• Keine Ergebnisse gefunden

3 Materials and methods

4.4 Protein assembly time line for spliceosomes by relative quantification

5.3.1. Methodical considerations

This work is the first proteome study that compares directly iTRAQ and SILAC for relative quantification of proteins. We performed relative quantification of proteins derived from spliceosomal B and C complexes and found that iTRAQ and SILAC yielded similar result.

However, when comparing the obtained protein ratios, iTRAQ showed in general slightly lower values for proteins that are enriched in the B complex and slightly higher values for proteins that are enriched in the C complex. This might be due to the fact that the precursor selection for MS/MS is not 100 % selective and co-eluting peptides thus might contribute to the iTRAQ reporter ion intensity as it has been recently discussed (Bantscheff et al., 2007).

Therefore, SILAC protein ratios show extremely high values or low values for proteins that are enriched or underrepresented, respectively, in one of the samples when compared to iTRAQ.

Protein ratios obtained from the SILAC experiments showed a lower standard deviation per se as compared to iTRAQ. Indeed, metabolic labeling should be a more reliable method for labeling proteins as compared to chemical labeling, as it should guarantee a 100 % labeling efficiency of proteins and thus of the peptides. Moreover, samples can be pooled at an earlier stage and therefore variation in the quantification ratios by e.g. sample losses, can be

neglected. Nonetheless, depending on the cellular system, sometimes complete labeling cannot or can only hardly be achieved by SILAC. In particular for those cells grown in culture that do not rapidly divide (e.g. embryonic stem cells; Graumann et al., 2008) or only proliferate in cell culture without dividing (e.g. primary neurons; Liao et al., 2008). Normally in such cells, a labeling efficiency of maximal 80 % is achieved. However, such incomplete labeling can be handled by novel computational methods (Liao et al., 2008).

Of course, chemical labeling always involves the risk of incomplete labeling. In our initial experiments a low labeling efficiency (< 80 %) was found to drastically influence the quantification results. Therefore, the labeling efficiency is one of the major issues for obtaining accurate quantification when using chemical labeling approaches and the initial experiments were performed to achieve the highest possible labeling efficiency.

Nonetheless, the great advantage of all chemical labeling approaches is that proteins from almost every source (cells, tissue, body fluids etc.) can be quantified.

An alternative chemical labeling approach to iTRAQ labeling has recently been introduced by Boersema et al., 2008. Dimethylation of the peptides’ N-termini and lysine side chains using different stable isotope labeled reagents allows relative quantification of three samples in one MS analysis. Dimethyl labeling is based on a simple chemical reaction without any observed byproducts and provides a 100 % labeling efficiency in almost all cases. In addition, it uses inexpensive reagents and is thus a cost-effective labeling technique in comparison to other stable isotope reagents. However, we have not performed this particular labeling strategy.

Among the different chemical labeling strategies, isobaric reagents, such as iTRAQ reagents or TMTs (Thompson et al., 2003), have the advantage that quantification is performed during MS/MS analysis so that sample complexity is not enhanced, as the differently labeled peptides show the same mass in the MS. As the reporter ions, which are used for quantification of the different samples, are released during MS/MS, the analytical depth of the analysis is higher as compared to analysis of differently labeled samples that show peak pairs in the MS (e.g. SILAC or dimethyl labeling).

Until now, isolated spliceosomal complexes have been compared by peptide count after LC-MS/MS analysis (Behzadnia et al., 2007; Bessonov et al., 2008; Deckert et al., 2006; Kuhn et al., 2009). The correlation between the relative protein abundances and the number of acquired tandem MS spectra (spectral count), the number of identified peptides (peptide count) and the obtained sequence coverage has recently been compared by Liu et al., 2004 in a study with defined standard proteins. A linear correlation was found between the relative protein amount and the number of acquired MS/MS spectra (spectral count), but not between the relative protein amount and the number of identified peptides (peptide count) or the

sequence coverage. In addition to iTRAQ and SILAC, we therefore evaluated spectral count for relative quantification of spliceosomal B and C complexes. Spectral count is a simple quantification technique that enables the comparison of almost every sample from a proteome study without additional sample preparation. As it does not require labeling of the peptides or proteins, no expensive labeling reagents are needed and as many samples as desired can be quantified relative to each other. However, it requires highly reproducible LC-MSMS analyses (see also below).

In our study, we found a good overall agreement between spectral count and quantification with iTRAQ and SILAC. In all three quantitative analyses, several proteins were clearly identified to be more abundant in spliceosomal B or C complexes and some proteins were found to be present in equal amounts within both complexes (see Tables 4.9 – 4.11). As discussed above, iTRAQ and SILAC yielded consistent results for relative quantification of B and C complexes. Strikingly, for most of the quantified proteins, the same quantitative trend, i.e. enrichment in one of the complexes or same abundance in both complexes, was also obtained by spectral count. Importantly, spectral count exhibits some discrepancies compared to iTRAQ or SILAC. In particular for small proteins (< 20 kDa), accurate quantification could not at all or only roughly be achieved by spectral count, as only a limited number of peptides were generated. For example, the relative quantification of the Sm proteins within both, the spliceosomal B and C complexes, using spectral count did not unambiguously reveal the expected ratios, i.e. a two-fold difference of Sm proteins in the B vs. C complex.

As the Sm proteins are common to all U snRNPs except for U6 snRNP, four copies of Sm proteins are expected in the B complex. Upon transition from the B to the C complex, only two copies of Sm proteins are left, because U1 and U4 snRNPs are destabilized/dissociated.

For all seven Sm proteins, protein ratios close to 2 are obtained by iTRAQ and SILAC, whereas spectral count yielded the correct value only for two of the Sm proteins (SmF and SmG). The other Sm proteins show protein ratios of approximately 1 by spectral count.

Another example for the lower accuracy of spectral count is the quantification of proteins that should be present in a 1:1 ratio or where iTRAQ and SILAC clearly showed such a ratio.

These are, for example, U5-220K, U5-40K, and CBP20 (for a complete list, see Tables 4.9 – 4.11). With spectral count, higher or lower protein ratios (approximately between 0.6 and 1.7) were obtained. These results are consistent with the observations of Liu et al., 2004 that small quantitative changes among proteins in different samples cannot be accurately monitored by spectral counting. The quantification of large proteins within the B and C complex by spectral count yielded results very similar to iTRAQ and SILAC (e.g. with the

SF3a and SF3b proteins), except when proteins are present in nearly equimolar amounts in both samples (e.g. U5-220K and U5-200K).

Another discrepancy between spectral count and labeling approaches is observed with the protein ratios obtained for proteins that are pre-dominantly associated with one of the two complexes. These, like the hPrp18, hPrp22, and DDX35 proteins, show extreme values after spectral count, thus suggesting their complete absence in one of the complexes. Such extreme protein ratios might be misleading regarding the presence or absence of proteins within the different samples. If a peptide is not selected for sequencing it does not necessarily mean that the peptide is not present in the sample. Such low abundance peptides, which might escape detection by spectral count, are still detectable by iTRAQ and SILAC quantification. During iTRAQ analysis, samples are pooled and the differently labeled peptides are isobaric and are thus selected for sequence analysis irrespective of whether the actual non-labeled peptides are of low or high abundance. Quantification is then performed on the MS/MS level, where even low abundance peptides produce the corresponding reporter ions. SILAC quantification is based on the correct assignment of the mass pairs that were generated upon the incorporation of stable isotopes. In this manner, the low abundance peptides – if present – will be recognized by the software (e.g. MSQuant, Schulze and Mann, 2004; or MaxQuant, Cox and Mann, 2008), through the assignment of the corresponding highly abundance peptide.

Spectral count has further limitations that are due to the following technical requirements: (i) A high reproducibility of the chromatography system is required to obtain comparable elution profiles of the peptides during sample separation. (ii) The spectral count response for every protein is not the same, i.e. due to the protein’s amino acid sequence and the different properties of the generated peptides (e.g. chromatographic behavior) the number of detectable spectra varies for every protein. As discussed above, smaller proteins generate only few peptides and relative quantification by spectral count is limited for these proteins.

(iii) Different co-eluting peptides in the respective samples can affect the acquisition of distinct MS/MS spectra and thus influence the quantification process. (iv) Dynamic exclusion of precursor masses - that is selection of a precursor that already has been selected for fragmentation before and is thus subsequently not selected again for fragmentation (within a certain time window) - is usually used during LC-MS/MS analyses. Although the analytical depth is enhanced by using dynamic exclusion, dynamic exclusion negatively influences accurate quantification by spectral count, because different peptides are always selected for MS/MS fragmentation. It is important to note that spectral count not only takes the number of unique peptides into account, but also the overall number of spectra, i.e. the same peptide is selected several times for fragmentation. Consequently, when working with dynamic

exclusion, the quantification of highly abundant peptides that show a longer retention time during their elution from the LC is not considered adequately. When working without dynamic exclusion, the spectral count is more reliable but, on the other hand, the analytic depth is drastically reduced and therefore only a limited number of proteins can be quantified.

The same holds true for SILAC and other labeling procedures that are based on peptide intensities on the MS level. Incorporation of stable isotopes (by metabolic or chemical labeling) generates peptides of different masses and the sample complexity is consequently increased. The analytical depth for the analysis of complex samples is thus reduced and only a limited number of proteins can be quantified.

Importantly, heavily modified proteins, protein isoforms and truncated proteins escape quantification. As the modified and the unmodified peptides show different masses, these proteins might yield false quantification values, in particular, when a protein becomes significantly modified during transition from the B to the C complex. Protein ratios of different protein isoforms within the quantified complexes might also not be correctly assigned and might thus affect the quantification for these proteins.

There are several reasons why (semi-quantitative) spectral count in our study, i.e. the comparison of highly purified spliceosomal complexes, is still applicable for relative comparison: (i) The analyzed spliceosomal complexes were highly purified under stringent conditions, thus minimizing the number of contaminating proteins during the analyses; (ii) the complexes are of moderate complexity and consist of only a limited number of proteins; (iii) The use of always the same LC system coupled front-end to the ESI Q-ToF mass spectrometer (Q-ToF Ultima, Waters) in all of the performed previous studies (Behzadnia et al., 2007; Bessonov et al., 2008; Deckert et al., 2006; Kuhn et al., 2009) fulfilled the prerequisite to generate comparable results and ensures high reproducibility such that analyses can be compared. However, when adapting the above described relative quantification approaches to biological systems other than spliceosomal complexes, the critical aspects discussed above have to be taken into account.