Signal-peptide-mediated ER localization of mRNAs encoding for membrane and secreted proteins, and RNA-zipcode-mediated intracellular targeting of mRNAs encoding for cytosolic proteins are two well-known mechanisms for mRNA localization. Here, we report a previously unidentified mechanism by which mRNA encoding for Dia1, a cytosolic protein without the signal peptide, is localized to the perinuclear ER in an RNA-zipcode-independent manner in fibroblasts. Dia1 mRNA localization is also independent of the actin and microtubule cytoskeleton but requires translation and the association of Dia1 nascent peptide with the ribosome–mRNA complex. Sequence mapping suggests that interactions of the GTPase binding domain of Dia1 peptide with active Rho are important for Dia1 mRNA localization. This mechanism can override the β-actin RNA zipcode and redirect β-actin mRNA to the perinuclear region, providing a new way to manipulate intracellular mRNA localization.
Numerous studies have demonstrated that mRNA localization plays an important role in cellular functions such as development, differentiation and cell migration (Corral-Debrinski, 2007; Czaplinski and Singer, 2006; Du et al., 2007; Holt and Bullock, 2009; Lecuyer et al., 2009; Martin and Ephrussi, 2009; Meignin and Davis, 2010; Mili and Macara, 2009; St Johnston, 2005). Among the mechanisms by which intracellular mRNAs are targeted to specific compartments, two are relatively well studied: co-translational enrichment of mRNAs encoding for secreted and membrane proteins on the endoplasmic reticulum (ER), and RNA-localization-sequence (zipcode) -mediated intracellular localization of mRNAs encoding for cytosolic proteins. In the first case, an mRNA, after initial translation to produce the signal peptide, is directed to the ER surface (Blobel and Dobberstein, 1975a; Blobel and Dobberstein, 1975b; Walter and Blobel, 1981). This translation-dependent and signal-peptide-mediated mRNA localization on the ER is believed to be specialized for the secreted and membrane proteins. This localized translation on the ER and insertion of the polypeptide products into the lumen ensure the subsequent specialized modifications for this group of proteins. The second mechanism is generally for the mRNAs encoding for cytosolic proteins, which lack the signal peptide. The localization of these mRNAs relies on one or multiple zipcode(s) on the RNA, often in the 3′-untranslated region (3′-UTR) (Chabanon et al., 2004; Hesketh, 2004; Moore, 2005; Oleynikov and Singer, 1998; St Johnston, 2005). Many transacting factors (e.g. zipcode binding proteins) interact with the zipcode(s) and regulate the transport, anchorage and translation of the corresponding mRNA. Motor proteins and the cytoskeleton also play a crucial role in these processes (Kiebler and Bassell, 2006; Lopez de Heredia and Jansen, 2004; Meignin and Davis, 2010; St Johnston, 2005), and in some cases membrane structures are also involved (Cohen, 2005; Kraut-Cohen and Gerst, 2010; Schmid et al., 2006).
Although ongoing translation is important for tethering the mRNAs encoding for secreted and membrane proteins on the ER, translation is usually not required for the localization of mRNAs encoding for cytosolic proteins. In fact, most of the mRNAs of the cytosolic proteins are translation-silent during transportation to their anchoring sites (Chartrand et al., 2002; Czaplinski and Singer, 2006; King et al., 2005; Lopez de Heredia and Jansen, 2004; St Johnston, 2005). The translation of these mRNAs does not occur automatically upon the mRNAs reaching their destination, but is strictly regulated by local signaling pathways (Bassell and Kelic, 2004; Huttelmaier et al., 2005; Paquin et al., 2007; Rodriguez et al., 2006). Recently, some mRNAs encoding for cytosolic proteins were found in the ER fraction (Lerner et al., 2003). The mechanism of sequestration of these mRNAs on the ER is unknown but their localization is translation-independent (Pyhtila et al., 2008). Mislocalization of mRNA often results in significant defects. In somatic cells, one of the best-studied examples of mRNA localization is β-actin, whose mRNA and protein are localized at the leading lamellae of motile cells and in the growth cone of developing neurons (Condeelis and Singer, 2005; Lawrence and Singer, 1986). β-actin mRNA localization is mediated by the zipcode in its 3′-UTR (Kislauskis et al., 1994) and its delocalization results in slower cell movement, loss of cell migration directionality and defects in neuron development (Eom et al., 2003; Kislauskis et al., 1997; Shestakova et al., 2001).
Dia1 is a member of the diaphanous-related formins (DRFs), which nucleate actin polymerization (Chesarone and Goode, 2009; Pollard, 2007). Unlike the Arp2/3 complex, which mediates the formation of highly branched actin networks (Mullins et al., 1998), DRFs mediate the formation of unbranched actin filaments and are responsible for the formation of stress fibers and filopodia (Faix and Grosse, 2006; Goode and Eck, 2007; Schirenbeck et al., 2005; Tominaga et al., 2000). The DRFs by themselves are in a folded and inactive conformation due to the intramolecular interactions between the diaphanous inhibitory domain (DID) near the N-terminus and the diaphanous autoregulatory domain (DAD) near the end of the C-terminus (Alberts, 2001; Li and Higgs, 2003). DRFs are activated through binding of Rho GTPase to their GTPase binding domain (GBD) to relieve them from intramolecular auto-inhibition (Alberts, 2001; Higgs, 2005; Tominaga et al., 2000; Watanabe et al., 1997). Knockdown of Dia in cells reduced both microtubule stability and polarized cell migration (Goulimari et al., 2005; Wen et al., 2004). Mice in which Dia1 is deleted exhibit defects in cell migration and proliferation and a higher risk of cancer (Eisenmann et al., 2005; Eisenmann et al., 2007; Peng et al., 2007). Overall, the DRFs are distributed uniformly throughout the cell and, in some instances, are concentrated at the cell periphery (Watanabe et al., 1997; Westendorf et al., 1999). DRFs can be found on membrane structures, presumably through their binding to Rho GTPases that associate with membranes (Cussac et al., 1996; Ridley, 2006), or through a Rho-independent mechanism (Seth et al., 2006). The intracellular distribution of mRNA for the DRFs is unknown.
As a prominent member of the Rho GTPase family, Rho plays an important role in many cell-signaling processes (Buchsbaum, 2007; Jaffe and Hall, 2005; Ridley, 2006). Rho is present in the cytoplasm but also on membrane structures owing to its post-translational modifications (Ridley, 2006). Through its downstream effectors such as formin and rock, Rho regulates the cytoskeleton, cell adhesion, migration and gene expression (Arai et al., 2002; Arthur et al., 2002; Fukata et al., 2003; Geneste et al., 2002; Hall, 2005; Malliri and Collard, 2003; Raftopoulou and Hall, 2004). Several reports also suggest that Rho might be involved in mRNA localization as manipulation of Rho activity altered mRNA localization (Latham et al., 2001; Mingle et al., 2009). However, these effects are likely to be indirect and probably result from the effects of Rho on the cytoskeleton and motor proteins as both are known to be required for the localization of these tested mRNAs. To date, a direct role for Rho in mRNA localization has not been reported.
We previously reported that all the mRNAs of the Arp2/3 complex, a seven-protein actin polymerization nucleator, are localized at the protrusions of fibroblasts (Mingle et al., 2005). The localization of these mRNAs is consistent with their protein localization at the leading edge of migrating cells (Machesky et al., 1997) and is predicted to facilitate complex assembly at the site of function through local translation (Mingle et al., 2005; Moore, 2005). During a test to see whether the mRNA encoding Dia1 is also localized at the cell protrusions, we were surprised to notice that Dia1 mRNA is localized at the perinuclear region of the cell. We sought to understand how and why Dia1 mRNA localizes so differently as compared with the mRNAs of the Arp2/3 complex. We now report that Dia1 mRNA is localized on the perinuclear ER membrane through a previously unidentified mechanism that is different from both signal-peptide-mediated ER localization of mRNAs of the membrane and secreted proteins, and the RNA-zipcode-mediated intracellular localization of mRNAs of cytosolic proteins.
Dia1 mRNA is localized to the perinuclear region of fibroblasts
We previously demonstrated that mRNAs of the actin polymerization nucleator Arp2/3 complex localize to the protrusions of fibroblasts (Mingle et al., 2005). Because DRFs also nucleate actin assembly, we wondered whether or not, and where, DRF mRNAs localized in cells. We examined the intracellular distribution of mRNA encoding for chicken Dia1 (cDia1) in chicken embryo fibroblasts (CEF). Unexpectedly, Dia1 mRNA localized in the perinuclear region of the cells (Fig. 1) even though Dia proteins were previously shown to be uniformly distributed or even slightly enriched at the cell periphery (Watanabe et al., 1997; Westendorf et al., 1999). A similar pattern of Dia1 mRNA was also observed in human foreskin fibroblasts (HFF). These results are intriguing because the mRNAs of the Arp2/3 complex and Dia1, the two major actin assembly nucleators, exhibit very different distribution patterns in the same cells (Fig. 1). To begin analyzing the mechanism and physiological significance of Dia1 mRNA localization, we transfected the CEF with a full-length cDia1 cDNA expression plasmid to test whether exogenously expressed cDia1 mRNA exhibits a similar distribution pattern. As shown in Fig. 2A, the exogenous cDia1 mRNA indeed exhibited perinuclear localization similar to that of the endogenous mRNA. Because the expression level of exogenous cDia1 mRNA was higher than the endogenous mRNA, the scoring of mRNA localization was facilitated and, unless otherwise indicated, ectopically expressed Dia1 was used for the rest of our studies.
Dia1 mRNA localization is independent of the actin- and tubulin-based cytoskeleton
Most of the localized mRNAs encoding for cytosolic proteins are associated with the cytoskeleton (Hesketh, 2004; Kiebler and Bassell, 2006; Lopez de Heredia and Jansen, 2004; Meignin and Davis, 2010; St Johnston, 2005). To test if Dia1 mRNA localization is also dependent on the cytoskeleton, we treated the CEF with cytochalasin D and colchicine to disrupt actin filaments and microtubules, respectively (Casella et al., 1981; Sternlicht and Ringel, 1979) (supplementary material Fig. S1). Unlike the mRNAs of the Arp2/3 complex, whose localization requires intact actin and microtubule cytoskeleton (Mingle et al., 2005), the localization of Dia1 mRNA was not altered by the treatment of either drug (Fig. 2B–D). These results clearly show that the cytoskeleton is not required for Dia1 mRNA localization.
Dia1 mRNA is localized on the ER
We next tested whether Dia1 mRNA is localized on intracellular membrane structures such as Golgi and ER because these membrane structures are relatively enriched in the cell body instead of the cell periphery. Sequential detergent extraction has been used to separate cytosolic and ER or Golgi membrane-bound transcripts from cells (Stephens and Nicchitta, 2007). This method is based on the differential sensitivity between the plasma membrane and the ER or Golgi membrane to detergents caused by their differences in membrane lipid components. After titration of the dose of detergent saponin for HFF (supplementary material Fig. S2), sequential extraction assays were performed. Marker proteins GAPDH (cytosolic) and calnexin (membrane-bound) showed expected distribution in these extraction samples (Fig. 3A). Dia1 protein was also predominantly in the cytosolic fraction. RNAs in these extractions were also analyzed (Fig. 3B). Unlike the predominant cytosolic distribution of GAPDH protein, there was a significant portion of the GAPDH mRNA in the membrane fraction, which has been observed previously by others (Lerner et al., 2003). This membrane fraction of GAPDH mRNA might represent the GAPDH mRNA molecules that were in the nucleus and were transported through the nuclear pores at the time of extraction. Nonetheless, even with such a background, calnexin mRNA is still a useful marker for ER-bound transcripts as there is a significant difference between the ratios of membrane-bound versus cytosolic mRNAs for GAPDH and calnexin, respectively (Fig. 3B). Like the mRNA of ER marker calnexin, Dia1 mRNA was dominant in the membrane fraction, indicating its association with the intracellular membranes. To further distinguish whether Dia1 mRNA is localized on the ER or Golgi or both, we detected Dia1 mRNA and Golgi and ER marker proteins in the same cells. As shown in Fig. 3C, Dia1 mRNA showed little colocalization with the Golgi marker protein golgin 97. Furthermore, we used brefeldin A to disrupt the Golgi and found no detectable alteration in Dia1 mRNA localization (supplementary material Fig. S3). In contrast to the case of the Golgi, Dia1 mRNA showed colocalization with ER marker proteins protein disulfide-isomerase (PDI; Fig. 3C) and calnexin (supplementary material Fig. S4). Quantitative analyses of Dia1 mRNA and calnexin protein distribution in HFF indicated that the vast majority of Dia1 mRNA colocalized with calnexin (supplementary material Table S1). Density gradient fractionation assays also demonstrated that Dia1 mRNA was cofractionated with the ER but not with the Golgi (Fig. 3D). In addition, we have also excluded the possibility that Dia1 mRNA localizes in the stress granule because stress granule marker protein TIAR and the Dia1 mRNA in HFF did not colocalize (Fig. 3E). The combined results from our multiple approaches consistently demonstrate that Dia1 mRNA is localized on the ER membrane.
The Dia1 coding region, not the 3′-UTR, contains the mRNA localization element
Most localized mRNAs contain one or multiple localization signal sequences (zipcode) in their RNA molecules, which can be recognized by zipcode binding proteins (Eom et al., 2003; Oleynikov and Singer, 1998). In most of the studied cases, the zipcode locates in the 3′-UTR (Chabanon et al., 2004; Moore, 2005; St Johnston, 2005), although a few zipcodes were also found in the coding region (CR) (Chartrand et al., 1999; Gonzalez et al., 1999). In plants, the localization of glutelin mRNA to the ER in rice endosperm cells is mediated by three RNA zipcodes that locate in the 5′ and 3′ ends of the CR and in the 3′-UTR, respectively (Washida et al., 2009). We first examined whether the 3′-UTR or CR of Dia1 mRNA contained a perinuclear localization zipcode. As shown in Fig. 4B and E, mRNA containing only the CR of cDia1 localized to the perinuclear region, whereas that containing a cDia1 3′-UTR did not. This suggested that the localization element was in the CR instead of the 3′-UTR of cDia1. We further tested whether a leading-edge localization zipcode of β-actin mRNA could override the CR of Dia1 and therefore alter the perinuclear localization of such chimeric mRNA. Interestingly, this chimeric mRNA of the Dia1 CR and β-actin 3′-UTR showed a perinuclear localization pattern that is indistinguishable from that of the cDia1 CR alone. These results indicate that the cDia1 CR contains the perinuclear localization element, which can even override the effect of a leading-edge localization zipcode of β-actin.
ER localization of Dia1 mRNA is mediated by nascent peptides that couple to the mRNA–ribosome complex in ongoing protein translation
Because the sequence for Dia1 mRNA localization is in the CR, it is possible that it is the polypeptide instead of the RNA sequence that is the element for the mRNA localization, even though previous examples showed that the RNA localization elements in the CR are RNA sequences (Chartrand et al., 1999; Gonzalez et al., 1999; Washida et al., 2009). To test this possibility, we converted the CR of cDia1 into a 3′-UTR downstream of the CR of GFP. Interestingly, this chimeric mRNA lost perinuclear localization completely (Fig. 5B,D), suggesting that the Dia1 polypeptide is required for its mRNA localization. To exclude the possibility that the loss of perinuclear localization of this chimeric mRNA was due to altered position and subsequently altered secondary structure of an RNA zipcode, we created a frame-shift mutation by inserting an ‘A’ immediately after the ATG start codon of the cDia1 CR. This mutation is expected to have minimal effects on the RNA secondary structure but to produce a protein unrelated to cDia1. Again, this frame-shift mutant mRNA lost perinuclear localization (Fig. 5C,D), confirming that the polypeptide, and not the RNA sequence, was the element for cDia1 mRNA localization.
The above results indicate a role for the Dia1 polypeptide in its mRNA localization and suggest that the translation of Dia1 mRNA into Dia1 polypeptides plays a key role in Dia1 mRNA localization. This reasoning led us to predict that nascent peptide coupling to the mRNA–ribosome complex is required for mRNA localization. To test this prediction, we used cycloheximide, a drug that prevents nascent peptide dissociation from the ribosome (Godchaux et al., 1967; Lodish, 1971), to treat the CEF. As predicted, Dia1 mRNA still localized to the perinuclear region in the drug-treated cells (Fig. 5E,G). We then used another drug, puromycin, which causes premature release of nascent peptide from the mRNA–ribosome complex (Joklik and Becker, 1965; Yarmolinsky and Haba, 1959). Puromycin-treated cells showed loss of perinuclear Dia1 mRNA localization (Fig. 5F,G). Thus, the association of the nascent polypeptides with the mRNA–ribosome complex is essential for Dia1 mRNA localization. Furthermore, to test whether the localizing Dia1 mRNA is being translated, we used a method of biarsenical dyes and tetracysteine tagging to detect translation sites in the cells as previously reported (Rodriguez et al., 2006). As shown in Fig. 5H, the newly translated Dia1 protein is localized in the perinuclear region, indicating an ongoing translation of Dia1 mRNA there.
The GBD and DID domains in the N-terminal coding region of Dia1 are required for Dia1 mRNA localization
We next mapped the minimal sequence in the CR of cDia1 required for Dia1 mRNA localization. A series of truncations of the CR were made and tested for their corresponding mRNA localization in transfected CEF. As summarized in Fig. 6A, an N-terminal fragment of approximately 334 amino acid residues appeared as the minimal region for perinuclear mRNA localization and any further truncation from either end led to dramatic loss of perinuclear mRNA localization. A corresponding N-terminal region of human DIA1 truncation also showed similar perinuclear mRNA localization. This region contains two previously identified domains, the Rho GTPase binding domain (GBD) and the diaphanous inhibitory domain (DID) (Higgs, 2005). To ensure that the DID is required to enhance GBD interaction with Rho GTP rather than functioning as a non-specific sequence to maintain a minimal time of translational elongation for the coupling of the nascent peptide with the mRNA–ribosome complex, we made two constructs: one (p-20) expressing both the GBD and DID plus a red fluorescence protein mCherry, and another (p-21) expressing the GBD and only part of the DID plus mCherry (Fig. 6B). As expected, mRNA from p-20 localized to the perinuclear region. However, mRNA from p-21 showed loss of perinuclear localization even though its polypeptide is predicted to be longer than the minimal sequence (GBD/DID) that is sufficient to localize its mRNA to the perinuclear region. These results indicated that, in addition to the GBD, the specific sequence of DID is also necessary for mRNA localization.
Dia1 mRNA localization is mediated through its nascent peptide interaction with Rho GTPase
Because the GBD and DID domains are known to interact directly with active Rho (Otomo et al., 2005; Rose et al., 2005), we went on to test whether the interaction of Dia1 nascent peptide with active Rho is crucial for Dia1 mRNA localization. Expression of C3-exoenzyme (C3), a Rho inhibitor (Chardin et al., 1989; Rubin et al., 1988), resulted in loss of Dia1 mRNA localization (Fig. 7B,E), suggesting that Rho activity is required for Dia1 mRNA localization. To ensure that the observed effect of C3 on Dia1 mRNA localization was due to the specific disruption of interactions between the Dia1 nascent peptides and the active Rho instead of a global effect of Rho inhibition to the cell, we created single point mutations on Dia1 to abolish the interaction. It was reported that amino acid residues V161 and N165 of mammalian Dia1 (mDia1) are crucial for Rho binding and that a single point mutation of either residue disrupted mDia1 binding to Rho–GTP (Otomo et al., 2005; Rose et al., 2005). We made two single point mutation constructs at the corresponding residues of cDia1 (V184D and S188D, respectively) and analyzed the distribution of these mutant mRNAs. As shown in Fig. 7C–E, both mutant mRNAs lost perinuclear localization, indicating that specific binding of Dia1 nascent polypeptides with active Rho is essential for Dia1 mRNA localization. To test whether there was Rho in this perinuclear region for the predicted interaction, we stained CEF and HFF and showed that RhoA protein is relatively enriched around the nucleus. Furthermore, RhoA can be activated in this region as indicated with a RhoA biosensor using fluorescence resonance energy transfer (FRET; Fig. 7H).
Dia1 mRNA localization overrides the RNA zipcode and mislocalizes β-actin mRNA
In Fig. 4, we showed that the chimeric Dia1-CR and β-actin 3′-UTR localized like the full-length Dia1 mRNA, suggesting that Dia1 nascent-peptide-mediated mRNA localization can override the RNA zipcode. To test whether this property can be used to alter RNA-zipcode-mediated mRNA localization, we chose the relatively well-studied β-actin mRNA as an example. By expressing it as a bi-cistronic mRNA, we mislocalized β-actin mRNA to the perinuclear region with high efficiency (Fig. 8). This result not only confirms the potential use of the system for manipulation of mRNA localization but also provides a hint that RNA-zipcode-mediated translation suppression might be subjected to regulation in the perinuclear compartment en route to its final destinations in the cytoplasm.
In this report, we present data demonstrating a new mechanism by which Dia1 mRNA is localized on the perinuclear ER. Although Dia1 is a cytosolic protein, its mRNA localization is independent of RNA zipcode and the cytoskeleton but requires translation. This makes this mechanism distinct from that for the intracellular localization of mRNAs encoding for cytosolic proteins, as the latter generally requires RNA zipcode and the cytoskeleton system but not translation, although there are exceptions. Furthermore, the mechanism for Dia1 mRNA localization is also different from that for the mRNAs encoding secreted and membrane proteins because Dia1 is a cytosolic protein without a signal peptide. The requirement of active Rho GTPase and its interaction with the GBD and DID domains further defines the mechanism of Dia1 mRNA localization, which was previously unidentified, as distinct from all currently known mechanisms.
Although Rho has been implicated in mRNA localization (Latham et al., 2001; Mingle et al., 2009), it is not clear how Rho mediates the process. Because of the well-known effects of Rho GTPases on the cytoskeleton and motor proteins (Arai et al., 2002; Arthur et al., 2002; Fukata et al., 2003; Geneste et al., 2002; Hall, 2005; Malliri and Collard, 2003; Raftopoulou and Hall, 2004), and that these downstream targets are required for the localization of many mRNAs, it is probable that the previously reported effects of Rho on the localization of those mRNAs are indirect. By contrast, the role for Rho in Dia1 mRNA localization reported here is to interact with the GBD and DID domains on the Dia1 nascent peptide. To the best of our knowledge, this is the first evidence that Rho plays a direct role in mRNA localization.
The precise mechanism through which Dia1 nascent peptide interacts with active Rho to mediate Dia1 mRNA localization on the ER has yet to be elucidated. We here propose two working models for the probable scenario. The first model suggests that Dia1 nascent peptide interacts with Rho–GTP, which is bound to the ER membrane, thereby tethering the mRNA–ribosome complex on the ER. This is supported by the observations that Rho can bind to membrane structures, including the ER (Cussac et al., 1996; Ridley, 2006), and that RhoA can be enriched and activated in the perinuclear region of a variety of cell types (Gadea et al., 2007; Liu and Horowitz, 2006; Picard et al., 2009; Su et al., 2009) (Fig. 7F–H). Furthermore, Dia1 was also reported to bind to RhoA in the perinuclear region as detected by FRET (Eisenmann et al., 2005). Our alternative model suggests that Dia1 mRNA is translated by the ribosomes that are already anchored on the ER and the interaction of the nascent peptide with active Rho (not necessarily ER-bound) prevents the mRNA–ribosome complex from releasing into the cytosol. This is based on previous reports that a subgroup of mRNAs encoding for cytosolic proteins could be targeted to the ribosomes on the ER and initially translated (Nicchitta et al., 2005; Potter and Nicchitta, 2000). However, unlike the mRNAs encoding for the secreted and membrane proteins, those encoding for the cytosolic proteins were subsequently released with the ribosomes into the cytosol owing to the lack of signal peptide. It is possible that interactions between RhoA–GTP and the Dia1 nascent peptide alter such a process and prevent the Dia1 mRNA–ribosome complex from releasing into the cytosol. Both models predict that Dia1 mRNA is promptly translated once it exits the nuclear pore. mRNA is transported out of the nuclear pore in a 5′-to-3′ direction and the ribosome can bind to the mRNA that is emerging (Cheng et al., 2006). We observed that exogenous Dia1 mRNA distributed in a narrower zone around the nucleus than did endogenous Dia1 mRNA (Figs 1, 2, 3 and 4). This might result from the synchronization of transfection-induced transcription of exogenous Dia1 mRNA and implies that these mRNA molecules were translated once they exit the nuclear pore.
The capability of Dia1 to override β-actin zipcode is intriguing. It is well accepted that the fate of mRNA in the cytoplasm begins in the nucleus because RNA binding proteins in the nucleus have a large impact on the subsequent processes (Moore, 2005). However, as the mRNA exits the nucleus, subsequent interactions with cytoplasmic factors also define its final destination. β-actin mRNA is known to bind to zipcode-binding proteins ZBP1 and ZBP2 in the nucleus and both factors are important for normal β-actin mRNA localization in the cell protrusions (Gu et al., 2002; Oleynikov and Singer, 2003; Pan et al., 2007). The translation of β-actin mRNA is suppressed by ZBP1 until such suppression is relieved by Src-mediated phosphorylation of ZBP1 in the cell periphery (Huttelmaier et al., 2005). It is therefore interesting to note that the β-actin zipcode could not inhibit the perinuclear localization of the Dia1–β-actin fusion transcripts (Figs 4, 8), suggesting that the zipcode failed to suppress translation. To directly confirm the translation-dependence of the Dia1 CR in the context of β-actin zipcode fusion, we made a Dia1–β-actin construct in which the translation of Dia1 is blocked by insertion of a stop codon at the 5′ end of the Dia1 coding region (supplementary material Fig. S5). mRNA from this construct lost perinuclear localization, supporting the translation-dependence of Dia1 to override β-actin zipcode (supplementary material Fig. S5). Why the β-actin zipcode failed to suppress the translation of Dia1 is unknown. Future studies on this question could provide insights as to how the translation of Dia1 and β-actin mRNA is regulated in the perinuclear region. On the application side, an immediate benefit of such overriding capability is the potential to effectively alter intracellular mRNA localization. Because this system can override the RNA zipcode, it does not require either previous knowledge about the zipcode or any mutation with the interested gene. Indeed, in addition to β-actin mRNA, we have successfully used this approach to redirect Arp2 mRNA (a subunit of the Arp2/3 complex) to the perinuclear region, which resulted in altered cell migration (Liao et al., 2010).
The identification of a new mechanism for Dia1 mRNA localization has raised many questions. For example, what is the physiological significance of targeting Dia1 mRNA to the perinuclear ER? It is intriguing to observe that mRNAs encoding the Arp2/3 complex and Dia1, the two major actin polymerization factors, exhibit such different localization in the same cells. Given that Dia1 mRNA localization requires the interaction of GBD and Rho GTP, it would be of great interest to determine how many other mRNAs that encode proteins containing a GBD domain would be similarly targeted. Beyond the scope of this report, answers to these questions from future studies are expected to provide new insights for mRNA targeting and localized protein translation.
Materials and Methods
Digoxigenin-11-dUTP (DIG-11-dUTP), sheep anti-DIG antibody (peroxidase-conjugated) and sheep anti-fluorescein antibody (peroxidase-conjugated) were from Roche (Indianapolis, IN). Antibodies against calnexin and PDI were from Stressgen (Ann Arbor, MI) and goat anti-58K Golgi protein antibody was from Abcam (Cambridge, MA). Antibodies against Dia1 and RhoA were from Santa Cruz (Santa Cruz, CA). Mouse anti-TIAR antibody was from BD Biosciences (San Jose, CA) and mouse anti-golgin-97 antibody was from Molecular Probes (Eugene, OR). Superase-in RNase inhibitor and mouse anti-GAPDH antibody were from Ambion (Austin, TX). RNasin was from Promega (Madison, WI). TC-FlAsH II and TC-ReAsH II In-cell Tetracysteine Tag Detection Kits were from Invitrogen (Carlsbad, CA). Tyramide signal amplification (TSA) reagents were purchased from Perkin Elmer (Boston, MA). Other general chemicals were from Sigma and Fisher (Pittsburgh, PA).
Cell culture and transfection
Primary chicken embryo fibroblasts (CEF) and human foreskin fibroblasts (HFF) were prepared and cultured as described previously (Mingle et al., 2005). For immunofluorescence staining (IF) or fluorescence in situ hybridization (FISH), cells were plated on glass cover slips coated with 0.5% gelatin for ~50–70% confluence 24 hours later. For transfection, cells on each cover slip were incubated with 0.3 μg DNA of each construct and Lipofectamine-PLUS (Invitrogen, Eugene, OR) for 2 hours before the medium was changed, followed by 22 hours of incubation. Cells were then fixed and subjected to FISH combined with IF for the detection of mRNA and proteins.
Construction of expression plasmids
Standard molecular biology techniques were used in cloning and plasmid construction. Accession numbers for the cDNA sequences used in this study are: AB025226 (chicken Dia1), AF051782 (human Dia1) and L08165 (chicken β-actin). Unless otherwise indicated, all the expression plasmids were constructed using the pNE plasmid backbone, which is under the control of a chicken β-actin promoter (courtesy of Stefan Kindler, Hamburg, Germany). To detect exogenous Dia1 protein and mRNA, we replaced the GFP in the pNE plasmid with a cassette that contains an HA-tag at the end of the coding region for protein detection and a fragment of LacZ in the 3′-UTR for mRNA detection. Point mutations were generated using Quikchange II XL Site-Directed Mutagenesis Kit (Stratagene, Santa Clara, CA). An improved version of the TC tag (Martin et al., 2005) was added to the N-terminus of Dia1 and cloned into the RSV-β plasmid (Kislauskis et al., 1993). For the chicken β-actin expression constructs, an internal ribosome entry site (IRES) was cloned into the pNE plasmid backbone. Red fluorescence protein mCherry alone or fused with chicken Dia1 were inserted upstream of the IRES, whereas chicken β-actin (full-length CR and 3′-UTR with an HA-tag at the N-terminus) was inserted at the 3′ end downstream. All the resulting expression plasmids were verified by DNA sequencing.
Pharmacological drug treatment
CEF transfected with a full-length cDia1 construct for 22 hours were treated with either vehicle control (methanol or DMSO), 0.4 μM cytochalasin D (CD), 5 μM colchicine (Col) or 5 μg/ml brefeldin A (BFA) for 90 minutes, or 10 μg/ml cycloheximide or 10 μg/ml puromycin for 60 minutes before being fixed for FISH and IF.
Nucleotides 62–1470 of chicken Dia1, 46–1699 of human Dia1 and 2159–2188 of LacZ (accession X65355) were cloned into pGEM-T Easy plasmids (Promega). These plasmids were linearized and transcribed in vitro in the presence of DIG- or fluorescein-labeled dUTP for RNA probes using a Maxiscript Transcription Kit (Ambion). Sense probes were similarly prepared and used for specificity control.
FISH for double mRNAs and FISH-IF for mRNA/protein detection
Sequential FISH was used to detect multiple mRNAs in the same cells as previously described (Mingle et al., 2005). Briefly, two RNA probes (DIG-labeled for Dia1 and fluorescein-labeled for Arp2) were hybridized to the same population of cells overnight then washed extensively. Sheep anti-DIG antibody (peroxidase-conjugated) was used and the fluorescence signal was amplified with TSA (using tetramethylrhodamine-tyramide). The antibody was washed-off and residual peroxidase activity was quenched (Liu et al., 2006), followed by PBS washes. Sheep anti-fluorescein antibody (peroxidase-conjugated) was then used to bind the second probe and the fluorescence signal was amplified with fluorescein-tyramide as the substrate. For the detection of mRNA and protein in the same cells, after FISH, the samples were treated with HCl then washed with PBS as described above, followed by normal IF procedures.
Sequential detergent extraction
This method was adopted from Stephens and Nicchitta (Stephens and Nicchitta, 2007) with modifications. To determine the optimal dose of saponin (a pure form of digitonin) for dissolving the plasma membrane, with minimal effect on the Golgi and ER membranes, we tested a series of saponin concentrations. Briefly, HFF cultured on glass cover slips were washed with cold PBS twice. Cold permeabilization buffer [110 mM KOAc, 25 mM K-HEPES, 2.5 mM MgCl2, K-EGTA, 1 mM dithiothreitol (DTT) and 1 mM phenylmethylsulfonyl fluoride (PMSF), pH 7.2] with different concentrations of saponin was added for incubation on ice for 5 minutes with gentle agitation. The permeabilization buffer was aspired and cold PBS was added for wash (twice). The cells were then fixed and processed for IF. GAPDH and calnexin were detected as markers for cytosolic proteins and ER-bound proteins, respectively. Based on the results of the titration (supplementary material Fig. S2), we chose 0.03% saponin to selectively permeabilize the plasma membrane. Briefly, cells were cultured in T25 flasks until 70–80% confluent and then briefly washed with cold PBS. For each flask, 0.4 ml of permeabilization buffer (with 40 U/ml RNasin and 0.03% saponin) was added to cover the cells, which were then incubated on ice for 5 minutes with gentle agitation. The solution was collected as a cytosolic fraction and the cells were washed once with cold permeabilization buffer containing 0.005% of saponin. The cells were further extracted with 0.4 ml of lysis buffer (400 mM KOAc, 25 mM K-HEPES, 15 mM MgCl2, 1% NP-40, 0.5% sodium deoxycholate, 1 mM DTT, 1 mM PMSF and 40 U/ml RNasin) for 5 minutes on ice. The solution was collected as the membrane-bound fraction. Both the cytosolic and membrane-bound fractions were clarified by centrifugation at 7500 g for 10 minutes at 4°C. Each supernatant was split equally into two portions, one for western blotting and another for RNA extraction using Trizol reagent (Invitrogen).
Iodixanol density gradient fractionation
HFF cultured in a 100 mm dish were washed once with cold PBS. One millilitre of homogenization medium (HM) (0.25 M sucrose, 1 mM EDTA, 10 mM Na-HEPES, pH 7.4, 100 U/ml Superasin RNase inhibitor, 0.5% protease inhibitor cocktail and 1 mM PMSF) was added and the cells were scraped off and collected into a 1.5 ml vial. The cells were homogenized by 10 passages through a 25-gauge needle then centrifuged at 3000 g for 10 minutes at 4°C. The supernatant was transferred onto the top of a 9 ml preformed 2.5–30% iodixanol gradient in HM. Iodixanol was used because it is a better medium than sucrose for fractionation (Graham, 2002). The samples were centrifuged at 402,000 rpm at 4°C for 3 hours with a Beckman SW 41Ti rotor. After centrifugation, fractions (0.7 ml each) were collected and each was split for protein extraction and RNA extraction using TCA precipitation and Trizol, respectively.
SDS-PAGE and western blotting
Standard processes of SDS-PAGE and western blotting were used. HRP-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) and SuperSignal enhanced chemiluminescence substrate (Pierce, Rockford, IL) were used for detection. The chemiluminescent signal was detected with films then scanned with the Bio-Rad gel documentation system (Bio-Rad, Hercules, CA) or directly imaged with the documentation system.
RT-PCR for mRNA in the fractions
RNA prepared from gradient fractions was reverse-transcribed using the Superscript III First-Strand Synthesis system from Invitrogen with an oligo(dT) primer. Resulting cDNA was then used in a PCR reaction with a primer pair of human Dia1 for 40 cycles. For the samples of sequential detergent extraction, the RNA samples were analyzed using the One-Step RT-PCR Kit (Invitrogen) over 25 cycles. The primer pairs used were: human GAPDH, 5′-AAGGTGAAGGTCGGAGTC-3′ and 5′-CGTCAAAGGTGGAGGAGT-3′; human calnexin, 5′-CTCCCAAGGTTACTTACA-3′ and 5′-TTCATCCCAGTCATCT-3′. The amplified products were then run on agarose gels, documented and analyzed with the Bio-Rad gel documentation system.
Samples were viewed and imaged with a BX60 Olympus microscope using a UPlanFl 60× oil objective (NA 1.25), a cooled charge-coupled device (CCD) camera (SensiCam; CooKe, Auburn Mills, MI) and Slidebook software (version 3.0, Intelligent Imaging Innovation, Denver, CO) or with a BX61 Olympus microscope using a UPlanApo 40× oil objective (NA 1.0), a cooled CCD camera (SensiCam) and IPLab software (version 3.6.5, Scanalytics Inc., Fairfax, VA). Additional image processing was performed using Adobe Photoshop (version 7.0, Adobe Systems, Mountain View, CA) and ImageJ (NIH).
Quantification of mRNA localization
The cell protrusion localization of the interested mRNAs was scored as described previously (Mingle et al., 2005). Two additional categories (perinuclear and uniform) of mRNA localization were used in this study: Briefly, cells were scored visually according to the following criteria: a cell with ≥80% of the total mRNA signal in the perinuclear region was scored as perinuclearly localized; a cell with >50% but <80% of the total mRNA signal or with uniformly distributed mRNA signal was scored as uniform; and a cell with >60% of the total mRNA signal at the leading protrusions or >70% of the total mRNA signal at both leading and trailing edges combined was scored as protrusion-localized. All the samples were scored blindly. At least 300–500 cells from three independent experiments were scored for each mRNA. Statistical analysis was performed using the Student's t-test.
Quantification of colocalization of Dia1 mRNA and ER markers
Although Dia1 mRNA is not expected to bind directly to ER markers and the fluorescence abundance of mRNA and an ER marker might not necessarily correlate, we do expect that the fluorescence signal of ER-localized Dia1 mRNA colocalizes with positive signal of the marker protein. On the basis of this concept and taking advantage of the punctate appearance of the Dia1 mRNA fluorescence signal, we chose to analyze the localization of Dia1 mRNA on the ER by determining the percentage of Dia1 mRNA particles that are colocalized with positive ER marker signal in a cell. Briefly, HFF were cultured, fixed and processed for the detection of Dia1 mRNA and calnexin protein as described above. Primary antibody for calnexin was omitted in control samples for fluorescence background. All the samples were imaged with the same exposure time (800 ms for the calnexin protein). Using the IPLab software, Dia1 mRNA particles (red) were selected by fragmentation and the corresponding green fluorescence pixel intensity (for calnexin protein) was quantified. To determine positive green fluorescence, two values of background green fluorescence pixel intensity were obtained from multiple spots around the nucleus (high fluorescence background) and cell periphery (low fluorescence background) in the control cells, respectively. These two values were used as the high-stringent and low-stringent cut-off to sort the mRNA particles in each cell based on their corresponding green fluorescence pixel intensity. The results were presented as the percentage of Dia1 mRNA particles with a green fluorescence intensity level above the cut-off in a cell.
Using biarsenical dyes and a tetracysteine tag (TC tag) to detect newly translated protein (for a translation site)
Membrane-permeable biarsenical dyes bind to a tetracysteine tag and then become fluorescent (Griffin et al., 2000; Griffin et al., 1998). Because the TC tag does not require folding for detection, an N-terminally tagged nascent peptide can be detected even before the full-length protein is synthesized. This advantage was taken to detect protein translation sites in living cells (Rodriguez et al., 2006). Using an improved TC motif (Martin et al., 2005; Rodriguez et al., 2006), we expressed cDia1 with a TC tag at the N-terminus to detect its mRNA translation site. Briefly, transfected CEF were incubated with green biarsenical dye FlAsH (final 1 μM) for 1 hour in MEM medium with 0.2% serum, followed by washes with 0.25 mM of 2,3-dimercaptopropanol (BAL) for a total of 30 minutes. The cells were then incubated in MEM with 10% FBS for 3 hours followed by cycloheximide (5 μg/ml) treatment for 1 hour. In the presence of cycloheximide, the cells were similarly stained with the red biarsenical dye ReAsH (final 1 μM) for 1 hour, washed with 0.25 mM BAL, then fixed for imaging. Green and red fluorescence indicates TC-tagged old and new proteins, respectively.
Fluorescence resonance energy transfer (FRET)
RhoA activity in CEF was similarly detected using a RhoA biosensor with FRET as previously described (Mingle et al., 2009), except that a wide-field fluorescence microscope was used here. The RhoA biosensor (pRaichu-1237x) was a gift from Michiyaki Matsuda (Osaka, Japan) (Yoshizaki et al., 2003). Transfected cells were fixed and processed using IF for histone H3 (nuclear marker). We used IF with a Cy5-conjugated secondary antibody instead of DAPI staining for the nucleus to avoid potential interference of DAPI with the CFP owing to their significant spectral overlap. Images were acquired using an Olympus IX70 microscope with independent excitation and emission filter wheels. Transfected cells were first identified using the YFP/YFP channel. With a multiple-channel beam splitter (dichroic mirror) at a fixed position, a CFPex/CFPem image was acquired, then immediately a CFPex/YFPem image (FRET) was obtained by rapidly switching to the YFP emission filter in the emission filter wheel with the same exposure time (usually 300 ms). An image from the Cy5 channel was also obtained for the nucleus. The acquired images were processed for FRET using our custom-written scripts for the IPLab software. Because the Raichu RhoA biosensor has both CFP and YFP in the same molecule, correction of spectral bleed-through is not necessary and we therefore used the FRET/CFP ratio directly for the relative activity of RhoA in the cells. To reduce the background noise outside the transfected cells, the areas of transfected cells were segmented, and then the images were processed with the binary function by designating the areas outside of the transfected cells as 0. The ratio of FRET/CFP intensity was calculated and presented as relative RhoA activity in the cells. Pseudo-color and the corresponding scale bar were created using a custom Color Look-Up Table in IPLab. The position of the nucleus was first traced on the Cy5 image then superimposed on the corresponding FRET image.
We thank Nataly Okuhama, Huijun Yao and Qingfen Li for technical support, Stefan Kindler for the pNE plasmid, Michiyaki Matsuda for the RhoA biosensor, Ceshi Chen for the pRSET-B mCherry, Lisa Mingle for suggestions on the project and Livingston Van De Water for critical reading of this manuscript. This work is supported by the NIH grant R01GM070560. Deposited in PMC for release after 12 months.