*Article* **Activation of Sonic Hedgehog Signaling Promotes Differentiation of Cortical Layer 4 Neurons via Regulation of Their Cell Positioning**

**Koji Oishi 1,2,3,\* , Kazunori Nakajima <sup>3</sup> and Jun Motoyama <sup>4</sup>**


**Abstract:** Neuronal subtypes in the mammalian cerebral cortex are determined by both intrinsic and extrinsic mechanisms during development. However, the extrinsic cues that are involved in this process remain largely unknown. Here, we investigated the role of sonic hedgehog (Shh) in glutamatergic cortical subtype specification. We found that E14.5-born, but not E15.5-born, neurons with elevated Shh expression frequently differentiated into layer 4 subtypes as judged by the cell positioning and molecular identity. We further found that this effect was achieved indirectly through the regulation of cell positioning rather than the direct activation of layer 4 differentiation programs. Together, we provided evidence that Shh, an extrinsic factor, plays an important role in the specification of cortical superficial layer subtypes.

**Keywords:** cerebral cortex; cortical subtype; layer; sonic hedgehog

## **1. Introduction**

The mammalian neocortex consists of two major types of neurons according to their usage of neurotransmitters. The majority of them are glutamatergic excitatory neurons derived from the dorsal telencephalon [1,2]. The other population is GABAergic inhibitory interneurons produced by the ventral telencephalon, which consist of 20–30% of all cortical neurons [3,4]. Both excitatory and inhibitory neurons can be further divided into a wider variety of types, which are recognized as neuronal subtypes, according to criteria other than neurotransmitters, such as morphology and gene expression profiles [5,6].

In the dorsal telencephalon, almost all subtypes of glutamatergic neurons are produced by common progenitor cells or neural progenitor/stem cells (NPCs) residing in the dorsal ventricular zone (VZ), which also give rise to glial cells including astrocytes and oligodendrocytes during late embryonic and early postnatal periods [7,8]. In the neurogenic period, NPCs sequentially generate different subtypes of neurons, which ultimately align in the cortical plate from the bottom to top parallelly to the pial surface and form cortical layers comprising anatomically distinguishable 6 layers [9]. Recent studies using unbiased approaches suggested that there are dozens of recognizable subtypes of cortical glutamatergic neurons in the motor area [10].

A plethora of efforts in recent years has provided evidence that determination of cortical subtypes is intrinsically regulated by specific transcription factors, such as Fezf2 [11–13], Bcl11b (aka Ctip2) [14], Rorb [15,16], Tbr1 [17], Brn1/2 [16], and Satb2 [18,19]. As subtypespecific features are established under the control of these factors, these factors are often called master regulators or subtype determinants. Given that NPCs sequentially generate different subtypes, temporal changes of NPC potentials, such as expression changes of

**Citation:** Oishi, K.; Nakajima, K.; Motoyama, J. Activation of Sonic Hedgehog Signaling Promotes Differentiation of Cortical Layer 4 Neurons via Regulation of Their Cell Positioning. *J. Dev. Biol.* **2022**, *10*, 50. https://doi.org/10.3390/ jdb10040050

Academic Editors: Hideyo Ohuchi, Tsutomu Nohno and Simon J. Conway

Received: 17 October 2022 Accepted: 23 November 2022 Published: 25 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

subtype determinants, which actually occur in Drosophila NPCs [20], have been postulated [21]. However, many subtype determinants start to be expressed in postmitotic neurons, although some are also expressed in NPCs [12,22,23], leaving the question of what mechanism regulates the sequential generation of different subtypes from NPCs.

Not only intrinsic factors but also extrinsic factors, such as extracellular environments, play important roles in determining cortical neuronal subtypes. We and others have provided evidence that the specification of L4 neurons is controlled by extracellular environments [24–27]. Moreover, fate regulation by extracellular environments could occur more generally than previously thought [25,28]. However, given that these results were obtained mostly from transplantation experiments, molecular mechanisms that underpin this notion remain to be determined.

As such environmental cues, sonic hedgehog (Shh) is a strong candidate. On top of its role in the patterning formation and resulting specification of ventral structures in the CNS, Shh signaling regulates a wide variety of biological processes [29], such as the proliferation of intermediate progenitor cells [30], induction and expansion of outer radial glial cells that compose the outer SVZ, a progenitor pool commonly observed in the gyrencephalic neocortex [31], and gliogenesis [32–34]. However, the role of Shh signaling in the specification of cortical subtypes has remained unknown; cortical subtype phenotypes observed in Shh signaling mutants are mostly attributable to alteration in dorsoventral patterning and progenitor proliferation [35]. Given that specification of cortical subtypes, especially the L4 subtype, utilizes environmental cues [24–27], we investigated the role of Shh signaling in L4 subtype generation.

#### **2. Materials and Methods**

#### *2.1. Mice*

Pregnant ICR mice were purchased from Japan SLC (Shimizu laboratory supplies, Kyoto, Japan). The morning of vaginal plug detection was designated as E0.5. Mice were maintained on a 12 h light/dark cycle with free access to food and water. All experiments were approved by the Doshisha University Animal Experiment Committee and conducted in accordance with guidelines established by the Doshisha University Ethics Review Committee.

#### *2.2. In Utero Electroporation*

Pregnant mice were deeply anaesthetized, and in utero electroporation was carried out as described previously [36]. In brief, an empty or Shh-encoding plasmid vector together with the pCAGGS vector carrying the enhanced GFP cDNA (1 mg/mL) was injected into the lateral ventricle of the intrauterine embryos, and electronic pulses (33 V, 50 ms, 4 times) were applied using an electroporator (CUY21 EDIT II, BEX, Tokyo, Japan) with a forceps-type electrode (CUY650P5, Nepagene, Chiba, Japan).

For expression of Shh, the gene-encoding, full-length mouse Shh obtained from mouse cDNA was cloned into the plasmid vector pCAGGS or pEF.

#### *2.3. Immunohistochemistry*

Brains removed from embryos and pups were fixed for 1 h in phosphate-buffered saline (PBS) containing 4% PFA (*w*/*v*), incubated overnight at 4 ◦C with 20% sucrose in PBS (*w*/*v*), embedded in OCT compound (Sakura Finetek, Torrance, CA, USA), and sectioned with a cryostat to obtain 14 µm-thick coronal sections.

For primary antibodies, we used chick antibody to EGFP (Abcam, Cambridge, UK, ab13970), mouse antibody to Rorb (Perseus Proteomics, Tokyo, Japan, N7927), goat antibody to Lhx2 (Santa Cruz, sc-19344), mouse antibody to Brn2 (Santa Cruz, Dallas, TX, USA, sc-393324), and rabbit antibody to Shh (Santa Cruz, c-9024). For some cases, antigen retrieval was performed by incubating the sections for 20 min at 80 ◦C in 0.01 M sodium citrate buffer (pH 6.0). Because EGFP fluorescence disappeared by the antigen retrieval treatment, EGFP was immunostained with chick antibody against EGFP for revisualization. Immune

complexes were detected with Alexa Flour-conjugated secondary antibodies (Invitrogen, Waltham, MA, USA). For nuclear staining, 1 µg/mL Hoechst 33,342 (Invitrogen) was used. Images were acquired using a confocal microscope (SP8, Leica, Wetzlar, Germany).

#### *2.4. Quantitative Analysis of the Cell Positioning*

To quantify the pattern of migration, the position of each GFP-positive cell relative to the total distance from the bottom of L4 or the subplate to the outer edge of the cortical plate (pial surface) was measured using the Image J software (National Institutes of Health shareware program), followed by sorting into 5 or 10 bins.

#### *2.5. Statistical Analysis*

Unless indicated otherwise, data are represented as means ± SEM of values from at least three embryos. For quantification of in vivo cell counting, all EGFP-positive cells were counted in the regions where rostrocaudal and mediolateral levels were carefully matched between animals. A representative section per electroporated embryo was quantified. The number of embryos analyzed was indicated in the figure legends. For two-group comparisons with equal variance as determined by the *F*-test, an unpaired Student's *t*-test was used. Welch's correction was used for unpaired t-tests of normally distributed data with unequal variance. Differences between groups were considered to be significant at *p* < 0.05. Each *p*-value was stated in figures or figure legends.

#### **3. Results**

According to the previous implication that the specification of L4 neurons may require environmental cues [24–27], we first investigated the role of Shh signaling in L4 subtype generation in mice. We chose to manipulate Shh signaling by introducing a Shh expression vector in NPCs at embryonic day (E) 14.5 by in utero electroporation (IUE) [36] because NPCs at this stage give rise to both L4 and L2/3 neurons. In fact, 55.6% of the EGFP-labeled cells at E14.5 were located in the upper part of L4 at postnatal day (P) P7, while 44.4% of them were located in the lower part of L2/3 in the controls, where only an EGFP vector was introduced (Figure 1A,E). On the other hand, when an Shh expression vector together with an EGFP expression vector was introduced, the electroporated cells were located more in L4 (79.3% in the upper part of L4 and 20.7% in the lower part of L2/3, Figure 1B,E). The expression of ectopic Shh was detected around the EGFP-positive cells, suggesting autocrine action (or short distance effect) of ectopic Shh (Figure 1F,G). Accordingly, we did not observe obvious differences in the overall thickness of L2/3 and L4 (Figures 1 and 2) although we cannot rule out the possibility of a non-cell-autonomous effect.

**Figure 1.** Shh overexpression at E14.5 increases the percentage of neurons in L4 at P7. (**A**,**B**). Empty (Control, (**A**)) or Shh expression vector (Shh, (**B**)) together with a GFP vector was electroporated into E14.5 brains, and then P7 brains were analyzed. The sections were counterstained with Hoechst (magenta). The boxed regions are shown at higher magnification in (**C**,**D**). (**E**). The percentages of the cells in L2/3 and L4 were determined in each condition. Quantitative data are presented (*n* = 3 for each group). \*\* *p* < 0.01. Note that the GFP-positive cells with Shh expression vector are located more in L4 than the control cells. (**F**,**G**). Expression of Shh was shown in the neurons treated as in (**A**,**B**). (Scale bars: 200 µm in (**A**,**C**,**F**)). **Figure 1.** Shh overexpression at E14.5 increases the percentage of neurons in L4 at P7. (**A**,**B**). Empty (Control, (**A**)) or Shh expression vector (Shh, (**B**)) together with a GFP vector was electroporated into E14.5 brains, and then P7 brains were analyzed. The sections were counterstained with Hoechst (magenta). The boxed regions are shown at higher magnification in (**C**,**D**). (**E**). The percentages of the cells in L2/3 and L4 were determined in each condition. Quantitative data are presented (*n* = 3 for each group). \*\* *<sup>p</sup>* < 0.01. Note that the GFP-positive cells with Shh expression vector are located morein L4 than the control cells. (**F**,**G**). Expression of Shh was shown in the neurons treated as in (**A**,**B**). (Scale bars: 200 µm in (**A**,**C**,**F**)).

The observed phenotype could be attributable not only to migration or positioning

failure but also to cell identity alteration. To distinguish these possibilities, we investigated the neuronal morphology, which often represents subtype-specific features [1,37], upon ectopic Shh expression. Magnified images showed that Shh overexpression decreased the neurons with a pyramidal shape that harbors an apical dendrite, a feature of L2/3 neurons, compared to those with a nonpyramidal shape, a feature of L4 neurons (Figure 1C,D) [15,16], suggesting that elevated Shh signaling modulates not only the positioning of superficial layer neurons but also their fate. These results suggest that elevated Shh signals enhance the generation of L4 neurons. To further investigate the identity of the neurons with elevated Shh expression, we The observed phenotype could be attributable not only to migration or positioning failure but also to cell identity alteration. To distinguish these possibilities, we investigated the neuronal morphology, which often represents subtype-specific features [1,37], upon ectopic Shh expression. Magnified images showed that Shh overexpression decreased the neurons with a pyramidal shape that harbors an apical dendrite, a feature of L2/3 neurons, compared to those with a nonpyramidal shape, a feature of L4 neurons (Figure 1C,D) [15,16], suggesting that elevated Shh signaling modulates not only the positioning of superficial layer neurons but also their fate. These results suggest that elevated Shh signals enhance the generation of L4 neurons.

analyzed the expression of molecular markers that distinguish L4 and L2/3 neurons. An immunohistochemical analysis for Rorb, an L4 subtype marker [38,39], revealed that 28.6% of the control cells electroporated at E14.5 differentiated into Rorb-positive neurons at P7 (Figure 2A,C). In contrast, 72.3% of the Shh-overexpressing cells became Rorb-positive (Figure 2B,C). Moreover, the percentage of neurons that expressed L2/3 markers, such as Lhx2 [40] and Brn2 [23,41,42], was decreased by ectopic Shh expression (Lhx2, 80.2% in control, 32.7% in Shh overexpressed (Figure 2E,F); Brn2, 72.3% in control, 50.2% in Shh overexpressed (Figure 2G–I)). These results suggest that elevated Shh signaling promotes L4 generation, at the expense of L2/3 neurons, at molecular levels. This observation implied a rather surprising scenario, in which high levels of Shh signaling reversed the sequence of subtype specification of cortical neurons (L6- > L5- > L4- > L2/3) [8,9]; elevated To further investigate the identity of the neurons with elevated Shh expression, we analyzed the expression of molecular markers that distinguish L4 and L2/3 neurons. An immunohistochemical analysis for Rorb, an L4 subtype marker [38,39], revealed that 28.6% of the control cells electroporated at E14.5 differentiated into Rorb-positive neurons at P7 (Figure 2A,C). In contrast, 72.3% of the Shh-overexpressing cells became Rorb-positive (Figure 2B,C). Moreover, the percentage of neurons that expressed L2/3 markers, such as Lhx2 [40] and Brn2 [23,41,42], was decreased by ectopic Shh expression (Lhx2, 80.2% in control, 32.7% in Shh overexpressed (Figure 2E,F); Brn2, 72.3% in control, 50.2% in Shh overexpressed (Figure 2G–I)). These results suggest that elevated Shh signaling promotes L4 generation, at the expense of L2/3 neurons, at molecular levels. This observation implied a rather surprising scenario, in which high levels of Shh signaling reversed the sequence of

subtype specification of cortical neurons (L6- > L5- > L4- > L2/3) [8,9]; elevated Shh can change the ultimate identity of neurons that are destined to become L2/3 neurons into an L4 fate, an earlier-born subtype than L2/3 subtypes. Shh can change the ultimate identity of neurons that are destined to become L2/3 neurons into an L4 fate, an earlier-born subtype than L2/3 subtypes.

*J. Dev. Biol.* **2022**, *10*, x FOR PEER REVIEW 5 of 12

**Figure 2.** Shh-overexpressing cells at E14.5 acquire L4 characteristics. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E14.5 brains, and then P7 brains were analyzed. The sections were immunostained for Rorb (**A**,**B**), Lhx2 (**D**,**E**), and Brn2 (**G**,**H**). The results of quantitative analysis for Rorb (**C**), Lhx2 (**F**), and Brn2 (**I**) are presented (*n* = 3 for each group). \*\* *p* < 0.01, \* *p* < 0.05. Note that Shh-overexpressing cells acquired expression of Rorb, but lost expression of Lhx2 and Brn2. (Scale bar: 200 µm in (**A**)). **Figure 2.** Shh-overexpressing cells at E14.5 acquire L4 characteristics. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E14.5 brains, and then P7brains were analyzed. The sections were immunostained for Rorb (**A**,**B**), Lhx2 (**D**,**E**), and Brn2 (**G**,**H**). The results of quantitative analysis for Rorb (**C**), Lhx2 (**F**), and Brn2 (**I**) are presented (*n* = 3 for each group). \*\* *p* < 0.01, \* *p* < 0.05. Note that Shh-overexpressing cells acquired expression of Rorb, but lost expression of Lhx2 and Brn2. (Scale bar: 200 µm in (**A**)).

To test the possibility of juvenilization of NPCs by high levels of Shh signaling, we overexpressed Shh in later NPCs (IUE at E15.5), which produced predominantly L2/3 neurons and analyzed if they produced earlier-born neurons, such as L4 neurons. Cell positioning analysis showed that Shh overexpression at E15.5 did not change the ultimate positioning of the electroporated cells (Figure 3A–C). In addition, an immunohistochemical analysis for Rorb revealed that only a small fraction of E15.5-electroporated neurons expressed Rorb with or without the ectopic expression of Shh (Figure 3D–F). These results indicate essentially no generation of L4 neurons from E15.5-electroporated cells even in high levels of Shh signaling, suggesting that Shh does not generally regulate the temporal production of different cortical subtypes. In contrast, Shh would play a role in the demarcation of L2/3 and L4 neurons in a rather specific temporal manner. To test the possibility of juvenilization of NPCs by high levels of Shh signaling, we overexpressed Shh in later NPCs (IUE at E15.5), which produced predominantly L2/3 neurons and analyzed if they produced earlier-born neurons, such as L4 neurons. Cell positioning analysis showed that Shh overexpression at E15.5 did not change the ultimate positioning of the electroporated cells (Figure 3A–C). In addition, an immunohistochemical analysis for Rorb revealed that only a small fraction of E15.5-electroporated neurons expressed Rorb with or without the ectopic expression of Shh (Figure 3D–F). These results indicate essentially no generation of L4 neurons from E15.5-electroporated cells even in high levels of Shh signaling, suggesting that Shh does not generally regulate the temporal production of different cortical subtypes. In contrast, Shh would play a role in the demarcation of L2/3 and L4 neurons in a rather specific temporal manner.

**Figure 3.** Shh overexpression at E15.5 does not alter cell positioning and Rorb expression. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E15.5 brains, and then P8 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of L4 to the outer edge of the cortical plate was measured, followed by sorting into 5 bins (Control, *n* = 3; Shh, *n* = 4). (**F**). The result of quantitative analysis for Rorb is presented (Control, *n* = 3; Shh, *n* = 4). (Scale bars: 200 µm in (**A**,**D**)). **Figure 3.** Shh overexpression at E15.5 does not alter cell positioning and Rorb expression. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E15.5 brains, and then P8 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of L4 to the outer edge of the cortical plate was measured, followed by sorting into 5 bins (Control, *n* = 3; Shh, *n* = 4). (**F**). The result of quantitative analysis for Rorb is presented (Control, *n* = 3; Shh, *n* = 4). (Scale bars: 200 µm in (**A**,**D**)).

As one of the mechanisms that determines L2/3 and L4 identity, we previously reported a cell position-dependent model, where L2/3 and L4 differentiation occurs along the ultimate positioning of the neurons in the superficial layer despite their birthdates [24]. Therefore, we hypothesized that Shh controls the cell positioning of E14.5-generated neurons. In this scenario, neurons that receive high Shh signals position the lower part of the superficial layer, where further differentiation processes occur. We then asked if Shh regulates the positioning of neurons before affecting the cell As one of the mechanisms that determines L2/3 and L4 identity, we previously reported a cell position-dependent model, where L2/3 and L4 differentiation occurs along the ultimate positioning of the neurons in the superficial layer despite their birthdates [24]. Therefore, we hypothesized that Shh controls the cell positioning of E14.5-generated neurons. In this scenario, neurons that receive high Shh signals position the lower part of the superficial layer, where further differentiation processes occur.

differentiation status. We examined the brains at E18.5, when most of the E14.5-electroporated cells had reached the pial side of the cortical plate and started maturating. We found that Shh-overexpressing cells were located in deeper regions than the control cells at this early time point (Figure 4A–C). Since normal 'future' L4 neurons (that are destined to become L4 neurons) reach the pia surface earlier than future L2/3 neurons and later align deeper regions, this observation suggested that cell positioning/migration was already affected at this stage. On the other hand, the expression of Rorb was not detected in the electroporated neurons even with or without the ectopic expression of Shh (Figure 4D,E). These results suggest that upregulated Shh signaling can control the positioning of neurons before affecting their subtype identity. We then asked if Shh regulates the positioning of neurons before affecting the cell differentiation status. We examined the brains at E18.5, when most of the E14.5-electroporated cells had reached the pial side of the cortical plate and started maturating. We found that Shh-overexpressing cells were located in deeper regions than the control cells at this early time point (Figure 4A–C). Since normal 'future' L4 neurons (that are destined to become L4 neurons) reach the pia surface earlier than future L2/3 neurons and later align deeper regions, this observation suggested that cell positioning/migration was already affected at this stage. On the other hand, the expression of Rorb was not detected in the electroporated neurons even with or without the ectopic expression of Shh (Figure 4D,E). These results suggest that upregulated Shh signaling can control the positioning of neurons before affecting their subtype identity.

**Figure 4.** Upregulated Shh signaling regulates cell positioning without affecting an L4 marker expression. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E14.5 brains, and then E18.5 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). The asterisk shows expression of Rorb in developing L5 neurons [16]. (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of the subplate to the outer edge of the cortical plate was measured, followed by sorting into 10 bins (*n* = 4 for each group). \* *p* < 0.05. (Scale bars: 200 µm in (**A**,**D**)). **Figure 4.** Upregulated Shh signaling regulates cell positioning without affecting an L4 marker expression. Empty (Control) or Shh expression vector (Shh) together with a GFP vector was electroporated into E14.5 brains, and then E18.5 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). The asterisk shows expression of Rorb in developing L5 neurons [16]. (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of the subplate to the outer edge of the cortical plate was measured, followed by sorting into 10 bins (*n* = 4 for each group). \* *p* < 0.05. (Scale bars: 200 µm in (**A**,**D**)).

The proposed mechanism, in which Shh promotes L4 subtype generation via neuronal positioning, predicts that the neurons that fail to position the lower part of the superficial layer do not differentiate L4 neurons even if they receive high Shh signaling. To directly test this possibility, we tried to reposition the Shh-overexpressing cells back to the upper part of the superficial layer by the knockdown of Pcdh20, which changed the positioning of future L4 neurons into more upper regions without affecting neuronal migration and early subtype specification [24]. We found that Pcdh20 knockdown was able to reposition Shh-overexpressing neurons back to more superficial regions (Figure 5A–C). We then investigated the expression of Rorb in these cells. Rorb staining revealed that 74.6% of Shh-overexpressing neurons were positive for Rorb, but this percentage was reduced to 18.9% by simultaneous Pcdh20 knockdown (Figure 5D–F). These results strongly support the notion that high levels of Shh signaling promote the specification of an L4 subtype via cell positioning. The proposed mechanism, in which Shh promotes L4 subtype generation via neuronal positioning, predicts that the neurons that fail to position the lower part of the superficial layer do not differentiate L4 neurons even if they receive high Shh signaling. To directly test this possibility, we tried to reposition the Shh-overexpressing cells back to the upper part of the superficial layer by the knockdown of Pcdh20, which changed the positioning of future L4 neurons into more upper regions without affecting neuronal migration and early subtype specification [24]. We found that Pcdh20 knockdown was able to reposition Shh-overexpressing neurons back to more superficial regions (Figure 5A–C). We then investigated the expression of Rorb in these cells. Rorb staining revealed that 74.6% of Shh-overexpressing neurons were positive for Rorb, but this percentage was reduced to 18.9% by simultaneous Pcdh20 knockdown (Figure 5D–F). These results strongly support the notion that high levels of Shh signaling promote the specification of an L4 subtype via cell positioning.

**Figure 5.** Shh overexpression promotes L4 fate acquisition in a cell positioning-dependent manner. Shh expression vector together with a control shRNA (Shh+shCon) or Pcdh20 shRNA vector (Shh+shPcdh20) was electroporated into E14.5 brains, and then P7 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of L4 to the outer edge of the cortical plate was measured, followed by sorting into 5 bins in each condition (*n* = 4 for each group). (**F**). The result of quantitative analysis for Rorb is presented (*n* = 4 for each group). \* *p* < 0.05, \*\*\* *p* < 0.001. (Scale bars: 200 µm in (**A**,**D**)). **Figure 5.** Shh overexpression promotes L4 fate acquisition in a cell positioning-dependent manner. Shh expression vector together with a control shRNA (Shh+shCon) or Pcdh20 shRNA vector (Shh+shPcdh20) was electroporated into E14.5 brains, and then P7 brains were analyzed. The sections were counterstained with Hoechst (**A**,**B**) or immunostained for Rorb (**D**,**E**). (**C**). Quantitative data of cell positioning are presented. The position of each GFP-positive cell relative to the total distance from the bottom of L4 to the outer edge of the cortical plate was measured, followed by sorting into 5 bins in each condition (*n* = 4 for each group). (**F**). The result of quantitative analysis for Rorb is presented (*n* = 4 for each group). \* *p* < 0.05, \*\*\* *p* < 0.001. (Scale bars: 200 µm in (**A**,**D**)).

#### **4. Discussion 4. Discussion**

In this study, we found a potential role of Shh signaling in the generation of L4 subtypes of the mouse cortical plate. Shh signaling appeared not to directly activate L4 specification programs but controlled the positioning of a subset of superficial layer neurons, thereby leading to the ultimate specification of L4 subtypes. A similar regulation of cell positioning by Shh signaling was reported previously in the chick optic tectum [43], suggesting a wider role of Shh in cell positioning/distribution. In this study, we found a potential role of Shh signaling in the generation of L4 subtypes of the mouse cortical plate. Shh signaling appeared not to directly activate L4 specification programs but controlled the positioning of a subset of superficial layer neurons, thereby leading to the ultimate specification of L4 subtypes. A similar regulation of cell positioning by Shh signaling was reported previously in the chick optic tectum [43], suggesting a wider role of Shh in cell positioning/distribution.

Shh has been shown to play roles in the proliferation and cell cycle control of progenitor cells in both positive and negative ways during neurogenesis [29]. We previously reported that ectopic expression of Shh in developing NPCs resulted in an increased proliferation of intermediate progenitor cells [30]. As the method used in the current study is similar to that in the previous one, high levels of Shh signaling might have also increased the proliferation rate of intermediate progenitor cells in the present study. However, although increased proliferation may increase the generation of later-born subtypes, altered proliferation did not account for the observed phenotypes that high levels of Shh signaling led to the generation of 'earlier-born' subtypes than the control. Shh has been shown to play roles in the proliferation and cell cycle control of progenitor cells in both positive and negative ways during neurogenesis [29]. We previously reported that ectopic expression of Shh in developing NPCs resulted in an increased proliferation of intermediate progenitor cells [30]. As the method used in the current study is similar to that in the previous one, high levels of Shh signaling might have also increased the proliferation rate of intermediate progenitor cells in the present study. However, although increased proliferation may increase the generation of later-born subtypes, altered proliferation did not account for the observed phenotypes that high levels of Shh signaling led to the generation of 'earlier-born' subtypes than the control.

What downstream effectors play a role in this type of subtype specification? Upon binding to its receptors, Shh influences a wide variety of signal transduction pathways including the activation of the transcription factor Gli1 [44]. It is well studied that Shh determines an oligodendrocyte fate through Gli1-dependent transcriptional regulation [45,46]. Gli1 directly upregulates Olig2, a master regulator of oligodendrocytes, allowing NPCs to differentiate into oligodendrocytes [45]. Thus, a similar mechanism, by which Shh determines the L4 subtype via direct transcriptional regulation, such as upregulation of L4 fate determinants, is conceivable. However, we are not in favor of this hypothesis due to mainly three reasons. First, we did not observe a premature expression of Rorb in early time points; if Shh directly activated the L4 specification program, the premature induction of downstream targets would be predicted. Second, the repositioning of Shhoverexpressing neurons to the more pial side in the superficial layer canceled the What downstream effectors play a role in this type of subtype specification? Upon binding to its receptors, Shh influences a wide variety of signal transduction pathways including the activation of the transcription factor Gli1 [44]. It is well studied that Shh determines an oligodendrocyte fate through Gli1-dependent transcriptional regulation [45,46]. Gli1 directly upregulates Olig2, a master regulator of oligodendrocytes, allowing NPCs to differentiate into oligodendrocytes [45]. Thus, a similar mechanism, by which Shh determines the L4 subtype via direct transcriptional regulation, such as upregulation of L4 fate determinants, is conceivable. However, we are not in favor of this hypothesis due to mainly three reasons. First, we did not observe a premature expression of Rorb in early time points; if Shh directly activated the L4 specification program, the premature induction of downstream targets would be predicted. Second, the repositioning of Shh-overexpressing neurons to the more pial side in the superficial layer canceled the expression of Rorb even in the presence of

Shh. Third, Gli1 transcriptional activation, a canonical downstream target of the Shh–Ptch1 axis [29,44], was hardly detected in the developing cortical neurons [47]. Accordingly, we did not detect *Gli1* and *Ptch1* mRNAs even in Shh-overexpressing neurons. Instead, another Shh receptor, Boc, which activates noncanonical Shh pathways, was strongly expressed [48]. These observations suggest that Shh may indirectly regulate the generation of L4 subtypes.

It remains to be determined what kind of intracellular events Shh signaling regulates to control cell positioning. A possible downstream is calcium signaling [49], which has been shown to play a role in the regulation of neuronal migration [50,51]. In addition, we recently found that Shh can activate calcium signaling (J.M. unpublished observation) [52], leading to a hypothesis that the Shh–calcium axis controls neuronal migration and positioning.

It is also to be determined how Shh specifically regulates L4 development. Shh may act as a limiting factor for immature neurons to be located in the future L4 region (bottom of the superficial layer); the amount of Shh is not abundant so that only a part of E14.5-born cells can receive Shh signaling, which accelerates the positioning of neurons in the future L4 region, where further L4 maturation processes occur. The neurons that do not receive enough Shh are positioned in the more superficial or future L2/3 region and differentiate into L2/3 subtypes. Such endogenous Shh might be provided from the marginal zone of the outermost cortical region, where cortical interneurons are migrating. In fact, interneurons were reported as one of the Shh sources in the developing cortex [53]. Ectopic Shh may have activated the population that normally does not receive Shh signaling, enabling them to be located in the future L4 region. Such differentiation plasticity is probably also regulated temporally in NPCs and/or immature neurons because ectopic Shh expression did not cause any alteration of cell positioning and subtype specification in the E15.5 electroporated cells. As Boc is expressed strongly in L4 (formed by mainly E14.0-born neurons) but very weak in L2/3 (formed by E15.5-born neurons), Boc expression levels could underlie this differential response to ectopic Shh [48]. In addition, given that E14.5 born neurons can respond to ectopic Shh, one might expect that Boc-high neurons exist in layer 2/3, presumably at its bottom. Further studies on the detailed expression pattern of Boc will clarify the difference.

Although Reelin, an extracellular protein, is well-known as a factor that controls positioning and/or migration of cortical neurons [54], cortical lamination has been regarded as a relatively intrinsic process, in which new neurons just pile up on the earlier-formed 'layers' according to their birthdates. Therefore, the extent to which the lamination process is regulated by extrinsic factors remains obscure [5]. Here, we reported on Shh as an extracellular regulator in the lamination of excitatory cortical neurons. A recent report showing the involvement of Cxcl12 in the positioning of cortical interneurons [55] would predict further roles of extrinsic cues in the regulation of neuronal migration/positioning and lamination.

**Author Contributions:** Conceptualization, K.O.; methodology, K.O. and K.N.; validation, K.O. and J.M; formal analysis, K.O.; investigation, K.O.; resources, K.O., K.N., and J.M.; data curation, K.O.; writing—original draft preparation, K.O.; writing—review and editing, K.O., K.N., and J.M.; visualization, K.O.; supervision, J.M.; project administration, K.O.; funding acquisition, K.O. and K.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Japan Society for the Promotion of Science (KAKENHI JP21K06381, JP20H05688).

**Institutional Review Board Statement:** All experiments were approved by the Doshisha University Animal Experiment Committee and conducted in accordance with guidelines established by the Doshisha University Ethics Review Committee.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Any data and original materials are available from the corresponding author upon reasonable request.

**Acknowledgments:** We would like to thank H. Hasegawa at Keio University for his expert advice. We thank the members of the Motoyama laboratory and the animal facility of Doshisha university for the discussions and technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Mikiko Kudo and Kunimasa Ohta \***

Department of Stem Cell Biology, Faculty of Arts and Science, Kyushu University, 744 Motooka, Nishi-Ku, Fukuoka 819-0395, Japan; kudo.mikiko.490@s.kyushu-u.ac.jp

**\*** Correspondence: ohta9203@artsci.kyushu-u.ac.jp

**Abstract:** In the central nervous system (CNS), which comprises the eyes, spinal cord, and brain, neural cells are produced by the repeated division of neural stem cells (NSCs) during the development of the CNS. Contrary to the notion that the CNS is relatively static with a limited cell turnover, cells with stem cell-like properties have been isolated from most neural tissues. The microenvironment, also known as the NSC niche, consists of NSCs/neural progenitor cells, other neurons, glial cells, and blood vessels; this niche is thought to regulate neurogenesis and the differentiation of NSCs into neurons and glia. Although it has been established that neurons, glia, and blood vessels interact with each other in a complex manner to generate neural tissues in the NSC niche, the underlying molecular mechanisms in the CNS niche are unclear. Herein, we would like to introduce the extracellular secreted protein, Akhirin (AKH; Akhi is the Bengali translation for eye). AKH is specifically expressed in the CNS niche—the ciliary body epithelium in the retina, the central canal of the spinal cord, the subventricular zone, and the subgranular zone of the dentate gyrus of the hippocampus—and is supposedly involved in NSC niche regulation. In this review, we discuss the role of AKH as a niche molecule during mouse brain formation.

**Keywords:** Akhirin; neurogenesis; vasculogenesis; LCCL domain; vWF domain; hydrocephalus

#### **1. Introduction**

The central nervous system (CNS; eyes, spinal cord, and brain) acts as a controller that receives, sorts, and organizes information from all over the body and sends the corresponding commands, which is vital for survival. The CNS is formed by the repeated division and proliferation of neural stem cells (NSCs) and neural progenitor cells (NPCs). Dividing immature neurons differentiate into mature neurons. A hallmark of the NSCs/NPCs in the CNS is that they have extremely high proliferation during development; however, once the respective tissues are formed, their rate of cell division reduces dramatically. Therefore, it is important to understand when neuron production is active and how tissue formation progresses during tissue development.

Several studies have identified various cellular sources of NSCs in the adult vertebrate eye [1]. Retinal NSCs are present in the ciliary body epithelium [2,3], iris pigment epithelium [4], peripheral margin of the retina [5], and Müller cells [6,7]. Müller cells are radial glial cells, with morphology and expression of glial markers similar to those of embryonic radial cells, which are used as progenitor cells in the CNS. To date, it is believed that Müller cells are the endogenous NSCs in the retina.

In the spinal cord, several cell types have been identified in the central canal, including cuboidal, tanycytic, and radial classes of lumen-contacted ciliated ependymal cells [8]. Numerous studies have indicated that ependymal cells localized in the dorsal central canal, originated from radial glial cells, show NSC activity [9]. Ependymal cells also contribute to the regeneration of oligodendrocytes and remyelination after spinal cord injury [10].

In the adult mouse brain, several studies have indicated two major neurogenic niches: the subventricular zone (SVZ) lining the lateral ventricle and the subgranular zone (SGZ) of the dentate gyrus (DG) of the hippocampus in the CNS [11,12]. In the SVZ, type B stem cells

**Citation:** Kudo, M.; Ohta, K. Regulation of the Brain Neural Niche by Soluble Molecule Akhirin. *J. Dev. Biol.* **2021**, *9*, 29. https://doi.org/ 10.3390/jdb9030029

Academic Editors: Tsutomu Nohno, Hideyo Ohuchi and Simon J. Conway

Received: 7 May 2021 Accepted: 20 July 2021 Published: 26 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

give rise to type C transit-amplifying cells which, in turn, produce type A neuroblasts [13]. Type B and C cells form a tubular network through which type A neuroblasts migrate into the rostral migratory stream toward the olfactory bulbs. In the SGZ, proliferating radial and nonradial precursors give rise to intermediate progenitors which, in turn, generate neuroblasts. Immature neurons migrate into the inner granule cell layer and differentiate into dentate granule cells in the hippocampus [13].

NSCs/NPCs are constantly maintained in a specific microenvironment (niche) since the time of development processes and throughout adulthood [14]. Typically, equilibrium between cell proliferation and differentiation between the two cell populations, NSCs and NPCs, is important for CNS development. Neurons emerge from a pool of NSCs/NPCs through neurogenesis, which is regulated by many extrinsic and intrinsic factors. The niche in which NSCs are maintained consists of a complex array of other neurons, blood vessels, and other glial cells. The division and self-renewal of NSCs are regulated by specialized niche regulators secreted by these cells. Despite the relevance of the fate of NSCs/NPCs, which is ultimately reflected in the final number of newly generated neurons, the timing and number of divisions of NSCs and their differentiation into neurons are flexible processes; moreover, in several cases, not all types of intermediate progenitors are generated in a clonal lineage [15,16]. Thus, the mechanisms that control their development processes are poorly understood.

Herein, we introduce the secreted protein, Akhirin (AKH; Akhi is the Bengali translation for eye), which was isolated from embryonic day 6 (E6) chick lens using signal sequence trap cDNA screening [17]. We have previously reported extensive expression of AKH in the niches of the CNS. AKH is specifically expressed in the ciliary marginal zone of the retina [18], the middle and ventral central canal of the spinal cord [19], and the SVZ and the DG of the hippocampus in the mouse CNS [20]. As AKH exhibits heterophilic cell adhesion activity, which has been confirmed by cell aggregation analysis [18], it is supposed to function as an extracellular adhesion factor regulating these niches in the CNS. In this review, after summarizing the molecular structure of AKH and its role in the eye and spinal cord, we mainly discuss its role in the brain as a niche regulator in the CNS.

#### **2. Possible Roles of the vWF-A and LCCL Domains in AKH**

The structure of AKH comprises two von Willebrand factor-A (vWF-A) domains and one Limulus factor C, Coch-5b2 and Lgl1 (LCCL) domain. The chick AKH has an open reading frame of 748 amino acid residues, and the mouse AKH has an open reading frame of 650 amino acid residues (Figure 1A). AKH has relatively high homology to vitrin [21] and cochlin [22]. In mice, AKH can be regarded as a factor almost identical to vitrin, and mouse cochlin in composed of 522 amino acid residues. Based on the two vWF-A domains in AKH, cell aggregation analysis using AKH-expressing transfectants indicated the role of AKH in cell adhesion [18]. As both the control cells and AKH-expressing transfectants adhered to immobilized AKH protein, we concluded that AKH has heterophilic cell adhesion activity [18]. It is one of great interest to identify the molecules that interact with AKH.

The vWF-A domain is involved in blood clotting. In vWF disease, the absence or malfunction of vWF-A results in bleeding which is difficult to control. Most of the vWF-A domains are synthesized in the endothelial cells of blood vessels. The vWF-A domain 1 directly binds to angiogenesis-inducing factors such as vascular endothelial growth factor-A (VEGF-A) and platelet-derived growth factor-BB (PDGF-BB). VEGF and PDGF-BB play an important role in vascularization [23] and neurogenesis, respectively, during early development [24].

Although the interaction of blood vessels with NSCs in the adult brain niche has already been widely reported [25–27], recent reports have shown that the timing of differentiation of NSCs/NPCs into neurons is implicated in the interaction between embryonic vascularization and neurogenesis. For instance, Di Marco et al. showed that nascent periventricular vessels interact with dividing apical neural progenitors by using vascular filopodia induced by the upregulation of VEGF-A in a cell-cycle-dependent manner [23].

They concluded that vascular filopodia helps in fine-tuning NSC behavior for proper brain development. As AKH contains two vWF-A domains, it is interesting to examine the molecular interactions between AKH and VEGF-A and PDGF-BB. *J. Dev. Biol.* **2021**, *9*, x FOR PEER REVIEW 3 of 9

**Figure 1.** (**A**): A simple schematic diagram of chick and mouse AKH, mouse cochlin protein structure. The amino acid number is presented for each domain. (**B**–**E**): The expression of *AKH* mRNA in SVZ (**B**,**D**) and the hippocampus region (**C**,**E**) of the mouse brain at E17 (**B**,**C**) and P20 (**D**,**E**). Arrows indicate SVZ niche areas along the LV regions (**B**,**D**) and the developing hippocampus (**C**). Arrowheads indicate the hippocampal CA2 region (**E**). **Figure 1.** (**A**): A simple schematic diagram of chick and mouse AKH, mouse cochlin protein structure. The amino acid number is presented for each domain. (**B**–**E**): The expression of *AKH* mRNA in SVZ (**B**,**D**) and the hippocampus region (**C**,**E**) of the mouse brain at E17 (**B**,**C**) and P20 (**D**,**E**). Arrows indicate SVZ niche areas along the LV regions (**B**,**D**) and the developing hippocampus (**C**). Arrowheads indicate the hippocampal CA2 region (**E**).

The vWF-A domain is involved in blood clotting. In vWF disease, the absence or malfunction of vWF-A results in bleeding which is difficult to control. Most of the vWF-A domains are synthesized in the endothelial cells of blood vessels. The vWF-A domain 1 directly binds to angiogenesis-inducing factors such as vascular endothelial growth factor-A (VEGF-A) and platelet-derived growth factor-BB (PDGF-BB). VEGF and PDGF-BB play an important role in vascularization [23] and neurogenesis, respectively, during early development [24]. Although the interaction of blood vessels with NSCs in the adult brain niche has al-The LCCL domain is found in the biodefense factor C of the horseshoe crab, which is a lipopolysaccharide (LPS)-binding protein [28]. LPS is a constituent of the outer membrane of the cell wall of gram-negative bacteria. LPS can bind to Toll-like receptor 4 (TLR4), which is present on the surface of the host cell membrane and works as an endotoxin [29]. TLR family members are involved in the expression of proinflammatory cytokines and play an important role in innate immunity. Recent reports have shown that the LCCL domain is cleaved from the cochin protein, which sequesters infiltrating bacteria and protects hearing in the organ of Corti [28,30].

ready been widely reported [25–27], recent reports have shown that the timing of differentiation of NSCs/NPCs into neurons is implicated in the interaction between embryonic vascularization and neurogenesis. For instance, Di Marco et al. showed that nascent periventricular vessels interact with dividing apical neural progenitors by using vascular filopodia induced by the upregulation of VEGF-A in a cell-cycle-dependent manner [23]. They concluded that vascular filopodia helps in fine-tuning NSC behavior for proper brain development. As AKH contains two vWF-A domains, it is interesting to examine the molecular interactions between AKH and VEGF-A and PDGF-BB. The LCCL domain is found in the biodefense factor C of the horseshoe crab, which is a lipopolysaccharide (LPS)-binding protein [28]. LPS is a constituent of the outer membrane of the cell wall of gram-negative bacteria. LPS can bind to Toll-like receptor 4 (TLR4), which is present on the surface of the host cell membrane and works as an endotoxin [29]. TLR family members are involved in the expression of proinflammatory cyto-Neurons, immune system cells, and blood vessels form a tight network with each other to maintain immune homeostasis in the brain. The LCCL domain binds to LPS and plays a role in immune defense. Infections of bacteria may cause encephalitis, meningitis, and various other diseases in the brain and, occasionally, cause disturbances in the bloodbrain barrier (between the neurons and vascular units) and the blood-cerebrospinal fluid barrier (BCSFB; between the cerebrospinal fluid (CSF) and choroid plexus (ChP)) [31]. The ChP is a vascular-rich tissue in the ventricle facing the NSC niche that produces and secretes CSF. Immune system activation following a bacterial infection in a pregnant mother increases the number of activated microglia in the ChP of the fetal brain, which disrupts the intercellular adhesion of ependymal cells, causing the disarrangement of the BCSFB and/or an increase in polarity, which biases the direction of the microglial neurites; this consequently results in abnormal cortical layer formation after birth [32,33]. The progression of layer formation is active during the embryonic and immediate postnatal

kines and play an important role in innate immunity. Recent reports have shown that the LCCL domain is cleaved from the cochin protein, which sequesters infiltrating bacteria periods. The neurons migrate toward the surface of the brain from the ventricular zones, forming the six-layered structure (inside-out). The ventricular zones consist of epithelial tissue, and the cell junctions are rigid. The disrupted regulation of cell junctions on the ventricular surface, especially cadherins, which are calcium (Ca2+)-dependent adhesion molecules between epithelial cells, causes abnormal layer formation [34]. In addition, since AKH has heterophilic cell adhesive activity [18], it is highly predictable that the loss of AKH will result in disruption of the tight junctions. Thus, it is likely that there is a mutual relationship between vascularization-neurogenesis and microglia during brain formation; however, the detailed molecular mechanism has not yet been elucidated. Therefore, the activation and neurite polarity of microglia may explain the cause of ventricle expansion in the *AKH*-/- mouse brain.

#### **3. AKH Localizes in the Niche of the Eye**

Different neuronal NSCs/NPCs exist in the vertebrate retina, and their proliferation and differentiation are influenced by a combination of intrinsic and extrinsic factors [35,36]. In chick peripheral retina, both *AKH* mRNA and protein are expressed through the ciliary epithelial layer in the embryonic stage. In the chick embryo, AKH expression is observed in the head ectoderm overlying the lens vesicle at stage 17 and in the retinal pigment epithelial layer at stage 22. Although AKH expression changes during the embryonic stage, AKH accumulates in the presumptive ciliary marginal zone at the postnatal stage where the NSCs/NPCs are localized [18]. As *AKH* mRNA and protein are co-localized in NSCs/NPCs, and AKH exhibits heterophilic cell adhesion activity, we hypothesized that AKH secreted from these cells is associated with other extracellular matrix components on their surface to regulate the niche [18]. Unfortunately, we could not observe the expression of AKH in E14 mouse retina.

#### **4. AKH Localizes in the Niche of the Spinal Cord**

The spinal cord is the caudal portion of the CNS and transduces information between the brain and the body. NSCs have been isolated from the ependymal zone surrounding the central canal of the spinal cord [37]. A recent study showed that NSCs are the most dorsally located glial fibrillary acid protein (GFAP)-positive cells lying ependymally [38]. AKH expression was observed in the spinal cord of mice on embryonic day 9.5 (E9.5), which disappeared by postnatal day 30 (P30). *AKH*-/- mice showed reduced spinal cord size compared to that in wild-type mice (*AKH+/+*). The expression patterns of ependymal niche molecules (nestin and GFAP) in *AKH*-/- mice were changed when compared with those of *AKH*+/+ mice in vivo [19]. In vitro culture of the spinal cord neurospheres showed significant reduction in the size of the neurospheres of *AKH*-/- mice compared with those of *AKH+/+* mice [19]. Interestingly, the distribution of ependymal proliferation factors (Cyclin D2 and vimentin) and proliferation markers (Ki67) in the neurospheres derived from *AKH*-/- was disturbed, indicating the involvement of AKH in NSCs/NPCs regulation.

In general, ependymal cells of the spinal cord are normally quiescent in adult mice. However, when the spinal cord is damaged, ependymal cells are rapidly activated and undergo differentiation to form astrocytes at the injured site [8]. Although the expression of *AKH* in the central canal ependymal cells is very low or not observed in the central canal at P30, AKH expression is rapidly upregulated in ependymal cells after spinal cord injury, suggesting that AKH is involved in post-injury neuronal neogenesis [19]. These observations suggest that AKH plays a crucial role in spinal cord formation in mice by regulating the ependymal niche in the central canal [38].

#### **5. AKH Is Exclusively Localized in Brain Neurogenic Niches**

The adult mammalian brain contains billions of neurons assembled in defined neural circuits, which are the essential components for mediating the higher functions of the CNS. During the development of the mouse brain, NSCs exist around the ventricles, and the newborn neurons migrate to their destination through various pathways. In the niches,

neurons and glia cells, such as microglia, astrocytes, and oligodendrocytes, emerge; they create a feedback interaction system via numerous secreted and contact-mediated signals for the regulation between quiescence and cell division of NSCs/NPCs. Although NSCs disappear from most parts of the brain after birth, they are still localized at the SVZ on the lateral wall of the LV and the SGZ of the hippocampal DG where they continue to produce neurons throughout life [38,39].

The expression of *AKH* at the SVZ was already observed at E17.5, which then disappeared by P20 (Figure 1B,D). In the hippocampal DG, *AKH* expression was observed in the entire hippocampal region immediately after birth, followed by a specific expression in the hippocampal CA2 region at P20 (Figure 1C,E). Thus, from the embryonic stage, AKH is expressed in the brain niche areas and disappears in tandem with the cessation of neurogenesis around P20 when brain formation is approximately complete; this observation suggests the involvement of AKH in neurogenesis and neuronal differentiation during early development but not in adult neurogenesis [20].

Newborn neurons migrate to other regions in the brain by forming special chainlike structures, which suggests that the interaction between newborn neurons and the extracellular matrix, such as AKH, is important in this process. NSCs are localized in the periventricular region, and nascent neurons migrate to the hippocampal region for hippocampal area formation. One possibility is that AKH might be involved in the migration of newborn neurons to the CA2 region because AKH secreted by NSCs is supposed to attach to the surface and interacts with the substrate during migration. The CA2 region of the hippocampus is a less well-characterized region than the CA1 and CA3 regions. In recent years, studies have reported the importance of the CA2 region in memory updating—a timeline of memory—and social memory [40–42]. When neuronal migration to the CA2 region is impaired due to the loss of AKH, psychiatric disorders, such as autism spectrum disorder, may occur. To examine the effect of AKH loss in the CA2 region, we are in the process of preparing a behavioral test battery with *AKH*-/- mice.

#### **6. Effects of** *AKH* **Knockout on Neurogenesis and Neuronal Differentiation in the Brain NSC Niche**

Self-renewal and differentiation potential of NSCs are dynamically regulated by various niche-derived factors, which relay signals in an autocrine or paracrine manner together with transcription factors that respond to those signals. To investigate whether AKH is a niche factor, we compared the brain morphology using *AKH*-/- and *AKH*+/+ mice and found that the ventricles of *AKH*-/- mice were widely expanded when compared with those of *AKH*+/+ mice (Figure 2A,B) [20]. The hippocampal DG region was reduced in *AKH*-/- mice when compared to that in *AKH*+/+ mice (Figure 2D,E) [20]. Furthermore, lower proportion of GFAP, SOX2, and Ki67 trip; e-positive (GFAP+/SOX2+/Ki67<sup>+</sup> ) cells was observed in *AKH*-/-, indicating reduced NSC proliferation, but higher population of GFAP and SOX2 double-positive (GFAP+/SOX2+/Ki67−) cells was increased in *AKH*-/- mice, indicating increase of quiescent NSCs. Finally, *AKH* deficiency inhibited the differentiation of NSCs into mature neurons and reduced the length of their neurites [20]. These results suggest that the loss of *AKH* causes NSCs to lose their proliferative capacity and become quiescent, resulting in a decrease in neurogenesis from NSCs, leading to the enlargement of the ventricles and reduction of the DG area during early development.

and SOX2 double-positive (GFAP+/SOX2+/Ki67-

cating increase of quiescent NSCs. Finally, *AKH* deficiency inhibited the differentiation of NSCs into mature neurons and reduced the length of their neurites [20]. These results suggest that the loss of *AKH* causes NSCs to lose their proliferative capacity and become quiescent, resulting in a decrease in neurogenesis from NSCs, leading to the enlargement

of the ventricles and reduction of the DG area during early development.

) cells was increased in *AKH-/-* mice, indi-

**Figure 2.** Histomorphology of the SVZ (**A**–**C**) and the hippocampal regions (**D**–**F**) in mouse brain on hematoxylin-eosin (HE) (**A**,**C**,**D**,**F**) and Nissl (**B**,**E**) staining. *AKH*+/+ (**A**,**D**), AKH*-/-* (**B**,**E**), and *AKH-/-* hydrocephalus-like mouse (**C**,**F**). **Figure 2.** Histomorphology of the SVZ (**A**–**C**) and the hippocampal regions (**D**–**F**) in mouse brain on hematoxylin-eosin (HE) (**A**,**C**,**D**,**F**) and Nissl (**B**,**E**) staining. *AKH*+/+ (**A**,**D**), *AKH*-/- (**B**,**E**), and *AKH*-/- hydrocephalus-like mouse (**C**,**F**).

#### **7. Relationship between AKH and Hydrocephalus 7. Relationship between AKH and Hydrocephalus**

tween these genes.

Hydrocephalus is a frequent neurological disorder caused by the expansion of the cerebral ventricles and is associated with high morbidity and mortality rates [43]. Different forms of hydrocephalus have been identified: Noncommunicating hydrocephalus is caused by a blockage in the ventricular system, mainly at the aqueduct level between the third and fourth ventricles. In contrast, the ventricular system is not obstructed in communicating hydrocephalus. Although multiple genes and environmental factors are involved in hydrocephalus development, the molecular mechanisms underlying this condition remain unclear. Due to insufficient knowledge regarding the molecular basis of hydrocephalus, its clinical treatment is limited to invasive methods, with failure rates close Hydrocephalus is a frequent neurological disorder caused by the expansion of the cerebral ventricles and is associated with high morbidity and mortality rates [43]. Different forms of hydrocephalus have been identified: Noncommunicating hydrocephalus is caused by a blockage in the ventricular system, mainly at the aqueduct level between the third and fourth ventricles. In contrast, the ventricular system is not obstructed in communicating hydrocephalus. Although multiple genes and environmental factors are involved in hydrocephalus development, the molecular mechanisms underlying this condition remain unclear. Due to insufficient knowledge regarding the molecular basis of hydrocephalus, its clinical treatment is limited to invasive methods, with failure rates close to 50%.

to 50%. Hydrocephalus-like enlargement of the ventricles was observed in the *AKH-/-* brain [20]. Interestingly, we found that approximately 3% of *AKH-/-* mice showed severely malformed brains that resembled a hydrocephalic brain morphology (Figure 2C,F). We were unable to analyze these fulminant *AKH-/-* mice in detail as these mutants died within the first month after birth. These phenotypes of *AKH-/-* hydrocephalic mice are similar to those of heterozygous *Nfib* and *Nfix* double mutant mice [44] and Yap mutant hydrocephalic mutant mice [45]. Although the ventricles expanded similarly to those in hydrocephalus because of these gene deletions, the molecular mechanisms underlying this condition remain unclear. Therefore, it is plausible to examine the molecular interactions be-Hydrocephalus-like enlargement of the ventricles was observed in the *AKH*-/ brain [20]. Interestingly, we found that approximately 3% of *AKH*-/- mice showed severely malformed brains that resembled a hydrocephalic brain morphology (Figure 2C,F). We were unable to analyze these fulminant *AKH*-/- mice in detail as these mutants died within the first month after birth. These phenotypes of *AKH*-/- hydrocephalic mice are similar to those of heterozygous *Nfib* and *Nfix* double mutant mice [44] and Yap mutant hydrocephalic mutant mice [45]. Although the ventricles expanded similarly to those in hydrocephalus because of these gene deletions, the molecular mechanisms underlying this condition remain unclear. Therefore, it is plausible to examine the molecular interactions between these genes.

#### **8. Conclusions**

**8. Conclusions**  *AKH* deficiency inhibits the proliferation of NSCs, resulting in a decrease in the total number of neurons in the brain niche area, thereby resulting in ventricular expansion in patients with hydrocephalus. It is of interest to examine the human *AKH* gene sequence in patients with hydrocephalus in the future. At present, the physiological functions of *AKH* deficiency inhibits the proliferation of NSCs, resulting in a decrease in the total number of neurons in the brain niche area, thereby resulting in ventricular expansion in patients with hydrocephalus. It is of interest to examine the human *AKH* gene sequence in patients with hydrocephalus in the future. At present, the physiological functions of AKH remain unclear. While we can predict the important functions of AKH during brain formation, such as neurogenesis, vasculogenesis, and immunocompetence, more studies are needed to clarify the functional consequences of the defects in brain formation in *AKH*-/- mice. To end this, it is necessary to determine the detailed expression and localization of AKH in developed mice brains, etc. Thus, we hope that a more detailed function of AKH will be discovered by analyzing the effect of *AKH* deficiency on the interaction between NSCs and blood vessels, and on microglia (Figure 3). To date, research

in the fields of neurogenesis, vasculogenesis, and immunocompetence has been specialized and advanced. We propose that AKH is an interesting molecule that links these three fields and regulates brain formation. tween NSCs and blood vessels, and on microglia (Figure 3). To date, research in the fields of neurogenesis, vasculogenesis, and immunocompetence has been specialized and advanced. We propose that AKH is an interesting molecule that links these three fields and regulates brain formation.

AKH remain unclear. While we can predict the important functions of AKH during brain formation, such as neurogenesis, vasculogenesis, and immunocompetence, more studies are needed to clarify the functional consequences of the defects in brain formation in *AKH- /-* mice. To end this, it is necessary to determine the detailed expression and localization of AKH in developed mice brains, etc. Thus, we hope that a more detailed function of AKH will be discovered by analyzing the effect of *AKH* deficiency on the interaction be-

*J. Dev. Biol.* **2021**, *9*, x FOR PEER REVIEW 7 of 9

**Figure 3.** AKH is expressed in the neural stem cell niche and acts as a regulator. AKH is involved in the maintenance of neural stem cell proliferation, vasculogenesis, and the immune system during neurogenesis. **Figure 3.** AKH is expressed in the neural stem cell niche and acts as a regulator. AKH is involved in the maintenance of neural stem cell proliferation, vasculogenesis, and the immune system during neurogenesis.

**Author Contributions:** M.K. and K.O. conceived the study and wrote the manuscript. All authors have read and agreed to the published version of the manuscript. **Author Contributions:** M.K. and K.O. conceived the study and wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding **Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable. **Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable. **Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable. **Data Availability Statement:** Not applicable.

**Acknowledgments:** We gratefully thank Arif Istiaq, Kousei Takashi, Gu Haoxuan, and Fucho Tan for their helpful assistance. This work was supported by Kumamoto University Advanced Research Project Stem Cell-Based Tissue Regeneration Research and Education Unit, the Joint Usage/Research Center for Developmental Medicine, IMEG, Kumamoto University. **Acknowledgments:** We gratefully thank Arif Istiaq, Kousei Takashi, Gu Haoxuan, and Fucho Tan for their helpful assistance. This work was supported by Kumamoto University Advanced Research Project Stem Cell-Based Tissue Regeneration Research and Education Unit, the Joint Usage/Research Center for Developmental Medicine, IMEG, Kumamoto University.

**Conflicts of Interest:** The authors declare no conflict of interest. **Conflicts of Interest:** The authors declare no conflict of interest.

#### **References References**


**Takuma Shinozuka 1,2,\* and Shinji Takada 1,2,3,\***


**Abstract:** The most dorsal region, or roof plate, is the dorsal organizing center of developing spinal cord. This region is also involved in development of neural crest cells, which are the source of migratory neural crest cells. During early development of the spinal cord, roof plate cells secrete signaling molecules, such as Wnt and BMP family proteins, which regulate development of neural crest cells and dorsal spinal cord. After the dorso-ventral pattern is established, spinal cord dynamically changes its morphology. With this morphological transformation, the lumen of the spinal cord gradually shrinks to form the central canal, a cavity filled with cerebrospinal fluid that is connected to the ventricular system of the brain. The dorsal half of the spinal cord is separated by a glial structure called the dorsal (or posterior) median septum. However, underlying mechanisms of such morphological transformation are just beginning to be understood. Recent studies reveal that roof plate cells dramatically stretch along the dorso-ventral axis, accompanied by reduction of the spinal cord lumen. During this stretching process, the tips of roof plate cells maintain contact with cells surrounding the shrinking lumen, eventually exposed to the inner surface of the central canal. Interestingly, Wnt expression remains in stretched roof plate cells and activates Wnt/β-catenin signaling in ependymal cells surrounding the central canal. Wnt/β-catenin signaling in ependymal cells promotes proliferation of neural progenitor and stem cells in embryonic and adult spinal cord. In this review, we focus on the role of the roof plate, especially that of Wnt ligands secreted by roof plate cells, in morphological changes occurring in the spinal cord.

**Keywords:** Wnt; roof plate; spinal cord; morphogenesis; central canal; dorsal collapse; dorsal median septum; neural crest

#### **1. Roof Plate Functions in the Early Developmental Stage of Spinal Cord**

#### *1.1. Generation of Trunk Neural Crest Cells*

In the beginning of neural development, the neural plate gradually invaginates and its lateral edges are transformed into the neural fold. The appearance of the neural fold is probably the first morphological indication of the dorsal region of neural tissues. Then, the tips of the neural fold fuse, resulting in formation of the neural tube, which develops into the brain in the head and the spinal cord in the trunk. In the mouse, anterior neural tube is generated by this process, called primary neurulation (Figure 1). On the other hand, the posterior neural tube is formed by a process called secondary neurogenesis, in which the neural tube is formed from precursors in the tail bud, followed by condensation of the mesenchyme and subsequent epithelialization [1]. After these processes, neuroepithelial cells adjacent to the lumen proliferate rapidly and differentiate into several distinct types of neuronal and glial cells. Roof plate cells are located in the most dorsal part of developing

**Citation:** Shinozuka, T.; Takada, S. Morphological and Functional Changes of Roof Plate Cells in Spinal Cord Development. *J. Dev. Biol.* **2021**, *9*, 30. https://doi.org/10.3390/ jdb9030030

Academic Editors: Tsutomu Nohno and Hideyo Ohuchi

Received: 5 July 2021 Accepted: 28 July 2021 Published: 30 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

spinal cord and serve as the organizing center for surrounding neuroepithelial cells, promoting their proliferation and specification [2]. Prior to functioning as an organizing center, these cells give rise to neural crest cells, which migrate to many different tissues, where they give rise to neurons and glial cells of the sensory, sympathetic, and parasympathetic nervous systems, pigment-containing cells of the epidermis, and chromaffin cells of the adrenal gland [3–6]. Thus, cells in the roof plate, have markedly different roles and their behaviors are dynamic in early embryonic stages. developing spinal cord and serve as the organizing center for surrounding neuroepithelial cells, promoting their proliferation and specification [2]. Prior to functioning as an organizing center, these cells give rise to neural crest cells, which migrate to many different tissues, where they give rise to neurons and glial cells of the sensory, sympathetic, and parasympathetic nervous systems, pigment-containing cells of the epidermis, and chromaffin cells of the adrenal gland [3–6]. Thus, cells in the roof plate, have markedly different roles and their behaviors are dynamic in early embryonic stages.

types of neuronal and glial cells. Roof plate cells are located in the most dorsal part of

*J. Dev. Biol.* **2021**, *9*, x FOR PEER REVIEW 2 of 13

**Figure 1.** Transition in morphology of neural tissue and functions of cells in the most dorsal region during early developmental stages. Schematic images of transverse sections of developing spinal cord before (**A**) and after (**B**) closure of the neural tube during mouse development. The edge of the neural plate elevates and becomes the neural fold (cyan), from which neural crest cells are generated (**A**). Then, the lateral tips of the neural fold fuse, generating the neural tube (**B**). Cells in the most dorsal region of the neural tube, or roof plate (cyan), act as progenitors of migrating neural crest cells (yellow) and also as the organizing center for dorsal neuroepithelial cells. Roof plate cells secrete signaling molecules such as Wnt and BMP, which govern development of both neural crest and dorsal neuroepithelial cells. By contrast, cells in the most ventral region, the floor plate (magenta), act as the ventral organizer, by producing Shh. **Figure 1.** Transition in morphology of neural tissue and functions of cells in the most dorsal region during early developmental stages. Schematic images of transverse sections of developing spinal cord before (**A**) and after (**B**) closure of the neural tube during mouse development. The edge of the neural plate elevates and becomes the neural fold (cyan), from which neural crest cells are generated (**A**). Then, the lateral tips of the neural fold fuse, generating the neural tube (**B**). Cells in the most dorsal region of the neural tube, or roof plate (cyan), act as progenitors of migrating neural crest cells (yellow) and also as the organizing center for dorsal neuroepithelial cells. Roof plate cells secrete signaling molecules such as Wnt and BMP, which govern development of both neural crest and dorsal neuroepithelial cells. By contrast, cells in the most ventral region, the floor plate (magenta), act as the ventral organizer, by producing Shh.

During development of the neural tube, premigratory neural crest cells first exist in the neural fold and undergo delamination and epithelial-mesenchymal transition to become migratory neural crest cells [7]. However, the timing and mechanism of fate determination in these neural crest cells remains controversial. Lineage tracing analysis using *R26R-Confetti* mice revealed that the vast majority of individual premigratory and even migratory neural crest cells are multipotent. Furthermore, in some clones with labeled progeny cells among neural crest derivatives, labeled progeny cells are retained in the dorsal neural tube, suggesting an asymmetric cell division of premigratory neural crest cells in the dorsal neural tube [6]. By contrast, lineage tracing analysis using avian embryos after electroporation with GFP reporter showed that pre-migratory neural crest cells generate progeny in single, rather than multiple derivatives, in most cases where delaminated neural crest cells are labeled. In these cases, no labeled cells remained in the neural tube [8]. This result with avian embryos suggests that premigratory neural crest cells are a distinct population from cells that remain in the neural tube, such as roof plate cells, which act as the organizing center. In addition, this transition from neural crest to roof plate accompanied by loss of responsiveness to BMP signaling in dorsal spinal cord [8– 10]. These discrepancies may be due to differences in labeling techniques, in the stage and location of labeling, or in mechanisms of lineage segregation between mammalian and avian systems. During development of the neural tube, premigratory neural crest cells first exist in the neural fold and undergo delamination and epithelial-mesenchymal transition to become migratory neural crest cells [7]. However, the timing and mechanism of fate determination in these neural crest cells remains controversial. Lineage tracing analysis using *R26R-Confetti* mice revealed that the vast majority of individual premigratory and even migratory neural crest cells are multipotent. Furthermore, in some clones with labeled progeny cells among neural crest derivatives, labeled progeny cells are retained in the dorsal neural tube, suggesting an asymmetric cell division of premigratory neural crest cells in the dorsal neural tube [6]. By contrast, lineage tracing analysis using avian embryos after electroporation with GFP reporter showed that pre-migratory neural crest cells generate progeny in single, rather than multiple derivatives, in most cases where delaminated neural crest cells are labeled. In these cases, no labeled cells remained in the neural tube [8]. This result with avian embryos suggests that premigratory neural crest cells are a distinct population from cells that remain in the neural tube, such as roof plate cells, which act as the organizing center. In addition, this transition from neural crest to roof plate accompanied by loss of responsiveness to BMP signaling in dorsal spinal cord [8–10]. These discrepancies may be due to differences in labeling techniques, in the stage and location of labeling, or in mechanisms of lineage segregation between mammalian and avian systems.

During development of neural crest cells, cells in the most dorsal region of the spinal cord produce secreted ligands such as BMP and Wnt [2,11,12] (Figure 1B). Several lines of evidence indicate that Wnt ligands, which activate Wnt/β-catenin signaling, are required for generation of neural crest cells. In the spinal cord of mouse embryos, two Wnt ligands, *Wnt1* and *Wnt3a*, are specifically expressed in roof plate cells [13–15]. These Wnt ligands activate a common signaling pathway, the Wnt/β-catenin pathway, and are functionally redundant in the dorsal spinal cord. Thus, neither *Wnt1* nor *Wnt3a* single KO mutants show any obvious defects in dorsal spinal cord development, although *Wnt1* KO mutant impairs the development of midbrain and cerebellum and *Wnt3a* KO mutant exhibits severe truncation of axis elongation [14,16–18]. On the other hand, *Wnt1* and *Wnt3a* doublemutant embryos exhibit a reduction of neural crest cell number and a marked deficiency of neural crest derivatives [19]. Similarly, Wnt/β-catenin signaling can induce and is required for neural crest formation in other species [20–24].

In addition to formation of neural crest cells, Wnt/β-catenin signaling also promotes segregation of sub-lineages of neural crest cells [25–29]. For instance, conditional loss of function of β-catenin in the mouse roof plate reduces melanocytes and *Ngn2*-positive sensory lineage cells, resulting in impaired formation of neurons and glial cells in dorsal root ganglia (DRG) [26]. On the other hand, studies in which β-catenin is activated at different time points in neural crest differentiation suggest the importance of this signaling in determining the fate of the neural crest sub-lineage [27,29]. In zebrafish, studies of gainand loss-of-function of Wnt/β-catenin signaling in pre-migratory cranial neural crest cells also indicate the importance of this signaling in determining the fate of the neural crest sub-lineage [25]. Furthermore, in addition to Wnt/β-catenin signaling, ligands of the BMP family are also involved in fate decision of neural crest cells. For instance, late emigrating neural crest cells in the roof plate are restricted to a sensory fate by *Gdf7* [30], and BMP2 antagonizes sensory specification induced by Wnt signaling [31].

#### *1.2. Specification of Dorsal Interneurons*

Direct evidence showing the requirement for roof plate cells in specification of dorsal neuroepithelial cells comes from genetic ablation of roof plate cells with *Gdf7-DTA.* Progenitors of dorsal interneurons are subclassified as dI1 to dI6, in dorsal-to-ventral order in developing spinal cord [32]. This ablation causes loss of progenitors of dorsal interneurons dI1-3 and compensatory occupation of a dorsal position by dI4-6 [33]. This specification, as well as proliferation of dorsal neuroepithelial cells, is regulated by roof plate-derived Wnt and BMP family proteins. For instance, *Wnt1* and *Wnt3a* double-mutant embryos exhibit impaired proliferation and specification of cells in the dorsal spinal cord [34]. A similar phenotype is also observed in mutant embryos in which components of the Wnt/β-catenin pathway, including *Wntless* and *β-catenin*, are impaired [35,36]. In addition, activation of Wnt/β-catenin signaling in developing spinal cord indicates that this signaling can regulate specification and proliferation of neuroepithelial cells [36–38]. In terms of Wnt-mediated proliferation, it has been proposed that several Wnt ligands expressed in the dorsal spinal cord generate a proliferation gradient along the dorso-ventral axis [39].

BMP family proteins, including BMP4, BMP6, BMP7, and Gdf7, are also expressed in the surface ectoderm and the dorsal spinal cord and are involved in differentiation of dorsal interneurons [40–44]. Manipulation of BMP signaling can promote differentiation of dorsal interneurons in vitro [45,46]. Genetic analyses with mutant embryos showed that *Gdf7* is required in formation of dI1 interneurons [42] and that *Bmp7* is similarly essential for several subtypes of dorsal interneurons [47]. In contrast, inhibitory Smad6 and Smad7 are expressed in the neural tube, restricting the action of BMP signaling in the dorsal neural tube [48]. *Bmp7* and *Gdf7* are also required for dI1 axon growth [49]. Since activation of Wnt/β-catenin signaling in roof plate induces expansion of BMP signaling activity in dorsal spinal cord [50], combinatorial Wnt and BMP signaling appears to regulate dorsal interneuron specification and proliferation. Moreover, recent study revealed that Notch signaling in dorsal neural tube is also required for roof plate and dI1 formation. Loss

of Notch function causes missing of roof plate and dI1 with a compensatory expansion of dI2 [51]. Taken together, there are several lines of evidence to show that signaling molecules secreted from roof plate cells are critical in neural crest formation and interneuron specification. On the other hand, there are few studies examining the role of roof platederived ligands in later developmental stages, as discussed below. *J. Dev. Biol.* **2021**, *9*, x FOR PEER REVIEW 4 of 13 Notch function causes missing of roof plate and dI1 with a compensatory expansion of dI2 [51]. Taken together, there are several lines of evidence to show that signaling mole-

#### **2. Morphological Transformation of the Spinal Cord and Central Canal Formation in Mouse Development** cules secreted from roof plate cells are critical in neural crest formation and interneuron specification. On the other hand, there are few studies examining the role of roof platederived ligands in later developmental stages, as discussed below.

#### *2.1. Morphological Transformation of the Spinal Cord*

During development of spinal cord, its size and morphology change dramatically. Neuroepithelial cells proliferate and give rise to migrating cells that accumulate around the original layer of neuroepithelial, or neuroprogenitor, cells. This accumulation results in formation of the mantle zone and the layer of neuroprogenitor cells remaining along the lumen, which is now called the ventricular layer. Cells in the mantle zone differentiate into neurons and glia, and these neurons are interconnected and extend their axons to the lateral region of the spinal cord, known as the marginal zone [11]. The mantle zone, which will form the gray matter of the spinal cord, gradually becomes a butterfly-shaped structure, surrounded by the marginal zone, which will form the white matter (Figure 2C). The spinal cord, as well as the brain, is surrounded by three layers of meninges, the pia mater, the arachnoid mater, and the dura mater. As the spinal cord develops, all three layers are generated from a mesenchymal sheath on the surface of the developing spinal cord, called the primary meninx [52]. **2. Morphological Transformation of the Spinal Cord and Central Canal Formation in Mouse Development**  *2.1. Morphological Transformation of the Spinal Cord*  During development of spinal cord, its size and morphology change dramatically. Neuroepithelial cells proliferate and give rise to migrating cells that accumulate around the original layer of neuroepithelial, or neuroprogenitor, cells. This accumulation results in formation of the mantle zone and the layer of neuroprogenitor cells remaining along the lumen, which is now called the ventricular layer. Cells in the mantle zone differentiate into neurons and glia, and these neurons are interconnected and extend their axons to the lateral region of the spinal cord, known as the marginal zone [11]. The mantle zone, which will form the gray matter of the spinal cord, gradually becomes a butterfly-shaped structure, surrounded by the marginal zone, which will form the white matter (Figure 2C). The spinal cord, as well as the brain, is surrounded by three layers of meninges, the pia mater, the arachnoid mater, and the dura mater. As the spinal cord develops, all three layers are generated from a mesenchymal sheath on the surface of the developing spinal cord, called the primary meninx [52].

**Figure 2.** Morphological transformation of the mouse spinal cord. Schematic images of spinal cord morphology and roof plate shape are shown at E10.5 (**A**), E13.5 (**B**), and E18.5 (**C**) during mouse development. The gray matter of the spinal cord is shown in gray. Roof plate cells, or their descendants, and their nuclei are shown in cyan and blue, respectively. The size and morphology of spinal cord change dramatically during development. In accordance with these changes, morphology of roof plate cells also transforms. In mouse embryos, the apical sides of roof plate cells gradually constrict, such that roof plate cells assume a wedge-shaped form at E10.5 (**A**). Then, the lumen of the spinal cord shrinks gradually. The reduction of lumen size is caused by gradual attrition of neuroepithelial cells adjacent to roof plate cells. As a result, cell reduction from the surface of the lumen proceeds progressively in a dorsal-to-ventral manner known as dorsal collapse. Accompanying the dorsal collapse and reduction of the lumen, roof plate cells are stretched along the dorso-ventral axis and line up along the midline, resulting in morphological transformation. Compared to the dorsal side, reduction of the ventral ventricular zone and transformation of floor plate (magenta) are comparatively smaller (**B**). The stretched roof plate cells are also known as dorsomedial Nestin-positive radial glia (dmNes+RGs). At E18.5, the diameter of the lumen finally shrinks to roughly a few cell diameters, resulting in the central canal, a cavity filled with cerebrospinal fluid and connected to the ventricular system of the brain. Quiescent neural stem cells locate around the central canal. At this stage, **Figure 2.** Morphological transformation of the mouse spinal cord. Schematic images of spinal cord morphology and roof plate shape are shown at E10.5 (**A**), E13.5 (**B**), and E18.5 (**C**) during mouse development. The gray matter of the spinal cord is shown in gray. Roof plate cells, or their descendants, and their nuclei are shown in cyan and blue, respectively. The size and morphology of spinal cord change dramatically during development. In accordance with these changes, morphology of roof plate cells also transforms. In mouse embryos, the apical sides of roof plate cells gradually constrict, such that roof plate cells assume a wedge-shaped form at E10.5 (**A**). Then, the lumen of the spinal cord shrinks gradually. The reduction of lumen size is caused by gradual attrition of neuroepithelial cells adjacent to roof plate cells. As a result, cell reduction from the surface of the lumen proceeds progressively in a dorsal-to-ventral manner known as dorsal collapse. Accompanying the dorsal collapse and reduction of the lumen, roof plate cells are stretched along the dorso-ventral axis and line up along the midline, resulting in morphological transformation. Compared to the dorsal side, reduction of the ventral ventricular zone and transformation of floor plate (magenta) are comparatively smaller (**B**). The stretched roof plate cells are also known as dorsomedial Nestin-positive radial glia (dmNes+RGs). At E18.5, the diameter of the lumen finally shrinks to roughly a few cell diameters, resulting in the central canal, a cavity filled with cerebrospinal fluid and connected to the ventricular system of the brain. Quiescent neural stem cells locate around the central canal. At this stage, roof plate cells (dmNes+RGs) are stretched, maintaining contact with the inner surface of the central canal and also with the pia mater, which covers the outer surface of the spinal cord (**C**).

#### *2.2. Morphological Transformation of the Lumen*

Contemporaneously with morphological transformation of the spinal cord, the lumen of the spinal cord, which is surrounded by the ventricular layer, gradually diminishes in size and finally becomes the central canal, a cavity filled with cerebrospinal fluid (CSF) that is connected to the ventricular system of the brain. In mouse embryos, this reduction starts on approximately embryonic day 13.5 (E13.5; Figure 2B). Prior to its reduction, the lumen extends over almost the entire dorsoventral axis of the spinal cord. The size of lumen, i.e., the dorso-ventral length of the ventricular layer, is dramatically reduced between E13.5 and E15.5. By E17.5 and E18.5, the diameter of the lumen has shrunk to roughly a few cell diameters, resulting in the central canal [53,54] (Figure 2).

The central canal is lined with the ependymal layer, which is composed of several distinct cell types, including ciliated ependymal cells, tanycytes (a subpopulation of radial glia), and CSF-contacting, neuron-like cells [55–58]. Although the functions of these cell types remain to be determined, some of them are likely responsible for homeostatic regulation of CSF in adults [59,60]. Moreover, the ependymal layer includes quiescent stem cells, which generate progeny that undergo glial fates after injury to the spinal cord [55,61,62].

Of note, at the dorsal pole of the central canal, neuron-like cells with extensive projections are observed. These cells express Nestin, and their projections extend from the apical side facing the central canal toward the superficial regions of the spinal cord, as far as the meninges [63–65]. These cells are referred to as dorsal midline Nestin (+) radial glia (dmNes+RG) [65].

#### *2.3. Origin of Ependymal Layer*

Lineage-tracing analyses reveal that most cells comprising the ependymal layer are from the ventral ventricular zone, especially in the subdomains called p2 and pMN [66–69]. Conditional knock-out of components of Shh signaling, including *Shh* and *Smo,* reveals that Shh signaling is required for formation of the ependymal zone [69]. By contrast, dmNes+RG cells which are derived from the roof plate and cells at the ventral pole of the ependymal layer are from the floor plate [54,70]. Thus, cells forming the ependymal layer are heterogenous in origin and reduction of the lumen does not progress proportionally along the dorso-ventral axis.

#### *2.4. Regulatory Mechanisms Governing Lumen Reduction*

Reduction of lumen size is caused by gradual attrition of the neuroepithelial cell population in the ventricular zone (Figure 2). Importantly, this process is promoted by a morphological phenomenon known as dorsal collapse [54,65]. In dorsal collapse, cell reduction proceeds progressively in a dorsal-to-ventral manner in the ventricular zone. In this process, cells adjacent to the dorsal midline down-regulate apical polarity proteins and delaminate in a stepwise manner. The loss of polarity and delamination can be promoted by a secreted form of Crumbs2 produced by dmNes+RGs [65]. In addition, loss of apical polarity protein, such as pard6γb, is involved in this process, because loss of *pard6γb* disrupts dorsal collapse and lumen reduction in zebrafish [71,72]. Moreover, ventricular layer attrition is accompanied by reduction of cell proliferation in the ventricular layer but not apoptosis [54]. Compared to the dorsal side, reduction of the ventral ventricular zone is comparatively smaller, but apparent over time. This ventral reduction may be pronounced by migration of glial cells differentiated from neuroepithelial cells in the ventricular layer [54].

#### **3. Development of Roof Plate Cells in Formation of the Central Canal**

## *3.1. Formation of dmNes+RGs*

As mentioned above, roof plate and floor plate cells remain at the dorsal and ventral poles of the ependymal layer, respectively. Labeling of zebrafish roof plate cells, as well as tracing of roof plate cells using mouse embryos carrying *Wnt1-creERT*, revealed that roof plate cells are actually elongated along the dorso-ventral axis and transformed into dmNes+RGs, accompanying the reduction of the lumen [35,72]. Low-frequency labeling of roof plate cells, which enables the morphology of each cell to be distinguished, revealed that dmNes+RGs contact the surface of the ependymal layer. dmNes+RGs also maintain contact with the outer surface of the spinal cord, the pia mater. Thus, it seems probable that roof plate cells are stretched, maintaining contact with the central canal and the pia mater (Figure 2C). dmNes+RGs eventually become part of the dorsal (or posterior) median septum, a thin, dense septum dividing the dorsal side of the spinal cord [35]. In contrast, floor plate cells do not exhibit dynamic morphological changes (Figure 2). Rather, only a subset of floor plate cells is retained around the central canal, whereas other floor plate cells separate from the ependymal layer during reduction of the lumen [54].

## *3.2. Morphology and Roles of dmNes+RGs*

Along the stretching of dmNes+RGs, cytoskeletal structures are well developed. Electron microscopic analysis revealed enrichment of intermediate filament structure during this process [53,73]. This is consistent with enrichment of Nestin, which is a component of intermediate filaments, and directional organization of actin filaments in this process [35,63]. In zebrafish, inhibition of Zic6 or Rock impairs the stretching morphogenesis of roof plate with disruption of the actin cytoskeleton [72,74]. Thus, these cytoskeletal structures apparently contribute to formation of the physically robust structure of the dorsal median septum.

In the dorsal median septum, dmNes+RGs seem to act as a physical and molecular barrier, preventing decussation of developing long tracts of commissural axons [75,76]. In addition, dmNes+RGs apparently serve several different functions. For instance, *dreher* (*Lmx1a*-deficient) mice, which impair roof plate formation, show that dmNes+RGs regulate growth of long-range dorsal column axons [77,78]. Furthermore, as described below, dmNes+RGs promote proliferation of ependymal cells by producing Wnt ligands [35,70]. Thus, in addition to serving as a signaling center in early spinal cord development, roof plate has additional roles in later spinal cord development [10].

#### **4. Wnt Signaling in Morphological Transformation of Roof Plate Cells**

#### *4.1. Expression of Wnt Ligands and Activation of Wnt Signaling*

As described above, *Wnt1* and *Wnt3a*, which are secreted by roof plate cells, participate in development of neural crest cells and dorsal interneurons [19,34]. Whereas expression of *Wnt1* and *Wnt3a* mRNA is detected in the roof plate of developing mouse spinal cord until E12.5 [15,79], using in situ hybridization with a digoxygenin-based probe, it had been difficult to judge whether expression of these Wnt genes is maintained after E12.5, because roof plate cells become long and thin. However, recent immunostaining analysis revealed that expression of both Wnt1 and Wnt3a proteins remains in elongating roof plate cells, dmNes+RGs, after E13.5 [35]. This persistent expression was also confirmed with knock-in mouse embryos in which endogenous *Wnt3a* is replaced by *egfp-Wnt3a.* Consistent with Wnt expression in dmNes+RGs, activation of Wnt/β-catenin signaling is evident in dmNes+RGs [35,70].

## *4.2. Functions of Wnt Ligands Secreted by dmNes+RGs*

Since most *Wnt1-* and *Wnt3a-*double mutants die before E12.5 [19], genetic studies of *Wnt1-* and *Wnt3a*-deficient mutant mouse embryos yield no information regarding their roles in later stages. Thus, the function of Wnt signaling in the morphological transformation from roof plate cell to dmNes+RG was investigated by generating roof plate-specific conditional knock-out of the *Wls/Evi/Sprinter* gene, which is specifically required for secretion of Wnt proteins [80–82]. In normal embryos, prior to elongation of the apical processes of roof plate cells, the apical surfaces of roof plate cells are constricted, causing cell shape to become wedge-like at E10.5 (Figure 2A). At E13.5, when reduction of the lumen starts, apical processes of roof plate cells begin to elongate in parallel in a dorsoventral direc-

tion. Subsequently, these processes continue to elongate along the midline, accompanying reduction of the lumen, and nuclei of these cells become aligned on the midline [35,54] (Figure 2B,C). In *Wls* cKO embryos, the apical end of each process is also attached to the dorsal pole of the central canal, but these processes are not aligned along the midline. Rather, they extend laterally, slightly away from the midline. As a result, dorso-ventral nuclear alignment is impaired in mutant embryos. In addition, the bundle of processes of dmNes+RGs appears thinner and is frequently branched in mutant embryos, although the cytoskeletons are properly oriented [35]. These data suggest that mechanical tension in roof plate cells regulated by Wnt signaling may control the coordinated rearrangement of roof plate. However, since the number of roof plate cells is slightly increased in *Wls* cKO embryos at E13.5, alternative possibility that the rearrangement of roof plate cells is disrupted by crowding of roof plate cells cannot be excluded. Thus, Wnt secreted by roof plate cells is required for the change in morphology of these cells along the midline (Figure 3B). lumen starts, apical processes of roof plate cells begin to elongate in parallel in a dorsoventral direction. Subsequently, these processes continue to elongate along the midline, accompanying reduction of the lumen, and nuclei of these cells become aligned on the midline [35,54] (Figure 2B,C). In *Wls* cKO embryos, the apical end of each process is also attached to the dorsal pole of the central canal, but these processes are not aligned along the midline. Rather, they extend laterally, slightly away from the midline. As a result, dorso-ventral nuclear alignment is impaired in mutant embryos. In addition, the bundle of processes of dmNes+RGs appears thinner and is frequently branched in mutant embryos, although the cytoskeletons are properly oriented [35]. These data suggest that mechanical tension in roof plate cells regulated by Wnt signaling may control the coordinated rearrangement of roof plate. However, since the number of roof plate cells is slightly increased in *Wls* cKO embryos at E13.5, alternative possibility that the rearrangement of roof plate cells is disrupted by crowding of roof plate cells cannot be excluded. Thus, Wnt secreted by roof plate cells is required for the change in morphology of these cells along the midline (Figure 3B).

roles in later stages. Thus, the function of Wnt signaling in the morphological transformation from roof plate cell to dmNes+RG was investigated by generating roof plate-specific conditional knock-out of the *Wls/Evi/Sprinter* gene, which is specifically required for secretion of Wnt proteins [80–82]. In normal embryos, prior to elongation of the apical processes of roof plate cells, the apical surfaces of roof plate cells are constricted, causing cell shape to become wedge-like at E10.5 (Figure 2A). At E13.5, when reduction of the

*J. Dev. Biol.* **2021**, *9*, x FOR PEER REVIEW 7 of 13

**Figure 3.** Multiple functions of Wnt ligands secreted from roof plate cells have multiple functions during spinal cord development. (**A**) Before stretching, roof plate cells exhibit a wedge-shaped form (cyan) and act as a source of neural crest cells (yellow), forming the dorsal organizing center. At this stage, Wnt signaling in roof plate regulates formation and fate **Figure 3.** Multiple functions of Wnt ligands secreted from roof plate cells have multiple functions during spinal cord development. (**A**) Before stretching, roof plate cells exhibit a wedge-shaped form (cyan) and act as a source of neural crest cells (yellow), forming the dorsal organizing center. At this stage, Wnt signaling in roof plate regulates formation and fate determination of neural crest cells (1), as well as proliferation and specification of dorsal interneuron progenitors (2). In the dorsal spinal cord, Wnt signaling is likely to be activated in a gradual manner (orange). (**B**) After stretching, roof plate cells extend long processes (cyan) in contact with the central canal and pial surface. During this later stage, Wnt signaling in stretching roof plate cells regulates transformation of roof plate cells themselves (3) and proliferation of ependymal cells (4). In ependymal cells, Wnt signaling is likely to be activated in a gradual manner (red).

#### **5. Wnt Signaling in Development of Cells Surrounding the Central Canal**

#### *5.1. Wnt Signaling in Ependymal Cells*

In the ependymal layer surrounding the central canal, Wnt/β-catenin signaling is activated [35,70]. This activation depends partly on Wnt secretion from dmNes+RGs, apical ends of which face the dorsal surface of the central canal. Actually, embryos in which *Wls* function is defective, specifically in stretched dmNes+RGs, exhibit a significant reduction in Wnt-active cells among ependymal cells at E15.5 and E18.5. Since this reduction is more severe in dorsal ependymal layer zone cells, it seems plausible that Wnt proteins from dmNes+RGs are specifically required for dorsal activation of Wnt/β-catenin signaling [35].

On the other hand, the Gene Expression Nervous System Atlas (GENSAT) Project revealed that several Wnt ligands, including *Wnt2b*, *Wnt3*, *Wnt7a* and *Wnt8b,* are expressed in spinal cord, except dmNes+RGs at E15.5 [83]. Thus, it also seems possible that Wnt proteins produced by some other cells, other than descendants of roof plate cells, are involved in activation of Wnt signaling in ependymal cells. In addition, since it was reported that a few Wnt-positive ependymal cells originated from the roof plate, it is also possible that these cells still maintain Wnt activity in the ependymal layer [70]. Actually, expression of *Wnt1* and *Wnt3a* is detectable in the ependymal layer shortly after birth, consistent with activation of Wnt/β-catenin signaling. In adults, many Wnt ligands, including *Wnt1*, *Wnt3a*, *Wnt5a* and *Wnt11*, are expressed in ependymal cells and expression of *Axin2*, a downstream target of Wnt/β-catenin signaling, is detected, indicating activation of Wnt/β-catenin signaling in these cells [70].

#### *5.2. Proliferation of Ependymal Cells*

In the spinal cord, a small number of ependymal cells constitutively proliferate [84], and these cells are concentrated dorsally [56]. Since Wnt/β-catenin regulates proliferation of neural progenitor/stem cells in the brain [85–87], it seems probable that this signaling is also involved in regulation of progenitor/stem cells in the spinal cord. A recent study showed that Wnt signaling actually appears to be involved in promoting proliferation of ependymal cells [35,70]. At E18.5, *Wls* cKO mutant embryos exhibited significant decreases in the frequency of Ki67-positive proliferating cells in the ependymal layer. Furthermore, this defect is more severe in the dorsal half, suggesting that Wnt secretion by dmNes+RGs is required for normal proliferation of ependymal cells in embryos [35] (Figure 3B).

#### *5.3. Adult Spinal Cord*

In adult mice, neurogenesis hardly occurs in the spinal cord. However, after injury to the spinal cord, ependymal cells proliferate and generate progeny that undergo multiple fates, suggesting that ependymal cells exhibit latent neural stem cell properties [55,61]. This injury-induced proliferation of ependymal cells is suppressed in *Wls* cKO mice, and as in embryos, cells in the dorsal half of the ependymal layer exhibit more severe defects [35]. Thus, Wnt secretion by roof plate cells is also required for proliferation of ependymal cells after injury to the spinal cord.

Moreover, secretion of Wnt and expression of *β-catenin* in ependymal cells is required for proliferation of ependymal cells in postnatal and adult mice. Disruption of *β-catenin* or *Wls* in *Axin2*-expressing ependymal cells in adult mice significantly reduces the frequency of Ki67-positive proliferating cells in the ependymal layer [70]. Thus, activation of Wnt/βcatenin signaling and secretion of Wnt ligands is required for proliferation of ependymal cells in adult mouse spinal cord.

#### **6. Conclusions and Future Perspectives**

In conclusion, in early development of the spinal cord, the most dorsal cells act as a source of neural crest cells and constitute the dorsal organizing center, the roof plate, that regulates interneuron specification and proliferation. These functions are regulated by Wnt signaling in roof plate (Figure 3A). After neural specification, roof plate cells transform to dmNes+RGs, controlled by Wnt/β-catenin signaling. Moreover, Wnt ligands

are still expressed in stretched roof plate cells and activate Wnt/β-catenin signaling in the ependymal layer (Figure 3B). Activation of Wnt/β-catenin signaling in ependymal cells promotes proliferation of neural progenitor and stem cells in embryonic and adult spinal cord. Thus, during spinal cord development, roof plate cells are a source of Wnt ligands and activate Wnt/β-catenin in themselves, as well as in surrounding cells that dramatically change morphology and function.

Despite the function of roof plate cells in each stage of spinal cord development, regulatory mechanism of morphological change of roof plate cells by Wnt signaling still remain unclear. An interesting question for future studies is whether roof plate cells elongate their processes or are pulled by surrounding cells during transformation of roof plate cells. Given that the cytoskeleton regulates transformation of roof plate cells, a remaining challenge is to understand the mechanism by which Wnt signaling controls the cytoskeleton. In general, *Wnt1* and *Wnt3a* act as ligands to activate Wnt/β-catenin signaling, which directly regulates transcription but not cytoskeletal reorganization. Thus, understanding of the regulatory mechanism of roof plate transformation may provide a new insight into the Wnt signaling pathway in cell biology.

In the adult spinal cord, Wnt/β-catenin signaling promotes cell proliferation in the ependymal layer in normal condition and after spinal cord injury. Pool of quiescent neural progenitor and stem cells in ependymal layer contribute to the regeneration in response to spinal cord injury. However, it remains unclear which type of the cells proliferate in ependymal layer in a Wnt/β-catenin signaling dependent manner and contribute to the regeneration after spinal cord injury. Elucidating the mechanism of cell proliferation and differentiation in the ependymal layer regulated by Wnt/β-catenin signaling may help to understand the mechanism of regeneration after spinal cord injury.

**Author Contributions:** T.S. and S.T. contributed to the manuscript. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JSPS) to T.S. (20K15809) and S.T. (17H05782, 18H02454, 19H04797).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We thank all members of S.T.'s laboratory for their helpful discussions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Craniofacial Phenotypes and Genetics of DiGeorge Syndrome**

**Noriko Funato**

Department of Signal Gene Regulation, Advanced Therapeutic Sciences, Medical and Dental Sciences, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University (TMDU), Tokyo 113-8510, Japan; noriko-funato@umin.ac.jp

**Abstract:** The 22q11.2 deletion is one of the most common genetic microdeletions, affecting approximately 1 in 4000 live births in humans. A 1.5 to 2.5 Mb hemizygous deletion of chromosome 22q11.2 causes DiGeorge syndrome (DGS) and velocardiofacial syndrome (VCFS). DGS/VCFS are associated with prevalent cardiac malformations, thymic and parathyroid hypoplasia, and craniofacial defects. Patients with DGS/VCFS manifest craniofacial anomalies involving the cranium, cranial base, jaws, pharyngeal muscles, ear-nose-throat, palate, teeth, and cervical spine. Most craniofacial phenotypes of DGS/VCFS are caused by proximal 1.5 Mb microdeletions, resulting in a hemizygosity of coding genes, microRNAs, and long noncoding RNAs. *TBX1*, located on chromosome 22q11.21, encodes a T-box transcription factor and is a candidate gene for DGS/VCFS. TBX1 regulates the fate of progenitor cells in the cranial and pharyngeal apparatus during embryogenesis. *Tbx1*-null mice exhibit the most clinical features of DGS/VCFS, including craniofacial phenotypes. Despite the frequency of DGS/VCFS, there has been a limited review of the craniofacial phenotypes of DGC/VCFS. This review focuses on these phenotypes and summarizes the current understanding of the genetic factors that impact DGS/VCFS-related phenotypes. We also review DGS/VCFS mouse models that have been designed to better understand the pathogenic processes of DGS/VCFS.

**Keywords:** 22q11.2 deletion syndrome; DiGeorge syndrome; velocardiofacial syndrome; cleft palate; skull base; cleidocranial dysplasia; hyoid bone; teeth abnormalities

#### **1. Introduction**

The 22q11.2 deletion syndrome is one of the most common chromosomal microdeletions, affecting approximately 1 in 4000 live births in humans [1]. A 1.5 to 2.5 Mb hemizygous deletion of chromosome 22q11.2 causes DiGeorge syndrome (DGS; OMIM #188400) and velocardiofacial syndrome (VCFS or Shprintzen VCF syndrome; OMIM #192430) [2]. DGS/VCFS appears to be a genomic disorder distinct from 22q11.2 distal deletion syndrome (OMIM #611867). The clinical phenotype of DGS/VCFS is a complex and variable congenital disability, including cardiovascular defects, thymic hypoplasia, parathyroid hypoplasia, and craniofacial malformations [3]. Craniofacial malformations occur in approximately 60% of patients with DGS/VCFS [4].

*TBX1*, located on chromosome 22q11.21, encodes a T-box transcription factor and is considered a candidate gene for DGS/VCFS since mutations in *TBX1* have been found in patients with DGS/VCFS [5]. Heterozygous *Tbx1*-mutant (*Tbx1*+/−) mice exhibit DGS/VCFSrelated cardiovascular, parathyroid, and thymic phenotypes, suggesting that *TBX1* dosage is critical for cardiovascular, parathyroid and thymic development [6–9]. *Tbx1*-null mice exhibit the most clinical features of DGS/VCFS, including craniofacial phenotypes, while *Tbx1***+/**<sup>−</sup> mice exhibit no significant craniofacial phenotypes [6–10].

There have been some excellent reviews on genetics and cardiovascular anomalies of DGS/VCFS [3,11–13]. However, information on the craniofacial anomalies of DGS/VCFS is limited. This review focuses on these phenotypes and summarizes the current understanding of the genetic factors that impact DGS/VCFS-related phenotypes. We also review

**Citation:** Funato, N. Craniofacial Phenotypes and Genetics of DiGeorge Syndrome. *J. Dev. Biol.* **2022**, *10*, 18. https://doi.org/ 10.3390/jdb10020018

Academic Editors: Hideyo Ohuchi and Tsutomu Nohno

Received: 21 April 2022 Accepted: 11 May 2022 Published: 13 May 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

DGS/VCFS mouse models that have been designed to better understand the pathogenic processes of DGS/VCFS.

#### **2. Craniofacial Phenotypes of Patients with DGS/VCFS**

Patients with DGS/VCFS manifest craniofacial anomalies involving the cranium, cranial base, jaws, pharyngeal muscles, ear-nose-throat, palate, teeth, and cervical spine (Figure S1, Tables 1 and 2). Frequently observed craniofacial phenotypes include velopharyngeal insufficiency (27–92%), enamel hypomineralization (39–41%), hearing loss (33–39%), platybasia (50–91%), and cervical spine anomalies (75%) (Table 1). Delayed development of the hyoid bone has also been reported [14,15].


**Table 1.** Craniofacial anomalies in patients with DGS/VCFS.

Data were summarized from the following references: [16–22].

**Table 2.** Craniofacial and skeletal phenotypes of DGS/VCFS and *Tbx1*-null mice.



**Table 2.** *Cont.*

3 August 2021) and the Monarch Initiative (https://monarchinitiative.org accessed on 3 August 2021).

In addition to morphological anomalies, infants and young children with DGS/VCFS often exhibit a high prevalence of functional difficulties in feeding and speech/language associated with cleft palate, laryngeal anomalies, and velopharyngeal dysfunction [37]. Even after cleft palate closure, children with DGS/VCFS sometimes present communication disorders related to speech-language problems, such as articulation disorders of speech sounds and vocal disorders [37]. They exhibit slower language acquisition than those with other disorders that may be associated with abnormal muscle development.

#### **3. Genetics of DGS/VCFS**

DGS/VCFS is caused by a 1.5 to 2.5 Mb hemizygous deletion of chromosome 22q11.2 (Figure 1). Chromosomal microdeletions at 10p14-p13 (the *DGS2* locus) in patients with DGS/VCFS phenotypes are defined as the DGS/VCFS complex 2. In this review, we focus on the 22q11.2 locus, its associated genes, and miRNAs. *J. Dev. Biol.* **2022**, *10*, x FOR PEER REVIEW 5 of 18

**Figure 1.** Proximal deletions of chromosome 22q11.2 are responsible for the clinical features of DGS/VCFS. Snapshot of the UCSC Genome Browser (http://genome.ucsc.edu accessed on 3 August 2021) in the hg38 assembly showing the genomic context in the proximal deletions of chromosome 22q11.2. Top, the 25 kb resolution Hi-C data in H1 human embryonic stem cell line (H1-hESC). Bottom, the coding (blue) and noncoding RNAs (green), including miRNAs and long noncoding RNAs, are shown. **Table 3.** DGS/VCFS-associated variants of *TBX1.* **Figure 1.** Proximal deletions of chromosome 22q11.2 are responsible for the clinical features of DGS/VCFS. Snapshot of the UCSC Genome Browser (http://genome.ucsc.edu accessed on 3 August 2021) in the hg38 assembly showing the genomic context in the proximal deletions of chromosome 22q11.2. Top, the 25 kb resolution Hi-C data in H1 human embryonic stem cell line (H1-hESC). Bottom, the coding (blue) and noncoding RNAs (green), including miRNAs and long noncoding RNAs, are shown.

**Anomalies References**

c.443T>A (F148Y) T-box Conotruncal anomaly face syndrome Yes [5]

c.292A>T N-terminal DiGeorge syndrome Yes ClinVar Variant: 526036 c.385G>A T-box Tetralogy of Fallot No ClinVar Variant: 488618

**Mutation Domain Condition Craniofacial** 

Most of the chromosomal deletions of the 22q11.2 locus are de novo, but inherited deletions of the 22q11.2 locus have been reported in 6–28% of patients as autosomal dominant [16,17]. The majority of clinical phenotypes of DGS/VCFS are caused by proximal 1.5 Mb microdeletions [3,22], resulting in a hemizygosity of approximately 30 coding genes, including *DGCR6, PRODH, DGCR2, ESS2, TSSK2, GSC2, FAM246C, SLC25A1, CLTCL1, UFD1, HIRA, CDC45, MRPL40, C22orf39, CLDN5, TBX1, SEPTIN5, SEPT5-GP1BB, GP1BB, GNB1L, RTL10, TXNRD2, COMT, ARVCF, TANGO2, TRMT2A, RANBP1, CCDC188, DGCR8, ZDHHC8, RTN4R, DGCR6L,* and *C007326*, as well as microRNAs (miRNAs) and long noncoding RNAs (Figures 1 and S2A). The Hi-C chromatin structure of the 1.5 Mb region indicates interactions between these loci and their neighboring regions (Figure 1).

#### *3.1. TBX1 Gene*

The proximal deletion of 1.5 Mb on the 22q11.2 locus includes *TBX1* (Figure 1). *TBX1* is considered a candidate gene of DGS/VCFS because haploinsufficiency of *TBX1* leads to the typical phenotypes of DGS/VCFS, conotruncal anomaly face syndrome (OMIM #217095), and tetralogy of Fallot (OMIM #187500) (Table 3). Identical mutations in *TBX1* present among patients resulted in distinct phenotypes, suggesting that genetic and epigenetic changes or environmental factors are involved in the clinical phenotypes [5]. The coding variants in the T-box and C-terminal domains of TBX1 showed high combined annotationdependent depletion (CADD) scores (Table S1); however, further investigation is required to confirm that the variants cause DGS/VCFS and how they impact the phenotypes.


**Table 3.** DGS/VCFS-associated variants of *TBX1.*

ClinVar (https://www.ncbi.nlm.nih.gov/clinvar accessed on 3 August 2021).

*3.2. DiGeorge Syndrome Critical Region (DGCR)*

DGCR8, DGCR6, and DGCR6L map to the commonly deleted 1.5 Mb region in DGS/VCFS (Figure 1). DGCR8 is a nuclear miRNA-binding protein required for miRNA biogenesis. Dgcr8 haploinsufficiency in mice reduces the expression of miRNAs in the brain [45]. DGCR6 and DGCR6L genes encode a protein with a sequence similar to the Drosophila gonadal [46] (Figure S2B). In a chicken model, targeting DGCR6 function resulted in a vascular phenotype [47]. Attenuation of DGCR6 affects the expression of three genes localized within the 1.5 Mb region, upregulating the expression of TBX1 and UFD1 and reducing the expression of HIRA in the heart and pharyngeal arches of the chicken embryos [47]. Thus, the haploinsufficiency of DGCR8 or DGCR6 may be linked to DGS/VCFS phenotypes when targeting DGS/VCFS-related genes and miRNAs.

#### *3.3. MicroRNAs*

The deleted 1.5 Mb on the 22q11.2 locus includes several miRNAs, such as miR-185, miR-4716, miR-3618, miR-1286, miR-1306, and miR-6816 (Figure 1). The TargetScan miRNA target prediction program (http://www.targetscan.org accessed on 3 August 2021) identified that the 30 UTR of *TBX1* includes conserved sites for miR-183-5p, miR-96-5p, miR-1271-5p, miR-182-5p, miR-144-3p, miR-139-5p, miR-101-3p, and miR-451. Two miR-NAs were confirmed to target the 30 UTR of *TBX1*. miR-96-5p represses *Tbx1* expression and, in turn, TBX1 suppresses the promoter activity and expression of miR-96 [48]. miR-451a, a tumor suppressor, also directly targets *TBX1* [49]. The expression of this gene is upregulated in cutaneous basal cell carcinoma, inversely to miR-451a [49]. miR-17-92 fine-tunes the expression of *Tbx1* in craniofacial development, suggesting miR-17-92 as a candidate genetic modifier for *Tbx1* [50]. Thus, miRNAs both inside and outside the 22q11.2 locus may influence the severity of the clinical phenotypes of DGS/VCFS.

#### **4. Craniofacial Phenotypes of DGS/VCFS Mouse Models**

Mouse models with DGS/VCFS help identify additional candidate genes or modifier genes that influence the penetrance and/or severity of DGS/VCFS-related phenotypes. According to the mouse genome informatics (MGI) database (http://www.informatics.jax.org accessed on 3 August 2021), DGS/VCFS-related anomalies concerning Tbx1, Chrd, Tgfbr2, Vegfa, Fgf8, Crkl, Aldh1a2/Raldh2, Hoxa3, Kat6a/Moz/Myst3, Dicer1, Plxnd1, Dock1, Ndst1, Prickle1, Trappc10, Zfp366, and Foxn1 have been reported in genetically altered mice (Tables 4 and S2). When these genes were analyzed according to biological process, "heart morphogenesis" and "cranial skeletal system development" were enriched (Table S3). Our enrichment analysis using ToppCluster [51] indicated that genes associated with DGS/VCFS phenotypes in mice are specifically enriched in the morphogenesis of craniofacial tissues and heart (Figure 2A). Interestingly, among these genes, only *Tbx1* and *Chrd* were specifically enriched in the morphogenesis of cricoid and thyroid cartilages (Figure 2A). Genes associated with DGS/VCFS phenotypes in mice also indicated that DGS/VCFS-related phenotypes involve the interaction of several signaling pathways, including bone morphogenetic protein (BMP), transforming growth factor (TGF)β, vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF), and retinoic acid signaling pathways (Figure 2B). Genes involved in the genetic pathway of *Tbx1* are likely to induce phenotypes similar to *Tbx1*-null mice (Figure 2B, Tables 4 and S2). These are described below.


Mouse models of DiGeorge syndrome with phenotypic similarity to human diseases can be found in the Mouse Genome Informatics (MGI) database (http://www.informatics.jax.org accessed on 3 August 2021). Data were summarized from the following references [6–10,29–36,52–61]. nr, not reported. A detailed description is provided in Table S2. *J. Dev. Biol.* **2022**, *10*, x FOR PEER REVIEW 8 of 18

**Figure 2.** Interaction network of genes associated with DGS/VCFS phenotypes in mice. (**A**) A genebased network where each gene connects to a feature. The network was constructed using ToppCluster (https://toppcluster.cchmc.org/ accessed on 6 May 2022). Mouse phenotypes are shown in the network. (**B**) The protein–protein interaction network was constructed using the STRING tool (https://string-db.org/ accessed on 6 May 2022). Genes associated with DGS/VCFS phenotypes in mice (Table 4) were the input. Different colors represent different types of evidence of a connection between proteins. **Figure 2.** Interaction network of genes associated with DGS/VCFS phenotypes in mice. (**A**) A gene-based network where each gene connects to a feature. The network was constructed using ToppCluster (https://toppcluster.cchmc.org/ accessed on 6 May 2022). Mouse phenotypes are shown in the network. (**B**) The protein–protein interaction network was constructed using the STRING tool (https://string-db.org/ accessed on 6 May 2022). Genes associated with DGS/VCFS phenotypes in mice (Table 4) were the input. Different colors represent different types of evidence of a connection between proteins.

Craniofacial structures with DGS/VCFS phenotypes are derivatives of the head mesenchyme and the first and second pharyngeal arches [62]. *Tbx1* is expressed in the mesoderm, ectoderm, and endoderm of the pharyngeal apparatus and head mesenchyme be-

and otic vesicle epithelium [63,64]. *Tbx1*-null mice exhibit the most clinical features of DGS/VCFS, while *Tbx1***+/**<sup>−</sup> mice exhibit no significant craniofacial phenotypes (Tables 2, 5 and S2). Information about ocular phenotypes in *Tbx1*-mutant mice is limited (Table 2), although these anomalies in patients with DGS/VCFS have been reported [16,17]. The Cre/loxP system has been used with *Tbx1* conditional knockout mice to examine the tis-

**Spine**

sue-specific function of TBX1 in craniofacial development (Table 5).

**Table 5.** Selected craniofacial phenotypes of *Tbx1*-mutant neonates.

**Type Tissue/Cell Cranium Cranial Base Palate Mandible Hyoid Bone Cervical** 

*Tbx1*+/− Entire body Normal Normal Normal Normal Normal Normal *Tbx1-*null Entire body Abnormal Abnormal CP Hypoplastic Hypoplastic Abnormal

Pharyngeal tissues \* Abnormal Abnormal CP Hypoplastic Hypoplastic NA

Epithelium Normal Normal Anterior CP Normal Normal Normal

*Tbx1***-Mutant Mice Craniofacial Phenotypes**

*4.1. Tbx1*

**Mutation** 

Deletion (*Foxg1-Cre*)

Deletion (*KRT14-Cre*)

#### *4.1. Tbx1*

Craniofacial structures with DGS/VCFS phenotypes are derivatives of the head mesenchyme and the first and second pharyngeal arches [62]. *Tbx1* is expressed in the mesoderm, ectoderm, and endoderm of the pharyngeal apparatus and head mesenchyme between embryonic day (E)9.5 and E11.5 in mice [62,63]. At E12.5, *Tbx1* is expressed in the oral epithelium, the myogenic core of the tongue, incisor tooth buds, pharyngeal muscles, and otic vesicle epithelium [63,64]. *Tbx1*-null mice exhibit the most clinical features of DGS/VCFS, while *Tbx1***+/**<sup>−</sup> mice exhibit no significant craniofacial phenotypes (Tables 2, 5 and S2). Information about ocular phenotypes in *Tbx1*-mutant mice is limited (Table 2), although these anomalies in patients with DGS/VCFS have been reported [16,17]. The Cre/loxP system has been used with *Tbx1* conditional knockout mice to examine the tissue-specific function of TBX1 in craniofacial development (Table 5).

**Table 5.** Selected craniofacial phenotypes of *Tbx1*-mutant neonates.


Data were summarized from the following references: [30,31,34–36,65,66]. \* pharyngeal pouches, otic and optic vesicles [30,31]; \* pharyngeal endoderm, ectoderm, and mesoderm [65]; CP, cleft palate; NA, not available.

#### 4.1.1. Cleft Palate

During palatogenesis, the palatal shelves develop bilaterally from the internal parts of the maxillary prominences and fuse above the tongue to form an intact oral cavity roof [67,68]. Because the palate consists of a bone-lined hard palate and a bone-free soft palate, cleft palate phenotypes include incomplete and submucosal cleft palates [67,68]. Ablation of *Tbx1*, which is expressed in the epithelium of the palatal shelves, results in abnormal intraoral epithelial fusions between the palatal shelves and the mandible, resulting in various degrees of the cleft palate phenotype (complete, incomplete, and submucosal cleft palate) [30,34,69]. Expression of *Pax9*, whose mutations lead to cleft palate and tooth agenesis [70], is downregulated in the palatal shelves and pharyngeal region of *Tbx1*-null embryos [34,71]. In *Tbx1*-null palatal shelves, muscle- and bone-related genes are downregulated, whereas neuron- and collagen biosynthesis-related genes are upregulated [72].

#### 4.1.2. Abnormalities in Craniofacial Bones

*Tbx1*-null mice display craniofacial bone abnormalities, including persistently open fontanelles, micrognathia, a short clavicle, a hypoplastic zygomatic arch, and the absence of the hyoid bone (Tables 2 and S2). Conditional deletion of *Tbx1* in the mesoderm or osteochondral progenitors recapitulates the calvarial and mandibular phenotypes of *Tbx1*-null mice [35,66], suggesting that *Tbx1* is required for morphogenesis and ossification of craniofacial bones. Although *Tbx1* expression has not been reported in the neural crest, conditional deletion of *Tbx1* here results in a hypoplastic hyoid bone [35] (Tables 2 and 5). These results indicate that *Tbx1* is required for the morphogenesis and ossification of mesoderm- and neural crest-derived membranous bones, although malformations observed in most neural crest-derived bones of *Tbx1*-null mice are secondary defects induced by non-neural crest

cells [35,66]. Interestingly, abnormalities in membranous bones observed in *Tbx1*-null mice are similar to those of cleidocranial dysplasia (OMIM #119600 and #216330) in humans, exhibiting hypoplastic membranous bones, including abnormal neurocranial morphology, a short clavicle, a hypoplastic zygomatic arch, and hyoid bone [73–75]. Cleidocranial dysplasia (OMIM #119600) is caused by heterozygous mutations in *RUNX2*, which encodes a master transcription factor for osteoblast differentiation [74,75]. Since ablation of *Tbx1* affects *Runx2* expression in calvarial bones, and TBX1 overexpression induces *Runx2* expression in vitro [35], TBX1 may act upstream of *Runx2* by maintaining cell populations that express *Runx2* at the onset of bone development. In addition, *TBX1* could be a candidate gene for recessive inheritance of cleidocranial dysplasia (OMIM #216330).

#### 4.1.3. Abnormalities in the Cranial Base and Cervical Spine

The spheno-occipital synchondrosis (SOS) in the cranial base is a vital growth center for the skull (reviewed in [76]). TBX1 is expressed in the mesoderm-derived cartilage primordium of the SOS and basioccipital bones, and *Tbx1* deletion in the mesoderm induces malformed basioccipital bones and precocious ossified SOS. This indicates that *Tbx1* is an essential regulator of chondrocyte differentiation and subsequent ossification at the SOS [36]. TBX1 inhibits the transcriptional activity of RUNX2 in vitro as well as the expression of RUNX2 target genes in SOS [36]. *Tbx1*-null mice also exhibit endochondral bone abnormalities in the atlas, axis, and xiphoid process [6,35]. There is potential to examine the phenotypes of cranial synchondroses in DGS/VCFS patients, as abnormalities in the SOS and basioccipital bones may induce cranial phenotypes of DGS/VCFS, such as dolichocephaly, basilar impression, and platybasia.

#### 4.1.4. Dental Anomalies

Dental abnormalities (single central incisors, enamel hypoplasia, and small teeth) have been reported in many patients [18,28]. Accordingly, in approximately 30% of *Tbx1*-null mice, the upper incisors are absent [6]. *Tbx1* is expressed in the cervical loops, which contain the dental stem cell niche in mice. The cervical loop region of the incisor is either severely reduced or completely absent in *Tbx1*-null mice, and cultured incisors of *Tbx1*-null mice are hypoplastic and lack enamel [77]. Ablation of *Tbx1* in the epithelium results in smaller teeth than in the wild type, suggesting that TBX1 regulates the proliferation of dental progenitor cells [48].

#### 4.1.5. Muscle Hypotonia

Branchiomeric muscles are derived from the mesoderm of the pharyngeal arch. In *Tbx1* null and *Tbx1*flox/-;*Mesp1*-*Cre* embryos, the masseter, pterygoid, and temporalis muscles are intermittently absent [78,79]. Accordingly, muscle-related genes are also downregulated in *Tbx1*-null palatal shelves [72]. *Tbx1* acts upstream of critical transcription factors to form branchiomeric muscles. These include LIM homeobox protein 2 (*Lhx2*), transcription factor 21 (*Tcf21/capsulin*), musculin (*Msc*), myogenic factor 5 (*Myf5*), myogenic differentiation 1 (*Myod1*), myocyte enhancer factor 2C (*Mef2c*), and GATA binding protein 4 (*Gata4*) [79–82]. *Tbx1* is in the downstream genetic pathways of *Tcf21*, paired-like homeodomain transcription factor 2 (*Pitx2*), and ISL LIM homeobox 1 (*Isl1*) [80,83,84]. Thus, TBX1 regulates the pattern and development of branchiomeric muscles through the transcriptional regulation of myogenic genes.

#### *4.2. Chordin (Chrd) and Transforming Growth Factor, Beta Receptor II (Tgfbr2)*

Mice lacking the *Chrd* gene encoding chordin, an antagonist of bone morphogenetic proteins (BMPs), exhibit recapitulating phenotypes in *Tbx1*-null mice [32,52] (Table S2). *Chrd*-null neonates exhibit most craniofacial phenotypes in the cranium, cranial base, maxilla, mandible, ears, and hyoid bone (Table S2). Both *Tbx1* and *Fgf8* were reduced in the endoderm of *Chrd*-null mice, indicating that *Chrd* acts upstream of *Tbx1* and *Fgf8* [52]. *Tbx1* acts upstream of SMAD family member 7 (*Smad7)*, an inhibitory Smad within the

BMP/TGFβ pathway, to regulate vascular smooth muscle and extracellular matrix investment of the fourth arch artery [85]. Conditional deletion of *Tgfbr2*, which encodes TGFβ receptor 2, in the neural crest resulted in DGS/VCFS-related cardiovascular defects [53]. These findings suggest a potential role of BMP/TGFβ signaling in the pathogenesis of DGS/VCFS.

#### *4.3. Vascular Endothelial Growth Factor A (Vegfa)*

VEGFA is an essential cytokine in angiogenesis and vascular development during embryogenesis [86]. *Vegfa*-null neonates exhibit a few aspects of DGS/VCFS-related craniofacial anomalies, including unfused cranial sutures, absent incisors, and short mandibles, as well as cardiovascular abnormalities [54] (Table S2). The deletion of *Vegfa* in mice reduces *Tbx1* expression, and the knockdown of *vegfaa/vegfa* levels in zebrafish enhances the pharyngeal arch malformations induced by *tbx1* knockdown [54]. In humans, low expression of the *VEGFA* haplotype increases the risk of a cardiac phenotype of DGS/VCFS, indicating that expression levels of *VEGFA* affect the severity of DGS/VCFS phenotypes [87]. These results suggest that VEGFA modifies DGS/VCFS-related phenotypes by regulating *TBX1* expression.

#### *4.4. Fibroblast Growth Factor 8 (Fgf8) and FGF Receptor 2 (Fgfr2)*

Ablation of *Fgf8* induces craniofacial, cardiovascular, thymic, and parathyroid phenotypes [55,88]. *Fgf8*-null neonates exhibit a few aspects of DGS/VCFS-related craniofacial anomalies, including cleft palate and abnormal outer ear morphology [55,88] (Table S2). *Fgf8*+/−;*Tbx1*+/<sup>−</sup> double heterozygous embryos show an increased penetrance of cardiovascular defects compared with *Tbx1*-heterozygous embryos [89]. Tissue-specific deletion of *Fgf8* in *Tbx1*-expressing domains results in cardiovascular anomalies [90]. TBX1 activates the *Fgf8* enhancer during cardiac development [9]. Deletion of the *Fgfr2* gene that encodes FGF receptor 2 decreases *Tbx1* expression in the dental epithelium, indicating a genetic link between FGF signaling and *Tbx1* in tooth development [91]. In addition, a *Tbx1-Six1/Eya1-Fgf8* genetic pathway is crucial for craniofacial morphogenesis [92,93]. These findings demonstrate that the FGF pathway and *Tbx1* interact genetically during pharyngeal arch development.

#### *4.5. CRK like Proto-Oncogene, Adaptor Protein (Crkl)*

*CRKL* maps to the 2.5 Mb region commonly deleted in DGS/VCFS (Figure 1). Variants in a predicted enhancer of *CRKL* are significantly associated with the risk of congenital heart defects in DGS/VCFS [94]. Approximately 12% of *Crkl*-null mice show mild cranial bone defects, such as small cranium and poor membranous ossification of the nasal bones [56]. Compound heterozygosity of *Crkl* and *Tbx1* in mice has revealed that *Crkl* deletion enhances DGS/VCFS-related abnormalities compared with *Tbx1*-heterozygous embryos [56], suggesting that *Tbx1* and *Crkl* genes act in the same genetic pathway. *CRKL* encodes an adaptor protein that promotes the intracellular response of FGF signaling. *Crkl*+/−;*Fgf8*+/<sup>−</sup> double heterozygous mice showed DGS/VCFS-related defects [95]. Thus, *CRKL* mutations cause or modify DGS/VCFS-related phenotypes and/or penetrance as a contiguous gene syndrome.

#### *4.6. Aldehyde Dehydrogenase Family 1, Subfamily A2 (Aldh1a2/Raldh2)*

Retinoic acid (RA), an active vitamin A derivative, is essential for various developmental processes in vertebrates. High levels of RA act as morphogens that cause phenocopies of DGS/VCFS by downregulating *Tbx1* expression in the pharyngeal apparatus [96,97]. RA levels are balanced by the RA-synthesizing enzyme aldehyde dehydrogenase (ALDH) and the Cyp26 RA-catabolizing enzyme [98,99]. Mouse embryos hypomorphic for *Aldh1a2/Raldh2* display DGS/VCFS-related cardiovascular, thymic, and parathyroid malformations [57]. Haploinsufficiency of *Aldh1a2/Raldh2* results in reduced embryonic synthesis of RA, increased levels of *Tbx1*, and accelerated recovery from arterial growth delay in *Tbx1*-heterozygous mice [100]. An inhibitor of the Cyp26 enzyme induces a phenocopy of DGS/VCFS in chick embryos [101]. In *Tbx1*-null mice, upregulated expression of *Aldh1a2/Raldh2* and downregulated expression of *Cyp26a1* have been observed [71].

Further interactions occur between RA signaling, *Crkl*, and *Tbx1*. The penetrance of thymic hypoplasia is reduced in *Crkl+/*−;*Tbx1+/*−;*Aldh1a2+/*<sup>−</sup> triple heterozygous embryos compared to *Crkl*+/−;*Tbx1*+/<sup>−</sup> mutants, suggesting that reducing the amount of RA may rescue the DGS/VCFS-related phenotype [102]. Thus, the levels of RA in embryogenesis could contribute to the phenotypic variability of DGS/VCFS.

#### *4.7. Homeobox A3 (Hoxa3)*

RA exposure increases the expression of *Hoxa3*, a gene which encodes a homeobox transcription factor, in the neural tube and pharyngeal apparatus [103]. Interestingly, *Hoxa3* null neonates show some aspects of the abnormalities of DGS/VCFS [58,104] (Table S2). Thus, *HOXA3* may be a genetic modifier of DGS/VCFS-related abnormalities.

#### *4.8. Kat6a/Moz/Myst3 (Lysine Acetyltransferase 6A) and Epigenetic Modifiers*

Homozygous mutation of *Kat6a/Moz/Myst3*, which encodes a histone acetyltransferase, leads to cardiovascular defects seen in DGS/VCFS and reduces *Tbx1* expression [59]. Treatment of pregnant mice with a histone demethylase inhibitor reportedly increased the methylation levels of histone H3 lysine K4 (H3K4) and partially rescued the cardiovascular phenotypes of *Tbx1*-heterozygous mice [105]. TBX1 regulates genes transcribed at a low level by recruiting lysine methyltransferase (KMT2C) and controlling monomethylation of H3K4 (H3K4me1) enrichment on chromatin [105]. In addition, TBX1 transcriptionally targets *Wnt5a* by interacting with SMARCD1/BAF60a, a component of the SWI/SNF-like BAF chromatin remodeling complex, along with the H3K4 monomethyltransferase SETD7 [106]. Microduplication in *KANSL1*, which encodes a member of the histone acetyltransferase complex, is associated with heart anomalies in individuals with DGS/VCFS [107]. In T cells of patients with DGS/VCFS, the status of transcriptional activation (H3K4me3 and H3K27ac) is globally increased [108]. Thus, epigenetic changes are involved in DGS/VCFSrelated phenotypes.

#### *4.9. Sonic Hedgehog (Shh)*

*Shh* encodes an SHH signaling molecule. In humans, *SHH* mutations lead to holoprosencephaly 3 (OMIM #142945), microphthalmia with coloboma (OMIM #611638), and single median maxillary central incisor (OMIM #147250). *Shh*-null embryos exhibit conotruncal and pharyngeal arch artery defects similar to those observed in DGS/VCFS and *Tbx1*-null embryos [109]. *Tbx1* expression is reduced in *Shh*-null embryos, and ectopic expression of *Shh* can result in the upregulation of *Tbx1*, suggesting that *Shh* is a possible modifier for DGS/VCFS [62,110]. *Shh* is also required for the expression of the Fox family of transcription factor genes, forkhead box A2 *(Foxa2*) and forkhead box C2 *(Foxc2*), in the head mesenchyme and the pharyngeal endoderm [62]. FOXA2 and FOXC2 bind to regulatory regions in the mouse and human *TBX1* loci [111].

#### *4.10. Paired-like Homeodomain Transcription Factor 2 (Pitx2)*

*Pitx2* gene encodes a bicoid-like homeodomain transcription factor. *Pitx2*-null mice show craniofacial defects, such as the arrest of tooth development, abnormal morphology of maxilla and mandible, and cleft palate. In humans, *PITX2* mutations lead to Axenfeld– Rieger syndrome, type 1 (OMIM #180500). Patients with Axenfeld–Rieger syndrome manifest dental and craniofacial anomalies involving the maxilla, mandible, and cranial base [112]. Both *Tbx1* and *Pitx2* are expressed in the early dental epithelium, oral epithelium, and secondary heart field [64,113,114]. *Tbx1*+/−;*Pitx2*+/<sup>−</sup> double heterozygous embryos exhibit increased penetrance of an extra premolar-like tooth [115] and DGS/VCFSrelated cardiovascular anomalies [114]. TBX1 directly activates the *Pitx2c* enhancer through

the synergistic action of the homeobox-containing transcription factor NK2 homeobox 5 (NKX2-5) [114]. TBX1 also interacts with PITX2 and represses PITX2 transcriptional activity [48,115]. Thus, *PITX2* may be a genetic modifier of DGS/VCFS-related abnormalities.

#### **5. Discussion**

The penetrance and severity of congenital anomalies are related to genetic and environmental factors. Recent studies have revealed the function of TBX1 and modifiers that impact the severity and penetrance of DGS/VCFS. Studies of DGS/VCFS mouse models have provided insights into signaling pathways and genes that interact with TBX1 and/or affect the DGS/VCFS phenotypes. In addition, mouse models with DGS/VCFS may help us to identify additional DGS/VCFS-related phenotypes. For example, there is potential to examine the phenotypes of cranial synchondroses, cranium, zygomatic arches, and pharyngeal muscles in DGS/VCFS patients. We also noted that information about ocular phenotypes in *Tbx1*-mutant mice is limited, although these anomalies in patients with DGS/VCFS have been reported [16,17]. Crosstalk with key embryonic signals, especially BMP, TGFβ, VEGFA, FGF, RA, and SHH, critically regulates DGS/VCFS-related pharyngeal development. Genes involved in these signaling pathways may modify the phenotypic spectrum of DGS/VCFS. Given the broad spectrum of DGS/VCFS disease phenotypes, other genes essential to craniofacial development could modify the phenotypic spectrum. Genetically engineered mice are useful for studying disease phenotypes; however, ablation of essential genes involved in cardiovascular development may cause early embryonic lethality, which would prevent observation of craniofacial phenotypes. For example, ablation of *Ufd1*, whose human ortholog has been mapped to the 1.5 Mb region, causes early embryonic lethality before organogenesis in mice [116]. It is also essential to identify novel proteins that interact with TBX1 and examine whether interacting partners may influence the phenotypes of mouse models.

#### **6. Conclusions**

Studies of *Tbx1*-mutant mice have provided insights into the underlying pathogenesis of DGS/VCFS and the knowledge to diagnose patients with DGS/VCFS. Genes, miRNAs, and epigenetics could change *Tbx1* expression. Polymorphisms, variations, and mutations in *TBX1* may induce the penetrance and severity of DGS/VCFS-like craniofacial phenotypes. The molecular basis of the variant sequence of *TBX1* will further define how *TBX1* contributes to the craniofacial and other phenotypes of DGS/VCFS. Since interactions with TBX1 and other molecules in transcriptional complexes or chromatin remodeling are crucial for TBX1 function, identifying and understanding these genetic and epigenetic modifiers individually for each patient may direct therapeutics to minimize the severity.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/jdb10020018/s1, Figure S1: Craniofacial and skeletal phenotypes of DGS/VCFS. Figure S2: Human genes in the proximal deletion of 1.5 Mb on the 22q11.2 locus. Table S1: Craniofacial and skeletal phenotypes of DGS/VCFS and Tbx1-null mice. Table S2: Craniofacial and skeletal phenotypes in mouse models of DGS/VCFS. Table S3: Classification of mouse genes associated with DGS/VCFS.

**Author Contributions:** N.F. contributed to the conceptual idea, performed the database searches, analyzed the data, and wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Japan Society for the Promotion of Science (JSPS) KAK-ENHI [20K09901].

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** We would like to thank Hiroshi Kurosaka, Cedric Boeckx, and Mizuki Funato for a critical reading of the manuscript.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**

