LIS1, a WD40 repeat scaffold protein, interacts with components of the cytoplasmic dynein motor complex to regulate dynein-dependent cell motility. Here, we reveal that cAMP-specific phosphodiesterases (PDE4s) directly bind PAFAH1B1 (also known as LIS1). Dissociation of LIS1–dynein complexes is coupled with loss of dynein function, as determined in assays of both microtubule transport and directed cell migration in wounded monolayers. Such loss in dynein functioning can be achieved by upregulation of PDE4, which sequesters LIS1 away from dynein, thereby uncovering PDE4 as a regulator of dynein functioning. This process is facilitated by increased intracellular cAMP levels, which selectively augment the interaction of long PDE4 isoforms with LIS1 when they become phosphorylated within their regulatory UCR1 domain by protein kinase A (PKA). We propose that PDE4 and dynein have overlapping interaction sites for LIS1, which allows PDE4 to compete with dynein for LIS1 association in a process enhanced by the PKA phosphorylation of PDE4 long isoforms. This provides a further example to the growing notion that PDE4 itself may provide a signalling role independent of its catalytic activity, exemplified here by its modulation of dynein motor function.
Platelet-activating factor acetylhydrolase IB subunit α (PAFAH1B1 also known and, hereafter, referred to as LIS1) has an N-terminal coiled–coiled domain for self-association (Ahn and Morris, 2001) and a C-terminal β-propeller structure formed from seven WD40 repeats (Tarricone et al., 2004). Deletions within its WD repeat region elicit the severe neurodevelopmental disorder, lissencephaly (Cardoso et al., 2000), a severe neurodevelopmental disorder characterised by loss of convolutions of the cortical brain surfaces resulting from defective neuronal migration (Dobyns et al., 1993; Reiner et al., 1993). The fundamental importance of LIS1 is further highlighted as its deletion in mice is embryonic lethal (Cahana et al., 2001; Hirotsune et al., 1998).
In interacting with components of the cytoplasmic dynein complex, LIS1 regulates dynein functioning as a microtubule-based motor (Hirokawa et al., 1998). It regulates microtubule dynamics, affecting microtubule stability and organisation (Caspi et al., 2000; Sapir et al., 1997; Shu et al., 2004). LIS1 localises with dynein at the centrosome, mitotic spindles, kinetochores and cell cortex, thereby regulating spindle orientation and chromosome alignment in mitosis (Faulkner et al., 2000; Siller et al., 2005), nuclear and centrosomal transport (Shu et al., 2004; Tanaka et al., 2004; Tsai et al., 2005) and directed cell migration (Dujardin et al., 2003; Kholmanskikh et al., 2003). Aberrant LIS1–dynein interactions are, therefore, likely to be important in the aetiology of lissencephaly.
Four genes encode the cAMP-specific phopshodiesterase-4 (PDE4) family (PDE4A, PDE4B, PDE4C, PDE4D) with alternative splicing generating >20 isoforms (Houslay and Adams, 2003). A key functional consequence of this diversity is that specific isoforms are sequestered by distinct signalling and scaffold proteins to underpin compartmentalised cAMP signalling (Houslay, 2010). PDE4 isoforms are sub-categorised into long forms with UCR1 and UCR2 regulatory regions; short forms lacking UCR1 and super short forms lacking UCR1 but with a truncated UCR2. Elevation of intracellular cAMP activates long PDE4 isoforms through the cAMP-dependent protein kinase A (PKA) phosphorylation of a target serine residue located in UCR1. This triggers conformational changes in the UCR1–UCR2 module, one of which engenders activation (Beard et al., 2000; Hoffmann et al., 1998; MacKenzie et al., 2002; Sette and Conti, 1996).
Here, we show that PDE4 isoforms interact directly with LIS1 so as to regulate LIS1–dynein interactions and that such interaction is facilitated in long isoforms upon their PKA phosphorylation.
LIS1 and PDE4 interact in a manner facilitated by PKA activation
HEK293 cells were transfected to express transiently GFP-tagged LIS1 with VSV-epitope-tagged versions of either of the long PDE4B and PDE4D isoforms, PDE4B1 and PDE4D3, respectively. A GFP-tagged LIS1 construct was used in these assays, as identifying LIS1 in immunoprecipitates can be problematic due to its migration proximity to the Ig heavy chains from the immunoprecipitating antiserum. PDE4 isoforms were immunopurified from lysates using anti-VSV agarose and immune complexes blotted for both the PDE4 proteins and LIS1–GFP. Specific co-capture of LIS1–GFP was observed in the immunoprecipitates (Fig. 1A).
Where indicated, cells were treated with a combination of the adenylyl cyclase activator forskolin and the non-selective cyclic nucleotide phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (IBMX) to maximise intracellular cAMP levels. This was done to evaluate whether LIS1–GFP and PDE4B1–PDE4D3 interactions were sensitive to elevated intracellular cAMP. Interestingly, elevated cAMP enhanced the levels of LIS1–GFP associating with PDE4D3 and PDE4B1 (Fig. 1A).
We hypothesised that the cAMP-stimulated increase in the interactions of LIS1 with these long PDE4 isoforms is triggered by their phosphorylation through PKA within the conserved UCR1 domain (Hoffmann et al., 1998; Sette and Conti, 1996). It seemed unlikely that PKA exerts a direct effect on LIS1, because it lacks any consensus PKA phosphorylation site. Thus, co-immunoprecipitation experiments were carried out with HEK293 cells overexpressing LIS1 and the PDE4B2 short isoform, which lacks UCR1 and is, thus, unable to be PKA phosphorylated (Huston et al., 1997; MacKenzie et al., 2002). PDE4B2 weakly associates with LIS1 and this interaction is not facilitated upon the challenge using forskolin and IBMX (Fig. 1B).
LIS1–PDE4 interactions can also be observed between endogenous species in non-transfected cells. HEK293 cells express the long PDE4D5 isoform, the short PDE4B2 isoform (Lynch et al., 2005) and endogenous LIS1 (Fig. 1C). Probing PDE4D5 and PDE4B2 immunoprecipitates we noticed the presence of endogenous LIS1 in both (Fig. 1C). Furthermore, challenge with forskolin and IBMX increased the association of endogenous LIS1 with PDE4D5, but had no effect on the magnitude of LIS1 interaction with PDE4B2, as determined by quantitative analysis of the immunoblots (Fig. 1C).
Uniquely, PDE4D3 has a second site for PKA phosphorylation (Ser13) located within its isoform-specific N-terminal region (Hoffmann et al., 1998), whose phosphorylation increases its ability to bind to the muscle-specific A-kinase anchoring protein (mAKAP) scaffold (Carlisle Michel et al., 2004). However, the PDE4D3-S13A mutant, which cannot be phosphorylated by PKA at serine residue 13, exhibited a cAMP-sensitive increase in the association with LIS1–GFP that was similar to the one in observed for wild-type PDE4D3. By marked contrast, the PDE4D3-S54A mutant, which cannot be phosphorylated by PKA at serine residue 54, showed no increase in association with LIS1–GFP upon treatment with forskolin and IBMX (Fig. 1Di).
Interestingly, we noticed that LIS1–GFP binding to PDE4-S54A was somewhat elevated in comparison with wild-type PDE4D3. We have previously shown that this mutation is not conformationally ‘silent’, as it elicits an increased sensitivity to inhibition by the PDE4-selective inhibitor rolipram that is thought to be due to disruption of a putative hydrogen-bond involving the side-chain hydroxyl group of Ser54 (Hoffmann et al., 1998). Such a conformational change might account for the increased binding between LIS1–GFP and the PDE4D3-S54A mutant. We also analysed coexpression of LIS1–GFP and the mutant PD4D3-R51A:R52A, which disrupts the Ser54-containing PKA phosphorylation motif but does not elicit a change in rolipram inhibition from wild-type PDE4D3 (Hoffmann et al., 1998). LIS1–GFP association with this mutant was unaltered by cAMP elevation, with observed levels of LIS1–GFP binding similar to those associated with wild-type PDE4D3 under basal conditions (Fig. 1Dii).
Long PDE4 isoforms can be phosphorylated within their catalytic unit by ERK2, thereby eliciting their inhibition (MacKenzie et al., 2000). However, EGF-dependent activation of ERK2 did not alter LIS1–GFP interaction with PDE4D3–VSV in transfected HEK293 cells (data not shown).
PDE4 and LIS1 interact in living cells: BRET analyses
The dynamics of LIS1–PDE4 interactions were examined in living cells by bioluminescence resonance energy transfer (BRET) (Prinz et al., 2006) where Renilla luciferase (Rluc) was fused to either wild-type PDE4D3 or PDE4D3-S54A, and the improved wild-type GFP, GFP2, was fused to LIS1 (GFP2–LIS1). No significant BRET signal was observed between either of the PDE4D3 donors and the LIS1 acceptor in unstimulated cells. However, the combined challenge of cells with forskolin and IBMX elicited a clear BRET signal between LIS1 and wild-type PDE4D3, but not between LIS1 and mutant PDE4D3-S54A (Fig. 1E). Thus PKA phosphorylation of UCR1 promotes interaction of PDE4D3 with LIS1 in living cells. However, the sensitivity of this methodology was insufficient to identify the small interacting pool in resting cells resolved by immunocapture (Fig. 1A).
PDE4 and LIS1 interact directly
To determine whether LIS1 and PDE4 interact directly, in vitro pull-down assays were performed using recombinant His-tagged LIS1 plus MBP-tagged PDE4B1 and PDE4D3 long isoforms, and the short PDE4B2 isoform. Successful pull-down of His–LIS1 protein was observed with all these species (Fig. 2A,B). Furthermore, phosphorylation of recombinant MBP–PDE4D3 by the activated PKA catalytic unit, followed by treatment with phosphorylation-specific antiserum (MacKenzie et al., 2002), caused increased association between MBP–PDE4D3 and His–LIS1 (Fig. 2B). Thus LIS1 and PDE4 interact directly, and PKA phosphorylation of long PDE4 enhances its interaction with LIS1.
We have previously employed scanning peptide array technology to map potential binding sites between PDE4 isoforms and various partner proteins (Baillie et al., 2007; Bolger et al., 2006; Collins et al., 2008; Murdoch et al., 2007). Here, we use purified MBP–PDE4D3 as a probe to interrogate a scanning peptide array of LIS1. A library of overlapping peptides (25-mers), sequentially shifted by five amino acids across the entire sequence of LIS1, was immobilized on cellulose membranes and probed with MBP–PDE4D3. This revealed two putative binding sites, corresponding to residues 41–80 and 216–250 in LIS1, which mapped to the coiled–coil region and WD repeats 3–4, respectively (Fig. 3A).
As a peptide formed from LIS1 residues 226–250 provided the strongest binding spot signal we subjected it to alanine-scanning substitution analysis. Its sequence was used to generate a family of peptide spots in which sequential residues were either substituted with alanine or, if the residue in the parent peptide was alanine, with aspartate. Substitution at several residues resulted in decreased interaction of PDE4D3 (Fig. 3A), with potentially important binding residues being Met239, Arg241, Asn243, Gln244 and Gly246 (Fig. 3A).
Stearoylated peptides can enter cells and disrupt protein–protein complexes by competing for binding partners (Hundsrucker et al., 2006). We have successfully employed these to disrupt PDE4 interactions with DISC1 (Murdoch et al., 2007) and βarrestin (Bolger et al., 2006) in intact cells and to disrupt MEK1–βarrestin interaction (Meng et al., 2009). Thus, to evaluate whether the LIS1 226–250 region is involved in PDE4 sequestration in vivo, we used a stearoylated LIS1 226–250 peptide in an effort to disrupt LIS1–PDE4 complexes in cells. As a control we generated a stearoylated (226–250)-LIS1 peptide where the residues indicated from scanning analyses were alanine substituted. Co-immunoprecipitation experiments with HEK293 cells ectopically expressing LIS1–GFP and PDE4D3 showed that, whereas treatment with the native LIS1 peptide clearly reduced LIS1–PDE4D3 interaction, treatment with the mutant peptide had no obvious effect (Fig. 3B).
We applied this approach to evaluate PDE4D3–LIS1 interaction in living cells that expressed GFP2–LIS1 and RlucPDE4D3 and were incubated with these stearoylated peptides prior to the challenge with forskolin and IBMX to elicit a BRET signal due to LIS1–PDE4D3 interaction (vide supra). Whereas in cells pre-treated with the stearoylated native LIS1 (226–250) peptide the challenge with forskolin and IBMX clearly failed to induce a BRET signal indicative of PDE4D3–LIS1 interaction, this was clearly evident in cells treated with the stearoylated mutant peptide (Fig. 3B). Thus, stearoylated native LIS1 (226–250) peptide disrupts PDE4D3–LIS1 interaction in living cells.
Identifying the LIS1 interaction site on PDE4
A PDE4D3 peptide library was probed with His–LIS1 to identify potential interaction sites on PDE4D3 for LIS1. This revealed three potential binding regions encompassing residues 26–60, 81–130 and 581–615 of PDE4D3 (Fig. 4A), which are located, respectively, within UCR1, LR1 (the region linking UCR1 and UCR2) and the C-terminal part of the catalytic domain.
Alanine-scanning substitution arrays of parent peptides of these regions were probed with His–LIS1 to gain insight into residues key for LIS1 interaction (Fig. 4B). Scanning of the first region indicated that substitution of either Arg51 or Arg52 ablated LIS1 interaction in this peptide. Scanning of the second region showed that substitution of any of the residues between Val100 and Phe104 severely attenuated LIS1 interaction. Scanning of the third region showed that substitution of Phe598 and Phe600 severely attenuated LIS1 interaction.
On the basis of these analyses, a series of VSV-epitope-tagged PDE4D3 constructs were generated with alanine mutations of key residues within each of the three predicted LIS1-interaction sites, namely R51A:R52A, R101A:N102A:N103A, and F598A:F600A. Constructs with single and combined mutations at these sites were made. Although none of the individual mutant constructs ablated PDE4D3 binding to LIS1, such interaction was clearly compromised with the ‘triple’ mutant, in which all three LIS1-binding sites had been mutated (Fig. 4C). This indicates that long PDE4 isoforms have multiple sites involved in LIS1 binding. That none of these mutations altered the catalytic activity of PDE4D3 (Fig. 4C), indicates that none trigger any gross, deleterious conformational change.
PDE4 modulates LIS1–dynein interaction
It has been widely reported that LIS1 localises with cytoplasmic dynein and its accessory complex dynactin, where it facilitates the functioning of the motor complex in various dynein-related cellular processes (Vallee and Tsai, 2006; Wynshaw-Boris, 2007). As LIS1 interaction with long PDE4 isoforms is sensitive to increased levels of cAMP, we set out to evaluate the role of cAMP–PDE4 signalling in the dynamics of LIS1–dynein interaction.
Endogenous dynein intermediate chain (DIC) was detected as a co-immunoprecipitating protein in LIS1 immune complexes (Fig. 5A) and endogenous LIS1 was visualised in DIC immune complexes (Fig. 5B). However, the interaction of either endogenous LIS1 or exogenous LIS1–GFP with endogenous DIC was clearly attenuated in cells treated with forskolin and IBMX together (Fig. 5A,B). Thus cAMP elevation elicits opposing effects on the interaction of LIS1 with dynein (negative) and with long PDE4 isoforms (positive). We thus set out to evaluate whether PDE4 and dynein provide competitive binding partners for LIS1. Particularly, whether cAMP elevation, in facilitating the recruitment of endogenous LIS1 to PKA-phosphorylated long PDE4 isoforms, might elicit the sequestration of LIS1 away from dynein with functional consequences for dynein.
Either knockdown or overexpression of LIS1 can have detrimental effects on various dynein-dependent functions (Faulkner et al., 2000; Shu et al., 2004; Smith et al., 2000; Tai et al., 2002; Tsai et al., 2005). It is, therefore, possible that alterations in available levels of free versus sequestered LIS1 alter the levels of LIS1–dynein complexes in cells and that this can be expected to impact on dynein cellular function. To determine whether PDE4 impacted on dynein function we set out to assess whether the transport of microtubules in cells is influenced by alterations in expression levels of PDE4 and by elevating intracellular cAMP. COS7 cells express endogenous long PDE4 isoforms of the PDE4A, PDE4B and PDE4D subfamilies (Fig. 5C), but do not express PDE4C isoforms (data not shown). Immunoblotting of COS7 cell lysates detected bands of molecular weights that have been observed for PDE4 long isoforms PDE4A4, PDE4A10 and PDE4A11, PDE4B1 and PDE4B3, and PDE4D4, PDE4D5, PDE4D7, PDE4D8 and PDE4D9. Approximately 96% of the total PDE4 activity in COS7 cells was inhibited by pre-treatment of cells with 10 μM rolipram (data not shown). As a main fraction of the total PDE4 activity within these cells is attributable to PDE4D (Fig. 5D), and we have characterised interaction of LIS1 with the PDE4D3 isoform in transfected cells, we decided to focus on the PDE4D sub-family. Thus co-immunoprecipitation studies demonstrate that endogenous LIS1–PDE4D complexes can be isolated from these cells, with LIS1 showing increased association with PDE4D upon elevation of intracellular cAMP by challenge with forskolin and IBMX (Fig. 5E). Dynein co-association in COS7 cell lysate PDE4D immunoprecipitates was not observed (data not shown).
The dynamic transport of microtubules outward from the microtubule organisation centre (MTOC) is a dynein-dependent process (Abal et al., 2002; Ahmad et al., 1998). Here, we used an established procedure to carry out microtubule transport assays (Shu et al., 2004), where COS7 cells were first treated with nocodazole to depolymerise the microtubule network within cells (Fig. 6A). Then, following removal of nocodazole, the synthesis and transport of new microtubules from the MTOC was allowed to recover for 3 minutes before cells were treated with vinblastine to prevent further microtubule synthesis. Using this procedure, newly formed microtubules can be visualised at the cell periphery (Fig. 6B).
Following the recovery period from nocodazole treatment we challenged cells with forskolin and IBMX together to maximally increase cAMP levels, or with rolipram to inhibit PDE4 activity alone. These challenges had a striking effect on microtubule transport. In control cells, microtubules radiated from the perinuclear region outward towards the cell periphery (Fig. 6C), whereas in cells treated with either forskolin and IBMX, or rolipram exhibited small microtubule asters at the MTOC, with few microtubules present beyond the perinuclear region (Fig. 6C). These data indicate that microtubule transport is severely impaired in cells where either PDE4 alone is inhibited or when cAMP levels are maximally increased.
Using this system we set out to evaluate whether the molecular event underpinning the dramatically altered microtubule transport caused by cAMP elevation through these routes is due to the sequestration of LIS1 away from dynein by increased complex formation between LIS1 and PKA-phosphorylated long PDE4 isoforms. To do this, we treated cells with the stearoylated native LIS1 (226–250) peptide to block PDE4 sequestration of LIS1, using the stearoylated mutant LIS1 peptide as a control. In both untreated and mutant LIS1 peptide-treated cells, addition of forskolin and IBMX blocked microtubule movement outward from the MTOC (Fig. 7Ai). In contrast to this, when cells were pre-treated with the native LIS1 peptide the forskolin and IBMX addition failed to prevent radial migration of microtubules towards the cellular periphery (Fig. 7Ai). We noticed, however, that the degree of microtubule transport visualised in the stearoylated native LIS1 peptide-treated cells was not quite as profound as the levels observed in unstimulated (control) cells (Fig. 7Ai). These data indicate that inhibiting PDE4 sequestration of LIS1 attenuates the ability of elevated cAMP levels to inhibit microtubule transport. To demonstrate that the effects of the native LIS1 (226–250) peptide on microtubule transport were not a consequence of disrupting LIS1–dynein interactions, DIC immunoprecipitates from COS7 cells were incubated with recombinant His–LIS1 protein followed by treatment with vehicle or native LIS1 (226–250) peptide or mutant LIS1 (226–250) peptide. In each instance, the co-capture of His–LIS1 in DIC immunoprecipitates was unaltered by incubation with the stearoylated LIS1 peptides when compared with vehicle-treated cells (Fig. 7Aii).
We next set out to determine whether increased PDE4 expression affects microtubule transport. To evaluate this, cells were transfected with GFP-tagged forms of either long PDE4B1 or short PDE4B2, or with GFP alone as a control. In GFP-expressing control cells, microtubules showed marked outward progression toward the edges of the cell in the absence of cAMP modulators (Fig. 7B) and, as expected, challenge with forskolin and IBMX together caused microtubule movement to be inhibited and aster formation at the perinuclear region (Fig. 7B). However, ectopic expression of either GFP–PDE4B1 or GFP–PDE4B2 clearly attenuated microtubule transport in both unstimulated cells and in cells treated with forskolin and IBMX together (Fig. 7B). As ectopically expressed PDE4 sequesters endogenous LIS1 (see above), it is to be entirely expected that an attenuated microtubule transport occurs if such sequestration elicits a reduction in the LIS1–dynein complexes required for microtubule transport. This result would be consistent with the notion that elevated cAMP levels achieve their inhibition by increasing LIS1 sequestration by PKA-phosphorylated long isoforms. Indeed, if inhibition of microtubule transport by elevated cAMP levels functions through another route, one intuitively predicts that ectopic expression of GFP–PDE4B1 and GFP–PDE4B2 enhances – rather than inhibits – microtubule transport in unstimulated cells as these enzymes actively degrade (inhibitory) cAMP.
Co-immunoprecipitation studies demonstrate that endogenous LIS1–dynein interaction is attenuated by increased intracellular cAMP (Fig. 5A). Thus, the observed inhibition of microtubule transport in response to the combined challenge with forskolin and IBMX is consistent with a decrease in LIS1–dynein complexes within the cells. In support of this, ectopic expression of LIS1 suffices to negate the cAMP-induced blockade of microtubule transport (Fig. 7C), indicating that LIS1 overexpression can restore functional LIS1–dynein complexes in cells treated with forskolin and IBMX together. The fact that overexpression of either GFP–PDE4B1 or GFP–PDE4B2 negatively regulates microtubule transport suggests that these exogenous PDE4s act to sequester endogenous LIS1, thereby lowering levels of the LIS1–dynein functional complex. To evaluate this, co-immunoprecipitation experiments were carried out using COS7 cells transfected with GFP alone, GFP–PDE4B1 or GFP–PDE4B2 and, where indicated, followed by challenge with forskolin and IBMX. Endogenous DIC was immunoprecipitated from the cell lysates and complexes probed for LIS1 capture. Immunoblotting revealed that LIS1 co-precipitated with DIC in GFP-transfected control cells under basal conditions. However, upon combined stimulation with forskolin and IBMX, LIS1 binding to DIC was ablated (Fig. 8). Furthermore, in cells that overexpressed either GFP–PDE4B1 or GFP–PDE4B2, no DIC–LIS1 complexes were detected in basal or cAMP-stimulated cells (Fig. 8).
Our hypothesis implies that PDE4 isoforms are able to regulate dynein function independently of their catalytic activity. It is possible to make a catalytically inactive form of PDE4 species by mutating, to alanine, a single aspartate group located deep within the catalytic site so as not to elicit any gross conformational change (Baillie et al., 2003; McCahill et al., 2005). The inhibitory action, on microtubule transport, of ectopic expression of such a catalytically inactive PDE4D3 was clearly evident (Fig. 9A).
If the inhibitory effect of PDE4 upregulation on microtubule transport were because LIS1 is sequestered away from dynein, we predict a PDE4 isoform (mutated so as to compromise binding to LIS1) to compromise its ability to negatively regulate microtubule transport. Using PDE4D3 as an example, although ectopic expression of the wild-type construct resulted in the blockade of microtubule transport, in cells that expressed similar levels of the LIS1-binding-compromised PDE4D3 triple mutant, microtubule transport was not inhibited to this extent – with long microtubules extending out from the MTOC to the periphery clearly evident (Fig. 9A). The overall conformational status of this mutant is unlikely to be adversely affected because it exhibited a similar catalytic activity compared with wild-type PDE4D3 (Fig. 4C).
Loss of inhibitory action on microtubule transport with this PDE4D3 triple mutant suggests that LIS1 is more accessible to dynein in cells transfected with this mutant, in comparison with cells transfected with wild-type PDE4D3. Indeed, in DIC immunoprecipitates, LIS1 co-capture was ablated in cells transfected with either wild-type PDE4D3 or catalytically inactive PDE4D3 (Fig. 9B). By marked contrast, in cells transfected with the PDE4D3 triple mutant DIC–LIS1 complexes at a level comparable with that observed in GFP-transfected cells were evident (Fig. 9B).
PDE4 modulates directed cell migration
LIS1 and dynein are functionally important in directed cell migration, with knockdown of dynein or LIS1 impeding forward migration of fibroblasts (Dujardin et al., 2003). Here, we set out to investigate the potential role of altered PDE4 expression in cell migration regulated by LIS1–dynein using wounded monolayers of A2780 ovarian cancer cells (Belotti et al., 1996). The movement of individual cells at the wound zone was tracked over a 20-hour period and the trajectories were plotted. Track plots of untreated cells showed that cells exhibited directed cell migration into the wound area to repair the monolayers (Fig. 10A). In contrast to this, track plots of cells treated with the PDE-resistant cAMP analogue, Sp-8-Br-CAMPS demonstrated a more-random pattern of cell migration at the wound edge (Fig. 10A). As above, we set out to determine whether such disruption of directed cell migration is attributed to increased levels of PDE4-sequestered LIS1, due to PKA-phosphorylated PDE4, by disrupting LIS1–PDE4 complexes with the stearoylated LIS1 (226–250) peptide. In cells pre-incubated with mutant LIS1 (226–250) peptide, which fails to disrupt such complexes (see above), the migration phenotype of Sp-8-Br-CAMPS-challenged cells was unaltered (Fig. 10A). However, when cells were pre-treated with the stearoylated native LIS1 (226–250) peptide the effects of Sp-8-Br-CAMPS treatment on cell migration were negated, with cells exhibiting directed movement within the wound (Fig. 10A). Measurements of the persistence of cell movement, defined as the ratio of the vectorial distance travelled to total path length, showed that elevation of cAMP levels led to loss of persistent cell migration (Fig. 10A). The decrease in persistence of cell migration in response to Sp-8-Br-CAMPS treatment was reversed when cells were pre-incubated with the stearoylated LIS1 (226–250) peptide (Fig. 10B). Thus the inhibitory effects of elevated cAMP on directed cell migration can be countered by blocking PDE4 sequestration of LIS1.
These data imply that ectopic PDE4 expression has a negative effect on cell migration in the wounded monolayers by sequestering LIS1. To test this, cells were transfected to express either wild-type PDE4D3, catalytically inactive PDE4D3 or the LIS1-binding-compromised PDE4D3 triple mutant. Track plots of mock-transfected cells showed that cells migrated with persistent movement into the wound area to repair the monolayers (Fig. 10C), whereas cells overexpressing wild-type PDE4D3 migrated more randomly. This effect was not simply linked to an increased intracellular PDE activity because a similar, random pattern of migration was observed in cells ectopically expressing the catalytically inactive PDE4D3 mutant (Fig. 10C). In marked contrast to this, tracks of cells exogenously expressing the LIS1-binding-compromised PDE4D3 triple mutant revealed a directed pattern of cell migration more akin to that observed in mock-transfected and control cells (Fig. 10C). Analysis of the persistence of cell migration showed that the LIS1-binding-compromised PDE4D3 triple mutant exhibited the most persistent migrational phenotype when compared with mock-transfected cells (Fig. 10D). The rescue of persistent cell migration in cells overexpressing the PDE4D3 triple mutant suggests that more of the LIS1 pool is available to facilitate dynein function in cells that express the LIS1-binding-defective PDE4D3 mutant, in comparison with cells with exogenous wild-type PDE4D3 expression.
Here, we show that the dynein-regulatory protein LIS1 interacts with cAMP-degrading PDE4 family members, where the binding of long PDE4 isoforms to LIS1 is dynamically facilitated by their cAMP-dependent PKA phosphorylation. Intriguingly, we demonstrate that PDE4 and dynein compete for interaction with LIS1. Thus, enhanced sequestration of LIS1 by PDE4 can be triggered either by increased expression of PDE4 isoforms or by elevation of intracellular cAMP levels causing PKA phosphorylation of long isoforms. This has a functional outcome, namely a reduction of LIS1–dynein complexes with consequent downregulation of dynein function. We demonstrate, for the first time, that PDE4, potentially, exerts a signalling (regulatory) role by virtue of its scaffolding ability in addition to its well-recognised enzymatic role as a terminator of cAMP signalling (Fig. 11).
We have shown that PDE4 interacts with sites located within both the N- and C-terminal parts of LIS1, and that a cell-permeable peptide encompassing the C-terminal site can be used to disrupt PDE4–LIS1 interactions in cells. The C-terminal interaction site for PDE4 lies within the β-propeller folds of LIS1, which are crucially involved in functional dynein binding (Tai et al., 2002; Tarricone et al., 2004). This is consistent with our proposal that PDE4 and dynein bind to LIS1 at mutually exclusive sites within LIS1, and our discovery that increased PDE4 sequestration of LIS1 disrupts LIS1–dynein complexes. LIS1 binds to both long and short PDE4 isoforms, independently of the sub-family. However, the regulatory UCR1 domain that characterises long isoforms allows them to be phosphorylated by PKA and, here – we have identified a so-far-unknown functional consequence of this – triggers enhanced association with LIS1.
LIS1 interacts directly with individual dynein subunits (McKenney et al., 2010; Mesngon et al., 2006; Sasaki et al., 2000; Tai et al., 2002) and we were able to co-immunoprecipitate endogenous LIS1 with DIC in COS7 cells. Elevation of intracellular cAMP elicited opposing effects on the interaction of LIS1 with endogenous long PDE4 isoforms and dynein, suggesting that PDE4 and dynein share overlapping binding sites on LIS1. This was further evidenced from the disruption of LIS1–dynein complexes in cells that exogenously express PDE4. As LIS1–dynein association was unaltered in cells that overexpress the LIS1-binding-compromised PDE4 mutant, we propose that PDE4 disrupts LIS1–dynein coupling by sequestering LIS1. PDE4 and dynein are, therefore, likely to compete with each other for LIS1 interaction, where modulation of cAMP levels and/or PDE4 expression is able to influence the levels of LIS1 associated with dynein. Indeed, MAP1B binding to LIS1 has also been reported to reduce LIS1–dynein coupling (Jimenez-Mateos et al., 2005) and the α1 and α2 subunits of PAFAH Ib have been shown to sequester LIS1, again decreasing the LIS1 pool associated with dynein (Ding et al., 2009).
Attenuation of LIS1 interaction with dynein disrupts dynein-dependent function in neuronal migration, mitosis, nucleokinesis, microtubule outgrowth and non-neuronal cell migration (Ding et al., 2009; Dujardin et al., 2003; Faulkner et al., 2000; Shu et al., 2004; Tai et al., 2002; Tsai et al., 2005). Using microtubule transport and cell migration assays to assess dynein activity, we showed that dynein function was decreased in cells with reduced LIS1–dynein complexes. Increased intracellular cAMP inhibited microtubule outgrowth and decreased the persistence of directed cell migration in wounded monolayers. However, these inhibitory effects of cAMP on dynein activation were reversed by either LIS1 overexpression or by using a membrane-permeable peptide that disrupts the PDE4–LIS1 complex so as to block LIS1 sequestration by PDE4, thereby restoring functional levels of the LIS1–dynein complex. PDE4-overxpressing cells exhibited phenotypes similar to those of cAMP-challenged cells, with decreased microtubule transport and more-random cell migration. Such downregulation of dynein function following PDE4 overexpression is not mediated by an increased intracellular PDE activity because similar effects were observed in cells that overexpress a catalytically inactive PDE4 species. Rather, compromised dynein function was due to increased sequestration of LIS1 by PDE4 with concomitant reduction in LIS1–dynein complexes. Consistent with this, no such changes in microtubule transport and directed cell migration were observed in cells transfected to express the LIS1-binding-defective PDE4 triple mutant.
Here, we have identified a so-far-unknown interaction between LIS1 and PDE4 isoforms, whereby PDE4 can modulate LIS1–dynein complexes and dynein-dependent processes within cells by sequestering LIS1. This may have functional consequences in indications where increased PDE4 expression has been observed, such as the increased levels in PDE4A4 long isoform noted in macrophages of patients with chronic obstructive pulmonary disease (Barber et al., 2004) and the upregulation of PDE4A4 and/or PDE4A5 in the hippocampus during sleep deprivation that leads to cognitive deficits (Vecsey et al., 2009). Our study prompts us to propose that PDE4 has a signalling role other than solely its ability to degrade cAMP, namely in this instance to act as a signalling scaffold able to regulate dynamically LIS1 cellular regulation of dynein. This concept may also allow insight into why the widely expressed catalytically inactive PDE4A7 is generated, because it might perform a regulatory scaffolding role in another system (Horton et al., 1995; Johnston et al., 2004).
Materials and Methods
Forskolin, 3-isobutyl-1-methylxanthine (IBMX), anti-VSV monoclonal antibody, anti-dynein intermediate chain antibody and cy3-β-tubulin antibody were from Sigma-Aldrich (UK), Sp-8-Br-cAMPS was from Biolog (Bremen, Germany), anti-LIS1 monoclonal and polyclonal and anti-GFP antibodies were from Abcam (Cambridge, UK), anti-MBP monoclonal antibody was from New England Biolabs (Hertfordshire, UK), PDE4B2 polyclonal antiserum was from Millipore (Watford, UK), antisera specific for PDE4A, PDE4B, PDE4C, PDE4D, PDE4D5 and phosphorylated UCR1 have been described previously (Bolger et al., 1997; Huston et al., 1996; Huston et al., 1997; MacKenzie et al., 2002; MacKenzie and Houslay, 2000). Stearoylated peptides (Hundsrucker et al., 2006) were stearoyl–CVKTFTGHREWVRMVRPNQDGTLIA (LIS1 226–250 peptide) and stearoyl–CVKTFTGHREWVRAVAPAADATLIA (LIS1 mutant peptide).
Mammalian expression constructs
pcDNA3.1-LIS1 construct was a kind gift from Li-Huei Tsai (Howard Hughes Medical Institute, MIT, MA). A C-terminally GFP-tagged LIS1 construct was generated by subcloning the ORF of LIS1 into pEYFP-N1 vector (Clontech). GFP2-LIS1 was generated by insertion of the LIS1 ORF into pGFP2-mcs-Rluc(h) (PerkinElmer, Boston, MA). Plasmids encoding PDE4B1, PDE4B2, PDE4D3, PDE4D3-S54A, PDE4D3-R51A,R52A, PDE4D3-S13A, RlucPDE4D3 and catalytically inactive PDE4D3 have been described previously (Collins et al., 2008; Hoffmann et al., 1998; Huston et al., 1997). Cloning PDE4B1 ORF into pEGFP-C1 (Clontech) and PDE4B2 ORF into pEGFP-C3 (Clontech) generated N-terminal GFP-tagged forms. PDE4D3 mutants were made using QuikChange (Stratagene).
Bacterial expression constructs
The PDE4B2 ORF was subcloned into pMALC2 vector (New England Biolabs) to generate N-terminal maltose-binding protein (MBP) fusion protein. MBP-tagged forms of PDE4B1 and PDE4D3 as described previously (Murdoch et al., 2007). The cDNAs for N-terminally hexahistidine-tagged full-length LIS1 and LIS1 truncates of residues 1–187 and 188–410 were generated by PCR using pcDNA3.1-LIS1 as a template and cloning into pET28c vector (New England Biolabs).
Cell culture and transient transfection
HEK293 and COS7 cells were maintained in DMEM plus 10% FCS, 10000 U/ml penicillin-streptomycin and 2 mM glutamine. A2780 cells were cultured in RPMI supplemented with 10% FCS and 2 mM glutamine. HEK293 and COS7 cell lines were transfected using Lipofectamine™ 2000 reagent (Invitrogen) according to recommended protocols. A2780 cells were transfected using an Amaxa nucleofector with solution T.
Immunoprecipitation and immunoblotting
Immunoprecipitation was performed as described previously (Collins et al., 2008; Murdoch et al., 2007). Western blotting was performed with the indicated primary and complementary horseradish peroxidase (HRP)-conjugated secondary antibodies and visualization with enhanced chemiluminescence (Pierce, Northumberland, UK). Immunoblots shown are representative images of at least three independent experiments.
Purification of MBPand His recombinant proteins and in vitro pull-down assays
MBP-tagged proteins were purified as described before (Murdoch et al., 2007). His-tagged LIS1 expression constructs were transformed into Escherichia coli BL21 cells and purified using Ni-NTA resin (Qiagen). For pull-down experiments, equimolar amounts of purified MBP-tagged PDE4 proteins were incubated with amounts of purified His-tagged LIS1 protein in molar excess. Protein complexes were isolated on amylose resin and assessed by SDS-PAGE and immunoblotting as described before (Murdoch et al., 2007).
SPOT synthesis of peptides and overlay experiments
Bioluminescence resonance energy transfer (BRET) analyses were done as described before (Collins et al., 2008). Control measurements of cells that express Renilla luciferase (Rluc) without a fusion partner were included in each experiment to determine the background BRET signal. For each condition, six wells were evaluated and experiments were repeated at least three times. Statistical analyses were carried out using GraphPad Prism 4 software.
Phosphodiesterase activity assay
Phosphodiesterase activity assay was done as described before, using 1 μM cAMP substrate and following linear rates (Marchmont and Houslay, 1980). Data shown are representative of three independent experiments.
Microtubule transport assay and confocal microscopy
Microtubule transport assays were performed using established methodology (Shu et al., 2004). COS7 cells grown on coverslips were transfected with various PDE4 constructs and/or GFP. Two days post-transfection, cells were incubated for 3 hours with 5 μg/ml nocodazole, washed three times in culture medium and allowed to recover for 3 minutes when 50 nM vinblastine was added for 20 minutes. Where indicated, forskolin and IBMX in combination (100 μM each) or rolipram (10 μM) were added together with vinblastine. Cells were fixed in 4% paraformaldehyde/TBS solution for 20 minutes at room temperature. Cy3-β-tubulin antibody [1:400 dilution in PBS/5% (v/v) donkey serum/1% (w/v) BSA] was used for immunostaining cells. Coverslips were mounted onto glass slides with Antifade/DAPI solution (Millipore) and staining visualised using a Zeiss Pascal LSM510 laser-scanning confocal microscope. At least 50 cells were imaged for each group in an individual experiment. Three independent experiments were performed.
A2780 cells were grown to confluence in six-well plates. A wound was made by scraping the cell monolayer with a micropipette tip. Cells were visualised every 15 minutes (C4742-95 digital camera) for 20 hours in an atmosphere of 5% CO2 at 37°C on a Carl Ziess microscope using a 20× phase-contrast objective. Stack movies and cell tracks were generated using Bioimaging software (Andor). For each condition, experiments were carried out at least three times. Statistical analyses were performed using GraphPad Prism 4 software.
This work was supported by grants from the Medical Research Council (U.K.; G0600765). Deposited in PMC for release after 6 months.