3.2. Protein Identifications in Skeletal Muscle by 1-D SDS-PAGE
We have previously used 1-D SDS-PAGE in conjunction with peptide mass fingerprinting and de novo
sequencing to identify the abundant proteins in carp skeletal muscle [26
]. Using this approach, it was only possible to confidently identify 24 proteins. Additional separation of the proteins by 2D-PAGE did not significantly increase the number of proteins identified. The protein composition was found to largely comprise creatine kinase isoforms and fragments and enzymes of the glycolytic pathway. In the current study, with a GeLC-MS/MS approach, the average number of identifications for each biological replicate was 120 ± 1 (Table 1
). Across the three biological replicates, a total of 182 carp skeletal muscle proteins were confidently assigned (Figure 2
and Supplemental Table S1
) indicating that increasing the number of replicates permits a more extensive coverage of the proteome.
Gene ontology (GO) analysis (Figure 3
and Supplemental Table S2
) revealed that proteins were predominantly metabolic (37%, GO Biological Process) followed by cellular process (23%) and developmental process (10%). The major metabolic functions of identified proteins were catalytic activity (44%), binding (25%), and structural molecule activity (20%). Proteins were largely intracellular and 35% were found to correspond to organelles. Almost 13% of identified proteins were found to be components of macromolecular complexes. There were a wide variety of protein classes identified (23 in total), which included cytoskeletal proteins (17%) and a range of enzymes including transferases (11%), hydrolases (11%), and oxidoreductases (10%). Enrichment analysis showed that glycolysis, TCA cycle, carbohydrate (including monosaccharide) metabolic processes, muscle contraction, and muscle organ development were all significantly enriched by more than five-fold compared to the complete Danio rerio
genome. Similarly, the molecular functions of actin binding, isomerase activity, cytoskeletal protein binding, and structural constituents of the cytoskeleton together with the cellular compartments related to the cytoskeleton and tubulin complexes were also enriched.
3.3. Proteome Equalization Increases the Number of Protein Identifications
The soluble muscle proteins were subjected to protein equalization and compared with unfractionated skeletal muscle. 1-D SDS-PAGE revealed that after treatment with the ProteoMiner system there was a substantial enrichment of the low-abundance proteins (Supplemental Figure S1
). This finding was reflected in the subsequent LC-MS/MS analysis, that led to a substantial increase in protein identifications (mean = 209 ± 15; Table 1
). The total number of proteins identified over the three biological replicates was 334 (Figure 2
and Supplemental Table 1
). A study by Rivers et al.
] achieved a similar number of protein identifications from chicken skeletal muscle (>350 unique proteins detected) using a protein equalization strategy. Whilst this approach was effective in reducing the most abundant proteins, the authors indicated that there was a different bias in the proteins detected, particularly when different ionic strength buffers were used during sample loading.
In the GO analysis of the ProteoMiner simplified protein fractions (Figure 3
and Supplemental Table S2
), the major molecular functions, biological process, and cellular compartment were similar to the 1-D SDS-PAGE analysis. However, there was a difference in the protein class distribution with nucleic acid binding proteins the predominant protein class (15%). Enrichment analysis of the proteins identified following ProteoMiner revealed that protein equalization covered a greater number of biological processes with the addition of protein complex assembly and biogenesis, protein folding, regulation of translation, and glycogen metabolic processes. However, muscle organ development and muscle contraction were significantly enriched to a lesser extent than in 1-D SDS-PAGE. Cellular localization enrichment revealed three components that were more than five-fold enriched. These were tubulin complex, ribosome, and cytosol. Nine specific pathways were significantly enriched including TCA cycle, glycolysis, cytoskeletal regulation by Rho GTPase, pyruvate metabolism, cell cycle, UPS, and both FGF and EGF signalling pathways.
3.4. Extending Proteome Coverage Through In-Solution Isoelectric Focusing
We also sought to pre-fractionate the protein complement of skeletal muscle through the use of OFFGEL. Separated protein fractions were recovered in the liquid phase and subjected to 1-D SDS-PAGE. The gel analysis displayed a clear increase in the number of observable proteins for each resolved fraction (Supplemental Figure S2
) and there was a near four-fold increase in the mean number of identifications to 425 ± 62 proteins (Table 1
). The total identified proteins across all biological replicates was 766 (Figure 2
and Supplemental Table S1
). This result demonstrates the resolving power of this fractionation strategy, that permits more extensive coverage of the skeletal muscle proteome although it extends sample preparation and requires additional LC-MS/MS run time.
GO analysis of the OFFGEL datasets indicated that the majority of proteins were involved in metabolic processes (30%) and cellular function (21%). Around 10% were involved in both developmental processes and biogenesis (Figure 3
and Supplemental Table S2
). Molecular functions and cellular component analysis revealed a similar profile to both 1-D SDS-PAGE and simplification by protein equalization. Twenty-six different protein classes were represented in the OFFGEL separated proteins, although, the major classes were similar to the previous analyses. Enrichment of biological processes revealed eight that were more than five-fold enriched, the majority of which overlapped with the ProteoMiner bead simplification of the proteome. The molecular function enrichment was however somewhat different and included amino acid kinase activity, phosphatase inhibitor activity, aminoacyl-tRNA ligase activity, translation elongation factor activity, phosphatase regulator activity, translation factor activity, nucleic acid binding, translation regulator activity, translation initiation factor activity and isomerase activity. Again, tubulin complex and ribosomal proteins were significantly enriched more than five-fold as were proteins located in the intermediate filament cytoskeleton. Pathway enrichment was similar to both 1-D SDS-PAGE and ProteoMiner simplification approaches, with the addition of the methylmalonyl pathway.
3.5. Improving the Confidence of Protein Identifications by FASP
We also evaluated the effect of pre-fractionation at the peptide level. In these experiments skeletal muscle homogenate was digested in solution through FASP. The resultant peptides were then separated by OFFGEL isoelectric focusing and the collected fractions were analysed by LC-MS/MS. Using this approach, the mean number of protein identifications was 118 ± 9 (Table 1
), with a total of 163 proteins confidently identified across the biological replicates (Figure 2
and Supplemental Table S1
). Analysis of the peptides derived from the FASP analysis alone (prior to OFFGEL separation) resulted in the identification of 75 ± 9 proteins (n
= 3 replicates), with a total of 91 proteins across the replicates. Although the FASP + OFFGEL and 1-D SDS-PAGE analyses resulted in a similar number of total identifications, there was an overlap of around only 50% (88 proteins) indicating that FASP analysis sampled a different sub-proteome of the skeletal muscle (Supplemental Figure S3
The GO analysis revealed that a comparable profile with respect to biological process and molecular function was observed (Figure 3
and Supplemental Table S2
). Whilst the cellular compartments of the proteins were broadly similar to the other simplification processes, the percentage distributions of the locations were different. Almost 35% of proteins were deemed to be cellular, with 30% attributed to organelles and over 12% to membranes. A further 10% of proteins were located to cell junctions. Twenty-three classes of proteins were identified and again cytoskeletal proteins were the major class although at an increased level of 22%. The next most abundant class was enzyme modulator (9%) and cell junction proteins (8%).
Thirty biological processes were found to be enriched with reference to the zebrafish genome, 18 of which were more than five-fold enriched. These were neuromuscular synaptic transmission, glycolysis, tricarboxylic acid cycle, sensory perception of sound, cytokinesis, muscle contraction, monosaccharide metabolic process, cellular component morphogenesis, generation of precursor metabolites and energy, muscle organ development, cellular component movement, anatomical structure morphogenesis, mitosis, mesoderm development, carbohydrate metabolic process, cellular component organization, cellular component organization or biogenesis, and sensory perception. Similarly, there were 12 biological processes and nine cellular components that were significantly enriched. There were however only four pathways found to be enriched-glycolysis, cytoskeletal regulation by Rho GTPase, nicotinic acetylcholine receptor signaling pathway, and the Wnt signaling pathway.
3.6. Comparison of Identification and Overlaps
The four different simplification strategies provided a different composition of identified proteins. Whilst there was some overlap between the different techniques in each approach, there were a number of proteins that were uniquely identified (Figure 4
When compared to 1-D SDS-PAGE analysis, simplification by proteome equalization technology yielded 159 proteins that were common to both analyses and 176 unique proteins. The OFFGEL analysis identified the largest number of unique proteins of any of the methodologies and identified all but 20 of those observed in the 1-D SDS-PAGE analysis alone. FASP yielded the lowest number of protein identifications. A FASP strategy together with OFFGEL fractionation has also recently been applied to the deep mining of murine skeletal muscle [27
] and single muscle fibre analysis [28
]. In our study the FASP analysis involved the use whole muscle homogenate rather than soluble muscle fraction and therefore would have sampled a different sub-proteome of the skeletal muscle. The lower number of protein identifications may result from losses during buffer exchange in the FASP protocol prior to OFFGEL analysis. In view of these results, this approach may need to be further refined for use with fish skeletal muscle. This could be achieved by employing different buffers and desalting methods. However, it should be noted that the mean number of peptides from FASP analysis was much higher than all other strategies at 20.5% compared to between 6% and 7% for 1-D SDS-PAGE, ProteoMiner, and OFFGEL. In total, 80 proteins were identified across all of the different strategies. GO analysis revealed this group was comprised of mostly catalytic activity and structural proteins that are typically present in high abundance in muscle tissue.
Enrichment analysis indicated clear differences in the classifications of the identified proteins. Eight biological processes (anatomical structure morphogenesis, carbohydrate metabolic process, generation of precursor metabolites and energy, glycolysis, metabolic process, monosaccharide metabolic process, muscle contraction, and tricarboxylic acid cycle) were enriched by all methods. Cell communication was only enriched following ProteoMiner simplification, whereas there were 11 categories specifically enriched following both OFFGEL separation and FASP. Only four molecular functions (actin binding, catalytic activity, structural constituent of cytoskeleton, and structural molecule activity) and five cellular compartments (actin cytoskeleton, cytoskeleton, intracellular, organelle, and tubulin complex) were enriched across all analyses. These results indicate that different approaches to sample pre-fractionation in proteomic experiments minimise bias and reinforce the need for a multi-layer approach to proteome characterization. Moreover, an analysis of each analytical approach showed that there was some variation in the proteins identified between technical replicates. This ranged from 46% overlap of identified proteins between OFFGEL experiments up to 66% for both 1D-SDS-PAGE and FASP (Table 2
). Further, each approach was assessed in terms of its experimental practicality. For a rapid screen of the most abundant proteins in skeletal muscle, 1D-SDS-PAGE would be appropriate, whereas ProteoMiner provides a robust analysis of lower abundance proteins. OFFGEL enables the most comprehensive profiling of the proteome but is both time and resource intensive and therefore may not be suitable for a large cohort study. Given the larger numbers of peptides identified per protein FASP might be best employed for quantitative studies.