Building bridges: formin1 of Arabidopsis forms a connection between the cell wall and the actin cytoskeleton


(fax +44 1865 483242; e-mail


Actin microfilament (MF) organization and remodelling is critical to cell function. The formin family of actin binding proteins are involved in nucleating MFs in Arabidopsis thaliana. They all contain formin homology domains in the intracellular, C-terminal half of the protein that interacts with MFs. Formins in class I are usually targeted to the plasma membrane and this is true of Formin1 (AtFH1) of A. thaliana. In this study, we have investigated the extracellular domain of AtFH1 and we demonstrate that AtFH1 forms a bridge from the actin cytoskeleton, across the plasma membrane and is anchored within the cell wall. AtFH1 has a large, extracellular domain that is maintained by purifying selection and that contains four conserved regions, one of which is responsible for immobilising the protein. Protein anchoring within the cell wall is reduced in constructs that express truncations of the extracellular domain and in experiments in protoplasts without primary cell walls. The 18 amino acid proline-rich extracellular domain that is responsible for AtFH1 anchoring has homology with cell-wall extensins. We also have shown that anchoring of AtFH1 in the cell wall promotes actin bundling within the cell and that overexpression of AtFH1 has an inhibitory effect on organelle actin-dependant dynamics. Thus, the AtFH1 bridge provides stable anchor points for the actin cytoskeleton and is probably a crucial component of the signalling response and actin-remodelling mechanisms.


Actin microfilaments (MF or F-actin) are a major component of the cytoskeleton and are essential for many cellular processes including division, expansion, differentiation, and are components of cellular response mechanisms. Within plant cells, most of the dynamic processes, e.g. organelle movement, vesicle trafficking and cytoplasmic streaming occur via interaction with MFs. Plant MFs themselves are extremely dynamic structures with a rate of polymerization of up to 1.7 μm sec−1 (Staiger et al., 2009; Smertenko et al., 2010). To make new filaments, actin nucleation proteins are required.

One of the main gene families involved in this function is formin. The formin gene family is very diverse, with more than 20 genes in A. thaliana (Deeks et al., 2002; Cvrckova et al., 2004b; Blanchoin and Staiger, 2010). They all share a formin homology domain (FH2) in the C-terminal half of the protein which is involved in nucleating MFs. In monocotyledonous and dicotyledonous plants, the formin gene family is organized into two classes. Most class I formins have a membrane targeting domain in the N-terminal region of the protein. This domain is composed of a signal peptide followed by a transmembrane domain and appears similar to that of type I membrane proteins in general. Formins in class I are in most cases targeted to the plasma membrane (PM) as has been demonstrated for AtFH1, AtFH5 and AtFH6 by GFP fusions (Cheung and Wu, 2004; Favery et al., 2004; Van Damme et al., 2004), and by immuno-localization for AtFH4 (Deeks et al., 2005). Class II formins (AtFH12AtFH21) are characterized by a phosphatase tensin (PTEN)-like domain at the N-terminus. This domain is hypothesized to mediate membrane targeting and probably determines subcellular localization of the protein (Cvrckova et al., 2004a; Vidali et al., 2009). Vidali et al. (2009) have shown by RNA silencing that class II formins are involved in polarized growth in moss (Physcomitrella pattens). In vascular plants, AtFH14 has recently been shown to be crucial for cell division and is able to induce co-alignment between MFs and microtubules (Li et al., 2010).

The formin homology domains (FH1 and FH2) of the Arabidopsis class I formin proteins AtFH1, AtFH3, AtFH4, AtFH5 and AtFH8, are able to nucleate actin filaments in vitro (Deeks et al., 2005; Michelot et al., 2005; Yi et al., 2005; Ye et al., 2009). The MF nucleation mechanism of AtFH1 has been described (Michelot et al., 2006). AtFH1 is a non-processive formin, it becomes localized to the side of the MF after a nucleation event and so, by repeated nucleation, AtFH1 causes formation of bundles of MFs during in vitro assays (Michelot et al., 2006). In vivo, overexpression of AtFH1, AtFH3 and AtFH4 induces formation of an excess of MFs at the cell cortex. Overexpression of AtFH8 induces aberrant root hair development (Deeks et al., 2005; Yi et al., 2005), and overexpression of AtFH1 and AtFH3 has an effect on pollen tube growth (Cheung and Wu, 2004; Ye et al., 2009). Down-regulation of AtFH3 by RNAi results in a decrease in the numbers of MFs and a reduction of cytoplasmic streaming in pollen tubes (Ye et al., 2009). More specific functions have been described for some formin genes. AtFH5 is localized at the cell plate and its loss of function results in impaired cytokinesis (Ingouff et al., 2005). More recently, AtFH5 has shown to be involved in pollen tube polarization (Cheung et al., 2010). Finally, RNA expression data show that AtFH1, AtFH6 and AtFH10 are up-regulated during nematode infection which predicts that Class I formins are involved in response to pathogen attack (Favery et al., 2004) and generally have a signalling function in environmental response pathways.

In animal cells, numerous studies describe protein complexes involved in interaction between the extracellular space of the cell and the actin cytoskeleton. One of the best described is the integrin/actin interaction, which is involved in mechanisms such as focal adhesion. There is no reason to suspect that interaction between cytoskeletal systems and the extracellular matrix – mediated by proteins with extracellular domains – does not occur in plant cells (Baluska et al., 2003). In fact, a rapid cytoskeleton, endoplasmic reticulum and Golgi apparatus reorganization occurs after extracellular stimulation (Hardham, 2007). Similar observations have been made during pathogen infection in which fungus attack induces a strong and fast reorganization of the actin cytoskeleton (Koh et al., 2005; Hardham, 2007).

According to protein topology prediction, class I formins have an extracellular domain. In the case of AtFH1, it is comprised of nearly 100 amino acids. For this reason, type I formins are good candidates for linking between the outside of cells and the actin cytoskeleton, and might be involved in signal transduction through the PM (Deeks et al., 2002). Alignment of formin 1 proteins from nine dicotyledonous and monocotyledonous plants enabled us to find a conserved region which suggests conserved functions for these domains (Figure S1). This study is focused on the potential role of the membrane targeting signal and extracellular domain in the function of formin1 of Arabidopsis (AtFH1). By resolving the topology, we first show that AtFH1 has an extracellular domain with subdomains conserved by purifying selection. Then, we show several proofs of interaction between this region and the extracellular matrix of the cell. Finally, we demonstrate a correlation between anchoring of AtFH1 within the extracellular matrix and its actin polymerization activity.


The extracellular domain of Arabidopsis formin 1 evolved under strong purifying selection

The amino acid alignment of a set of plant formin1 proteins clearly reveals a conserved region in the potential extra cellular domain (Figure S1). Estimation of the nonsynonymous to synonymous substitution ratio (ω = dN/dS ratio) was used to quantify the selective pressure acting on the AtFH1 gene and orthologous sequences in plants. Genes or domains with no function evolve neutrally (ω = 1), while genes evolving under selection have ω≠1. Positive selection (ω > 1) is observed when amino acid replacement is favored by natural selection. Purifying selection (ω < 1) is observed when amino acid replacement is counter-selected. Evolution under strong purifying selection pressure tends to have low values of ω, typically a few percent.

The ω estimation of the putative extracellular domain studied here (amino acids 1–107) was 0.070, that of the intermediate protein region (amino acids 108–520) was 0.200, that of the 3′ part of the gene which included the FH1 and FH2 conserved domains (amino acids 521–1051) was 0.078. This result indicates that the putative extracellular domain evolved under strong purifying selection of the same order of magnitude as the intracellular region which contains the conserved FH1 and FH2 actin-interacting domains and demonstrates the evolutionary importance of the extracellular domain for AtFH1 function in plants.

Arabidopsis formin1 is targeted to the plasma membrane and is excluded by cortical microtubules

We decided to confirm the sub-cellular localization of AtFH1 by fluorescent protein fusion. AtFH1 cDNA was cloned as a C-terminal fusion with fluorescent proteins (GFP, YFP and RFP) under control of the 35S promoter and expressed transiently in tobacco (Nicotiana tabacum) leaf tissue (Figure 1a). This experimental system was selected as it was already known that AtFH1 has a dramatic effect on MF organization and it has proven difficult to generate stably transformed Arabidopsis plants (Banno and Chua, 2000; Cheung and Wu, 2004). AtFH1–GFP labelled very clearly the outline of leaf epidermal cells, produced a sheet of fluorescence at the cell surface (Figure 1b) and co-localized with a known plasma membrane protein, PIP2;1–CFP (Figure S2) thus demonstrating that AtFH1 is targeted to the PM. We also tested whether AtFH1 is targeted to the PM via the secretory pathway. Inhibition of ER/Golgi transport with brefeldin A or with Sar1-GTP locked (daSilva et al., 2004) induced an accumulation of AtFH1 in the ER (Figure S3). AtFH1 is therefore co-translated inserted into the ER membrane and then targeted to the PM. Curiously, AtFH1–GFP does not have an homogeneous distribution at the PM as unlabeled stripes were visible (Figure 1b, arrow). The pattern of non-fluorescent stripes in the PM resembled microtubules (MT) in shape and quantity. To verify this hypothesis, AtFH1–YFP was co-expressed with the MT labelling construct CFP–TuA6. Microtubules were present exactly underlying the non-fluorescent pattern of AtFH1–YFP (Figure 1c). Moreover, when the AtFH1–GFP tobacco leaves were incubated with 20 μm oryzalin, a microtubule-depolymerizing drug, the non-fluorescent stripes disappeared (Figure S4). Actin MFs are not involved in formation of this pattern as MF depolymerization with latrunculin B did not abolish the stripes (Figure S4).

Figure 1.

 AtFH1 is localized to the plasma membrane and marks it in a non-homogeneous way.
(a) Schematic drawing of the different constructs used in the study. FORMIN1 of Arabidopsis thaliana (AtFH1) contains a transmembrane domain (TM) and formin homology domains (FH1 and FH2) within the intracellular part of the protein (not to scale), and a series of conserved domains (Domains A–D) within the extracellular region. SP, signal peptide; FP, green, yellow, or red fluorescent protein.
(b) Cortical section of AtFH1–GFP expressed in tobacco leaf epidermal cells. Arrow points to one of the unlabeled structures that resemble microtubules in the fluorescent membrane.
(c) The stripe pattern of AtFH1–YFP labelling (magenta) and the stripes visible when microtubules are labelled with CFP–TuA6 (green) are exactly coincidental when overlain. Scale bars = 10 μm.

AtFH1 has a large extracellular domain which interacts with the cell wall

The predicted intracellular part of the protein was removed by cloning the first 140 amino acids of the N-terminus of the protein as a fusion with GFP (AtFH1-dFH–GFP) (Figure 1a). AtFH1-dFH–GFP-expressing cells were plasmolyzed and GFP fluorescence was observed in the withdrawn PM and in Hechtian strands but not in the apoplastic space (Figure 2a, *). AtFH1-dFH–GFP labelled the PM and proved that the first 140 amino acids are sufficient for PM targeting and protein anchoring in the PM. We then removed the predicted transmembrane domain by fusing only the N-terminal 107 amino acids of AtFH1 with RFP (AtFH1-107–RFP). As predicted, the truncated protein was not able to remain fixed within the PM and was mainly localized instead at the cell wall (Figure 2b). We conclude that AtFH1 has an extracellular domain at its N-terminus of around 100 amino acids. Interestingly, in plasmolyzed tissue, AtFH1-107–RFP cells show labelling of the cell wall (as opposed to the apoplastic space created between cell wall and PM by plasmolysis) which was never observed when cells expressed the secreted form of RFP (compare Figure 2b,c, arrows). This finding suggests a potential interaction between the extracellular domain of AtFH1 and the cell wall.

Figure 2.

 AtFH1 interaction with the cell wall in plasmolysis experiments.
(a–c) Cells expressing different fluorescent protein labelled constructs were incubated with 0.5 m mannitol to induce plasmolysis.
(a) The formin deletion construct with intracellular domain removed (AtFH1-dFH–GFP) labels the plasma membrane and Hectian strands within the apoplastic space created by plasmolysis (*). The cell wall (arrow) is unlabelled.
(b) In cells expressing only the N-terminal 107 amino acids of AtFH1 (AtFH1-107–RFP) which does not include the transmembrane domain, the protein remains for the most part within the cell wall (arrow) and marks the apoplastic space (*) and plasma membrane weakly.
(c) Secreted RFP (Sec–RFP) filled the apoplastic space (*) while fluorescence was not apparent within the cell wall (arrow). Scale bars = 10 μm.

AtFH1 is immobilized within the PM by a cell wall/plasma membrane connection

To test the lateral mobility of AtFH1 as a measure of its binding to other cellular components, we performed Fluorescence Recovery After Photobleaching (FRAP) experiments over a short time scale (1 or 2 mim) to exclude any side effects that might result from endo-, or exocytotic removal or insertion of protein from the membrane (Kleine-Vehn et al., 2008; Men et al., 2008). AtFH1–GFP produced a very unusual FRAP recovery curve. Approximately 20% of initial fluorescence was recovered in the bleached region within 10–20 sec but then nearly no fluorescence increase was recorded for the next 100 sec (Figure 3a,c; dashed blue line). The plateau value of fluorescence intensity recovery was designated I100s and taken to represent the mobile fraction of AtFH1–GFP. AtFH1–GFP, therefore, has a surprisingly low mobile fraction (I100s = 20.2 ± 5.1%, Figure 3c and Table 1). Indeed, we expected from observations of other plasma-membrane anchored proteins (such as GFP-LTI6b in Figure 3b) that AtFH1 would laterally diffuse in the membrane and have a much higher I100s. One hypothesis to explain the low mobility of AtFH1–GFP is that it interacts with other proteins or structures which constrain its mobility. Formins are known to interact with the cell cytoskeleton (Michelot et al., 2005; Deeks et al., 2010; Li et al., 2010) so interaction with MF or microtubules might reduce the lateral mobility of AtFH1. To test this hypothesis, we treated AtFH1–GFP-expressing cells with latrunculin B and oryzalin to depolymerise both actin and microtubule networks. These treatments had no effect on protein mobility (Table 1). Furthermore, the truncated construct AtFH1-dFH–GFP, which we predict to be unable to interact with cytoplasmic elements, remained largely immobile in FRAP experiments (Figure 3c; solid blue line and Table 1). The relative immobility of AtFH1 within the PM is not due to an intracellular interaction so we turned to investigation of its extracellular domain. We analyzed lateral mobility of AtFH1-dFH–GFP in cells which were plasmolyzed with various concentrations of mannitol to separate the plasma membrane from the cell wall and found that it more than doubled in plasmolyzed cells compared with untreated cells (Figure 3c; green and red lines and Table 1). This increase of I100s was not due to a general effect of osmotic treatment on protein mobility within the PM as no increase of I100s was observed for PIP2a–CFP under similar conditions (Table 1). Spatial proximity between cell wall and PM is required for AtFH1 to remain largely immobile within the PM.

Figure 3.

 Lateral mobility of Formin is constrained by the cell wall.
(a, b) Fluorescence Recovery After Photobleaching experiments to compare the lateral mobility of AtFH1–GFP with another plasma membrane marker, GFP-LTI6b.
(a) AtFH1 is very immobile during 80 sec recovery within the plasma membrane due to the cell wall/plasma membrane connection. Fluorescence is slow to recover when compared with that of (b) GFP-LTI6b. Scale bar = 5 μm.
(c) FRAP curves of AtFH1 and AtFH1-dFH–GFP before and after plasmolysis. Mannitol treatment causes the PM to separate from the cell wall and results in a significant increase in mobility of AtFH1-dFH–GFP.
(d) Cellulose stained with Calcofluor white M2R (blue). At time 0 no staining is apparent. By 48 h after making protoplasts the cell wall is well developed and stains strongly. Scale bar = 10 μm.
(e) FRAP curves show decreasing mobility of AtFH1-dFH–YFP during the 48 h of cell wall regeneration in protoplasts as the connection between cell wall and the extracellular domain of AtFH1 is re-established.
(f, g) Examples of FRAP of AtFH1-dFH protoplasts without (e) and with (f) cell wall. Fluorescence recovery at 110 sec is much less in protoplasts with primary cell wall (f) Pixels values are rainbow coloured so that high intensity fluorescence is indicated with red. The rectangle shows the bleaching area.

Table 1.   Mobility of AtFH1 within the plasma membrane
ConstructTreatmentMobile fraction (%)
  1. aMannitol treatments compared with no treatment for AtFH1-dFH–GFP, < 0.05 Tukey HSD test, n = 10 cells. PIP2a-CFP did not differ in lateral mobility between control and plasmolyzed cells (n = 9 cells). Mobile fraction is expressed as mean ± standard deviation.

AtFH1–GFP20.2 ± 5.1
AtFH1–GFPLatrunculin b/oryzalin19.2 ± 6.6
AtFH1-dFH–GFP21.4 ± 3.6
AtFH1-dFH–GFP0.35 m mannitol42.0 ± 11.7a
AtFH1-dFH–GFP0.5 m mannitol48.0 ± 13.7a
PIP2;1–CFP15.8 ± 5.7
PIP2;1–CFP0.5 m mannitol14.2 ± 5.7

Cell wall regeneration inhibits the lateral mobility of AtFH1-dFH

Interaction with the cell wall seems to reduce the lateral mobility of AtFH1-dFH–GFP. To support these observations, we recorded protein mobility during cell wall regeneration of protoplasts. AtFH1-dFH–YFP-expressing protoplasts were used in FRAP experiments to record lateral mobility of AtFH1-dFH–YFP at times 0, 24, and 48 h during development of the new primary cell wall (Figure 3d). Protein lateral mobility decreased gradually from t0 to t48 (Figure 3e–g) (Tukey test of I100s, t0 versus t24, = 0.046; t0 versus t48, P = 0.001; Table 2). During cell wall regeneration, a concomitant stabilization of AtFH1-dFH–YFP occurs in the PM. The agent of this stabilization is absent or reduced in newly made protoplasts and gradually appears during the early phase of cell-wall deposition.

Table 2.   Lateral mobility of AtFH1-dFH–YFP decreases during cell wall regeneration
Hours post cell wall digestionMobile fraction (%)
  1. *P < 0.05 Tukey HSD test, n = 12 cells. Mobile fraction is expressed as mean ± standard deviation.

029.9 ± 15.9
2419.3 ± 5.7*
4812.3 ± 6.7*

A conserved subdomain of the proline-serine-rich extracellular domain is involved in anchoring of AtFH1 to the cell wall

The extracellular domain of AtFH1 has four highly-conserved sub-domains: Domain A residue 23–32, Domain B residue 37–56, Domain C residue 66–80, and Domain D residue 83–96 (Figure 1a). Complete removal of all four domains resulted in a fluorescent construct that remained in the ER (Figure S5b). Proper targeting to the PM was achieved when Domain A was included in the deletion constructs. To determine which of Domains B, C or D are involved in anchoring of AtFH1 to the cell wall, three deletion constructs were made from the full length protein. AtFH1-TR1 in which Domains B, C and D are removed, AtFH1-TR2 in which Domain B and C are removed, and AtFH1-TR3 in which only Domain B is removed (Figure 1a). All three constructs labelled the PM strongly without any obvious differences compared with the full length protein (Figure S5a,c–e). FRAP was used to record the lateral mobility of these different deletion constructs. None of them had a fluorescence-intensity recovery value that differed from that of the control protein AtFH1–YFP (P = 0.785; Table 3). Because formins are known to interact with cytoskeletal elements (Michelot et al., 2006; Deeks et al., 2010), we treated cells with 25 μm latrunculin B and 20 μm oryzalin to remove MF and microtubules, respectively. In these conditions, the full length construct and AtFH1-TR3–YFP did not behave differently between control and treated cells, but fluorescence-intensity recovery of AtFH1-TR1 and AtFH1-TR2 increased significantly compared with that in untreated cells (Tukey HSD test, untreated versus treated, for AtFH1–YFP, P = 0.889; for AtFH1-TR1–YFP, P = 0.004; for AtFH1-TR2–YFP, P = 0.008, for AtFH1-TR3–YFP, P = 0.284; Table 3). The only difference between AtFH1-TR2 and AtFH1-TR3 is the addition of the 18 amino acids of Domain C (PFFPLYPSSPPPPSPASF). This sequence is sufficient for keeping AtFH1 immobilized and contains the motif SPPPP which is homologous with the extensin glycosylated domain (Banno and Chua, 2000). Consequently, this domain appears to be involved in a cell wall/AtFH1 interaction.

Table 3.   Lateral mobility of AtFH1 is affected in extracellular domain deletion constructs when the cytoskeletons are removed
 Mobile fraction (I100%) (cytoskeleton present)Mobile fraction (%) (cytoskeleton absent)
  1. *Tukey HSD test, untreated versus latrunculin B/oryzalin treated, n = 9 cells. anova Between constructs, P = 0.785; anova Between treatments (cytoskeleton removal), P = 0.001. Mobile fraction is expressed as mean ± standard deviation.

AtFH1–YFP15.2 ± 7.915.7 ± 4.6
AtFH1-TR1–YFP12.7 ± 4.219.6 ± 4.0*
AtFH1-TR2–YFP13.3 ± 5.921.1 ± 5.8*
AtFH1-TR3–YFP15.3 ± 5.717.7 ± 2.4

AtFH1 overexpression induces an increase in the number of actin filaments when AtFH1 interacts with the cell wall

Overexpression of AtFH1 in tobacco pollen tubes induces an increase in the number of actin filaments at the cell cortex (Cheung and Wu, 2004). To determine if overexpression of AtFH1-xFP induces any similar effect in tobacco leaf epidermal cells, it was co-expressed with one of the actin labelling constructs Lifeact–RFP or Lifeact–GFP (Era et al., 2009; Riedl et al., 2008; Roca et al., 2010). Long, thick bundles of MFs, as would be observed without AtFH1 overexpression, disappeared and a dense meshwork of thin, mainly parallel MFs was observed (Figure 4a–c). This finding was also observed with other actin labelling constructs (Figure S6a,b). To confirm that this effect was not due to the fluorescent-protein tag fused to AtFH1, we expressed a non-tagged AtFH1 with Lifeact–GFP and the same altered MF pattern was observed (Figure 4d). When the FH domain was removed from AtFH1, the actin network appeared as in AtFH1 non-overexpressing cells (Figure 4e). These two experiments indicate that AtFH1-xFP is functional in our system. But, in a somewhat unexpected manner, when the extracellular domain truncation constructs AtFH1-TR1–YFP and AtFH1-TR2–YFP were co-expressed with Lifeact–RFP, the MF pattern did not resemble the actin structure described for AtFH1 overexpressing cells (Figure 4f,g). Rather, for these deletions (TR1 and TR2), the effect on actin network morphology was relatively weak with only 31% of cells for AtFH1-TR1–YFP and 34% of cells for AtFH1-TR2–YFP having an increase in the number of MFs (n = 100 transformed cells from four independent experiments). In contrast, the Lifeact–RFP pattern in AtFH1-TR3–YFP-expressing cells was similar to that in cells expressing the full length protein AtFH1–YFP (Figure 4h). These results show that presence of Domain C – and its cell-wall anchoring function – is required to produce the increase in MF number that results in an actin sheet in the cell cortex. Unfortunately, it is difficult to quantify precisely the effect of the AtFH1 truncation constructs on the MF network, especially as we use actin labelling tags that might also affect actin morphology and dynamics. For this reason, we decided to follow Golgi body motility as a reporter of the AtFH1-overexpression effect on the actin cytoskeleton.

Figure 4.

 Overexpression of AtFH1 induces a re-organization of the actin cytoskeleton.
(a–h) Actin structure labelled with Lifact–GFP (green) or Lifact–RFP (magenta) in presence or absence of different AtFH1 constructs.
(a) Actin cytoskeleton in the absence of AtFH1 overexpression.
(b) A cell co-expressing AtFH1–GFP and Lifeact–RFP (arrow) has a dense mat of actin MFs very different to the cells which surround it that only express Lifact–RFP (stars).
(c) The same effect was recorded with co-expression of Lifeact–GFP and AtFH1–RFP.
(d) Overexpression of un-tagged AtFH1 resulted in formation of a dense array of MFs similar to that induced by the tagged version of AtFH1.
(e) Overexpression of the AtFH1 construct with no actin-binding intracellular domain (AtFH1-dFH–YFP) had no effect on organization of the actin cytoskeleton.
(f–h) Actin cytoskeleton morphology in cells overexpressing various deletions of the AtFH1 extracellular domain.
(f) Overexpression of the construct in which extracellular subdomains B, C, and D are deleted (AtFH1-TR1–YFP), or (g) the construct in which extracellular subdomains B and C are deleted (AtFH1-TR2–YFP), had very little apparent effect on actin cytoskeleton morphology.
(h) When subdomain C was added back into the extracellular domain deletion (AtFH1-TR3–YFP) the actin cytoskeleton appeared as if the full length AtFH1 was being overexpressed [compare with (c–h)]. Scale bars = 10 μm.

Overexpression of AtFH1 inhibits Golgi body motility

Golgi bodies are highly motile organelles in plant cells, and this movement is dependent on presence of the actin cytoskeleton and myosin proteins (Brandizzi et al., 2002; Sparkes et al., 2008). We expect that changes in MF organization might perturb Golgi motility. Sialyl transferase-mRFP (ST-mRFP), a Golgi body fluorescent tag, was co-expressed with our set of AtFH1 constructs. Golgi body motility was recorded by time lapse imaging. The Meandering Index (MI – Golgi displacement:track length) was calculated. The value of MI varies from close to 0 when the Golgi body is immobile or moves erratically, to close to 1, when a Golgi body follows a straight trajectory (Runions et al., 2006; Sparkes et al., 2008). When ST-mRFP is expressed on its own, MI = 0.41 ± 0.26 (standard deviation) (Figure 5a,h). This value is consistent with previous studies and, as others have shown, has a high standard deviation because of the co-occurrence of moving and non-moving Golgi bodies in the same cell (Runions et al., 2006; Sparkes et al., 2008). After removal of the actin cytoskeleton by latrunculin B treatment, Golgi body motility was significantly reduced (MI = 0.07 ± 0.07) (Figure 5b,h). This reduction in MI is because Golgi bodies can no longer move along straight trajectories when the actin cytoskeleton is depolymerized. When AtFH1–YFP is co-expressed with ST-mRFP in cells with intact actin cytoskeletons, Golgi movement is strongly reduced (MI = 0.20 ± 0.16; Figure 5c,h), but not stopped as with latrunculin B treatment (Tukey HSD test, ST-mRFP alone versus ST-mRFP + AtFH1–YFP, P = 0.001). AtFH1 expression, because of its effect on actin network reorganization, inhibits Golgi body motility. Expression of AtFH1-dFH–YFP, which does not contain the intracellular FH1 and FH2 domains that bind actin, had no effect on Golgi body movement (MI = 0.42 ± 0.26; Figure 5d,h). AtFH1 constructs with various truncations in the conserved extracellular domains all reduced Golgi body movement with AtFH1-TR3–YFP having the most severe effect and confirming the requirement of cell wall anchoring by extracellular Domain 3 for producing the observed overexpression phenotype (Figure 5e–h).

Figure 5.

 Overexpression of AtFH1 inhibits the mobility of Golgi bodies.
(a–g) Time series data of ST-mRFP-expressing cells was used for tracking Golgi-body mobility. Golgi body paths are indicated with colored lines.
(a) Control cells.
(b) Latrunculin b treated cells. Golgi-body mobility is inhibited when the actin cytoskeleton is depolymerized as can be seen by the reduced path length.
(c–g) ST-mRFP co-expressed with different AtFH1 constructs, (c) the full length protein AtH1–YFP, (d) AtFH1-dFH–YFP, (e) AtFH1-TR1–YFP, (f) AtFH1-TR2–YFP and (g) AtFH1-TR3–YFP. Golgi bodies are immobilized when either the FH domain or the extracellular domain of AtFH1 is present.
(h) Golgi-body Meandering Index (MI), calculated as the ratio between Golgi displacement and total track length. MI is lower when mobility is inhibited. Removal of the actin cytoskeleton by treatment with latrunculin B almost completely eliminates Golgi movement. Overexpression of both the full length AtFH1 and the AtFH1 extracellular deletion construct with Domain C present (AtFH1-TR3) significantly reduced Golgi mobility. When the intracellular, actin binding domains of AtFH1 were removed (AtFH1-dFH), or when Domain C of the extracellular domain was absent (AtFH1-TR1 or AtFH1-TR2), Golgi mobility was significantly increased relative to that in AtFH1 expressing cells. Mean ± SEM; * mobility of Golgi in presence of AtFH1 compared with other constructs, P < 0.05 t-test; n = 10 cells. Scale bar = 10 μm.


Actin microfilaments are extremely dynamic structures (Staiger et al., 2009). Many physiological processes, e.g. pathogen response or re-arrangement during cell division illustrate the plasticity of this network. The constant re-modelling activity of actin microfilaments (nucleation, bundling and severing) is mainly driven by actin interacting proteins (Staiger et al., 2009). Regulation of actin cytoskeleton morphology is crucial and a mechanism for coordinating these re-arrangements in response to signals from the external medium is necessary. Many different actin-interacting proteins have been postulated as the physical link between cell wall and the actin cytoskeleton (Baluska et al., 2003) and we can now demonstrate that the plasma membrane protein Formin1 connects the cell wall to the actin cytoskeleton.

AtFH1 forms a link between the actin cytoskeleton and the cell wall

We believed that to organize and coordinate a complex structure, having spatially stable anchor points is crucial. For instance, the endoplasmic reticulum has a constantly changing shape. Recently, stable ER anchoring points were described (Sparkes et al., 2009). Our observations suggest that Arabidopsis Formin1 (AtFH1) has a similar function in that it forms a bridge across the plasma membrane between actin filaments and the extracellular matrix (Figure 6). AtFH1 and most members of the class I plant formins have been predicted to be type I integral membrane proteins (Cvrckova, 2000). In this study, we have confirmed that AtFH1 is a plasma membrane protein and have shown that it has a conserved N-terminal domain, which is localized within the extracellular space.

Figure 6.

 A model of AtFH1 bridging between the cell wall and the actin cytoskeleton. AtFH1 (yellow) spans the plasma membrane (blue) connecting the actin cytoskeleton (red) and the cell wall (indicated as cellulose microfibrils in brown). AtFH1 is an actin nucleating protein that binds the side of actin microfilaments via its intracellular formin homology domains. When AtFH1 is overexpressed, dense sheets of actin form. This actin modelling function requires that AtFH1 link to both the cell wall and to actin.

Four experimental techniques have been used to demonstrate the presence of a link between the extracellular part of AtFH1 and the extra-cellular matrix (ECM). Firstly, when plasma membrane/cell wall proximity was disrupted by plasmolysis, AtFH1 remained in the PM and its lateral mobility within the plasma membrane increased. This finding suggests that a physical link with a protein-anchoring function between AtFH1 and the ECM was disrupted by plasmolysis. Secondly, the PM lateral mobility of AtFH1-dFH was high in protoplasts and steadily decreased during 48 h of cell-wall regeneration. Thirdly, removal of the majority of the extracellular N-terminus of full length AtFH1 induced an increase in the protein’s lateral mobility when the cytoskeleton was removed. Finally, removal of the transmembrane domain (AtFH1-107–RFP) resulted in labelling of the cell wall specifically in plasmolyzed cells. As a consequence of these observations, we can say that AtFH1 forms a relatively immobile link from the cell wall through the PM and acts as an anchor point for the actin cytoskeleton. This model can be related to the biochemical model of AtFH1’s actin-nucleating mechanism. Formin1 is a non-processive formin which means that it does not move with the growing barbed end of a microfilament (Michelot et al., 2006). Rather, the barbed end is spatially motile and grows away from the stable MF nucleation site of formin attachment.

The extracellular region of AtFH1 has four highly conserved subdomains and the cell-wall anchoring role of this region could explain why the amino acid sequences of these subdomains are strongly conserved by purifying selection within the plant kingdom. The extracellular domain of AtFH1 that seems most responsible for immobilization of the protein is Domain C in construct AtFH1-TR3. In the absence of this domain, AtFH1 was mobile within the PM. Domain C is only comprised of 15 amino acids that, intriguingly, include an SPPPP motif which is the signature peptide of extensin, a class of cell-wall associated proteins (Banno and Chua, 2000; Showalter et al., 2010) so it seems reasonable that this motif confers the cell-wall interacting function.

The biochemical nature of the cell wall anchoring interaction is not yet understood. It might be either a protein–protein interaction or an interaction with other cell wall components. AtFH1’s extracellular region contains a high percentage of serine and proline residues, which could carry O-linked oligosaccharide modifications. In this case, carbohydrates might also be responsible for the cell wall interaction.

Interestingly, AtFH4 and AtFH5 have specific patterns of plasma membrane localization – AtFH4 strongly accumulates in cell-to-cell contact zones and AtFH5 accumulates at the pollen-tube tip (Deeks et al., 2005; Cheung et al., 2010). The fact that formins are targeted to a sub-domain of the PM might reflect heterogeneity of cell wall composition.

The cell wall connection is required for AtFH1-induced remodelling of the actin cytoskeleton

The bridging topology of AtFH1 makes us think that this protein may participate in signal transduction between outside stimuli and the actin cytoskeleton. Here, we show that the extracellular region of AtFH1 – especially Domain C – is required to effect a change in shape of the actin network when AtFH1 is overexpressed (compare Figure 4g,h). Similarly, only AtFH1 with intact Domain C was able to strongly reduce long distance movement of Golgi bodies (compare Figure 5f,g) and, as stated in the previous section, Domain C seems to be involved in anchoring of AtFH1 to the ECM. These results, taken together, show that anchoring to the ECM via Domain C is needed for inducing changes in the actin network as observed in AtFH1-overexpressing cells.

Several molecular mechanisms can be proposed for explaining the link between anchoring in the extracellular space and actin nucleation or actin bundling within the cell. Animal cell formin needs to form a dimer to be functional (Harris et al., 2004). Isolated intracellular domains FH1 and FH2 of AtFH1 are able to form dimers in in vitro assays, and dimers are probably necessary for formin1 activity (Michelot et al., 2005, 2006). Consequently, we can extrapolate that the full protein is also able to form functional dimers. In the plasma membrane, both non-functional monomer and functional dimers may co-exist. We speculate that anchoring of AtFH1 dimers in the cell wall has an effect on their stabilization and conformation. In absence of anchoring, the dimer might be less stable, less likely to form or might not possess the active conformation and consequently the protein will not induce the observed meshwork-like array of microfilaments.

AtFH1 labelling forms a negative microtubule pattern on the plasma membrane

Several animal formins have been described as interacting directly with microtubules (Chesarone et al., 2010). This phenomenon is involved in coordination of the two main cytoskeletal networks: actin microfilaments and microtubules. Recently, it was reported that plant formin also interacts with microtubules (Deeks et al., 2010; Li et al., 2010). The microtubule shadow observed in the plasma membrane when AtFH1 is overexpressed may also be the result of a physical interaction between AtFH1 and microtubules (Figure 1c). But more likely the exclusion of AtFH1 from PM regions overlying the microtubules is because the large cytoplasmic domain of AtFH1 is corralled by, and cannot cross, cortical MTs which are also anchored to the plasma membrane (Pesquet et al., 2010). The exact nature and the biological significance of this phenomenon are yet to be determined.

Experimental Procedures

Constructs and cloning procedures

The cDNA of At3g25500 (AtFH1), ABRC clone number U11142 was used as a template for all AtFH1 constructs. For AtFH1 (nucleotide 1–3156), AtFH1-dFH (nucleotide 1–420) and AtFH1-107 (nucleotide 1–323) PCR fragments were cloned in pENTR-D-TOPO (Invitrogen). For AtFH1-dECM (nucleotide 1–72 fused to nucleotide 315–3156), AtFH1-TR1 (nucleotide 1–102 fused to nucleotide 294–3156), AtFH1-TR2 (nucleotide 1–102 fused to nucleotide 240–3156) and AtFH1-TR3 (nucleotide 1–102 fused to nucleotide 189–3156) overlapping PCR were performed before the cloning via TOPO reaction. For CFP–TuA6 or YFP-TuA6, TUBULINa6 was amplified from GFP6-TuA6 (Ueda et al., 1999) and cloned into the pENTR-D-TOPO vector. The following Gateway (Invitrogen) binary vectors were selected to do N- and C-terminal fusions of cloned PCR products with various fluorescent proteins: pB7RWG2, pB7FWG2, pB7GWC2, pB7CWG2, pB7GWY2, pB7YWG2 (Plant Systems Biology, Ghent University). pB2GW7 was used to make the non-fluorescent protein tagged AtFH1 (Plant Systems Biology, Ghent University). All final clones were verified by sequencing.

Plant transformation and confocal microscopy

Nicotiana tabacum cv. Petit Havana plants were grown and used in transient transformation experiments as described in Sparkes et al. (2006). All the formin constructs, sialyl transferase signal anchor sequence fused to mRFP (ST-mRFP) (Saint-Jore et al., 2002), GFP-LTI6b (Runions et al., 2006) and Sec-mRFP (Samalova et al., 2006) were infiltrated using Agrobacterium tumefaciensGV3101 at 0.1 OD600 (Sparkes et al., 2006). Cytoskeleton labelling probes; Lifeact–GFP (Deeks et al., 2010), Lifeact–RFP (Berepiki et al., 2010), GFP-FABD2 (Sheahan et al., 2004), GFP-mTn (Kost et al., 1998) and C/YFP-TuA6, were infiltrated at 0.01 OD600 to prevent side effects of overexpression (Runions et al., 2006). Two to three days after transformation leaf tissue was observed with a Zeiss ( LSM 510 META confocal system.

Drugs treatments, cell wall labelling and plasmolysis assays

Leaf pieces (approximately 0.25 cm2) were perfused by immersion in 25 μm latrunculin B (30–180 min) to remove the actin cytoskeleton (stock solution 2.5 mm in DMSO), 20 μm oryzalin (30 min) to remove microtubules (stock solution 50 mm in DMSO), or with 357 μm Brefeldin A (BFA, 3 h for ST-mRFP and 6 h for AtFH1) to block retrograde secretion. For protoplast experiments (see below), primary-cell-wall regeneration was monitored by staining with 1 μg ml−1 of Calcofluor White M2R from a stock solution at 1 mg ml−1 in DMSO. For the plasmolysis assay, leaf sections (approximately 0.25 cm2) were incubated with a hypertonic solution of 0.5 m mannitol for 1 h.

Quantification of the relative mobile fraction of formin constructs

The relative mobile fraction of formin fused to different fluorescent proteins was assessed by Fluorescence Recovery After Photobleaching (FRAP). A 63× (1.4 NA) objective was used at a zoom setting of 5. Pre-bleaching and post-bleaching excitation was done using either the 488 nm line of an argon-ion laser (GFP) or the 514 nm line (YFP) set at 50% output and 3% transmission. Five scans of the entire field of view were made to establish the pre-bleach intensity of the fluorescent protein and then a circular region of interest (ROI) 8 μm2 in a median optical section of the fluorescent plasma membrane was bleached. Ten iterations of the 488 or 514 nm laser set at 100% transmission were used for bleaching. Recovery of fluorescence was recorded during 115 sec with a delay of 1.5 sec between frames. Images were 256 × 256 pixels and were made with a scan speed of 0.46 sec per frame. We confirmed that the energy of the 488 or 514 laser used to record post-bleach data had no bleaching effect by recording a control region outside of the bleaching ROI. Fluorescence intensity data were normalized using the equation:


Where In is the normalized intensity, It is the intensity at any time t, Imin is the minimum post-photobleaching intensity and Imax is the mean pre-photobleaching intensity.

Non-linear regression was used to model the normalized FRAP data. In this case, a two-phase exponential association equation was used:


Where Y(t) is normalized intensity, A, B, C, K1 and K2 are parameters of the curve, and t is time.

For each treatment, 7–13 cells were analysed. The value of the fluorescence intensity recovery plateau (t = 110 sec) was calculated and used as an approximation of the relative mobile fraction.

Protoplast preparation

Fusion-protein-expressing tobacco leaf pieces were submerged in a 20× dilution of DOMESTOS commercial bleach solution (Unilever) for 1 min and rinsed several times with distilled water. The leaf tissue was then incubated for 15 h in 0.5 m mannitol, 1% cellulase R10, 0.1% macerozyme, 1 mm CaCl2 and 10 mm MES pH 5.6. After filtration, protoplasts were rinsed twice with 0.5 m mannitol, 10 mm glucose, 1 mm CaCl2 and 10 mm MES pH 5.6. An aliquot of cells was used immediately as a time 0 sample. Protoplasts were left at room temperature in light for 24 and 48 h time points.

Golgi body mobility

For recording the mobility of Golgi bodies, ST-mRFP-expressing cells were imaged (0.45 sec frame−1) for 49 sec. The data were analysed with Volocity 4 (Improvision), according to Runions et al. (2006). The meandering index (MI) was calculated as the ratio between Golgi body displacement (length of the straight line connecting start to end point) and track length (length of the actual path followed). For each treatment 54–138 Golgi bodies were followed for 10 different cells.

Calculation of selection acting on AtFH1–PAML analysis

The AtFH1 sequence from A. thaliana (NC_003074) was used to retrieve coding sequences of AtFH1 orthologous genes in public databases from Ricinus communis (XM_002532408), Populus trichocarpa (XM_002310361), Vinis vitifera (XM_002274965), Medicago truncatula (AC171534), N. tabacum (AF213695), Oryza sativa v. Japonicum (AP003349) and Sorghum bicolour (XM_002458874). Nucleotide alignment was performed using Translator X (Abascal et al., 2010) using the default settings of the software. Alignment of the whole gene was then partitioned into three alignments, according to putative functional domains of the protein. Part 1 is from the 5′ part to the end of the putative extracellular domain of AtFH1 (nt 1–323 in A. thaliana sequence). Part 2 is from the transmembrane domain to the beginning of the FH1 domain (nt 324–1560 in A. thaliana sequence). Part 3 is the 3′ part of the gene and contains FH1 and FH2 domains (nt 1561–3156 in A. thaliana sequence).

Estimations of the nonsynonymous to synonymous substitution rate (ω = dN/dS) were performed using Phylogenetic Analysis by Maximum Likelihood (PAML) with the codeml program implemented in paml 4.3 (Yang, 2007). Phylogenetic trees required for PAML were then inferred from protein sequence data of each of the three parts of the protein. To do so, the software Gblocs (Talavera and Castresana, 2007) was used to eliminate poorly aligned positions and divergent regions of an alignment of protein sequences. These positions may not be homologous or may have been saturated by multiple substitutions and it is convenient to eliminate them prior to phylogenetic analysis (Castresana, 2000). Phylogenies were inferred by Phyml (Guindon and Gascuel, 2003). The GTR substitution model was selected assuming a proportion of invariant sites (0.01). To account for rate heterogeneity across sites, the gamma distribution with four rate categories was used.


The authors would like to thank Dr Michael Deeks for discussions and comments on an early version of the manuscript and Dr Jens Tilsner who provided us with the binary vector for Lifeact–RFP. This research was funded by BBSRC Grant number BB/F014074/1.