Proteomic Analysis of Human Immunodeficiency Virus Using Liquid

A major challenge to studying virus-incorporated host proteins is the fact that they are not ... Our comparison of wild-type and mutant viruses demons...
0 downloads 0 Views 284KB Size
Proteomic Analysis of Human Immunodeficiency Virus Using Liquid Chromatography/Tandem Mass Spectrometry Effectively Distinguishes Specific Incorporated Host Proteins Andrew C. S. Saphire, Philippe A. Gallay, and Steven J. Bark* The Center for Protein Sciences and the Department of Immunology, The Scripps Research Institute, 10550 N. Torrey Pines Road, La Jolla, California 92037 Received August 19, 2005

A major challenge to studying virus-incorporated host proteins is the fact that they are not encoded by the viral genome. We used Liquid Chromatography/Tandem Mass Spectrometry (LC-MS/MS) on whole virions to obtain a snapshot of the HIV-1 proteome. We identified known viral and host-cellular proteins and also identified novel components of HIV-1 and confirm these by traditional biochemical methods. Our comparison of wild-type and mutant viruses demonstrates that LC-MS/MS has the specificity to distinguish the presence/absence of a single host protein in intact virions. Keywords: proteomics • HIV • tandem mass spectrometry • liquid chromatography

Introduction Liquid chromatography/tandem mass spectrometry (LCMS/MS) is an analytical method for identifying multiple component proteins comprising a highly complex biological matrix.1,2 Despite this capability, the analysis of viruses has only recently been recognized.3 Perhaps the greatest potential of LC-MS/MS lies in its capacity to directly identify proteins without reference to the viral genome. The importance of this potential can be recognized by the fact that many viruses have been found to contain host cellular proteins (i.e., Vesicular Stomatitis Virus, Cytomegalovirus and HIV).4-10 While many host proteins are taken up nonspecifically, some such as Cyclophilin A (CypA), have been shown to be incorporated specifically and to play an important role in viral replication.4,11-14 These proteins could provide attractive targets for antiviral therapies because; (1) they can perform critical functions in the virus lifecycle and (2) they are not subject to the same mutational pressures as viral proteins because they are not encoded by the viral genome. Existing methods for identifying host cellular proteins in viruses have a number of drawbacks, including the need for large quantities of radiolabeled material, subsequent methodologies to positively identify proteins, and the fact that only relatively abundant proteins can be distinguished. We evaluated the potential of LC-MS/MS to directly analyze the contents of whole enveloped viruses. The objective of this study is to define the scope, capabilities and limitations of this (LC-MS/ MS) technology for studying enveloped viruses using samples of known composition. Furthermore, we performed our LCMS/MS experiments using the most basic and easily reproduced experimental protocols (i.e., no reduction and alkylation and single-phase C18 capillary HPLC) to confirm that such * To whom correspondence should be addressed. Phone: (858) 784-9941. Fax: (858) 784-2684. E-mail: [email protected].

530

Journal of Proteome Research 2006, 5, 530-538

Published on Web 02/11/2006

methods would be capable of identifying host-cellular proteins in whole virus samples. Furthermore, we utilize a model system amenable to biochemical confirmation of any discoveries made in our proteomic analyses. Our experiments demonstrate that LC-MS/MS has many advantages that are complementary to traditional methods. HIV-1 is almost unique in that it has defined cell culture methods that produce both wild-type and mutant viruses of known composition. Several host cellular proteins are already known to be associated with HIV-1 virions including CypA (CypA),4,7 Elongation Factor 1-alpha (EF-1R),8 and Heat Shock Protein 70 (Hsp70).9 CypA is known to be specifically incorporated into HIV-1 and defined mutants that fail to incorporate this protein without other perturbations of the virus are available.4,7 Furthermore, biochemical methods for confirming proteomic results are readily available. Finally, because of the global impact of HIV-1, there is an urgent need to develop methodologies to identify novel targets for the development of antiviral therapeutics. In this study we examined the potential of LC-MS/MS to identify the known viral protein components of HIV-1, the known HIV-1-associated hostcellular proteins previously identified by traditional methods, and previously un-recognized host-derived proteins incorporated into HIV-1. We assess the specificity of this methodology by comparing wild-type viruses with mutants known to fail to incorporate a single host protein, Cyp A. Our study demonstrates that LC-MS/MS provides a rapid and sensitive means of characterizing the proteome of whole retroviruses and provides a basis for extending these studies to other viruses important in human disease.

Materials and Methods HIV Isolation and Digestion. All viruses in this study were generated by standard methods previously identified for pro10.1021/pr050276b CCC: $33.50

 2006 American Chemical Society

research articles

Proteomic LC-MS/MS Analysis of HIV

ducing known wild-type or mutant HIV virions using either transfection of 293T cells, or by propagation in the human CD4+ Jurkat T-cell line. Specifically, 40 µg of the proviral clones R915 (NL4.3 derivative) and R9 G89V as described previously20 were used to transiently transfect 293T cells as described previously16 or into Jurkat T-cells by electroporation. Viral supernatants were harvested 48h post-transfection and filtered through a 0.2 µm-pore-size filter to remove cellular debris. Intact virus was further purified by isolation to buoyant density on a 20-70% sucrose gradient (approximately 1.17 g/mL). Collected fractions were subsequently re-pelleted through a sucrose cushion. Total yield of virus was measured via antiHIV-1 capsid p24 ELISA (Perkin-Elmer Life Sciences). After isolation, 16 µg samples (p24) of viral particles were subjected to exhaustive proteolysis at 37 °C with 20 units of sequencing grade trypsin (Promega) in 25 mM ammonium bicarbonate buffer at pH 7.5 and 0.5% octylglucopyranoside. No reduction and alkylation was used in the preparation of these samples. After 16-24 h, the virus digests were either directly analyzed or stored at -80 °C until analysis. Protein Analyses and Antibodies. Purification and immunoblot analysis of viruses produced from 293T transfected cells were conducted as previously described.16 Anti-CA IgG was obtained through the AIDS Research and Reference Program. Rabbit anti-CypA serum was obtained by immunization with recombinant human CypA protein generated using the previsouly described GST-CypA plasmid.17 Anti-CypA IgG was purified on CypA affinity and protein A columns. Anti-CD48 was obtained from Chemicon. HIV was quantified using an established elisa assays for the capsid p24 (LifeScience). Sample Preparation. For each experiment, 16 µg of crude virus digest was prepared by C18 Ziptip (Millipore) with standard elution into 20 µL of 60% acetonitrile/40% water with 1% formic acid. The eluent was concentrated under vacuum to a volume of approximately 4-5 µL, diluted with 15 µL 1% formic acid in water, then 5 µL wasloaded onto the capillary column for LC-MS/MS analysis. LC-MS/MS Acquisition. LC-MS/MS acquisition was performed on a Thermoquest LCQ Deca Ion Trap Mass Spectrometer fitted with a nanospray source. Liquid chromatography was performed using a 100 µm I.D. capillary column with a 5 µm tip (pulled using a Sutter Instruments Laser Capillary Puller, Model P-2000). The column was packed with 8-12 cm of reversed-phase silica (Targa C18, Higgens Analytical). Peptides were eluted using a 90 min gradient of 5-55% acetonitrile acid in water, each containing 0.1% formic acid. The applied column voltage was 2.60 kV using a liquid junction. The mass spectrometer was set for two stages of mass spectral analysis (MS and MS/MS) with dynamic exclusion set at a 3 amu window with a 1 min time limit. LC-MS/MS Data Analysis. LC-MS/MS data was processed using Sequest18 to query observed data against the NCBI-nr, human and HIV protein specific databases. Proteins identified were first ranked according to the statistical criteria correlation value (Xcorr) and difference between the best and next best identification (∆Cn). Xcorr value filters for peptides were set to >1.5 for +1 peptides, >2.0 for +2 peptides and >3.5 for +3 peptides, while the ∆Cn value was set to >0.085. The filters Xcorr and ∆Cn values were marginally lower than standard (>1.8 for +1 peptides and >2.5 for +2 peptides and ∆Cn >1.000 for some experiments) to increase the probability of identifying low abundance proteins and increase our opportunities for spurious identifications. These statistical thresholds were adopted

prior to obtaining experimental results. Identifications were considered only for peptides exhibiting appropriate statistics, fully tryptic or partially tryptic sequences, and satisfactory manual analysis of the MS/MS data demonstrating good ion isolation, fragmentation and signal-to-noise (see Figure 1). Heparin Affinity Chromatography. Purified HIV viruses were lysed with 0.5% octylglucopyranoside in 100 mM ammonium acetate buffer and incubated for 2 h at 4 °C with 50 µL of packed BSA- or heparin-Sepharose beads (Sigma). Beads were settled by gravity and unbound viral proteins were removed by washing eight times with cold PBS. Gravity settling limits the level of nonspecifically associated proteins in the sample. Bound proteins were eluted from the affinity column with a step NaCl gradient of 50, 150, 500, and 1000 mM salt. The eluted proteins were dialyzed against 25 mM ammonium acetate and lyophilized. The lyophilized proteins were digested for 24 h with 20 units of sequencing grade trypsin in 25 mM ammonium bicarbonate, pH 7.5 prior to LC-MS/MS analysis. As described for the digest of whole virus, no reduction and alkylation was used in the preparation of these samples.

Results LC-MS/MS Accurately Identifies Known HIV-1 Encoded Proteins from Complex Mixtures. Previous studies have used a variety of biochemical methods, such as 2D electrophoresis,19 to isolate individual proteins or small subsets of proteins from viruses for subsequent analysis via mass spectrometry. However, no study has employed LC-MS/MS as a proteomic tool to analyze whole retroviruses without the use of other biochemical methods to separate proteins prior to analysis. Given that whole proteomic analysis of nonenveloped viruses using LC-MS/MS has been performed with success,3 we reasoned that this approach could be used to analyze lipid-enveloped viruses such as retroviruses. First we asked whether LC-MS/ MS can be used to identify the known virus-coded proteins of HIV-1 from intact purified viruses. 293T cells, a commonly used producer cell for molecular analysis of HIV-1, or Jurkat cells were transfected with the proviral construct NL4.3, which encodes a wild-type T-cell tropic CXCR4 virus, via calcium phosphate precipitation. After 48 h, viral supernatants were harvested and filtered, and viral particles were purified by sucrose gradient density purification as described previously.20 To generate HIV-1 peptides for analysis, pelleted virus was digested for 24 h at 37 °C using trypsin. To ensure full access of the protease to the viral interior, digestions were performed in the presence of the detergent n-octylglucopyranoside, which permeabilizes the lipid viral membrane. Viral digests were prepared by C18 Ziptip and eluted with 60% acetonitrile/40% water with 1% formic acid, concentrated under vacuum and then loaded onto the capillary column for LC-MS/MS analysis using a Thermoquest LCQ Deca Ion Trap Mass Spectrometer fitted with a nanospray source. LC-MS/MS data was processed using Sequest to query observed data against the NCBI-nr, human and HIV protein specific databases. Proteins identified were first ranked according to the statistical criteria correlation value (Xcorr) and difference between the best and next best identification (∆Cn). Identifications were considered only for peptides exhibiting appropriate statistics, fully tryptic or partially tryptic sequences, and satisfactory manual analysis of the MS/MS data demonstrating good ion isolation, fragmentation and signal-to-noise (see Figure 1). Remarkably, whole digests of the viral pellet from both 293T and Jurkat cells resulted in the positive identification of virtually Journal of Proteome Research • Vol. 5, No. 3, 2006 531

research articles

Saphire et al.

Figure 1. MS/MS spectra for the EF-1a peptide DNVGFNVK in the +1 (top) and +2 (bottom) charge states. Statistical data for these identifications are Xcorr ) 2.068 and ∆Cn ) 0.1000 for the +1 charge state and Xcorr ) 2.282 and ∆Cn ) 0.1724 for the +2 charge state. Spectra are cut at m/z ) 905 as no significant data is present above this value. The lower table lists the predicted y and b ion series from the identified sequence.

all known HIV-1 encoded proteins (Table 1A). In each of three separate experiments, we detected envelope protein, all of the components of HIV-1 Gag including nucleocapsid, matrix, capsid, P6 and P2, as well as the components of Pol including reverse transcriptase, and protease. Testifying to the sensitivity of this approach, we also detected a number of the accessory proteins, which are far less abundant than the HIV-1 structural proteins, such as Vpr. Interestingly, in one of the preparations, we also detected the presence of Vif. This is consistent with early studies which suggest that Vif can be incorporated into HIV-1.21 However, no HIV-1 protein has been demonstrated to serve as a site for the specific incorporation of Vif in HIV1.22 It is possible that the presence of Vif is either due to the co-purification of immature viruses, which have been shown to incorporate Vif,23 or due to nonspecific incorporation.24 LC-MS/MS Correctly Identifies Host Cellular Proteins Known to be Incorporated into HIV-1. Although LC-MS/MS has been used before to identify subset components of viruses such as in adenovirus,3,25 it has not been applied to the analysis of enveloped proteins. We asked if host cellular proteins known 532

Journal of Proteome Research • Vol. 5, No. 3, 2006

to be incorporated into HIV-1 can be detected using LC-MS/ MS. By this method we identified all of the host proteins that have been shown to be incorporated into HIV-1 via other methods, including HSP70, CypA, EF-1R,4,8,11,26 Parvulin,27 glyceraldehyde-3-phosphate dehydrogenase,27 Lck,27 Ubiquitin,28 and SUMO-129 in our viral preparations. These host proteins/ modifications were found in HIV-1 viruses derived from both 293T and Jurkat cells (Table 1B). Importantly, control supernatants which include mock transfected 293T and Jurkat cells prepared in parallel did not detect these human proteins, suggesting that their detection arises exclusively from purified virions. This finding demonstrates that LC-MS/MS can be used to recognize proteins associated with the retrovirus that cannot be predicted by scrutiny of the viral genome. One class of host cellular protein not recognized to be incorporated into HIV-1, histones, was identified in all LCMS/MS experiments. We consistently identified all of the histone subunits known to comprise packaged chromatin including histones H1, H2A, H3, and H4. Indeed, Figure 2 demonstrates that, for histone H2A, it is probable that the entire

research articles

Proteomic LC-MS/MS Analysis of HIV Table 1. Comparative LC/MS/MS Can Accurately Components of HIV-1 in 293T and Jurkat Cells: Examples of HIV-1 Proteins (A) and Host Cellular Proteins (B) Identified in Both 293T and Jurkat Producer Cells by LC-MS/MS A

B

GAG polyprotein precursora HIV-1 Integrase POL polyprotein precursora

Cyclophilin Aa Heat Shock 70kDa Human Elongation Factor-1 Alpha (EF-1R)* Histones H1,aH2A,a H3,a H4a beta-globina Trypsin Precursor* Parvulin Glyceraldehyde-3-phosphate dehydrogenasea Lck, Ubiquitina SUMO-1 CD48

Capsida Nucleocapsida p17 matrix* p6a p2a VPR Vif

Figure 2. Sequence of Histone H2A from Homo sapiens. Peptides identified in LC-MS/MS experiments for Histone H2A (Table 2) are represented in BOLD underlined. Sequence coverage is 37% with 48 of 129 amino acids represented in identified peptides.

a Denotes that the protein was identified by multiple entries and identifier numbers in the FASTA databases. Databases used were NCBI-nr, Human, and HIV-specific.

protein is incorporated. While it is reasonable to suppose that the identification of histones represents a bona fide association with viral genomic RNA, there is reason to regard this conclusion with caution. Specifically, it is known that compact RNA/ Histone replication complexes derived from HIV infected cells are known to co-sediment with HIV-1 through sucrose gradients,30 which is consistent with our findings. Interestingly, comparison of host-cell contents of HIV-1 produced in 293T revealed little difference with those derived from Jurkat cells. Recent evidence indicates that the cell type used to produce HIV-1 has a significant impact on the infectivity of the virus.31 However, our LC-MS/MS data from 293T and Jurkat cell experiments did not demonstrate significant differences in the primary proteins identified. While these data would be consistent with the known similar infectivity for HIV derived from these cell lines, our limited sensitivity from using a singlephase C18 LC-MS/MS experiment may also preclude our observation of differences in lower abundance proteins. LC-MS/MS Accurately Discriminates between Incorporated Host-Cellular Proteins Known to be Incorporated into HIV-1. Although the experiments above demonstrate that LCMS/MS can be used to detect incorporated host proteins, it is possible that the high sensitivity of LC-MS/MS may misleadingly over-represent trace amounts of protein adventitiously co-purified with viruses in addition to those specifically incorporated. Thus, we next asked if LC-MS/MS can correctly distinguish between a wild-type virus and one genetically modified to exclude the incorporation of a single host protein. Specifically, we compared the contents of wild-type viruses with those that fail to incorporate CypA due to a genetic modification of the CypA incorporation site in the HIV-1 capsid. This mutation in the capsid protein (CA) at Glycine 89 to Valine or Alanine completely blocks CypA incorporation into HIV-1 and has no other effect on the protein content of HIV-1.7 We generated wild type and mutant G89V viruses in parallel by transfecting either 293T or Jurkat cells with proviral constructs as described above, and purified these viruses simultaneously for analysis by LC-MS/MS and Western blot. As shown by Western blot (Figure 3), the mutation completely abrogates CypA incorporation. Comparison of the peptide sequences generated from the wild type and G89V mutant virus exactly parallel the findings by western blot: in multiple independent

Figure 3. Western blot of wild-type HIV-1 and G89V mutant probed with antibodies to capsid (top) or CypA (bottom), showing that the mutant excludes CypA without altering capsid production. Viruses were produced via transfection of 239T cells and purified by sucrose gradient.

runs using virus from 293T cells and electroporated Jurkat cells, numerous peptides corresponding to CypA were detected from wild type virus, yet no peptide derived from CypA was identified from the G89V mutant. This finding is particularly significant given that CypA is abundant in serum-supplemented growth media and in cell cytosol. Furthermore, these findings demonstrate that standard HIV-1 purification methods are sufficient to generate LC-MS/MS data that can be used to compare viral compositions. Specifically, we asked whether other host proteins bind preferentially to the G89V mutant capsid in the absence of CypA. Conversely, other proteins may be recruited to the HIV-1 capsid complexed with CypA. Given that HIV-1 virions lacking CypA are less infectious than wild-type virus,4,12,13,26 the presence or absence of other host proteins may be directly related to HV-1 infectivity. Importantly, comparison of the LC-MS/MS data from wild-type and the G89V virus revealed a subset of proteins present in the mutant, but not in the native virus, including the 60s Ribosomal Protein, CD99, Probable Transcription Activator EDRF1, and RNA Helicase A (data not shown). Further experiments are underway to examine whether these proteins contribute to HIV-1 pathogenesis. Altogether, these data demonstrate that LC-MS/MS can reproducibly and accurately detect the presence or absence of a single protein in a complex mixture of proteins derived from an intact virus. Journal of Proteome Research • Vol. 5, No. 3, 2006 533

research articles

Saphire et al.

Figure 4. Western blot reveals that CD48 is incorporated into HIV-1 derived from both 293T cells Jurkat T-cells (Thick arrows at 45 kDa bands). Jurkat or 293T cells were lysed via freeze thaw in the presence of detergent lysis buffer, normalized for protein content and probed with anti-CD48. Viral pellets corresponding to approximately 1 µg p24 from viruses produces either from Jurkat cells or 293T cells were probed on the same blot. Note that we have included strips from a much longer exposure for the viral pellets, as the intense signal from the Jurkat cell lysate required an extremely brief exposure. The high molecular weight bands (Thin arrows at ∼75 kDa) present in the Jurkat viral pellet are likely a rare multimer form of CD48 present in the preparation as CD48 is highly abundant in Jurkat cell membranes. 293T cells are not derived from immune cells and have significantly less CD48 abundance.

LC-MS/MS Identification a Host-Derived Heparan Sulfate Binding Protein in the HIV-1 Envelope. In enveloped viruses, transmembrane proteins are taken up from the host plasma membrane during budding and comprise part of the viral surface. For example, the host cell-surface proteins HLA-DR, ICAM-1, CD40, CD40L, and CD86 have all been shown to be incorporated into the HIV-1 envelope.32 Furthermore, although gp120 is the principle ligand for HIV-1 attachment to primary cells, ICAM-1 has been shown to enhance HIV-1 entry.33,34 We previously reported that although HIV-1 produced from 293T cells lacking gp120 fails to bind T-cells and T-cell derived cell lines, it nevertheless remains able to bind to adherent HeLa cells in a heparan sulfate dependent manner.35 This effect is producer cell dependent, as 293T cell derived virus attaches in a CypA dependent manner, whereas T-cell derived virus does not.35 This points to the possibility that the supplemental factors are taken from the host cell membrane and is consistent with the notion that the producer cell influences HIV-1 pathogenicity.31 A challenge to the identification of these hostderived viral factors is the fact that they are not encoded by the virus, but are likely taken up by the virus at the time of budding. Thus, we asked if LC-MS/MS could be used to identify membrane associated host cellular proteins incorporated into HIV that could mediate binding to heparin sulfated proteoglycans (HSPG’s). We analyzed HIV proteins isolated from affinity chromatography using heparin as an affinity reagent bound to sepharose beads. Specifically, purified HIV-1 viruses were lysed with 0.5% octylglucopyranoside and incubated for 2 h at 4 °C with heparin-Sepharose beads or control BSA- Sepharose beads. Beads were washed, and bound proteins were eluted from the affinity column with a step NaCl gradient of 50 mM, 150 mM 500 mM and 1000 mM salt. Eluted proteins were dialyzed against ammonium acetate buffer, digested with trypsin and analyzed by LC-MS/MS as described above. After extensive washing, three proteins eluted from the heparin beads at 50mM salt which did not elute from the control BSA beads. 534

Journal of Proteome Research • Vol. 5, No. 3, 2006

Table 2. Sample of Proteins Observed in Wild-Type HIV Grown in 293T Cells protein IDa

peptides in identification (charge state)b

Pol (Reverse PISPIETVPVK (+1) Transcriptaste) DLIAEIQK (+1) IVDIIATDIQTK (+1) SESELVSQIIEQLIK (+2) Vpr EWTLELLEEIK (+2) HSP 70 DLGGGTF (+1) DAGVIAGLNVLR (+1) GVPQIEVTFD (+1) DVSILTIEDGIFEVK (+2) DIEIDSLFEGIDFYTSITR (+2) Cyclophilin A EGMNIVEAMER (+1) VSFELFADK (+1) EF-1R DNVGFNVK (+1) DNVGFNVK(+2) QLIVGVNK(+1) Histone H2A VTIAQGGVLPN(+1) AGLQFPVGR(+2) TAEILELAGNAAR (+1) VGAGAPVYLAAVLEYLTAEILELAGNAAR (+3) VGAGAPVYLAAVLEYLTAEILELAGNAA (+2) CD48 LDPQSGALYISK (+2)

trypticc

Xcorr (∆Cn)d

P

2.207(0.2401)

T T T P N T P P P T T P P T P T P T

2.366(0.0938) 3.317(0.3268) 4.771(0.1011) 3.515(0.1039) 1.712(0.1244) 2.426(0.1871) 1.594(0.1162) 4.587(0.0850) 5.249(0.4684) 2.507(0.2108) 2.418(0.1899) 2.068(0.1000) 2.282(0.1724) 2.580(0.1449) 1.888(0.1703) 2.552(0.081) 2.843(0.2209) 4.985(0.2904)

P

5.796(0.4336)

T

3.151(0.3154)

a

The protein identities based on the peptides observed. bThe peptide sequences and charge state (in parentheses) are reported. cThe tryptic status of the peptide identified. T ) completely tryptic, P ) one cleavage site tryptic. d The statistical correlation values for each peptide in column 2. Xcorr is reported first, then Delta Cn is reported in parentheses. The statistical threshold for reporting a peptide as significant was set at Xcorr)1.5 for a +1 charge state, Xcorr)2.0 for a +2 charge state and Xcorr)3.5 for a +3 charge state. Delta Cn threshold was set at 0.085. Proteins identified by a single peptide with appropriate statistical significance were reported.

Two of these proteins were already known to be HIV-associated and known to bind to heparin: the HIV envelope glycoprotein gp12036,37 and the host cell protein CypA.20 Interestingly, the third protein that specifically bound to heparin beads was a protein not previously known to be associated with the HIV-1 envelope- the CD48 antigen (Table 2). This is a membrane-

Proteomic LC-MS/MS Analysis of HIV

linked 45 kDa protein which is widely recognized to be found on the surface of leukocytes and the major HIV target, Tlymphocytes, including Jurkat cells.38,39 Confirming the relevance of our finding, CD48 is known to specifically bind to heparin.40 Western blot analysis of purified HIV isolated from 293T and Jurkat cells confirmed that CD48 was incorporated into virus derived from these cell types (Figure 4), confirming our LC-MS/MS data. Note that while CD48 is known to be abundantly expressed in T-cells and T-cell lines, it has not been previously reported to be expressed in 293T cells. This may in part be due to the fact that the relative expression of CD48 is considerably higher in Jurkat cells than in 293T cells (Figure 4).

Discussion In this study, we demonstrate the applicability of a simple LC-MS/MS technique to the proteomic analysis of HIV-1. Even using a single-phase HPLC system with all of the inherent limitations, we identified almost all of the known viral, and known incorporated host-cell derived proteins of HIV-1. In addition, we show that LC-MS/MS, in combination with affinity chromatography, is suitable for the identification of viral membrane-associated proteins (i.e., CD48). Highlighting the efficacy of this approach to identify novel HIV-1-associated proteins, we demonstrate for the first time that the cell surface protein CD48 is incorporated into HIV-1 via LC-MS/MS and affinity chromatography, and verify this finding by western blot. Most importantly, we demonstrate the specificity of this methodology by comparing wild-type viruses with mutants known to fail to incorporate a single host protein, Cyclophilin A. This is the first time LC-MS/MS has been shown to have the capacity to distinguish the presence or absence of a single protein from a complex mixture of retroviral proteins. The objective of our experiments was to determine the scope, capabilities, and limitations of LC-MS/MS proteomic methods by studying virus samples of known composition. Furthermore, we attempted these experiments using a simple C18 HPLC separation without further fractionation and without reduction and alkylation. An important difference in our application of LC-MS/MS methods when compared to most current literature on this subject is that we utilize this technology to define possible incorporated proteins within intact virus particles. Despite our “lower confidence” in our protein identifications compared to more stringent identification criteria, we can directly validate our results in a relevant biological system. Most other examples of LC-MS/MS applied to biological systems use both highly stringent identification criteria and more complex separation procedures. Highly stringent identification criteria will increase the certainty of identification for defined proteins, but could eliminate many protein identifications that would be possible and testable in a biological context. More complex experiments using multiple fractionation methods and sample preparation would certainly improve data quality and improve our sensitivity for very low abundance proteins. However, these techniques are beyond the scope of the limited studies described here. Our data demonstrates that the capabilities of even basic LC-MS/MS experiments are considerable and most proteins known to be incorporated into HIV virus particles were identified. Our choice of experimental system was critical for these experiments because it permits direct biochemical confirmation of identified proteins. HIV-1 was a logical choice because both wild-type and mutant viruses produced in 293T and Jurkat cell

research articles lines have been exhaustively studied. Jurkat cells, in particular, are well suited for this purpose being derived from CD4+ T-cells, the principal physiological target for HIV-1. 293T cells and Jurkat cells are routinely used to produce HIV-1, as it permits comparison of a wide range of strain and mutants from a common cellular context. The use of these cells is critical to our analysis as it allows direct comparison of our proteomic results with the vast body of biochemical data in the literature. In particular, wild-type and Cyp A deficient G89V HIV-1 viruses were produced via transient transfection. This was necessitated by the fact that the loss of Cyp A in the G89V mutant renders the virus incapable of infection, and thus cannot be propagated in PBMCs.4,7,12-14,41-47 However, it must be emphasized that HIV-1 produced in 293T cells and Jurkat cells are fully infectious, and thus by definition these viruses contain all the hostcell derived proteins necessary for HIV-1 infectivity. Two new host-cellular proteins were identified in our virus preparations, CD48 and histones, specifically histone H1, H2, H3, and H4 (Table 1B). While histones were identified directly in our experiments, CD48 required a secondary purification step by affinity chromatography. CD48 is a complementary determinant protein found on the surface of immune derived cells and has not been identified previously as being associated with or co-purifying with HIV-1 virus. The presence of this protein in our experiments was demonstrated by both LCMS/MS (Table 2) and by western blot with anti-CD48 antibodies (Figure 4). While the mass spectral data does suggest the presence of CD48, only a single peptide was identified which absolutely requires confirmation by traditional biochemistry methods. There are several relevant points to consider in Figure 4. First, the bands present in Figure 4 for the cell lysates demonstrate the extraordinarily high concentration of CD48 found in Jurkat cells as compared to 293T cells. CD48 is expressed by a subset of cells of the immune system, and it considered a specific marker for this cell type. Jurkat cells are derived from immune cells, whereas 293T cells are not. Thus, it is expected that Jurkat expression of CD48 will be much higher than 293T cells. This disparity of expression forced longer exposures to show CD48 from the 293T cells than would have been necessary for Jurkat cell lysate. The higher molecular weight bands in the Jurkat viral pellet in Figure 4 (Western blot of CD48) reflect (1) the very high levels of expression of CD48 in these cells compared to 293T cells, (2) the rarer multimeric form CD48 which occurs in the presence of CD248,49 and (3) the a long exposure revealing some cross-reacting proteins. As might be expected, higher quantity of CD48 expressed in Jurkat cells results in greater incorporation of CD48 in HIV-1 derived from Jurkat cells compared to viruses derived from 293T cells. Our observation of histones as associated with HIV-1 virus must be approached with caution. We emphasize that the experimental association between HIV-1 and histones is consistent because control experiments without virus demonstrate no histones (data not shown). Indeed, Figure 2 demonstrates that the identified peptides for histone 2A are found across much of the defined sequence (amino acids 21-29, 42-71, and 100-110). However, RNA/histone complexes derived from virally infected cells are known to co-sediment with HIV-1 through sucrose gradients.30 Therefore, the critical importance of histones as incorporated proteins in HIV-1 must be confirmed by other biochemical methods and functional analysis rather than relied upon based solely on our data. Journal of Proteome Research • Vol. 5, No. 3, 2006 535

research articles It must be emphasized that despite the considerably wide proteomic “snapshot” of HIV-1 proteins afforded by LC-MS/ MS, not all HIV-1 proteins were identified in any single experiment. Most notably, Vpr was identified with only a single peptide and Nef and gp120 in only some experiments. As stated in the Introduction, expansion of the experimental design to include reduction and alkylation steps and multiple fractionation methods would likely improve the data quantity and quality concerning these low abundance protein identifications. Nevertheless, it must be emphasized that our findings directly reflect the known relative abundance of these proteins in mature HIV-1. The primary consideration for identification is the abundance of particular proteins in a given sample. As is evident in Table 2, higher abundance proteins were identified by multiple peptides for each protein, while lower abundance proteins were identified by fewer peptides. Very low abundance proteins were identified by only a single peptide (Vpr) or in some experiments and not others (Nef). In the case of gp120, although it is critical to HIV-1 infectivity, quantitative studies demonstrate that it is actually not abundant in mature HIV-1, both because of low incorporation50 and shedding of gp120 proteins from the viral surface.52 Mature HIV-1 virions are estimated to have 7 to 10 gp120 envelope spikes (21 to 30 molecules gp120) per mature virion.50,51 By comparison there are 1200 to 2500 Gag molecules per mature virion.50 The known molar ratio of HIV-1 capsid to gp120 is 60:1. Furthermore, gp120 is a heavily glycoslylated, conferring considerable protection from the exhaustive proteolytic degradation necessary for LC-MS/MS analysis.53 Thus the low levels of gp120 we observe directly reflect the known biology of HIV-1. All experiments described above queried the acquired LCMS/MS data (∼5000 MS/MS scans) against the NCBI-nr database (>500 000 protein sequences) with the idea that using one of the largest current databases would increase the probability of obtaining spurious positive protein identifications. A report on precisely this problem has been recently published.54 In addition, we used marginally lower statistical filtering criteria, which should have also exacerbated false positive identifications. Despite this, the vast majority of proteins identified in our experiments were HIV-1 and human proteins that were known to be incorporated into HIV-1. Most of the few questionable proteins observed were not from biologically relevant species and did not have the appropriate statistical information to confirm their identification. In addition, careful analysis of the experimental data and that represented in Table 2 indicates that most of the identified peptides were above the default criteria for both Xcorr and ∆Cn. Our marginally lower statistical thresholds were likely unnecessary for identification of lower abundance proteins. However, it is critical that novel proteins identified in LC-MS/MS experiments must be confirmed by other biochemical methods.

Conclusions In summary, this study demonstrates the applicability of LCMS/MS techniques to the proteomic analysis of enveloped viruses using HIV-1 as a model system. We identified almost all of the known viral, and known incorporated host-cell derived proteins of HIV-1. In addition, we show that LC-MS/ MS, in combination with affinity chromatography, is suitable for the identification of viral membrane-associated proteins. Highlighting the efficacy of this approach to identify novel HIV1-associated proteins, we demonstrate for the first time that the cell surface protein CD48 is incorporated into HIV-1 via 536

Journal of Proteome Research • Vol. 5, No. 3, 2006

Saphire et al.

LC-MS/MS, and verify this finding by Western blot. Most importantly, we demonstrate the specificity of this methodology by comparing wild-type viruses with mutants known to fail to incorporate a single host protein, Cyp A. This is the first time LC-MS/MS has been unequivocally demonstrated to have the capacity to distinguish the presence or absence of a single protein from a complex mixture of retroviral proteins. These results validate LC-MS/MS as a tool for analysis of viral proteomes and provide a framework for extending this technology to study the infection processes of this and other enveloped viruses. We anticipate that LC-MS/MS proteomic methods will contribute significant new understanding to the role host-cellular proteins play in the viral lifecycle.

Acknowledgment. This study was supported by The Center for Protein Sciences and the Department of Immunology at The Scripps Research Institute and instrumentation grant 1S10RR13896-01A1 from the National Institutes of Health. Part of this work was supported by an amfAR scholar award to A.S. S.J.B. would like to acknowledge Hayes McDonald and David Tabb for helpful discussions. References (1) Vazquez, N.; Greenwell-Wild, T.; Marinos, N. J.; Swaim, W. D.; Nares, S.; Ott, D. E.; Schubert, U.; Henklein, P.; Orenstein, J. M.; Sporn, M. B.; Wahl, S. M. Human immunodeficiency virus type 1-induced macrophage gene expression includes the p21 gene, a target for viral regulation. J. Virol. 2005, 79 (7), 4479-4491. (2) Florens, L.; Washburn, M. P.; Raine, J. D.; Anthony, R. M.; Grainger, M.; Haynes, J. D.; Moch, J. K.; Muster, N.; Sacci, J. B.; Tabb, D. L.; Witney, A. A.; Wolters, D.; Wu, Y.; Gardner, M. J.; Holder, A. A.; Sinden, R. E.; Yates, J. R.; Carucci, D. J. A proteomic view of the Plasmodium falciparum life cycle. Nature 2002, 419 (6906), 520-526. (3) Chelius, D.; Huhmer, A. F.; Shieh, C. H.; Lehmberg, E.; Traina, J. A.; Slattery, T. K.; Pungor, E., Jr. Analysis of the adenovirus type 5 proteome by liquid chromatography and tandem mass spectrometry methods. J. Proteome Res. 2002, 1 (6), 501-513. (4) Thali, M.; Bukovsky, A.; Kondo, E.; Rosenwirth, B.; Walsh, C. T.; Sodroski, J.; Gottlinger, H. G. Functional association of cyclophilin A with HIV-1 virions. Nature 1994, 372 (6504), 363-365. (5) Lodish, H. F.; Porter, M. Specific incorporation of host cell surface proteins into budding vesicular stomatitis virus particles. Cell 1980, 19 (1), 161-169. (6) Grundy, J. E.; McKeating, J. A.; Griffiths, P. D. Cytomegalovirus strain AD169 binds beta 2 microglobulin in vitro after release from cells. J. Gen. Virol. 1987, 68 (Pt 3), 777-784. (7) Braaten, D.; Ansari, H.; Luban, J. The hydrophobic pocket of cyclophilin is the binding site for the human immunodeficiency virus type 1 Gag polyprotein. J. Virol. 1997, 71 (3), 2107-2113. (8) Cimarelli, A.; Luban, J. Translation elongation factor 1-alpha interacts specifically with the human immunodeficiency virus type 1 Gag polyprotein. J. Virol. 1999, 73 (7), 5388-5401. (9) Gurer, C.; Cimarelli, A.; Luban, J. Specific incorporation of heat shock protein 70 family members into primate lentiviral virions. J. Virol. 2002, 76 (9), 4666-4670. (10) Grundy, J. E.; McKeating, J. A.; Ward, P. J.; Sanderson, A. R.; Griffiths, P. D. Beta 2 microglobulin enhances the infectivity of cytomegalovirus and when bound to the virus enables class I HLA molecules to be used as a virus receptor. J. Gen. Virol. 1987, 68 (Pt 3), 793-803. (11) McKeating, J. A.; Griffiths, P. D.; Grundy, J. E. Cytomegalovirus in urine specimens has host beta 2 microglobulin bound to the viral envelope: a mechanism of evading the host immune response? J. Gen. Virol. 1987, 68 (Pt 3), 785-792. (12) Braaten, D.; Franke, E. K.; Luban, J. Cyclophilin A is required for the replication of group M human immunodeficiency virus type 1 (HIV-1) and simian immunodeficiency virus SIV(CPZ)GAB but not group O HIV-1 or other primate immunodeficiency viruses. J. Virol. 1996, 70 (7), 4220-4227. (13) Braaten, D.; Franke, E. K.; Luban, J. Cyclophilin A is required for an early step in the life cycle of human immunodeficiency virus type 1 before the initiation of reverse transcription. J. Virol. 1996, 70 (6), 3551-3560.

Proteomic LC-MS/MS Analysis of HIV (14) Braaten, D.; Luban, J. Cyclophilin A regulates HIV-1 infectivity, as demonstrated by gene targeting in human T cells. Embo. J. 2001, 20 (6), 1300-1309. (15) Gallay, P.; Hope, T.; Chin, D.; Trono, D. HIV-1 infection of nondividing cells through the recognition of integrase by the importin/karyopherin pathway. Proc. Natl. Acad. Sci. USA 1997, 94 (18), 9825-9830. (16) von Schwedler, U.; Song, J.; Aiken, C.; Trono, D. Vif is crucial for human immunodeficiency virus type 1 proviral DNA synthesis in infected cells. J. Virol. 1993, 67 (8), 4945-4955. (17) Luban, J.; Bossolt, K. L.; Franke, E. K.; Kalpana, G. V.; Goff, S. P. Human immunodeficiency virus type 1 Gag protein binds to cyclophilins A and B. Cell 1993, 73 (6), 1067-1078. (18) Eng, J. K.; McCormack, A. L.; Yates, J. R., III. J. Am. Soc. Mass. Spectrom. 1994, 5, 976-989. (19) Misumi, S.; Fuchigami, T.; Takamune, N.; Takahashi, I.; Takama, M.; Shoji, S. Three isoforms of cyclophilin A associated with human immunodeficiencyvirus type 1 were found by proteomics by using two-dimensional gel electrophoresis and matrix-assisted laser desorption ionization mass spectrometry. J. Virol. 2002, 76 (19), 10000-10008. (20) Saphire, A. C.; Bobardt, M. D.; Gallay, P. A. Host cyclophilin A mediates HIV-1 attachment to target cells via heparans. Embo. J. 1999, 18 (23), 6771-6785. (21) Karczewski, M. K.; Strebel, K. Cytoskeleton association and virion incorporation of the human immunodeficiency virus type 1 Vif protein. J. Virol. 1996, 70 (1), 494-507. (22) Camaur, D.; Trono, D. Characterization of human immunodeficiency virus type 1 Vif particle incorporation. J. Virol. 1996, 70 (9), 6106-6111. (23) Sova, P.; Volsky, D. J.; Wang, L.; Chao, W. Vif is largely absent from human immunodeficiency virus type 1 mature virions and associates mainly with viral particles containing unprocessed gag. J. Virol. 2001, 75 (12), 5504-5517. (24) Simon, J. H.; Miller, D. L.; Fouchier, R. A.; Malim, M. H. Virion incorporation of human immunodeficiency virus type-1 Vif is determined by intracellular expression level and may not be necessary for function. Virology 1998, 248 (2), 182-187. (25) Trauger, S. A.; Wu, E.; Bark, S. J.; Nemerow, G. R.; Siuzdak, G. The identification of an adenovirus receptor by using affinity capture and mass spectrometry. Chembiochem. 2004, 5 (8), 10951099. (26) Franke, E. K.; Yuan, H. E.; Luban, J. Specific incorporation of cyclophilin A into HIV-1 virions. Nature 1994, 372 (6504), 359362. (27) Ott, D. E.; Coren, L. V.; Johnson, D. G.; Kane, B. P.; Sowder, R. C., 2nd; Kim, Y. D.; Fisher, R. J.; Zhou, X. Z.; Lu, K. P.; Henderson, L. E. Actin-binding cellular proteins inside human immunodeficiency virus type 1. Virology 2000, 266 (1), 42-51. (28) Ott, D. E.; Coren, L. V.; Copeland, T. D.; Kane, B. P.; Johnson, D. G.; Sowder, R. C., 2nd; Yoshinaka, Y.; Oroszlan, S.; Arthur, L. O.; Henderson, L. E. Ubiquitin is covalently attached to the p6Gag proteins of human immunodeficiency virus type 1 and simian immunodeficiency virus and to the p12Gag protein of Moloney murine leukemia virus. J. Virol. 1998, 72 (4), 2962-2968. (29) Gurer, C.; Berthoux, L.; Luban, J. Covalent modification of human immunodeficiency virus type 1 p6 by SUMO-1. J. Virol. 2005, 79 (2), 910-917. (30) Karageorgos, L.; Li, P.; Burrell, C. Characterization of HIV replication complexes early after cell-to-cell infection. AIDS Res. Hum. Retroviruses 1993, 9 (9), 817-823. (31) Olinger, G. G.; Saifuddin, M.; Hart, M. L.; Spear, G. T. Cellular factors influence the binding of HIV type 1 to cells. AIDS Res. Hum. Retroviruses 2002, 18 (4), 259-267. (32) Martin, G.; Tremblay, M. J. HLA-DR, ICAM-1, CD40, CD40L, and CD86 are incorporated to a similar degree into clinical human immunodeficiency virus type 1 variants expanded in natural reservoirs such as peripheral blood mononuclear cells and human lymphoid tissue cultured ex vivo. Clin. Immunol. 2004, 111 (3), 275-285. (33) Bounou, S.; Leclerc, J. E.; Tremblay, M. J. Presence of host ICAM-1 in laboratory and clinical strains of human immunodeficiency virus type 1 increases virus infectivity and CD4(+)-T-cell depletion in human lymphoid tissue, a major site of replication in vivo. J. Virol. 2002, 76 (3), 1004-1014. (34) Tardif, M. R.; Tremblay, M. J. Presence of host ICAM-1 in human immunodeficiency virus type 1 virions increases productive infection of CD4+ T lymphocytes by favoring cytosolic delivery of viral material. J. Virol. 2003, 77 (22), 12299-12309.

research articles (35) Saphire, A. C.; Bobardt, M. D.; Gallay, P. A. Cyclophilin a plays distinct roles in human immunodeficiency virus type 1 entry and postentry events, as revealed by spinoculation. J. Virol. 2002, 76 (9), 4671-4677. (36) Mondor, I.; Ugolini, S.; Sattentau, Q. J. Human immunodeficiency virus type 1 attachment to HeLa CD4 cells is CD4 independent and gp120 dependent and requires cell surface heparans. J. Virol. 1998, 72 (5), 3623-3634. (37) Moulard, M.; Lortat-Jacob, H.; Mondor, I.; Roca, G.; Wyatt, R.; Sodroski, J.; Zhao, L.; Olson, W.; Kwong, P. D.; Sattentau, Q. J. Selective interactions of polyanions with basic surfaces on human immunodeficiency virus type 1 gp120. J. Virol. 2000, 74 (4), 19481960. (38) Brown, M. H.; Preston, S.; Barclay, A. N. A sensitive assay for detecting low-affinity interactions at the cell surface reveals no additional ligands for the adhesion pair rat CD2 and CD48. Eur. J. Immunol. 1995, 25 (12), 3222-3228. (39) Popik, W.; Alce, T. M.; Au, W. C. Human immunodeficiency virus type 1 uses lipid raft-colocalized CD4 and chemokine receptors for productive entry into CD4(+) T cells. J. Virol. 2002, 76 (10), 4709-4722. (40) Ianelli, C. J.; DeLellis, R.; Thorley-Lawson, D. A. CD48 binds to heparan sulfate on the surface of epithelial cells. J. Biol. Chem. 1998, 273 (36), 23367-23375. (41) Braaten, D.; Aberham, C.; Franke, E. K.; Yin, L.; Phares, W.; Luban, J. Cyclosporine A-resistant human immunodeficiency virus type 1 mutants demonstrate that Gag encodes the functional target of cyclophilin A. J. Virol. 1996, 70 (8), 5170-5176. (42) Saphire, A. C.; Bobardt, M. D.; Gallay, P. A. trans-Complementation rescue of cyclophilin A-deficient viruses reveals that the requirement for cyclophilin A in human immunodeficiency virus type 1 replication is independent of its isomerase activity. J. Virol. 2002, 76 (5), 2255-2262. (43) Rosenwirth, B.; Billich, A.; Datema, R.; Donatsch, P.; Hammerschmid, F.; Harrison, R.; Hiestand, P.; Jaksche, H.; Mayer, P.; Peichl, P.; et al. Inhibition of human immunodeficiency virus type 1 replication by SDZ NIM 811, a nonimmunosuppressive cyclosporine analogue. Antimicrob Agents Chemother. 1994, 38 (8), 1763-1772. (44) Billich, A.; Fricker, G.; Muller, I.; Donatsch, P.; Ettmayer, P.; Gstach, H.; Lehr, P.; Peichl, P.; Scholz, D.; Rosenwirth, B. SDZ PRI 053, an orally bioavailable human immunodeficiency virus type 1 proteinase inhibitor containing the 2-aminobenzylstatine moiety. Antimicrob. Agents Chemother. 1995, 39 (7), 1406-1413. (45) Franke, E. K.; Luban, J. Inhibition of HIV-1 replication by cyclosporine A or related compounds correlates with the ability to disrupt the Gag-cyclophilin A interaction. Virology 1996, 222 (1), 279-282. (46) Steinkasserer, A.; Harrison, R.; Billich, A.; Hammerschmid, F.; Werner, G.; Wolff, B.; Peichl, P.; Palfi, G.; Schnitzel, W.; Mlynar, E.; et al. Mode of action of SDZ NIM 811, a nonimmunosuppressive cyclosporin A analogue with activity against human immunodeficiency virus type 1 (HIV-1): interference with early and late events in HIV-1 replication. J. Virol. 1995, 69 (2), 814824. (47) Mlynar, E.; Bevec, D.; Billich, A.; Rosenwirth, B.; Steinkasserer, A. The nonimmunosuppressive cyclosporin A analogue SDZ NIM 811 inhibits cyclophilin A incorporation into virions and virus replication in human immunodeficiency virus type 1-infected primary and growth-arrested T cells. J. Gen. Virol. 1997, 78 (Pt 4), 825-835. (48) van der Merwe, P. A.; Brown, M. H.; Davis, S. J.; Barclay, A. N. Affinity and kinetic analysis of the interaction of the cell adhesion molecules rat CD2 and CD48. Embo. J. 1993, 12 (13), 49454954. (49) van der Merwe, P. A.; McNamee, P. N.; Davies, E. A.; Barclay, A. N.; Davis, S. J. Topology of the CD2-CD48 cell-adhesion molecule complex: implications for antigen recognition by T cells. Curr. Biol. 1995, 5 (1), 74-84. (50) Chertova, E.; Bess, J. W., Jr.; Crise, B. J.; Sowder, I. R.; Schaden, T. M.; Hilburn, J. M.; Hoxie, J. A.; Benveniste, R. E.; Lifson, J. D.; Henderson, L. E.; Arthur, L. O. Envelope glycoprotein incorporation, not shedding of surface envelope glycoprotein (gp120/SU), is the primary determinant of SU content of purified human immunodeficiency virus type 1 and simian immunodeficiency virus. J. Virol. 2002, 76 (11), 5315-5325. (51) Layne, S. P.; Merges, M. J.; Dembo, M.; Spouge, J. L.; Conley, S. R.; Moore, J. P.; Raina, J. L.; Renz, H.; Gelderblom, H. R.; Nara, P. L. Factors underlying spontaneous inactivation and susceptibility

Journal of Proteome Research • Vol. 5, No. 3, 2006 537

research articles to neutralization of human immunodeficiency virus. Virology 1992, 189 (2), 695-714. (52) Moore, J. P.; McKeating, J. A.; Weiss, R. A.; Sattentau, Q. J. Dissociation of gp120 from HIV-1 virions induced by soluble CD4. Science 1990, 250, (4984), 1139-1142. (53) Zhu, X.; Borchers, C.; Bienstock, R. J.; Tomer, K. B. Mass spectrometric characterization of the glycosylation pattern of

538

Journal of Proteome Research • Vol. 5, No. 3, 2006

Saphire et al. HIV-gp120 expressed in CHO cells. Biochemistry 2000, 39 (37), 11194-11204. (54) Cargile, B. J.; Bundy, J. L.; Stephenson, J. L., Jr. Potential for false positive identifications from large databases through tandem mass spectrometry. J. Proteome Res. 2004, 3 (5), 1082-1085.

PR050276B