High-Performance Workflow for Identifying Site-Specific Crosslinks Originating from a Genetically Incorporated, Photoreactive Amino Acid

Abstract

In conventional crosslinking mass spectrometry, proteins are crosslinked using a highly selective, bifunctional chemical reagent, which limits crosslinks to residues that are accessible and reactive to the reagent. Genetically incorporating a photoreactive amino acid offers two key advantages: any site can be targeted, including those that are inaccessible to conventional crosslinking reagents, and photoreactive amino acids can potentially react with a broad range of interaction partners. However, broad reactivity imposes additional challenges for crosslink identification. In this study, we incorporate benzoylphenylalanine (BPA), a photoreactive amino acid, at selected sites in an intrinsically disordered region of the human protein HSPB5. We report and characterize a workflow for identifying and visualizing residue-level interactions originating from BPA. We routinely identify 30 to 300 crosslinked peptide spectral matches with this workflow, which is up to ten times more than existing tools for residue-level BPA crosslink identification. Most identified crosslinks are assigned to a precision of one or two residues, which is supported by a high degree of overlap between replicate analyses. Based on these results, we anticipate that this workflow will support the more general use of genetically incorporated, photoreactive amino acids for characterizing the structures of proteins that have resisted high-resolution characterization.

Keywords: crosslinking; disorder; photo-crosslinking; protein structure; small heat shock proteins.

Figure 1.

The BPA-containing variant proteins were first purified without the use of antibodies or purification tags, as described previously. Then, samples are divided into two fractions prior to the crosslinking reaction to have a non-crosslinked control. The left illustrates the workflow for the analysis of the monomeric reactants to create the validated-protein database. Alternatively, a validated-protein database can be created using the LC-MS data from the dimeric products. The right describes the workflow for the analysis of the dimeric products, which was used to identify site-specific crosslinks.

Figure 2.

(A) The number of crosslink PSMs found and the corresponding search times are shown as a function of database size and using Kojak, StavroX, or MeroX. These searches were all performed on the same LC-MS data of trypsin-digested, dimeric products of W9B. StavroX identified 10 or fewer crosslink PSMs across database sizes. StavroX results are only reported for up to the 250-protein database because searches timed out for larger database sizes. MeroX identifies 32 crosslink PSMs across database sizes. MeroX results are only reported for up to the 500-protein database because the search timed out at larger database sizes. We used a 2-PSM minimum at a 1% FDR as the criteria for the validated-protein database. In this plot, the 12-protein database is the validated-protein database. (B) At the peptide level, this Venn Diagram shows that using StavroX or MeroX resulted in the identification of a small subset of the crosslinked peptides identified using Kojak. (C) At the residue level, this Venn diagram shows that using MeroX or StavroX resulted in the identification of a smaller number of residue-level crosslink sites than Kojak, many of which were assigned with greater precision using Kojak.

Figure 3.

The results of nine replicates of trypsin-digested, dimeric products of W9B at pH 6.5 (Figure S3 and Table S3) are compared here. The number of crosslink sites identified (defined as the number of residues assigned a frequency value greater than zero in Figure S3) across differing numbers of replicates is indicated. A value of nine indicates that a given crosslink was identified in all nine datasets, whereas a value of one indicates that a given crosslink was only identified in a single dataset. Of the 128 total crosslink sites identified 54 are identified in all nine replicates.

Figure 4.

The number of HSPB5-HSPB5 crosslinks, decoy crosslinks, and other crosslinks identified as a function of FDR. The search represented in this data uses the validated-protein database (12 proteins) on the same dataset of trypsin-digested, dimeric products of W9B represented in Figure 2.

Figure 5.

Above panels A and B, the BPA position and domain boundaries are shown for the structural regions of HSPB5: a disordered N-terminal region (NTR), an α-crystallin domain (ACD), and a disordered C-terminal region (CTR). All panels represent the same analysis of trypsin-digested W9B at pH 6.5, which is injection 3 in Figure S3 and Table S3. Panel A shows a peptide-level representation of the results, in which every residue in a crosslinked peptide received a frequency of one. Panel B shows a residue-level representation of the results, in which the PSM for the crosslinked peptide was distributed among the potential crosslinking sites to account for ambiguity as described in the text. Crosslink results are reported as both a histogram and a rolling average of three because of ambiguity in the crosslinking site. The frequency axis corresponds to the number of PSMs, and the relative population axis corresponds to the percent of total crosslink PSMs. Panel B has 293 crosslink PSMs. In Panel C, the number of possible crosslink sites (x-axis) indicates how many potential equivalent crosslinking sites a PSM has. Over 70% of the PSMs have no ambiguity in the crosslink site assignment, and over 80% of the crosslink PSMs have two possible crosslink sites or fewer.

Figure 6.

Above panels A and B, the top row shows the BPA position and domain boundaries. The following rows show the expected cleavage sites for trypsin or GluC. Panels A and B depict W9B samples with varying digestion enzymes. Panel A has 277 crosslink PSMs from trypsin-digested W9B at pH 6.5. Panel B has 195 crosslink PSMs from the trypsin-GluC-digested W9B at pH 6.5. The large tryptic peptide spans from sites 23 to 56, and we observe crosslinks to that region when using a trypsin-GluC, parallel digestion. The raw file used for the analysis in panel B has been reported previously. In panel C, the Venn diagram illustrates the overlap in crosslink sites identified in trypsin-digested W9B at pH 6.5 (panel A) and trypsin-GluC-digested W9B at pH 6.5 (panel B).

Figure 7.

Panels A and B depict trypsin-GluC-digested samples with differing sites of BPA incorporation. Panel A has 66 crosslink PSMs from F24B. Panel B has 62 crosslink PSMs from L33B. The Venn diagram (panel C) illustrates the overlap in crosslink sites identified in F24B (panel A) and L33B (panel B). About half of the crosslink sites identified are found in both samples. The raw files used for this analysis have been reported previously.