LIMK1 and LIMK2 (LIMKs, LIM kinases) are kinases that play a crucial role in cytoskeleton dynamics by independently regulating both actin filament and microtubule remodeling. LIMK1 and, more recently, LIMK2 have been shown to be involved in cancer development and metastasis, resistance of cancer cells to microtubule-targeted treatments, neurological diseases, and viral infection. LIMKs have thus recently emerged as new therapeutic targets. Databanks describe three isoforms of human LIMK2: LIMK2a, LIMK2b, and LIMK2-1. Evidence suggests that they may not have completely overlapping functions. We biochemically characterized the three isoforms to better delineate their potential roles, focusing on LIMK2-1, which has only been described at the mRNA level in a single study. LIMK2-1 has a protein phosphatase 1 (PP1) inhibitory domain at its C-terminus which its two counterparts do not. We showed that the LIMK2-1 protein is indeed synthesized. LIMK2-1 does not phosphorylate cofilin, the canonical substrate of LIMKs, although it has kinase activity and promotes actin stress fiber formation. Instead, it interacts with PP1 and partially inhibits its activity towards cofilin. Our data suggest that LIMK2-1 regulates actin cytoskeleton dynamics by preventing PP1-mediated cofilin dephosphorylation, rather than by directly phosphorylating cofilin as its two counterparts, LIMK2a and LIMK2b. This specificity may allow for tight regulation of the phospho-cofilin pool, determining the fate of the cell.

Introduction

LIM kinases (LIMKs) are serine/threonine and tyrosine kinases involved in regulating cytoskeleton dynamics. They phosphorylate cofilin, resulting in its inhibition [1]. Cofilin is a member of the actin-depolymerizing factor (ADF) family and regulates actin polymerization dynamics by promoting rapid turnover of actin filaments [24]. LIMKs also control microtubule dynamics, independently from their regulation of actin remodeling. The molecular mechanism by which LIMKs regulate microtubule dynamics is still unknown [5]. LIMKs thus play a crucial role in cytoskeleton remodeling, contributing to many cellular functions, such as cell motility, morphogenesis, division, differentiation, apoptosis, neuronal morphology, neuritogenesis, and oncogenesis.

The LIMK family consists of only two members: LIMK1 and LIMK2. They share 50% overall identity and 70% identity in their kinase domain. They have a unique sequence organization consisting of two LIM (Lin11, Isl-1, and Mec-3) domains at their N-terminus, a central PDZ domain, and a C-terminal kinase domain.

Three isoforms of human LIMK2 are described in databanks: LIMK2a, LIMK2b, and LIMK2-1. These three isoforms differ only at their extremities. The first LIM motif is truncated at the N-terminus of LIMK2b and LIMK2-1, whereas LIMK2-1 has an extra C-terminal domain identified as a protein phosphatase 1 inhibitory (PP1i) domain by sequence homology (Figure 1). Only one paper has shown the existence of LIMK2-1 mRNA [6]. Very little is known about the functional differences between these three isoforms and most studies did not indicate which isoform was under investigation. However, some data suggest that they probably do not have completely overlapping functions. The tissue distribution of LIMK2a and LIMK2b is different and they are differentially expressed according to the developmental stage [7,8]. They also show different subcellular localization: LIMK2a is found in the cytoplasm and the nucleus, whereas LIMK2b is mainly localized to the cytoplasm [8]. LIMK2a is an extremely stable protein with a half-life of ∼24 h, whereas LIMK2b has a much shorter half-life of ∼6 h [9]. LIMK2b and LIMK2-1, but not LIMK2a, are targets of p53. Upon DNA damage, p53 induces up-regulation of LIMK2b and LIMK2-1 mRNA expression, thus modulating G2/M arrest [6,10]. LIMK2 has been described as a cancer cell-survival factor as p53 plays a major role in cancer development [11]. These data further connect LIMK2 to cancer. Indeed, increasing evidence shows a role for LIMK2 in cancer development and metastasis formation. LIMK2 is deregulated in many cancers (such as pancreatic and breast cancers) [1214], and LIMK2a and 2b deregulation differ depending on cancer type [7,8]. Furthermore, LIMK2 also interacts with Aurora A, a kinase overexpressed in many cancers, and believed to promote tumorigenesis. LIMK2 and Aurora A regulate each other in a positive feedback loop in which LIMK2 appears to be a key oncogenic effector of Aurora A [15]. Recently, LIMK2a and LIMK2b were also shown to play a role in microtubule organization, but the molecular mechanisms involved are not yet known [5,9,16,17]. Gamell et al. [9] showed that LIMK2a and LIMK2b differentially regulate G2/M cell cycle arrest induced by microtubule-targeted drugs. Moreover, LIMK2 expression is elevated in neuroblastoma cells resistant to microtubule-targeted drugs. Thus, LIMK2 may be a possible target to overcome resistance to microtubule-targeted drugs [9,16,17]. We and others have shown that LIMK2 may also be involved in neurofibromatosis type I, a common genetic disease, of which the main manifestations are cancer development and cognitive defects [18,19]. LIMK2 is also involved in neurodevelopmental disorders [20]. Indeed, LIMK2 plays a critical role in growth cone establishment, neurite guidance and outgrowth, synapse plasticity, and neuronal death [21]. More recently, a possible role for LIMK2 in schizophrenia was suggested by gene expression studies in an animal model [22].

Evidence for the existence of the LIMK2-1 protein.

Figure 1.
Evidence for the existence of the LIMK2-1 protein.

(A) Sequence alignment and motif representation of the three isoforms of human LIMK2. LIMK2 isoforms are described in Entrez Gene: LIMK2-1 (NP_001026971.1), LIMK2a (NP_005560.1), and LIMK2b (NP_057952.1). The various domains of LIMK2 are shown: LIM, PDZ, SP (serine proline-rich), KIN (kinase), and PP1i (protein phosphatase 1 inhibitory) domains. * indicates the activatable threonine that modulates LIMK2 activity towards cofilin. The PP1-binding consensus sequence is underlined in purple. The sequence chosen for anti-PP1i antibody design is underlined in orange. (B,C) Validation of the anti-LIMK2-1 antibody. (B) HEK-293 cells were transfected with untagged LIMK2-1 (pCMV-LIMK2-1) and HA-tagged isoforms of LIMK2. Lysates were analyzed by western blotting using the indicated antibodies. (C) HEK-293 were transfected with LIMK2 siRNA or control siRNA. Lysates were analyzed by western blot. LIMK2-1 is expressed in various human cell lines (D) and tissues (E). HEK-293 and HeLa cells were disrupted in 1% Triton X-100 lysis buffer. Tissue extracts were directly purchased. Samples were analyzed by western blotting.

Figure 1.
Evidence for the existence of the LIMK2-1 protein.

(A) Sequence alignment and motif representation of the three isoforms of human LIMK2. LIMK2 isoforms are described in Entrez Gene: LIMK2-1 (NP_001026971.1), LIMK2a (NP_005560.1), and LIMK2b (NP_057952.1). The various domains of LIMK2 are shown: LIM, PDZ, SP (serine proline-rich), KIN (kinase), and PP1i (protein phosphatase 1 inhibitory) domains. * indicates the activatable threonine that modulates LIMK2 activity towards cofilin. The PP1-binding consensus sequence is underlined in purple. The sequence chosen for anti-PP1i antibody design is underlined in orange. (B,C) Validation of the anti-LIMK2-1 antibody. (B) HEK-293 cells were transfected with untagged LIMK2-1 (pCMV-LIMK2-1) and HA-tagged isoforms of LIMK2. Lysates were analyzed by western blotting using the indicated antibodies. (C) HEK-293 were transfected with LIMK2 siRNA or control siRNA. Lysates were analyzed by western blot. LIMK2-1 is expressed in various human cell lines (D) and tissues (E). HEK-293 and HeLa cells were disrupted in 1% Triton X-100 lysis buffer. Tissue extracts were directly purchased. Samples were analyzed by western blotting.

These findings all suggest LIMK2 as an emerging therapeutic target to treat cancer, neurofibromatosis type I, and neuronal disorders, and to overcome resistance to chemotherapy treatment with microtubule-targeting drugs. Nonetheless, no exhaustive studies of the three LIMK2 isoforms have yet been performed.

We biochemically characterized the three LIMK2 isoforms, particularly LIMK2-1, to delineate their specific properties. We showed that LIMK2-1 can be found in various cell lines and tissues. All three LIMK2 isoforms interact with each other. They show different subcellular localization with LIMK2-1 being mostly cytoplasmic. They are all phosphorylated by the upstream kinase ROCK on the catalytic regulatory threonine, Thr505/484. However, LIMK2-1 does not phosphorylate cofilin, the canonical substrate of LIM kinases, even though it exhibits kinase activity towards myelin basic protein (MBP) and promotes stress fiber formation. LIMK2-1 interacts with PP1 and partially inhibits it, resulting in an increase in the level of phospho-cofilin. Our data suggest that the three LIMK2 isoforms regulate actin polymerization dynamics by increasing the global pool of phospho-cofilin, either by direct phosphorylation by LIMK2a and LIMK2b or by preventing its dephosphorylation through the action of LIMK2-1. LIMK2 isoforms may provide fine tuning of the balance between cofilin and phospho-cofilin, which may be crucial, depending on the biological context.

Materials and methods

Materials

Antibodies anti-LIMK2 (sc-8390) and anti-GFP (sc-9996) were from Santa Cruz Biotechnology, Inc., anti-cMyc (MA1-21316) from Invitrogen, anti-HA (11687423001) from Roche Applied Science, and anti-phospho-LIMK2 (3841), cofilin (3312, 5175), and phospho-cofilin (3313) from Cell Signaling Technology. Antibody against a 12 amino acid peptide (DKIRAMQKLSTP) belonging to the PP1i domain of LIMK2-1 was developed by Eurogentec. EZview™ Red anti-HA affinity gel (E6779) and anti-Flag®-M2 affinity gel (A2220) were from Sigma–Aldrich Co., as well as antibodies anti-actin (A1978) and anti-Flag (M2 antibody F3165). GFP-trap beads were from Chromotek. Lipofectamine 2000 and Lipofectamine RNAiMAX were from Invitrogen, Opti-MEM from Gibco. Recombinant GST-fused cofilin and MBP were purchased from Upstate Cell Signaling, Inc. The different tissue extracts were from Biochain. Y27632 was from Tocris Bioscience. Plasmids used in this study are listed Table 1.

Table 1
Plasmids used in the present study
Plasmid Description Source/reference 
pcDNA3-(HA)2-LIMK2-1 PCMV-(HA)2-LIMK2-1 The present study 
pcDNA3-(HA)2-LIMK2-1_ΔPP1i PCMV-(HA)2-LIMK2-1_ΔPP1i The present study 
pcDNA3-(HA)2-LIMK2-1_D430N PCMV-(HA)2-LIMK2-1_D430N The present study 
pcDNA3-(HA)2-LIMK2a PCMV-(HA)2-LIMK2a [19
pcDNA3-(HA)2-LIMK2b PCMV-(HA)2-LIMK2b The present study 
pcDNA3-(HA)2-KIN2 PCMV-(HA)2-KIN2 [19
pcDNA3-(HA)2-KIN1-PP1i PCMV-(HA)2-KIN1-PP1i The present study 
pcDNA3-(HA)2-PP1i PCMV-(HA)2-PP1i(AA581-686) The present study 
pCMV-LIMK2-1 PCMV-sport6-LIMK2-1 Open Biosystem 
pcDNA3-(HA)2-Larp6 PCMV-(HA)2-LARP6 The present study 
pXJN-myc-LIMK2b pXJN-myc-LIMK2b [10
pE-YFP-LIMK2-1 pE-YFP-N1-LIMK2-1 The present study 
pE-YFP-LIMK2a pE-YFP-N1-LIMK2a The present study 
pE-YFP-LIMK2b pE-YFP-N1-LIMK2b The present study 
pcDNA3-(HA)2-LIMK2-T505A PCMV-(HA)2-LIMK2-2a-T505A The present study 
pCAG ROCK1 ROCK1-myc [47
p3x-Flag-PP1 PCMV-(Flag)-PP1cα The present study 
Plasmid Description Source/reference 
pcDNA3-(HA)2-LIMK2-1 PCMV-(HA)2-LIMK2-1 The present study 
pcDNA3-(HA)2-LIMK2-1_ΔPP1i PCMV-(HA)2-LIMK2-1_ΔPP1i The present study 
pcDNA3-(HA)2-LIMK2-1_D430N PCMV-(HA)2-LIMK2-1_D430N The present study 
pcDNA3-(HA)2-LIMK2a PCMV-(HA)2-LIMK2a [19
pcDNA3-(HA)2-LIMK2b PCMV-(HA)2-LIMK2b The present study 
pcDNA3-(HA)2-KIN2 PCMV-(HA)2-KIN2 [19
pcDNA3-(HA)2-KIN1-PP1i PCMV-(HA)2-KIN1-PP1i The present study 
pcDNA3-(HA)2-PP1i PCMV-(HA)2-PP1i(AA581-686) The present study 
pCMV-LIMK2-1 PCMV-sport6-LIMK2-1 Open Biosystem 
pcDNA3-(HA)2-Larp6 PCMV-(HA)2-LARP6 The present study 
pXJN-myc-LIMK2b pXJN-myc-LIMK2b [10
pE-YFP-LIMK2-1 pE-YFP-N1-LIMK2-1 The present study 
pE-YFP-LIMK2a pE-YFP-N1-LIMK2a The present study 
pE-YFP-LIMK2b pE-YFP-N1-LIMK2b The present study 
pcDNA3-(HA)2-LIMK2-T505A PCMV-(HA)2-LIMK2-2a-T505A The present study 
pCAG ROCK1 ROCK1-myc [47
p3x-Flag-PP1 PCMV-(Flag)-PP1cα The present study 

Cell culture and transfection

HEK-293 and HeLa cells were cultured under 5% CO2 at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. HEK-293 cells were transiently transfected with 10 µg of plasmid/100-mm dish with calcium phosphate method, and HeLa cells with Lipofectamine 2000 according to manufacturer's recommendations. Further experiments were conducted 48 h after transfection. For Y27632 experiments, cells were treated for 30 min with 10 µl of a 10 mM stock solution of Y27632 in ethanol (10 µM final concentration) and then lysed.

Cell lysates for endogenous LIMK2-1 detection

HEK-293 and HeLa cells were lysed in 1% Triton X-100 lysis buffer (50 mM Tris–HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA, 50 mM NaF, 10 mM sodium pyrophosphate, 1 mM Na3VO4, 20 mM p-nitrophenyl phosphate, 20 mM β-glycerophosphate, 10 µg/ml aprotinin, 0.05 µg/ml okaidic acid, 1 µg/ml leupeptin, and 1 mM PMSF).

LIMK2-1 RNAi transfection

LIMK2 siRNA (#s8191) and control siRNA (#AM4611) were from Ambion®. HEK-293 cells were transfected with 2 nM siRNA with Lipofectamine RNAiMAX for 48 h, using manufacturer's instructions.

Cell fractionation

Nuclear and cytoplasmic fractions were prepared as described by Smolich et al. [23]. Briefly, HEK-293 cells were transfected or not with expression plasmids as described above and cultured for 48 h. They were scraped off the dish in PBS and resuspended in 10 mM Tris–HCl (pH 7.6), 10 mM NaCl, 3 mM MgCl2, 0.5% Triton X-100 and a cocktail of protease inhibitors. Cells were homogenized with 20 strokes in a Dounce homogenizer and the nuclei pelleted by centrifugation in a microfuge at 2000×g for 10 min. Supernatant was collected as the cytoplasmic fraction. The nuclear pellet was washed twice in the previous buffer, and the final pellet was solubilized in 10 mM Tris–HCl (pH 7.0), 150 mM NaCl, 1% NP40, 1% Na-deoxycholate, 0.1% SDS, and a cocktail of protease inhibitors.

Immunoprecipitation

HEK-293 cells were transfected with expression plasmids as described above and cultured for 48 h. Cells were lysed in 0.5 ml of 0.1% Triton X-100 lysis buffer (50 mM Tris–HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA, 50 mM NaF, 10 mM sodium pyrophosphate, 1 mM Na3VO4, 20 mM p-nitrophenyl phosphate, 20 mM β-glycerophosphate, 10 µg/ml aprotinin, 0.05 µg/ml okaidic acid, 1 µg/ml leupeptin, and 1 mM PMSF) and incubated on ice for 10 min. After centrifugation, the supernatants were incubated for 2–3 h at 4°C either with anti-HA affinity gel for HA-LIMK2s or with GFP-trap beads for YFP-LIMK2s. Beads were washed five times with lysis buffer and then eluted with Laemmli sample buffer.

Kinase assay

Immunoprecipitates bound to HA-beads or GFP-beads, as described above, were washed twice with lysis buffer and then three times with kinase buffer (50 mM HEPES-NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl2, 5 mM MnCl2, 50 mM NaF, 1 mM Na3VO4, 20 mM β-glycerophosphate, 1 µg/ml leupeptin, and 1 mM PMSF). Immunoprecipitates were incubated for 20 min at 30°C in 22 µl of kinase buffer containing 50 µM ATP, 5 µCi of γ[32P]ATP (3000 Ci/mmol), and 2.5 µg of GST-fused cofilin or 10 µg of MBP. The reaction was terminated by heating 5 min at 90°C in Laemmli sample buffer. Samples were then subjected to SDS–PAGE and analyzed by autoradiography.

Cell staining

HeLa cells were fixed with 4% paraformaldelyde in PBS for 20 min and permeabilized with 0.5% Triton-X100 in PBS for 15 min at room temperature. After blocking with 1% fetal calf serum in PBS for 30 min, these cells were incubated with anti-HA antibodies for 1 h and subsequently with FITC-conjugated anti-rat IgG, and simultaneously with AlexaFluor568-conjugated phalloidin for 1 h. The cells were then washed three times with PBS, mounted on glass slides, and then analyzed by confocal microscopy using a Zeiss Axiovert 200 M microscope coupled with a Zeiss LSM 510 scanning device (Carl Zeiss Co. Ltd., Iena, Germany).

Statistics

Statistical significance was determined using one-way ANOVA (***P < 0.001, **P < 0.01, *P < 0.05).

Results

The LIMK2-1 protein is synthesized

LIMK2-1 is mentioned in databanks, but only at the mRNA level by one publication [6]. LIMK2-1 has a PP1i domain at its C-terminus, which LIMK2a and LIMK2b do not (Figure 1A). We first designed an antibody that targets this domain, amino acids 671–684 (Figure 1A, underlined in orange). A blast search of this 12 amino acid sequence against human protein databases showed only one strong similarity within the sequence of the protein phosphatase PHI-1. PHI-1 migrates at 23 kDa on SDS–PAGE gels [24], and thus, it should not interfere with the detection of endogenous LIMK2-1, expected to migrate at ∼75 kDa. We validated this antibody firstly on HEK-293 cells transfected with LIMK2-1, LIMK2a, or LIMK2b. The anti-PP1i antibody recognized transfected LIMK2-1 (pCMV-LIMK2-1) and HA-tagged LIMK2-1, but did not cross-react with transfected HA-tagged LIMK2a or LIMK2b (Figure 1B). We observed a band of endogenous LIMK2-1 in HEK cells transfected with HA-tagged versions of the LIMK2 isoforms (indicated by an arrow in the anti-PP1i blot; Figure 1B). Secondly, we checked if the signal induced by our anti-PP1i antibody was specific to LIMK2-1 using siRNA targeting the three spliced variants of LIMK2 (Figure 1C). In the presence of LIMK2 siRNA, we could observe a reproducible and significant decrease in our band of interest (indicated by an arrow) compared with control conditions, suggesting that our antibody is specific to LIMK2-1. We then used the anti-PP1i antibody to detect LIMK2-1 in two cell line extracts from HEK-293 and HeLa cells and various human tissues (Figure 1D,E). LIMK2-1 appeared to be expressed in HEK-293 and HeLa cell lines (Figure 1D). LIMK2-1 protein levels varied depending on the tissue: we found the highest levels in liver, somewhat less in pancreas, and the lowest in testis and lung. LIMK2-1 was barely detectable in brain (Figure 1E). However, we observed lower molecular mass bands in all tissue samples except liver, suggesting degradation of the full protein probably due to the lysis conditions of these commercial samples (this lysis buffer contains a cocktail of inhibitors not specified on the data sheet, which may be less efficient than our many protease and phosphatase inhibitors, we used in our home-made buffer). These data show that human LIMK2-1 protein is synthesized and differently expressed in the tested tissues.

Subcellular localization

Previous studies have shown that the subcellular localization of LIMK2a and LIMK2b is different. Both are found in the cytoplasm and the nucleus, but LIMK2b to a lesser extent in the nucleus [8,23].

We first assessed the subcellular localization of each LIMK2 isoform in HeLa cells transfected with a HA-tagged version of each by immunofluorescence. We found both LIMK2a and LIMK2b in the cytoplasm and the nucleus, whereas LIMK2-1 was exclusively localized to the cytoplasm (Figure 2A).

Subcellular localization of LIMK2s.

Figure 2.
Subcellular localization of LIMK2s.

(A) Immunofluorescence. HeLa cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b). Cells were fixed and stained with anti-Draq5 (to label the nucleus) and anti-HA antibodies (to detect the LIMK2 isoforms). Scale bar represents 20 µm. (B) Cell fractionation on transfected cells. Upper panel: HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b). Cells were then fractionated as described in Materials and Methods. Lower panel: Quantification of three independent experiments. The values obtained for nuclear and cytosolic fractions were first normalized with U1 and RKIP, respectively, and then summed and normalized to 100%. (C) HEK cell fractionation. Upper panel: HEK-293 cells were fractionated as above. Lower panel: Quantification of three independent experiments. The values obtained for nuclear and cytosolic fractions were first normalized with U1 and RKIP, respectively, and then summed and normalized to 100%.

Figure 2.
Subcellular localization of LIMK2s.

(A) Immunofluorescence. HeLa cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b). Cells were fixed and stained with anti-Draq5 (to label the nucleus) and anti-HA antibodies (to detect the LIMK2 isoforms). Scale bar represents 20 µm. (B) Cell fractionation on transfected cells. Upper panel: HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b). Cells were then fractionated as described in Materials and Methods. Lower panel: Quantification of three independent experiments. The values obtained for nuclear and cytosolic fractions were first normalized with U1 and RKIP, respectively, and then summed and normalized to 100%. (C) HEK cell fractionation. Upper panel: HEK-293 cells were fractionated as above. Lower panel: Quantification of three independent experiments. The values obtained for nuclear and cytosolic fractions were first normalized with U1 and RKIP, respectively, and then summed and normalized to 100%.

We further characterized the subcellular localization of LIMK by cell fractionation of HEK cells transfected with each of the HA-tagged LIMK2 isoforms. The results were consistent with those obtained by immunofluorescence. We found LIMK2a and LIMK2b in both the nuclear and cytoplasmic fractions, with less LIMK2b found in the nuclear fraction, whereas LIMK2-1 was mostly found in the cytoplasmic fraction. Quantification of three independent experiments showed that a mean of 3.4% of LIMK2-1 was localized to the nuclear fraction versus 28.4% for LIMK2a and 24.6% for LIMK2b (Figure 2B).

We then checked endogenous LIMK2-1 subcellular localization by repeating this subcellular fractionation on non-transfected HEK cells. LIMK2-1 distribution was analyzed using anti-PP1i antibody (Figure 2C). Endogenous LIMK2-1 appeared mainly localized in the cytoplasmic fraction (95.2% versus 4.8% in the nuclear fraction).

LIMK oligomers

LIMK1 and LIMK2 can form heterodimers [25,26]. We determined whether the various isoforms of LIMK2 could form homo and/or heterodimers. We performed co-transfection experiments on HEK-293 cells using vectors encoding the three HA-tagged isoforms of LIMK2 and a non-tagged LIMK2-1, cMyc-tagged LIMK2b, or YFP-tagged LIMK2a isoform. Cell extracts were then immunoprecipitated using anti-HA antibodies and the co-immunoprecipitated LIMK2 isoforms analyzed by western blotting using anti-LIMK2, anti-cMyc, and anti-GFP antibodies. Each isoform interacted with itself and with its two other counterparts (Figure 3).

LIMK2 oligomers.

Figure 3.
LIMK2 oligomers.

(A) LIMK2-1 interacts with itself and with LIMK2 isoforms 2a and 2b. HEK-293 cells were co-transfected with LIMK2-1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the empty parental vector pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. * indicates non-tagged LIMK2-1. (B) LIMK2b interacts with itself and with LIMK2a. HEK-293 cells were co-transfected with cMyc-tagged LIMK2b and HA-tagged LIMK2a or 2b or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (C) LIMK2a interacts with itself. HEK-293 cells were co-transfected with YFP-tagged LIMK2a and HA-tagged LIMK2a or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (D) The three isoforms interact together. HEK-293 cells were co-transfected with HA-tagged LIMK2-1 or the empty parental vector, pcDNA3, YFP-tagged LIMK2a, and cMyc-tagged LIMK2b. Lysates and anti-HA immunoprecipitates were subjected to western blotting.

Figure 3.
LIMK2 oligomers.

(A) LIMK2-1 interacts with itself and with LIMK2 isoforms 2a and 2b. HEK-293 cells were co-transfected with LIMK2-1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the empty parental vector pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. * indicates non-tagged LIMK2-1. (B) LIMK2b interacts with itself and with LIMK2a. HEK-293 cells were co-transfected with cMyc-tagged LIMK2b and HA-tagged LIMK2a or 2b or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (C) LIMK2a interacts with itself. HEK-293 cells were co-transfected with YFP-tagged LIMK2a and HA-tagged LIMK2a or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (D) The three isoforms interact together. HEK-293 cells were co-transfected with HA-tagged LIMK2-1 or the empty parental vector, pcDNA3, YFP-tagged LIMK2a, and cMyc-tagged LIMK2b. Lysates and anti-HA immunoprecipitates were subjected to western blotting.

Regulation by ROCK

LIMK2 belongs to the Rho/ROCK/LIMK2/cofilin signaling pathway. ROCK activates LIMK2a by direct phosphorylation of Thr505 [27,28]. We first assessed whether the three LIMK2 isoforms can interact with ROCK by co-immunoprecipitation. HEK cells were co-transfected with cMyc-tagged ROCK1 and one of the HA-tagged LIMK2 isoforms, followed by anti-HA immunoprecipitation. ROCK1 specifically interacted with each isoform of LIMK2 (Figure 4A). We then determined whether ROCK1 can activate each of the three isoforms using an antibody that specifically targets phospho-Thr505 of LIMK2a. HEK cells were transfected by one of the LIMK2 isoforms or the non-phosphorylatable mutant LIMK2a-T505A. Lysates were analyzed with the anti-phospho-Thr505 LIMK2 antibody. This antibody exhibits background signal, as we observed a band for the T505A mutant (Figure 4B). This background signal is probably due to the fact that this antibody was produced by immunization with a synthetic phospho-peptide corresponding to residues surrounding Thr505; these residues may be also slightly recognized. The three isoforms of LIMK2 exhibited basal phosphorylation, which was reduced by treating cells with the ROCK inhibitor Y27632 (Figure 4B, left panels). This signal of phosphorylation increased drastically when cells were co-transfected with ROCK1. These data suggest that ROCK1 phosphorylated the three LIMK2 isoforms on Thr505 for LIMK2a and on the corresponding Thr484 for LIMK2b and LIMK2-1 (Figure 4B).

The three isoforms of LIMK2 interact with ROCK1, which specifically phosphorylates them on Thr505/484.

Figure 4.
The three isoforms of LIMK2 interact with ROCK1, which specifically phosphorylates them on Thr505/484.

(A) ROCK1 interacts with the three isoforms of LIMK2. HEK-293 cells were co-transfected with cMyc-tagged ROCK1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (B) ROCK1 activates the three isoforms of LIMK2. Left panels: HEK-293 were transfected with one of the three HA-tagged LIMK2 isoforms 2-1, 2a, 2b, or the inactivatable mutant, LIMK2a-T505A. Cells were treated for 30 min with 10 µl of a 10 mM stock solution of Y27632 in ethanol (10 µM final concentration), or with 10 µl of ethanol (negative control). Cells were then lysed, and lysates were analyzed by western blotting. Right panels: HEK-293 cells were co-transfected with cMyc-tagged ROCK1 or empty parental plasmid, pCAG, and one of the three HA-tagged LIMK2 isoforms 2-1, 2a, 2b, or the inactivatable mutant, LIMK2a-T505A. Lysates were subjected to western blotting.

Figure 4.
The three isoforms of LIMK2 interact with ROCK1, which specifically phosphorylates them on Thr505/484.

(A) ROCK1 interacts with the three isoforms of LIMK2. HEK-293 cells were co-transfected with cMyc-tagged ROCK1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the empty parental vector, pcDNA3. Lysates and anti-HA immunoprecipitates were subjected to western blotting. (B) ROCK1 activates the three isoforms of LIMK2. Left panels: HEK-293 were transfected with one of the three HA-tagged LIMK2 isoforms 2-1, 2a, 2b, or the inactivatable mutant, LIMK2a-T505A. Cells were treated for 30 min with 10 µl of a 10 mM stock solution of Y27632 in ethanol (10 µM final concentration), or with 10 µl of ethanol (negative control). Cells were then lysed, and lysates were analyzed by western blotting. Right panels: HEK-293 cells were co-transfected with cMyc-tagged ROCK1 or empty parental plasmid, pCAG, and one of the three HA-tagged LIMK2 isoforms 2-1, 2a, 2b, or the inactivatable mutant, LIMK2a-T505A. Lysates were subjected to western blotting.

Kinase activity

Stress fiber formation

We next focused on LIMK2 activity. LIMK2 phosphorylates cofilin on serine 3, resulting in the inhibition of this actin depolymerization factor. Cellular overexpression of LIMK2 inhibits cofilin-mediated actin depolymerization, leading to the accumulation of stress fibers [19,28,29]. We tested whether this is true for each of the three LIMK2 isoforms by immunofluorescence experiments on fixed intact cells. We transfected HeLa cells with one of the HA-tagged LIMK2 isoforms and visualized the actin filaments by phalloidin staining. Expression of each LIMK2 isoform resulted in the formation of more actin stress fibers than in cells transfected with an unrelated control construct (Larp6) (Figure 5A). LIMK2-1-induced stress fibers appeared to be slightly different from the one induced by LIMK2a or 2b: they are thinner and not present all over the z-plans of the cell (Supplementary Figure S1). Stress fibers were quantified by categorizing them into two batches: (i) thick and numerous stress fibers and (ii) thin or no stress fibers (Figure 5A, right panel). LIMK2-1-transfected cells exhibited less thick and numerous stress fibers (batch (i)) compared with LIMK2a- or 2b-transfected cells; nevertheless, they exhibited significant stress fibers compared with the negative control (Larp6) (Figure 5A, right panel). These data suggest that LIMK2-1, as well as LIMK2a and LIMK2b, plays a role in actin cytoskeleton organization by promoting stress fiber formation.

LIMK2-1 remodels the actin cytoskeleton, but has no kinase activity towards cofilin, although it phosphorylates MBP.
Figure 5.
LIMK2-1 remodels the actin cytoskeleton, but has no kinase activity towards cofilin, although it phosphorylates MBP.

(A) Stress fibers induced by LIMK2s. HeLa cells were transfected with one of the HA-tagged LIMK2 isoform. Cells were fixed and stained with phalloidin (main picture), and anti-HA antibodies (bottom left corner picture). The scale bar represents 50 µm. Right panel: quantification of observed stress fibers by categorizing them into two batches. Four independent experiments were performed, each time 70–90 transfected cells were counted, (B) HEK cells transfected with LIMK2a or LIMK2b exhibit a higher level of endogenous phospho-cofilin than HEK cells transfected with LIMK2-1. HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (1, 2a, and 2b) or the empty parental vector, pcDNA3 (left panel), or with one of the three YFP-tagged LIMK2 isoforms or YFP alone (right panel). Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells was normalized to 100. Each value represents the mean ± SE of three independent experiments. (C) LIMK2-1 does not phosphorylate cofilin in vitro. HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or an unrelated HA-tagged protein, Larp6, as a negative control. Anti-HA immunoprecipitated proteins and GST-cofilin were used in the kinase assay. The anti-HA immunoprecipitates were also subjected to anti-HA immunoblotting and Coomassie blue staining. (D) The three LIMK2 isoforms have kinase activity towards MBP. HEK-293 cells were transfected with one of the three YFP-tagged LIMK2 isoforms (2-1, 2a, and 2b) or YFP alone. Anti-GFP-immunoprecipitated LIMK2 isoforms and cofilin or MBP were used in the kinase assay. The anti-GFP immunoprecipitates were also subjected to anti-GFP immunoblotting and Coomassie blue staining. Quantification of phospho-cofilin and phospho-MBP is shown in the bottom graph. Phospho-cofilin levels obtained with anti-GFP-immunoprecipitated LIMK2a were normalized to 100. Each value represents the mean ± SE of three independent experiments.

Figure 5.
LIMK2-1 remodels the actin cytoskeleton, but has no kinase activity towards cofilin, although it phosphorylates MBP.

(A) Stress fibers induced by LIMK2s. HeLa cells were transfected with one of the HA-tagged LIMK2 isoform. Cells were fixed and stained with phalloidin (main picture), and anti-HA antibodies (bottom left corner picture). The scale bar represents 50 µm. Right panel: quantification of observed stress fibers by categorizing them into two batches. Four independent experiments were performed, each time 70–90 transfected cells were counted, (B) HEK cells transfected with LIMK2a or LIMK2b exhibit a higher level of endogenous phospho-cofilin than HEK cells transfected with LIMK2-1. HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (1, 2a, and 2b) or the empty parental vector, pcDNA3 (left panel), or with one of the three YFP-tagged LIMK2 isoforms or YFP alone (right panel). Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells was normalized to 100. Each value represents the mean ± SE of three independent experiments. (C) LIMK2-1 does not phosphorylate cofilin in vitro. HEK-293 cells were transfected with one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or an unrelated HA-tagged protein, Larp6, as a negative control. Anti-HA immunoprecipitated proteins and GST-cofilin were used in the kinase assay. The anti-HA immunoprecipitates were also subjected to anti-HA immunoblotting and Coomassie blue staining. (D) The three LIMK2 isoforms have kinase activity towards MBP. HEK-293 cells were transfected with one of the three YFP-tagged LIMK2 isoforms (2-1, 2a, and 2b) or YFP alone. Anti-GFP-immunoprecipitated LIMK2 isoforms and cofilin or MBP were used in the kinase assay. The anti-GFP immunoprecipitates were also subjected to anti-GFP immunoblotting and Coomassie blue staining. Quantification of phospho-cofilin and phospho-MBP is shown in the bottom graph. Phospho-cofilin levels obtained with anti-GFP-immunoprecipitated LIMK2a were normalized to 100. Each value represents the mean ± SE of three independent experiments.

Phospho-cofilin in intact cells

We then studied LIMK2 kinase activity more directly by measuring the level of endogenous phospho-cofilin in HEK cells overexpressing one of the LIMK2 isoforms. HEK cells were transfected with one of the HA-tagged LIMK2 isoforms or the corresponding empty vector, and lysates were analyzed by western blotting using an antibody specifically targeting phospho-Ser3 cofilin. The presence of LIMK2a or LIMK2b induced a significant and reproducible increase in phospho-cofilin levels over that of control conditions, whereas the presence of LIMK2-1 had no detectable effect (Figure 5B, left panel).

We repeated the same experiment with a C-terminal YFP-tagged version of the three LIMK2 isoforms and an untagged version of LIMK2-1 to rule out possible interference by the N-terminal HA tag. The results were identical with those obtained using the HA-tagged versions of the three isoforms (Figure 5B, right panel for YFP-tagged isoforms).

In vitro kinase tests

We then studied the kinase activity of the LIMK2 isoforms by in vitro labeling with γ[32P]ATP. HEK cells were transfected with each of the HA-tagged versions of the LIMK2 isoforms and the kinase activity of the anti-HA-LIMK2 immunoprecipitates measured using recombinant GST-cofilin as a substrate in the presence of γ[32P]ATP. Immunoprecipitated LIMK2a and LIMK2b phosphorylated cofilin, whereas LIMK2-1 did not (Figure 5C). We obtained similar results using YFP-tagged versions of the three isoforms immunoprecipitated with GFP-trap beads in the presence of recombinant cofilin and γ[32P]ATP (Figure 5D).

We then tested whether LIMK2-1 had no kinase activity or if its activity on cofilin was impaired. We repeated the in vitro labeling experiment using MBP as a substrate instead of cofilin. MBP is an efficient substrate for numerous protein kinases. There was a high background signal under the control conditions when the assay was performed in the presence of the HA-tagged versions of the LIMK2 isoforms (data not shown). We overcame this problem by using the YFP-tagged version of these proteins. Under these conditions, the background in the control was low, allowing further studies. GFP-trapped YFP-LIMK2a, LIMK2b, and LIMK2-1 showed kinase activity towards MBP, although the activity of LIMK2-1 was lower (Figure 5D). However, LIMK2-1 was also less efficiently immunoprecipitated under these conditions (see CBB staining and western blotting). Thus, the three isoforms showed comparable activity on MBP when phospho-MBP was normalized to immunoprecipitated LIMK2 levels (by CBB staining) (Figure 5D, lower panel).

Overall, these data show that LIMK2a and LIMK2b have similar activities on cofilin and MBP. Although LIMK2-1 has kinase activity towards MBP comparable to that of the other two isoforms, cofilin is not a good substrate for it.

LIMK2-1 and PP1

LIMK2-1 interacts with PP1

LIMK2-1 has a PP1i domain at its C-terminus, identified by sequence homology, whereas its two counterparts do not. We hypothesized that LIMK2-1 may regulate actin cytoskeleton dynamics via its interaction with PP1. Indeed, PP1 dephosphorylates cofilin [3033]. LIMK2-1 may indirectly increase the pool of phospho-cofilin by inhibiting its dephosphorylation by PP1.

PP1 is known to interact with many proteins via a consensus motif R/K-V/I-X-F, which serves as an anchor for the initial binding between PP1 and its partner. Furthermore, a secondary interaction site is often present and affects the activity and substrate specificity of PP1 [3436]. The K-V-R-F motif is present in the sequences of the three LIMK2 isoforms (Figure 1A, underlined in purple and Figure 6A). We tested the interaction between PP1 and each of the three LIMK2 isoforms by co-immunoprecipitation. Flag-tagged PP1 was co-transfected with one of the HA-tagged LIMK2 isoforms or unrelated Larp6 protein and immunoprecipitated with anti-Flag beads. Each isoform of LIMK2 specifically interacted with PP1 (Figure 6B).We then look for the potential secondary interaction site. First, we focused on the potential role of the C-terminal portion of the LIMK2 isoforms by using N-terminal truncated versions of the LIMK2 isoforms, consisting of the kinase domain of LIMK2a/2b, referred to as KIN2, and the kinase-PP1i domain of LIMK2-1, referred to as KIN1-PP1i (Figure 6C, right panel). Flag-tagged PP1 was co-transfected with HA-tagged KIN2 or HA-tagged KIN1-PP1i and immunoprecipitated with anti-Flag beads. PP1 interacted with KIN1-PP1i and a weak band of KIN2 was observed (Figure 6C). It is possible that the interaction between PP1 and KIN1-PP1i is mediated by the extra PP1i domain of LIMK2-1, which may be the secondary interaction site of LIMK2-1 with PP1. Flag-PP1 was then co-transfected with the restricted PP1i domain of LIMK2-1 and immunoprecipitated with anti-Flag beads. PP1i alone interacted with PP1 (Figure 6D).

The three LIMK2 isoforms interact with PP1.
Figure 6.
The three LIMK2 isoforms interact with PP1.

(A) The consensus PP1 interaction motif is present in the sequence of the three LIMK2 isoforms. (B) Interaction between LIMK2 isoforms and PP1. HEK-293 cells were co-transfected with Flag-tagged PP1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the unrelated Larp6 protein. Lysates and anti-Flag immunoprecipitates were subjected to western blotting. (C) Interaction between LIMK2 C-terminal fragments and PP1. Left panel: Scheme of KIN1-PP1i and KIN2. Right panel: HEK-293 cells were co-transfected with Flag-tagged PP1 and either the C-terminal domain of LIMK2a/2b (KIN2) or the C-terminal domain of LIMK2-1 (KIN1-PP1i). Lysates and anti-Flag immunoprecipitates were subjected to western blotting. (D) Interaction between the PP1i domain of LIMK2-1 and PP1. HEK-293 cells were co-transfected with Flag-tagged PP1 and the HA-tagged PP1i domain of LIMK2-1. Lysates and anti-Flag immunoprecipitates were subjected to western blotting.

Figure 6.
The three LIMK2 isoforms interact with PP1.

(A) The consensus PP1 interaction motif is present in the sequence of the three LIMK2 isoforms. (B) Interaction between LIMK2 isoforms and PP1. HEK-293 cells were co-transfected with Flag-tagged PP1 and one of the three HA-tagged LIMK2 isoforms (2-1, 2a, and 2b) or the unrelated Larp6 protein. Lysates and anti-Flag immunoprecipitates were subjected to western blotting. (C) Interaction between LIMK2 C-terminal fragments and PP1. Left panel: Scheme of KIN1-PP1i and KIN2. Right panel: HEK-293 cells were co-transfected with Flag-tagged PP1 and either the C-terminal domain of LIMK2a/2b (KIN2) or the C-terminal domain of LIMK2-1 (KIN1-PP1i). Lysates and anti-Flag immunoprecipitates were subjected to western blotting. (D) Interaction between the PP1i domain of LIMK2-1 and PP1. HEK-293 cells were co-transfected with Flag-tagged PP1 and the HA-tagged PP1i domain of LIMK2-1. Lysates and anti-Flag immunoprecipitates were subjected to western blotting.

In conclusion, the three isoforms of LIMK2 interact with PP1. LIMK2-1 seems to have a secondary interaction site present in its PP1i domain.

LIMK2-1 partially inhibits the dephosphorylation of cofilin by PP1

The LIMK2-1 PP1i domain has very strong sequence similarity with the C-terminal portion of PHI-1, a protein which inhibits PP1 (Figure 7A) [24]. We tested whether LIMK2-1 inhibits cofilin dephosphorylation by PP1. We co-transfected HEK-293 cells with various combinations of plasmids and analyzed lysates by western blotting using an antibody that specifically targets phospho-Ser3 cofilin. All the phosphatase inhibitors were omitted in the lysis buffer. The level of phospho-cofilin was much lower in the presence of PP1 than under the control conditions (Gal3 + Larp6, non-related proteins) (Figure 7B): there was detectable PP1 activity towards cofilin in these in cellulo conditions. The level of phospho-cofilin was slightly, but significantly, higher when LIMK2-1 was co-transfected with PP1 than when PP1 was transfected with the unrelated Larp6 protein. Moreover, when PP1 was co-transfected with LIMK2-1 missing its PP1i part (LIMK2-1_ΔPP1i), the level of phospho-cofilin was comparable to the one of PP1 co-transfected with Larp6, suggesting a role of the PP1i domain in this modulation of phospho-cofilin level. Altogether, these data suggest that LIMK2-1 partially inhibits cofilin dephosphorylation by PP1 under these in cellulo conditions.

LIMK2-1 regulates cofilin phosphorylation via its PP1i domain.
Figure 7.
LIMK2-1 regulates cofilin phosphorylation via its PP1i domain.

(A) Alignment between the C-terminal portions of LIMK2-1 and PHI-1. (B) LIMK2-1 partially inhibits cofilin dephosphorylation by PP1. HEK-293 cells were co-transfected with the indicated plasmid pairs, resulting in overexpression of the indicated proteins. Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells (Gal3 + Larp6) was normalized to 100. Each value represents the mean ± SE of three independent experiments. (C) LIMK2a and LIMK2-1 act synergistically to increase the pool of phospho-cofilin. HEK-293 cells were co-transfected with the indicated plasmid pairs, resulting in overexpression of the indicated proteins. Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells (Larp6 + pcDNA3) was normalized to 100. Each value represents the mean ± SE of three independent experiments. (D) LIMK2-1_D430N, the kinase dead mutant of LIMK2-1, induces stress fibers. HeLa cells were transfected with HA-tagged LIMK2-1_D430N. Cells were fixed and stained with phalloidin (main picture), and anti-HA antibodies (top right corner picture). The scale bar represents 10 µm.

Figure 7.
LIMK2-1 regulates cofilin phosphorylation via its PP1i domain.

(A) Alignment between the C-terminal portions of LIMK2-1 and PHI-1. (B) LIMK2-1 partially inhibits cofilin dephosphorylation by PP1. HEK-293 cells were co-transfected with the indicated plasmid pairs, resulting in overexpression of the indicated proteins. Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells (Gal3 + Larp6) was normalized to 100. Each value represents the mean ± SE of three independent experiments. (C) LIMK2a and LIMK2-1 act synergistically to increase the pool of phospho-cofilin. HEK-293 cells were co-transfected with the indicated plasmid pairs, resulting in overexpression of the indicated proteins. Lysates were subjected to western blotting. Quantification of the ratio of phospho-cofilin versus cofilin is shown in the bottom graph. Phospho-cofilin versus cofilin ratio of mock-transfected cells (Larp6 + pcDNA3) was normalized to 100. Each value represents the mean ± SE of three independent experiments. (D) LIMK2-1_D430N, the kinase dead mutant of LIMK2-1, induces stress fibers. HeLa cells were transfected with HA-tagged LIMK2-1_D430N. Cells were fixed and stained with phalloidin (main picture), and anti-HA antibodies (top right corner picture). The scale bar represents 10 µm.

LIMK2-1 and LIMK2a synergistically phosphorylate cofilin

Our data suggest that LIMK2-1 regulates phospho-cofilin levels and thus cytoskeleton dynamics, not by directly phosphorylating cofilin like LIMK2a and LIMK2b, but by preventing cofilin dephosphorylation by PP1. The three LIMK2 isoforms may have a complementary action, resulting in an increase in the phospho-cofilin pool.

We tested this possibility by cotransfecting HEK-293 cells with LIMK2-1 or LIMK2a alone, or both, and analyzing lysates by western blot using an antibody specific for phospho-Ser3 cofilin. All the phosphatase inhibitors were omitted in the lysis buffer. Phospho-cofilin levels were significantly and reproducibly higher when cells were co-transfected with both LIMK2-1 and LIMK2a than with LIMK2a alone (Figure 7C). When LIMK2a was co-transfected with a kinase dead version of LIMK2-1, LIMK2-1_D430N, this increase was even higher. This mutant was also able to induce thin stress fiber formation (Figure 7D). These data suggest that these effects of LIMK2-1 are not due to its kinase activity. LIMK2a and LIMK2-1 act synergistically to increase the pool of phospho-cofilin in the cell.

Discussion

Recent research has shown LIMK2 to have a role in cancer development, metastasis formation, tumor resistance to microtubule-targeting drugs, and neurological disorders [1,21]. Three isoforms of human LIMK2 are described in databanks: LIMK2a, LIMK2b, and LIMK2-1. Most studies have focused on LIMK2a and, to a lesser extent, LIMK2b. LIMK2-1, which differs from its two counterparts by the presence of a C-terminal PP1i domain, is very poorly studied. Here, we focused on this isoform and further characterized the three LIMK isoforms to better delineate their role in the cell.

We first showed the existence of the LIMK2-1 protein. We developed an antibody against a 12 amino acid peptide of the PP1i domain of LIMK2-1. We could detect endogenous LIMK2-1 expression in HEK293 and HeLa cell lines, as well as in various human tissues (liver, pancreas, testis, and lung). LIMK2-1 expression varied depending on the tissue.

We biochemically characterized the three LIMK2 isoforms. LIMK2 subcellular localization has already been investigated in former studies. Osada et al. [8] showed that both LIMK2a and LIMK2b were located in the cytoplasm and the nucleus, but LIMK2b to a lesser extent in the nucleus, by immunofluorescence of COS cells overexpressing cMyc-tagged LIMK2a or LIMK2b. Smolich et al. [23] showed that LIMK2b was present in both the cytoplasmic and nuclear fractions by fractionation experiments on HEK-293 cells overexpressing LIMK2b. Our data using HeLa and HEK cells are in accordance with these published results. We also showed that transfected as well as endogenous LIMK2-1 is mostly located in the cytoplasm (Figure 2). This result is surprising, given the LIMK2 isoform sequences. Goyal et al. [37,38] identified several nuclear localization signals (NLS) in regions 480–503 and 280–286 of LIMK2a. These domains are identical in all three isoforms (Figure 1) and thus cannot explain the different localization of the LIMK2 isoforms. It is possible that the longer C-terminal domain of LIMK2-1 hides the NLS, thereby preventing its import into the nucleus. LIMK2-1 may also use this extra domain to interact with another partner, segregating it to the cytoplasm. These differences in subcellular localization may also arise from the N-terminal domain of the LIMK2 isoforms. Indeed, LIMK2-1 and LIMK2b are shorter at their N-terminus and the first 18 first amino acids of LIMK2a are different from those of the other two isoforms (their first LIM domain is truncated, Figure 1). Osada et al. [8] suggested that this N-terminal domain could be involved in an interaction with a partner, blocking LIMK2b in the cytoplasm. Alternatively, the extra N-terminal portion of LIMK2a may also interact with a partner, facilitating its import into the nucleus, or retaining it there. Furthermore, LIMK2b and LIMK2-1 have been shown to regulate cell cycle progression via G2/M transition in tumoral cells [6,9,10]. At this cell cycle stage, LIMK2b and LIMK2-1 are most likely present in the nucleus. Thus, shuttling of LIMK2 between the cytoplasm and the nucleus is probably tightly regulated according to environmental conditions. This was suggested by Goyal et al. [37,38] who showed that LIMK2a phosphorylation by PKC on both Ser283 and Thr494 completely inhibited LIMK2a nuclear import in human umbilical vein endothelial cells.

LIMK2a belongs to the Rho/ROCK/LIMK2/cofilin pathway [28,29]. We showed that the three LIMK2 isoforms bind to ROCK1, which specifically phosphorylates them on Thr505 for LIMK2a and Thr484 for LIMK2b and LIMK2-1. Surprisingly, however, LIMK2-1 did not phosphorylate cofilin, either in cellulo or in vitro, whereas LIMK2a and LIMK2b did. Nevertheless, LIMK2-1 has kinase activity towards MBP, a substrate broadly used to test kinase activity. The role of LIMK2-1 phosphorylation by ROCK is not immediately clear, as LIMK2a phosphorylation by ROCK activates LIMK2 kinase activity towards cofilin [27,28]. Phosphorylation of LIMK2-1 by ROCK may trigger its activation towards another substrate. It is unclear why LIMK2-1 did not phosphorylate cofilin, the canonical substrate of LIM kinases, as the three isoforms of LIMK2 have kinase domains with almost identical sequences. There are several possible explanations for these results: (i) the extra PP1i domain at the LIMK2-1 C-terminus may hide the cofilin-binding site, (ii) LIMK2-1 is missing a few amino acids at the end of the kinase domain with respect to that of LIMK2a/2b, which may perturb kinase activity or cofilin binding, or (iii) a second protein that interacts with or processes LIMK2-1 (maybe by cleavage of the extra PP1i domain) may be necessary for LIMK2-1 to phosphorylate cofilin. The crystal structure of a complex between the kinase domain of LIMK1 and cofilin has been recently resolved [39]. The LIMK1 residues involved in the interaction with cofilin are conserved in the three LIMK2 isoforms, and thus, the LIMK2–cofilin interaction should be preserved. However, we cannot exclude that the PP1i domain of LIMK2-1 may destabilize this interaction, although the C-terminal portion of LIMK1 is distant from the cofilin interaction site in the crystal structure of the complex.

Although LIMK2-1 is unable to phosphorylate cofilin, it still remodels the actin cytoskeleton and promotes stress fiber formation. The mechanism by which LIMK2-1 regulates cytoskeleton dynamics may reside in the C-terminal PP1i domain. This domain receives its name in the databanks because it shows strong homology with PP1 inhibitor proteins, especially PHI-1, which shares 93% identity with this domain (Figure 7A). The mammalian genome encodes far fewer Ser/Thr phosphatases than Ser/Thr kinases (∼40 versus ∼400) [35]. Phosphatases get their diversity by forming holoenzymes. This is particularly true for the Ser/Thr phosphatase, PP1. PP1 holoenzymes consist of a catalytic subunit and one or two variable regulatory (PPP1R-) subunits that determine the substrate specificity, allosteric regulation, and subcellular compartmentalization of PP1. One family of such PPP1Rs, the PPP1R14s, encode members of the PHI/CPI-17 family, also called phosphatase holoenzyme inhibitors, comprising CPI-17, PHI1/2, KEPI, and GBPI [40,41]. These specific regulators of PP1 complexes act in addition to, not instead of, regulatory subunits and result in PP1 holoenzyme inhibition. LIMK2-1 may belong to this family, as the PP1i domain of LIMK2-1 exhibits 93% sequence identity with PHI-1. Here, the three LIMK2 isoforms interacted with PP1, consistent with the fact that they all possess the consensus sequence R/K-V/I-X-F, which constitutes the anchoring motif between PP1 and its partners. Moreover, PP1i domain of LIMK2-1 alone interacts with PP1, suggesting that this domain may constitute a second PP1 interaction site. The second binding site has been shown to be crucial, as it brings PP1 into close proximity with its PPP1R, resulting in additional interactions that determine the activity and substrate specificity of the holoenzyme [42]. PP1 has been shown to dephosphorylate cofilin in various cell lines [3033]. Our data are in good accordance with these results as we observed a strong decrease in phospho-cofilin levels in HEK-293 cells transfected with PP1 catalytic subunit α (Figure 7B). Moreover, co-transfection of PP1 with LIMK2-1 resulted in a weak but relevant increase in phospho-cofilin levels, suggesting slight inhibition of PP1-mediated cofilin dephosphorylation by LIMK2-1 in our in cellulo conditions. This effect seems to be correlated with PP1i domain as it was abolished when LIMK2-1 missed its PP1i domain. The faintness of the effect of LIMK2-1 on PP1 activity may come from the fact that we may have missed activation of this inhibition, as reported for PHI/CPI-17 family member PPP1R14s. Indeed, PHI/CPI-17 family members become potent inhibitors of PP1 holoenzymes after phosphorylation on a conserved threonine by diverse kinases including PKC, ROCK, MYPT-associated kinase, or integrin-linked kinase [24,4345]. LIMK2-1 has this conserved threonine (Thr 596). We attempted to mimic this threonine phosphorylation by substituting it with aspartic acid residue. However, this mutant no longer inhibited cofilin dephosphorylation by PP1 (data not shown). These results are not surprising as substituting Thr38 of CPI-17 by Asp or Glu did not result in an increase in inhibition, but rather destabilized the interaction between CPI-17 and PP1 [46]. We are currently studying whether LIMK2-1 can be phosphorylated by a kinase on Thr596, and whether such phosphorylation enhances LIMK2-1-mediated inhibition of cofilin dephosphorylation by PP1. We also attempted to develop an in vitro assay to further characterize this inhibition, but were unsuccessful. We probably missed regulatory unit(s) necessary to trigger catalytic activity of the PP1 holoenzyme towards cofilin. It may be necessary to identify this or these regulatory subunits to further understand the action of PP1 on cofilin. PP1 appears to act as a hub as it interacts with specific partners to form holoenzymes, according to specific signaling mechanisms (normal or tumorigenic conditions, cell types). Multiple holoenzyme-specific strategies have evolved for fine and acute regulation of phosphatase activity. The underlying molecular mechanisms are mostly still poorly understood. This field will provide many challenges in coming years as phosphatases are emerging as new therapeutic targets. Indeed, the role of phosphatases in many diseases is well established and new therapies that target phosphatases may be an attractive alternative to those that target kinases.

Here, we have characterized a new isoform of human LIMK2, LIMK2-1. LIMK2-1 regulates actin cytoskeleton dynamics, as its two counterparts, LIMK2a and LIMK2b, but through a distinct signaling pathway. LIMK2a and LIMK2b directly phosphorylate cofilin, whereas LIMK2-1 appears to partially prevent its dephosphorylation by PP1 (Figure 8). The cofilin/phospho-cofilin balance mediates actin dynamics, which, in turn, control many physiological mechanisms and are involved in many pathological processes. Currently, small molecules inhibiting the kinase activity of LIMK2 are developed as LIM kinases appear to be new therapeutic targets. As LIMK2-1 has no kinase activity towards cofilin, it may constitute a by-pass for these inhibitors. Our work provides a better understanding of the regulation of actin dynamics by the cofilin/phospho-cofilin balance via the action of LIMK2s, providing new elements for the development of future therapies.

Schematic representation of our findings.

Figure 8.
Schematic representation of our findings.

LIMK2a, LIMK2b, and LIMK2-1 form a complex that regulates the balance between cofilin and phospho-cofilin. LIMK2a and LIMK2b directly phosphorylate cofilin, whereas LIMK2-1 prevents cofilin dephosphorylation by the phosphatase PP1.

Figure 8.
Schematic representation of our findings.

LIMK2a, LIMK2b, and LIMK2-1 form a complex that regulates the balance between cofilin and phospho-cofilin. LIMK2a and LIMK2b directly phosphorylate cofilin, whereas LIMK2-1 prevents cofilin dephosphorylation by the phosphatase PP1.

Abbreviations

     
  • LIMKs

    LIM kinases

  •  
  • MBP

    myelin basic protein

  •  
  • NLS

    nuclear localization signals

  •  
  • PP1

    protein phosphatase 1

  •  
  • PP1i

    protein phosphatase 1 inhibitory

Author Contribution

B.V., H.C., M.D., and F.G. performed experiments. D.G. did immunofluorescence image treatment. B.V. and H.B. were involved in the concept, design, and interpretation of data. P.V. and C.A. discussed the manuscript. B.V. and H.B. wrote the manuscript.

Funding

This work was supported by La ligue contre le Cancer, l'Association Neurofibromatoses et Recklinghausen, and la Région Centre Val de Loire.

Acknowledgments

We thank S. Shieh who kindly provided cMyc-LIMK2b plasmid. Many thanks to Aurélie Cosson and Déborah Casas for their support.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Scott
,
R.W.
and
Olson
,
M.F.
(
2007
)
LIM kinases: function, regulation and association with human disease
.
J. Mol. Med.
85
,
555
568
2
Bamburg
,
J.R.
and
Bernstein
,
B.W.
(
2010
)
Roles of ADF/cofilin in actin polymerization and beyond
.
F1000 Biol. Rep.
2
,
62
3
Mizuno
,
K.
(
2013
)
Signaling mechanisms and functional roles of cofilin phosphorylation and dephosphorylation
.
Cell. Signal.
25
,
457
469
4
Pollard
,
T.D.
and
Borisy
,
G.G.
(
2003
)
Cellular motility driven by assembly and disassembly of actin filaments
.
Cell
112
,
453
465
5
Prunier
,
C.
,
Prudent
,
R.
,
Kapur
,
R.
,
Sadoul
,
K.
and
Lafanechère
,
L.
(
2017
)
LIM kinases: cofilin and beyond
.
Oncotarget
8
,
41749
41763
6
Croft
,
D.R.
,
Crighton
,
D.
,
Samuel
,
M.S.
,
Lourenco
,
F.C.
,
Munro
,
J.
,
Wood
,
J.
et al. 
(
2011
)
p53-mediated transcriptional regulation and activation of the actin cytoskeleton regulatory RhoC to LIMK2 signaling pathway promotes cell survival
.
Cell Res.
21
,
666
682
7
Nomoto
,
S.
,
Tatematsu
,
Y.
,
Takahashi
,
T.
and
Osada
,
H.
(
1999
)
Cloning and characterization of the alternative promoter regions of the human LIMK2 gene responsible for alternative transcripts with tissue-specific expression
.
Gene
236
,
259
271
8
Osada
,
H.
,
Hasada
,
K.
,
Inazawa
,
J.
,
Uchida
,
K.
,
Ueda
,
R.
,
Takahashi
,
T.
et al. 
(
1996
)
Subcellular localization and protein interaction of the human LIMK2 gene expressing alternative transcripts with tissue-specific regulation
.
Biochem. Biophys. Res. Commun.
229
,
582
589
9
Gamell
,
C.
,
Schofield
,
A.V.
,
Suryadinata
,
R.
,
Sarcevic
,
B.
and
Bernard
,
O.
(
2013
)
LIMK2 mediates resistance to chemotherapeutic drugs in neuroblastoma cells through regulation of drug-induced cell cycle arrest
.
PLoS ONE
8
,
e72850
10
Hsu
,
F.F.
,
Lin
,
T.Y.
,
Chen
,
J.Y.
and
Shieh
,
S.Y.
(
2010
)
p53-Mediated transactivation of LIMK2b links actin dynamics to cell cycle checkpoint control
.
Oncogene
29
,
2864
2876
11
Rak
,
R.
and
Kloog
,
Y.
(
2014
)
Targeting LIM kinase in cancer and neurofibromatosis
.
Cell Cycle
13
,
1360
1361
12
Manetti
,
F.
(
2012
)
Recent findings confirm LIM domain kinases as emerging target candidates for cancer therapy
.
Curr. Cancer Drug Targets
12
,
543
560
13
Suyama
,
E.
,
Wadhwa
,
R.
,
Kawasaki
,
H.
,
Yaguchi
,
T.
,
Kaul
,
S.C.
,
Nakajima
,
M.
et al. 
(
2004
)
LIM kinase-2 targeting as a possible anti-metastasis therapy
.
J. Gene Med.
6
,
357
363
14
Vlecken
,
D.H.
and
Bagowski
,
C.P.
(
2009
)
LIMK1 and LIMK2 are important for metastatic behavior and tumor cell-induced angiogenesis of pancreatic cancer cells
.
Zebrafish
6
,
433
439
15
Johnson
,
E.O.
,
Chang
,
K.H.
,
Ghosh
,
S.
,
Venkatesh
,
C.
,
Giger
,
K.
,
Low
,
P.S.
et al. 
(
2012
)
LIMK2 is a crucial regulator and effector of Aurora-A-kinase-mediated malignancy
.
J. Cell Sci.
125
,
1204
1216
16
Po'uha
,
S.T.
,
Shum
,
M.S.
,
Goebel
,
A.
,
Bernard
,
O.
and
Kavallaris
,
M.
(
2010
)
LIM-kinase 2, a regulator of actin dynamics, is involved in mitotic spindle integrity and sensitivity to microtubule-destabilizing drugs
.
Oncogene
29
,
597
607
17
Prudent
,
R.
,
Vassal-Stermann
,
E.
,
Nguyen
,
C.H.
,
Pillet
,
C.
,
Martinez
,
A.
,
Prunier
,
C.
et al. 
(
2012
)
Pharmacological inhibition of LIM kinase stabilizes microtubules and inhibits neoplastic growth
.
Cancer Res.
72
,
4429
4439
18
Starinsky-Elbaz
,
S.
,
Faigenbloom
,
L.
,
Friedman
,
E.
,
Stein
,
R.
and
Kloog
,
Y.
(
2009
)
The pre-GAP-related domain of neurofibromin regulates cell migration through the LIM kinase/cofilin pathway
.
Mol. Cell. Neurosci.
42
,
278
287
19
Vallée
,
B.
,
Doudeau
,
M.
,
Godin
,
F.
,
Gombault
,
A.
,
Tchalikian
,
A.
,
de Tauzia
,
M.L.
et al. 
(
2012
)
Nf1 RasGAP inhibition of LIMK2 mediates a new cross-talk between Ras and Rho pathways
.
PLoS ONE
7
,
e47283
20
Cuberos
,
H.
,
Vallée
,
B.
,
Vourc'h
,
P.
,
Tastet
,
J.
,
Andres
,
C.R.
and
Bénédetti
,
H.
(
2015
)
Roles of LIM kinases in central nervous system function and dysfunction
.
FEBS Lett.
589
,
3795
3806
21
Manetti
,
F.
(
2012
)
LIM kinases are attractive targets with many macromolecular partners and only a few small molecule regulators
.
Med. Res. Rev.
32
,
968
998
22
Genis-Mendoza
,
A.D.
,
Gallegos-Silva
,
R.I.
,
Lopez-Casamichana
,
M.
,
Lopez-Rubalcava
,
C.
and
Nicolini
,
H.
(
2013
)
Gene expression profiles of nucleus accumbens, prefrontal cortex and hippocampus in an animal model of schizophrenia: proposed candidate genes
.
Actas Esp. Psiquiatr.
41
,
154
163
PMID:
[PubMed]
23
Smolich
,
B.
,
Vo
,
M.
,
Buckley
,
S.
,
Plowman
,
G.
and
Papkoff
,
J.
(
1997
)
Cloning and biochemical characterization of LIMK-2, a protein kinase containing two LIM domains
.
J. Biochem.
121
,
382
388
24
Eto
,
M.
,
Karginov
,
A.
and
Brautigan
,
D.L.
(
1999
)
A novel phosphoprotein inhibitor of protein type-1 phosphatase holoenzymes
.
Biochemistry
38
,
16952
16957
25
Acevedo
,
K.
,
Moussi
,
N.
,
Li
,
R.
,
Soo
,
P.
and
Bernard
,
O.
(
2006
)
LIM kinase 2 is widely expressed in all tissues
.
J. Histochem. Cytochem.
54
,
487
501
26
Hiraoka
,
J.
,
Okano
,
I.
,
Higuchi
,
O.
,
Yang
,
N.
and
Mizuno
,
K.
(
1996
)
Self-association of LIM-kinase 1 mediated by the interaction between an N-terminal LIM domain and a C-terminal kinase domain
.
FEBS Lett.
399
,
117
121
27
Amano
,
T.
,
Tanabe
,
K.
,
Eto
,
T.
,
Narumiya
,
S.
and
Mizuno
,
K.
(
2001
)
LIM-kinase 2 induces formation of stress fibres, focal adhesions and membrane blebs, dependent on its activation by Rho-associated kinase-catalysed phosphorylation at threonine-505
.
Biochem. J.
354
,
149
159
28
Sumi
,
T.
,
Matsumoto
,
K.
and
Nakamura
,
T.
(
2001
)
Specific activation of LIM kinase 2 via phosphorylation of threonine 505 by ROCK, a Rho-dependent protein kinase
.
J. Biol. Chem.
276
,
670
676
29
Maekawa
,
M.
,
Ishizaki
,
T.
,
Boku
,
S.
,
Watanabe
,
N.
,
Fujita
,
A.
,
Iwamatsu
,
A.
et al. 
(
1999
)
Signaling from Rho to the actin cytoskeleton through protein kinases ROCK and LIM-kinase
.
Science
285
,
895
898
30
Ambach
,
A.
,
Saunus
,
J.
,
Konstandin
,
M.
,
Wesselborg
,
S.
,
Meuer
,
S.C.
and
Samstag
,
Y.
(
2000
)
The serine phosphatases PP1 and PP2A associate with and activate the actin-binding protein cofilin in human T lymphocytes
.
Eur. J. Immunol.
30
,
3422
3431
31
Li
,
G.B.
,
Cheng
,
Q.
,
Liu
,
L.
,
Zhou
,
T.
,
Shan
,
C.Y.
,
Hu
,
X.Y.
et al. 
(
2013
)
Mitochondrial translocation of cofilin is required for allyl isothiocyanate-mediated cell death via ROCK1/PTEN/PI3K signaling pathway
.
Cell Commun. Signal.
11
,
50
32
Oleinik
,
N.V.
,
Krupenko
,
N.I.
and
Krupenko
,
S.A.
(
2010
)
ALDH1L1 inhibits cell motility via dephosphorylation of cofilin by PP1 and PP2A
.
Oncogene
29
,
6233
6244
33
Samstag
,
Y.
and
Nebl
,
G.
(
2003
)
Interaction of cofilin with the serine phosphatases PP1 and PP2A in normal and neoplastic human T lymphocytes
.
Adv. Enzyme Regul.
43
,
197
211
34
Bollen
,
M.
,
Peti
,
W.
,
Ragusa
,
M.J.
and
Beullens
,
M.
(
2010
)
The extended PP1 toolkit: designed to create specificity
.
Trends Biochem. Sci.
35
,
450
458
35
Ceulemans
,
H.
,
Stalmans
,
W.
and
Bollen
,
M.
(
2002
)
Regulator-driven functional diversification of protein phosphatase-1 in eukaryotic evolution
.
BioEssays
24
,
371
381
36
Wakula
,
P.
,
Beullens
,
M.
,
Ceulemans
,
H.
,
Stalmans
,
W.
and
Bollen
,
M.
(
2003
)
Degeneracy and function of the ubiquitous RVXF motif that mediates binding to protein phosphatase-1
.
J. Biol. Chem.
278
,
18817
18823
37
Goyal
,
P.
,
Pandey
,
D.
,
Behring
,
A.
and
Siess
,
W.
(
2005
)
Inhibition of nuclear import of LIMK2 in endothelial cells by protein kinase C-dependent phosphorylation at Ser-283
.
J. Biol. Chem.
280
,
27569
27577
38
Goyal
,
P.
,
Pandey
,
D.
and
Siess
,
W.
(
2006
)
Phosphorylation-dependent regulation of unique nuclear and nucleolar localization signals of LIM kinase 2 in endothelial cells
.
J. Biol. Chem.
281
,
25223
25230
39
Hamill
,
S.
,
Lou
,
H.J.
,
Turk
,
B.E.
and
Boggon
,
T.J.
(
2016
)
Structural basis for noncanonical substrate recognition of cofilin/ADF proteins by LIM kinases
.
Mol. Cell
62
,
397
408
40
Eto
,
M.
(
2009
)
Regulation of cellular protein phosphatase-1 (PP1) by phosphorylation of the CPI-17 family, C-kinase-activated PP1 inhibitors
.
J. Biol. Chem.
284
,
35273
35277
41
Eto
,
M.
and
Brautigan
,
D.L.
(
2012
)
Endogenous inhibitor proteins that connect Ser/Thr kinases and phosphatases in cell signaling
.
IUBMB Life
64
,
732
739
42
Verbinnen
,
I.
,
Ferreira
,
M.
and
Bollen
,
M.
(
2017
)
Biogenesis and activity regulation of protein phosphatase 1
.
Biochem. Soc. Trans.
45
,
89
99
43
Deng
,
J.T.
,
Sutherland
,
C.
,
Brautigan
,
D.L.
,
Eto
,
M.
and
Walsh
,
M.P.
(
2002
)
Phosphorylation of the myosin phosphatase inhibitors, CPI-17 and PHI-1, by integrin-linked kinase
.
Biochem. J.
367
,
517
524
44
Liu
,
Q.R.
,
Zhang
,
P.W.
,
Lin
,
Z.
,
Li
,
Q.F.
,
Woods
,
A.S.
,
Troncoso
,
J.
et al. 
(
2004
)
GBPI, a novel gastrointestinal- and brain-specific PP1-inhibitory protein, is activated by PKC and inactivated by PKA
.
Biochem. J.
377
,
171
181
45
Liu
,
Q.R.
,
Zhang
,
P.W.
,
Zhen
,
Q.
,
Walther
,
D.
,
Wang
,
X.B.
and
Uhl
,
G.R.
(
2002
)
KEPI, a PKC-dependent protein phosphatase 1 inhibitor regulated by morphine
.
J. Biol. Chem.
277
,
13312
13320
46
Ohki
,
S.
,
Eto
,
M.
,
Shimizu
,
M.
,
Takada
,
R.
,
Brautigan
,
D.L.
and
Kainosho
,
M.
(
2003
)
Distinctive solution conformation of phosphatase inhibitor CPI-17 substituted with aspartate at the phosphorylation-site threonine residue
.
J. Mol. Biol.
326
,
1539
1547
47
Ishizaki
,
T.
,
Naito
,
M.
,
Fujisawa
,
K.
,
Maekawa
,
M.
,
Watanabe
,
N.
,
Saito
,
Y.
et al. 
(
1997
)
p160ROCK, a Rho-associated coiled-coil forming protein kinase, works downstream of Rho and induces focal adhesions
.
FEBS Lett.
404
,
118
124

Supplementary data