Systematic Mapping of Posttranslational Modifications in Human Estrogen Receptor-α with Emphasis on Novel Phosphorylation Sites*S

A systematic study of posttranslational modifications of the estrogen receptor isolated from the MCF-7 human breast cancer cell line is reported. Proteolysis with multiple enzymes, mass spectrometry, and tandem mass spectrometry achieved very high sequence coverage for the full-length 66-kDa endogenous protein from estradiol-treated cell cultures. Nine phosphorylated serine residues were identified, three of which were previously unreported and none of which were previously observed by mass spectrometry by any other laboratory. Two additional modified serine residues were identified in recombinant protein, one previously reported but not observed here in endogenous protein and the other previously unknown. Although major emphasis was placed on identifying new phosphorylation sites, N-terminal loss of methionine accompanied by amino acetylation and a lysine side chain acetylation (or possibly trimethylation) were also detected. The use of both HPLC-ESI and MALDI interfaced to different mass analyzers gave higher sequence coverage and identified more sites than could be achieved by either method alone. The estrogen receptor is critical in the development and progression of breast cancer. One previously unreported phosphorylation site identified here was shown to be strongly dependent on estradiol, confirming its potential significance to breast cancer. Greater knowledge of this array of posttranslational modifications of estrogen receptor, particularly phosphorylation, will increase our understanding of the processes that lead to estradiol-induced activation of this protein and may aid the development of therapeutic strategies for management of hormone-dependent breast cancer.

A systematic study of posttranslational modifications of the estrogen receptor isolated from the MCF-7 human breast cancer cell line is reported. Proteolysis with multiple enzymes, mass spectrometry, and tandem mass spectrometry achieved very high sequence coverage for the full-length 66-kDa endogenous protein from estradiol-treated cell cultures. Nine phosphorylated serine residues were identified, three of which were previously unreported and none of which were previously observed by mass spectrometry by any other laboratory. Two additional modified serine residues were identified in recombinant protein, one previously reported but not observed here in endogenous protein and the other previously unknown. Although major emphasis was placed on identifying new phosphorylation sites, N-terminal loss of methionine accompanied by amino acetylation and a lysine side chain acetylation (or possibly trimethylation) were also detected. The use of both HPLC-ESI and MALDI interfaced to different mass analyzers gave higher sequence coverage and identified more sites than could be achieved by either method alone. The estrogen receptor is critical in the development and progression of breast cancer. One previously unreported phosphorylation site identified here was shown to be strongly dependent on estradiol, confirming its potential significance to breast cancer. Greater knowledge of this array of posttranslational modifications of estrogen receptor, particularly phosphorylation, will increase our understanding of the processes that lead to estradiol-induced activation of this protein and may aid the development of therapeutic strategies for management of hormonedependent breast cancer.

Molecular & Cellular Proteomics 8:467-480, 2009.
The ␣-isoform of the estrogen receptor (ER␣) 1 is a 66-kDa nuclear transcription factor that mediates transcriptional regulation of genes involved in cell proliferation and differentiation and plays a pivotal role in the development and progression of breast cancer (1)(2)(3). Consequently the development of therapies for management of hormone-dependent breast cancer has targeted signaling pathways based on modulation of ER␣ activity (4 -6). Current knowledge and understanding of ER␣ activity is derived from over 30 years of accumulated scientific evidence that has sought to delineate ER-mediated mechanisms and signaling pathways under pathological conditions modeled in cultured breast cancer cell lines. Because the function or activity of a protein may depend strongly on the presence of posttranslational modifications (PTMs), significant research has focused on detection and quantitation of such modifications in ER␣. The constellation of PTMs on a protein constitutes a molecular code that may dictate protein conformation, localization, and function. Thus biological inference based on ER␣ structure requires a comprehensive study of all possible PTMs on the constituent amino acids.
Modifications to ER␣ reported to date are listed in Table I. Two of the most common and important PTMs that modulate the activity of ER␣ are redox-based modifications of cysteine residues (7)(8)(9)(10)(11)(12) and phosphorylation of serine, threonine, and tyrosine residues (13). Biochemical techniques used previously to map ER␣ phosphorylation sites utilized radioactive labeling with 32 P (14 -16), Edman degradation, deletion and/or point mutations, and Western blots (17)(18)(19). Mutational analysis in combination with estrogen response element-luciferase reporter assays can validate the functional relevance of phosphorylation sites (16,20). Although these techniques have provided substantial information regarding PTMs in ER␣, they have several disadvantages. Autoradiography using radioisotope labeling ( 32 P) lacks specificity, Edman degradation requires large amounts of purified protein and is not applicable to proteins/peptides with N-terminal acetylation, and quantitation of PTMs by Western blot analysis relies on prior knowledge of the type and position of specific modifications and on the availability of high quality antibodies. Deletional and mutational analyses on transfected rather than endogenous protein are laborious and time-consuming. Moreover discrepancies in the literature concerning ER␣ phosphorylation may arise from the use of different promoters in the reporter constructs (13).
Because of its selectivity and sensitivity, MS/MS has emerged as a powerful tool for the analysis of PTMs such as phosphorylation (21)(22)(23). The analysis of phosphopeptides benefits from the use of multiple ionization methods and instrumentation platforms as ESI and MALDI interfaced with different tandem mass analyzers provide complementary information (21). Orthogonal Q-TOF and linear quadrupole trap (LTQ) instruments are widely used for ESI analysis, and several tandem instruments have been developed for or adapted to MALDI methods, including the LTQ (24 -26). We previously utilized vacuum (v) MALDI-LTQ to optimize quantitation of cysteine oxidation within the DNA-binding domain (DBD) of recombinant ER␣ (7). More recently we enriched and immunoaffinity-purified endogenous ER␣ derived from the human breast cancer cell line MCF-7 in quantities compatible with tandem MS analysis and identified a novel phosphorylation site, Ser 154 . vMALDI-MS n confirmed Ser 154 phosphorylation in human breast cancer cells grown under both ligand-dependent and ligand-independent conditions (27). Coincidentally we also observed known phosphorylation sites Ser 118 and Ser 167 not previously confirmed by MS methods.
Here we report a comprehensive MS study of ER␣ to achieve very high sequence coverage for the full-length 66-kDa endogenous protein from estradiol-treated MCF-7 cell cultures with particular emphasis on identifying new phosphorylation sites. Using multiple reaction monitoring (MRM)/MS we demonstrated that one previously unreported phosphorylation site is strongly dependent on estradiol, confirming its potential significance to ER activation and breast cancer. This study is the first to describe extensive mass spectrometric sequencing and phosphopeptide mapping of full-length ER␣ derived from a human breast cancer cell line.

EXPERIMENTAL PROCEDURES
Reagents and Chemicals-These were as described previously (27) except that human recombinant ER␣ (rER␣) was from Invitrogen, and sequencing grade chymotrypsin, Lys-C, Asp-N, and Glu-C were from Roche Diagnostics.
Cell Culture and Sample Preparation-The human breast cancer cell line MCF-7 was obtained from the American Type Culture Collection (Manassas, VA). Protocols for cell culture and extraction, immunopurification, and in-gel digestion of ER␣ from 20 15-cm plates of cells were as described previously (27) except that for each experiment in-gel digestion was with one of the following enzymes: chymotrypsin (300 ng; 37°C), trypsin (150 ng; 37°C), Glu-C (250 ng; 25°C), Lys-C (400 ng; 37°C), Asp-N (280 ng; 37°C), or a combination of the last two. The resultant peptides were extracted by vortexing and sonication with 50% ACN, 5% formic acid; vacuum centrifuged to remove most of the solvent; reconstituted in 20 l of 0.1% formic acid; and stored at Ϫ80°C.
On-bead Proteolytic Digestion-After immunoprecipitation and PBS washes the immune complex was transferred to a siliconized 1.5-ml Eppendorf tube and washed twice in 100 mM Tris, pH 7, by gentle rotation for 2 min at room temperature and three times in 25 mM NH 4 HCO 3 with centrifugation between washes. The pelleted beads were suspended in a further 150 l of NH 4 HCO 3 and 170 ng of sequence grade trypsin (10 l of 17 ng/l stock solution) and then incubated overnight at 37°C with shaking at 1,200 rpm. Eppendorf tubes were centrifuged at 12,000 rpm, and the supernatant was transferred to low binding polymer technology 0.65-ml microcentrifuge tubes (PGC Scientifics). Ten microliters of acetonitrile and 1 l of 10% formic acid were added to the peptide solution, which was then evaporated down to ϳ15 l. Each sample was divided into three 5-l aliquots and stored at Ϫ80°C until used for mass spectrometry.
Peptide Mass Fingerprinting by MALDI-TOF-Peptides to be analyzed by MALDI were desalted with C 18 ZipTips (Millipore), and then 1 l of peptide solution and 1 l of 2,5-dihydroxybenzoic acid or ␣-cyano-4-hydroxycinnamic acid matrix (50 mg/ml in 4:1 ACN, 0.6% phosphoric acid) were mixed, deposited on a stainless steel sample plate, and air-dried. Alternatively 1 l of peptide solution was mixed with 1 l of 50% ACN, 0.5% TFA; spotted on the MALDI plate; and dried. This addition of solvent and drying was repeated three more times, and then 1 l of matrix solution was added and dried. 1 l of 0.1% TFA was added and aspirated off by applying a tissue. Mass spectra were recorded on a Voyager-DE STR Plus instrument (Applied Biosystems, Framingham, MA) with 337 nm laser radiation, positive ion delayed extraction, reflector mode, and 20-kV accelerating voltage.
Peptide Identification by Tandem Mass Spectrometry-Peptide mixtures were further analyzed on a vMALDI-LTQ linear ion trap (Thermo Fisher, San Jose, CA) using both manual and automated data acquisition. Both methods used "Tune plus" using the "crystalpositioning system" and low threshold settings for "auto spectrum filter." MS 2 spectra were typically collected using threshold values of 500 counts for MS and 150 counts for MS 2 with "auto gain control" turned on, limiting the number of ions admitted to the trap. A parent mass isolation width of 3 m/z units and fragmentation settings of activation Q ϭ 0.25, activation time of 30 ms, and relative collision energy of 35% were used. For phosphopeptide mapping, potential peptide peaks were interrogated by MS/MS for the neutral loss of phosphoric acid, and (MH Ϫ 98) ϩ peaks were selected for MS 3 . Loss of phosphoric acid converted serine residues to dehydroalanine (dS). Reversed phase nano-HPLC-MS/MS (LC-MS/MS) was performed using an Ultimate HPLC instrument (Dionex, Sunnyvale, CA) with a C 18 analytical column directly connected to a Q-STAR Pulsar I quadrupole orthogonal TOF mass spectrometer (MDS Sciex, Concorde, Canada) as described previously (27).
LC-MS/MS and vMALDI-LTQ fragment ion data were searched against a database constructed for human full-length ER␣ using Mascot in-house licensed version 2.2 (Matrix Sciences, London, UK) with the "no enzyme" option. LC-MS/MS data were also searched with Paragon (Protein Pilot 2.0, Applied Biosystems).
LC-MRM/MS-Samples were analyzed by nano-LC-MRM/MS on a 4000 QTRAP hybrid triple quadrupole/linear ion trap mass spectrometer (Applied Biosystems). Chromatography was performed using an Eksigent (Dublin, CA) NanoLC-2D LC system with buffer A (0.1% formic acid) and buffer B (90% acetonitrile in 0.1% formic acid). Digests were loaded at 20 l/min in buffer A onto a 5-mm ϫ 300-m reversed phase C 18 column (5 m, 100 Å; Dionex) and eluted at 300 nl/min with a 75-m-inner diameter Integrafrit column (New Objective, Woburn, MA) packed in house with 10 -12 cm of ReproSil-Pur C 18 -AQ 3-m reversed phase resin (Dr. Maisch, GmbH) with a gra-dient of 2-70% B over 32 min. Peptides were ionized using a PicoTip emitter (75 m, 15-m tip; New Objective). Data acquisition was performed with ion spray voltage at 2450 V, curtain gas pressure at 10 p.s.i., nebulizer gas at 20 p.s.i., and interface heater temperature at 150°C. Collision energy, declustering potential, and collision cell exit potential were optimized using recombinant ER for maximum sensitivity. MRM transitions were monitored and acquired at unit resolution both in the first and third quadrupoles (Q1 and Q3) to maximize specificity. The y 7 transition was monitored at Q1 670.9 m/z and Q3 785.5 m/z for the unmodified peptide and at Q1 710.9 m/z and Q3 865.4 m/z for the phosphorylated peptide with dwell times of 10 and 320 ms, respectively. Furthermore although the y 7 transition gave the most intense signal, MRM transitions y 8 and y 7 Ϫ phosphoric acid were also assayed to confirm the retention time of the Ser 294 peptides. A minimum of nine data points were collected per peak.

RESULTS
The overall analytical strategy ( Fig. 1A) required optimizing immunoaffinity enrichment of human ER␣ from estradioltreated MCF-7 cells for peptide analysis by tandem mass spectrometry. Estradiol treatment of confluent MCF-7 cells for 30 min, overnight incubation of cell lysates with agarose-coupled ER␣ antibody, and one-dimensional gel electrophoresis separation yielded endogenous ER␣ protein at 66 kDa ( Fig. 1B). Recombinant protein (rER␣) was used to confirm this identification and to model in-gel digestions. Comparison with 13 pmol of rER␣ indicates a yield of ϳ8 -10 pmol of protein/20 plates of cultured MCF-7 cells. Western blots of cell lysates before and after immunoprecipitation indicate the ER␣ immunocapture to be as much as 98% efficient (Fig. 1C).
Peptidecutter (ExPASy) was used to predict peptide fragments of a size and nature suitable for ionization and sequencing by LC-MS/MS and/or vMALDI-MS n , showing that multiple enzymes would be required to achieve comprehensive coverage of the primary sequence. Digests were prepared using trypsin, chymotrypsin, Glu-C, Lys-C, Asp-N, and the combination of Lys-C and Asp-N and analyzed as unseparated mixtures by MALDI-TOF. Fig. 2 shows a typical peptide mass fingerprint derived from a chymotryptic digest of immunopurified MCF-7 ER␣. Internally calibrated peak lists were searched against the ER␣ database on Swiss-Prot using Protein Prospector Version 4.5.1. Peptide sequence information was then obtained by LC-MS/MS and vMALDI-MS n . All spectra were interrogated by database searching, and sequence identifications were confirmed by visual inspection. All identified ER␣ peptides are listed in supplemental Table 1  as supplemental data. Data for Ser(P) 118 , Ser(P) 154 , and Ser(P) 167 have been reported previously and are not repeated here except for a Ser(P) 154 peptide that models an unexpected MS/MS fragmentation.

Monitoring of Cysteine Residues
To facilitate detection of the 13 ER␣ cysteine residues, protein samples were reduced and alkylated. For the important DBD, consecutive digestion by Lys-C and Asp-N gave four major peptides, each containing two cysteine residues. Cys 381 and Cys 417 were not identified in Lys-C/Asp-N digests of MCF-7-derived ER␣, although Cys 417 was identified in rER␣ following chymotryptic digestion. Cys 530 was observed only in Glu-C digests, and Cys 381 was not observed at all. Our studies on the oxidation of cysteine residues in the DBD have been described elsewhere (8,10,11).

Observation of Acetylated Amino Acids
No identified peptides contained the N-terminal methionine residue predicted by the gene sequence. Broadening the search criteria to allow for loss of this residue together with N-terminal acetylation led to the identification of Ac-2 TMTLHTK in tryptic and Lys-C/Asp-N digests. The LC-MS/MS spectrum of the MH 2 2ϩ is illustrated in Fig. 4A: b 1 and b 2 ions confirm N-terminal acetylation, whereas y 2 and y 3 exclude the possibility of the acetyl group being located on Lys 8 . Ac-2 TMTLHTKASGMALLHQIQGNE was also identified with a Mascot score of 72 (Glu-C digest; spectrum not shown). Thus there is strong evidence that endogenous MCF-7 ER␣ undergoes this common modification, which would render it blocked for Edman sequencing.
Two tryptic peptides having similar MS/MS spectra differed in mass by 42 Da, suggesting another acetyl group. Alternatively this mass difference could correspond to trimethylation, which could not be differentiated with the instruments used here. Several ER␣ acetylation sites have been reported (Table  I), but a methylation was also detected recently (28). For the current discussion, it is assumed the modification is an acetylation. vMALDI-LTQ fragmentation of the unmodified MH ϩ (m/z 1263.4) identified it as 468 SLEEKDHIHR with sequence ions y 4 , y 6 , y 7 , y 7 ϪNH 3 , y 8  consistent with a centrally located modification. MS 3 of both y 4 ions gave identical spectra for both the non-acetylated and acetylated forms (inset). Thus the acetyl group is within EKD, placing it on the side chain of Lys 472 .

Phosphopeptide Identification by MS/MS
Several phosphopeptides were identified in this study (Table II), some confirming previously known sites as well as several that were previously unreported. As a potential aid to phosphopeptide analysis, three Web-based algorithms, Scansite, NetPhos, and Disphos, were used to predict phosphorylation sites, but the correlation between prediction and experiment proved weak (results not shown). MS/MS analysis was performed on digested ER␣ peptides without any prior phosphopeptide enrichment step. Consequently it was often possible to observe both the phosphorylated and unphosphorylated species. In LC-MS/MS the phosphopeptides were consistently eluted later than the unphosphorylated forms, providing additional chromatographic confirmation of their existence. Phosphopeptides usually resulted in pairs of pep- tides separated by 80 Da plus y n Ϫ 98 ions for the higher mass species. Intact phosphate-containing y n ions were seen infrequently.
Complementary peptide information was obtained by automated vMALDI-LTQ-MS/MS experiments, initially based on MALDI-TOF-generated inclusion lists. Manual data acquisition was also used to sequence peptides with low ion counts not initially observed by MS and to isolate weak precursor ions and monitor peptides not detected by LC-MS/MS analysis. vMALDI-MS n also allowed repeated analysis of the same sample as well as extensive MS n interrogation of ions, whereas LC analysis was limited to the brief time during which a peptide eluted.
All ER␣ peptides containing unmodified serine, threonine, or tyrosine residues were interrogated by vMALDI-LTQ-MS/MS at 80 m/z units above the observed molecular ion even if no ions were seen at this mass. Additionally hypothetical phosphopeptide m/z values were generated by in silico proteolysis of ER␣ and probed for putative phospho-Ser, -Thr, and -Tyr residues in both recombinant and endogenous ER␣ by multiple stage vMALDI-MS n . The loss of phosphoric acid (98 Da) in the linear ion trap provided a sensitive readout for the presence of a phosphopeptide; thus m/z regions that might correspond to (MH Ϫ 98) ϩ ions were examined for indications of phosphorylated residues using MS 3 to probe for fragment ions that correlated with the MS 2 spectra derived from the corresponding unphosphorylated parent ion.

Systematic Search for Phosphorylation Sites
The N-terminal Domain-Serine phosphorylation within the N-terminal domain containing activation factor-1 (AF-1) con-tributes to ER␣ activation and may lead to drug resistance in breast cancer treatment. We recently reported Ser 154 as a new phosphorylation site within the N-terminal domain of ER␣ from cultured human breast cancer cells. Here we found no other new sites but observed known sites at Ser 102/104/106 , Ser 118 , and Ser 167 . Data for Ser(P) 118 and Ser(P) 167 were reported previously (27) and are not presented here.
We sought to confirm whether each of the previously reported sites Ser 102/104/106 can be phosphorylated; the data is presented in supplemental Fig. 1. The analysis was on an MCF-7 ER␣ chymotryptic peptide containing all three residues, 90 GSNGLGGFPPLNSVSPSPL. LC-MS/MS spectra in supplemental Fig. 1, A 22) in the phosphorylated case, localizing the modification on Ser 106 . All y ions above y 2 show the loss of phosphoric acid, but y 4 alone definitively locates the site of the modification. Supplemental Fig. 1C shows a weaker MS/MS spectrum for an additional isomer eluting later (28.59 min). The y 3 /y 4 ions show no phosphorylation, whereas y 5 has lost phosphoric acid, suggesting phosphorylation of Ser 104 .
Crystal heterogeneity of unseparated digests can cause some separation of isomers, and different regions of the MALDI spot may yield distinct spectra. Several spectra from phosphorylated 90 GSNGLGGFPPLNSVSPSPL indicated multiple isomeric forms. Supplemental Fig. 1D presents a vMALDI-LTQ-MS/MS spectrum with (MH Ϫ 98) ϩ as the base  Peptides possessing both two and three phosphate groups prove that multiple serine residues can be phosphorylated. Supplemental Fig. 1E shows vMALDI-MS/MS of a doubly phosphorylated peptide of m/z 1956.5. The y 11 and y 11 Ϫ 98 ions confirm two phosphates C-terminal to Phe 97 not involving Ser 91 , whereas b 15 Ϫ 98 and b 16 Ϫ 98 Ϫ H 2 O ions suggest that Ser 106 is not phosphorylated. Supplemental Fig. 1F shows a triply phosphorylated species of m/z 2036.6. The stoichiometry for triple phosphorylation is clearly low, but three consecutive neutral losses of 98 Da confirm all three serine residues can be phosphorylated simultaneously.
The DBD-A new phosphorylation site was detected at Ser 212 within the loop between the two zinc fingers of this domain. Two Asp-N peptides separated by 80 Da gave similar fragments in vMALDI-LTQ-MS/MS. The unphosphorylated MH ϩ (m/z 1778.9) showed losses of water and ammonia and weak backbone fragments even at increased collision energy (Fig. 5A). Fragment ions b 9 , b 13 , b 14 , y 7 , y 12 Ϫ NH 3 , y 13 , and y 13 Ϫ NH 3 identified the peptide as 203 EGCKAFFKRSQGHN with cysteine modified by iodoacetamide. The heavier mass MH ϩ (1858.8) gave a weaker spectrum (Fig. 5B) but showed loss of phosphoric acid (Ϫ98 Da) and additional losses of water and ammonia. An N-terminal fragment common to both spectra was identified as b 14 Ϫ H 2 O for the unmodified peptide and b 14 Ϫ H 3 PO 4 for the phosphopeptide. The heavier molecular ion also retained phosphate but lost combinations of water, ammonia, and carbon dioxide. The peptide is clearly phosphorylated, and the only possible location for a phosphate group in this peptide is Ser 212 .
It was hypothesized that such a phosphopeptide with Nterminal glutamate might undergo the combined loss of phosphoric acid and water to form a new ionic moiety containing dehydroalanine and pyroglutamate. Identical fragmentation was observed for an analogous peptide incorporating the known Ser 154 site. Fig. 5C shows MS/MS of the unmodified peptide of m/z 1305.6 with (MH Ϫ NH 3 Ϫ H 2 O) ϩ as the base peak plus b 11 , b 9 , b 9 Ϫ H 2 O, b 7 Ϫ H 2 O, b 6 Ϫ H 2 O, y 11 , y 10 , and y 9 that confirm this peptide as 143 EAGPPAFYRPNS. Fig.  5D shows the spectrum of the phosphopeptide at m/z 1385.6. Again the base peak corresponds to combined losses of phosphoric acid and water, and other prominent peaks combine losses of phosphoric acid, H 2 O, NH 3 , and CO 2 . The only significant backbone fragment (b 9 ) also shows water loss; thus such fragmentation appears to characterize phosphoserine peptides with N-terminal glutamate.
A previously reported phosphorylation site falling within the second zinc finger of the DBD at Ser 236 was confirmed here in rER␣ only. LC-MS/MS of a Lys-C digest of rER␣ identified the triply charged peptide 236 SCQACRLRK (with iodoacetamidederivatized cysteines) with and without serine phosphorylation; the modified form eluted half a minute later. Supplemental Fig.   2A  . Supplemental Fig. 2B shows the corresponding spectrum of the phosphorylated MH 3 3ϩ (m/z 420.2) with base peak corresponding to the loss of phosphoric acid (triply charged) and y 6 2ϩ , y 7 2ϩ , and y 8 2ϩ ions that locate the modification at Ser 236 .
The Hinge Region between the DBD and AF-2-Although no phosphorylation has been reported for this region, residues 263-301, we identified Ser 294 as a phosphorylation site in MCF-7 cells. Fig. 6A shows the vMALDI-LTQ-MS/MS of MH ϩ of a chymotryptic peptide (m/z 1124.8) for which b 4 Ϫ NH 3 , a 5 Ϫ NH 3 , b 5 Ϫ NH 3 , a 6 Ϫ NH 3 , b 6 , b 6 Ϫ NH 3 , b 8  NH 3 , b 9 , b 9 ϩ H 2 O, and MH ϩ Ϫ 2NH 3 identify 287 RAANLW-PSPL. The N-terminal arginine is consistent with predominant b ions. Fig. 6B for the analogous species 80 Da heavier (m/z 1204.5) shows (MH Ϫ 98) ϩ at m/z 1106.5, which was selected for MS 3 analysis (Fig. 6C). This gives identical fragments to MS 2 of the unmodified peptide for ions b 6 and below, but b 8 , b 8 Ϫ NH 3 , b 9 , b 9 ϩ H 2 O, and (MH Ϫ 2NH 3 ) ϩ are all 18 Da lower, confirming that the phosphate is situated between the b 6  No phosphorylation sites were identified within AF-2, but we found two novel phosphorylation sites in the F-domain at Ser 554 and Ser 559 . The assignment of Ser 554 was based only on observations made with recombinant protein, not with MCF-7 protein, so it is illustrated in supplemental Fig. 3. Panel A shows a vMALDI-MS/MS spectrum, and panel B shows an LC-MS/MS spectrum for a tryptic phosphopeptide of sequence 549 LHAPTSR phosphorylated at either Thr 553 or Ser 554 . Both spectra show strong fragment ions, including the loss of phosphoric acid. Only the vMALDI spectrum contains a strong b 5 fragment ion that confirms the phosphate group to be on Ser 554 rather than Thr 553 . Fig. 7 shows MS/MS spectra that identify phospho-Ser 559 in MCF-7-derived ER␣ digested with trypsin. Fig. 7A shows the LC-MS/MS spectrum of the unmodified peptide MH 3 3ϩ (m/z 862.4) eluting at 26.4 min and identified as 556 GGAS-VEETDQSHLATAGSTSSHSLQK. Fig. 7B shows the corresponding spectrum of the modified form (m/z 889.1) eluting at 27.1 min. This 26-residue peptide contains six serines and three threonines, any one of which could potentially be the phosphorylation site. Comparing the two spectra, the similarity of the y-series ions indicates that the modification is not in the C-terminal region, whereas strong b 4 -b 7 ions show the loss of phosphoric acid, which unambiguously establishes Ser 559 as the site of the modification. The same peptide was also observed by vMALDI. Fig. 7C shows the vMALDI-MS 2 spectrum of the unmodified peptide at m/z 2585.3, Fig. 7D shows the species 80 Da heavier at m/z 2665.3, and Fig. 7E represents the MS 3 spectrum of the (M Ϫ 98) ϩ at m/z 2567.2. Although these spectra show numerous fragment ions that clearly define the sequence, no unique ions identify the specific location of the phosphate group, unlike the LC-MS/MS data.

Quantitation of Ser 294 Phosphorylation by LC-MRM/MS
To determine the biological relevance of these previously unreported phosphorylation sites an MRM/MS assay was developed to monitor any change in Ser 294 phosphorylation following estradiol treatment of MCF-7 cells. MRM is the most sensitive and robust quantitative mass spectrometric method because of (i) coincident detection of both the precursor ion (Q1) and a designated fragment ion (Q3) and (ii) rapid cycling that allows each transition to be sampled multiple times during chromatographic elution of the selected species. Fragment ions with m/z values greater than that of the precursor ion lead to lower noise levels, which requires that the fragment ion has a lower charge state than the precursor. The LC-MS/MS spectra in Fig. 8, A and B, of tryptic peptide 288 AAN-LWPSPLMIK with and without phosphorylation of Ser 294 show that the most abundant fragment ions that fit this criterion are y 7 Ϫ phosphoric acid (m/z 767.48) from the doubly charged Ser 294 phosphorylated peptide (m/z 710.9) and y 7 (m/z 785.47) from the doubly charged unmodified peptide (m/z 670.9).
Because of the high specificity of MRM, gel isolation of the immunoprecipitated ER␣ from background antibody proteins, protein A, and other contaminants was not required before trypsin digestion. This eliminated the sample losses typical of in-gel digestion and also minimized methionine oxidation, significantly increasing the peptide signal strength as this was not split between the oxidized and non-oxidized forms. Fig.  8C illustrates MRM data for the doubly protonated molecular ion fragmenting to y 7 for unmodified and phosphorylated Ser 294 peptides from immunoprecipitated ER␣ obtained from control conditions and following 30-min incubation with 10 nM estradiol. As expected, the phosphopeptide eluted 1-2 min later than the unmodified peptide. The increased peak heights following estradiol treatment relative to control conditions clearly indicate significant induction of phospho-Ser 294 . The -fold induction shown in Fig. 8D was obtained by dividing the phospho-Ser 294 integrated peak areas by the corresponding Ser 294 integrated peak areas normalized to the value obtained for the control sample. Three biological replicates demonstrated a mean 27.8-fold induction of phosphorylation following estradiol treatment relative to control with a standard deviation of 9.8.

DISCUSSION
Previous comprehensive mapping of ER␣ and its PTMs by mass spectrometry had been hampered by the inability to purify/enrich sufficient endogenous protein. The optimization of ER␣ immunoaffinity enrichment from MCF-7 cells for MS analysis reported here is a significant achievement in ER␣ biochemical studies as demonstrated by our recent observation of the novel phosphorylation site Ser 154 in the N-terminal domain of ER␣ (27).
Several PTMs of ER␣ summarized in Table II and Fig. 9 were identified in this tandem mass spectrometry study, which focused on detection and characterization of phosphorylation sites. The majority of phosphorylation sites identified here occur in endogenous protein isolated from the estradiolstimulated MCF-7 human breast cancer cell line (Ser 102 , Ser 104 , Ser 106 , Ser 118 , Ser 154 , Ser 167 , Ser 212 , Ser 294 , and Ser 559 ), although two were observed only in recombinant protein (Ser 236 and Ser 554 ). Ser 154 was a novel phosphorylation site previously detected by us using the same methods, but several other sites had not been reported previously: namely Ser 212 , Ser 294 , Ser 554 , and Ser 559 . Most or all of these serine phosphorylations are likely estradiol-dependent, although evaluation of the induction level and functional consequences for these various residues are not yet completed. However, in regard to Ser 294 , Fig. 8 demonstrates that estra- diol stimulation of MCF-7 cells produced an approximate 28-fold phosphorylation enhancement of this residue, strongly implicating its functional relevance to ER activation. It is interesting to note that an early study examining the phosphorylation of ER expressed in COS-1 cells failed to detect Ser 294 phosphorylation (29). We note that our studies differ from these initial studies in several key aspects including our inclusion of the potent phosphatase inhibitor okadaic acid in the immunoprecipitation buffer; the use of MCF-7 cells, which endogenously express ER instead of the more artificial system Previously reported phosphorylation sites Ser 305 , Thr 311 , and Tyr 537 were not confirmed here, and Ser 236 phosphorylation was observed only in recombinant protein. However, Ser 236 is phosphorylated by cAMP-dependent protein kinase (ligand-independent) so 30-min treatment with estradiol may not induce phosphorylation in MCF-7 cells, although 1-5-min exposure to estradiol might activate the cAMP-dependent protein kinase pathway through membrane ER activation (15). Residue 305 is in a region for which no peptide was observed so no data were obtained to support or refute its phosphorylation. However, it is phosphorylated by PAK1 (growth factors heregulin and epidermal growth factor induce PAK1 activation), which again is ligand-independent (30). Unmodified Thr 311 was identified by LC-MS/MS in chymotryptic peptide 309 SLTADQMVSALL with good mass accuracy (an error of 19 ppm) and a significant Mascot score of 35 and probably would be observed as phosphorylated if that actually occurred. This residue was reported phosphorylated by p38 kinase and was estradiol-dependent in cells overexpressing ER and tagged p38 using in vitro kinase assays (31). We observed unmodified Tyr 537 peptides with trypsin (532-548), Glu-C (524 -538), and Asp-N (527-537), so any phosphorylation of this residue would probably be observed, but phosphorylation may not be induced by 30-min estradiol treatment as it may be Src-dependent (32).
This study demonstrates the sensitivity of tandem mass spectrometric methods for detection of phosphorylation sites in low level proteins such as ER␣, and it reiterates the advantages of using complementary methods of ionization and analysis as no single method was able to detect all the modifications reported here. Additional modifications detected in ER␣ from MCF-7 cells were N-terminal cleavage of methionine accompanied by acetylation of Thr 2 and side chain acetylation or possibly trimethylation of Lys 472 .