The in Vivo Phosphorylation Sites of Rat Brain Dynamin I*

Dynamin I (dynI) is phosphorylated in synaptosomes at Ser774 and Ser778 by cyclin-dependent kinase 5 to regulate recruitment of syndapin I for synaptic vesicle endocytosis, and in PC12 cells on Ser857. Hierarchical phosphorylation of Ser774 precedes phosphorylation of Ser778. In contrast, Thr780 phosphorylation by cdk5 has been reported as the sole site (Tomizawa, K., Sunada, S., Lu, Y. F., Oda, Y., Kinuta, M., Ohshima, T., Saito, T., Wei, F. Y., Matsushita, M., Li, S. T., Tsutsui, K., Hisanaga, S. I., Mikoshiba, K., Takei, K., and Matsui, H. (2003) J. Cell Biol. 163, 813–824). To resolve the discrepancy and to better understand the biological roles of dynI phosphorylation, we undertook a systematic identification of all phosphorylation sites in rat brain nerve terminal dynI. Using phosphoamino acid analysis, exclusively phospho-serine residues were found. Thr780 phosphorylation was not detectable. Mutation of Ser774, Ser778, and Thr780 confirmed that Thr780 phosphorylation is restricted to in vitro conditions. Mass spectrometry of 32P-labeled phosphopeptides separated by two-dimensional mapping revealed seven in vivo phosphorylation sites: Ser774, Ser778, Ser822, Ser851, Ser857, Ser512, and Ser347. Quantification of 32P radiation in each phosphopeptide showed that Ser774 and Ser778 were the major sites (up to 69% of the total), followed by Ser851 and Ser857 (12%), and Ser853 (2%). Phosphorylation of Ser851 and Ser857 was restricted to the long tail splice variant dynIxa and was not hierarchical. Co-purified, 32P-labeled dynIII was phosphorylated at Ser759, Ser763, and Ser853. Ser853 is homologous to Ser851 in dynIxa. The results identify all major and several minor phosphorylation sites in dynI and provide the first measure of their relative abundance and relative responses to depolarization. The multiple phospho-sites suggest subtle regulation of synaptic vesicle endocytosis by new protein kinases and new protein-protein interactions. The homologous dynI and dynIII phosphorylation indicates a high mechanistic similarity. The results suggest a unique role for the long splice variants of dynI and dynIII in nerve terminals.

called the dephosphins (1). The dephosphins are constitutively phosphorylated in nerve terminals, and their collective rephosphorylation after SVE is necessary for maintaining synaptic vesicle recycling and thus synaptic transmission. One such dephosphin is dynamin I (dynI), a large GTPase enzyme that is crucial for the fission stage of SVE (2). During SVE, dynI is dephosphorylated by the calcium-dependent phosphatase calcineurin (3) and is subsequently rephosphorylated by cyclindependent kinase 5 (cdk5) on Ser 774 and Ser 778 (4,5). These phosphorylation sites, located in the proline-rich domain (PRD), are thought to regulate the interaction with the src-3 homology (SH3) domain-containing proteins involved in SVE. A long list of SH3 domain-containing proteins has been shown to bind the dynI PRD in vitro. Recently, we have identified syndapin I (sdpnI) as the phosphorylation-regulated dynI partner in vivo and that its interaction with dynI is crucial to SVE (6).
Apart from phosphorylation of Ser 774 and Ser 778 by cdk5, there have been a number of reports on other dynI phosphorylation sites and their potential protein kinases, both in vivo and in vitro. DynI is phosphorylated on Ser 857 by minibrain kinase/ Dyrk1A in vitro, and the phosphorylation regulates binding of dynI to amphiphysin I (amphI) and Grb2 (7). This phosphorylation was shown to occur in PC12 cells and was responsive to depolarizing stimuli, strongly suggesting a physiological relevance. It has also been reported that dynI is phosphorylated at Thr 780 by cdk5, which apparently regulates its binding to amphI (8). This is in conflict with our previous in vivo phosphorylation site analysis of dynI, which did not detect phosphorylation at this site (5). Other studies have suggested that there may be other kinases capable of phosphorylating dynI in vitro, but the in vivo relevance of these events has not been established (9 -12). DynI and dynII are also substrates for the tyrosine kinase c-Src in non-neuronal cells on Tyr 231 and Tyr 597 (13,14). The two sites are highly conserved between all three dynamin genes. However, the tyrosine phosphorylation of dynI has only been reported in transfected non-neuronal cells where it is not known to be normally expressed (13), whereas that of dynII occurs under endogenous conditions (13,14). This suggests that in vivo phosphorylation of these tyrosine phospho-sites in dynI remains an open question.
In this study, we sought to determine whether there were other in vivo dynI phosphorylation sites, in addition to Ser 774 , Ser 778 , and Ser 857 , that may prove to be functionally important for SVE. We established a method that ensures maximum purification of 32 P-labeled dynI so that none of the relevant phosphorylation sites were missed. We show that dynI was exclusively phosphorylated on serine residues by 32 P-phosphoamino acid analysis. We identified two new dynI phosphorylation sites at Ser 347 and Ser 512 and two new stimulation-dependent dynI phosphorylation sites at Ser 822 and Ser 851 . We also showed that phosphorylation of Thr 780 by cdk5 was restricted to in vitro conditions, did not occur in vivo (within detection limits), and did not regulate binding of amphI.

EXPERIMENTAL PROCEDURES
DNA Constructs and Protein Expression-The dynI PRD (rat, amino acids 746 -864) was amplified from the green fluorescent protein-tagged dynamin and subcloned into pGEX4T-1 as described previously (6). DynI point mutants were generated using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by DNA sequencing. GST-amphI SH3 domain was from Pietro de Camilli (Yale, New Haven, CT). GST-sdpnI SH3 domain was from Markus Plomann (University of Cologne, Germany). GST-endophilin I (endoI) SH3 domain was from Peter McPherson (McGill, Canada). GST-cdk5 and GST-p25 were from Jerry Wang (Hong Kong University, Hong Kong) and Li-Huei Tsai (Massachusetts Institute of Technology, Cambridge, MA), respectively. All GST fusion proteins were expressed in Escherichia coli and purified using glutathione-Sepharose beads (Amersham Biosciences) according to the manufacturer's instructions.
Synaptosomal DynI Purification-Crude (P2) synaptosomes were prepared from rat brain and labeled with 32 P as described previously (10). Briefly, synaptosomes were incubated with 0.75 mCi/ml [␥-32 P]P i for 1 h at 37°C. The synaptosomes were briefly depolarized with 41 mM KCl and immediately lysed as described previously (4). DynI was purified from synaptosome lysates using the GST-amphI SH3 domain bound to glutathione-Sepharose beads as described previously (4). GST-endoI SH3 domain and GST-sdpnI SH3 domain were also used as bait to pull down dynI from synaptosome lysates, as described previously (6). Beads were washed extensively, eluted in 2ϫ SDS sample buffer, and resolved on 7.5-15% gradient SDS gels.
Tryptic Digestion and Phosphopeptide Enrichment-DynI gel bands, each containing ϳ2 g of purified protein, were excised and diced from colloidal Coomassie Blue-stained SDS gels. Four untreated and four KCl-treated 32 P-labeled dynI bands were used in each two-dimensional map. Twenty unlabeled dynI bands were TiO 2 -enriched (see below) and added to the untreated sample for two-dimensional mapping. Two dynI bands were used in the nano-LC-MS/MS experiments (see below). The bands were destained in three 1-ml washes of 25 mM ammonium bicarbonate in a 50% acetonitrile solution with 1 h of vortexing between washes. The gel pieces were digested in 25 mM ammonium bicarbonate aqueous solution containing 12.5 ng/l trypsin at 37°C overnight. The digest solution was made up to 50% acetonitrile, and the tryptic peptides were extracted after 15 min of vortexing. A second extraction was obtained with 50% acetonitrile solution after 15 min of vortexing. A final extraction was obtained with 80% acetonitrile, 5% formic acid solution after 15 min of vortexing. The combined extract was dried down to 2.5 l.
Phosphopeptides from a tryptic digest of non-radioactive dynI were enriched using TiO 2 as described earlier (15). Briefly, the tryptic digest was concentrated into 5 l and added to 25 l of loading solution (300 mg/l dihydroxybenzoic acid in 0.1% trifluoroacetic acid 80% acetonitrile or 5% trifluoroacetic acid in 80% acetonitrile in the absence of dihydroxybenzoic acid). The sample was loaded onto a GELoader tip (Eppendorf), converted to a microcolumn, packed with TiO 2 , and washed twice with loading solution and then twice with a wash solution of 0.1% trifluoroacetic acid/80% acetonitrile, without the presence of dihydroxybenzoic acid. The sample was eluted in 20% ammonium hydroxide/20% acetonitrile solution and immediately dried.
Phosphoamino Acids Analysis and Two-dimensional Phosphopeptide Mapping-The 32 P-labeled dynI tryptic digest was analyzed by phosphoamino acid analysis as described previously (4,10). Two-dimensional phosphopeptide mapping was done on 20 ϫ 20 cm cellulose plates (Merck) by electrophoresis at pH 4.7 (16) followed by ascending chromatography in 35% 1-butanol, 20% pyridine, 7.5% acetic acid, 2.5% acetonitrile, and 35% water. The radioactive spots were detected by quantitative phosphorimaging (Storm 860, Amersham Biosciences), scraped from the plate, and extracted from the cellulose as described (16), except that the extraction was done in 5% formic acid solution. The phosphopeptides were concentrated using C18 or graphite microcolumns and detected by matrix-assisted laser desorption ionization-mass spectrometry (MALDI-MS, Voyager DE-PRO, Applied Biosystems) as described previously (5).
A theoretical two-dimensional phosphopeptide map was generated firstly by determining the theoretical electrophoretic mobility by predicting the charge at pH 4.7 and dividing by the mass of the phosphopeptide as described by Meisenhelder et al. (16). The mobility in the second dimension, which is based on relative hydrophobicity, was estimated from the reversedphase HPLC index (17) as calculated using GPMAW version 7.01 (Lighthouse Data, Denmark). The theoretical HPLC index was decreased by two units for each additional phosphate group to account for the observed decreased mobility of multiply phosphorylated peptides (Fig. 2, B and C).
Nano-LC-MS-An 8-l aliquot (estimated at 0.5-1.5 pmol) of TiO 2 -enriched dynI phosphopeptides was loaded onto the nano-HPLC system (LC Packings Ultimate HPLC system, Dionex, Netherlands) with a 75-m inside diameter pre-column of C18 reversed-phase material (ReproSil-Pur 120 C18-AQ, 3-m beads, Dr. Maisch, Germany) in 5 min. It was then eluted through a 50-m inside diameter C18 column of the same material at 100 nl/min. The gradient was from 100% phase A (0.1% formic acid in water) during loading, then increased to 10% phase B (90% acetonitrile, 0.1% formic acid, and 9.9% water) in 3 min, then to 50% phase B in 28 min, then to 60% phase B in 3 min, and finally to 100% phase B in 1 min. The eluate was sprayed through a 10-m inside diameter distal coated SilicaTip (New Objective) into a QSTAR XL quadrupole-TOF (QqTOF) MS (Applied Biosystems) or a QTOF Ultima MS (Micromass/Waters) using 1900 V on the tip. For the detection of phosphopeptides of a known molecular mass the precursor ion selection was fixed at the specific m/z of the phosphopeptide in its most abundant charge state with a wide m/zsetting(2-3units)usingconsecutive2-sscans.Informationdependent data acquisition was done by using a 1-s survey scan from which the three most abundant doubly, triply, or quadruply charged peptides were selected for product ion scans (2 s). The data for the phospho-dynI 343-364 was from experiments using stable isotope labeling, the full details of which will be published elsewhere. As a consequence, the N terminus of a phospho-dynI 343-364 was labeled with iTRAQ reagent (Fig. 4C). In this instance, the TiO 2 -enriched digest was first treated with iTRAQ 116 reagent (Applied Biosystems) according to the manufacturer's instructions before nano-LC-MS/MS analysis. All phosphopeptides reported in this study were detected in at least three independent experiments.
In Vitro Phosphorylation-GST-dynI PRD (1 g) immobilized on glutathione-Sepharose beads was phosphorylated by recombinant cdk5/p25 in a total volume of 40 l containing 30 mM Tris-HCl, pH 7.4, 1 mM EGTA, 1 mM MgCl 2 , 0.05% Tween 20 and 40 M [␥-32 P]ATP (6.9 ϫ 10 6 cpm/nmol) at 37°C. Phosphorylation proceeded for 10 min, and reactions were terminated by addition of 3ϫ SDS sample buffer and boiling. Samples were analyzed by SDS-PAGE, and phosphoproteins were detected by autoradiography. A single band each of in vitro phosphorylated GST-dynI PRD wild-type and double mutant Ser to Ala were digested as above and enriched for phosphopeptides using Fe 3ϩ -immobilized metal affinity chromatography as described previously (5). The phospho-box phosphopeptides were sequenced by nano-LC-MS/MS by fixing the precursor ion selection at a specific m/z as described above.
Pulldown Experiments-A rat brain or synaptosomal extract was prepared as described previously (6). Various GST-dynI PRD or GST-amphI SH3 recombinant proteins were then incubated with the same amounts of tissue lysate at 4°C for 1 h. All pulldown experiments were done in the presence of 150 mM NaCl unless stated otherwise. Beads were washed extensively, eluted in 2ϫ SDS sample buffer and analyzed by SDS-PAGE.
Antibodies and Western Blots-The anti-amphiphysin I monoclonal antibody was from Pietro de Camilli. The anti-dynI antibodies and phosphospecific antibodies to phospho-Ser 774 and phospho-Ser 778 in dynI were reported previously (4). Immunoprecipitations from rat brain synaptosomes were performed as described previously (10). Protein samples were separated by SDS-PAGE on 10% or 12% acrylamide gels and transferred to nitrocellulose membrane. Western blots were analyzed by enhanced chemiluminescence method using the SuperSignal West Pico Chemiluminescent Substrate (Pierce).

Quantitative Extraction of DynI by GST-amphI SH3 Domain-
To catalogue all in vivo phosphorylation sites in dynI from rat brain synaptosomes, we used a method capable of extracting and purifying all dynI from synaptosome lysates, as judged by SDS-PAGE analysis and autoradiography. We have previously reported that the SH3 domain of amphI quantitatively extracts synaptosomal dynI independently of its phosphorylation status (6). However, another study has proposed that the interaction of GST-amphI SH3 with dynI is phospho-dependent (8). Here, we compared the efficiency of GST fusion proteins of amphI, endoI, and sdpnI SH3 domains to quantitatively extract dynI by three sequential pulldown experiments. Initial pulldown experiments with GST-amphI SH3 and GST-sdpnI SH3 recovered a large amount of endogenous dynI protein (Fig. 1A) and 32 Plabeled phospho-dynI (Fig. 1B), whereas GST-endoI SH3 bound less dynI. A second pulldown of the same extracts with GST-amphI SH3 was performed to capture residual dynI missed in the first pulldown assay. There was Ͻ2% residual dynI protein and phospho-dynI following amphI or sdpnI in the first pulldown experiment (Fig. 1, A and B). In contrast, there was significant dynI and phospho-dynI captured following endoI SH3 in the first pulldown experiment (Fig. 1, A and B, middle panels). This is in agreement with a previous report that endoI SH3 is partly sensitive to dynI phosphorylation in vitro (6) and shows that amphI SH3 can indeed capture this small pool. A third sequential pulldown assay with sdpnI SH3 or endoI SH3 domains recovered no residual dynI protein or phospho-dynI after two previous extractions with amphI SH3, endoI SH3, or sdpnI SH3 (Fig. 1, A and B). Therefore the SH3 domain of amphI and sdpnI quantitatively extracts synaptosomal dynI from synaptosomes.
We independently performed the similar GST-amphI SH3 triple pulldown experiments using synaptosome extracts and blotted with anti-dynI antibodies. Two rounds of GST-amphI SH3 pulldown assay was enough to quantitatively extract all synaptosomal dynI, because no trace of unbound dynI was found in the second or third pulldown experiment (Fig. 1C). We thus used this system of double GST-amphI SH3 pulldown experiments in all subsequent experiments to purify all synaptosomal dynI, regardless of its in vivo phosphorylation status.
DynI Is Phosphorylated Exclusively on Serine Residues-DynI phosphorylation on Thr 780 has been controversial (18). It was not detected in a previous in vivo study (5) but was claimed to be found after in vitro phosphorylation with cdk5 (8). A simple way to address whether a protein is phosphorylated on serine, threonine, or tyrosine residues is by phosphoamino acid analysis after 32 P labeling. There may be some phosphorylation sites that are not labeled with 32 P, however, in dynI these sites are unlikely to be relevant to SVE, because they do not turnover 32 P in 1 h of labeling. The first published report of phosphoamino acid analysis on synaptosomal dynI showed that only serine was phosphorylated (19). The source of dynI in that experiment was the bands cut from an SDS gel of a synaptosome lysate, which may be contaminated by another underlying protein with similar size. The same result was reported when dynI was purified by a single pulldown experiment with the GST-amphI SH3. Phosphorylation at threonine or tyrosine was absent (4). In this study, we performed phosphoamino acid analysis using dynI purified by GST-amphI-SH3 pulldown experiments from 32 Plabeled synaptosomes and again found only serine was phosphorylated ( Fig. 2A). We conclude that synaptosomal dynI is exclusively phosphorylated on serine residues with no detectable 32 P labeling on threonine or tyrosine residues.
Separation and Detection of in Vivo 32 P-Labeled DynI Phosphopeptides-A tryptic digest of dynI purified from 32 Plabeled synaptosomes was subjected to two-dimensional phos- MAY 18, 2007 • VOLUME 282 • NUMBER 20 JOURNAL OF BIOLOGICAL CHEMISTRY 14697 phopeptide mapping (Fig. 2B). Similarly, a tryptic digest of dynI from synaptosomes depolarized with 41 mM KCl for 10 s was also analyzed in parallel (Fig. 2C). The 32 P-labeled phosphopeptides were detected by autoradiography. We observed 15 clearly distinct spots on the two-dimensional map (A-O). The majority of the radiation was in spots A-D (Fig. 2B). These spots have previously been shown to contain overlapping tryptic phosphopeptides (including tryptic cleavages and missed cleavages) from the dynI phospho-box ( 772 RRSPTSSPTPQRR 784 ) (6). Spots C and D were previously shown to contain almost exclusively phospho-Ser 774 , and spots A and B were shown to contain doubly phosphorylated peptides that are equal parts phosphorylated at Ser 774 and Ser 778 (5).

MS/MS Identification of Phosphorylation Sites in DynI
Phosphopeptides-To account for all the 32 P that is distributed among the various dynI phosphorylation sites, the phosphopeptides contained within each spot on the two-dimen-FIGURE 1. AmphI SH3 domain quantitatively extracts dynI from nerve terminals. A, GST-amphI SH3, GST-sdpnI SH3, or GST-endoI SH3 bound to GSH-Sepharose were used in pulldown experiments from 32 P-labeled P2 synaptosomes lysed in Triton X-100 in the presence of 150 mM NaCl. Bound proteins were analyzed by SDS-PAGE and visualized by Coomassie Blue staining. DynI (arrow) was extracted in duplicate pulldown experiments (top panel). GST-amphI SH3 was then used sequentially in the second pulldown experiment to recover any remaining unbound dynI in the initial extract (second panel). In the third sequential pulldown experiment of this extract GST-sdpnI SH3 or GST-endoI SH3 were used to recover any remaining unbound dynI (third panel). The amounts of GST-SH3 domains used in the pulldown experiment are shown (lower panel, which is from the same gel as in the top panel). B, phospho-dynI was visualized by autoradiography of the same pulldown experiments. Results are representative of three independent experiments. C, GST-amphI SH3 bound to GSH-Sepharose was used in the triple pulldown experiments from synaptosome lysates. 10% each of the starting material (input), bound proteins from the first, second, and third pulldown experiments, and the unbound proteins from each pulldown experiment were subjected to Western blotting and probed with specific anti-dynI and anti-phospho-Ser 774 antibodies (top two panels). The amount of protein loaded onto the SDS gel was visualized by Coomassie Blue staining (bottom panel). FIGURE 2. Phosphoamino acid analysis and two-dimensional tryptic phosphopeptide mapping. A, autoradiograph of the phosphoamino acid analysis of dynI from 32 P-labeled synaptosomes. dynI was purified using multiple sequential pulldown assays with the GST-amphI SH3 domain. The migration of a standard mixture (std) of phosphoserine, -threonine, and -tyrosine is shown alongside, as determined by ninhydrin staining. B, autoradiograph of a two-dimensional tryptic phosphopeptide map of purified dynI from control synaptosomes. Radioactive spots are labeled alphabetically above and to the left of each spot position. The phosphopeptides of the dynI phospho-box ( 772 RRSPTSSPTPQRR 784 ) are enclosed in a rectangle. These phosphopeptides were previously identified by tandem MS (5,6). C, autoradiograph of the twodimensional phosphopeptide map of purified dynI from depolarized synaptosomes.
sional map were extracted from the cellulose and analyzed by MALDI-MS and tandem MS. Spots A-D contained the dynI phospho-box peptides ( 772 RRSPTSSPTPQRR 784 , data not shown) that we identified previously (6). Spots E-H had low relative radiation levels and were analyzed by MALDI-TOF MS, but no phosphopeptides were detectable (Fig. 2B). Similarly, no phosphopeptides were detected in spots J, M, and N, which had even lower levels of total radiation (ϳ2%, 1%, and 1% respectively, Fig. 2B).
Mono-and di-phosphopeptides matching the tail of the long splice variant dynIxa, 847 SGQASPSRPESPRPPFDL 864 , were found in spots K and I, respectively. The mono-phosphorylated peptide in spot K was determined to be dynIxa 847-864 , where either Ser 851 or Ser 857 is phosphorylated (data not shown). Neither phosphorylation site was grossly dominant in this mixture of two phosphopeptides of equal molecular mass. The di-phosphorylated peptide in spot I was determined to be dynIxa 847-864 where both Ser 851 and Ser 857 are phosphorylated (Fig. 3A). Phospho-Ser 851 was detected between y 13 and y 14 and also as a dehydroalanine between b 4 and b 5 . Phospho-Ser 857 was revealed as the only possible position for the second phosphorylation site in the sequence 855 PES 857 between the b 8 /y 10 and b 11 /y 7 ions. In contrast to the two sites in the dynI phospho-box (5,6), there appears to be no hierarchy between phospho-Ser 851 and -Ser 857 , suggesting independent regulation and potentially independent protein kinases.
A weak signal, that was sensitive to phosphatase treatment, was detected by MALDI-TOF MS in spot O. Nano-LC-MS/MS was used to sequence this peptide by selecting the quadruply charged precursor at m/z 606.3. The sequence was determined to be dynIIIxxb 849 RPPPSPTRPTIIRPLESSLLD 869 where Ser 853 is phosphorylated (Fig. 3B). DynIII was previously found to be co-purified with dynI from P2 synaptosomes (5). Small y-type ions and both small and large b-type ions were detected, with little information on the middle part of the sequence. However, the position of the phosphorylation site was unequivocally determined by crucial ions such as the b 2 and b 5 ions, which limit the phosphorylation site to the sequence 851 PPS 853 , of which only Ser 853 can be phosphorylated. In addition to the homologous phosphorylation of the dynI and dynIII phosphobox (5), we have now revealed that dynIIIxxb has a third in vivo phosphorylation site in a sequence location that is homologous to the dynIxa sequence containing phospho-Ser 851 . As for dynI, this site is present only in the long-tailed splice variant of dynIII.
Spot L was analyzed by MALDI-TOF MS, and a phosphopeptide was observed, as confirmed by dephosphorylation with Antarctic phosphatase (data not shown). Nano-LC-MS/MS was required to sequence this phosphopeptide. The sequence was dynI 797 GPAPGPPPAGSALGGAPPVPSRP-GASPDPFGPPPQVPSRPNR 838 where Ser 822 is phosphorylated (Fig. 4A). Small b-type ions and y-type ions throughout the sequence were used to determine the phosphorylation site and the identity of the phosphopeptide. Y-type ions up to y 16 and b-type ions up to b 16 (and then a weak signal at b 25 ) were not phosphorylated. Because y 17 and higher ions were phosphorylated, the phosphorylation site can be unequivocally localized to Ser 822 . This is the fifth proline-directed serine (i.e. SP sequence) in the dynIxa PRD that was found to be phosphorylated in vivo.
Finally, we used another independent approach to maximize detection of any relatively minor phosphorylation sites. Extraction of phosphopeptides from cellulose plates after two-dimensional peptide mapping may not be completely efficient due to adsorption to surfaces. Therefore to maximize the number of phosphopeptides detected, without relying on 32 P labeling, we used TiO 2 to enrich for all phosphopeptides from a dynI tryptic digest and analyzed them using nano-LC-MS/MS with information-dependent data acquisition. Two new dynI phosphorylation sites were found. Automatic selection of the precursor at m/z 712.8 produced a unique fragmentation spectrum (Fig. 4B). This phosphopeptide matched the sequence of dynI 511 TSGNQ-DEILVIR 522 where Ser 512 is phosphorylated. The sequence was entirely described by y-type ions. The phosphorylation site was detected as the dehydroalanine residue between y 11 and y 10 . In the same experiment, we also sequenced phospho-dynI 510 -522 and found that Ser 512 was phosphorylated in this slightly larger phosphopeptide (data not shown). Automatic selection of the precursor at m/z 750.0 produced another unique fragmentation spectrum (Fig. 4C). This phosphopeptide matched dynI 343 RIEGSGDQIDTYELSGGARINR 364 where Ser 347 is phosphorylated. Non-phosphorylated y-type ions ruled out the possibility of phosphorylation near the C-terminal half of this phosphopeptide. The phosphorylation site was deduced to be at Ser 347 , before the residues producing phosphorylated (b 5 to b 11 ) fragments and after the residues producing non-phosphorylated fragments (b 3 ). This is the first time that in vivo dynI sites have been detected outside of the PRD. It is possible that these sites were 32 P-labeled but were not detected by MALDI-TOF MS.
It can be predicted where the peptides encompassing Ser 512 and Ser 347 would appear in the two-dimensional map in Fig. 2B, if they were 32 P-labeled. We plotted the theoretical HPLC index versus the estimated electrophoretic mobility for each of the detected dynI and dynIII peptides (Fig. 5A). Only peptides with up to one missed trypsin cleavage are shown. The theoretical electrophoretic mobility correlated with the experimentally observed mobility of the sequenced phosphopeptides (Figs. 2B, 2C, and 5A). In the second dimension, the HPLC index was a good predictor of mobility, except for the large peptide in spot L (Figs. 2B, 2C, and 5A). This may reflect an overemphasis on peptide molecular mass when using the HPLC index. DynI 511-522 has an HPLC index greater than dynI 774 -783 (spots A-D) and less than dynI 847-864 (spots K and I, Fig. 5A). The phosphopeptides dynI 511-522 , dynI 511-523 , and dynI 510 -522 are predicted to migrate in the second dimension near the unidentified spots at E-H (Fig. 2B). Therefore, phosphorylated dynI 510 -523 theoretically accounts for at least two of the spots in the region of E-H. There are two unassigned spots (M and N, Fig. 2B) above dynI 847-864 (spots K and I). Spot N (Fig. 2B) might be accounted for by dynI 839 -864 , which would be expected to be phosphorylated at Ser 851 and/or Ser 857 . The theoretical mobility of dynI 343-361 predicts migration to a spot above and to the left of dynI 847-864 (spots K and I). Therefore, dynI 343-361 theoretically accounts for spot M. However, direct evidence is lacking for Ser 347 and Ser 512 being 32 P-labeled in 1 h during synaptosome labeling or depolarization. Moreover, the possibility of non-or semi-tryptic dynI proteolysis products and minor contaminating non-dynI phosphopeptides in the two-dimensional map cannot be ruled out.
Distribution of the 32 P between Phospho-sites-We next quantified the amount of phosphate in each spot (Table 1). Quantitative analysis allowed a determination of the potential in vivo significance of each site in terms of its abundance. Note that "cold phosphorylation" (i.e. that occurring prior to 32 P labeling) was not quantifiable in this way. The 32 P-labeled pep- was selected for sequencing and produced a sequence that matches the sequence of dynI 511-522 where Ser 512 is phosphorylated. C, same as for panel B except that the precursor at m/z 750.0 was selected for sequencing and produced a sequence that matches the sequence of dynI 343-361 where Ser 347 is phosphorylated. The N terminus was labeled with ITRAQ 116 reagent affecting the mass of the b-type ions. tides were detected by autoradiography, and the intensity of each spot was quantified and expressed as a percentage of the total radiation in all spots (Fig. 5B). Using quantitative phosphorimaging data, plus the knowledge of which phosphopeptide accounted for each spot it was possible to deduce that phospho-Ser 774 and -Ser 778 were 32 P-labeled in a ratio of two to one (6). We now extend this analysis to cover all phosphorylation sites. The phosphopeptides identified in each spot are shown (Fig. 5B). Spots A-C each contained about 20% of the total 32 P and contained Ser 774 or Ser 774 plus Ser 778 . Therefore these previously reported sites contain 69% of the total 32 P in dynI. Note that a small, unknown amount of this radiation is derived from dynIII (5), but this amount was ignored in the present calculation. MALDI-MS spectra have indicated that dynIII phospho-box phosphopeptides are between 5-and 10-fold less abundant than the homologous dynI phosphopeptides (4,5). Spot O was dynIII Ser 853 and had 2% of the total 32 P (Fig. 5B). Spots K and I together accounted for 12% of the total radiation. The majority of the 32 P was in spot K, which was a mix of Ser 851 or Ser 857 phosphopeptides. The minority of this label was in spot I, which was doubly phosphorylated on both sites; therefore the double phosphorylation is relatively rare. Spot L had 5% of the radiation and contained Ser 822 . Spots E-H together contained a combined value of 8% of the total 32 P radiation. Although likely to include Ser 512 , the presence of other sites or phosphopeptides from contaminating proteins cannot be ruled out. Therefore Ser 512 must represent less (or much less) than 8% of the total dynI phosphorylation. Spots M and N contained ϳ2% of the total 32 P. No sites were identified, but Ser 347 is predicted to migrate in this region. Assuming this to be the case, Ser 347 has Ͻ2% of the dynI total phosphorylation. Overall, all the sites identified to date can be ranked in decreasing order of abundance: 774, 778, 857 (and 851), 822, 512, and then 347.
To predict the potential biological significance of each phosphorylation site we next determined which sites were responsive to depolarization (Fig. 5C). This provides a crude measure FIGURE 5. Quantitative analysis of dynI phosphorylation sites. A, theoretical two-dimensional map produced by plotting HPLC index versus predicted electrophoretic mobility. Known dynI and dynIII phosphopeptides with up to one missed trypsin cleavage are shown. Phosphopeptides that were detected from the two-dimensional map in Fig. 2B are shown with the same lettering. B, quantitative analysis of the radiation for each spot in Fig. 2 (B and C) normalized to the total counts detected in all spots in panel b (defined as 100%). Filled bars are from synaptosomes at rest, and open bars are after depolarization for 10 s with 41 mM KCl. The phosphorylation site(s) associated with each spot are shown. The sites are as identified previously (5) or in this study. C, quantitative analysis of the total radiation for each phosphorylation site in dynI purified from control (solid bars) and depolarized synaptosomes (open bars). Data for Ser 774 , Ser 778 , and Ser 822 are averages and errors (ϮS.E.) from three phosphopeptide maps. Data for Ser 851 and Ser 857 are the average of two phosphopeptide maps, and the errors are the ranges for these two experiments. Note that the phosphorylation of Ser 851 and Ser 857 could not be individually apportioned, as was done for Ser 774 and Ser 778 (6), so they are grouped together.

TABLE 1 Distribution of 32 P radiation on all dynI and dynIII phosphorylation sites detected from two-dimensional tryptic phosphopeptide mapping and nano-LC-MS/MS
Phosphorylation sites identified by MS from phosphopeptides extracted from the two-dimensional map were correlated with the 32 P radiation detected in each spot to determine the percentage of total radiation. Phosphorylation sites that were not extracted from the two-dimensional map, rather by nano-LC-MS/MS alone, were correlated with the spots in the two-dimensional map that match the theoretically predicted migration of the phosphopeptides (see ЉExperimental ProceduresЉ and Fig. 5A). Note that it is likely that an unknown phosphopeptide contributed to part of this measured radiation predicted to be associated with Ser 512 . Note also, that a small, but unknown amount of radiation is likely to be contributed by dynIII Ser 769 and Ser 773 , reducing the proportion of radiation from dynI Ser 774 and Ser 778 . of the relevance of each site to depolarization-induced SVE. The amount by which each phosphopeptide in spots A-D was decreased upon depolarization was quantified (Fig. 5B) and correlated to the phosphorylation site identified (found here and previously). The amount of 32 P-labeled phospho-Ser 774 was decreased by 30% (Ϯ2% S.E.), and phospho-Ser 778 was decreased by 47% (Ϯ3% S.E.) following depolarization (Fig. 5C).
Again, the small amount of radiation from co-purified dynIII phospho-Ser 769 and Ser 773 in spots A-D was ignored. Although the phosphopeptide containing the sites Ser 851 and Ser 857 were not highly 32 P-labeled, the phosphorylation was reduced by 34% (range of Ϯ1%) following depolarization (Fig. 5C). Therefore, the two C-terminal phosphorylation sites in dynIxa are dephosphorylated upon depolarization. The phosphopeptide containing Ser 822 was significantly dephosphorylated by 56% (Ϯ0.5% S.E.) following depolarization (Fig. 5, B and C). No depolarization-dependent change in the phosphorylation of Ser 512 could be measured, because there was no 32 P label directly associated with this site. However, if it was part of spots E-H as we surmise, then it was depolarization-sensitive ( Fig.  5B). There was no depolarization-dependent change in the phosphopeptide surmised to contain Ser 347 .

Phosphorylation of Thr 780 on DynI Is Restricted to in Vitro
Conditions-It was previously reported that dynI is phosphorylated by cdk5 only at Thr 780 in vitro and that all phosphorylation is blocked by mutation of Thr 780 to Ala (8). These data strongly suggest Thr 780 is the sole (cdk5-mediated) phosphorylation site in the dynI PRD and conflicts with our main findings. We previously found no evidence for Thr 780 phosphorylation in vivo (5). Mass spectrometry is not a good tool for resolving this discrepancy because it cannot be used to definitively prove the absence of a phosphorylation site. Therefore we used a site-directed mutagenesis strategy to determine how a single mutation on Thr 780 might abolish [␥-32 P]ATP phosphorylation of dynI on Ser 774 or Ser 778 by cdk5 in vitro. The results contradict those of Tomizawa et al. (8). We generated a series of single, double, or triple point mutations of Ser 774 , Ser 778 , and/or Thr 780 to Ala in GST-dynI PRD for the purpose of testing the circumstances under which phosphorylation of the phosphobox by recombinant cdk5/p25 could be prevented (Fig. 6A). Single or double mutation of Ser 774 and Ser 778 had little effect on the extent of phosphorylation by cdk5, as determined by autoradiography (Fig. 6B) and by quantitative phosphorimaging analysis (Fig. 6C). A similar result was obtained with a single mutation on Thr 780 (Fig. 6, B and C). However, when all three sites in the phospho-box were mutated to Ala, phosphorylation was greatly reduced (32 Ϯ 3% of Dyn1 WT, p Ͻ 0.001, Fig. 6, B and C). It appears that mutation of any one of these three sites results in a compensatory increase in phosphorylation at one or more of the remaining sites. So it is only when all three sites are mutated that a significant drop in phosphorylation was detected in the PRD. However, note that 30% of the total phosphorylation remained in the triple mutant, suggesting a major contribution from other sites in the PRD. The dynI PRD construct used in these studies was based on the long splice variant dynIxa, therefore there are at least three additional sites available for phosphorylation by cdk5; Ser 822 , Ser 851 , and Ser 857 . Some or all of these may account for the remaining phosphorylation in Fig. 6B. Using tandem mass spectrometry, we confirmed that the phospho-box of the in vitro phosphorylated wild-type GST-dynI PRD was indeed phosphorylated at Ser 774 , Ser 778 , and Thr 780 (supplemental Fig. S1). As expected, in vitro phosphorylation at Thr 780 was most easily observed by analyzing the double mutant dynI PRD (Fig. 6D). The phosphorylation at Thr 780 was determined by the y 3 and y 4 -98 ions. The relatively strong signal for the phosphorylated y 5 fragment ion, owing to the favorable cleavage of an X-P bond (20), was a clear signal that the phosphorylation site on this peptide was C-terminal to Pro 779 , i.e. it must only be Thr 780 , because no other part of 779 PTPQR 783 can be phosphorylated. This y 5 ion was also a strong signal in the spectra of the wild-type sequences (supplemental Fig. S1), alternately conferring and denying phosphorylation to Thr 780 . Curiously, the phosphorylated y 5 ion was absent from the spectrum of a similar in vitro analysis by Tomizawa et al. (8).
The results demonstrate that a single mutation in a protein that is multiply phosphorylated is insufficient to detect reduced overall in vitro phosphorylation. While Thr 780 is easily detected as an in vitro phosphorylation site, it does not necessarily follow that Thr 780 is phosphorylated in vivo. Our systematic approach contradicts the overall results and conclusions of Tomizawa et al. (8).
AmphI Binding Is Independent of DynI Phosphorylation-There are several reports that amphI binding to dynI is regulated by the in vitro phosphorylation status of dynI (8,21). To extend these in vitro observations, we also generated a series of Glu mutations at Ser 774 , Ser 778 , and Thr 780 in GST-dynI PRD (Fig. 6A). Pulldown experiments with GST-dynI PRD wild-type and GST-dynI Ala mutants (non-phosphorylatable) or Glu mutants (pseudo-phosphorylation) did not significantly alter the binding of native full-length amphI as shown by Western blot (Fig. 7A). The results were confirmed by quantitative densitometry analysis of multiple experiments (Fig. 7B).
We then performed a reverse pulldown experiment, using recombinant GST-amphI SH3 to capture native dynI in the presence of different sodium chloride concentrations. As the salt concentration increased from 0 to 1 M, the amount of total dynI protein bound to GST-amphI SH3 decreased (Fig. 7C). However, the amount of phosphorylated dynI bound was unchanged, even at high salt concentration (Fig. 7C). This suggests that amphI SH3 binding to dynI is not regulated by dynI phosphorylation (see Fig. 1). The observation that GST-amphI SH3 has an apparently higher binding affinity toward phospho-dynI than non-phospho-dynI is interesting but not likely to have any physiological relevance, because it only occurred at 0.5-1 M NaCl.
We have shown how the dynI PRD interacts with full-length amphI and how amphI SH3 interacts with full-length dynI. Next, we used an immunoprecipitation experiment to ask whether the association of native full-length dynI with native full-length amphI might be phosphorylation-regulated. Pretreatment of synaptosomes with Ba 2ϩ for 1 h was previously shown to produce massive dynI dephosphorylation (22) by chronic depolarization (23). We prepared Ba 2ϩ -treated synap- FIGURE 7. DynI phosphorylation has no effect on amphI interaction. A, interaction of pseudo-phosphorylated dynI PRD with amphI. GST-dynI PRD WT or Glu mutants were bound to GSH-Sepharose and used in pulldown experiments from rat brain lysates. The samples were blotted with anti-amphI antibodies (top panel). The amount of GST-dynI PRDs used in the pulldown experiments are shown (bottom panel). B, the amount of amphI bound to GST-dynI PRD mutants was quantified by densitometric analysis of Western blots such as in panel A (n ϭ 4). Data were expressed as a percentage of dynI PRD WT Ϯ S.E. One-way analysis of variance was applied, but no statistically significant differences were found. C, GST-amphI SH3 binds phospho-dynI with high affinity. GST-amphI SH3 was used in pulldown experiments from P2 synaptosomes lysed in Triton X-100 in the absence or presence of 0.15, 0.5, or 1 M NaCl. DynI was detected by Coomassie Blue staining of the gel (top panel). Lysates from the same experiment were probed with antibodies to phospho-Ser 774 and phospho-Ser 778 . Immunoblots are displayed (bottom two panels). Data are representative of two independent experiments. D, interaction of phospho-dynI with amphI. Synaptosomes were incubated for 60 min in Krebs-like buffer containing 0.1 mM Ca 2ϩ or 2.5 mM Ba 2ϩ , lysed, and immunoprecipitated with anti-dynI antibodies. The complexes were subjected to Western blotting analysis with antibodies to amphI and phospho-Ser 774 (top two panels). The amount of dynI immunoprecipitated by the dynI antibodies was visualized by Ponceau staining of nitrocellulose membrane. Blots are representative of three independent experiments. E, quantitative analysis from Western blots of the amount of amphI co-immunoprecipitated with dynI (D). Results are from three independent experiments. tosomes and compared them to resting (0.1 mM Ca 2ϩ ) synaptosomes. DynI was immunoprecipitated from each of these preparations. DynI was massively dephosphorylated upon Ba 2ϩ treatment, as shown by Western blot analysis with the antiphospho-Ser 774 antibody (Fig. 7D). However, the interaction between amphI and dynI was not affected (Fig. 7, D and E). From these interactions studies we conclude that amphI is not a phosphorylation-dependent partner for dynI, in vivo or in vitro.

DISCUSSION
DynI Has Seven Phosphorylation Sites-An exhaustive strategy was used to identify and rank all the phosphorylation sites in dynI. Mass spectrometry of synaptosomal 32 P-labeled dynI phosphopeptides separated by two-dimensional tryptic mapping revealed seven in vivo phosphorylation sites at Ser 774 , Ser 778 , Ser 822 , Ser 851 , Ser 857 , Ser 512 , and Ser 347 (Fig. 8A). Thr 780 phosphorylation was not detected. It was shown previously that a peptide containing Ser 857 was phosphorylated in vitro by Mnb/Dyrk1A (7), whereas a peptide containing Ser 851 was not. In this study, we confirmed the existence of Ser 857 as an in vivo phosphorylation site (7) and identified Ser 851 as a new in vivo phosphorylation site within the same tryptic peptide. These sites are restricted to the tail of the longer splice variant, dynIxa (Fig. 8B). We found that Ser 851 and Ser 857 were phosphorylated to a similar extent, and there was no hierarchy to their phosphorylation. It has been shown that Ser 857 phosphorylation reduced dynI binding to amphI SH3 domain (7). However, our data suggested that dynI phosphorylation does not affect binding of amphI. It may be difficult to discern the true effect of phosphorylation at Ser 857 on amphI binding from our experiments, because it is Ͻ12% of the total radiation and we focused mainly on cdk5 substrates within the phospho-box.
The phosphorylation site found at Ser 512 is near the start of the pleckstrin homology domain (Fig.  8A). It is not known how much the phosphate on this site is turning over, because we were unable to correlate a radioactive spot with the sequenced phosphopeptides. Nonetheless, some of the unidentified spots were good candidates for containing phospho-Ser 512 , based on peptide mobility prediction. If Ser 512 is involved in SVE, then it is most likely involved in regulating dynI interaction with phospholipids. The phosphorylation at Ser 347 is nearest to the GTPase domain and may have some influence on this domain. Along with part of the assembly domain, this domain is thought to be involved in dynI tetramerization and higher order self-assembly (24). The lack of a crystal structure or functional role for the sequence encompassing this site precludes further comment on its function. This site is predicted to have low or no turnover of 32 P in a 1-h incubation of synaptosomes, and it is therefore difficult to envisage an important role for Ser 347 in SVE. Both Ser 347 (RIEGSGDQID) and Ser 512 (KKKTSGNQDE) are not situated in front of a proline residue (Table 1). Ser 512 is more clearly in a basophilic context than Ser 347 , on the N-terminal side of the phosphorylation site, so they may not be phosphorylated by the same protein kinase. However, both Ser 347 and Ser 512 share some common amino acid sequences (SGXQ) followed closely by an acidic residue on the C-terminal side. Hence, we conclude that there is at least one other protein kinase, apart from cdk5 (and potentially minibrain kinase/Dyrk1A (7)), that phosphorylates dynI in vivo.
Ser 822 (RPGASPDPF) is located in front of a proline and might be phosphorylated by one of the same protein kinases that phosphorylate Ser 774 , Ser 778 , Ser 851 , or Ser 857 . Ser 822 is flanked on either side by PXXP motifs, which bind SH3 domain-containing proteins. The motifs immediately C-terminal to Ser 822 are involved in binding grb2 and amphI (these are called Site 8 and Site 9, respectively, Fig. 8B) (25). Therefore, it is possible that this binding may be phospho-regulated by Ser 822 .
New information about the hierarchical nature of Ser 774 and Ser 778 was revealed when the extent of dephosphorylation of each site was quantified. These data showed that following depolarization, phospho-774 and phospho-778 are dephosphorylated to a similar extent, perhaps phospho-778 slightly more. More importantly, the pool of doubly phosphorylated dynI (with both Ser 774 and Ser 778 phosphorylated) was dephosphorylated, in preference to the singly phosphorylated pool (with only Ser 774 ). This implies a greater relative importance on the removal of phospho-Ser 778 . We propose that dynI phosphorylation at Ser 774 has a different regulatory role to phosphorylation at Ser 778 in SVE.
Phosphorylation at Thr 780 Is an in Vitro Artifact-Controversially (18), it was reported that cdk5 phosphorylates dynI at Ser 774 and Ser 778 in two studies (4,5) and solely at Thr 780 in another study (8). However, we have now demonstrated that, although Thr 780 is phosphorylated in vitro by cdk5 and phosphorylation mutants may be able to produce a functional effect in vivo (8), it is unlikely that dynI is phosphorylated at Thr 780 in vivo in synaptosomes within the ϳ100-fold detection limits of our methods. The most likely explanation is that phosphorylation at Thr 780 is restricted to in vitro conditions; i.e. it is simply an artifact of cdk5 phosphorylation in vitro. Three experiments supported our conclusion that Thr 780 is not phosphorylated by cdk5 in synaptosomes. Firstly, phosphoamino acid analysis of purified dynI from 32 P-labeled synaptosomes revealed exclusive phosphorylation of dynI on serine residues. This experiment alone, within the limits of detection, restricts Thr 780 or any tyrosine phosphorylation to being a static phosphorylation site (i.e. not labeled with 32 P i after 1 h) in synaptosomes or to not being present in synaptosomes at all. Secondly, we were unable to detect any Thr 780 phosphorylation by exhaustive two-dimensional phosphopeptide mapping of 32 P-labeled dynI and MS analysis. This experiment, within the limits of detection, rules out the presence of a major static pool of dynI phosphorylated on Thr 780 . Any such pool must contain Ͻ1% of the phosphorylation in dynI. Thirdly, a single mutation of Thr 780 to Ala did not abolish cdk5 phosphorylation in the dynI-PRD, in contrast to observations reported by Tomizawa et al. (8). Like Thr 780 , a single Ala mutation on Ser 774 and Ser 778 , or in combination, had no effect on dynI phosphorylation by cdk5 in vitro. It was only when all three sites at Ser 774 , Ser 778 , and Thr 780 were mutated to Ala that phosphorylation by cdk5 was significantly reduced. This suggests that there is a compensatory phosphorylation of alternative sites within the PRD when any single site is mutated. In fact, we have previously shown in phosphoamino acid analysis that phospho-threonine only appeared when dynI was phosphorylated by cdk5 in vitro but was absent when dynI was purified from 32 P-labeled synaptosomes (4). Potential differences between the two PRD splice variants are possible. The long form was used in our study (containing six SP or TP motifs, while there are only four in the short form), but the form used by Tomizawa et al. (8) was not reported (see Fig. 8). Placing further focus on in vitro phospho-rylation versus in vivo phosphorylation site analysis does not appear to be constructive.
The remaining discrepancy between our study and that of Tomizawa et al. relates to the use of their claimed phosphospecific antibody to Thr 780 . With this antibody it was shown that the phosphorylation of Thr 780 occurred in synaptosomes and brain slices and was stimulus-sensitive (see Fig. 7 of Tomizawa et al. (8)). Although the data appear compelling at face value, it assumes there is specificity for this antibody toward phospho-Thr 780 , which was not reported. Rather, that antibody was raised against a synthetic peptide that encompasses all three phospho-sites, Ser 774 , Ser 778 , and Thr 780 , and was phosphorylated in vitro prior to immunization of rabbits. It is not possible to prevent phosphorylation at Ser 774 or Ser 778 under these conditions, and indeed we showed that all three are phosphorylated in vitro in the PRD (supplemental Fig. S1). The site specificity of the antibody has not been tested, and it is highly likely to crossreact with both phospho-serines. Therefore, the results obtained with its use cannot be ascribed to Thr 780 but match previous data for phosphorylation at the serines, particularly if Thr 780 is not phosphorylated in vivo.
DynIII Has Three Phosphorylation Sites-We previously found that a small amount of dynIII co-purifies during the dynamin purification from P2 synaptosomes and were able to identify the homologous phospho-box phosphorylation sites to Ser 774 and Ser 778 for dynIII (5). The amino acid sequences around these sites are highly conserved between dynI and dynIII. In this study we found a new dynIII phosphorylation site at Ser 853 that was 32 P-labeled. DynIII has previously been reported to be mainly expressed in the post-synapse (26). It is unlikely that the P2 synaptosome preparation contained postsynaptic dynIII that could acquire 32 P label. Therefore, we conclude that the 32 P-labeled dynIII was presynaptic. Ser 853 is homologous to Ser 851 of dynI. Like dynI, this site is only present in the long splice form. We note that Ser 857 from dynI is not present in dynIII (Fig. 8B). This third homologous phosphorylation site, Ser 853 , suggests that dynIII may perform an analogous function to dynI in the presynaptic nerve terminal. Its phosphorylation on the same sites in synaptosomes suggests that the same protein kinases may be involved and that there is a physiologically significant role for Ser 853 (dynIII) and Ser 851 (dynI). The amino acid sequence of the tail of the long splice variants of dynI and dynIII are not similar except for five amino acids, SPXRP, that encompass the phosphorylation site motif (Fig. 8B).
It remains to be determined whether all the phosphorylation sites in dynIII have now been identified, although the results suggest that the main dynIII sites have been found. Among the seven dynI sites, only serines 774, 778, 851, and 347 are conserved in the sequence of dynIII. Three of them have now been identified, and although phosphorylation at Ser 347 in dynIII is possible, it would be well below detection limits in our current approaches. No sites were detected that were not homologous to dynI. However, the amino acid sequences of dynIII are sufficiently different as to indicate that different protein partners are phospho-regulated by these sites. The results suggest dynIII has a role in nerve terminals; however, it must be a highly specialized role. MAY 18, 2007 • VOLUME 282 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 14705
DynII Phosphorylation?-Perhaps surprisingly, we found no evidence for phosphorylation of dynII in this study. DynII was present in the synaptosomes, as judged by Western blot and mass spectrometry data (data not shown). The GST-amphI SH3 domain pulldown experiment also extracts dynII and III (data not shown). Although we have not directly determined that the extraction of dynII or III is quantitative, it is likely to be so. In fact, our laboratory purifies dynII by pulldown experiment on GST-amphII SH3 beads. The absence of any dynII phosphorylation sites in synaptosomes is not conclusive evidence that none are present but is compelling because sites in dynIII were detectable at low levels. Only two of the seven dynI phospho-sites are conserved in the sequence of dynII. Ser 774 is missing in dynII, although Ser 778 and Ser 347 are conserved. Ser 822 and Ser 512 are missing in dynII, although the surrounding amino acid sequences are highly conserved. The tail of dynII is not subjected to the same alternative splicing as dynI and III, however it is highly related in sequence to the long tail of dynIII, both of which are missing Ser 857 (dynI). The main difference is that dynII specifically lacks Ser 851 (dynI, or Ser 853 in dynIII), hence it has none of the sites present in either of the other dynamins and would not be expected to be phosphorylated by a proline-directed protein kinase in this region. The absence of detection of phosphorylation of dynII at Tyr 231 and Tyr 597 (13,14) is not surprising, because the synaptosomes were not stimulated with any growth factor receptors. However, these sites may be phosphorylated in other cellular compartments or after the appropriate stimulation.
A Complete Description of dynI in Vivo Phosphorylation?-We have attempted to map all the in vivo phosphorylation sites in dynI. However, no technology is yet sufficiently sensitive to allow unequivocal conclusions that all sites in any protein have been identified. Other previously reported sites were not found: Thr 780 (8), Tyr 231 and Tyr 597 in dynI (13,27) and dynII (14), or Ser 795 (28). We conclude that phosphorylation of Thr 780 and Ser 795 is restricted to in vitro conditions only. We found no evidence for phosphorylation on tyrosine but cannot rule out several possibilities. Phosphorylation of Tyr 231 and Tyr 597 was first observed in dynI transfected into non-neuronal cells, thus phosphorylation occurred in an inappropriate in vivo context. Although we found no evidence for phosphorylation of dynI on Tyr, it, and other sites, cannot be ruled out completely for a variety of reasons. Firstly, other sites in dynI may be phosphorylated in different subcellular compartments or at different stages of development. Secondly, it is possible (although unlikely) that some phosphopeptides do not bind TiO 2 or Fe 3ϩ -immobilized metal affinity chromatography and therefore would have been missed. A number of phosphopeptides are not mass spectrometry "friendly" (i.e. do not produce sufficient signal), although the specific properties of such phosphopeptides have yet to be described. Within such limitations we conclude that any new sites discovered in the future must represent Ͻ2% of the total phosphorylation, because the majority of the 32 P label on dynI has been accounted for. The caveats on this conclusion are the possibility that a comigrating, poorly detected phosphopeptide could have been missed in one of the spots or the prior existence of high stoichiometry phosphorylation sites that may not be labeled with 32 P after 1 h. The potential physiological significance of the latter hypothetical sites would be questionable in the context of SVE. A thorough and deliberate strategy of maximized phosphoprotein capture, 32 P labeling, and highly sensitive mass spectrometry has been utilized to avoid missing significant in vivo phosphorylation sites.
The discovery of four new in vivo dynI phosphorylation sites at Ser 512 , Ser 822 , Ser 851 , and Ser 347 in addition to the three already identified sites (Ser 774 , Ser 778 , and Ser 857 ) provides a basis for further study of the phosphoregulation of dynI and SVE. It remains to be shown whether these new phosphorylation sites in the PRD regulate binding of other SH3 domaincontaining proteins, besides syndapin I (6), or whether they have a supplementary role in the same process of endocytosis or perhaps a major part in a mechanistically distinct mode of endocytosis.