Created: 28th February 1999, last updated: 7th April 1999, © 1999 ABRF
aKathryn Stone, bJoseph Fernandez, cArie Admon, dWilliam Henzel, eWilliam Lane, fMichael Rohde, and gLaurey Steinke
a W.M. Keck Foundation Biotechnology Resource Laboratory, Yale University School of Medicine, New Haven, CT 06510; b Protein/DNA Technology Center, The Rockefeller University, New York, NY 10021; c Department of Biology, Technion, Haifa 32000 Israel; d Genentech Inc., South San Francisco, CA; e Harvard University, Cambridge, MA; f Amgen Inc., Thousand Oaks, CA; and g Protein Structure Core Facility, University of Nebraska Medical Center, Omaha, NE.
Reprinted from the electronic version of JBT available at http://www.abrf.org/JBT/JBT.html accession #0011.
Address correspondence and reprint requests to Kathyrn Stone, ABRF Protein Sequence Research Committee, 9650 Rockville Pike, Bethesda, MD 20814.
The ABRF-97SEQ sample was the 10th in a series of studies designed to aid ABRF participant laboratories in determining their abilities to obtain amino acid sequence data. The sample for 1997 was a mixture of two peptides at an approximate picomole ratio of 10:2 and was indicative of a peak that might be obtained by reverse-phase high-performance liquid chromatography (HPLC) of a tryptic digest. Participants were asked to use Edman sequencing or a combination of Edman sequencing and serial mass spectrometry (MS/MS) or postsource decay (PSD) sequence analysis to identify the primary amino acid sequence of the peptides. Cysteine was derivatized to Cys-S-propionamide (Cys-S-PAM) before sending out the sample, and a phenylthiohydantoin (PTH)-Cys-S-PAM standard was included to locate the elution position of this derivative on the participant's HPLC system. An internal standard containing norleucine and succinylated lysine was sent as a control for testing the instrument's performance; it was co-sequenced with the unknown sample. A total of 50 responses were returned, with all participants performing Edman sequencing. The accuracy of the positive correct identifications was 91.5%, with eight participating laboratories correctly identifying the major sequence. Cysteine identification improved from the ABRF-95SEQ sample, which contained an unmodified cysteine, and tryptophan identification was similar to that observed in previous studies. (J Biomol Tech 1999;10:26-32)
Key words: cysteine identification, tryptophan identification, MS/MS, mass spectrometry, protein sequencing, peptide sequencing.
ABRF-97SEQ represents the 10th in a series of unknown samples that have been distributed to members of the Association for Biomolecular Resource Facilities that perform protein sequencing. These samples are designed to provide member laboratories a mechanism with which to evaluate themselves by using a sample that presents some of the problems the facilities routinely encounter. Previous samples allowed facilities to examine the issues of sensitivity of protein sequencing,1,2 sample heterogeneity,3,4 protein-bound peptides on polyvinylidene difluoride (PVDF) membrane or in solution,5,6 posttranslational modifications,7 identification of cysteine and tryptophan,4,8,9 length of sequence assignment,4 and accuracy of sequence calls from a dataset.9 The current sample was designed to examine the ability of participating laboratories to sequence low-level mixtures of peptides--a situation often encountered in core facilities--and their ability to identify tryptophan and derivatized cysteine and be compatible with serial mass spectrometry (MS/MS) and postsource decay (PSD) techniques. The sample was designed to be similar in composition and quantity to the peptide mixture used as dataset B in the ABRF-96SEQ sequence calling study. This allowed some comparison of sequence calling expertise with the technical aspect of sequencing peptides.
Another goal of this study was to determine the success rate of cysteine determination when a prederivatized sample was provided. The cysteines in the sample were modified using acrylamide to Cys-S-propionamide (Cys-D-PAM) before sending out the sample. A phenylthiohydantoin (PTH)-Cys-S-PAM standard was supplied along with the sample so that participant laboratories could first determine the separation of this cysteine derivative on their sequencing systems. An internal sequencing standard was also supplied. This standard, which participants were asked to co-sequence with the ABRF-97SEQ sample, allowed independent monitoring of the sequencer performance. Another goal of the study was to help participants correctly use the information in a matrix-assisted laser desorption and ionization (MALDI)-mass spectrometry (MS) spectrum of ABRF-97SEQ that was provided. Participants were asked to use these data to assist in determining the length of the peptide and to help verify that the correct sequence had been called.
The composition of ABRF-97SEQ was similar to ABRF-96SEQ B, with the major 21-mer peptide containing two tryptophans, one early in the sequence and the other later, and two cysteines. The sequence was designed to reduce the number of lysines in the sample (for simplification of MS/MS and PSD sequencing) and for ease of synthesis. The composition of the minor 14-mer sequence was identical to the minor sequence in ABRF-96SEQ B but had one arginine removed. Both peptides (Fig. 1) were synthesized by Janet Crawford at the W.M. Keck Foundation Biotechnology Resource Laboratory at Yale University on a Rainin Symphony Multiple Peptide Synthesizer. Double coupling and standard 9-fluorenyl-methyl-carbonyl chemistry were used. The protonated average mass of the 21-mer was 2509.8, and the 14-mer was 1502.62.
| Major Sequence | IWTCMEGANSYQCASWAGLFK (21-mer) |
| Minor Sequence | HYAEGDESVATKPAR (14-mer) |
FIGURE 1. Major and minor sequences used in the ABRF-97SEQ sample. The two synthetic peptides were mixed at a picomolar ratio of 10:2 (major:minor) for the ABRF-97SEQ sample.
Crude peptides were purified on a Waters 600E system using a 20 X 250 mm YMC reverse-phase high-performance liquid chromatography (RP-HPLC) column (YMC Inc., Wilmington, NC) eluted at 7 mL/min. Buffer A consisted of 0.06% trifluoroacetic acid and water, and buffer B was 0.052% trifluoroacetic acid, 80% acetonitrile, and 20% water. A linear gradient from 0% to 98% of B was used over 300 minutes.
The 21-mer peptide used as the major sequence contained two cysteines and was derivatized using acrylamide by Joe Fernandez at Rockefeller University before sending out the samples. The dried peptide was reconstituted in 200 mM Tris-HCl (pH 8.0) and 100 mM DTT and was reduced by incubation at 55°C for 30 minutes. After the sample was cooled, 6 M acrylamide was added for a final acrylamide concentration of 2 M, and the sample was incubated at 37°C for 1 hour. After alkylation, the 21-mer and 14-mer peptides were purified on a 4.6 250 mm Vydac C18 reverse-phase column (Separations Group, Hesperia, CA). The gradient used was as follows with the buffers previously described: 0 to 15 minutes (2% to 37% B), 15 to 125 minutes (37% to 45% B), and 125 to 128 minutes (75% to 98% B).
Amino acid analysis was performed by standard techniques to determine the peptide concentrations. The peptides were then mixed at an approximate 10:2 ratio of the 21-mer to the 14-mer. Sequencing was performed using Edman, MS/MS, and PSD techniques by the Protein Sequence Research Committee members to ensure sample quality.
Samples were distributed to 215 ABRF member laboratories, which were asked to sequence the mixture by Edman chemical sequencing, MS/MS, PSD, or a combination of these techniques. In the case of Edman sequencing, 5 pmol of an internal standard was to be added when loading the sample on the sequencer. This 17-mer internal sequencing standard contained norleucine in positions 1, 6, 11, and 16 and succinylated lysine in the remaining positions.10 Results were reported to a third party, who removed identifying marks and forwarded the data to the sequencing committee for analysis.
Of the 215 facilities that were mailed the ABRF-97SEQ sample, 50 returned results. These sequence calls were scored as positive correct (PC), positive wrong (PW), tentative correct (TC), tentative wrong (TW), overcall (OC), or no assignment. PW, TC, and TW included calls of "end" before the end of the peptide. Table 1 summarizes the assignments made for the ABRF-97SEQ major and minor sequences. To download these data, complete with facility identification codes, http://www.abrf.org/ABRF/researchcommittees/97psrcposter/prseq97poster.html can be accessed.
TABLE 1.
Summary of Sequence Assignments for ABRF-97SEQ Compared With ABRF-96SEQ
Dataset B
|
|
||||||||||
Characteristic |
Sequence Callsa |
ABRF-97SEQ Major |
ABRF-97SEQ Minor |
ABRF-96SEQ B Major |
ABRF-96SEQ B Minor |
|||||
|
|
||||||||||
| Total # cycles | PC+TC+PW+TW |
|
|
|
|
|||||
| Avg. # cycles/R | Total # cycles/R |
|
|
|
|
|||||
| Total # correct | PC+TC |
|
|
|
|
|||||
| Total # wrong | PW+TW |
|
|
|
|
|||||
| Total # positive | PC+PW |
|
|
|
|
|||||
| Total # tentative | TC+TW |
|
|
|
|
|||||
| Total # OC | Number past end |
|
|
|||||||
| Total # unassigned | Residues not called |
|
|
|
|
|||||
| Avg. # correct | (PC+TC)/R |
|
|
|
|
|||||
| Avg. # positive | (PC+PW)/R |
|
|
|
|
|||||
| Avg. # tentative | (TC+TW)/R |
|
|
|
|
|||||
| Avg. # incorrect | (PW+TW)/R |
|
|
|
|
|||||
| Accuracy of PC assignments | PC/(PC+PW) |
|
|
|
|
|||||
| Accuracy of TC assignments | TC/(TC+TW) |
|
|
|
|
|||||
|
|
||||||||||
aAssignments were categorized as positive correct (PC), positive wrong (PW), tentative correct (TC), and tentative wrong (TW). The total number of responses (R) was 50 for ABRF-97SEQ and 95 for ABRF-96SEQ B. Overcalls (OC) are indicated, and positive or tentative assignments made in the overcalls are reflected in the total PC, PW, TC, and TW numbers.
The overall accuracy for the ABRF-97SEQ major sequence was 91.8% for positive calls, with 27 of the 50 responses containing no positive incorrect calls. Eight responses called the entire sequence correctly (PC + TC included), and two responses had no positive correct calls. One of these responses did use the internal sequencing standard, which sequenced well and indicated that the problem was caused by the sample, not by a sequencer malfunction. Either the sample did not get onto the sequencer, or the Eppendorf tube contained no sample.
The average number of cycles called was 17 of a total of 21 residues. The residues that were identified correctly (PC + TC) most often were alanine in cycle 8 (48 of 50 responses), methionine in cycle 5 (47 of 50), glutamic acid in cycle 6 (46 of 50), and glycine in cycle 7 (45 of 50). The residues that were incorrectly identified (PW + TW) most often were isoleucine in cycle 1 (10 of 50), tryptophan in cycle 16 (15 of 50), and serine in cycles 10 (10 of 50) and 15 (6 of 50). No assignment was made for 54 residues, and 11 responses (22%) had 16 residues of overcalled sequences (Fig. 2).
FIGURE 2. Bar graph results of sequence calls for ABRF-97SEQ major. The correct sequence for ABRF-97SEQ major is indicated on the bottom of the graph. Sequence calls are indicated as described in the footnote of Table 1.
For the minor sequence, there was a 74.2% accuracy for positive calls, with 12 of 50 responses making 100% correct calls. Three responses of the 50 called the entire minor sequence correct, and 22 of 50 responses made no positive correct calls. This lower percentage of accurate positive calls probably reflects the low amount (~2 pmol) of the minor peptides. Four respondents did not assign any minor sequence. The valine in residue 8 was identified correctly (PC + TC) most often (26 of 50), followed by glutamic acid in position 4 (22 of 50). The histidine at residue 1 was called incorrectly most frequently (22 of 50), and 19 had "no assignment" responses. Only one respondent identified His in cycle 1 as positive correct. Because free amino acids associated with the sample elute in the first cycle, this residue is often difficult to call. This problem, combined with the low amount of the minor sequence and the low recovery of histidine compared with other amino acids, added to the difficulty of assigning this residue. Residue 9 of the minor sequence was called incorrectly in 21 of 50 responses. The assignment of alanine at residue 9 was complicated because of carryover of the alanine from residue 8 in the major sequence.
Three responses called the major and minor sequences entirely correctly (PC + TC). Because of the low level of the minor sequence, it was not anticipated that both sequences could be called 100% correctly. However, with the use of MS, which all three laboratories routinely use, these respondents were able to correctly identify both of the peptide sequences (Fig. 3).
FIGURE 3. Bar graph results of sequence calls for ABRF-97SEQ minor. The correct minor sequence for ABRF-97SEQ is indicated along the bottom of the graph. Correct sequence assignments are indicated as described in the footnote for Table 1.
The 1997 sample (ABRF-97SEQ) was designed to be similar to the specimen used in the previous study (ABRF-96SEQ dataset B), in which a sequencing dataset, instead of a sample, was sent out. The goal was to determine whether participating facilities were having difficulties with sample handling, instrumentation, and data interpretation. This information was sought because the positive accuracy (PC/(PC + PW) in the ABRF-95SEQ study was only 78%, which was about 20% less than the positive accuracy observed in previous years' studies.
As shown in Table 1, a greater number of amino acid residues were positively correctly called by participants in 1996 for the major and minor sequences. This difference resulted in part from having approximately one half of the number of responses and one less amino acid in the ABRF-97SEQ study. The average number of correct cycles called in the major sequence was 14.5 residues in the ABRF-97SEQ study and 20.9 residues in the ABRF-96SEQ study. The minor sequence had an average number of correct calls of 4.8 residues (ABRF-97SEQ) and 10.7 residues (ABRF-96SEQ B). The peptide compositions were similar, suggesting that participating laboratories are having difficulty with sample handling, instrument performance, or both. The positive correct accuracy for the major sequence was slightly lower in the ABRF-97SEQ (91.5%) than in the ABRF-96SEQ (95.8%) study. The minor sequence also had a higher accuracy in the ABRF-96SEQ B (87.6%) than in the ABRF-97SEQ B (71.7%) study. The tentative calls were approximately equal for both major and minor studies.
The ABRF-96SEQ B dataset was obtained from a sequencing run done on an Applied Biosystems/Perkin Elmer Procise (Model 494) protein sequencer operated in the gas phase mode. The 21 responses in the ABRF-97SEQ study that were run on ABD 49X-HT sequencers (similar to the model used for the ABRF-96SEQ B study) had about the same percentage of positive accuracy for the major and minor sequences (96.5% and 80.8%) as that observed in the ABRF-96SEQ B study (95.8% and 87.6%). The average number of positive correct calls using ABD 49X-HT sequencers (ABRF-97SEQ) was 17.0, compared with 20.9 for the ABRF-96SEQ B study. When all types of sequencers from the ABRF-97SEQ study are included for determining the average number of correct calls, this number drops to 14.5.
To further study cysteine identification, the two cysteines in the major peptide were modified using acrylamide to Cys-S-propionamide.11 A PTH-Cys-S-PAM standard was supplied so that participants could first optimize the PTH separation of this derivative on their systems. Many participants, however, had the PTH-Cys-S-PAM derivative eluting with DMPTU or PTH amino acids, such as PTH-Glu. Acrylamide was chosen as the modification reagent because many proteins are isolated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) before submission to core facilities. These samples, whose cysteines might have been modified by acrylamide to Cys-S-propionamide, are then sequenced directly on PVDF membranes or digested in situ in the gel or membrane and internally sequenced. There is the possibility of sequencing a cysteine that has been modified to Cys-S-propionamide.
The positive accuracy of cysteine determination in the ABRF-97SEQ major sequence was 88% and 97% for the two cysteines at cycles 4 (C4) and 13 (C13), respectively. This level of accuracy is significantly better than for any previous ABRF study in which a sequencable sample was distributed, as can be seen in Table 2. In the ABRF-95SEQ study, for which laboratories were asked to derivatize cysteines in house before sequencing, the positive accuracy was only 63%, with 35% of the respondents reducing and alkylating the sample. The ABRF-94SEQ study emphasized cysteine reduction and alkylation and supplied various protocols to use.7 In the latter study, the positive accuracy was 82% for C10 and only 59% for C20. In the ABRF-96SEQ B study, a dataset was provided for sequence calling and contained carboxamidomethylated cysteine. The cysteine accuracy was 99% for C6 and for C13, which is excellent. Results of these studies indicate that cysteine derivatization is the problem with identifying cysteine, not identification of the cysteine derivative.
TABLE 2.
Summary of Cysteine and Tryptophan Identification
in the Past Four ABRF Sequencing Studies
|
|
||||||||
Sample |
Distributed |
Accuracy |
Accuracy |
Accuracya |
||||
|
|
||||||||
| ABRF-94SEQ | 50 pmol | 95% | C10 = 82% C20 = 59% |
W9 = 81% W23 = 44% |
||||
| ABRF-95SEQ | 45 pmol | 78% | C15 = 63% | W19 = 65% C20 = 61% |
||||
| ABRF-96SEQB Major |
10 pmol | 96% | C6 = 99% C13 = 99% |
W4 = 83% W19 = 92% |
||||
| ABRF-97SEQ | 10 pmol | 92% | C4 = 88% C13 = 97% |
W2 = 90% W16 = 64% |
||||
|
|
||||||||
aThe positive accuracy for cysteine and tryptophan is listed as the amino acid (C = Cys, W = Trp), followed by the cycle number of occurrence.
Positive identification of tryptophan was similar (90% and 83%) for the ABRF-97SEQ and ABRF-96SEQ B major sequences when it appeared early in the peptide. When tryptophan appeared late in the peptide, positive accuracy of identification was lower in the ABRF-97SEQ peptide sequenced in member laboratories (64% versus 92%). This finding is consistent with results observed in previous studies (see Table 2), in which ABRF-94SEQ W9 (W = Trp) had 82% positive accuracy, and W23 had only 44% accuracy. ABRF-95SEQ contained two tryptophans late in the sequence that were identified at lower accuracies of 65% (W19) and 61% (W20).
The internal sequencing standard was included to provide an independent means of determining the performance of the sequencers used in the study. This standard was used by 42 (84%) of 50 respondents, with 33 respondents finding it useful. In the case of one reply, in which no sequence was observed, the internal sequencing standard was observed. The problem was with the sample, not with the instrument. Another response that had no positive correct calls did not make use of the internal sequencing standard, and the instrument performance therefore could not be determined. The main problem encountered with the internal sequencing standard was that norleucine was already used as a standard.
The PTH-Cys-S-PAM standard was used by 40 (80%) laboratories. On most systems, this derivative eluted between PTH-Glu and PTH-His for ABD systems and a few on top of DMPTU. One laboratory placed it between PTH-Gly and PTH-Glu. Most Hewlett Packard users found it eluted before PTH-His.
Provided with the ABRF-97SEQ sample was a MALDI mass spectrum of the sample. The average masses observed were 1502.1 and 2510.0 ± 1 atomic mass unit (amu). Participants were asked to use this information to assist them in determining the approximate length of the peptides and to help verify that the correct sequence had been called. A total of 41% of the participating laboratories routinely use MS for estimating sample purity and peptide length or to eliminate Edman ambiguities. Seven laboratories (14%) routinely use MS/MS or PSD, and nine laboratories (18%) ran their own MS analysis of the ABRF-97SEQ sample. Six of these runs used MALDI-MS, and three used electrospray ionization-MS (ESI-MS).
There were 11 responses that overcalled the sequence. Of these, two responses indicated routine use of MS in their laboratories. These two sequence responses matched the predicted mass of 2510 within 1 or 2 amu, but researchers had made errors earlier in the sequence that led to an overcall of the sequence.
ABRF-97SEQ was designed to be compatible with MS/MS and PSD analysis. It was anticipated that a few laboratories (particularly the seven that indicated they routinely use MS/MS or PSD) would sequence the ABRF-97SEQ using these techniques or a combination of MS and Edman sequencing. Because only 1 of the 50 participating laboratories used MS/MS or PSD on this sample, the committee is led to believe that these techniques are not routinely used for obtaining primary sequence information.
Figure 4 contains the MS/MS spectra of ABRF-97SEQ obtained on approximately 1 pmol of material using a Sciex triple quadrupole mass spectrometer operated by John Stults at Genentech. The triply charged species at 837 was fragmented. As indicated in the figure, most of the Y and B series ions were observed. This offered sufficient data to enable the entire sequence of ABRF-97SEQ major to be determined, with a tentative assignment of Glu in position 6. Knowing the mass of the peptide, residue 6 was confirmed to be Glu. These data have been included as an example of the manner in which MS/MS can be used to obtain or confirm peptide sequence. In many cases, however, the MS/MS or PSD data are not of sufficient quality to allow interpretation of the entire primary sequence. A combination of Edman and MS/MS or PSD data often can be used to fill in holes obtained in an Edman sequence.
FIGURE 4. Nanospray MS/MS of ABRF-97SEQ. The MS/MS data were obtained on a Sciex triple quadrupole mass spectrometer that was equipped with a nanospray source. The Y and B ions observed are indicated.
In the ABRF-97SEQ study, eight laboratories were able to report the entire major peptide sequence correctly, and three of these laboratories also assigned the minor sequence correctly. The latter three laboratories all used ABD 494-HT sequencers. As expected, the positive accuracy of the major 10-pmol sequence (91.8%) was higher than the minor 2-pmol sequence (74.2%). However, the positive accuracy for the ABRF-97SEQ study was lower than that observed in the ABRF-96SEQ dataset B (95.8%). In light of the better accuracy and number of residues called in the ABRF-96SEQ B 10-pmol sample compared with the ABRF-97SEQ 10-pmol sample, it appears that problems at this level result from instrumentation and sample handling rather than data interpretation.
Cysteine derivatization appears to be the difficult step in cysteine determination, because the accuracy for determination of an early and a late cysteine in ABRF-97SEQ was quite high (ie, C4 = 88% and C13 = 97%). Both of these cysteines were derivatized using acrylamide to Cys-S-propionamide before sending out the ABRF-97SEQ sample.
The positive accuracy of tryptophan identification was about the same as seen in previous studies in which five times more material was sequenced. The 10-pmol amount did not appear to interfere with tryptophan identification.
Surprisingly, little use was made of MS in analyzing the sample. The survey responses indicate that MS is used only in a minority of protein sequencing facilities. The committee considered this is a valuable tool and expected wider use. MS analysis can complement Edman sequencing by providing an estimate of the number of Edman cycles needed before performing Edman degradation. A comparison of the observed mass of a peptide with the mass calculated from the Edman assignment can provide increased confidence about the sequence assignment. MS analysis can often determine a nonquantitative estimate of the number of species in the sample.
Peptide sequencing by MS/MS has some advantages over Edman sequencing. MS/MS analysis often requires less material, and blocked peptides usually can be sequenced. Modified amino acids and amino acids that often result in low recovery by Edman degradation (eg, tryptophan, serine, threonine, underivitized cysteine) are often easier to assign by MS/MS analysis. This technique can also yield sequence information from peptide mixtures, and holes or ambiguities in Edman assignments can sometimes be determined. Sequencing by Edman degradation also has some advantages over MS/MS analysis, because good mass accuracy is needed to differentiate Asn from Asp and Gln from Glu. Most MS/MS methods cannot differentiate amino acids that have the same mass, such as Ile and Leu or Gln and Lys. The major problem of MS/MS analysis of peptides is that often not all peptide bonds are observed, resulting in a gap in the sequence. Edman and MS data are complementary and essential techniques in protein characterization and should be used together whenever possible.
1. Niece RL, Williams KR, Wadsworth CL, et al. In Hugli TE (ed): Techniques in Protein Chemistry. San Diego: Academic Press, 1989:89-101.
2. Rush J, Andrews PC, Crimmins DL, et al. In Crabb JW (ed): Techniques in Protein Chemistry, V. San Diego: Academic Press, 1994:453-461.
3. Speicher DW, Grant GA, Niece RL, Blacher RW, Fowler AV, Williams KR. In Hugli TE (ed): Current Research in Protein Chemistry. San Diego: Academic Press, 1990:89-101.
4. DeJongh KS, Fernandez J, Gambee JE, et al. In Marachak D (ed): Techniques in Protein Chemistry, VII. San Diego: Academic Press, 1996:347-358.
5. Yuksel KU, Grant GA, Mende-Muller LM, Niece RL, Williams KR, Speicher DW. In Villafranca JJ (ed): Techniques in Protein Chemistry, II. San Diego: Academic Press, 1991:151-162.
6. Crimmins DL, Grant GA, Mende-Muller LM, et al. In Angeletti RH (ed): Techniques in Protein Chemistry, III. San Diego: Academic Press, 1992:35-35.
7. Mische SM, Yuksel KU, Mende-Muller LM, Matsudaira P, Crimmins DL, Andrews PC. In Angeletti RH (ed): Techniques in Protein Chemistry, IV. San Diego: Academic Press, 1993:453-461.
8. Gambee JE, Andrews PC, Grant GA, Merrill B, Mische SM, Rush J. In Crabb JW (ed): Techniques in Protein Chemistry, VI. San Diego: Academic Press, 1995:209-217.
9. Fernandez J, Admon A, De Jongh K, et al. In Marshak DR (ed): Techniques in Protein Chemistry, VIII. San Diego: Academic Press, 1997:69-78.
10. Elliott JI, Stone KL, Williams KR. Anal Biochem 1993;211:94-101.
11. Brune DC. Anal Biochem 1992;207:109-116.