Dual spatio-temporal regulation of axon growth and microtubule dynamics by RhoA signaling pathways Free

Studies on cultured hippocampal and cortical neurons that develop in situ have established that the generation of an axon from a rather symmetric array of short and highly dynamic undifferentiated neurites is one of the early events underlying the establishment of morphological polarity (Caceres et al., 2012; Funahashi et al., 2014; Wilson et al., 2022). Current evidence favors the view that an increase in actin dynamics and growth cone size is behind the selection of a minor neurite to become the axon (Bradke and Dotti, 1997, 1999; Kunda et al., 2001). It is believed that dynamic microtubules (MTs) penetrate the loose growth cone microfilament network allowing transport of polarizing and/or growth-promoting factors to the distal domain that in turn activate membrane-associated signaling pathways (Dent and Baas, 2014; Wojnacki et al., 2014) leading to MTs stabilization (Witte et al., 2008; Montenegro-Venegas et al., 2010). The newly formed axon would then initiate a phase of rapid elongation and branching. Not surprisingly, a considerable effort has been devoted to identifying the molecular machinery involved in driving MT stability during axon formation (Conde and Caceres, 2009; van Beuningen and Hoogenraad, 2016). Nevertheless, key questions and mechanisms remain unsolved.

RhoA, a conspicuous small RhoGTPase family member (Hall and Lalli, 2010; Gonzalez-Billault, et al., 2012), has been implicated in neuronal polarization. RhoA, and its downstream effector Rho kinases (ROCK1 and ROCK2; hereafter collectively ROCK), exert an inhibitory tone preventing minor neurites from engaging in axon-like growth (Da Silva et al., 2003; Chuang et al., 2005; Kollins et al., 2009; Takano et al., 2019); accordingly, suppression of Lfc (also known as ARHGEF2), a RhoA guanosine-nucleotide exchange factor (GEF), promotes axon initiation whereas its upregulation has the opposite effect (Conde et al., 2010; Wilson et al., 2020a). Inhibition of RhoA–ROCK signaling also impairs oriented axon formation in situ (Xu et al., 2015). Thus, inactivation of RhoA appears to be sufficient to trigger axon formation and hence the establishment of neuronal polarity. However, a recent study has provided evidence suggesting that although RhoA restrains axon initiation and growth, it has no effect on axon specification (Dupraz et al., 2019).

Another set of evidence suggests a more complex function of RhoA during axonogenesis. For example, in cultured cerebellar granule neurons, high RhoA activity downstream of stromal cell-derived factor-1α (SDF-1α; also known as CXCL12) inhibits axonal elongation through ROCK. Surprisingly mild or low RhoA activity has been reported to have the opposite effect, promoting axonal elongation through a different effector, namely mDia1 (also known as DIAPH1) (Arakawa et al., 2003). A dual effect has also been reported in NGF-stimulated PC12 cells, where RhoA activation inhibits neurite outgrowth but, if activated afterwards, it promotes neurite elongation (Sebok et al., 1999). These results suggest that the function of RhoA is regulated both spatially and temporally. So far, the spatio-temporal activation pattern of RhoA during neuronal polarization remains unknown. There is no information regarding distinct RhoA effectors regulating different stages of axon development or over different neuronal compartments. Additionally, RhoA regulates MT stability and dynamics in other systems (Cook et al., 1998; Palazzo et al., 2001; Bartolini and Gundersen, 2010; Bartolini et al., 2012), but has not been tested for regulating MT dynamics or stability during axon growth.

Here, we report evidence of a dual role of RhoA signaling during de novo axonal growth. First, by using Förster resonance energy transfer (FRET)-based biosensors (Fritz et al., 2013; Li et al., 2017), we found that RhoA is highly active in growth cones of minor neurites and elongating axons; by contrast, low growth cone RhoA activity parallels the transformation of a minor neurite into an axon. Interestingly, ROCK activity is high in minor neurites, but low in both sprouting and elongating axons. Lysophosphatidic acid (LPA), a strong RhoA activator, prevents axon formation in stage 2 neurons (e.g. neurons with several minor neurites but no axon), but enhances elongation in neurons with established axon (e.g. stage 3 neurons); accordingly, RhoA suppression enhances axon outgrowth in early polarizing neurons but reduces axonal elongation in stage 3 polarized neurons. We find that diverging signaling downstream of RhoA is achieved by means of different effector proteins acting on MTs dynamics. Although ROCK mediates the inhibitory effects of RhoA on initial axon growth, mDia1 stimulates axonal extension by stabilizing MTs in the axonal growth cone.

In the first set of experiments, we used a first-generation RhoA-FRET biosensor (RhoA1G), based on cyan fluorescent protein (CFP) as donor and yellow fluorescent protein (YFP) as acceptor (Fig. S1; see also Fritz et al., 2013), to examine RhoA activity during the so-called second (growth) phase of polarization (Caceres et al., 2012), when a single axon is generated from an array of multiple short and highly dynamic minor neurites.

To this end, neuronal cultures were transfected at 2, 18, 24 and 48 h after plating with RhoA1G, fixed 18 h later and RhoA activity evaluated by FRET using ratiometric imaging. We validated the FRET assay using CHO cells transfected with constitutively active (Q63L), dominant-negative (T19N) and wild-type (WT) versions of the RhoA biosensor (Fig. S1A–C). In agreement to previous reports with this biosensor (Fritz et al., 2013), our measurements yielded an average relative-FRET difference between Q63L and T19N of ∼20% (Fig. S1D). In addition, the mutant biosensors showed no discernible spatial activation patterns, whereas the WT variant displayed the highest RhoA FRET signal at the leading edge of the cells (Fig. S1C, arrows), similar to what has been described for randomly migrating mouse embryonic fibroblasts (MEFs; Machacek et al., 2009).

A similar analysis performed in 1–3 days in vitro (DIV) hippocampal neurons, revealed that high RhoA activity was spatially confined to specific cellular domains (Fig. 1), despite RhoA1G YFP fluorescence being found homogenously distributed throughout the neuron, including the cell body, the neurites and their growth cones (Fig. S2; see also Gonzalez-Billault et al., 2012). A representative image showing color-coded RhoA activity before axon formation (multipolar stage-2 neurons) is shown in Fig. 1A. In such neurons (∼20 h in vitro) high RhoA activity was predominantly detected at tips (Fig. 1A, arrowheads) or growth cones (Fig. 1A, arrows) of minor processes. Remarkably, this pattern of RhoA activity changed dramatically upon onset of the transition from stage 2 to stage 3 (i.e. during axon sprouting) (stages in this paper are according to Caceres et al., 2012). The transformation of a minor neurite into an axon has been associated with an increase in growth cone size, accumulation of growth promoting factors and reorganization of the MTs and actin cytoskeletons (Bradke and Dotti, 1997, 1999; Kunda et al., 2001; Witte et al., 2008; Caceres et al., 2012). A considerable decrease of RhoA activity was detected in the enlarged growth cones (Fig. 1B; Fig. S2A,B), concomitant with high Rac1 (Fig. 1C, arrow; Fig. S2C,D) and Cdc42 (Chuang et al., 2005) activities.

Low RhoA activity was also detected in more than 80% of newly formed axons (early stage 3; 24–30 h in vitro); however, by 36 h in vitro it was possible to distinguish axonal populations that were similar in length but with differences in RhoA activity (Fig. 1D–G). Interestingly, at later time points (late stage 3), when most of the axons become longer than 150 µm and began to extend collateral branches, the percentage of cells displaying low axonal RhoA activity decreased significantly (Fig. 1H), with almost 70% of them showing a high RhoA FRET signal along the distal axonal segment, including the growth cone. Quantitative analysis revealed that axonal growth cones of early stage 3 neurons have on average a 50% decrease in RhoA activity compared to equivalent ones from minor processes of the same cell or of stage 2 neurons (Fig. 1I). It also revealed that RhoA activity in axonal growth cones of late stage 3 neurons increased significantly, becoming similar to that of the remaining minor neurites of the same neuron or of stage 2 neurons (Fig. 1I). By 3 DIV, almost all axons (∼300 µm in length) displayed high RhoA activity at their tips and/or along the distal third of the axonal shaft.

The FRET experiments showed that active RhoA could be found in neurites that exhibit little if any net growth, such as minor processes of either stage 2 or stage 3 neurons, and in elongating axons of stage 3 neurons. These observations raised the possibility that, in actively growing axons, RhoA might act as an inhibitory brake to prevent excessive growth or alternatively that it has a stimulatory effect.

To begin testing this second idea, we first evaluated the effect of the acute downregulation of RhoA on axon growth by the use of an siRNA targeting RhoA (denoted RNAi-RhoA; Dupraz et al., 2019 and see Materials and Methods). To examine RhoA function on initial axon differentiation, freshly plated neurons were electroporated with the RNAi-RhoA at the time of plating and analyzed 1 day later. In agreement with Dupraz et al. (2019), the early suppression of RhoA led to neurons displaying longer Tau-1-positive axon-like neurites when compared with control cells (Fig. 2A–C). Also, it produced an increased percentage of neurons reaching stage 3 of neurite development by 1 DIV (Fig. 2D). A similar phenomenon was observed when neurons were treated 4 h after plating and analyzed 1 day later with C3 toxin, a well-known RhoA inhibitor (Tsuge et al., 2015; Fig. S3A–D). By contrast, when the RNAi-RhoA treatment was applied 2 days after plating and neurons examined 24 or 48 h later, a significant decrease in axonal length was detected (Fig. 2E–G). Interestingly, such an effect was not observed in minor processes, which are destined to become dendrites (Fig. 2H).

To further validate these observations, we also analyzed the effect of LPA, a well-known RhoA activator (Li and Gundersen, 2008; Gonzalez-Billault et al., 2012; Choi and Chun, 2013; Quassollo et al., 2015), on neurite formation and extension at different stages of neuronal development. In the first group of experiments, LPA was added to the culture medium at 4–6 h after plating. After 1 day of treatment, the cultures were fixed and stained for Tau-1 and tyrosinated α-tubulin (Tyr-Tub) (Fig. 2I–K). LPA (2.5 µM) treatment resulted in a significant decrease (almost 40%) in the number of neurons capable of sprouting an axon (Fig. 2K). Halted cells failed to localize Tau-1 to the distal third of a single neurite or axon, a typical feature of polarized neurons (Fig. 2I,J); instead, most Tau-1 immunofluorescence was found at the cell body (Fig. 2J). In addition, all these cells lacked large growth cones distinctive of prospective axons. A 10 µM LPA dose had a more profound effect, with more than 90% of the cells failing to develop an axon and with minor neurites significantly shorter than equivalent ones from control neurons.

To further test our hypothesis, we evaluated the effect of LPA in older cultures (2 DIV), when more than 80% of cells have reached stage 3 of neurite development. When such neurons were treated for 24 h with different doses of LPA, axonal length increased significantly compared with that in control neurons (Fig. 2L–O). The highest concentration tested (5.0 µM) increased axonal length as much as 40% above the value of the control group (Fig. 2P). By contrast, all tested LPA doses (1.0, 2.5 and 5.0 µM) significantly decreased the length of minor neurites compared with non-treated ones (Fig. 2P). A similar response was observed when the LPA treatment was initiated at 3 DIV and analyzed 1 day later (data not shown). Finally, we also observed that in 2 or 3 DIV neurons C3 toxin significantly reduced axonal length and LPA-stimulated axonal growth (Fig. S3E–I). FRET experiments revealed that LPA treatment (5.0 or 10.0 µM) significantly increased RhoA activity in the distal part of the axon, including their growth cones (Fig. 2Q,R). Typically, this increase occurred within 30 min and lasted for at least 90 min, the final time point analyzed (Fig. 2Q,R).

Together, these observations support the existence of a RhoA inhibitory tone on initial axon outgrowth (Caceres et al., 2012; Dupraz et al., 2019; Takano et al., 2019) but more importantly reveal for the first time that, once axons are formed, RhoA activity no longer has a negative influence and instead becomes a promoter of axon elongation. The latter results raise the possibility that distinct effector pathways operate at different developmental stages (e.g. axon differentiation versus axon elongation) and/or cell domains (minor neurites versus axons).

ROCK is a RhoA effector with a wide variety of functions including regulation of motility, morphology and neuronal polarity (Amano et al., 2010; Caceres et al., 2012; Takano et al., 2017, 2019). As expected, inhibition of ROCK with Y27632 enhanced axonal growth when applied 6 h after plating and examined at 1 DIV (Fig. 3A,B). This treatment also increased the percentage of neurons with several axon-like (Tau-1-positive) neurites or multipolar neurons (Fig. 3C,D) and reverted the inhibitory effect of LPA on axonal differentiation (Fig. 3D).

To determine whether ROCK continues operating at later stages of axon development (e.g. during axon elongation), 2 DIV cultures were treated with Y27632 for different time periods. Neurites were arranged according to their length from longest to shortest, always defining the axon as the longest one (total length of at least 150 µm) (Fig. 3E). In these neurons, a 12 or 24 h treatment with Y27632 caused a 2- to 3-fold increase in minor neurite length compared with control cultures (Fig. 3E). Remarkably, the length of the differentiated axon (the longest neurite) was not significantly increased by Y27632 treatment (Fig. 3E). Thus, ROCK is neither inhibiting nor promoting axon elongation once the axon has been formed and is elongating. One possible explanation could be that elongating axons have low ROCK activity. To test this idea, we used Eevee-ROCK, a FRET-based biosensor with high sensitivity and specificity for ROCK activity (Li et al., 2017; see also Takano et al., 2017). FRET revealed that ROCK activity is restricted to shafts and growth cones of minor neurites (Fig. 3F,G), declining progressively towards the distal axonal end. The lowest ROCK activity was detected in the axon growth cone (Fig. 3F′). The polarized distribution of ROCK activity contrasts with that of RhoA, suggesting that other downstream effectors mediate its functioning during axonal elongation.

In a scratch-wound migration assay with fibroblasts, LPA–RhoA–mDia signaling induces rapid polarized formation of stable MTs required for directed migration (Li and Gundersen, 2008; Bartolini and Gundersen, 2010; Etienne-Manneville, 2013; Wojnacki et al., 2014). Whether or not this signaling pathway could induce similar changes in elongating axons remained to be established. Therefore, in the next series of experiments we first tested whether LPA could alter neuronal MT organization and/or dynamics. We focused our analysis on axonal growth cones of late-stage 3 neurons, given that this is the major site of MT assembly and stabilization, and membrane addition required for axonal elongation (Quiroga et al., 2018). LPA (10 µM) rapidly (30 min) induced an increase in the detyrosinated (Glu-Tub; Fig. 4A,B, left panel) and acetylated α-tubulin (Acetyl-Tub; Fig. S4) fluorescence signals that extend into the central and peripheral growth cone regions, which are usually devoid of stable MTs. These changes in MT organization were exclusively found in growth cones (not observed in axonal shafts or minor neurites). The morphology of the growth cones remained unchanged; no modifications in growth cone area (Fig. 4B, middle panel) or perimeter (Fig. 4B, right panel) were detected. This LPA-induced effect appears to be mediated by RhoA activation, as treatment with the specific RhoA inhibitor C3 toxin prevents the increase in acetylated tubulin fluorescence (Fig. S5A,B). In accordance with these observations, a significant decrease in acetylated α-tubulin immunofluorescence was also detected in axons of RhoA-suppressed neurons; this effect was quite prominent at the distal end of the axon, including the growth cone (Fig. 4C–E).

To better visualize the presence and location of Glu- or acetylated-MTs in axonal growth cones of LPA-treated neurons, we used stimulated emission depletion (STED) nanoscopy. Super-resolution imaging clearly revealed more stable MTs in the central growth cone domain of LPA-treated neurons, with some of them reaching the transition and growth cone peripheral domains (Fig. 4F and G, left panel, arrows). A completely different situation was observed in axonal growth cones of control neurons, where most stable MTs only reached the base or neck of the growth cone (Fig. 4G, right panel). The effect of LPA on stable MTs was not observed in growth cones of minor neurites of either stage 2 or stage 3 neurons (Fig. 4H). As an additional, and more functional test of the LPA-mediated enhancement of MT stabilization, we evaluated the activation of KIF17, a MT-based motor that preferentially associates with stable MTs and becomes activated upon MT-binding (Jaulin and Kreitzer, 2010; Espenel et al., 2013). Although in mature neurons, KIF17 preferentially transports cargo to dendrites due to a selective filter at the axon initial segment (AIS, Song et al., 2009), in young neurons (less than 6 DIV) it also traffics to axonal tips (Franker et al., 2016). Using a KIF17 FRET biosensor (Espenel et al., 2013) that reports on active (open conformation: no tail-head interaction, MT-bound, low FRET) or inactive (close conformation: tail-head interaction, no MT-binding, high FRET) conformations, we evaluated whether LPA-induced MT-stabilization correlated with a decrease in KIF17 biosensor activity in axonal growth cones. As expected, the results showed that LPA treatment results in a significant and selective decrease in KIF17 FRET efficiency at the axonal growth cone (Fig. 4I–K).

It has been proposed that stable MT ends serve as seeds for the assembly of new MT polymers within the axonal growth cone (Mitchison and Kirschner, 1988; Witte et al., 2008; Conde and Caceres, 2009); therefore, we decided to evaluate the effect of LPA on the distribution and abundance of dynamic MTs in axonal growth cones of stage 3 hippocampal neurons. LPA (10 µM) rapidly induced an increase in the tyrosinated α-tubulin fluorescence signals that extend into the peripheral growth cone region, an effect that was prevented when neurons were also incubated with C3 toxin (Fig. S5C,D). Moreover, STED nanoscopy revealed that dynamic MTs, labeled with monoclonal antibodies (mAbs) that recognize tyrosinated α-tubulin, a marker of recently assembled polymer, are more abundant in LPA-treated neurons than in control ones (Fig. 4L,M). Quantitative analyses of super-resolved images showed that the number and length of axonal growth cone tyrosinated MTs increased significantly after LPA treatment (Fig. 4N), despite no changes in growth cone area being detected between LPA- and DMSO-treated cultures.

To further analyze the stimulatory effect of LPA on MT assembly we used an EB3–GFP (EB3 is also known as MAPRE3) construct that binds to the plus-ends of growing MTs and thus serves to monitor MT dynamics (van de Willige et al., 2016). Live imaging of growth cone EB3 comets (see Materials and Methods) revealed that it was possible to follow them for longer periods of time in LPA-treated neurons than in control ones (Fig. 4O). A quantitative analysis of trajectories of EB3 comets in axonal growth cones with the software plusTipTracker (Applegate et al., 2011; Stout et al., 2014) confirmed this observation, revealing that LPA treatment produced significant increases in EB3 comet speed (µm/min), of the mean number of EB3 comets per/unit area and of the mean number of nucleation events per/unit area (Fig. 4O). These observations are fully consistent with those derived from the immunofluorescence data obtained using either confocal microscopy or nanoscopy (Fig. 4A–H).

Formins are another type of RhoA effector that regulates the actin and MT cytoskeletons (Goode and Eck, 2007; Bartolini and Gundersen, 2010; Kuhn and Geyer, 2014); some of them, like mammalian diaphanous (mDia), have been implicated in neuronal tangential migration (Shinohara et al., 2012), axonal guidance (Toyoda et al., 2013), dendrite complexity and spine density in differentiated hippocampal neurons (Qu et al., 2017), as well as axonal regeneration (Pinto-Costa et al., 2020) and SDF-1α-promoted axonal elongation in cerebellar neurons (Arakawa et al., 2003). Given that we did not detect any effect on axonal elongation upon ROCK inhibition, it became of interest to evaluate whether the RhoA stimulatory effect on axon elongation involves mDia formins. Therefore, in the final set of experiments, we decided to specifically address mDia participation in axon elongation. mDia formins have a prominent role in stabilizing MTs in nonneuronal (Palazzo et al., 2001; Wen et al., 2004; Bartolini et al., 2008) and neuronal cells (Qu et al., 2017), but have not been implicated in axon specification or elongation. We silenced mDia1 by transfecting neurons with a specific shRNA targeting mDia1 (shmDia1; Qu et al., 2017; see Materials and Methods) at 24 h after plating. After 3 days in vitro (DIV) we measured a substantial reduction in axonal length when compared to control shScramble-transfected neurons. Notably, the length of minor processes remained relatively unaffected (Fig. 5A–C).

Consistent with the increase in stable MTs after LPA treatment (Fig. 4), quantitative analysis revealed that neurons expressing shmDia1 exhibited a lower acetylated α-tubulin fluorescence signal compared to that in control neurons (Fig. 5D–F). These results suggest that mDia1 is necessary for microtubule polymerization and stabilization within the growth cones, consequently facilitating axonal elongation.

Diaphanous-related formins are auto inhibited through intramolecular binding of a diaphanous autoregulatory domain (DAD) to a conserved N-terminal diaphanous inhibitory domain (DID) (Rose et al., 2005). Binding of active RhoA to DID displaces DAD from the N-terminal region, releasing auto-inhibition and therefore activating mDia1 (Lammers et al., 2005; Otomo et al., 2005). Transfection with a GFP-tagged mDia1 DAD domain fragment activates the endogenous protein by relieving auto-inhibition (Alberts, 2001; Palazzo et al., 2001; Eng et al., 2006). Two different GFP-tagged mDia1 DAD domains, C′ terminal and core domains (Alberts, 2001; Wallar et al., 2006), were used to explore the effect of endogenous mDia1 activation on axonal length. To this end, axonal length was quantified in cultured hippocampal neurons transfected with GFP or GFP–DAD constructs (GFP–DAD core or GFP–DAD-C′) at 2 DIV and evaluated 24 h later. DAD-mediated mDia activation significantly increased axonal length when compared to control (GFP-transfected) cultures (Fig. 5H,I and J, left panel). By contrast, ectopic expression of either GFP–DAD core or GFP–DAD-C′ did not affect the length of minor neurites (Fig. 5J, right panel). Together, these results point towards mDia1 being the formin activated downstream of RhoA during axonal elongation.

mDia is best known for stimulating actin nucleation and polymerization (Pruyne et al., 2002). However, in recent years it has become increasingly evident that it also promotes MT stabilization in non-neuronal cells and in primary neurons (Palazzo et al., 2001; Bartolini et al., 2008; Bartolini and Gundersen, 2010; Bartolini et al., 2012; Qu et al., 2017). The MT-stabilizing domain of mDia2 (also known as DIAPH3) maps within the same region involved in actin polymerization, namely the FH2 domain (Xu et al., 2004). Two different point mutations (K853A and I704A) within the FH2 domain of constitutively active mDia2 prevent actin polymerization activity but retain the MT-stabilizing activity (Harris et al., 2006; Bartolini et al., 2008). Because an LPA–RhoA–mDia MT stabilization signaling pathway is required for polarized cell migration in fibroblasts (Bartolini et al., 2008; Morris et al., 2014), and because LPA-promoted axonal elongation also involves enhanced MT stabilization, it became of interest to test the consequences of expressing theses mutants on axon extension. To this end, 2 DIV hippocampal cell cultures were transfected with either K853A or I704A mDia mutants or a control plasmid and axonal length evaluated 18 h later. Ectopic expression of any one of these mutants resulted in a significant increase in axonal length (Fig. 5K–N). It is noteworthy that this effect was similar to the one observed after endogenous mDia activation induced by transfection of DAD domains (Fig. 5G–J).

Then, we addressed the question of whether mDia1 could promote axonal elongation by modifying the dynamic behavior of MTs. To test this idea 2 DIV hippocampal neurons were transfected with GFP–mDia1 (WT) or the I704A mDia mutant, and 1 day later changes in the relative amount of stable MTs were evaluated by quantitative immunofluorescence with antibodies against Glu- or acetyl-α-tubulin. The results obtained revealed that the ectopic expression of either mDia1 (Fig. 5O–R) or the MT-stabilizing mutant I704A (data not shown) significantly increase the fluorescent signal generated by both antibodies (Fig. 5S). Together, this information supports the possibility that mDia-promoted MT stabilization mediates RhoA-enhanced axonal elongation.

It is now well established that small RhoGTPases play major roles in neuronal morphogenesis by controlling cytoskeletal organization or dynamics and membrane trafficking (Gonzalez-Billault et al., 2012; Wojnacki and Galli, 2016; Quiroga et al., 2018). A long-lasting view based on pharmacological studies and evaluation of the up- or down-regulation of their expression and/or activity during neuronal development, has established the idea of RhoA being a major inhibitory factor for de novo and regenerative axon growth (Arimura and Kaibuchi, 2007; McKerracher et al., 2012; Schelski and Bradke, 2017; Takano et al., 2017; Dupraz et al., 2019; Takano et al., 2019; Wilson et al., 2020a). Additionally, RhoA negatively regulates minor process outgrowth through a ROCK–myosin light chain (MLC)–myosin II signaling pathway (Kollins et al., 2009). The present results put forward a far more complete picture of the function of RhoA during axon formation and growth. First, FRET imaging assays conclusively challenge the current dogma by revealing not only high RhoA activity in undifferentiated neurites exhibiting little if any growth but also, and surprisingly, in elongating axons. Second, along with RNAi and pharmacological experiments, our observations demonstrate that RhoA can function as a positive or negative regulator of axon growth depending on various factors, such as neurite type (minor neurite versus prospective axon versus elongating axon), stage of development (axon differentiation versus axon elongation) and downstream effectors (e.g. ROCK versus mDia; see next section). Moreover, we reveal for the first time that dual inhibitory and stimulatory functions of RhoA co-exist at the same time within the same cell but at different domains, as in the case of late stage 3 cultured hippocampal pyramidal neurons, where active RhoA promotes axonal elongation while preventing minor neurite growth, and during the subsequent generation of multiple axon-like neurites (Fig. 6A).

All this information is consistent with data from another important morphogenetic event, namely cell migration, where Rho GTPases also serve as master regulators (Etienne-Manneville, 2013). In this setting, early studies prompted the idea of Rac and/or Cdc42 acting as stimulatory factors promoting actin protrusive activity and cell locomotion at the leading edge, and RhoA being a negative regulator triggering acto-myosin contractility and cell retraction at the trailing edge (Raftopoulou and Hall, 2004). However, it soon became evident that this impression was an oversimplification and that, depending on cell type and environmental factors, considerable variations exist in the spatio-temporal activation pattern and function of RhoA (Kurokawa and Matsuda, 2005). For example, in migrating HeLa cells, FRET experiments have revealed that there is RhoA activation at both the leading and trailing edges, whereas in migrating MDCK cells RhoA activity is only detected at the leading edge (Kurokawa and Matsuda, 2005); interestingly, in growth factor-stimulated migrating fibroblasts RhoA activity is low at the plasma membrane, but high at membrane ruffles in growing lamellipodia, where high Rac activity is also detected (Kurokawa and Matsuda, 2005). Subsequent studies with a new generation of Rho-GTPase FRET biosensors and computational multiplexing have revealed that RhoA activation at the leading edge of migrating fibroblasts is synchronized with protrusion and precedes Rac and Cdc42 activation (Pertz et al., 2006; Machacek et al., 2009). Furthermore, it has been proposed that this early activation of RhoA might serve to initiate formin-mediated actin polymerization. Altogether, this information clearly indicates that, in other cell types, RhoA has both stimulatory and inhibitory activities that are spatially confined to particular cellular domains, with both being required for proper cell growth and polarization.

It has been proposed that different subcellular pools of a given RhoGTPase associated with distinct downstream effectors and regulatory proteins comprise a ‘spatio temporal signaling module’ (Pertz, 2010) that performs different functions. Our experiments show that in cultured hippocampal neurons, the inhibitory effect of RhoA on axon formation involves ROCK, whereas a different effector, namely mDia, has an opposite influence stimulating axon elongation (Fig. 6A).

Previous pharmacological studies have established that inhibition of ROCK in early developing cultured hippocampal neurons promotes axon formation and/or the generation of multiple axon-like neurites (Takano et al., 2019; see also references cited in the Introduction). More recently, an optogenetic approach has shown that activation of RhoA or ROCK at the cell body or distal end of newly formed axons results in retraction of minor neurites or inhibition of axon outgrowth, respectively (Takano et al., 2017). Consistent with this, our FRET experiments revealed high activity of RhoA and ROCK in growth cones of minor neurites and a simultaneous decrease of activity selectively located at the large growth cones of prospective and newly formed axons. However, although ROCK activity decreases significantly and permanently from elongating axons, RhoA activity show a second rise after axon sprouting becoming high in axonal shafts and growth cones of late stage 3 neurons.

Activation of endogenous mDia1 by transfection of cultured hippocampal neurons with a GFP-tagged diaphanous auto inhibitory domain (DAD) revealed that mDia1 mediates the stimulatory effect of RhoA on axonal elongation, whereas its downregulation with a specific shRNA had the opposite effect. These results are consistent with a previous report showing that, in cultured cerebellar granule cells, SDF-1α promotes axonal elongation by activating a RhoA–mDia signaling pathway (Arakawa et al., 2003). We have now extended this observation by showing not only that RhoA–mDia stimulates axonal elongation in the absence of exogenous SDF-1α,but also by providing evidence about its mechanism of action.

Our results indicate that RhoA–mDia signaling regulates axonal extension by promoting MT stability and assembly at growth cones of elongating axons (Fig. 6B). Several lines of evidence support this notion. First, using different experimental approaches, we showed that LPA increases the number of stable MTs entering the central and peripheral domains of axonal growth cones. Second, this effect was prevented by shmDia1, unaffected by Y27632 and absent in growth cones of minor neurites. Third, expression of mDia mutants lacking actin polymerization activity, but retaining their ability to promote MT stabilization, stimulate axonal elongation. Finally, these observations are in full agreement with studies in scratched monolayers of NIH-3T3 fibroblasts showing that LPA-induced RhoA activation drives mDia mediated-MT stabilization at the leading edge, which is required for polarized cell migration (Cook et al., 1998; Palazzo et al., 2001; Gundersen et al., 2005; Bartolini and Gundersen, 2010; Bartolini et al., 2012). The precise mechanism by which mDia stimulates MT stabilization in elongating axons remains to be established. In this regard, it is worth noting that KIF4, a kinesin superfamily member highly enriched in growth cones of growing axons and required for L1-stimulated axonal elongation (Peretti et al., 2000), interacts with EB1 to mediate mDia-induced MT stabilization in migrating fibroblasts (Morris et al., 2014). Future studies should explore this, among other possibilities.

Our study provides a comprehensive analysis of the dual functions of RhoA in axonal growth, significantly advancing our understanding of neuronal morphogenesis. We demonstrate that RhoA exhibits both inhibitory and stimulatory roles in axonal development, contingent on the stage of neuronal development, the type of neurite and the specific downstream effectors involved, illustrating its spatial and temporal regulation within the same cell. Overall, this study reshapes the current understanding of RhoA involvement in axon growth, highlighting its capacity to both hinder and promote axonal development through distinct molecular pathways.

Pregnant Wistar rats were born in the vivarium of INIMEC-CONICET-UNC (Córdoba, Argentina). Wistar rat lines were originally provided by Charles River Laboratories International Inc (Wilmington, USA). All procedures and experiments involving animals were approved by the Animal Care and Ethics Committee (CICUAL http://www.institutoferreyra.org/en/cicual-2/) of INIMEC-CONICET-UNC (Resolution numbers 014/2017 B, 015/2017 B, 006/2017 A and 012/2017 A) and were in compliance with approved protocols of the National Institute of Health Guide for the Care and Use of Laboratory Animals (SENASA, Argentina).

All methods were carried out in accordance with relevant guidelines and regulations.

The following cDNA constructs were used in this study: cDNA coding for small RhoGTPases biosensors Rac1-Raichu (Itoh et al., 2002; kindly provided by Dr Michiyuki Matsuda, Institute of Advanced Energy, Kyoto University, Japan); RhoA 1G and RhoA 2G (Fritz et al., 2013; made in the laboratory of O.P.); cDNA coding for a FRET reporter for ROCK activity designated as EeveeROCK (Li et al., 2017; kindly provided by Dr Michiyuki Matsuda); cDNA coding for a FRET reporter for KIF17 activity (Espenel et al., 2013; kindly provided by Dr Geri Kreitzer, Department of Cell and Developmental Biology, Cornell University, New York, USA); cDNA coding for EB3–GFP (kindly provided by Dr Leticia Peris, Grenoble Institute Neurosciences, Grenoble, France); cDNA coding for mDia1–GFP, the point mutated K853A mDia–GFP, the point mutated T1704A mDia–GFP, the DADc′ mDia–GFP domain and the DAD core mDia–GFP domain (Bartolini et al., 2008; made in the laboratory of G.G.G.); cDNA coding eGFP protein (Clontech); short hairpin RNA (shRNA) against mDia1 and a scrambled sequence as described and validated previously (Qu et al., 2017) targeting the following sequence of rat mDia1 5′-GCGACGGCGGCAAACATAA-3′ and control scramble shRNA 5′-GGCAAATCTTCTAGTCTAT-3′ (made in the laboratory of F.B.); for RhoA gene silencing experiments, siRNA oligonucleotides and siRNA negative control were purchased from GenBiotech as described and validated previously (Dupraz et al., 2019), targeting the following sequence of rat RhoA 5′-TAGTTTAGAAAACATCCCAGA-3′.

Embryonic day (E)18 rat embryos (euthanized by CO₂ overdose) were used to prepare primary hippocampal cultures as previously described (Bisbal et al., 2018; Wilson et al., 2020b). Briefly, hippocampi from E18 fetal rats were dissected and incubated with trypsin (0.25% for 15 min at 37°C) (Thermo Fisher Gibco; cat. no. 15090-046) and mechanically dissociated by trituration with a Pasteur pipette. Cells were plated on cover glasses circles (12 mm; Marienfeld Superior; cat. no. 633029) coated with 1 mg/ml poly-L-lysine (Sigma Chemical Co.; cat. no. P2636) at a density of 2000 cells/cm² in minimum essential medium (MEM, Thermo Fisher Gibco; cat. no. 61100-061) supplemented with CTS GlutaMAX I Supplement (Thermo Fisher Gibco; cat. no. A1286001), sodium pyruvate (Thermo Fisher Gibco; cat. no. 11360070), penicillin (100 U/ml), streptomycin (100 mg/ml) (Thermo Fisher Gibco; cat. no. 15140122) and 10% horse serum (Thermo Fisher Gibco; cat. no. 16050122). After 2 h, the coverslips were transferred to dishes containing serum-free neurobasal medium (Thermo Fisher Gibco; cat. no. 21103049) with B-27 Plus supplement (Thermo Fisher Gibco; cat. no. A3582801) and CTS GlutaMAX I supplement (Thermo Fisher Gibco; cat. no. A1286001).

CHO cells (ATCC) were plated on round 12-mm cover glasses (Marienfeld Superior; cat. no. 633029) and cultured in Minimum Essential Medium (MEM, Thermo Fisher Gibco; cat. no. 61100-061) supplemented with 10% fetal bovine serum (Internegocios S.A., cat. no. 000008), 1% CTS GlutaMAX I supplement (Thermo Fisher Gibco; cat. no. A1286001), 1% sodium pyruvate (Thermo Fisher Gibco; cat. no. 11360070) and 1% penicillin-streptomycin (Thermo Fisher Gibco; cat. no. 15140122) at 37°C and 5% CO₂ until fixation.

For transient electroporation, cultured neurons were electroporated using the Lonza Nucleofector II device (cat. no. AAD-1001N, Lonza; program O-003) and the mouse neuron Nucleofector^® Kit (cat. no. VPG-1003, Lonza), according to the manufacturer's instructions. 0.5×10⁶–1×10⁶ cells were used for each transfection with 3 µg of plasmid DNA. For the silencing experiments, 0.5×10⁶–1×10⁶ cells were transfected with a mix of 0.2 pmol of the siRNA and 1 µg of GFP-expressing plasmids (pEGFP) as carrier DNA and transfection reporter.

For transient lipofection, cultured cells were transfected with Lipofectamine 2000 (Thermo Fisher Gibco; cat. no. 11668027), following the manufacturer's instructions. Briefly, for a 35 mm dish, 4 μl of Lipofectamine 2000 and selected plasmids (3 μg) were diluted in Opti-MEM medium (Thermo Fisher Scientific) to a final volume of 250 μl and incubated for 20 min at room temperature (RT). Cells were incubated for 2 h in Opti-MEM containing the Lipofectamine and plasmids mix and returned to Neurobasal-B-27 plus medium until analysis. For the silencing experiments, neurons were transfected with a mix of 0.2 pmol of the siRNA and 1 µg of GFP-expressing plasmids (pEGFP) as carrier DNA and transfection reporter.

Cells were fixed with 4% paraformaldehyde (Sigma-Aldrich, cat. no. 441244) and 4% sucrose diluted in phosphate buffered saline (PBS) for 20 min at RT as previously described (Bisbal et al., 2018; Pesaola et al., 2021). For STED and confocal microscopy, neurons with growth cones for MT analysis were simultaneously fixed and permeabilized in PHEM buffer [60 mM PIPES, 25 mM HEPES, 5 mM EGTA, 1 mM MgCl (pH 6.9)] containing 0.25% glutaraldehyde (Sigma-Aldrich, cat. no. G6257), 3.7% paraformaldehyde, 3.7% sucrose and 0.1% Triton X-100, and quenched in 0.1 M glycine/PBS for 10 min as previously described (Unsain et al., 2018).

Fixed cells were washed three times with PBS, permeabilized in 0.2% Triton X-100 in PBS at RT for 5 min and again washed in PBS before antibody incubation. Cells were incubated in blocking buffer (5% bovine serum albumin in PBS) for 1 h. Cells were then incubated with primary antibodies diluted in blocking buffer for 1 h, washed with PBS three times and incubated with fluorescent-conjugated secondary antibodies for 1 h. Finally, cells were washed with PBS three times and the coverslips mounted using FluorSave (Millipore Calbiochem; cat. no. 34-578-9).

The following primary antibodies were used for immunofluorescence (IF) in this study: a monoclonal antibody (mAb) against tau protein (clone Tau-1; Millipore, cat. no. MAB3420, RRID: AB_94855) diluted 1:1000; a rat monoclonal against tyrosinated α-tubulin (clone YL1/2, Abcam, cat. no. ab6160, RRID: AB_305328) diluted 1:2000; a mAb against tyrosinated α-tubulin (clone TUB-1A2; Sigma, cat. no. T9028, RRID: AB_261811) diluted 1:1000 or 1:200 for STED microscopy; a mAb against α-tubulin (clone α-3A1; Sigma-Aldrich, cat. no. T5168, RRID:AB_477579) diluted 1:1000; a mAb against acetylated α-tubulin (clone 6-11B-1; Sigma-Aldrich, cat. no. T7451, RRID: AB_609894) diluted 1:1000 or 1:200 for STED microscopy, and a rabbit polyclonal antibody against detyrosinated (Glu) tubulin (made in in-house by the laboratory of G.G.G.; Gundersen et al., 1984) diluted 1:1000 or 1:100 for STED microscopy.

The following secondary antibodies and dyes were used: for widefield, the corresponding secondary antibodies conjugated to Alexa Fluor 488, Alexa Fluor 568 (Thermo Fisher Scientific, cat. nos A11001, A11008, 11006, A11004, A11031 and A11077), phalloidin–Rhodamine and phalloidin–Alexa Fluor 647 (Thermo Fisher Scientific, cat. nos R415 and A22287); and for STED microscopy anti-mouse-IgG conjugated to Atto 647N, anti-rabbit-IgG conjugated to Atto 594 (Sigma-Aldrich, cat. nos 50185 and 77671) and phalloidin–Atto 594 (Sigma-Aldrich, cat. no. 51927).

LPA was purchased from Sigma-Aldrich (cat. no. L7260), Y27632 from Calbiochem (cat. no. 688000) and C3 toxin (Rho Inhibitor I) from Cytoskeleton (cat. no. CT04).

Cells were visualized using either a conventional LSM 800 (Zeiss, Germany) and TCS FV300 (Olympus, Japan) or a spectral FV1200 (Olympus, Japan) inverted confocal microscope. For high-magnification analysis, z stack confocal imaging was carried out with a plan-apochromat 60x or 63x/1.4NA oil-immersion objective. Low magnification images were acquired with a plan-apochromat 20x/0.75NA oil-immersion objective. Pixel size and Z-step were set to fulfil Nyquist criterion. In all cases, lasers and spectral bands were chosen to maximize signal recovery while avoiding signal bleed-through.

Post-imaging analysis and measurements were done using Fiji-ImageJ software (NIH, USA).

Stimulated emission depletion (STED) nanoscopy was performed in a custom-built nanoscope at the Center for Bionanoscience Research (CIBION, CONICET). A detailed description of the setup was provided in a previous publication (Szalai et al., 2021). Briefly, two linearly polarized lasers pulsed at 640 nm (200 ps pulse width, PicoQuant LDH-P-C-640B) and at 594 nm (100 ps pulse width, PicoQuant LDH-D-TA-595) were used for fluorescence excitation, both operating at 40 MHz. In order to obtain circular polarization, both beams passed through broadband (400–800 nm) quarter-wave plates (Thorlabs AQWP05M-600) and broadband (400–800 nm) half-wave plates (Thorlabs AHWP05M-600).

For depletion, a linearly polarized laser pulsed at 775 nm (800 ps pulse width, Onefive Katana HP) operating at 40 MHz was used. The depletion beam passed through a 2π vortex phase plate (Vortex Photonics, V-775-70) in order to obtain a doughnut-shaped beam. Circular polarization was obtained using a broadband (690-1200 nm) quarter-wave plate (Thorlabs AQWP05M-980) and a broadband (690-1200 nm) half-wave plate (Thorlabs AHWP05M-980). The three lasers were combined using suitable dichroic mirrors, and light was focused on the sample with an objective of 1.4 NA (Leica HCX PL APO 100×/1.40-0.70 oil CS).

Fluorescence arising from the sample was split with a long-pass dichroic mirror (FF649-Di01-25×36, Semrock). The reflected light passed through a band-pass filter (FF01-623/24-25, Semrock) and was focused on an avalanche photodiode (APD, SPCM-AQR-13, PerkinElmer Optoelectronics). The transmitted light passed through a band-pass filter (FF01-676/37-25, Semrock) and was focused on a second avalanche photodiode detector (APD, SPCM-AQR-13, PerkinElmer Optoelectronics).

The depletion beam was delayed 0.4 ns with respect to the excitation pulses, in order to maximize the depletion. Time gating of fluorescence photons was also performed, using a custom-made electronic board (MPI for Biophysical Chemistry).

The power of the lasers was adjusted to obtain the best possible resolution in a given set of stained samples, and the imaging conditions were maintained within samples of the same experiment. The optimum values finally obtained were 3 μW and 1.5 μW for 594 nm and 640 nm, respectively. For the depletion beam, a laser pulse energy of 5 nJ was used.

The lateral resolution reached was 50–60 nm in both detection channels. This value was obtained by imaging isolated single molecules and measuring the full width at half maximum (FWHM) of the obtained images.

For Förster resonance energy transfer (FRET), cells were transfected using Lipofectamine 2000 as described above. After 10 to 16 h, cell cultures were fixed with 4% paraformaldehyde and 4% sucrose in PBS for 20 min, washed three times (4 min each) in PBS and mounted on a microscope slide with FluorSave. Cells were visualized with a FV300 or FV1000 confocal Olympus microscopes or with an Olympus IX81 inverted microscope equipped with a total internal reflection fluorescence Cell^TIRF module. For FRET detection, the cyan or teal fluorescent protein (FRET donor) was excited with a continuous laser of 458 nm while simultaneously acquiring donor and acceptor (yellow or Venus fluorescent protein) emissions signals. For ratio imaging FRET calculation, the donor channel (donor emission) and FRET channel (acceptor emission) images were smoothed with a median filter (1.5 pixel ratio), background subtracted (50.0 pixels rolling ball radius) and aligned. FRET map images were generated by dividing the processed FRET channel images over donor channel images. To remove out-of-cell pixels from the analysis, a 0–1 intensity value binary mask was created using the FRET channel images and multiplied by the FRET map images (Quassollo et al., 2015; Rozes Salvador et al., 2016). Finally, pixel values of the FRET maps images were color coded using a custom look-up table (LUT). All image processing was coded in an ImageJ macro so all images were processed in the same way (available upon request).

Early stage 3 neurons were defined as those in which we could clearly identify a single axon determined as a neurite which is at least twice as long as all other neurites but shorter than 60 µm. Late stage 3 neurons were defined as those having a clearly defined single axon of 60 µm or longer and at least twice as long as the remaining neurites.

Neurons were immuno-stained for Tau1 and tyrosinated tubulin for morphometric analysis as described above and by Wojnacki et al. (2021a,b). The maximum intensity projection of all optical slices of each neuron was used for morphometric analysis. Axons were defined as any process longer than 100 µm and with Tau1-positive staining. Unipolar neurons were defined as those with only one axon. Multipolar neurons were defined as those bearing two or more axons. Total axonal length was measured as the cumulative lengths of the longest uninterrupted process and all shorter branches of the axon. The length of minor processes was also measured by determining the cumulative length of the longest neurite and all branches if present.

Fiji-ImageJ and the Simple Neurite Tracer plugin were used to measure all neurites lengths (Schindelin et al., 2012).

Fluorescence images of living neurons were acquired as previously described (Bisbal et al., 2016). Briefly, transfected neurons expressing EB3–GFP were captured with a charge-coupled device camera (Andor iXon3; Oxford Instruments) using an 60×/NA 1.4 oil immersion objective in an Olympus IX81 inverted microscope equipped with a Disk Spinning Unit (DSU) with epifluorescence illumination (150 W Xenon Lamp). Neurons were imaged in neurobasal medium supplemented with 30 mM HEPES buffer (pH 7.2) and maintained at 37°C. Time-lapse sequences were acquired at a continuous rate of 1 frame per second during 5–10 min inside a stage top incubator (INU series, TOKAI HIT). Images were processed offline using a plusTipTracker from µ-Track software Version 2.0 (https://github.com/DanuserLab/u-track; Applegate et al., 2011; Jaqaman et al., 2008). Movies were color coded using a temporal-color code look up table (LUT) to generate an xy 2D image.

For the images where the fluorescence intensity was quantified, once the microscope was configured on the first acquired image, the same configuration was maintained for the acquisition of all subsequent images. The maximal intensity projection of all optical slices of each neuron was used for fluorescence intensity analysis. By adjusting the intensity level, we excluded all pixels with values less than 200 and cells with saturated pixels from the analysis. Subsequently, the regions of interest (axonal growth cones or minor processes) were selected and the average fluorescence intensity value per region was quantified.

For quantification of the number and length of MTs in growth cones, STED images were used. All the images between experiments were acquired with same microscope configuration. Growth cones areas were defined by making a region of interest (ROI) in the phalloidin channel. Individual microtubules were quantified within the ROI and the length measured from the base of the growth cone to the tip of the microtubule.

Statistical analysis was performed with the R software and Infostat software. When the experimental design required a single pairwise comparison, unpaired two-tailed Student's t-tests were applied. If the experimental design required, the comparison of multiple conditions a one-way ANOVA test was used followed by a Tukey's HSD test. A Shapiro–Wilks test for normality and Levene's test to evaluate the homogeneity of variance were performed for all data. In all cases, the assumption of normality was maintained. The exact value of n, what it represents, and center and dispersion measures are described in the figures and figure legends. In experiments where we measured neurite length or immunofluorescence intensity, each cell was considered as an independent statistical observation. Statistical observations come from at least three independent experiments. Significance was defined as a P<0.05 unless otherwise stated. The lower and upper hinges of all boxplots correspond to the first and third quartiles, respectively (the 25th and 75th percentiles). The upper whisker extends from the hinge to the largest value no further than 1.5×IQR from the hinge (where IQR is the inter-quartile range, or distance between the first and third quartiles). The lower whisker extends from the hinge to the smallest value at most 1.5×IQR of the hinge. Data beyond the end of the whiskers were defined as ‘outliers’ and were plotted individually. All researchers were aware of the experimental conditions for tests in this study; sample calculation and randomization were not conducted in this study.

The authors greatly acknowledge the Centro de Micro y Nanoscopía de Córdoba (CEMINCO), a Microscopy and Imaging Core Facility from UNC-CONICET, for technical and imaging assistance. The authors thank Laura Montroull, Andrea Pellegrini, Romina Maiorano, Jesica Piovano, Milagros Nigro, Marisa Gigena and Silvina Ferrer for technical assistance.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261970.reviewer-comments.pdf