INF2 is an endoplasmic reticulum-associated formin protein

Formins are a class of proteins that accelerate filament nucleation and subsequently modulate filament elongation (Higgs, 2005). A distinguishing feature of formins, as opposed to other assembly factors such as the Arp2/3 complex, is their diversity. Most eukaryotes possess multiple formin genes, with mammals having 15 (Higgs and Peterson, 2005). This diversity provides the potential to act in a myriad of cellular functions. Despite this potential, little is known of the function or even the cellular localization of many mammalian formins (Chhabra and Higgs, 2007).

The two core sequence domains defining formins are the formin homology 1 (FH1) and formin homology 2 (FH2) domains. The FH2 domain is dimeric and directly interacts with actin. In addition to accelerating nucleation, the FH2 domain binds the actin filament barbed end, and moves processively with the barbed end as additional monomers are added (Higgs, 2005; Kovar, 2006). The FH1 domain is proline-rich, and binds the actin monomer-binding protein, profilin. In the presence of profilin, the FH1 accelerates filament elongation by facilitating monomer addition to the FH2-bound barbed end (Pollard, 2007).

The mammalian formin INF2 has the unique ability to accelerate both actin polymerization and depolymerization (Chhabra and Higgs, 2006). The depolymerization activity is mediated both by its FH2 domain and by its unique C-terminal region, which includes an actin monomer-binding WASP homology 2 (WH2) motif. The switch between polymerization and depolymerization is triggered by the nucleotide state of actin. Similarly to other formins, INF2 accelerates polymerization of ATP-actin monomers. Upon addition to the filament, actin subunits hydrolyze their bound ATP and release the phosphate product. Phosphate release triggers INF2 to accelerate depolymerization in a two-step mechanism (Chhabra and Higgs, 2006).

The regulatory mechanisms controlling INF2 polymerization and depolymerization activity are unknown. Several other formins, including mDia1 and mDia2, are regulated by an auto-inhibitory mechanism (Alberts, 2001; Li and Higgs, 2003; Li and Higgs, 2005; Wallar et al., 2006), whereby an N-terminal diaphanous inhibitory domain (DID) binds to a C-terminal diaphanous auto-regulatory domain (DAD). The DID-DAD interaction blocks the ability of the FH2 to interact with actin. For mDia1, autoinhibition is relieved by Rho family GTPase binding to a region overlapping the DID, disrupting the DID-DAD interaction, and allowing FH2 activity. INF2 possesses a putative DID, and the WH2 motif of INF2 has hallmarks of a DAD. Thus, INF2 might be regulated by autoinhibition.

No information is available on the cellular function of INF2. Here, we localize both endogenous INF2 and GFP-fusions of INF2 to the cytoplasmic face of the endoplasmic reticulum (ER). ER binding is partially mediated by a C-terminal farnesyl group, and partially by ionic interactions. We show that the DID in INF2 binds to the DAD, and that the DID-DAD interaction inhibits INF2 depolymerization activity. However, INF2-mediated polymerization is not inhibited by the DID-DAD interaction, in stark contrast to other formins. Mutations to the DAD/WH2 disrupt both the DID-DAD interaction and the interaction between the WH2 domain and actin monomers. Overexpression of GFP-INF2 containing the DAD/WH2 mutation causes ER to collapse around the nucleus, with the collapsed ER surrounded by abundant actin filaments.

We raised antibodies against two regions of INF2: the FH1FH2 domain and INF2-(C) (Fig. 1A). Western blotting with Swiss 3T3 cell extract revealed a single major band with an apparent molecular mass of 200 kDa with both antibodies (Fig. 1B), which is considerably larger than the predicted mass (138 kDa). However, tryptic digest and mass spectrometric analysis of the immunoprecipitated 200 kDa band resulted in a peptide footprint that was remarkably similar to that predicted for INF2 (unpublished results), confirming this band as INF2.

INF2 was detected in all tissues and cell lines tested, with highest levels in the brain and Swiss 3T3 cells (Fig. 1C). We assessed INF2 levels in 3T3 cells by quantitative western blotting (supplementary material Fig. S1). By this analysis, 3T3 cells contain 381,000±32,000 molecules of INF2 per cell. From calculations of total cell and nuclear volumes (see Materials and Methods), we estimated the cytoplasmic INF2 concentration to be about 300 nM. For comparison, we found the concentration of the Arp2/3 complex to be 1.5 μM (Nicholson-Dykstra and Higgs, 2008) and that of mDia1 to be at <100 nM (unpublished observations) in these cells.

To assess whether INF2 is cytosolic or membrane-bound, we fractionated 3T3 cell extract by low speed (2000 × g) centrifugation followed by high-speed (436,000 × g) centrifugation of the low-speed supernatant. INF2 partitions mainly to the low-speed pellet, with lower levels in the high-speed pellet and high-speed supernatant (Fig. 1D). Several transmembrane proteins displayed similar ratios in the low- and high-speed pellets, including syntaxin 6 (Golgi) (Fig. 1D), ERV46 (ERGIC, not shown) and calnexin (ER, not shown), and were not present in the high-speed supernatant. By contrast, tubulin (Fig. 1D) and actin (not shown) partition almost exclusively to the high-speed supernatant. This result suggests that INF2 is membrane-bound, but that a small pool of cytosolic INF2 exists.

To test the nature of the INF2 interaction with membranes, we treated the low-speed supernatant fraction with high ionic strength, high pH, or non-ionic detergent (Fig. 1E). High ionic strength causes a portion of INF2 to shift to the high-speed supernatant fraction. High pH, which has been shown to be more effective than high ionic strength at disrupting electrostatic interactions between proteins and membranes (Fujiki et al., 1982), causes all detectable INF2 to shift to the supernatant. Under these conditions, the transmembrane protein, syntaxin, remains in the pellet. These results suggest that ionic interactions are important for INF2 association with membranes.

When analyzed on 5% SDS-PAGE, which gives optimal resolution in the 200 kDa range, INF2 migrates as a doublet. Although the reason for this doublet is unclear, we doubt that it results from proteolysis during extract preparation, because lysis of cells directly in SDS-PAGE sample buffer containing a broad range of protease inhibitors does not change this pattern (not shown).

By immunofluorescence microscopy, INF2 displays a distinct reticular staining pattern in 3T3 cells, which is characteristic of ER. To examine this staining further, we transfected 3T3 cells with GFP-tagged human sec61β (GFP-Sec61), an ER membrane protein, and immunostained for INF2. GFP-Sec61 and INF2 colocalize extensively around the nucleus as well as in a reticular pattern at the cell periphery (Fig. 2A,B). We obtained similar results when we costained cells with INF2 and the ER luminal protein, GRP94 (supplementary material Fig. S2). Both anti-INF2 antibodies gave similar results. Although INF2 did not colocalize extensively with microtubules (Fig. 2C), treatment of cells with the microtubule-depolymerizing drug, nocodozole, resulted in loss of the peripheral reticular INF2 (Fig. 2D), GFP-Sec61 (data not shown) and GRP94 (supplementary material Fig. S2) staining patterns. Since microtubules have a vital role in the maintenance of peripheral ER structure (Terasaki et al., 1986; Waterman-Storer and Salmon, 1998), this result further suggests an ER localization for INF2.

Exogenously expressed GFP-INF2 showed a similar localization pattern to endogenous INF2 in fixed cells (supplementary material Fig. S3). To rule out the possibility that INF2 localization was due to a fixation artifact, we examined GFP-INF2 localization and dynamics by live-cell microscopy in Swiss 3T3 cells. GFP-INF2 showed similar localization and dynamics as mCherry-Sec61, with clear emergence and disappearance of ER tubules at the cell periphery (supplementary material Movie 1).

INF2 does not contain any putative transmembrane or ER localization or retention signals, and its membrane-extraction properties suggest that it is peripherally bound. We used an established differential membrane permeabilization technique (Lin et al., 1999) to determine whether INF2 associates with the lumenal or cytoplasmic face of the ER. We treated formaldehyde-fixed cells with either digitonin or Triton X-100. Under these conditions, digitonin permeabilizes the plasma membrane but not ER, whereas Triton X-100 permeabilizes both plasma membrane and ER. We costained cells with anti-INF2 and either anti-tubulin (as cytoplasmic control) or anti-calreticulin (as ER lumen control). Calreticulin staining was only visible in Triton-X-100-treated cells, whereas cells displayed a strong immunofluorescence signal for INF2 (Fig. 3) and tubulin (not shown) under both permeabilization conditions. This result suggests that INF2 is bound to the cytoplasmic leaflet of the ER.

We modified the digitonin permeabilization technique to examine the strength of the association of INF2 with the ER. Before fixation, we extracted Swiss 3T3 cells with digitonin for variable times, then fixed and stained for INF2, actin filaments and DNA. Separate experiments show that GFP is depleted within 2 minutes by this procedure (see Fig. 6). After 20 minutes, actin filament staining in stress fibers was undetectable, suggesting depolymerization followed by leaking of actin monomers (supplementary material Fig. S4). By contrast, INF2 staining was not diminished after 120 minutes of digitonin extraction (supplementary material Fig. S4), suggesting a strong association with the ER.

INF2 possesses a CAAX box at its C-terminus (C1270VIQ), which is indicative of post-translational protein prenylation, either with the 15 carbon farnesyl or 20 carbon geranylgeranyl group. To identify the prenyl moiety, we analyzed native INF2 by mass spectrometry. INF2 was immunoprecipitated from mouse brain and subjected to tryptic digest followed by LC-MS analysis. By this procedure, we found that INF2 is farnesylated on C1270. INF2 is subjected to two additional modifications associated with prenylation. First, the VIQ residues are removed; second, the C-terminus is carboxy-methylated. Farnesylation and VIQ proteolysis appear complete, as we detected no peptide population without these modifications, although lack of positive evidence is admittedly not negative evidence in LC-MS/MS experiments such as this. Carboxy-methylation is incomplete, as we estimate 20-30% free C-terminus from extracted ion chromatographic peak heights for each species (supplementary material Fig. S5). Incomplete carboxy-methylation occurs for prenylated proteins (Magee and Seabra, 2005). To test the importance of farnesylation to the ER localization of INF2, we mutated Cys1270 (C1270S) in full-length GFP-INF2. Transfected cells displayed apparently normal ER morphology by mCherrry-Sec61 staining, but GFP-INF2 C1270S was localized to the cytoplasm (Fig. 4). This result suggests that prenylation of INF2 is required for ER localization.

A C-terminal construct of INF2 (amino acids 993-1273) was sufficient for ER localization (Fig. 5). However, the ER staining of GFP-INF2-(C) appeared to overlay a hazy background, suggesting the presence of some cytoplasmic GFP-INF2-(C). To examine the strength of the INF2-(C) to ER interaction, we coexpressed GFP-INF2 constructs and mCherry-Sec61, then permeabilized cells with digitonin before fixation. GFP alone was depleted from cells after 2 minutes of digitonin permeabilization. Full-length GFP-INF2 maintained robust ER localization after 40 minutes of digitonin permeabilization (Fig. 6), similarly to endogenous INF2 (supplementary material Fig. S4). By contrast, GFP-INF2-(C) was depleted from ER after 10 minutes of digitonin permeabilization (Fig. 6), but maintained some ER staining after 2 minutes, suggesting weaker association than for full-length GFP-INF2.

Fractionation of 3T3 cell extracts further suggested a weak association of GFP-INF2-(C) with ER. In these experiments, we used a modified extraction procedure to that used in Fig. 1. Homogenization using a Type B dounce in buffer containing 50 mM KCl caused endogenous INF2 to partition entirely to the low-speed pellet, with undetectable levels in the high-speed pellet or high-speed supernatant. GFP-INF2 partitioned similarly (Fig. 7). By contrast, ∼50% of GFP-INF2-(C) partitioned to the high-speed supernatant. GFP alone partitioned almost 100% to the high speed supernatant (Fig. 7). These results suggest that GFP-INF2-(C) is more weakly bound to ER than is the full-length protein.

Previous studies have revealed little clear physical or functional association of actin with ER in mammals. Since INF2 is a potent actin polymerization and depolymerization factor, we investigated its cellular localization with respect to actin. The preponderance of other actin-based structures (notably stress fibres) hindered detection of possible phalloidin staining on the ER (Fig. 8A). Treating cells with latrunculin A for 30 minutes did not disrupt the reticular INF2 staining pattern (Fig. 8B), suggesting that actin filaments are not necessary for either ER integrity or INF2 reticular localization in these cells. Given the potency of INF2 polymerization and depolymerization activity in vitro, we sought other means to test whether INF2 could induce actin dynamics on the ER.

As for other formins, INF2 has an N-terminal DID and a C-terminal DAD, which are postulated to bind in an autoinhibitory interaction. Previously, we showed that the DAD is also an actin-monomer-binding WH2 motif, and that mutating three conserved leucine residues in DAD/WH2 abolished the actin depolymerization ability of INF2 without having any effect on actin polymerization (Chhabra and Higgs, 2006). To address INF2 regulation, we examined the DID-DAD interaction biochemically using bacterially expressed proteins. The INF2-(DID) construct binds the DAD-containing INF2-(C) construct with a K_d^app of 1.1 μM, from fluorescence anisotropy measurements, whereas the triple Leu→Ala mutation of DAD/WH2 rendered DID binding undetectable (Fig. 9A). Thus, the DAD/WH2 serves two roles: to accelerate actin depolymerization and to bind to DID. In actin polymerization assays, the presence of saturating DID had no effect on actin polymerization by INF2-(FH1-FH2-C), but strongly inhibited depolymerization (Fig. 9B). When depolymerization activity was examined directly, addition of DID inhibited FH1-FH2-C with an IC₅₀ of approximately 3 μM (Fig. 9C). These results suggest that the DID-DAD interaction of INF2 is fundamentally different from those of mDia1 or mDia2. Although DID-DAD binding blocks the polymerization activities of mDia1 and mDia2, it does not inhibit polymerization for INF2.

We next examined the effect of the DAD/WH2 mutation in full-length GFP-INF2 (GFP-INF2-W) expressed in 3T3 cells. GFP-INF2-W accumulated in crumpled tubule-like structures around the nucleus, with no peripheral staining in 55% of the cells at 24 hours after transfection (n=200 cells from three independent experiments) (Fig. 10A; supplementary material Fig. S6A). Most GFP-INF2-W-expressing cells round up and detach from the substratum after 48 hours. The collapsed GFP-INF2-W-containing tubules were Sec61-positive, suggesting that they originated from ER (Fig. 10A; supplementary material Fig. S6B). Peripheral Sec61 tubules were almost entirely abolished in GFP-INF2-W-expressing cells. However, rare regions of Sec61-positive and GFP-INF2-W-negative tubules existed in some cells (Fig. 10A).

Since the DAD/WH2 mutation abolishes actin depolymerization but not polymerization by INF2, we predicted that actin filaments might accumulate around the collapsed tubules. Indeed, the collapsed tubules stain heavily with rhodamine-phalloidin, with concomitant reduction of staining in other cellular actin-based structures (Fig. 10B,C). By contrast, wild-type GFP-INF2 localized largely in a reticular pattern, and actin filaments did not accumulate in regions positive for GFP stain (Fig. 10D).

To probe the role of INF2 in maintenance of ER morphology, we suppressed INF2 expression in 3T3 cells using shRNA. After 78 hours of shRNA treatment, INF2 levels were reduced to background, as determined by immunofluorescence (Fig. 11). A control shRNA construct did not result in INF2 reduction (not shown). INF2 knockdown had no discernible effect on ER morphology (Fig. 11) or dynamics (data not shown), as judged by mCherry-Sec61 staining in fixed or live cells. INF2 knockdown also had no discernable effect on actin filament or microtubule distribution, as judged by staining with TRITC-phalloidin or anti-tubulin, respectively (not shown).

In this paper, we show that INF2 is peripherally bound to the cytoplasmic leaflet of the ER. This interaction depends on post-translational modification by a C-terminal farnesyl group. An INF2 fragment containing the prenylation site localizes to ER but is not tightly bound, suggesting that stable association requires additional interactions. Membrane extraction experiments suggest that this interaction may be ionic. Our biochemical studies suggest that regulation of INF2 differs from that previously studied for mDia1 in that its polymerization activity is not autoinhibited by the DID-DAD interaction. However, the DID-DAD interaction does inhibit INF2's unique depolymerization activity, probably due to the fact that DAD is also an actin monomer-binding WH2 motif that is necessary for depolymerization. A point mutant that is inhibited for both the depolymerization activity and DID-DAD interaction causes ER collapse and actin filament build-up around the collapsed ER. However, suppression of INF2 does not result in altered ER morphology.

To our knowledge, this work represents the first evidence linking an actin polymerization factor to the ER in mammals. Previous work shows that myosin V localizes to ER fractions in neuronal cells (reviewed by Wagner and Hammer, 2003), and another study suggests a role for actin in ER retrograde flow (Terasaki and Reese, 1994). These findings provide hints that actin might polymerize and be used on the ER surface. We will discuss our results by addressing three questions.

The staining patterns of both endogenous INF2 and GFP-INF2 clearly indicate ER localization. We find no evidence of localization to other membranes. Membrane-extraction experiments show that the vast majority of INF2 is membrane bound. As with many peripheral membrane proteins (McLaughlin and Aderem, 1995), there appear to be two components to the membrane interaction: an ionic component and a hydrophobic component. The fact that INF2 can be extracted from membranes by treatments that disrupt ionic interactions (high ionic strength, high pH) suggest the ionic component, whereas the importance of the farnesyl group suggests the hydrophobic component.

Both endogenous INF2 and GFP-INF2 are strongly associated with ER, because both remain in the pellet fraction after homogenization and neither is extracted after extensive digitonin treatment that efficiently extracts cytoplasmic proteins. The GFP-INF2-(C) construct, containing the farnesylation site as well as a poly-basic region, localizes to ER but is easily extracted by digitonin treatment. Two possibilities for stable association are: (1) dimerization of full-length INF2 through the FH2 domain, because dimerization is expected to increase binding affinity exponentially for a membrane-associated protein (Klein et al., 1998); or (2) additional interaction affinity provided by regions outside the C-terminus.

We were surprised that the INF2 DID was unable to inhibit polymerization acceleration by the DAD-containing INF2 (FH1-FH2-C), even though DID binds tightly to DAD. This result is fundamentally different from the autoregulatory mechanism established for mDia1, in which the DID-DAD interaction inhibits the FH2 domain polymerization activity (Li and Higgs, 2003; Li and Higgs, 2005). One possibility is that there is an additional interaction between DID and FH2 for mDia1, this second interaction being inhibitory, whereas the DID-DAD interaction provides high affinity. Indeed, mDia1 DID is able to inhibit a DAD-less mDia1 FH2 construct, but with an IC₅₀ 10,000-fold higher than that observed when DAD is present (Li and Higgs, 2005). If a direct DID-FH2 interaction is necessary for inhibition, we would predict that this interaction does not occur for INF2.

However, it is unlikely that INF2's polymerization activity is constitutively active in cells, so other inhibitory mechanisms are likely to exist, such as inhibitory binding by another molecule. We do not rule out the possibility that our biochemical data might not recapitulate the situation in the full-length protein, in that the inhibitory interaction might depend on the N-terminus being in cis with the FH2 domain. Alternatively, inhibition could depend on farnesylation or another post-translational modification not present in our bacterially expressed proteins.

In contrast to the INF2 polymerization activity, its depolymerization activity is inhibited by the DID-DAD interaction. This inhibition is reasonable in view of our previous biochemical studies showing that the DAD also serves as an actin-monomer-binding WH2 motif, which is crucial for the depolymerization activity (Chhabra and Higgs, 2006). DID binding to the DAD/WH2 would compete for actin-monomer binding, thus blocking depolymerization.

Since DID and DAD do interact and block the depolymerization activity, a mechanism presumably exists for disruption of DID-DAD. The mDia formins bind Rho family GTPases, the most well studied being the interaction of RhoA or C with mDia1 (Goode and Eck, 2007; Lammers et al., 2005; Li and Higgs, 2003; Rose et al., 2005). This interaction involves regions in the DID as well as a region N-terminal to DID (Lammers et al., 2005; Li and Higgs, 2005; Rose et al., 2005). Similar N-terminal sequences exist in several other formins shown or expected to bind Rho GTPases, including mDia2, mDia3, FRL1-FRL3 and DAAM1 and DAAM2. In INF2, the DID is at the extreme N-terminus, starting at amino acid 5 by our sequence analysis. Therefore, INF2 appears to lack the N-terminal Rho interaction region found in these other formins. For this reason, we predict that relief of the DID-DAD interaction takes place by a different mechanism. In that context, the N-terminus of DAAM1 does bind RhoA, but relief of autoinhibition occurs by Dishevelled binding to the DAD region (Liu et al., 2008).

When expressed in cells, full-length INF2 containing the DAD/WH2 mutation causes ER collapse and a build-up of actin filaments around the collapsed ER. Our biochemical results provide a possible interpretation for this cellular result. The mutation blocks both actin monomer binding by WH2 and DID-DAD interaction biochemically. In the absence of the actin-monomer binding, INF2 is not able to depolymerize filaments. A fraction of the mutant INF2 gets activated for polymerization, through relief of the as yet unknown inhibitory interaction on the FH2 domain. This activated INF2 is unable to depolymerize the filaments it assembles, which might cause filaments to build up on ER, disrupt its intricate reticular structure, and cause it to collapse on the nucleus. Although this result provides further evidence that INF2 is an ER-binding protein, it does not address INF2 function, because the ER-collapse phenotype probably represents the consequence of overexpressing an unregulated actin-polymerization factor on the ER surface.

We suspect that the function of INF2 is related to its unique actin polymerisation and actin depolymerization activity, providing it with the ability to assemble highly transient filaments. One intriguing issue is how polymerisation and depolymerization would occur in the context of ER-bound INF2, in which ER binding depends partly on the C-terminal prenyl group. How does ER binding affect the depolymerization activity? Does the same INF2 molecule that nucleates a filament also control this filament's depolymerization? Answers to these questions would have strong implications for cellular function.

Our shRNA result suggests that INF2 is not involved in maintenance of gross ER structure, because the reticular network appears to be intact in these cells. One possibility is that INF2 is functionally redundant. Mammals possess two INF isoforms, INF1 and INF2. The classification of both proteins as INF proteins is based on phylogenetic similarity between their FH2 domains, compared with the 13 other mammalian formins (Higgs and Peterson, 2005). However, INF1 and INF2 differ significantly outside their FH2 domains in several respects: (1) INF1 contains no DID or other sequence N-terminal to its FH1 domain; (2) INF1 contains a significantly longer region C-terminal to the FH2 (671 amino acids in mouse INF1 versus 299 amino acids in mouse INF2); and (3) INF1 does not possess a DAD or a WH2. Recent work shows that INF1 localizes not to ER but to microtubules in cells, by virtue of two microtubule-binding motifs not present in INF2 (Young et al., 2008). We find no clear evidence of microtubule localization for INF2. Thus, we doubt that INF1 and INF2 are functionally redundant.

A number of possible INF2 functions remain to be tested, including a role in retrograde ER movement, because actin appears to have a role in this process (Terasaki and Reese, 1994; Waterman-Storer and Salmon, 1998). Finally, another actin polymerization factor, the Arp2/3 complex, is implicated at a number of locations in the secretory pathway, including the Golgi (Egea et al., 2006; Hehnly and Stamnes, 2007), the trans-Golgi network (Cao et al., 2005) and the ERGIC and cis-Golgi (Campellone et al., 2008). Future work will examine the role of INF2 in ER-related processes.

INF2 constructs (Fig. 1A) were generated by reverse transcriptase PCR from RNA isolated by the TRIzol method (Invitrogen) from 300.19 murine Abelson leukaemia-virus-transformed pre-B cells (Alt et al., 1981). cDNA was synthesized using oligo (dT) primer and SuperScript II reverse transcriptase (Invitrogen). INF2-(DID) (BamHI site) and INF2-(C) (HindIII-EcoRI sites) was cloned into pGEX-KT vector (Hakes and Dixon, 1992) for bacterial expression as glutathione S-transferase fusion proteins. INF2-(FH1-FH2-C) was cloned into the NdeI-EcoRI sites of pET22b (Novagen) for expression as an untagged protein. Constructs to be expressed in 3T3 cells were cloned into the BglII site of the eGFP-C1 vector (Clontech). Point mutagenesis was conducted using the QuikChange kit (Stratagene) or the Change-It kit (USB). Three conserved leucine residues (amino acids 1008, 1009 and 1018) in the WH2 motif were converted to alanine in the WH2 mutants, and C1270 was converted to serine in the CAAX mutant. The GFP-Sec61β and mCherry-Sec61β constructs were kind gifts from Thomas Rapoport (Harvard Medical School) and Jennifer Lippincott-Schwartz (NIGMS), respectively.

Rosetta 2 DE3 Escherichia coli (Novagen) were used for expression of all constructs and expression was induced as described previously (Chhabra and Higgs, 2006). INF2-(C) and INF2-(FH1-FH2-C) were purified as previously described (Chhabra and Higgs, 2006). INF2-DID was purified by glutathione-Sepharose 4B chromatography (GE Healthcare) followed by removal of the glutathione S-transferase. Purified INF2-(DID) was dialyzed in Na150MEPD and stored at 4°C or -70°C.

Rabbit muscle actin was purified from acetone powder and labeled with pyrenyliodoacetamide (Spudich and Watt, 1971). Both labeled and unlabeled actin were gel-filtered on an S-200 column (MacLean-Fletcher and Pollard, 1980), which was essential for obtaining reproducible kinetic results.

The following buffers were used frequently: G buffer (2 mM Tris, pH 8, 0.5 mM DTT, 0.2 mM ATP, 0.1 mM CaCl₂, and 0.01% NaN₃); G-Mg buffer (same as G buffer but with 0.1 mM MgCl₂ instead of CaCl₂); 10 × KMEI: 500 mM KCl, 10 mM MgCl₂, 10 mM EGTA, and 100 mM imidazole, pH 7.0; and polymerization buffer (G-Mg buffer plus 1 × KMEI).

Pyrene-labeled and unlabeled actin was mixed in G buffer to produce an actin stock of 5 μM containing 5% pyrene-labeled actin. This stock was converted to Mg²⁺ salt by incubation at 23°C for 2 minutes in 0.1 volume of 10E/1M immediately before polymerization. Polymerization was initiated by the addition of INF2 in a solution of KMEI and G-Mg to result in a final concentration of 1 × KMEI. Care was taken to ensure that none of the added proteins significantly altered the ionic strength or pH of the reaction. Pyrene fluorescence (excitation 365 nm, emission 407 nm) was monitored in a PC1 spectrofluorometer (ISS, Champaign, IL) or an LS50B spectrofluorometer (PerkinElmer Life Sciences). The time between final mixing and the start of data collection ranged between 10 and 15 seconds for each assay. For measuring the effect of DID on FH1-FH2-C polymerization, FH1-FH2-C was pre-mixed with desired concentration of INF2-DID in polymerization buffer, prior to mixing with actin monomers.

Fluorescence labeling techniques used were described previously (Chhabra and Higgs, 2006). INF2-(C) was dialyzed for 2 hours in 50 mM NaCl, 2 mM NaPO₄, pH 7, 0.5 mM MgCl₂, 0.5 mM EGTA. Fluorescein 5′-maleimide (Molecular Probes) was added to 100 μM, and the reaction was incubated on ice for 5 minutes. Labeling was stopped by adding DTT to a concentration of 10 mM. Labeled protein was separated from free dye by Superdex 200 10/30-size exclusion chromatography (GE Healthcare). For measuring INF2-(C) interaction with INF2-(DID), fluorescein-labeled INF2-(C) was diluted to 200 nM in 10 × K50MEI-G-Mg containing 50 μM thesit (Sigma). A stock of INF2-(DID) was prepared in N2MPED (2 mM NaCl, 0.3 mM DTT, 0.066 mM MgCl₂, 0.066 mM EGTA). To 10 μl of FL-INF2-(C), 90 μl of INF2-(DID) solution was added, and fluorescence anisotropy was measured using a PC1 spectrofluorometer (ISS) with 495 nm excitation wavelength and a 520±5 band-pass emission filter (Omega Optics).

Actin (1.05 μM, 5% pyrene) was polymerized for 24 hours at 23°C in polymerization buffer in the dark. Actin stock (95 μl) was gently mixed with 5 μl of INF2 solution in polymerization buffer. Pyrene fluorescence was monitored within 15 seconds of dilution. For measuring the effect of DID on depolymerization, FH1-FH2-C was pre-mixed with desired concentration of INF2-(DID) in polymerization buffer.

Swiss 3T3 cells line was obtained from Jonathan Chernoff (Fox Chase Cancer Centre) and cultured in DMEM with 10% calf serum and 50 IU/ml penicillin-streptomycin solution. Cells were grown at 37°C in 5% CO₂.

Cells were either electroporated or transfected using Effectene Transfection Reagent (Qiagen). For electroporation cells were mixed with desired DNA in a ratio of 1.5 × 10⁶ cells per 50 μg DNA. 400 μl cells in 1 × DPBS (containing Ca²⁺ and Mg²⁺) was added to cuvettes containing DNA and gently pipetted up and down twice. Cells were electroporated at 300V, 0.975 capacitance (BioRad Gene Pulser II). Immediately after electroporation 1 ml pre-warmed 3T3 media was added to the electroporated cells. After 5 minutes of recovery, cells were plated on 12-mm-diameter acid-washed glass coverslips placed in 24-well plates or on 10 mm Petri plates, as required. After 4 hours of incubation at 37°C and 5% CO₂, medium was refreshed.

For Effectene transfection, 3T3 cells were trypsinized, washed twice in 1 × PBS and 2 × 10⁴ cells seeded onto 24-well plates on 12 mm-diameter glass acid-washed coverslips. Cells were incubated overnight at 37°C, 5% CO_2. Next day 0.8 μg DNA was mixed with Effectene reagent according to the manufacturer's instructions and added to cells. After 6 hours of incubation, cells were washed 1 × PBS and supplemented with fresh medium. Cells were incubated for another 6 hours under normal growth conditions followed by fixation and staining.

Cells were fixed with formaldehyde (4% formaldehyde, 0.1% saponin in 1 × PBS) for 1 hour at room temperature. After washing with PBS and blocking coverslips with blocking buffer (10% calf serum, 0.1% saponin, 1 × PBS, 0.02% sodium azide) for 30 minutes at room temperature, cells were stained for INF2, actin, tubulin or GRP94 (as required). For INF2 staining, either unlabeled rabbit anti-INF2 at 5 μg per coverslip or Cy3-labeled rabbit anti-INF2 at 0.5 μg/coverslip was used. Actin was stained using 0.1 μM TRITC-phalloidin (Sigma). Tubulin was stained using the DM1-alpha mouse monoclonal (Sigma) at 1:10,000 dilution. GRP94 was stained using a rabbit polyclonal antibody from Stressgen (SPA-851). All primary staining was conducted for 1 hour at room temperature. Secondary antibodies used were FITC-conjugated anti-rabbit IgG (1:600, Invitrogen), FITC and Alexa Fluor 680-conjugated anti-mouse IgG (1:600). For nuclear staining DAPI (4′,6-diamino-2-phenylindol) was used for 10 minutes at 1:250,000 dilution. In cases where Cy3-labeled anti-INF2 was used, cells were treated subsequent to the secondary antibody incubation for 1 hour. Nikon-Eclipse TE-2000 microscope with an X-Cite Exfo 120 watt mercury arc lamp and a 60 × 1.4 numerical aperture objective was used and images were acquired using a Roper Cool-Snap camera using MetaVue software (Universal Imaging). The filter sets used were as follows, with excitation wavelength, emission wavelength, dichroic: UV, 350/50, 460/50, 400 DCLP (31000v2 Chroma); `green', 480/30, 535/40, 505 DCLP; (31001 Chroma); `red', 540/25, 620/60, 565 DCLP (31002, Chroma); `far red', 620/60, 700/75, 660 LP (41008, Chroma).

The method to partially permeabilize ER (Fig. 3) was adapted from a published method (Lin et al., 1999). 3T3 cells were plated onto 12 mm glass coverslips and incubated overnight. Coverslips were washed several times in PBS, then fixed with 4% formaldehyde in PBS for 10 minutes at 23°C. After three more washes with PBS, cells were permeabilized in buffer (0.3 M sucrose, 2.5 mM MgCl₂, 0.1 M KCl, 1 mM EDTA, and 10 mM PIPES at pH 6.8) containing either 5 μg/ml digitonin (50% by TLC, Sigma D1407) for 5 minutes on ice or 0.25% Triton X-100 for 10 minutes at 23°C. After permeabilization, coverslips were processed for fluorescence microscopy as described above.

The method to remove cytosolic molecules (Fig. 6) used digitonin prefixation. 3T3 cells were plated onto 12 mm glass coverslips and incubated overnight. Coverslips were washed several times in iced PBS, then extracted with EB with 0.15 mM digitonin (Calbiochem 300410) on ice for the indicated time. Cells were then fixed and processed for fluorescence microscopy as described above.

To generate polyclonal antibodies against INF2, purified INF2-(C) (amino acids 993-1273) was injected into rabbits or chickens by Covance (Denver, PA). A second antibody against INF2-(FH1-FH2) (amino acids 538-993) was raised in chickens by Aves. Polyclonal antibodies were affinity purified by using INF2-(C) or INF2-(FH1-FH2) coupled to Sulfolink (Pierce). Serum was mixed with resin overnight in the presence of 0.15 thesit. After washing with PBS with 500 mM NaCl, then with PBS, antibodies were eluted with 200 mM glycine-HCl pH 2.5. Eluted antibodies were dialyzed into PBS with 0.02% sodium azide. Purity of antibody was verified by western blot with both recombinant protein and 3T3 cell extract. Antibody was directly labeled with Cy3-succinimide following manufacturer's protocol (GE Biosciences).

To make Swiss 3T3 extract, cells at 80-90% confluency were lysed by adding 0.5 ml lysis buffer (50 mM HEPES pH 7.5, 4% SDS, 300 mM NaCl, 1 mM EDTA, 5 mM DTT, 1 tablet/50 ml Roche Complete protease inhibitors without EDTA) followed by scraping with a rubber policeman. Solution was sheared through a 25-gauge needle and boiled for 5 minutes. This solution was cooled to 23°C, then 1/10 volume of 300 mM of freshly made N-ethylmaleimide (NEM, Pierce) in water was added. Just before SDS-PAGE, the protein sample was mixed 1:1 with 2 × DB (250 mM Tris-HCl pH 6.8, 2 mM EDTA, 20% glycerol, 0.8% SDS, 0.02% bromophenol blue, 1000 mM NaCl, 4 M urea). Proteins were then separated on 7.5% SDS gel and transferred to a PVDF membrane (polyvinylidine difluoride membrane-Millipore). The membrane was blocked with 3% BSA containing TBS-T (20 mM Tris-HCl, pH 7.6, 136 mM NaCl, and 0.1% Tween-20) for 30 minutes, then incubated with anti-INF2 antibody solution at 4°C overnight. After washing with TBS-T, the membrane was incubated with horseradish peroxidase (HRP)-conjugated secondary antibody for 1 hour at room temperature. Signals were detected by Chemiluminescence (Pierce).

To make subcellular membrane fractions (entire procedure conducted on ice or at 4°C), 3T3 cells at approximately 80% confluence were washed several times in iced PBS, then once with EB (50 mM HEPES pH 7.5, 5 mM EGTA, 1 mM MgCl₂, 1 mM DTT, 1 tablet/50 ml Roche complete protease inhibitors without EDTA), then scraped in 1 ml EB per 100 mm plate. Cells were homogenized 30 times in a stainless steel Dura-Grind (Wheaton). KCl was added to 50 mM after homogenization. The homogenate was centrifuged for 10 minutes at 2000 × g max in a swinging bucket rotor. The supernatant (LSS) was centrifuged for 20 minutes at 100,000 rpm (436,000 × g max) in a TLA120.2 centrifuge (Beckman). Both pellets from the 2000 × g spin (LSP) and 436,000 × g spin (HSP) were washed in EB with 50 mM KCl, then resuspended in EB with 50 mM KCl. In the experiments conducted in Fig. 7, the extraction procedure was modified in two ways: (1) 50 mM KCl was present in EB throughout the procedure; and (2) a Type B dounce (Wheaton) was used instead of the Dura-Grind. To prepare samples for SDS-PAGE, 50 μl sample was mixed with 34 μl of 10% SDS and 1 μl of 1 M DTT, boiled for 5 minutes, cooled to 23°C, then subjected to NEM and subsequent treatments as described above. We found that these sample preparation conditions were essential to obtain a single protein band by western blotting. If extracted in normal SDS-PAGE buffer (containing 0.4% SDS, with no cysteine alkylation), we obtained multiple bands of apparent lower mass, as well as significant staining at very high apparent mass.

Mouse tissues were removed from freshly killed adult mice and homogenized in EB (10 vol EB per g tissue) using a polytron followed by Type B dounce (Wheaton). Homogenate was subjected to SDS-NEM treatment as described for 3T3 fractions.

A similar procedure to that used recently for Arp2/3 complex in 3T3 cells was used (Nicholson-Dykstra and Higgs, 2008). Briefly, protein extracts were made from a known number of Swiss 3T3 cells grown to subconfluence. Bradford assays of these extracts revealed that Swiss 3T3 cells contain 0.7 ng protein/cell (0.685, 0.693 and 0.710 ng/cell from three independent extracts). Western blots were run using 1 μg 3T3 cell extract and varying ng amounts of mouse INF2-(C), which was the epitope against which the antibody was raised. INF2-(C) was included in the same lanes as the extract, to serve as an internal control. Densitometric analysis revealed that western blot bands were linear for endogenous INF2 from 0.25-2 μg cell extract, and for INF2-(C) from 0.025-0.2 μg. We used 1 μg cell extract for quantification. Mean cellular (2.85 × 10^-12 l) and nuclear volume (0.59 × 10^-12 l) were calculated from mean diameters of cells (17.6 μm) and nuclei (10.4 μm) measured from DAPI and TRITC-phalloidin images of trypsinized cells.

Swiss 3T3 cells were cotransfected with full-length GFP-INF2 and mCherry Sec61 by electroporation. 16 hours after transfection, cells were trypsinized and plated on to 12-mm-diameter glass coverslips coated with 0.1% poly-L-lysine. Coverslips were sealed in a modified Rose chamber using VALAP sealant. Cells were supplemented with fresh phenol red lacking 3T3 medium. Oxyrase was added to the medium at 1:100 dilution, to reduce photo bleaching. The imaging chamber was then sealed from the top with another coverslip. Images were taken at 15 second intervals for 3 or 5 minutes using Nikon-Eclipse TE-2000 microscope with a 60 × 1.4 NA objective. Images were acquired using a Roper Cool-Snap camera and processed using MetaVue software (Universal Imaging).

In Swiss 3T3 cells, INF2 gene was silenced by shRNA strategy, using the pSuperior-retro-neo-GFP vector (OligoEngine). The target sequence used was 5′-GAATTTCCTCAACTACGGC-3′. Cells were electroporated with either empty plasmid or plasmid containing the shRNA. At 56 hours post-electroporation, cells were transfected with mCherry-Sec61 using Effectene transfection reagent (Qiagen). Eighty hours post-shRNA transfection; cells were fixed with formaldehyde and stained for INF2 using anti-INF2. Alternately, cells were imaged live for mCherry-Sec61 dynamics, as described above.

Mouse brain (12 adult brains) was extracted and fractionated by differential centrifugation as described above. The high-speed pellet was solubilized in EB containing 2% thesit, then immunoprecipitated using 50 μg anti-INF2 and 50 μl protein-A-Sepharose (GE Biosciences). INF2 was further purified by 4-15% reducing SDS-PAGE, visualized by Coomassie staining, and the band corresponding to INF2 was excised from the gel and digested with trypsin per established in-gel digestion procedures. The resulting gel extracts were dried in a vacuum centrifuge, resuspended in 2.5% formic acid and 15% acetonitrile solution and introduced to the LC-MS/MS platform by way of a capillary autosampler (FAMOS, LC Packings, San Francisco, CA). The peptide mixture was separated online by in-house fabricated microcapillary (180 mm × 0.125 mm, Sutter Instruments P-2000, San Francisco, CA) reverse-phase (C18-AQ Reprocil-Pur 3 μm, 200 Å pore, Dr Maisch, Ammerbuch-Entringen, Germany) chromatography (Agilent 1200 capLC, Aglient, Palo Alto CA) from 20% acetonitrile, 0.1% formic acid to 80% acetonitrile, 0.1% formic acid over 60 minutes. The capillary LC eluent was directed into an LTQ-Orbitrap mass spectrometer set to acquire five different scans in looped series: a full MS1 scan (200-1000 m/z, R≥60,000), and four directed ion trap MS2 scans corresponding to each of four different predicted tryptic peptide forms of C-terminal, farnesylated INF2 ([RLfC-OH + 2H⁺]2+: 298.2 m/z; [RLfC-OCH3 + 2H⁺]²⁺: 305.2 m/z; [LfC-OH + H⁺]⁺: 439.3 m/z and [LfC-OCH3 + H⁺]⁺: 452.3 m/z, where `f' represents farnesyl group). The resultant Orbitrap MS1 ion chromatograms were extracted to high precision (±1.5 p.p.m.) of the theoretical exact masses (NIST, http://physics.nist.gov/) of the parent ions described above.