INF1 Is a Novel Microtubule-associated Formin

Analysis of INF1 Evolutionary Relationships

The FH2 domains of INF1, INF2, nematode inft-1, and invertebrate FH2 sequences that were found to be most similar to the FH2 domains of INF1 and INF2 were compared with the FH2 domains of several other animal formins. The invertebrate formin proteins similar to INF1 and INF2 (named INFX in this study) were initially retrieved from GenBank using TBLASTN (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi) with either the INF1 zebrafish or INF2 pufferfish sequence. The same invertebrate sequences were found when searching with either the INF1 or INF2 sequence. Specific sequences used are listed in Figure 1. The FH2 regions used from each formin were determined with SMART analysis (http://smart.embl-heidelberg.de/). A phylogenetic comparison was performed with ClustalX 2.0 run locally. We used a bootstrap analysis with 1000 replicates after initial alignment of the sequences. An unrooted tree based on this analysis was generated with NJplot, and redrawn in Adobe Illustrator (San Jose, CA). The overall domain structure of these proteins was then compared using SMART analysis.

Plasmid Constructs

INF1 expression plasmids were generated using the pEYFPN1, pEGFPC3 (Clontech, Palo Alto, CA), pGex6p2 (Amersham Pharmacia Biotech, Piscataway, NJ), and EFplink vectors with either myc or Flag epitope tags (Sotiropoulos et al., 1999). The full-length human INF1 open reading frame was generated using DNA from the KIAA1727 cDNA clone (obtained from the Kazusa DNA Research Institute), with missing 5′ sequence being added on using oligonucleotides encoding amino acids 1-25. C-terminal truncated INF1 fusions were made in pEYFP by removing sequence coding for the region downstream from amino acid 1055 (INF1ΔC1-YFP) or amino acid 959 (INF1ΔC2-YFP) using internal BglII or SacII sites, respectively. The INF1 C-terminal fusion, GFP-INF1C (codons 958-1143), was made by removing a SacII fragment from the full-length INF1FL-YFP plasmid to insert into the pEGFPC3 vector. This same SacII fragment was also used to generate a GST-INF1C fusion in pGex6p2. GFP-ΔMTB1 and GFP-ΔMTB2 plasmids were generated by removing codons 958-1001 or 1033-1042, respectively. Myc-NT plasmid encoded amino acids 1-485, myc-FH2 encoded amino acids 85-485, and myc-CT encoded amino acids 486-1143. Codon numbering is given for the human INF1 protein in GenBank accession number NP_203751. For mouse INF1, the full open reading frame was cloned using overlapping fragments generated with RT-PCR, shown in Supplemental Figure S1. The full insert was cloned as a BglII-SalI insert into pEGFPC1 (GFP-mINF1) using standard cloning techniques. Constructs used for SRF activation assays, 3DA.Luc and MLV-LacZ, have been described previously (Geneste et al., 2002 and Sotiropoulos et al., 1999, respectively). The inserts of all newly generated plasmids were sequenced to ensure their accuracy, and appropriate expression of the fusion proteins was confirmed by immunoblotting.

Cell Culture and Animals

The Cos-1 monkey kidney cell line, N2A mouse neuroblastoma cells, F11 rat/mouse sensory neuron cell line, H9C2 rat cardiomyocyte line, and HeLa human carcinoma line were each maintained in DMEM containing 10% FBS and 1% pen/strep in a 37°C incubator with 5% CO₂. Subconfluent cells were passaged in 10-cm plastic dishes. For differentiation of the N2A cells, serum containing DMEM was replaced with DMEM containing 0.5 mM dibutyryl cAMP (Calbiochem, Mississauga, ON, Canada) and 1% penicillin/streptomycin for 24–48 h. Mouse NIH 3T3 fibroblast cells were maintained in DMEM containing 10% FBS and 1% pen/strep in a 37°C incubator with 10% CO₂. Subconfluent cells were passaged in 10-cm plastic dishes for up to 13 passages.

Cos-1 cells were transfected with PEI (polyethyleneimine; Polysciences, Warrington, PA). This was done by mixing 5 μl of a 1 μg/μL PEI solution with 1.5 μg of plasmid DNA in 50 μl of OptiMEM (Invitrogen, Carlsbad, CA) for 15–30 min. This was then added to one well of cells in a six-well plate in 1 ml of OptiMEM. After a 5- to 8-h incubation, the transfection medium was replaced with normal growth medium. NIH 3T3 cells were transfected with either PEI, or in the case of the SRF activation assays, Lipofectamine (Invitrogen). Drug treatments with nocodazole (Sigma, Oakville, Ontario, Canada) or latrunculin A (Sigma) were performed as indicated in Results. Both nocodazole and latrunculin A stock solutions were prepared in DMSO (Sigma); DMSO alone was used for control samples. For microscopy, all cells were plated on sterile glass coverslips.

Mouse tissues were collected from 5-d-old and 2-mo-old C57BL/6 mice. Tissues use for immunohistology were immersed in OCT compound (Sakura, Tokyo, Japan) in plastic molds, frozen using liquid nitrogen, and stored at −80°C. Frozen sections were cut in a cryostat at an 8-μm thickness and placed on gelatin-coated Superfrost Plus slides (Fisher Scientific, Pittsburgh, PA). The sections were air-dried and either stained immediately or stored at −20°C. Tissue used for immunoblotting were collected in lysis buffer containing 50 mM Tris (pH 7.5), 100 mM NaCl, 5% glycerol, 1 mM EDTA, and 1% Triton X-100. The protein content was determined by Bradford analysis, and samples were diluted in a standard SDS buffer for SDS-PAGE analysis.

INF1 Antibody Generation

INF1 antibody was generated against a protein corresponding to the human INF1 FH2 domain (amino acids 85–541). Briefly, protein expressed in BL21 bacteria was purified by PreScission Protease (GE Healthcare, Waukesha, WI) cleavage of GST-INF1 FH2 protein bound to glutathione Sepharose 4B beads. The purified FH2 protein was dialyzed against phosphate-buffered saline (PBS), and used for injection into New Zealand White rabbits (Cedarlane Laboratories, Burlington, ON, Canada). Antibody from the immune serum was affinity-purified against protein A, followed by purification against GST-INF1 FH2, using standard techniques. Antibody was eluted with a glycine buffer (pH 2.8) and then buffered with pH 9.0 Tris before dialysis into PBS.

Immunoblotting

Protein lysates were collected from cells grown in culture by two methods. Initially, equal numbers of cells were washed twice in PBS, scraped in Laemmli buffer, and boiled for 5 min before cooling on ice. As a second method, cells were washed twice in PBS and then scraped in a high salt lysis buffer (50 mM HEPES, pH 7.5, 4% SDS, 300 mM NaCl, 1 mM EDTA) supplemented with a protease inhibitor cocktail (Roche, Indianapolis, IN), and dithiothreitol (5 mM) was added directly before use. These samples were immediately boiled for 5 min and then cooled in room temperature water for 1 min, before adding 50 μl of 300 mM iodoacetamide per 500 μl of sample. The samples were centrifuged and sheared through a 27-gauge needle. To load these samples on a gel, they were added to an equal volume of gel-loading buffer containing 100 mM Tris, pH 6.8, 4% SDS, 20% glycerol, 5% β-mercaptoethanol, 1 M NaCl, 4 M urea, and 0.1% bromophenol blue. Lysates were run on standard SDS-PAGE gels and transferred onto PVDF membrane (Millipore, Bedford, MA). The blots were blocked with 5% nonfat dairy milk (NFDM) in Tris-buffered saline (TBS). Primary and secondary antibody incubations were performed in TBS containing 0.5% Tween 20 (TBST). All washings were performed using TBST. HRP-labeled secondary antibodies were detected with chemiluminescence reagent (Perkin Elmer-Cetus, Boston, MA) before blots being exposed to autoradiography film. Affinity-purified anti-INF1 antibody was incubated at 1:500 overnight at 4°C, anti-α-tubulin antibody was incubated at 1:5000 and anti-lamin B at 1:200 at room temperature for 1 h.

INF1 Small Interfering RNA

INF1 was knocked down in NIH 3T3 cells using an RNA oligonucleotide duplex (Integrated DNA Technologies, Coralville, IA) with the following sequences: 5′-GCUAUAGCACCAAAGAGAAAUUCCT-3′ and 5′-AGGAAUUUCUCUUUGGUGCUAUAGCAU-3′. This duplex was found to reliably reduce GFP-mINF1 levels when cotransfected in Cos-1 cells (Supplemental Figure S2). NIH 3T3 cells were transfected with the small interfering RNA (siRNA) duplex using Dharmafect 1 (Dharmacon Research, Boulder, CO) following the manufacturer's instructions. We examined INF1 levels in cell lysates collected at 2, 3, 5, 7, and 10 d after transfection by immunoblotting. Lysates were collected in high-salt lysis buffer as described above. Antibodies to detect acetylated tubulin (Sigma), α-tubulin (Sigma), and lamin B (Santa Cruz Biotechnology, Santa Cruz, CA) were used to probe the blots following stripping in a pH 2.5 glycine (0.1 M) and 0.5% SDS buffer between each probing.

Serum Response Factor Activation Assay

SRF activation assays were performed as described previously (Copeland et al., 2002). Briefly, NIH 3T3 cells were transfected with 50 ng of the reporter construct 3DA.Luc, along with 250 ng of the reference plasmid, MLV-LacZ, and the expression plasmid encoding the fusion protein to be tested. Empty pEF Flag was cotransfected to a total of 1.5 μg of DNA per well in six-well plates. After the transfection, cells were maintained in 0.5% FBS in DMEM with 1% penicillin/streptomycin. After 24 h, cells were harvested for a standard luciferase assay. Transfection efficiency was standardized by a β-galactosidase assay. Data are shown relative to reporter activation by the constitutively active SRF derivative, SRFVP16 (50 ng).

In Vitro Microtubule Association and Bundling Assays

GST-INF1C fusion protein was purified from BL21 bacteria induced with 0.2 mM IPTG at 37°C. Purified protein incubated on glutathione Sepharose 4B beads (GE Healthcare) was recovered by PreScission Protease (GE Healthcare) cleavage of the glutathione S-transferase (GST) moiety, followed by dialysis into PEM buffer (20 mM PIPES, pH 7, 20 mM KCl, 1 mM DTT, and 5% glycerol). Microtubule spin-down assays were performed using a microtubule spin-down kit (Cytoskeleton, Denver, CO). Microtubules were assembled for 20 min at 35°C in a general tubulin buffer (80 mM PIPES, pH 7.0, 2 mM MgCl₂, and 0.5 mM EGTA) in the presence of GTP and taxol. INF1C or control protein was incubated with 0.83 μM tubulin dimers for 30 min. Microtubules and associated proteins were then pelleted at 100,000 × g through a 50% glycerol cushion buffer. Supernatant and pellet were combined with Laemmli loading buffer, and equal amounts of sample were run on a 10% SDS-PAGE gel for analysis by Coomassie blue staining.

To examine microtubule bundling by immunofluorescence, INF1C or control protein was incubated with stabilized microtubules at room temperature for 30 min. Samples were then placed on gelatin-coated Superfrost Plus slides (Fisher) and incubated for 10 min before fixing with 4% paraformaldehyde. Microtubules were visualized with anti-α-tubulin antibody and an Alexa Fluor 594–labeled secondary antibody (Invitrogen). The preparations were photographed on an AxioImager microscope (Zeiss, Thornwood, NY) with a 63× Plan Apochromat objective.

Immunofluorescence

Tissue culture cells and mouse brain sections were stained using a basic immunofluorescence protocol. Briefly, cells or tissue sections were washed twice in PBS and then fixed with 4% paraformaldehyde in PBS for 10 min. The cells or tissue sections were washed three times for 5 min in PBS and permeabilized and blocked with PBS containing 0.4% Triton X-100 and 10% FBS. Primary antibody diluted in PBS containing 0.04% Triton X-100, and 5% FBS was then incubated on the samples either overnight at 4°C or for 1 h at room temperature. The samples were washed three times for 5 min and then incubated for 1 h at room temperature with the appropriate secondary antibody. After a final wash of three times for 5 min in PBS, the samples were mounted using Vectashield with DAPI (Vector Laboratories, Burlingame, CA). Primary antibodies used included affinity-purified anti-INF1 polyclonal at 1:50, anti-α-tubulin monoclonal at 1:1000, anti-acetylated tubulin monoclonal at 1:1000 (Sigma), anti-detryrosinated tubulin at 1:250 (Chemicon, Temecula, CA), and anti-βIII-tubulin (tuj1) at 1:1000 (Covance Laboratories, Madison, WI). Phalloidin-rhodamine or phalloidin-Alexa Fluor 488 (1:20; Invitrogen), incubated with the secondary antibody, were used to detect F-actin.

Cells and tissues were imaged primarily with a Zeiss AxioImager equipped with an apotome. Apotome optical sections were imaged with a 63× Plan Apochromat objective producing 0.7-μm sections. Confocal imaging was done with a LSM 5 Pascal confocal system attached to an Axiovert 200M inverted microscope (Zeiss). Confocal sections of 1 μm were produced using a 63× Plan Apochromat objective with 488- and 543-nm laser lines being used for excitation. A bandpass 505–530-nm emission filter was used to collect the green emission and exclude nonspecific emission from the red fluorophore, and a long-pass 560 filter was used for the red emission.

INF1 Is a Unique Microtubule-associated Formin That Affects F-Actin Organization

To understand the origin of the unique INF1 domain structure, we analyzed the phylogenetic relationship between INF1 and other animal formins. The INF1 and INF2 proteins diverged in the early chordates from an ancestral form containing both a diaphanous-related region upstream from the FH1 and FH2 domains, and an extensive C-terminal end downstream. Related sequences for previously undescribed Animalia formins with this type of structure are found in various insect species, as well as the sea urchin, Strongylocentrotus purpuratus (Figure 1). We have named these the INFX proteins. These proteins resemble the INF2 protein (Higgs and Peterson, 2005), but have longer C-terminal regions downstream from the FH2 domain. The INFX FH2 domains are intermediate to INF1 and INF2 in their primary sequences. Regions coding for N-terminal diaphanous-related formin domains are absent 5′ to the FH1/FH2 coding region in the INF1 gene from human, mouse, and zebrafish (as assessed with MIT Genscan). Sequences resembling the diaphanous autoregulatory domain (DAD) and diaphanous inhibitory domain (DID) do not appear to be present in INF1. INF1 does contain several regions within its C-terminal end that are conserved between vertebrate orthologues (Katoh and Katoh, 2004), including regions we have labeled as MTB1 and MTB2 (Figure 1C). INF1 is therefore likely to be regulated in a manner fundamentally different from the better-characterized diaphanous-related formins and may contain novel functions conferred on it by its unique C-terminal end.

To examine if INF1 exhibited functions characteristic of other formins, we assessed whether expression of INF1 fusion proteins could induce the formation of stress fibers and induce SRF activation in serum-starved NIH 3T3 fibroblasts. Transfection of cells with plasmids encoding full-length and N-terminal INF1 fusion proteins could induce the formation of stress fibers in the 3T3 cells after 1 d of expression (Figure 2 and Supplemental Figure S3). Induction of stress fiber formation by full-length INF1 was similar to that of the isolated FH2 domain from mDia1 (Figure 1E). The N-terminal fusion proteins of INF1 were, however, less efficient at inducing thick stress fiber formation, producing mainly an increase in thin filaments. Interestingly, the full-length INF1 localized discretely with microtubules (insets in Figure 2, A and C).

Full-length and N-terminal INF1 fusion proteins also induced SRF activation in serum-starved NIH 3T3 cells (Figure 2F). Expression of the FH2 domain alone (myc-FH2) was sufficient to induce activation of the SRF reporter gene, consistent with our previous results obtained with mDia1 (Copeland et al., 2002). Also similar to mDia1, expression of protein with both the FH1 and FH2 domains (myc-Nterm and myc-INF1FL) induced an increased response in the SRF activation assay. Expression of the C-terminal end of INF1 had no effect on SRF activation. Of note is that the full-length INF1 fusion protein was constitutively active in the SRF and stress fiber formation assays. This is consistent with INF1 activity being regulated differently than the diaphanous-related formins.

Endogenous INF1 Is Predominantly Microtubule-associated

To examine expression of the INF1 protein, we produced affinity-purified polyclonal antibody against the human INF1 FH2 domain. This antibody recognized both mouse and human INF1 fusion proteins (Figure 3). By immunofluorescence, the INF1 antibody specifically recognized full-length INF1FL-YFP fusion protein expressed in Cos-1 cells (Figure 3A). Untransfected Cos-1 cells did not display any staining. By immunoblot analysis of NIH 3T3 lysates, we observed a prominent band of ∼125 kDa, which could be knocked down by transfecting cells with an siRNA duplex targeted to the INF1 transcript (Figure 3B). Unexpectedly, full-length INF1 fusion proteins ran appreciably higher than the 125-kDa mark on standard SDS-PAGE gels (Figure 3C). We were able to observe a band corresponding to the size of full-length INF1 fusion protein, present in several cell lines analyzed, though only with the use of cell lysis buffer containing protease inhibitors with a high salt concentration (see Materials and Methods), and longer exposures of the blots. Detection of this second band, as well as the other bands recognized by the antibody, was greatly reduced by preincubating the INF1 antibody with purified GST-INF1 (data not shown). These data suggest that the predominant lower molecular weight band may be due to protein modification of the endogenous INF1. Alternatively, rapid protein degradation or modification upon cell lysis may occur. No evidence of alternative splicing for mouse INF1 was found by RT-PCR analysis (Supplemental Figure S1).

By immunofluorescence analysis of endogenous INF1 in NIH 3T3 cells, the INF1 antibody detected protein discretely coaligning with a subset of microtubules (Figure 3, D–G). Filamentous INF1 staining was present in a majority of cells, with some punctate staining also being commonly observed. Both punctate and filamentous INF1 staining was primarily microtubule-associated in all cells. The same was true in a second cell line with a fibroblast-like morphology, H9C2 cardiomyoblasts (Supplemental Figure S4). Because the INF1 FH2 domain may be involved in actin filament association and polymerization, we also investigated whether INF1 colocalized with F-actin staining. There was, however, no consistent pattern of INF1 localization with F-actin in NIH 3T3 (Figure 3) or H9C2 cells. Correspondingly, disruption of microtubule polymerization with nocodazole in NIH 3T3 cells resulted in the diffuse localization of the endogenous INF1 (Figure 4, E–H). The disruption of actin filament polymerization with latrunculin A had little effect on INF1 distribution (Figure 4, I–L). Endogenous INF1 is therefore a predominantly microtubule-associated protein in these cells.

INF1 Expression in Mouse

To gain some insight into the types of structures INF1 might normally associate with in animal tissues, we screened several mouse tissues for INF1 expression. By immunoblot analysis, we observed INF1 to be readily detectable in brain, heart, and lung tissues (Figure 5A). Correspondingly, INF1 was easily detected by immunofluorescence staining in ventricular muscle sections of the heart, but not in kidney sections (Figure 5, B and C). In the brain, we found INF1 to be most strongly expressed in Purkinje neurons of the cerebellum. In sections from 5-d-old animals, INF1 was predominantly found in the cell bodies of the immature Purkinje neurons (Figure 5D), which do not fully mature until the second week after birth (Millen et al., 1994). In the adult animal, INF1 staining was primarily found in the white matter of the cerebellum, with weak staining in projections from the Purkinje cell layer (Figure 5E). This staining appeared to be axonal and was localized in proximity to neuronal βIII-tubulin staining (Figure 5, F–H).

The INF1 C-Terminus Contains a Microtubule-binding Domain and Is Required for INF1 Microtubule Association

To examine which domain of the INF1 protein might target INF1 to microtubules, we investigated whether the conserved regions in the C-terminal half may be necessary for proper localization. We initially examined the importance of the two conserved regions located toward the C-terminal end of the protein (labeled MTB1 and MTB2 in Figure 1). Full-length INF1FL-YFP fusion protein primarily coaligned with microtubules in Cos-1 cells (Figure 6, A–C). Transfection with a full-length mouse GFP-mINF1 construct additionally resulted in a noticeable bundling of the microtubules (Supplemental Figure S1). The INF1ΔC1-YFP fusion, which had sequence removed downstream of the MTB1 and MTB2 regions, also localized in a filamentous pattern along the length of the microtubules. This filamentous localization occurred in fewer than half of the cells compared with the full-length INF1 fusion protein, however, indicating a reduced ability to associate with microtubules (Figure 6, D–F). More cells expressing INF1ΔC1-YFP displayed puncta or diffusely localized fusion protein. The INF1ΔC2-YFP protein, which lacks the MTB1 and MTB2 regions, was primarily diffusely localized or localized in puncta (Figure 6, G–I). A C-terminal fusion protein containing MTB1 and MTB2, GFP-INF1C, localized discretely with bundled microtubules in Cos-1 cells (Figure 6, J–L). Prominent nuclear localization in most of the cells may have been an artifact of the high expression of this small fusion protein and was not observed with the full-length fusion protein. The failure of INF1ΔC2-YFP to associate with microtubules, as well as the association of GFP-INF1C with bundled microtubules, demonstrates that the C-terminal end of INF1 is both necessary and sufficient for the microtubule association of INF1.

To further define the regions in the C-terminal end of INF1 that are necessary for microtubule association, we generated fusion proteins lacking the conserved MTB1 (GFP-ΔMTB1) or MTB2 (GFP-ΔMTB2) regions. In Cos-1 cells, these proteins were diffusely localized, with limited microtubule association (Figure 7, C–F). Additionally, neither GFP-ΔMTB1 nor GFP-ΔMTB2 induced any microtubule bundling, whereas microtubule bundles were apparent in one-third of the Cos-1 cells expressing GFP-INF1C (Figure 7, A and B). Therefore, both the MTB1 and MTB2 regions play a role in the association of INF1 with microtubules.

To determine if the INF1 C-terminus is able to bind microtubules directly, we produced a GST-INF1C protein. The INF1 C-terminal end was purified by glutathione affinity chromatography and protease cleavage of the GST tag. The purified INF1C protein was then used in an in vitro microtubule spin-down assay and in a fluorescence-based microtubule-bundling assay. Proteins that are normally soluble will cosediment with the microtubules in the spin-down assay when centrifuged at 100,000 × g if they bind directly to the microtubules. In this assay, soluble INF1C incubated on its own remained in the supernatant after ultracentrifugation. In contrast, INF1C incubated with purified microtubules cosedimented with the microtubules and was found primarily in the pellet after the high-speed spin (Figure 7G). To assess the effect of INF1C on microtubule bundling, we incubated INF1C with purified microtubules that were then fixed and stained on microscope slides. Incubation with INF1, but not control GST protein, induced microtubule bundling in this assay (Figure 7, H and I). Thus, the INF1 C-terminus is able to bind and bundle microtubules directly.

INF1 Stabilizes Microtubules

The induction of bundled basket-like microtubules by INF1C is consistent with INF1 promoting microtubule stabilization. To test this, we treated Cos-1 cells with 1 or 10 μM nocodazole for 1 h. Treatment with 1 μM was sufficient to depolymerize most microtubules outside of the centrosomes and midbodies, whereas 10 μM nocodazole additionally caused centrosome and midbody depolymerization. Cells expressing the full-length INF1FL-YFP were able to retain an extensive microtubule network at 1 μM nocodazole as well as at the 10 μM concentration, albeit to a lesser extent (Figure 8, A, B, and G). Cells expressing the C-terminal GFP-INF1C protein also retained extensive microtubule networks, with the microtubules commonly retaining a bundled appearance (Figure 8, E and F). INF1ΔC2-YFP expressing cells contained primarily diffuse microtubule staining (Figure 8, C and D). A few cells expressing INF1ΔC2-YFP displayed a very limited number of microtubule filaments (Figure 8G), though at the 1 μM nocodazole concentration this was not significantly different from GFP expressing control cells (p = 0.34, two-tailed t test). There was, however, a marginally significant difference observed between INF1ΔC2-YFP and GFP expressing cells for the 10 μM concentration (p = 0.04).

INF1 Induces the Formation of Acetylated Microtubules

We next examined whether the expression of INF1 fusion proteins could be associated with the formation of acetylated microtubules, a marker for microtubule stabilization. We expressed INF1FL-YFP, INF1ΔC2-YFP, or GFP-INF1C in confluent NIH 3T3 cells. The confluent NIH 3T3 cells contained weak acetylated microtubule labeling, with strong labeling limited to the spindles of dividing cells and midbodies. As expected, expression of GFP-INF1C induced the robust formation of bundled, acetylated microtubules (Figure 9, E and F). The INF1FL-YFP protein also induced acetylated microtubule formation, though to a lesser extent than the isolated C-terminus (Figure 9, A and B). Surprisingly, expression of the INF1ΔC2-YFP protein, which did not obviously associate with microtubules, also induced the accumulation of acetylated microtubules in approximately one-third of cells (Figure 9, C and D). Unlike the full-length and C-terminal fusions, however, the INF1ΔC2-YFP did not necessarily colocalize with the acetylated microtubules. Transfection of a control GFP plasmid did not cause any increase in acetylated microtubule staining in the expressing cells. Thus the effect is specific to expression of INF1. The FH2 domain of other formins has been implicated in microtubule interactions (Ishizaki et al., 2001; Rosales-Nieves et al., 2006; Bartolini et al., 2008). Therefore, we tested the ability of the INF1 FH2 domain to induce formation of acetylated microtubules. Expression of a Flag-tagged derivative of the INF1 FH2 domain failed to induce the accumulation of acetylated microtubules (Figure 7, G and H).

Microtubule detyrosination is another marker of microtubule stability, and therefore we examined whether INF1 overexpression could be associated with an increase in detyrosinated microtubules. In contrast to microtubule acetylation, we were unable to detect an INF1FL-YFP–induced increase in detyrosinated microtubules (Supplemental Figure S5).

Because INF1 appeared to promote the formation of acetylated microtubules in NIH 3T3 cells, we next examined if the knockdown of INF1 in these cells would result in decreased levels of acetylated microtubules. As assessed by immunoblot analysis, we observed a decrease in acetylated tubulin levels in cells with reduced INF1 levels (Figure 9J). Expression of α-tubulin, which is the acetylated monomer (Eddé et al., 1991), was not decreased, indicating that it is only tubulin acetylation that is reduced and not tubulin expression levels. Finally, to determine if endogenous INF1 was targeted to acetylated microtubules in NIH 3T3 cells, we immunostained cells for both and looked at their colocalization. INF1 was found to be associated with acetylated microtubules (Figure 9K, L). Together these data show that INF1 is able to bind to microtubules and induce microtubule stabilization and acetylation.