Roland S. Annan
SmithKline Beecham Pharmaceuticals
Phosphorylated proteins are known to play a critical role in transmitting signals from outside the cell into the nucleus. In fact, protein phosphorylation is probably the single most common intracellular signal transduction event. Among the thousands of proteins expressed in a typical mammalian cell, as many as one-third are now thought to be phosphorylated. Knowledge of the specific residues phosphorylated in a protein can help to elucidate the signaling pathway. Identifying phosphopeptides in protein digests usually requires labeling the protein with radioactive [32P]-phosphate. Several rounds of HPLC purification may then be necessary to isolate a single peptide and prepare it for Edman sequencing. Unfortunately, the Edman chemistry itself is not particularly well suited to identifying phosphorylated residues. Mass spectrometry is a tool that has proven to be very useful in solving protein sequence problems. In recent years, the Research Mass Spectrometry Laboratory at SmithKline Beecham Pharmaceuticals and others have developed mass spectrometry methods for mapping phosphorylation sites in proteins and peptides (1-4). This article describes our experience with a simple approach which utilizes positive-ion matrix-assisted laser desorption/ionization (MALDI) time-of-flight (TOF) mass spectrometry for identifying and sequencing phospho-peptides.
Time-of-flight mass spectrometers can be classified into two broad categories,
linear and reflector type instruments. To appreciate how these instruments can
be useful for phosphopeptide analysis, it is worthwhile to describe their
principles of operation. A linear instrument (Figure 1A) 
(16k)consists of an ion
source (where the MALDI process takes place), a flight tube (where ions of
different masses are separated from one another), and a detector. After the
ions are formed by the MALDI process, they are accelerated out of the source,
reaching their final kinetic energy before entering the flight tube. They
travel the length of the flight tube at full accelerating potential, until they
strike the detector. The classical equation for kinetic energy, E:
E = (mv2)/2 equation 1
where m is mass and v is velocity, tells us that because all ions have been accelerated to the same energy, their velocities will be inversely proportional to the square root of their molecular masses:
v = (2E/m)1/2 equation 2
Thus ions of greater mass will travel slower than lighter ones and therefore reach the detector later. Assuming they have the same charge, the flight times of two different ions are directly proportional to the square root of their masses.
The resolution of a mass spectrometer can be defined simply as its ability to
separate two ions of different mass. Simple linear MALDI-TOF instruments are
rather low resolution mass spectrometers (5). For peptides, this means they
cannot measure the distribution of the various naturally occurring isotopes for
each peptide or separate peptides that differ in mass by only a few daltons
(see Figure 2A).
(16k) One primary reason for this is the initial kinetic energy
spread of individual ion populations. Ions of any particular mass will have a
small distribution of energies centered around the accelerating voltage (for
instance, Vacc = 25 kV). Because of this, they will also have a small spread in
velocities. This velocity spread will give rise to a broader peak for that ion,
because members of the same ion population will arrive at the detector at
slightly different times.
A reflectron TOF instrument (Figure 1B) corrects for the energy spread created during acceleration by using an ion mirror or reflector. The back of the reflector is at a voltage slightly higher than the source accelerating voltage. The reflector works by slowing an ion until it stops, is turned around, and is re-accelerated back to a second detector. Ions with an initial kinetic energy lower than Vacc (and a lower velocity) will not penetrate the reflector as deeply and therefore will turn around sooner, catching up to ions with a full 25 kV of kinetic energy. Ions with energies greater than Vacc (and thus higher velocities) will penetrate more deeply into the reflector and be turned around later. Their flight is retarded, allowing other ions to catch up. All this gives a distribution of ions with the same mass-to-charge ratio more similar flight times, thus improving resolution. Figure 2 shows the molecular ion region of the linear and reflector spectrum of a synthetic tyrosine phosphorylated peptide TRDIYETDpYYRK (monoisotopic Mr = 1701.8), recorded in the positive-ion mode with [[alpha]]-cyano-4-hydroxycinnamic acid ([[alpha]]HCA) as matrix. The isotopically resolved ion cluster in Figure 2B shows the contribution that all the isotopes of C, H, N, O, and P make to the elemental composition of this peptide. In addition to resolution, mass accuracy is also improved when spectra are recorded in the reflector mode (6-7). Using external calibrations one can routinely expect mass accuracies on the order of 0.01-0.05% (1-5 Da per 10,000). On the other hand, in the linear mode of operation, if one carefully records both sample and reference spectra at the same laser irradiance and ion intensities are similar, a mass accuracy of only about 0.1% can be achieved (8).
Figure 3 
(16k)shows the positive-ion linear and reflector MALDI-TOF spectra of the
phosphoserine-containing tryptic peptide VPQLEIVPNpSAEER (monoisotopic Mr =
1659.7), isolated from [[alpha]]-casein S1. The presence of the phosphorylated
and dephosphorylated forms of the peptide in the reflector spectrum is the
result of post source decay (PSD). PSD is a process whereby a precursor ion
that is metastable (that is, which is sufficiently stable to be transported out
of the ion source, but insufficiently stable to survive the flight to the
detector) decomposes in the flight tube. It is believed that ions acquire
excess internal energy via multiple collisions with matrix ions in the source.
In Figure 3, a percentage of the phosphopeptide ion population has decomposed
in the flight tube, creating [MH-H3PO4]+ and [MH-HPO3]+
fragment ions. This series of ions is characteristic of serine and threonine
phosphorylated peptides. If these ions disappear when the spectrum is recorded
in the linear mode, one can be reasonably certain the peptide is
phosphorylated.
Once in the flight tube, all ions will have reached full accelerating voltage; therefore, when the precursor phospho-peptide ion decomposes, the fragments will each have a fraction of the full accelerating voltage. In order to maintain conservation of energy, the fragment ions must have kinetic energies proportional to the ratio of their masses to the mass of the precursor ion and totaling 25 kV. Recalling equation 2, we can see that because the ratio of energy to mass has remained the same, all the fragments will have the same velocity as the precursor. This means that the precursor and fragments will all reach the linear detector at the same time and are thus all detected at the same apparent mass. For this reason, the linear spectrum in Figure 3 shows only an [M+H]+ ion. However, in the reflector-TOF instrument, the ions will be energy focused by the reflector. Thus the dephosphorylated fragments, with less than the full 25 kV of energy, do not penetrate as deeply into the ion mirror as the phosphorylated precursor and thus have their flight times decreased relative to the precursor. The fragments are detected at a lower mass. [H3PO4] and [HPO3], having no charge, are not affected by the mirror and strike detector 1.
The major fragment ion that is observed for serine and threonine phosphorylated peptides is [MH-H3PO4]+ or MH-98 daltons. It appears in the reflector spectrum as MH-97, because the reflector is calibrated for ions with the full accelerating energy (in other words, the precursor ion). The fragment ion, with less than full accelerating energy, is not focused properly and thus appears at the wrong mass. The deviation between expected and observed mass increases as the fragment ion masses decrease and their energies lower (thus being less and less well focused by the reflector). The [MH-H3PO4]+ fragment ion does not correspond to the dephosphorylated peptide but contains dehydroalanine in place of the phosphoserine residue; it is therefore not observed at the same mass as the non-phosphorylated peptide. However, if the non-phosphorylated peptide is present in the same HPLC fraction or sample as the phosphopeptide, the [MH-H3PO4]+ ion from the phosphopeptide will overlap with the [MH-NH3]+ fragment ion from the non-phoshorylated peptide. If this is suspected, a linear spectrum should be recorded. With regards to phosphorylated peptides purified from protein digests, our experience has been that, if of average length (8-15 residues), the non-phosphorylated peptide will usually elute in a later HPLC fraction.
A second fragment ion, common to serine, threonine, and tyrosine phosphorylated peptides is [MH-HPO3]+ or MH-80 daltons. This ion is much less abundant than [MH-H3PO4]+ and is, of course, equivalent to the non-phosphorylated peptide, but for the reasons discussed above it will appear in the spectrum at MH-79 daltons. In the case of serine and threonine phosphorylation, the [MH-HPO3]+ ion is always observed with the more abundant [MH-H3PO4]+ ion.
The reflectron spectrum of a phosphotyrosine containing peptide from vSrc,
LIEDAEpYAAR (monoisotopic Mr = 1229.5), is shown in Figure 4, inset.
Phosphotyrosine containing peptides fragment to lose HPO3 (M-80) as discussed
above. However, since phosphotyrosine is much more stable than either
phosphoserine or phosphothreonine, this ion is normally not very abundant. The
loss of H3PO4, typical of serine and threonine phosphorylation, is not favored
in the case of a phosphotyrosine- containing peptide because it would require
cleaving a bond adjacent to an aromatic ring, leaving a radical on an aromatic
ring. [MH-98]+ ions observed in the spectra of phosphotyrosine-
containing peptides are likely the result of sequential losses of HPO3 and H2O.
Our experience is that the reflectron spectra of tyrosine phosphorylated
peptides rarely display this ion, but we almost always observe it to a limited
extent when we record a full post source decay product ion spectrum with
precursor ion selection (See Figure 4, main panel).
(16k)
It should be noted that the ratio of fragment to parent ion is matrix dependent. [[alpha]]-Cyano-4-hydroxycinnamic acid ([[alpha]]HCA) yields the most intense fragment ions while 2,5-dihydroxy-benzoic acid (DHB) gives the least, as is shown in Figure 3B and 3C. In spite of what we originally suspected, the ratio of fragment to parent ion seems to show little dependence on laser irradiance. In the range of 300-900 fmol, the ratio of intact phosphopeptide to fragment ion remained rather constant, no matter how energetically we irradiated the sample. In practice, spectra are always recorded using the lowest possible laser irradiance, because under these conditions resolution will be the greatest. For serine and probably for threonine phosphorylated peptides, however, the ratio of fragment to parent ion does appear to be sequence dependent, because some peptides yielded almost entirely fragment ions while others gave abundant phosphopeptide precursor ions. We have examined only a limited number of multiply phosphorylated peptides, but in all cases (serine and tyrosine phosphorylation) the reflector spectra showed sequential losses of neutral phosphate. In these cases, recording a linear spectrum is essential to making some sense out of the data, because in linear mode all fragment ions will revert to a single MH+ ion. We have found that we can reduce fragmentation by lowering the initial field strength of the acceleration region, but unfortunately this also degrades the overall resolution. It is also likely that the final accelerating voltage, the length of the flight tube, and the residual gas pressure in the instrument will affect the degree of fragmentation.
This simple method for identifying phosphopeptides is especially useful because it can easily be done at the femtomole level. We typically remove 1 ul from a 30-40 ul HPLC fraction, mix it 1:1 with matrix, and spot 0.5 ul on the target. With 20-50 pmol of protein digest loaded onto a 1 mm column, the fractions (not dried down) usually give very good spectra even though the phosphorylated peptide may be present in very low yield relative to the non-phosphorylated peptide. Of course all the usual caveats apply: the ionization efficiency of the peptide is somewhat sequence dependent, it may not be detected it in a complex mixture because of suppression effects (again this is sequence dependent), and because the dynamic range of MALDI is not particularly good, the peptide may not be seen if it is present in very low abundance relative to another peptide in the same fraction.
An additional advantage of reflectron type mass spectrometers, just recently realized, is the potential to sequence peptides (9). The same PSD process that gives rise to the loss of phosphate from phosphorylated peptides will fragment the peptide backbone in a predictable way. Recall that all fragment ions will have kinetic energies less than the precursor. Therefore, if the mass spectrometer is to detect these fragment ions at their correct masses, the reflector voltage must be lowered to bring these ions into energy focus. The reflector voltage is normally set to optimally focus ions of full accelerating voltage. A complete PSD product ion spectrum can be acquired by successively lowering the reflector voltage, thereby bringing lower and lower mass fragment ions into focus, and recording the desired mass range in segments. The segments may be calibrated and assembled afterward. PSD spectra resemble low-energy collision-induced dissociation spectra, similar to those commonly recorded on triple quadrupole mass spectrometers. The spectra contain mainly bn and yn fragments with abundant internal fragments and low mass immonium ions (for a review of peptide fragment ion nomenclature see reference 10).
Figure 4 (main panel) shows the complete PSD spectrum of the tyrosine phosphorylated tryptic peptide (LIEDAEpYAAR) from Figure 4, inset. An ion gate (11) with a resolution of approximately 100 was used to select the MH+ precursor at m/z 1230.5. If this was an unknown sequence, the large MH-80 ion (MH-HPO3) and small MH-98 ion (in this case MH-HPO3-H2O) would strongly suggest that this is a tyrosine, not serine or threonine phosphorylated peptide. The presence of a strong phosphotyrosine immonium ion (labeled pY in Figure 4) at m/z 216 supports this. Because the peptide has a C-terminal Arg, the spectrum is dominated by a yn series. An m/z 175 is characteristic of a y1 ion in peptides where the C-terminal residue is Arg. The mass difference between successive y ions relates to an amino acid residue mass. For example, the mass difference of 71 between y1 and y2 indicates the second C-terminal residue is Ala. The residue mass between y3 and y4 is 243 Da and tells us the fourth C-terminal amino acid is pTyr (163+80). A table of residue masses can be found in reference 10. The yn series stops at residue 8 (from the C-terminus), leaving 226 Da of unassigned sequence. A short bn series provides some sequence overlap, and the b2 ion confirms the first two amino acids as Leu-Ile. If this were an unknown sequence, the first two amino acids could be readily identified through a database search using the molecular weight and the known portion of the sequence. In this case, the spectrum provides some additional information about the identify of the first two N-terminal residues. The b2 ion at m/z 227 suggests a composition of either Xle, Xle (where Xle is either Leu or Ile) or Pro, Glu--because these are the only di- or tri-peptides that fit this mass--but does not indicate the order. Using these compositions as starting points, the spectrum can be searched for internal fragments. The series XE, XED, and XEDA (where X is either Leu or Ile) confirms that the second N-terminal residues is Xle, and therefore that the first two residues must be Xle, Xle. Unfortunately, because Leu and Ile have the same residue mass, it is not possible to distinguish them by PSD. The reflectron and PSD spectrum shown in Figure 4 were recorded on 1 pmol of peptide loaded onto the target. Similar spectra have been recorded in our laboratory on 250 fmol loadings (determined by quantitative amino acid analysis) of unfractionated protein digest recovered from two-dimensional gels and as little as 100 fmol of peptide standard (12).
The data presented here indicate that phosphopeptides can be identified in positive-ion MALDI reflectron spectra based on the presence of characteristic fragment ions. This suggests that it may be generally possible to distinguish tyrosine phosphorylation from serine or threonine phosphorylation--at least for singly phosphorylated peptides--based on the type of fragment ions present. Any ambiguity in assigning of ion structures can be resolved by re-recording a linear-mode spectrum. Because fragment ions are not detected at their correct masses in the normal mode of reflector operation, in some cases it may be difficult to assign a structure to an ion because of uncertainty in the mass assignments. This can usually be alleviated by recording one segment of a PSD spectrum and measuring the ions at their correct masses. Previously, Talbo and Mann (13) demonstrated that sulfated peptides could be distinguished from phosphorylated peptides based on their negative-ion MALDI reflectron spectra. They found that in negative-ion mode, tyrosine, serine, and threonine phosphorylated peptides all gave essentially identical reflectron spectra, showing an ion at [M-H-97]-. The extent of fragmentation we have observed in the positive-ion reflectron mode is dependent primarily on the matrix used and the strength of the initial acceleration region. Other reflectron instruments have either single-stage acceleration regions or do not allow control of the initial acceleration voltage. Therefore, the extent of the fragmentation described above may vary from instrument to instrument.
Experimental
A Fisons VG TofSpec SE single-stage reflectron time-of-flight mass spectrometer was used to record the spectra shown in this manuscript. In reflector mode it has a maximum resolution of about 6000 (FWHM). This instrument has a 3.4 meter effective path length and is co-axial in geometry. The three-element ion source provides a variable ion extraction region, useful for obtaining high fragment ion yields. Typical final accelerating voltage is 25 kV. Samples are irradiated by 337 nm photons from a pulsed nitrogen laser. Spectra are calibrated externally using two peptide standards. Precursor ions for PSD are selected using a Bradbury-Nielsen type ion gate (11), which has a resolution of about 100. PSD spectra are typically recorded under computer control in either 8 or 10 segments. Segments are joined and calibrated under computer control using the VG Opus data system, a process that takes 2-4 seconds. After reviewing the initial spectrum, weak segments can be recorded again and added to or substituted for the original segments. Product ion resolution of more than 2000 (FWHM) has been achieved on this instrument, making monoisotopic mass determinations possible up to m/z 1800. This greatly reduces the uncertainty in assigning structures to ions.
Sample preparation was discussed above. [[alpha]]HCA (Aldrich) matrix is prepared by dissolving 10 mg in 1 ml of 50:50 ethanol:acetonitrile. DHB (Fluka) is prepared by dissolving 10 mg in 1 ml of 50:50 ethanol:water.
The reader who would like more background on MALDI and PSD is referred to references 14-15.
Acknowledgments
The author would like to acknowledge the other members of the Research MS group at SmithKline, most especially Steven Carr, Mike Huddleston, and Gerry Roberts.
References
1. Huddleston, M. J., Annan, R. S., Bean, M. F., and Carr, S. A. (1994) J. Am. Soc. Mass Spectrom. 4, 710.
2. Affolter, M., Watts, J. D., Krebs, D. L., and Aebersold, R. (1994) Anal. Biochem. 223, 74-81.
3. Nuwaysir, L. M. and Stults, J. T. (1993) J. Am. Soc. Mass Spectrom. 4, 662-669.
4. Liao, P. C., Leykam, J., Andrews, P. C., Gage, D. A., and Allison, J. (1994) Anal. Biochem. 219, 9-20.
5. A technique known as "time-lag energy focusing" (Wiley and McLaren, Rev. Sci. Instr. (1953) 26, 1150-1157) or "delayed extraction" has recently been shown to provide significant resolution improvements for MALDI on linear TOF instruments. This technique has been successfully implemented on one commercial TOF-MS, showing resolution of 7000 (FWHM) for bovine insulin on a 2.1 meter linear instrument (M. Vestal, P. Juhasz, and S. Martin (1995), Rapid Commun. Mass Spectrom. 9, 1044-1050). Several other vendors are currently attempting to adapt this technique to their existing mass spectrometers.
6. Vorm, O. and Mann, M. (1994) J. Am. Soc. Mass Spectrom. 5, 955-958.
7. Talbo, G., Morand, K., Margis, L., and Mann, M. (1993) Proceedings of the 41st ASMS Conference on Mass Spectrometry and Allied Topics, San Francisco, CA, p. 947.
8. Beavis, R. C. and Chait, B. T. (1990) Anal. Chem. 62, 1836-1840.
9. Kaufmann, R., Spengler, B., and Lutzenkirchen, F. (1993) Rapid Commun. Mass Spectrom. 7, 902-910.
10. Biemann, K. (1990) in Methods in Enzymology, Volume 193, Mass Spectrometry (McCloskey, J. A., ed.) pp. 886-888, Academic Press, New York; and references contained therein.
11. Bradbury, N. E. and Nielsen, R. A. (1936) Phy. Rev. 49, 388.
12. Carr, S. A., Roberts, G., Annan, R. S., Hemling, M. E., and Hoyes, J. (1995) Proceedings 43rd ASMS Conference on Mass Spectrometry and Allied Topics, Atlanta GA, p. 620.
13. Talbo, G. and Mann, M. (1994) in Techniques in Protein Chemistry V (Crabb, J. W., ed.), Academic Press, San Diego, pp. 105-113.
14. Kaufmann, R. (1995) J. Biotechnol. 41, 155-175.
15. Karas, M., Bahr, U., Ingendoh, A., Nordhoff, E., Stahl, B., Strupat, K., and Hillenkamp, F. (1990), Anal. Chim. Acta, 241, 175-185.
The author may be contacted at SmithKline Beecham Pharmaceuticals, Dept. of Physical and Structural Chemistry, 709 Swedeland Rd., King of Prussia, PA 19406.
Return to the The ABRF Home Page