Cardiac light chain is phosphorylated by + Ca2 /calmodulin-dependent and -independent activities

Audrey N. Changa,1, Pravin Mahajanb, Stefan Knappb,c,d, Hannah Bartone, H. Lee Sweeneye,f, Kristine E. Kamma, and James T. Stulla

aDepartment of Physiology, University of Texas Southwestern Medical Center, Dallas, TX 75390; bNuffield Department of Clinical Medicine, Structural Genomics Consortium, University of Oxford, Oxford OX3 7DQ, United Kingdom; cInstitute for Pharmaceutical Chemistry, Goethe University Frankfurt, 60438 Frankfurt, Germany; dBuchmann Institute for Life Sciences, Goethe University Frankfurt, 60438 Frankfurt, Germany; eDepartment of Physiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104; and fDepartment of Pharmacology and Therapeutics, University of Florida College of Medicine, Gainesville, FL 32610

Edited by James A. Spudich, Stanford University School of Medicine, Stanford, CA, and approved May 6, 2016 (received for review January 13, 2016) The well-known, muscle-specific smooth muscle myosin light chain of MLCK4 protein has not been reported. Without available crystal kinase (MLCK) (smMLCK) and skeletal muscle MLCK (skMLCK) are structures of any of the MLCKs, the mechanism of autoregulation + dedicated protein regulated by an autoregulatory segment has been defined by structural studies of other Ca2 /calmodulin C terminus of the catalytic core that blocks myosin regulatory light (CaM)-dependent kinases combined with biochemical inves- chain (RLC) binding and phosphorylation in the absence of Ca2+/cal- tigations of different MLCKs (2, 3). The smMLCK and skMLCK are + modulin (CaM). Although it is known that a more recently discovered Ca2 /CaM-dependent protein kinases (CaMKs) with autoregulatory cardiac MLCK (cMLCK) is necessary for normal RLC phosphorylation in segments C-terminal of the catalytic core. In the absence of 2+ vivo and physiological cardiac performance, information on cMLCK bio- Ca /CaM, the autoinhibitory sequence N-terminal to the CaM chemical properties are limited. We find that a fourth uncharacterized binding sequence binds to the surface of the kinase C domain to- 2+ MLCK, MLCK4, is also expressed in cardiac muscle with high catalytic ward the catalytic cleft, blocking RLC binding. Ca /CaM binding domain sequence similarity with other MLCKs but lacking an autoinhi- to the CaM-binding sequence of the autoregulatory segment dis- bitory segment. Its crystal structure shows the catalytic domain in its places the autoregulatory segment, exposing the catalytic cleft for active conformation with a short C-terminal “pseudoregulatory helix” RLC binding (4, 5). Thus, the catalytic activity of these two MLCKs 2+ that cannot inhibit catalysis as a result of missing linker regions. are completely dependent on Ca /CaM (1). As cMLCK also has + MLCK4 has only Ca2 /CaM-independent activity with comparable an autoregulatory segment, and RLC is dephosphorylated in non- 2+ V K V beating hearts, presumably because of reduced [Ca ]i (6, 7), ex- max and m values for different RLCs. In contrast, the max value + pectations are that it is Ca2 /CaM-dependent like smMLCK of cMLCK is orders of magnitude lower than those of the other + and skMLCK. However, conflicting reports on the Ca2 /CaM- three MLCK family members, whereas its Km (RLC and ATP) and dependency of cMLCK activity necessitate further investiga- KCaM values are similar. In contrast to smMLCK and skMLCK, which + lack activity in the absence of Ca2 /CaM, cMLCK has constitutive tion (8, 9). activity that is stimulated by Ca2+/CaM. Potential contributions of The extent of RLC phosphorylation in muscle cells is deter- autoregulatory segment to cMLCK activity were analyzed with chi- mined by balanced activities of MLCK and myosin light chain meras of skMLCK and cMLCK. The constitutive, low activity of cMLCK phosphatase(s) (1). When heart extracts are processed by using appears to be intrinsic to its catalytic core structure rather than an autoinhibitory segment. Thus, RLC phosphorylation in cardiac muscle Significance may be regulated by two different protein kinases with distinct bio- chemical regulatory properties. Chronic heart failure is associated with decreased cardiac myosin light chain kinase (MLCK; cMLCK) expression and myosin regulatory cMLCK | MLCK4 | kinase | calcium/calmodulin | crystallography light chain (RLC) phosphorylation, similar to heart failure associated with mutations in numerous sarcomeric proteins. Although abla- nlike most smooth and skeletal muscles, the muscle of a tion of cMLCK expression reduces RLC phosphorylation sufficiently Ucontinuously beating heart never relaxes for an extended to cause heart failure, the residual phosphorylation indicates that + period as a result of cyclic increases and decreases in [Ca2 ]with another kinase also phosphorylates RLC. We find that MLCK4 is also expressed abundantly in cardiac muscle, and structural analyses activation/deactivation of myofilament contractile proteins. The heart + indicate that it is a Ca2 /calmodulin (CaM)-independent kinase, in is intricately linked to signaling modules, which regulate cardiac 2+ performance to meet the circulatory demands of the body. Within contrast to Ca /CaM-stimulated cMLCK. Biochemical kinetic the contractile apparatus of the cardiac myocyte, fine tuning of car- analyses confirmed these structural predictions. These studies diac contractile performance is partly achieved by posttranslational define distinct regulation of cMLCK and MLCK4 activities to modifications of myofilament proteins. The main driver of muscle affect RLC phosphorylation, and lay the foundation for RLC contractions is the ATPase activity of myosin molecules, which are phosphorylation as a therapeutic target for heart failure. hexamers comprised of two each of heavy chain, essential light chain, Author contributions: A.N.C., S.K., H.L.S., K.E.K., and J.T.S. designed research; A.N.C., and regulatory light chain (RLC) subunits. RLCs of are P.M., and S.K. performed research; H.B. and H.L.S. prepared and contributed new phosphorylated by myosin light chain kinases (MLCKs) to activate or reagents; A.N.C., S.K., K.E.K., and J.T.S. analyzed data; and A.N.C. wrote the paper. modulate the myosin ATPase activity (1). The authors declare no conflict of interest. The MLCK family is comprised of four distinct kinases, MLCK1, This article is a PNAS Direct Submission. -2, -3, and -4, which are encoded by distinct (1). Based on Data deposition: The atomic coordinates and structure factors have been deposited in the their muscle type-specific expression and activities, MLCK1 is Protein Data Bank, www.pdb.org (PDB ID code 2X4F). known as smooth muscle MLCK (smMLCK), MLCK2 as skeletal 1To whom correspondence should be addressed. Email: [email protected]. muscle MLCK (skMLCK), and MLCK3 as cardiac muscle MLCK This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. (cMLCK). MLCK4 remains uncharacterized, and tissue distribution 1073/pnas.1600633113/-/DCSupplemental.

E3824–E3833 | PNAS | Published online June 20, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1600633113 Downloaded by guest on October 2, 2021 procedures that minimize loss of RLC phosphorylation and RLC S15 phosphorylation for normal cardiac performance PNAS PLUS maximize solubilization of myofilament proteins, cardiac RLC (29), and have confirmed that cMLCK is the primary protein (cRLC) shows an average phosphorylation of 0.40 ± 1mol kinase responsible for phosphorylation of RLC in ventricular phosphate/mol RLC in a variety of animals (6, 7, 10–20). Se- muscle, information on the regulation of RLC phosphorylation lective ablation of the for cMLCK, but not skMLCK, and the enzymatic properties of cMLCK is conflicting (8, 9). In decreases cRLC phosphorylation from 0.45 to 0.10 mol of addition, residual RLC phosphorylation (20–25% of normal phosphate/mol of RLC, showing kinase-specific effects in the RLC phosphorylation) in cMLCK-KO animals indicates that heart (6, 11, 12, 21). The decreased cardiac performance and another MLCK could phosphorylate RLC. dilation of the adult mouse heart associated with the attenua- Here we report that MLCK4 protein is expressed in the mouse tion of RLC phosphorylation in the cMLCK-KO animals is heart at levels higher than in skeletal and smooth muscles. We similartotheresultsobtainedwithknock-inmutantmice also report the resolved structure of MLCK4, to our knowledge containing a nonphosphorylatable cRLC (18). the first MLCK family member to be crystallized, and biochemically RLC phosphorylation is reduced in human heart failure (22– establish MLCK4 as a kinase for cRLC. Distinct from other 25) and in heart failure animal models (8, 26, 27). Over- MLCK family members, MLCK4 has a “pseudoregulatory helix” expression of MLCK in the heart increases the extent of RLC (PRH) in place of a full autoregulatory segment. From the phosphorylation, which is associated with improved cardiac crystal structure of MLCK4 and sequence comparisons with performance and attenuation of hypertrophic responses to other MLCK family members, we hypothesized that MLCK4 + stress (15, 27). In addition, overexpression of an RLC phos- would have Ca2 /CaM-independent kinase activity and cMLCK + phomimetic mutation S15D in the background of a hypertro- would be Ca2 /CaM-dependent. In vitro kinase assays with pu- phic cardiomyopathy-associated mutation D166V was recently rified MLCK4 showed that it is a constitutive kinase as predicted shown to prevent the development of a disease phenotype as- by the crystal structure, but purified recombinant cMLCK and sociated with the overexpression of D166V mutant RLC (28). endogenous cMLCK immunoprecipitated from mouse hearts + Although animal studies have solidified the requirement of showed that cMLCK has Ca2 /CaM-independent and -dependent

A C

fragment

lung spleen edl soleusheart liver MLCK4 50 -MLCK4

40 BIOCHEMISTRY -GAPDH

B

100

D E fragment 50 non-musclemyocyteheart soleusheartliver MLCK4 50 100-cMLCK -MLCK4

MLCK4 protein (AU) 50 40 0 -MLCK4 -GAPDH t 50 dl 40 -pre-adsorbed en e us ar -GAPDH lung e e liver MLCK4 pl ol he 40 -GAPDH s s

Fig. 1. Tissue expression of MLCK4. (A) Representative immunoblot of MLCK4 and GAPDH expression in supernatant fractions of mouse tissue homogenates. Purified MLCK4 (0.02 pmol) was used as control. (B) Comparison of MLCK4 protein expression in mouse tissues (mean ± SE; n = 3). (C) Immunohistochemistry for MYLK4 in adult mouse myocardium. Micrograph of basal junction of right ventricular wall and ventricular septum shows MYLK4 protein expression conserved to cardiomyocytes (brown DAB Chromogen). In the micrograph, staining is absent in fibroblasts (arrowheads), wall of aorta (marked as “A”), endothelium and basal lamina of cardiac veins (CV), aortic valve spongiosa (AVS), and basal brown fat (BF). Nuclei are counterstained with Harris hematoxylin (blue). (Scale bar: 40 μm.) (D) Representative immunoblot of cMLCK, MLCK4, and GAPDH protein expression in homogenates of mouse heart, isolated cardiac myocytes, and cardiac nonmuscle cells. (E) Immunoadsorption test for MLCK4 antibody. MLCK4 antibody (Upper) and a solution of the same antibody after preadsorption with purified MLCK4 protein (Lower) were used to blot MLCK4 protein in tissues and the purified MLCK4 fragment. Tissues are indicated. Apparent molecular weights are noted.

Chang et al. PNAS | Published online June 20, 2016 | E3825 Downloaded by guest on October 2, 2021 activities, contrary to predictions. MLCK chimeras were generated assumed an extended conformation containing also an extra-short from cMLCK and skMLCK sequences, and activities compared β-sheet segment. Sheet β1 was highly twisted, including a short turn. with WT kinases to determine the contribution of the specific The activation segment was fully ordered. The inhibitor bound to + autoregulatory segment sequences to Ca2 /CaM-dependency. the ATP formed a hydrogen bond to the hinge main We report that cMLCK has high affinity for cRLC, ATP, and chain (residue V183). A structural overview showing the main A CaM, comparable to other MLCKs, and that the Vmax value of structural elements is shown in Fig. 2 . cMLCK is orders of magnitude lower than those of other MLCK The C terminus formed a short helix typical for autoinhibited family members. CaMKs. Interestingly, in contrast to autoinhibited CaMKs, this helix was oriented away from the kinase domain. Superimposi- Results tion with the autoinhibited structure of CaMK1D (Fig. 2B) MLCK4 Is Highly Expressed in Cardiac Myocytes. Tissue extracts from showed that the extended linker region that allows the regulatory WT mice were prepared, and MLCK4 protein and GAPDH in helix to interact with a deep groove formed by αD and αFis lung, spleen, extensor digitorum longus muscle, soleus muscle, heart lacking in MLCK4, whereas at least two of the three hydrophobic apex, and liver were compared. There was one band in the region of residues that anchor the amphipathic regulatory helix to the ki- expected mass (44 kDa; Fig. 1A), and the relative MLCK4 ex- nase core (L366, L370) are still present in MLCK4. Thus, a short pression normalized to GAPDH was greatest in the heart (Fig. 1B). C-terminal helix is also conserved in MLCK4. This helix, how- Compared with the purified MLCK4 protein standard, MLCK4 was ever, cannot inhibit the catalytic domain because of missing expressed at 0.12 μM based on previously published assumptions linker regions. We therefore named this structural element PRH. (6). Expression of MLCK4 in cardiac myocytes was confirmed by In comparison with the autoinhibited CaMK1D structure, helix immunohistochemistry (Fig. 1C) of mouse heart, and immuno- αD shifted inward, closing the binding groove of the regulatory blotting isolated adult mouse cardiac myocytes and cardiac non- helix. This rearrangement of αD has been observed in CaMKII muscle cells (Fig. 1D). Relative to the GAPDH-normalized MLCK4 as a result of CaM binding and is consistent with a constitutively content in heart tissue extracts, 98 ± 7% of the MLCK4 is expressed active kinase (30). in cardiac myocytes. Preadsorption of the antibody mixture with Structure-based alignments revealed the main residue con- C purified MLCK4 removed MLCK4 band from all samples (Fig. servation shared among CaMKs (Fig. 2 ). This analysis revealed “ ” 1E), confirming specificity of the antibody. a conserved helix insertion ( HPW ) present in all compared CaMKs and a diverse linker region. The autoregulatory helices MLCK4 Assumes an Active Conformation in the Absence of Ca2+/CaM. present in MLCKs were predicted based on the structure-based To obtain structural insight into the regulation of MLCKs, we alignment and conserved hydrophobic residues anchoring the autoregulatory helix. Also, the missing linker region and the first crystallized MLCK4 and determined its structure at 2.7-Å reso- 2+ lution. By using several truncation constructs, an N-terminal trun- turn of the autoregulatory helix is evident, suggestive of Ca /CaM- cation starting at residue K40 yielded stable protein and readily independent activity. formed crystals. The structure was refined, maintaining favorable TheKinaseActivityofMLCK4IsDistinctfromskMLCK.Kinase rates were bond geometry to an R/Rfree of 19.2%/24.1% (Table 1). The final 2+ model included residues D80–Q373; thus. 40 residues at the N ter- measured to compare the Ca /CaM-dependent and -independent activities of MLCK4 to skMLCK. MLCK4 activity was not dif- minus as well as 11 residues at the C terminus were not visible in 2+ electron density maps and have been assumed to be disordered. In ferent in the presence and absence of Ca /CaM. The aver- aged V value was 129 ± 6 mol phosphate/min/mol kinase, addition, no density was visible for the tip of the phosphate binding max and the RLC Km value was 7.6 ± 1.1 μM(Fig.3A). The loop (P-loop) G113–Q119, and these residues have therefore + skMLCK has catalytic activity only in the presence of Ca2 /CaM; not been modeled. The protein was crystallized in complex the V value was 1,132 ± 54 mol phosphate/min/mol kinase and with a nonspecific ATP mimetic inhibitor [4-(2 amino-4-methyl- max the K value was 11.6 ± 2 μM(Fig.3B). 1,3-thiazol-5-yl)-n-(3-dioxaziridin-3-yl phenyl)pyrimidine-2-amine], m which was used to stabilize the protein during crystallization. Immunoprecipitated cMLCK fromMouseVentriclesHasbothCa2+/CaM- As expected, the structure of MLCK4 revealed the typical bilobal Dependent and -Independent Activities. To resolve the discrepancy in + architecture of protein kinase catalytic domains. The N terminus the reported Ca2 /CaM dependency of cMLCK activity, a new custom antibody was generated against a peptide fragment N ter- minus of the catalytic core (Fig. 4A). Comparison of cMLCK in Table 1. Data collection and refinement statistics for the input and unbound fractions showed an immunoprecipitation effi- crystallization of MLCK4 ciency 85% (Fig. 4B). Comparison of immunoprecipitated proteins from the hearts of cMLCK-KO and WT mice by Coomassie-stained Data collection MLCK4 gel analysis showed specific immunoprecipitation of cMLCK Space group C 1 2 1 C Cell dimensions a, b, c (Å) 148.44, 70.65, 96.57 expressed in the heart (Fig. 4 ). Coimmunoprecipitation of other α, β, γ (°) 90.0, 118.96, 90.0 proteins specifically with cMLCK was not observed. Immunopre- Resolution (Å) 2.67 (2.67–2.81) cipitated cMLCK was assayed for activity by measurement of 32P incorporated over time, with purified cRLC (20 μM) in the Unique observations 23,801 (1,729) 2+ D Completeness (%) 99.78 (99.8) presence and absence of Ca /CaM (Fig. 4 ). The initial kinase rate in EGTA normalized to cMLCK contents in assay mixtures was Redundancy 3.7 (3.7) ± 2+ I/σI 8.4 50 8% of the rate in the presence of Ca /CaM. Refinement The Properties of GST-Tagged Truncated cMLCK Purified from Sf9 Insect R /R (%) 19.0/24.1 work free Cell Expression System Are Similar to Full-Length cMLCK Found in Mouse No. of atoms (protein/other/solvent) 4,568/79/252 Hearts. To facilitate rigorous measurements of the enzymatic prop- Mean B-factors (Å2)20.7 + erties of cMLCK in vitro, and to confirm Ca2 /CaM-independent rmsd bond (Å) 0.015 activity measured with immunoprecipitated endogenous cMLCK, rmsd angle (o) 1.526 a truncation mutant was generated whereby the sequences PDB ID code 2X4F N-terminal of the catalytic core were replaced with a GST-tag. The Values in parentheses are last shell. kinase was overexpressed by using the Sf9 insect cell system and

E3826 | www.pnas.org/cgi/doi/10.1073/pnas.1600633113 Chang et al. Downloaded by guest on October 2, 2021 PNAS PLUS A MLCK4 B CAMK1D

4 2 1 Autoregulatory 5 Helix (CAMK1D) 3 A-loop P-loop C hinge D F

PRH G A-loop C I

D H F E C Autoregulatory Helix ...Linker.... N CAMK1D HPWIAGDTALNKNI HESVSAQIRKN.FA G CAMK2A HPWISHRSTVASCMHR QETVDC.L..KKFNARR CAMK4 HPWVTGKAANFVH MDTAQKKLQE..FNA C I MLCK4 HPWLSD...... HKLHS.RLNAQ

smMLCK HPWLMKDT.. ...K NMEAKK.LSKDRMKK cMLCK HPWLNNLPAKASRSK TRLKSQLLLQKYIA PRH H skMLCK HPWLNNLAEKAKRCN RRLKSQILLKKYLMKR

Fig. 2. Structure of MLCK4. (A) Ribbon diagram of the catalytic domain. The main structural elements are labeled. The cocrystallized inhibitor is shown in ball-and-stick representation. The C-terminal PRH is highlighted in magenta. (B) Superimposition of the MLCK4 structure with autoinhibited CaMK1D (PDB ID code ID 2JC6) colored in blue. (C) Structure-based sequence alignment of the linker and the regulatory helices of CaMK1D, CaMK2A, CaMK4, and

MLCK4. The alignment of smMLCK, cMLCK, and skMLCK was based on sequence only. Conserved hydrophobic residues that anchor the regulatory helix BIOCHEMISTRY are highlighted in red.

affinity-purified in tandem with glutathione-agarose and CaM- rate of autophosphorylation was 0.008 mol 32P incorporated/min/ Sepharose resins. cMLCK was purified to >90% purity (Fig. 5A) mol kinase. The low extents and rates of autophosphorylation, along + and assayed for activity in the presence and absence of Ca2 /CaM with the absence of autophosphorylation in the presence of RLC, (Fig. 5B). Kinase that selectively bound to CaM-Sepharose had indicate that cMLCK is probably not significantly autophosphory- + Ca2 /CaM-dependent and -independent activities, consistent with lated in vivo. the activities of full-length cMLCK immunoprecipitated from V K K mouse ventricles. Kinase activity measured after thermal de- cMLCK Has a Low max Value with RLC and ATP m and CaM Values 2+ naturation showed equivalent sensitivity to increased temper- Comparable to Other MLCKs. In the presence of Ca /CaM and 2+ V ± ature in the presence and absence of Ca /CaM (Fig. 5C). cRLC, cMLCK had a max value of 2.6 0.1 mol phosphate/min/mol These results are not consistent with the possibility of two dis- tinct populations of purified kinase accounting for the observed 2+ Ca /CaM-dependent and -independent activities. Validating the AB use of the truncated kinase to gain insight into the properties of 0.03 0.006 + cMLCK found in vivo, the purified kinase had Ca2 /CaM- dependent and -independent activities when whole myosin or cardiac myofibrils were used as substrate (Fig. 5 D and E). 0.02 0.004 Phosphorylation of cMLCK was detected by autoradiograph from purified cMLCK incubated alone in assay conditions (Fig. E

5 , lanes 1 and 4). Extent of maximum cMLCK phosphoryla- MLCK4 1/V 0.01 skMLCK 1/V 0.002 tion was measured by two methods, autoradiography of proteins on the same film as a 32P standard curve and liquid scintillation spectrometry of incorporated 32Pinanextendedassay(5–150 min) F of kinase phosphorylation (Fig. 5 ). Autoradiography and calcula- -0.150.000.150.30-0.15 0.00 0.15 0.30 tion from a standard curve showed that 7.5% of total cMLCK + proteininCa2 /CaM and 6.6% in EGTA were phosphorylated in 1/[RLC] 1/[RLC] + the absence of other specific substrates. Addition of BSA did Fig. 3. Comparison of Ca2 /CaM dependency of MLCK4 activity to not affect the extent of phosphorylation. Liquid scintillation skeletal MLCK. Representative Lineweaver–Burk plot of (A)MLCK4and + spectrometry measurements showed that the maximum extent (B) skMLCK rates in the presence of EGTA (open circle) and Ca2 /CaM of phosphorylation was 2.6% of total kinase, and the calculated (filled circle).

Chang et al. PNAS | Published online June 20, 2016 | E3827 Downloaded by guest on October 2, 2021 ABN Catalytic Core C input unbound an-cMLCK LATRDWRDETVGTT Peptide antigen sequence 100% 15%

C D CKO WT 60 Ca2+/CaM - + - + WT-Ca2+/CaM WT-EGTA 40 CKO-Ca2+/CaM -140 CKO-EGTA

-115 P incorporated 20 32 IP cMLCK-

-80 pmol 0 0 20406080 -70 Time (minutes)

Fig. 4. Kinase activity of endogenous cMLCK immunoprecipitated from mouse hearts. (A) Illustration of cMLCK depicting the location of the peptide antigen sequence used to generate the custom antibody used in immunoprecipitations. C terminus of the catalytic core (gray bar) represents the autoregulatory segment. (B) Immunoblot of cMLCK detected from equivalent volumes of heart homogenate before (input) and after (unbound) immunoprecipitation. Quantified amount of cMLCK in unbound fraction is shown as percentage of input. Representative gel bands are shown, taken from the same immunoblot image. (C) Coomassie-stained gel of immunoprecipitated proteins from cMLCK-KO (CKO) and WT hearts. Proteins used to assay for kinase activity in the presence (+) or absence (−)ofCa2+/CaM are shown. Apparent molecular weights are noted. (D) Representative time course of 32P incorporated into purified cRLC in vitro by immunoprecipitated proteins shown in C.

kinase and a Km of 3.4 ± 0.4 μM. In EGTA, the Vmax value was the significant kinase activity in EGTA suggests that replacement of 0.7 ± 0.1 mol phosphate/min/mol kinase and Km was 4.4 ± the skMLCK autoregulatory segment with cardiac sequences partially 0.9 μM (Fig. 6A). The ATP Km value in the presence and absence activates the kinase, mimicking the constitutive activity of cMLCK. + of Ca2 /CaM was not significantly different, at 6.3 ± 0.5 μMin 2+ Discussion Ca /CaM and 6.2 ± 0.1 μM in EGTA (Fig. 6B). KCaM was 2.3 ± 0.4 nM for cMLCK (Fig. 6C), comparable to values obtained for Increased cRLC phosphorylation in vivo is associated with im- smooth and skeletal MLCK, which are absolutely dependent on proved cardiac performance and resistance to maladaptive hy- + Ca2 /CaM for activation (1). pertrophy (15, 27). Moreover, RLC phosphorylation is decreased in several models of heart failure, underscoring the importance + Low Ca2 /CaM-Dependent and Independent Activities of cMLCK Are of cRLC phosphorylation for normal cardiac function. With the Not Caused by the Autoregulatory Sequence. Chimeras of cMLCK exceptions of two recent reports (31, 32), ventricular cRLC is and skMLCK were designed to determine contributions of distinct reported to be phosphorylated in normal beating hearts, at ap- autoregulatory sequences to the unique properties of cMLCK proximately 0.40 mol phosphate/mol RLC when tissue extraction (Fig. 7). The cMLCK chimeras were generated whereby the minimizes phosphatase activity with reasonable recovery of total whole autoregulatory segment, or CaM-binding sequence only, myofilament proteins (6, 7, 11–13, 15–20, 24, 27, 33–39). We and were exchanged with that of skMLCK. In addition, skMLCK others have previously reported that cMLCK is the primary ki- chimeras with cardiac autoregulatory or CaM binding sequences nase responsible for RLC phosphorylation (8, 9, 40), but, in the were generated to determine effects of the cardiac sequences on conventional KO animal hearts, RLC phosphorylation at S15 is + the Ca2 /CaM-dependency of skMLCK. Chimeras were com- not completely abolished (6). Thus, another kinase is present in pared with WT cMLCK and skMLCK. the heart that phosphorylates RLC in the absence of cMLCK. Comparisons of Vmax and Km values of cMLCK chimeras vs. We know skMLCK is not expressed in the heart (21), and WT controls showed that the low activity of cMLCK did not smMLCK is not a kinase for cRLC (9). After careful evaluation increase with skMLCK sequences (Fig. 8 and Table 1). In ad- of commercially available antibodies toward MLCK4 (Fig. S1), dition, the presence of skMLCK autoregulatory sequences did we have found that MLCK4 is a soluble protein kinase signifi- + not affect the Ca2 /CaM-independent activity of cMLCK (Fig. cantly expressed in the heart, so we explored its biochemical 8). Comparisons of Vmax and Km values of skMLCK chimeras vs. properties. We solved a crystal structure of MLCK4, providing WT controls showed no effect of cardiac CaM binding sequence what is, to our knowledge, a first model for the MLCK family. on skMLCK activity, consistent with the comparable KCaM values Surprisingly, we found that a short regulatory helix is still present for cMLCK and skMLCK. Exchange of the whole autoregulatory in MLCK4. As a result of deletions in the linker region to helix segment of skMLCK for cardiac sequence caused a 100-fold αI and the first turn of the regulatory helix, this structural ele- decrease in Vmax, from 3,204 ± 283 mol phosphate/min/mol ki- ment forms a PRH that does not inhibit catalysis. Additionally, + nase. When only the CaM binding sequence was from cMLCK, there is no Ca2 /CaM binding sequence. Enzymatic assays con- the Vmax value was 29 ± 2 mol phosphate/min/mol kinase. The firmed that MLCK4 is indeed a constitutively active kinase with 2+ Km increased from 7.4 ± 2 μM with cardiac CaM-binding se- no Ca /CaM-dependency. The amount of MLCK4 expressed in quence only to 24 ± 5 μM with the whole cardiac autoregulatory cardiac myocytes is 20-fold less than cMLCK (6), but MLCK4 ’ segment. Exchange of the skMLCK s autoregulatory sequence has a Vmax value that is 50-fold higher than cMLCK in vitro + for that of cMLCK sequence caused the appearance of Ca2 /CaM- (Table 1). Thus, MLCK4 is potentially a significant contributor independent activity (Fig. 8). Because of the high Km value of to RLC phosphorylation in intact cells, indicating that MLCK4 2+ >200 μMinEGTA,theVmax value for Ca /CaM-independent ac- is a candidate kinase responsible for the residual 20–25% of the tivity could not be accurately measured by using cRLC. However, 0.4 mol phosphate/mol RLC in cMLCK-KO hearts. This hypothesis

E3828 | www.pnas.org/cgi/doi/10.1073/pnas.1600633113 Chang et al. Downloaded by guest on October 2, 2021 A D PNAS PLUS µL purified Ca2+/CaM Ca2+/CaM EGTA/CaM kinase pmol BSA no kinase + kinase + kinase 140 2 4 4 8 12 16 115 0.7 0.6 0.6 0.4 0.3 0.3 mol phos/mol RLC 80 pRLC * 70 RLC 50 1 2 5 1 2 5 1 2 5 µM myosin 40 30 E Ca2+/CaM EGTA 25 140 115 B 80 70 -kinase 2+ 50 100 Ca /CaM 40 EGTA 30 25 no substrate -RLC 15 Coomassie-Stained Gel 10 50 140 115 80 70 -kinase 50 40 30 0 25 0204060 -RLC 15 Time (min) Autoradiograph 10 C 1 2 3 4 5 6 7 0.4 F Ca2+/CaM 0.3 0.4 EGTA

P/min 0.2 32 0.2 P incorporated mol 0.1 32

0.0 pmol 0.0 1.4 1.5 1.6 1.7 0 50 100 150 Log( C) Time (minutes)

Fig. 5. Kinase activity of expressed cMLCK purified from Sf9 cells. (A) Coomassie-stained gel of purified GST-cMLCK (residues 447–795) with BSA loading

+ + BIOCHEMISTRY curve. (B) Kinase activity of purified GST-cMLCK in EGTA or Ca2 /CaM. (C) Kinase activity of purified GST-cMLCK in EGTA (open circle) or Ca2 /CaM (filled circle) after thermal denaturation. (D) Immunoblot for total cRLC after separation of proteins by Phos-tag-PAGE. Cardiac myosin purified from mouse ventricles was + phosphorylated by purified GST-cMLCK in vitro in EGTA or Ca2 /CaM. Numbers above bands are moles of phosphate incorporated per mole of RLC. (*Faster migration by phosphorylated RLC in EGTA reaction set is affected by EGTA in the manganese-Phos-tag acrylamide gel system.) (E) Phosphorylation of RLC in mouse cardiac myofibrils. Kinase assay mixtures of GST-cMLCK alone (lanes 1 and 4), myofibrils alone (lanes 2 and 5), or combined (lanes 3 and 6) were incubated in EGTA or Ca2+/CaM. Assay mixtures were separated by SDS/PAGE, then Coomassie-stained (Upper) and autoradiographed (Lower) to identify + phosphorylated proteins. (F) Representative time course of 32P incorporated into GST-cMLCK by autophosphorylation in EGTA (open circles) or Ca2 /CaM (filled circles).

needs to be tested, and potential regulatory mechanisms for ciently immunoprecipitates cMLCK directly from heart extract for MLCK4 remain to be investigated. Potential roles of MLCK4 in measurement of cMLCK for kinase activity. The immunoprecipi- noncardiac tissues also need exploration. tation of cMLCK provides purification to control for any nonspe- RLC is dephosphorylated in nonbeating hearts and rephos- cific protein kinase activity that may be comparable to the low phorylated with restoration of rhythmic contractions (6, 7, 41). Thus, cMLCK activity in total cell lysates. With both approaches, we + cMLCK activity may be regulated by cytosolic [Ca2 ], similar to the measured low cMLCK activity in EGTA that was stimulated by 2+ regulation of smMLCK and skMLCK activities (1). However, there Ca /CaM. The Km values for RLC and ATP were similar to those + are two conflicting reports on the Ca2 /CaM-dependency of cMLCK reported for skMLCK and smMLCK (3, 42–44). Additionally, the 2+ activity (8, 9). One study added no magnesium to form MgATP as KCaM value of the Ca /CaM stimulated activity of cMLCKwas also a kinase substrate, and thus addition of EGTA probably inhibited similar to values reported for skMLCK and smMLCK (3). The + + cMLCK activity as a result of chelation of all divalent metal ions primary difference in Ca2 /CaM-independent and Ca2 /dependent (9). The other study used RLC concentrations too low to accurately activities are in the Vmax values. We were careful to use RLC con- V 2+ assess max values (8). Whether cMLCK is Ca /CaM-dependent is centrations that were greater and less than the Km values to accu- important to resolve biochemically to identify and understand the rately determine these Vmax values (45). regulatory mechanisms for RLC phosphorylation in vivo. Comparison of the Coomassie-stained proteins that were We took two approaches to investigate whether cMLCK activity immunoprecipitated from WT and cMLCK-KO hearts did not + is Ca2 /CaM-dependent. First, we expressed and purified the ki- reveal any coimmunoprecipitated proteins associated specifically nase domain of cMLCK, which is comprised of the catalytic core with cMLCK. This result is consistent with our previous reports and autoregulatory segment. Control experiments confirmed ac- that cMLCK is soluble in heart homogenates, and that ectopically tivities measured were from the expressed kinase (Fig. S2). Kinase overexpressed GFP-tagged cMLCK diffuses out of detergent- assays were performed with purified RLC, intact cardiac myosin treated live adult cardiac myocytes (6, 12). Collectively, these and cardiac myofibrils. Second, we generated a custom antibody results suggest that cMLCK does not form a complex with an- raised to an epitope N-terminal of the catalytic core which effi- other protein in the heart that might affect its activity.

Chang et al. PNAS | Published online June 20, 2016 | E3829 Downloaded by guest on October 2, 2021 cMLCK was reported to be autophosphorylated (8), and we have also observed autophosphorylation. skMLCK is known to be rapidly autophosphorylated by an intramolecular mechanism, A but the phosphorylation does not affect its catalytic activity, and a 6 potential function remains unrecognized (13). Measurements of the low extent and rate of cMLCK autophosphorylation in vitro, and lack of autophosphorylation in the presence of its physiological substrate RLC, indicates that it is not likely to be a significant physiological regulatory mechanism. This conclusion does not ne- 4 gate regulation of phosphorylation from other protein kinases. 2+ The low kinase activity (i.e., Vmax value) and Ca /CaM- 2+

1/V independent activity that is stimulated by Ca /CaM are two properties that distinguish cMLCK from smMLCK and skMLCK 2 (Table 2). Unlike smooth and skeletal muscles with defined contraction and relaxation mechanisms, the heart beats contin- uously and never fully relaxes for any extended period. Based on + the apparent high affinity of Ca2 /CaM for cMLCK, the rate of dissociation of CaM from cMLCK is predicted to be slow (on the − order of sec 1) with diffusion-limited binding and activation (46). + -0.2 -0.1 0.0 0.1 0.2 0.3 In a beating heart, cMLCK may thus be saturated with Ca2 /CaM, with maximal activity. However, limited RLC phosphorylation in 1/[RLC] beating hearts suggest that the low activity of cMLCK is in balance with slow dephosphorylation by myosin light chain phosphatase activity, consistent with a slow RLC phosphate turnover rate (14). In B + 3 nonbeating hearts, cMLCK activity in the absence of Ca2 /CaM is not sufficient to exceed the phosphatase activity to maintain 40% + phosphorylation. Thus, stimulation by Ca2 /CaM allows fine tuning of RLC phosphorylation by cMLCK. A shared property between autoregulated MLCKs is the 2 presence of an autoinhibitory sequence N-terminal of the CaM binding sequence. We argued that, if cMLCK was constitutively partially autoinhibited, it would have low kinase rates and activity in 2+

1/rate the absence of Ca /CaM. This possibility was tested by comparing kinase rates of chimeric cMLCK and skMLCK, whereby the + 1 autoregulatory sequences were exchanged. The Ca2 /CaM- independent activity of cMLCK did not disappear when the autoregulatory sequence was changed to sequences of skMLCK, and the Vmax value did not increase, which suggests that the low kinase rate and constitutive activity is intrinsic to the catalytic core sequence. Interestingly, exchanging the autoregulatory se- -0.2 -0.1 0.0 0.1 0.2 quence of skMLCK for cardiac sequences caused an increase in 2+ V 1/[ATP] Ca /CaM-independent activity and a large decrease in max for the skMLCK catalytic domain. The distinct effects of exchanging the autoregulatory sequences of cMLCK and skMLCK could be C caused by differential destabilization of the catalytic core. In the + 30 absence of Ca2 /CaM, the autoinhibitory sequence of skMLCK may bind into the catalytic cleft, stabilizing the catalytic core in a closed conformation (5, 47). The cardiac autoregulatory sequence in the skMLCK chimera may not effectively inhibit the catalytic domain + in the absence of Ca2 /CaM. For both kinases, detailed struc- 20 tural information is necessary to determine the intermolecular interaction between the catalytic core and autoregulatory segment. In summary, the low constitutive activity of cMLCK stimulated 2+ 1/V by Ca /CaM is a distinct property that is not shared by other MLCK family members, and it appears to be intrinsic to the + 10 catalytic core sequence. This Ca2 /CaM-stimulated activity over + Ca2 /CaM-independent cMLCK appears to maintain RLC phosphorylation at 0.40 mol phosphate/mol RLC, which is nec- essary for normal cardiac performance. Additionally, MLCK4 may also contribute to the extent of RLC phosphorylation in -0.5 0.0 0.5 1.0 1.5 cardiac muscle. 1/[CaM] Materials and Methods Animals. All procedures were performed in accordance with the institutional Fig. 6. Measurement of the enzymatic properties of cMLCK. Line- animal care and use guidelines at University of Texas Southwestern Medical weaver–Burk plots of cMLCK rates in the presence of EGTA (open circle) Center, with all animal experimental procedures reviewed and approved by + and Ca2 /CaM (filled circle) with (A)RLC(inμM), (B)ATP(inμM), and the institutional animal care and use committee. Animals were housed under (C)CaM(innM). standard conditions in the rodent facility.

E3830 | www.pnas.org/cgi/doi/10.1073/pnas.1600633113 Chang et al. Downloaded by guest on October 2, 2021 Full length kinase N Catalyc Core C PNAS PLUS

Truncated kinase NCGST Catalyc Core

Autoregulatory sequence Chimeras Ca-Ca-Sk LKHEWLSHLPAKASGSNVRLRSQQLLQKYMAKRRWKKNFIAVSAANRFKKISSSGALMALGV Ca-Sk-Sk LAHPWLNNLAEKAKRCNRRLKSQILLKKYLMKRRWKKNFIAVSAANRFKKISSSGALMALGV Ca-Ca-Ca LKHEWLSHLPAKASGSNVRLRSQQLLQKYMAQSKWKKHFHVVTAVNRLRKFPTCP------Sk-Sk-Ca LAHPWLNNLAEKAKRCNRRLKSQILLKKYLMQSKWKKHFHVVTAVNRLRKFPTCP------Sk-Ca - Ca LKHEWLSHLPAKASGSNVRLRSQQLLQKYMAQSKWKKHFHVVTAVNRLRKFPTCP------Sk-Sk-Sk LAHPWLNNLAEKAKRCNRRLKSQILLKKYLMKRRWKKNFIAVSAANRFKKISSSGALMALGV MLCK4 LKHPWLSDKLHSRLSAQKKKNRGSDAQDFVTK------Calmodulin binding sequence

Fig. 7. Scheme of chimeric kinases designed to test contributions of the autoregulatory and CaM-binding sequences to the Ca2+/CaM-independent activity of cMLCK. Names of chimeras correspond to sequence origin for interchanged segments. Ca-Ca-Sk is truncated cMLCK with skMLCK CaM binding sequence; Ca- Sk-Sk is truncated cMLCK with skMLCK autoinhibitory and CaM binding sequence; Ca-Ca-Ca is WT cMLCK; Sk-Sk-Ca is truncated skMLCK with cMLCK CaM binding sequence; Sk-Ca-Ca is truncated skMLCK with cMLCK autoinhibitory and CaM binding sequence; and Sk-Sk-Sk is WT skMLCK.

Statistical Analyses. Data are expressed as mean ± SE. Statistical evaluation Immunoblot of Phosphorylated Myosin. Myosin was purified from mouse was carried out in GraphPad Prism using ANOVA with Dunnett’s posttest for ventricles by using low-salt precipitation steps at 4 °C, similar to the original comparison with control. Significance was accepted at a value of P < 0.05. protocol by Murakami and Uchida (52). Purified mouse cardiac myosin was phosphorylated in vitro with purified GST-cMLCK for 15 min at 30 °C. Quantification of MLCK4 Protein. Tissues from WT anesthetized mice were Reactions were terminated with addition of 10% trichloroacetic acid × homogenized in 30 volume of homogenization buffer (50 mM Tris, pH 8.0, containing 10 mM DTT. Precipitated protein was washed free of acid with 50 mM NaF, 1% Nonidet P-40, 2 mM EGTA, 0.1% sodium deoxycholate, 0.1% three 5-min washes in ethyl ether and resuspended by vigorous agitation Brij-35, 2× Halt Protease Inhibitor mixture, 10 μM E-64) and lysed on ice for 15 min, and the supernatant fraction was collected after centrifugation at 20,000 × g for 2 min. Adult cardiac myocytes and cardiac nonmuscle cells were isolated as previously described (6). Cells were lysed in the tissue ho- mogenization buffer. Protein concentration was determined by Bradford A assay, and 10 μg of total protein was boiled in 1× LDS Buffer (Invitrogen) 10000 2+ with reducing reagent (Invitrogen) and separated by 4–12% Bolt gradient Ca /CaM gel (Invitrogen). Separated proteins were immunoblotted for MLCK4 and EGTA GAPDH. Antibody to MLCK4 was from Abcam (Ab179395), and antibody to 1000

GAPDH was from Santa Cruz (sc25778). Antibody to cMLCK was previously BIOCHEMISTRY described (6). 100 Tissue Harvest and Preparation. Heart for immunohistochemistry was harvested from anesthetized mice and fixed via retrograde perfusion with 4% (wt/vol) paraformaldehyde, freshly prepared in PBS solution. Subsequent paraffin 10

processing, embedding, and sectioning were performed by standard pro- P/minute/mol kinase

cedures (48, 49). 32 1

Immunohistochemistry. Rabbit anti-sera used for MYLK4 immunolabeling of mol paraffin heart sections was obtained from Abcam (Ab179395). Following N.A. 0.1 N.A. deparaffinization and heat antigen retrieval with 10 mM Tris/1 mM EDTA, Ca-Ca-Sk Ca-Sk-Sk Ca-Ca-Ca Sk-Sk-Ca Sk-Ca-Ca Sk-Sk-Sk 0.05% Tween-20 (pH 9.0), sections were blocked against endogenous peroxidase activity and secondary antibody host-serum affinity. Serial sec- B tions were then subjected to primary antibody (1:33 dilution of commercially supplied stock) or normal rabbit serum and incubated overnight at 4 °C. 1000 Ca2+/CaM Subsequent biotin/streptavidin HRP detection of bound primary was con- EGTA ducted the following day according to previously described immunoperox- idase methods (50, 51).

Immunoprecipitation. Ventricles from WT or cMLCK-KO anesthetized mice 100 were rapidly frozen in liquid nitrogen and stored at −80 °C. Frozen ventricles were homogenized/thawed for 1 min by using a ground-glass homogenizer in 10× volume of homogenization buffer (50 mM Tris, pH 8.0, 50 mM NaF, 1% Nonidet P-40, 2 mM EGTA, 0.1% sodium deoxycholate, 0.1% Brij-35, 2× μ M cardiac RLC 10 Halt Protease Inhibitor mixture, 10 M E-64). Homogenates were lysed on ice µ for 15 min, and then supernatant fraction was collected after centrifugation at 20,000 × g for 2 min. Protein-A agarose (Thermo Fisher) prebound with a polyclonal antibody raised to a peptide N terminal to the catalytic core of N.A. N.A. mouse cMLCK, designed and produced by Genscript, was used to immuno- 1 precipitate endogenous cMLCK from the supernatant fraction. Antibody- Ca-Ca-Sk Ca-Sk-Sk Ca-Ca-Ca Sk-Sk-Ca Sk-Ca-Ca Sk-Sk-Sk bound beads were incubated with the supernatant fraction for 2 h, rocked

at 4 °C, then washed three times in PBS solution. Immunoprecipitated pro- Fig. 8. Comparison of chimera activities with cRLC. Comparison of (A) Vmax × 2+ teins were eluted by boiling in 1 LDS buffer (Invitrogen) with reducing values and (B)cRLCKm values in the presence of EGTA (gray) or Ca /CaM reagent (Invitrogen) and separated by 4–12% Bolt gradient gel (Invitrogen). (black). Bars are plotted on a logarithmic scale to accommodate values that Separated proteins were visualized by staining with Coomassie (Sigma). vary by multiple orders of magnitude.

Chang et al. PNAS | Published online June 20, 2016 | E3831 Downloaded by guest on October 2, 2021 Table 2. Kinetic parameters of skMLCK, smMLCK, and cMLCKs skRLC smRLC cRLC ATP

MLCK Km Vmax Vmax/Km Km Vmax Vmax/Km Km Vmax Vmax/Km Km KCaM

smMLCK 100 600 6 9 2,055 228 63 1,800 50 50 1 skMLCK 10 3,300 330 15 4,817 321 9 3,563 396 340 1 cMLCK 14 0.5 0.04 5 0.8 0.17 3 2.6 0.76 6 2 cMLCK (EGTA) 54 0.3 0.01 21 0.4 0.02 4 0.7 0.08 6 — MLCK4 14 42 3 8 134 16.75 8 129 14.44 ——

Average Km values for skeletal RLC (skRLC), smooth (smRLC), and cRLC (in μM) and ATP (in μM) and Vmax (in mol phosphate/min/mol kinase) and KCaM (in nM) values were obtained from double-reciprocal plots of at least three independent experiments for cMLCK and MLCK4. For cMLCK, Ca2+/CaM-dependent and independent (EGTA) values are shown. Assays were performed as described in Materials and Methods. Measured values are summarized in the + table without SE values. Ca2 /CaM-dependent values for smMLCK and skMLCK were previously published (3, 42–44).

in urea sample buffer (8 M urea, 20 mM Tris base, 23 mM glycine, 0.2 mM mixture; Promega). The cell suspension was then lysed by sonication, and the EDTA, 10 mM DTT) by using an orbital shaker (IKA Vibrax VXR) set at lysate was cleared by centrifugation in a Beckman JA-17 rotor at 36,000 × g 1,400 rpm for 30 min at room temperature. Complete denaturation and for 45 min. The supernatant fraction was mixed with 10 mL of 50% Ni-NTA solubilization was achieved by further addition of urea crystals and pro- slurry. The mixture was rotated for 1 h at 4 °C and then loaded on a column, μ longed agitation. Solubilized proteins were subjected to 30 M Phostag-10% which was washed with 150 mL of wash buffer (50 mM Hepes, 300 mM NaCl, SDS/PAGE after boiling in Laemmli buffer and transferred to PVDF (Immo- 5% glycerol, pH 7.5) and wash buffer with 25 mM imidazole. The protein bilon-P; Millipore). Proteins were fixed onto the PVDF membrane with 0.4% was eluted using 300 mM imidazole in wash buffer. The eluate of the nickel- glutaraldehyde/PBS solution for 15 min at room temperature. The mem- affinity column was diluted in 50 mM Hepes, 5% glycerol to achieve a final brane was then rinsed three times in PBS solution and immunoblotted with NaCl concentration of 150 mM and loaded onto an anion-exchange HiTrap antibody to cRLC (Enzo). Q FF column. The protein was eluted by using a liner gradient ranging from 150 mM to 2 M NaCl in 50 mM Hepes, 5% glycerol, pH 7.5. The fractions Kinase Activity Assay. Kinase activities were assayed in 10 mM Mops, pH 7.4, containing recombinant MLCK4 were pooled, concentrated, and applied to 5 mM MgCl , 100 mM NaCl, 0.3 mM CaCl or 3 mM EGTA, 1.5 μM CaM, 1 mM 2 2 a Superdex 75 GF column equilibrated in GF Buffer (50 mM Hepes, 300 mM DTT, and 0.2 mM [γ-32P]ATP (100–300 cpm/pmol) with purified proteins in NaCl, 5% glycerol, pH 7.5). The eluted protein was more than 95% pure as 40 μL total volume. Reaction mixtures were preincubated for 5 min, and the kinase activity was measured at 30 °C by the addition of [γ-32P]ATP judged by SDS/PAGE. Liquid chromatography electrospray ionization MS (time-of-flight) revealed the expected mass of the protein (42,433 Da) as as described previously (53). For measurement of ATP Km values, 0.1 μM cMLCK was assayed with 15 μM RLC in 1–150 μM[γ-32P]ATP. For measure- predicted from the expressed sequence after TEV cleavage. This purified kinase was used for kinase assays herein. ment of RLC Km and Vmax values, 1–50 nM kinase (empirically determined for each kinase used) was assayed with 0.1–120 μM RLC. Km and Vmax values Crystals were grown at 4 °C in 300-nL sitting drops from a 2:1 ratio of were calculated by nonlinear fit to the Michaelis–Menten equation or linear protein (10 mg/mL) to reservoir solution containing 2 M ammonium sulfate fit to Lineweaver–Burk plots by using GraphPad Prism 6.0 software. Kinase and 2.5 wt/vol PEG400 in Hepes buffer (50 mM), pH 8.0. The ATP mimetic concentrations were verified by silver stain of assay mixtures. For measure- inhibitor [4-(2 amino-4-methyl-1,3-thiazol-5-yl)-n-(3-dioxaziridin-3-yl phenyl)pyrim- 2+ ment of Ca /CaM required for half-maximal activation of cMLCK (KCaM idine-2-amine] was added to the concentrated protein from a 50-mM DMSO values), assays were performed at 25 °C for 4–8 h with 0.5 nM cMLCK to stock solution. For data collection, crystals were cryoprotected by using the measure the rate of cMLCK activity at a concentration below the measured well solution supplemented with 2 M Li2SO4 and flash-frozen in liquid ni- KCaM value (46). trogen. Diffraction data were collected from a single crystal on Diamond beamline IO2 at a single wavelength of 0.9802 Å, and the structure was Expression and Purification of Kinases. All kinases were expressed in Sf9 cells refined to 2.8 Å. Indexing and integration was carried out by using MOSFLM and affinity-purified. Cells were lysed for 20 min in 20 mM Mops, pH 7.4, (54), and scaling was performed with SCALA (55). The structure was solved 0.5 mM EGTA, 1% Nonidet P-40, 1 mM DTT, 1× Halt Protease Inhibitor mixture by molecular replacement (PHASER) using the structure of Protein Data Bank (Pierce), and 10 μM E-64. Lysates were centrifuged for 30 min at 20,000 × g (PDB) ID code 1KOB as a starting model (56). COOT was used for model (Beckman Coulter), and supernatant fractions were collected for purification building, and refinement was carried out in REFMAC5 (57, 58). Thermal procedures. GST-tagged cMLCK and chimeras were purified with glutathione- motions were analyzed using TLSMD, and defined domains were used in agarose (ThermoFisher) and CaM-Sepharose (GE Healthcare) by using proce- dures found in the instruction manual. skMLCK was a gift from Kathy Trybus, later cycles of refinement. The coordinates have been deposited in the PDB University of Vermont, Burlington, VT. with the accession code 2X4F.

Crystallization of MLCK4. MYLK4 (residues K40–K388) in frame with an ACKNOWLEDGMENTS. We thank John Shelton for help with immunohisto- chemistry images of MLCK4 expression in mouse hearts and Kathy Trybus for N-terminal His tag and TEV protease site (MGHHHHHHSSGVDLGTENLY the generous gift of purified skMLCK. This work was supported by FQ^SM) was expressed in Sf9 cells at 27 °C. After 48 h infection by the virus, an American Heart Association Postdoctoral Fellowship (to A.N.C.), the the cells were harvested by centrifugation, washed once with PBS solution, Leducq Foundation (H.L.S.), NIH Grant HL080536 (to J.T.S.), the Moss Heart and resuspended in 3 three volumes of lysis buffer (50 mM Hepes, 300 mM Fund (J.T.S.), and the Fouad A. and Val ImmBashour Distinguished Chair in NaCl, 5% glycerol, pH 7.5, 1 mM TCEP, 1:1,000 dilution of protease inhibitor Physiology (J.T.S.).

1. Kamm KE, Stull JT (2011) Signaling to myosin regulatory light chain in sarcomeres. JBiol 6. Chang AN, et al. (2015) Constitutive phosphorylation of cardiac myosin regulatory light Chem 286(12):9941–9947. chain in vivo. JBiolChem290(17):10703–10716. 2. Stull JT, Nunnally MH, Michnoff CH (1986) Calmodulin-dependent protein kinases. The 7. High CW, Stull JT (1980) Phosphorylation of myosin in perfused rabbit and rat hearts. Am J , Control by Phosphorylation Part A (Academic, London), 3rd revised Ed, pp 113–166. Physiol 239(6):H756–H764. 3. Gallagher PJ, Herring BP, Trafny A, Sowadski J, Stull JT (1993) A molecular mechanism for 8. Chan JY, et al. (2008) Identification of cardiac-specific myosin light chain kinase. Circ Res autoinhibition of myosin light chain kinases. J Biol Chem 268(35):26578–26582. 102(5):571–580. 4. Krueger JK, et al. (1997) Structures of calmodulin and a functional myosin light chain 9. Seguchi O, et al. (2007) A cardiac myosin light chain kinase regulates sarcomere assembly in kinase in the activated complex: A neutron scattering study. Biochemistry 36(20): the vertebrate heart. J Clin Invest 117(10):2812–2824. 6017–6023. 10. Akiyama K, et al. (1997) Myocardial infarction and regulatory myosin light chain. J Mol Cell 5. Krueger JK, Zhi G, Stull JT, Trewhella J (1998) Neutron-scattering studies reveal fur- Cardiol 29(10):2641–2652. ther details of the Ca2+/calmodulin-dependent activation mechanism of myosin light 11. Chang AN, et al. (2013) The effects of neuregulin on cardiac Myosin light chain kinase gene- chain kinase. Biochemistry 37(40):13997–14004. ablated hearts. PLoS One 8(6):e66720.

E3832 | www.pnas.org/cgi/doi/10.1073/pnas.1600633113 Chang et al. Downloaded by guest on October 2, 2021 12. Ding P, et al. (2010) Cardiac myosin light chain kinase is necessary for myosin regu- heart ventricle of the European hamster during hibernation and in summer. J Muscle PNAS PLUS latory light chain phosphorylation and cardiac performance in vivo. J Biol Chem Res Cell Motil 13(1):64–70. 285(52):40819–40829. 38. Morano I, Osterman A, Arner A (1995) Rate of active tension development from rigor 13. Fitzsimons DP, Bodell PW, Baldwin KM (1990) Myocardial functional correlates of in skinned atrial and ventricular cardiac fibres from swine following photolytic release cardiac myosin light chain 2 phosphorylation. J Appl Physiol 68(6):2426–2433. of ATP from caged ATP. Acta Physiol Scand 154(3):343–353. 14. Herring BP, England PJ (1986) The turnover of phosphate bound to myosin light 39. Wei B, Wei H, Jin JP (2015) Dysferlin deficiency blunts β-adrenergic-dependent lusitropic chain-2 in perfused rat heart. Biochem J 240(1):205–214. function of mouse heart. J Physiol 593(23):5127–5144. 15. Huang J, Shelton JM, Richardson JA, Kamm KE, Stull JT (2008) Myosin regulatory 40. Ding HL, Ryder JW, Stull JT, Kamm KE (2009) Signaling processes for initiating smooth light chain phosphorylation attenuates cardiac hypertrophy. JBiolChem283(28): muscle contraction upon neural stimulation. J Biol Chem 284(23):15541–15548. 19748–19756. 41. Silver PJ, Buja LM, Stull JT (1986) Frequency-dependent myosin light chain phos- 16. Klaiman JM, Fenna AJ, Shiels HA, Macri J, Gillis TE (2011) Cardiac remodeling in fish: phorylation in isolated myocardium. J Mol Cell Cardiol 18(1):31–37. Strategies to maintain heart function during temperature change. PLoS One 6(9): 42. Nunnally MH, Rybicki SB, Stull JT (1985) Characterization of chicken skeletal muscle e24464. myosin light chain kinase. Evidence for muscle-specific isozymes. J Biol Chem 260(2): 17. Morano I (1992) Effects of different expression and posttranslational modifications of 1020–1026. myosin light chains on contractility of skinned human cardiac fibers. Basic Res Cardiol 43. Padre RC, Stull JT (2000) Functional assembly of fragments from bisected smooth – 87(Suppl 1):129 141. muscle myosin light chain kinase. J Biol Chem 275(35):26665–26673. 18. Sheikh F, et al. (2012) Mouse and computational models link Mlc2v dephosphorylation to 44. Stull JT, Nunnally MH, Moore RL, Blumenthal DK (1985) Myosin light chain kinases – altered myosin kinetics in early cardiac disease. J Clin Invest 122(4):1209 1221. and myosin phosphorylation in skeletal muscle. Adv Regul 23:123–140. 19. Taniguchi M, et al. (2015) New isoform of cardiac myosin light chain kinase and the 45. Segel IH (1993) : Behavior and Analysis of Rapid Equilibrium and α role of cardiac myosin phosphorylation in 1-adrenoceptor mediated inotropic re- Steady-State Enzyme Systems (Wiley, New York). sponse. PLoS One 10(10):e0141130. 46. Blumenthal DK, Stull JT (1980) Activation of skeletal muscle myosin light chain kinase 20. Toepfer C, et al. (2013) Myosin regulatory light chain (RLC) phosphorylation change as a by calcium(2+) and calmodulin. Biochemistry 19(24):5608–5614. – modulator of cardiac muscle contraction in disease. J Biol Chem 288(19):13446 13454. 47. Krueger JK, et al. (2001) Activation of myosin light chain kinase requires translocation 21. Zhi G, et al. (2005) Myosin light chain kinase and myosin phosphorylation effect of bound calmodulin. J Biol Chem 276(7):4535–4538. frequency-dependent potentiation of skeletal muscle contraction. Proc Natl Acad Sci 48. Sheehan DC, Hrapchak BB (1980) Theory and Practice of Histotechnology (Mosby, USA 102(48):17519–17524. St. Louis). 22. Morano I (1999) Tuning the human heart molecular motors by myosin light chains. 49. Woods AE, Ellis RC (1994) Laboratory Histopathology: A Complete Reference J Mol Med (Berl) 77(7):544–555. (Churchill Livingstone, London). 23. van der Velden J, et al. (2003) The effect of myosin light chain 2 dephosphorylation 50. Borvak J, et al. (1998) Functional expression of the MHC class I-related receptor, FcRn, on Ca2+ -sensitivity of force is enhanced in failing human hearts. Cardiovasc Res 57(2): in endothelial cells of mice. Int Immunol 10(9):1289–1298. 505–514. 51. Cianga P, Medesan C, Richardson JA, Ghetie V, Ward ES (1999) Identification and 24. van der Velden J, et al. (2003) Increased Ca2+-sensitivity of the contractile apparatus function of neonatal Fc receptor in mammary gland of lactating mice. Eur J Immunol in end-stage human heart failure results from altered phosphorylation of contractile 29(8):2515–2523. proteins. Cardiovasc Res 57(1):37–47. 52. Murakami U, Uchida K (1985) Contents of myofibrillar proteins in cardiac, skeletal, 25. Riise J, et al. (2012) Prostanoid-mediated inotropic responses are attenuated in failing and smooth muscles. J Biochem 98(1):187–197. human and rat ventricular myocardium. Eur J Pharmacol 686(1-3):66–73. 53. Chang AN, Chen G, Gerard RD, Kamm KE, Stull JT (2010) Cardiac myosin is a substrate 26. AitMou Y, et al. (2011) Beneficial effects of SR33805 in failing myocardium. for zipper-interacting protein kinase (ZIPK). J Biol Chem 285(8):5122–5126. Cardiovasc Res 91(3):412–419. 54. Leslie AGW, Powell HR (2007) Processing diffraction data with mosflm. Evolving 27. Warren SA, et al. (2012) Myosin light chain phosphorylation is critical for adaptation Methods for Macromolecular Crystallography, NATO Science Series, eds Read RJ, to cardiac stress. Circulation 126(22):2575–2588. – 28. Yuan CC, et al. (2015) Constitutive phosphorylation of cardiac myosin regulatory light Sussman JL (Springer, Dordrecht, The Netherlands), Vol 245, pp 41 51. chain prevents development of hypertrophic cardiomyopathy in mice. Proc Natl Acad 55. Evans P (2007) SCALA-Scale Together Multiple Observations of Reflections (MRC Sci USA 112(30):E4138–E4146. Laboratory of Molecular Biology, Cambridge, UK), 3.3.0 Ed. 29. Sheikh F, Lyon RC, Chen J (2015) Functions of myosin light chain-2 (MYL2) in cardiac 56. McCoy AJ, Grosse-Kunstleve RW, Storoni LC, Read RJ (2005) Likelihood-enhanced fast

– BIOCHEMISTRY muscle and disease. Gene 569(1):14–20. translation functions. Acta Crystallogr D Biol Crystallogr 61(Pt 4):458 464. 30. Rellos P, et al. (2010) Structure of the CaMKIIdelta/calmodulin complex reveals the 57. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta – molecular mechanism of CaMKII kinase activation. PLoS Biol 8(7):e1000426. Crystallogr D Biol Crystallogr 60(pt 12 pt 1):2126 2132. 31. Kampourakis T, Irving M (2015) Phosphorylation of myosin regulatory light chain 58. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular struc- controls myosin head conformation in cardiac muscle. J Mol Cell Cardiol 85:199–206. tures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 53(pt 3): 32. Peng Y, et al. (2014) Top-down proteomics reveals concerted reductions in myofila- 240–255. ment and Z-disc protein phosphorylation after acute myocardial infarction. Mol Cell 59. Bogomolovas J, et al. (2014) kinase is an inactive pseudokinase scaffold that Proteomics 13(10):2752–2764. supports MuRF1 recruitment to the sarcomeric M-line. Open Biol 4(5):140041. 33. Fewell JG, et al. (1997) A treadmill exercise regimen for identifying cardiovascular 60. Shymanets A, Ahmadian MR, Nürnberg B (2009) Gbetagamma-copurified lipid kinase phenotypes in transgenic mice. Am J Physiol 273(3 pt 2):H1595–H1605. impurity from Sf9 cells. Protein Pept Lett 16(9):1053–1056. 34. Fitzsimons DP, Bodell PW, Baldwin KM (1989) Phosphorylation of rodent cardiac 61. Herring BP, Gallagher PJ, Stull JT (1992) Substrate specificity of myosin light chain myosin light chain 2: Effects of exercise. J Appl Physiol 67(6):2447–2453. kinases. J Biol Chem 267(36):25945–25950. 35. Holroyde MJ, Small DA, Howe E, Solaro RJ (1979) Isolation of cardiac myofibrils and 62. Zhi G, Abdullah SM, Stull JT (1998) Regulatory segments of Ca2+/calmodulin-dependent myosin light chains with in vivo levels of light chain phosphorylation. Biochim Biophys protein kinases. J Biol Chem 273(15):8951–8957. Acta 587(4):628–637. 63. Zhi G, Herring BP, Stull JT (1994) Structural requirements for phosphorylation of 36. Mizutani H, et al. (2010) Overexpression of myosin phosphatase reduces Ca(2+) sen- myosin regulatory light chain from smooth muscle. J Biol Chem 269(40):24723–24727. sitivity of contraction and impairs cardiac function. Circ J 74(1):120–128. 64. Ryder JW, Lau KS, Kamm KE, Stull JT (2007) Enhanced skeletal muscle contraction 37. Morano I, Adler K, Agostini B, Hasselbach W (1992) Expression of myosin heavy and with myosin light chain phosphorylation by a calmodulin-sensing kinase. J Biol Chem light chains and phosphorylation of the phosphorylatable myosin light chain in the 282(28):20447–20454.

Chang et al. PNAS | Published online June 20, 2016 | E3833 Downloaded by guest on October 2, 2021