Combinatorial Modification of Human Histone H4 Quantitated by Two-dimensional Liquid Chromatography Coupled with Top Down Mass Spectrometry*

Quantitative proteomics has focused heavily on correlating protein abundances, ratios, and dynamics by developing methods that are protein expression-centric (e.g. isotope coded affinity tag, isobaric tag for relative and absolute quantification, etc.). These methods effectively detect changes in protein abundance but fail to provide a comprehensive perspective of the diversity of proteins such as histones, which are regulated by post-translational modifications. Here, we report the characterization of modified forms of HeLa cell histone H4 with a dynamic range >104 using a strictly Top Down mass spectrometric approach coupled with two dimensions of liquid chromatography. This enhanced dynamic range enabled the precise characterization and quantitation of 42 forms uniquely modified by combinations of methylation and acetylation, including those with trimethylated Lys-20, monomethylated Arg-3, and the novel dimethylated Arg-3 (each <1% of all H4 forms). Quantitative analyses revealed distinct trends in acetylation site occupancy depending on Lys-20 methylation state. Because both modifications are dynamically regulated through the cell cycle, we simultaneously investigated acetylation and methylation kinetics through three cell cycle phases and used these data to statistically assess the robustness of our quantitative analysis. This work represents the most comprehensive analysis of histone H4 forms present in human cells reported to date.

Histones are a class of proteins around which DNA is wrapped and packaged inside a eukaryotic nucleus. Two molecules of each core histone H2A, H2B, H3, and H4 together with ϳ146 bp of DNA form the fundamental unit of chromatin called the nucleosome. These proteins are heavily modified, with combinations of these enzymatic modifications thought to form a "histone code" orchestrating epigenetic processes such as long-term gene silencing and gene activation (1), higher level chromatin packaging (2), and DNA repair mechanisms (3). All of these activities change with relation to the cell cycle, a sequence of events during which a cell commits to DNA replication (G 1 ), replicates its DNA (S), prepares for mitosis (G 2 ), and undergoes cell division (M) (4). Histone synthesis and deposition are largely coupled to DNA replication during S phase (5). As a cell doubles its nuclear DNA, there is a concomitant doubling of the content of histones and nucleosomes. Even though antibodies have been used to track single modifications, the fate of preexisting histone modifications and the acquisition of new histone modifications during the cell cycle is not well understood because this approach is unable to distinguish previously modified forms from newly modified ones (6 -8). However, an epigenetic mechanism presumably exists to faithfully transmit patterns of histone modification and chromatin structure to ensure normal cellular function over successive generations (9).
Dynamic changes in the PTMs 4 affecting the N-terminal tails of the core histones, which comprise ϳ25-30% of their individual masses, are thought to modulate their interactions with both DNA and other proteins (10). Great efforts have been made to catalogue the types and sites of modifications and ascertain their biological contexts. For example, it has been asserted that methylation of lysine 9 in human H3 is a mark for heterochromatin by specifically interacting with heterochromatin-associated protein 1 (11,12). Nevertheless, the true biological significance of such observations remains to be defined in the absence of quantitative data regarding modification type, co-occurrence with other PTMs on the same histone, and abundance in bulk chromatin. Antisera have been used to provide qualitative descriptions of the genomic distributions of specific histone PTMs in various organisms (e.g. Refs. 13 and 14). However, a quantitative profile of the intact histone forms that populate the ϳ32 million nucleosomes packaging the human genome has yet to be reported.
There are many known sites and types of post-translational modification on human histone H4: N-terminal acetylation (a␣Ser-1), Ser-1 phosphorylation, Arg-3 mono-and dimethylation (mArg-3, 2mArg-3), lysine 5,8,12, and 16 acetylation (aLys-5, aLys-8, aLys-12, aLys- 16), and lysine 20 mono-, di-, and trimethylation (mLys . Current methods used to characterize and quantify histone H4 PTMs paint only part of the nuclear picture. For example, antibodies, which are specific to one modification, can detect single-site modifications (e.g. mArg-3), the abundance of which in some cell types may be below the detection limit of MS. "Bottom-up MS" allows the detection of combinations of modifications only when they reside on the same peptide (15). The use of peptide mass fingerprinting relies on mass accuracy instead of fragmentation for PTM detection (16). None of these approaches is able to provide a simple measure of percent occupancy of a given PTM or the percentage of its parent histone form in a molecular population. To unambiguously determine the combinations of PTMs on histone proteins, the intact protein must be processed and analyzed such that the identity and regiospecificity of the PTM(s) is restricted to a single known, pre-selected mass. Recent advances in Top Down mass spectrometry (TDMS) and dissociation methods have made this approach possible (17,18). For example, TDMS has shown that the intact MS profile of reverse-phase purified H2A and H2B are comprised of mainly unmodified variants (19,20), whereas H3 and H4 are comprised mainly of forms harboring PTMs (21,22).
Hydrophilic interaction chromatography (HILIC) has been shown to separate allelic variants of histone H1 (23) and modified forms of murine H4 (24,25). Lindner and co-workers (24,25) successfully identified the major H4 PTMs present in the most abundant HILIC peaks using bottom-up MS, categorizing and localizing some of the PTMs on H4. More recently, our laboratories have successfully utilized a HILIC/TDMS approach to identify more than 150 different combinations of post-translational modifications on histone H3.2 (26).
Here, we extend previous work by identifying additional H4 forms of human H4 via Top Down MS and validating a quantitative method used to assess abundance of human histone H4 from asynchronous and cycling cells. We extend our quantitative approach (27) to combinatorially modified forms of histone H4, detected at Ͻ0.02% of the total population. In this way we are able to assess abundance levels of 42 unique combinations of PTMs on H4 in asynchronous HeLa cells. In addition, we report the cell cycle dynamics of seven histone H4 protein forms, identify two novel H4 forms, and uncover correlations between methylation and acetylation in cycling cells.

EXPERIMENTAL PROCEDURES
Histone Preparation: Reverse Phase and Hydrophilic Interaction Liquid Chromatography-RP-HPLC-purified H4 was prepared by chromatography of ϳ100 g of crude HeLa S3 histone protein (with or without prior oxidation as described previously (28,29)). Fractions of H4 resolved by HILIC from three individual preparations of asynchronous HeLa cells were analyzed by Fourier transform MS (FTMS). The first experiment was performed without intentional oxidation. For the second experiment, each individual HILIC fraction was intentionally oxi-dized before MS analysis (29). For the third experiment, crude histone was intentionally oxidized before RP-HPLC (28), and the fully oxidized histone H4 was collected and used for the subsequent HILIC purification. HILIC fractions of H4 were prepared by fractionating ϳ150 g of RP-HPLC purified H4 (pooled from several RPLC runs) as described previously (24)  Mass Spectral Analysis by Electrospray Ionization-FTMS/MS-All data were acquired on a custom 8.5 Telsa Quadrupole-FTMS hybrid with an electrospray ionization source operated in positive ion mode as described previously (29). Residual chloride or phosphate contaminating some samples was partially removed by applying Ϫ25 V on the accumulation octupoles. Electron capture dissociation (ECD) was performed by applying 5 A through a dispensor cathode filament (Heatwave Technologies, Crescent Valley, BC). During the ECD event, ϳ10 V were applied on the grid potential whereas ϳ9 V were sent through the filament for optimal ECD. Typically, 300 cycles of ECD were performed, with individual irradiation times of 5 ms and a 10-ms relaxation time between cycles. All relative molecular weight (M r ) values and fragment masses are reported as neutral, monoisotopic species. ECD MS/MS spectra were internally calibrated using 3-4 identified z . ions from the unmodified C-terminal region.
Quantitation of Modified Forms of Intact H4-The molecules remain intact so that all possible combinations of modifications are present when analyzed by TDMS. With this in mind, the HILIC chromatogram was integrated, and each fraction was assigned a percent area relative to the entire elution area of H4. This relative area was then multiplied by the percent of each observed mass in the mass spectrum associated with that fraction. If multiple species were present, the MS intensity values were used to calculate the relative ratios for all protein ions observed (so called "PIRR" (27)). Each PIRR was then multiplied by the relative HILIC area percentage. This value, the intact mass percentage, reports solely on the relative percentage of each observed intact mass and does not depend on the absolute abundances in the MS. The intact mass percentage can be further dispersed if isobaric or isomeric components exist (e.g. a mass corresponding to monoacetylated H4 could theoretically be a mixture of acetylation at Lys-5, -8, -12, or -16). Fragmentation of a single intact mass comprised of isobars or isomers will produce fragment ions pairs. As previously demonstrated in our laboratory, the MS/MS intensity values of these fragment ion pairs (so called "FIRRs") can be used to determine the iso-meric composition with an error of Յ5% for species generating moderate to high MS signals (27). After isomeric composition is determined, the relative percent of each protein form can be determined for that one fraction. If a form elutes over multiple fractions, multiplying the percentages from MS (PIRRs and FIRRS) and chromatography gives a value for total percentage for that form. A comprehensive list of all observed histone H4 modifications and their respective abundances from Fig. 1 is shown in Table 1.
For calculating PIRR values for histone H4, the intensity of the four most abundant isotopes ( 13 C 5 , 13 C 6 , 13 C 7 , and 13 C 8 ) were integrated to calculate intact abundance ratios for the 14ϩ and/or 12ϩ charge states. When possible, these were combined into a weighted average to report a final PIRR value for the sample, which was then used to calculate the relative amount of the corresponding intact H4 form relative to the sum of the total. ECD generated 1ϩ, 2ϩ, and 3ϩ fragment ions, from which the 13 C 0 , 13 C 1 , and 13 C 2 (when observed) isotopes were integrated and used to calculate the FIRRs.
Software-A Web-based software and data base suite, Pro-Sight PTM (21,30), was used to accelerate the characterization of histone H4 protein forms as previously reported (21). An adapted version of THRASH (31) was used to convert raw data into monoisotopic mass values.
Analysis of H4 from Synchronized Cells-HeLa S3 cells were grown in Jokliks medium supplemented with 10% newborn calf serum and maintained at a density of 2-3 ϫ 10 5 /ml before synchronization using the double thymidine block procedure as described previously (29). The degree of synchrony was assessed by flow cytometry every hour after the second release. The representative G 1 /S, mid-S, late S/G 2 , M/G 1 , and mid-G 1 phases were determined to occur 0, 4, 8, 10.5, and 14 h postrelease, respectively. To enable a rigorous comparison of H4 from different cell cycle stages, H4 was prepared from cells collected at 0, 8, and 14 h post-release in three independent synchronizations. Approximately 150 g of RP-HPLC-purified H4 was then further purified for each time point by HILIC as described above. The HILIC chromatogram, MS, and MS/MS data were used to calculate the abundance of unacetylated and monoacetylated H4 as described above. In a few cases, a stored wave-form inverse Fourier transform (SWIFT) isolated window (Ͻ1.5 m/z) contained salt-adducted forms of a lower mass species. The amount of contamination was determined by fragment ions larger than c 20 . The FIRRs were then corrected by subtracting out this contamination. Of the 93 HILIC fractions analyzed, one showed no measurable protein within the detection limit. The chromatographic area of this peak, representing an estimated ϳ0.7% of the most abundant peak, was subtracted of the whole and removed from further calculations and analysis.

RESULTS AND DISCUSSION
Hydrophilic-interaction Liquid Chromatography-Using HILIC conditions similar to those described previously (24,25), we separated multiply modified histone H4 forms from asynchronous HeLa S3 cells. To reduce sample complexity, each 1-min fraction was analyzed individually (shaded bands in Fig.  1A). The MS of each fraction contained on average two distinct masses typically 14 Da apart (Fig. 1B). A cursory examination of the intact mass in each HILIC fraction revealed an acetylationand methylation-dependent elution profile. The most highly modified (e.g. the tetraacetylated, dimethylated H4) and, therefore, most hydrophobic form eluted earliest, whereas the least hydrophobic form (e.g. N-terminal acetylation only) eluted last. In general, H4 separated primarily by acetylation state, generating five groups of resolved peaks (Fig. 1A). Each one of these groups was further broken down into a series of peaks which included (in order of elution) tri-, di-, mono-, and unmethylated Lys-20. Mono-and dimethylated Arg-3 (with 2mLys-20) were observed to coelute with Lys-20 trimethylation (see Fig. 1 and Table 1). Intact mass isolation using a SWIFT followed by ECD fragmentation identified the modifications on and relative abundances of 42 H4 forms (Fig. 2, Table 1, and see below).
Unacetylated, Methylated H4-It is well known that all H4 molecules are co-translationally acetylated on the ␣-NH 2 of Ser-1 (56,57), resulting in a ⌬m of at least ϩ42 Da for all forms of H4. This ϩ42 Da form will be referred to as the "unacetylated" form of H4 (a␣Ser-1), whereas the term "monoacety-FIGURE 1. PTM-dependent fractionation of histone H4 by HILIC enables detection of minor multiply modified forms. RP-H4 prepared from asynchronously growing HeLa S3 cells was fractionated on a Poly CAT A column essentially as described previously (25). A, this chromatogram is representative of three HILIC experiments performed. Top Down MS analysis of every fraction throughout the elution volume of H4 forms revealed that they generally eluted in order of increasing hydrophilicity with relative retention inversely related to their degree of ⑀-N-acetylation and Lys-20 methylation. The entire complement of H4 forms characterized and quantitated after HILIC fractionation is listed in Table 1. B, TDMS of fractions 36 -57 show multiple intact masses corresponding to 0, 1, or 2 internal acetylations and 0, 1, 2, or 3 methylations.
lated" H4 will be referred to as acetylation at only one internal acetylation site (Lys-5, -8, -12, or -16). The a␣Ser-1 molecules eluted the latest and ranged from ϳ51 to 58 min ( Fig. 1). Three unique masses with mass shifts (⌬m values) of ϩ42, 56, and 70 Da, corresponding to pure (non-isomeric) forms of a␣Ser-1, a␣Ser-1 ϩ mLys-20, and a␣Ser-1 ϩ 2mLys-20, respectively, were identified. The Lys-20, mLys-20, and 2mLys-20 forms comprised 2, 6, and 65, respectively, in bulk chromatin ( Table  1). The a␣Ser-1 ϩ 2mLys-20 form was the most abundant (ϳ65%), in close agreement with previously recorded FTMS spectra of asynchronous HeLa RP-H4 (21,29). A fourth mass was found to have a ⌬m of ϩ84 Da, which is listed as 3m in fraction 51 (Fig. 1). In addition to the a␣Ser-1 (ϩ42 Da), the other ϩ42 Da could be a mixture of monoacetylation (42.011 Da) or three methylations (42.047 Da). SWIFT-isolation followed by ECD fragmentation of the ϩ84 Da form generated two forms of the c 16 2ϩ ion, which contains all possible acetylation sites, with observed ⌬m values of ϩ42 (a␣Ser-1) and ϩ56 (a␣Ser-1, mArg-3), indicating that no internal acetylations were present (data not shown). Furthermore, the c 22 3ϩ ion was observed to have a ⌬m of ϩ84 Da, indicating that by the last modifiable residue, Lys-20, all of the mass difference was taken into account, which necessitated that all modifications must be on the first 20 residues. Therefore, taking all mass shifts into consideration, we can say with confidence that this isomeric species contained a ϳ1:1 mixture of a␣Ser-1 ϩ mArg-3 ϩ 2mLys-20 and a␣Ser-1 ϩ 3mLys-20. As expected from the area underneath their HILIC peak, the mArg-3 ϩ 2mLys-20 and the 3mLys-20 were the rarest of the unacetylated but methylated H4, each accounting for ϳ1% of the global abundance of H4 PTMs. Strikingly, no forms with phosphorylated Ser1 were observed. M-phase arrest of HeLa with colchicine does build the occupancy of this PTM up to ϳ5% (29). Considering this PTM, there are at most ϳ84 forms of histone H4 present in human cells, with 75 of these detected by the Coon Laboratory from differentiating embryonic stem cells. 5 Monoacetylated, Methylated H4-The second most abundant forms of H4 are those carrying one internal acetylation, which eluted between ϳ43 and 50 min ( Fig. 1; see Fig. 5). Surprisingly, many chromatographic peaks were observed, unlike the unacetylated region where the four major chromatographic peaks corresponded to the four methylated states of unacetylated H4. The molecular masses did not change by greater than 5 J. Coon, personal communication. ϩ28 Da over the course of elution (Fig. 1B, middle panel), suggesting that the location of the single acetyl group at either Lys-5, -8, -12, or -16 may have an effect on elution of different monoacetylated isomers of H4. Indeed, distinct elution differences between the four monoacetylated sites (with the same number of methylations) were observed (supplemental Fig.  2A). We also found that the abundance of aLys-5 Ϸ aLys-8 Ͻ aLys-12 Ͻ aLys-16, a trend also noted in other studies of asynchronous and cycling cells (15,(32)(33)(34). The "zipper model" states that acetylation occupancies are highest at aLys-16 followed by aLys-12 Ͼ aLys-8 Ͼ aLys-5. This trend is clearly evident from the data shown in Table 1 and Fig. 5 (15). However, when the acetylation site occupancy is considered relative to the degree of Lys-20 methylation in the same molecule (Fig. 3), it becomes obvious that the widely accepted zipper model of H4 acetylation may not hold the same weight for different Lys-20 methylation states. The most striking example of this is shown by Fig. 3A when comparing the degree of acetylation occupancy of unmethylated and dimethylated Lys-20 molecules. Although the aLys-16 Ͼ aLys-12 Ͼ aLys-8 Ͼ aLys-5 hierarchy exists for both un-and dimethylated Lys-20 molecules, the magnitude of this trend differs dramatically between these two methylated forms. From these data it appears that the frequency of acetylation at sites more distal to Lys-20 (i.e. Lys-5, -8, or -12) is inversely related to the degree of Lys-20 methylation. In contrast, the frequency of acetylation at Lys-16 appears to be directly related to the degree of Lys-20 methylation (Fig. 3). This trend is also observed for monoacetylated forms during G 1 /S, S/G 2 , and G 1 (discussed below). These findings suggest that site preferences for either acetyltransferases or deacetylases are influenced by the Lys-20 methylation state. Methylation at Arg-3 (with dimethylation at Lys-20) and trimethylation at Lys-20 are observed at low levels on unacetylated H4 molecules ( Table 1). Methylation of H4 at Arg-3 is associated with transcriptional coactivation by protein arginine methyltransferase 1 (PRMT1) (35,36), but the nature of this mark has not been unequivocally established. Conventional MS/MS of H4 peptides prepared from asynchronously growing 293T cells detected only monomethyl Arg-3 (35). The extent to which mono-and dimethylation of Arg-3 is attributable to PRMT1 or other PRMTs is unknown. However, because most PRMTs are thought to catalyze dimethylation (37), the monomethyl Arg-3 we detected presumably arises either from a previously unrecognized ability of PRMT1 or other PRMTs to catalyze monomethylation or via partial demethylation of dimethylated Arg-3. Because our analyses revealed that methylation itself was the only alteration made to the structure of arginyl side chains in the a␣Ser-1 ϩ mArg-3 ϩ 2mLys-20, a␣Ser-1 ϩ mArg-3 ϩ aLys-16 ϩ 2mLys-20, and a␣Ser-1 ϩ mArg-3 ϩ aLys-12 ϩ aLys-16 ϩ 2mLys-20 forms, the latter case would have to involve a mechanism other than deimination or demethylimination FIGURE 2. RP-HPLC followed by HILIC allows a >10 4 dynamic range for detection of H4 forms. A, fraction 41 contains ϳ0.07% of the total observed histone H4 area in the HILIC chromatogram (Fig. 1). B, FTMS of fraction 41 from A shows the presence of at least three different intact masses. C, ECD of the isolated mass 11387 Da in B identifies at least two isomeric forms present at ϳ50% each. Using the MS and chromatographic signal levels, a dynamic range of ϳ7 ϫ 10 4 is possible, translating to the ability to detect a H4 form present in ϳ500 nucleosomes of the ϳ32 million total per cell. S/N, signal-tonoise ratio. described for protein arginine deiminase 4 (PAD4) (38,39). The finding that Arg-3 can be mono-or dimethylated (see below) suggests that these forms may be functionally distinct, as proposed for mono-, di-, and trimethylation of histone lysines (40,41). The fact that all six methylated Arg-3 forms (see Table 1) observed were dimethylated at Lys-20 may reflect co-operativity between methylases targeting Arg-3 and Lys-20. However, further analysis is necessary to exclude the possibility that Arg-3-methylated molecules with no or monomethylation at Lys-20 exist in lower abundance. Our finding that many Arg-3-methylated molecules were also mono-or diacetylated is consistent with the association of methylation at Arg-3 with transcriptional activation. Together with evidence from immunoblot analyses that acetylation inhibits Arg-3 methylation (36) and the recent observation that PRMT1-mediated methylation of H4 at Arg-3 is a prerequisite for acetylation of H3 and H4 in the chicken ␤-globin gene locus (42), our findings suggest that Arg-3 methylation by PRMT1 may facilitate subsequent acetylation, whereas prior acetylation may inhibit Arg-3 methylation. Lys-20 trimethylation was readily detected in unacetylated (a␣Ser-1 ϩ 3mLys-20), monoacetylated (a␣Ser-1 ϩ Lys-16 ϩ 3mLys-20), and diacetylated (a␣Ser-1 ϩ Lys-12 ϩ Lys-16 ϩ 3mLys-20) forms of asynchronous HeLa H4, consistent with previous analyses that detected Lys-20 trimethylation on unacetylated and monoacetylated forms of H4 from mouse, rat, and human cells in the absence of histone deacetylase inhibition (24,25). Together with our evidence that Lys-16 acetylation and Lys-20 mono-and dimethylation are not mutually inhibitory in vivo (Tables 1 and 2), the absence of Lys-20 trimethylation on tri-and tetra-acetylated forms of H4 lends some support for the hypothesis that the effects of Lys-20 trimethylation are distinct from those of mono-and dimethylation. Lindner and co-workers (25) have proposed that Lys-20 trimethylation is coordinated with gene silencing by histone deacetylases to recruit heterochromatin promoting factors involved in maintaining silencing, as suggested previously for trimethylation at Lys-9 and -27 of H3 (43)(44)(45). The coincidence of Lys-20 trimethylation and hypoacetylation of H4 observed here is consistent with the enrichment of Lys-20 trimethylated and hypoacetylated forms of H4 in constitutive heterochromatin observed in immunocytochemical analyses (46 -48). However, the co-occurrence of Lys-20 trimethylation and Lys-16 and -12 acetylation argues against a heterochromatin-promoting role for the 3mLys-20 mark. Despite these conflicting interpreta-tions, our finding that Lys-20-trimethylated H4 lacks Arg-3 methylation (Tables 1 and 2) further supports the notion that Lys-20-trimethylated H4 is enriched in silent chromatin as Arg-3 methylation is thought to be associated with transcriptional activation (35,36). The lack of 3mLys-20 and mArg-3 co-occurring on the same H4 molecule suggests that Arg-3 methylation is either precluded from Lys-20-trimethylated forms or is removed upon Lys-20 trimethylation (although it should be noted that we did not identify citrulline or other residues derived from arginine in any of the Lys-20-trimethylated species analyzed).
Achieving Ͼ10 4 Dynamic Range-One limitation for TDMS detection of H4 forms from reverse-phase purified H4 is its dynamic range of ϳ100. This dynamic range translates to a detection limit that precludes complete characterization of forms representing less than ϳ3% of the H4 pool (29). By using HILIC to separate rare forms (e.g. fraction 41, Figs. 1 and 2A) from the most highly abundant forms (e.g. fraction 52, Fig. 1), an enrichment of nearly 500fold was achieved. In fraction 41, a H4 form of mass 11,387 Da was observed with a ϳ75 signal-tonoise ratio (S/N, Fig. 2B). After SWIFT isolation and ECD fragmentation, this mass was determined to consist of two isomeric components, resulting in a further increase of the effective dynamic range (Fig. 2C, inset). Using the signal-to-noise ratio of the lowest level forms detected here, the approximate dynamic range of the combined 4-dimensional analysis reported here (RPLC, HILIC, MS, and MS/MS) was determined to be ϳ7 ϫ 10 4 . This translates to a projected ability to detect an H4 form present in ϳ500 of 32 million total nucleosomes in the human nucleus.  B and D, respectively. B, a monoisotopic mass of 11359 Da (corresponds to a ⌬m of ϩ98 Da, shaded gray) was subsequently isolated and fragmented by ECD. C, the MS/MS data show that a ⌬m of ϩ70 Da on fragment ions less than c 20 correspond to an N-terminal acetylation (ϩ42) and two methylations (ϩ28), with the latter localized to Arg-3. Fragment ions containing the first 20 residues or more were found to have a ⌬m of ϩ98 Da, localizing ϩ28 Da to a dimethlyated Lys-20. D, a monoisotopic mass of 11,289 Da (shaded gray) corresponds to a ⌬m of 28 Da, which is equivalent to 2 methylations. E, in a similar fashion to C, all fragments before Lys-20 showed no modification, whereas fragments after Lys-20 exhibit a ϩ28 Da shift. If an N-terminal acetylation was present, fragments before Lys-20 would exhibit a ϩ42 Da mass shift.
Two Novel Modifications: 1) 2mLys-20 and 2) a␣Ser-1 ϩ 2mArg-3 ϩ 2mLys-20-With the increased dynamic range described above, we then interrogated H4 from cells arrested at the G 1 /S border of the cell cycle (see below), generating the HILIC chromatogram shown in Fig. 4A. Of the many forms described above, two additional forms were observed; one with ⌬m ϭ 28 Da and the other with ⌬m ϭ 98 Da (Fig. 4). The former comprised dimethylation at Lys-20 with no N-terminal acetylation (Fig. 4, C and E). More experimentation needs to be done to probe possible functions for this unique form, which represented only 0.28% of total histone H4 at the G 1 /S border ( Table  2). A second rare H4 form, ⌬m ϭ 98 Da, was determined by ECD to be a␣Ser-1 ϩ 2mArg-3 ϩ 2mLys-20 and constituted 0.2% of the total H4 population (Figs. 4, B and D, and Table 2). However, although the a␣Ser-1 ϩ 2mArg-3 ϩ 2mLys-20 species were observed throughout the cell cycle, only at the G 1 /S time point was this purely unacetylated ϩ 2mLys-20 form observed.
Analysis of Unacetylated, Multiply Methylated H4 during the Cell Cycle-The fraction-by-fraction analysis of each HILIC separation revealed many unique H4 forms, some present at Ͻ1% levels in bulk chromatin. Of these forms, the identification of a high amount of Lys-20 methylation and isomeric acetylation begged the question of how these forms change during the cell cycle, as other groups have reported a correlation between the two events (6,49). For example, transcription of PR-Set7, the histone methyltransferase responsible for Lys-20 monomethylation, increases during G 2 /M phase and declines during the transition to G 1 (6). HeLa cells were synchronized by a double thymidine treatment and released from G 1 /S arrest after resuspension in fresh media. Our data and others have shown that after being released from thymidine, HeLa cells progress from G 1 /S (0 h) to S phase by 4 h, late S/G 2 by 8 h, M by ϳ10.5 h, and finally G 1 by 14 h (50). Because of the time requirements for each HILIC, MS, and tandem MS experiment, we limited our analysis to the following three cell cycle phases: G 1 /S (0 h), late S/G 2 (8 h), and G 1 (14 h). Three separate cell synchronies (i.e. harvesting at the same time points) were performed to assess the statistical soundness of our quantitative method. By applying our method for quantitation to the unacetylated region of the HILIC chromatogram, we have rigorously quantitated seven forms of histone H4 (Table 2). These forms include 2mLys-20 (i.e. no N-terminal acetylation), a␣Ser-1, a␣Ser-1 ϩ mLys-20, a␣Ser-1 ϩ 2mLys-20, a␣Ser-1 ϩ 3mLys-20, a␣Ser-1 ϩ mArg-3 ϩ 2mLys-20, and a␣Ser-1 ϩ 2mArg-3 ϩ 2mLys-20. Upon release from thymidine, newly synthesized histone H4 was identified by the 3.5ϫ increase of unacetylated, unmethylated H4 (a␣Ser-1) at 8 h (Fig. 5A, compare 0 Ac regions of G 1 /S and late S/G 2 chromatograms). As the cell cycles from 8 to 14 h, the a␣Ser-1 form decreased by 14ϫ until it existed as only 1% of the total histone H4 present in the nucleus (Fig. 5B).
Additionally, there was a 2.3ϫ increase in the amount of unacetylated, monomethylated H4 (a␣Ser-1 ϩ mLys-20) at 8 h, and it remained constant at 14 h. Unacetylated, dimethylated H4 (a␣Ser-1 ϩ 2mLys-20) decreased from 65 to 53% during the passage from the G 1 /S border to late S/G 2 . Between late S/G 2 and mid-G 1 there was a return of a␣Ser-1 ϩ 2mLys-20 to the level in asynchronous cells. These results indicate an increase in methylation of new, unmethylated H4, creating monomethylated Lys-20 and, subsequently, dimethylated Lys-20 during the cell cycle (29). Both a␣Ser-1 ϩ 3mLys-20 and a␣Ser-1 ϩ mArg-3 ϩ 2mLys-20 represented less than 1% each of the total histone H4 present in the nucleus. These species were present at their highest levels at the G 1 /S border and subsequently decreased during late S/G 2 ( Table 2). There was no significant difference in the level of Lys-20 trimethylation or Arg-3 monoand dimethylation from 8 to 14 h.
Statistical Significance-To test the robustness of HILIC/ FTMS as an assay, all cell cycle time points (0, 8, and 14 h) were repeated 3 times. Fractions were collected every minute to allow for the utmost precision in characterization and quantitation of histone H4 forms. After MS and MS/MS analysis, data were generated from each biological replicate for the five major H4 forms listed in Table 2 (species with ⌬m ϭ 42 to ⌬m ϭ 84 Da). Table 2 contains the mean values of the three replicate experiments. All analysis was conducted in SAS (SAS Institute, Cary, NC). To assess statistical significance, a square root transformation was performed (supplemental Fig. 3B). After this transformation, analysis of variance was conducted to test the differences between the square root of the percent abundance of the observed modification isoforms across the measured time points. The time by protein form interaction effect (i.e. the effect of the cell cycle on the presence of all five forms of H4) was found to be significant (p Ͻ 0.0001). Therefore, leastsquares-adjusted means were calculated for the main effects.
For the particular result of a␣Ser-1 progressing from 14 to 1% in the 8-to 14-h time points (Table 2), a statistically significant difference (F ϭ 151.37 with 1 degree of freedom, Bonferroni corrected p Ͻ ϭ 0.0001) was found. This finding validates the reproducibility of our method and can be understood by the entrance of the a␣Ser-1 form of histone H4 into the nucleus during S phase (thereby being detected at late S/G 2 ) and then decreasing due to the progressive methylation of Lys-20 as the cell cycles.
Analysis of Monoacetylated, Multiply Methylated H4 during the Cell Cycle-The coupling of HILIC with Top Down mass spectrometry enabled a dynamic range in excess of Ͼ10 4 , which allowed identification and quantitation of rare histone forms. Whereas unacetylated histone H4 represented the most abun-   Table 3). This drop in abundance does not correspond to a drop in complexity. Indeed, twice as many monoacetylated isomers were identified as unacetylated isomers. Changing patterns of methylation were also observed on monoacetylated H4, specifically at lysine 20 (0mLys-20, 1mLys-20, 2mLys-20, and 3mLys-20) and arginine 3 (mArg-3). Of the 24 combinations of acetylation-methylation isomers possible, 14 were identified in this study (Table 3). Multiple monoacetylation isomers were detected on molecules with un-, mono-, and dimethylated Lys-20 throughout the cell cycle. The acetylation site occupancies followed a general trend whereby the abundance of aLys-5 Ϸ aLys-8 Ͻ aLys-12 Ͻ Ͻ aLys-16 as Lys-20 became mono-, di-, and -trimethylated (Fig. 5). Indeed, the 3mLys-20 population had Lys-16 as the sole site of acetylation. Furthermore, aLys-16 was the only acetylation site occupied throughout each of the three cell cycle phases. Moreover, this trend remains constant during the cell cycle despite changing abundances, suggesting that H4 acetylation rearrangement occurs rapidly upon chromatin maturation, consistent with other reports (52,53). However, other factors may contribute to the distribution of acetylated isomers within each Lys-20 methylation class, such as the possibility that deacetylation at Lys-12, -8, or -5 is enhanced through increased deacetylase recruitment by increasing degrees of Lys-20 methylation, whereas Lys-16 deacetylation is simultaneously restricted due to steric factors.
Nested within the data of Table 3 are the dynamics of both histone H4 monoacetylation and methylation as the cell cycles. Looking first at acetylation, a number of observations can be made (Fig. 5). From 0 to 8 h, a large increase in monoacetylated, unmethylated molecules (3.3ϫ increase), and monoacetylated, monomethylated molecules (3ϫ increase) was observed. However, from 8 to 14 h, the monoacetylated, unmethylated molecules decreased dramatically (15ϫ decrease), and a small decrease (1.4ϫ) was observed for the monoacetylated, monomethylated molecules. The monoacetylated, dimethylated molecules also decreased by 1.2-fold from 8 to 14 h. These data indicate that methylation at Lys-20 on un-and monoacetylated molecules is progressive as the cell cycles. Interestingly, similar magnitudes of increasing Lys-20 methylation were observed on un-and monoacetylated H4 during the three cell cycle phases studied (compare Fig. 5, B and C, middle graphs). This suggests that acetylation does not have an effect on Lys-20 methylation in vivo, contrary to in vitro data showing mutual antagonism between the two modifications (51). To date there has been no histone methyltransferase identified to be responsible for dimethylating H4 Lys-20. Nevertheless, there is a clear change of Lys-20 during the cell cycle. We were surprised to observe that the frequency of Lys-20 methylation on unacetylated molecules was the same as on monoacetylated molecules, as others have reported that acetylation inhibits Lys-20 methylation (51). To further investigate whether hyperacetylation might antagonize Lys-20 methylation, we released synchronized cells into media containing 10 mM sodium butyrate. The RP-HPLC H4 fraction was then analyzed directly by TDMS (supplemental Fig. 1A). Newly synthesized H4 was observed as highly acetylated, unmethylated molecules (supplemental Fig. 1A, black isotopes). As the cells cycled, the abundance of the unmethylated mono-, di-, tri-, and tetraacetylated molecules decreased due to increasing Lys-20 dimethylation. Interestingly, at 10.5 h the corresponding intensity ratios of the un-, mono-, and dimethylated forms of highly acetylated molecules were very similar (supplemental Fig. 1). Finally at 15 h virtually all histone H4 was dimethylated. These results suggest that hyperacetylation of H4 has no differential effect on Lys-20 methylation rate in vivo.
This study presents new insights into the abundances of 42 unique H4 forms and their dynamic changes for three cell cycle time points. Future work examining the biological significance of the dynamic methylation at Lys-20 through the cell cycle and specific combinations of acetylation may benefit from the quantitative determinations of H4 PTMs described here. The delineations of the types and abundances of combinatorially modified forms of histone H4 will illuminate future research in chromatin biology.