| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Molecular & Cellular Proteomics 5:671-685, 2006.
© 2006 by The American Society for Biochemistry and Molecular Biology, Inc.
Subunit Processing Heterogeneity*,S
From the Department of Cell and Molecular Pharmacology and Experimental Therapeutics, Medical University of South Carolina, Charleston, South Carolina 29425
 |
ABSTRACT |
|---|
 TOP ABSTRACT EXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
subunit isoforms associated with bovine brain G proteins, a primary mediator of cellular communication. G subunits were isolated from purified brain G proteins and characterized by Edman sequencing, by MALDI MS, by chemical and/or enzymatic fragmentation assayed by MALDI MS, and by MS/MS fragmentation and sequencing. Multiple forms of six different G isoforms were detected. Significant variation in processing was found at both the amino termini and particularly the carboxyl termini of the proteins. All G isoforms contain a carboxyl-terminal CAAX motif for prenylation, carboxyl-terminal proteolysis, and carboxymethylation. Characterization of these proteins indicates significant variability in the normal processing of all of these steps in the prenylation reaction, including a new variation of prenyl processing resulting from cysteinylation of the carboxyl terminus. These results have multiple implications for intracellular signaling mechanisms by G proteins, for the role of prenyl processing variation in cell signaling, and for the site of action and consequences of drugs that target the prenylation modification.
G proteins are heterotrimeric GTP-binding proteins involved in signal transduction from receptors to intracellular effectors (6–10). They consist of a guanine nucleotide-binding 
subunit that reversibly associates with a ß dimer (11). Together these two components regulate numerous downstream effectors such as adenylyl cyclase, ion channels, and phospholipase C (6–8, 11–13). G proteins are a component of a ubiquitous cellular signaling system. More than 600 of the 25–30,000 genes found in the human genome code for G protein-coupled receptors (14, 15). These receptors are estimated to be the targets of 50% of clinically useful drugs (16) and an equally impressive array of endogenous regulatory compounds (17). The functional implication of these statistics is that the G proteins in cells process an immense amount of incoming information. Not surprisingly, the G proteins themselves are also structurally diverse. Over 20 subunits (the products of 16 genes), five ß subunits, and 12 subunits have been identified (9, 18). All of the known subunits of these proteins are substrates for prenylation, which appears to be essential for the normal function of the heterotrimeric G proteins.
About 2% of cellular proteins end in a CAAX motif that is a signal for prenylation (19–21). These include a broad range of proteins, many of which are involved in cell signaling and growth regulation, for example the Ras proto-oncogene products and the subunits of the heterotrimeric G proteins (20, 22–24). Prenylation is a complex modification that involves attachment of a geranylgeranyl or a farnesyl group to a Cys residue 4 amino acids from the carboxyl terminus of target proteins (19, 23). Which prenyl group is attached is determined by the carboxyl-terminal amino acid. Geranylgeranyl is added if this residue is a Leu or Phe, whereas Farnesyl is added if it is a Ser, Met, Gln, Cys, or Ala (19, 25). Following prenylation, the three carboxyl-terminal amino acids are removed by a specific protease, and the protein is carboxymethylated. The fidelity and variability of these multiple enzymatic steps is not clear, particularly for native proteins, and may play a role in both the normal biochemistry of these proteins and the pathological consequences of their aberrant processing. These modifications and the enzymes involved are potential targets for drug development. For example, farnesyltransferase and geranylgeranyltransferase I inhibitors have potential as antitumor agents and are being used in clinical trials (26–28), but the exact mechanism of action of these inhibitors remains unclear. This is in part due to uncertainty about their real targets (29, 30) but also due to lack of information about the processes with which they interfere. In addition, the prenylation reaction turns out to be a heretofore unsuspected target of other antitumor agents, such as the statins that interfere with generation of the prenyl moiety (30) and methotrexate that interferes with the carboxymethylation reaction (31). What is lacking, in particular, is a clear understanding of the normal biology of this series of modifications and how variation in each step might affect function of the target proteins.
Brain cortex is a rich source of a diverse population of G proteins, which are regulated by many different signaling molecules. Past biochemical and molecular biology studies provide evidence for at least six G protein subunits in brain (32–41). When purified brain G proteins were analyzed by MALDI-TOF mass spectrometry, seven prominent signals in the G subunit molecular weight range were found (32). Four of these molecular weights could be assigned to G subunits known to be in bovine brain, G2 (41), G3 (38), G5 (33), and G7 (40), provided specific modifications of those G subunits were proposed. Previously we described the characterization in these samples G2 (42), G10 (43), G5 (44), a novel G5 that is not proteolytically processed after prenylation (44), and a G2 subunit truncated and arginylated at the amino terminus as a signal for ubiquitylation (45). Throughout the study of all of these proteins, however, it was clear that the number of potential G protein subunits associated with brain Go/Gi preparations, due to variable processing, is much larger than originally appreciated.
Here we characterize the variability and processing of G protein subunits associated with G proteins purified from bovine brain. Significant variation in the carboxyl-terminal processing was found, including probable cysteinylation as a newly identified carboxyl-terminal modification of the prenylation site. In light of past studies characterizing the functional role for a number of these modifications, these results suggest that carboxyl-terminal variability could affect G protein signaling in a number of ways, including receptor recognition and fidelity, cellular localization, and even turnover of G protein functional components. These results also indicate the potential site of action of drugs that interfere with selective reactions in the prenylation process.
|
EXPERIMENTAL PROCEDURES |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
S1 binding assays, protein quantitation, and Coomassie Blue staining polyacrylamide gels indicated that G and Gß (and presumably G) subunits were purified in roughly stoichiometric amounts. Protein was stored at –80 °C until use in these studies in a buffer containing 25 mM Tris-HCl, pH 8.0, 1 mM EDTA, 1 mM DTT, 100 mM NaCl, and 0.1% Thesit at a protein concentration of 2–4 mg/ml.
HPLC Separation of G Isoforms—
The G subunit isoforms were separated from G and Gß subunits by reverse-phase HPLC as described previously (48). The region of the HPLC elution that contained the G subunits was identified by immunoblotting with specific antisera to G2, G3, G5, and G7 (48). In many preparations proteins were eluted from the column in line with a Finnigan LCQ mass spectrometer, and fractions were identified by the presence of previously characterized intact G subunit masses. The region of the HPLC separation where G eluted was well separated from that region where G and Gß eluted from the column. Fractions were stored at –20 °C until further analysis. For separation of Asp-N-digested subunits, aliquots from the general HPLC separation (48) were dried under vacuum and resuspended in 5 µl of Solvent A (0.102% trifluoroacetic acid in water). Peptides were separated at 5 µl/min on a C18, 300-Å, 150 x 0.5-mm-inner diameter column from Brownlee equilibrated with Solvent A for 30 min before injection and eluted with a gradient of 5–75% Solvent B (0.083% trifluoroacetic acid in 100% acetonitrile) over 140 min with a final hold at 75% B for 40 min. Eluted peptides were monitored at 210 nm. Peptides were either eluted onto PVDF membrane for Edman sequencing or hand-collected and stored at –20 °C for MS analysis.
MALDI Mass Spectrometry—
MALDI-TOF analysis was performed on a PerSeptive Biosystems Voyager-DE mass spectrometer as described previously (48). Generally the matrix used was -cyano-4-hydroxycinnamic acid, although sometimes results were verified using the alternative matrix sinapinic acid. Approximately 75–150 scans were averaged for a single representative MALDI spectrum. Spectra were analyzed under conditions that favored evaluation of the singly charged mass of each protein. Masses reported from MALDI MS were determined from spectra containing internal calibrants of known mass bracketing the signals being evaluated. Calibrants normally used and their associated [M + H]+ were horse cytochrome c (12,361), bovine insulin (5,735), and a synthetic G2 peptide (1,701) with the sequence PASENPFREKKFFC. Average mass estimates come from multiple G protein preparations independently separated by HPLC and analyzed at different times with independent calibration of the mass spectrometer.
Aspartate-Proline Bond Cleavage and MALDI Analysis—
HPLC fraction aliquots (50–100 µl) containing G subunits were dried under vacuum, resuspended in 1:1 acetonitrile:water and 1–3% TFA. Samples were incubated at room temperature for at least 12 h and analyzed by MALDI mass spectrometry as described previously (48).
Protein Sequencing—
Edman sequencing of 50-µl aliquots of HPLC fractions was performed in the Medical University of South Carolina Proteogenomics Facility on a Procise 494 Applied Biosystems instrument as described previously (45).
Electrospray Ion Trap Mass Spectrometry—
ESI-MS/MS was performed either in line with HPLC on a Finnigan LCQ mass spectrometer or by direct infusion into a custom built nanospray source attached to the LCQ ion trap. Nanospray (flow, 
30 nl/min) was performed as described previously using a window of 1 or 2 m/z units for precursor ion selection (49).
Statistical Analysis—
Masses detected in G protein preparations were quantified in multiple independent MALDI MS experiments using internal standards bracketing the measured mass. Measurements are reported as mean ± S.E. Typically S.E. was 1–2 Da for a 7–9,000-Da protein. A t test was used to evaluate consistency of observed masses with predicted masses compiled in a table of all possible combinations of plausible modifications of all known G protein subunits. This table contained about 2,500 total entries. Data are reported as the probability (p) that the theoretical mass is not different from the observed mass using the t distribution with n – 1 degrees of freedom and the observed S.E. Given the precision of mass spectrometry, this represents a fairly stringent test of an assignment given the large number of possible assignments. In effect, this test was used to support the assignment of a proposed structure to a mass only if the predicted and measured masses were within about 2–3 Da of 7–9,000 Da. Typically our assignments are consistent with only one possible structure of about 2,500 possible structures considered.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
Subunits of Brain G Proteins subunits from bovine brain were separated from other components of the G protein heterotrimers by reverse-phase HPLC. The G protein preparations used in these studies were highly purified and contained protein with a specific activity of 12.4 ± 0.7 nmol of GTPS binding activity/mg of protein (mean ± S.E. for 17 independent G protein preparations). Compared with a theoretical activity of 13.0 nmol/mg for a predicted 84,000-Da heterotrimer, this suggests greater than 95% purity of the proteins analyzed; this could be underestimated because of protein inactivity or to persistent binding of endogenous guanine nucleotides (50). These preparations are expected to contain stoichiometric amounts of G subunits. HPLC fractions containing G subunits were analyzed by SDS-PAGE and identified by silver staining and by immunoblotting with G subunit-specific antisera (48). Fractions containing G subunits were well separated in the elution profile from G and Gß proteins. Fig. 1 shows MALDI analysis of the G subunit mass range (6,500–9,000 Da) of fractions from a typical HPLC separation of bovine brain G subunits. Fig. 1 shows spectra collected under constant ionization conditions where laser intensity was modulated to generate an ion current less than saturated for the completely processed G2 (Fig. 1, Mass 25), which is consistently the G isoform with the strongest signal in these G protein preparations. Although there is not a simple relationship between MALDI signal intensity and the amount of sample because of the complex variables affecting ion generation, the abundance, distribution, and intensity of masses seen in Fig. 1 are typical of more than a dozen HPLC separations of bovine brain G proteins. Additional spectra were collected from the analysis of each fraction to provide high resolution mass estimates (data not shown but see supplemental data) based upon internal standards bracketing the signals being evaluated. Data on masses consistently seen in multiple G Protein preparations and numbered in Fig. 1 are summarized in Table I.
|
|
subunit isoforms. Significantly very few masses were smaller than m/z 6,500, arguing against the idea that many are proteolytic products of G subunits. If most of these masses represent nonspecific proteolytic products of G or Gß or of any protein for that matter, a fairly random distribution of masses might be expected, but most of the masses present are coincident with the mass range predicted for G subunits. Finally Edman sequencing readily identified G3 (Fig. 1, Mass 21), which has an unblocked amino-terminal Met residue, in appropriate HPLC fractions but, in general, failed to find appreciable amounts of protein that could be sequenced in most of our samples (see below for some significant exceptions). In a separation such as that shown in Fig. 1, samples of all fractions were sequenced, and multiple fractions yielded sequences of 15 cycles (maximum cycles processed) of G3 with an estimated total of 1.1 nmol recovered from the separation. In comparison, our recently characterized N-arginylated G2 protein (45) was detected in a single fraction at a level of about 70 pmol (Fig. 1, Mass 15). Altogether 1.4 nmol of sequenceable material was recovered from the separation; this, if all represents G subunits of about 7.5 kDa, would account for about 10% of the initial G protein preparation (if no correction is made for recovery). Recognizing that more than half of the G3 appears to be N-acetylated on Met1 (Fig. 1, Mass 26) and allowing for recovery of protein, this would be a reasonable estimate of the abundance of G3 in brain G proteins. G2 and G3 would be expected to be nearly equivalent in abundance in brain G protein preparations, but nearly all G2 variants appear to be more readily ionizable than their corresponding G3 variants. Sequencing of the unblocked G3, which would be thought to be in high abundance, as well as the RDTASIA-G2, which would be thought to be in much lower abundance if rapidly turned over, demonstrates that appropriate proteins in the HPLC fractions are readily sequenceable by Edman sequencing and that proteins in general are not nonspecifically blocked. The failure to find significant amounts of other sequenceable protein also argues against proteolytic generation of the masses present because at least one-half of all generated peptides should be unblocked and amenable to sequencing if significant random proteolysis occurs.
Identification and Characterization of G Subunit Isoforms and Their Modifications
Newly described proteins and their structural assignments are summarized in Table II. All of these, as well as modified proteins described previously (42–45), correspond to modified forms of G2, G3, G5, G7, G10, or G12 (Table II and Fig. 2), six isoforms previously reported in brain. In Table II we characterize the relationship between the observed and expected masses for an assignment based on the probability that the predicted mass is within the error distribution of the observed mass calculated from the t distribution given the observed S.E. and its associated degrees of freedom. Generally the accuracy of the internally calibrated Voyager-DE MALDI-TOF instrument is 1 in 2,000 or for a potential G subunit of mass 8,000 ± 4 Da. Because these values were obtained entirely from experiments using internal standards bracketing the characterized masses we expect the measured masses to be within the instrument tolerance. Furthermore estimates of precision can be obtained from errors determined by repeated independent measurements of our masses, which are generally in the range of 1–2 Da (Table I). These measurements have 95% confidence intervals of about 6 Da (i.e. ±3 Da on either side of the observed mass).
|
|
subunit isoforms in HPLC fractions or modifications of those G subunits (Table II and Fig. 2). The proteins summarized in Table II have multiple lines of support for their structural assignments. One way that we generated an initial hypothesis for many of these structures was to compare the observed mass with a library of about 2,500 masses generated from all known G sequences modified by all variations of known modifications (such as presence or absence of N-acetylation or geranylgeranylation versus farnesylation) or plausible modification such as phosphorylation. Generally there is only one plausible structure within the 95% confidence intervals for each observed mass. If an observed mass corresponded to one of the possible masses, we used the techniques below to seek supporting evidence for this assignment. In some cases, observed masses were not compatible with the predicted mass of any known G isoform modified by any known combination of posttranslational modifications. In such cases, we used the strategies below to elucidate the structure but still imposed the requirement that the final proposed structure had to have a mass within the 95% confidence intervals of observations in Table II.
One of the first strategies used to identify masses in Table I was Edman sequencing. As noted above, however, this was often not a useful approach because most masses contained blocked amino termini. In addition, in general this approach cannot be used for identifying protein modifications. Nevertheless prominent signals were obtained for G3 (Fig. 1 and Table I, Mass 21), for an N-arginylated G2 variant previously characterized (45) (Fig. 1 and Table I, Mass 15), and for a G2 fragment related to the arginylated protein but starting with Thr6 (Mass 20). We also found a sequence co-eluting with the major G3 sequence that could be G4, another G subunit predicted to contain an unprotected amino terminus. This signal was very weak, however, and we are unable to identify a specific mass corresponding to a modified bovine G4 in our preparation, although this could be a minor signal not seen consistently.
A second approach to confirm structures for masses of predicted G subunits and to obtain information for identifying unknown masses, took advantage of the characteristic single aspartate-proline bond found in all known G isoforms except G10 (43). This Asp-Pro bond is susceptible to hydrolysis at low pH, generating amino- and carboxyl-terminal products (4,500–6,000 and 2,000–3,000 Da, respectively) that can be compared with known G subunit sequences with different possible modifications (48). This technique was used previously to characterize G2 (42). Because all of the modifications described here were found at the amino or carboxyl terminus of the proteins, this technique was particularly useful for confirming the site of variable modification, in the case of proteins that corresponded to predicted masses, or for providing clues to the G isoform corresponding to a mass for proteins with unknown or novel modifications. Fig. 3 shows a MALDI spectrum of an HPLC fraction containing a complex set of G subunit species and the analysis of the acid digest of this fraction to identify amino-terminal and carboxyl-terminal fragments of G subunits with newly described modifications, a farnesylated species of G2, and a cysteinylated species of G2.
|
subunits (49) or MS/MS and MS/MS/MS sequencing of proteolytic fragments generated with trypsin or Asp-N and separated by in-line HPLC (48). In many cases, this approach was used to confirm hypothesized structures of modified G isoforms based upon application of the earlier described approaches and generated from consistency of the observed MALDI mass with reasonable modifications of known G isoforms. In particular, this approach was useful to show the presence of an isoprenyl group on many of the subunits listed in Table II. This thioether bond is readily fragmented by CID, and the existence of an isoprenyl group can be inferred from the loss of the mass (272 Da for geranylgeranyl and 204 Da for farnesyl) from the intact protein (49). An example of this is shown in Fig. 4 where a farnesylated G2 subunit was characterized. One of the advantages of the ion trap mass spectrometer is that ions can be sequentially selected for multiple rounds of fragmentation. In general the molecular ions of the intact G subunits appear to have preferred charge states in the MS of 5+ to 8+. Although the LCQ has insufficient resolution for unambiguous assignment of the charge state from isotopic distribution of fragment ions, assignments were facilitated by identification of signals compatible with multiple charge states of fragment ions and by the ability to associate nearly all major signals with predicted b and y ions of structures proposed by multiple independent lines of evidence as summarized above.
|
isoforms with their predicted modifications (Tables I and II and Fig. 2). For example, signals indicating the presence of G2 (Mass 25, m/z 7,751.3), G5 (Mass 11, m/z 7,169.8), and G7 (Mass 8, m/z 7,409.6) were routinely found and agree with previous predictions (32). In addition, a signal consistent with G12 (Mass 4, m/z 7,925.0) was found in a fraction that was immunoreactive with the antiserum to G7 (data not shown). (The antiserum was generated against an amino-terminal G7 peptide, and cross-reactivity is due to their similar amino-terminal sequences.) There was also a G10 isoform (Mass 28, m/z 7,134.4) found (43). All of these masses were compatible with proteins that are fully processed, i.e. acetylated at the amino terminus after Met1 cleavage and prenylated at the carboxyl terminus followed by proteolytic removal of the last three residues and carboxymethylation of the new carboxyl terminus. In addition, we previously found a broad MALDI signal that could not differentiate between two differently modified forms of G3 (32). In the analysis here, both of these forms were elucidated, one with an unblocked amino terminus (Mass 21, m/z 8,295.6) and the other N-acetylated on Met1 (Mass 26, m/z 8,337.6). Aside from these seven proteins, however, many of the masses could be accounted for by versions of these proteins with alternative processing patterns. All but two of these alternatively processed proteins (Masses 29 and 30 appear to be exceptions but were not satisfactorily resolved in these studies, but see below) were modified at either the amino terminus or the carboxyl terminus, which appear to be regions of hypervariability.
Amino-terminal Variants of G Isoforms and Related Modifications
Although most of the variation in G protein modifications described here occurs in association with the carboxyl-terminal prenylation site, there was also variation in amino-terminal processing detected, particularly for the most abundant isoform, G2. Most brain G protein isoforms are predicted to be acetylated after removal of Met1. This was found to be the case for G2, G5, G7, G10, and G12; for G3, more than one-half was acetylated on Met1 according to MALDI signal intensity, and the rest was unmodified (Fig. 1 and Table II). In addition to G3, Edman sequencing identified a low level of a peptide homologous to the unblocked amino terminus of G4 (MKEGMSNNST). The G4 sequence, like that of G3, could remain unacetylated at the amino terminus in the cell because its amino-terminal sequences are similar. However, neither MALDI nor ESI analysis identified the expected G4 mass in that fraction.
Recently we described (45) an N-arginylated variant of the G2 protein that seems to be generated by a complex processing pathway leading to a G2 species with an N-end rule ubiquitylation signal (51). The N-arginylated G2 protein (45) is missing its first five residues (MASNN), which have been replaced by the sequence "RD." Portions of an enzymatic pathway that could generate this protein have been described previously (51). These include an amino-terminal deamidase that could use as a substrate the Asn at position 5 in G2 protein (after proteolysis of the MASN sequence) and a protein arginyltransferase that could arginylate the resulting amino-terminal Asp. The RD-G2 protein that we previously described is the first protein identified that would be a product of the complete pathway, and so this pathway has not yet been established definitively. Proteins characterized here that are likely to be part of this pathway include a truncated form of G2 identified by Edman sequencing with the following amino-terminal sequence: TASIAQARKLVEQLK. This protein is identical in sequence to processed G2 starting with Thr6 (Tables I and II and Fig. 1, Mass 20, m/z 7,324.2). This protein is very similar to another previously identified G2 from bovine spleen that is truncated on the fifth amino acid (Asn) (52). It is important to state that this is the only other truncated G subunit found in our bovine brain analyses; its relationship to the RD-G2 is conspicuous and suggests that this is not a random event. It is not clear whether this is a precursor or a product of the RD-G2. Also related may be the protein represented in Fig. 1 and Table I as Mass 30 (m/z 7,752.2), which has a 1-Da increase in mass from the correctly processed G2 protein. This could be deamidated G2, an early point in the RD-G2 pathway, but attempts to identify specific sites of deamidation in this protein have not been successful. A similar situation exists for a "deamidated G3" (Table I and Fig. 1, Mass 29, m/z 8,339.8). Thus, we believe that there is physiologically significant modification of the amino terminus of at least G2, but the processes and mechanisms involved are not clear from the present studies.
Carboxyl-terminal Variants of G Isoforms
Unmethylated G2—
Carboxymethylation is the final reaction in the progressive steps involved in prenylation of the CAAX motif proteins. Although expectations were that blocking the carboxymethylase would have relatively modest effects on the function of Ras (53), this turned out not to be the case. Deletion of the carboxymethylase gene is early embryonic lethal (53) and appears to block the transforming activity of activated K-Ras and N-Ras (54). Surprisingly inhibition of carboxymethylation appears to be a secondary effect of antitumor agents such as methotrexate and may be a primary mode of action in its suppression of tumor cell growth (31). For the G proteins, carboxymethylation is a reversible reaction (55), possibly regulated by receptors (56), that may influence turnover (57) and membrane association of the Gß dimer (58, 59). Thus, it is of considerable interest whether or not our studies provide evidence of variation in carboxymethylation of native G protein subunits. In one case, a signal at m/z 7,737.0 (Tables I and II, Mass 16) had an Asp-Pro fragment ([M + H]+, 5,159.8 ± 3.8) compatible with the amino terminus predicted for G2 ([M + H]+, 5,161.9), but the corresponding carboxyl-terminal fragment ([M + H]+, 2,590.2 ± 2.0) was compatible with a G2 lacking a carboxymethyl group at the carboxyl terminus ([M + H]+, 2,594.2) and not the predicted G2 isoform ([M + H]+, 2,608.2). ESI-MS/MS experiments supported this assignment (supplemental data). The generality of these putative structures is suggested by identification of intact masses compatible with unmethylated forms of G3 (Mass 18) and G12 (Mass 2) (Table I and Fig. 1), although we do not now have additional supporting data on these for assignment of a structure. The presence of unmethylated G2 was not a nonspecific phenomenon either because the apparent amounts of unmethylated and methylated forms of the proteins varied substantially from being a large and prominent signal for G2 (Mass 16, unmethylated versus Mass 25, methylated) to not being observable as a significant signal for nearly equally prominent G7 (Mass 8).
Farnesylated G2—
The carboxyl-terminal residue of CAAX motif (X) is the primary determinant of the type of lipid group added to the cysteine (19, 25). A 20-carbon geranylgeranyl group is added if the final residue is a Leu (or Phe), and a 15-carbon farnesyl group is added if the final residue is a Ser, Met, Gln, Cys, or Ala (21). The G2 isoform contains CAIL as its carboxyl-terminal signal for geranylgeranylation, which is the major 2 isoform observed (Fig. 1, Mass 25). Surprisingly, however, a MALDI signal at m/z 7,684.5 (Fig. 1 and Table II, Mass 12) was compatible with a farnesylated G2 isoform. Acid hydrolysis to cleave the predicted Asp-Pro bond in G2 generated confirming fragments at m/z 5,160.7 and 2,539.1 (compared with predicted m/z 5,161.9 and 2,540.1) (Fig. 3). MS/MS sequencing on an LCQ ion trap mass spectrometer confirmed this structural assignment, generating peaks compatible with b and y ions for the sequence of G2 as well as evidence supporting the presence of a farnesyl prenylation modification (Fig. 4). The 7+ charge state of farnesylated G2, m/z 1,098.5, was selected for MS/MS fragmentation. The MS/MS spectrum in Fig. 4B shows signals of m/z 1,247.2 and 1,069.3, compatible with the 6+ and 7+ charge states of G2 subunit after loss of 204 mass units, the mass of the farnesyl group. In contrast, the fragmentation spectrum did not generate identifiable ions consistent with the loss of geranylgeranyl group from this protein; loss of geranylgeranyl group was readily identifiable in most proteins predicted to contain this prenyl group (supplemental data).
As far as we know, the variability in the prenylation pattern of native proteins has not previously been systematically investigated. Previous studies have shown that recombinant G protein subunits (60) and expressed RhoB (61–63) can be prenylated with either geranylgeranyl, their predicted prenyl group, or farnesyl, albeit at substantially different levels under comparable conditions. Further Ras, predicted to be farnesylated, can also be geranylgeranylated in cells treated with a farnesyl protein transferase inhibitor (64–67). This could be due in part to acceptance of alternative substrates by geranylgeranyltransferase I and farnesyltransferase (68), but the sequence determinants of this appear to be complex and may differ between proteins (69). Given the controversy over the site of action of farnesyl protein transferase inhibitors and the increasingly apparent complexity of the enzymology associated with protein prenylation (as described here), it is important to know which pathways and processes are biologically relevant to normal cells and tissues, which are pharmacologically induced, which are due to pathologic processes, and which are possibly related only to the processing of overexpressed recombinant proteins. An early protein identified as variably prenylated as a recombinant protein was RhoB (61), which is predicted to be geranylgeranylated but was also found to be farnesylated. Although different mechanisms of farnesylation of this protein have been proposed (62, 63), immunologic evidence suggests that the farnesylated form is expressed both as a recombinant and as an endogenous protein in cells (70). Those studies also suggested, however, that this dual prenylation is sequence-specific and coded not just by the carboxyl-terminal amino acid but all three carboxyl-terminal amino acids of the protein (CKVL), which are somewhat unique for RhoB among prenylated proteins (70). Here it is clear that variant prenyl group incorporation also occurs in biological samples in a subunit, G2, with a classical CAAX motif coded for by CAIL. Although farnesylated G2 is a minor component of the total G2 in brain, it is likely of functional significance. The prenyl group of the G subunit can dictate receptor coupling and specificity (71, 72), and specific receptors may preferentially bind heterotrimers containing G subunits with a particular type of prenyl group (73, 74). For example, the A1 adenosine receptor prefers G subunits with a geranylgeranyl group (73). Thus, the farnesylated G2 may have different receptor specificity than the major form, which is geranylgeranylated.
Unprenylated Carboxyl Terminus—
Protein prenylation through a thioether linkage is a relatively stable bond. Studies of the complexity of metabolizing prenylcysteine degradation products of these proteins resulted in the description of a lysosomal prenylcysteine lyase that is an FAD-dependent thioether oxidase (75, 76). This enzyme, however, will not process prenylated peptides (75), only the free prenylcysteine, and as far as we know, no enzyme activity has been described that specifically removes a prenyl group from a functional prenylated protein. Nevertheless we characterized in purified bovine brain G2 (Fig. 1 and Table II, Mass 22) and G7 (Fig. 1 and Table II, Mass 6) proteins that were otherwise processed normally but lacked a prenyl group. When fragments of expressed G subunits are assayed by ESI-MS, unprenylated proteins can be generated by CID (60), similar to our recently described characterization of intact G subunits (49). This does not explain the results reported here because the prenylated and unprenylated proteins travel with different average retention times (Table I) on HPLC. It is possible to generate an unprenylated protein by acid hydrolysis as during the HPLC separation in the presence of TFA. However, masses corresponding to unprenylated variants were not generated in samples containing intact G subunits and stored in HPLC buffer before analysis, and acid hydrolysis at higher TFA concentrations to produce Asp-Pro fragments (Fig. 3) did not generate unprenylated proteins even though these conditions readily cleaved the acid-sensitive Asp-Pro bond. Finally although prenylcysteine lyase has not been shown to metabolize prenyl peptides (75), perhaps it could do so at a low level if released from lysosomes during tissue preparation. Presumably such release would be nonspecific, however, and other enzymes, such as carboxymethylases, would nonspecifically cleave all G targets, something that we did not observe (see above). Thus, although we cannot rule out absolutely generation of unprenylated proteins during isolation, there is also the real possibility that this is a physiological process. Given what is generally thought to be the role of the prenyl group in membrane localization, such a reaction would produce cytosolic Gß dimers and would provide a potential intracellular signaling pattern in G protein responses.
Unproteolytically Processed Carboxyl Terminus of G5 and G7—
Generally it is believed that prenylation of CAAX motif proteins involves a sequence of reactions including prenylation of Cys at the –4 position, proteolysis of the carboxyl-terminal three amino acids, and carboxymethylation of the new carboxyl terminus. The proteolysis step is mediated by one of two enzymes in yeast (77) and by one known enzyme in mammalian cells, referred to as Ras-converting enzyme 1 (Rce1). Absence (knock-out) of this enzyme is embryonic lethal in mice and disrupts Ras localization (78) and leads to abnormal cardiac function and eventual death if disrupted cardioselectively (79). Rce1 has been targeted for development of antitumor drugs (29). Previously (44) we showed that the major G5 protein expressed in bovine brain is an exception to this rule; it retains its three carboxyl-terminal amino acids after prenylation without carboxymethylation (Fig. 1 and Table II, Mass 14), and a minor fraction of the protein is processed as is predicted by current understanding of this reaction (Mass 11). One thing that we looked for in our data was the generality of this observation. The only other G subunit for which we could find evidence of lack of Rce1 processing was G7, which was found to be prenylated and to retain its terminal -IIL sequence in a minor fraction of the protein observed (Table II and Fig. 1, Mass 10, m/z 7,736.2). Significantly we did not find a mass compatible with this processing pattern for G2, the most abundant G subunit in brain; this would tend to rule out ineffective proteolysis as a general explanation for the lack of processing of G5 and G7. In addition, the and ß subunits of farnesylated rabbit skeletal muscle glycogen phosphorylase kinase have been shown to have a similar processing pattern (80). There is no obvious carboxyl-terminal motif common to G5 and G7 that might direct this alternative processing pattern, and it is not clear that previous surveys of the specificity of, for example, the yeast Rce-related enzymes would adequately predict these observations in mammals (77). Nevertheless given the results with Rce1 knock-out mice and the mislocalization of Ras in cells from these animals (78), it is a reasonable hypothesis that these alternatively processed forms of G subunits would provide for alternative targeting sites for their associated G proteins.
Carboxyl-terminal Cysteinylation of G Protein Subunits—
The array of G subunits described above can be explained as variations of previously described prenylation-related modifications. In addition to these, however, a G2 species was found that is 89 mass units higher than completely processed G2 (Tables I and II and Fig. 1, Mass 13, m/z 7,840.1). That this is related to G2 is based upon analysis of Asp-Pro cleavage patterns (Fig. 3) and ESI-MS analysis (Fig. 5 and supplemental data). The Asp-Pro analysis indicated that the 89-Da addition was associated with the carboxyl-terminal fragment of the protein. We isolated the peptide fragment analogous to this by HPLC after Asp-N digestion of the protein. Edman sequencing of this peptide showed that its amino-terminal sequence was that predicted for G2 (data not shown). MS/MS followed by MS/MS/MS fragmentation of this peptide disclosed a fragmentation pattern characteristic of a Cys residue at the carboxyl terminus of the protein (Fig. 5). The data were compatible with the Cys linked either through a peptide bond or as a thioester. Attempts to alkylate the protein with several different reagents (N-ethylmaleimide and vinyl pyridine) did not modify any residues other than the free Cys in the protein predicted by the sequence of G2. These data would support the thioester linkage structure, but this conclusion is not yet fully substantiated. Intact masses were found that were compatible with similarly processed species of other G subunits, including G3 (Table I and Fig. 1, Mass 19, predicted mass 8,426.1 compared with 8,425.4 observed) and G12 (Mass 3, predicted mass 8,017.3 compared with 8,021.3 observed), although we have no data at this time to confirm these assignments. These data might suggest that carboxyl-terminal cysteinylation is a common modification of prenylated G proteins, possibly an alternative to carboxymethylation. To the best of our knowledge this is a novel protein modification with unknown enzymology and determinants. An interesting contrast to this is cysteinylation via disulfide bond formation, which has been a long known modification of serum albumin (81). Albumin contributes 80% of the free thiols in plasma and provides the major buffer for absorbing oxidized Cys through cysteinylation of Cys34, the only free Cys on albumin (82), a reaction of clinical significance (83, 84). In another case, cysteinylation of protein kinase C isoforms through disulfide bond formation may regulate different protein kinase C isoforms in response to oxidative stress (85, 86). In these cases the mechanism is likely very different from the carboxyl-terminal addition of Cys described here. In the case of disulfide bond formation this could be a chemically mediated response to oxidative stress, whereas the modification described here, whether through a thioester bond or peptide bond, would have to be enzymatic, as would removal of the Cys in a normal cellular environment, if the modification is readily reversible.
|
Subunit Isoforms subunit in bovine brain is G2. Seven of the masses described here or previously represent differently processed forms of G2 that vary at either the amino or carboxyl termini (Tables I and II and Fig. 1, Mass 12, farnesylated m/z 7,684.5; Mass 13, cysteinylated m/z 7,840.1; Mass 16, unmethylated m/z 7,737.0; Mass 20, truncated TASIA-G2 m/z 7,324.2; Mass 15, RDTASIA-G2 m/z 7,593.7; Mass 6, unprenylated m/z 7,481.1; Mass 25, fully processed G2 m/z 7,751.3). Most of the processed forms of this protein represent variations at the site of prenylation. It is possible that the variable processing of G2 is indicative of what would also be found for the other G protein subunits if expressed at a comparable level, but some forms seem to be more prominent for some G isoforms than for others.
Although some alternatively processed forms, such as the unproteolyzed G5 and unmethylated G2, are major forms of their respective proteins, others of these signals suggest that the modified forms are minor components of the total subunit present. Although this might minimize the biological role of these variants, this would also be expected if these represent transitional variants responding to incoming signals of one kind or another. For example, this might apply to the potentially unmethylated isoforms found at low levels (i.e. G3, Mass 18, and G12, Mass 2), which are in contrast to the unmethylated 2 (Mass 16) that appears to represent a sizable fraction of the protein. This would be compatible with previous suggestions that carboxymethylation of these proteins is regulated (55, 56), and our present data would suggest that G2 is preferentially affected by this at least under the specific conditions with which these proteins were isolated. Similar considerations would apply to proteins that have lost their prenyl group or are amino-terminally modified in transit to the ubiquitinylation pathway. Although found at low levels, these proteins may be indicative of important regulatory processes altering the distribution and level of G proteins in cells.
In contrast to low levels of modified forms that may represent transient states of the protein, low levels of farnesylated G subunits that are usually geranylgeranylated may result from the nonspecificity of the transferases involved (68), although even this seems to be sequence-specific even if complex (69) and suggests specificity to this variation. Regardless of the mechanism of generation, however, a population of such G subunits (farnesylated when predicted to be geranylgeranylated) do exist in cells and should have altered signaling responses to receptors (60). Similarly the unprenylated variants, if generated in cells, would be predicted to lead to altered membrane association of the Gß dimer and almost assuredly altered signaling properties. Thus, the processing variants described here, when viewed in light of past studies of the role of prenylation in cell, could have a dynamic interplay existing among them that will contribute to the responses in cells following regulation of these proteins. It is important also to recognize that level of expression is a relative value. Whereas many of the modified forms of G2, in particular, appear to be low relative to the predicted form of this protein, G proteins in brain are expressed at truly remarkable levels, approaching 1% of particulate protein. In most cells G proteins would constitute on the order of 0.01% of even membrane protein. Thus, although many of the modified forms described here are low compared with the high levels normally found in brain, they are comparable to the protein levels found for G proteins in other cells, and so they may be predominant forms of the protein in restricted compartments even within brain.
Evaluation of the proteome of different cells and disease states is a challenging endeavor. To begin with there is the enormous variation in protein sequence and chemistry magnified further by posttranslational processing of the proteins that can be, as demonstrated here, quite variable for what has been believed to be a well defined processing reaction. Ultimately characterization of the complete complement of proteins in cells will have to take into account all of these factors. A major step in the process will be defining the fidelity and variability associated with posttranslational processing events as is reported here for the prenylation reaction associated with the subunit of the heterotrimeric G proteins. A primary conclusion of our work is that signaling proteins, such as G proteins, that are posttranslationally modified at potentially important functional sites may vary greatly at these sites, and their role in normal or pathological function will not be apparent without taking these posttranslational variations into account. These results indicate too that the characterization of the proteome expressed in cells will have to ultimately incorporate a strategy to account for functionally significant variation in protein processing patterns.
|
FOOTNOTES |
|---|
Published, MCP Papers in Press, December 6, 2005, DOI 10.1074/mcp.M500223-MCP200
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 The abbreviations used are: GTPS, guanosine 5'-3-O-(thio)triphosphate; Rce, Ras-converting enzyme. 
* This work was supported in part by National Institutes of Health Grant DK37219, by National Science Foundation Grant EPS9630167, and by institutional support of the Mass Spectrometry and the Proteogenomics Facility and the Protein Sequencing Facility of the Medical University of South Carolina.
S The on-line version of this article (available at http://www.mcponline.org) contains supplemental material.
To whom correspondence should be addressed: Dept. of Cell and Molecular Pharmacology, Medical University of South Carolina, 171 Ashley Ave., Charleston, SC 29425. Tel.: 843-792-3209; Fax: 843-792-2475; E-mail: hildebjd{at}musc.edu
|
REFERENCES |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
dimers as well as subunits.
Cell 71, 1069
–1072[CrossRef][Medline] dimers in transmembrane signalling.
Nature 365, 403
–406[CrossRef][Medline]
