Application of the peptide-focused approach on cross-linked bacteria

3.3.1 Performance comparison of the conventional and peptide-focused approach

The peptide-focused database search approach was benchmarked on purified cross-linked complexes and it was shown that the results obtained with the approach are similar to the conventional approach.

Next, the approach was applied on cross-linked bacteria to test its performance on more complex samples. Bacteria have the largest proteomes that current algorithms for non-cleavable cross-linkers can evaluate in a reasonable time frame^{41, 140}.

Bacillus subtilis and cereus were grown in LB medium and cross-linked in vivo during exponential growth phase without prior cell disruption. After XL-MS analysis, a conventional and a peptide-focused database search approach was applied in comparison to each other. Escherichia coli was also employed for in vivo cross-linking, but the migration pattern of DSS and DSP cross-linked proteins differed in SDS-PAGE (data not shown). It was hypothesized that the outer membrane of gram-negative bacteria interferes with cross-linker diffusion into the cell, which was also described in the past¹⁴¹.

Table 7 summarizes the identifications obtained with both approaches. The proteomes of the two species were in silico digested and peptides with a lysine within the sequence besides the C-terminus were considered. These peptides constitute the cross-linking database search space of theoretically linkable peptides. Bacillus subtilis and cereus contain 202,109 and 246,013 theoretically cross-linkable peptides. The generated peptide databases were much smaller in comparison but are based on experimental data (table 7). Nonetheless, peptide databases covered 60 to 73 % of the reference proteome with at least one peptide per protein, which approximately corresponded to the expressed proteome for bacteria during exponential growth¹⁴². Proportional to the ratio of theoretically cross-linkable peptides, the database search was five to ten times faster than the conventional approach.

The number of identified CSMs, cross-linked residues and inter-protein cross-linked residues were similarly high for Bacillus subtilis for both approaches. However, the peptide-focused approach outperformed the conventional approach for Bacillus cereus. Approximately 9,000 more CSMs, 200 more unique cross-linked residues and 62 more inter-protein cross-linked residues were identified with the peptide-focused approach. Bacillus cereus has approximately 1,000 protein sequences more than Bacillus subtilis. Hence, approximately 50,000 additional theoretically cross-linkable peptides need to be considered for the database search space.

By narrowing down the search space to peptides that were proven cross-link candidates in a parallel experiment, it was possible to obtain similar or more identified CSMs and cross-linked residues for complex samples like bacteria in a much shorter analysis time.

3.3 Application of the peptide-focused approach on cross-linked bacteria

Table 7: Comparison of a conventional and peptide-focused database approach applied on Bacillus subtilis and cereus. Cells were processed as described in chapter 2.2.3. Data was searched with pLink 1 against the reference proteome (conventional approach) and against a peptide database generated in parallel.

Bacillus subtilis (4,260 proteins) Conventional approach Peptide-focused approach

Theoretically cross-linkable peptides 202,109 29,689

Proteins covered in database 4,260 3,115

Identified CSMs 12,786 14,187

Identified unique cross-linked residues 1,197 1,167

Inter-protein cross-linked residues 271 262

Relative average search time 100 % 18.18 %

Bacillus cereus (5,240 proteins)

Relative average search time 100 % 10.77 %

3.3.2 Ribosomal interaction surfaces of EF-Tu

In vivo cross-linking of bacteria as a benchmarking experiment has shown that the peptide-focused approach yielded similar or more identified cross-linked residues than the conventional approach.

Cross-linked residues identified with the peptide-focused approach were subsequently manually inspected without a fixed CSM or pLink score cut-off. Inter-protein cross-links were evaluated whether an identification was described in the past or made sense in terms of biological function. A map of protein interactions based on cross-links in exponentially growing bacilli was obtained (supplemental figure 2). Most cross-links involved the ribosome and RNA polymerase. Their dominance makes sense, since biomass production is a focus of bacteria during exponential growth phase. After all, these complexes belong to the most abundant proteins in the cell. Cross-links within multimeric complexes of the tricarboxylic acid cycle like pyruvate dehydrogenase and succinate-CoA ligase, and the aspartyl/glutamyl-tRNA amidotransferase were identified in both species (supplemental figure 2).

In addition, numerous hitherto unknown cross-linked protein interactions were identified (supplementary table 1). Exemplarily, cross-links obtained for elongation factor Tu (EF-Tu) to the ribosome were highlighted (figure 5). A model structure from E. coli is shown in figure 5, because there is no structure available for Bacilli that includes EF-Tu. The positioning of ribosomal subunits in the ribosome is identical for Bacillus subtilis, cereus and Escherichia coli. EF-Tu served as a central interaction node in the interaction map of both bacilli species. Cross-links to ribosomal proteins RL14, RL17, and RL19 were either in cross-link distance or were located very close to EF-Tu bound to the tRNA in the aminoacyl site. Therefore, they could have been cross-linked together when EF-Tu was approaching the ribosome. Surprisingly, several cross-links of EF-Tu to ribosomal proteins RS5, RS8, RS9, and RS10 of the small subunit were identified, which mostly cluster on one surface (figure 5, central right view of the ribosome). These cross-links are not explainable by the function of EF-Tu delivering aminoacyl-tRNAs to the A-site. Some cross-links to e.g. RS9 and RS10 could be disregarded as a singular linked residue, but especially RS5 and RS8 exhibited three and two unique linked residues, respectively, and links to RS8 and RS9 were identified in both species. The

cross-Figure 5: Interaction surfaces of EF-Tu on the ribosome. Bacteria were treated as described in chapter 2.2.3. Cross-links from EF-Tu (red circle and cartoon) to the ribosome (pdb: 5UYK, from E. coli)¹⁴³ are shown as interaction maps in the bottom left and right corner. Those ribosomal subunits that cross-linked to EF-Tu are colored blue in the crystal structure. The number of