Adenylyl cyclase AC8 directly controls its micro-environment by recruiting the actin cytoskeleton in a cholesterol-rich milieu

AC8 enhances the cortical actin signal

Initial experiments examined the distribution of the actin cytoskeleton using phalloidin in untransfected HEK293 cells and cells expressing AC8–HA, AC1–FLAG or AC2–HA. Remarkably, there was a marked difference in the localisation of the actin cytoskeleton when AC8–HA was expressed (Fig. 1A). Filaments of the actin cytoskeleton extended throughout the cytosol in untransfected cells and in cells expressing AC1 or AC2; however, cortical actin at the PM was significantly enhanced when AC8 was expressed (Fig. 1A). Because phalloidin only detects F-actin, the signal from GFP–actin was also examined in cells expressing AC1, AC2 or AC8 (supplementary material Fig. S1A). Highly dispersed G-actin and F-actin were identifiable throughout the cytosol in all cells, however, the GFP–actin signal was enhanced at the PM in cells expressing AC8. Thus the observed redistribution of the cytoskeleton was not the result of the phalloidin staining (supplementary material Fig. S1A). It is not surprising that there are major differences between AC1, AC2 and AC8 given that their N-terminal sequences are quite dissimilar (supplementary material Fig. S1B).

To ensure that this observation was not a function of the HA epitope of AC8, GFP–AC8 was also examined (Fig. 1B). Pearson's coefficient (Rr) was calculated (supplementary material Fig. S2) in order to ascertain the degree of colocalisation between the GFP signal and cellular structural elements. The PM residence of AC8 meant that GFP–AC8 colocalised well with the PM marker wheat germ agglutinin (WGA; Rr=0.77±0.02; Fig. 1C). Interestingly, GFP–AC8 also consistently colocalised well with the actin cytoskeleton marker phalloidin (Rr=0.72±0.02), but much less so with the microtubule marker tubulin (Rr=0.25±0.03; Fig. 1C). Lyn–GFP, a PM-targeted control peptide (Park et al., 2008) (Fig. 1D) also colocalised with WGA (Rr=0.78±0.02), but less so with phalloidin (Rr=0.50±0.03; Fig. 1E). Thus the Rr data underline the specificity of AC8 in enhancing the cortical actin signal. In order to analyse the relative distribution of the phalloidin signal in cells expressing GFP–AC8 or Lyn–GFP the ratio of the signal intensity was determined at the PM and in the cytosol (Ratio_PM/Cyt; supplementary material Fig. S2). Ratio_PM/Cyt confirmed that the phalloidin signal was greater at the PM in cells expressing GFP–AC8 (21.2±3.4) than in those expressing Lyn–GFP (6.9±1.1; Fig. 1F). Thus, an association between AC8 and the actin cytoskeleton is strongly suggested.

The enhancement of cortical actin depends on the N-terminus of AC8

To explore the association between AC8 and actin, two mutants were studied: an N-terminally truncated construct, AC8M1, and a catalytically inactive, full-length mutant, AC8 D416N (Fig. 2A). AC8M1 lacks the first 106 amino acids of the N-terminus, a hyper-variable region of ACs that contains several protein binding sites, and thus might be a candidate domain to interact with actin. AC8 D416N was used to elucidate a role for cAMP in actin remodelling; a point mutation at the invariant Mg²⁺-binding aspartic acid residue (D416) within the C1a catalytic domain renders the enzyme inactive (Howe, 2004; Nadella et al., 2009). All constructs were tagged with GFP at their N-termini (Fig. 2A) and stably expressed (Fig. 2B).

Activity assays were performed in order to define the activity profiles of each construct in vitro (Fig. 2C) and in vivo (Fig. 2D). These experiments confirmed the inactivity of AC8 D416N and demonstrated that GFP–AC8 and GFP–AC8M1 have similar Ca²⁺-activation profiles in vitro (Fig. 2C), whereas in the intact cell, GFP–AC8 was stimulated 4.0±0.3 fold by 4 mM external Ca²⁺ following triggering of CCE, whereas GFP–AC8M1 required 8 mM external Ca²⁺ for comparable (3.6±0.3 fold) stimulation (Fig. 2D). The reduced responsiveness of AC8M1 to CCE has been attributed to the deletion of an N-terminal calmodulin (CaM) binding domain (residues 34–51), which precludes pre-recruitment of CaM (although additional factors are not excluded) (Gu and Cooper, 1999; Simpson et al., 2006) (Fig. 2D).

The expression and relative distribution of GFP–AC8M1 and GFP–AC8 D416N was determined by confocal image analysis. Although GFP–AC8M1 and GFP–AC8 D416N were both targeted to the PM, the enhancement of cortical actin was only observed in cells expressing GFP–AC8 D416N (Fig. 2E). Indeed the degree of colocalisation between GFP–AC8M1 and phallodin was significantly reduced compared with when GFP–AC8 or GFP–AC8 D416N was expressed (Fig. 2F) and, Ratio_PM/Cyt analysis confirmed that the distribution of the actin cytoskeleton was not enhanced at the PM in cells expressing GFP–AC8M1 (Fig. 2G). Thus the redistribution of the cytoskeleton and enhancement of cortical actin by AC8 depends on the N-terminus of AC8; however, a role for cAMP emanating from AC8 is not obvious from these measurements.

GST pull-down and co-immunoprecipitation (co-IP) experiments confirm an association between AC8 and actin

In order to investigate biochemically the interaction between AC8 and the actin cytoskeleton, GST pull-down experiments were performed. In the absence of Ca²⁺, only full-length 8NT (residues 1–179) significantly pulled down actin, compared with the level using unconjugated GST beads (Fig. 3A–C; n=5). Interestingly, the separate halves of the N-terminus of AC8 did not pull down actin, suggesting that the N-terminus must be intact and adopt an integrated secondary structure to associate with actin. The interaction between 8NT and actin was lost in the presence 10 μM Ca²⁺ (Fig. 3A–C; n=5). The presence of Ca²⁺ allows endogenous CaM to bind to both known CaM binding domains in the 8NT and 8C2b fragments. To further explore the interaction between actin and the N-terminus of AC8, HA co-IP experiments were performed with cell lysates prepared from cells expressing full-length AC8–HA or AC8M1–HA. AC8–HA, but not AC8M1, pulled down actin (Fig. 3D,E; n=3). Therefore, both biochemical and single cell confocal experiments indicated that the N-terminus of AC8 associates with the actin cytoskeleton. Furthermore, the AC8–actin interaction is affected by Ca²⁺–CaM and requires the complete N-terminus.

The distribution and regulation of GFP–AC8 depends on the intact actin cytoskeleton

To explore the functional significance of the AC8–actin interaction in a cellular context, the cytoskeleton was disrupted with latrunculin B (LatB; 2 μM). This pharmacological tool binds to G-actin to produce shorter, thicker stress fibres and thereby promote the complete passive disruption of the actin filaments. Treatment with LatB led to the aggregation and partial internalization (37%) of GFP–AC8 (Fig. 4A–C; supplementary material Fig. S3A and Fig. S4) without inducing necrotic or apoptotic characteristics, such as blisters and blebs (supplementary material Fig. S3A), and indeed, several of the AC8 aggregates colocalised with the phalloidin signal of the disrupted cytoskeleton (Fig. 4B). These data indicate that the AC8–actin interaction persists despite cytoskeletal disruption.

The regulation of voltage-gated ion channels (Schubert and Akopian, 2004) and Lck (Chichili and Rodgers, 2007) are both compromised when the actin cytoskeleton is disrupted. In order to investigate whether the intact actin cytoskeleton is essential for regulation of GFP–AC8, activity assays were performed using the Förster resonance energy transfer (FRET)-based cAMP sensor, Epac2-camps, in single cells expressing AC8–HA. Untreated cells displayed a robust response to CCE (Fig. 4D). However, disrupting the integrity of the actin cytoskeleton with LatB resulted in the deregulation of GFP–AC8; these cells produced significantly more cAMP under basal conditions (Fig. 4E), but were consequently less stimulated by CCE (Fig. 4F). These Epac2-camps results were mirrored in whole-cell cAMP accumulation experiments (supplementary material Fig. S3B–E). To discount the possibility that the deregulation of AC8 by LatB is due to a diminution or compromising of CCE, CCE was directly assessed by measuring the internal Ca²⁺ concentration ([Ca²⁺]_i) using Fura-2 (supplementary material Fig. S3F). LatB did not significantly alter [Ca²⁺]_i before or following the induction of CCE (supplementary material Fig. S3G,H). Thus the cytoskeleton is not only essential for the organization of AC8, but it also contributes to the proper regulation of AC8 by CCE. Thus, disrupting the cytoskeleton appears to remove an inhibitory influence that results in a more active enzyme under basal conditions, but, perhaps because of the disruption of an association of the enzyme from the CCE apparatus, the enzyme is less responsive to CCE.

Membrane cholesterol is also essential in the distribution and regulation of AC8

To address the role of membrane rafts in the function and distribution of AC8, rafts were disrupted using methyl-β-cyclodextrin (MβCD) or sphingomyelinase (SMase). MβCD is a chelator of cholesterol, which alters cell morphology by extruding vesicles of cholesterol from the cell, which consequently reduces the cell surface area and rounds up the cells (Ilangumaran and Hoessli, 1998). The enzyme SMase, hydrolyses sphingomyelin into ceramide and phosphorylcholine to stimulate the internalization of cholesterol from the PM to the ER (Contreras et al., 2003; Staneva et al., 2009). Therefore, whereas MβCD entirely extrudes cholesterol from the cell, SMase internalizes the cholesterol.

Treatment of cells with MβCD resulted in the extrusion of GFP–AC8 with cholesterol in vesicles, and the small proportion of AC8 retained within the cell was expressed at the PM to increase Ratio_PM/Cyt. By contrast, SMase dramatically internalized GFP–AC8 (80%; Fig. 4A,C; supplementary material Fig. S4). It is worth noting that neither MβCD nor SMase disrupted the integrity of the actin cytoskeleton because actin filaments could still be identified (Fig. 4A; supplementary material Fig. S3A and Fig. S4). MβCD did not alter the distribution of the actin cytoskeleton because the phalloidin signal remained enhanced at the PM, although some of the actin cytoskeleton was also extruded in the cholesterol–AC8 vesicles. Conversely, SMase, significantly internalized (41%) the actin cytoskeleton, along with AC8 and cholesterol (Fig. 4C). These data underline the strength of the interaction between AC8 and the cytoskeleton, even following raft disruption with MβCD or SMase.

Disrupting membrane rafts dramatically altered the distribution of AC8. In order to ascertain whether this redistribution affects the regulation of AC8, single-cell cAMP determinations were performed as above. As previously reported (Pagano et al., 2009), MβCD and SMase (like LatB) consistently and dramatically increased the basal accumulation of cAMP by AC8–HA (Fig. 4E), and the response to CCE was significantly reduced (Fig. 4F). To eliminate the possibility that MβCD or SMase altered AC8 activity because of a compromised Ca²⁺ response, [Ca²⁺]_i transitions were measured using Fura-2 (supplementary material Fig. S3). Following treatment with MβCD or SMase, the thapsigargin (TG)-triggered [Ca²⁺]_i rise differed somewhat although the subsequent CCE response (which regulates AC8) was unaltered.

The N-terminus of AC8 interacts with actin and cholesterol

The experiments above demonstrate that the N-terminus of AC8 interacts with actin, and this interaction controls the distribution and regulation of AC8 (Fig. 2G, Fig. 4). In support of this conclusion, the distribution of the actin cytoskeleton is not altered when GFP–AC8M1 (which lacks the N-terminus) is expressed (Fig. 5A). Consequently, it was not surprising that the distribution of GFP–AC8M1 was not altered when the actin cytoskeleton was disrupted with LatB (Fig. 5A,B; supplementary material Fig. S5) or that LatB did not alter the regulation of GFP–AC8M1 under basal conditions or following CCE (Fig. 5C,D). These data reinforce the association between the N-terminus of AC8 and actin.

Membrane cholesterol is essential for the distribution and regulation of GFP–AC8M1

In order to determine whether the association of AC8 with the actin cytoskeleton is essential for the residence of AC8 in membrane rafts, cholesterol was depleted from cells expressing GFP–AC8M1 (which does not associate with actin) with MβCD or SMase. Similar to GFP–AC8, GFP–AC8M1 was extruded in vesicles, along with cholesterol, when cells were treated with MβCD and was internalized by SMase (Fig. 5A,B; supplementary material Fig. S5). The relative distribution of the actin cytoskeleton was not altered by MβCD or SMase treatment in these cells, reinforcing the role of the complete N-terminus in the interaction with cortical actin (Fig. 5B).

Given the dramatic effect of cholesterol depletion on the regulation of AC8, similar experiments using Epac2-camps were performed using AC8M1. Both MβCD and SMase significantly increased the basal activity of GFP–AC8M1, which precluded further stimulation by CCE (Fig. 5C,D). Thus, intact membrane rafts are essential for the proper regulation of both AC8 and the truncated AC8M1.

Given the similarity in the effect of MβCD and SMase on the distribution and regulation of AC8 and AC8M1, it might be concluded that the hydrophobic associations that drive AC8 into membrane rafts are independent of the first 106 amino acid residues. In order to detect a direct association between AC8 and the fluorescent cholesterol surrogate, Ergosta-5,7,9(11),22-tetraen-3β-ol (ETO), acceptor bleaching FRET experiments were performed on cells expressing GFP–AC8 or GFP–AC8M1. Strikingly, the bleaching of the acceptor (GFP) led to an increase in the fluorescence emitted from the donor ETO in cells expressing GFP–AC8 (Fig. 6A,C) indicating intimate communication between the pair. However, somewhat unexpectedly, GFP–AC8M1 showed no FRET with ETO (Fig. 6B,D). Thus GFP–AC8 associates with ETO, but GFP–AC8M1 does not (Fig. 6E). However, because the previous confocal analysis (Fig. 4A, Fig. 5A) showed that membrane cholesterol contributes substantially to the distribution and regulation of both AC8 and AC8M1, it is probable that additional mechanisms consolidate the associations of AC8 in membrane rafts rather than simply an affinity for sterols.

cAMP production is essential in the AC8 association with actin, but not cholesterol

Because the full-length, but inactive, mutant GFP–AC8 D416N also enhanced cortical actin (Fig. 2G), we explored possible contributions of local cAMP production on the maintenance of the AC8–actin interaction. Following LatB treatment, GFP–AC8 D416N was significantly internalized (88%; Fig. 5E,D; supplementary material Fig. S6). Indeed, GFP–AC8 D416N was significantly more sensitive to cytoskeletal disruption than was GFP–AC8 (37% internalization), which suggests that cAMP plays, at least, a partial role in maintaining AC8 at the PM, perhaps through the known actions of PKA and/or EPAC on the polymerization of the actin cytoskeleton (Nadella et al., 2009). Thus the robust interaction between actin and AC8 that occurs through the N-terminus is reinforced by the local production of cAMP.

In order to explore the contribution of cAMP production to the residence of AC8 in membrane rafts, cholesterol was depleted from the PM using MβCD or SMase. Similar to GFP–AC8 and GFP–AC8M1, MβCD-mediated chelation of cholesterol led to the extrusion of GFP–AC8 D416N in cholesterol-containing vesicles, and SMase internalized GFP–AC8 D416N (Fig. 5E,D; supplementary material Fig. S6). Thus no obvious role was played by cAMP in the association of AC8 with lipid rafts.

The mobility of AC8 is influenced by its N-terminus

Clearly, the distribution and function of AC8 is tightly regulated by an association with actin as well as cholesterol. We felt that FRAP and fluorescence correlation spectroscopy (FCS) analysis would discern the quantitative underpinnings of this cellular association. Spot-bleach FRAP was used to quantify the diffusion coefficient (D_FRAP) and mobile fraction (MF) of GFP–AC8, GFP–AC8M1 or GFP–AC8 D416N by measuring GFP recovery into bleached regions of interest (ROI, 1.54 μm; Fig. 7A,B) and MF from data collected up to 200 seconds post-bleach (Fig. 7C). D_FRAP was calculated by recording the mean fluorescence intensity along the bleached membrane (Fig. 7D) to provide Gaussian recovery curves (Fig. 7E). When the radius of the Gaussian curves are squared and plotted against time, the slope of the subsequent straight-line graph is proportional to D_FRAP (5–20 seconds post-bleach, where R²>0.9; Fig. 7F). To validate the data collection and analysis methods, Lyn–GFP was used as a control; the D_FRAP value measured was 0.81±0.08 μm/second², which agrees with the previously published value of 0.79±0.06 μm/second², from HEK293 cells (Hammond et al., 2009). Given the proposed interaction between the N-terminus of AC8 and the actin cytoskeleton, we anticipated that AC8M1 might move more freely in the membrane. Indeed, FRAP data confirmed that the mobility of GFP–AC8M1 differed significantly from its full-length counterparts, however, a larger proportion of GFP–AC8M1 was in fact immobile, but the mobile population diffused more quickly than GFP–AC8. Interestingly, the mobility profiles of GFP–AC8 and GFP–AC8 D416N were indistinguishable (Fig. 7C,G) suggesting that the local production of cAMP does not have any substantial influence on the mobility of AC8 in the intact cell. These data reiterate the importance of the N-terminus in constraining AC8.

FCS is an additional useful method to assess molecular diffusion in a very small confocal volume (~0.1 μm²) by recording fluorescence fluctuations. Autocorrelation analyses of time-dependent fluctuations indicates the dwell time within the measurement volume, and subsequently, the diffusion coefficient, D_FCS and the number of fluorescent particles (N; and ultimately concentration, N/μm²). Alternatively, the photon counting histogram (PCH) method analyses the amplitude of the same fluctuations and, in addition to particle number, gives a value for molecular brightness (ε), indicating the number of photons per second emitted by each molecule and therefore the extent of molecular aggregation or oligomerization. Thus, whereas FRAP accurately measures the mobility of a population of molecules within 1.54 μm² of a juxtanuclear region of the PM, FCS measures the diffusion of tens of molecules in real time, within a smaller area of the apical membrane (~0.1 μm²) (Edidin, 1992; Kusumi et al., 2005). Furthermore, FCS distinguishes fast moving molecules within a microdomain, whereas FRAP detects long-range diffusion between several microdomains in the PM (Hancock and Parton, 2005; Haustein and Schwille, 2004; Lingwood et al., 2008; Pucadyil et al., 2007). However, unlike FRAP, FCS cannot reliably resolve very slow moving or immobile proteins and hence cannot determine the mobile fraction.

FCS produced comparable D values to those obtained with FRAP analysis and in addition determined that GFP–AC8M1 diffused more rapidly than GFP–AC8 and GFP–AC8 D416N (Fig. 7G,H). The modest variation in D reflects differences in the measurement area and the temperature at which the experiments were performed. All three constructs yielded a similar particle number (Fig. 7I), but PCH analysis showed that the molecular brightness of GFP–AC8M1 was significantly higher than that of either GFP–AC8 or GFP–AC8 D416N (Fig. 7J), suggesting that GFP–AC8M1 is either present in a higher multimeric form or confined within a microdomain containing more than one molecule of AC8M1.

AC8 mobility depends on membrane cholesterol

In accordance with several other studies on GFP-tagged raft proteins, such as the class II MHC (Nishimura et al., 2006), the removal of membrane cholesterol with MβCD reduced the MF and D_FRAP of GFP–AC8, GFP–AC8M1 and GFP–AC8 D416N (Fig. 8A–C) (Nishimura et al., 2006). However, a reduction in mobility of non-raft proteins following treatment with MβCD has also been observed (Kenworthy et al., 2004; Shvartsman et al., 2006; Zidovetzki and Levitan, 2007). These authors have drawn attention to additional effects mediated by MβCD that can include cytoskeletal disruption, depletion of cholesterol from non-raft regions of the PM and effects on other phospholipids. Thus caution is warranted in placing the simplest possible interpretation on the mode of action of MβCD. Treatment with SMase, like MβCD, also significantly decreased the MF and D_FRAP of GFP–AC8 (Fig. 8A–C), which reinforces the conclusion that these constructs reside in regions of the PM rich in cholesterol. However, SMase induces the formation of large aggregates of GFP–AC8 at the PM, rendering further analysis impractical (Fig. 8A).

Following treatment with MβCD, FCS, like FRAP, revealed a reduced mobility of GFP–AC8, GFP–AC8M1 and GFP–AC8 D416N (Fig. 8D). Furthermore, treatment with MβCD decreased the molecular brightness of all three constructs (Fig. 8E), whereas the particle number remained unchanged (Fig. 8F). These data support the observation from confocal imaging that all three constructs are extruded from the cell by MβCD, leaving pockets of AC8 molecules with restricted diffusion at the PM because of the absence of movement as promoted by cholesterol on lipid ordering (Nishimura et al., 2006).

AC8 mobility depends on the intact cytoskeleton

In general, if the mobility of a raft protein is curtailed by the actin cytoskeleton, then disrupting it would increase the MF and D_FRAP because of the protein is released from the constraints of the cytoskeletal corrals. This effect was seen for Na⁺/K⁺-ATPase (Paller, 1994). However, FRAP studies showed a decrease in the MF and D_FRAP of GFP–AC8 following disruption of the actin cytoskeleton with LatB (Fig. 8B,C), suggesting that AC8 is not simply released from corrals. Rather, the AC8-actin association persists, forming large aggregates of GFP–AC8 and actin as seen at the PM in confocal studies. These aggregates reduce the D_FRAP and thus prevent inhibitory and stimulatory constraints from interacting with AC8, precluding normal regulation. In contrast to FRAP analysis, treatment with LatB significantly increased the D_FCS of GFP–AC8 (Fig. 8D). However, the molecular brightness and particle number were not altered by LatB, indicating that the aggregation or oligomerization state of GFP–AC8 did not change (Fig. 8E,F). Therefore the GFP–AC8 aggregates detected at the PM in confocal studies are slow moving and so become bleached. Only non-aggregated material is measured by FCS, resulting in an overall increase in the diffusion of AC8.

Because AC8 enhances the cortical actin, but the N-terminally truncated mutant AC8M1 does not, the effect of LatB treatment on the mobility of AC8M1 was investigated. In contrast to GFP–AC8, LatB treatment increased the MF and D_FRAP of GFP–AC8M1 (Fig. 8B,C). Interestingly, consistent with the FRAP data, D_FCS of GFP–AC8M1 also increases following LatB treatment and the brightness and particle numbers are unaltered (Fig. 8D–F). Collectively, these data are consistent with the idea that GFP–AC8M1 is transiently trapped within actin cytoskeleton corrals and is simply released upon cytoskeletal disruption with LatB (Fig. 9). Thus, we propose that the N-terminus of AC8 associates with the actin cytoskeleton, enhancing cortical actin by consequently recruiting the filamentous meshwork to the PM, which restricts the diffusion of AC8 to tracking the path of the filaments that lie beneath the PM (Fig. 9B). However, without the N-terminus, in a similar manner to Kv2.1 channels (O'Connell et al., 2006; Tamkun et al., 2007), AC8M1 is transiently trapped within actin cytoskeleton corrals, constrained by the perimeters created by the actin meshwork (Fig. 9B). The large cytosolic domain of AC8M1 (C1 and C2 regions of approximately 72 kDa), could be sterically hindered by the cytoskeleton. Such restrictions are also seen in the transport protein, band 3, where cleavage of the cytosolic domain (approximately 40 kDa) allows the protein to move freely (Tomishige et al., 1998). However, at this stage, the model presented (Fig. 9) is speculative and not detailed; in particular, the corrals and membrane domains might, in part, be established through the spectrin membrane skeleton. The relationship between the membrane skeleton and cortical actin rim should be experimentally addressed in future studies as should the impact of disruptors such as LatB on the discrete pools of actin, which might have differing effects on the gelation state of the cytoskeleton at the PM.

The contribution of cAMP to the AC microdomain

The production of cAMP by ACs offers an increased opportunity for ACs to sculpt their microdomain. Careful observation revealed colocalisation ‘hot spots’ between AC8 and phalloidin (Fig. 1B, Fig. 2E, Fig. 4A), which was far more prevalent than in cells expressing AC8 D416N (Fig. 2E, Fig. 5E). Given that confocal image analysis indicated that local cAMP contributes to the maintenance of AC8 at the PM from the increased sensitivity of the catalytically inert AC8 mutant to cytoskeletal disruption (Fig. 4B, Fig. 5E), the mobility of GFP–AC8 D416N following treatment with LatB was investigated. Treatment with LatB dramatically increased the D_FRAP of GFP–AC8 D416N, whereas the MF was not altered. These data indicate that cAMP production selectively influences AC8 mobility (Fig. 8B,C). Interestingly, the D_FCS of GFP–AC8 D416N was not altered with LatB (Fig. 8D). However, there was a significant reduction in molecular brightness (Fig. 8E), with a corresponding increase in particle number (Fig. 8F). The FCS and FRAP data, when considered together, confirm the observation that aggregates of GFP–AC8 D416N were dramatically internalized (88.2%) following treatment with LatB, leaving only monomeric proteins at the membrane. These monomers are likely to be present in two populations: individual proteins that diffuse quickly, and slower moving proteins bound to small linear actin filaments or aggregates. Because FRAP records the diffusion of AC8 D416N over larger distances, it is unlikely to detect the cytoskeletal hindrance distinguishable by FCS (Fig. 8C). The discrepancy between the effects of LatB on GFP–AC8 and GFP–AC8 D416N therefore reflects the basal cAMP produced by GFP–AC8 within its microdomain. There is a large body of literature that describes the effect of cAMP, through the actions of PKA and EPAC, on the structure, remodelling and polymerization of the actin cytoskeleton (Howe, 2004; Nadella et al., 2009). Our data are in agreement with these findings and illustrate that even following cytoskeleton disruption, local cAMP produced by AC8 could induce the polymerization of actin, perhaps through the LIM kinase-1 pathway (Nadella et al., 2009), forming actin aggregates around AC8, thus trapping it and reducing its mobility.

PKA inhibition alters AC8 mobility following actin disruption

Because the cAMP target, PKA, has a major function in the structure and integrity of the actin cytoskeleton, we wondered whether the mobility of AC8 would resemble that of AC8 D416N following PKA inhibition. PKA inhibition by H89 or KT5720 did not significantly alter the MF and D_FRAP of GFP–AC8 in the intact cell (Fig. 10A). However, PKA inhibition prevented the decrease in the MF of GFP–AC8 following cytoskeleton disruption (Fig. 10A) and significantly increased D_FRAP, compared with control cells treated with LatB (Fig. 10B). Therefore, the mobility of GFP–AC8 in the presence of the PKA inhibitors resembles that of GFP–AC8 D416N when cells were treated with LatB. Consequently AC8 by producing cAMP, and thereby (through PKA) remodelling the actin cytoskeleton, influences its mobility by altering the cytoskeletal framework of its microdomain.

The expression of AC8 and cAMP analogues partially ‘rescue’ GFP–AC8 D416N mobility

To pursue further the role of cAMP in modulating the AC8 microdomain, we asked whether exogenous cAMP analogues or the co-expression of untagged (but live) AC8 would ‘rescue’ the mobility of inactive GFP–AC8 D416N. Because cAMP activates both PKA and EPAC, but there are no available inhibitors of Epac, two cAMP analogues were used: chlorophenylthio–cAMP (CPT–cAMP), which specifically activates EPAC, and dibutyryl–cAMP (DB–cAMP), which is a potent PKA activator (Enserink et al., 2002). CPT–cAMP significantly decreased the mobility of GFP–AC8 D416N when the actin cytoskeleton was disrupted with LatB (Fig. 10C,D) and DB–cAMP reversed the effect of LatB on the D_FRAP of GFP–AC8 D416N so that disrupting the cytoskeleton no longer significantly increased the diffusion (Fig. 10D). These data suggest that these cAMP analogues can indeed alter the GFP–AC8 D416N microdomain, partially converting the dynamic behaviour of GFP–AC8 D416N to be similar to that of GFP–AC8, when the actin cytoskeleton is disrupted. A similar effect was recorded in cells coexpressing AC8–HA and GFP–AC8 D416N; the local cAMP produced by AC8 partially ‘rescued’ the mobility of AC8 D416N when cells were treated with Lat B (Fig. 10E,F). These experiments support a role for local cAMP in shaping the architecture of the AC8 microdomain.

AC8 interacts with the plus end of the actin filament

In order to define the nature of the AC8–actin association, actin perturbants other than LatB were used. Cytochalasin D (CytD) directly binds to and terminates the plus end of the actin fibres to prevent polymerization of actin filaments, inducing them to contract, whereas jasplakinolide (Jasp) binds the length of the actin filament and induces polymerization of G-actin onto growing filaments leading to the overall disruption of F-actin. (Note, however, that Jasp competes with phalloidin for actin, precluding the monitoring of actin disruption using phalloidin.)

To compare the effects of CytD and Jasp, with that of LatB, on the distribution, regulation and mobility of GFP–AC8, GFP–AC8M1 and GFP–AC8 D416N, cells were incubated with 30 μM CytD or 1 μM Jasp and analyzed as before. Somewhat unexpectedly, neither CytD nor Jasp altered the distribution of GFP–AC8, GFP–AC8M1 or GFP–AC8 D416N (supplementary material Fig. S7A–C). Relative distribution analysis confirmed that all three constructs remained at the PM and were not internalised (supplementary material Fig. S7D). Because LatB compromised the regulation of GFP–AC8 by CCE, cAMP accumulation assays were performed on cell populations expressing GFP–AC8 or GFP–AC8M1 following treatment with CytD or Jasp. Interestingly, neither treatment altered the activity of GFP–AC8 or GFP–AC8M1 (supplementary material Fig. S7E). Furthermore, FRAP analysis revealed that CytD and Jasp did not alter the mobility of GFP–AC8, GFP–AC8M1 or GFP–AC8 D416N (supplementary material Fig. S7F,G).

Because neither CytD nor Jasp altered AC8 mobility, association with the cytoskeleton is unlikely to impede the diffusion of AC8 at the PM. It is not uncommon for different cytoskeletal disruptors to have differing consequences; for example, Lat B, but not CytD, increased the diffusion of MHC-II (Umemura et al., 2008). The intriguing differences in the effects of each actin cytoskeleton disruptor must be attributed to their varying mechanisms of action. CytD and Jasp associate with the plus end of the actin filament, so it is probable that AC8 interacts with this region of the filament and is perhaps displaced by CytD and Jasp. However, LatB binds to G-actin and passively degrades the actin filaments around AC8, permitting the accumulation of short filaments around AC8, as seen in confocal images (Fig. 4B), and prevents access to the catalytic site, impeding regulation by CCE (Fig. 4F).

Reagents and constructs

The rabbit polyclonal anti-AC8 antibody was a kind gift from Jim Cali (Promega, Madison, WI, USA). Epac2-camps was a gift from Martin Lohse (Würzberg University, Germany) (Nikolaev et al., 2004).

GFP–AC8 was generated by cloning AC8 cDNA between the ApaI and XbaI restriction sites of pEGFP-C1. GFP–AC8 D416N (Tesmer et al., 1999) was introduced by site-directed mutagenesis according to the QuickChange protocol (Stratagene, La Jolla, CA) using the fusion high-fidelity polymerase kit (Finnzymes) according to manufacturer's instructions, using the forward and reverse primer sequence, 5′-CGAGAACGTCAGTATTCTTTTTGCAAATGTCAAAGGATTTAC-3′ and 5′-GAGAGGTTGGTAAATCCTTTGACATTTGCAAAAAGAATAC-3′, respectively. Lyn–GFP, the modified N-terminus of Lyn kinase (MGCIKSKGKDS), was a gift from T. Meyer (Stanford University, Stanford, CA) and was cloned into pEGFP-N1 (Clontech Laboratories) between the EcoRI and BamHI sites.

Cell culture and immunocytochemistry

HEK293 cells were cultured in Eagle's minimal essential medium (EMEM) supplemented with 10% (v/v) fetal bovine serum, 50 μg/ml penicillin, 50 μg/ml streptomycin and 2 mM L-glutamine, and maintained at 37°C in a humidified atmosphere of 5% CO₂. The Lipofectamine 2000 method of transfection was used according to the manufacturer's instructions. After 48 hours, 800 μg/ml G-418 disulphate was added to select for transfected cells, which was reduced to 400 μg/ml after 48 hours.

Where appropriate, cells were incubated with 2 μM LatB, 10 mM MβCD or 200 mU/ml SMase in cell growth medium, for 1 hour at 37°C. Live cells were incubated with WGA (5 μg/ml), 3 minutes before fixing in 4% (v/v) paraformaldehyde for 10 minutes. Cells stained with phalloidin or anti-tubulin were permeabilized [0.1% (v/v) Triton X-100 for 10 minutes] and blocked [10% (v/v) goat serum for 1 hour];. Cells were incubated with phalloidin (1 IU/ml, 2 hours) or mouse anti-tubulin antibody [6 μg/ml, diluted in 1.5% (v/v) goat serum for 1 hour]. After rinsing in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.76 mM KH₂PO₄, pH 7.4), cells were incubated with Alexa-Fluor-555-conjugated goat anti-mouse antibody (10 μg/ml for 1 hour), followed by a final rinse in PBS. Non-specific binding could not be detected, nor could a significant fluorescent signal, when either the primary or the secondary antibody was added alone. All coverslips were mounted in Vectashield with DAPI.

AC in vitro activity assay

Crude membranes were prepared and AC activity was measured in vitro as described previously (Nakahashi et al., 1997; Boyajian et al., 1991).

Single cell cAMP measurements

Cells transiently expressing Epac2-camps were equilibrated at room temperature in HEPES-buffered saline (HBS; 140 mM NaCl, 4 mM KCl, 0.2 mM MgCl₂, 11 mM D-glucose and 10 mM HEPES, pH 7.4) containing 0.5 mM Ca²⁺. CCE was triggered following a pretreatment with 200 nM TG in HBS in the absence of Ca²⁺ (plus 100 μM EGTA) followed by the addition of 0.5 mM or 2 mM CaCl2 for AC8 and AC8M1, respectively, and in the presence of 10 nM FSK to potentiate CCE-dependent AC activity. A saturating FRET signal was stimulated [10 μM FSK, 2 mM Ca²⁺, 100 μM 3-isobutyl-1-methylxanthine (IBMX) and 10 μM isoproterenol] at the end of each experiment. Fluorescence was measured using an Andor CCD Ixon+ camera and Optosplit (505DC, Cairn Research, Faversham, UK) as described previously (Wachten et al., 2010) and analyzed using Metamorph software (Molecular Devices). Cells in which the CFP and YFP fluorescence intensity was less than twice the background signal were excluded, as were cells with excessive expression of the probe. Data were plotted as changes in the 470 nm/535 nm (CFP/YFP) emission ratio normalized to maximum FRET change for each cell.

Single cell Ca²⁺ measurements

Cells were loaded with 2 μM Fura-2/AM and 0.02% (v/v) Pluronic F-127 for 40 minutes with 2 μM LatB, 10 mM MβCD or 200 mU/ml SMase in cell growth medium at 37°C. Cells were incubated for a further 20 minutes in fresh pharmacological disruptor. Cells were washed in HBS with 0.5 mM Ca²⁺ before pre-treating with 200 nM TG in 0 mM Ca²⁺ HBS (100 μM EGTA), and CCE was triggered with 0.5 mM external Ca²⁺. Maximum fluorescence (F_Max) was obtained with 5 μM ionomycin and 10 μM Ca²⁺ to saturate the Fura-2 signal. Ca²⁺ traces are relative to the fluorescence ratio at 0 minutes (F/F₀) and normalised to the maximum ratio change for each cell. Cells were imaged using a CoolSNAP-HQ CCD camera (Photometrics) and monochromator system (Cairn Research) and analysed using MetaFluor software (Molecular Devices) (Wachten et al., 2010).

Acceptor bleaching FRET

The FRET efficiency between GFP–AC8 or GFP–AC8M1 (donor) and Ergosta-5,7,9(11),22-tetraen-3β-ol (ETO; acceptor) a cholesterol fluorescent analogue, was monitored using an FV1000 Olympus spectral confocal microscope in combination with the acceptor bleaching FRET technique. ETO was purchased from Sigma (cat. no. E2634). The cholesterol analogue was excited at 420 nm, peak emission occurs at 488 nm, which coincides with the excitation maxima for GFP, making this cholesterol analogue an excellent FRET pair. HEK293 cells expressing GFP–AC8 or GFP–AC8M1 were incubated for 10 minutes with the cholesterol fluorescent analogue used at a final concentration of 5 μM. Images were obtained at a rate of 1 every 5 seconds for the duration of the experiments. Pre- and post-bleaching fluorescence was determined for each individual experiment in the same cells. FRET efficiency was calculated as previously described (Salgado et al., 2008). All experiments were corrected for fluorescence bleed-trough before calculating FRET efficiency.

Confocal imaging and spot-bleach FRAP

Confocal images (1 μm optical slice thickness) were captured with a Leica SP5 TCS laser scanning confocal microscope attached to a DM16000 inverted microscope, equipped with a 63× plan-apochromatic 1.4 NA oil immersion objective (Leica) running the Leica Application Suite (LAS) AF Leica software, version 1.8.2. Cellular DAPI and GFP were visualized using the 405 nm and 488 nm line of an argon ion laser, and collected at 415–455 nm and 495–545 nm, respectively. Cells stained with Alexa Fluor 555 or Alexa Fluor 568 were visualized using the 543 nm line of a HeNe laser and collected between 615 and 700 nm. Cells expressing single fluorophores were used to adjust the laser emission and collection spectra to eliminate bleed-through into different channels. Graphical representations of product of the differences from the mean (PDM) values (n) were calculated using ImageJ where PDM = (red intensity – mean red intensity) × (green intensity – mean green intensity). Pseudocolour yellow represents colocalisation, whereas pseudocolour blue is where colocalisation is absent.

Cells were plated on glass coverslip inserts coated with poly-L-lysine for 24 hours and imaged on a Leica SP5 confocal microscope, in an environmental chamber at 37°C (Solent Scientific, Segensworth, UK). For maximum light acquisition and to defocus the bleaching light beam to effectively bleach GFP either side of the focal plane, the pinhole was fully opened to an optical section of approximately 5 μm. FRAP experiments were performed according to the FRAP wizard in the LAS AF software. Images were scanned bidirectionally at maximum speed (1400 Hz) with a single line average to reduce noise, using the 488 nm line of an argon ion laser and collected between 500 and 575 nm. Images (256×256 pixels) were acquired at a rate of 1 frame/second. Ten pre-bleach frames were acquired with a laser power of 5–8% before bleaching the membrane proteins (100 mseconds, 80% argon laser at 100% transmission) using a spot bleach protocol with the laser focused on the PM. Fluorescence recovery was followed for 200 seconds, with a laser power of approximately 5–8% to ensure full recovery.

Estimating the MF and D_FRAP

The mobile fraction (MF) and diffusion coefficient (D)_FRAP were calculated as previously described (Hammond et al., 2009; Oancea et al., 1998). In brief, Leica (lif) image stacks from FRAP spot bleach experiments were imported into ImageJ. The 10 pre-bleach frames were averaged and used as a baseline to normalize all post-bleach frames. Recovery curves of fluorescence intensity within the bleached region of the PM were determined using an ROI with a diameter of 1.54 μm over the bleached spot for all frames. The MF was determined by the maximal projected value, Y_MAX, of the recovery curves.

D_FRAP was determined using the segmented line tool in ImageJ to trace the bleached region across the PM and fluorescence intensity recorded in the normalized image stack for 5–20 frames (1 frame/second) using the ‘record profile’ macro. Total cellular fluorescence was adjusted for experimental photobleaching. The normalized and adjusted profiles were analyzed in Prism (GraphPad Software) and fitted independently with the Gaussian function: F(x)=1–Bxe–[(x–c)²/r²] where F is the normalized fluorescence intensity, c is the centre of the bleached profile (of distance x, in μm), B is the depth of the Gaussian profile and r is the Gaussian radius at e⁻¹. The values of r² were plotted against time (goodness of fit, R²>0.9) and the gradient was proportional to D_FRAP, obtained from: r²=4Dt+r₀², where t is time in seconds.

FCS

Cells were plated onto poly-L-lysine coated eight-well chambered coverglasses (Nunc Nalgene), and equilibrated to room temperature to minimize artefacts from temperature-induced PM fluctuations, before FCS measurements were taken on a Zeiss LSM510 ConfoCor 3 inverted confocal microscope using a 40× c-Apochromat 1.2 NA water-immersion objective. The confocal volume was calibrated on the day of each experiment using 20 nM Rhodamine 6G, the confocal volume was positioned over the cell nucleus, (488 nm excitation, with emission detected through a LP505 emission filter) and the focal plane was positioned precisely on the upper PM peak, using an intensity z-scan. Data were collected using 488 nm excitation with emission collected through a BP505-610IR emission filter, for 1×20 seconds at a laser intensity of 0.08 kW/cm² following a 10-second pre-bleach at the same laser power. Data were analyzed using Zeiss AIM4.2 software. Autocorrelation data was fitted using non-linear regression to a two-component, two-dimensional diffusion model, incorporating a pre-exponential component to account for fast GFP blinking. The first of these two-dimensional components was accounted for by photophysics of the GFP, and the second was representative of AC8 diffusion. The confocal volume dimensions were calculated from a calibration FCS read using 20 nM Rhodamine 6G, as previously described (Briddon et al., 2004). D_FCS were calculated using the equation D=ω₀²/4.τ_D, where ω₀ is the radius of the beam waist of the detection volume and τ_D is the average dwell time of the AC8 in the volume as determined by autocorrelation analysis. Particle number was calculated directly from the fractional contribution of the GFP diffusing component to the total particle number as determined by autocorrelation. This was subsequently expressed in particles per μm² (Nμ/m²), by normalizing to πω₀². Photon counting histogram (PCH) data were fitted to a single component curve using a bin time of 100 μseconds, with a first-order correction fixed to that obtained for a calibration read using 20 nM Rhodamine 6G. Careful examination of the data led to the exclusion of eight diffusion outliers (out of 50 recordings) for MβCD-treated cells, which were up to 240 times faster than the average. These very fast diffusions might represent the photophysical blinking of the GFP, as they were in the correct time range.

Western blot analysis

GFP-tagged AC8-based proteins were resolved using a 6% (w/v) SDS-polyacrylamide gel, before being transferred onto a nitrocellulose membrane and electrophoresed for 90 minutes at 300 mA. Membranes were blocked in TBS-T (20 mM Tris, pH 7.5, 150 mM NaCl and 0.05% (v/v) Tween 20) containing 5% (w/v) non-fat dry milk (16 hours at 4°C). Membranes were rinsed in TBS-T before incubating with anti-GFP antibody (2 μg/ml) for 1 hour and then rinsed again. Membranes were incubated with goat anti-mouse IgG conjugated to horseradish peroxidase (100 ng/ml) for 1 hour before rinsing in TBS-T then TBS (20 mM Tris, pH 7.5 and 150 mM NaCl). Membranes were visualized using ECL Plus reagent.

Statistics

Statistical significance of colocalisation, Ratio_PM/Cyt, cAMP accumulation and Epac2-camps experiments were determined using Student's t-tests with Welch's correction. Densitometry, FRAP and FCS data were analysed by one-way ANOVA with Neuman–Keuls multiple comparison tests. In each case, the number in brackets (n) refers to the number of cells measured in at least three separate experiments, or the number of times the experiment was repeated, as appropriate. Data are presented as the means ± s.e.m.; n.s., not significant; asterisks: *P<0.05, **P<0.01 and ***P<0.001 in comparison to untreated control or otherwise indicated.