Site-specific Microtubule-associated Protein 4 Dephosphorylation Causes Microtubule Network Densification in Pressure Overload Cardiac Hypertrophy*

In severe pressure overload-induced cardiac hypertrophy, a dense, stabilized microtubule network forms that interferes with cardiocyte contraction and microtubule-based transport. This is associated with persistent transcriptional up-regulation of cardiac α- and β-tubulin and microtubule-stabilizing microtubule-associated protein 4 (MAP4). There is also extensive microtubule decoration by MAP4, suggesting greater MAP4 affinity for microtubules. Because the major determinant of this affinity is site-specific MAP4 dephosphorylation, we characterized this in hypertrophied myocardium and then assessed the functional significance of each dephosphorylation site found by mimicking it in normal cardiocytes. We first isolated MAP4 from normal and pressure overload-hypertrophied feline myocardium; volume-overloaded myocardium, which has an equal degree and duration of hypertrophy but normal functional and cytoskeletal properties, served as a control for any nonspecific growth-related effects. After cloning cDNA-encoding feline MAP4 and obtaining its deduced amino acid sequence, we characterized by mass spectrometry any site-specific MAP4 dephosphorylation. Solely in pressure overload-hypertrophied myocardium, we identified striking MAP4 dephosphorylation at Ser-472 in the MAP4 N-terminal projection domain and at Ser-924 and Ser-1056 in the assembly-promoting region of the C-terminal microtubule-binding domain. Site-directed mutagenesis of MAP4 cDNA was then used to switch each serine to non-phosphorylatable alanine. Wild-type and mutated cDNAs were used to construct adenoviruses; microtubule network density, stability, and MAP4 decoration were assessed in normal cardiocytes following an equivalent level of MAP4 expression. The Ser-924 → Ala MAP4 mutant produced a microtubule phenotype indistinguishable from that seen in pressure overload hypertrophy, such that Ser-924 MAP4 dephosphorylation during pressure overload hypertrophy may be central to this cytoskeletal abnormality.

Although many important alterations have been described in the properties of hypertrophied myocardium, the mechanisms responsible for contractile dysfunction and many other maladaptive changes of cardiac muscle cells, or cardiocytes, have yet to be fully defined. Although most research in this area has focused on structural and regulatory changes within the myofilament, it has also been found that changes in the microtubule component of the extra-myofilament cytoskeleton may lead both to contractile dysfunction by increasing the internal resistance to sarcomere motion (1,2) and to disordered cellular homeostasis by impeding cytoskeleton-based intracellular transport (3)(4)(5)(6).
Our major original finding was that, in severe pressure overload cardiac hypertrophy with increased ventricular wall stress, there is the early appearance and then the persistence of a dense microtubule network and associated contractile dysfunction (7,8). We have now found several synergistic bases for this dense microtubule network. First, during hypertrophy there is persistent transcriptional up-regulation of ␣-tubulin (9), ␤1-tubulin, i.e. the ␤-tubulin isoform that is otherwise expressed only during cardiac development (10), and MAP4, 3 the major cardiac microtubule-binding and -stabilizing microtubule-associated protein (11). Second, microtubules in hypertrophied cardiocytes are heavily decorated by MAP4 and thus greatly stabilized when compared with those in control cells (11). Third, overexpression of cardiocyte MAP4 via adenoviral infection and cardiac MAP4 via transgenesis are both associated with striking increases in ␣and ␤-tubulin and the generation of a dense, stabilized microtubule network that is structurally and functionally indistinguishable from that found in severe pressure overload cardiac hypertrophy (12).
These data would suggest that the microtubule network changes that occur in hypertrophied myocardium are caused by transcriptionally driven increases in the relevant structural proteins. However, an early hint that this view might not be sufficient to be a full explanation was given by our finding that ␤-tubulin mRNA stability is unchanged in hypertrophied myocardium despite a major increase in tubulin protein levels (see Fig. 5 in Ref. 10) and despite our finding that the co-translational negative feedback control, which tubulin exerts on its own rate of synthesis via reduced mRNA stability (13,14), is intact in normal and hypertrophied cardiocytes in vitro and in myocardium in vivo (10). A more recent but similar hint was given by our finding that, even cardiac-restricted ␤1-tubulin overexpression, despite being driven by the strong ␣-myosin heavy chain promoter, does not cause an increase in myocardial ␤-tubulin protein levels (see Fig. 3 in Ref. 15). This too is presumably due at least in part to the tubulin co-translational control mechanism, because we find that the level of cardiac ␤-tubulin mRNA is ϳ4.5-fold greater in these transgenic mice than it is in wild-type mice.
However, given that dephosphorylation within the microtubule-binding domain of several MAPs is the major determinant of greater MAP-microtubule affinity (16), a transcriptionally driven up-regulation of the tubulins and MAP4 need not be the primary cause of this cardiac cytoskeletal pathology. Rather, it might instead be secondary, via negative feedback control (17), to the hyperstabilized microtubules and their associated MAP4 being effectively isolated from their ordinarily dynamic intracellular pools. That is, what we are hypothesizing here is the possibility that the primary determinant of the microtubule network changes seen in hypertrophied myocardium could be persistent, site-specific MAP4 dephosphorylation. The findings of the present study tend to support this hypothesis.

EXPERIMENTAL PROCEDURES
Right Ventricular Pressure and Volume Overloading-Pressure overload hypertrophy of the feline right ventricle (RV) without associated RV failure was created as before (18,19) by placement of a 3.2-mm inner diameter pulmonary artery band (PAB) in adult cats of random sex weighing 2.0 -3.4 kg. Volume overload hypertrophy of the feline RV was created as before (20) by creating an atrial septal defect (ASD) via the Blalock-Hanlon procedure (21). The normally loaded left ventricle (LV) served as a same-animal control in all cases. Because the RV mass increase stabilizes by ϳ4 weeks after a step increase in load (9), at 2 or 4 weeks after surgery intravascular pressures were measured in these and in normal control cats; values in the systemic circulation were the same for all groups. All operative procedures were carried out under full surgical anesthesia; all procedures and the care of the cats were in accordance with institutional guidelines.
Isolation of Myocardial MAP4-The goal here was to isolate MAP4 from fresh tissue samples with its native state of phosphorylation intact. We took advantage of the thermostability of MAP4 to heat-denature the great bulk of non-MAP4 myocardial proteins. The remaining soluble proteins, which included MAP4, were then concentrated and separated. Thus, immediately after the above physiological measurements were complete, the heart was extirpated under deep anesthesia, and separate 3-to 4-g samples of the RV and LV from each heart were immediately frozen in liquid nitrogen and pulverized in the frozen state. The powdered frozen tissue was brought immediately to 100°C in a relatively large volume (1 g of tissue/10 ml) of boiling high salt buffer (350 mM NaCl, 100 mM Tris, pH 7.4, 10 mM EGTA, 2 mM dithiothreitol) for 5 min; after cooling on ice, the suspension was homogenized with a Polytron homogenizer at maximum speed for 1 min. Insoluble material was removed by centrifugation at 37,000 ϫ g for 1 h. The clear supernatant was concentrated using an ultrafiltration membrane (Amicon Centriprep-50; molecular mass ϭ 50 kDa) followed by dialysis for 24 h in MEM buffer (20 mM MES, pH 6.8, 1 mM EGTA, 0.5 mM NaCl, 1 mM dithiothreitol) containing 1 M (NH 4 ) 2 SO 4 and centrifuged at 7,000 ϫ g for 30 min to remove insoluble material. The clarified concentrate was then fractionated via FPLC on a hydrophobic column (ToyoScreen Butyl-650M 5-ml column, Tosoh Bioscience) equilibrated with the same MEM/1 M (NH 4 ) 2 SO 4 buffer. After washing the column in 10 ml of the MEM/(NH 4 ) 2 SO 4 buffer, adsorbed materials were eluted with a decreasing linear gradient of (NH 4 ) 2 SO 4 (from 1 to 0 M, 42-ml total volume) in the MEM buffer, followed by an additional 20 ml of the MEM buffer without (NH 4 ) 2 SO 4 . About seventy 1.0-ml fractions were collected at a flow rate of 0.5 ml/min. All of the fractions were screened for MAP4 via Western blot analysis using our MAP4 antibody (12); peak concentration of the 220-and 350-kDa MAP4 isoform fractions was found to be in FPLC fractions 51-57 (Fig. 1A). The peak concentration of the very abundant and potentially interfering 250-kDa ␤-myosin heavy chain was found via Western blotting to be in FPLC fractions Ն 59 (Fig. 1A). FPLC fractions 50 -60 were subjected to SDS-PAGE using a pre-cast 3-8% gradient Tris acetate gel (NuPAGE, Invitrogen). To achieve clear separation of the proteins, electrophoresis was continued for 2.5 h at 130 volts after the dye front had run off of the gel. Proteins were visualized  (12), showing the muscle-specific 350-kDa mMAP4 and ubiquitous 220-kDa uMAP4 isoforms. Lower blot, immunoblot of myosin heavy chain in the same FPLC fractions, made using a monoclonal antibody (clone BA-D5, American Type Culture Collection). B, separation of FPLC MAP4 fractions on a 3-8% Tris acetate SDS-PAGE gel stained with Simply Blue Safe Stain (Invitrogen). The position of the 220-kDa uMAP4 isoform is indicated. In-gel Digestion-The excised gel plugs were pooled, placed in an Eppendorf tube, washed with 50 mM NH 4 CO 3 for 10 min, and de-stained twice using 25 mM NH 4 CO 3 in 50% acetonitrile for 15 min. The plugs were then dehydrated with 100% acetonitrile for 15 min, dried in a SpeedVac, covered with proteomics-grade trypsin (Sigma), and incubated at 37°C overnight. The supernatant was collected in a clean, dry Eppendorf tube. Peptides were further extracted with 1 wash of 25 mM NH 4 CO 3 for 20 min and 3 washes of 5% formic acid, 50% acetonitrile for 20 min each. The supernatants were collected and pooled after each wash, then dried down in a SpeedVac to ϳ1 l.
Sequencing of Feline MAP4-Total RNA was extracted from feline myocardium using TRIzol (Invitrogen) according to the manufacturer's protocol. Total RNA (2 g) was reverse transcribed using the SuperScript III first-strand synthesis system in the presence of oligo(dT) primer and Moloney-murine leu-kemia virus reverse transcriptase (Invitrogen). The cDNA synthesis was carried out at 55°C, providing increased specificity, higher yields of cDNA, and more full-length product. PCR was performed using primers based on the human gene sequence for the uMAP4 isoform (24). A forward primer, 5Ј-ATG-GCTGACCTCAGTCTTGCAG-3Ј, and a reverse primer, 5Ј-TTAG-ATGCTTGTCTCCTGGATCT-3Ј, were used for synthesizing a full-length cDNA with Platinum TaqDNA Polymerase (Invitrogen). The PCR profile was as follows: initial denaturation at 94°C for 5 min, denaturation at 94°C for 30 s, annealing at 60°C for 4 min, and extension at 72°C for 7 min for a total of 35 cycles. A cDNA of the expected size (3.5 kb) for the uMAP4 isoform was identified, ligated into the pTOPO vector (Invitrogen), and sequenced bidirectionally. Sequencing (supplemental Fig. S1 and GenBank TM accession number EU921827.1) revealed the presence of a single open reading frame of 3408 nucleotides that encoded 1135 amino acid residues. At the nucleotide level, this had a 62% identity with murine MAP4 (24), a 76% identity with human MAP4 (24), and a 78% identity with bovine MAP4 (24,25). At the amino acid level, feline MAP4 had a 56%, 72%, and 66% identity with murine, human, and bovine MAP4, respectively (24,25). Referring to Fig. 2A and the legend, the closest approach to identity occurs in the extreme N terminus, in a region adjacent to the C-terminal end of the KDM repeats, in the C-terminal microtubule-binding domain, and in the extreme C-terminal acidic tail.
Mass Spectroscopic Analysis of MAP4 Phosphorylation-Using the derived amino acid sequence of the feline 220-kDa uMAP4 isoform given in supplemental Fig. S1, mass spectrometry methods that we had developed for the quantitative analysis of G protein-coupled receptor phosphorylation (26 -28) were applied here to the analysis of MAP4 phosphorylation. The gel-extracted, FPLC-purified cardiac MAP4 samples were dried, resolubilized in 1 l of 0.2% formic acid, and separated by nanoflow reversed-phase capillary HPLC (LC Packings). Buffer A was 0.2% formic acid in 2% acetonitrile, and buffer B was 0.2% formic acid in 98% acetonitrile. The samples were loaded on a loop and transferred to a capillary column (Micro-Tech Scientific, 75 m ϫ 15 cm, C18, 3 m, 100 Å) for 9 min. The peptides were eluted over a 60-min gradient from 0 to 60% buffer B at a flow rate of 180 nl/min; no trap column or desalting was utilized. The column effluent was directed into the nanospray source of a Finnigan LTQ XL linear ion-trap mass spectrometer The microtubule-binding domain (amino acids 671-1135) is also divided into three subdomains: a region rich in both proline and basic residues (P-rich region), an assembly-promoting (AP) region consisting of five 18-amino acid imperfect repeats containing a KXGS motif (amino acids 918 -1068), and the hydrophobic C-terminal tail. B, MAP4 levels in normal and hypertrophied myocardium. The three immunoblots were prepared using our polyclonal anti-MAP4 antibody (12) from the LV and RV of a normal control cat, a cat at 2 weeks, and a cat at 4 weeks after RV pressure overloading via a PAB, and a cat at 4 weeks after RV volume overloading via an ASD. For each cat, an equal amount of protein was loaded for the LV and RV lanes. The uMAP4 isoform ran at 220 kDa. The mMAP4 isoform, which has a single 3.2-kb coding region insertion within the projection domain of uMAP4 between the "b region" and the "P-rich region," ran at 350 kDa. Hypertrophyrelated increases in MAP4 expression were confined to the uMAP4 isoform and were seen only in the pressureoverloaded RV of the PAB cats: for three experiments such as that shown here, the densitometric ratio of RV/LV uMAP4 was 1.04 Ϯ 0.22 for "Control," 1.82 Ϯ 0.26 for "2 weeks PAB," 2.20 Ϯ 0.31 for "4 weeks PAB," and 0.95 Ϯ 0.33 for "4 weeks ASD." For the mMAP4 isoform, these RV/LV values were 1.01 Ϯ 0.02 for "Control," 1.10 Ϯ 0.04 for "2 weeks PAB," 1.15 Ϯ 0.01 for "4 weeks PAB," and 1.17 Ϯ 0.04 for "4 weeks ASD." (Thermo-Finnigan Instrument Systems) through a pre-cut Sili-caTip for the Finnigan nanospray ion source (5 cm long, 360-m optical density, and 5-m inner diameter, New Objective, Woburn, MA). Data-dependent analysis was used to perform one MS and five MS/MS scans automatically in each cycle, with MS/MS/MS being triggered by the loss of phosphate in the MS/MS scan. This method was carried out using Xcalibur software (version 1.4 SR2, Finnigan) with dynamic exclusion enabled (repeat count of 2). After MS/MS had been acquired on an ion twice, it was excluded from sequencing events for 180 s. The collected data were analyzed with the aid of the TurboSequest unit of the Bioworks 3.3 software (Finnigan), taking into account two possible missed trypsin cleavages per fragment. Peptide identifications with a cross-correlation score versus charge state value Ն1.0, 2.0, and 2.5 for ϩ1, ϩ2, and ϩ3 ions and probability score Ͻ0.1 values were accepted automatically.
Site-directed Mutagenesis of MAP4 and Adenovirus Construction-Based on mass spectroscopic identification of MAP4 dephosphorylation at Ser-472, Ser-924, and Ser-1056 in the pressure-overloaded RVs of PAB cats, primers were designed against our full-length N-terminal Myc-tagged feline MAP4 wild-type cDNA to generate the desired point mutations of MAP4 S472A, S924A, S1056A, and S924D (Table 1) through PCR reactions using the QuikChange site-directed mutagenesis kit (Stratagene). The entire cDNA sequence was confirmed for each of the MAP4 mutant constructs. The wild-type and mutated MAP4 cDNAs were then cloned into the Dual GFP-CCM adenoviral shuttle vector from Vector Biolabs and sent to them for the generation of adenovirus. This vector expresses enhanced GFP under one cytomegalovirus promoter and expresses the gene insert of interest using a second cytomegalovirus promoter. Adult cardiocytes were then infected on day 1 in culture by adding titered adenovirus to the culture medium at different multiplicities of infection (m.o.i.) for 8 h. Based on GFP expression, infection with an m.o.i. of one resulted in gene transfer to Ͼ85% of the plated cells. The duration of infection for each adenovirus was then titrated in isolated feline cardiocytes to determine via immunoblotting the optimum level of MAP4 expression.
Immunoblots-For MAP4, myocardial tissue was homogenized in a high salt buffer (100 mM Tris-HCl, pH 7.4, 10 mM EGTA, 0.35 M NaCl), boiled, and centrifuged at 16,000 ϫ g, 4°C for 30 min. The supernatant was mixed with an equal volume of SDS-sample buffer and boiled for 3 min. For MAP4 analyses in isolated cardiocytes, extracts were prepared in the same way as for the tissue samples with the exception that 1% Nonidet P-40 was added to the high salt buffer, and a protease inhibitor mixture (Sigma #P8340), as well as phosphatase inhibitor mixtures 1 (#P2850, Sigma) and 2 (#P5726, Sigma) were added to the buffer just before use. The lysate was centrifuged at 16,000 ϫ g, 4°C for 15 min, and the supernatant was mixed with an equal volume of SDS-sample buffer and boiled for 3 min. For the tubulin heterodimer and microtubule fractions, samples were homogenized in a microtubule stabilization buffer (7, 29) (1% Nonidet P-40, 50% glycerol, 5% DMSO, 0.5 mM GTP, 10 mM Na 2 HPO 4 , 0.5 mM EGTA, 0.5 mM MgSO 4 , 25 mM Na 4 P 2 O 7 ). Protease and phosphatase inhibitor mixtures were included in this buffer, and the lysate was centrifuged at 100,000 ϫ g, 25°C for 20 min. This supernatant was saved as the tubulin heterodimer fraction. The pellet was resuspended in 1% SDS buffer and boiled for 5 min to dissolve the pellet. This was saved as the microtubule fraction. The tubulin heterodimer and microtubule fractions were each mixed with an equal volume of SDSsample buffer and boiled for 3 min.
Processed protein samples in SDS-sample buffer were adjusted for equal BCA-based protein loading for each sample and used for SDS-PAGE analyses using pre-cast 3-8% gradient Tris acetate gels (NuPAGE, Invitrogen). Separated proteins were transferred to polyvinylidene difluoride membranes (Immobilon, Millipore), which were blocked with 1ϫ Animal-Free Blocker (#SP-5030, Vector Laboratories) and incubated overnight at 4°C with the primary antibodies at a 1:5000 dilution of our MAP4 antibody, 1:1000 dilution of our non-phospho-MAP4 antibodies and 1:1000 dilution of an anti-␤-tubulin antibody (clone DM-1B, Abcam). After incubating with biotinylated secondary antibody, specific protein bands were detected using avidin-biotinylated horseradish peroxidase in conjunction with enhanced chemiluminescence.
Quantitation of MAP4 Phosphorylation at Ser-924 and Ser-1056 by Western Blot Analysis-Although we could reliably identify MAP4 Ser-472, Ser-924, and Ser-1056 dephosphorylation by MS/MS, because of a low detection rate for the Ser-924 and Ser-1056 sites we could use MS/MS to quantitate site-specific MAP4 dephosphorylation only for the Ser-472 site. Thus, to characterize any changes at the Ser-924 and Ser-1056 sites in terms of the extent of MAP4 dephosphorylation, site-specific anti-phospho and anti-non-phospho antibodies were made for the Ser-924 and Ser-1056 regions of MAP4 using the deduced amino acid sequence for feline MAP4 shown in supplemental Fig. S1. The peptide synthesis and the production and purification of antibodies were performed for us by Antagene, Inc. of Mountain View, CA. The sequences of the peptide immunogens used to generate site-specific anti-phosphopeptide antibodies, which were also used for affinity purification and as blocking peptides to validate antibody specificity, were Cys-PDLKNVRSKVG (p) STENIKHQPG for pSer-924 and Cys-AQAKVG (p) SLDNVGHLPA for pSer-1056. For the purification of site-specific antibodies for the non-phosphorylated form of MAP4 at Ser-924 and Ser-1056, we used these peptides: Cys-PDLKNVRSKVGSTENIKHQPG for Ser-924 and Cys- AQAKVGSLDNVGHLPA for Ser-1056. As shown in Fig. 4 (C and D), the specificity of each of the two anti-non-phospho antibodies was then validated by neutralizing it with the respective phosphopeptide and non-phosphopeptide having the above sequences. For this purpose, 2.5 g of primary antibody was incubated with 25 g of either the respective phosphopeptide used for the antibody production or its non-phosphopeptide counterpart for 1 h in 1 ml of 1% nonfat dry milk in TBST at room temperature. The preincubated primary antibody/blocking peptide was then added to 9 ml of 1% nonfat dry milk in TBST. Thereafter, the antibody alone or the antibody plus one of the two peptides was used for immunoblotting. The nonphosphopeptide antibodies were not competed off by the phosphopeptides, indicating that they are specific only for the nonphosphorylated form of MAP4 in each case.
Immunofluorescence Confocal Microscopy-The cells were fixed and stained as described previously (15). Briefly, cardiocytes plated on coverslips were extracted for 1 min in a 1% Triton X-100 buffer containing 2 mM EGTA, 0.1 mM EDTA, 1 mM MgSO 4 , and 100 mM MES, pH 6.75. They were then rinsed three times in the same buffer with no detergent and fixed in 3.7% formaldehyde in this same buffer for 30 min. After each coverslip was blocked with 10% donkey serum in 0.10 M glycine in phosphate-buffered saline for 30 min at room temperature, the cells were incubated at 4°C overnight in a 1:200 dilution of primary antibody in 2% normal donkey serum in phosphatebuffered saline. After being washed three times with phosphate-buffered saline, the cells were incubated at room temperature for 2 h in a 1:200 dilution of fluorescein-conjugated secondary antibody in 2% normal donkey serum in phosphatebuffered saline. Optical sections (0.1 m) were acquired using a Zeiss LSM510META confocal microscope equipped with a 30-milliwatt argon laser (458, 477, 488, and 514 nm), a 1-milliwatt helium-neon laser (543 nm), a 5-milliwatt helium-neon laser (633 nm), and a Plan-Apochromat 63ϫ/1.40 differential interference contrast objective to obtain high resolution images. Adobe Photoshop 7.0 software was used for superimposing the laser channels and for cropping and rotating images. To derive semiquantitative data from cardiocyte micrographs, the "Lasso" tool in Photoshop was used to outline the cell boundary, and the mean pixel intensity within this boundary was then determined using the "Histogram" tool.

RESULTS
Characteristics of the Experimental Animals-At 2 weeks after RV pressure overloading via PAB, there was a 156 Ϯ 7% (S.E.) increase in RV systolic pressure and a 55 Ϯ 6% increase in the ratio of RV to body weight; these cats had no evidence of congestive heart failure. At 4 weeks after PAB, there was a 207 Ϯ 24% increase in RV systolic pressure and a 117 Ϯ 26% increase in the ratio of RV to body weight; these cats had moderate RV failure, with increases in RV end-diastolic pressure and in the ratio of liver to body weight. Note that, although some data suggest that ϳ1% of the ventricular cardiocytes of young cats may be mitotically competent (30), in our hands during an extensive characterization of this PAB model in young and older cats (31), evidence for cardiocyte mitosis in terms of nuclear [methyl-3 H]thymidine incorporation in vivo was confined to the normal neonatal cat (1.07% of LV cardiocytes, 0.16% of RV cardiocytes), and we found only rare RV cardiocytes showing nuclear labeling in vivo either at baseline or at any time after PAB in either young or old animals. Thus, because in this same model there is a 44% increase in RV cardiocyte cross-sectional area after PAB (31), what we are studying here is cardiocyte hypertrophy rather than hyperplasia. At 4 weeks after RV volume overloading via ASD, RV systolic pressure was unchanged, and there was a 75 Ϯ 4% increase in the ratio of RV to body weight, a value midway between that of the 2-and 4-week PAB groups; in no case was there an increase in RV end-diastolic pressure or any other evidence of congestive heart failure. In all cases for all groups, the mass of the normally loaded same-animal control LV was unchanged.
MAP4 Isoform Levels in Normal and Hypertrophied Myocardium-A schematic of the structure and domain organization of uMAP4 is shown in Fig. 2A. The levels of the myocardial mMAP4 and uMAP4 isoforms are shown in Fig. 2B. The only change that occurred was that in the pressure-overloaded RVs from the two PAB groups the level of MAP4 was increased at 2 weeks and further increased at 4 weeks. This was not a result of cardiac hypertrophic growth per se, because the level of MAP4 was unchanged in the similarly hypertrophied ASD RVs.
Site-specific Phosphorylation of MAP4 from Normal and Hypertrophied Myocardium-The sample mixture was very complex, as a trypsin digest of MAP4 is expected to produce 105 complete cleavage fragments and additional incomplete fragments. To maximize detection of phosphopeptides, no trap column or sample desalting procedures were utilized. The resultant MAP4 protein coverage was relatively high, with coverages of 53  lated Ser-472. The fragmentation pattern of this peptide is consistent with phosphorylation at residue Ser-472. The dominant peak at mass/charge (m/z) 932.5 corresponds to the loss of phosphoric acid (m/z 49) from the phosphopeptide upon fragmentation and is characteristic of doubly charged phosphopeptides. The presence of the y 9 ion at m/z 969.5 and the y 10 ion at m/z 1136.5 is consistent with phosphorylation at residue Ser-472. The presence of the y 7 ion at m/z 743 further pinpoints that Ser-472 was phosphorylated, and Thr-475, which could be another potential phosphorylation site, was not. The sequence pattern was the same in all of the data collected, indicating that Ser-472 is the site of phosphorylation. Met-469 was oxidized in both spectra (M ox ); however, in other trypsin fragments the non-oxidized methionine form was also seen. The bottom panel of Fig. 3 shows the ratio of trypsin fragments with and without phosphorylated Ser-472 relative to the total detected trypsin fragments that contained Ser-472. In the LV and RV of normal control cats, in the LV and volume overload-hypertrophied RV of ASD cats, and in the LV of PAB cats with RV pressure overload hypertrophy, the ratio of non-phosphorylated to phosphorylated Ser-472 was ϳ1:1. However, in hypertrophied RVs of PAB cats, the singly phosphorylated species was progressively reduced, and by 4 weeks only non-phosphorylated Ser-472 was present.
Pressure  Fig. S3). However, due to a low detection rate of these peptides from the hydrophobic microtubulebinding domain of MAP4, the phosphorylation state of Ser-924 and Ser-1056 could not be quantitated reliably by MS/MS. We therefore used site-and phosphorylation state-specific antibodies to determine the amount of unphosphorylated MAP4 at the Ser-924 and Ser-1056 sites in myocardial homogenates from the RV and LV of normal control cats, RV volume overload-hypertrophied ASD cats, and RV pressure overload-hypertrophied PAB cats. As shown in Fig. 4A, solely in the pressure-overloaded RV, wherein we consistently find microtubule network density to be increased (1, 7), total MAP4 FIGURE 4. Total uMAP4 and uMAP4 Ser-924 and Ser-1056 non-phosphorylation in the LV and RV of normal cats and cats with RV hypertrophy. A, levels of total MAP4, non-phospho-Ser-924 MAP4, and nonphospho-Ser-1056 MAP4 in RV and LV myocardium from normal control, 4 weeks RV volume overload (ASD), and 4 weeks RV pressure overload (PAB) cats. Myocardial homogenates for these blots were prepared as described under "Experimental Procedures." For total MAP4 we used our polyclonal anti-MAP4 antibody (12), and for the loading control we used a monoclonal anti-glyceraldehyde-3-phosphate dehydrogenase antibody (clone 6C5, Upstate Biotech). The sequences of the peptide immunogens used to generate site-specific antiphospho MAP4 antibodies, and the peptides used for the purification of non-phospho MAP4 antibodies are detailed under "Experimental Procedures." B, to facilitate direct visual comparison in the RV versus LV PAB samples of the increase in total MAP4 versus the increase in MAP4 that is not phosphorylated at either Ser-924 or Ser-1056, the loading volume ratio of the RV and LV samples used for the top panel of Fig. 4A was adjusted to produce an equal amount of total MAP4 in the top blot of Fig. 4B. This same loading ratio was then used to prepare the MAP4 Ser-924 and Ser-1056 blots. C and D, validation of non-phospho-Ser-924 and non-phospho-Ser-1056 antibody specificity. For the upper blots, the purified primary non-phospho antibody was incubated with either the respective phosphopeptide used for the antibody production or its non-phosphopeptide counterpart as described under "Experimental Procedures." Western blots were then performed with the antibody alone or the antibody plus one of the two peptides. For the lower blots, normal feline cardiocytes were infected at an m.o.i. of ϳ1 for 72 h with Myc-tagged adenoviruses expressing either wild-type MAP4 or one of the two dephosphomimetic MAP4 mutants at the sites of interest: Ser-924 3 Ala or Ser-1056 3 Ala. The immunoblots were then probed with the specified non-phospho antibodies as well as with a polyclonal anti-Myc tag antibody (#06 -549, Upstate Biotech) to confirm equal MAP4 expression and loading.
is greatly increased. However, as shown by antibodies that detect only MAP4 in which Ser-924 or Ser-1056 is not phosphorylated, there is an even greater proportional increase in both non-phospho-Ser-924 and non-phospho-Ser-1056 MAP4. Fig. 4B was then prepared to facilitate this comparison. That is, for each of the three MAP4 blots in Fig. 4A, equal protein loading was employed for each lane as shown by the glyceraldehyde-3-phosphate dehydrogenase blot. However, for Fig. 4B the MAP4 loading volume for the PAB RV and LV samples was adjusted to give a ratio of ϳ1:1 of overexpressed MAP4 protein in each lane; this same volume ratio was then used for loading the RV and LV samples blotted with the anti-non-phospho-Ser-924 MAP4 and anti-non-phospho-Ser-1056 MAP4 antibodies. Thus, Fig. 4B shows that, for a given amount of total MAP4, both non-phospho-Ser-924 and non-phospho-Ser-1056 MAP4 levels are much higher in pressure overload-hypertrophied RV myocardium. Together, these data indicate that, in addition to increased levels of MAP4, pressure overloading results in greatly increased levels of the unphosphorylated Ser-924 and Ser-1056 MAP4 pools. In Fig. 4 (C and D), the specificity of the non-phospho-Ser-924 and non-phospho-Ser-1056 antibodies is validated.
Effect of MAP4 Ser-472, Ser-924, or Ser-1056 Dephosphorylation or Ser-924 Phosphorylation on Cardiocyte Microtubules-These three MAP4 serine residues, which showed a decrease in the level of phosphorylation in pressure overload-hypertrophied myocardium, were mutated to alanine to mimic constitutive dephosphorylation, just as has been done before for Tau, a neuronal homologue of MAP4 (32), or were mutated to aspartate to mimic constitutive phosphorylation. This allowed us to test the significance of each sitespecific phosphorylation state in terms of its effect on microtubule properties.
As shown in Fig. 5, equivalent levels of adenovirus-mediated expression of Myc-tagged wild-type MAP4 and Ser-472 3 Ala MAP4 in adult feline cardiocytes caused similar modest shifts of free tubulin to the microtubule fraction, and expression of Ser-1056 3 Ala MAP4 caused a somewhat greater shift of free tubulin to the microtubule fraction. However, the Ser-924 3 Ala MAP4 mutant caused a major shift of tubulin into the microtubule fraction, just as we see in severe pressure overload cardiac hypertrophy (1). Further, the phosphomimetic Ser-924 3 Asp MAP4 mutant produced the opposite effect: here, the ratio of free to polymerized tubulin is quite similar to that seen with no viral infection. Note that expression of wildtype or mutant MAP4 introduced by adenoviruses, which is controlled by a strong viral promoter, results in an increased quantity of each of these MAP4 proteins. However, as expected from our previous data (33)(34)(35) showing that the rate of synthesis of native structural proteins is quite low in quiescent adult cardiocytes maintained in mitogen-free medium, the stable ␣␤-tubulin heterodimers shift between the pre-existing free and polymerized pools but do not increase in total quantity. Therefore, the differential effect of these MAP4 mutants on the ratio of free/polymerized tubulin is based on differing MAP4microtubule affinity and thus microtubule stability rather than on altered tubulin synthesis.
The data in Fig. 6 confirm this differential effect of the MAP4 mutants on microtubule stability. Tyr-tubulin is the predominant isoform of ␣-tubulin in dynamic microtubules (upper panel); after microtubule assembly, microtubule-incorporated ␣-tubulin undergoes two time-dependent posttranslational changes: reversible C-terminal detyrosination (Tyr-tubulin 7 Glu-tubulin, middle panel) and then irreversible deglutamination (Glu-tubulin 3 ⌬2-tubulin, lower panel), such that Gluand ⌬2-tubulin are markers for stable microtubules (11). Although the extent of post-translational modification of ␣-tubulin is rather modest after only 48 h of transgene expression, MAP4 overexpression increases the Glu-and ⌬2-tubulin mark- ers of microtubule stability in each group of virus-infected cardiocytes, and this is especially apparent for the Ser-924 3 Ala MAP4 mutant.
The appearance of the microtubule network in the confocal micrographs shown in Fig. 7 is consistent with the immunoblots shown in Fig. 5. That is, compared with the uninfected cell, there is a modest increase in microtubule network density for the cardiocytes infected with wild-type feline MAP4 or the single MAP4 point mutations at Ser-472 3 Ala or Ser-1056 3 Ala, and coincident labeling shows that transfected MAP4 is associated to only a limited extent with the microtubules of these cells. In contrast, however, for the Ser-924 3 Ala cardiocyte there is a much greater increase in microtubule network density, and the extent of coincident labeling with the transfected MAP4 is also much greater. Finally, the opposite result to this is seen for the Ser-924 3 Asp cardiocyte, as would be predicted if constitutive phosphorylation at this MAP4 site reduces MAP4microtubule affinity and thus microtubule stability.

DISCUSSION
Studies to date have shown increased density of a stabilized myocardial microtubule network in a number of cardiac disease states (reviewed in Refs. 1 and 36), and this has been especially striking in our hands when the hypertrophic growth response to pressure overloading is not sufficient to normalize ventricular wall stress (1,2). However, none of this research fully explained why this cytoskeletal abnormality was occurring, because there was no apparent reason for the apparently unique coincidental transcriptional up-regulation of the tubulins and MAP4 that is seen in this situation (9,10), and especially because this contrasts strikingly with the usual lack of correlation in a variety of tissues between tubulin and MAP4 mRNA abundance (24). Thus, the goal here was to seek the basis for the formation and persistence of this dense, stabilized microtubule network.
In considering this, the very well characterizedmicrotubule-MAPabnormality seen in Alzheimer disease provided an intriguing context. Here, the very opposite of what has been found in cardiac hypertrophy occurs: a neuronal MAP, Tau, dissociates from microtubules and forms neurofibrillary tangles of paired helical filaments (37). This is caused by phosphorylation of Tau, which reduces its affinity for microtubules, and this occurs both at sites within the microtubule-binding domain and in the projection domain (38). Thus, the question here was whether a similarly persistent but directionally opposite MAP dephosphorylation in hypertrophied myocardium might increase MAP4-microtubule affinity, thereby explaining both the extensive MAP4 decoration of microtubules and their stabilization (11). In this setting, transcriptional up-regulation of the tubulins and MAP4 reported in our previous studies (9,11) would become secondary rather than primary events.
The MAP4 phosphorylation data in Figs. 3 and 4 lend support to this idea, because opposite to the multiple-site hyperphosphorylation of Alzheimer disease Tau (38), we find dephosphorylation sites both within the microtubule-binding domain and within the projection domain of cardiac hypertrophic disease MAP4. Further, the data in Figs. 5-7 show that, when wild-type MAP4 and the mutants mimicking constitutive dephosphorylation of MAP4 at the sites identified by MS as dephosphorylated in hypertrophied myocardium are expressed at equivalent levels in normal cardiocytes, the microtubule phenotype of hypertrophied cardiocytes is reproduced by the mutant MAP4 protein in which constitutive dephosphorylation at Ser-924 is mimicked by the site-specific dephosphomimetic (39) mutation of that serine to alanine. Finally, to confirm this phosphorylation state-dependent mechanism, we found that a site-specific phosphomimetic (39) mutation of Ser-924 to aspartate produced as expected the opposite effect on the microtubule phenotype.
The finding of the novel and complete myocardial MAP4 Ser-472 dephosphorylation within the KDM repeat region ( Fig.  2A) of the MAP4 projection domain during pathological hypertrophy was unexpected. That is, apart from the assembly-promoting repeat region of the MAP4 microtubule-binding domain, there are within the C-terminal portion of MAP4 basic domain multiple other important determinants of MAP4 affinity for microtubules (40). Thus, phosphorylation of Ser-696 and FIGURE 6. Unmodified and post-translationally modified ␣-tubulin. These blots were prepared using the same microtubule protein samples used to prepare the "Polymerized Tubulin" blot in Fig. 5. An equal amount of microtubule protein sample was loaded in each lane of each of these three blots; this was verified by Coomassie Blue staining (data not shown), because glyceraldehyde-3-phosphate dehydrogenase is not present in the microtubule protein fraction. The blots were then probed with our peptide antibodies for ␣-tubulin isoforms (11) as specified. Integrated optical density measurements of lightly exposed blots for three experiments such as that shown here showed that the amount of unmodified Tyr-tubulin was the same in all samples. For the post-translationally modified Glu-tubulin, the ratio relative to "No Virus" was 1.29 Ϯ 0.39 for "WT," 1.27 Ϯ 0.38 for "S472A," 1.61 Ϯ 0.84 for "S924A," and 1.34 Ϯ 0.47 for "S1056A"; for the post-translationally modified ⌬2-tubulin, the ratio relative to "No Virus" was 1.30 Ϯ 0.42 for "WT," 1.32 Ϯ 0.45 for "S472A," 1.58 Ϯ 0.81 for "S924A," and 1.18 Ϯ 0.24 for "S1056A." Ser-787 within the proline-rich region greatly reduces the ability of MAP4 to increase microtubule polymerization (41)(42)(43), and even the short, acidic C-terminal tail appears to modify the interaction of MAP4 with microtubules (44). In contrast, there is no evidence to date for a major effect of the MAP4 projection domain on MAP4 affinity for microtubules, and we did not find any such effect for the feline Ser-472 MAP4 site here. Indeed, fusion proteins consisting solely of the MAP4 projection domain do not bind to microtubules in vitro (44,45). What is known about the role of the projection domain in MAP4 interactions with microtubules is that, as might be expected, it affects the spacing and packing of the microtubule array (46), and its presence also appears to be important for normal microtubule dynamic instability (47). Further, evidence derived from an expression construct of the MAP4 projection domain indicates that, although it does not bind to microtubules, it does have a major effect on the activity of the MAP4 microtubulebinding domain; of special interest here, this has been thought to be mediated by phosphorylation of one or more potential sites in the MAP4 projection domain (44).
What then is the normal role of the KDM repeat region wherein we find Ser-472 dephosphorylation? This region constitutes the bulk of the N-terminal MAP4 acidic projection domain and is thought to consist of numerous short ␣-helices separated by proline residues, which would predict a filamentous, flexible projection domain (24). As has been suggested by Olmsted (24), because Tau phosphorylation causes it to become long and stiff (48), perhaps phosphorylation of MAP4 within the KDM repeat region changes the physical properties of the molecule in such a way as to alter its activity. Again, however, in terms of the major goal of this study, i.e. defining the control of MAP4-microtubule affinity in hypertrophied myocardium, we did not find any such role for MAP4 Ser-472 phosphorylation state in the adult cardiocyte.
In contrast, the finding of MAP4 Ser-924 and to a lesser extent Ser-1056 dephosphorylation in hypertrophied myocardium, wherein we find a striking increase in MAP4 decoration of microtubules (11), was of much greater interest in terms of the goal of this study, because serine dephosphorylation within the KXGS motif of the assembly-promoting repeats of the microtubule-binding domain of MAPs has been shown to increase the affinity of MAPs for microtubules in a different biological context (16). That is, the homologous C-terminal regions in the various MAPs that have been identified as binding to the acidic surface of microtubules have a net basic charge (49 -51), and MAP4 contains a typical microtubule-binding domain at the C terminus of the protein. Indeed, an in vitro mixture of tubulin and MAP4 fragments containing only the assembly-promoting region forms microtubules that lack projections but are otherwise morphologically normal (52), and a change in phosphorylation state of a single KXGS motif serine within the assemblypromoting repeat region is sufficient to cause a major change in MAP-microtubule binding both for Tau (52), for the several forms of MAP2 (53), and for MAP4 (53). Finally, mutagenesis of a single one of these serines to glutamic acid to mimic constitutive phosphorylation disrupts MAP2-microtubule interactions in living cells (54).
The most critical aspect of the present data is that they show control of increases in 1) MAP4-microtubule affinity, 2) microtubule network stability, and 3) microtubule network density by the phosphorylation state of Ser-924. Note that these increases are the three characteristic features of the microtubule phenotype seen in severe pressure overload cardiac hypertrophy (1,2). Further, this abnormal microtubule network may well be an important contributor to the contractile and growth defects that lead to the crucial FIGURE 7. Confocal micrographs of adult feline cardiocytes. These micrographs were prepared from the same groups of cells used to generate the data shown in Fig. 5, where the treated cells were infected 48 h earlier at an m.o.i. of ϳ1 with bicistronic adenoviruses expressing GFP and either Myc-tagged wild-type feline MAP4 or Myc-tagged feline MAP4 having the dephosphomimetic single point mutations Ser-472 3 Ala, Ser-924 3 Ala, or Ser-1056 3 Ala or the phosphomimetic single point mutation Ser-924 3 Asp. At 48 h after infection, the GFP signal as estimated by fluorescence microscopy was equivalent among these five groups of infected cardiocytes. A polyclonal anti-Myc tag antibody (#2272, Cell Signaling) (green), which binds to the Myctagged MAP4 in the cells expressing each of the five transgenes, was used to identify the transfected protein. An isoform-non-selective monoclonal anti-␣-tubulin antibody (clone B-5-1-2, Santa Cruz Biotechnology) (red) was used to identify the microtubules. The yellow signal shows co-localization of these two proteins. For the red ␣-tubulin channel in which the microtubules are labeled, the mean pixel intensity within the boundary of each cardiocyte was 38.61 for "No Virus," 46.82 for "WT," 44.01 for "S472A," 59.84 for "S924A," 48.27 for "S1056A," and 26.65 for "S924D." For the green Myc tag channel in which the microtubule-bound Myc-tagged MAP4 is labeled, the mean pixel intensity within the boundary of each cardiocyte was 00.00 for "No Virus," 9.34 for "WT," 7.39 for "S472A," 31.21 for "S924A," 12.44 for "S1056A," and 3.48 for "S924D." Scale bar ϭ 20 m. clinical issue of decompensation of initially compensatory cardiac hypertrophy into the heart failure state (2).
To put these data in the context of the hypothesis driving this study, phosphorylation at human Tau Ser-262, which is homologous to feline MAP4 Ser-924, strongly inhibits the interaction of Tau with microtubules (16,52), and Tau phosphorylation at this site is a hallmark of Alzheimer disease (38,(55)(56)(57)(58)(59). Tau Ser-262 is also a critical site for phosphorylation-dependent control of Tau-microtubule affinity via the family of microtubule affinity-regulating kinases (16), as well as protein phosphatase 2A (39). Thus, it now will be very interesting to define the role of kinase and phosphatase interactions with MAP4 Ser-924 as potential etiologic base(s) for the abnormal microtubule array seen in hypertrophied myocardium.