The versatility of RhoA activities in neural differentiation

ABSTRACT

In this commentary we discuss a paper we published recently on the activities of the GTPase RhoA during neural differentiation of murine embryonic stem cells, and relate our findings to previous studies. We narrate how we found that RhoA impedes neural differentiation by inhibiting the production as well as the secretion of noggin, a soluble factor that antagonizes bone morphogenetic protein. We discuss how the questions we tried to address shaped the study, and how embryonic stem cells isolated from a genetically modified mouse model devoid of Syx, a RhoA-specific guanine exchange factor, were used to address them. We detail several signaling pathways downstream of RhoA that are hindered by the absence of Syx, and obstructed by retinoic acid, resulting in an increase of noggin production; we explain how the lower RhoA activity and, consequently, the sparser peri-junctional stress fibers in Syx^−/− cells facilitated noggin secretion; and we report unpublished results showing that pharmacological inhibition of RhoA accelerates the neuronal differentiation of human embryonic stem cells. Finally, we identify signaling mechanisms in our recent study that warrant further study, and speculate on the possibility of manipulating RhoA signaling in combination with other pathways to drive the differentiation of neuronal subtypes.

Introduction

Stem cell research has been dominated, understandably, by a focus on transcriptional regulation¹ and transcription factor (TF) wiring diagrams.² The relevance of the small GTPase RhoA³ to this field may have appeared, therefore, unlikely. The RhoA signaling pathway received attention primarily because one of its well-studied effectors, Rho-associated protein kinase (ROCK),⁴ was identified as the main culprit in the apoptosis of dissociated human embryonic stem cells (hESCs),⁵ a major impediment to research in this field. But even before their detrimental effect on ESC survival was recognized, RhoA and ROCK had been known for their antagonism of neurite growth⁶ – the first step of the morphological polarization of differentiating neural precursor cells (NPCs). Whereas the former is an acute and fast-acting event, the latter is a gradual and long-term process. Though the importance of RhoA in neural development had been well established,⁷ RhoA-dependent suppression of the differentiation of NPCs into neurons is a more recent observation.^8,9 The molecular mechanism that underlies the dependence of hESCs survival on RhoA/ROCK has been partially deciphered,^10,11 but whether a similar mechanism accounts for the inhibitory effect of RhoA on neural differentiation, or an entirely different one, remained unknown.

In a recent study, we provided evidence to substantiate multiple simultaneous molecular mechanisms that can account for the RhoA-dependent inhibition of neural differentiation of murine ESCs into neuronal precursor cells.¹² This has been the result of an unexpected observation on the differentiation of embryoid bodies (EBs) formed by ESCs of the Syx^−/− genotype, where the gene coding for the RhoA-specific guanine exchange factor (GEF) Syx^13,14 (also named Plekhg5 or Tech) had been disrupted¹⁵ We noticed that in comparison to Syx^+/+, the former EBs sprouted longer linear extensions, which, upon analysis with appropriate markers, were identified as NPCs. The differentiation into NPCs is not surprising in itself because it has been long known as the ESC default differentiation once BMP signaling is terminated.¹⁶ However, we realized that the accelerated neural differentiation of cells that are lacking Syx, a RhoA activator, is the opposite of the inhibitory effect of RhoA on neural differentiation, and sought to identify the signaling pathways that transduce RhoA activity into inhibition of ESC to NPC differentiation. Surprisingly, we found that the inhibitory effect of RhoA was exerted by several unrelated mechanisms that suppressed both the production and secretion of the BMP antagonist noggin (Nog),¹⁷ a differentiation-promoting factor.¹⁸ In the following sections we narrate how this study evolved, what questions we asked, and what are the next questions it raises.

Inhibition of RhoA increases Nog production

The choice whether to remain in the pluripotent state or to differentiate into NPCs is determined by the balance between bone morphogenetic protein (BMP) 4 and Nog.¹⁷ The latter interferes with the former's binding to its cognate cell surface receptors, thus preventing the downstream phosphorylation and activation of the small mothers against decapentaplegic (Smad1) transcription factors.¹⁹ Neural differentiation is driven also by retinoid acid (RA),²⁰ a vitamin A metabolite that promotes neural gene expression by binding to nuclear RAR-RXR heterodimers, which then induce gene transcription.²¹ Since we observed that neural differentiation was substantially faster in Syx^−/− ESCs, our objective was to find functional links between RhoA activity and the Nog and RA signaling pathways, though the differentiation of Syx^−/− ESCs was faster than that of Syx^+/+ ESCs even in the absence of RA. To be able to manipulate differentiation, we used EBs, an established in vitro model of this process.^22,23

Increase of Nog production

Because of the major role of Nog in neural differentiation, we addressed the possible relation between RhoA activity and the abundance of Nog. Differentiating Syx^−/− cells secreted more than twice the amount of Nog into the culture medium than Syx^−/− cells, supporting the existence of a RhoA-dependent effect on the production of Nog. We confirmed the connection between RhoA activity and the abundance of Nog by comparing Nog levels in differentiating ESCs that overexpressed a constitutively active (CA) RhoA mutant. The abundance of Nog was more than halved in the latter cells compared with cells overexpressing wild-type RhoA. Since we established that RARγ is required for Nog production, we asked if RhoA could reduce the production of RARγ, and confirmed that it does. We next asked how can RhoA inhibit RARγ signaling. Though no functional connection is known between the two, the RhoA effector rhophilin-2 (Rhpn2)²⁴ was shown to bind RARγ²⁵²⁵. Since the functional significance of Rhpn2 binding to RAR is unknown, we tested and confirmed that knocking down Rhpn2 increases Nog expression (Fig. 1A), suggesting that the binding inhibits RAR transcriptional activity.

Figure 1.

Schemes of the effects of disrupting Syx on Nog production in Syx^−/− cells, as described in the text. The schemes describe the effects of Syx disruption on Nog production through RARγ(A), pSmad1 (B), and Sirt1 (C). Solid lines represent direct regulatory events, whereas dashed lines represent events that either are indirect or have not been shown to be direct. X represents deactivation of a signaling step by the disruption of Syx. See text for details.

Reduction of Smad1 phosphorylation

Given the antagonistic effect of Nog on BMP4, the next signaling pathway we tested was the effect of RhoA on the level of Smad1 phosphorylation. This level was lower in RA-treated differentiating Syx^−/− ESCs by 70% in comparison to Syx^+/+ ESCs. How then can RhoA and RA signaling relate to each other? They may signal through entirely separate pathways, both of which inhibit Smad1 phosphorylation, or be interconnected. RA is known to cause degradation of pSmad1 through an increase in the production of growth arrest and DNA damage-inducible protein (GADD) 45, resulting in the activation of mitogen-activated protein kinase kinase (MAPKK), and, subsequently, of MAPK²⁶ (Fig. 1B). Our finding that pMapk was more abundant in Syx^−/− cells supported this premise. A separate RhoA-dependent pathway is suggested by previous findings whereby RhoA can increase Smad1 phosphorylation through the activation of ROCK and the resulting increase in stress-fiber density and mechanical tension in the cell.²⁷ Stress-fiber density was indeed lower in differentiating Syx^−/− cells (see next section), providing a possible connection between the reduced RhoA activity due to the absence of Syx, and the diminished pSmad1 abundance in these cells (Fig. 1B). The manner by which mechanical tension reduces pSmad1 abundance is unknown. However, we also found that RAR abundance was higher in Syx^−/− ESCs, suggesting that RhoA activity can reduce RAR production, thus interconnecting the RA and RhoA signaling pathways. Further experiments showed an antagonistic relation between pSmad1 and Nog levels, whereby knockdown of Nog increased pSmad1 abundance more than 2-fold, thus implicating the low pSmad1 production in the higher Nog abundance in differentiating Syx^−/− cells than in their Syx^+/+ counterparts (Fig. 1B).

Probable increased shuttling of RA into the nucleus

RA requires binding to the cellular retinoic acid-binding protein 2 (CRABP2) to be transported into the nucleus.²⁸ In turn, CRABP2 is deacetylated and inhibited by the NAD⁺-dependent deacetylase Sirtuin-1 (Sirt1).²⁹ Since Sirt1 abundance was lower in differentiating Syx^−/− cells, whereas its production was increased upon expression of CA RhoA, it is likely that the lower RhoA activity in Syx^−/− cells increased RA shuttling into the nucleus, thus augmenting the production of Nog and the transcription of the NPC marker Pax6,³⁰ as well as that of other known target genes (Fig. 1C).

Inhibition of RhoA increases Nog secretion

RhoA activity increases stress-fiber assembly by activating myosin-2 through ROCK,³¹ and by assembling actin filaments through diaphanous homolog (Diaph) 1.³² Our previous studies showed that Syx activates RhoA proximal to the cell junctions. The lower RhoA activity in Syx^−/− cells suggested, therefore, that the peripheral stress-fiber density could be lower in these cells compared with Syx^+/+ cells. Immunofluorescence imaging (Fig. 2) confirmed that stress fibers surrounding Syx^−/− cells were 5 times sparser than their Syx^+/+ counterparts. Nog is exocytosed by membrane trafficking dependent on the GTPase Rab3d.³³ Concordantly, the 2 proteins colocalized in Syx^+/+ and Syx^−/− cells. Nog and Rab3d were 4 and 6-fold more abundant, respectively, in Syx^+/+ cells, indicating that Nog exocytosis was greater from the Syx^−/− than from the Syx^+/+ cells. This possibility was substantiated by our finding that the abundance of Nog was more than 2-fold higher in the medium of Syx^−/− cells than in its Syx^+/+ counterpart. It is likely that the thicker band of peripheral stress-fibers in the Syx^+/+ cells impeded the outward movement of Nog-carrying vesicles, thus accounting for the lower Nog secretion from these cells.

Figure 2.

Immunofluorescence images of filamentous actin (F-actin) and Rab3d in RA-treated differentiating mESCs of the indicated genotypes. Note the thicker circumferential stress fibers and higher Rab3d abundance in the Syx^+/+ cells. Bar, 25 μm.

Extension to human ESCs

An obvious next step is to test if RhoA has a similar role in the differentiation of hESCs. The closest hESC model to the genetically modified mESCs we had used could be generated by CRISPR-Cas9 disruption of Syx. However, aside from the Syx-specific activation of RhoA close to the cell junctions,^34,35 and the ensuing impediment to Nog secretion, the inhibition of Nog production by RhoA may not necessarily be limited to activation of RhoA by Syx. As an initial approach, we used pharmacological inhibition of RhoA by exoenzyme C3 transferase.³⁶ C3 transferase treatment markedly increased the expression of several neural differentiation marker genes up to 75 or 10 fold at the end of the ESC to NPC differentiation and fate specification stage, and at the end of midbrain dopaminergic NPC expansion and maturation stage (stages 2 and 3, 7 and 14 d after the application of C3 transferase, respectively). C3 transferase-treated hESCs changed their morphology at the end of stage 2 from the tightly-packed polygonal shape typical of stem cells, to a more dispersed and polarized shape (Fig. 3A). Immunofluorescence imaging of the NPC differentiation marker tubulin β3 at the end of stage 2 revealed that by 14 days, C3 transferase-treated cells extended long axons, whereas a far smaller number of the untreated cells grew only short neurites (Fig. 3B).

Figure 3.

Phase (A) and immunofluorescence (B) mages of C3 exotransferase-treated and untreated (control) hESCs at the end of differentiation stage 2 (7 d after the start of treatment). Note the polarized shape (A) and the elongated axons (B) of the C3 exotransferase-treated cells. Bars, 25 μm in A, 100 μm in B.

Open questions

Several open questions remain concerning missing nodes in the signaling pathways we constructed, and the future implications of the current state of our study. The inhibition of RARγ by Rhpn2 (Fig. 1A) requires a more direct confirmation and a molecular mechanism to explain how Rhpn2 binding inhibits the transcriptional activity of RARγ. We have established that RARγ increases Nog production. (Fig. 1A), but we have not yet shown directly that RARγ promotes Nog transcription. There is still no mechanism to explain how ROCK increases the abundance of pSmad1 (Fig. 1B). It is unknown how Nog depletes pSmad1, and if the reverse effect occurs (Fig. 1B). It is similarly unknown how RhoA sustains Sirt1 abundance (Fig. 1C). In our study, the decrease in RhoA activity was imposed extraneously, either by disruption of Syx or by its pharmacological inhibition. It is tempting to ask, therefore, if normal neural differentiation requires endogenous inhibition of RhoA, and if so how it is achieved. It is similarly important to establish whether the inhibition of one GEF (e.g. Syx) is sufficient, or if more GEFs need to be inhibited to optimize neural differentiation. Further, the direct inhibition of RhoA (e.g., by C3 transferase as in our experiments) could be more effective than the inhibition of their GEFs to drive neural differentiation.

Interestingly, a very recent study by García-Mariscal et al. published in this journal observed that RhoA expression and ROCK activity were negatively related to RAR expression,³⁷ opposite to our findings.¹² The study of García-Mariscal et al. attributed the inhibitory effect of RhoA on RA signaling to an impediment in retinol metabolism. However, both studies found that reduced RhoA expression and activity increased the expression of genes involved in keratinocyte or in NPC differentiation, respectively. While García-Mariscal et al. found no change in RAR expression in RhoA-null cells, we found that the expression of CA-RhoA reduced RAR abundance. This difference could reflect dissimilarity between the expression profiles of keratinocytes and stem cells. Furthermore, a gene coding for a RhoA-specific GEF (Syx) was disrupted in the ESCs we used, rather than RhoA. Therefore, RhoA activity was only partially reduced in Syx^−/− ESCs, while it was completely absent in RhoA^−/− keratinocytes. Lastly, since García-Mariscal et al. found no difference in the expression level of genes involved in retinoid signaling and metabolism, they attributed the inhibition of RA metabolism to putative RhoA-dependent post-translational modifications, such as phosphorylation of retinol dehydrogenase by ROCK. In contrast, we detected substantial changes in the expression levels of several established RA target genes.

Our recent study¹² focused on the differentiation of murine and human ESCs into NPCs, but did not address the next step of differentiation into neuronal subtypes. Generation of such subtypes has been intensively studied, but current differentiation protocols produce heterogeneous neuron populations that consist not only of the desired subtype. Though the pharmacological inhibition of RhoA could possibly be further optimized, it is probably unlikely that manipulation of RhoA signaling alone can optimize the generation of a single neuronal subtype. It is more likely that combination of RhoA inhibition with the inhibition or activation of one or more signaling pathways – possibly Wnt signaling in the case of dopaminergic differentiation,³⁸ for example, may be necessary to optimize NPC differentiation into specific subtypes.

Abbreviations

BMP

bone morphogenetic protein

CRABP

cellular retinoic acid-binding protein

CA

constitutively active

Diaph

diaphanous homolog

EB

embryoid bodies

ESC

embryonic stem cell

GADD

growth arrest and DNA damage-inducible protein

GEF

guanine exchange factor

Mapkk

mitogen-activated protein kinase kinase

NPC

neural precursor cells

RA

retinoid acid

ROCK

Rho-associated protein kinase

Rhpn

rhophilin

Sirt

sirtuin

Smad

small mothers against decapentaplegic

TF

transcription factor

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

This study was supported by the USA National Institute of Health grants HL119984 (AH) and NS075839 (LI).