An ordered pathway for the assembly of fungal ESCRT-containing ambient pH signalling complexes at the plasma membrane

Underlying the ability of many fungi to thrive over a broad range of pH is the existence of a transcriptional regulatory system ensuring that the synthesis of molecules mediating adaptation of cells to acidic or alkaline pH is tailored to the needs imposed by the environment. This mechanism has been most extensively studied in Aspergillus nidulans and Saccharomyces cerevisiae (Peñalva et al., 2008; Peñalva and Arst, 2004), but the robust phenotypes resulting from inactivation of this regulatory system and the larger multinucleated cells in A. nidulans make it ideally suited for studies correlating gene function with subcellular localisation of the cognate gene products.

The key player in this ambient-pH-mediated regulatory system is the zinc-finger transcription factor PacC. In A. nidulans PacC undergoes two-step proteolytic processing activation in response to alkaline ambient pH (Díez et al., 2002; Espeso et al., 2000; Hervás-Aguilar et al., 2007; Orejas et al., 1995), but only the first of these sequential proteolytic steps, known as signalling proteolysis, is regulated by ambient pH. Current evidence indicates that the signalling protease is the calpain-like PalB, which directly interacts with endosomal sorting complex required for transport (ESCRT)-III (Peñas et al., 2007; Rodríguez-Galán et al., 2009). However, it is not known how and where PalB is activated by extracellular alkaline ambient pH.

In addition to PalB, five dedicated proteins form the Pal signalling pathway mediating PacC activation. Three appear to be components of an alkaline-pH-sensing module in the plasma membrane: the helper transmembrane protein PalI, the seven-transmembrane (7-TMD) receptor PalH and the PalH-coupled arrestin-like protein PalF (Calcagno-Pizarelli et al., 2007; Herranz et al., 2005; Hervás-Aguilar et al., 2010a). Similar to mammalian β-arrestins, PalF is ubiquitylated in a receptor (PalH)- and signal (alkaline pH)-dependent manner (Herranz et al., 2005; Lefkowitz and Shenoy, 2005) and indeed ubiquitylation of arrestin-like PalF leads to constitutive signalling and PacC processing, bypassing the need for the PalH receptor (Hervás-Aguilar et al., 2010a). The remaining two proteins, PalA and PalC have been shown to interact with Vps32, the main component of ESCRT-III (Galindo et al., 2007; Vincent et al., 2003). PalA additionally interacts with YPx[L/I] motifs in PacC and thus appears to mediate recruitment of the transcription factor to the pH signalling machinery (Vincent et al., 2003; Xu and Mitchell, 2001).

In agreement with the known interactions of PalA, PalB and PalC, work in S. cerevisiae established that ESCRT-I, ESCRT-II and two ESCRT-III components, Vps20 and Vps32, are required for pH signalling (Xu et al., 2004). These findings supported the hypothesis that pH signalling would be mediated by a specialised class of ESCRT- and Pal-containing complexes assembled on the surface of endosomes, perhaps connected to the plasma membrane sensor module by endocytic trafficking (Peñalva et al., 2008), in agreement with the fact that mammalian β-arrestin ubiquitylation promotes the endocytic internalisation of arrestin-receptor complexes (Lefkowitz and Shenoy, 2005).

However, evidence argues against endocytosis and endosomal ESCRT complexes playing a role in pH signalling, because: (1) both in S. cerevisiae (Xu et al., 2004) and in A. nidulans (Calcagno-Pizarelli et al., 2011) the key ESCRT-0 component Vps27 linking multivesicular body biogenesis to phosphatidylinositol 3-phosphate (PtdIns3P)-containing endosomes (PtdIns3P is a landmark of multivesicular endosomes) is fully dispensable for signalling, and indeed PacC processing is unaffected in rabB deletion (rabBΔ) mutants unable to recruit the PtdIns3-kinase Vps34 to endosomes (Abenza et al., 2010; Calcagno-Pizarelli et al., 2011); (2) in vivo, alkaline pH drives the Vps32 interactor PalC to plasma-membrane-associated structures rather than to endosomes (Galindo et al., 2007); (3) artificial ubiquitylation of arrestin-like PalF leads to receptor-independent signalling and promotes recruitment of PalC to these cortical structures under acidic (non-signalling) conditions; (4) PalF promotes the plasma membrane, rather than endosomal, localisation of the PalH receptor (Hervás-Aguilar et al., 2010a). The finding that the S. cerevisiae PalF orthologue Rim8p interacts with Vps23p (Herrador et al., 2009) explained the conundrum posed by the crucial involvement of ESCRT-I, -II and -III in pH signalling and yet the complete dispensability of Vps27, suggesting a model in which receptor-activated arrestin-like PalF (Rim8p in S. cerevisiae) recruits Vps23 (ESCRT-I), thus priming the recruitment of ESCRT-II and -III to pH signalling complexes (Herrador et al., 2009; Hervás-Aguilar et al., 2010a). PalC recruitment to cortical structures (Galindo et al., 2007) and the finding that overexpression of Rim8p relocalises overexpressed Vps23p to the plasma membrane (Herrador et al., 2009) lend credence to the possibility that pH signalling occurs in a class of specialised ESCRT-containing complexes associated with the plasma membrane. This possibility would gain strong credibility if, for example, Vps23 expressed at physiological levels were shown to be recruited to the plasma membrane at alkaline pH.

The severely debilitating phenotype of A. nidulans null mutations previously precluded investigation of ESCRT in pH signalling in this organism. However, the recent finding (Calcagno-Pizarelli et al., 2011) that vigorous growth of null ESCRT strains can be rescued by mutations in the sltA regulatory gene, which by themselves do not affect pH signalling, allowed examination of the localisation of fluorescently tagged reporters, expressed at physiological levels, in wild-type and deficient ESCRT and pal (i.e. pH signalling) backgrounds. Our results provide strong evidence for a model in which PalF-dependent recruitment of Vps23 to the plasma membrane and the subsequent formation of ESCRT-containing cortical complexes enable the ordered recruitment of downstream pH signalling proteins at these locations to ultimately mediate activation of ESCRT-associated calpain-like PalB. Thus in these pH signalling complexes ESCRT-III would play signal transduction roles without facilitating membrane scission.

Because pH signalling is independent of Vps27, current models assume that Vps23 plays the ESCRT-recruiting role in pH signalling complexes that Vps27 plays in the multivesicular body pathway, and Vps32 acts as a scaffold for these signalling complexes. We examined the involvement of ESCRT proteins in the alkaline-ambient-pH-dependent recruitment of PalC–GFP to cortical structures. Although cells carrying single sltA⁻ mutations, used to rescue the severely debilitating ESCRT-null mutant phenotype, behaved in the same manner as the wild-type, the additional presence of vps23Δ or vps32Δ virtually abolished PalC recruitment (Fig. 1A,C). By contrast, vps27Δ did not. Western blots (Fig. 1B) ruled out the possibility that the sharp decrease in PalC–GFP cortical structures resulting from vps23Δ or vps32Δ reflects increased turnover of the fluorescent protein. Thus cortical structures to which PalC is recruited meet the requirements for bona fide pH signalling complexes.

To determine whether the second Vps32 interactor PalA is also recruited to these cortical structures we constructed a gene-replaced strain expressing physiological levels of PalA–GFP. Under acidic conditions, PalA localises to the cytosol, to a few motile cytosolic structures (almost certainly early endosomes) and to the spindle pole bodies (SPBs) (supplementary material Fig. S1; SPB localisation will be addressed elsewhere). Upon shifting cells to alkaline conditions, PalA–GFP was clearly recruited to cortical structures resembling those seen with PalC (Fig. 1D–F). As for PalC, PalA recruitment was dependent on the 7-TMD receptor PalH (Fig. 1F, Fig. 2E). Fig. 2A shows that, in cells expressing physiological levels of fluorescently tagged PalA and PalC from gene-replaced alleles, PalA–mCherry and PalC–GFP colocalise in the same alkaline-pH-induced cortical structures, strongly indicating that activation of the pH signal transduction pathway involves the recruitment of both PalC and PalA to plasma-membrane-associated signalling complexes. Anti-GFP western blots showed that the physiological levels of PalA and PalC are very similar (Fig. 2C), as are the fluorescence levels of cortical pH signalling foci, irrespective of whether they are labelled with PalA–or PalC–GFP (data not shown).

PalA, like PalC, is a Vps32 interactor (Vincent et al., 2003; Xu and Mitchell, 2001). Thus we addressed whether PalA cortical recruitment also involves this ESCRT-III principal component. Alkaline pH shift experiments demonstrated that PalA cortical foci were essentially absent in vps32Δ cells (Fig. 2B), despite the fact that neither of the mutant backgrounds used (sltA54 or sltA54 vps32Δ) affected the steady-state levels of PalA–GFP (Fig. 2C). The dependence of PalA and PalC cortical recruitment on Vps32 strongly implicates Vps32 in the formation of the plasma-membrane-associated signalling complexes.

PalC–GFP recruitment is dependent on PalF and independent of PalA (Galindo et al., 2007; Hervás-Aguilar et al., 2010a). By contrast, PalA recruitment is fully dependent on PalC (Fig. 2D,E, palCΔ) and independent of PalB [Fig. 2D,E, palB38 is a classical palB null allele (Peñas et al., 2007)]. These key data show that PalC acts downstream of PalF and upstream of PalA and that PalA acts upstream of the calpain-like PalB, thus establishing the order of action of Pal proteins (see Discussion).

Previous work (Galindo et al., 2007) suggested that PalC recruitment to cortical structures is transient. To analyse this transient recruitment, we cultured hyphae expressing PalC–GFP under acidic (pH 5) conditions in a microscopy chamber and shifted them to alkaline conditions (pH 7.2) before starting to acquire time-lapse sequences in the GFP channel (supplementary material Movies 1 and 2). PalC–GFP at these cortical structures appeared as recurring fluorescent dots (Fig. 3A,C). These dots, unlike endocytic sites labelled with AbpA^Abp1 (Araujo-Bazán et al., 2008), are essentially static. Therefore, the arrival and departure of PalC at these structures could be tracked with montages made of individual frames (Fig. 3C) or with kymographs, in which PalC–GFP gave rise to discontinuous lines perpendicular to the distance scale. The length of these lines reflects the time of residence of PalC–GFP (the only source of PalC) in the structures (Fig. 3A, kymograph horizontal lines), which was determined to be ~36 seconds (n=50 structures). PalA–GFP-containing cortical structures also appeared as recurring foci (supplementary material Movie 3; Fig. 3B,D), leading to short lines in kymographs (Fig. 3B). The half-life of ‘alkaline’ PalA–GFP foci (~68 seconds; n=50) was longer than that of PalC–GFP foci. For either PalC–GFP or PalA–GFP foci, fluorescence recurrently ‘oscillated’ at the same cortical locations (Fig. 3A–D), suggesting the existence of specialised signalling domains in the plasma membrane. Thus, all the above experiments strongly suggest that activation of the ambient pH signalling pathway results in the formation of membrane-associated complexes transiently containing the Bro1 domain protein PalA and the Bro1-domain-like protein PalC.

We next used a strain co-expressing (at physiological levels) PalC–GFP with PalA–mCherry to acquire dual-channel time series of cells shifted to alkaline conditions ‘on the stage’. The time resolution achieved was 4.4 frames/second. Frame-by-frame visual inspection of single foci showed that PalC precedes PalA in its arrival at and departure from these structures (supplementary material Movie 4, and Fig. 3E, lower panel, showing one example as a dual channel composition). This was confirmed by determining average signals in the corresponding GFP and mCherry frames. For any given structure these values were normalised to the frame showing the maximal signal in each channel (i.e. the ‘peak’ intensity for each channel). These relative intensity values were plotted versus time. To allow comparison between different series and/or structures, the time point immediately preceding that at which the PalA (mCherry) signal for any given structure was increased over background was arbitrarily chosen as the zero time point. The plots illustrating the individual data showed that the initiation of GFP fluorescence clearly preceded that of mCherry fluorescence (supplementary material Fig. S2 shows primary data for six structures). The conclusion that PalC precedes PalA in arrival at the pH signalling structures was further reinforced after averaging data for 50 structures (Fig. 3E). This conclusion agrees with the above finding that PalA localisation is dependent on PalC but independent of PalB.

ESCRT-III forms a complex on membranes that is disassembled by Vps4 (Teis et al., 2008). The complete dependence that pH signalling complexes have on the major ESCRT-III component Vps32 is consistent with a mechanism in which plasma-membrane-associated foci containing ESCRT-III polymers mediate recruitment of the Vps32 interactors PalA and PalC. If so, reduced levels of the Vps4 ATPase would be expected to increase the time of residence of PalC and PalA in the foci. To test this prediction, we replaced the resident gene encoding Vps4 with two regulatable expression alleles of Vps4, using the nitrite reductase promoter niiA^p (Hervás-Aguilar and Peñalva, 2010), which is nitrate-inducible and ammonium repressible. These two alleles, denoted vps4-1 and vps4-ha3, drive expression of wild-type Vps4 and (HA)3-Vps4, respectively (Fig. 4). They are the only source of Vps4 in strains carrying them.

ESCRT gene deletions cause a severely debilitating phenotype (Calcagno-Pizarelli et al., 2011). Thus, Vps4 downregulation was expected to impair growth markedly. Indeed vps4-1 strains grew very poorly on ammonium, indicating that, under such conditions, Vps4 levels are limiting (Fig. 4A). By contrast, vps4-1 strains grew as well as the wild type on urea, a ‘neutral’ (i.e. non-inducing and non-repressing) nitrogen source that results in basal levels of expression of niiA (Fig. 4A). Unlike urea, full induction of vps4-1, with nitrate, impaired growth (Fig. 4A). Therefore, these data strongly suggest that in vps4-1 cells the downregulated levels of Vps4 on ammonium are insufficient to support growth, whereas those on nitrate are excessively high, leading to toxicity. Western blot analysis of vps4-1 cells using a polyclonal anti-Vps4 antiserum (Fig. 4B) demonstrated that in cells cultured on ammonium Vps4 was virtually undetectable, whereas on nitrate the level was markedly higher than in the wild type. On urea, levels were detectable but below wild-type levels, indicating that Vps4 is normally in excess, as urea levels suffice for virtually normal growth.

vps4-ha3 mutants did not grow on ammonium either (Fig. 4A). However, vps4-ha3 strains, unlike vps4-1 strains, showed impaired growth on urea, strongly indicating that attachment of (HA)3 to the N-terminal MIT domain reduces Vps4 function. To confirm this and to monitor the extent of (HA)3-Vps4 downregulation achieved with vps4-ha3, relative to the physiological levels, we constructed a transgene driving (HA)3-Vps4 expression under the control of vps4^p. In single copy, this vps4^p::(HA)3-Vps4 transgene does not complement the vps4-1 growth defect on ammonium, indicating that (HA)3-Vps4 is hypofunctional (data not shown). Anti-HA western blots showed that, as for untagged Vps4, the tagged (HA)3-Vps4 urea-cultured levels obtained with niiA^p (vps4-ha3 allele) were lower than those attained with the physiological vps4^p promoter (in a strain carrying the wild-type vps4 allele, to maintain viability; Fig. 4C). The level of vps4-ha3 in ammonium-cultured cells was barely detectable (Fig. 4C). In contrast to vps4-1 strains, vps4-ha3 strains grew like wild-type cells on nitrate, correlating with elevated (HA)3-Vps4 levels (Fig. 4C), strongly indicating that overexpression compensates for impaired function. In summary, vps4-1 and vps4-ha3 cells cultured on ammonium are markedly deficient in Vps4 function. vps4-ha3 also behaves as a partial loss-of-function allele on urea but resembles the wild type on nitrate. By contrast, the levels of untagged Vps4 (vps4-1) on urea are functionally sufficient but the levels on nitrate impair growth because of overexpression toxicity.

We next addressed the effects of Vps4 downregulation in the alkaline-pH-induced recruitment of PalC–GFP to cortical foci. None of the tested conditions (i.e. levels of expression) affected the formation of PalC–GFP foci induced by alkaline pH (supplementary material Fig. S3; note that vps4-1 spores were able to germinate and give rise to germlings on ammonium). However, kymographs from time-lapse sequences (Fig. 4D) showed that whereas recruitment was transient under Vps4-sufficient (nitrate or urea) conditions, Vps4-deficient conditions (ammonium) dramatically increased the time of residence of PalC–GFP at these structures, such that they were seen as continuous vertical lines for the complete duration of the kymographs (Fig. 4D). vps4-ha3 cells behaved similarly to wild-type cells on nitrate (data not shown). However, both ammonium (Fig. 4D) and urea (data not shown) led to an effect similar to that caused by ammonium downregulation of wild-type Vps4. These data are strong evidence that ESCRT-III components localise to cortical sites, where they mediate the transient recruitment of PalC and PalA until disassembled by Vps4.

PacC undergoes two-step proteolytic activation. Alkaline ambient pH results in the Pal-pathway-dependent conversion of full length PacC72 to PacC53, a committed intermediate that is converted into the fully processed, functional form PacC27 by the proteasome (Fig. 5). We hypothesised that inefficient disassembly of pH signalling structures resulting from Vps4 deficiency should lead to PacC proteolytic activation under inappropriate circumstances. In vps4-1 cells cultured on urea or nitrate, or in vps4-ha3 cells cultured on nitrate, PacC proteolytic processing proceeded as in wild-type cells (Fig. 5). In marked contrast, in vps4-1 cells cultured on ammonium, or in vps4-ha3 cells cultured on urea or ammonium, conditions leading to deficient Vps4 activity, processing assays unambiguously established that PacC72 is converted into PacC53 and PacC27 in a pH-signalling-independent manner. Under acidic conditions PacC53 and PacC27 levels, normally below detection in the wild type, were high, apparently at the expense of PacC72, which was very low or even undetectable (Fig. 5, red boxes). Vps4-deficient conditions do not increase the number of signalling foci under acidic conditions (their abundance being alkaline-pH-dependent as in the wild type; supplementary material Fig. S3). Therefore, these results indicate that the abnormally long-lived pH signalling foci in the mutants amplify the basal levels of PacC processing to an extent that normally occurs exclusively under alkaline pH conditions.

An important question is what determines the plasma membrane recruitment of ESCRT-III. Experiments in S. cerevisiae pointed to Vps23 as one connection between the alkaline pH sensing module and ESCRTs (Herrador et al., 2009). Pull-down experiments using the A. nidulans Vps23 UEV domain strongly supported this conclusion. GST fusion proteins carrying either full length Vps23 (not shown) or the Vps23^UEV domain (residues 1–159 of A. nidulans Vps23) pulled down arrestin-like PalF-(HA)3 from extracts as efficiently as a PalH cytosolic tail construct (residues 349–760 including the two arrestin-binding regions) (Herranz et al., 2005), but did not pull-down the unrelated bait Vps41 (Fig. 6A,B). PalF-(HA)3 extracts prepared after growth in acidic and alkaline (thus leading to PalF ubiquitylation) conditions showed that although GST–Vps23^UEV pull-down material would appear to be very slightly enriched in the ubiquitylated forms of PalF, the non-ubiquitylated forms were also efficiently bound (Fig. 6B). In agreement, GST–Vps23^UEV pulled down a PalF–Ub protein fusion apparently as efficiently as PalF (Fig. 6C, note that PalF–Ub pull-down lanes contain twice the loading of PalF ones). Therefore these assays establish that, as in S. cerevisiae, A. nidulans Vps23 links the ESCRT machinery to the receptor–arrestin complex. However, in view of the lack of preference shown by Vps23^UEV for PalF–Ub, they do not satisfactorily explain the crucial role that PalF–Ub must play as alkaline ambient pH transducer [note that ubiquitin attachment to PalF leads to constitutivity (Hervás-Aguilar et al., 2010a)] (see Discussion).

The above experiments cannot discriminate differences in relative affinities, and thus their results might simply reflect that the ‘basal’ affinity of GST–Vps23 for PalF in vitro, almost certainly mediated by the SxP motifs of the latter (see Discussion), is sufficiently high to efficiently pull-down the arrestin, whether ubiquitylated or not. Thus we analysed, by anti-HA western blotting, anti-GFP immunoprecipitates of a PalF-(HA)3 strain also expressing, at physiological levels, Vps23–GFP (see below), using extracts from cells cultured in acidic or alkaline conditions (Fig. 7). In this strain and in the PalF-(HA)3 control (i.e. without Vps23 tagging), PalF was ubiquitylated in an alkaline-pH-dependent manner, resulting in a smear of PalF bands showing lower mobility than non-ubiquitylated PalF (Fig. 7, lanes 1, 2, 4, 5, upper panels) (Herranz et al., 2005; Hervás-Aguilar et al., 2010a). Despite the large predominance of non-ubiquitylated PalF, only ubiquitylated PalF species present in cells shifted to alkaline conditions co-immunoprecipitated with Vps23–GFP (Fig. 7, lane 7, arrow). No PalF bands were co-immunoprecipitated from extracts of acidic-grown cultures (Fig. 7, lane 6), in which PalF is not ubiquitylated, nor from extracts from acidic- or alkaline-grown cells that did not express the Vps23–GFP bait (Fig. 7, lanes 9 and 10). Thus these experiments establish that ubiquitylated PalF and Vps23–GFP occur together in alkaline-pH-induced protein complexes in vivo.

Next we studied whether the presence of ubiquitylated PalF in Vps23-containing complexes correlates with recruitment of ESCRTs to the plasma membrane. Recruitment of Vps32 to cortical structures cannot be studied in vivo because tagging with fluorescent proteins prevents its function (Hervás-Aguilar et al., 2010b; Nickerson et al., 2006). Thus, in view of the crucial role that Vps23 appears to play in pH signalling, we investigated this ESCRT-I protein. We found that strains in which the resident vps23 gene had been replaced by the GFP-tagged version (Materials and Methods) grew normally, indicating that Vps23–GFP is functional [as Vps23 is virtually essential (Calcagno-Pizarelli et al., 2011)]. This allowed study of the localisation of Vps23 expressed at physiological levels, avoiding overexpression artefacts. In acidic conditions Vps23 localises to the cytosol and to punctate cytosolic structures that undeniably represent endosomes because many showed the characteristic movement of early endosomes (EEs) in kymographs (Fig. 8A), and because in the absence of Vps27, Vps23–GFP does not localise to them (Fig. 8B). Studies on the potential recruitment of Vps23 to cortical sites were hampered by its presence in these endosomes, which often move along microtubule tracks in the vicinity of the cortex (supplementary material Movie 5). Thus, we converted time-lapse sequences of Vps23–GFP into sum projections, which essentially flattened the signal of moving endosomes and enhanced that of static structures, allowing us to detect clearly Vps23–GFP cortical foci in cells that had been shifted to alkaline conditions (Fig. 8C). Vps23–GFP foci were rare in cells continuously cultured under acidic conditions (Fig. 8B). Indeed recruitment of Vps23–GFP to cortical foci is PalF dependent (Fig. 8C), demonstrating that it requires activation of the ambient pH-signalling pathway (Vps23–GFP levels were unaffected by vps27Δ, sltA59, sltA59 vps27Δ or palF15 mutations, data not shown). The finding that such alkaline-pH-dependent recruitment was seen in vps27Δ cells indicated that this cortical localization is unrelated to the role that Vps23 plays in the multivesicular body pathway. Because vps27Δ eliminates the endosomal Vps23–GFP signal, we used these mutant cells to demonstrate unambiguously that pH-signalling-related Vps23 cortical structures clearly overlap with the plasma membrane stained with the lipophilic dye FM4-64 (Fig. 8D).

Finally we constructed a vps27Δ strain expressing Vps23–GFP and PalA–mCherry and acquired time-lapse sequences in the red and green channels. The Vps23–GFP signal obtained under such conditions was very weak, precluding any dynamic analysis. Sum projections showed that Vps23–GFP-containing cortical structures were more abundant than PalA–mCherry-containing structures. However, 77% of 448 PalA-containing structures from 27 hyphae colocalised with Vps23 (supplementary material Fig. S4), clearly linking cortical Vps23–GFP-containing structures and pH signalling complexes.

A model for the initial steps of the ambient pH-signalling pathway is shown in Fig. 9. The 7-TMD protein PalH that localises to the plasma membrane with the aid of the helper PalI (not depicted) must be the ambient pH sensor. The cytosolic PalH tail binds strongly to arrestin-like PalF and this interaction discourages trafficking of PalH away from the plasma membrane (Hervás-Aguilar et al., 2010a). In alkaline conditions PalF is ubiquitylated in a PalH-dependent manner. Ubiquitylated PalF mediates recruitment of Vps23 (Fig. 9) and perhaps of other ESCRT components (see below) to plasma membrane signalling complexes, thus linking ambient pH reception with the plasma membrane assembly of ESCRTs. Once at the plasma membrane, Vps23 promotes the recruitment of ESCRT-III components, possibly with ESCRT-II involvement, as suggested by the findings that at least ESCRT-II vps36Δ prevents pH signalling in Aspergillus and that vps36Δ, vps22Δ and vps25Δ prevent signalling in yeast (Calcagno-Pizarelli et al., 2011; Xu et al., 2004).

Following ESCRT-III assembly on the plasma membrane, its major component Vps32 (Teis et al., 2008) recruits, or crucially cooperates in, recruitment of one of its Pal interactors, the Bro1 domain-like PalC (Galindo et al., 2007). PalA arrives to these complexes after PalC incorporation (Fig. 3) and PalA recruitment to cortical pH signalling sites is PalC dependent. Finally, calpain-like protease PalB, which is dispensable for PalA recruitment (Fig. 2E), is itself recruited to these complexes through interaction of its MIT domain with Vps24 and possibly through additional, as yet uncharacterised interactions with Vps32 or other ESCRT components (Ito et al., 2001; Rodríguez-Galán et al., 2009). What are the different roles that Vps32 binders PalA and PalC play in these complexes, how is ESCRT-III polymerisation controlled to prevent budding events (Hurley and Hanson, 2010), and what is the stoichiometry of ESCRT and Pal proteins in these complexes, are questions for future investigation. Finally a major unanswered question is how this ordered chain of events leads to PalB protease activation. It is worth noting that Maki and coworkers have recently reported that the MIT domain-containing calpain 7 protease, the human PalB orthologue, is activated by recruitment to ESCRT components (Osako et al., 2010).

In this speculative model the molecular details of the intracellular switch(-es) triggering this cascade are insufficiently understood. It is clear that Vps23 plays an important role in association with PalF, as shown by the strong binding detected between A. nidulans Vps23 and PalF (Fig. 6) and between S. cerevisiae Vps23p and Rim8p (Herrador et al., 2009), and also by the inability of PalF null mutants to recruit Vps23–GFP to cortical sites (Fig. 8C). Also well established is the key role that PalF ubiquitylation must play, as expression of PalF–Ub at physiological levels results in constitutive PacC processing and recruitment of PalC to cortical sites even under acidic pH conditions (Hervás-Aguilar et al., 2010a). Moreover, ubiquitylated, rather than non-ubiquitylated S. cerevisiae Rim8p is preferentially recovered from Vps23 immunoprecipitates (Herrador et al., 2009), a finding also true for A. nidulans PalF (Fig. 7). However, in GST pull-downs, the A. nidulans Vps23 bait did not show any marked preference for PalF–Ub over PalF (Fig. 6). These results are consistent with the possibility that another Ub-binding module different from the Vps23 ubiquitin E2 variant (UEV), such as the carboxyl Npl4 zinc-finger (NZF) of Vps36 (Teo et al., 2006), contributes to a network of reinforcing interactions stabilising ESCRT proteins in cortical pH signalling complexes. Also not sufficiently understood is the fact that in contrast to A. nidulans PalF, whose ubiquitylation is strictly dependent on alkaline ambient pH and the 7-TMD receptor, Rim8p ubiquitylation is independent of both (Herrador et al., 2009), suggesting that Rim8p might be ubiquitylated through alternative inputs or receptors.

In S. cerevisiae, Vps23p binds to a SxP motif (E⁵³³SDP⁵³⁶) in Rim8p (Herrador et al., 2009) that resembles two Vps27p PSDP sequences containing a negatively charged residue that are recognised by a type-II-designated binding site in Vps23p (Ren and Hurley, 2011). This type II site is different from the type I binding site by which human Tsg101 (Vps23) binds P[S/T]AP motifs (Im et al., 2010). A. nidulans Vps23 lacks the basic binding pocket that accommodates the Asp residue in the Vps23p type II binding site (data not shown) (Ren and Hurley, 2011). In agreement, the equivalent of the E⁵³³SDP⁵³⁶ Rim8p SXP motif (Herrador et al., 2009) is PalF A⁷²⁷SAP⁷³⁰ (lacking the negatively charged residue). PalF has additional SXP motifs such as P⁶³⁴SQP⁶³⁷, S⁷⁰⁰SAP⁷⁰³, P⁷⁰⁷SRP⁷¹⁰ and the ‘type I-like’ P⁷⁴⁵SAP⁷⁴⁸ that might be potentially recognised by A. nidulans Vps23. However, the physiological role(s) of these motifs (including A⁷²⁷SAP⁷³⁰) remains to be analysed.

An intriguing aspect of this model is the different roles that PalC and PalA, both direct Vps32 binders (Galindo et al., 2007; Rodríguez-Galán et al., 2009), play in plasma-membrane-associated signalling complexes. For both, recruitment to cortical sites is Vps32 dependent, but PalC acts upstream of PalA and precedes PalA in arrival at the complexes. These findings are consistent with the predicted ‘late’ role of PalA in pH signalling, as PalA is also able to bind PacC through two conserved YPx[L/I] motifs located either site of the signalling protease cleavage site in the transcription factor (Vincent et al., 2003). Thus, the current view is that PalA links PacC to Vps32-containing complexes, ‘presenting’ the PacC substrate to the signalling protease PalB (Rim20p), itself an ESCRT-III interactor (Peñas et al., 2007; Rodríguez-Galán et al., 2009; Xu and Mitchell, 2001). Our data strongly indicate that signalling necessitates the Vps32-dependent PalA recruitment to the plasma membrane [in agreement with the finding that endosomal association of the PalA orthologue Rim20p is not sufficient to promote pH signalling (Boysen et al., 2010)]. They are consistent with the model in Fig. 9 in which most if not all these ‘late’ reactions would take place within ESCRT-containing pH signalling complexes associated with the plasma membrane, rather than endosomes as previously suggested (Mitchell, 2008; Peñalva et al., 2008).

PalA (Rim20p) is closely related to Bro1p, the founding member of the Bro1 domain-containing proteins (Kim et al., 2005), but does not play any role in the multivesicular body (MVB) pathway. Conversely, Bro1p does not play any role in pH signalling (Xu et al., 2004). PalC is less closely related to Bro1p although it also shows considerable similarity to the Bro1 domain across its Bro1-like domain, including a patch of residues conserved with Bro1p in which substitutions impair Vps32 binding and pH signalling (Galindo et al., 2007). A notable aspect of the model in Fig. 9 is that ESCRTs can be recruited to the plasma membrane without being involved in a membrane scission event. Experiments with liposomes strongly implicated ESCRT-III as the executioner of membrane scission during the formation of ILVs (Wollert and Hurley, 2010). A recent report showed that the Bro1 domain has the ability to regulate the membrane scission activity of ESCRT-III by regulating its stability (Wemmer et al., 2011). Thus it is tempting to speculate that the roles of PalA and PalC are related to the need to regulate the organisation and/or stability of ESCRT polymers in pH signalling complexes in a manner that is not conducive to membrane scission.

A. nidulans minimal (MM) and complete medium (MCA) were used. MM usually contained 1% glucose and 5 mM ammonium tartrate as the carbon and nitrogen source, respectively. C-terminally tagged PalC-GFP, PalA-GFP, PalA-mCherry and Vps23-GFP were encoded by gene-replaced alleles, following established procedures (Szewczyk et al., 2006; Yang et al., 2004). vps4-1 and vps4-ha3 are gene-replacement alleles driving expression of Vps4 and (HA)3-Vps4 under the control of the niiA^p promoter, which is inducible by nitrate and repressible by ammonium (Hervás-Aguilar and Peñalva, 2010). Thus, where indicated 10 mM nitrate (inducing conditions for niiA^p) or 5 mM urea (non-repressing conditions for niiA^p) was used instead of ammonium as nitrogen source. Watch minimal medium (WMM), used for all microscopy experiments, is a modified version of MM (Hervás-Aguilar et al., 2010a). Genotypes of the strains used in this work are given in supplementary material Table S1. Primers are listed in supplementary material Table S2.

Microscopy, image analysis and image manipulation was carried out essentially as previously detailed (Hervás-Aguilar et al., 2010a; Pantazopoulou and Peñalva, 2009; Pantazopoulou and Peñalva, 2011). We used a Nikon Eclipse epifluorescence microscope equipped with a 100× 1.40 NA plan-apochromat objective and an inverted Leica DMI6000B microscope with 63× 1.4 NA or 100× 1.4 NA plan-apochromat objectives and an incubation chamber coupled to the microscope stage. Images were acquired with a Hamamatsu ORCA ER digital camera driven by Metamorph software (Molecular Dynamics), using Semrock Brightline GFP-3035B and TXRED-4040B (mCherry) filter sets. Recruitment of fluorescent versions of PalA, PalC and Vps23 to cortical structures was monitored as described previously (Galindo et al., 2007; Hervás-Aguilar et al., 2010a). Cells were cultured overnight in WMM containing 0.1% glucose (w/v) as the carbon source and adjusted to acidic pH with 25 mM NaH₂PO₄ (pH 5.2-5.3) before being washed with and shifted to the same medium adjusted to acidic (5.2), neutral (7-7.2; with 12.5 mM NaH₂PO₄ plus 12.5 mM Na₂HPO₄) or alkaline (8.2-8.3; with 25 mM Na₂HPO₄) pH. Image acquisition was started immediately after the shift. For quantification, PalC-GFP-, PalA-GFP- or Vps23-GFP-containing punctate structures were counted in at least 25 germlings per strain and pH condition, and the resulting data were normalized to the germling length.

To study the order of PalA and PalC arrival at cortical structures, cells were cultured for 16 hours at 25°C in eight-well uncoated chambers (Ibidi, Germany) containing pH ~5 WMM and shifted to pH 7 WMM before immediately starting acquisition of time-lapse sequences of the GFP and mCherry channels using a Leica DMI6000 microscope. Time resolution, determined by the exposure times (1 seconds for GFP, 1.2 seconds for mCherry) and automated filter exchange was 4.4 seconds. Sequences for single PalA-mCherry and PalC-GFP transient fluorescence bursts (n=50) were then ‘isolated’ as 1.65 μm square regions of interest (ROIs) containing the cortical fluorescence. The red and green channels in each movie were analysed to determine the background, which was used individually to threshold the series and reduce the area corresponding to each fluorescent structure to a 0.9 μm diameter circle. Average intensity values within these circles were calculated and subsequently considered relative to the maximal average value in any given sequence, which was set as 100%. Paired PalA-mCherry and PalC-GFP data were plotted versus time, setting the ‘zero’ time point to the frame immediately preceding the start of the PalA-mCherry burst. Means and standard errors were then calculated for 50 structures.

Contrast adjustment, movie construction, thresholding, and colour alignment were made using Metamorph (Molecular Dynamics). Images were converted to 8-bits before being transferred to CorelDraw for annotation.

GST-Vps23, GST-Vps23^UEV(1-159) and GST-PalH(507-714) were expressed in Escherichia coli JM109-pRIL and purified as described previously (Abenza et al., 2010). Strains expressing PalF-HA₃ or PalF-HA₃::Ub^K48R (Hervás-Aguilar et al., 2010a) were cultured overnight in acidic Aspergillus fermentation medium (MFA) medium (final pH 5.1-5.4) or, for alkaline pH shifts, cultured for 16 hours in pH ~4.3 MFA medium before transferring mycelia to pH ~8.3 MFA medium for 10 minutes (Hervás-Aguilar et al., 2007). Protein extracts were prepared from lyophilized mycelia as described previously (Abenza et al., 2010). Powdered mycelia were resuspended and lysed in binding buffer (BB): 20 mM Tris-HCl pH 8, 110 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol (DTT), 10% glycerol, 0.1% Triton X-100 2.5 μM Pefabloc, 2 μM pepstatin, 1.2 μM leupeptin, 5 μM MG132 and 5 mM N-ethylmaleimide. ~5 μl glutathione-Sepharose beads containing ~50 μg of bait proteins (GST-Vps23^UEV, GST-PalH(507-714) or GST-Vps23 were mixed with 1 mg of A. nidulans protein extract in 0.8 ml of BB, using Handee-Spin Columns (Pierce). The mix was rotated for 2 hours at 4°C before collecting the beads, which were washed three times with 20 mM Tris-HCl pH 8, 175 mM KCl, 5 mM MgCl₂, 1 mM DTT and 0.1% Triton X-100, with a 15 minute rotation at 4°C between the second and third washes. Bound proteins were eluted with SDS-PAGE loading buffer. Aliquots (1/5) were run in 7% (29:1 acrylamide:bis-acrylamide) SDS-PAGE gels that were blotted using anti-HA antibody as reported (Hervás-Aguilar et al., 2010a). Separate aliquots were run in parallel for Coomassie staining.

Cell extracts [2 mg protein in IP buffer (25 mM Tris-HCl, pH 7.2, 150 mM KCl, 5 mM MgCl₂, 1 mM DTT, 0.1% Triton X-100, 10% glycerol, 10 mM Pefabloc, 5 μM pepstatin, 5 μM leupeptin, 5 μM MG-132, 15 mM NEM and 1× EDTA-free Roche's inhibitor cocktail)] were incubated with 1.2 μg of mouse anti-GFP cocktail of monoclonal antibodies (Roche) for 3 hours at 4°C, before being mixed with Pierce Protein G plus agarose (12 μl per sample) and incubated for an additional 2 hours. Beads containing the immunocomplexes were recovered by centrifugation at 1000 g and washed with IP buffer before eluting bound materials with Laemmli loading buffer prior to their analysis by western blotting. Anti-GFP western blots were as described previously (Galindo et al., 2007).

Plasmid p1672 (pQEzzHis::Vps4) drives expression in E. coli of a fusion protein containing two tandem copies of the protein A ‘Z’ domain fused to the N-terminus of A. nidulans Vps4. A 30,000 g supernatant obtained from a lysate from overexpressing bacteria was mixed with 0.4 ml IgG-Sepharose and purified following recommendations of the supplier (GE Healthcare). The fusion protein (~2.5 mg) was used to immunise rabbits (Davids Biotechnology, Regensburg, Germany). This antiserum was used at 1/5000 in western blots, after blocking non-specific binding sites with 3% (w/v) skimmed milk and 0.03% (w/v) of A. nidulans mycelial powder. Peroxidase-coupled anti-rabbit IgG (GE Healthcare) was used as the secondary antibody (1:3000). The specificity of the antibody was demonstrated by the fact that the band corresponding to Vps4 was shifted to a position of lesser mobility after (HA)3 tagging.

Cells were cultured in 50 ml MM containing 10 mM (NH₄)₂SO₄, 5 mM urea or 10 mM NaNO₃, as required, adjusted to pH 4.4 with 50 mM sodium citrate buffer, pH 3.5. After a 15-hour incubation at 30°C, mycelia were collected by filtration (cellulose membranes, 0.45 μm, 47 mm diameter) and transferred to medium buffered to pH 8.5 with HEPES-NaOH, containing the same nitrogen sources. Mycelial samples were taken before and at different times after the pH shift, pressed dry, quick-frozen, lyophilised and processed for PacC western blot analysis (Hervás-Aguilar and Peñalva, 2010).

The genes and proteins used in this work can be found in the Aspergillus genome database (AspGD, http://www.aspergillusgenome.org/) under the following database entry numbers: PalA, AN4351; PalB, AN0256; PalC, AN7560; PalH, AN6886; PalF, AN1844; Vps32, AN4240; Vps23, AN2521; Vps27, AN2701; Vps4, AN3061.

We thank Elena Reoyo for technical assistance.