As they initiate migration in vertebrate embryos, neural crest cells are enriched for methylation cycle enzymes, including S-adenosylhomocysteine hydrolase (SAHH), the only known enzyme to hydrolyze the feedback inhibitor of trans-methylation reactions. The importance of methylation in neural crest migration is unknown. Here, we show that SAHH is required for emigration of polarized neural crest cells, indicating that methylation is essential for neural crest migration. Although nuclear histone methylation regulates neural crest gene expression, SAHH and lysine-methylated proteins are abundant in the cytoplasm of migratory neural crest cells. Proteomic profiling of cytoplasmic, lysine-methylated proteins from migratory neural crest cells identified 182 proteins, several of which are cytoskeleton related. A methylation-resistant form of one of these proteins, the actin-binding protein elongation factor 1 alpha 1 (EF1α1), blocks neural crest migration. Altogether, these data reveal a novel and essential role for post-translational nonhistone protein methylation during neural crest migration and define a previously unknown requirement for EF1α1 methylation in migration.
Neural crest cells migrate throughout vertebrate embryos using stereotyped behaviors (Kulesa and Gammill, 2010). After neural crest cells undergo an epithelial-to-mesenchymal transition to emigrate from the neural tube, they extend lamellipodia and filopodia to contact their neighbors and form chain arrangements (Kulesa and Fraser, 2000; Teddy and Kulesa, 2004; Berndt et al., 2008). Complement component 3a–mediated co-attraction allows these formations to move collectively, while environmental cues stabilize N-cadherin–dependent cell–cell contacts in a polarized manner to promote their directionality (Theveneau and Mayor, 2010; Carmona-Fontaine et al., 2011). Inside the cell, noncanonical Wnt signaling orients neural crest cell protrusive activity through effects on Rho GTPases that coordinate actin cytoskeletal remodeling (De Calisto et al., 2005; Matthews et al., 2008; Clay and Halloran, 2011; Theveneau and Mayor, 2012). Nevertheless, based upon analysis of other migratory cell types, we still have much to learn about neural crest cell migration.
Post-translational modifications regulate protein activity and interactions to affect actin polymerization and cell adhesion in migrating cells (Ammer and Weed, 2008; Rottner and Stradal, 2011; Boulter et al., 2012; Schaefer et al., 2012). For example, signaling through Rho GTPases and the activity of their targets is regulated post-translationally (Boulter et al., 2012). In fact, the phosphorylation status of a downstream target of Rho GTPases, the actin-depolymerizing factor cofilin, regulates neural crest cell directional migration (Y. Zhang et al., 2012). In spite of this, the post-translational control of neural crest migration has been largely ignored.
Just before migration, enzymes of the core methylation cycle are enriched in neural crest cells, suggesting that methylation regulates early neural crest development (Gammill and Bronner-Fraser, 2002; Adams et al., 2008). Indeed, in order for neural crest specification to proceed, the DNA methyltransferase DNMT3A must silence expression of Sox2 and Sox3 that drive neural fate (Hu et al., 2012). In addition, removal of repressive (Strobl-Mazzulla et al., 2010) and addition of activating histone methylation (unpublished data) are required for spatiotemporal regulation of neural crest gene expression that specifies neural crest cell identity (Bajpai et al., 2010; Prasad et al., 2012). Although neural crest cells with methylation-dependent specification defects also fail to migrate, it has been difficult to distinguish a role for methylation during neural crest migration from indirect effects of failed specification. Moreover, as methylation also has transcriptional, post-transcriptional, and post-translational consequences for cell function (Dricu, 2012), nonepigenetic roles for methylation in the neural crest await discovery.
Lysine methylation of nonhistone proteins modulates their stability (e.g., p53 [Chuikov et al., 2004]), localization (e.g., SF2/ASF [Sinha et al., 2010]; HSP70 [Cho et al., 2012]), protein–protein interactions (e.g., BRCA1 [Guendel et al., 2010]), and/or enzymatic activities (e.g., RIP140 [Huang and Berger, 2008; Huq et al., 2009; X. Zhang et al., 2012]). There is also evidence to suggest that nonhistone protein methylation regulates cell migration. For example, S-adenosylhomocysteine hydrolase (SAHH), which breaks down the feedback inhibitor of methylation reactions and is essential for further methylation reactions to proceed, redistributes to the leading edge of motile, chemotaxing Dictyostelium and is necessary to maintain cell polarity that drives chemotaxis (Shu et al., 2006). Meanwhile, Slit-Robo GTPase-activating protein 2 (srGAP2), which promotes protrusive activity to negatively regulate neuronal cell migration, must be arginine methylated for localization into protrusions (Guerrier et al., 2009; Guo and Bao, 2010). In addition, the lysine methyltransferase, enhancer of zeste homologue 2 (Ezh2), is cytoplasmically required for actin polymerization during fibroblast membrane ruffling (Su et al., 2005), and valosin-containing protein lysine methyltransferase is required for invasive cell migratory behaviors in cultured human cells (Kernstock et al., 2012). Moreover, nonhistone protein methylation is essential for bacterial chemotaxis and Toxoplasma parasite motility (Vladimirov and Sourjik, 2009; Heaslip et al., 2011). Nevertheless, the specific nonhistone proteins that are methylated in migratory eukaryotic cells, as well as the possibility that nonhistone protein methylation regulates neural crest migration, have not been investigated.
Here, we analyze the role of methylation in neural crest migration. Chick neural crest cells express SAHH mRNA and protein, and SAHH is required for the emigration of polarized migratory neural crest cells, suggesting that methylation is essential for neural crest migration. In contrast with the established role of nuclear histone methylation in neural crest gene expression (Bajpai et al., 2010; Strobl-Mazzulla et al., 2010; and unpublished data), SAHH and lysine-methylated proteins are abundantly cytoplasmic in migratory neural crest cells. This led us to postulate that cytoplasmic protein methylation regulates the dynamic process of neural crest migration. We profiled cytoplasmic proteins with mono- and dimethylated lysines in migratory neural crest cells, identifying a number of cytoskeleton-associated factors. To test the functional relevance of this methylation during neural crest migration, we focused on one target in particular, elongation factor 1-α 1 (EF1α1), which binds actin filaments and β-actin mRNA to localize actin translation to the leading edge of migratory cells (Liu et al., 2002; Condeelis and Singer, 2005). Mutating the methylated lysines in EF1α1 inhibits neural crest migration. This is, to our knowledge, the first function to be ascribed to EF1α1 methylation. Altogether, our work defines the novel requirement for methylation during neural crest migration, and specifically reveals the importance of nonhistone lysine methylation in migratory neural crest cells.
Neural crest cells express SAHH
Gene expression profiling of neural crest cells identified several enzymes that regulate methylation reactions, including SAHH (Gammill and Bronner-Fraser, 2002; Adams et al., 2008). SAHH hydrolyzes S-adenosylhomocysteine (SAH), which is a byproduct of trans-methylation reactions and a potent methyltransferase inhibitor (Fig. S1 A; Chiang et al., 1996). SAHH mRNA was broadly expressed at varying levels throughout early chicken embryos, but was particularly abundant in premigratory neural crest precursors in the neural folds of cranial, hindbrain, and trunk domains (Fig. S1, B–E and I, white arrowheads), as well as in the nonneural ectoderm (Fig. S1, C and G, black arrows; nne). SAHH expression persisted in HNK-1–positive (Fig. S1 G′, white arrow) cranial migratory neural crest cells (Fig. S1, F–H, black arrowheads). The robust expression of SAHH in neural crest cells suggests that methylation is important for early stages of neural crest development.
SAHH is required for neural crest migration
Because SAHH is essential to clear the methyltransferase feedback inhibitor SAH, one way to prevent methylation is to block SAHH (Fig. S1 A; Fabianowska-Majewska et al., 1994). We first examined the requirement for SAHH during neural crest migration in vivo using a translation-blocking antisense morpholino oligonucleotide (SAHH MO). We unilaterally targeted SAHH MO into neural crest precursors by electroporation at late gastrula, at the time of neural crest induction (Basch et al., 2006; Gammill and Krull, 2011). This allowed SAHH, which is a stable protein (Ueland and Helland, 1983), sufficient time to turn over in targeted cells. After incubation for 8 or 14 h to 4 or 8 somites, cells targeted with fluorescein-modified SAHH MO exhibited reduced or absent SAHH immunofluorescence, indicating that the MO effectively knocked down SAHH protein (Fig. S2, A′′ and B′′, circles).
Although we were interested in migration, sustained SAHH knockdown could also affect specification once sufficient time elapsed for SAH to accumulate and inhibit methyltransferase activity. In particular, DNA methyltransferase 3A (DNMT3A) and the lysine methyltransferase, nuclear SET-domain containing protein 3 (NSD3), are required for neural crest specification (Hu et al., 2012; unpublished data). To gain temporal insight, we evaluated two key neural crest transcription factors: Snail2, a target of neural plate border specifiers, and Sox10, a downstream target of other neural crest transcription factors (Prasad et al., 2012). Expression of Snail2 (Fig. 1, A–C) and Sox10 (Fig. 1, D–F) was scored in SAHH MO–targeted (white arrowhead) compared with untargeted (black arrowhead) sides of electroporated embryos with 4–6 somites in regions of maximal MO targeting. Relative to control MO–electroporated embryos (CO MO), Snail2 expression was minimally altered in SAHH MO–electroporated embryos (Fig. 1 B, white arrowhead; Fig. 1 C, P = 0.07). In contrast, the Sox10 expression domain was mildly to severely reduced in the majority of the SAHH MO–electroporated embryos (Fig. 1 E, white arrowhead; Fig. 1 F, P = 3.44 × 10−3; phenotype examples in Fig. S3). Thus, methyltransferases became inhibited during neural crest specification in SAHH MO–electroporated embryos.
Next, we determined the impact of SAHH knockdown on neural crest migration. First, we examined the distance migrated by the Sox10-expressing cells that formed on the SAHH MO–targeted side compared with the untargeted side at 8–10 somites (Fig. 2, A–C). SAHH MO–targeted neural crest cells exhibited mildly to severely reduced neural crest migration (Fig. 2 B, white arrowhead; Fig. 2 C, P = 2.05 × 10−4; phenotype examples in Fig. S3). As Sox10 expression was affected by SAHH knockdown (Fig. 1 E), we also assayed the general migratory neural crest cell marker, HNK-1, and found that migration of HNK-1–positive cells was similarly reduced (Fig. 2 E, white arrowhead; Fig. 2 F, P = 0.04). Reduced migration was not a result of increased cell death or decreased proliferation (Fig. S2, D–G). Moreover, this phenotype could be rescued by adding back SAHH (Fig. S4). Altogether, these data support a requirement for methylation during neural crest cell specification, and suggest that SAHH, and thus methylation, are crucial for neural crest migration. However, sustained knockdown does not distinguish whether migration is disrupted as an indirect consequence of defective specification, or due to a direct requirement for SAHH during migration.
To block SAHH activity specifically during neural crest migration, we treated neural crest cultures with the efficient SAHH inhibitor tubercidin, or 7-deaza-adenosine (Fig. S1 A; Fabianowska-Majewska et al., 1994). After 16–24 h of culture, carrier-treated neural tube explants showed characteristic migration of HNK-1–positive neural crest cells (Fig. 3 A, white arrowheads) in large numbers (Fig. 3 D) at a distance around the neural tube (Fig. 3 C). In contrast, neural tube explants incubated in 1.0 µM tubercidin produced fewer HNK-1–positive migratory neural crest cells (Fig. 3, B [white arrowhead] and D) that traveled shorter distances away from the neural tube (Fig. 3 C). Interestingly, tubercidin-treated neural crest cells extended processes in all directions (Fig. 3 B′) whereas carrier-treated migratory neural crest cells exhibited polarized protrusions (Fig. 3 A′). Indeed, length/width ratios of tubercidin-treated cells were significantly reduced and closer to 1 (symmetrical) compared with control-treated cells (Fig. 3 E). Inhibition of migration was tubercidin dose dependent, indicating specificity (Fig. S2, H–K). Moreover, tubercidin treatment did not increase cell death in the explants (Fig. S2 L), and tubercidin-treated cells extended protrusions over the full 16–24-h time course of the experiment. This suggests that these effects were not due to inhibition of the alternate tubercidin target, adenosine deaminase, as blocking this enzyme is toxic to cells (Hershfield and Krodich, 1978; Kozlowska et al., 1999). Altogether, these results indicate that neural crest cell emigration, migration, and polarization require SAHH, and thus methylation, independent of its role during neural crest specification.
SAHH localizes to the cytoplasm of migratory neural crest cells
In characterizing SAHH expression, we noted a striking subcellular localization of SAHH protein in neural crest cells. In most cells of the embryo, SAHH was nuclear (Fig. 4 A); however, SAHH immunoreactivity in HNK-1–positive migratory neural crest cells (Fig. 4 B′′, outline) was more diffuse (Fig. 4 B′). High magnification views revealed that this was because neural crest cells (ncc) exhibit cytoplasmically localized SAHH protein (Fig. 4, D and D′, white arrowheads) in addition to nuclear SAHH. In contrast, SAHH was primarily nuclear in surrounding cell types such as head mesenchyme (Fig. 4, D and D′, mes, white arrow) and nonneural ectoderm (Fig. 4 E, nne, white arrow). To further assess this localization, we cultured individual migratory cranial neural crest cells. SAHH was expressed throughout HNK-1–positive neural crest cells (Fig. 4, F and F′′′), but was most concentrated outside the nucleus, particularly in protrusions (Fig. 4 F′′, white arrowheads). SAHH immunofluorescent and Western blot detection are SAHH protein dependent (Fig. S2, A–C), indicating the antibody is specific for SAHH. Notably, SAHH nuclear versus cytoplasmic localization is stage-dependent in Xenopus embryos, and there is some evidence that SAHH interacts with methyltransferases, potentially to localize SAHH to the site of methylation reactions and SAH production (Radomski et al., 1999, 2002; Lee et al., 2012). Thus, the cytoplasmic localization of SAHH led us to postulate that SAHH, and methylation, act outside of the nucleus in migratory neural crest cells.
Migratory neural crest cells contain cytoplasmic methylated proteins
If nonnuclear methylation regulates neural crest migration, methylated proteins should be present in the cytoplasm of neural crest cells. Because other work in our laboratory has identified a neural crest–essential lysine dimethylase (unpublished data), we focused on lysine methylation and obtained an antibody against mono- or di-methylated lysines (K-me1/2; Ab23366 [Abcam]). This antibody immunofluorescently detects only lysine-methylated peptides in arrays of variously modified and unmodified peptides and is a methyl lysine–specific reagent (Levy et al., 2011). In sections of nine-somite chick embryos, K-me1/2 immunoreactivity (Fig. 5, A and B, green) largely surrounded DAPI-stained nuclei (Fig. 5, A and B, blue), consistent with known lysine methylation of diverse and abundant cellular proteins including ribosomal proteins (Ong et al., 2004; Iwabata et al., 2005; Pang et al., 2010). Strikingly, cytoplasmic K-me1/2 immunoreactivity (Fig. 5, A′′ and B′′, white arrowheads) was particularly prominent in HNK-1–postitive cranial migratory neural crest cells (Fig. 5, A′′′ [black arrowheads] and B′′′). To better assess the subcellular localization of lysine-methylated proteins, we evaluated individual cultured cranial migratory neural crest cells (Fig. 5 C). In addition to bright puncta and diffuse staining in the nucleus (Fig. 5 C′′, black arrow), K-me1/2 immunoreactivity was pronounced in the cytoplasm of migratory cranial neural crest cells, particularly in the periphery and in protrusions (Fig. 5 C′′, white arrowheads). Although nuclear lysine methylation was expected given extensive histone methylation, and some cytoplasmic methyl lysine immunoreactivity was anticipated, the abundance of cytoplasmic lysine methylation in migratory neural crest cells suggested a role during neural crest migration.
Proteomic analysis of neural crest cytoplasmic-methylated proteins identified numerous cytoskeletal-associated proteins
Motivated by these results, we devised a proteomic screen to identify cytoplasmic lysine-methylated proteins in migratory neural crest cells (Fig. 6 A). We tested several strategies to isolate neural crest cells for profiling, including expansion of neural crest cells in culture (Etchevers, 2011) and antibody-based separation (Lee and Lwigale, 2008); however, in our hands, cultured neural crest cells differentiated and immune selection produced a mixed population of cells. Instead, we found that manual dissection of dorsal neural folds provided the most controlled harvest of neural crest tissue. Cranial neural folds were cultured to obtain neural crest cells at two time points during migration: emigrating (Fig. 6, “E”; 3 h in culture) and actively migrating (Fig. 6, “A”; 16–36 h in culture and remaining neural fold discarded). Approximately 400 µg of cytoplasmic protein lysate was prepared from each tissue population (E, 400 explants; A, 1,200 explants), and proteins with mono- and di-methylated lysines were immunopurified and separated by SDS-PAGE. Tryptic peptides were identified by liquid chromatography coupled to tandem mass spectrometry with electrospray ionization (LC/ESI/MS/MS). Two replicates of this method identified 182 proteins with high confidence (Table S1). This list included several known lysine-methylated proteins, such as β-actin, α-tubulin, and ribosomal proteins (Iwabata et al., 2005; Pang et al., 2010; Xiao et al., 2010), validating the outcome of the screen. Of these 182 proteins, 19 are known to participate in or regulate the cytoskeleton and were of particular interest. This category included several proteins that are important for regulating cell migration, such as actin-depolymerizing factor, myosin 9, tropomyosin α-1, and several forms of tubulin (Fig. 6 B; Vermillion et al., 2013).
Elongation factor 1-α 1: A lysine-methylated protein in migratory neural crest cells
Due to the limits of embryonic sample collection and the quantity of protein necessary for mass spectrometry (MS), as well as the known challenge of identifying methyl modifications by MS (Ong et al., 2004; Moore et al., 2013), our proteomic screening generally did not identify methylated peptides; one exception was eukaryotic elongation factor 1-α 1 (EF1α1). MS analysis showed that EF1α1 was methylated at five lysine residues in chick neural crest cells (Lys 79 and Lys 316 were tri-methylated, whereas Lys 55, Lys 165, and Lys 290 were dimethylated). Methylation of lysines 55, 165, and 316 was identified with high confidence. Four of these five methylated residues (lysine 55, 79, 165, and 316) have been previously identified in rabbit and human cells, whereas two (Lys 79 and 316) have been identified in yeast, supporting their designation as methylated lysines in chick (Dever et al., 1989; Cavallius et al., 1993; Magrane and Consortium, 2011). Lys 290 is a novel methylated residue, not previously identified as methylated in any organism. As EF1α1 methylation is not required for translation (Sherman and Sypherd, 1989; Cavallius et al., 1997), the function of EF1α1 methylation is unknown.
EF1α1 is primarily known as a component of the translation machinery that shuttles tRNA into the ribosomal A position (Slobin, 1980); however, due to the molar excess of EF1α1 compared with other translation components in the cell, it also serves as an actin-binding protein (Dharmawardhane et al., 1991; Collings et al., 1994; Edmonds et al., 1995). In particular, EF1α1 binds both actin filaments and actin mRNA at the leading edge of polarized migratory cells (Liu et al., 2002). β-Actin mRNA targeting to the leading edge is required for migratory cell polarity and movement (Kislauskis et al., 1994, 1997) and is thought to enable localized actin translation to facilitate actin polymerization for motility (Condeelis and Singer, 2005). Based on this established role of EF1α1 in motility, we examined EF1α1 localization in cultured cranial migratory neural crest cells. EF1α1 immunoreactivity is concentrated around the nucleus, but is also readily apparent as filamentous staining in the periphery (Fig. 7, A–A′, white arrowhead). Co-staining with phalloidin shows EF1α1 colocalization with actin filaments (Fig. 7 A′′) and at higher magnification shows colocalization along actin filaments in lamella (Fig. 7, B–B′′, white arrowheads) and filopodia (Fig. 7, C–C′′, white arrowheads). That EF1α1 exhibits strong, filamentous, peripheral localization and is methylated on several lysines in neural crest cells, along with the established role of EF1α1 in motility, made EF1α1 an excellent candidate cytoplasmic protein to be regulated by methylation during neural crest migration.
EF1α1 methylation is required for neural crest migration
To determine whether EF1α1 methylation is functionally relevant for neural crest migration, we created methylation-resistant EF1α1 by mutating to alanine the six lysines that are methylated in chick and/or human (EF1α1-6x-methyl mutant, or EF1α1-6xMM). We reasoned that, when overexpressed, EF1α1-6xMM would compete with endogenous EF1α1 to incorporate into complexes and bind actin. If EF1α1 methylation is required for an event in neural crest development, the inclusion of methylation-resistant EF1α1 should disrupt that event, even in a background of wild-type EF1α1. Knocking down and replacing EF1α1 with the 6x-methyl mutant was another option; however, we did not pursue this approach because EF1α1 is an essential component of the basal translation machinery (Riis et al., 1990). Moreover, EF1α1 can either prevent or promote actin polymerization depending on its cellular concentration (Murray et al., 1996) and EF1α1 levels are developmentally regulated (Gao et al., 1997), thus exact replacement by overexpression would be difficult.
First, we evaluated whether EF1α1 was correctly localized when fused to GFP and driven from a chick expression construct in cultured neural crest cells. EF1α1-GFP and EF1α1-6xMM-GFP localization resembled the pattern of endogenous EF1α1 immunofluorescence (Fig. 7): GFP fusion proteins were abundant around the nucleus and formed strands within the cell body and in protrusions (Fig. S5, A–F). Thus, EF1α1-GFP and EF1α1-6xMM-GFP were found in the same locations as endogenous EF1α1, suggesting GFP fusion and lysine mutations do not prevent EF1α1 from incorporating into migration-related structures.
Next, we overexpressed EF1α1-6xMM to disrupt EF1α1 methylation in neural crest cells. Vector only, EF1α1-GFP and EF1α1-6xMM-GFP were unilaterally targeted into chick neural crest precursors at late gastrula. Embryos were incubated 8–12 h to 4–6 somites to evaluate specification, or 14–20 h to 8–10 somites to assay effects on migration. In either case, neural crest cells were visualized by Sox10 in situ hybridization. In contrast to vector-only electroporation (pMES; Fig. 8, A, D, E, and H), overexpressing wild-type EF1α1 had variable effects on neural crest development. In specified neural crest cells, Sox10 expression ranged from increased to severely decreased, with the majority of embryos showing no phenotype compared with the untargeted side (Fig. 8, B [white arrowhead] and D; P = 0.02). Neural crest migration distance was also variable, with the majority of wild-type EF1α1-targeted neural crest cells exhibiting mild migration defects when comparing the targeted to untargeted side of the embryo (Fig. 8, F [white arrowhead] and H; P = 0.09). Thus, cells are sensitive to EF1α1 dose, consistent with the observation that EF1α1 can either promote or inhibit actin polymerization at different concentrations (Murray et al., 1996). At 4–6 somites, EF1α1-6xMM elicited a similar range of neural crest specification phenotypes as wild-type EF1α1 (Fig. 8, C [white arrowhead] and D; P = 0.02 compared with pMES, P = 0.72 compared with EF1α1). This suggests the K to A mutations did not disrupt EF1α1 activity during specification, much as a methylation-resistant EF1α does not affect translation in yeast (Cavallius et al., 1997). In contrast, at 8–10 somites, electroporation of the EF1α1-6xMM blocked neural crest migration in most embryos (Fig. 8, G [white arrowhead] and H; P = 4.66 × 10−7 compared with pMES, P = 6.82 × 10−4 compared with EF1α1). This migration defect was not due to cell death (Fig. S5, H and I). Moreover, EF1α1-6xMM did not disrupt down-regulation of the cranial neural crest epithelial cadherin6B, indicating that this feature of epithelial-to-mesenchymal transition preceding migration took place on schedule (Fig. S5 J). Together, these data reveal that EF1α1-methylated lysines are essential for neural crest migration and demonstrate a novel role for nonhistone protein methylation during neural crest migration.
Although data in the literature are consistent with a role for nonhistone protein methylation in eukaryotic cell migration, direct evidence, particularly during development, is lacking. This study defines the importance of nonhistone protein methylation during neural crest migration. We show that the methylation cycle enzyme SAHH is required for neural crest cells to migrate away from the neural tube, revealing for the first time that methylation is essential for neural crest migration. Cytoplasmic localization of SAHH and lysine-methylated proteins in migratory neural crest cells motivated a proteomic screen, which identified an extensive list of methylated and putatively methylated nonhistone proteins. Characterization of one of these proteins, the actin-binding protein EF1α1, showed that its methylated lysines are required for neural crest migration. Taken together, our data show that nonhistone protein methylation, in particular methylation of EF1α1, is essential for neural crest migration.
Although the focus of our analysis was migration, our experiments also give insight into neural crest specification. In SAHH MO–electroporated embryos assayed at four somites, we noted a striking difference: Snail2 expression was minimally disrupted (Fig.1, A–C), while Sox10 expression was moderately or severely reduced in half the embryos analyzed (Fig. 1, D–F). First of all, this indicates that methylation became inhibited in SAHH MO–electroporated embryos during specification. Neural crest specification requires DNA methyltransferase 3A (DNMT3A) to silence neural transcription factors Sox2 and Sox3 that repress Snail2 and Sox10 expression (Hu et al., 2012). Because Snail2 was expressed (Fig. 1 B), SAHH protein turnover (Fig. S2) and the resulting accumulation of SAH must inhibit methyltransferases after the critical period for DNMT3A activity. Second, the distinct effects of SAHH knockdown on Snail2 and Sox10 expression support their placement in the neural crest transcriptional hierarchy: Snail2 is regulated by neural plate border specifiers, while Sox10 is turned on later in response to other neural crest transcription factors (Prasad et al., 2012). SAHH knockdown-dependent inhibition of methyltransferase activity likely occurs after Snail2 up-regulation, but before activation of Sox10 expression. Third, our results are consistent with differential epigenetic regulation of Sox10 and Snail2, as the lysine methyltransferase NSD3 is required for histone H3 lysine 36 methylation of Sox10 but not Snail2 (unpublished data). The idea that these genes are regulated by different methyltransferases or marked with different combinations of methyl marks that decay at different rates in a state of methyltransferase inhibition (Barth and Imhof, 2010) is consistent with our data. Further characterization of NSD3, identification of other neural crest methyltransferases, and analysis of methylation is needed to evaluate these possible mechanisms.
The neural crest methyl proteome offers an unprecedented view of cytoplasmic protein methylation in a migratory cell type. Given the importance of post-translational modifications like phosphorylation in migration (Rottner and Stradal, 2011; Y. Zhang et al., 2012), we were intrigued by the abundance of lysine-methylated proteins in the cytoplasm of migratory neural crest cells (Fig. 5 C′′, white arrow) and set out to profile these proteins. Our efforts identified 182 methylated and putatively methylated proteins, providing the first view of this level of regulation during cell migration. Although extensive, this collection is likely not exhaustive; for most proteins only a few peptides were identified, and these peptides generally did not contain the methylated residue. Peptide coverage for each protein and the number of proteins identified were undoubtedly limited by the amount of embryonic sample we were able to obtain. As a result, failure to identify a peptide in the spectra does not necessarily mean it is absent from the tissue, and comparison between time points is not possible. Moreover, it is a known challenge to identify methyl modifications by mass spectrometry, due to the size of the resulting mass shift and three possible states (mono, di, tri), as well as the fact that lysine methylation can prevent trypsin digestion and affect subsequent detection (Ong et al., 2004). Even in yeast, where sample collection is nonlimiting and metabolic labeling strategies can be applied, profiling has identified limited methylated peptides (Moore et al., 2013). Despite these challenges, we identified five methylated lysines in EF1α1, four of which are validated by work in other systems (Dever et al., 1989; Cavallius et al., 1993; Magrane and Consortium, 2011), lending support to our analysis.
In addition to the identification of EF1α1-methylated lysines, several features indicate that this is a high quality profile. First, all proteins were identified with high confidence. Moreover, we identified many known lysine-methylated proteins. Ribosomal proteins formed the largest category (Table S1), and ribosomal proteins, including the 60S ribosomal protein L27 and the 40S ribosomal protein S13 we identified, are lysine methylated (Pang et al., 2010; Levy et al., 2011). β-Actin, α-tubulin, and EF1α1 are also known lysine-methylated proteins (Cavallius et al., 1993; Iwabata et al., 2005; Xiao et al., 2010). Finally, although we prepared cytoplasmic extracts with an established, commercially available reagent, cross-contamination with nuclear proteins is expected to be ∼10% (Thermo Fisher Scientific). Thus, we identified histone proteins, including all four core histones that are well characterized for their lysine methylation (Young et al., 2010) and further support our profile. Putting this all together, it is likely we were able to detect EF1α1 because it is methylated at multiple lysines and due to its sheer abundance within cells. This is consistent with and undoubtedly accounts for much of the elevated cytoplasmic K-me1/2 immunoreactivity in migratory neural crest cells (Fig. 5). In sum, this collection of putatively methylated, nonhistone proteins provides an exciting first glimpse of cytoplasmic nonhistone protein methylation during neural crest migration and is an important resource for future studies to determine the function, extent, and stability of methyl marks on nonhistone proteins.
As a functional validation of the screen and our hypothesis that nonhistone protein methylation regulates neural crest migration, we showed that neural crest cells expressing methylation-resistant EF1α1 fail to migrate (Fig. 8). Although EF1α1 methylation was first detected three decades ago, and EF1α1-specific methyltransferases have been identified in yeast, the function of EF1α1 methylation is unclear (Hiatt et al., 1982; Lipson et al., 2010; Couttas et al., 2012). All evidence suggests that it is not required for translation (Sherman and Sypherd, 1989; Cavallius et al., 1997; Polevoda and Sherman, 2007). Given this, and the fact that EF1α1 regulates the actin cytoskeleton in migrating cells, it was intriguing to identify it as a cytoplasmic methylated protein in migratory neural crest cells (Fig. 6). Indeed, in comparison to EF1α1 overexpression, which consistently elicited a wide range of effects on specification and migration, a methylation-resistant form of EF1α1 specifically blocked neural crest migration (Fig. 8). This phenotype is particularly striking because methylated and unmethylated EF1α1 are equivalent in translation assays (Sherman and Sypherd, 1989), and yeast with methylation-resistant EF1α1 are phenotypically normal (Cavallius et al., 1997). Methyl mutant EF1α1 did not disrupt down-regulation of the cranial epithelial cadherin, cadherin6B (Fig. S5 J), and some neural crest cells emerged from EF1α1-6xMM-GFP–electroporated neural folds (Fig. S5, D–F), implying that epithelial-to-mesenchymal transition can take place when EF1α1 methylation is defective. This suggests that EF1α1 methylation is required for motility, potentially to regulate EF1α1’s role in the actin cytoskeleton (Condeelis and Singer, 2005). In combination with SAHH MO knockdown (Fig. 2) and tubercidin experiments (Fig. 3), these data also reveal that cytoplasmic protein methylation is essential for neural crest migration.
What is the function of EF1α1 methylation during migration? Methylation-resistant EF1α1 was not mislocalized and still found in strand formations (Fig. S5, D–F); thus, EF1α1 methylation is unlikely to affect binding to actin filaments. One possibility is that methylation-resistant EF1α1 disrupts EF1α1 protein–protein interactions that are required for local actin translation, nucleating actin polymerization, or other actin-related functions. Indeed, there is evidence to suggest that EF1α1 methylation alters its protein–protein interactions (Sherman and Sypherd, 1989). EF1α1 methylation could also be required for β-actin mRNA binding and targeting (Condeelis and Singer, 2005). In favor of this, β-actin mRNA targeting to the leading edge is required for cell polarization and motility (Kislauskis et al., 1994, 1997), as is methylation (Figs. 2 and 3), although disrupting EF1α1 methylation does not lead to a statistically significant reduction in neural crest cell polarity (Fig. S5 G). EF1α1 was lysine methylated at both E and A migration time points, suggesting it is involved throughout neural crest migration. Whatever the role of EF1α1 methylation, it is a migration-specific function, as neural crest specification was equivalently affected by overexpression of wild-type EF1α1 or EF1α1-6xMM (Fig. 8), consistent with previous reports that EF1α1 methylation is dispensable for translation-related activities (Sherman and Sypherd, 1989; Cavallius et al., 1997). These findings expand our view to include nonhistone methylation as a novel layer of post-translational control in the neural crest, and open new avenues of research to understand the mechanism by which EF1α1 methylation regulates cell migration.
Materials and methods
Fertilized chicken embryos were obtained from local sources. Eggs were incubated at 37°C in a humidified incubator until the desired stage of development (Hamburger and Hamilton, 1992), judged by counting somite pairs.
Morpholinos and DNA constructs
FITC-tagged morpholinos (MOs) were synthesized by GeneTools, LLC with the following sequences: translation blocking SAHH MO 5′-CAGCCTGTCCGACATGCTGGAGGCA-3′; and standard control MO 5′-CCTCTTACCTCAGTTACAATTTATA-3′ (CO MO). For SAHH overexpression and rescue, full-length SAHH was subcloned into pMES-mCherry (Roffers-Agarwal et al., 2012). Full-length chick EF1α1 was PCR amplified to include terminal EcoRI and SspI sites. The internal ribosome entry site in pMES (Swartz et al., 2001) was excised by EcoRI–MscI digest, and chick EF1α1 was inserted to create an in-frame EF1α1-GFP fusion. Methylation site mutations were made using the QuikChange Multi Site-Directed Mutagenesis kit (Agilent Technologies).
Ex ovo early embryo electroporation was performed on late gastrula stage 4–5 embryos as described previously (Gammill and Krull, 2011; Roffers-Agarwal et al., 2012). In brief, embryos were adhered to Whatman filter paper, isolated from the yolk, rinsed in chick Ringer’s saline, and placed in fresh Ringer’s in an electroporation cuvette with a 4-mm gap. 1.0 mM MO (SAHH or control) or 3 µg/µl DNA (EF1α1 constructs or pMES vector) was injected from the ventral side into the subvitelline space adjacent to neural crest precursors and electroporated using five square-wave 7 V 50-msec pulses with 100-msec gaps. Embryos were then cultured on agar-albumin plates until the desired stage.
Neural crest cultures
Cranial neural folds were dissected from 4–7 somite embryos. Stage 16 trunk neural tubes were prepared by explanting the region of the last 10 formed somite pairs, incubating the tissue in dispase solution (1.2 U/ml; Life Technologies) on ice for 15 min followed by 10 min at 37°C, and aspirating through a fire-polished Pasteur pipette to separate the neural tube from surrounding tissues. Cranial or trunk explants were then cultured for 16–24 h at 37°C on 10–100 µg/ml fibronectin-coated glass coverslips (Thermo Fisher Scientific) in neural crest complete media (L15, 1% l-glutamine, 0.1% penicillin/streptomycin [all from Life Technologies], 10% FBS [VWR Scientific], and 10% chick embryo extract [Bronner-Fraser and García-Castro, 2008]). For evaluation, cultures were fixed in 4% paraformaldehyde and permeabilized in PBS + 1.0% Triton X-100.
A 150-mM stock of tubercidin (Sigma-Aldrich) was prepared in 8.7 M (50%) acetic acid. Trunk neural tubes were incubated in chamber slides (Thermo Fisher Scientific) with neural crest complete media containing 58 µM acetic acid (carrier) or 0.1–1.0 µM tubercidin (58 µM final concentration of acetic acid). Well dividers were removed and slides were fixed and immunostained as indicated.
In situ hybridization
Embryos were infiltrated with 5% and 15% sucrose, embedded in gelatin in 15% sucrose, frozen in liquid nitrogen, and sectioned with a cryostat (model CM1900; Leica) at 10–20 µm. Gelatin was removed from the sections by incubating for 30 min in 42°C PBS.
Sections, cultures, and whole embryos were blocked in PBS + 10% fetal bovine serum + 0.1% Triton X-100 for 30 min at room temperature and stained with anti-HNK-1 (ATCC; Manassas, VA), anti-SAHH (anti-32-5B6; DSHB, Iowa City, IA), anti-K-me1/2 (Ab23366; Abcam), anti-EF1α1 (Abgent), anti-cleaved caspase3 (rabbit anti-cl.casp3; Cell Signaling Technology), and anti–phospho-histone H3 (rabbit anti-pH3; EMD Millipore) followed by the appropriate secondary antibody (mouse AF488, rabbit AF568, mouse AF568 [Life Technologies]; or Cy2 anti–mouse/rabbit IgG, Cy5 anti–mouse/rabbit IgG, RRX anti–mouse IgM [Jackson ImmunoResearch Laboratories, Inc.]) as indicated. For some assays the signal was amplified using a mouse anti–rabbit IgG (Jackson ImmunoResearch Laboratories, Inc.).
Microscopes and image acquisition
To visualize electroporation targeting or in situ hybridization results, whole-mount embryos were imaged in PBS + 0.1% Tween using a stereoscope (Discovery V8; Carl Zeiss) with an Achromat S 1.0× lens and fluorescence module outfitted with a GFP 500 filter cube. Images were acquired with a digital camera (AxioCam MRc5; Carl Zeiss) using Axiovision 4.8.2 software (Carl Zeiss). Slides and cultures were mounted with Permafluor (Thermo Fisher Scientific) containing 1 mg/ml DAPI. In situ hybridized sections were imaged on a microscope (Axioimager A1; Carl Zeiss) with a Plan Apochromat 20×/0.8 NA lens and acquired with the AxioCam MRc5 camera using Axiovision 4.8.2 software. Immunofluorescently labeled cells and sections were imaged using a confocal system (LSM 710; Carl Zeiss) attached to an inverted microscope (Observer Z.1; Carl Zeiss) outfitted with Plan Apochromat 10×/0.45 NA, 20×/0.8 NA, 40×/0.95 NA, and 100×/1.46 NA lenses and FS38, FS43, and FS49 filter sets. Confocal images were acquired using ZEN 2010 software (Carl Zeiss), and two-dimensional maximum intensity projections were produced from three-dimensional z-stack images using the data from the highest intensity pixels along the projection axis. All imaging was performed at room temperature. Image files were assembled in Photoshop (Adobe).
Quantification of migratory neural crest cell numbers, migration distance, and polarity
Seven neural tube explants in each condition (carrier or 1.0 µM tubercidin) were immunostained with anti–HNK-1 antibody. HNK-1–positive cells surrounding the neural tubes were counted and their distance migrated designated by their position relative to concentric boundaries drawn 120 µm, 240 µm, or 360 µm away from the borders of the neural tube. To evaluate polarity, 29 carrier-treated cells and 21 cells treated with 1.0 mM tubercidin were measured at their longest length and perpendicular width using ImageJ (National Institutes of Health; Schneider et al., 2012). Ratios of length to width were calculated in Excel (Microsoft).
Because electroporated constructs are inherited mosaically, phenotype was scored at the cranial axial level with maximal fluorescent MO or GFP signal in the dorsal neural tube. Effects on premigratory gene expression were analyzed in whole mount by judging the relative amount of colorimetric in situ hybridization signal on the targeted and untargeted side of the embryo (Fig. S3, A–D). A reduction encompassed both lower levels of mRNA and/or fewer specified neural crest cells. Effects on migration were evaluated by comparing the distance migrated on targeted and untargeted sides (Fig. S3, E–H). All phenotypes were scored blind by three independent judges and statistics were performed using two-sided Fisher’s exact test in R (R Development Core Team, 2011).
Cell death and proliferation
For the tubercidin assay, cultures were fixed in 4% paraformaldehyde for 15 min and cell death was detected using the In-Situ Cell Death Detection kit (Roche) according to the manufacturer’s instructions. For SAHH MO and EF1α1 overexpression assays, embryos were sectioned and immunostained with anti-cleaved caspase3 to assay cell death and anti–phospho-histone H3 to assay proliferation. Immunoreactive cells were counted in the dorsal half of the cranial neural tube on both the targeted and untargeted sides of the embryo for at least five sections per embryo (n = 5) and significance was evaluated by t test in Excel (Microsoft).
Emigrating neural crest cells (E) were collected from 4–7 somite cranial neural folds after they were cultured in neural crest complete media for 3 h and harvested. Actively migrating (A) neural crest cells were prepared by culturing cranial neural folds for 16–36 h, removing the neural fold, and collecting the migratory neural crest cells by 3-min treatment with trypsin-EDTA (Life Technologies). Cytoplasmic protein extracts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction kit, and protein concentration was determined by BCA Assay (both from Thermo Fisher Scientific). Cytoplasmic lysates were split equally into four tubes and incubated for 30 min at room temperature with rotation in control goat IgG antibody (5 µg; R&D Systems) to preclear, and rabbit polyclonal to mono-and di-methylated lysine (K-me1/2, 5 µg; Abcam) mixed with Protein G magnetic beads (Life Technologies) to immunoprecipitate. After incubation, beads were washed three times with PBS. Proteins were eluted from the beads by boiling in water. Eluates were combined and concentrated by speed vacuum. After addition of SDS-PAGE sample buffer containing Bromophenol blue and 5% β-mercaptoethanol, protein samples were boiled. Immunoprecipitated proteins were resolved using a Mini-PROTEAN TGX 4–15% gel (Bio-Rad Laboratories) and then silver stained (SilverQuest; Life Technologies) or stained with Imperial Protein stain (Thermo Fisher Scientific) in accordance with the manufacturers’ instructions. Lanes were cut into five sections, and destained according to the manufacturers’ instructions.
Excised gel pieces were trypsin digested (Shevchenko et al., 1996) using an Investigator ProPrep (Genomic Solutions) and lyophilized. For the first screen, digested peptide mixtures were desalted using C18 columns and mass spectrometry was performed on an LTQ mass spectrometer (Thermo Fisher Scientific) as described previously (Beckmann et al., 2013). Sequest (Thermo Finnigan) was set up to search the NCBI nonredundant Gallus gallus (September 03, 2010 version) database. Search parameters were: cysteine iodoacetamide; trypsin; instrument LTQ; and variable modifications-oxidized methionine. For the second screen, digested peptide mixtures were desalted with C18 resin according to the “Stage Tip” procedure (Rappsilber et al., 2003). Using a mass spectrometry–based approach (Lin-Moshier et al., 2013), peptides were analyzed using a mass spectrometer (Velos Orbitrap; Thermo Fisher Scientific). ProteinPilot 4.5 (Ab Sciex) searches were performed against the NCBI nonredundant Gallus gallus database (September 03, 2010 version) to which a contaminant database (www.thegpm.org/crap/index.html) was appended. Search parameters were: cysteine iodoacetamide; trypsin; instrument Orbi MS (1–3 ppm) Orbi MS/MS; biological modifications ID focus; special modifications-purified histones, thorough search effort; and False Discovery Rate analysis (with reversed database). Proteins included in Table S1 had two or more peptides identified with 95% confidence.
Online supplemental material
Fig. S1 shows the mRNA expression pattern of the methylation cycle enzyme SAHH. Fig. S2 shows efficacy and specificity controls for SAHH knockdown and inhibition. Fig. S3 shows neural crest phenotype categories. Fig. S4 shows SAHH MO rescue. Fig. S5 shows EF1α1 overexpression controls. Table S1 lists proteins identified by proteomic analysis.
We are grateful to Y.-C. Cheng for the gift of Sox10 plasmid. We would like to thank M. Murphy, Y. Kawakami, and the members of the Gammill laboratory for experimental suggestions, advice, and comments over the progress of this research. Many thanks to W. Gong for assistance with statistical analysis. Special thanks to the University of Minnesota Mass Spectrometry and Proteomics Facility, in particular L. Anderson, L. Higgins, and B. Wittuhn, for logistical advice and performing mass spectrometry, and T. Markowski for mass spectrometry sample preparation. We also acknowledge the resources of the Minnesota Supercomputing Institute. The monoclonal antibody 32-5B6, developed by C. Dreyer, was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242.
This work was supported by National Institutes of Health Individual NRSA F31 DE019755-01 to K.L. Vermillion; and National Science Foundation Research grant IOS-1052101, March of Dimes Basil O’ Connor Starter Scholar Award, and University of Minnesota Academic Health Center Seed Grant awards to L.S. Gammill.
K.L. Vermillion’s present address is Department of Biology, University of Minnesota Duluth, Duluth, MN 55812.