) the expression and cell positivity level of 10 developmental genes (Figure 5C and Supplemental Figure 2A), (b) the number of conserved genes expressed in each cluster (Figure 5F and Supplemental Table 1), and (c) the top 20 genes expressed in each cluster (Supplemental Table 2):

scRNA-Seq identification of cells in hepatic organoid at day 45. (A) PCA of the 2 first principal components (PC1 versus PC2) of hepatic organoids at day 45 (n = 10,887 cells). (B) UMAP visualization with a SSN at 0.8 resolution. (C) UMAP distribution plot of differentiation markers: POU5F1 and SOX2 (pluripotency markers); CER1, OTX2, and FOXA2 (DE markers); CDX2 and TBX3 (PFG markers); HNF4A and ALB (hepatocyte markers); KRT7 (cholangiocyte marker); and DIO2 and DIO3. (D) UMAP visualization after grouping the clusters. (E) scRNA-Seq violin plots of markers HNF4A, KRT7, and ALB in the cell groups. 1 indicates hepatocyte-like cells; 2 indicates cholangiocyte-like cells; 3 indicates cholangiocyte-like precursor cells; and 4 indicates hepatoblast-like cells. (F) A plot of the number of conserved genes shared across the clusters. Colored squares (green, dark blue, sky blue, turquoise) indicate the cluster of cells that make up every group. Black dotted and dashed lines indicate a resemblance among groups.

Hepatocyte-like cells. This group consisted of 6 cell clusters (clusters 0, 2, 3, 7, 9, and 16) expressing HNF4A (26), CDX2, and the typical hepatocyte markers (i.e., FGG, FGB, APOC3, APOA4, and GSTA1) (Figure 5C, Supplemental Figure 2A, and Supplemental Table 1) as well as a high number of conserved genes (Figure 5F).

Cholangiocyte-like cells. Using a similar rationale, we identified 4 clusters (clusters 1, 4, 5, and 18), all of which expressed KRT7; other typical cholangiocytes markers — e.g., KRT8, PROM1, SOX9, and HNF1β — were not highly expressed (Supplemental Table 1).

Hepatoblast-like cells. The next 8 most similar clusters (clusters 8, 10, 11, 12, 13, 14, 17, and 20) expressed the early developmental markers SOX2 (iPSC), HNF4A, and FOXA2 (DE, PFG) not present in the hepatocyte- or cholangiocyte-like groups of cells (Figure 5C). These clusters also expressed OTX2 (DE), CDX2 (PFG), and ALB (all expressed in hepatocyte-like cells but not in cholangiocyte-like cells; Figure 5, C–E).

Cholangiocyte-like precursor cells. The last 3 unassigned clusters (clusters 6, 15, and 19) expressed KRT7 and SOX2 (Figure 5, C and E, and Supplemental Figure 2A). This profile suggested an intermediate group identity between hepatoblast-like cells and cholangiocyte-like cells.

Cell composition of hepatic organoids is affected by local D2-mediated T3 production

Local D2-mediated T3 production influenced the hepatoblast maturation process, greatly affecting the proportion of the cell clusters (χ2 test, P < 0.0001) (Figure 6, A and B). The presence of either a functional D2-T3 pathway (χ2 test, P < 0.0001) or, to a lesser extent, T3 (χ2 test, P < 0.0001) was a critical determinant of the fate of the hepatoblasts (Figure 6, A and B). Within hepatocyte-like cells, all cell clusters were affected by TH signaling (χ2 test, P < 0.0001, Figure 6C). The presence of TH, with D2-T3 (χ2 test, P < 0.0001) or T3 (χ2 test, P < 0.0001), influenced the hepatocyte-like cell proportion. As an example, cluster 7 was the largest in V-HOs but dropped to one of the smallest with the addition of TH to the medium (Figure 6C).

scRNA-Seq quantification of cells and gene expression in hepatic organoids at day 45. (A) UMAP visualization of V-HOs, T3-HOs, and T4-HOs. The number of cells is denoted in each group. (B) Histogram of the relative number of cells in groups treated with V-HOs, T3-HOs, and T4-HOs. (C–F) Histograms of the relative number of cells in clusters treated with V-HOs, T3-HOs, and T4-HOs. The identification number of each cell cluster is indicated at the bottom right corner of each rectangle. V-HOs (vehicle), T3-HOs (free T3 = ~10 pM), T4-HOs (free T4 = ~15 pM). The χ2 test for multiple comparisons and pairwise cell proportion. **P < 0.01; ****P < 0.0001.

TH effects were also present in cholangiocyte-like cells (χ2 test, P < 0.0001) and cholangiocyte-like precursor cells (χ2 test, P < 0.0001), but they were relatively less dramatic (Figure 6, D and E). Cholangiocyte-like cell clusters were similarly affected by T4 (χ2 test, P < 0.0001) but not by T3 (χ2 test, P < 0.093) (except for cluster 18, which responded prominently to TH; Figure 6D). Cholangiocyte-like precursor cells were relatively more responsive to TH, with clusters 19 (which expresses THRB; Supplemental Table 2) and 15 responding to TH (Figure 6E). As in hepatocyte-like cells, the intensity of the D2-T3 effects on the proportion of cells was greater than with T3 (χ2 test, P < 0.0001). As precursor cells, the hepatoblast-like cell clusters indicated a high response to TH (χ2 test, P < 0.0001). D2-T3 (χ2 test, P < 0.0001) and T3 (χ2 test, P < 0.0001) actions dramatically affected the hepatoblast-like cell clusters, except for clusters 12 and 13 (Figure 6F).

Effects of local D2-mediated T3 production on the organoid transcriptome

Local D2-T3 production also markedly affected the organoids’ gene expression, as shown in the volcano plot distribution (Supplemental Figure 2, B–E) and in the cellular heatmaps (Supplemental Figures 3–5). The hepatocyte-like cell group was the most affected by D2-T3 (Supplemental Figure 2B), which modified the expression of 361 genes (302 were >1.5-fold upregulated). The changes in gene expression with T3 were less intense but qualitatively similar (Supplemental Figure 2B). As a group, these genes were involved in metabolic and developmental processes (Supplemental Table 3). The main pathways affected by D2-T3 included lipoprotein transport (fold-enrichment [FE]: 21), localization (FE: 20), alcohol metabolism (FE: 17), cholesterol metabolism (FE: 17), and sterol metabolism (FE: 17) (Supplemental Table 3). Other specific biological pathways were related to the estrogen metabolism and to the cholesterol and sterol homeostasis, with FE values of 11, 7.2, and 7.1, respectively (Supplemental Table 3). Of note, 71 differentially expressed genes (DEGs) (χ2 test, P < 0.00001) in the hepatocytes lacking D2-T3 (mainly involved in lipid metabolism) overlapped with the group of hepatic genes differentially expressed after the Dio2 knockdown, confirming the relevance of these findings in organoids (Supplemental Figure 2F).

Hepatoblast-like cells were also sensitive to D2-T3, and this sensitivity affected the expression of 209 genes (121 were >1.5-fold upregulated); similarly to hepatocyte-like cells, incubation with T3 had qualitatively similar but less intense effects (Supplemental Figure 2C). Cholangiocyte-like cells and cholangiocyte-like precursor cells were much less sensitive to incubation with TH (Supplemental Figure 2, D and E). D2-T3 modified the expression of only 28 genes (18 were >1.5-fold upregulated), whereas, in cholangiocyte-like precursor cells, that number was 74 (55 were >1.5-fold upregulated). The effects of T3 were less substantial but qualitatively similar (Supplemental Figure 2, D and E).