Localization of cofilin mRNA to the leading edge of migrating cells promotes directed cell migration

The family of VICKZ RNA-binding proteins has been implicated in regulating several aspects of post-transcriptional RNA fates – intracellular RNA localization, translation and stability (Yisraeli, 2005; Bell et al., 2013). Through the use of various techniques that reduce expression or activity of VICKZ proteins, a wide variety of functions have been associated with these proteins, such as regulation of cell proliferation, membrane dynamics and cell motility in vivo and in vitro, elongation, regeneration and growth cone turning of axons, invasion and metastasis of cancer cells, and spine formation in nerve dendrites. The search for direct VICKZ–RNA interactions that causatively influence a particular cell behavior, however, has proven to be complicated. First, VICKZ proteins can bind a plethora of RNA targets that can affect the cell in opposing ways (Jønson et al., 2007; Hafner et al., 2010). Second, known targets include RNAs encoding transcription factors and signaling molecules that could affect VICKZ target RNAs themselves. In addition, although VICKZ proteins help to localize certain RNA transcripts to particular intracellular sites, the mere localization of the RNA does not constitute a proof of its functionality. To demonstrate that VICKZ-mediated localization of a given mRNA plays a role in the function of the protein encoded by the RNA, one would, ideally, want to show that it is possible to rescue a knockdown phenotype with the RNA only when it contains the cis-acting element directing its localization through VICKZ proteins.

Cofilin is an actin depolymerizing factor that plays an important role in helping to promote lamellipodium formation by regenerating a pool of G-actin and creating new barbed ends of actin filaments at the periphery of cells (Wang et al., 2007). The local release of caged, constitutively active cofilin (resistant to LIM kinase) has been shown to initiate lamellipodium formation adjacent to its site of release (Ghosh et al., 2004). Indeed, small interfering RNA (siRNA)-mediated knockdown of cofilin leads to a lower G-actin versus F-actin ratio, decreased cell motility and reduced lamellipodium formation (Ghosh et al., 2004; Hotulainen et al., 2005). Its activity appears to synergize with Arp2/3 in generating lamellipodial protrusions, and is held in check by RhoC-mediated activation of LIM kinase that forms a boundary around invadopodia in invasive cells and protrusive lamellipodia in migrating cells (Bravo-Cordero et al., 2011; Zhang et al., 2011; Bravo-Cordero et al., 2013b). Constitutively active cofilin enhances the formation of lung tumors in nude mice injected with PC-3 prostate cancer cells in their tail vein (Collazo et al., 2014). Cofilin mRNA has been reported to be a target of VICKZ proteins (Piper et al., 2006; Jønson et al., 2007).

We hypothesized that cofilin would be a good candidate for a protein whose function might be mediated not only by post-translational processes but also by localization of its mRNA by VICKZ. We demonstrate here that cofilin-1 (hereafter referred to as cofilin) mRNA is rapidly localized to the leading edge of migrating lung carcinoma cells in culture, and this localization is mediated by its 3′UTR to which VICKZ proteins bind. VICKZ1 (also known as IGF2BP1) and cofilin mRNA colocalize at the leading edge, and expression of a dominant-negative VICKZ construct prevents localization of endogenous cofilin mRNA. Analysis of live imaging and phagokinetic tracks of cells in which the amount of cofilin mRNA has been reduced by the use of short hairpin RNA (shRNA) show that reduction in the amount of cofilin protein affects both lamellipodium formation and directed cell motility in these cells. Although lamellipodium formation can be rescued by the introduction of cofilin mRNA with or without its 3′UTR, only the full-length cofilin mRNA is capable of also fully rescuing directed cell motility. These results argue for a functional role for VICKZ-directed mRNA localization in cofilin-mediated regulation of cell motility.

To determine whether cofilin mRNA is localized in migrating cells, we performed fluorescent in situ hybridization (FISH) on serum-starved H1299 lung carcinoma cells (Fig. 1). In the absence of serum, lung carcinoma cells lose their lamellipodia and become sessile, due to upregulation of E-cadherin (Dong et al., 2014). Upon the addition of fresh medium containing serum, the cells begin to form large flat lamellipodia. Cofilin mRNA accumulated at the leading edge of these lamellipodia within 30 minutes of serum addition (Fig. 1B). Virtually no accumulation of cofilin mRNA was observed at the periphery of cells that had not been induced, nor in induced cells that had not yet formed a lamellipodium; furthermore, cofilin mRNA appeared to be homogenously distributed in cells lacking a lamellipodium (Fig. 1A). A control mRNA, GAPDH (Park et al., 2012), showed no significant accumulation at the periphery of either starved or fed cells. Significantly, GAPDH mRNA was predominantly homogeneously distributed even in induced cells with well-formed lamellipodia (Fig. 1C,D). These results have been quantified by calculating the percentage of cells demonstrating mRNA localization to lamellipodia, as determined by the presence of a peripheral band of FISH signal that is clearly separated from the rest of the signal in the cell. Only a small percentage of cells that have been starved overnight contained either cofilin or GAPDH mRNA that had been localized (12% and 0%, respectively; Fig. 1E). After only 30 minutes of incubation with serum, however, localized cofilin mRNA was observed in >60% of the cells, whereas GAPDH mRNA localization was detected in only 15%. Thus, cofilin mRNA is specifically and rapidly localized to lamellipodia in cells induced to migrate.

Many intracellularly localized RNAs contain cis-acting sequence elements in their 3′UTR that are required for their targeting. To enable detection and comparison of exogenous cofilin mRNA constructs, we tagged cofilin mRNA transcripts by cloning a sequence containing 24 MS2 repeats into the beginning of the 3′UTR and performed FISH using a probe that recognizes the repeats. When full-length cofilin mRNA was transfected into H1299 cells and the cells were induced to form lamellipodia, exogenous cofilin mRNA transcripts accumulated at the periphery of lamellipodia, in a pattern very similar to that of the endogenous mRNA (Fig. 2B). Cofilin mRNA transcripts lacking the 3′UTR, however, were not observed at the periphery, even in cells with well-formed lamellipodia (Fig. 2A). Although the percentage of cells with localized full-length exogenous transcripts was lower than that observed for endogenous transcripts (compare Fig. 1), the presence of the 3′UTR made a significant difference to the distribution of the transfected mRNA; 1.8-fold more cells localized the exogenous cofilin mRNA when it was full length as opposed to when it lacked the 3′UTR (Fig. 2C). These results suggest that the 3′UTR of cofilin mRNA is necessary for its localization to lamellipodia.

In the genome-wide study looking for RNAs present in human (h)VICKZ1-containing complexes in human HEK293 cells, Jønson et al. (Jønson et al., 2007) identified cofilin mRNA as a candidate target. In Xenopus lysates made from the heads of stage 35/36 embryos, Piper et al. (Piper et al., 2006) found that cofilin mRNA coimmunoprecipitated with antibodies against Xenopus (x)VICKZ. Reasoning that hVICKZ paralogs might be interacting directly with cofilin mRNA, we immunoprecipitated VICKZ-containing ribonucleoprotein (RNP) complexes from H1299 cell extracts and tested for the presence of cofilin mRNA using quantitative RT-PCR (Fig. 3A,B). An antibody recognizing all three VICKZ paralogs enriches cofilin mRNA concentration over 15-fold compared to pre-immune serum; conversely, mRNA for the housekeeping gene HPRT is virtually non-detectable in the immunoprecipitates with either pre-immune or pan-VICKZ antibody. To initially characterize the region in cofilin mRNA that is recognized by VICKZ paralogs, we incubated recombinant hVICKZ1 individually with eight different 200-nucleotide-long RNA oligonucleotides spanning the 5′UTR, open reading frame and 3′UTR of cofilin mRNA. Oligonucleotides were synthesized in the presence of biotinylated UTP, enabling pull-down of labeled RNA and associated protein with streptavidin beads. Eluted proteins were electrophoresed on SDS-polyacrylamide gels and blotted with the pan-VICKZ antibody. Overlapping of oligonucleotides ensured that RNA sequences would be present in different parts of two distinct oligos, minimizing any potential problems arising from splitting a binding site at the end of an oligo. As seen in Fig. 3C, two overlapping oligonucleotides in the 3′UTR were indeed bound directly by hVICKZ1. No other binding sites for hVICKZ1 were observed throughout the rest of the cofilin mRNA.

To test whether other VICKZ paralogs recognize the 3′UTR of cofilin mRNA, we employed an alternative affinity purification technique involving the use of an RNA aptamer that binds streptavidin with high affinity (Srisawat and Engelke, 2001; Dienstbier et al., 2009). The fragment that most strongly bound recombinant hVICKZ1 (fragment 7, Fig. 3C) was fused to the streptavidin aptamer, and the fusion transcript was immobilized on streptavidin beads and then incubated with an H1299 protein extract. Following extensive washing, bound proteins were eluted and analyzed for the presence of the VICKZ paralogs, using specific antibodies. As seen in Fig. 3D, all three VICKZ paralogs bound fragment 7. Thus, the 3′UTR of cofilin mRNA is a direct target of VICKZ proteins.

To determine whether VICKZ proteins might be involved in helping to mediate cofilin mRNA localization, we compared VICKZ and cofilin mRNA distribution in cells induced to form lamellipodia. H1299 cells, stably infected with lentivirus expressing GFP–hVICKZ1, were serum starved overnight, and then fixed after a 30-minute incubation in fresh medium with serum (Fig. 4). Cells expressing the chimeric protein (Fig. 4A) generated well-formed lamellipodia similar to non-transfected cells, with endogenous cofilin mRNA localizing to these structures as well (compare with Fig. 1). Strikingly, exogenous hVICKZ1 also accumulated at these same regions within the same short period after serum induction. To more causally implicate VICKZ in cofilin mRNA localization, we then made use of a previously described hVICKZ1 deletion construct (termed hVICKZ1ΔKH4 or hVICKZ1-DN). This construct (lacking most of the KH4 RNA-binding domain) functions in a dominant-negative fashion to inhibit RNA binding of all the endogenous VICKZ paralogs (Oberman et al., 2007). When GFP–hVICKZ1-DN was introduced into H1299 cells by lentivirus infection, cofilin mRNA showed a localized pattern predominantly in those cells that did not express the exogenous GFP-labeled construct (Fig. 4B). By defining the degree of cofilin mRNA localization (as shown in Fig. 4D), the extent of this inhibition could be quantified and compared to the effect of expressing GFP alone (Fig. 4C), GFP–hVICKZ1 (Fig. 4A) or GFP–hVICKZ1-DN (Fig. 4B) in the same cell line. Although cells overexpressing GFP–hVICKZ1 showed no significant difference in the amount of clearly localized cofilin mRNA (49% of the cells) as compared to cells expressing GFP alone (44%), cofilin mRNA localization was observed in only 12% of the cells expressing GFP–hVICKZ1-DN (Fig. 4E). The percentage of cells demonstrating localized cofilin mRNA was equivalent to the percentage of cells showing localization under serum starvation (Fig. 1). Taken together, these results argue for a direct role of VICKZ in the localization of cofilin mRNA to nascent lamellipodia.

Local activation of caged, constitutively active cofilin protein has been shown to polymerize actin, generate protrusions and determine directionality of migration at the site of activation in rat mammary carcinoma cells (Ghosh et al., 2004). Knockdown of cofilin mRNA causes a lower G-actin to F-actin ratio and reduces cell motility and lamellipodium formation in breast cancer cells (Ghosh et al., 2004; Hotulainen et al., 2005). To test the effect of interfering with cofilin activity in H1299 cells, cofilin protein expression was reduced by >70% with the introduction of either of two different shRNAs into these cells (Fig. 5A). Both shRNAs were effective at inhibiting cell motility (see below); the sh11 knockdown (termed CFL1-KD) was chosen for subsequent experiments.

Real-time movies comparing serum-induced H1299 wild-type or CFL1-KD cells indicated that reduced levels of cofilin protein were associated with impaired membrane dynamics (supplementary material Movies 1 and 2). To quantify these results, we used ImageJ software to generate from each movie, in an unbiased fashion, kymographs at ten evenly spaced positions along the periphery of the cell (Fig. 5B,C). The net area that the membrane extended, beyond a ring of cortical vesicles present in these cells, was determined for each kymograph and is summarized in Fig. 5D. Areas of <8000 pixels represent a kymograph with no significant lamellipodium formation, such as the CFL1-KD kymograph shown in Fig. 5C. Consistent with previously observed enhanced membrane protrusion and directionality at sites of locally activated cofilin, global reduction of cofilin protein levels in the cell caused a significant (P<0.05) inhibition of the extent of membrane protrusion of ∼30%. The membrane was also much less dynamic, with lamellipodia limited to specific points along the perimeter of CFL1-KD cells (46%); in wild-type cells, lamellipodia extended and retracted in a more even manner along the entire perimeter (77%). Thus, global reduction of cofilin inhibited the ability of the cell membrane to create dynamic lamellipodia.

A high-throughput phagokinetic assay for determining different cell motility parameters has been described previously (Naffar-Abu-Amara et al., 2008). Cells are seeded on a layer of latex beads, and the tracks cleared by motile cells are analyzed using a multi-parametric, morphometric, automated approach. As an initial calibration of this system, we compared wild-type H1299 cells with cells transfected with either full-length or dominant-negative hVICKZ1. Although transfected GFP–hVICKZ1-DN was not expressed at high levels in the H1299 cells (supplementary material Fig. S1A), nevertheless a significant reduction in parameters reflecting both general lamellipodium formation (perimeter, roughness and solidity) and directed motility (axial ratio, area and net path) was observed (supplementary material Fig. S1B–D). These studies confirm and extend our previous findings (Oberman et al., 2007; Vainer et al., 2008) that hVICKZ1 helps mediate cell migration by influencing membrane dynamics. Furthermore, these results demonstrate that this assay can be used to quantitatively compare the effects of different constructs on membrane dynamics.

The high throughput motility assay was then used to quantify the effects of knocking down cofilin protein expression with shRNA. As seen in Fig. 6A, reduction in the levels of cofilin caused the tracks to be rounder, smoother and more solid, with much less movement. A histogram analysis of the distribution of mobility parameters demonstrates that, although the wild-type cells contained a wide range of motilities, the CFL1-KD tracks were more uniform, with reduced mobility (Fig. 6B). Parameters reflecting both lamellipodium formation (Fig. 6C) and directed cell movement (Fig. 6D) were all very significantly (P<0.0001) reduced by the RNA interfence (RNAi) knockdown. Very similar results were observed with the sh13 knockdown (Fig. 4; supplementary material Fig. S2), confirming that the reduction in all of the cell motility parameters was not the result of off-site effects. These results confirm the influence of cofilin in H1299 cells on membrane dynamics, which was observed in the real-time movies (Fig. 5), and demonstrate how the effects of cofilin on different aspects of cell movement (i.e. general lamellipodium formation and directed movement) can be quantified.

To assay the ability of cofilin mRNA constructs to rescue CFL1-KD motility phenotypes, we needed to engineer a cofilin mRNA that would be resistant to the shRNA used to knockdown cofilin mRNA but that would still encode the same protein (Materials and Methods). In addition, to enable normalization of the amount of exogenous cofilin protein expressed in the cells, the transfected cofilin mRNA was fused to the GFP coding sequence. CFL1-KD cells transfected with this construct produced tracks that were indistinguishable from those of wild-type cells (supplementary material Fig. S3).

As shown above, endogenous cofilin mRNA is localized to the leading edge of migrating cells, and this localization requires sequences in the 3′UTR. To test whether this RNA localization plays a role in regulating cell motility, we assayed the effects of rescuing cells containing reduced endogenous cofilin protein levels with cofilin mRNA that does or does not localize. CFL1-KD cells were transfected with shRNA-resistant, GFP-fused cofilin mRNA, either with or without its 3′UTR. The cells were sorted by fluorescence-activated cell sorting (FACS), and those expressing similar low levels of GFP were collected (supplementary material Fig. S4). When the amounts of cofilin were analyzed using an anti-cofilin antibody, similar levels of total cofilin protein (endogenous plus GFP–cofilin) were present in the wild-type cells, in the CFL1-KD cells transfected with full-length RNA and in the CFL1-KD cells transfected with truncated RNA (Fig. 7A,B).

The ability of the cofilin constructs to rescue cell motility parameters was assayed using the high-throughput approach described above (Fig. 7C). The full-length cofilin RNA construct was able to rescue both the general lamellipodium formation parameters (perimeter, roughness and solidity) as well as the parameters that measure directional movement (axial ratio, area and net path; Fig. 7D,E). When the 3′UTR-truncated cofilin mRNA construct was used in the rescue assay, however, the results were mixed. Lamellipodium formation was indeed restored by transfection of the Δ3′UTR mRNA construct, to levels that were very similar to those attained by transfection of the full-length construct (Fig. 7D). Indeed, no statistically significant difference was observed between the rescue with either construct in any of these parameters. These results suggest that the presence of the 3′UTR in the rescuing RNA is apparently not important for cofilin activity associated with general non-polarized membrane dynamics. A different picture was observed, however, when directed cell movement parameters were examined. In these assays, the truncated construct was not able to fully restore directed movement to the levels that the full-length construct did, despite the fact that wild-type levels of cofilin protein were present in the cell (Fig. 7B,E). Although a partial rescue was observed with the truncated construct in all cases, in two of the three parameters (area and net path), there was also a statistically significant difference between the rescue of the full-length as opposed to the truncated construct. In the case of the third directed movement parameter (axial ratio), although the full-length construct showed a statistically significant rescue of the knockdown phenotype, the differences between the full-length and truncated constructs were too small to reach statistical significance, given the intermediate level of rescue provided by the truncated construct. These results suggest that, in order for cofilin to fully enhance directed motility, it must be encoded by an RNA that is localized to the leading edge of the cell.

The data presented here argue that VICKZ proteins play an important role in localizing cofilin mRNA to the leading edge of migrating cells, and that this mRNA localization is important for the ability of the cofilin protein to help mediate directed cell motility. If cofilin mRNA is not localized to the leading edge of the cell, then directed cell movement is impaired, even if cofilin protein is present at normal, endogenous levels. Although the expression of cofilin protein is also required for the efficient generation of lamellipodia in cells, its mRNA need not be intracellularly localized in order to rescue this function in cells with reduced cofilin protein levels.

Cofilin mRNA is localized in 65% of cells within 30 minutes of induction of lamellipodium formation. Analysis of RNA localization was performed on the lowest optical section immediately above the plate, which ensures that there is no optical artifact due to the lamellipodium being raised above the surface of the plate (as observed in the Nomarski images). Using criteria similar to those used to categorize β-actin mRNA localization [scoring cells as localized if there is a significant reduction in signal or empty space between the bulk cytoplasmic signal and the signal at the periphery (Ben-Ari et al., 2010)], the percentage of cells localizing cofilin mRNA is equal to or greater than that observed for β-actin mRNA [whether comparing endogenous cofilin mRNA to endogenous β-actin mRNA (Oleynikov and Singer, 2003) or exogenous cofilin mRNA to exogenous β-actin mRNA (Ben-Ari et al., 2010)]. GAPDH mRNA was chosen as a control for the localization because it has been reported to be uniformly distributed throughout the cell and does not undergo localization (Park et al., 2012). Indeed, in the absence of serum, no cells showed localization of GAPDH mRNA. Upon addition of serum, a low but detectable fraction (15%) of cells show GAPDH mRNA localized at the newly formed lamellipodium. Because these images are not deconvolved, the increase in fluorescence at the lamellipodium might be the result of out-of-focus light from other optical sections in a few cells. Whether or not this represents some low level of localization, this is essentially the same percentage of cells in which cofilin mRNA is scored as localized prior to serum addition, and for this reason, we speculate that this represents the basal background level of localization. Following the addition of serum, however, there is an almost fivefold increase in cells showing clear localization of cofilin mRNA in only 30 minutes.

Several lines of evidence argue for a role for VICKZ proteins in cofilin mRNA localization. First, cofilin mRNA is highly enriched in VICKZ-containing RNPs, and sequences in the 3′UTR of cofilin mRNA are specifically bound by all three VICKZ paralogs (Fig. 3). Second, exogenous cofilin transcripts lacking a 3′UTR show little or no localization as compared to exogenous full-length cofilin transcripts (Fig. 2). Third, expression of the dominant-negative GFP–hVICKZ1-DN construct significantly impairs localization of endogenous cofilin mRNA, as compared to either GFP or wild-type GFP–hVICKZ1 (Fig. 4). Fourth, GFP–hVICKZ1 protein also localizes quickly to the lamellipodia of the serum-fed cells, in the same time frame and to the same location as cofilin mRNA (Fig. 4). Taken together, these data strongly suggest that VICKZ proteins play a role in the localization of cofilin mRNA to nascent lamellipodia in H1299 cells.

The F-actin severing activity of cofilin has been intensively studied over the last few years and has been implicated in promoting both polymerization and depolymerization of actin filaments. Its activity is carefully regulated at the leading edge of migrating cells through a variety of dynamic mechanisms that include phosphorylation and dephosphorylation by locally activated enzymes, association with and dissociation from the plasma membrane, or sequestration by and release from a cortactin-containing complex. Common to all these mechanisms is that activation of cofilin protein at the leading edge helps to produce membrane protrusions in migrating cells, and inhibition of activation leads to defective cell movement (Bravo-Cordero et al., 2013a).

Our results suggest that an additional level of control of cofilin function can be exercised by localization of its mRNA. Despite the distribution of the cofilin protein throughout the cell, localized mRNA could lead to a localized source for cofilin protein that would spatially influence its activity. We observed subtle but clear effects of mild shRNA-mediated knockdown of cofilin protein, using both real-time movie and phagokinetic analyses. The assays look at cell dynamics in the absence of a directed cytokinetic signal and thus reflect the normal processive movement of the cells.

These results, inasmuch as they suggest that the maintenance of a particular level of active cofilin protein at the leading edge is of utmost importance for motility, might also explain why there are several levels of control. We envision that the existence of multiple mechanisms regulating cofilin enables a delicate modulation of its activity. Indeed, in order to detect the effect of cofilin mRNA localization on directed movement in the rescue experiments, it was crucial to ensure that the total amount of cofilin protein in the cell was maintained; given the ability of unlocalized cofilin mRNA to rescue general lamellipodium formation, high expression levels of rescuing cofilin protein could mask differences between the rescuing potential of the two RNA constructs. It is interesting to note that in Drosophila egg chambers, reduction of cofilin gene dosage can partially rescue the initiation of border cell migration inhibited by defective phosphorylation of cofilin protein (Zhang et al., 2011). These results raise the possibility that multiple layers of control involving not only post-translational mechanisms work together to precisely regulate cofilin function at the leading edge of the cell.

A number of studies have demonstrated the importance of VICKZ proteins in helping to regulate cell migration (Farina et al., 2003; Oberman et al., 2007; Vainer et al., 2008; Katz et al., 2012; Stöhr et al., 2012). Among the validated VICKZ target mRNAs are many that encode proteins involved in regulating β-actin polymerization at the leading edge of migrating cells (Gu et al., 2009; Gu et al., 2012). Huttelmaier and colleagues have shown that both stabilization as well as translational repression of mRNAs by VICKZ can influence microfilament formation and how cell movement occurs (Stöhr et al., 2012). The results here show that there can also be significance to VICKZ-mediated localization of an mRNA to the leading edge of a migrating cell that goes beyond simply expressing the protein product in the cell.

Previous reports from a number of labs have suggested that intracellular localization of an mRNA can influence the behavior of the protein product. Integrin α3 mRNA is localized to adhesion plaques through a 3′UTR zipcode sequence recognized by the RNA-binding protein muscleblind-like 1 (MLP1, also known as MBNL2). Using an mRNA encoding a GFP reporter fused to the 3′UTR of integrin α3, Adereth et al. (Adereth et al., 2005) observed a punctate GFP distribution, suggesting that the mRNA localization led to retention of the GFP protein product in plaques. In chick embryo fibroblasts, RNAs encoding all of the subunits of the Arp2/3 complex are localized to the protruding edge of migrating cells (Mingle et al., 2005). Liao et al. (Liao et al., 2011) have shown that when Arp2 mRNA is mislocalized in these cells, cell motility is affected and directionality is lost. Normal cell migration can be restored by introducing into these cells Arp2 mRNA that localizes to protrusions. In neurons, a large number of mRNAs are known to be localized in axons during axon elongation, turning and regeneration (Holt and Schuman, 2013). Twiss and colleagues, in a series of elegant experiments, have demonstrated the importance of sufficient ZBP1 (VICKZ1) levels to ensure adequate localization of specific mRNAs into axons; their experiments also demonstrate that the localized translation of these mRNAs is crucial for axon elongation and regeneration following injury (Donnelly et al., 2011; Donnelly et al., 2013; Yoo et al., 2013). The experiments described here, in which the effects of localizing or not localizing mRNA to its intracellular target are compared under conditions where the levels of protein product are similar, extend these previous experiments and emphasize how trafficking mRNAs to particular locations can influence the effect of the protein product on cell behavior. Although Condeelis and Singer (Condeelis and Singer, 2005) have argued that intracellular localization of β-actin mRNA to the leading edge of migrating cells can have a mass action effect leading to enhanced microfilament polymerization, in the case of cofilin mRNA, the effects appear to be much more subtle. Trafficking of mRNA to particular locations within the cell allows for additional levels of regulation.

Cofilin1 constructs were cloned from human cDNA (NM_005507.2) into the pGEM-T-Easy vector (Promega). The full-length construct contained a portion of the 5′UTR, the entire open reading frame (ORF) and the 3′UTR with a short A-rich stretch of 26 bp at the 3′ end (nucleotides 112–1234). The cofilin-Δ3′UTR construct contained a portion of the 5′UTR and the entire ORF (nucleotides 112–735) but lacked the 3′UTR. For MS2-FISH experiments, a fragment containing 24 MS2 repeats from pSL-MS2-24 [a gift of the laboratory of Yaron Shav-Tal (Fusco et al., 2003)] was inserted after the ORF. The entire fragment containing the 5′UTR, ORF, MS2 repeats and 3′UTR was cloned into pCDNA3 (Invitrogen). shRNAs recognizing human cofilin1 were purchased from Sigma: sh11 (TRCN0000029711) targets nucleotides 460–480 and sh13 (TRCN0000029713) targets nucleotides 498–518. The sh11-resistant fragment was generated using site-directed mutagenesis to create silent point mutations designed to make the construct resistant to the shRNA: 5ʹ-GCCAGATAAGGACTG-CCGCTAT-3ʹ, encoding the amino acid sequence LPDKDCRY was mutated to 5ʹ-CCCCGACAAAGATTGTAGGTAT-3ʹ. The construct was cloned into pEGFP-C1.

The shLUC viral plasmid, with a PLKO backbone and an shRNA sequence (5′-CTTACGCTGAGTACTTCGA-3ʹ) directed against firefly luciferase, was a gift from the laboratory of Shlomo Rotshenker (IMRIC, Hebrew University of Jerusalem). The GFP viral plasmid had a PLKO backbone and was purchased from Sigma (SHCO14). For expression in cultured cells, sequences encoding GFP–hVICKZ1 and GFP–hVICKZ1-DN (Oberman et al., 2007) were cloned into a lentiviral backbone plasmid, a kind gift of Prof. Yinon Ben-Neriah (Hebrew University of Jerusalem). The MS2 coat protein plasmid with a pCDNA3.1+ backbone (Invitrogen) was a gift of the Shav-Tal laboratory. The dsRed viral plasmid, with a PLKO backbone, was a gift from the laboratory of Ittai Ben-Porath (IMRIC, Hebrew University of Jerusalem).

Cell extracts were prepared using Passive Lysis Buffer (Promega), with the addition of 1× Protease Inhibitor (Sigma) and protein concentrations were determined using Bradford reagent (Bio-Rad). 5–20 µg of total protein per well was loaded on 8% (for VICKZ and tubulin) or 12% (for cofilin1) SDS-PAGE gels and transferred to nitrocellulose membranes. The primary antibodies used were rabbit anti-pan xVICKZ (1:20,000; Natkunam et al., 2007), rabbit anti-cofilin (1:1000; Cell Signaling Technology, 3312), chicken anti-rat cofilin1 (1:1000; Sidani et al., 2007) anti-tubulin conjugated to horseradish peroxidase (1:20,000; Abcam, AB40742), mouse anti-VICKZ1 (D-9) (1:1000, Santa Cruz Biotechnology, sc166344), rabbit anti-VICKZ2 (1:10,000, a gift from Finn Nielsen, University of Copenhagen) and goat anti-VICKZ3 (N-19) (1:1000, Santa Cruz Biotechnology, sc47893).

Lysates were prepared from H1299 cells and RNA-protein complexes (RNPs) were immunoprecipitated with 5 µg of either pre-immune serum or anti pan-VICKZ antibody, using Protein A Dynabeads (Life Technologies), according to the RNA immunoprecipitation (RIP) method described in Jayaseelan et al. (Jayaseelan et al., 2014). Total and immunoprecipitated RNAs were isolated using the EZ-RNA Isolation Kit (Biological Industries) and reverse-transcribed with random primers using the ImPromII Reverse Transcription System (Promega) in 20 µl reactions. cDNAs were analyzed by quantitative PCR reaction using the iQ SYBR Green Supermix and the iCycler system (Bio-Rad). Primers used for the amplification were as follows: cofilin1 forward, 5′-AACGACATGAAGGTGCGTAA-3ʹ; cofilin1 reverse, 5′-CTCACTCAGGCAGAAGAGCA-3ʹ; HPRT1 forward, 5′-CCT-GGCGTCGTGATTAGTGA-3ʹ; HPRT1 reverse, 5′- CGAGCAAGACG-TTCAGTCCT-3ʹ. RIP experiments were performed twice. Quantitative RT-PCR was performed on each biological repeat, with reactions done in triplicate in each case.

The NSCLC H1299 line (a gift of the Geiger lab) was maintained in RPMI 1640 (Biological Industries, Beit Haemek) containing 10% fetal calf serum (Biological Industries) and 10 µg/ml ciprofloxacin (Bayer). Transient transfections were performed using TransIT (Mirus) according to the manufacturer's instructions.

Viruses for dsRed, shLuc, shcfl1, GFP, GFP–hVICKZ1 and GFP–hVICKZ1-DN were generated by transfecting third generation viral constructs (a kind gift of Dr Gustavo Mostoslavsky, Boston University School of Medicine, Boston, MA), into HEK 293 cells and collecting the supernatants to infect target wild-type H1299 cells as described previously (Vainer et al., 2008). To create stable lines expressing shLuc and shRNA against cofilin mRNA, the infected cells were incubated with puromycin (1 µg/ml; Sigma) for 2 weeks. Cofilin knockdown was assayed by western blot analysis. For dsRed, GFP–hVICKZ1 or GFP–hVICKZ1-DN infections, efficiency was assayed by analyzing expression of dsRed or GFP.

Stable transfections of GFP–cofilin and GFP–cofilin-Δ3′UTR on a background of the cofilin-knockdown line were sorted using FACSAria (Becton Dickinson). Cells were sorted into four categories: high, medium, low and negative. The same gates were used for each population of cells. The low cells from each stable transfection were used for the high throughput motility assay and western blotting, and were frozen for future use.

To prepare templates for making biotinylated RNA transcripts, PCR reactions were performed on the cofilin-ORF-pGEM and cofilin-UTR-pGEM plasmids using GoTaq (Promega) (supplementary material Table S2). In addition to the cofilin sequences, each primer included a binding site for the T7 promoter.

The PCR products were cleaned with the High Pure PCR product purification kit (Roche). Biotinylated RNA transcripts were prepared by in vitro transcription of the DNA templates with T7 RNA polymerase (Promega) by replacing a sixth of the cold UTP in the reaction with biotin-16-UTP (Roche). RNAs were cleaned on Mini Spin RNA columns (Roche).

Streptavidin MagneSphere paramagnetic particles (Promega; 300 µl per reaction) were pre-washed five times with binding buffer (20 mM Tris-HCl pH 8, 140 mM KCl, 4 mM MgCl₂, 0.1% NP40, 0.75 mM DTT). For the binding reaction, 1 µg of biotinylated RNA was incubated with 10 ng of recombinant VICKZ1 protein in 250 µl of binding buffer, in the presence of 50 µg of tRNA, 20 µg of BSA, 250 µg of heparin and RNasin (Promega), for 25 minutes at room temperature. The pre-washed beads were added to the binding reaction and incubated at room temperature on a rotator for an additional 30 minutes. The magnetic beads were then isolated on a magnet (Dynal), washed and boiled in Laemmli buffer. Eluates were loaded onto a 10% SDS-PAGE gel for western blotting.

The 200-bp fragment from the 3′UTR of cofilin, fragment 7, was cloned into the pTRAP version 5 vector (Cytostore, Inc.), and fragment 7 RNA fused to two copies of a streptavidin-binding aptamer were generated by in vitro transcription of this construct using T7 RNA polymerase. Aptamer-fused RNA (1.5 µg) was immobilized on Streptavidin MagneSphere paramagnetic particles and then incubated with H1299 cell extracts (50 µg) as described previously (Dienstbier et al., 2009). Protein eluted from beads was analyzed by western blotting as described above.

Probes to human cofilin1 mRNA and GAPDH mRNA were synthesized in the Robert Singer laboratory (Albert Einstein College of Medicine, Yeshiva University) according to the online protocol (http://www.singerlab.org/protocols). For each gene, five regions of interest, each 50 base pairs in length, were chosen as probes (supplementary material Table S1). The cofilin and GAPDH probes were labeled with Cy5 and Cy3, respectively. The probes were synthesized with modified Ts about 10 bases apart that were coupled to fluorophores after synthesis.

FISH was performed according to http://www.singerlab.org/protocols on H1299 wild-type cells or stable lines expressing the GFP–VICKZ constructs with the following modifications. To induce lamellipodium formation, cells were starved overnight in complete medium without FCS, followed by addition of complete medium containing 10% FCS for 30 minutes. 20 ng of cofilin probe or 40 ng of GAPDH probe were used for each slide. A hybridization chamber was prepared using a 6-well plate with pre and post hybridization wash in the spaces between the wells to keep the slides moist. The plate was incubated at 37°C overnight. The cells were counterstained with DAPI.

The in situ technique is based on the protocol found in Chartrand et al. (Chartrand et al., 2000) with the following additions. A 50-mer Cy3-labeled DNA probe against the MS2 repeats, 5ʹ-CTAGGCAATTAGGTACC-TTAGGATCTAATGAACCCGGGAATACTGCAGAC-3ʹ, was a gift of the Shav-Tal laboratory. At 1 day after seeding, cells were transfected with either cofilin-MS2-full or cofilin-MS2-Δ3′UTR. After 4 hours, cells were starved in medium without FCS overnight. In the morning, complete medium with 10% FCS was added for 30 minutes to induce lamellipodia.

For FISH of endogenous cofilin mRNA, cells were imaged using the Fluoview FV1000 confocal microscope (Olympus). Pictures were taken at 40×, 60× and 100×. Stacks of 0.5 µm step size were taken, and localization scoring was done on the lowest z, where the lamella was in sharp focus. All slices were examined to ensure that localization was not due to uneven distribution of cytoplasm. Images with the cofilin probe were pseudo-colored red, images with the GAPDH probe were pseudo-colored green. Cells were scored as localized if they met all of the following criteria: (1) the signal was perinuclear; (2) there was a large cytoplasmic gap in the signal; and (3) a thin region of signal was observed along the lamellipodia. Cells were scored as intermediate if the signal diffused to, and even if it was enriched at, the edge of the cell but without a cytoplasmic gap, or if enrichment was confined to a small region of the cell. Experiments comparing cofilin and GAPDH mRNA localization under starved and serum-induced conditions were performed four times; the data shown (Fig. 1) are from one representative experiment (number of cells in the experiment: cofilin starved, 39; cofilin induced, 45; GAPDH starved, 17; GAPDH induced, 13). Experiments assaying the effects of the dominant-negative VICKZ1 on cofilin mRNA localization were performed twice; the data shown (Fig. 4) are from one representative experiment (number of cells in the experiment: GFP, 32; hVICKZ1, 37; hVICKZ1-DN; 34).

In order to quantify signal distribution of the MS2-transfected cells, ImageJ was used to trace the cell body using the transmitted light channel, and the nucleus, using the DAPI channel. These regions of interest (ROIs) were used to measure integrated density and area of the same slice in the Cy3 channel. Statistical analysis was performed using Graphpad software. Fisher's exact test was employed for analyzing the endogenous FISH and the MS2 FISH, and the Chi-squared test was employed for the FISH with the VICKZ constructs. Experiments were performed three times; data shown (Fig. 2) are from one representative experiment (number of cells in the experiment: full length, 56; Δ3′UTR, 37).

This protocol is based on the assay developed by the Geiger laboratory (Naffar-Abu-Amara et al., 2008) with the following modifications. For 30 wells, 1 ml of surfactant-free CML white polystyrene latex beads with a diameter of 0.4 or 0.5 µm (Interfacial Dynamics Corp Batch number 1049.1 and 1307-PHO-1) was used. The lines used for the assay were H1299 cells expressing GFP–hVICKZ1, GFP–hVICKZ1-DN, GFP, shRNA to cofilin mRNA and shLuc, and rescue lines with either cofilin GFP–Δ3′UTR or full-length cofilin–GFP. The plates were incubated under standard conditions for 4–6 hours. Cells were imaged using an Axiovert inverted light microscope (Zeiss) or a LSM 700 confocal microscope (Zeiss) for the rescue experiment. The tracks in each well were analyzed using Phagotracker, a program designed specifically to analyze these types of experiments (Sharon et al., 2006; Naffar-Abu-Amara et al., 2008). After tracks were automatically detected, they were manually approved or rejected. All comparisons between motility experiments were analyzed for statistical significance by determining P-values using a one-tailed Student's t-test. Briefly, the parameters assayed were as follows: for general lamellipodia measurements, track perimeter, roughness [perimeter²/(4π×area)] and solidity (track area/area of the convex hull enclosing the track) were calculated; for directed cell movement, track area, net path and axial ratio (major axis/minor axis) were determined (for further clarification, see Naffar-Abu-Amara et al., 2008). Experiments analyzing the effects of the shRNA knockdowns were performed three times; the data shown are from one representative experiment (wild type, 66 cells; sh11, 100 cells; sh13, 67 cells). The cofilin rescue experiment was performed twice; data shown in Fig. 7 are from one representative experiment (CFL1-KD, 111 cells; CFL1-KD+full, 113 cells; CFL1-KD+UTR-, 137 cells).

In order to study membrane dynamics, H1299 wild-type and cofilin-knockdown cells were compared. The wild-type cells were stably infected with dsRed in order to allow for direct comparison of both cell lines in the same plate. The cells were seeded in 60-mm plates with a glass coverslip, embedded and allowed to settle overnight. The next day, they were starved for 3 hours and treated with PMA (50 ng/ml) for 45 minutes. They were then imaged at one frame per four seconds for a total of four hundred frames, using a LSM 700 confocal microscope (Zeiss). Each movie was 25 minutes in length.

In order to create kymographs, a circle covering the maximum extension of the membrane was drawn around the cell. Using a macro created by Johannes Schindelin (Max Planck Institute of Molecular Cell Biology and Genetics), this circle was divided into ten slices with equal angles. These lines were selected to create kymographs using the kymograph plugin for ImageJ (http://www.embl.de/eamnet/html/body_kymograph.html). Kymographs were created for all of the ROIs where there was no attachment of the membrane to another cell. Using the vesicles in H1299 cells as a marker of the cell body, areas from the lowest point in the membrane to the highest were traced using ImageJ for each kymograph. Regions where there was no lamellar extension were given an area score of zero. Statistics were done using the Mann–Whitney analysis to compare medians. Experiments were performed twice. Statistical analysis was performed on four wild-type and four cofilin-knockdown cells, with ten kymographs per/cell.

We would like to thank the Geiger laboratory at the Weizmann Institute of Science, and particularly Suha Naffar Abu-Amara, for help with the motility assay, and Ofra Golani for her support with the Phagotracker program. We would like to thank the Singer laboratory at the Albert Einstein College of Medicine for preparing the FISH probes, Finn Nielsen (University of Copenhagen) for the anti-VICKZ2 antibody and the Shav-Tal laboratory at Bar-Ilan University for the MS2 probe and helpful discussions. We would also like to thank Yael Feinstein Rotkopf in the Interdepartmental Equipment Facility of the Faculty of Medicine of the Hebrew University for her help with the microscopy.