Regulation of Kalirin by Cdk5

In mammals, flies and worms, members of the Trio/Kalirin family of Rho guanine nucleotide exchange factors (GEFs) play essential roles in neuronal process outgrowth (Awasaki et al., 2000; Newsome et al., 2000; Bateman et al., 2000; Steven et al., 1998; Ma et al., 2003; Debant et al., 1996). In the adult nervous system, Kalirin-7, the major Kalirin (official protein symbol KALRN) isoform expressed, is essential for the maintenance of dendritic spines and the dendritic arbor (Ma et al., 2003; Ma et al., 2008). Kalirin and Trio are unique in having two Rho-GEF domains and multiple spectrin-like repeats (Rossman et al., 2005; Johnson et al., 2000; Debant et al., 1996). The spectrin-like repeat region of Kalirin is half the size of spectrin itself and affects cell morphology in a manner independent of GEF activity (Schiller et al., 2008). As is the case for many other Rho-GEFs, the factors that control the catalytic activity and actions of Kalirin are poorly understood (Rossman et al., 2005; Schiller et al., 2006). During development, the interaction of Trio with the protein `deleted in colorectal cancer' (DCC) allows it to mediate the effects of netrin-1 on axonal growth (Briancon-Marjollet et al., 2008). In mature neurons, receptor tyrosine kinases, such as EphB, and signaling proteins, such as CamKIIα, affect the ability of Kalirin-7 to participate in the formation of dendritic spines (Penzes et al., 2003; Xie et al., 2007).

Cdk5, a proline-directed kinase, phosphorylates substrates that regulate axon guidance, cytoskeletal dynamics and the formation and function of dendritic spines (Bibb et al., 2001b; Cheung et al., 2006; Dhavan and Tsai, 2001; Humbert et al., 2002; Lee et al., 1996; Nikolic et al., 1998), the same events in which Kalirin plays a key role. Cdk5 is active only when bound to a noncyclin cofactor, p35 or p39 (Dhavan and Tsai, 2001). Neuronal growth cones are enriched in the Cdk5-p35 complex, along with Kalirin, Trio and Rac1 (Nikolic et al., 1998; Awasaki et al., 2000; Johnson et al., 2000; Humbert et al., 2002; Estrach et al., 2002). Trio can be phosphorylated by Cdk5, and inhibition of Cdk5 results in a decrease in the ability of Trio to activate Rac (Xin et al., 2004). In cells that secrete neuropeptides, inhibition of Cdk5 results in the accumulation of filamentous actin beneath the plasma membrane and in a decrease in the regulated release of peptide-containing granules (Xin et al., 2004). Since Kalirin has several potential Cdk5 phosphorylation sites, and Cdk5 and Kalirin are involved in many of the same physiological processes, we sought an experimental system in which we could determine whether Cdk5 plays a role in controlling Kalirin function.

We chose rat PC12 pheochromocytoma cells to evaluate this possibility. Although NGF treatment of PC12 cells increases the expression of p35 and the activation of Cdk5 (Harada et al., 2001), nondifferentiated PC12 cells express Cdk5, p35 and p39 (Harada et al., 2001; Sharma et al., 1999; Tang et al., 1995). Antisense-mediated reduction of Kalirin expression in PC12 cells demonstrated a role for Kalirin in neurite extension (Chakrabarti et al., 2005). We did not examine the actions of Kalirin in NGF-differentiated PC12 cells because levels of Trio protein increase rapidly in response to treatment with NGF (Estrach et al., 2002). The dose and timing of NGF application are critical determinants of the response observed and Kalirin interacts directly with TrkA, resulting in activation of TrkA in the absence of ligand (Chakrabarti et al., 2005).

We chose to explore the effects of Cdk5 on Kalirin-7 because it is the major isoform in the adult rodent brain, has a single Rho-GEF domain and a single potential Cdk5 phosphorylation site, Thr1590 (Johnson et al., 2000; Penzes et al., 2000). Thr1590 is situated in a 60 amino acid region that follows the GEF domain and precedes the 20 amino acid sequence unique to Kalirin-7 (Fig. 1A) (Penzes et al., 2000). The same site is found in the larger isoforms of Kalirin as well as the N-terminally truncated Δ-isoforms. The larger isoforms of Kalirin are expressed early in development (Johnson et al., 2000) and cause axon initiation and extension when overexpressed in sympathetic neurons (May et al., 2002). Expression of Kalirin-7 is first seen when synapses begin to form during postnatal development. Since antisense-mediated reductions in Kalirin-7 expression in hippocampal pyramidal neurons and interneurons reduced the size of the dendritic tree (Ma et al., 2003; Ma et al., 2008), we were hopeful that exogenous Kalirin-7 would produce a morphological response that could be quantified in PC12 cells.

We expressed Kalirin-7 in nondifferentiated PC12 cells and found that it causes a GEF-activity dependent decrease in cell roundness and formation of cytoplasmic protrusions. These effects are blocked by dominant-negative Cdk5. We then demonstrated that Cdk5 phosphorylates Kalirin-7 on Thr1590, increasing its GEF activity slightly and changing its solubility properties. Inhibitors of protein phosphatase 1 (PP1) stabilize phosphorylation at this site. A Kalirin-7 mutant that cannot be phosphorylated by Cdk5 does not cause the formation of protrusions whereas the phosphomimetic mutant, Kalirin-7-Asp1590, does. The response of PC12 cells to Kalirin-7-Asp1590 is not blocked by dominant-negative Cdk5, placing Kalirin-7 downstream of Cdk5. To verify the importance of this interaction in neurons, we expressed Kalirin-7 containing an Ala1590 mutation in cortical neurons; although dendritic spines formed, they exhibited aberrant morphology.

Endogenous Kalirin plays an essential role in the ability of PC12 cells to produce neurites in response to stimulation with NGF (Chakrabarti et al., 2005). A direct interaction between TrkA and the PH domain of Kalirin-7 can lead to downregulation of TrkA, complicating the study of exogenous Kalirin in differentiated PC12 cells (Chakrabarti et al., 2005). PC12 cells expressing exogenous Kalirin-7 are unable to produce neurites in response to NGF (50 ng/ml; data not shown); however, in the absence of NGF, exogenous Kalirin-7 causes extension of broad cytoplasmic protrusions (Fig. 1B). Control cells (expressing GFP) are compact polygons, with filamentous actin concentrated at the periphery; filopodia are found at the vertices and lamellipodia are rare (Fig. 1B). The cytoplasmic protrusions extended in response to Kalirin-7 are decorated with small lamellipodia; filamentous actin accumulates at the margins of the cell, often forming spokes within the lamellipodia. Punctate staining for Kalirin-7 is apparent in the perinuclear region and near the tips of the protrusions. Staining is also sometimes apparent in the nucleus; the reasons for this are not clear.

The protrusions formed in response to Kalirin-7 are easily distinguished from neurites formed in response to NGF (Black et al., 1986) (Fig. 1B). To quantify the morphological effects of Kalirin-7, cell perimeter and cell roundness (see Materials and Methods) were measured (Fig. 1E). Because of the large protrusions, cell perimeter is increased over threefold in PC12 cells expressing Kalirin-7. A perfectly round cell has a roundness indicator of 1.0; cells expressing Kalirin-7 adopt a more polarized morphology, resulting in an approximately fivefold decrease in cell roundness (Fig. 1E). Although this unusual response to Kalirin-7 resembles neither axonal extension nor spine formation, the fact that it is quantifiable allowed us to use it to test the role of Cdk5.

Like Kalirin, Cdk5 is involved in both neurite outgrowth and the formation of dendritic spines (Aoki et al., 2004; Harada et al., 2001; Dhavan and Tsai, 2001; Tan et al., 2003; Xin et al., 2004). Kalirin-7 contains a single consensus Cdk5 phosphorylation site, Thr1590, which is highly conserved (Fig. 1A). To test the role of Cdk5, which is active in nondifferentiated PC12 cells (Harada et al., 2001), in the morphological response of PC12 cells to Kalirin-7, we evaluated the ability of dominant-negative (DN) Cdk5 to block the response. We first expressed DN Cdk5 alone, which should inhibit endogenous Cdk5 (Tan et al., 2003); no significant morphological changes were observed in nondifferentiated PC12 cells (Fig. 1C). Coexpression of DN Cdk5 and Kalirin-7 reduced protrusion formation by Kalirin-7 (Fig. 1D). Cell perimeter increased less than twofold compared with the GFP control and was significantly less than that observed in cells expressing Kalirin-7 alone (Fig. 1E). Cells expressing Kalirin-7 and DN Cdk5 were more rounded than cells expressing Kalirin-7 alone, but are not as round as cells expressing GFP. Since DN Cdk5 inhibited Cdk5 without affecting the activity of Cdk1 or Cdk2 (Meikrantz and Schlege, 1996), our data suggest that Cdk5 plays a role in the morphological effects of Kalirin-7 on nondifferentiated PC12 cells.

In addition to His-Myc-tagged Kalirin-7 and ΔKalirin-7 (Fig. 1A), we generated mammalian and bacterial expression vectors encoding the DH/PH (GEF) domain followed by the C-terminus of Kalirin-7 (KGEF1-7end) (Fig. 2A). Vectors encoding the GEF domain (KGEF1) and the region unique to Kalirin-7 (Kal7-CT), which both lack Thr1590, were generated previously (Penzes et al., 2001; Schiller et al., 2006; Schiller et al., 2005). We first used synthetic peptides and recombinant Cdk5-p25 to explore the hypothesis that Thr1590 can be phosphorylated by Cdk5 (Fig. 2B). Cdk5-p25-dependent phosphorylation of the Thr1590-Pro-Ala-Lys (TPAK) peptide was observed. To verify the importance of Pro1591 in the ability of Cdk5-p25 to recognize this site, it was changed to Ala (TAAK); no Cdk5-p25-mediated phosphorylation was detected with the TAAK peptide. The K_m and V_max values determined for the TPAK peptide were comparable with those reported for the Cdk5 site in histone H1 (Qi et al., 1995). We next asked whether Cdk5-p25 could phosphorylate GST-KGEF1-7end (Fig. 2C). Robust phosphorylation was observed; no signal was detected without the addition of Cdk5-p25. As controls, and to verify the phosphorylation site, GST fusion proteins with point mutations of Thr1590 to Ala or Asp, and a control, GST-KGEF1 (which terminates at Arg1573, before the TPAK site) were incubated with Cdk5-p25. None of these proteins served as a Cdk5-p25 substrate. The smallest naturally occurring variant of Kalirin, ΔKalirin-7, which includes Thr1590, was also phosphorylated by Cdk5-p25. Despite the presence of similar amounts of KGEF1-7end and ΔKal7, KGEF1-7end is phosphorylated much more extensively, suggesting that access to Thr1590 is restricted by the presence of spectrin-like repeats 5 through 9.

We sought a cell system in which we could readily study the phosphorylation of Kalirin by Cdk5. Like PC12 cells, pEAK Rapid cells express Cdk5 (Fig. 2D), but whereas PC12 cells express p35, pEAK Rapid cells express p39. Cdk5 and p35 are expressed at substantially higher levels in the adult rat striatum. To increase the amount of enzyme available, Cdk5 and p35 were coexpressed in pEAK RAPID cells (Xin et al., 2004); antibody to p35 was used to isolate the Cdk5-p35 complex. In agreement with the results obtained using recombinant Cdk5-p25 complex, the TPAK peptide, but not the TAAK peptide, was phosphorylated by immunoprecipitated Cdk5-p35 complex (Fig. 2E, top). Phosphorylation was inhibited when roscovitine, an inhibitor of Cdk1, Cdk2 and Cdk5, was added to the immunoprecipitate (Zhang et al., 2006; Canduri et al., 2004); since immunoprecipitation was carried out using an antibody to p35, which does not interact with Cdk1 or Cdk2 (Dhavan and Tsai, 2001; Lee et al., 1996), we can conclude that roscovitine acts by inhibiting Cdk5 in the complex (Fig. 2E, bottom left). Consistent with this conclusion, when coexpressed with DN Cdk5, phosphorylation of the TPAK peptide by the p35-Cdk5 complex was inhibited (Fig. 2E, bottom right).

We next sought evidence that Kalirin-7 is phosphorylated by Cdk5 in cells. Expression of exogenous Kalirin-7 yielded a 191 kDa band detected by antibody to Myc and by antibody specific for the dipeptide Thr-Pro-P (Fig. 2F); this band was not detected in GFP-transfected control cells. There are three Thr-Pro sequences in Kalirin-7, only one of which is a potential Cdk5 site. Cotransfection of vectors encoding DN Cdk5 and Kalirin-7 yielded similar levels of Myc-tagged Kalirin-7, but the 191 kDa band was no longer detected by the Thr-Pro-P antibody, suggesting that endogenous Cdk5-p39 present in pEAK RAPID cells phosphorylates Thr1590 in Kalirin-7.

To explore the role of Cdk5-mediated phosphorylation of Kalirin-7, we generated vectors encoding Kalirin-7 in which Thr1590 was mutated to Ala1590 or Asp1590; if phosphorylation of Thr1590 plays a critical role in its morphological effects, Kalirin-7 T/A, which cannot be phosphorylated, and Kalirin-7 T/D, which might mimic phosphorylated Kalirin-7, would be expected to affect PC12 cell morphology differently. When expressed in nondifferentiated PC12 cells, Kalirin-7 T/D mimicked the effects of Kalirin-7 whereas Kalirin-7 T/A did not (Fig. 3A). Expression of Kalirin-7 T/A was not without effect; PC12 cells expressing Kalirin-7 T/A were typically surrounded by extensive lamellipodia filled with spokes of filamentous actin; cytoplasmic protrusions were not observed (Fig. 3A, left). In cells expressing Kalirin-7 T/D, broad, flat cytoplasmic protrusions decorated with lamellipodia were prominent (Fig. 3A, right, arrowhead). Filamentous actin was localized near the plasma membrane, especially in lamellipodial spokes, often overlapping sites of Kalirin-7 T/D expression. Nuclear staining for both Kalirin-7 T/A and Kalirin-7 T/D was often apparent, although its significance is not known.

To compare the morphological effects of the Kalirin mutants, cell perimeter and roundness were quantified (Fig. 3B). Kalirin-7 T/D increased cell perimeter as much as Kalirin-7. Kalirin-7 T/A, which caused extensive lamellipodial formation, caused a lesser, but still significant, increase in cell perimeter. Kalirin-7 and Kalirin-7 T/D caused a similar decrease in cell roundness; by contrast, Kalirin-7 T/A did not alter cell roundness. Mutation of the single Cdk5 phosphorylation site in Kalirin-7 alters the ability of Kalirin-7 to affect PC12 cell morphology.

Since coexpression of DN Cdk5 diminished the ability of Kalirin-7 to increase cell perimeter and decrease cell roundness, we determined whether DN Cdk5 altered the response of PC12 cells to Kalirin-7 T/A or Kalirin-7 T/D mutants. Coexpression of DN Cdk5 did not alter the response of PC12 cells to either Kalirin mutant (Fig. 3C,D). The fact that Kalirin-7 T/D alters cell roundness and perimeter in the presence of DN Cdk5 supports the conclusion that Cdk5 functions upstream of Kalirin-7. The fact that DN Cdk5 is without effect on the morphological response of PC12 cells to Kalirin-7 T/D or T/A supports the hypothesis that it is acting by blocking Cdk5-mediated phosphorylation of Thr1590 in Kalirin-7.

We next assessed the role of GEF activity in these changes in cell perimeter and roundness. Kalirin-7 mutated at two sites near the C-terminus of its DH domain (Asn1443-Asp1444 to Ala-Ala; Kalirin-7 ND/AA) has greatly diminished GEF activity (Penzes et al., 2003; Chakrabarti et al., 2005; Schiller et al., 2006). Expression of Kalirin-7 ND/AA in PC12 cells failed to generate any distinguishable morphological change; cytoplasmic protrusions were not extended and lamellipodia were not formed (Fig. 4A). Kalirin-7 ND/AA was localized along the plasma membrane and staining was again apparent in the nucleus. Quantitative analysis indicated that expression of the GEF-inactive mutant of Kalirin-7 had no significant effect on cell perimeter or roundness (data not shown).

Since the GEF activity of Kalirin-7 is required for the induction of cytoplasmic protrusions in nondifferentiated PC12 cells, we evaluated the morphological response of nondifferentiated PC12 cells to expression of constitutively active Rac or RhoG (Fig. 4B). The effects of these two Kalirin substrates were similar; cells exhibited an approximately twofold increase in cell perimeter (not shown). Unlike cells expressing Kalirin-7, cells expressing constitutively active Rac or RhoG produced massive lamellipodial structures filled with spokes of filamentous actin. Cell roundness was decreased slightly in cells expressing constitutively active Rac but was unchanged by expression of constitutively active Rho G (data not shown).

Since KGEF1-7end is an active GEF, includes Thr1590 and is phosphorylated by Cdk5, we next asked whether it produced a morphological response similarly to that of Kalirin-7 or constitutively active Rac and RhoG when expressed in nondifferentiated PC12 cells (Fig. 4C). PC12 cells expressing KGEF1-7end adopted a round shape, with massive lamellipodial structures surrounding the entire cell, as observed in cells expressing constitutively active Rac; protrusions like those extended in response to Kalirin-7 were never observed. KalGEF1 produced a similar response (data not shown). We next asked whether mutation of Thr1590 to Asp or Ala altered the morphological response. Non-differentiated PC12 cells responded to KGEF1-7end T/D and KGEF1-7end T/A in the same manner; protrusions such as those formed in response to Kalirin-7 were not observed. Our data indicate that GEF activity, the spectrin-like repeat region and the Cdk5 site in Kalirin-7 each play essential roles in the morphological response observed.

Since Kalirin is a Cdk5 substrate both in cells and in vitro, we generated an antibody against the phosphorylated TPAK peptide (Thr1590-P Ab) (Fig. 5). The specificity of this antibody for the phosphopeptide was first established by ELISA (Fig. 5A). Plates were coated with equal amounts of phosphorylated TPAK (TPAK-P) or TPAK peptide; only the phospho-peptide yielded a detectable signal. Western blotting was used to further characterize the Thr1590-P antibody. Equal amounts of GST-KGEF1-7end and GST-KGEF1-7end T/A mutant were analyzed directly or after incubation with 2 ng or 10 ng recombinant Cdk5-p25 (Fig. 5B). Coomassie Brilliant Blue staining demonstrated that a similar amount of each protein was present (Fig. 5B, lower panel). The Thr1590-P antibody detected GST-KGEF1-7end only after phosphorylation by Cdk5-p25 and failed to detect the T/A mutant protein even after exposure to Cdk5-p25 (Fig. 5B, upper panel).

To determine the ability of the Thr1590-P antibody to detect phosphorylation of Kalirin in a cellular environment, duplicate wells of pEAK RAPID cells were transfected with vectors encoding Kalirin-7 or its parent vector as control; one well was treated with roscovitine for 4 hours before harvest and the other was not (Fig. 5C). Lysates were visualized with antibody to Myc or to Thr1590-P. Expression of Kalirin-7 was not affected by roscovitine. Under basal conditions, phosphorylated Kalirin-7 was detected; signal intensity was reduced substantially after treating the cells with 10 μM roscovitine (Fig. 5C, upper panel), supporting the hypothesis that phosphorylation of Thr1590 is catalyzed by endogenous Cdk5-p39, Cdk1 or Cdk2 under basal conditions in these cells. After normalizing the intensity of the Thr1590-P signal to the Myc signal, roscovitine treatment reduced phosphorylation of Thr1590 to 30% of that in the control (Fig. 5D).

Since a basal level of Kalirin phosphorylation at Thr1590 was detectable, we assessed the ability of a panel of protein phosphatase inhibitors to stabilize this modification. pEAK RAPID cells transiently expressing Kalirin-7 were incubated with inhibitors of protein phosphatase 1 (PP1), protein phosphatase-2A (PP2A), and protein phosphatase-2B (PP2B or calcineurin) and extracted for analysis of Thr1590 phosphorylation (Fig. 6A). Incubation with calyculin A, an inhibitor of PP1 and PP2A, resulted in a substantial increase in the amount of Kalirin-7 phosphorylated at Thr1590. None of the more selective PP2A inhibitors (cantharidin, endothall) resulted in a similar increase in phosphorylation at this site. Thus PP1, which plays a critical role in synaptic plasticity (Morishita et al., 2001), is likely to play a major role in the dephosphorylation of Kalirin Thr1590. Inhibitors of calcineurin did not affect the level of Thr1590 phosphorylation.

To verify the ability of calyculin A to increase phosphorylation of Kalirin at Thr1590, Kalirin-7, ΔKalirin-7 and KGEF1 were transiently expressed in duplicate wells of pEAK RAPID cells; one set of cells was treated with calyculin A before extraction. Western blot analysis with the Thr1590-P antibody (Fig. 6B, upper panel) and the Myc antibody (Fig. 6B, lower panel) showed that calyculin A treatment of cells expressing Kalirin-7 or ΔKalirin-7 substantially enhanced the signal detected by the Thr1590-P antibody. No signal was detected for KGEF1, which does not include Thr1590. The substantial increase in the signal detected by the Thr1590-P antibody after 30 minutes of calyculin A treatment indicates that phosphorylation-dephosphorylation of Thr1590 occurs rapidly under basal conditions.

As shown previously, levels of Cdk5, p35/p39 and PP1 vary widely in pEAK Rapid cells, PC12 cells and striatum. The ability of roscovitine and calyculin A to affect Thr1590 phosphorylation of exogenous Kalirin-7 in PC12 cells and endogenous Kalirin-7 in primary cultures of neonatal rat striatum was assessed (Fig. 6C). Basal phosphorylation of Thr1590 was not detected when exogenous Kalirin-7 was expressed in PC12 cells; after pretreatment with calyculin A, Thr1590-P could be detected (Fig. 6C, upper). Basal phosphorylation of endogenous Kalirin-7 on Thr1590 was detected in mature cultures of striatal neurons (Fig. 6C, lower). Pretreatment with roscovitine, which should only affect Cdk5 in these nondividing cells (Dhavan and Tsai, 2001), caused a slight decrease in Thr1590 phosphorylation whereas pretreatment with calyculin A caused a substantial increase in Thr1590 phosphorylation. Steady-state phosphorylation of Thr1590 by Cdk5 followed by rapid PP1-mediated dephosphorylation suggests a role for this site in regulating Kalirin function.

Since the GEF activity of Kalirin is critical for its effect on PC12 cell morphology as well as its role in the maintenance of dendritic spines, the dendritic arbor and axonal elongation (Ma et al., 2003; May et al., 2002), we asked whether Thr1590 phosphorylation altered GEF activity. We used both in-cell assays based on the binding of GTP-Rac to the CRIB domain of PAK (May et al., 2002; Schiller et al., 2006) and assays with purified KGEF1-7end to address this issue.

We first utilized the Pak-CRIB pull-down assay to determine whether the ability of cells transiently expressing various Kalirin constructs to activate Rac was altered following pre-incubation with roscovitine or calyculin A (Fig. 7). Cells expressing GFP, KGEF1-7end or KGEF1-7end T/A, were treated with roscovitine or calyculin A and harvested for analysis of active Rac (Fig. 7A). Active Rac was not detectable in extracts of control cells (Fig. 7A, left). Pre-incubation of cells expressing KGEF1-7end with roscovitine reduced levels of active Rac, and preincubation with calyculin A increased levels of active Rac (Fig. 7A, middle). Although expression of KGEF1-7end and Rac was not altered by drug treatment, levels of Thr1590-P were decreased by roscovitine and increased by calyculin A. Expression of KGEF1-7end T/A activated Rac, but neither drug affected the level of Rac activation (Fig. A, right). Since the effects of roscovitine and calyculin A on GEF activity were abrogated when Thr1590 was mutated to Ala, we can conclude that GEF activity is enhanced by Thr1590 phosphorylation. Although it cannot be activated by phosphorylation of Thr1590, KGEF1-7end T/A retains significant GEF activity.

Similar results were obtained in cells expressing Kalirin-7 or ΔKalirin-7 (Fig. 7B,C). Expression of Kalirin and Rac was unaltered by roscovitine or calyculin A treatment. For both Kalirin-7 and ΔKalirin-7, levels of active Rac were increased approximately twofold in the presence of calyculin A. To quantify the effect, Rac activation in cells expressing Kalirin-7 or ΔKalirin-7 was set to 100%. In cells pre-incubated with calyculin A, the same amount of Kalirin-7 or ΔKalirin-7 produced significantly more Rac-GTP (Fig. 7D). Pre-incubation with roscovitine produced a slight decrease in Rac activation by ΔKalirin-7; its effect on Rac activation by Kalirin-7 was not significant. These data suggest that phosphorylation of Thr1590 increases the catalytic activity of the GEF domain in KGEF1-7end, ΔKalirin-7 and wild-type Kalirin-7 or increases their access to Rac in this cell-based assay.

To distinguish between these possibilities, we compared the catalytic activities of purified GST-KGEF1-7end, GST-KGEF1-7end T/A, and GST-KGEF1-7end T/D using GST-Rac1 loaded with GDP-Mant. Representative data from one assay are shown in Fig. 8A; initial rates were calculated from data obtained over the first 10-15 minutes using a semi-logarithmic plot (Fig. 8B). Data from at least three independent assays for each enzyme/substrate combination are summarized in Fig. 8C. Although Thr1590 lies beyond the C-terminus of the PH domain, mutation of this residue to Asp affects GEF activity. With Rac1 as the substrate, KGEF1-7end T/D was significantly more active than KGEF1-7end T/A or KGEF1-7end. It could be that replacing Thr1590 with Asp mimics phosphorylation of Thr1590, making this result consistent with the ability of calyculin A to increase both Thr1590 phosphorylation and Rac activation by Kalirin in cell-based assays. Substitution of Ala for Thr1590 prohibits phosphorylation and an effect on catalytic activity was not anticipated.

Since phosphorylation of Thr1590 produced only a modest increase in GEF activity, we sought evidence for other ways in which phosphorylation at this site might alter Kalirin-7 function. Protein phosphorylation is well recognized as one means of targeting proteins to specific sites (Gulli et al., 2000). Endogenous Kalirin-7 in adult rat cortex is largely insoluble (Penzes et al., 2001). When rat brain cytosol and organelle-enriched fractions were separated from the remaining Triton X-100 (TX-100)-insoluble material, almost all of the Kalirin-7 was recovered from the TX-100 insoluble pellet (Fig. 9A). Because of the many indirect effects of treating cells with roscovitine or calyculin A, we utilized Kalirin-7 T/A and Kalirin-7 T/D to assess solubility in PC12 cells. Although the three proteins were expressed at similar levels, a greater percentage of the Kalirin-7 T/D was recovered from the soluble fraction; Kalirin-7 T/A was less soluble than wild-type Kalirin-7 (Fig. 9B). Access of RhoGEFs to their Rho-GTPase substrates is critical and could be altered by a solubility change of this type.

Thr1590 is present in each of the major isoforms of Kalirin, suggesting regulatory roles for Cdk5 because Kalirin-9 and Kalirin-12 participate in process outgrowth (May et al., 2002) and Kalirin-7 participates in spine formation (Ma et al., 2003; Ma et al., 2008). Since methods for assessing the effects of Kalirin-7 on spine formation have been optimized, we explored the possibility that phosphorylation of Thr1590 affects this response. Rat cortical neurons were nucleofected simultaneously with vectors encoding GFP and Kalirin-7 or its T/A and T/D mutants (Fig. 10). GFP fills each transfected cell, making it possible to visualize dendritic processes and spines (Fig. 10); exogenous Kalirin-7 was visualized using Myc antibody. After 16 days in vitro, the dendritic processes shown in Fig. 10A were covered with spines, most of which had a readily discernible spine head. Myc staining was not detectable in these GFP-expressing neurons and the spine density observed is typical of neurons of this age (Ma et al., 2003; Ma et al., 2008). As expected, spine density was increased substantially in cortical neurons expressing exogenous Kalirin-7 (Fig. 10B). Although dendritic spines were more numerous in neurons expressing Kalirin-7 T/A (Fig. 10C) than in neurons expressing GFP, very few of the spines had a clear spine head. Kalirin-7 T/A is localized to the tips of the spine-like structures that form, but the clusters are noticeably smaller than those formed by Kalirin-7 and the spines are shorter and thinner than normal. Expression of Kalirin-7 T/D also caused an increase in spine number (Fig. 10D); many of the spines formed in response to this protein, which may mimic Kalirin-7 phosphorylated at Thr1590, had large spine heads. Thr1590 clearly plays a critical role in the pathway through which Kalirin-7 affects spine structure. Further studies will be required to determine whether Cdk5 and PP1 are the key factors controlling the phosphorylation state of Kalirin-7 in dendritic spines and how phosphorylation of Thr1590 alters the actions of Kalirin-9 and Kalirin-12.

Cdk5 plays an important role during neuronal development and in mature neurons (Dhavan and Tsai, 2001; Hahn et al., 2005; Humbert et al., 2002). To date, over 25 proteins with diverse functions have been identified as Cdk5 substrates (Hahn et al., 2005; Humbert et al., 2002; Nikolic, 2002; Takashima et al., 2001; Dhavan and Tsai, 2001; Castillo-Liuva et al., 2007; Fu et al., 2007; Kesavapany et al., 2006). Four GEFs, Kalirin, Trio, RasGRF1 and RasGRF2, are phosphorylated by Cdk5 (Kesavapany et al., 2004; Xin et al., 2004; Kesavapany et al., 2006). The Cdk5 site in Kalirin-7 (T1590PAK) lies just beyond the first PH domain; because it precedes the 20 amino acid sequence unique to Kalirin-7, this site also occurs in Kalirin-8, Kalirin-9 and Kalirin-12, the Kalirin isoforms most prevalent early in development. This site is conserved in mammals and chickens and Ser substitutes for Thr in Drosophila Trio; by contrast, the Caenorhabditis elegans and Takifugu rubripes homologues lack a (T/S)PAK motif. Recombinant Cdk5-p25 phosphorylates Thr1590 more efficiently in GST-KGEF1-7end than in ΔKalirin-7. Although access of the kinase complex to Thr1590 may be restricted, exogenous Kalirin-7 expressed in pEAK RAPID cells and endogenous Kalirin-7 in striatal neurons is phosphorylated at this site, based on our antibody staining.

Since expression of exogenous Kalirin-7 disrupted NGF signaling in PC12 cells (Chakrabarti et al., 2005), we used nondifferentiated cells to establish a role for Cdk5 in regulating the effects of Kalirin on the cytoskeleton. Expression of full-length Kalirin-7 in nondifferentiated PC12 cells caused the extension of broad cytoplasmic protrusions. Kalirin-7 in which Thr1590 was mutated to Ala to prevent phosphorylation, failed to cause this response; lamellipodia continued to form, yielding a phenotype similar to that produced by expression of the isolated GEF domain of Kalirin-7 or constitutively active Rac or RhoG. Kalirin-7 in which Thr1590 was mutated to Asp in an attempt to mimic phosphorylation caused the formation of protrusions. Unlike Kalirin-7, the morphological effects of Kalirin-7 T/D were not blocked by DN Cdk5. The ability of Cdk5, which is ubiquitously expressed, to phosphorylate Kalirin-7 may be controlled by activation and/or localization of Cdk5-p35-p39, localization of Kalirin-7 or a conformational change in Kalirin-7 that unmasks the Cdk5 site (Nikolic, 2002; Dhavan and Tsai, 2001).

Although the protrusions formed in undifferentiated PC12 cells are distinctly different from the neurites formed in NGF-differentiated PC12 cells, their formation indicates that the cells have adopted a somewhat polarized morphology. Cdk5 plays an essential role in establishing cell polarity in organisms ranging from yeast to human. For example, Rac1 and laminin form an autocrine pathway that orients the apical pole during cyst development by polarized epithelial cells (O'Brien et al., 2001). Cdk5 plays an essential role in localizing Cdc24 and activated Rac1 in Ustilago maydis cells infected by the corn smut fungus; without active Cdk5, these cells are unable to sustain the dramatic polar growth required for formation of infective structures (Castillo-Liuva et al., 2007). Inhibition of endogenous Cdk5 by DN Cdk5 blocked the ability of Kalirin-7 to cause the formation of protrusions, identifying Cdk5 as an upstream regulator of Kalirin.

Kalirin-7 is highly expressed only in mature neurons (Ma et al., 2003; Ma et al., 2008) and Cdk5 plays an important role in spine formation and synaptic plasticity (Bibb et al., 2001a; Dhavan and Tsai, 2001; Humbert et al., 2002; Lee et al., 1996; Nikolic et al., 1998; Cheung et al., 2006). Primary cortical neurons respond to exogenous Kalirin-7 by producing additional dendritic spines (Ma et al., 2003; Ma et al., 2008) and mutation of Thr1590 to Ala or to Asp does not eliminate the ability of Kalirin-7 to promote spine formation. The fact that exogenous Kalirin-7 T/A causes the appearance of short, thin spine-like structures that lack a clear spine head strongly suggests that Kalirin-7 is one of the Cdk5 substrates that plays an essential role in spine morphogenesis.

Trio, a Kalirin homologue (McPherson et al., 2004), causes neurite outgrowth in PC12 cells; its GEF activity is required for this effect (Estrach et al., 2002). Likewise, the GEF activity of Kalirin-7 is required for its effect on protrusion formation. Expression of constitutively active Rac or RhoG causes the formation of lamellipodia, but does not cause the formation of protrusions or produce polarity. Expression of the isolated GEF1 domain of Kalirin, KalGEF1-7end or ΔKalirin-7 mimics expression of constitutively active Rac; protrusions are not formed. The fact that protrusion formation requires the presence of the spectrin-repeat region of Kalirin-7 suggests a special role for this noncatalytic region. Expression of a fragment of Kalirin-7 that contains only its Sec14p domain followed by its nine spectrin-like repeats completely disrupts normal cytoskeletal organization in non-neuronal cells; the exogenous protein and filamentous actin accumulate under the plasma membrane (Schiller et al., 2008). Both the ability of the Sec14p domain to bind specific phosphoinositides and the ability of the spectrin repeat region to oligomerize are essential for this response. In addition, proteins such as HAP1, DISC1, iNOS and PAM, which are known to interact with the spectrin-repeat region of Kalirin (Alam et al., 1997; Colomer et al., 1997; Ratovitski et al., 1999; Takaki et al., 2005), may play a role in the response. A yeast two-hybrid study identified the N-terminal region of Kalirin as an interactor with the tetramerization domain of αII-spectrin (Oh and Fung, 2007); this interaction has not yet been verified in tissue.

Phosphorylation and dephosphorylation are dynamic processes. A survey of phosphatase inhibitors indicates that protein phosphatase 1 (PP1) is largely responsible for the dephosphorylation of Thr1590-P in pEAK RAPID cells. Phosphorylation of Thr1590 in exogenous Kalirin-7 expressed in PC12 cells and in endogenous Kalirin-7 in striatal neurons is increased following a 30 minute treatment with calyculin A. PP1, with its multiple substrates (Cohen, 2002), regulates the phosphorylation state of many receptors, activates Na⁺ channels (Greengard et al., 1999) and plays an essential role in long-term depression (Morishita et al., 2001). PP1 is found in a complex with another Rho-GEF, Lfc (also known as ARHG2 or Rho/Rac-GEF2). This complex, which regulates Rho-dependent organization of F-actin in spines, involves the interaction of Lfc with neurabin and spinophilin (Ryan et al., 2005). Both neurabin and spinophilin were identified as interactors with the C-terminal PDZ-binding motif of Kalirin-7 (Penzes et al., 2001). Phosphorylation state often regulates protein localization and specific protein-protein interactions (Gulli et al., 2000). The fact that Kalirin-7 T/D is more soluble than Kalirin-7 suggests a similar role in this case.

The first GEF domain of Kalirin has specificity for Rac and RhoG (Penzes et al., 2001; May et al., 2002; Schiller et al., 2005). With a cell-based assay, we demonstrate that treatment with calyculin A, which increases phosphorylation of Thr1590, increases the ability of KalGEF1-7end, Kalirin-7 and ΔKalirin-7 to activate Rac. Calyculin A has no effect on Rac activation in cells expressing GFP or KGEF1-7end T/A, supporting the conclusion that calyculin A acts by stabilizing Thr1590-P. Differences in the level of expression of KGEF1-7end and KGEF1-7end T/A or KGEF1-7end T/D make a direct comparison of their GEF activities using the cell-based assay difficult. Whether the increase in activity upon phosphorylation of Thr1590 is caused by a change in K_m or V_max or a change in subcellular localization requires further study. The magnitude of the calyculin A effect on Thr1590 phosphorylation (five- to tenfold increase in 30 minutes) suggests a high turnover rate for phosphorylation at this site, allowing control by activation of Cdk5 or inhibition of PP1. Cdk5-p35-mediated phosphorylation of Ras guanine nucleotide releasing factor (RasGRF2), a widely expressed Rho-GEF that includes a Ras-exchanger motif, reduces its ability to activate Rac in cell-based assays (Kesavapany et al., 2004). RasGRF1 also interacts with and is phosphorylated by Cdk5 on Ser731. Phosphorylation on this site leads to RasGRF1 degradation through a calpain-dependent mechanism. A reduction of RasGRF1 levels leads to nuclear condensation in neurons (Kesavapany et al., 2006). In agreement with the cell-based assays, in vitro assessment of catalytic activity demonstrated that KGEF1-7end T/D was twice as active as KGEF1-7end T/A using Rac1 as the substrate. In a cellular environment, differences of this magnitude are likely to be of functional significance.

Inserts encoding His₆-Myc-tagged Kalirin-7 and ΔKalirin-7 were constructed in the pEAK vector (Ma et al., 2003; Johnson et al., 2000; Penzes et al., 2000). In pCMS.EGFP.His-myc-KGEF1-7end, His₆-Myc precedes KGEF1 and the protein extends to the C-terminus of Kalirin-7 (Asp1250-V1654). After changing Thr1590 to Ala1590 or Asp1590 (Quikchange; Stratagene, La Jolla, CA), inserts were sequenced in their entirety and subcloned into pCMS.EGFP.His-Myc-KGEF1-7end, pGEX.GST-KGEF1 and pEAK.His₆-Myc-Kalirin-7. ΔKalirin-7 was subcloned into pVL1393 (BD Biosciences) for expression in the Baculovirus system. Cdk5 and p35 were subcloned as described (Xin et al., 2004). A dominant-negative Cdk5 mutant (Lys33 to Thr, DN Cdk5) was generated as described above (Tan et al., 2003).

pGEX.GST-KGEF1, pGEX.GST-KGEF1-7end and its mutants were purified as described (Alam et al., 1997; Penzes et al., 2001; Schiller et al., 2006). The pGEX-GST-Rac1 vector was a gift from Richard Cerione (Cornell University, Ithaca, NY). pVL1393-ΔKalirin-7 was cotransfected with Baculo-Gold linearized virus into Sf9 cells (BD Biosciences). Cells were extracted into TES-mannitol (TM) buffer containing protease inhibitor cocktail (Xin et al., 2004). ΔKalirin-7 bound to Talon Superflow Metal Affinity Resin (BD Biosciences) was eluted with 50 mM NaH₂PO₄, pH 7.0, 300 mM NaCl, 1 M imidazole, 0.05% Triton X-100 containing 1 mM PMSF.

Transfection and extraction was as described (Xin et al., 2004). Western blots used: Myc monoclonal 9E10 (1:10); Cdk5 polyclonal (1:500), p35 polyclonal (1:1000) and PP1 monoclonal (1:1000) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA); Rac monoclonal antibody (1:1000; BD Biosciences); Thr/Pro-P polyclonal antibody 1:1000 (Cell Signaling Technology, Beverly, MA); Kalirin-7 C-terminal specific polyclonal antibody JH2959 1:1000 (Penzes et al., 2000). Antibody to p39 was a gift from Amy Fu (Hong Kong University, Science/Technology, Hong Kong, China). Kalirin Thr1590-P antibody CT188 (Thr1590-P) was generated at Covance (Hazleton, PA). Specificity for Thr1590-P was assessed by solid-phase ELISA; no crossreactivity with nonphosphorylated peptide was observed at >1:300 dilution.

To determine whether phosphorylation of Kalirin by Cdk5 was proline-directed, wild-type peptide was compared with mutant peptide with a Pro to Ala mutation: T1590PAK, P1584IQLPKTPAKLRNNSK; T1590AAK, PIQLPKTAAKLRNNSK. Kinase assay with recombinant Cdk5-p25 (Upstate) was performed based on the company protocol and cell derived immunoprecipitated Cdk5-p35 complex was assayed (Xin et al., 2004).

In vitro Rac activation assays using GST-KGEF1, GST-KGEF1-7end, GST-KGEF1-7end mutants and GST-Rac1 were performed as described (Schiller et al., 2005). Data were analyzed using the formula d/dt(fluorescence)=–k^*([Rac-GDP-MANT]-[EDTA-endpoint]); the initial rate gives the reaction velocity. The extent of GDP-MANT loading was calculated using ϵ₃₅₆=5700 (Invitrogen). Cell-based Rac activation assays were described (May et al., 2002; Xin et al., 2004).

PC12 cells were maintained in DMEM-F12 (Mediatech, Inc., Herndon, VA) containing 10% fetal calf serum (Hyclone, Logan, UT) and 10% NuSerum (Collaborative Research, Waltham, MA) with 25 mM HEPES, penicillin and streptomycin (Xin et al., 2004). Cells (60% confluent) plated onto poly-L-lysine-coated coverslips were transfected and fixed for immunocytochemistry 24 hours later (Xin et al., 2004). Coexpression of EGFP and DN Cdk5 from the dual promoter pCMS.EGFP vector was verified by staining cells with rabbit polyclonal antiserum to Cdk5 (1:250) (Santa Cruz Biotechnology) followed by Alexa Fluor 633-conjugated donkey anti-rabbit IgG (1:500) (Invitrogen) and TRITC-phalloidin; GFP-positive cells always expressed Cdk5. Cotransfection of Kalirin-7 and DN Cdk5 was established using Myc antibody with Alexa Fluor 633-conjugated goat anti-mouse IgG and GFP fluorescence. Images were taken with a Zeiss LSM 510 Meta confocal microscope (CCAM, University of Connecticut Health Center).

Image quantification was performed with SimplePCI Imaging software (Compix, Sewickley, PA). Transfected cells were identified based on coexpression of GFP or staining of epitope-tagged exogenous protein. The transfection rate for PC12 cells was 2-3%; at least 15 fields of cells were analyzed for each vector. Well-separated transfected cells were identified using the SimplePCI thresholding software. Cell perimeter [horizontal pixels + vertical pixels + (1.4142 × diagonal pixels surrounding the object)] and cell roundness [(4 × π × area)/(perimeter)²] were measured based on this threshold; for a circle, roundness=1.0.

Adult rat striatum extracted into SDS lysis buffer (Xin et al., 2004) was fractionated (20 μg total protein) by SDS-PAGE and subjected to Western blot analysis. Striatal cultures were prepared from P1 rat pups after dissociating striata with 0.25% trypsin; 2-3×10⁶ cells in DMEM-F12, 10% FCS, 2 mM Glutamax, 0.5 mg/ml gentamycin were plated per well of a 12-well plate. Medium was changed every 5 days. After 21days in vitro, cells treated with 10 μM roscovitine for 4 hours or 25 nM calyculin A for 30 minutes were extracted into SDS lysis buffer; 10% of each sample was subjected to western blot analysis. Neonatal rat brain cortical cultures were prepared and nucleofected as described (Ma et al., 2003; Ma et al., 2008).

PC12 cells were nucleofected (3 μg plasmid DNA/2-3×10⁶ cells) with the manufacturer's protocol (AMAXA, Germany, Cat. no. VCA-1003). Cells were extracted in 20 mM NaTES, 10 mM mannitol, pH 7.4 containing protease and phosphatase inhibitor cocktails (Xin et al., 2004); centrifugation at 430,000 g for 15 minutes was used to separate soluble from insoluble proteins, which were resuspended using SDS-lysis buffer. Aliquots of the soluble fraction and the resuspended pellet were subjected to western blot analysis using antibody to Myc.

We thank Martin Schiller for sharing data, Jacqueline Sobota for assistance with imaging, Atul Deshpande for initiating phosphatase inhibitor studies and Darlene D'Amato for expert technical assistance. Support: DA-015464.