Genome-wide analysis demonstrates conserved localization of messenger RNAs to mitotic microtubules

Localization of mRNAs within cells functions to spatially restrict gene expression, which is critical for cell fate determination and proper body patterning during embryogenesis (St Johnston, 2005). Localized mRNA translation is also important for many processes in differentiated somatic cells (Kislauskis et al., 1997; Adereth et al., 2005; Huttelmaier et al., 2005) and is used extensively in neuronal cells to restrict the expression of specific proteins to synapses (Kiebler and Bassell, 2006). In Xenopus laevis and Drosophila melanogaster, mRNAs coding for proteins involved in mitotic progression are enriched on the mitotic spindle (Raff et al., 1990; Groisman et al., 2000). Furthermore, cytological and biochemical studies of X. laevis egg extract spindles and sea urchin microtubules have shown that ribosomes are tightly associated with mitotic microtubules, suggesting that localized translation might be a key regulator of mitosis (Suprenant, 1993; Liska et al., 2004; Mitchison et al., 2004). However, the extent of mRNA targeting to specific subcellular structures, such as the spindle, has not been comprehensively investigated and its underlying mechanisms and functions are poorly understood.

To gain insight into the role of localized mRNAs during mitosis, we comprehensively identified mRNAs that localize to microtubules in meiotic X. laevis egg extracts and mitotic human cell extracts using Affymetrix microarrays. We found that a specific and conserved group of mRNAs, termed microtubule-bound mRNA (MT-mRNA), is enriched on microtubules. We observed that only a subset of MT-mRNAs was associated with microtubule-bound polyribosomes, indicating that mitotic spindles contain both translationally active and inactive mRNAs. Furthermore, we found that active translation is not required for targeting of endogenous or exogenous mRNAs to mitotic microtubules. We propose that localization of specific mRNAs to the spindle is a conserved mechanism for enhancing protein localization and for segregation of translationally inactive mRNAs.

To address whether mRNA localization contributes to cell division, we sought to identify mRNAs that localize to microtubules using extracts prepared from X. laevis eggs, which are naturally arrested in metaphase of meiosis II by cytostatic factor (CSF) activity. To isolate mRNAs that are targeted to the spindle apparatus, we prepared a small-scale cDNA library from MT-RNAs from Xenopus tropicalis egg extracts (Brown et al., 2007) and identified ∼50 mRNAs and 50 ribosomal RNAs (rRNAs) by sequence analysis. X. tropicalis was used because its sequenced genome facilitated comprehensive identification of cloned RNAs. To test whether the mRNAs identified were enriched on meiotic microtubules, we evaluated a subset of randomly chosen MT-mRNAs from the library by comparing their levels in total extract to the amount in the microtubule-bound fraction using semiquantitative RT-PCR (Fig. S1). 8 out of 10 mRNAs tested were enriched on microtubules compared with total extract, suggesting that specific mRNAs are targeted to microtubules during mitosis.

To comprehensively identify mRNAs that associate with meiotic microtubules, we took advantage of the available Affymetrix X. laevis microarrays. We purified taxol-stabilized microtubules from metaphase-arrested X. laevis egg extracts by centrifugation through a glycerol cushion and observed that ∼2–5% of the total mass of the RNA in the extract cofractionated with microtubules (unpublished data). Copurification of RNA with microtubules was specific to pelleted microtubules, as no RNA pelleted without inducing microtubule polymerization (Blower et al., 2005; unpublished data), and similar amounts of RNA cofractionated with microtubules regardless of how microtubule polymerization was induced (taxol, DMSO, RanQ69L, or sperm DNA; unpublished data). We then prepared cRNA from microtubule-associated RNAs and total-extract RNAs, hybridized the cRNA to Affymetrix microarrays, and compared the relative abundance of mRNAs on microtubules to total extract. Extract mRNAs exhibited a continuum of binding to microtubules, with ∼5% of mRNAs >1.5-fold enriched on microtubules compared with total extract (Fig. 1 A). We found the cyclin B mRNA, which was previously reported to localize to the mitotic spindle (Groisman et al., 2000), several developmentally regulated mRNAs that are targeted to the animal or vegetal cortex of the X. laevis egg (Kloc et al., 2002; Table S2), and several mRNAs enriched on microtubules that were also identified in our small-scale library (e.g., centromere-associated protein E [CENP-E] and dynein). The known MT-mRNA cyclin B was ∼1.5-fold enriched on microtubules, suggesting that even low levels of enrichment on microtubules might be biologically relevant. These findings demonstrate that our genome-wide approach accurately identifies microtubule-associated mRNAs.

Analysis of a subset of the enriched MT-mRNAs using the Affymetrix Netaffx analysis package demonstrated that mRNAs annotated to function in mitosis, DNA metabolism (replication, repair, and topology), and germ cell and body axis determination were overrepresented on mitotic microtubules (Table S5). Furthermore, examination of the 3′ UTRs of MT-mRNAs revealed that the cytoplasmic polyadenylation element (CPE) was present in 7% of MT-mRNAs, which represents a 4.1-fold enrichment compared with all mRNAs in the database. These results are consistent with and extend previous findings that ribosomes and specific mRNAs are tightly associated with polymerized microtubules in early sea urchin embryos (Suprenant et al., 1989; Suprenant, 1993; Hamill et al., 1994) and they demonstrate that particular classes of mRNAs are targeted to microtubules during cell division.

To determine if the MT-mRNAs identified by microarray analysis localize to mitotic spindles, we transcribed and fluorescently labeled four different MT-mRNAs (Xpat, Xrhamm, Smc3, and xDia) and two control mRNAs (net1 and mCherry) in vitro, added labeled mRNAs to preformed meiotic spindles in X. laevis egg extracts, and assayed the localization of the added mRNA 15 min later. All of the MT-mRNAs tested localized to a high proportion of the spindles, whereas the controls did not (Fig. 1, B and C), demonstrating that our microarray analysis accurately identified mRNAs that can localize to the spindle.

We next asked whether localization of specific mRNAs to microtubules during mitosis is conserved among different cell types and species. Synchronized HeLa cells were used to prepare mitotic extracts that assemble radial microtubule asters when treated with taxol, reproducing many aspects of mitotic microtubule organization (Gaglio et al., 1995; Mack and Compton, 2001). RNA localized in a granular pattern on taxol microtubule asters in HeLa extracts, and both rRNA and mRNA specifically copurified with microtubules (Fig. 2, A and B). To identify MT-mRNAs in mitotic HeLa extracts, we hybridized MT-RNA– derived cRNAs to Affymetrix microarrays and compared the abundance of mRNAs in the microtubule sample to the total extract. Similar to our results with X. laevis, we observed that there was a continuum of mRNA binding to microtubules, with ∼10% of all mRNAs enriched 1.5-fold or more on mitotic microtubules (Fig. 2 C), and that mRNAs encoding proteins involved in various aspects of mitosis and DNA metabolism (replication, repair, and topology) were overrepresented in the MT-mRNA fraction (Table S5). The CPE was also overrepresented in the UTRs of MT-mRNAs with 5.7% of them possessing this element, which is a 4.1-fold enrichment compared with all mRNAs (Table S3).

X. laevis oocytes contain both polyadenylated and deadenylated mRNAs (Mendez and Richter, 2001), and lengthening of the poly-A tail is a major mechanism of translational control during early development and is also important in cell cycle control. The cyclin B mRNA contains a CPE that regulates its cytoplasmic polyadenylation via the CPEB protein in developing X. laevis embryos (Groisman et al., 2002). In addition, the CPEB protein targets the cyclin B mRNA to microtubules, which leads to an enrichment of the cyclin B protein on the mitotic spindle (Groisman et al., 2000). However, most MT-mRNAs do not contain a CPE, suggesting that there are additional pathways that mediate mRNA targeting to the mitotic spindle.

The targeting to microtubules of mRNAs involved in mitosis and DNA metabolism suggested that spindle mRNAs might regulate cell cycle events through localized translation. To directly test whether local translation is occurring on the X. laevis meiotic spindle, we took advantage of a series of recently developed analogues of the protein synthesis inhibitor puromycin that label sites of active mRNA translation in cultured cells (Starck et al., 2004). We added fluorescent puromycin derivatives to X. laevis egg extract reactions containing replicated sperm chromosomes as they were induced to enter metaphase and fixed the spindles at various time points. Fluorescent puromycin formed spots all along spindle microtubules and concentrated near the spindle poles (Fig. 3 A). To confirm that the puromycin spots were sites of active protein synthesis (and not truncated nascent peptides dissociated from ribosomes), we costained spindles with antibodies against the ribosomal protein S6. We found extensive colocalization between the fluorescent puromycin and ribosomal protein S6 (Fig. 3 A), which confirms that local protein synthesis occurs on the meiotic spindle in X. laevis egg extracts and is consistent with previous electron microscopy data showing a tight association of ribosomes with spindle microtubules in X. laevis egg extracts (Liska et al., 2004; Mitchison et al., 2004). To test whether newly synthesized proteins were incorporated into meiotic spindles, we added biotinylated lysine tRNA to extract reactions, purified spindles through a glycerol cushion, separated the proteins by gel electrophoresis, and tested for incorporation of biotin into newly synthesized proteins using a streptavidin blot. We found that biotinylated lysine was incorporated into many spindle-associated proteins, demonstrating that mitotic translation contributes to the protein content of the mitotic apparatus (Fig. 3 B).

To determine which MT-mRNAs were locally translated on microtubules during mitosis, we purified mRNAs that associated with microtubule-bound polyribosomes. Both rRNA and ribosomal proteins pelleted from microtubule preparations in the presence of cycloheximide and Mg²⁺, conditions that stabilize polysomes, but not in the presence of EDTA or puromycin, which dissociate polyribosomes (Fig. 4, A–D), confirming that our protocol efficiently isolated microtubule-bound polysomes. To identify the MT-mRNAs that are translated during mitosis, we hybridized the microtubule polysomal mRNAs to Affymetrix microarrays. The levels of MT-mRNAs associated with polyribosomes varied considerably, suggesting that some MT-mRNAs are translated locally on the spindle, whereas others are translationally inactive (Fig. 4 E). Many mRNAs highly represented on microtubule-bound polyribosomes were only marginally enriched on microtubules (e.g., many of the mitotic cyclins; Fig. 4 E, red box), indicating that these mRNAs might not be highly enriched on microtubules but their translation may be spatially restricted to microtubules. Several mRNAs annotated to have a role in development or body patterning that were enriched on microtubules were not highly represented on MT polysomes. This may indicate that these mRNAs are on the spindle as passive cargo, potentially in a form similar to a P-granule, raising the possibility that targeting of inactive mRNAs to the mitotic spindle may serve as a mechanism for their segregation during cell division.

The observation that many MT-mRNAs are associated with polyribosomes suggests that active translation might be required for targeting some mRNAs to microtubules. In this scenario, as mRNAs are translated into protein the short peptides produced would interact with components of the mitotic apparatus, such as chromosomes or microtubules, thereby localizing polysomes to the spindle. An alternative hypothesis is that mRNAs are actively transported to microtubules in a translation-independent manner, where they can interact with spindle-bound ribosomes and undergo translation. To distinguish between these two possible targeting mechanisms, we examined the effects of the translation inhibitor puromycin on mRNA localization to mitotic microtubules. Because puromycin is a tRNA analogue that blocks peptide bond formation, causing premature chain termination and mRNA release from the ribosome (Azzam and Algranati, 1973), it should disrupt mRNA targeting to microtubules if translation is required. Addition of puromycin to X. laevis egg extracts had no effect on the copurification of two endogenous actively translating mRNAs with microtubules (Fig. 5 B). To determine if ribosomes associate with microtubules in a translation-dependent manner, we examined the proportion of spindle-associated ribosomal protein that copurified with polysomes. We found that ∼10–15% of microtubule-bound ribosomal protein copurified with polysomes, which is comparable to the amount of ribosomal protein that copurifies with polysomes in total extract (Fig. 5 C), demonstrating that microtubules bind both translationally active and inactive ribosomes. These data demonstrate that translation is not required for the localization of endogenous mRNAs or ribosomes to mitotic microtubules.

To determine if translation inhibition affected the spindle localization of exogenously added mRNAs, we inhibited mitotic translation for 45 min with puromycin, and then added Alexa 488–labeled mRNAs and assayed localization after 15 min. We analyzed mRNAs from three classes: cyclin B1, which is tightly associated with microtubule-bound polysomes and whose protein product is known to associate with the mitotic apparatus; Xpat, which is highly represented on microtubule-bound polysomes but whose protein product is not known to associate with the mitotic apparatus; and Xdia, which is enriched on microtubules but is not significantly represented on polysomes and whose protein product is not known to interact with the mitotic apparatus. Translation inhibition did not impair the localization of any of these three mRNAs to the mitotic spindle, demonstrating that translation is not a mechanism of their localization to mitotic microtubules (Fig. 5 D). Translation inhibition caused subtle morphological changes in localization pattern of the mRNA on the spindle, suggesting that although translation does not regulate spindle mRNA targeting, it might affect localization within subdomains of the spindle.

In conclusion, we have found that a specific subset of mRNAs is enriched on mitotic microtubules in both X. laevis and humans (Table S4). Mechanisms of mRNA localization to subcellular structures are thought to occur through the use of zipcode sequences in the 3′ UTR (Kislauskis and Singer, 1992), which are recognized by trans-acting factors that mediate intracellular targeting. There are a few well-documented cases of mRNA localization through a zipcode, but it remains unclear how many mRNAs find their subcellular locations. Our identification of a significant fraction of mRNAs in both X. laevis and human mitotic extracts as microtubule targeted during mitosis suggests that this is a conserved and widely used mechanism for enhancing protein localization to regulate mitotic events and delivering translationally inactive mRNAs to daughter cells. We speculate that asymmetric targeting of developmentally regulated mRNAs to one spindle pole could be a mechanism by which organismal asymmetry is established early in X. laevis development, and it is worth noting that asymmetric localization of specific mRNAs to one mitotic centrosome during early mollusc development has been reported (Lambert and Nagy, 2002). The next challenge is to identify the cis-acting sequences and trans-acting factors that mediate mRNA localization to microtubules and to understand how translation of specific MT-mRNAs is controlled in space and time, both within the context of the mitotic cell cycle and during development.

X. laevis egg extracts were prepared exactly as previously described (Desai et al., 1999). Microtubule polymerization was induced in 100 μl CSF extract by the addition of taxol to 10 μM. Microtubules were allowed to polymerize for 15 min at 20°C. Extract was then diluted with 1 ml BRB80 (80 mM Pipes, pH 6.8, 1 mM MgCl₂, and 1 mM EGTA) + 30% glycerol + 0.1% Triton X-100. Diluted extract was layered over a 10-ml cushion of BRB80 + 60% glycerol. Samples were then centrifuged for 10 min at 17,000 g. The supernatant was aspirated and the interface was washed twice with water. The microtubule pellet was dissolved in RLT solution (RNeasy kit; QIAGEN) and the RNA was purified using the RNeasy kit. The typical yield of microtubule-associated RNA was ∼5% of the total RNA present in the extract. Total RNA was purified from untreated CSF extract using the RNeasy kit.

Mitotic HeLa S100 was prepared from HeLa cells that were synchronized using 2 mM thymidine for 24 h, followed by release into unsupplemented media for 6 h and a 14-h treatment with 10 μM nocodazole, essentially as previously described (Gaglio et al., 1995). Microtubule polymerization was induced by the addition of taxol to the extract to 100 μM. Extracts were incubated at 33°C for 45 min. Extracts were diluted and centrifuged as described for X. laevis extracts. RNA was purified from the microtubule pellet or total extract using the RNeasy kit.

For cytological analysis of HeLa mitotic extract asters, asters were formed as described and fixed in solution with 1 ml BRB80 + 0.1% glutaraldehyde for 5 min at 33°C. Samples were then centrifuged onto coverslips as described for X. laevis egg extract reactions and fixed to the coverslip by a brief treatment with −20°C methanol. Asters were then stained with SYTOX RNA select (Invitrogen) as described by the manufacturer.

Purification of polysome RNA was performed according to a modification of Fritz and Sheets (2001). Cycloheximide or puromycin was added to the extract at a concentration of 0.1 mg/ml and incubated for 45 min, and then microtubules were polymerized and isolated as described in the preceding paragraph. The resulting microtubule pellets were resuspended in 2.5 ml PB (20 mM Tris, pH 7.5, 300 mM KCl, 0.5% NP-40, and 2 mM MgCl₂ [where indicated]) supplemented with cycloheximide or puromycin and incubated on ice for 15 min. The solubilized microtubule pellet was then layered over a 2.5-ml cushion of PB + 25% sucrose ± cycloheximide or puromycin and centrifuged at 200,000 g for 2 h in a rotor (SW55Ti; Beckman Coulter). Polysome-associated RNA was purified from the pellet using the RNeasy kit and used for microarray analysis or RT-PCR. For protein analysis, polysome pellets were dissolved in 1× SDS-PAGE loading buffer, separated on a 12% gel, and either silver stained or blotted for the ribosomal proteins S6 or L7a (Cell Signaling Technology). For RT-PCR analysis of pelleting of MT-mRNAs, RNA isolated from MT or polysome pellets was reverse transcribed using an oligo-dT primer and SuperscriptIII reverse transcriptase. A dilution series (1:10, 1:100, and 1:1,000) of the resulting product was used as input for 25 cycles of PCR.

5 μg each of microtubule-associated RNA or total extract RNA was used as a template for the one-cycle cRNA synthesis kit (Affymetrix). cRNA was hybridized to X. laevis or human U133 Plus2 arrays. For each experiment, RNA was prepared from two to three different extract samples.

Microarrays were normalized using the Affymetrix RT controls and all data points with a Log₂ <5 were eliminated. The ratio of the signal of a given message on microtubules to total extract was compared using the Log₂ of the normalized data. Statistical significance of differences between data was compared using a two-tailed Student's t test. The correlation between two trials of polysome-associated mRNAs from two different extracts was 0.91, indicating good reproducibility.

All microarray data was deposited in the ArrayExpress database (www.ebi.ac.uk/arrayexpress/) and is available under accession no. E-MEXP-1320.

mRNAs that were 1.5-fold enriched on microtubules with a p-value <0.05 were analyzed manually for the presence of known RNA sequence elements using the UTR database (www.ba.itb.cnr.it/srs7bin/cgi-bin/wgetz?-page+top).

mRNAs enriched as described by the criteria in the preceding paragraph were used as input to Netaffx analysis software package (Affymetrix) to generate gene ontology (GO) graphs. The GO graphing package calculates whether a particular functional class of genes is overrepresented in the designated subset of genes as compared with the representation of this class of genes in all genes with GO annotations. P-values of <0.05 are statistically overrepresented.

1,000 sperm nuclei per microliter were added to CSF-arrested egg extracts that were cycled through interphase by the addition of calcium to 0.6 mM and by a 90-min incubation at 21°C. Extracts were driven back into mitosis by the addition of an equal volume of CSF-arrested extract containing either a buffer control or 100 μg/ml puromycin. For the addition of fluorescent puromycin, F2PMe (provided by S. Starck and R. Roberts, California Institute of Technology, Pasadena, CA) was added to extracts at a concentration of 10 μM.

All cDNA clones were obtained from Open Biosystems (clone numbers are listed in parentheses). Full-length clones of XRHAMM (3402708), XDia (4684250), Xpat (5155227), Smc3 (6865177), cyclin B1 (4683804), and mCherry were amplified by PCR and cloned into pCR2.1 downstream of the T7 promoter, and plasmids were verified by sequencing. Plasmids were linearized by KpnI digestion and used to produce Alexa 488– labeled RNA (0.5 μl of 1 mM Alexa 488 UTP [Invitrogen] was included in each 10-μl reaction) using the mMessage Machine T7 in vitro transcription kit (Ambion). The same kit was used to produce RNA encoding net1 (8949944; an mRNA that was dramatically underrepresented on microtubules) directly from cDNA clones. RNAs were precipitated using lithium chloride and resuspended in RNase-free water. Labeled RNA was stored at −80°C until use.

To assess localization of RNAs in X. laevis extract spindles, spindles were formed around replicated sperm DNA and 0.02 pmol of labeled RNA was added to the preformed spindles. In extracts where translation had been inhibited, puromycin was added to 100 μg/ml as extracts were cycled into mitosis, and the extract was incubated in the presence of puromycin for 45 min before RNA localization was assessed. RNA was allowed to bind to spindles for 15 min and then structures were diluted into BRB80 + 4% formaldehyde, 0.1% glutaraldehyde, and 0.5% Triton X-100, incubated for 5 min at room temperature, and centrifuged onto coverslips and fixed in −20°C methanol for 5 s.

To quantify RNA localization to spindles, localization of each RNA to spindles was counted for 100 structures by visual assessment. This was repeated for three independent extract experiments.

Images were acquired using a microscope (BX61; Olympus) with 60×/1.42NA or 40×/1.2NA objectives and a charge-coupled device camera (ORCA; Hamamatsu) controlled by Slidebook software (3i). All coverslips were mounted in Vectashield containing 1 μg/ml DAPI. Tubulin was visualized by including rodamine-labeled tubulin in all extract reactions at a concentration of 200 nM. All images were acquired at room temperature (∼21°C). Images in Figs. 1–5,²³⁴⁵ were acquired as stacks of images spaced 0.25 μm in the z axis. Images were deconvolved using a constrained iterative algorithm using Slidebook.

Transcend biotinylated lysine tRNA (Promega) was added to the extract at a 1:10 or 1:20 dilution as the extract entered mitosis. Spindles were pelleted through a 40% glycerol cushion, resuspended in 1× SDS sample buffer, and analyzed for incorporation of biotin by Western blotting with streptavidin HRP (Jackson ImmunoResearch Laboratories).

Fig. S1 shows that various MT-mRNAs are enriched on meiotic microtubules in X. tropicalis extracts. Table S1 contains all the microarray data for mRNAs enriched on mitotic microtubules and microtubule-associated polyribosomes in X. laevis. Table S2 contains the relative microtubule enrichment for a group of mRNAs that are localized during X. laevis development. Table S3 contains all the microarray data for mRNAs enriched on mitotic microtubules in HeLa extracts. Table S4 contains a list of genes conserved between X. laevis and HeLa cells, and their relative enrichment on microtubules. Table S5 contains a list of all the classes of genes that are overrepresented on mitotic microtubules in both X. laevis and HeLa cells.

The authors would like to thank Vivian Peng of the University of California, Berkeley microarray core facility for microarray analysis, Shelly Starck and Richard Roberts for the gift of fluorescent puromycin analogues, and Judith Sharp and members of the Heald, Weis, and Blower laboratories for helpful discussions.

M.D. Blower is supported by the Damon Runyon Cancer Research Fund and Burroughs Wellcome Fund Career Award. R. Heald and K. Weis were supported by a grant from the National Institutes of Health (GM065232).