SALL2 regulates neural differentiation of mouse embryonic stem cells through Tuba1a – Cell Death & Disease

Expression of SALL2 protein during mouse embryo development

We set to survey the expression of Sall2 during mouse embryo development. First, we retrieved and analyzed the published RNA-seq datasets of mouse embryos from zygote to E15.5 (GSE222357, GSE98150, GSE214161, GSE216492, and https://www.informatics.jax.org/gxd/marker/MGI:1354373), which showed that the transcription of Sall2 initiated at E5.5 and was restricted to the epiblast of the embryo (Supplementary Fig. S1A). With embryo development, the expression of Sall2 gradually increased and was mainly detected in the brain, kidney, and other tissues (Supplementary Fig. S1B). These results were consistent with the reported ISH of Sall2 mRNA expression during mouse embryo development [11]. Next, in order to determine SALL2 protein expression during mouse embryo development, we applied immunofluorescence staining of SALL2 from zygote to E15.5 embryos.

For preimplantation embryo, from zygote to blastocyst, there was no SALL2 protein expression observed (Fig. 1A). After implantation, SALL2 was detected in the epiblast of E5.5-E6.5 embryos, then gradually increased (Fig. 1B). At gastrulation stage embryo (E7), SALL2 mainly expressed in neuroectoderm and primitive streak, and a few SALL2+ cells were detected in amnion and chorion (Fig. 1B). In E8-E8.5 embryos, a critical period for the formation and closure of neural tube (NT) [22], SALL2 was extensively expressed in neural fold, neuroectoderm, forebrain vesicle and NT (Fig. 1B and Supplementary Fig. S1C, F). From E9.5-E10.5, constant SALL2 protein was detected in the nervous system, such as NT and brain (Fig. 1B and Supplementary Fig. S1C, F).

Fig. 1: The expression pattern of SALL2 protein in mouse E0.5-E15.5 embryos.
figure 1

A Expression of SALL2 from mouse zygote to blastocyst. The embryos were immunostained for SALL2 and SOX2. DAPI stained the nuclei. The arrowheads indicated the embryo, and the dashed circle outlined the ICM of the blastocyst. Scale bar, 50 μm. B Expression of SALL2 in mouse E5.5-E9.5 embryos. The embryos were immunostained for SALL2. DAPI stained the nuclei. Dashed circles outlined the Epi of embryos. E5.5, scale bar, 75 μm; E6.5, scale bar, 250 μm; E7-E9.5, scale bar, 250 μm. C cell, ICM inner cell mass, TE trophectoderm, Epi epiblast, Ch chorion, Am amnion, Hf headfold, Ps primary streak, Nf neural fold, Fv forebrain vesicle, Ne neural ectoderm, F forebrain, M midbrain, H hindbrain, A anterior, P posterior, D dorsal, V ventral.

It has been reported that in the cerebral cortex of the adult mouse, SALL2 was translocated from the nucleus into the cytoplasm [23]. Interestingly, for E11.5 embryos, in the midbrain limbic system, the cytoplasm expression of SALL2 was observed in neural cells characterized by the expression of neural marker TUBB3. Again, SALL2 is mainly expressed in the forebrain, diencephalon, midbrain, hindbrain, and NT (Supplementary Fig. S1D, E, G). At E12.5, except its expression in the telencephalon, midbrain, and pons, SALL2 is also expressed in the spinal cord and preoptic neuroepithelium (Supplementary Fig. S1D, G).

With the development and maturation of organs in mouse embryo, from E13.5 to E15.5, SALL2 expression decreased with its expression pattern being stable in telencephalon, diencephalon, midbrain, pons, hindbrain, spinal cord, preoptic neuroepithelium as well as pituitary gland and olfactory epithelium (Supplementary Fig. S1D, E, G).

Therefore, by combining immunofluorescence staining results with RNA-seq data, we charted a complete SALL2 expression pattern during mouse embryo development (E0.5-E15.5) and found that SALL2 mainly expressed in the developing nervous system, indicating its potential role in neurogenesis.

Impaired pluripotency of Sall2 deficient ESCs

PSCs have been widely used as in vitro model [24] to investigate gene function in development and diseases. To interrogate the role of SALL2 in early embryo development, we first examined SALL2 expression in ESCs, while there was almost no SALL2 expression in naïve ESCs cultured in 2i/LIF, ESCs cultured in serum/LIF medium (M15), which contain differentiating cells, expressed SALL2 (Supplementary Figs. S1H, S2C). These results mirrored our in vivo SALL2 immunostaining data, as naïve ESCs resemble epiblast in blastocyst without SALL2 expression, while SALL2 was detected in post-implantation embryos [25, 26] (Fig. 1A, B).

The pluripotency marker REX1 tagged GFP (REX1:GFP) ESC reporter line has been broadly used to monitor the self-renewal and pluripotency of ESCs [27, 28]. We applied clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 technology to knock out Sall2 in REX1:GFP ESCs [29]. Since at least two Sall2 isoforms have been discovered, we designed guide RNA (gRNA) targeting the common Exon 2 of Sall2 (Fig. 2A). After transfection and drug selection, ESC colonies were picked up for characterization. By Sanger sequencing, we verified at least two independent Sall2 KO ESC lines with different deletions (Supplementary Fig. S2A, B). Results of qRT-PCR, western blot, and immunofluorescence staining showed Sall2 deletion in these lines (Fig. 2B, C and Supplementary Fig. S2C). Moreover, by checking the predicted off-target sites, no mutation was detected (Supplementary Fig. S2D). Thus, we have successfully established Sall2 KO REX1:GFP ESC lines (REX1:GFP Sall2 KO).

Fig. 2: Construction and characterization of REX1:GFP Sall2 KO ESC lines.
figure 2

A Schematic diagram of sgRNA design for Sall2 deletion. B mRNA expression of Sall2 in REX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). C SALL2 protein expression in REX1:GFP Sall2 KO and WT ESCs. GAPDH served as a loading control. D Morphological analysis of REX1:GFP Sall2 KO ESCs. Scale bar, 250 μm. E Flow cytometry analysis of REX1:GFP+ cells in REX1:GFP Sall2 KO ESCs (n = 3, biological replicates). Statistical significance was determined by one-way ANOVA with Tukey’s test. ns not significant. F qRT-PCR analysis of pluripotency markers (Nanog, Sox2, Rex1, Klf2, Klf4) in REX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test. *P < 0.05, **P < 0.01, ns not significant. G Expression of OCT4 in REX1:GFP Sall2 KO and WT ESCs. The cells were immunostained for OCT4. DAPI stained nuclei. Scale bar, 250 μm. H qRT-PCR analysis of Sall2 in REX1:GFP Sall2 KO and WT ESCs during EB differentiation. Relative to Gapdh expression (n = 3, technical replicates). I qRT-PCR analysis of pluripotency markers (Rex1, Oct4, Klf4) during EB differentiation. Relative to Gapdh expression (n = 3, technical replicates). J qRT-PCR analysis of meso-endoderm markers (Flk1, Gata6, T, Gata4) during EB differentiation. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between Sall2 KO cell lines (KO1, KO2) and WT cells. ****P < 0.0001. K Observation of REX1:GFP during RA-induced EB differentiation in REX1:GFP Sall2 KO and WT ESCs. Scale bar, 250 μm. L qRT-PCR analysis of ectoderm markers during RA-induced EB differentiation. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between Sall2 KO cell lines (KO1 + RA, KO2 + RA) and WT cells (WT + RA). ****P < 0.0001.

Next, we characterized the REX1:GFP Sall2 KO ESC lines. Morphologically, the Sall2 KO ESCs displayed a dome shape, similar to wild-type (WT) ESCs (Fig. 2D). When observed under fluorescence microscope, the REX1:GFP expression in Sall2 KO ESCs was comparable to WT cells, which was further confirmed by flow cytometry analysis of REX1:GFP (Fig. 2D, E). We then determined the expression of some pluripotency markers Nanog, Sox2, Rex1, Klf2, and Klf4. For Nanog, Sox2, Klf2, and Klf4 expression, there was no significant difference between WT and KO1 ESCs, while a slight increase was detected in KO2 ESCs; Rex1 expression increased moderately in both KO1 and KO2 ESCs (Fig. 2F). The difference in pluripotency gene expression may be attributed to clone difference [30, 31]. Meanwhile, by immunostaining, the OCT4 protein expression was not affected in Sall2 KO ESCs as well (Fig. 2G).

We then tested whether Sall2 KO ESC lines could be established from blastocysts with Sall2 deficiency. Zygotes were injected with Sall2 single guide RNA (sgRNA) and Cas9 protein, then cultured in vitro until the blastocyst stage [32, 33]. From 91 zygotes injected, 77 progressed to blastocysts (84.6%), as confirmed by SOX2 expression in the inner cell mass (ICM), which was not significantly lower than that of the noninjected embryos (92.3%) (Supplementary Fig. S2E–G). Meanwhile, eight embryos were collected for PCR and sequencing, which showed about 94.2% KO efficiency (Supplementary Fig. S2H, I). From 16 blastocysts cultured in M15, nine outgrowths formed, after passaging, nine ESC clones were established (56.25%), all of which showed Sall2 deletion (Supplementary Fig. S2J–L). We further performed qRT-PCR and immunostaining to check the expression of pluripotency markers, which displayed comparable levels between WT and Sall2 KO ESCs, though no SALL2 protein was detected (Supplementary Fig. S2M, N). These data demonstrated that SALL2 was not required for the acquisition of ESC state.

As Sall2 expression increased in heterogenous ESC culture with M15 and in post-implantation embryos, we performed embryoid body (EB) differentiation to determine the effects of Sall2 on the pluripotency of ESCs [34, 35]. Hang-drop protocol was applied to form EBs [36, 37], during the differentiation, Sall2 expression increased in WT cells, and the expression of Rex1, Oct4, and Klf4 reduced in both Sall2 KO and WT cells (Fig. 2H, I). However, the meso-endoderm markers Flk1, Gata6, T, and Gata4 significantly decreased in Sall2 KO cells (Fig. 2J). As the differentiated cells in hang-drop EBs were mainly from meso-endoderm lineage, we treated the EBs with retinoic acid (RA), which efficiently induces neuroectoderm differentiation [38, 39]. Upon RA induction, the REX1:GFP fluorescence gradually faded away in both WT and Sall2 KO cells, meanwhile, the neural markers Nestin, Pax6 and Tubb3 significantly upregulated in WT cells, in contrast, in Sall2 KO cells, their expression was still at very low level (Fig. 2K, L). Together, these data indicated that Sall2 deficiency impaired the pluripotency of ESCs.

Sall2 was critical for the neural differentiation of ESCs

Regarding SALL2, mainly expressed in the nervous system during embryo development and Sall2 KO EBs manifested impaired neural differentiation, we next focus on its role in neural differentiation [40]. To this end, we selected the SOX1:GFP ESC line to perform monolayer neural differentiation. Sox1 is a key neural progenitor marker, by monitoring SOX1:GFP expression, the neural differentiation efficiency can be determined conveniently [41, 42]. First, we generated SOX1:GFP Sall2 KO ESC lines by utilizing the same strategy as we did for REX1:GFP Sall2 KO ESCs. The deletion of the Sall2 coding sequence was confirmed by sequencing the Sall2 loci (Supplementary Fig. S3A, B). Then, the mRNA and protein level of Sall2/SALL2 was determined by qRT-PCR, immunostaining, and western blot, which showed deficient Sall2 expression (Fig. 3A, B and Supplementary Fig. S3C). Similarly, pluripotency markers were not significantly affected in SOX1:GFP Sall2 KO ESCs (Fig. 3C, D).

Fig. 3: Sall2 KO inhibited neural differentiation of ESCs.
figure 3

A mRNA expression of Sall2 in SOX1:GFP Sall2 KO ESCs. Relative to Gapdh expression (n = 3, technical replicates). B Expression of SALL2 in SOX1:GFP Sall2 KO and WT ESCs. The cells were immunostained for SALL2. DAPI stained nuclei. Scale bar, 250 μm. C qRT-PCR analysis of pluripotency markers (Oct4, Sox2, Rex1, Klf2, Klf4) in SOX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by an unpaired t-test. **P < 0.01, ns not significant. D Expression of OCT4 in SOX1:GFP Sall2 KO and WT ESCs. The cells were immunostained for OCT4. DAPI stained nuclei. Scale bar, 250 μm. Cells treated with secondary antibodies only served as control. E Phase and fluorescence images of SOX1:GFP Sall2 KO and WT ESCs during monolayer neural differentiation. Scale bar, 250 μm. F Flow cytometry analysis of SOX1:GFP+ cells during monolayer neural differentiation of Sall2 KO and WT ESCs (n = 3, biological replicates). Statistical significance was determined by an unpaired t-test. ****P < 0.0001. G qRT-PCR analysis of Sall2, pluripotency markers (Rex1, Oct4), and neuroectoderm markers (Nestin, Pax6, Tubb3) during monolayer neural differentiation of SOX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by unpaired t-test. **P < 0.01, ***P < 0.001, ****P < 0.0001. H Expression of TUBB3 and SALL2 during monolayer neural differentiation of SOX1:GFP Sall2 KO and WT ESCs at day 8. The cells were immunostained for TUBB3 and SALL2. DAPI stained nuclei. Scale bar, 250 μm. I Western blot analysis of SALL2 and TUBB3 expression during monolayer neural differentiation of SOX1:GFP Sall2 KO and WT ESCs at days 2, 4, 6, and 8. GAPDH served as a loading control. J Flow cytometry analysis of REX1:GFP+ cells in REX1:GFP OE-E1, KO1, and WT ESCs with or without DOX induction during monolayer neural differentiation (n = 3, biological replicates). K qRT-PCR analysis of neural markers (Nestin, Pax6, Tubb3) in REX1:GFP OE-E1, KO1, and WT ESCs with or without DOX induction during monolayer neural differentiation. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between REX1:GFP OE-E1 cells with (E1+) and without (E1−) DOX induction. ***P < 0.001, ****P < 0.0001. L mRNA Expression of E1, E1A Sall2 in REX1:GFP WT ESCs during monolayer neural differentiation. Relative to Gapdh expression (n = 3, technical replicates).

We then performed neural differentiation with SOX1:GFP Sall2 KO ESCs. By observing the GFP signal under a fluorescence microscope, we found the SOX1:GFP+ cells were significantly fewer in Sall2 KO cells than in WT cells (Fig. 3E). We further quantified the dynamic change of SOX1:GFP+ population during differentiation by flow cytometry. As early as day 4, some SOX1:GFP+ cells emerged in WT cells, whereas no GFP signal was observed in Sall2 KO cells. Next, the SOX1:GFP+ population reached the peak on day 6, and was relatively stable from days 6 to 8 in WT cells, while SOX1:GFP still weakly expressed during the neural differentiation in Sall2 KO cells (Fig. 3F).

Meanwhile, we examined the expression of Sall2, pluripotency and lineage marker genes during differentiation. In WT cells, Sall2 mRNA consistently increased until day 6, then went down at day 8, similarly, high SALL2 protein level was detected at day 4 (Fig. 3G, I). The expression of Rex1 and Oct4 quickly lost in both Sall2 KO and WT cells. However, the expression level of Nestin, Pax6, and Tubb3 was not upregulated in Sall2 KO cells, consistent with RA-induced EBs differentiation (Fig. 3G). Also, both immunostaining and western blot confirmed the reduction of TUBB3 expression in Sall2 KO cells at differentiation day 8 (Fig. 3H, I).

To corroborate these findings, we further performed monolayer neural differentiation with REX1:GFP Sall2 KO ESC lines. The REX1:GFP+ population diminished quickly during the first 2 days of differentiation, as revealed by flow cytometry (Supplementary Fig. S3D). Concomitant with SOX1:GFP ESCs, the Sall2 level increased during the differentiation of REX1:GFP ESCs (Supplementary Fig. S3E). While the transcription level of Rex1 and Oct4 reduced in both Sall2 KO and WT cells, again, low expression of Nestin, Pax6, and Tubb3 was detected in Sall2 KO cells (Supplementary Fig. S3F, G). Also, by immunostaining, very fewer TUBB3+ cells were observed in Sall2 KO cells at differentiation day 8 (Supplementary Fig. S3H). Thus, the impaired neural differentiation due to Sall2 deletion was validated in two independent Sall2 KO ESC lines.

E1 isoform of Sall2 restored neural differentiation of Sall2 KO ESCs

To further validate the function of Sall2 in neural differentiation, we designed rescue experiments by transgenic overexpression of Sall2 in REX1:GFP Sall2 KO ESCs. We constructed PB-TRE-E1, PB-TRE-E1A vectors, the expression of these two Sall2 isoforms can be induced by doxycycline (DOX) [10, 43, 44] (Supplementary Fig. S3I). After transfection and colony picking, the integration of transgene was confirmed by genomic PCR and western blot verified transgenic Sall2 expression in the presence of DOX (Supplementary Fig. S3J, K). We named the transgenic Sall2 expressing cells as REX1:GFP OE-E1 and REX1:GFP OE-E1A ESCs, respectively.

We then set up neural differentiation for REX1:GFP WT, Sall2 KO, OE-E1, and OE-E1A ESCs with and without DOX. Flow cytometry analysis showed that during neural differentiation, REX1:GFP+ populations decreased significantly in 2 days in all groups (Fig. 3J and Supplementary Fig. S3L). In terms of neural markers, intriguingly, upon E1 Sall2 expression, the level of Nestin, Pax6, and Tubb3 was comparable to that of WT cells at differentiation day 8 (Fig. 3K), whereas, overexpressing E1A Sall2 did not elevate their expression (Supplementary Fig. S3M). We therefore analyzed the expression of E1 and E1A isoforms during neural differentiation, which showed that E1 level was much higher than E1A (Fig. 3L). That may be one of the reasons that only E1 overexpression can restore defected neural differentiation of Sall2 KO ESCs. Taken together, Sall2 played a critical role during neural differentiation from ESCs and E1 isoform may be the main determinant.

SALL2 was indispensable for the derivation of NSCs from ESCs

SOX1 expression was low in Sall2 KO cells during neural differentiation, and SALL2 was one of the factors that can reprogram differentiated glioma cells into glioma stem cells [14, 15]. During early neural development, NSCs played an important role in neural lineage differentiation [45, 46]. NSCs have the ability to proliferate, migrate, and differentiate into neurons, astrocytes, and oligodendrocytes, providing cell sources for nerve repair [47, 48]. We speculated that Sall2 may participate in the establishment of NSCs. We thus leveraged the protocol for NSCs derivation from ESCs on SOX1:GFP Sall2 KO ESCs.

After 3 days of suspension culture, neural spheres formed, which were then transferred into an adhesive plate. When NSCs migrated out from the neurospheres, the cells were dissociated and plated into an NSC medium for expansion and characterization [49]. NSCs derived from SOX1:GFP WT ESCs displayed elongated morphology whereas the cells from SOX1:GFP Sall2 KO ESCs were more differentiated with large and flat morphology (Fig. 4A). Subsequently, after second passage, the Sall2 KO cells all died while WT NSCs continuously grew and could be passaged; also, the expression of NSC markers, NESTIN and SOX2, confirmed their identity (Fig. 4A, B). To understand this phenotype, we performed immunostaining of NESTIN and SOX2 on cells after the first passage. Surprisingly, no NESTIN and SOX2 expression was observed in Sall2 KO cells, indicating that in the absence of Sall2, NSCs could not be derived from ESCs (Fig. 4C). We then performed qRT-PCR to characterize the passage 1 (P1) of WT and Sall2 KO cells. After Sall2 KO, the expression of Nestin, Sox1, Sox2, and Pax6 were repressed (Fig. 4D). We also tried to derive NSCs from REX1:GFP Sall2 KO ESCs. Similarly, at P1, the Sall2 KO cells lost NESTIN and SOX2 expression, then died after the second passage, while NSCs were successfully derived from REX1:GFP WT ESCs, verified by SOX2 expression and the expression of Nestin, Sox1, Sox2 and Pax6 (Supplementary Fig. S4A–C).

Fig. 4: Sall2 was indispensable for the derivations of NSCs from ESCs.
figure 4

A Phase images of NSCs derivation from SOX1:GFP Sall2 KO and WT ESCs. EBs and Neurospheres, scale bar, 500 μm; P1 and P2, scale bar, 250 μm. B Expression of NESTIN and SOX2 in NSCs (P7) derived from SOX1:GFP WT ESCs. DAPI stained nuclei. Scale bar, 250 μm. C Expression of NESTIN and SOX2 in NSCs (P1) derived from SOX1:GFP Sall2 KO and WT ESCs. The cells were immunostained for NESTIN and SOX2. DAPI stained nuclei. Scale bar, 250 μm. D qRT-PCR analysis of Sall2 and NSC markers (Nestin, Sox1, Sox2, Pax6) in NSCs (P1) derived from SOX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by one-way ANOVA with Tukey’s test. ***P < 0.001, ****P < 0.0001. E Phase images of NSCs derivation from REX1:GFP OE-E1, KO1, and WT ESCs with or without DOX induction. Scale bar, 250 μm. F Expression of SALL2 and SOX2 in NSCs (P1) derived from REX1:GFP OE-E1, KO1, and WT ESCs with or without DOX induction. The cells were immunostained for SALL2 and SOX2. DAPI stained nuclei. Scale bar, 250 μm. G Phase images of NSCs derived from REX1:GFP OE-E1 and WT before and after DOX removal (P2, P3). Scale bar, 250 μm.

In order to further verify the function of SALL2 in NSCs, we conducted rescue experiments by overexpressing E1 Sall2 in REX1:GFP Sall2 KO ESCs. During NSCs derivation, at P1, there were NESTIN and SOX2 positive cells upon DOX induction, which could be passaged steadily (Fig. 4E, F). However, when E1 Sall2 transgene expression was shut off after second passage, the cells gradually died (Fig. 4G). In all, E1 Sall2 played a vital role in the derivation of NSCs from ESCs.

SALL2 deletion impacted NTOs generation

Defects in NT development have been documented in Sall2 KO mice [11], we also showed that Sall2 was essential for NSCs derivation from ESCs. To get insights of Sall2 in NT development, we generated NTOs from ESCs as previously reported [50].

The naïve ESCs were transiently converted to primed epiblast stem cells (EpiSCs), then differentiated to NSCs and self-organized into NTOs. At day 5, the NT like morphology was observed in both WT and SOX1:GFP Sall2 KO ESCs accompanied by some SOX1:GFP+ cells. At day 6, the WT cells formed a long, narrow tube-like structure resembling NT, with SOX1:GFP+ cells clustering in NTOs (Fig. 5A). However, the number of NTOs as well as SOX1:GFP+ cells were fewer in Sall2 KO cells compared with WT cells under the microscope, which was further validated by tracking SOX1:GFP expression with flow cytometry. When measuring SOX1:GFP and Sall2 transcription at different time points (days 4, 5, and 6) during NTOs formation, Sall2 gradually increased and reached a peak at day 4 and then decreased (Fig. 5B, C). Moreover, the level of Sox1, Sox2, Nestin, and Pax6 was considerably lower in Sall2 KO NTOs (Fig. 5D). Additionally, immunofluorescence assay showed the co-localization of SALL2/SOX1:GFP and SOX1:GFP/SOX2 in WT NTOs while fewer positive signals were detected in Sall2 KO NTOs (Fig. 5E). Meanwhile, we collected WT and Sall2 KO NTOs and determined PAX6 protein expression by western blot, which also showed that PAX6 was much lower in Sall2 KO NTOs (Fig. 5F).

Fig. 5: Sall2 impaired NTOs formation from ESCs.
figure 5

A Phase and fluorescence images of NTOs formation from SOX1:GFP Sall2 KO and WT ESCs. The dashed box indicated enlarged images. Scale bar, 500 μm (4–6, 10–12); 250 μm (1–3, 7–9, 13–18). B Flow cytometry analysis of SOX1:GFP+ cells during NTOs formation from SOX1:GFP Sall2 KO and WT ESCs at days 4, 5, 6 (n = 3, biological replicates). Statistical significance was determined by unpaired t-test. ****P < 0.0001. C Sall2 expression during NTOs formation from SOX1:GFP Sall2 KO and WT ESCs at days 2, 4, 6. Relative to Gapdh expression (n = 3, technical replicates). D qRT-PCR analysis of NTO markers (Sox1, Sox2, Nestin, Pax6) during NTOs formation from SOX1:GFP Sall2 KO and WT ESCs at days 2, 4, 6. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by unpaired t-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. E Immunofluorescence of SALL2 and SOX2 expression in SOX1:GFP Sall2 KO and WT NTOs at differentiation day 6. The NTOs were immunostained for SALL2 and SOX2. DAPI stained nuclei. Scale bar, 250 μm. F Western blot analysis of PAX6 and SALL2 expression in SOX1:GFP Sall2 KO3 and WT NTOs at differentiation day 6. GAPDH served as a loading control. G Flow cytometry analysis of SOX1:GFP/SOX2 positive cells in SOX1:GFP Sall2 KO and WT NTOs at differentiation day 6 (n = 3, biological replicates). Statistical significance was determined by unpaired t-test. ****P < 0.0001. H Flow cytometry analysis of SOX1:GFP/PAX6 positive cells in SOX1:GFP Sall2 KO and WT NTOs at differentiation day 6 (n = 3, biological replicates). Statistical significance was determined by unpaired t-test. ****P < 0.0001.

We then dissociated WT and Sall2 KO NTOs into single cells to perform flow cytometry analysis. The percentages of SOX1:GFP/SOX2 and SOX1:GFP/PAX6 positive cells in WT NTOs were significantly higher than those of Sall2 KO NTOs (Fig. 5G, H). Consistently, fewer SOX1:GFP and SOX1:GFP/SALL2 positive cells were detected in Sall2 KO NTOs (Supplementary Fig. S5A).

The capability of generating NTOs was also examined on REX1:GFP Sall2 KO ESCs. In line with SOX1:GFP ESCs, fewer NTOs formed from REX1:GFP Sall2 KO ESCs, and the expression of Sox1, Sox2, and Pax6 decreased as well (Supplementary Fig. S5B). By immunostaining, the expression of SOX2 was much lower in Sall2 KO NTOs as well (Supplementary Fig. S5C). We then induced Sall2 transgene expression during NTOs generation. By qRT-PCR, E1 Sall2 led to increased levels of Sox1 and Pax6 (Supplementary Fig. S5D). Meanwhile, we performed flow cytometry analysis of REX1:GFP Sall2 E1 overexpressing NTOs, which showed increased percentages of SOX2, PAX6, and SOX2/PAX6 positive cells (Supplementary Fig. S5E). These data not only verified the previously reported NTDs in Sall2 KO mice, but also demonstrated the indispensable role of Sall2 in NT development.

Tuba1a mediated SALL2 regulated neural differentiation

SALL2 was well known as a transcription factor to modulate gene transcription [51]. To figure out the potential downstream targets of SALL2 in regulating early neural differentiation, we constructed a hemagglutinin (HA)-tagged Sall2 (HASall2) ESC line to perform ChIP-seq. ChIP-seq profiles obtained with HA-tag antibody were distinct from input samples around the transcription start site (TSS) (Fig. 6A), validating the efficiency of ChIP. In total, we identified 131 SALL2 targets, with their genomic distribution mainly located at promoters (41.98%), intergenic (12.21%), exon (9.16%), transcription termination site (TTS, 9.16%), and small nuclear RNA (snRNA, 7.63%) (Fig. 6B). Gene Ontology (GO) analysis revealed that SALL2 targets were enriched in chromatin or gene regulatory functions (nuclear chromatin, protein−DNA complex, nucleosome, DNA packaging complex, and RNA polymerase complex), regulation of translation (negative regulation of translation, cytosolic ribosome, cytosolic large ribosomal subunit). Other than the cellular machinery functions, GO terms related to nervous systems, such as myelin sheath and presynaptic cytosol, were also enriched (Fig. 6C). Next, we overlapped the SALL2 targets with genes related to neural differentiation, and found 14 potential candidate genes (Apoe, Ddc, Fus, Gas5, Lef1, Malat1, Marcksl1, Myc, Pim1, Polr2a, Snhg1, Snord118, Terc, Tuba1a) (Fig. 6D). Further literature review showed that tubulin alpha 1a (Tuba1a) was specifically expressed in the central nervous system (CNS) and peripheral nervous system (PNS) of mouse embryo at E13.5. Knock-down of Tuba1a reduced the production of neural progenitors, while overexpression of WT TUBA1A could promote neurogenesis [52, 53]. PIM1 was a cell cycle regulator that could be recruited as a lineage determinant by enhancer. Also, as one of the downstream targets of STAT3, PIM1 supported self-renewal and inhibited endoderm differentiation of ESCs [54, 55]. These studies suggested that Tuba1a and Pim1 might play important roles in neurogenesis.

Fig. 6: SALL2 regulated neural differentiation of ESCs through Tuba1a.
figure 6

A Metagene showing ChIP-seq signal profiles of HA-Sall2 (pink) and Input (gray) around the TSS regions (±3.0 Kb). Each group has two replicates. B The genomic distribution of HA-Sall2 binding regions identified by HA-Sall2 ChIP-seq. C GO analysis of SALL2 targets identified by HA-Sall2 ChIP-seq. D Venn diagram of overlaps of SALL2 binding genes (Sall2 targets) with genes related to neural differentiation. E Genome browser view of HA-Sall2 ChIP-seq analysis in HA-Sall2 (pink) and Input (gray) samples at Pim1 and Tuba1a gene loci. The promoter regions of Pim1 and Tuba1a were marked by dashed boxes. F ChIP-qPCR analysis of Pim1 and Tuba1a. The binding of SALL2 to the promoter region of Pim1 (Pim1 prom.) and Tuba1a (Tuba1a prom.) was measured by ChIP-qPCR, and the intergenic nonbinding region was amplified as negative control (Neg.). Relative to 5% input (n = 3, technical replicates). Statistical significance was determined by unpaired t-test. **P < 0.01. G Expression of Tuba1a during monolayer neural differentiation of SOX1:GFP Sall2 KO and WT ESCs. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by unpaired t-test. ***P < 0.0001. H Expression of Tuba1a during monolayer neural differentiation of REX1:GFP OE-E1, Sall2 KO1, and WT ESCs with or without DOX induction. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between REX1:GFP OE-E1 cells with (E1+) and without (E1−) DOX induction. ****P < 0.0001. I Flow cytometry analysis of SOX1:GFP during monolayer neural differentiation of SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between SOX1:GFP OE-Tuba1a cells with (Tuba1a+) and without (Tuba1a−) DOX induction. ****P < 0.0001. J Expression of neural markers (Nestin, Pax6, Tubb3) during monolayer neural differentiation of SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction. Relative to Gapdh expression (n = 3, technical replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between SOX1:GFP OE-Tuba1a cells with (Tuba1a+) and without (Tuba1a−) DOX induction. ****P < 0.0001. K Immunofluorescence of TUBB3 at day 8 during monolayer neural differentiation of SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction. The cells were immunostained for TUBB3. DAPI stained nuclei. Scale bar, 250 μm. L Phase images of NSCs derived from SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction. Scale bar, 250 μm. M Flow cytometry analysis of SOX1:GFP+ cells during NTOs formation from SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction at days 4, 5, 6 (n = 3, biological replicates). Statistical significance was determined by two-way ANOVA with Tukey’s test, indicating significant changes between SOX1:GFP OE-Tuba1a cells with (Tuba1a+) and without (Tuba1a−) DOX induction. ****P < 0.0001. N Flow cytometry analysis of SOX1:GFP/SOX2, SOX1:GFP/PAX6 positive cells at day 6 of NTOs formation from SOX1:GFP OE-Tuba1a, Sall2 KO3, and WT ESCs with or without DOX induction (n = 3, biological replicates). Statistical significance was determined by one-way ANOVA with Tukey’s test, indicating significant changes between SOX1:GFP OE-Tuba1a cells with (Tuba1a+) and without (Tuba1a-) DOX induction. ****P < 0.0001.

In order to validate whether SALL2 regulated Tuba1a and Pim1 during neural differentiation, we first analyzed our ChIP-seq data and found that SALL2 bound to the promoter regions of Tuba1a and Pim1 (Fig. 6E), which was confirmed by ChIP-qPCR (Fig. 6F). We next investigated mRNA expression of Tuba1a and Pim1 during monolayer neural differentiation by qRT-PCR. Sall2 KO inhibited the expression of Tuba1a compared with the WT cells, but it barely affected Pim1 expression (Fig. 6G and Supplementary Fig. S6A). After E1 Sall2 induction, the expression of Tuba1a but not Pim1 could be restored (Fig. 6H and Supplementary Fig. S6B). Therefore, we selected Tuba1a for further validation.

Next, we constructed DOX inducible Tuba1a expressing cell lines in SOX1:GFP Sall2 KO ESCs (Supplementary Fig. S6C, D). We then examined Tuba1a expression during monolayer neural differentiation, which showed a similar level of Tuba1a to WT cells after DOX induction in Sall2 KO cells (Supplementary Fig. S6E). Further flow cytometry analysis at different time points revealed that SOX1:GFP expression was also restored in the presence of Tuba1a (Fig. 6I). In addition, by qRT-PCR analysis, we found that the expression of Nestin, Pax6, and Tubb3 inhibited by Sall2 KO, was rescued by Tuba1a overexpression (Fig. 6J). We then collected differentiation day 8 cells to perform immunofluorescence staining of TUBB3, which displayed a comparable level with that of WT cells when Tuba1a was overexpressed (Fig. 6K).

Since Sall2 KO ESCs failed to produce NSCs and decreased the NTOs forming efficiency, we determined the NSCs derivation and NTOs generation in Tuba1a overexpressing Sall2 KO ESCs. After DOX induction, SOX2+ cells were detected in P1 NSCs, which could be further passaged (Fig. 6L and Supplementary Fig. S6F). For NTOs formation, in the presence of TUBA1A, the efficiency was significantly improved in SOX1:GFP Sall2 KO cells, evidenced by morphology and flow cytometry analysis of SOX1:GFP (Fig. 6M and Supplementary Fig. S6G, H). In addition, the SOX1:GFP/SOX2 and SOX1:GFP/PAX6 positive cells were recovered to the levels of WT NTOs (Fig. 6N). Taken together, Tuba1a mediated SALL2 in regulating NSCs derivation and NTOs formation from ESCs.