Evolutionary conserved role of eukaryotic translation factor eIF5A in the regulation of actin-nucleating formins

Elongation factor eIF5A is required for the translation of consecutive prolines, and was shown in yeast to translate polyproline-containing Bni1, an actin-nucleating formin required for polarized growth during mating. Here we show that Drosophila eIF5A can functionally replace yeast eIF5A and is required for actin-rich cable assembly during embryonic dorsal closure (DC). Furthermore, Diaphanous, the formin involved in actin dynamics during DC, is regulated by and mediates eIF5A effects. Finally, eIF5A controls cell migration and regulates Diaphanous levels also in mammalian cells. Our results uncover an evolutionary conserved role of eIF5A regulating cytoskeleton-dependent processes through translation of formins in eukaryotes.

mutants in yeast haploid cells were unable to grow a mating projection, termed shmoo, towards a pheromone gradient during mating 15 . Despite the functional connection between eIF5A and translation of polyPro-containing proteins in yeast, it is unclear whether eIF5A can regulate cytoskeletal remodeling through the translation of formins in other systems.
In this study, we identified a role of Drosophila eIF5A (Dm eIF5A) in dorsal closure (DC), the last major morphogenetic rearrangement during embryogenesis. DC involves the formation of an actomyosin contractile supracellular cable together with the migration, spreading and fusion of lateral epidermal cellular sheets 16,17 . We have found that Dm eIF5A is the functional homolog of Sc eIF5A and its depletion in Drosophila causes defects in actomyosin cable formation in the embryonic epidermis during DC. Our results also revealed that eIF5A regulates the protein levels of Diaphanous (Dia), and that this formin is required during DC downstream of eIF5A. Lastly, depletion of active eIF5A decreases the levels of Dia and migration in mouse neural stem cells (NSCs). Together, our results indicate an evolutionary conserved function of eIF5A in the translation of actin-nucleating formins with important consequences on the regulation of cytoskeleton-mediated processes from yeast to mammals.

Results and Discussion
The translational regulator eIF5A is required for actin cable assembly during Drosophila DC. During Drosophila embryogenesis, the dorsal epidermis appears discontinuous due to germ band retraction and is transiently covered by an extraembryonic tissue, the amnioserosa (AS). The hole is subsequently sealed by a complex process called DC, which involves the formation of an actomyosin contractile cable at the leading edges of the dorsal-most epidermal (DME) cells, the emission of cellular protrusions from opposite sides of the epidermis that interdigitate and zip the dorsal opening at its corners, and the apical constriction and apoptosis of AS cells that facilitates the occlusion of the hole. As a result, the two lateral epidermal sheets converge towards the dorsal midline while the AS reduces its surface area until it disappears inside the embryo 16,17 . The cabut (cbt) gene encodes a transcription factor with high similarity to the vertebrate TIEG (TGF-β-inducible early-response genes) proteins and is involved in several key events during DC including actin dynamics 18,19 . However, its direct target genes during this process are unknown. Using genome-wide ChIP-on-chip analysis for Cbt binding sites in extracts from 9-14 h embryos we identified eIF5A as a Cbt putative target gene (V.M.-S. and N.P., in preparation). Reduced eIF5A function affects cell growth and causes autophagy in several Drosophila tissues and increases polysome size in S2 cells 20 , suggesting a potential role in translation elongation. However, the role of the only Dm eIF5A gene in DC has never been investigated. To clarify this issue, we reduced the expression of Dm eIF5A specifically in the embryonic epidermis using a UAS-iReIF5A RNAi line and the arm-GAL4 or 69B-GAL4 epidermal drivers (Fig. 1A,B) which resulted in late embryonic lethality. Cuticle preparations revealed that most eIF5A mutant (eIF5A KD ) embryos displayed DC defects with holes in the dorsal and anterior epidermal regions (Fig. 1C,D). Phalloidin stainings were performed to examine the integrity of the actin cable at the leading edge of DME cells and revealed an irregular and discontinuous distribution of actin in eIF5A mutants (Fig. 1E,F). Defective attachments between AS and epidermal cells, as well as misalignments at the dorsal midline, were also observed (Fig. 1F). Taken together, these results indicated that Dm eIF5A plays an essential role in actomyosin organization during embryogenesis. Drosophila eIF5A is a functional homolog of yeast eIF5A by rescuing the translation of polyPro-containing formins in yeast TIF51A mutants. Human eIF5A can restore growth in yeast cells lacking their own eIF5A 21 . To investigate whether Dm eIF5A was also a functional homolog of Sc eIF5A, we cloned the Dm eIF5A cDNA and the ORF of the TIF51A yeast gene in yeast plasmids under the control of the constitutive TEF promoter and with the CYC1 terminator. We then transformed these plasmids, or an empty plasmid as control, in two temperature sensitive yeast mutants, tif51A-1 and tif51A-3, which contain one (P83S) or two (C39Y, G118D) amino acid substitutions, respectively, and are unable to grow at restrictive temperature (37 °C) 22,23 . Yeast mutants transformed with plasmids containing either Dm eIF5A or Sc eIF5A exhibited restored growth at the restrictive temperature ( Fig. 2A). With regards to the defect in pheromone-induced shmoo formation observed in eIF5A-depleted yeast haploid cells, tif51A-1 mutants containing empty, Dm eIF5A or Sc eIF5A plasmids showed reduced shmoo formation under permissive temperature, suggesting that the Tif51A-1 mutated version of yeast eIF5A plays a dominant role. However, upon the introduction of Sc eIF5A and Dm eIF5A shmoo formation was restored to full or almost full (77% for Dm eIF5A) wild-type levels at 37 °C (Fig. 2B). These data indicated that Dm eIF5A is a functional homolog of yeast eIF5A, allowing normal growth and pheromone-induced polarized growth in eIF5A-deficient yeast cells.
Shmoo formation during yeast mating requires Bni1 24 , a formin containing several stretches of consecutive prolines (Fig. 3A). Since we have recently shown that translation of polyPro stretches of Bni1 requires eIF5A 15 , we therefore investigated whether Dm eIF5A also is able to replace Sc eIF5A in the translation of a full-length genomic HA-tagged BNI1 (Fig. 3A). BNI1-HA levels were importantly reduced in tif51A-1 cells after 4 h at 37 °C 15 , and barely detectable after 6 h (Fig. 3B). However, no reduction in Bni1 protein was observed in tif51A-1 cells expressing either Sc eIF5A or Dm eIF5A (Fig. 3B). We measured BNI1-HA mRNA expression in the same cultures by RT-qPCR and calculated the protein/mRNA ratios as an indication of translation efficiency, which was compared to that of hexokinase 2, a protein with no polyPro motifs. Translation efficiency of BNI1 was dramatically reduced after the temperature shift to 37 °C in tif51A-1 cells carrying an empty plasmid whereas wild-type translation efficiencies were observed in Sc eIF5A or Dm eIF5A-transduced tif51A-1 cells (Fig. 3C). Therefore, fly eIF5A is able to allow the progression of yeast ribosomes through the polyPro encoding codons of BNI1, which otherwise stall ribosomes in the absence of endogenous eIF5A 8 . These results indicated that both Sc eIF5A and Dm eIF5A are interchangeable in its regulation of formin translation in yeast.  32 . Among them, only the diaphanous (dia) gene has been analyzed in DC (reviewed in ref. 32). Embryos lacking dia display cell misalignments at the dorsal midline, suggesting a role in coordinating actomyosin contractibility and cell adhesion during DC 33 . Actin cable assembly in dia mutant embryos wounded by laser ablation is also inhibited at wound edges 34 , further supporting a role for Dia in actin regulation. Reduced expression of formin3, Fhos, cappuccino, and frl in embryonic epidermis using the arm-GAL4 and 69B-GAL4 drivers and the corresponding RNAi lines resulted in fully viable embryos, suggesting that they were not required for DC (not shown). Similar results have been obtained in DAAM mutants (J. Mihaly, personal communication). Therefore, we tested whether Dia could be the translational target of Dm eIF5A during DC. Dia protein levels in arm-GAL4/UAS-iReiF5A or 69B-GAL4/ UAS-iReiF5A embryos were apparently lower than in control embryos (arm-GAL4/+ and 69B-GAL4/+) by immunoblot detection (data not shown). The differences were, however, very small, most likely because the strategy used deletes eIF5A in only a small fraction of cells. Therefore, we decided to analyze Dia expression at the single cell level in embryos in which eIF5A RNAi is driven in a mosaic pattern by the en-GAL4 line (Fig. 4A,B). Previous analyses showed that Dia has a dynamic cellular localization during DC, being cortical in elongating epidermal cells 33 . Accordingly we detected Dia by immunofluorescence in the contours of all epidermal cells in both wild-type and eIF5A mutant regions. Interestingly, quantifications by image analysis (see Materials and Methods) showed that Dia levels were reduced in eIF5A mutant cells compared to nearby wild-type cells (Fig. 4C), consistent with the possibility that Dia translation is regulated by eIF5A.
Since eIF5A depletion in the epidermis during DC likely reduce Dia translation, we studied whether an increase of Dia levels in that tissue would be able to suppress the embryonic lethality phenotype of eIF5A mutants. First, we generated a fly line harboring two constructs, UAS-iReIF5A and UAS-Dia, to simultaneously reduce eIF5A endogenous levels and increase Dia levels in a given tissue. This line was crossed with the 69B-GAL4 epidermal driver and the percentage of embryonic lethality in the progeny was determined. We observed reduced lethality in Dia-overexpressing eIF5A mutant embryos (77% dead embryos, n = 816) with respect to that without Dia (94% dead embryos, n = 1,412). Similarly, partial suppression was observed in yeast for the shmooing defects of eIF5A mutants when increasing Bni1 formin levels 15 , which could indicate that either higher levels of this formin would be necessary for a complete rescue of the phenotype or that additional proteins involved in DC might also be targets of eIF5A. In summary, phenotypic analysis and immunofluorescence visualization indicate that Dia is a functional target of Dm eIF5A during DC and suggest that defects in actomyosin cable assembly in eIF5A mutant embryos are likely the result of low Dia levels.
Protein levels of Dia and cell migration are regulated by eIF5A in primary mammalian cells. Mammalian Dia1 (mDia1) catalyzes actin polymerization and regulates microtubule dynamics, playing an important role in cell polarization, adhesion and migration downstream of GTPase RhoA 35,36 . A requirement for eIF5A in cytoskeletal-mediated processes such as cell migration has also been reported 37 , however a link between mDia and eIF5A has not been established. We found that NSCs isolated from the subependymal zone of the adult mouse brain and grown as neurospheres express detectable levels of eIF5A1 but not eIF5A2 (mRNA expression normalized to Gapdh, in mean arbitrary units ± SD × 10 −3 : Eif5a1, 132 ± 4; Eif5a2, 0.02 ± 0.09; n = 3); in addition, we also found detectable levels of mDia mRNA (45 ± 3; n = 3). Therefore, we decided to test whether mDia protein could be regulated by eIF5A in NSCs. Hypusination is a post-translational modification selectively found in eIF5A that is essential for its function 6 . A 3-day treatment of neurospheres with the cell-permeable inhibitor of deoxyhypusine synthase GC7 at 5 or 10 µM did not affect cell viability (not shown), but resulted in a 3-fold reduction in the levels of hypusinated eIF5A that could be completely reversed by washing the drug out for 3 additional days (Fig. 5A,B). Individual neurospheres of similar sizes that had been grown with or without 10 µM GC7 for 5 days were then plated onto Matrigel. Quantitation of cell dispersion from each neurosphere at different times after plating indicated that inhibition of hypusination reduces the extent of migration (Fig. 5C,D). More interestingly, mDia1 levels were decreased by inhibition of eIF5A hypusination and rescued after drug wash-out (Fig. 5E,F) without changes in mRNA expression (mean fold-change relative to control condition ± SD: 1.00 ± 0.04 after 10 µM GC7 and 1.03 ± 0.09 after wash-out, n = 3).
Our results together indicate that eIF5A is required for Dia translation in both vertebrate and invertebrate cells. We demonstrate that structural and sequence conservation of eIF5A across evolution correlates with its conserved role in cytoskeletal dynamics through the regulation of formin translation. Actin nucleation is a tightly regulated process 1 , therefore the unique modulation of eIF5A activity by polyamine-dependent hypusination opens exciting possibilities for the pharmacological targeting of cytoskeletal remodeling in different biological systems.

Cuticle preparations and lethality assays in Drosophila embryos. For cuticle analyses, embryos
were collected and processed as reported 19 and examined using Nomarski DIC optics. Lethality assays were performed at 25 and 29 °C, except when using UAS-iReIF5A and UAS-iReIF5A; UAS-Dia-GFP that were performed only at 25 °C.

Drosophila embryos and imaginal disc immunofluorescence. For immunohistochemistry, embryos
were fixed with 3.7% formaldehyde diluted in phosphate buffer saline (PBS) and 1:1 heptane-methanol for devitellinization. Third instar larval wing imaginal discs were dissected, fixed and stained as described 38 . Primary antibodies used were: rabbit anti-Dia (1:500 39 , mouse anti-GFP (1:250, Roche) and rabbit anti-eIF5A (1:100, Abcam). Secondary antibodies coupled to fluorochromes were purchased from Invitrogen and used at a 1:200 dilution. For phalloidin staining, embryos were fixed with 4% paraformaldehyde in PBS and devitellinization was done with in 1:1 heptane-80% ethanol. After washing in 100% ethanol, embryos were incubated twice in PBS with Triton-X100 (PBS-Tx)−1% BSA for 10 min each time, twice in PBS-Tx-2% BSA for 10 min each time and 1 h in Alexa Fluor 568 phalloidin reagent (1:20, Invitrogen). Phalloidin incubation was followed by 15 min incubations of PBS-Tx with decreasing concentrations of BSA (2-0%). Specimens were mounted in Mowiol and images were acquired either on a LeicaTCS-NT confocal laser-scanning microscope or on an Olympus FV1000MPE confocal microscope. Several Z sections of each embryo were taken and processed with ImageJ. Imaginal discs images were acquired in a Leica DMI2500 optical fluorescence microscope. For quantification of fluorescence intensity in en>GFP> iReIF5A embryos, images of anti-Dia stainings were converted to grayscale. The Mean Gray Values of ten regions of interest of similar areas were measured per segment with ImageJ. The fluorescence intensity was calculated as the average of the Mean Gray Values obtained from three eIF5A mutant and three control segments.

Yeast strains, growth conditions and plasmids. Saccharomyces cerevisiae haploid strains
wild-type BY4741 (MATa his3∆0 leu2∆0 met15∆0 ura3∆0) and temperature-sensitive eIF5A mutants tif51A-1 and tif51A-3 derivatives of the BY4741 strain 23 were grown in liquid YPD or SC complete medium when carrying plasmids at 25 °C until the exponential phase and then treated as indicated. The Sc and Dm eIF5A genes were cloned in the yeast centromeric heterologous expression plasmid p416TEF (URA3 marker), under the control of the constitutive TEF promoter and using the CYC1 terminator (PA300 and PA304 plasmids, respectively). For this, we used a GAP repair strategy 40 . The ORF of the yeast TIF51A gene (encoding the essential copy of eIF5A) was PCR amplified using primers GAPR 5′-eIF5A-Sc (5′-GTTTTCTAGAACTAGTGGATCCCCCGGGCTGCAGGAATTCGATATGTCTGACGAAGAACA-3′) and GAPR 3′-eIF5A-Sc (5′-ACTAATTACATGACTCGAGGTCGACGGTATCGATAAGCTTGATTTAATCG GTTCTAGCAG-3′) and yeast genomic DNA as template. The Dm eIF5A unique gene was amplified with primers GAPR 5′-eIF5A-Dm (5′-GTTTTCTAGAACTAGTGGATCCCCCGGGCTGCAGGAATTCGATAT GGCTGAGTTGGACGA-3′) and GAPR 3′-eIF5A-Dm (5′-ACTAATTACATGACTCGAGGTCGACGGTATCG ATAAGCTTGATCTATTTGTCCAGAGCAG-3′) using the GH16179 cDNA clone from the Drosophila Genomics Resource Center (DGRC) EST collection as template. PCR cassettes were co-transformed with EcoRI linearized p416TEF plasmid 41 in BY4741 yeast cells and cells with GAP repaired plasmids were selected in SC-ura media. Plasmids were recovered from yeast cells and directly transformed in E. coli competent cells, then plasmids were extracted from bacteria and the correct cloning was confirmed by DNA sequencing. NSCs culture and migration assays. Subependymal NSCs were isolated from 2 month-old C57BL/6J mice and cultured as described 42 . Viability was analyzed using a commercial MTS assay (Promega). Deoxyhypusine synthase inhibitor GC7 (Calbiochem) was dissolved in sterile distilled water. For migration assays, neurospheres were individually selected according to their size and plated on Matrigel ® -coated coverslips. Migration area was measured with ImageJ software and normalized to the initial neurosphere area.  (1:800, Jackson Immunoresearch) secondary antibodies for 1 h. DAPI (1 μg/ml) was used to counterstain nuclei. Pictures were taken using a confocal microscope (Olympus FV10) and all images were processed equally with Photoshop.
Quantitative RT-PCR. RNA extraction, reverse transcription and quantitative PCR (qRT-PCR) analyses in yeast were used to quantify BNI1-HA and HXK2 mRNA amount and normalized against endogenous ACT1 expression as described 45 . Translation efficiencies of BNI1-HA and HXK2 were estimated as the protein/ mRNA ratio, using the Western and RT-qPCR data (normalized against ACT1 expression) obtained from cell samples of the same experiment. BNI1-HA translation efficiency is expressed relative to HXK2 translation efficiency and represented as a fraction against 25 °C for each strain. For NSCs, RNA was isolated using RNeasy Mini Kit (Qiagen), cDNA was synthesized using PrimeScript ™ RT reagent Kit (Takara) and qRT-PCR was performed using TaqMan ® probes (Life Technologies). For qRT-PCR in yeast, BNI1-3/ BNI1-4, HXK2-1/HXK2-2, and ACT1-1/ACT1-2 primers were used, which are described in ref. 15. For qRT-PCR in NSCs, the following probes were used: Eif5a1 (Mm01971736_g1), Eif5a2 (Mm00812570_g1), Cyb5r3 (Mm00504077_m1) and Gapdh (Mm99999915_g1).