| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Molecular & Cellular Proteomics 5:949-958, 2006.
© 2006 by The American Society for Biochemistry and Molecular Biology, Inc.
,
,,¶,||
From the Department of Proteomics and Signal Transduction, Max Planck Institute of Biochemistry, Am Klopferspitz 18, 82152 Martinsried, Germany and ¶ Center for Experimental BioInformatics (CEBI), Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark
 |
ABSTRACT |
|---|
 TOP ABSTRACT EXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
CID is known to efficiently fragment proteins in ion traps, but these mass spectrometers lack sufficient resolution to resolve large protein fragment ions and their charge states. With the hybrid ion trap-FTICR mass spectrometer (LTQ-FT, Thermo Electron Corp., Bremen, Germany) it is possible to fragment large peptides or even protein ions in the ion trap and detect them with high resolution and accuracy by FT-ICR (7, 8). However, in our experience, and as shown here, the LTQ-FT is less suitable for detection of fragments produced in the LTQ due to lower sensitivity and time-of-flight effects. On the other hand, the ions can be fragmented in the ICR cell using methods like ECD and infrared multiphoton dissociation. Furthermore although the LTQ-FT is a commercial and robust instrument, the necessity for a high magnetic field detector and relatively high maintenance costs tend to limit its use to specialized laboratories.
Very recently a new hybrid mass spectrometer, the LTQ-Orbitrap (Thermo Electron), was introduced (9). It consists of a linear quadrupole ion trap (LTQ) coupled to a novel mass analyzer, the orbitrap, invented by A. Makarov (10–12). In the orbitrap, ion packages circle between two concentric electrodes, and their axial motion is detected, as in the FT-ICR instrument, by recording their image currents followed by Fourier transformation of the time domain signal to obtain the mass spectrum. Importantly the orbitrap is very compact and requires no magnetic field or special maintenance. LTQ and orbitrap are coupled via the C-trap, an intermediate radio frequency-only storage device, which can also be used to store background ions of known composition. When analyte ions are added and analyzed together with this "lock mass," sub-ppm mass accuracy for peptides is achievable (13).
No systematic analysis dealing with intact proteins and their fragments in the LTQ-Orbitrap has been reported so far. In this study, we explored the utility of the LTQ-Orbitrap mass spectrometer for top-down analysis of proteins ranging in mass from 10 to 25 kDa. Our results show that the instrument is capable of routinely achieving high sensitivity (attomole to femtomole range), high mass accuracy (low ppm), and isotope resolution of small proteins. In addition, selected multiply charged protein ions can be successfully fragmented in two stages, MS2 and MS3, in the linear ion trap, and their fragments can be transferred and measured in the orbitrap. We demonstrate ready identification of modified and unmodified proteins by MS2 and MS3 data.
|
EXPERIMENTAL PROCEDURES |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
-crystallin A and B chains (catalog number C-41633), bovine ß-lactoglobulin (catalog number L-3908), bovine ß-casein (catalog number C-6905), and recombinant human ubiquitin (catalog number U-5382). Proteins were dissolved in methanol/water with 0.5% formic acid immediately before analysis. Sample concentration ranged from 100 fmol/µl (acquisition of whole protein spectra) to 1 pmol/µl (acquisition of MSn spectra). A sample volume of 0.5 µl was delivered to the mass spectrometer using a NanoMate 100 system (Advion Biosciences, Ithaca, NY). A NanoMate low flow 2.5-µm ESI chip was used as static nanoelectrospray emitter, providing a stable flow of 20–30 nl/min. Voltages of 1.3–1.6 kV were applied to the chip through the NanoMate power supply, while the mass spectrometer source voltage was set to zero. Samples were infused using nitrogen gas at a pressures of 0.7–1.0 p.s.i.
LTQ-Orbitrap Mass Spectrometry—
Measurements were performed on an LTQ-Orbitrap mass spectrometer (Thermo Electron) in the positive ion mode. Mass accuracy (FT) was calibrated immediately before measurements according to the manufacturer’s instructions. CID of selected protein charge states was performed in the LTQ, and fragments were subsequently transferred and measured in the orbitrap. Selected protein charge states (typically the most abundant ones) were isolated with a width of m/z = 6–10 and activated for 30 ms using 30% normalized collision energy and an activation q of 0.25. The instrument was controlled using TunePlus 2.0 (beta 3), and the acquired spectra were evaluated using Xcalibur 2.0 software.
The orbitrap automated gain control (AGC) targets were set to 2 x 106 charges for full scan and 2 x 105 for MSn scan. Protein mass spectra were acquired at a resolving power of 60,000, and MSn spectra were acquired at 60,000, 30,000, or 15,000. Lock mass option was enabled in all measurements unless otherwise stated, and polydimethylcyclosiloxane (PCM) background ions (at m/z 445.120025 and 429.088735) were used for real time internal calibration as described previously (13). Unless otherwise stated, all orbitrap scans consisted of 10 microscans (see below).
Protein masses were determined either by deconvolution using the integrated Xcalibur Extract software (Thermo Electron) or by direct calculation from the peak positions and charge states. Expected protein fragment masses were calculated using PILGrinder software (developed in house by Peter Mortensen), Protein Prospector software (prospector.ucsf.edu/) (14), or GPMAW (General Protein/Mass Analysis for Windows) (Lighthouse data) (15). MS2 spectra were searched with the web-based ProSight PTM (prosightptm.scs.uiuc.edu/) (16) against either the human UniProt or a custom bovine database in "Absolute Mass" mode. A wide tolerance of up to 2000 Da was used for the protein mass to allow for differences between measured and theoretical masses due to protein modification. Fragment ion mass tolerance was in all cases set to 5 ppm, and at least five matched fragments were required for protein identification.
MS3 spectra were searched against the National Center for Biotechnology Information non-redundant (NCBInr) protein database (April 15, 2005; 244,0425 sequences) using the Mascot search engine (Matrix Science, London, UK) (17). Search criteria were as follows: no enzyme specificity; precursor mass tolerance, 5 ppm; and fragment mass tolerance, 0.01 Da. Because Mascot cannot handle MS3 data, we manually added the mass of H2O (18.0106 Da) to all precursor ions that gave good quality MSn spectra but did not result in protein identification. This formally turns b-type ions or internal fragments into peptide precursors. To match MS3 fragments by Mascot when the precursor was a b-ion, y-ions were allowed to match only with H2O loss.
LTQ-FTICR Mass Spectrometry—
Measurements were performed on an LTQ-FT mass spectrometer (Thermo Electron) in the positive ion mode. Bovine cytochrome c, bovine ß-lactoglobulin, and bovine ß-casein were prepared for measurement and delivered to the mass spectrometer using the NanoMate 100 system in a manner as identical as possible to the LTQ-Orbitrap measurements. The instrument was fully calibrated prior to all measurements according to the manufacturer’s instructions.
CID of selected protein charge states was performed in the LTQ, and fragments were subsequently transferred and measured in the ICR cell. An isolation width of m/z = 10–15 was used for selected protein charge states, which were subsequently activated for 30 ms using 30% normalized collision energy and an activation q of 0.25. The instrument was controlled using TunePlus 1.1 (beta 4), and the acquired spectra were evaluated using Xcalibur 1.4 software.
The AGC target values were set to 2 x 106 for full scan and 2 x 105 for the MSn scan. A resolving power of 100,000 was used in acquisition of protein spectra, whereas 50,000 was used for MSn spectra. All FT scans consisted of 10 microscans.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
First we wanted to establish optimal conditions for protein measurement on the LTQ-Orbitrap. Nanoelectrospray (18, 19) is commonly used in top-down proteomics because it allows detailed investigation of complex samples for extended periods of time and because it is very sensitive. Here we used a newly developed very low flow nanoelectrospray chip (NanoMate, 2.5-µm nozzle inner diameter, Advion Bioscience), which supports stable flow rates in the true nanoelectrospray range of 20–30 nl/min. This allowed us to routinely use a volume of 0.5 µl and still acquire data for more than 15 min, more than sufficient time for all measurement sequences.
We investigated optimal parameters for acquisition of whole protein spectra in the orbitrap mass analyzer. The best sensitivity, S/N, and accuracy were obtained when each scan consisted of 10 microscans. In this regime, transients of 10 consecutive microscans are added to form a final transient on which Fourier transformation is performed. Data acquisition time was between 10 and 30 s at the resolution chosen (see below), so it was still very short compared with total available spray time. At the concentrations used, fill times for the 106 target value chosen was between 0.2 and 4 s, comparable to the transient time of 750 ms.
In non-mass-resolved mode, target values of up to 107 are possible in the LTQ part of the instrument, values much higher than the limit of about 106 of the C-trap. However, in mass-resolved mode, for example when storing a charge state for subsequent dissociation, only about 105 ions can be accumulated. In this mode, the ability to sequentially fill the C-trap would be useful as pointed out previously (13).
Measurement of Intact Proteins
Sensitivity—
At the intact protein level, well defined "envelopes" arising from detection of multiple charge states were routinely obtained in the orbitrap for all investigated proteins at concentrations of 100–500 fmol/µl and total protein amounts of less than 250 fmol (Table I). The lowest amount analyzed in this study was 50 fmol of cytochrome c (Fig. 1). Note that this was the total protein amount used for measurement and that high S/N protein mass spectra were obtained even after one MS scan (10 microscans) on a population of about 2 x 107 ions (low attomole range) and with an acquisition time of 6 s. The same amount of cytochrome c was detected on the LTQ-FT instrument under the same measurement conditions albeit with lower intensity and S/N (Supplemental Fig. 1).
|
|
Resolution—
All orbitrap protein spectra presented here were recorded at 60,000 resolution at m/z 400, which is the specified resolving power of the orbitrap mass analyzer. This resolution requires "1-s" measurements (750-ms transients), whereas the maximum possible measurement time of 1.8 s leads to a resolution of 100,000. Although this value is less than the maximum possible with high magnetic field FT-ICR analyzers, it was sufficient for isotope resolution of all investigated proteins (Figs. 1 and 2) and provided a good duty cycle. Furthermore resolution in the orbitrap is inversely proportional to the square root of the m/z value rather than to the m/z value directly as it is in the FT-ICR analyzers (10), resulting in slower decrease of observed resolution across the mass range. In practice this means that at m/z values above 1111.1 resolution power of 60,000 (m/z 400, 750-ms transient) of the orbitrap analyzer exceeds resolution power of 100,000 of the 7-tesla LTQ-FTICR instrument. This unique feature is useful in analysis of intact protein mass spectra where protein charge states are often observed above m/z 1000. It has to be noted, however, that longer transient acquisition in the LTQ-FT instrument leads to significantly higher resolution (specified to up to 500,000), whereas in the orbitrap mass analyzer the transient acquisition time is limited to 1.8 s, and consequently its resolving power is limited to about 100,000. Isotope resolution directly enables the detection of modifications such as disulfide bridges (
m = 2 Da) or deamidation (m = 1 Da) of small proteins, whereas the ability to resolve isotopic clusters of different protein charge states across the mass range is important for proper charge state assignment in complex spectra caused by overlapping protein charge distributions.
|
Masses and standard deviations for each protein standard were calculated from up to 10 members of each isotopic distribution in at least two different charge states and two different scans. For larger proteins, where the monoisotopic peak could not readily be determined, the isotope state was calculated using the top-fitting method described by Zubarev et al. (22) with the cutoff level of h = 50%. Alternatively the protein mass was determined from the protein spectra deconvoluted by Xcalibur software. As expected, all protein standards were measured with very high mass accuracy, between 0.92 ppm (S.D., 0.79 ppm) and 2.80 ppm (S.D., 1.02 ppm) (Table I). The overall (combined) absolute mass accuracy was 2.25 ppm with average S.D. of 1.46 ppm. The mass accuracy of the cytochrome c measured in the LTQ-FTICR analyzer was 5.1 ppm (S.D., 2.7 ppm), which was lower than that measured in the orbitrap, probably due to relatively low signal intensity.
Although FT-ICR MS is capable of achieving extremely high mass accuracies, there are only a few reports where accuracy was better than 10 ppm at the protein level. The main reason for this is the space-charging effect, which depends on the ion population in the ICR cell. Although several strategies have been proposed to overcome this effect (23, 24), the best results were obtained when the ion population was well controlled or when internal calibration was performed. For example, Lee et al. (25) have used a dual ESI source to infuse 
-endorphin as internal calibrant and enabled "mass locking" to its doubly charged ion. In a high throughput LC-FT-ICR analysis of intact proteins of the yeast large ribosomal subunit they have identified about a hundred proteins with average absolute accuracy of about 2.5 ppm and S.D. of 2 ppm in cases when protein monoisotopic mass could be measured (25).
Although a detailed study of space-charging effects in the orbitrap mass analyzer has not yet been published, we have not observed systematic mass shifts as a function of ion number up to the limit imposed by the C-trap of about 106 charges. Thus in our experiments mass accuracies for all six standard proteins were never worse than 7.5 ppm even without lock mass option. In addition, protein masses were measured with extremely high precision, almost always better than 2 ppm (Table I). Together these data demonstrate that the LTQ-Orbitrap can routinely achieve extremely high mass accuracy at the protein level without any software or hardware modification.
Analysis of PTMs at the Protein Level—
One of the strengths of the top-down approach in MS-based proteomics is potentially comprehensive characterization of PTMs. Three of the proteins analyzed in this study, namely ß-casein and -crystallin A and B chains are known to be phosphorylated. In the case of ß-casein, five phosphorylation sites have been reported, and the mass measured in the orbitrap corresponded to the mass of the protein form containing all five phosphate groups. In contrast, both A and B chains of -crystallin were detected in several forms; A chain was detected as unmodified and modified with one phosphate group; B chain was detected as unmodified and modified with one, two, and three phosphate groups (Fig. 2). Stoichiometries of these modifications were in agreement with those observed in a previous MS study (26). Note that mass differences arising from modifications were measured with high precision (10 mDa, or two significant decimal places), potentially enabling discrimination of modifications with the same nominal mass, such as acetylation and trimethylation (m = 35 mDa), at the intact protein level. In addition, it is clear from Fig. 2 that sensitivity for detection of phosphoproteins was excellent in this case because 250 fmol of total protein mixture had been loaded, and the singly and doubly phosphorylated crystallin B chain makes up less than 15% of this amount.
Measurement of Protein MS2 Fragments in the Orbitrap
Accurate protein mass alone would be of little use in the analysis of unknown proteins or protein mixtures. As discussed before, various fragmentation methods have therefore been applied to intact proteins (or their isolated charge states) to obtain at least partial information on protein primary structure. In the LTQ-Orbitrap fragmentation must be performed outside of the orbitrap mass analyzer; therefore two fragmentation types are currently possible. Proteins can be fragmented by nozzle-skimmer (in-source) dissociation or CID in the LTQ with subsequent detection of fragment ions either in the ion trap or in the orbitrap. Our initial studies using nozzle-skimmer fragmentation resulted in very little fragmentation without useful sequence information (data not shown); therefore all fragmentation was performed by CID in the LTQ.
As already reported in experiments on the LTQ-FTICR instrument, intact proteins readily fragment in the LTQ under the same conditions normally used for peptide sequencing (7, 8). Various charge states of all analyzed proteins produced multiply charged fragment ions that were isotopically resolved and showed excellent S/N (Figs. 3 and 4). As in the case of intact protein measurements, spectra were acquired as the sum of 10 microscans. Efficient transfer from the LTQ to the C-trap and orbitrap and absence of time-of-flight effects insured that fragments could be acquired in a single mass range. MS2 fragments were matched by ProSight (Single Protein mode) (16). Because ProSight considers only b- and y-ions, internal fragments were calculated using Protein Prospector (14) or GPMAW software (15). As demonstrated for ß-lactoglobulin and ß-casein, fragments were measured with average absolute accuracy better than 2 ppm in the orbitrap (Supplemental Table 1).
|
|
Fragmentation Patterns—
Fig. 3A shows an MS2 spectrum of ß-lactoglobulin upon isolation and CID fragmentation of its [M + 15H]15+ charge state in the LTQ and subsequent detection of fragment ions in the orbitrap. The fragmentation patterns of proteins in our study were in agreement with the ones reported previously for ion trap CID of intact proteins (7, 8, 27). Cleavages C-terminal to charged residues, in particular Asp, Glu, and Lys, as well as N-terminal to Pro were the major fragmentation channels, although other cleavages were observed as well (Figs. 3B and 4B). In MS2 spectra of cytochrome c, ubiquitin, and ß-lactoglobulin, complementary b/y fragments were observed. However, ß-casein fragmented to 
70 residues from the N terminus and 25 residues from the C terminus, leaving the central portion of the molecule uncovered (Fig. 4). Importantly all detected b-ions formed by CID of its [M + 22H]22+ charge state that encompassed the five predicted phosphorylation sites indeed had all five phosphate groups attached. Together with low mass, unmodified b-ions, this locates all phosphorylation sites to between residues 9 and 57 of the sequence. Because the loss of phosphoric acid from the peptide backbone, commonly observed in the "bottom-up" approach, was not observed after CID of intact proteins, modified fragments could be fragmented further to give insight into the exact locations of protein modifications.
It is well known that CID fragmentation of whole proteins depends largely on the protein structure (such as the positions of disulfide bonds) and size and that larger proteins (>20 kDa) tend to fragment mostly in terminal regions. It is both an advantage and a disadvantage of CID that it concentrates fragmentation products into a few preferred channels. This increases the sensitivity for detecting these fragments but will often preclude complete characterization with single residue resolution. ECD can potentially cleave almost all peptide bonds in a protein (28); however, it suffers from relatively low efficiency, which results in lower sensitivity and increased acquisition times. Therefore, these methods are complementary rather than competing. Electron transfer dissociation, a recently developed fragmentation technique for ion traps, would be very suitable for the LTQ-Orbitrap because it would combine the advantages of nonergodic fragmentation (high sequence coverage, preservation of labile bonds, etc.) with the high resolution, sensitivity, and accuracy of the orbitrap mass analyzer demonstrated here.
Protein Identification Using MS2 Data—
High accuracy measurement, typically better than 2 ppm, and straightforward assignment of the charge states of fragment ions together with the accurately determined protein mass were used for protein identification. Monoisotopic masses of the MS2 fragments were submitted to ProSight PTM (Absolute Mass mode), currently the only publicly available search engine for top-down proteomics. MS2 spectra of recombinant human ubiquitin were searched against the human UniProt database, whereas all other proteins were searched against both a custom made bovine database and the human UniProt database. The protein mass tolerance was set to 2000 Da to allow for potential modifications, and a minimum of five matched fragments with accuracy of 5 ppm or better were required for a hit. These stringent criteria led to unambiguous identification of all analyzed proteins. In the analysis of FLAG-tagged ubiquitin, the protein was identified with 10 fragment ions (average absolute mass accuracy, 2.45 ppm; S.D., 0.92 ppm) and probability score of 7.4 x 10–19. As expected, its mass was 1597.42 Da higher than the theoretical mass, corresponding to the difference of the FLAG tag (DYKDDDDKKLMV) plus linker sequence and Met at its N terminus (m = 0.030 Da). In the case of cytochrome c, the measured mass differed from the theoretical mass by 616.172 Da, which corresponds to the mass of the heme group (m = 0.006 Da). This not only shows the potential of the top-down approach to conclusively identify proteins but to point to modifications not considered in the database as well. -Crystallin A and B chains, ß-casein, and ß-lactoglobulin were also identified as unique, high probability hits; however, their scores were not representative because the custom made bovine database contained only eight entries. When searched against the human database MS2 spectra of these proteins did not lead to any hits. This demonstrates that high accuracy MS2 spectra of whole proteins acquired in the orbitrap can routinely be used for high stringency database search in top-down proteomics, leading to high confidence identification and greatly constraining the nature of a possible modification.
Comparison to LTQ-FT Performance in MS2—
Detection of LTQ-derived CID fragments of bovine ß-lactoglobulin and ß-casein was directly compared between the orbitrap and the FT-ICR mass analyzers. As expected, under the same measurement conditions, fragment ions produced in the LTQ and transferred to the ICR cell resulted in spectra of lower S/N (Supplemental Fig. 2). This resulted in a lower number of identified CID fragments as compared with the orbitrap and hence demonstrated higher sensitivity of the orbitrap mass analyzer in the MSn mode The low and high mass regions of the FT-ICR mass spectra were particularly weak due to the time-of-flight effect the ions experience in the about 1-m-long flight tube between the LTQ and the ICR cell. However, longer accumulation times and higher AGC MSn target values significantly improved the appearance of the FT-ICR MS2 spectra (not shown).
Top-down Proteomics Using MS3 on the Orbitrap
We have shown recently that an additional level of peptide fragmentation (MS3) is feasible on the LTQ-FTICR at high sensitivity and chromatographic time scale (29). Together with a new scoring algorithm, these data significantly improved confidence of peptide identifications. With this background and the excellent quality of the MS2 spectra, we decided to explore the feasibility of the LTQ-Orbitrap for top-down MS3 experiments.
Fig. 5A shows the MS2 spectrum of ubiquitin. The predominant ion is y587+ caused by proline-directed cleavage. This ion was isolated and dissociated in the LTQ, and resulting fragments were analyzed in the orbitrap (Fig. 5B). Even a single microscan acquisition (0.6 s) led to a good S/N spectrum, which was further improved by summing 21 spectra (Fig. 5B). The spectrum contains many informative ^b and y ions (^b refers to the MS3 fragment ion generated from the end of the sequence truncated by the MS2 cleavage, see Ref. 29). To generate MS3 spectra similar in quality to Fig. 5B typically required selecting one of the dominant ions of the MS2 spectrum and a total acquisition time less than 1 min. Thus, several different MS3 spectra can be acquired of a single protein loaded at a few hundred femtomoles into the NanoMate. We also explored higher stages of MSn. Fig. 5C shows an MS4 spectrum analyzed in the LTQ and demonstrates that interpretable spectra can still be obtained after four rounds of isolation and fragmentation.
|
Finally we decided to investigate whether small, peptide-like MS2 fragments of proteins could be fragmented and used for protein identification in a manner similar to protein identification in the bottom-up approach. The advantage of such a strategy would be that small fragments are less likely to be modified and that there are relatively few possible fragmentation channels. We chose a small fragment of ß-casein that was easily determined to be doubly charged based on its isotope spacing (see Fig. 4A). This fragment was accumulated in the LTQ, fragmented, and analyzed in the orbitrap. Fig. 6 shows a relatively simple MS3 spectrum similar to the MS2 spectrum of a small peptide (see Supplemental Table 2 for the peak list). Several changes were made to the normal Mascot modus to be able to search these data. Because the MS3 precursor could be a b or y ion of the protein, both possibilities were checked. If a y ion did not lead to any matches, then the mass of water was added to the precursor mass to formally convert it from a b ion to a peptide precursor. Mascot was directed to match only b or y – H2O ions (these are the ^y MS3 fragments generated from b ion precursors (29)). Despite the short peptide length, a "non-enzyme" specificity search by Mascot in the non-redundant database (NCBInr) yielded only two significant matches. The top match was located in the N-terminal region of the ß-casein precursor. A check of the database confirmed that Mascot had identified the eight N-terminal residues of the mature protein. The second significant hit occurred on a different protein and had a related peptide sequence accounting for matched fragments despite the high mass accuracy but was located close to the center of that protein sequence and was therefore discarded. Because the protein was modified with five phosphogroups, this example shows an interesting additional way of identifying modified proteins.
|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Published, MCP Papers in Press, February 13, 2006, DOI 10.1074/mcp.T500042-MCP200
1 The abbreviations used are: PTM, posttranslational modification; LTQ, linear quadrupole ion trap; AGC, automatic gain control; PCM, polycyclodimethylsiloxane; ECD; electron capture dissociation; S/N, signal-to-noise ratio. 
* Work on this project at Max Planck Institute of Biochemistry was supported in part by "Interaction Proteome," a 6th Framework grant from the European Union research directorate. Work at CEBI was supported by a generous grant from the Danish National Research Foundation. 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.
S The on-line version of this article (available at http://www.mcponline.org) contains supplemental material.
Both authors contributed equally to this work.
|| To whom correspondence should be addressed. Tel.: 49-89-8578-2557; Fax: 49-89-8578-2219; E-mail: mmann{at}biochem.mpg.de
|
REFERENCES |
|---|
|
TOPABSTRACTEXPERIMENTAL PROCEDURESRESULTS AND DISCUSSIONREFERENCES |
|---|
-crystallin A chain.
Anal. Chem.
75, 6509
–6516[Medline]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| All ASBMB Journals | Journal of Biological Chemistry |
| Journal of Lipid Research | ASBMB Today |
