Cell migration is driven by pushing and pulling activities of the actin cytoskeleton, but migration directionality is largely controlled by microtubules. This function of microtubules is especially critical for neuron navigation. However, the underlying mechanisms are poorly understood. Here we show that branched actin filament networks, the main pushing machinery in cells, grow directly from microtubule tips toward the leading edge in growth cones of hippocampal neurons. Adenomatous polyposis coli (APC), a protein with both tumor suppressor and cytoskeletal functions, concentrates at the microtubule-branched network interface, whereas APC knockdown nearly eliminates branched actin in growth cones and prevents growth cone recovery after repellent-induced collapse. Conversely, encounters of dynamic APC-positive microtubule tips with the cell edge induce local actin-rich protrusions. Together, we reveal a novel mechanism of cell navigation involving APC-dependent assembly of branched actin networks on microtubule tips.
Directional migration of individual cells in multicellular organisms plays key roles in multiple aspects of normal development and disease. The precision with which cells find their destinations is especially critical for neurons, which extend their processes, axons and dendrites, over enormous distances and need to find exact targets in order to establish functional neural circuitry during development and following regeneration after injury. Any imperfection in this daunting task leads to neuropsychiatric disorders (Blanquie and Bradke, 2018; Eira et al., 2016; Engle, 2010; Muñoz-Lasso et al., 2020).
The cytoskeleton is the main cellular machinery that generates force for migration and executes cellular decisions in response to guidance cues. The cytoskeleton components playing major roles in cell motility, microtubules and actin filaments, are responsible for different aspects of cell motility, but also extensively cooperate with each other. The actin cytoskeleton can produce large pushing and pulling forces to drive cell protrusion and contraction, respectively (Svitkina, 2018). Leading edge protrusion is typically driven by polymerizing actin filaments in the form of Arp2/3 complex-dependent branched actin networks. Microtubules, on the other hand, largely function as “managers” by coordinating pushing and pulling activities of the actin cytoskeleton in space and time to enable directional cell migration (Vasiliev et al., 1970) and neuron navigation (Tanaka and Kirschner, 1995). However, it remains poorly understood how microtubules accomplish their “supervising” roles. Besides serving as tracks for the intracellular transport, the best known function of microtubules, microtubules can also manipulate the actin cytoskeleton by regulating Rho family GTPases (Krendel et al., 2002; van Haren et al., 2014) and cell–matrix adhesions (Kaverina et al., 1999), and by inducing assembly of unbranched actin filaments in cooperation with formin family actin nucleators (Henty-Ridilla et al., 2016).
Actin-microtubule cooperation is particularly important in neurons (Cammarata et al., 2016; Pacheco and Gallo, 2016), where such cross-talk ensures precise connectivity in the neural circuitry. Growth cones, the actin-rich structures at the tips of microtubule-rich neurites, drive neurite extension and guide them toward proper targets. Bidirectional actin-microtubule cross-talk plays important roles in growth cone motility (Dogterom and Koenderink, 2019). On one hand, the actin cytoskeleton can regulate the distribution of microtubules. For example, microtubules can be captured and guided to the cell periphery by actin filament bundles in filopodia (Burnette et al., 2007; Dent and Kalil, 2001; Schaefer et al., 2002; Zhou et al., 2002) or kept away from the leading edge by the actin retrograde flow (Myers et al., 2006; Schaefer et al., 2002). These phenomena likely require physical interaction between actin filaments and microtubules, which can be mediated, for example, by septins (Hu et al., 2012), doublecortin (Tint et al., 2009), drebrin (Geraldo et al., 2008), and/or spectraplakins (Alves-Silva et al., 2012). On the reverse side, microtubules also can regulate actin dynamics. In striking examples of this kind in neurons, “pioneer” microtubules that approach the growth cone periphery induce growth cone advance or turning (Bearce et al., 2015; Buck and Zheng, 2002; Dent et al., 2003), whereas microtubules entering dendritic spines, actin-rich protrusions that form the postsynaptic component of excitatory synapses, induce spine expansion and synapse strengthening (Hu et al., 2011; Jaworski et al., 2009). Since both growth cone motility and dendritic spine expansion depend on the assembly of the Arp2/3-dependent branched actin networks (Korobova and Svitkina, 2008, 2010), these observations suggest that microtubules may be able to locally induce assembly of branched actin networks. However, this possibility so far remains purely hypothetical, if considered at all.
A starting point for this study was our serendipitous discovery by platinum replica electron microscopy (PREM), a technique uniquely suited for visualization of fine architecture of the cytoskeleton (Svitkina, 2017), that branched actin networks can be physically associated with microtubules in neuronal growth cones, suggesting that a microtubule might be able to directly “tell” the growth cone which way to go. In this study, we investigated the molecular mechanism of microtubule-dependent branched network assembly and uncovered key roles of adenomatous polyposis coli (APC) in this process.
Although best known as a tumor suppressor (Zhang and Shay, 2017), APC is also a cytoskeletal protein that regulates cell motility and polarity (Nelson and Näthke, 2013), as well as neurite outgrowth (Koester et al., 2007; Yokota et al., 2009). APC can interact with the microtubule lattice directly and also bind growing microtubule ends by interacting with end-binding (EB) proteins (Jiang and Akhmanova, 2011; Zumbrunn et al., 2001), the core family of plus-end tracking proteins (+TIPs; Bearce et al., 2015). APC can also bind actin (Moseley et al., 2007) and nucleate unbranched actin filaments, either by itself or in cooperation with formins (Breitsprecher et al., 2012; Okada et al., 2010). These APC interactions depend on the sequences located within the C-terminal region of APC. The central unstructured region of APC is responsible for tumor suppressor functions of APC by binding components of the β-catenin destruction complex (Zhang and Shay, 2017). The N terminus of APC containing Armadillo repeats, among other interaction partners, binds and activates the Cdc42 guanine nucleotide exchange factor, Asef (Kawasaki et al., 2000; Mitin et al., 2007), thus forming a potential link through Cdc42 (Gotthardt and Ahmadian, 2007; Hamann et al., 2007) and its effector N-WASP (Rohatgi et al., 1999) to activation of the Arp2/3 complex and formation of branched actin networks. APC is highly enriched in neuronal tissues (Bhat et al., 1994), whereas APC mutations have been linked to intellectual disabilities and neurological disorders (Gonzalez et al., 2015; Mohn et al., 2014; Onouchi et al., 2014). Therefore, our findings not only uncover a fundamental cell biological mechanism but also shed light on the etiology of some human diseases.
Association of branched actin networks with microtubule tips in neuronal growth cones
We used PREM to study the cytoskeletal structure of growth cones of cultured rat hippocampal neurons (Fig. 1). In PREM images, actin filaments and microtubules can be distinguished by their thickness, which is ∼10 nm for actin filaments and ∼25 nm for microtubules (including platinum coating), whereas immunogold labeling of microtubules improves detection of microtubules within dense actin networks. Several types of actin filament arrays could physically associate with microtubule tips in neuronal growth cones (Fig. 1, A and I). Specifically, 24.1% of clearly distinguishable microtubule tips were associated with filopodial bundles (Fig. 1 A, green arrowheads), as observed previously (Dent et al., 2007; Schaefer et al., 2002). An identical fraction (24.1%) of tips was associated with individual linear actin filaments that extended from the microtubule into the surrounding actin network (Fig. 1 A, yellow arrowhead). The most striking actin-microtubule arrangement, however, was the association of branched actin filaments with the distal microtubule regions (Fig. 1, A, B, D, F, and H; blue shade), which was seen at 34.5% of microtubule ends (Fig. 1 I). Of these cases, 73.3% (25.3% of total) of branched actin filaments could be clearly traced back to the microtubule surface, whereas in the remaining 26.7% of cases (9.2% of total), microtubule tips ended within a dense local patch of branched actin network (Fig. 1 A, dark blue arrowhead). In 17.2% of cases, microtubules ended within a complex actin network that could not be explicitly categorized. Branched actin networks associating with microtubule tips were also observed in nonneuronal COS7 cells (Fig. S1).
Branched actin networks are generated by the Arp2/3 complex, which nucleates new actin branches on the side of preexisting “mother” actin filaments creating a ∼70° angle between the barbed (plus) ends of mother and daughter filaments (Mullins et al., 1998; Svitkina and Borisy, 1999). The orientation of this angle, which corresponds to the direction of applied force, suggested that the actin network grew from the microtubule surface toward the nearest edge (Fig. 1, B, D, and F). The branched networks often sprouted from a few microtubule-associated unbranched actin filaments (Fig. 1, F and H) that appeared to serve as initial mother filaments for branched nucleation. We could occasionally detect additional material at microtubule tips (Fig. 1 F, purple), which likely corresponded to a collection of +TIPs, which mediate many microtubule functions (Bearce et al., 2015). The observed position and geometry of the microtubule-associated branched networks are favorable for driving local plasma membrane protrusion near the microtubule tip.
Availability of mother filaments, which also serve as Arp2/3 complex coactivators (Machesky et al., 1999), may be a rate-limiting step for starting new branched networks. The existence of microtubule-associated branched networks suggests that microtubules may be able to somehow overcome this problem. If so, microtubules are expected to be more extensively employed in this capacity when availability of preexisting actin filaments is decreased, for example, following treatment with actin-depolymerizing drugs. We tested this prediction by culturing hippocampal neurons in the presence of low concentrations (1–5 µM) of cytochalasin D (CytoD) or latrunculin B (LatB), which still permit actin-dependent neurite outgrowth (Chia et al., 2016). PREM analysis revealed the actin-rich growth cone-like structures at the neurite tips of CytoD- or LatB-treated neurons, which contained dense networks of branched actin filaments often associated with microtubules (Fig. 2).
Localization of APC at the interface between microtubules and branched actin networks
Microtubules cannot directly bind or nucleate actin filaments, suggesting that some cross-linkers connect branched actin filaments and microtubules. Among multiple candidates proposed to serve as actin–microtubule cross-linkers (Cammarata et al., 2016; Dogterom and Koenderink, 2019), APC uniquely combines promising biochemical properties (see Introduction) with an ability to regulate neurite outgrowth (Koester et al., 2007; Yokota et al., 2009), which makes it an excellent candidate to initiate the formation of branched actin networks in a microtubule-dependent manner.
In cultured neurons, APC was expressed throughout in vitro development but at slightly higher levels in more mature neurons (Fig. 3 A). By fluorescence microscopy, APC in young neurons cultured for 2 d in vitro (DIV) was often distributed in the form of small scattered puncta, whereas growth cones of mature DIV14 neurons mostly contained bright APC streaks (Fig. 3, B and C). Both types of APC structures colocalized with microtubule ends (Fig. 3 B). In growth cones of young DIV2–3 neurons treated with CytoD or LatB, APC showed prominent clustering analogous to that in untreated mature DIV14 growth cones (Fig. S2, A and B). Interestingly, F-actin in these inhibitor-treated neurons almost perfectly colocalized with APC, suggesting that APC accumulations could promote local actin assembly in conditions of hindered actin assembly.
If APC connects microtubules and branched actin networks, it should localize at the microtubule-actin interface in these structures. Using PREM of DIV3 and DIV14 neurons in combination with APC immunogold labeling (Fig. 4), we indeed detected APC at such interfaces. Consistent with the immunofluorescence data (Fig. 3 B), APC immunogold particles were more abundant at the microtubule-actin junctions in DIV14 growth cones (Fig. 4, E–I) and in DIV3 growth cones of CytoD-treated neurons (Fig. S2, C and D) as compared with untreated DIV3 growth cones (Fig. 4, A–D). APC immunogold particles were also present along the length of microtubules and on actin filaments, as expected based on the ability of APC to bind the microtubule lattice and actin filaments. Notably, the immunogold labeling was specific, because we observed virtually no labeling when we omitted the primary antibody (Fig. S3, A and B). Thus, APC is properly localized to serve as a linker between microtubules and branched actin networks in growth cones.
Induction of local actin-rich protrusions by APC-positive microtubules
Using coexpression of fluorescently tagged APC with either F-tractin, a marker of F-actin, or cortactin, a marker of branched actin networks, in REF52 or COS7 cells, we correlated APC dynamics with actin-based protrusions (Fig. 5, A–C and E; and Fig. S4). As a control, we expressed EGFP-EB3 with F-tractin or cortactin (Fig. 5, D and E). As shown previously (Kita et al., 2006; Mimori-Kiyosue et al., 2000), full-length APC often formed relatively stable accumulations in the vicinity of cell edges. Within these accumulations, individual APC clusters, likely associated with microtubule tips, made relatively short runs toward the cell edge, roamed around within a small area, or remained nearly stationary. During roaming behavior, APC clusters usually remained distant from the cell edge. No obvious correlation between APC motility and cell edge protrusion was observed in such cases. The stationary APC clusters often persisted at the cell edge for the duration of several protrusion-withdrawal cycles of the cell edge, which made these events unsuitable for correlative analysis. However, when an APC cluster hit the cell edge after a detectable run toward the edge, such encounters were often followed by a local F-tractin– or cortactin-rich protrusion (Fig. 5, B and C; Fig. S4; and Videos 1, 2, 3, and 4). Quantification of such events in REF52 cells revealed that 64.1 ± 7.0% of the APC-edge encounters were followed within ∼30 s by a local F-actin–rich or cortactin-rich protrusion, as compared with only 30.1 ± 3.9% of such events in the case of EB3 tracks (Fig. 5 E). The protrusion distance following the encounter was similar for both proteins (0.84 ± 0.50 µm for APC-edge encounters and 0.84 ± 0.62 µm for EB3-edge encounters; P = 0.14; n = 367 for APC and 168 for EB3; Mann–Whitney two-tailed test). These data show that the branched actin networks observed at the APC-positive microtubule tips by PREM are indeed able to generate local membrane protrusions at the sites where the microtubule-delivered APC contacts the cell edge.
Loss of branched actin networks from neuronal growth cones after APC knockdown
We used RNAi to determine whether APC is needed for assembly of branched actin networks in neurons. We used two lentivirus-encoded shRNAs (sh1 and sh2), which efficiently depleted the APC protein in cultured rat hippocampal neurons, as validated by Western blotting (Fig. 5 F) and immunofluorescence (Fig. S5 A). Phalloidin staining showed that APC knockdown led to a significant loss of F-actin in neurons (Fig. 5 G) and in individual growth cones at DIV6 (Fig. S5, C and D) and DIV14 (Fig. 5, H and I), whereas the total expression levels of actin remained unchanged (Fig. 5 F). In neurons transfected with control shRNA, both axonal and dendritic growth cones exhibited typical palm-like morphology, whereas APC-depleted growth cones gained an atypical elongated spear-like shape (Fig. 5 H and Fig. S5 C). These respective shapes were maintained during growth cone motility (Fig. 5 J and Video 5). Fluorescence staining of APC-depleted DIV14 neurons with phalloidin and APC antibody revealed that the remaining growth cones in APC-depleted neurons were associated with weak accumulations of residual APC (Fig. S5 B), thus further emphasizing an important role of APC for actin assembly in neuronal growth cones.
We next used PREM to determine which subsets of actin filament arrays are impaired in APC-depleted growth cones (Fig. 6). As expected (Korobova and Svitkina, 2008), control growth cones (Fig. 6, A–D) were filled with branched actin networks abutting the growth cone edge and also contained filopodia extending beyond the leading edge. In contrast, APC-depleted growth cones (Fig. 6, E–J) were dramatically deprived of branched actin networks and mostly contained long unbranched actin filaments that extended far beyond the most distal microtubule tips in the neurites (Fig. 6, F, H, I, and L). Branched actin networks could be observed only in some APC-depleted growth cones and only as very small patches (Fig. 6, G and I–K). They were usually located at the base of the long unbranched actin extensions and often in the vicinity of the distal microtubule tips (Fig. 6 I, inset; Fig. 6 J). Quantification of the most salient PREM features of the APC knockdown phenotype, the area of the distal cohesive regions of branched actin networks (Fig. 6 K) and the length of the longest unbranched actin extension (Fig. 6 L), in individual growth cones revealed highly significant differences between control and APC-depleted neurons. The severe loss of branched actin networks after APC depletion suggests that APC is important even for steady-state assembly of branched actin networks in growth cones, but not for assembly of unbranched actin filaments, despite an ability of APC to nucleate linear actin filaments in vitro (Breitsprecher et al., 2012; Okada et al., 2010).
Abrogation of growth cone recovery after repellent-induced collapse by APC deficiency
We next tested physiological significance of APC functions in the context of growth cone behavior. We hypothesized that APC can be particularly important to jump-start actin assembly at quiescent cell edges. For example, APC could help the growth cone to resume protrusive activity after its collapse induced by repulsive guidance cues. We mimicked such conditions by using a repulsive hormone semaphorin 3A (Sema3A; Raper, 2000) in the treatment-and-washout experiments (Fig. 7).
After Sema3A treatment, growth cones of control neurons collapsed and acquired an elongated morphology with significantly reduced F-actin contents (Fig. 7, A, C, and D). Following Sema3A washout, F-actin recovery within individual growth cones was not reliably detectable after 5 min but was nearly complete by 20 min, as the F-actin intensity at this time was not significantly different from either the original F-actin level or that at the 60-min time point, despite a trend to increase further (Fig. 7 D). It also appeared that more growth cones were present in the neuronal culture by 60 min washout (Fig. 7 C). In contrast to control neurons, growth cones of APC-depleted neurons showed neither an additional loss of F-actin after Sema3A treatment beyond that achieved by APC knockdown alone nor recovery after Sema3A washout (Fig. 7). An unexpected decrease in F-actin intensity after 5 min washout (Fig. 7 D) was reproducible in two independent experiments. It can potentially be explained by increased susceptibility of growth cones double-challenged by APC knockdown and Sema3A treatment to the mechanical impact of the washing procedure. In any case, these data show that APC is essential for recovery of growth cones after repellent-induced collapse.
Taken together, we show that APC plays key roles for the assembly of branched actin networks in association with microtubules, which is important for initiation of membrane protrusions at quiescent cell edges and for augmentation of preexisting protrusions.
The general concept that microtubules regulate directional cell migration (Vasiliev et al., 1970) and navigation of neuronal growth cones (Tanaka and Kirschner, 1995) was established decades ago, but the underlying mechanisms are still poorly understood. Theoretically, a most straightforward scenario would be that microtubules induce actin polymerization, especially in the form of branched actin networks, which represent the main protrusive machinery in the cell that can reorient the cell leading edge in response to guidance cues (King et al., 2016; Suraneni et al., 2012; Wu et al., 2012). Recent findings support the idea that some microtubule +TIPs, such as APC (Breitsprecher et al., 2012; Okada et al., 2010) and CLIP170 (Henty-Ridilla et al., 2016), can induce assembly of unbranched actin filaments in cooperation with formins. However, a mechanism that would allow microtubules to induce assembly of Arp2/3 complex-dependent branched actin networks was previously neither supported by experimental evidence nor even seriously considered, except for the possibility that microtubules can regulate Rac1 activity, although by not yet fully defined mechanisms (van Haren et al., 2014; Waterman-Storer et al., 1999).
Our results now fill this void by showing that microtubules, with the help of APC, can locally induce assembly of branched actin networks in order to drive local protrusion or launch new protrusions after growth cone collapse. This mechanism adds a conceptually new aspect to the multifaceted actin–microtubule cross-talk in cells. It can explain not only directional cell motility (Vasiliev et al., 1970) and growth cone guidance (Tanaka et al., 1995) but also expansion of dendritic spines following the entry of microtubules (Hu et al., 2011; Jaworski et al., 2009). Although we focused on growth cones in this study, we also noticed that dendritic spines are largely absent in the APC-depleted neurons at DIV14, whereas they are normally abundant in cultured neurons of this age (Korobova and Svitkina, 2010).
Our data suggest that APC is an important coordinator between microtubule dynamics and actin-based protrusion. Among numerous proteins previously proposed to mediate actin-microtubule cross-talk (Dogterom and Koenderink, 2019; Etienne-Manneville, 2013), APC appears as a best candidate. In cells, microtubule-associated APC preferentially localizes to protruding cell regions (Etienne-Manneville and Hall, 2003; Koester et al., 2007; Näthke et al., 1996), whereas APC deficiency leads to abnormal growth cone morphology and perturbed neurite outgrowth in neurons (Koester et al., 2007; Yokota et al., 2009), as well as to impaired polarized migration in nonneuronal cells (Kroboth et al., 2007; Watanabe et al., 2004). However, these phenotypes were previously explained by destabilization of microtubules in the absence of APC (Dent et al., 2011; Gordon-Weeks and Fournier, 2014), even though a link between microtubule stability and cell migration remained largely undefined. Although our results do not exclude a role of APC in regulating microtubule dynamics, we show that APC has an additional role of stimulating formation of branched actin networks in association with microtubules, which bears important novel implications for cell motility.
The mechanism by which APC induces the assembly of branched actin networks can potentially involve both of its actin-regulating activities: nucleation of unbranched actin filaments by the C terminus (Breitsprecher et al., 2012; Okada et al., 2010) and activation of the Cdc42 guanine nucleotide exchange factor Asef (Kawasaki et al., 2000; Mitin et al., 2007), and eventually of the Arp2/3 complex via Cdc42 (Gotthardt and Ahmadian, 2007; Hamann et al., 2007) and N-WASP (Rohatgi et al., 1999), by its N terminus. In this appealing scenario, unbranched actin filaments nucleated by the APC’s C terminus can serve as initial mother filaments to start branched actin nucleation by the Arp2/3 complex activated by the N terminus. However, we were surprised to find that unbranched actin filaments remain abundant in growth cones after APC depletion. These data suggest that APC is not necessary for nucleation of linear actin filaments in association with microtubules. In hindsight, given that binding of microtubules and actin to the C-terminal region of APC is mutually exclusive (Okada et al., 2010), this result is not totally unexpected. We propose that the main contribution of the C terminus of APC in this context is to bind microtubules, so that branched actin nucleation induced by the N terminus of APC was initiated in a spatially defined manner in association with the microtubule that delivered a cluster of APC to the plasma membrane. However, actin-nucleating activity of APC can be important in other physiological contexts. Indeed, this mechanism has been found to be involved in regulation of focal adhesion dynamics in nonneuronal cells (Juanes et al., 2017; Juanes et al., 2019).
Our results show that rather than nucleating unbranched actin filaments, APC is clearly critical for starting branched nucleation on the preexisting unbranched filaments that appear to be nucleated by other nucleators. This function of APC most likely depends on its N terminus, which has a potential to stimulate the Asef–Cdc42–N-WASP–Arp2/3 signaling pathway. This idea is consistent with the previous findings showing that the N terminus of APC has gain-of-function effects in cells (Nelson et al., 2012) and could restore normal morphology in APC knockout neurons (Chen et al., 2011). Our results that APC depletion suppresses growth cone formation appear inconsistent with previous reports showing growth cone expansions in APC-deficient neurons (Chen et al., 2011; Yokota et al., 2009). However, these authors used a conditional knockout mouse (Apc580S; Shibata et al., 1997) in which Cre-mediated recombination is predicted to produce an N-terminal fragment of APC (residues 1–580). Although usually considered to be nonfunctional, this fragment actually retains the most critical residues for the interaction with Asef (Zhang et al., 2012) and thus might be able to activate Asef. Indeed, we were able to reproduce the growth cone expansion and increased branching phenotypes by transfecting Cre recombinase into cultured neurons from newborn Apc580S/580S mice (Fig. S5, E and F). In contrast, our shRNA-mediated strategy is not expected to produce any gain-of-function APC fragments and thus causes a real loss-of-function phenotype.
In conclusion, we reveal a new mechanism by which microtubules can regulate directional migration of neurons or nonneuronal cells. This mechanism involves stimulation of local actin-based protrusions in a microtubule-dependent manner when microtubules approach the cell leading edge. In neuronal growth cones, APC plays a central role in this process by inducing assembly of branched actin networks at microtubule tips. The N-terminal domain of APC appears to be most essential for this activity, probably by stimulating signaling that leads to Arp2/3 complex activation. This signaling is likely triggered by an APC-positive pioneer microtubule when it hits the plasma membrane, where it can encounter the downstream effectors. As a result, the growth cone or the cell leading edge would move in the direction pointed out by the pioneer microtubule. An important aspect of this model is that microtubule-associated actin assembly is not an obligatory mechanism for the formation of branched actin networks, but is employed only in special cases. For example, it can be used to achieve high precision of protrusive activity in response to external signals or be employed when a new protrusion needs to be launched at a quiescent cell edge. Both conditions are expected to frequently occur in neurons. Since precise navigation is of critical importance for neurons, it can explain the high expression levels of APC in neuronal tissues, the APC accumulation in neuronal growth cones, and the link between APC mutations and intellectual disabilities and neurological disorders (Gonzalez et al., 2015; Mohn et al., 2014; Onouchi et al., 2014).
Materials and methods
Dissociated rat embryo hippocampal neurons isolated as described previously (Wilcox et al., 1994) were obtained from the Neurons R Us Cell Service Core (University of Pennsylvania). In brief, hippocampi were dissected from brains of Sprague–Dawley rat embryos at embryonic day 18–20 and dissociated into individual cells by incubating in a trypsin-containing solution. The cells were then washed and plated on poly-L-lysine-coated (0.5 mg/ml) glass coverslips (for fixed cell experiments) or MatTek dishes (for live cell imaging) at concentration of 100,000 cells per 35-mm dish in 1.5 ml neurobasal medium (21103049, Thermo Fisher Scientific) with 2% B27 supplement (17504044, Thermo Fisher Scientific). Live-cell imaging was performed in an environmental chamber at 37°C in the presence of 5% CO2. LatB (428020, Calbiochem) or CytoD (C8273, Sigma-Aldrich) was added to cultures at 12–48 h after plating from stock solutions in DMSO and cultured for an additional 24 h.
All mouse experiments were performed in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania. Apc 580S mice (Shibata et al., 1997) were maintained on a C57/BL6N background. Genotyping was done as described previously (Shibata et al., 1997) using P3 (5′-GTTCTGTATCATGGAAAGATAGGTGGTC-3′) and P4 (5′-CACTCAAAACGCTTTTGAGGGTTGATTC-3′) primers. Mouse hippocampal neurons were isolated from brains of newborn mice and plated on poly-L-lysine–coated (1 mg/ml) coverslips in Neurobasal A medium (10888-022, Thermo Fisher Scientific) with 2% B27 supplement, 10% FBS, 1% Glutamax, and penicillin/streptomycin. After 4 h, when cells were attached, the medium was replaced with one without FBS. At DIV1, neurons were transfected with EGFP-tagged Cre recombinase in a lentiviral vector (Ho et al., 2006), kindly provided by T. Sudhof (Stanford University, Stanford, CA). At DIV6, hippocampal neurons from wild-type and Apc580S/580s mice were fixed with 4% formaldehyde in PBS and stained with Alexa Fluor 568–labeled phalloidin. Cre-expressing neurons were identified by EGFP fluorescence in the nucleus.
REF52 rat embryo fibroblasts and COS-7 green monkey kidney cells were cultured in high glucose DMEM (10569010, Thermo Fisher Scientific) supplemented with 10% FBS (F2442, Sigma-Aldrich) at 37°C and 5% CO2. Cells were plated on glass coverslips for fixed cell experiments or on MatTek glass-bottomed dishes (MatTek Inc.) for live cell imaging. CO2-independent phenol-red-free L15 medium (21083027, Thermo Fisher Scientific) supplemented with 10% FBS was used for live cell imaging.
The following rabbit polyclonal primary antibodies were used: APC-M2 against amino acids 959–1338 of APC (Wang et al., 2009), MAP2 (AB5622, Millipore), and GFP (ab6556, Abcam). The following primary mouse monoclonal antibodies were used: MAP2 (M2320, Sigma-Aldrich), pan-axonal marker SMI312 (SMI-312R, Covance), βIII tubulin (clone Tuj1, MMS-435P, Covance), α-tubulin (clone DM1α, T-6199, Sigma-Aldrich), and pan-actin (ab-5, MS-1295-P0, Thermo Fisher Scientific). Secondary antibodies and phalloidin fluorescently labeled with Alexa Fluor 488, 594, and 680 were from Invitrogen. Secondary gold-conjugated anti-rabbit (711-215-152, 18 nm; 111-205-144, 12 nm) and anti-mouse (115-215-068, 18 nm; 115-205-146, 12 nm) IgG antibodies were from Jackson ImmunoResearch Laboratories. Horseradish peroxidase–linked ECL anti-mouse (NA931V) and anti-rabbit (NA934V) IgG antibodies were from GE Healthcare.
Immunofluorescence staining of neuronal cultures and nonneuronal cells was performed after fixation with 4% formaldehyde (Electron Microscopy Sciences) in PBS for 15 min followed by permeabilization with 0.3% Triton X-100 for 5 min. Confocal fluorescence microscopy was performed using a Nikon Eclipse Ti inverted microscope equipped with PlanApo 20 × 0.75 NA, CFI60 Apochromat TIRF 100 × 1.49 NA oil objective, and QuantEM 512SC digital camera (Photometrics) driven by NIS-Elements software (Nikon). Wide-field epifluorescence microscopy was performed using a Nikon Eclipse TE2000 inverted microscope equipped with Plan Apo 20 × 0.75 NA and 100 × 1.3 NA oil objectives and Cascade 512B CCD camera (Roper Scientific) driven by MetaMorph imaging software (Molecular Devices).
Sample preparation for PREM and immunogold PREM was performed as described previously (Svitkina, 2007; Svitkina, 2009) with minor modifications. In brief, detergent-extracted samples were sequentially fixed with 2% glutaraldehyde in 0.1 M Na-cacodylate buffer, pH 7.3, 0.1% tannic acid, and 0.2% uranyl acetate; critical point dried; coated with platinum and carbon; and transferred onto electron microscopic grids for observation. Detergent extraction was done with 0.5% Triton X-100 in 100 mM Pipes-KOH, pH 6.9, 1 mM MgCl2, and 1 mM EGTA (PEM buffer) containing 2% polyethelene glycol (molecular weight, 35,000), 4 µM phalloidin, and 10 µM taxol for 3 min at room temperature.
For APC immunogold PREM, primary antibody in PEM buffer containing 4 µM unlabeled phalloidin and 10 µM taxol was added for 30 min either to extracted unfixed cells or to cells extracted with the same extraction solution supplemented with 0.5% formaldehyde. Samples were rinsed twice in PEM, incubated in the third change of PEM for 15 min, and fixed with 0.2% glutaraldehyde in 0.1 M cacodylate buffer for 20 min. Samples were quenched with 2 mg/ml NaBH4 in PBS, incubated overnight with secondary gold-conjugated antibodies, fixed with 2% glutaraldehyde, and processed for PREM, as described above. Samples were analyzed using a JEM 1011 transmission electron microscope (JEOL) operated at 100 kV. Images were captured with an ORIUS 832.10W CCD camera (Gatan) and are presented in inverted contrast. Gold particles in PREM samples were identified at 200,000× magnification after contrast enhancement to distinguish them from other bright objects. Color labeling and image overlays were performed using Adobe Photoshop (Adobe Systems), as described previously (Shutova et al., 2012). For color segmentation, microtubules were identified by their unique thickness of ∼25 nm; branched actin networks were recognized by a combination of some or all of the following characteristics: the presence of frequent Y-shaped configurations, short filament lengths, a wide range of filament orientations, and abundance of actin filament ends within a local area of actin network.
Rat hippocampal neurons grown on plastic dishes coated with 0.5 mg/ml poly-L-lysine were lysed in a buffer containing 20 mM Hepes (pH 6.9), 150 mM KCl, 2 mM MgCl2, 1 mM DTT, and 0.5% Triton X-100 or 30 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1% CHAPS. After clarification of lysates by centrifugation at 13,000 g for 10 min at 4°C, the protein concentration in supernatants was determined using Protein Assay (500-0006, BIO-RAD). Protein samples were mixed with the NuPAGE sample buffer (NP0007, Invitrogen) supplemented with 200 mM DTT and 10% glycerol and heated at 75°C for 10 min. About 20 µg (for cell lysates) of total protein was loaded onto 3–8% NuPAGE Tris-acetate 1.0 gradient gels (EA0375, Invitrogen), resolved by SDS-PAGE, and transferred to 0.45 µm nitrocellulose membrane in NuPAGE Transfer Buffer for 2 h at 20 V using the XCell SureLock Mini-Cell Electrophoresis and Western Blotting System (Invitrogen). The membrane was blocked with 5% milk in TBS (170-6435, BIO-RAD) for 1 h, incubated overnight with primary antibodies in 0.5% TBS/Tween 20 (P7949, Sigma-Aldrich) at 4°C for 2 h, washed in TBS/Tween 20, incubated with secondary antibodies (1/20,000 dilution) for 40 min, and developed with ECL Western Blotting Detection Reagent (RPN2232, GE Healthcare).
shRNA and expression constructs and transfection
The following shRNA sequences were used to target APC: 5′-ACGAGCACAGCGAAGAATAG-3′ (sh#1; Mills et al., 2016) and 5′-GCACACTGCACTGAGAATA-3′ (sh#2; Demir et al., 2013). The luciferase-targeting sequence 5′-GCCATTCTATCCTCTAGAG-3′ (shC) was used as nontargeting control. Oligos for shRNA construction were obtained from Integrated DNA Technologies, annealed, and cloned into HpaI and XhoI restriction enzyme sites of the lentiviral pLL-Cerulean 5.0 vector (kindly provided by A. Efimov, Fox Chase Cancer Center, Philadelphia, PA; Smalley-Freed et al., 2010). For producing lentiviral particles, expression plasmids containing shRNA sequences were cotransfected together with helper plasmids MD2G and Pax2 into HEK-293T cells using FuGENE6 (E2691, Promega). Virus was harvested on the third day after transfection, filtered through 0.45 µm filter, and applied to hippocampal neurons at DIV1 together with 5 µg/ml protamine sulfate (P4020, Sigma). Culture medium was replaced after 6–8 h of incubation with the virus. Neurons were cultured for additional five or more days before analysis.
The following expression constructs were used: YFP-APC was a gift from E.M. Wenzel (The Norwegian Radium Hospital, Oslo, Norway) and J. Behrens (University of Erlangen-Nürnberg, Erlangen, Germany; Schneikert et al., 2013), cortactin-dsRed (mouse cortactin cloned into EcoR1-Sal1 sites of pDsRed2-N1 vector) was a gift from S. Kojima and G.G. Borisy (Northwestern University Feinberg School of Medicine, Chicago, IL), cortactin-pmCherryC1 (Addgene plasmid 27676) was a gift from C. Merrifield (Medical Research Council Laboratory of Molecular Biology, Cambridge, UK; Taylor et al., 2011), pLL-mCherry and pLL-EGFP vectors (Smalley-Freed et al., 2010) and pLL-EGFP-EB3 (human) construct were gifts from A. Efimov (Fox Chase Cancer Center, Philadelphia, PA). Full-length pLL-EGFP-APC was generated by PCR from the YFP-APC template and cloned into pLL-EGFP-5.0 vector. The primers were 5′-TTCGTGACCGCGGCCAGATCTG-3′ (forward with SacII restriction site) and 5′-CAAGGGATCCACAGATGTCACAAGGTAAGACCC-3′ (reverse with BamHI restriction site). The PCR product was treated with DpnI enzyme and transformed into DH5α cells.
F-tractin–tdTomato (amino acids 9–40 from inositol-1,4,5-trisphosphate-3-kinase A fused to tdTomato; Johnson and Schell, 2009) was a gift of M. Schell (Uniformed Services University, Bethesda, MD). To generate F-tractin–mCherry and F-tractin–EGFP in lentiviral expression vectors, F-tractin prototype (amino acids 9–52) was amplified from the F-tractin–pEGFP construct using the 5′-GCTCGAATTCATGGGCATGGCG-3′ and 5′-CGGTGGATCCGACCCTGC-3′ primers, digested with EcoRI and BamHI, and recloned into pLL-mCherry and pLL-EGFP vectors, respectively. Similarly, LifeAct (Riedl et al., 2008) was recloned from LifeAct-EGFP (gift from D. Vignjevic, Institute Curie, Paris, France) into pLL-mCherry vector after amplification using the 5′-CTTCGAATTCATGGGTGTCGCA-3′ and 5′-CGGTGGATCCCCTTCTTCC-3′ primers and digestion with EcoRI and BamHI. For transfection, subconfluent cells cultured for 15–24 h were trypsinized, collected by centrifugation at 1,200 rpm for 5 min, and electroporated using the Neon TM transfection system (MPK5000, Invitrogen) with the following parameters: voltage, 1,300 V; width, 10 ms; pulses, 3 (for COS7 cells); and voltage, 1,350 V; width, 30 ms; and pulse, 1 (for REF52 cells).
Sema3A treatment and washout
Sema3A plasmid (gift from J. Raper, University of Pennsylvania, Philadelphia, PA) was transfected into HEK-293T cells using Fugene6 transfection reagent. Medium from transfected and nontransfected (control) cell cultures was collected after 48 h of expression, filtered through a 0.45 µm syringe filter, aliquoted, and stored at –80°C. Expression efficiency was tested by Western blotting before use. The optimal dilution of Sema3A-containing medium that induced maximal reduction of F-actin staining at the lowest concentration was determined in preliminary experiments. For subsequent experiments, DIV7 neurons transfected at DIV1 with control or APC-targeting shRNAs were treated for 1 h with either control or Sema3A-containing medium. To induce recovery, Sema3A was washed out with warm neurobasal medium with B27 supplement. After different time periods of recovery, samples were fixed with 4% PFA in PBS, permeabilzed with 0.3% Triton X-100 for 5 min, and stained with phalloidin and MAP2 antibody for fluorescence microscopy.
Image analysis and statistics
Image analyses and measurements were done using Fiji (ImageJ, National Institutes of Health) or Adobe Photoshop software. Statistical analyses were performed using GraphPad Prism version 8.1.1 for Windows (GraphPad Software).
Quantification of the percentage of different actin filament arrays associated with microtubule tips in growth cones was performed using PREM images of DIV2 neurons stained with mouse DM1α tubulin antibody and 18-nm gold-conjugated anti-mouse antibody. Only clearly distinguishable microtubule tips were included in the analysis. Microtubules were categorized as the following. They were considered to be associated with (a) “branched filaments” if any actin filament physically associated with the microtubule exhibits side branch(es) along its length; (b) “branched patch” if the microtubule terminated in a local enrichment of branched actin filaments, but high network density did not allow visualization of the physical interaction of individual actin filaments with the microtubule; (c) “linear filaments” if actin filaments physically associated with the microtubules, but did not exhibit branches; (d) “filopodial bundles” if microtubules touched and were approximately aligned with bundles of long parallel actin filaments regardless of whether these bundles extended beyond the leading edge of the growth cones or not; and (e) “complex network” if microtubules were entangled with random actin filaments without obvious physical association with any of them. “Naked” microtubule tips not interacting with any actin filaments were not included into quantification.
The APC pattern in growth cones was categorized as “streaks” if immunofluorescence intensity was predominantly concentrated in elongated features with at least a 2:1 aspect ratio, as “puncta” if it represented a collection of nearly round dots, and as “mixed” if both types of clusters were present in growth cones.
Quantification of actin cytoskeleton architecture at neurite tips was performed using PREM images of DIV6 hippocampal neurons treated with control or APC-targeting shRNA starting from DIV1. Every neurite tip without technical defects or extensive overlap with other structures was included in the analysis. For determination of the area of branched actin network at the neurite tip, only the most distal cohesive region occupied by a branched network was measured excluding the filopodial regions that extended beyond the network. The distance from the extreme neurite tip to the most distal branched network was measured along the set of distal unbranched actin filaments. In control neurons, this parameter usually corresponded to the length of the longest filopodium extending beyond the branched network. If there were no branched actin networks at the neurite tip of APC-knockdown neurons, the length of the set of linear filaments was measured up to the most distal microtubule tip. Large, apparently pausing, growth cones observed in control cultures were excluded from the analysis.
Quantification of fluorescence intensities of phalloidin (F-actin marker) and APC immunofluorescence staining in growth cones was performed using Fiji software after background subtraction and thresholding of individual growth cones based on phalloidin staining. For quantifications shown in Fig. 2, neurons were imaged by epifluorescence microscopy at 100× magnification. The same regions of interest were used to determine the growth cone area and fluorescence intensities. For Figs. 4 and 7, neurons were imaged by spinning disk confocal microscopy at 20× magnification, and confocal slices were summed up in Fiji. Dendritic and axonal growth cones were identified based on staining with MAP2 antibody (dendritic marker) and in some cases also with SMI312 antibody (axonal marker).
Quantification of protrusions induced by APC-positive microtubules was performed using time-lapse sequences of REF52 cells coexpressing full-length APC or EB3 with either F-tractin or cortactin in different colors. Kymographs were generated along a straight or segmented line drawn along the path of a moving APC or EB3 signal. Only tracks that reached the cell edge or an actin-rich zone (lamellipodium) at the cell periphery were taken into consideration. If the APC signal lingered at the edge after arrival, it is still considered, but not if there was no preceding track. No lingering of the EB3 signals was observed. Cell protrusion events were scored positive if they began simultaneously with or within 30 s after the APC or EB3 track touched the cell edge or the actin-rich zone. The protrusion distance was measured using kymographs from the position of the edge at the time when the APC or EB3 track touched the edge to the point of maximal advance before the edge position plateaued or the edge began retraction or the next bout of protrusion.
Online supplemental material
Fig. S1 shows microtubule-associated branched actin networks in COS7 cells. Fig. S2 shows localization of APC in growth cones of rat hippocampal neurons treated with actin polymerization inhibitors. Fig. S3 shows a PREM control for specificity of immunogold labeling in a growth cone. Fig. S4 shows induction of membrane protrusions upon encounters of dynamic APC clusters with the cell edge. Fig. S5 shows additional fluorescence microscopy data on down-regulation of APC in hippocampal neurons. Videos 1, 2, 3, and 4 show dynamics of fluorescently tagged APC with either F-tractin or cortactin. Video 5 shows dynamics of growth cones in control and APC-depleted neurons.
We thank Drs. E.M. Wenzel, J. Behrens, S. Kojima, G.G. Borisy, A. Efimov, and T. Sudhof for generous gifts of reagents and Dr. M.S. Shutova for constructing the control luciferase-targeting shRNA.
This work supported by National Institutes of Health grants R01 GM 095977 to T.M. Svitkina, R01 CA 168654 to C.J. Lengner, R50 CA 221841 to N. Li, and R01 CA 109220 to K.L. Neufeld.
The authors declare no competing financial interests.
Author contributions: N. Efimova, C. Yang, and J.X. Chia performed experiments and analyzed data; K.L. Neufeld provided reagents; C.J. Lengner supervised mouse work; N. Li performed mouse breeding and genotyping; T.M. Svitkina oversaw research and wrote the manuscript.