Cyclic AMP (cAMP) binding to G-protein-coupled receptors (GPCRs) orchestrates chemotaxis and development in Dictyostelium. By activating the RasC–TORC2–PKB (PKB is also known as AKT in mammals) module, cAMP regulates cell polarization during chemotaxis. TORC2 also mediates GPCR-dependent stimulation of adenylyl cyclase A (ACA), enhancing cAMP relay and developmental gene expression. Thus, mutants defective in the TORC2 Pia subunit (also known as Rictor in mammals) are impaired in chemotaxis and development. Near-saturation mutagenesis of a Pia mutant by random gene disruption led to selection of two suppressor mutants in which spontaneous chemotaxis and development were restored. PKB phosphorylation and chemotactic cell polarization were rescued, whereas Pia-dependent ACA stimulation was not restored but bypassed, leading to cAMP-dependent developmental gene expression. Knocking out the gene encoding the adenylylcyclase B (ACB) in the parental strain showed ACB to be essential for this process. The gene tagged in the suppressor mutants encodes a newly unidentified HECT ubiquitin ligase that is homologous to mammalian HERC1, but harbours a pleckstrin homology domain. Expression of the isolated wild-type HECT domain, but not a mutant HECT C5185S form, from this protein was sufficient to reconstitute the parental phenotype. The new ubiquitin ligase appears to regulate cell sensitivity to cAMP signalling and TORC2-dependent PKB phosphorylation.
Dictyostelium discoideum development is characterized by chemotaxis-driven aggregation of starving cells and subsequent differentiation of multicellular aggregates into fruiting bodies (Kessin, 2001). Cyclic AMP (cAMP) plays a morphogenetic role all over development (Gerisch, 1987; Kessin, 2001; Dormann et al., 2001). During the first hours of starvation, cAMP acts as chemoattractant by binding to the serpentine receptor cAR1, and stimulating adenylyl cyclase A (ACA) through the Gα2βγ protein (van Haastert and Devreotes, 2004). ACA stimulation triggers cAMP accumulation, which acts as a second messenger to regulate gene expression. Most cAMP is, however, released extracellularly, where it serves to relay the signal to distal cells (Gerisch, 1987; Devreotes, 1989). G-protein-dependent ACA activation requires the activity of two cytosolic proteins, Crac and Pianissimo (Pia) (Insall et al., 1994; Chen et al., 1997). Pia is the ortholog of Rictor, a subunit of the target of rapamycin complex 2 (TORC2), together with the serine-threonine kinase TOR, Lst8 and Rip3 (Lee et al., 2005). TORC2 is also responsible for the phosphorylation of the PKB (also known as AKT in mammals) proteins PKBR1 and PKBA (Lee et al., 2005; Kamimura et al., 2008). The TORC2–PDKA–PKB pathway is activated at the cell leading edge, where it regulates actin recruitment, and thus cell polarization and chemotaxis (Liao et al., 2010; Kamimura and Devreotes, 2010; Kamimura et al., 2008). Homologs of these proteins also function in metazoan chemotaxis, hence the importance of Dictyostelium as model organism for studying the mechanisms regulating chemotaxis and development (Bozzaro, 2013; Artemenko et al., 2014).
To identify new actors involved in chemotaxis signalling pathways, we applied saturation mutagenesis to the Dictyostelium temperature-sensitive aggregation-null mutant HSB1 (Bozzaro et al., 1987a). In this mutant, a point mutation in the piaA gene results in a single aminoacid replacement (G917D) in the Pia protein. Owing to this mutation, the cells fail to activate ACA, and thus cannot produce and relay cAMP, and aggregate, at temperatures above 18°C (Pergolizzi et al., 2002).
Here, we performed mutagenesis of the HSB1 genome by random insertion of a plasmid bearing the blasticidin resistance, leading to identification of suppressor mutants that were capable of aggregating and undergoing development to fruiting bodies. In two of these mutants, the tagged gene encoded a new protein with three conserved domains: a SPRY, PH and HECT domain. The latter displays the highest homology with the HECT domain of mammalian HERC1 ubiquitin ligases, thus we name the gene hephA (for HERC and PH domain), and the encoded protein HectPH1. Gene knockout by homologous recombination confirmed the rescue. We further showed that hephA knockout in HSB1 cells restored chemotaxis, PKBR1 and PKBA phosphorylation, short-range cAMP relay, cAMP-dependent gene expression, but not Pia-dependent adenylyl cyclase activation by cAMP pulses. Thus, hephA suppression rescues the HSB1 phenotype, but bypasses Pia (TORC2) and adenylyl cyclase signalling. By generating a double HSB1 and acrA-knockout (HSBacrA–) mutant, we further show that the acrA gene product, adenylyl cyclase B (ACB), plays an essential role in this rescue. A model is proposed whereby inactivating the HectPH1 ubiquitin ligase increases cellular sensitivity to cAMP, allowing cell development in response to very low cAMP levels, thus suggesting that HectPH1 is involved in the desensitization of cAMP signalling.
Selection of HSB1 mutant suppressors by REMI saturation genetics
To identify new components involved in cAMP signalling networks, we applied a genetic suppression approach to the HSB1 aggregation-deficient mutant (Bozzaro et al., 1987a). The aggregation defect in HSB1 depends on inability to activate ACA and, thus, produce cAMP. Although the cells respond to exogenously applied cAMP pulses by enhanced expression of cAMP-dependent developmentally regulated genes, and they are able to chemotax toward cAMP diffusing from a microcapillary, the HSB1 cells fail to relay cAMP. Thus, the final phenotype consists of a single cell monolayer (Fig. 1). We found by serendipity that this phenotype was temperature dependent, with the mutant being able to develop at temperatures up to 17°C, but being totally blocked above 18°C. The defective phenotype depended solely on a point mutation in the gene encoding Pia, resulting in a G917D amino acid change (Pergolizzi et al., 2002).
Since HSB1 was generated chemically, the strain is suitable for saturation mutagenesis mediated by random insertion of the blasticidin resistance sequence to generate genetic suppressors of the Pia mutation phenotype. Suppressors can be easily selected, based on their ability to form developing plaques on a bacterial lawn. Approximately 30,000 independent blasticidin-resistant transformants were generated in several rounds of transfection by electroporation, plated clonally with E. coli B/2 on agar and visually screened for their ability to rescue the aggregation-deficient HSB1 phenotype. Four positives clones were selected, one that was blocked at mound stage, and three that developed to fruiting bodies (Fig. 1). Two clones, #9.2 and #10.2, were further characterized genotypically and phenotypically. Clones #3.3 and #1.3 are being characterized presently.
Recovery of the flanking DNA sequences in #9.2 and #10.2 shows that gene DDB_G0286931 has been hit
To identify the genes responsible for the observed phenotype, genomic DNA from both clones was digested, and the flanking regions of the inserted plasmid were recovered and sequenced. BLAST analysis displayed sequence identity with the gene DDB_G0286931, which is 16,053 bp long and encodes a 5222-amino-acid protein (Fey et al., 2009). The gene harbours two introns and three exons. Protein analysis predicts the presence of four putative functional domains: a SPRY domain (amino acids 2620–2753), PH domain (amino acids 3834–3980), CUB domain (amino acids 4427–4512) and a HECT domain at the C-terminus (amino acids 4855–5212) (Fig. S1A). The insertion sites of pUCBsrΔBam in #9.2 and #10.2 were upstream of the PH-domain-encoding sequence, very close to each other (Fig. S1A), confirming the independent origin of the clones. The HECT domain displays 57% similarity and 38% identity with the HECT domain of mammalian HERC1 E3 ubiquitin ligases (Fig. S1B). The HERC1 family includes giant proteins, which in addition to the HECT domain all contain one or more RLD domain(s), with facultative SPRY and/or other domains (Garcia-Gonzalo and Rosa, 2005). The RLD domain is missing in the Dictyostelium protein. On the other hand, no HERC or HECT ubiquitin ligases have been described, to our knowledge, containing a PH domain. Thus, we named the gene hephA and the encoded protein HectPH1, to highlight the presence of the PH and HECT domain.
To confirm that #9.2 and #10.2 phenotype was due to REMI insertion into the DDB_G0286931 gene, a knockout strain was created by homologous recombination (Fig. S2A). Upon HSB1 transfection, colonies forming fruiting bodies on agar were obtained, and recombination in the DDB_G0286931 gene was confirmed by Southern blotting (Fig. S2B). Thus, we name the double mutant HSB1HectPH1−. The same approach was used to generate knockout mutants in the parental AX2 strain. HephA disruption in AX2 led to a 3–4 h delay in tight aggregate formation, and asynchronous postaggregative development (Fig. 2). Starving cells were also plated on agar at different densities, to test to what extent aggregate formation depended on cell density. The aggregation efficiency declined to a similar extent for AX2, AX2HectPH1− and HSB1HectPH1− cells with decreasing cell density (Fig. 2B). HSB1 failed to aggregate at all densities tested, consistent with previous data (Bozzaro et al., 1987a). Thus, inactivating HectPH1 in HSB1 restores the ability of cells to spontaneously aggregate by chemotaxis, with the cells able to make short streams (see Movie 1). It is worth mentioning that HSB1 cell aggregates formed under shaking after 5 to 8 h cAMP pulsing, once deposited on glass or agar, slowly disaggregate and fail to re-aggregate and complete development (Bozzaro et al., 1987a).
We cloned the hephA gene fragment encoding the HECT domain, and fused it with GFP, to test whether this fragment was sufficient for rescuing the HSB1HechtPH1− mutant. Sequence alignment with other HECT ubiquitin ligases highlights a conserved cysteine residue that is predicted to be necessary for transferring the ubiquitin moiety to target proteins (Fig. S1B; Scheffner et al., 1995). We thus generated a mutated HECT fragment by site-directed mutagenesis, using the wild-type (wt) pDEX-HECTwt-GFP vector as template to construct the vector pDEX-HECTC5185S-GFP. Both vectors were transfected into the HSB1HectPH1− strain, and G418-resistant cells were cloned on agar plates with bacteria to assess the phenotype. Most colonies of cells transfected with pDEX-HECTwt-GFP failed to form fruiting bodies, in sharp contrast to cells transfected with pDEX-HECTC5185S-GFP (Fig. 3A,D). Thus, overexpressing the wild-type HECT domain of HectPH1 is sufficient to abolish the rescue of aggregation seen in HSB1HectPH1−, with cells showing a HSB1 aggregation-less phenotype, whereas the C5185S mutation does not, suggesting impairment of the enzymatic activity. Similarly, in AX2HechtPH1−, expression of the HECTwt domain, but not the HECTC5185S domain, also led to cells displaying the parental AX2 phenotype (data not shown). Cells were also observed for GFP labelling, and for both plasmids a nuclear localization, confirmed by DAPI staining, was evident, although the mutant form was also found in smaller or larger clumps dispersed in the cytosol (Fig. 3B,C,E,F). Surprisingly, expression of both plasmids in the parental AX2 or HSB1 strains led only to transient fluorescence in the cell population, and selection of stable clones was unsuccessful, despite repeated attempts.
Cell polarity and chemotaxis are restored in the suppressor mutant HSB1HectPH1−
Spontaneous HSB1HectPH1− cell aggregation is accompanied by the ability to form streams (Movie 1), although these are shorter than in the AX2 strain. Since the HSB1HectPH1− mutant was able to aggregate even at low density (Fig. 2), we examined whether chemotaxis and cAMP responses were fully recovered. Upon stimulation with cAMP diffusing from a microcapillary, 5-h starved HSB1HectPH1− cells displayed an elongated morphology, moved smoothly towards the microcapillary and formed short streams, resembling AX2 wild-type cells (Fig. 4A; Table S1). cAMP-pulsed HSB1 cells, though responding chemotactically, failed to form streams and moved with reduced speed towards the capillary as single cells (Fig. 4A; Table S1), in agreement with their inability to relay cAMP (Pergolizzi et al., 2002).
To assess the chemotactic efficiency, HSB1HectPH1− and AX2HectPH1− cells were exposed to different cAMP gradients. At 0.1 mM cAMP diffusing from the capillary, the chemotaxis index (i.e. directionality for both mutants) was comparable to AX2, with its efficiency decreasing gradually with increasing distance from the microcapillary. At 0.01 mM cAMP, directionality had already decreased drastically for AX2 at a distance between 150 and 450 µm, remaining constant thereafter (Fig. 4B), very likely because cells relay the cAMP signal (McCann et al., 2010). AX2HectPH1− and HSB1HectPH1− cells displayed a higher directionality, with a gradual decrease with increasing distance up to 750–900 µm, with HSB1HectPH1− showing random motility at this latter distance range (Fig. 4B). Indeed, HSB1 cells, which are unable to relay, displayed reduced, although chemotactically still significant, directionality at both cAMP concentrations, but a rapid decrease to values corresponding to random motility (Fig. 4B). After 5 h starvation, cells were also tested for chemotaxis in the small population assay (Kamimura et al., 2009). AX2 cells were less responsive than AX2HectPH1− and HSB1HectPH1− at concentrations below 100 nM (Fig. 4C). Taken together, these results suggest that inactivating HectPH1, both in the AX2 and HSB1 genetic background, increases cell sensitivity to cAMP signals.
To assess whether this differential sensitivity to cAMP could be due to altered cAMP receptors, cAMP-binding assays were performed in 5-h starved cells. cAMP binding kinetics were roughly similar, with a maximal cAMP binding (Bmax) that was comparable for all strains except HSB1, as expected, since cells were not stimulated with cAMP pulses, and thus expressed lower levels of cAR1 receptors (Pergolizzi et al., 2002). The range of dissociation constants (Kd) was comparable for AX2, HSB1 and AX2HectPH1−, but was 2.64±0.28-fold (mean±s.e.m.) higher for HSB1HectPH1− (Fig. 4D). The higher dissociation constant displayed by HSB1HectPH1− indicates a lower affinity of membrane receptors for cAMP, which could result in increased sensitivity to cAMP (Xiao et al., 1999).
Chemotactic cell motility is regulated by a TORC2–PDKA–PKB (PKBA and PKBR1) signalling network, that transduces G-protein and RasC- or RasG-linked membrane signals to the actin cytoskeleton, leading to cell polarization and oriented movement (Lim et al., 2001; Sasaki et al., 2004; Cai et al., 2010; Zhang et al., 2008; Lee et al., 2005; Kamimura et al., 2008). PKBA and PKBR1 are transiently phosphorylated within seconds after cAMP stimulation (Meili et al., 2000; Kamimura et al., 2008; Kamimura and Devreotes, 2010). In addition PKB-dependent phosphorylated proteins, involved in cytoskeletal reorganization, have been identified (Kamimura et al., 2008). PKBR1 and PKBA activation appears to require sequential phosphorylation by TORC2 and PDKA, which phosphorylate, respectively, the hydrophobic motif (HM) and the activation loop (AL) in both proteins (Kamimura et al., 2008; Kamimura and Devreotes, 2010; Liao et al., 2010).
To assess whether this network was restored in HSB1HectPH1−, cells were stimulated with cAMP and phosphorylation events followed with antibodies recognizing specifically the phosphorylated forms of the HM and AL of PKBR1 and PKBA, as well as their phosphorylated substrates (Kamimura et al., 2009). cAMP stimulation triggered transient phosphorylation of PKBR1, PKBA and their substrates in AX2 but not in HSB1 cells (Fig. 5A). Thus, the point mutation in HSB1 piaA abrogates PKBR1 and PKBA phosphorylation and their activity, as in the Pia-null mutant (Kamimura et al., 2008; Liao et al., 2010), confirming that Pia–TORC2 kinase activity is a pre-requisite for full phosphorylation of PKBR1 and PKBA. Remarkably, the phosphorylation pattern of PKBR1, PKBA and their substrates was restored in the HSB1HectPH1− suppressor mutant (Fig. 5A). Thus, HectPH1 deletion rescued both chemotactic cell polarity and the underlying PKB phosphorylation and kinase activity. PKBR1 and PKBA phosphorylation was also analysed in the AX2HectPH1− mutant. Compared to parental AX2 cells, the phosphorylation pattern followed a similar kinetics, but phosphorylation was sustained for longer in the mutant (Fig. 5B). Taken together, these results are consistent with TORC2 activity being restored in HSB1HectPH1–, a different kinase replacing TORC2 being activated or, alternatively, a phosphatase being inhibited upon HectPH1 deletion.
G protein-linked activation of ACA is not rescued in the suppressor mutant
cAMP relay depends on GPCR-linked ACA stimulation, which requires Pia activity (Chen et al., 1997) and is defective in HSB1 (Pergolizzi et al., 2002). To test whether ACA stimulation was restored in HSB1HectPH1−, cells were synchronized with periodic cAMP pulses for 5 h under shaking and subjected to a cAMP assay. Under these conditions, in response to a cAMP pulse, AX2 cells produce a transient burst of cAMP (Fig. 6), due to transient stimulation of adenylyl cyclase (Gerisch, 1987; Devreotes, 1989). As expected, in HSB1 cells this response was absent (Fig. 6A; Bozzaro et al., 1987a). Surprisingly, no cAMP increase was detectable in HSB1HectPH1− cells as well (Fig. 6A). The experiment was repeated four times between 5 and 8 h of cAMP pulsing, with a similar trend (Fig. S3A). We also measured changes in cAMP accumulation in starving cells. In AX2, cAMP accumulated more than 10-fold during starvation, whereas in both HSB1 and HSB1HectPH1− cAMP concentration remained at vegetative level (Fig. 6B). Thus, it appears that in HSB1HectPH1−, G protein- and Pia-dependent ACA stimulation is not restored.
cAMP-dependent developmental gene expression and PKA activity in HSB1HectPH1− and in the double mutant HSB1acrA−
We investigated expression of the early aggregation genes carA and csaA, encoding the cAMP receptor cAR1 and the cell adhesion molecule csA, respectively. Expression of both genes is induced at low level by starvation and strongly stimulated by cAMP pulses (Bozzaro et al., 1987a; Mann and Firtel, 1989). Consistent with a defect in ACA activation, expression of both genes is low in HSB1, with no difference between 3 and 5 h starvation time, whereas a higher expression is observed between 3 and 5 h both in AX2 and HSB1HectPH1− (Fig. 7A; Fig. S3B). cAMP pulsing also stimulates gene expression in HSB1, in agreement with previous results (Bozzaro et al., 1987a). Thus, inactivating HectPH1 in HSB1 appears to fully restore expression of genes required for aggregation, despite spontaneous cAMP pulsing being undetectable.
The finding that developmentally regulated cAMP-dependent genes were expressed normally in HSB1HectPH1− cells suggested that PKA activity was restored. PKA is the major downstream effector of adenylyl cyclase signalling inside the cell, it is required for developmental gene expression and overexpressing the PKA catalytic subunit is sufficient to induce development in ACA-null cells (Wang and Kuspa, 1997; Mann et al., 1997; Schulkes and Schaap, 1995; Williams et al., 1993). We measured PKA activity in cell extracts by assessing phosphorylation of the substrate Kemptide. As shown in Fig. 7B, cAMP stimulated PKA activity at a comparable level in aggregation-competent AX2 and HSB1HectPH1− cell extracts, in sharp contrast to HSB1, where PKA activity remained at vegetative levels, unless the cells were pulsed with cAMP for 5 h. Thus we conclude that, similar to with cAR1 and csA, PKA fails to accumulate in HSB1 cells that are not treated with cAMP, but accumulates normally in the double HSB1HectPH1− mutant.
The findings that inactivating HectPH1 in HSB1 reconstitutes development, and that exogenous cAMP pulses rescue developmental gene expression in HSB1, but in both cases without detectable activation of the adenylyl cyclase ACA, led us to study whether adenylyl cyclase B (ACB or ACR), the product of the acrA gene (Kim et al., 1998; Soderbom et al., 1999; Meima and Schaap, 1999), might play a role in both processes. ACB is present at a low level during aggregation and increases during the postaggregative stage, in contrast to ACA, which is maximally expressed in the pre-aggregation and aggregation stage. Inactivating the acrA gene results in delayed ACA accumulation, delayed cell aggregation and formation of fruiting bodies devoid of viable spores (Soderbom et al., 1999; B.P. and S.B., unpublished results).
We generated a double mutant HSB1acrA− (Fig. S2D), treated the cells with cAMP pulses and checked for developmental gene expression. As depicted in Fig. 7A and Fig. S3B, carA and csA were expressed at an extremely low level in HSB1acrA−, well below the level found in HSB1 cells, and cAMP pulses failed to elicit any increase in gene expression. In contrast to HSB1 cells, which displayed chemotaxis to cAMP diffusing from a microcapillary, although without forming streams, cAMP-pulsed HSB1acrA− cells moved randomly, with very little if any orientation toward the cAMP source (Fig. 4A; Table S1).
We expected starving HSB1acrA− cells to display very low basal ACA activity and no ACB activity. ACA or ACB enzymatic activities can be distinguished from each other due to their differential sensitivity to Mn2+ or Mg2+, with Mn2+ activating ACA and Mg2+ preferentially ACB (Pitt et al., 1992; Meima and Schaap, 1999; Soderbom et al., 1999). Furthermore, G protein-dependent ACA stimulation can be assayed by challenging a cell lysate with the non-hydrolyzable analog GTPγS (Pitt et al., 1992). We measured adenylyl cyclase activity, and its induction with GTPγS, in HSB1acrA− or control cells at different developmental times. Extracts were prepared from all cell lines at the beginning of starvation (t0), after cAMP pulsing for 5 h under shaking (aggregation-competent cells), or from AX2 and HSB1 at mound and pre-culminant stages (both cell strains were incubated at 13°C to allow development to proceed in HSB1 cells). As HSB1acrA− cells fail to develop also at 13°C, cell extracts were prepared in parallel with the HSB1 extracts). Mn2+-dependent ACA activity increased sharply in both AX2 and HSB1 cells during the first 5 h of starvation under shaking, and at the mound stage on agar, decreasing at the pre-culminant stage. In HSB1acrA− cells, a 10- to 20-fold lower ACA activity was measured (Fig. 7C). When assayed in the presence of Mg2+, no ACB activity was detected in HSB1acrA− cells, as expected, whereas in AX2 and HSB1 there was a comparable steady increase from t0 to the pre-culminant stage (Fig. 7C). GTPγS stimulated adenylyl cyclase activity more than 10-fold in cAMP-pulsed AX2, but only minimally in HSB1 and HSB1HechtPH1− cell extracts, consistent with the requirement of Pia for G protein-dependent ACA stimulation. No stimulation was observed in HSB1acrA− cells (Fig. 7D).
In conclusion, the parental mutant strain HSB1, although having comparable basal activities of both adenylyl cyclases ACA and ACB to that in AX2, is strongly inhibited in G protein-dependent ACA activation, consistent with the temperature-sensitive defect in Pia. In contrast, HSB1acrA− fails to express ACB activity, due to ACB disruption, and displays less than 10% of the ACA activity of parental strain, even after cAMP pulsing, due to the additional defect in Pia-dependent ACA stimulation. The inability to detect GTPγS stimulation of adenylyl cyclase in these latter cells is likely due to the negligible level of basal ACA activity.
Suppression, by random mutagenesis, of a pre-existing mutation is a powerful tool for examining gene function or interactions. In this paper, we exploited REMI-mediated random insertion of blasticidin-resistance in the genome of the nitrosoguanidine aggregation-deficient mutant HSB1 to generate revertant mutants, thus identifying suppressor genes. In two clones in which development was fully restored, the same gene was disrupted, which encoded a newly discovered HECT E3 ubiquitin ligase, which had a ubiquitin ligase domain that was homologous to the HECT domain of mammalian HERC1. HERC1 belongs to the class 1 family of HECT E3 ubiquitin ligases, which also includes HERC2 and the small HERC proteins, and which usually contain a SPRY domain (Grau-Bove et al., 2013; Scheffner and Kumar, 2014). Although HECT E3 ubiquitin ligases appear to regulate many physiological processes, including membrane receptor and transporter trafficking, mTOR signalling, and transcription or chromatin remodelling, the exact function of HERC1 and HERC2 remain unclear (Sanchez-Tena et al., 2016; Rotin and Kumar, 2009; Garcia-Gonzalo and Rosa, 2005). The HECT domain of HERC1 has been shown to conjugate ubiquitin through its active site cysteine, indicating that it is very likely a functional ubiquitin ligase, but no clear substrates have been identified so far (Sanchez-Tena et al., 2016). HERC2 has been shown to regulate the stability of several proteins involved in DNA damage repair. Additionally, it targets the deubiquitinating enzyme USP33, involved in cancer cell migration, and β2-adrenergic receptor signalling (Chan et al., 2014).
Similar to mammalian HERC1 and HERC2, Dictyostelium HectPH1 is a giant protein with a conserved HECT domain at the C terminus, and a PH and a SPRY domain upstream, but it does not possess the RLD domains typical of HERC1 and HERC2. The isolated PH domain fused with GFP displays cytosolic distribution, enrichment in the nucleus and sometimes in the plasma membrane (Fig. S4) suggesting that HectPH1 can transitorily bind plasma membrane phosphoinositides, where it could display its ubiquitin ligase activity. Interestingly, mammalian HERC1 binds to phosphatidylinositol 4,5-bisphosphate sites via the RLD1 domain (Garcia-Gonzalo and Rosa, 2005). The SPRY domain could mediate binding of potential ubiquitylation substrates (Nishiya et al., 2011) or facilitate HectPH1 interaction with other proteins (Tae et al., 2009). The 2500-amino-acid N-terminal stretch upstream of the SPRY domain does not display any recognizable domains, but harbours several motifs that could be involved in regulation, including GSK3, PKA and Ca2+/calmodulin kinase phosphorylation sites.
The HECT domain contains a conserved cysteine residue (LPEAQTCFFTL) that is essential for activity (Scheffner et al., 1995; Huang et al., 1999). We have shown that transfecting the HECTwt domain is sufficient to abrogate the rescue of aggregation in HSB1HectPH1−, restoring the aggregation-less HSB1 phenotype, whereas replacing the cysteine residue with serine (HECTC5185S) results in an inactive HECT, when overexpressed in the suppressor background.
The HSB1HectPH1− mutant displays almost complete reversion of the aggregation-less HSB1 phenotype, despite the finding that Pia-dependent ACA activation was not rescued, thus confirming that Pia is still inactive in HSB1HectPH1− cells. Although Pia, like the other interacting subunits of the TORC2 complex, fails to form a stable complex with TOR (Cai et al., 2010), ACA activation in Dictyostelium appears to require a pre-formed TORC2 complex (Lee et al., 2005).
How can this complex phenotype be explained? It is worth remembering that exogenously applied cAMP pulses rescue developmentally regulated gene expression in HSB1, similar to in the piaA-null mutant (Chen et al., 1997), but the aggregates formed under shaking disaggregate once deposited on glass, and fail to proceed further in development (Bozzaro et al., 1987a). On the other hand, HSB1 cells can aggregate and form fruiting bodies on agar if mixed with 10–20% AX2 cells (Bozzaro et al., 1987a), suggesting that synergy with even a few wild-type cells acting as a autonomous, long-lasting source of cAMP is sufficient to rescue HSB1 cells, despite their inability to relay cAMP signals. This does not occur if the acrA gene, encoding adenylyl cyclase B, is inactivated in HSB1. HSB1acrA− cells also fail to respond to exogenous cAMP pulses, in contrast to parental HSB1 cells, suggesting that ACB is essential for transducing exogenous cAMP signals, at least when G protein–ACA stimulation is impaired. This notwithstanding, there is only a negligible increase in cAMP accumulation in HSB1HectPH1− compared to cAMP-pulsed HSB1 cells, despite HSB1HectPH1− being able to aggregate and complete development.
If cAMP concentration remains at a very low level in HSB1HectPH1−, an intriguing possibility is that disruption of the HectPH1 ubiquitin ligase could lead to hypersensitivity to cAMP signals, such that low concentrations of cAMP could activate downstream pathway(s) regulating developmental gene expression and chemotaxis, thus allowing cells to aggregate and form fruiting bodies. In favour of this hypothesis, both HSB1HectPH1− and AX2HechPH1− displayed a more efficient chemotactic index than AX2, particularly at lower cAMP concentrations. Hypersensitivity to cAMP could also explain the observed effect of HectPH1 disruption in the AX2 background, namely a delay of few hours in the beginning of aggregation and a lower efficiency of aggregation. In contrast to HSB1HectPH1−, the AX2HectPH1− strain would resemble AX2 cells exposed to high concentrations of cAMP, which is known to inhibit, rather than stimulate, cAMP-dependent developmentally-regulated gene expression as well as cAMP relay (Rossier et al., 1979; Mann and Firtel, 1987; Brzostowski et al., 2013).
Hypersensitivity may occur at different levels, starting with the cAMP receptors to downstream pathways. Desensitization of the cAMP receptors could, for example, be altered in the suppressor mutant. Little is known about cAMP receptor desensitization. Upon cAMP binding, the cAR1 receptors are phosphorylated, with phosphorylation inducing loss of ligand binding (Kim et al., 1997). Inhibiting phosphorylation results in unaltered ligand binding, which leads to formation of smaller aggregates and disruption of cell streaming (Brzostowski et al., 2013), a phenotype resembling the HSB1HectPH1− mutant. It is possible that HectPH1 ubiquitylates the cAR1 receptors or arrestins (Cao et al., 2014), with its disruption favouring membrane exposure of the receptors, thus increasing sensitivity to cAMP. A few E3 ligases attaching ubiquitin to specific GPCRs have been identified in other systems (Haglund and Dikic, 2012) (Alonso and Friedman, 2013; Marchese and Trejo, 2013). Persistent signal sensitivity could also result from altered receptor degradation due to impaired ubiquitylation of proteins involved in endosome-to-lysosome trafficking (Feinstein et al., 2011; Haglund and Dikic, 2012; Holleman and Marchese, 2014; Alonso and Friedman, 2013). The finding that the Kd of cAMP receptors in cAMP-binding assays is higher in HSB1HectPH1− may point in this direction, suggesting that two sequential events, linked to Pia and HectPH1 being both defective, are required for changing the affinity of the receptors. More experiments are required to unravel the dynamics of cAMP receptors and its regulation, and analysis of both mutants should prove to be very useful in this regard. An alternative possibility is that Pia could be a direct substrate of HectPH1, such that inactivating the ubiquitin ligase could result in increased accumulation of the protein. Overexpression of the mutated protein resulted in partial recovery of the mutant phenotype (Pergolizzi et al., 2002), thus this possibility cannot be excluded.
HectPH1 could also regulate pathways downstream of cAR1. By excluding the G protein- and Pia-dependent ACA activation, which is not rescued in the HSB1HectPH1− mutant, and is not essential for stimulating developmental gene expression, as it is bypassed by exogenous cAMP pulsing in HSB1 and piaA-null cells, the postulated increased sensitivity to cAMP signalling could depend on a pathway parallel to ACA. A potential candidate is an ACB-linked pathway to PKA or its downstream effectors regulating expression of developmental genes. The contribution of ACB in early Dictyostelium development is debated (Anjard et al., 2001; Pitt et al., 1992; Meima and Schaap, 1999). Our results clearly show that the HSB1 mutant, deficient in ACA activation, is able to respond to exogenous cAMP pulses inducing expression of cAMP-dependent genes. Inactivating the ACB-encoding acrA gene in these cells totally inhibits both chemotaxis toward cAMP and cAMP-dependent gene expression. Thus, we suggest that ACB plays a role in mediating both processes, although this role is obscured in wild-type cells by the activity of ACA, whose expression is in any case delayed in wild-type cells in which acrA has been deleted (Soderbom et al., 1999; B.P. and S.B., unpublished results). As depicted in Fig. 8, if HectPH1 downregulates a component of the ACB-linked pathway to PKA and gene expression, its inactivation would resemble ACA-minus cells overexpressing the PKA catalytic subunit, which are able to develop (Wang and Kuspa, 1997). PKA could phosphorylate the GATA family transcription factor GataC (Loomis, 2014), which has been recently shown to be phosphorylated also by the GSK3 ortholog GskA (Cai et al., 2014). Periodic cAMP oscillations coordinate GataC phosphorylation with its nucleo-cytoplasmic shuttling, thus modulating its transcriptional activity. Stable nuclear localization of GataC induces precocious expression of developmentally regulated genes, including carA and csA (Cai et al., 2014). Interestingly, the activity of mammalian GATA transcription factors is regulated by phosphorylation, acetylation and ubiquitylation (Nakajima et al., 2015; Kitagawa et al., 2014). Whether GataC is a potential substrate of HectPH1 is under investigation.
PKBR1, and to a lesser extent PKBA, appear to be required for chemotactic cell polarization (Meili et al., 1999; Meili et al., 2000). Phosphorylation of PKBR1 and PKBA has been shown to depend on sequential activity of TORC2 and PDK1, which phosphorylate PKB hydrophobic motifs and activation loops, respectively (Kamimura et al., 2008; Kamimura and Devreotes, 2010; Liao et al., 2010). In cAMP-pulsed HSB1 cells, similar to in the piaA-null mutant, PKBR1 and PKBA are not phosphorylated, in agreement with Pia–TORC2 signalling being inactive. Both are, however, phosphorylated in the HSB1HectPH1− suppressor mutant, leading to phosphorylation of PKB substrates. It is possible that HectPH1 inactivation in the suppressor mutant stabilizes a putative alternative kinase to TORC2, or that its inactivation results in inhibition of a phosphatase that is antagonistic to TORC2, in the assumption that TORC2 is operating in HSB1 at a basal low level (Fig. 8). It is worth remembering that PKB regulation is a complex event involving multiple Ras proteins and downstream pathways working in parallel, cooperatively and antagonistically (Meili et al., 1999; Kamimura and Devreotes, 2010; Cai et al., 2010; Liao et al., 2010; Rodriguez Pino et al., 2015). Analysis of the HSB1HectPH1− mutant could contribute to a better understanding of the pathways regulating PKB activity.
Like many HERC1 ubiquitin ligases, HectPH1 is a giant protein, but differs from large and small HERCs owing to the absence of RLD motifs and the presence of a PH domain. We have no direct evidence for ubiquitin ligase activity, but overexpressing the HECTwt, in contrast to HECTC5185S, domain in HSB1HectPH1− restored the HSB1 phenotype. It is possible that, in the absence of the long N-terminal sequence, the overexpressed HECTwt domain binds E2 enzymes, transferring the ubiquitin moiety indiscriminately to specific substrates responsible for the HSB1 phenotype in addition to non-specific substrates (Weiss et al., 2010; Park et al., 2009).
It is intriguing that the HECTwt domain fused with GFP is concentrated exclusively in the nucleus, both in vegetative and aggregating HSB1HectPH1− and AX2HectPH1− cells, whereas the mutated HECTC5185S domain is also found in the cytosol. The fusion protein is 70 kDa in size, thus nuclear enrichment cannot be due to passive diffusion. Since the plasmid constructs do not contain nuclear localization signals typical of Dictyostelium (Catalano and O'Day, 2012), it is possible that the HECT domain is co-transported to the nucleus bound to a potential substrate. To what extent the nuclear localization is an artefact of the isolated HECT fragment is an open question. It is worth remembering that mammalian HERC2 is enriched in the nucleus, where it ubiquitylates several substrates. Future investigations will be directed to devising strategies to clone and express, if not the full protein, at least the entire region encompassing the SPRY, PH and HECT domains that could be used as bait to capture potential HectPH1 substrates and for biochemical and molecular genetic studies.
MATERIALS AND METHODS
All strains were cultured in AX2 medium (Watts and Ashworth, 1970) at 23°C under shaking at 150 rpm in a Kuehner climoshaker (Birsfelden, Switzerland) (Bozzaro et al., 1987b). Blasticidin (InvivoGen, Toulouse, France) at 10 µg/ml final concentration was added to knockout mutants. Cells expressing GFP-fused proteins were cultured in the presence of 20 µg/ml G418 (Sigma-Aldrich, Milan, Italy). For growth on bacteria, spores or cells were mixed with E. coli B/2 and plated on nutrient agar plates (Bozzaro and Merkl, 1985; Bozzaro et al., 1987b).
For development, cells were washed twice in 0.017 M Soerensen Na/K-phosphate buffer, pH 6.1, resuspended at 107 per ml and plated on non-nutrient agar (Bozzaro et al., 1987b). For development under shaking, cells were resuspended at a concentration of 107 per ml in Soerensen phosphate buffer and incubated in the Kuehner climoshaker.
REMI mutagenesis, mutant suppressor screening and plasmid rescue
HSB1 cells were mutagenized by restriction enzyme-mediated insertion (REMI) of BamHI-linearized pUCBsrΔBam (Adachi et al., 1994), electroporated in the presence of MboI (ThermoFisher Scientific, Waltham, MA, USA), and treated with 10 µg/ml blasticidin for 10 days (Shaulsky et al., 1996). Drug-resistant cells were plated clonally on nutrient agar in association with E. coli B/2. Colonies were screened visually for rescue of the HSB1 phenotype, and positive clones transferred into liquid culture for growth with blasticidin. Plasmid rescue was performed as described by Kuspa and Loomis (1992), using NdeI and EcoRV restriction enzymes for re-circularization of genomic DNA. Primers matching the bsr-cassette were used to sequence the genomic flanking regions, and corresponding genes were searched using the NCBI and the Dictyostelium database (www.dictybase.org) with the BLAST server. Protein sequence analysis was performed with the Pfam database (pfam.xfam.org). Macvector software was used for DNA sequence analysis and restriction map construction.
Generation of knockout strains
The hephA-knockout vector pBLS-hephA-bsr was constructed as depicted in Fig. S2A. After digestion with EcoRI and XbaI, the linearized DNA (10 μg) was electroporated in HSB1 or AX2 cells (Pang et al., 1999). For generating the HSB1acrA− mutant, HSB1 cells were transfected with the pDG1100 plasmid (Soderbom et al., 1999). In both cases, blasticidin-resistant cells were cloned in 96-wells plates and checked for gene disruption by Southern blotting or PCR analysis (Fig. S2).
Generation of HECTwt–GFP, HECTC5185S–GFP, and GFP–PH(HectPH1)
The AX2 hephA fragment, encoding the HectPH1 HECT domain, was amplified using HD_FWD and HD_REV primers (Table S1) and cloned into the pGemT vector. A NcoI blunt-ended fragment was then inserted into the gfp 5′-end sequence in the original pDEX-H plasmid (Westphal et al., 1997), previously digested with EcoR I and blunt-ended, generating the plasmid pDEX-HECTwt-GFP. This vector was used as template for site-directed mutagenesis. Cysteine residue 5185 in the HECT domain was mutated into a serine residue with the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA), using the primers C5185S_FWD and C5185S_REV (Table S2). The resulting plasmid was named pDEX-HECTC5185SGFP.
To generate GFP–PH(HectPH1), the PH-encoding fragment was amplified using PH.D_FWD and PH.D_REV primers (Table S2) and cloned into pGemT vector. A EcoRi/ClaI fragment was inserted into the gfp 3′end sequence of pDEX-H, generating the plasmid named pDEX-GFP-PH(HectPH1).
Nucleic acid analysis
Total RNA was purified using TRIzol reagent (Life Technologies, Gaithersburg, MD). RNA electrophoresis, northern and Southern blotting were performed as described previously (Bracco et al., 1997).
Starving cells were disaggregated by vortexing and plated onto 35-mm diameter glass-based dishes (Iwaki, Tokyo, Japan) at a density of 105 cell/cm2. Chemotaxis was evaluated by local stimulation with a microcapillary (Femtotips 1, Eppendorf, Milan, Italy), filled with cAMP, using an Eppendorf micromanipulator (Peracino et al., 1998). Images were captured with intervals varying between 0.66 and 1.8 s and recorded in a Panasonic video-recorder connected to a ZVS-47DE camera, mounted on Axiovert 200 microscope (Zeiss, Oberkochen, Germany). Alternatively, images were acquired digitally with intervals of 15 s with a Lumenera Infinity 3 camera (Lumenera Corporation, Ottawa, Canada) mounted on the same microscope. Movies were analysed with ImageJ Manual Tracking and Chemotaxis/Migration plugins for determining the chemotaxis index (i.e. the directionality; the ratio between Euclidean and accumulated distance). Motility speed (accumulated distance over time) and cell polarity (ratio between length and wide) were calculated manually in at least 30 cells per movie.
The chemotaxis small population assay was performed as described previously (Kamimura et al., 2009), except that 0.8% agar containing 5 mM caffeine and Soerensen phosphate buffer were used.
cAMP binding and Scatchard analysis
cAMP binding analysis was performed as described by van Haastert (2006). Briefly, cells were incubated with 5 mM caffeine for 10 min under shaking, washed and resuspended at 108 cell/ml in Soerensen phosphate buffer. Aliquots of 0.08 ml were incubated with a mixture containing 0.3 mM [H3]cAMP (Perkin Elmer, Milan, Italy), 50 mM dithiothreitol, 5 mM caffeine, and 50 to 9700 nM cAMP. After 45 s incubation at room temperature, cells were centrifuged at 14,000 g for 30 s, and the pellet was treated with 0.1 ml of 0.1 M acetic acid and dissolved in 1.3 ml scintillation fluid. Radioactivity was measured with a LS-6500 Multi-Purpose Scintillation Counter (Beckman, Indianapolis, USA). Curve fitting for cAMP saturation binding data and Scatchard plots was achieved by non-linear regression, using Prism software GraphPad (GraphPad Inc., San Diego, CA).
For cAMP-stimulated adenylyl cyclase activity, starving cells at 2×107 cells/ml were treated with 40 nM cAMP pulses every 6 min for 5 to 8 h. Immediately before and after a cAMP pulse, cell aliquots were collected at every minute, lysed in 3.5% perchloric acid, neutralized and assayed for total cAMP in cell lysate (Bussolino et al., 1991), using the Biotrack cAMP 125I assay kit (GE Healthcare Europe, Life Sciences, Buckingamshire, UK).
The in vitro Mg2+- or Mn2+-dependent adenylyl cyclase assay was performed as described previously (Kim et al., 1998). GTPγS stimulation of adenylyl cyclase was assayed as described previously (Lilly and Devreotes, 1994; Pergolizzi et al., 2002), except that IBMX and DTT were added to inhibit cAMP phosphodiesterases.
For the PKA assay, starving cells were resuspended in 0.5 ml of cold extraction buffer (20 mM Tris-HCl pH 7.5, 4 mM MgCl2, 10 mM β-mercaptoethanol, 1 µg/ml leupeptin and aprotinin) and lysed by pressing through 3 µm-pore Nucleopore filters. The lysates were clarified by centrifugation and assayed by using the SignaTECT cAMP-dependent protein kinase assay system (Promega, Madison, WI), according to the manufacturer's instructions.
PKBR1, PKBA and PKB substrate phosphorylation was assayed as described previously (Kamimura et al., 2009), after pulsing the cells with cAMP for 5 h.
Fluorescence microscopy imaging
Cells expressing GFP-fused proteins were transferred onto 36-cm2 glass coverslips equipped with plastic rings for observation in a confocal Zeiss LSM510 microscope equipped with a 100× objective. Confocal series images were taken as described previously (Peracino et al., 2010; Buracco et al., 2015).
We thank the late W. F. Loomis for plasmid pDG1100, and A. Kamimura for helpful suggestions on PKB phosphorylation assays.
B.P. and E. B. planned and conducted the experiments, collecting the data. E.B. and B.P. wrote the initial draft. S.B. conceived and supervized the study, revising the final manuscript.
This work was supported by a research grant of the Compagnia di San Paolo (12-CSP-C03-065) to S.B. and research funding of the Università degli Studi di Torino to B.P. and S.B.
The authors declare no competing or financial interests.