Recent years have seen unprecedented progress in mass spectrometry-based proteomics 1. This has enabled development of various new methodologies for interrogating the proteome. These include assessment of relative protein expression 2 , detection of protein ligand interactions 3 , 4 , monitoring changes in the abundance of post-translational modifications 5 , protein half-life determinations 6 , 7 , 8 , 9 , and many others.
In order to continue improving proteome-wide characterization of proteostasis 6 , 7 , 10 , 11 , a further development of experimental and computational 12 , 13 quantitative mass spectrometry 14 work flows is required. For instance, when using dynamic SILAC stable isotope labeling by amino acids in cell culture to measure global protein turnover 6 , 15 , precise and accurate peptide ion intensity quantification is needed, since even small deviations in the accuracy of measured fold changes can have a pronounced effect on the half-life measurement.
In particular, when measuring protein turnover in non-dividing cells 16 , many proteins will exhibit very-slow turnover because the continuous replication of the entire proteome, which occurs in exponentially growing cells is not required.
As primary cells can only be kept in culture for a limited amount of time before adapting to the cell culture conditions or going into senescence, protein turnover determinations have to be based on relatively short-term treatments with stable isotope-encoded amino acids. Consequently, accurate and precise quantification is required in order to allow accurate determination of protein half-lives.
We, therefore, developed procedures based on a better utilization of the isotopic distributions of ionized peptides to improve the accuracy and precision of peptide ion intensity-based quantification. We applied this peptide ion intensity quantification strategy to analyze mass spectrometry data from dynamic SILAC experiments 17 performed in five different, non-dividing cell types: B-cells, monocytes, natural killer NK cells, hepatocytes, and mouse embryonic neurons to calculate protein half-lives as previously described 6.
We used this data set to validate and extend the previous observation 18 of coherent subunit turnover of protein complexes, but also observed complex architecture-dependent protein half-life distributions.
To demonstrate the usefulness of our data as a resource, we examined some exemplifying protein complexes in more detail. In agreement with previous literature 19 , 20 , we found that histone proteins, aside from some notable exceptions in hepatocytes, have extremely slow turnover. Both, proteasomes and nuclear pore complexes NPCs , show a clear subcomplex-dependent turnover of their subunits. The extreme longevity of the NPC previously reported in vivo for brain tissue 16 , is not observed for any of the cell types investigated in vitro in this study.
These results emphasize that slow NPC turnover is not a general phenomenon occurring in all non-dividing cells, but that specific NPC turnover mechanisms might exist. We conclude that our data set is a useful resource for the scientific community and our method can be broadly applied in the future. Protein half-life determination in non-dividing cells requires precise and accurate measurement of protein fold changes.
In non-dividing cells the incorporation of heavy isotope labels will be very slow for some proteins, resulting in very-low new-to-old protein ratios because only a very-small portion of the isotope has yet been incorporated. As a consequence, the ratio determination is error prone, particularly at the early time points. Such data might be stringently filtered to select for high-confidence measurements, but at the cost of coverage, specifically affecting long-lived proteins.
To achieve accurate protein half-life measurements with good coverage for long-lived proteins in primary cell systems, we investigated and optimized the parameters, which are relevant for determining very reproducible and accurate protein fold changes for the greatest possible number of proteins. We introduced two innovations into the data analysis workflow.
First, we use the exact elemental composition of the identified peptides for calculating the theoretical isotopic envelope see methods. We established that this method yields a more accurate representation of peptide ion intensities as compared to the previously used averagine model 21 Supplementary Figure 1 that relies on a scaled version of the average amino acid Averagine , but does not take into account that the composition of an individual peptide might well differ from the average.
Second, we used a measure to detect overlapping isotopic distributions of different peptides 22 , 23 see methods , to improve the quality of quantification Supplementary Figures 1 — 3. Comparison of this quantification strategy with the widely used MaxQuant software shows that our methodology performs significantly better for cases when very-low protein ratios need to be quantified, and more accurately determines the protein fold changes Supplementary Figure 4.
For proteins where the fold changes are less pronounced both softwares perform equally well. This quantification strategy is, thus, well suited for determining protein half-lives of long-lived proteins in non-dividing cells. Taken together across all five cell types, protein half-lives were determined for a total of unique protein groups Supplementary Data 2. The coverage in individual cell types ranged from protein groups identified in NK cells to protein groups identified in mouse neurons.
This criterion was fulfilled by proteins taken together across the five cell types. The mean half-life was calculated for proteins present in both biological replicates for each cell type. The comparison of protein half-lives between biological replicates Fig. The R 2 of the log 10 -transformed half-lives between replicates was 0. The acquired data set, thus, represents the most extensive, high-quality catalog of protein half-lives in primary cells. Experimental workflow for protein half-life determination using dynamic SILAC and quantitative mass spectrometry.
Five non-dividing primary cell types comprising B-cells, NK cells, monocytes, hepatocytes, and mouse embryonic neurons were adapted to light SILAC medium. To label newly synthesized proteins, the cells were exposed to heavy SILAC medium and collected at different time points. Peptides of pre-existing and newly synthesized proteins were distinguished by their mass due to incorporation of light or heavy arginine and lysine.
Protein fold changes at different time points were calculated using the intensity ratios of heavy vs. Assessment of protein half-lives in different cell types. On the diagonal the half-lives determined in biological replicate experiments in the same cell type are compared. Red dots indicate nuclear pore complex components.
The plots below the diagonal compare the average protein half-lives determined in one cell type against other cell types as indicated. The median protein fold changes are denoted by the slope of the line. The same is shown for three nuclear pore complex members NUPs as for b , c , d. The other cell types monocytes, hepatocytes, B-cells, and mouse embryonic neurons had, respectively, only 7, 17, 15, and 4 such proteins.
Despite having an overall slower turnover rate, the relative log 10 transformed half-lives between NK cells and B-cells, as well as NK cells and monocytes were in good agreement: R 2 of 0. The same holds true for monocytes and B-cells: R 2 of 0. In contrast, hepatocytes, which are not of hematopoietic lineage, showed the weakest correlation among the human cells, R 2 of 0. Half-lives determined in the mouse embryonic neurons agreed slightly better with B-cells, NK cells, and monocytes than with hepatocytes R 2 of 0.
We conclude that a general categorization of the proteome in terms of protein half-lives is to quite some extent preserved over all cell types investigated here.
All protein half-lives determined in this study are summarized in Supplementary Data 2. Amongst the fast turnover proteins, we find members of the Janus family of kinases. Our data is in good agreement with a previous study that used radioactive labeling of the long-lived cerebral histone fractions in mice and reported half-lives of 50— days We also assessed the turnover behavior within the different cellular compartments, Supplementary Figure 6.
Within each compartment a broad range of protein half-lives was observed and there is a large overlap in turnover behavior between all compartments.
However, significant differences in general turnover behavior are observed between the different compartments, although the effect size is small, Supplementary Figure 6.
The most prominent and significant trend is the slower turnover of the mitochondrial proteins, which is present in all cell types. Proteins from the Golgi apparatus and nucleus had reproducibly the fastest turnover, while endoplasmic reticulum and cytoplasmic proteins were located in the middle with close to identical turnover distributions.
In summary, we observe a wide spread turnover behavior within the different cell types, with particularly long half-lives observed for histones. We also observe small, but significant and reproducible differences for protein turnover in different cellular compartments.
Our extensive data set enabled us to assess the turnover of protein complexes on a proteome-wide scale. In an earlier protein turnover study spanning proteins using 15 N-pulsed labeling in mice and mass spectrometry, Price et al 18 observed a coherent turnover behavior of several protein complexes in mouse liver and brain. We performed a statistical analysis to evaluate if the observation of coherent protein complex turnover holds true when considering all complexes from an in-depth proteome turnover analysis across different non-dividing cell types.
We calculated the standard deviations of the half-life values of proteins that are subunits of the same annotated complex, and compared those to the standard deviations obtained for the same complexes after the subunits were reshuffled across all complexes preserving the number of proteins in each complex group see methods.
The chaperonin complex has the most tightly controlled turnover of the individual subunits in all different cell types. Looking at two larger complexes with more intricate architecture, the NPC and the 26S proteasome—we observe a much less tightly controlled turnover, except for the 26S proteasome in mouse neurons, where the turnover behavior is very coherent.
Half-life variability among members of protein complexes is smaller than expected by chance. Distributions of standard deviations SD of half-lives from proteins in complexes as annotated in the CORUM database compared to SD of the half-lives of the same proteins shuffled across the different complexes, while preserving the number of proteins in each complex group.
Differences in the log 10 half-lives of true protein complex members vs. Looking in more detail at the 26S proteasome Fig. Interestingly, there is a significant trend for the 20S core complex to be more stable than the 19S regulatory complex in B-cells, monocytes, and NK cells, but a clear and significant opposite behavior is observed for hepatocytes. Clustering of the proteasome subunits according to the similarity between their half-lives across the human cells also leads to distinct separation between the core and regulatory subunits Fig.
Architecture-dependent turnover of the proteasome subunits. For each cell type, half-lives were averaged over biological replicates, except for rare cases where only one half-life value was available, and converted to a color gradient as explained in the Methods. The median, minimum, and maximum half-lives are indicated together with the color bars.
Subunits with undetermined half-lives are colored green. Lower panel: distributions of the reproducibly measured half-lives of the regulatory and the non-exchangeable core subunits of the proteasome in the different cell types. Hierarchical clustering leads to separation of the regulatory subunits from the non-exchangeable core subunits.
Next, we looked at the NPC, one of the largest protein complexes in the cell. This complex has been studied in rat brain by Toyama 19 et al, also using an in vivo 15 N labeling strategy and nucleoporins were shown to be particularly long-lived in rat brain tissue The authors also observed that the half-life of nucleoporins correlates with their allocation in subcomplexes within the NPC. To the best of our knowledge, the data set by Toyama et al. In the non-dividing cells in our study, we observe a relatively quick turnover of Nups that is almost one order of magnitude faster as compared to histones Fig.
Generally, the half-lives of the Nups are approximately located in the middle of the distribution of all other protein half-lives of the same cell type. Mouse embryonic neurons are an exception where Nups are turning over slightly slower than most other proteins, but still much faster than histones Fig.
We do not observe pronounced differences between the inner ring and Y-complex members; e. In our comprehensive data set, we do, however, observe a general clustering of half-lives into known subcomplexes Fig. The half-lives of members of the Nup complex, and to some extent also the Nup complex, are generally shorter when compared to the inner ring and Y-complexes.
The half-lives of members of the Nup62 complex, although spatially positioned in the inner ring complex, appear to be uncoupled from the latter.
In hepatocytes and monocytes it is more short-lived but in B-cells more long-lived when compared to other inner ring Nups Fig. Interestingly, the turnover of Nup is in line with those of the Nup complex and Nup98, and an association of which has been proposed Architecture-dependent turnover of the nuclear pore subunits.
Nups are shown color-coded as gradient from red short half-life to blue long life-life. An architectural model of the nuclear pore 42 , 43 is shown as seen from top top panel , cut in half middle panel , and a subcomplex scheme bottom panel. The nucleoplasmic side is at the bottom in all cases.
For each cell type, half-lives were averaged over two biological replicates, except for rare cases where only one half-life value was available, and converted to a color gradient as explained in the Methods. The median and minimum and maximum half-lives are indicated together with the color bars. Nups with non-determined half-lives are colored green. Lower panel: distributions of the reproducibly measured half-lives of the scaffold and peripheral subunits of the nuclear pore in the different cell types.
Nups of the inner ring are colored blue, of the outer Y-complex rings—orange, trans-membrane nucleoporins—brown, Nup and Nup—green, nuclear basket nucleoporins—yellow, Nup62 subcomplex—magenta, Nup subcomplex—salmon, and Nup complex—red. Partitioning the nucleoporins into a scaffold and a peripheral group Supplementary Data 4 and comparing the half-life distributions between the two groups shows a statistically significant trend for faster turnover of the nucleoporins in the peripheral group for all cell types Fig.
In agreement with previous work 19 , we find that Nup98 turns over considerably more quickly than Nup96, although both proteins are synthesized as a single fusion protein prior to autoproteolytic cleavage. This might be explained by the existence of an additional transcript encoding only Nup Nup, Nup50, and the transmembrane Nup gp have been shown to have short mean residence times at the NPC 26 , although this of course does not mean that they turn over once they disassociate.
Interestingly, both Nup and Nup50 have relatively short half-lives, e. In striking contrast, gp generally lives at least as long as scaffold Nups, e. In order to accurately quantify protein turnover in non-dividing cells, we introduced and validated new concepts for peptide ion intensity-based quantification.
In total we determine half-lives for proteins taken together from five different non-dividing cell types; B-cells, monocytes, NK cells, hepatocytes, and mouse neurons and make these publicly available as a resource to the scientific community. In our extensive study of protein turnover in non-dividing cells, we observed very-long half-lives for histone proteins in accordance with previous in vivo work in non-dividing brain cells 16 on a relative scale. Similarly, the turnover of proteasomes observed in vivo 18 , is well in line with our data.
For NPCs, Toyama et al. This is not true for any of the non-dividing cell types investigated in in vitro in this study, but relatively short half-lives of Nups, both on an absolute scale and relative to histones are observed. The technical details and also the biological context of both experiments are very difficult to compare, and either of them have its own benefits.
The half-lives measured in non-dividing cells in vitro should yield more accurate values at the cost of losing the endogenous in vivo context, which, of course, is also important.
We can nevertheless conclude that the very-slow NPC turnover is not a general phenomenon of all non-dividing cells. This might have important biological implications. Although an NPC surveillance pathway for defective NPC assembly intermediates has been described for yeast 32 , the turnover of fully assembled NPCs is believed to be extremely rare 33 and a corresponding pathway has, to the best of our knowledge, not yet been identified.
One might thus speculate that NPC-specific turnover or maintenance pathways might exist that remain to be characterized. It will thus be interesting to look out for NPC turnover in non-dividing cells in the future. One would predict that both should occasionally occur. Hopper , , Springer Protocols. Gomes et al. References Hershko, A. Carafoli, E. Pillay, C. Johnston, M. Harlow, E. Patrick, G. Amon, A. Cell 77 , — Yaglom, J. Hershko, A. Carafoli, E. Pillay, C. Johnston, M.
PubMed Google Scholar. Harlow, E. Google Scholar. Patrick, G. Amon, A. Cell 77 , —
0コメント