Cell type-specific filamin complex regulation by a novel class of HECT ubiquitin ligase is required for normal cell motility and patterning

Multicellular development requires the differentiation of different cell types and their correct positioning within the embryo. In most organisms, this developmental patterning process requires some cells to move from their site of differentiation to different positions. For directed cell movement to occur, cells must be able to sense their position and orient accordingly, organise their motile machinery and maintain appropriate contacts with neighbouring cells. When these processes fail it can lead to severe developmental disorders and disease. Therefore, understanding how cell movements are coordinated within a three-dimensional multicellular environment represents one of the major challenges in developmental biology.

Cell movements are fundamental throughout development and are known to underpin processes such as gastrulation and the formation of the nervous system. One less-studied example, however, is illustrated by a mode of pattern formation in which different cell types differentiate randomly intermingled with each other, but then subsequently move and sort out to form discrete patterns. Cell sorting has been the subject of much research since it was demonstrated that dissociated cells from amphibian embryos adhere to form random aggregates that subsequently sort out according to the germ layer of their origin (Townes and Holtfreter, 1955). However, it is now widely recognised that a similar mechanism also operates during normal development and is broadly evolutionarily conserved (Kay and Thompson, 2009). For example, during the earliest stages of mouse embryonic development the inner cell mass is divided into the pluripotent epiblast, which gives rise to the embryo proper, and the primitive endoderm, which forms the yolk sac. Prior to the segregation of these lineages, the precursors can already be detected by the expression of lineage-specific transcription factors. However, rather than the distribution of the precursor cells being position dependent, they show a ‘salt and pepper’ distribution. These two different cell types subsequently sort out into their appropriate layers by the late blastocyst stage (Chazaud et al., 2006; Chisholm and Houliston, 1987; Dietrich and Hiiragi, 2007; Kurimoto et al., 2006; Plusa et al., 2008).

Perhaps the best illustration of patterning by sorting out is during the developmental cycle of the social amoeba Dictyostelium discoideum. Dictyostelium development is triggered by the onset of starvation, resulting in the aggregation of separate amoebae in response to extracellular cAMP, which acts as a chemoattractant (Kessin, 2001). The resulting ‘mound’ then develops into the motile ‘slug’ and finally, when conditions are favourable, a ‘fruiting body’ is formed. This terminally differentiated structure consists of spore cells, which are alive and are dispersed into the soil, and stalk cells, which are dead and function to hold aloft the spore cells (Kessin, 2001). The prestalk cells can be further subdivided into prestalk A cells (pstA, found at the very tip of the slug), prestalk AB cells (pstAB, a cone of cells within the A region), prestalk B cells (enriched at the prestalk/prespore boundary and dotted around the prespore region) and prestalk O cells (pstO, a collar of cells behind the pstA region) (Williams, 2006). A number of signalling molecules are essential for proper cell type specification. The most studied of these is differentiation-inducing factor 1 (DIF-1) (Kay, 1998), which is essential for the specification of pstO and pstB cells (Keller and Thompson, 2008; Saito et al., 2008; Thompson and Kay, 2000). DIF-1 is known to affect gene transcription in prestalk cells via the activation of a number of transcription factors, including STATc (Fukuzawa et al., 2003), DimA (Thompson et al., 2004a), DimB (Huang et al., 2006; Zhukovskaya et al., 2006) and GATAc (Keller and Thompson, 2008).

It is now generally accepted that the different cell types are initially specified intermingled, without positional information, followed by subsequent cell sorting and pattern formation (Kay and Thompson, 2009; Thompson et al., 2004a; Williams et al., 1989a). For example, the cells destined to become stalk cells (prestalk cells) first differentiate in the mound, intermingled with prespore cells, and then sort to the front of the slug, whereas the prespore cells are found in the rear of the slug (Williams et al., 1989b). Furthermore, if cell movement is blocked at the mound stage, when differentiation begins, the cell types remain intermingled and no pattern is formed (Kay and Thompson, 2009; Thompson et al., 2004b). Different models have been proposed to explain how differential motility and cell sorting are regulated in Dictyostelium and other organisms. For example, it has been suggested that cells could sort due to their specific adhesive properties or surface tension (Clow et al., 2000; Jiang et al., 1998; Siu et al., 2004). Furthermore, there is good evidence that differential chemotactic cell motility may play an important role. Indeed, Dictyostelium prestalk and prespore cells exhibit different speeds of movement in response to a cAMP stimulus (Early et al., 1995; Lam et al., 1981; Matsukuma and Durston, 1979; Siegert and Weijer, 1995; Traynor et al., 1992).

Consequently, it has been proposed that the sorting of prestalk cells is likely to require the coordination of differentiation, motility and adhesion. As DIF-1 regulates pstO cell differentiation, we therefore reasoned that DIF-1 targets would represent candidate genes for pstO cell sorting. However, relatively few DIF-1-regulated genes have been identified (Maruo et al., 2004). Furthermore, none of these has been shown to be required for cell fate choice or sorting (Morrison et al., 1994). It is of vital importance to identify downstream DIF-1 targets to elucidate the mechanisms that drive patterning. We have found that Dictyostelium HfnA is induced by DIF and is expressed in pstO cells. The hfnA gene encodes a protein with significant sequence homology to both HECT ubiquitin ligases and filamin.

HECT domain-containing ubiquitin ligases target proteins for destruction via the proteasome. They play widespread biological roles and have been extensively linked to disease states in humans (Scheffner and Staub, 2007). However, relatively few of their targets have been identified (Scheffner and Staub, 2007). HECT ligases typically consist of a C-terminal HECT domain and a variable N-terminal domain (Huibregtse et al., 1995). The HECT domain is responsible for the addition of ubiquitin subunits to specific proteins (Scheffner et al., 1995), whereas the N-terminal domain is thought to confer target specificity by binding to its particular substrate (Scheffner and Staub, 2007). As HfnA contains N-terminal sequences similar to those found in filamin, we hypothesised that HfnA might be involved in filamin complex regulation during Dictyostelium development.

Filamins are F-actin-binding proteins that are thought to function as adaptor proteins, bringing together elements of the cytoskeleton and signal transduction proteins (Bandala-Sanchez et al., 2006; Popowicz et al., 2006). Filamins contain actin-binding domains at their N-termini, followed by several filamin repeats. Each filamin repeat is ~100 amino acids long and contains an immunoglobulin fold (Noegel et al., 1989). The repeats function as binding domains, through which filamins can dimerise (Fucini et al., 1999). Over 30 diverse protein interactions with vertebrate filamins have been identified, which include transcription factors such as FOXC (Berry et al., 2005), many cytoskeletal and adhesion proteins such as migfilin (Tu et al., 2003), GTPases (Ohta et al., 1999), kinases and receptors (Popowicz et al., 2006). Unsurprisingly for a protein with such complex interactions, filamins are associated with a number of disease states and developmental disorders. In humans, there are three filamins (A, B and C), with filamin A (FLNA) being the most widely studied. Loss-of-function mutations in FLNA lead to periventricular nodular heterotopia, a developmental disorder that results from defective neuronal migration (Fox et al., 1998). However, other mutations that alter filamin function also result in a large number of different developmental malformations, indicating that the precise regulation of filamin is essential (Robertson, 2005). Filamin has multiple modes of regulation, including phosphorylation and degradation by calpain (Chen and Stracher, 1989); although regulation by ubiquitylation has not been reported, the identification of an evolutionarily conserved filamin domain-containing HECT ligase has suggested that ubiquitylation could play a vital role in filamin complex regulation.

Dictyostelium contains one filamin (Fln) homologue, which consists of an actin-binding domain at its N-terminus followed by six filamin repeat domains. Filamin functions as a dimer, and dimerisation requires repeats 5 and 6 (Fucini et al., 1999). In Dictyostelium, filamin has been shown to play roles in the organisation of the actin cytoskeleton, in pseudopod formation and in the speed of movement of single cells. During development, however, the major role of filamin appears to be in slug phototaxis (Fisher et al., 1997; Khaire et al., 2007), which requires repeats 2-6 (Annesley et al., 2007). Filamin has been shown to function in the tip of the slug, in the pstA region, as expression of full-length filamin under the control of a pstA-specific promoter is able to completely rescue the Fln-null phototaxis phenotype (Khaire et al., 2007). Filamin has been proposed to act as an adaptor protein at the front of the slug during this process, as it is found within a large signalling complex that comprises many proteins, including RasD, ERKB, GRP125 and PKB (Bandala-Sanchez et al., 2006). Whereas numerous binding partners of filamin have been identified in mammalian cells (Popowicz et al., 2006), relatively few have been identified in Dictyostelium. The best characterised is filamin-interacting protein (FIP), which is developmentally regulated, whereas filamin is not (Knuth et al., 2004). FIP has been demonstrated to interact with filamin in vivo and is required for the correct regulation of filamin (Knuth et al., 2004). FIP colocalises with filamin in Dictyostelium cells and has been shown to directly bind to filamin between repeats 2 and 4 (Knuth et al., 2004). Although attempts to make an FIP-null mutant have been unsuccessful (Knuth et al., 2004), overexpression of a C-terminal fragment of FIP (cFIP) has been reported to cause defects consistent with a role in filamin regulation (Knuth et al., 2004).

We demonstrate here that the DIF-1-induced, filamin domain-containing HECT ubiquitin ligase HfnA affects filamin complex activity. Furthermore, we show that HfnA is necessary for normal development and correct pstO cell motility in a three-dimensional environment during pattern formation. These findings suggest that differential HfnA expression and filamin complex activity provide a key regulatory step in the mechanism by which heterogeneous salt-and-pepper gene expression can lead to organised tissue patterning.

Dictyostelium discoideum strains were grown in axenic medium (HL5) (Sussman, 1987) or on Klebsiella lawns. For development, axenically growing cells were washed in KK2 (16.1 mM KH₂PO₄, 3.7 mM K₂HPO₄) and plated at 2×10⁷ cells per 10-cm 1.5% KK2 agar plate and left in the dark. For analysis of Dictyostelium plaque ‘slugging’, Dictyostelium cells were spotted onto Klebsiella lawns on SM agar plates and left in a humid chamber for 5 days.

Microarray experiments were performed as described (Van Driessche et al., 2002). Each sample was analysed by a two-colour assay in which test samples were compared with a common pooled reference sample of total RNA from several wild-type developmental time points. The samples were hybridised to a microarray of 5669 unique targets. A single-chip normalisation procedure was used to allow for multi-array comparisons. Each hybridisation target was printed twice on the array, allowing for single-chip normalisation. At least two hybridisation experiments were carried out for each RNA extraction. Two biological replications were also performed. Technical and biological replicates were averaged to generate normalised timecourse data (Katoh et al., 2004; Van Driessche et al., 2002).

Primers flanking the upstream and downstream regions of hfnA (5′-GATGAGCGGGGAGAATTCTTG-3′ and 5′-AGAAACCATCTCCTTCAGTGA-3′) were used to generate a 3.5 kb genomic fragment that was cloned into TOPO TA 2.1 vector (Invitrogen). A gene-disruption construct was generated by in vitro transposition (Abe et al., 2003) of a blasticidin resistance cassette into the HECT domain. This construct was introduced into wild-type AX4 cells by electroporation and mutants were selected at 10 μg/ml blasticidin. Blasticidin-resistant colonies were screened by PCR for homologous recombination.

The filamin gene abpC was disrupted in the hfnA⁻ background using hygromycin as a selectable marker (Egelhoff et al., 1989). The C-terminal FIP fragment expression construct (Knuth et al., 2004) was introduced into cells by electroporation and selected at 10 μg/ml G418.

A 0.9 kb fragment upstream of the hfnA transcription start site was cloned into vector pDGal17 (Harwood and Drury, 1990). This construct and ecmO-lacZ, ecmA-lacZ and ecmAO-lacZ were electroporated into AX4 and mutant cells and the resultant transformants selected at 10 μg/ml G418. lacZ whole-mount staining was then performed as previously described (Dingermann et al., 1989) as well as measurements of β-galactosidase activity upon DIF-1 induction in monolayer culture, as described (Thompson et al., 2004a). For each experiment, several thousand transformants were pooled and the result confirmed with duplicate electroporations.

For analysis of movement during multicellular development, transformants were developed to the mound stage. A thin strip of agar was cut out and transferred to a humid chamber on its side to view structures ‘side on’. Cell tracking was performed using ImageJ software and the Manual Tracking plug-in. To measure chemotaxis of dissociated mound stage cells, mounds were transferred to KK2 medium containing 0.1% pronase, and dissociated by gently syringing through a 20G needle. Then 2-5×10⁴ cells were plated onto glass-bottomed dishes (World Precision Instruments) and allowed to settle for 30 minutes. Cells were then stimulated with a micropipette filled with KK2 containing 2 mM MgSO₄ and 100 μM cAMP. Centroid analysis was performed using ImageJ software. In addition to cell speed, chemotaxis index and persistence were calculated. Persistence was calculated from the cosine of the angle between two consecutive movement steps. Chemotaxis index was determined as the ratio of net cell displacement over the course of the movie and total distance travelled.

Qualitative phototaxis tests were performed as described (Darcy et al., 1994). A toothpick scraping of amoebae was transferred from a colony growing on a Klebsiella lawn to the centre of charcoal agar plates (5% activated charcoal, 1% agar). The plates were incubated for 48 hours at 21°C with a lateral light source and the phototaxis was scored. For quantitative phototaxis tests, amoebae were harvested from mass plates, thoroughly washed free of bacteria, and suspended in saline at the appropriate dilutions. Then 20 μl of the amoebal dilutions were inoculated onto a 1 cm² area in the centre of each charcoal agar plate. The resulting cell densities ranged from ~2×10⁶ to 2.5×10⁷ cells/cm². Phototaxis was again scored after 48 hours incubation at 21°C with a lateral light source. For both the qualitative and quantitative assays, slug trails were transferred to PVC discs and stained with Coomassie Blue before being digitised. Qualitative phototaxis test results are presented as complete, digitised trails plotted from a common origin. For statistical analysis of quantitative phototaxis experiments, the start and end points of slug trails were digitised and the orientation was analysed using directional statistics (P. R. Fisher, PhD thesis, Australian National University, 1981) (Fisher and Annesley, 2006).

The Dictyostelium hfnA gene was identified as a DIF-1 target from a microarray study in which the transcriptional profiles of wild type and DIF signalling mutants were compared (our unpublished results; see Materials and methods). HfnA (DDB_G0283983) contains a C-terminal HECT domain and an N-terminal filamin domain. Searches indicate that proteins containing both filamin and HECT domains are present in most organisms and adopt the same modular architecture (Fig. 1A). Alignments of the filamin domains of these proteins indicate that they have clear homology to those found in filamin proteins from humans and Dictyostelium (Fig. 1B). Furthermore, it is also apparent that these proteins share more similarities with each other than with true filamin proteins (Fig. 1B,C), indicating that this is an evolutionarily conserved group of related proteins.

In order to confirm that hfnA exhibits DIF-dependent developmental expression, we generated an hfnA reporter construct in which upstream hfnA promoter sequences were used to drive the expression of lacZ. First, hfnA expression was induced by DIF-1 in a monolayer culture of Dictyostelium cells (Fig. 2A). Secondly, we used upstream hfnA promoter sequences to drive lacZ expression to test whether hfnA exhibits DIF-1-dependent cell type-specific developmental regulation. Previous studies have shown that DIF-1 acts on pstO cells, which are mainly localised to the collar region of the slug (Thompson and Kay, 2000). Accordingly, hfnA expression was highly enriched in the pstO cell regions at the slug stage of wild-type development (Fig. 2B). Although expression levels are much reduced in culminants, any cells that did stain were found in the largely pstO cell-derived upper cup (Fig. 2B).

In order to determine the role of HfnA during development, we generated two independent hfnA⁻ mutants by homologous recombination. One mutant line was generated with a 3.5 kb insertion at the beginning of the HECT domain and another with a large deletion within the hfnA coding sequence that removes both the HECT and filamin domains (Fig. 3A). The mutants exhibit indistinguishable phenotypes and are thus likely to be null alleles. When developed on non-nutrient agar plates, hfnA⁻ slugs appeared smaller and misshapen compared with wild type (Fig. 3B). The hfnA⁻ cells also showed a so-called ‘slugger’ phenotype (Sussman et al., 1978), in which culmination is delayed such that the slugs were still migrating after 24 hours of development on non-nutrient agar plates (Fig. 3B). When grown on Klebsiella lawns, the hfnA⁻ mutants also exhibited another manifestation of the slugger phenotype in that a large number of slug trails could be seen extending out from the edge of a growing plaque (Fig. 3C).

In order to investigate the cause of the hfnA⁻ mutant developmental phenotypes, we examined the expression of cell type-specific markers. As hfnA expression is DIF-1 induced and hfnA is expressed in pstO cells, this suggested a role in pstO cell differentiation or patterning. We found that ecmAO-lacZ (a marker of both pstA and pstO cells – the entire length of the prestalk region) strongly stained the slug tip, but was expressed more weakly behind this, in the pstO region (although scattered cells were also visible within the prespore zone), when hfnA⁻ slugs were compared with wild type (Fig. 4). This staining pattern suggested a specific defect in pstO cell differentiation. Consistent with this, no defect in ecmA-lacZ expression (which normally stains the very tip of the slug – the pstA cells) could be detected (Fig. 4). Most importantly, ecmO-lacZ expression (a specific pstO cell marker that is normally expressed in the collar region behind the pstA cells) was clearly aberrant (Fig. 4). In the hfnA⁻ mutant, pstO cells were still visible but strikingly mislocalised towards the rear of the slugs. This defect is therefore distinct from that described for other DIF signalling mutants, in which pstO cells fail to differentiate.

As pstO cells fail to occupy the collar region at the early slug stage of development in the hfnA⁻ mutant, these findings suggest that HfnA is important for targeting or keeping the pstO cells in the correct place. To address this, we carefully followed the behaviour of pstO cells over a developmental time series. At the tight mound stage, the earliest stage at which ecmO-lacZ expression can be detected, staining was scattered in both wild type and mutant (Fig. 5). However, as soon as a tip could be detected, pstO cells were seen to accumulate below the tip in the wild type but not the hfnA⁻ mutant. This phenotype was most apparent at the early finger and slug stages. However, when slugs were examined over a timecourse, pstO cells were found in a pattern indistinguishable from that seen in wild type at the late slug stage (Fig. 5A). These findings therefore suggest that HfnA is needed for the correct early targeting or motility of pstO cells to the collar region of the developing finger and early slug, as pstO cells are mislocalised in the hfnA⁻ mutant (Fig. 5B).

It is thought that prestalk cell sorting is driven by differential adhesion and/or chemotaxis towards a localised source of cAMP produced by the developing tip cells. If so, the delay in pstO cell sorting could be due to their failure to sense and orient to the chemoattractant stimulus, or to a simple failure to move within the three-dimensional environment in the mound (or some combination of the two). To distinguish between these possibilities, we first directly measured the speed and directionality of prestalk and prespore cell movement, using real-time imaging, during the transition from mound to finger stages. In order to follow pstO cell movement, co-transformant lines were generated that express ecmAO-RFP and ecmO-GFP. Cells expressing both markers were defined as pstO cells. For prespore cells, psA-RFP transformants were used. PstO and prespore cells were tracked to determine their speed and directionality of movement. From this, we found that mutant pstO cells move at a significantly slower speed, which was almost half that of wild-type controls (Fig. 6A). This is consistent with the finding that complete pstO sorting takes ~6-8 hours in the mutant compared with 3-4 hours in the wild type (Fig. 5A). More importantly, this does not reflect a general motility defect in the mutant, as the speed of prespore movement was indistinguishable between the wild type and mutant (Fig. 6B). Finally, and as expected, the speed of pstO cell movement was significantly higher than that of prespore cells, although the difference was not as marked in the hfnA⁻ mutant as in the wild type (Fig. 6A,B). Importantly, however, when the directionality of movement of each cell type was compared, no differences between wild-type and mutant cells were detected (Fig. 6A,B).

Since the speed, but not directionality, of movement of mutant pstO cells was affected, this suggested a defect in cell motility but not chemoattractant sensing. In order to confirm this we compared the ability of dissociated wild-type and hfnA⁻ pstO cells to move towards a micropipette containing cAMP. Under these conditions, we detected no defect in the speed or directionality of movement of the mutant pstO cells, supporting the idea that the ability of hfnA⁻ cells to sense a chemoattractant is unaffected (Fig. 6C). Furthermore, these findings suggest that the defect is specific to movement in a three-dimensional environment. Since both differential adhesion and chemotaxis have been implicated in prestalk cell sorting during normal development, we tested whether adhesion might be affected. Wild-type and hfnA⁻ cells expressing the prestalk marker ecmAO-RFP were dissociated at the finger stage of development and allowed to reaggregate in suspension. The extent of prestalk cell aggregation was indistinguishable between wild-type and mutant cells (Fig. 7). This indicates that neither the adhesive properties nor the chemotactic responses of mutant prestalk cells are affected. Taken together, these findings suggest that the defect in pstO cell sorting in the hfnA⁻ mutant is most likely due to a failure to drive movement in the three-dimensional environment of multicellular development.

HECT domain-containing ubiquitin ligases often contain a proposed N-terminal substrate-targeting domain (Scheffner and Staub, 2007). HfnA contains an N-terminal filamin domain, which is rarely found in Dictyostelium proteins. We therefore reasoned that filamin, or filamin complex members, could be targets of HfnA during development. Indeed, a role for filamin in developmental patterning has previously been suggested because Fln-null cells fail to sort to the tip of slugs when developed in chimera with wild-type cells. A simple hypothesis, therefore, is that aspects of the hfnA⁻ phenotype are caused by increased filamin, or filamin complex, activity. To test whether filamin levels were altered, we analysed the amount of filamin present in hfnA⁻ versus wild-type cells at different developmental stages (Fig. 8). However, we found no significant differences in filamin levels between the wild type and hfnA⁻ mutant.

Another possibility is that HfnA targets a filamin-interacting protein, thereby affecting the activity of filamin complexes. To address this, we tested for genetic interactions by examining the effects of increasing or decreasing filamin complex activity in wild-type or hfnA⁻ genetic backgrounds. First, we found that disruption of the filamin gene abpC (Fln⁻; loss of filamin complex activity) in the hfnA⁻ background (hfnA⁻/Fln⁻) effectively rescues the slug morphology and delayed culmination defects of the hfnA⁻ mutant (Fig. 9A). Second, we found that filamin disruption results in the ‘opposite’ pstO cell phenotype to that caused by hfnA disruption: PstO cells were more strongly localised, resulting in a tight collar with few, if any, scattered cells in the rear of slugs (Fig. 9B). Furthermore, filamin disruption in an hfnA⁻ background resulted in an intermediate ‘rescued’ phenotype, thus demonstrating a genetic interaction between hfnA and Fln (Fig. 9B). Finally, overexpression of filamin or the C-terminal portion of the known filamin binding partner FIP (Knuth et al., 2004) in the wild type led to pstO mislocalisation (Fig. 9C). The pstO sorting defect seen in the hfnA⁻ mutant is thus phenocopied when filamin complex activity is artificially elevated. It is also interesting that the effects of increasing filamin complex activity through FIP overexpression do not depend on the presence of filamin, as they are observed in a Fln-null background (Fig. 9C). This supports the idea that the effects are due to misregulation of filamin complex activity rather than of filamin activity itself. Most importantly, as filamin is an actin-binding adapter protein that is known to be involved in cell movement and adhesion (Popowicz et al., 2006), altered filamin complex activity in the hfnA⁻ mutant provides an explanation for the observed mislocalisation of pstO cells.

Filamin is essential for slug phototaxis and filamin-deficient strains exhibit severe phototaxis defects (Fisher et al., 1997; Khaire et al., 2007; Knuth et al., 2004). It is thought that filamin is present in a ‘phototaxis complex’ that is involved in sensing light and in signal transduction at the tip of the slug (Bandala-Sanchez et al., 2006). We tested whether HfnA is also required to regulate normal responses to light. Interestingly, the slugging phenotype of the hfnA⁻ mutant requires a lateral light source, as incubating the mutants in the dark resulted in no slugging at all from the edge of the plaque (Fig. 10). As wild-type slugs do not usually migrate out of the plaque, even in the light, hfnA⁻ slugs appear more sensitive to light. This phenotype is consistent with hyperactivity of the filamin complex, as filamin has been shown to be required for slug phototaxis. Consistent with this idea, we found that the slugging phenotype was rescued in the hfnA⁻ mutant by inactivation of filamin (Fig. 9). To confirm this, we tested the ability of hfnA⁻ slugs to orient accurately during phototaxis. We found that phototaxis by hfnA⁻ slugs was enhanced compared with the parental wild-type controls (Fig. 11). In quantitative experiments, the accuracy of phototaxis by hfnA⁻ slugs was ~1.5-fold greater than that of the wild type (Fig. 12A and Table 1). Since phototaxis accuracy depends quantitatively on filamin complex activity (Annesley et al., 2007), this suggests signalling hyperactivity by the filamin complex when HfnA is absent. As expected, hfnA⁻/Fln⁻ slugs, like Fln-null mutants, exhibited severe phototaxis defects. Finally, although cFIP has previously been reported to show dominant-negative effects on filamin complex activity and thus impair phototaxis, we found that overexpression of cFIP in the genetic background used in this study improves phototaxis efficiency (Figs 11 and 12, Table 1). Together, these findings provide further support for the idea that HfnA inhibits the signalling activity of the filamin complex.

Our findings provide a crucial mechanistic advance in our understanding of pattern formation by salt-and-pepper differentiation and sorting out in Dictyostelium. It has been proposed that heterogeneities in sensitivity to the DIF-1 signal within cell populations generate cell type-specific gene expression, which in turn drives cell sorting (Kay and Thompson, 2009). Our study identifies the first such gene, hfnA. HfnA is the first member to be characterised of a novel group of evolutionarily conserved proteins, the HECT ligases, which contain an N-terminal filamin domain. In Dictyostelium, HfnA is DIF induced, expressed in pstO cells and essential for normal development. We have demonstrated that HfnA is important for the timing of culmination, the regulation of slug movement and localisation of pstO cells. HfnA appears to regulate the activity of filamin complexes, which are in turn needed for phototactic slug migration and correct cell type motility and patterning during development. These studies provide the first step in defining the molecular relationship between the input DIF-1 signal and output cell type-specific behaviours that lead to sorting out.

Pattern formation is a complex and coordinated process. HfnA, through its downregulation of specific targets, is essential for normal development. We have demonstrated that filamin is a possible target for HfnA, as filamin disruption rescues, and filamin hyperactivation phenocopies, the pstO sorting and slug phototaxis phenotypes of the hfnA⁻ mutant. Despite this genetic link, we found that filamin levels are unchanged in the hfnA⁻ mutant. One possibility is that any change in filamin level might be extremely small as it may be restricted to the pstO cell population of the slug. However, we found that the expression of a C-terminal fragment of FIP phenocopies a lack of HfnA. FIP levels are, however, unchanged in the hfnA⁻ mutant (data not shown). Therefore, our data support an alternative explanation in which neither filamin nor FIP is the direct target of HfnA. Instead, other filamin interactors might be targeted, leading to altered filamin complex activity. The identification of these putative targets is likely to further reveal the roles of HfnA in DIF-1-dependent patterning events.

Cell type-specific sorting in Dictyostelium is thought to require both differential cell motility and adhesion. Cells must respond to the cAMP chemoattractant produced by the tip and navigate their way through the three-dimensional environment of other cells. Furthermore, the different prestalk subtypes must also behave differently in order for relatively complex cell type organisation to arise. A possible mechanism to explain this is that all cells have an equal ability to sense the chemoattractant but move at different speeds or with different strengths. For example, by modulating adhesion or intercellular contacts, it is possible to envisage a situation in which the most adhesive cells move more slowly because they are, in effect, tethered. Alternatively, each cell type could differ in the motile force generated in response to a directional signal. Our data support a role for the filamin complex in the regulation of cell type-specific speed of movement, but not in adhesion or direction sensing. Since the filamin complex is known to regulate actin cytoskeleton dynamics, we suggest that this might underlie the pstO sorting defects observed. Indeed, some support for this idea comes from the finding that other mutants with defects in cytoskeletal organisation exhibit cell type sorting defects, especially when developed in chimera with wild-type cells (Chisholm and Firtel, 2004).

Ubiquitylation represents an evolutionarily conserved regulatory mechanism. RING finger ubiquitin ligases and ancillary proteins have been shown to be important for normal Dictyostelium development (Clark et al., 1997; Ennis et al., 2000; Nelson et al., 2000; Wang and Kuspa, 2002; Whitney et al., 2006) and even for slug phototactic migration (Whitney et al., 2006). Searches indicate that the Dictyostelium genome is likely to encode six HECT ligases, each containing a C-terminal HECT domain and most an identifiable N-terminal domain that might be responsible for binding to its specific target. These N-terminal domains include PH domains (DDB0187197), protein kinase and RCC1 domains (DDB0219986) and HEAT/Armadillo repeats (DDB0235186). Our studies provide the first characterisation of a Dictyostelium HECT ligase and suggest that in this case an N-terminal filamin domain specifically targets filamin complexes. It will be of great interest to define the role of other HECT ligases in Dictyostelium, as these studies might also uncover evolutionarily conserved regulatory mechanisms.

Filamins are adaptor proteins that have numerous binding partners and are able to bring together many signalling and cytoskeletal elements. The precise regulation of filamin activity is of crucial importance for normal filamin complex function. Consequently, filamin has multiple modes of regulation. Protein stability is central to the regulation of filamin complex activity. For example, filamin itself is subject to degradation by calpain (Onji et al., 1987). Filamin is also phosphorylated by a number of serine-threonine kinases, and its phosphorylation by protein kinase A inhibits its cleavage by calpain (Chen and Stracher, 1989). The importance of protein stability in its regulation is further illustrated by its mammalian binding partner FILIP (filamin A interacting protein), which also downregulates filamin through degradation (Nagano et al., 2002). Despite stability being a core theme in the regulation of filamin complex activity, interactions with ubiquitin ligases and the ubiquitylation system have not previously been demonstrated. For example, although ubiquitin ligases that contain both a RING finger and a filamin domain are found in some organisms (Schaefer et al., 2000; Wan et al., 2000), no role in filamin complex regulation has been described. Most importantly, we have found that a family of HECT ligase- and filamin repeat-containing proteins closely related to HfnA is widely conserved through evolution. We therefore propose that the degradation of filamin interactors via the proteasome might represent a novel and widespread mechanism for regulating filamin complex activity.

We thank Prof. Gad Shaulsky for help in conceiving the project and encouragement throughout, Neil Buttery for help with statistical analyses, and Dr Angelika Noegel for providing the filamin antibody and the cFIP construct. This work was supported by project grants from the Wellcome Trust and Medical Research Council, a Lister Institute of Preventive Medicine Research Prize to C.R.L.T. and funds to P.R.F. from the Thyne Reid Memorial Trusts. Deposited in PMC for release after 6 months.