stage MS and tandem MS (MS/MS) proteomic analyses. The technology enables broad protein identification from nanogram-size proteomics samples and ...
Anal Bioanal Chem (2004) 378 : 1037–1045 DOI 10.1007/s00216-003-2329-8
O R I G I N A L PA P E R
Y. Shen · N. Tolić · C. Masselon · L. Paša-Tolić · D. G. Camp II · M. S. Lipton · G. A. Anderson · R. D. Smith
Nanoscale proteomics
Received: 31 July 2003 / Revised: 29 September 2003 / Accepted: 7 October 2003 / Published online: 29 November 2003 © Springer-Verlag 2003
Abstract Efforts to develop a liquid chromatography (LC)/mass spectrometry (MS) technology for ultra-sensitive proteomics studies (i.e., nanoscale proteomics) are described. The approach combines high-efficiency nanoscale LC (separation peak capacity of ≈103; 15-µm-i.d. packed capillaries with flow rates of 20 nL min–1, the optimal separation linear velocity) with advanced MS, including high-sensitivity and high-resolution Fourier transform ion cyclotron resonance MS, to perform both singlestage MS and tandem MS (MS/MS) proteomic analyses. The technology enables broad protein identification from nanogram-size proteomics samples and allows the characterization of more abundant proteins from sub-picogramsize samples. Protein identification in such studies using MS is demonstrated from 1-microgram) samples. Additional technical details related to the developments described here are given in several recent and future publications.
Experimental Nanoscale proteomics using LC/MS The nanoscale proteomics studies were performed using an LC/MS system described elsewhere [18]. Briefly, this system includes a 4 cm×50-µm i.d.×5 µm C18 microSPE pre-column for sample concentration and an 85 cm×15-µm i.d.×3 µm C18 packed capillary column for high efficiency gradient reversed-phase nanoscale LC separation of extremely small samples. The micro-SPE stage allows solution to be loaded onto the nanoLC column at approximately 8 µL min–1 which requires 10-fold variation in ESI-MS signal levels generally observed for different peptides from the same protein (which may reflect differences in proteolysis, selective peptide losses, as well as possible variations in ionization and detection effi-
1041 Fig. 2 Nanoscale protein identifications from a 50-pg D. radiodurans sample using the method described in Fig. 1. Experimental conditions are given in the text. Annotations for proteins are referenced by an assigned peptide (tXXX represents the XXXth tryptic peptide counting from the amino terminus, while XXX. XXX represents the starting and ending positions of the non-tryptic digestion cleavage) [13]
Fig. 3 Protein identification dynamic range from nanoscale proteomics. The 5-ng D. radiodurans sample contained 75 fmol cytochrome C and 75 zmol BSA tryptic digests [18]
1042
Fig. 4 Nanoproteomics coverage. For FTICR MS results, proteins identified were detected using 15N/14N-labeled peptide pairs; for ion trap MS/MS only 14N peptides were assigned by searching against the database. The number of ORFs identified is given for each sample size analyzed by the two instrumental approaches; for larger samples only a subset of the ORFs are indicated
ciency in the MS). Figure 3 shows the range of relative protein abundances detectable for a 5-ng D. radiodurans tryptic digest sample spiked with tryptic digests of the standard proteins cytochrome c and BSA [22], which have a 106-fold difference in relative abundances. The peptides in the 5-ng sample provided a similar intensity range to that obtained for the added standard proteins with a 106 dynamic range in protein content, allowing the proteome measurement dynamic range to be estimated. These results indicate that the protein content in a single mammalian cell (≈50 pg) should allows proteins varying by 103 to 104 in relative abundance to be studied. The comprehensiveness of proteome measurements depends on a combination of factors: the measurement dynamic range, the sample amount, LC separation quality, the LC flow rate and the mass spectrometric performance. Figure 4 shows the number of proteins identified from various D. radiodurans sample sizes using FTICR MS
and ion trap MS/MS. FTICR single-stage MS allowed identification of 428 different ORFs (i.e., 13.7% coverage of the 3,116 predicted ORFs) from a 10-ng sample, compared with 141 ORFs (i.e., 4.5% coverage) from a single 3-h LC-ion trap MS/MS analysis. One observation from these studies is that the number of proteins that could be characterized was significantly different for MS and MS/MS methods. Additionally, the ORFs detected with the greatest sensitivity varied for the two approaches, a result primarily attributed to the different MS/MS efficiencies and dissociation patterns for peptides. However, the overlap for identified proteins was typically approximately 75% for both methods. The proteins not common to both the MS and MS/MS analyses were generally identified through assignment of a single peptide. Consistently, these single peptides were typically detected with low FTICR MS intensities or with ion trap MS/MS peptide identification confidence scores close to the acceptable limits. Protein identification throughput and accurate quantitation When a proteomic separation is sufficiently fast, the proteome measurement throughput is controlled by the MS
1043
Fig. 5 Protein measurement throughput using nanoscale singlestage MS proteomics. The spectrum was obtained at one point in the gradient reversed-phase capillary LC/FTICR MS analysis of a 2.5-ng D. radiodurans tryptic digest sample
data acquisition rate. Single-stage MS has the capability of being able to detect >103 molecular species per spectrum (each scan typically requires 1–7 s depending upon experimental details) [10]; however, only a very small portion of this capability is effectively utilized to identify low-abundance species in typical proteomics analyses. Figure 5 shows the peptide/protein identification throughput in a single MS scan (spectrum) for a D. radiodurans proteome sample. In this experiment fourteen proteins per scan (or ≈2 proteins s–1) were identified, and the non-uniform distributions averaged to an overall proteome measurement throughput of approximately 150 protein pairs h–1 and about 3 h sample–1 for automated operation and protein identification. (Higher rates are achieved when larger samples result in detection of more peptides.) While far short of the theoretical maximum rate, this corresponds to a much higher throughput protein identification compared with conventional proteomics analyses. The corresponding LC/MS/MS nanoscale proteomics throughput is approximately 50 proteins h–1, a level primarily dictated by its lower sensitivity and slower data acquisition. Multi-
plexed MS/MS using LC/TOF (or FTICR) MS, where multiple peptides are simultaneously dissociated and assigned in single scans [23, 24], offers an avenue for improving throughput. Ultimately, we expect that there will be significant improvements to both approaches, but note that single-stage MS analyses will inevitably retain significant advantages for throughput, sensitivity, and (less directly) quantitation in comparison to approaches using MS/MS. Most proteomics applications carry the need for accurate or precise protein quantitation. Figure 6 shows an example of relative protein quantitation based upon 14N/15Nstable isotope labeling [25] of a D. radiodurans sample. The protein relative abundance ratio of 1:1.13 was obtained by summing ion intensities from the set of individual MS spectra obtained during peptide elution, since slight differences in elution for the different labeled species could give rise to significant errors for quantitation based on any one spectrum. An advantage of very low flow rates from very small i.d. capillary columns is the much more uniform ionization efficiencies and the minimization (or even elimination) of the ionization suppression (and matrix) effects [17, 18] that also can contribute significant error. While operation in the nanoflow regime, as described here, addresses many ESI-MS-related issues, one has to consider that many other issues will impact relative abundances (e.g., selective sample losses in processing and incomplete proteolysis). However, with such issues in mind,
1044
Fig. 6 An example of relative abundance protein quantitation [14] using the LC with single-stage FTICR MS analysis approach. A 0.5-ng D. radiodurans tryptic digest sample was used and accurate quantitation involved summing ions detected in the set of spectra in which the peptide is detected
study of, for example, micro-dissected tissues. To fully exploit these new capabilities, more effective sample processing is needed for comparably small sample sizes and is likely to be realized by using microfabricated devices.
it appears that the basis now exists for much improved absolute quantitation.
Acknowledgments We thank the US Department of Energy’s Office of Biological and Environmental Research and the National Cancer Institute (Grant CA 86340) for their support of portions of this research. Pacific Northwest National Laboratory is operated by the Battelle Memorial Institute for the US Department of Energy through Contract DE-ACO6–76RLO 1830.
Conclusions References The new LC/MS technology described here provides a basis for broad protein identification from nanogram-quantity proteomics samples (i.e., nanoscale proteomics), and has been shown to enable identifications of the higher abundance proteins from samples as small as 0.5 pg. The protein content of a single eukaryotic cell (≈50 pg) is thus sufficient for protein identification within a dynamic range of >103. The use of high-efficiency LC/single-stage MS analyses simultaneously enables broad protein identification coverage and high throughput. Thus the nanoscale proteomics technology provides a new tool for studying proteome heterogeneity in tissues that extends to the single cell level, thereby opening avenues for quantitative
1. Devine KM, Wolfe K (1995) Trends Genet 11:429; http://www. ebi.ac.uk/research/cgg/genomes.html; http://www.tigr.org 2. Pandey A, Mann M (2000) Nature 405:837 3. Heppner GH, Miller FR (1998) Int Rev Cytol 177:1 4. Emmert-Buck MR et al (1996) Science 274:998; Bonner RF et al (1997) Science 278:1481 5. Sgroi DC et al (1999) Cancer Res 59:5656; Roth MJ et al (2001) Cancer Res 61:4098 6. Shen Y et al (2001) Anal Chem 73:1766; 3011 7. Washburn MP et al (2001) Nat Biotechnol 19:242; Wolters DA et al (2001) Anal Chem 73:5683 8. Fann JB et al (1989) Science 246:64 9. Marshall AG et al (1998) Mass Spectrom Rev 17:1; Martin SE et al (2000) Anal Chem 72:4266; Quenzer EL et al (2001) Anal Chem 73:1721; Belov et al (2003) Anal Chem 75:4195
1045 10. Anderson GA et al (1998) Proceedings of the 46th ASMS conference on mass spectrometry and allied topics, p 1270; Bruce JE et al (1999) Anal Chem 71:2595; Goodlett DR et al (2000) Anal Chem 72:1112; Harkewicz R et al (2002) J Am Soc Mass Spectrom 13:144 11. Bruins AP, Covey TR, Henion JD (1987) Anal Chem 59:2642 12. Shen Y et al (2002) Anal Chem 74:4235 13. Martin SE, Shabanowitz J, Hunt DF, Marto JA (2000) Anal Chem 72:4266 14. Quenzer TL et al (2001) Anal Chem 73:1721 15. Haskins WE et al (2001) Anal Chem 73:5005 16. Chen J et al (2003) Anal Chem 75:3145
17. Shen Y et al (2003) Anal Chem 75:3596 18. Shen Y et al (2003) Anal Chem (in press) 19. Anderson GA, Bruce JE (eds) (1995) ICR-2LS, Tolić N (ed) (2001) LaV2DG. Pacific Northwest National Laboratory, Richland, WA 20. ftp://ftp.tigr.org/pub/data/d_radiodurans/GDR.pep. Cited 31 July 2003 21. Lipton MS et al (2002) Proc Natl Acad Sci USA 99:11049 22. Shen Y et al (2003) Anal Chem (submitted) 23. Masselon CD et al (2000) Anal Chem 72:1918 24. Li L et al (2001) Anal Chem 73:3312 25. Oda Y et al (1999) Proc Natl Acad Sci USA 96:6591
