Arabidopsis Class II Formins AtFH13 and AtFH14 Can Form Heterodimers but Exhibit Distinct Patterns of Cellular Localization

Formins are evolutionarily conserved multi-domain proteins participating in the control of both actin and microtubule dynamics. Angiosperm formins form two evolutionarily distinct families, Class I and Class II, with class-specific domain layouts. The model plant Arabidopsis thaliana has 21 formin-encoding loci, including 10 Class II members. In this study, we analyze the subcellular localization of two A. thaliana Class II formins exhibiting typical domain organization, the so far uncharacterized formin AtFH13 (At5g58160) and its distant homolog AtFH14 (At1g31810), previously reported to bind microtubules. Fluorescent protein-tagged full length formins and their individual domains were transiently expressed in Nicotiana benthamiana leaves under the control of a constitutive promoter and their subcellular localization (including co-localization with cytoskeletal structures and the endoplasmic reticulum) was examined using confocal microscopy. While the two formins exhibit distinct and only partially overlapping localization patterns, they both associate with microtubules via the conserved formin homology 2 (FH2) domain and with the periphery of the endoplasmic reticulum, at least in part via the N-terminal PTEN (Phosphatase and Tensin)-like domain. Surprisingly, FH2 domains of AtFH13 and AtFH14 can form heterodimers in the yeast two-hybrid assay—a first case of potentially biologically relevant formin heterodimerization mediated solely by the FH2 domain.


Introduction
The eukaryotic cytoskeleton represents a dynamic network of protein filaments and tubules that has been extensively studied in a variety of model systems. In addition to basic cellular functions, including nuclear division, cytokinesis, organelle positioning, membrane trafficking and cell expansion, the cytoskeleton plays an important role in processes such as polar cell growth, cell division plane positioning, or cell to cell communication, which are essential for proper morphogenesis and thus the development of multicellular bodies in both metazoans [1] and plants [2]. This dynamic network relies on a large ensemble of cytoskeleton-associated proteins controlling organization, remodeling, and crosstalk of cytoskeletal systems, as well as coordination of the cytoskeleton with cell membranes and organelles. While the cellular structures and organismal body plans vary greatly among the eukaryotes, many molecular and cellular mechanisms involving the two basic cytoskeletal systems-the actin microfilaments and microtubules-are conserved among divergent groups such as, e.g., plants and metazoans, which are connected only through the last eukaryotic common ancestor [3,4]. To examine in vivo subcellular localization of these formins, we constructed vectors expressing full length AtFH13 (Uniprot Acc. No. Q9LVN1) or a previously characterized variant of AtFH14 [23] tagged by C-terminal yellow fluorescent protein (YFP) or red fluorescent protein (RFP) fusion under the control of the constitutive ubiquitin 10 promoter (UBQ10) and used these constructs for transient transformation of N. benthamiana leaves. To examine contribution of individual domains of both proteins to their localization, we also prepared constructs expressing YFP-tagged N-terminal fragments of either protein containing the PTEN-like and C2 domains, as well as green fluorescent protein (GFP) and YFP-tagged isolated FH2 domains of AtFH13 and AtFH14. In addition, we also generated an YFP fusion of the N-terminal (PTEN-like and C2 domain-containing) fragment from a fortuitously cloned AtFH13 variant missing 21 amino acids within the PTEN-like domain (denoted PTENΔ) as a putative inactive variant of the PTEN domain ( Figure 1c).

Both AtFH13 and AtFH14 Associate with Microtubules and the ER in Tobacco Epidermis
As usual in transient N. benthamiana leaf epidermis transformation, individual transformed cells exhibited variable signal intensity. Confocal imaging of YFP-tagged AtFH13 in cells with a relatively weak signal showed fibers, suggesting cytoskeletal association, as well as punctate structures of varying size in the cortical cytoplasm (Figure 2a), whereas cells which overexpressed the construct To examine in vivo subcellular localization of these formins, we constructed vectors expressing full length AtFH13 (Uniprot Acc. No. Q9LVN1) or a previously characterized variant of AtFH14 [23] tagged by C-terminal yellow fluorescent protein (YFP) or red fluorescent protein (RFP) fusion under the control of the constitutive ubiquitin 10 promoter (UBQ10) and used these constructs for transient transformation of N. benthamiana leaves. To examine contribution of individual domains of both proteins to their localization, we also prepared constructs expressing YFP-tagged N-terminal fragments of either protein containing the PTEN-like and C2 domains, as well as green fluorescent protein (GFP) and YFP-tagged isolated FH2 domains of AtFH13 and AtFH14. In addition, we also generated an YFP fusion of the N-terminal (PTEN-like and C2 domain-containing) fragment from a fortuitously cloned AtFH13 variant missing 21 amino acids within the PTEN-like domain (denoted PTEN∆) as a putative inactive variant of the PTEN domain ( Figure 1c).

Both AtFH13 and AtFH14 Associate with Microtubules and the ER in Tobacco Epidermis
As usual in transient N. benthamiana leaf epidermis transformation, individual transformed cells exhibited variable signal intensity. Confocal imaging of YFP-tagged AtFH13 in cells with a relatively weak signal showed fibers, suggesting cytoskeletal association, as well as punctate structures of varying size in the cortical cytoplasm (Figure 2a), whereas cells which overexpressed the construct exhibited large, brightly fluorescent aggregates (Figure 2b). In case of AtFH14-YFP the cortical signal was mainly of fibrous character (Figure 2c), suggesting association with microtubules, consistent with previously observed localization of this protein's FH1-FH2 domain construct to the preprophase band, mitotic spindle, and phragmoplast in tobacco BY2 cells [23]. exhibited large, brightly fluorescent aggregates (Figure 2b). In case of AtFH14-YFP the cortical signal was mainly of fibrous character (Figure 2c), suggesting association with microtubules, consistent with previously observed localization of this protein´s FH1-FH2 domain construct to the preprophase band, mitotic spindle, and phragmoplast in tobacco BY2 cells [23]. To establish the relationship between the formin-labeled fibers and cytoskeletal structures, we co-expressed both tagged formins with the actin and microtubule markers LifeAct-RFP or KMD (kinesin motor domain)-RFP, respectively. Neither AtFH13 nor AtFH14 showed association with actin fibers (Figure 3) but both proteins differed in their localization patterns. AtFH13 exhibited patchy distribution following the microtubules, whereas AtFH14 decorated the microtubule network more homogeneously (Figure 4, Video S1, Video S2). An obvious, though statistically non-significant, trend towards co-localization with microtubules rather than microfilaments was detected also by quantitative image analysis ( Figure S1).  To establish the relationship between the formin-labeled fibers and cytoskeletal structures, we co-expressed both tagged formins with the actin and microtubule markers LifeAct-RFP or KMD (kinesin motor domain)-RFP, respectively. Neither AtFH13 nor AtFH14 showed association with actin fibers (Figure 3) but both proteins differed in their localization patterns. AtFH13 exhibited patchy distribution following the microtubules, whereas AtFH14 decorated the microtubule network more homogeneously ( Figure 4, Video S1, Video S2). An obvious, though statistically non-significant, trend towards co-localization with microtubules rather than microfilaments was detected also by quantitative image analysis ( Figure S1).  (Figure 2b). In case of AtFH14-YFP the cortical signal was mainly of fibrous character (Figure 2c), suggesting association with microtubules, consistent with previously observed localization of this protein´s FH1-FH2 domain construct to the preprophase band, mitotic spindle, and phragmoplast in tobacco BY2 cells [23]. To establish the relationship between the formin-labeled fibers and cytoskeletal structures, we co-expressed both tagged formins with the actin and microtubule markers LifeAct-RFP or KMD (kinesin motor domain)-RFP, respectively. Neither AtFH13 nor AtFH14 showed association with actin fibers (Figure 3) but both proteins differed in their localization patterns. AtFH13 exhibited patchy distribution following the microtubules, whereas AtFH14 decorated the microtubule network more homogeneously (Figure 4, Video S1, Video S2). An obvious, though statistically non-significant, trend towards co-localization with microtubules rather than microfilaments was detected also by quantitative image analysis ( Figure S1).   To confirm functional connection between the AtFH13-labelled foci and the microtubule cytoskeleton, we examined the effect of the microtubule depolymerizing drug oryzalin on localization of AtFH13-YFP co-expressed with KMD-RFP. The oryzalin treatment caused partial redistribution of the cytoplasmic dots and increased their mobility, which may be due to cytoplasmic streaming, not affected by microtubule disruption ( Figure 5). This suggests that integrity of microtubules is essential to maintain subcellular distribution of AtFH13.  To confirm functional connection between the AtFH13-labelled foci and the microtubule cytoskeleton, we examined the effect of the microtubule depolymerizing drug oryzalin on localization of AtFH13-YFP co-expressed with KMD-RFP. The oryzalin treatment caused partial redistribution of the cytoplasmic dots and increased their mobility, which may be due to cytoplasmic streaming, not affected by microtubule disruption ( Figure 5). This suggests that integrity of microtubules is essential to maintain subcellular distribution of AtFH13.  To confirm functional connection between the AtFH13-labelled foci and the microtubule cytoskeleton, we examined the effect of the microtubule depolymerizing drug oryzalin on localization of AtFH13-YFP co-expressed with KMD-RFP. The oryzalin treatment caused partial redistribution of the cytoplasmic dots and increased their mobility, which may be due to cytoplasmic streaming, not affected by microtubule disruption ( Figure 5). This suggests that integrity of microtubules is essential to maintain subcellular distribution of AtFH13.  Since both AtFH13 and AtFH14 contain the PTEN-like domain known to associate with membranes [20,24], we hypothesized that the fluorescent cytoplasmic punctate structures not associated with the cytoskeleton may be related to compartments of the endomembrane system. We thus examined co-localization of AtFH13-YFP and AtFH14-YFP with the ER marker ER-rk (ER red kanamycin). Remarkably, the two proteins exhibited obviously different localization patterns, with AtFH13-labelled structures peripherally associated with the ER network with minimal co-localization, while AtFH14-labeled structures exhibited noticeable overlap with the ER, which, however, could have been due to microtubule-ER co-alignment ( Figure 6, Video S3, Video S4). Better co-localization of AtFH14 with the ER was supported also by a statistically non-significant trend detected quantitatively ( Figure S1). largely depolymerized). Single channel images are shown in grayscale, merged images display an overlay of the YFP channel in green with RFP in magenta. Particle tracks show trajectories of individual AtFH13-YFP dots over 50 s with insets magnified 2.5 times compared to the main image. (b) Box plot of mean velocities of AtFH13-YFP particles without, after 10 and after 90 min of oryzalin treatment. Scale bar is 10 µm. Two asterisks indicate a statistically significant difference (ANOVA, p < 0.001), NS: Non-significant. Construct abbreviations are defined in the text and in Figure 1.
Since both AtFH13 and AtFH14 contain the PTEN-like domain known to associate with membranes [20,24], we hypothesized that the fluorescent cytoplasmic punctate structures not associated with the cytoskeleton may be related to compartments of the endomembrane system. We thus examined co-localization of AtFH13-YFP and AtFH14-YFP with the ER marker ER-rk (ER red kanamycin). Remarkably, the two proteins exhibited obviously different localization patterns, with AtFH13-labelled structures peripherally associated with the ER network with minimal colocalization, while AtFH14-labeled structures exhibited noticeable overlap with the ER, which, however, could have been due to microtubule-ER co-alignment ( Figure 6, Video S3, Video S4). Better co-localization of AtFH14 with the ER was supported also by a statistically non-significant trend detected quantitatively ( Figure S1).

Isolated FH2 Domains of AtFH13 and AtFH14 Exhibit Distinct Patterns of Microtubule Association
Next, we compared the localization of isolated FH2 domains of AtFH13 and AtFH14 in transiently transformed epidermal cells. GFP-tagged FH2 domain of AtFH13 was predominantly cytoplasmic with a few small punctate structures; in contrast, YFP-tagged FH2 domain of AtFH14 exhibited prominent fibrous distribution similar to the full-length protein (Figure 7a). Interestingly, upon co-expression with LifeAct-RFP, the GFP-tagged FH2 domain of AtFH13 organized into fibrous structures partially co-localizing with the actin marker, whereas the YFP-tagged FH2 domain of AtFH14 retained fibrous organization clearly distinct from the actin meshwork, similar to the fulllength protein (Figure 7b). The microtubular nature of the network decorated by the YFP-tagged FH2 domain of AtFH14 was also confirmed by co-expression with the microtubule marker KMD-RFP. Remarkably, also the rare cytoplasmic dots labeled by GFP-tagged FH2 domain of AtFH13 were microtubule-associated (Figure 7c, Video S2).

Isolated FH2 Domains of AtFH13 and AtFH14 Exhibit Distinct Patterns of Microtubule Association
Next, we compared the localization of isolated FH2 domains of AtFH13 and AtFH14 in transiently transformed epidermal cells. GFP-tagged FH2 domain of AtFH13 was predominantly cytoplasmic with a few small punctate structures; in contrast, YFP-tagged FH2 domain of AtFH14 exhibited prominent fibrous distribution similar to the full-length protein (Figure 7a). Interestingly, upon co-expression with LifeAct-RFP, the GFP-tagged FH2 domain of AtFH13 organized into fibrous structures partially co-localizing with the actin marker, whereas the YFP-tagged FH2 domain of AtFH14 retained fibrous organization clearly distinct from the actin meshwork, similar to the full-length protein (Figure 7b). The microtubular nature of the network decorated by the YFP-tagged FH2 domain of AtFH14 was also confirmed by co-expression with the microtubule marker KMD-RFP. Remarkably, also the rare cytoplasmic dots labeled by GFP-tagged FH2 domain of AtFH13 were microtubule-associated (Figure 7c     also strong accumulation in the nucleus and weak cytoplasmic background. In contrast, the deletion derivative of the AtFH13 PTEN-like domain showed only diffuse cytoplasmic distribution, consistent with the deletion altering the protein´s folding and consequently its subcellular localization (Figure 8a). Co-expression with the ER marker ER-rk shows that the YFP-tagged PTEN-like domain of AtFH13 peripherally associates with the ER, while its deletion derivative is cytoplasmic. The particles decorated by YFP-tagged PTEN-like domain of AtFH14 associated, and even partly overlapped, with the ER (Figure 8b). background. In contrast, the deletion derivative of the AtFH13 PTEN-like domain showed only diffuse cytoplasmic distribution, consistent with the deletion altering the protein´s folding and consequently its subcellular localization (Figure 8a). Co-expression with the ER marker ER-rk shows that the YFP-tagged PTEN-like domain of AtFH13 peripherally associates with the ER, while its deletion derivative is cytoplasmic. The particles decorated by YFP-tagged PTEN-like domain of AtFH14 associated, and even partly overlapped, with the ER (Figure 8b).

FH2 Domains of AtFH14 and AtFH13 Can Heterodimerize
It is well established that formins can form homodimers through their FH2 domains [13] but there is so far only limited evidence regarding their possible heterodimerization (see Introduction). The distinct localization patterns of AtFH13 and AtFH14 suggested that if heterodimerization between these formins does take place, it is unlikely to affect a substantial part of the cellular pool of both proteins. We nevertheless tested whether FH2 domains of AtFH13 and AtFH14 can heterodimerize by means of the yeast two hybrid assay.
In the yeast two hybrid system, an isolated FH2 domain of AtFH13 can form homodimers as expected. However, in our two-hybrid setup the FH2 domain of AtFH13 also readily formed heterodimers with the FH2 domain of AtFH14 (Figure 9). This is, at the first glance, surprising, given the distinct cellular localization of these two proteins. Nevertheless, when co-expressed, full length YFP-tagged AtFH13 and RFP-tagged AtFH14 exhibited partial co-localization in tobacco epidermal

FH2 Domains of AtFH14 and AtFH13 Can Heterodimerize
It is well established that formins can form homodimers through their FH2 domains [13] but there is so far only limited evidence regarding their possible heterodimerization (see Introduction). The distinct localization patterns of AtFH13 and AtFH14 suggested that if heterodimerization between these formins does take place, it is unlikely to affect a substantial part of the cellular pool of both proteins. We nevertheless tested whether FH2 domains of AtFH13 and AtFH14 can heterodimerize by means of the yeast two hybrid assay.
In the yeast two hybrid system, an isolated FH2 domain of AtFH13 can form homodimers as expected. However, in our two-hybrid setup the FH2 domain of AtFH13 also readily formed heterodimers with the FH2 domain of AtFH14 (Figure 9). This is, at the first glance, surprising, given the distinct cellular localization of these two proteins. Nevertheless, when co-expressed, full length YFP-tagged AtFH13 and RFP-tagged AtFH14 exhibited partial co-localization in tobacco epidermal cells, in spite of their different overall localization patterns (Figure 10a). In particular, the puncta of AtFH13 often co-localized with fibers decorated by AtFH14, as documented by pixel intensity analyzes of the florescence signal (Figure 10b), consistent with partial co-localization which may involve FH2 domain-mediated heterodimerization. Quantitative estimate of co-localization of the two formins yielded values significantly higher than those obtained for either formin with the non-colocalizing actin marker LifeAct and also higher than those observed for microtubules and the ER, although in these cases the differences mostly were not statistically significant ( Figure S1). cells, in spite of their different overall localization patterns (Figure 10a). In particular, the puncta of AtFH13 often co-localized with fibers decorated by AtFH14, as documented by pixel intensity analyzes of the florescence signal (Figure 10b), consistent with partial co-localization which may involve FH2 domain-mediated heterodimerization. Quantitative estimate of co-localization of the two formins yielded values significantly higher than those obtained for either formin with the noncolocalizing actin marker LifeAct and also higher than those observed for microtubules and the ER, although in these cases the differences mostly were not statistically significant ( Figure S1).   (Figure 10a). In particular, the puncta of AtFH13 often co-localized with fibers decorated by AtFH14, as documented by pixel intensity analyzes of the florescence signal (Figure 10b), consistent with partial co-localization which may involve FH2 domain-mediated heterodimerization. Quantitative estimate of co-localization of the two formins yielded values significantly higher than those obtained for either formin with the noncolocalizing actin marker LifeAct and also higher than those observed for microtubules and the ER, although in these cases the differences mostly were not statistically significant ( Figure S1).

Discussion
Formins are an evolutionarily ancient family of proteins [10][11][12] that can nucleate, cap, and bundle actin filaments, but also bind microtubules, resulting in an ability to modulate actin and microtubule organization and dynamics [7,9], well documented also in plants [8,[25][26][27]. Angiosperm plants possess two clades of formins whose members often exhibit characteristic domain organization. While Class I formins are usually, though not always, integral membrane proteins [11,28], their Class II counterparts often harbor an N-terminal domain related to the mammalian PTEN (phosphatase and tensin homolog) lipid/protein phosphatase that was proposed to mediate membrane binding [11] and later shown to be responsible for membrane association and specific intracellular localization of Class II formins in the moss Physcomitrella patens [19,24] and in rice [21]. Thus, typical formins of both Class I and Class II are capable of membrane localization (reviewed in References [25,29]). In case of Class I formins, plasmalemma or endomembrane localization, as well as cytoskeletal, especially microtubule, association is well documented (e.g., References [6,[30][31][32][33][34][35]). Much less is, however, known about in vivo intracellular localization of Class II formins.
In this study, we use a heterologous in planta expression system to document in vivo subcellular localization of fluorescent protein-tagged derivatives of two Arabidopsis Class II formins, the hitherto uncharacterized AtFH13 and somewhat characterized AtFH14 (see below). Although protein tagging by fusion with a relatively large polypeptide such as the YFP or GFP might influence its intracellular localization, such fusion proteins, including a plant formin [35], have been documented to be functional. Unfortunately, we could not confirm our observations by an independent method since we were unable to obtain specific antibodies suitable for immunostaining of formins, including AtFH13 [36]. Another theoretical alternative, expression of small tags visualized using cell permeable ligands, is well established in mammalian cell biology but practically restricted to the study of extracytoplasmic proteins in plants [37], with the exception of a single report with no published experimental follow-up [38].
Both formins studied exhibit the typical domain composition involving a N-terminal PTEN-like domain followed by a repetitive, proline-rich FH1 domain believed to mediate profilin-actin binding and by the C-terminal conserved dimerizing and actin-nucleating FH2 domain [11,18]. Despite the shared overall domain organization, the two proteins only exhibit limited similarity. Phylogenetic analysis of combined PTEN-like and FH2 domain sequences documented that they represent two distinct branches of the Class II formin clade that have separated no later than in basal Brassicaceae, and possibly earlier. While inner topology of the Class II formin clade in previous single-domain phylogenetic analyzes [11,18] had poor statistical support, it was nevertheless consistent with possible divergence of the clades represented by AtFH13 and AtFH14 as early as in ancestral angiosperms. The two proteins thus may exhibit different behavior and function, including distinct patterns of intracellular localization, which, indeed, turns out to be the case.
Formins, as known actin nucleators, can be expected to associate with actin filaments. Indeed, although AtFH14 was shown to preferentially bind microtubules, some co-localization of AtFH14 and actin was reported upon heterologous expression in mitotic tobacco BY2 cells [23]. In addition, using single molecule total internal reflection fluorescence microscopy, the FH1 and FH2 domains of AtFH14 were documented to directly bind the microfilament barbed end and act as a processive formin in vitro [22]. Nevertheless, in our in planta system co-expression of YFP-tagged full length AtFH14 or AtFH13 with actin markers did not reveal co-localization with the actin cytoskeleton, consistent with the strength, or indeed presence, of formin-microfilament association depending on the cellular context and molecular environment. Similar behavior has also been observed for the Class I formin AtFH4 which can bind and nucleate actin in vitro but does not decorate actin filaments in vivo [6]. Surprisingly, while the isolated FH2 domain of AtFH14 behaved comparably to the full length protein in co-expression with an actin marker, i.e., did not co-localize with actin fibers, the FH2 domain of AtFH13 showed a limited overlap with the actin network, suggesting differences in microfilament affinity between the two Class II formins.
Several Class II plant formins including AtFH14 were shown to interact with microtubules, sometimes more readily than with actin filaments [21,23,39,40]. We confirmed that YFP-tagged AtFH14 can co-localize with microtubules in vivo, and documented microtubule co-localization also for AtFH13. Remarkably, AtFH13 decorated the microtubules in a noticeably more discontinuous manner than AtFH14, often forming distinct particles along the microtubules. Mobility of these particles was greatly increased by the MT-depolymerizing drug oryzalin, proving that their organization is microtubule-dependent. Moreover, while the isolated FH2 domain of AtFH14 readily localized to the microtubules, the FH2 domain of AtFH13 was predominantly cytoplasmic, clearly indicating functional differences between the two FH2 domains. Analogously, metazoan formins have been documented to vary in their abilities to interact with microtubules and showed distinct localization patterns [41,42].
Interestingly, while both AtFH13 and AtFH14 associated with microtubules, only in the later the FH2 domain, which is also responsible for actin binding, was sufficient for microtubule interaction. Direct binding of FH2 domains to microtubules, or at least a specific contribution of FH2 to microtubule interaction, has been reported for several metazoan formins [41,[43][44][45], as well as for plant formins of both Class I, such as rice FH15 [46], and Class II, namely rice FH5 [39]. In other cases, notably including Arabidopsis Class I formins AtFH4 and AtFH8 [6], the formin-microtubule interaction is mediated by specific domains or motifs outside FH2. In some cases, the association between formins and microtubules may be indirect, i.e., mediated by microtubule-binding proteins [47]. Together with previous reports, the observed different localization of isolated FH2 domains derived from AtFH13 and AtFH14 documents that while microtubule association appears to be a general feature of formins, its mechanism can vary substantially even within a relatively narrow clade such as the plant Class II formins.
In addition to microtubule co-localization, we observed partial overlap of AtFH14-labeled structures with the endoplasmic reticulum, as well as peripheral association of AtFH13-labelled particles with the ER network. Some derivatives of the Arabidopsis Class I formins, AtFH4 [6] and AtFH5 [25], as well as the non-conventional Class II formin AtFH16, which lacks the PTEN domain, can localize to the ER or to ER-like intracellular structures [25], and the transmembrane Class I formin AtFH1 was recently documented to pass through several endomembrane system compartments in differentiating root tissues [35]. The metazoan INF2 formin also associates with the ER periphery [48]. Our observations, which are consistent with the PTEN-like domains of AtFH13 and AtFH14 mediating peripheral association with the ER, are thus not surprising. However, the membrane nature and compartment identity of the observed formin-positive intracytoplasmic structures remains to be established.
Crystallographic studies revealed that the functional form of formin's FH2 domain is an antiparallel dimer capable of actin nucleation or barbed end capping [13]. In the case of AtFH14, dimerization and binding of the dimer to the barbed end of microfilaments has been documented for the FH1-FH2 domain combination, as well as for the isolated FH2 domain [22]. Using the yeast two hybrid assay, we confirmed that also the FH2 domain of AtFH13 can form homodimers. Although the localization patterns of AtFH13 and AtFH14 or their FH2 domains in transiently transformed N. benthamiana epidermis differed noticeably, we nevertheless observed co-localization of a fraction of the signal, and documented that the FH2 domains of these two Class II formins can form heterodimers in the yeast two-hybrid assay.
Heterodimerization of paralogous formins might theoretically provide a possible mechanism of formin activity regulation, as well as means of increasing functional diversity of fomins [18]; however, heterodimer formation may be, and often probably is, limited by structural divergence among formin paralogs [49]. Very few experimental studies address this topic, and so far, the only evidence of FH2 domain-mediated formin heterodimerization comes from metazoa. The FH2 domains of mammalian FRL2 and FRL3 are namely able to form hetero-oligomers when combined with N-terminal dimerization domains [16]. Other reported cases of mammalian Diaphanous-like formin heterodimerization [14,15] involve interactions mediated by domains other than FH2. Our finding that isolated FH2 domains of AtFH13 and AtFH14 can form heterodimers thus presents, to our knowledge, the first documented case of exclusively FH2-mediated formin heterodimerization that might further increase the functional diversity of/not only plant formins.

Materials and Methods
For phylogenetic analysis, closest AtFH13 and AtFH14 relatives were identified by BLAST in the RefSeq section of GenBank after taxonomic restriction of the database to Arabidopsis lyrata or to Brassica sp. Only full-length sequences that retrieved the original query upon reverse BLAST search as the best match were considered. Missing exons in the predicted sequence of the AtFH13 ortholog from A. lyrata were added manually based on an experimentally sequenced partial cDNA (GenBank EFH40810). Multiple sequence alignment was constructed, manually cleaned of unreliably aligned segments and used to compute a maximum likelihood phylogenetic tree as described previously [50] except that Gamma distribution was used to model evolutionary rate differences among sites.
Constructs for transient expression in planta were prepared as follows. Vectors were generated using conventional restriction cloning or the Gateway (GW) cloning system (Invitrogen TM , Carlsbad, CA, USA). For construction of entry clones, desired inserts were amplified using a PCR reaction with specific primer sets and templates summarized in Table S1. To amplify the AtFH14, the previously published cDNA clone [23] has been used as a template. Enzymatic cloning was used to create all entry clones derived from the GW-compatible donor vector pENTRA TM 1A (Invitrogen TM ) for further construction of expression vectors. Amplified inserts were digested with appropriate restriction enzymes (Fermentas, Vilnius, Lithuania; Table S1) and further ligated into the corresponding sites of the donor vector using T4 DNA ligase (Fermentas). For PCR-based entry clone construction, the pDONR™221 (Invitrogen TM ) vector was used. The PCR products were amplified by specific primers (Table S1) and transferred into donor vector by BP Clonase II (Invitrogen TM ). To obtain expression vectors, the resulting entry clones and binary destination vector pUBC-YFP-DEST/-RFP-DEST/-GFP-DES [51], which allows C-terminal fusion of the desired protein with YFP/RFP/GFP tag were recombined using LR Clonase II (Invitrogen TM ). The combinations of entry clones and destination vectors are summarized in Table S2. All newly cloned inserts where verified by sequencing.
For Agrobacterium-mediated transient transformation, expression vectors were transferred by electroporation into Agrobacterium tumefaciens strain GV3101 or C58C1. Plasmids carrying fluorescent protein markers, namely LifeAct-RFP [52], KMD-RFP [6], ER-rk [53], were used in co-localization experiments. Leaves from 2-3 weeks old Nicotiana benthamiana (native Australian tobacco) plants grown on peat pellets (Jiffy, Zwijndrecht, Netherlands) under the 8 h dark/16 h light photoperiod at 22 • C were used. Overnight Agrobacterium cultures were diluted in the infiltration medium (50 mM MES pH 5,6; 2 mM Na 3 PO 4 ; 20 mM MgSO 4 ; 0.5% glucose, 100 µM acetosyringone). A mixture containing equal amounts of bacterial suspensions carrying each construct of interest, corresponding to final optical density (OD 600 ) 0.07 each, together with an amount of a suspension of agrobacteria expressing the viral silencing suppressor p19 [54] corresponding to OD 600 0.05, was infiltrated into the abaxial side of N. benthamiana leaves using a syringe. The fluorescent proteins were observed 48 h post-infiltration using confocal microscopy.
For the oryzalin treatment, oryzalin (Sigma-Aldrich, St Louis, MO, USA) was dissolved in dimethylsulfoxide (DMSO) as a 20 mM stock solution, stored at −20 • C, and later diluted to 10 µM final concentration with distilled water. Before observation, discs cut from infiltrated tobacco leaves were incubated in oryzalin solution (or in water for controls).
Microscopic images were obtained using the Zeiss LSM880 (Carl Zeiss, Oberkochen, Germany) confocal laser scanning microscope with a C-Apochromat 40×/1.2 W Korr FCS M27 objective. Fluorophores were excited with the 488 nm argon laser (YFP, GFP) and 561 nm laser (RFP), respectively, and detected by sensitive 32-chanell Gallium arsenide phosphide (GaAsP) spectral detector. All images were processed and analyzed using the Fiji software. All typical localization patterns that have been regularly observed in two to four independent biological replicates, each involving several leaves, are shown. Particle tracking was performed using the Fiji plugin TrackMate [55] v3.8.0 with the following settings: LoG detector, 2 µm diameter; no median filter; no sub-pixel localization; thresholding values were set to discard false-positive spots while the correctly identified spots were retained; simple LAP tracker; linking max distance = 5 µm; gap closing max distance = 5 µm; gap-closing max frame gap = 2. Data from at least five cells for each treatment with more than 200 trajectories detected were proceeded into a boxplot using the BoxPlotR tool [56]. To quantify protein co-localization, Pearson's coefficient was calculated from at least five images per construct pair using the Fiji/ImageJ plugin JACoP (Just Another Colocalisation Plugin) with manual threshold correction [57]. For statistic evaluation, an online tool [58] was used to perform one-way ANOVA with post-hoc Tukey HSD (honestly significant difference) test.
Constructs for the yeast two-hybrid assay were prepared by restriction cloning from previously prepared AtFH13 and AtFH14 clones (see above) and introduced into vectors pGADT7 (containing the GAL4 activation domain) and pGBKT7 (containing the GAL4 DNA binding domain). Primers, templates, and destination vectors are summarized in Table S3. The yeast two-hybrid assay was performed using Matchmaker GAL4 two-hybrid system 3 (Clontech, Mountain View, CA, USA) as described previously [50].
Supplementary Materials: The following are available online at http://www.mdpi.com/1422-0067/21/1/348/s1. Figure S1: Quantification of fluorescent protein-tagged AtFH13 and AtFH14 derivatives co-localization with other markers examined in this study; Table S1: Summarization of cloning methods used for each entry clone; Table S2: The combinations of entry clones and destination vectors used for generation of expression vectors used in this study; Table S3: Summarization of primers, templates, destination vectors, and restriction enzymes used to generate expression vectors for the yeast two-hybrid assay; Video S1: Co-expression of AtFH13-YFP with the microtubule marker KMD-RFP; Video S2: Co-expression of AtFH14-YFP with the microtubule marker KMD-RFP; Video S3: Co-expression of AtFH13-YFP with the endoplasmic reticulum marker ER-rk; Video S4: Co-expression of AtFH14-YFP with the endoplasmic reticulum marker ER-rk.
Funding: This research was funded by the Ministry of Education of the Czech Republic, project NPUI LO1417.

Acknowledgments:
We thank Viktor Žárský, Denisa Oulehlová, and Michal Hála for helpful discussion, Denisa Oulehlová, Haiyun Ren, and Jan Petrášek for sharing clones and vectors, MartaČadyová for technical assistance and Ondřej Šebesta from the Faculty of Science Microscopy Facility for help with confocal microscopy.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.