Enabling Photoactivated Cross-Linking Mass Spectrometric Analysis of Protein Complexes by Novel MS-Cleavable Cross-Linkers

Cross-linking mass spectrometry (XL-MS) is a powerful tool for studying protein–protein interactions and elucidating architectures of protein complexes. While residue-specific XL-MS studies have been very successful, accessibility of interaction regions nontargetable by specific chemistries remain difficult. Photochemistry has shown great potential in capturing those regions because of nonspecific reactivity, but low yields and high complexities of photocross-linked products have hindered their identification, limiting current studies predominantly to single proteins. Here, we describe the development of three novel MS-cleavable heterobifunctional cross-linkers, namely SDASO (Succinimidyl diazirine sulfoxide), to enable fast and accurate identification of photocross-linked peptides by MSn. The MSn-based workflow allowed SDASO XL-MS analysis of the yeast 26S proteasome, demonstrating the feasibility of photocross-linking of large protein complexes for the first time. Comparative analyses have revealed that SDASO cross-linking is robust and captures interactions complementary to residue-specific reagents, providing the foundation for future applications of photocross-linking in complex XL-MS studies.

Cross-linking mass spectrometry (XL-MS) is a powerful tool for studying protein-protein interactions and elucidating architectures of protein complexes. While residuespecific XL-MS studies have been very successful, accessibility of interaction regions nontargetable by specific chemistries remain difficult. Photochemistry has shown great potential in capturing those regions because of nonspecific reactivity, but low yields and high complexities of photocross-linked products have hindered their identification, limiting current studies predominantly to single proteins. Here, we describe the development of three novel MS-cleavable heterobifunctional crosslinkers, namely SDASO (Succinimidyl diazirine sulfoxide), to enable fast and accurate identification of photocrosslinked peptides by MS n . The MS n -based workflow allowed SDASO XL-MS analysis of the yeast 26S proteasome, demonstrating the feasibility of photocross-linking of large protein complexes for the first time. Comparative analyses have revealed that SDASO cross-linking is robust and captures interactions complementary to residue-specific reagents, providing the foundation for future applications of photocross-linking in complex XL-MS studies.
Protein-protein interactions (PPIs) are fundamental to the assembly, structure, and function of protein complexes, which in turn exert control over a diverse array of biological processes integral to cell biology. Cross-linking mass spectrometry (XL-MS) is a unique structural tool capable of studying PPIs because of its ability to simultaneously capture and identify PPIs with interaction contacts from native cellular environments (1)(2)(3)(4)(5). In addition, the residue-specific cross-linkable distances defined by cross-linkers can function as restraints to assist structural modeling and to elucidate architectures of large protein complexes (6)(7)(8). To date, amine-reactive homobifunctional NHS ester cross-linkers have been the most popular reagents in XL-MS studies. This is because of the relatively high occurrence of lysines-particularly at the surfaces of protein structures-as well as the specificity and efficiency of amine-reactive chemistries. Although effective, these reagents alone cannot yield complete PPI maps, as profiling of interaction regions lacking lysines would be difficult. Thus, to complement lysine-reactive reagents, additional amino acid-specific cross-linkers have been developed, including carboxyl-residue (9)(10)(11), sulfhydryl-residue (12,13), arginine-residue (14), and multiresidue targeting ones (15)(16)(17), clearly expanding PPI coverage. In addition, integration of multiple cross-linkers has improved characterization of PPIs and increased the depth and accuracy of structural analysis (7,8,18,19), demonstrating the benefits of multichemistrybased combinatory XL-MS approaches. However, despite these successes, mapping interaction regions lacking targetable residues by specific chemistry remains challenging.
In recent years, photochemistry has shown great potential in capturing regions inaccessible to residue-specific crosslinkers because of its nonspecific reactivity (2,3,20,21). Various types of photoreactive reagents have been explored in XL-MS studies (13,(22)(23)(24)(25)(26)(27)(28)(29)(30), almost all of which have been heterobifunctional cross-linkers with an amine-reactive specific end and a nonspecific end. Among the commonly used photoreactive groups, alkyl diazirine is most attractive because of its small size, long excitation wavelength, photostability, reactivity, and proven success in XL-MS studies (22,(24)(25)(26)(27)(28)(29)(30). Diazirines are activated by UV light to yield highly reactive carbenes, which then react with an X-H bond (X: C, N, O, S) of any proximal amino acids (24,25,27,(29)(30)(31). While promising, the indiscriminate nature of photocross-linking often results in highly complex and low abundance crosslinked products that complicate MS analysis and database searching, thus limiting its application predominantly to single proteins (24)(25)(26)(27)(28)30). Therefore, to advance photoreactive XL-MS studies for complex PPI mapping, it is essential to develop novel reagents that permit effective MS detection and accurate identification of photocross-linked peptides. MS-cleavable cross-linking reagents have significantly facilitated MS analysis of cross-linked peptides in complex mixtures, because of their unique capability of eliminating the "n-square" problem and permitting effective sequencing of cross-linked peptides (2,32). To enable robust MScleavability, we have previously developed a series of sulfoxide-containing MS-cleavable cross-linking reagents (e.g., disuccinimidyl sulfoxide [DSSO]) (Fig. 1A) (10,12,(33)(34)(35)(36). The MS-labile C-S bonds adjacent to the sulfoxide can be preferentially fragmented before peptide backbone cleavage upon collision-induced dissociation (CID), physically separating the two cross-linked peptide constituents for individual sequencing. Notably, this predictable fragmentation occurs independent of cross-linking chemistry, peptide charge, and peptide sequence. These unique characteristics allow straightforward and unambiguous identification of crosslinked peptides by MS n analysis coupled with conventional database searching tools. Sulfoxide-containing MS-cleavable cross-linkers have been successfully applied to not only study PPIs in vitro (33,(37)(38)(39) and in vivo (34,39) but also to dissect structural dynamics of protein complexes (8,40,41). Thus, to expedite the identification of photocross-linked peptides, we have developed three sulfoxide-containing MS-cleavable heterobifunctional NHS-diazirine cross-linkers with varied lengths, namely, SDASO (Succinimidyl diazirine sulfoxide)-L (long), -M (medium) and -S (short). These SDASO reagents represent the first class of sulfoxide-containing MS-cleavable heterobifunctional photoreactive cross-linkers. To illustrate their capabilities, we have characterized SDASO cross-linkers with a standard protein bovine serum albumin (BSA) and applied them to map PPIs of affinity purified yeast 26S proteasome. Our results demonstrate that MS-cleavability enables accurate identification of photocross-linked peptides and that the SDASO-based XL-MS workflow is well-suited for probing PPIs in complex samples. In addition, comparison with residue-specific XL-MS data has determined that SDASO cross-linking is robust and captures PPIs complementary to existing reagents.

Synthesis and Characterization of SDASO Cross-Linkers
Three SDASO cross-linkers were designed, synthesized, and analyzed in this work (Fig. 1), including SDASO-L, SDASO-M, and SDASO-S. Their synthesis and characterization are described in supplemental Fig. S1 and supplemental Methods.

XL-MS Analysis of BSA and 26S Proteasome
Protein cross-linking was performed similarly to previous studies with some modifications (12,25). Briefly, for SDASO cross-linking of BSA, 50 μl of 50 μM protein solution in PBS buffer (pH 7.4) was reacted in triplicate with SDASO-L, SDASO-M, or SDASO-S in molar ratio of 1:50, respectively, for 1 h at 25 • C in the dark. The NHS reactive ends were quenched with the addition of ammonium bicarbonate at a 50-fold excess for 10 min at 25 • C in the dark. Then NHS ester labeled proteins were transferred into Millipore Microcon Ultracel PL-30 (30-kDa filters) and washed three times with 300 μl PBS buffer. Diazirine cross-linking was activated by UV irradiation, which was carried out on ice~5 cm from the light source in an UV light chamber (Analytikjena UVP Cross-linker CL-1000L) and irradiated at 365 nm for 30 min.
The affinity purified yeast 26S proteasome was (supplemental Methods) cross-linked by SDASO linkers similarly as described above. To determine the optimal SDASO cross-linking conditions, we have performed initial XL-MS experiments of the yeast 26S proteasome using 5, 10, 20, and 40 mM SDASO, respectively. As a result, 20 mM SDASO yielded the highest number of cross-link identifications and was determined as the optimal cross-linking condition for this work. Specifically, 100 μg of the 26S proteasome in PBS buffer (pH 7.4) was cross-linked in triplicate with 20 mM SDASO-L, SDASO-M, and SDASO-S, respectively. In addition, 100 μg of the yeast 26S proteasome in PBS buffer (pH 7.4) was cross-linked with 2.5 mM or 5 mM DSSO for 1 h at 25 • C temp similarly as described (8), and the reactions were quenched with the addition of ammonium bicarbonate at a 50-fold excess for 10 min. Then cross-linked proteins were transferred into Millipore Microcon Ultracel PL-30 (30-kDa filters) for digestion.

Digestion of Cross-Linked Proteins
The resulting cross-linked products were subjected to enzymatic digestion using a FASP protocol (43). Briefly, cross-linked proteins on FASP filters were reduced/alkylated and digested with Lys-C/trypsin or chymotrypsin as described (8,33). The resulting digests were desalted, and cross-linked peptides were enriched by size-exclusion chromatography before LC MS n analysis (10,44).

Experimental Design and Statistical Rationale
Three SDASO cross-linkers were designed and characterized in this work with a standard protein BSA and an affinity purified yeast proteasome. Each SDASO XL-MS experiment was performed in biological triplicate under optimized conditions. To evaluate the effect of enzymatic digestion on SDASO results, chrymotrypsin digestion was performed for SDASO-L cross-linked 26S proteasome in biological triplicate as well. In total, nine SDASO XL-MS experiments for BSA analysis and 12 SDASO XL-MS experiments for the yeast 26S proteasome. Cross-validation was carried out among the results obtained from the three SDASO linkers. To further evaluate the SDASO results, DSSO XL-MS experiments of the yeast 26S proteasome were performed in two biological replicates. Reproducibility in XL-MS experiments was assessed at the level of cross-linked peptide sequences and sites, respectively.

LC-MS n Analysis and Identification of Cross-Linked Peptides
Cross-linked peptides were analyzed by LC-MS n using a Thermo Scientific Dionex UltiMate 3000 system online coupled with an Orbitrap Fusion Lumos mass spectrometer (8). A 50 cm × 75 μm Acclaim PepMap C18 column was used to separate peptides over a gradient of 1% to 25% ACN in 106 min for BSA and in 166 min for the 26S proteasome at a flow rate of 300 nl/min. MS 1 scans (375-1500 m/z, resolution at 120,000) were performed with the AGC target set to 4e5 in top speed mode with a cycle time of 5 s. For MS n analysis, 3+ and up charged ions were selected for MS 2 -CID in FT mode, followed by top four data-dependent MS 3 acquisition method (45). A targeted MS 3 acquisition was also used for DSSO cross-linked peptides by utilizing the mass difference between alkene-and thiol-modified ion pairs (31.9721 Da) (45). For MS 2 scans, the resolution was set to 30,000, the AGC target 5e4, the precursor isolation width was 1.6 m/z, and the maximum injection time was 100 ms for CID. The CID-MS 2 normalized collision energy was 25%. For MS 3 scans, CID was used with a collision energy of 35%, the AGC target was set to 2e4, and the maximum injection time was set to 120 ms.

Identification of Cross-Linked Peptides
MS n data were extracted using MSConvert (ProteoWizard 3.0.10738) and analyzed similarly as previously described (8). Briefly, the extracted MS 3 data were subjected to a developmental version of Protein Prospector (v.6.0.0) for database searching, using Batch-Tag against a custom random concatenated database (a total of 988 entries) derived from BSA and 493 Saccharomyces cerevisiae protein sequences that were identified from the affinity purified yeast 26S proteasomes. The mass tolerances for parent ions and fragment ions set as ±20 ppm and 0.6 Da, respectively. Trypsin or chymotrypsin was set as the enzyme with three or four maximum missed cleavages allowed, respectively. A maximum of four variable modifications were allowed, including cysteine carbamidomethylation, protein N-terminal acetylation, methionine oxidation, and N-terminal conversion of glutamine to pyroglutamic acid. In addition, three defined modifications representing alkene on uncleaved lysines, thiol and sulfenic fragment moieties on any amino acid (AAs) were selected for each respective SDASO cross-linker. Specifically, for SDASO-L cross-links: alkene (C 3 H 2 O; +54 Da), sulfenic acid (C 7 H 13 NO 2 S; +175 Da), and thiol (C 7 H 11 NOS; +157 Da). For SDASO-M cross-links: alkene (C 3 H 2 O; +54 Da), sulfenic acid (C 5 H 10 OS; +118 Da), and thiol (C 5 H 8 S; +100 Da). For SDASO-S cross-links: alkene (C 3 H 2 O; +54 Da), sulfenic acid (C 4 H 8 OS; +104 Da), and thiol (C 4 H 6 S; +86 Da). For DSSO cross-links, three defined modifications on uncleaved lysines are: alkene (C 3 H 2 O; +54 Da), sulfenic acid (C 3 H 4 O 2 S; +104 Da), and thiol (C 3 H 2 SO; +86 Da) (33). Owing to the conversion of the SDASO sulfenic acid moiety to the thiol moiety alongside backbone fragmentation during MS 3 analysis, we have incorporated such neutral loss in Batchtag to facilitate the identification of sulfenic acid-modified peptides during database searching using Protein Prospector. The in-house program xl-Tools was used to validate and summarize cross-linked peptides based on MS n data and database searching (33,39). To ensure the confidence in cross-link identification, we examined whether peptide sequences with ambiguous diazirine labeling sites have been identified repeatedly and found that the majority of those were verified by redundant identifications of same peptide sequences but different site localizations. Owing to the labeling capability of diazirine, we cannot exclude the possibility of the ambiguous sites being targeted. Further manual inspection was performed to examine peptide identification and site localization. Following integration of MS n data, there were no decoy hits found in the final lists of identified cross-linked peptides for all XL-MS experiments except for the tryptic digests of SDASO-L cross-linked 26S proteasome with a FDR ≤0.08%. To ensure the reliability of the identified cross-links, crossvalidation was performed among the three biological replicates for each linker and across the three SDASO reagents reported here.

Analysis of Amino Acid Preference for Diazirine Labeling
The unique K-X linkages identified for both BSA (supplemental Table S1B) and 26S (supplemental Table S2B) were used to assess diazirine labeling frequency at specific amino acids, in which only the peptide constituents labeled by diazirine were used for evaluation. The weighted occurrence values of diazirine-labeled AAs were determined based on their localization precision similarly as described (31). Briefly, for a given cross-linked peptide identified with n possible ambiguous sites, the weighted score Wx of each site x is determined as ax,r, which is the preference of reagent r toward residue at site x. So, Wx = ax,r/n. Assuming the preference for any AA in a given peptide is equal to 1, then Wx = 1/n. For all cross-linked peptides identified from the three biological replicates for each SDASO linker, the total weighted score for a given site x was calculated as Wx Wx, in which m is the total number of x in the identified cross-linked peptides. Then, the likelihood of carbene insertion at any site x was calculated as: Wi ′ (sum of all weighted scores for every x sites).

Distribution of Random Cross-Links
XWalk (46) was utilized to generate random cross-link distribution. Alpha carbon distances from lysine residues to all other residues (X) were generated individually using Euclidean distances only, skipping solventpath-distance calculations. The maximum distance was set to 100 Å for BSA and 300 Å for 26S proteasome to capture all possible residue linkage combinations in each protein/protein complex. Individual data for all residue combinations were compiled to generate histograms corresponding to random distributions for BSA, 26S, 20S, and 19S, respectively.

Designs of MS-Cleavable NHS-Diazirine Heterobifunctional Cross-Linkers
To advance photoreactive cross-linkers for complex PPI mapping, we sought to create novel sulfoxide-containing MScleavable NHS-diazirine heterobifunctional cross-linking reagents to cross-link lysines to any nearby AAs. It is noted that all of our previous sulfoxide-containing MS-cleavable crosslinkers are homobifunctional and carry two symmetric MScleavable C-S bonds adjacent to the central sulfoxide ( Fig. 1, A and E) (10,12,(33)(34)(35). Owing to the structural differences in reactive groups and their targeted residues, this symmetry is not retained in heterobifunctional cross-linkers. Recently, we have explored effects of spacer arm structures on MS-cleavability of sulfoxide-containing cross-linkers and identified an asymmetric spacer arm structure (47) that maintains the characteristic and predictable fragmentation expected of symmetric sulfoxide-containing MS-cleavable cross-linkers (10,12,(33)(34)(35)(36). This unique asymmetric spacer arm region carries a sulfoxide group that divides the spacer arm into two halves, i.e., a fixed half identical to DSSO with the sulfoxide and carbonyl group separated by '3' bond lengths, and a flexible half. Based on this design, we constructed three MS-cleavable heterobifunctional SDASO cross-linkers composed of a fixed NHS ester end and a flexible diazirine side with varying lengths from the center sulfoxide (i.e., long, 12.5 Å; medium, 10.2 Å; short, 7.7 Å), well within the distance range suited for studying PPIs (2) (Fig. 1, B-D). The synthesis routes and chemical analyses of SDASOs were detailed here (supplemental Fig. S1 and supplemental Methods).

Fragmentation Characteristics of SDASO Cross-Linked Peptides
Based on our recent studies on asymmetric sulfoxidecontaining cross-linkers (47), only the C-S bond at the NHS ester end in SDASO should be preferentially cleaved during CID. Thus, a single pair of MS 2 fragment ions is expected for all three SDASO cross-linkers (Fig. 1F). For an SDASO interlinked peptide (α-β), cleavage during CID physically separates the two crosslinked constituents and thus leads to the detection of two characteristic fragment ions (α A /β S ) carrying remnants of SDASO. The α A fragment contains a cross-linked lysine modified with the alkene (A) moiety, whereas the β S fragment contains a photocross-linked amino acid modified with a sulfenic acid (S) moiety. Because the NHS ester side of all three SDASO reagents are identical to half of DSSO, the expected alkene moieties are the same as seen in DSSO cross-linked peptides (Fig. 1E). In contrast, the three SDASO cross-linkers yield three different sulfenic acid moieties because of spacer arm differences in the diazirine end (Fig. 1F). As previously noted for other sulfoxide-containing crosslinkers (10,12,(33)(34)(35)47), the sulfenic acid moiety typically undergoes dehydration to become a more stable and dominant unsaturated thiol (T) moiety, leading to the detection of β T (supplemental Fig. S2A). To examine whether SDASO crosslinked peptides produce the expected fragmentation, standard protein BSA was cross-linked by the three SDASO cross-linkers separately, and the resulting peptide digests were analyzed by LC MS n . As illustrated (Fig. 2), each MS n analysis of the same BSA peptides interlinked by the three SDASO reagents yielded a dominant MS 2 fragment pair (α A /β T ) as predicted. These resultant MS 2 fragment ions representing single peptide chains were then subjected to individual MS 3 analyses, permitting unambiguous identification of both cross-linked peptide sequences and crosslinking sites. As a result, the respective cross-links between BSA:K155 and BSA:E41 were identified for all SDASO linkers.
Similar to residue-specific cross-linkers, SDASO crosslinking can also result in dead-end and intralinked peptides. For SDASO cross-linkers, two types of dead-end peptides are expected as both reactive ends can be hydrolyzed (supplemental Fig. S2, B and C). For NHS ester dead-ends, the resulting fragment ions would carry thiol moieties (supplemental Fig. S2B), whereas the MS 2 fragment ion of diazirine dead-end peptides would be labeled with an alkene moiety (supplemental Fig. S2C). These predicted MS 2 fragmentations were demonstrated by respective SDASO deadend peptides of BSA (supplemental Fig. S3). Similarly, for SDASO intralinked peptides, a single fragment would be expected, containing both an alkene and thiol modification (supplemental Fig. S2D). Exemplary MS n spectra of the three SDASO intralinked peptides of BSA further demonstrated the anticipated fragmentation (supplemental Fig. S4).
Collectively, the three types of SDASO cross-linked peptides fragment as predicted during CID to generate characteristic and predictable MS 2 products, which enable their simplified and accurate identification by MS n analysis in the same way as other sulfoxide-containing cross-linked peptides (10,12,(33)(34)(35)47).

SDASO XL-MS Analysis of BSA
To evaluate the performance of the three SDASO crosslinkers, we first carried out XL-MS analyses of BSA with three biological replicates each. Based on the general workflow (supplemental Fig. S5), LC MS n analyses resulted in a total of 556 unique SDASO-L, 405 SDASO-M, 324 SDASO-S interlinked BSA peptides, encompassing 427, 338, 306 unique K-X linkages, respectively (supplemental Table S1, A and B). Here, X represents any of the 20 common AAs. Although the three SDASO cross-linkers produced similar
Our results indicate that the three SDASO linkers have similar efficiency in cross-linking BSA and mapped a considerable number of shared regions but also yielded unique cross-linked peptides and sites.

Evaluation of SDASO Cross-Links of BSA
To explore the interaction coverage of BSA by SDASO cross-linking, we derived both 2-D and 3-D XL-maps based on the identified K-X linkages (Fig. 3, C and D). In comparison with our published XL-MS data of BSA using DSSO (aminereactive), DHSO (acidic residue-reactive), and BMSO (cysteine-reactive) cross-linkers (supplemental Fig. S7, A-D, the generation of the most extensive interaction coverages. As shown, interactions within the central core of BSA are broadly mapped by all types of linkers, while interactions at the N and C termini of BSA are best profiled by the SDASO linkers (Fig. 3, C-E, supplemental Fig. S7, A-D). These results demonstrate that SDASO cross-linking is effective for mapping interactions of single proteins and generates structural information complementary to residue-specific cross-linkers.
Among the 20 common AAs that can be targeted by diazirine, arginine has the longest side-chain. Considering the spacer arm lengths of SDASOs (i.e., SDASO-L [12. Although the spacer arm lengths are comparable, SDASO cross-links displayed higher satisfaction rates and lower average distances than those of DSSO and DHSO crosslinks of BSA (10). This may be due to the fact that amino acids other than arginine would result in distances less than the expected upper limits (29). To further validate, we compared distance distributions of SDASO data with that of random cross-links in BSA (supplemental Fig. S8, A-C), which displayed statistically significant differences, demonstrating that SDASO cross-links do not represent purely random cross-links.

SDASO-Based XL-MS Analysis of the Yeast 26S Proteasome Complex
To access the feasibility of photoactivated cross-linking for complex PPI mapping, we performed SDASO XL-MS analyses of affinity purified yeast 26S proteasome complex. This 33subunit protein degradation machine consists of two subcomplexes, the 19S regulatory particle (RP) and 20S core particle (CP) (49). The 19S RP contains 19 subunits that are assembled into the lid (i.e., Rpn3, Rpn5-9, Rpn11, Rpn12, Rpn15/Sem1) and base (Rpt1-6, Rpn1-2, Rpn10, Rpn13) subcomplexes, whereas the 20S CP is composed of 14 subunits (α1-7, β1-7) that form four stacked 7-member ring structures in the order of αββα. With three biological replicates for each linker, LC MS n analyses of tryptic digests of SDASO cross-linked complexes resulted in the identification of 1165 SDASO-L, 1133 SDASO-M, and 902 SDASO-S unique cross-linked peptides within the 26S proteasome (supplemental Table S2A), representing 1094 SDASO-L (496 intersubunit and 598 intrasubunit), 871 SDASO-M (416 intersubunit and 455 intrasubunit), and 777 SDASO-S (255 intersubunit and 522 intrasubunit) unique K-X linkages (supplemental Table S3A). As a result, 43% of SDASO-L, 52% of SDASO-M, and 60% of SDASO-S cross-linked peptide sequences (supplemental Fig. S9, A-C), as well as 29% of SDASO-L, 37% of SDASO-M, and 38% of SDASO-S K-X linkages were found reproducible among their respective biological replicates (supplemental Fig. S9, D-F), comparable to BSA data. These results further support the robustness of SDASO cross-linking. When comparing XL-MS data among the three linkers, we found that the number of SDASO cross-links of proteasomes increased with spacer arm lengths of the linkers, similar to BSA data. However, the resulting cross-link data among the three linkers shared considerably fewer in common for proteasomes than for BSA, with overlaps of 16% versus 37% for cross-linked peptide sequences and of 11% versus 29% for K-X linkages (Figs. 3, A and B and 4A, supplemental Fig. S10).These results suggest that spacer arm lengths of SDASO linkers play a more significant role in capturing interactions within protein complexes, most likely attributed to the presence of both interprotein and intraprotein interactions. Thus, the use of the three SDASO linkers is beneficial not only for result cross-validation but also for comprehensive PPI mapping of protein complexes.
As additional enzymatic digestions are known to increase sequence coverage in XL-MS analyses using residue-specific cross-linkers (44), we expected that similar results would be obtained for SDASO linkers. To test this, we performed chymotrypsin digestion of SDASO-L cross-linked proteasomes with three biological replicates. LC MS n analyses of chymotryptic digests resulted in the identification of a total of 776 unique SDASO-L cross-linked peptides of the 26S proteasome (supplemental Table S2, A and B), representing 804 SDASO-L unique K-X linkages, comparable to the trypsin XL-MS data as described above (supplemental Table S3A). While the reproducibility of XL-MS data was somewhat similar for both chymotryptic and tryptic digests of SDASO-L crosslinked proteasomes (supplemental Figs. S8, A and D and S11, A and B), their overlaps of cross-linked peptide sequences and K-X linkages were quite limited (~10%) (supplemental Fig. S11, C and D). This confirms that additional enzymatic digestion could facilitate the expansion of PPI coverages. Thus, tryptic and chymotryptic datasets of SDASO-L were combined, yielding a total of 1711 unique SDASO-L K-X linkages for subsequent analyses (supplemental Table S3A).

Validation of Proteasome Cross-Links by Structural Mapping
It is known that the 26S proteasome is a dynamic entity and possesses multiple conformational states to fulfill its function (49,50). To validate SDASO cross-links, we mapped the identified K-X linkages onto the four known structures of the yeast 26S proteasome that represent its progression through
To eliminate the possibility of the identified SDASO cross-links being random, we have compared their distance distributions with that of random cross-links of the 26S proteasome (supplemental Fig. S13, A-C). As shown, SDASO distributions are significantly different from the random distribution, similar to a previous report on the yeast 26S proteasome using residue-specific cross-linkers (53), further demonstrating the reliability of our identified cross-links.
Additionally, we noticed a group of SDASO linkages that appeared to fit better with a subset of models (supplemental Table S3A), suggesting the presence of conformational heterogeneity in the sample. To examine this, we classified a total of 159 SDASO cross-links as structural state-specific, because they were satisfied only by one, two or three out of the four models. We then grouped these differentially satisfied cross-links into 14 state-specific combinations to infer the presence of preferred structural states. As illustrated in Figure 4C, among all combinations, two major categories were detected for the three SDASO linkers, representing 82% of the total state-specific SDASO cross-links. One of them contained cross-links (54%) satisfied only by s1-s3 states but not by the s4 state, implying the presence of s1, s2 and/or s3 states in the purified proteasome. The other described crosslinks (~28%) satisfied only by the s4 state, indicating presence of that state. These two groups of state-specific cross-links represent 28 protein interactions, half of which describe connectivity within the 20S CP. The remaining half embody interactions within the 19S, particularly concerning Rpn11 and Rpn1. The results correlate well with the fact that these regions are expected to undergo significant conformational changes during state conversions of the 26S proteasome (51,52).
When considering intersubunit and intrasubunit cross-links separately, the latter has a slightly higher distance satisfaction when mapped to known structures (intrasubunit: SDASO-L: 98%, SDASO-M: 96%, and SDASO-S: 88% versus intersubunit: SDASO-L: 79%, SDASO-M: 86%, and SDASO-S: 86%) (supplemental Fig. S14, A-M). This is expected as intersubunit interactions are typically more dynamic. Coincidentally, the majority of nonsatisfied intersubunit linkages also localized to the 19S RP (supplemental Fig. S15, A-M), which is known to have diverse conformations (50). Collectively, structural mapping supports the validity of the identified SDASO cross-links and suggests the existence of multiple states in our purified proteasome.

Comparison of SDASO XL-Maps of the 26S Proteasome
To further evaluate the performance of SDASO in complex PPI mapping, we generated 2-D XL-maps of the 26S proteasome based on unique K-X linkages identified by each SDASO linker (Fig. 4D). A total of 135 nonredundant PPIs (103 intersubunit and 32 intrasubunit) within the 26S proteasome were determined based on 2427 K-X linkages identified by the three SDASO linkers, including 119 from SDASO-L (79 intersubunit and 30 intrasubunit), 81 from SDASO-M (53 intersubunit and 28 intrasubunit), and 61 from SDASO-S (32 intersubunit and 29 intrasubunit) (supplemental Table S3C). While~20% of intersubunit interactions were identified across all three linkers  Figure 4E. The differences in the PPIs captured by SDASO linkers are most likely related to their spacer arm lengths. Nevertheless, these results indicate that SDASO cross-linking covers a diverse range of protein interactions and that each SDASO linker contributes to mapping the comprehensive interaction network within the 26S proteasome.

DSSO XL-MS Analysis of the 26S Proteasome
To better assess SDASO cross-link data, we performed a set of XL-MS experiments on the yeast 26S proteasome using DSSO for comparison. LC MS n analyses identified a total of 2254 unique DSSO cross-linked peptides of proteasomes from two biological replicates, representing 1115 K-K linkages (659 intersubunit and 456 intrasubunit) and describing 107 intersubunit and 30 intrasubunit interactions (supplemental Tables S2C and S3C). While the overlap (65%) of DSSO cross-linked peptide sequences between the two biological replicates was comparable to those of SDASO data (57% 70%) (supplemental Fig. S16A), the reproducibility of DSSO residue-to-residue (i.e., K-K) linkages was higher (~65%) (supplemental Fig. S16B) than those of SDASO data (29% 38%). The increased variation in identified SDASO cross-link sites is expected as nonspecific cross-linking chemistry is inherently more variable. Nonetheless, these comparisons further demonstrate that SDASO cross-linking is robust on targetable interaction regions.
Next, we mapped DSSO cross-links onto the four conformational states (s1-s4) of the yeast 26S proteasome (51,52) and determined that on average~75% of DSSO K-K linkages were satisfied (≤30 Å) across all four models (supplemental Fig. S17, A and B). Interestingly, a total of 114 DSSO crosslinks were also found to be state-specific cross-links, as described above. However, the distribution of cross-links across the 14 state-specific combinations was somewhat different from SDASO data (supplemental Figs. S4C and S17C). In addition to the notable representations of s4 (30%) and s1-s3 states (15%) as seen in SDASO data, respective state-specific DSSO cross-links satisfied only by s1 state (~9%), s3 state (~8%), and s2-s3-s4 states (~14%) were markedly detected. These DSSO state-specific cross-links further support the presence of multiple conformational states of the 26S proteasome. Similar to SDASO data, intrasubunit DSSO cross-links were much better satisfied than intersubunit linkages for all four models (intra: 89% versus inter: 62%) (supplemental Fig. S17, D, F-I), and most of the violating DSSO intersubunit cross-links were attributed to the 19S RP (supplemental Fig. S17, E, J-M). Taken together, DSSO XL-MS data corroborate well with SDASO results, confirming the structural heterogeneity of affinity purified 26S proteasome and the dynamic nature of the 19S RP.

Comparison of SDASO and DSSO Cross-Linking of Proteasomes
To delineate the interactions captured by residue-specific and nonspecific cross-linkers, we took the cross-links identified in at least two biological replicates from all of our XL-MS experiments and combined SDASO data for further comparison. As a result, we obtained a total of 2186 SDASO crosslinks (959 intersubunit, 1227 intrasubunit) and 1098 DSSO cross-links (649 intersubunit, 449 intrasubunit) of the 26S

Identification of Proteasome Interacting Proteins
Besides interactions within the 26S proteasome, we also examined physical contacts with co-purified proteasomeinteracting proteins (PIPs). Considering only cross-links that were identified in at least two biological experiments from all of our XL-MS experiments, we obtained a total of 125 unique SDASO cross-linked peptides (175 K-X linkages) and 90 unique DSSO cross-linked peptides (90 K-K linkages), representing 44 interprotein and four intraprotein pair-wise interactions. This resulted in the identification of 24 PIPs (21 SDASO and seven DSSO) with direct contacts to the 26S proteasome, including 22 known (https://thebiogrid.org/) and two novel ones (Fig. 6, D and E), in which only four PIPs (Ecm29, Ubp6, Fzo1, and Rlf2) were found by both types of linkers (Fig. 6A, supplemental Table S3C). The four shared PIPs were identified with a total of 17 PPIs, of which only three (Rpt2-Ubp6, Rpt3-Ecm29, and Rpn2-Rlf2) were captured by both SDASO and DSSO. Among the known PIPs, Ecm29 is a key regulator of the 26S proteasome, and human Ecm29 has been shown to interact with Rpt1, Rpt4, Rpt5, Rpn1, and Rpn10 by DSSO cross-linking (54). Similarly, the interactions of yeast Ecm29 with Rpt1, Rpt4, Rpt5, and Rpn1 were confirmed by DSSO XL. In addition, Ecm29-Rpt3 and Ecm29-Rpn6 interactions from DSSO were identified for the first time. Furthermore, SDASO validated Ecm29-Rpt3 interaction and identified Ecm29-Rpt6 interaction (supplemental Fig. S19A). These results demonstrate extensive contacts between Ecm29 and the 26S proteasome, corroborating well with previous observation of its human orthologue (54). Ubp6 is a proteasome-associated deubiquitinase that interacts with the 26S proteasome through Rpn1 (55,56). While DSSO caught Ubp6-Rpt1 and Ubp6-Rpt2 interactions as reported (56) subunits including Rpn1, Rpn2, Rpn8, and Rpt2 (supplemental Fig. S19B). Overall, SDASO XL-MS analyses identified higher number of PIPs than DSSO, illustrating its capability of capturing interacting proteins in affinity purified samples.

Relative Specificity of Diazirine Cross-Linking
Diazirine photoactivation leads to not only the production of reactive carbene for AA labeling through X-H bond insertion, but also isomerization to form diazo compound to specifically react with carboxyl groups (27). Recent studies have suggested that diazirine labeling shows preferences for acidic residues (27,31). To examine this, we sought to determine whether any AA preference was observed in SDASO cross-linking of the 26S proteasome. On average,~26% of residues cross-linked by SDASO linkers were determined precisely at a single site, whereas the rest were localized ambiguously at one out of two (~34%), three (~20%), or four and more (~20%) possible sites (supplemental Fig. S20A). Similar precisions in SDASO crosslinked site localization was also observed in BSA data (supplemental Fig. S20B), consistent with conventional diazirine linkers (29). To prevent overestimation due to site ambiguity, we calculated the weighted AA occurrence to assess the preference of diazirine labeling in the 26S proteasome, similarly as described (31) (See Experimental Procedures section). Our results suggest that glutamic acid was the most favored by diazirine cross-linking, representing~30% of the targeted residues for all three SDASO linkers (supplemental Fig. S21A). In comparison, four additional residues, i.e., alanine (7.2%), aspartic acid (6.8%), leucine (7.3%), and tyrosine (6.4%) were targeted relatively favorably by SDASOs, as they had an average frequency well above those of the remaining AAs (2.7%). The dominant preference of glutamic acid displayed by diazirine cross-linking in proteasome samples was also detected in BSA, in which~25% of SDASO cross-linked sites were glutamic acids (supplemental Fig. S21B, supplemental Table S1D). Interestingly, five relatively favorable diazirine cross-linked sites in BSA contained aspartic acid, histidine, threonine, valine, and tyrosine with an average frequency of 6.8~8.4%, in which only aspartic acid and tyrosine residues showed similar preference in proteasome samples. This discrepancy is more likely attributed to the occurrence of common AAs in close proximity to cross-linkable lysines at interaction interfaces within proteins of interest as well as MS detectability and identification of the resulting cross-linked peptides. Nonetheless, while diazirine reactivity is nonspecific, our results suggest that it preferably targets a subset of AAs with glutamic acid as its most favored one. the first generation of sulfoxide-containing MS-cleavable heterobifunctional cross-linkers. The unique designs of the SDASO linkers enable a single labile bond to be preferentially cleaved over peptide backbone, leading to only one pair of MS 2 fragment ions and enhancing analysis sensitivity (47). Importantly, SDASO cross-linked peptides possess robust and predictable MS 2 fragmentation characteristics similar to sulfoxide-containing homobifunctional cross-linkers, thus permitting their fast and accurate identification using MS nbased XL-MS workflow (10,12,(33)(34)(35). Although MS 2 -based approaches have been widely used in XL-MS studies (2), it is important to note that MS n analysis is critical for effective database searching to identify photocross-linked peptides and localize nonspecific cross-linked sites with speed and accuracy, especially for complex samples. Owing to their unique capabilities, the SDASO cross-linkers have been successfully employed to study PPIs of not only a single protein BSA but also the affinity purified yeast 26S proteasome complex. To the best of our knowledge, this work represents the first application of photoactivated cross-linking on PPI mapping of large protein assemblies. The development of SDASO cross-linkers further demonstrates the robustness and potential of our XL-MS technology based on sulfoxidecontaining MS-cleavable cross-linkers and provides a viable analytical platform for the expansion of new MS-cleavable reagents to generate a complete PPI map of cellular systems in the future.
Although photoinduced diazirine labeling is nonspecific, the observed reproducibility of cross-linked peptide sequences was comparable for SDASO and residue-specific crosslinkers (8), supporting the reliability of photoactivated crosslinked products. While all of the 20 common AAs were detected as SDASO cross-linked sites in this work, SDASO displays preferential labeling of glutamic acids, corroborating well with previous reports on diazirine favoring acidic residues (27,31). Although aspartic acids are in comparable abundance to glutamic acids in BSA and proteasomes, they were targeted noticeably less by SDASO. In comparison, acidic residuereactive cross-linkers such as DHSO do not appear to have noticeable differences in reactivity toward these two AAs (8,10). Therefore, the preferential labeling of glutamic acids over aspartic acids displayed by diazirine may be because of differences in physiochemical properties of their side-chains and short-lived photoactivated reaction. In addition to acidic residues, several AAs including tyrosine, valine, leucine, threonine, and histidine have been detected as SDASO crosslinked sites more often than other AAs, in which tyrosine and histidine residues have exhibited favored carbene insertion in the past (31). The preferred reactivity of SDASO cross-linkers toward a subset of AAs including ones that cannot be easily targeted by specific cross-linking chemistries is beneficial to XL-MS studies, as it helps enhance the analysis of the resulting photoactivated cross-linked peptides and expand PPI coverage.
The complementarity in PPI mapping among the three SDASO linkers appears to be much more pronounced in the XL-MS analyses of proteasomes than BSA, implying the benefits of variable linker lengths for complex PPI profiling. In comparison to residue-specific cross-linkers such as DSSO, DHSO, and BMSO (10,12), SDASO XL-MS analyses of BSA has yielded the highest number of cross-linked peptides and the most comprehensive interaction maps. The high-density SDASO XL-maps of BSA illustrates the effectiveness of the heterobifunctional photocross-linkers for mapping a diverse range of interactions, which is in good agreement with previous reports (25,26). Intriguingly, while SDASO XL-MS analysis of the yeast 26S proteasome identified extensive intersubunit and intrasubunit interactions, the overall scopes of PPIs obtained from all three SDASO linkers is only comparable to those by DSSO and other residuespecific cross-linkers (53). Although DSSO produced a higher number of cross-linked peptides of the 26S proteasome than SDASO, comparisons of their cross-linked peptide sequences have revealed limited overlaps. Owing to diazirine nonspecificity, SDASO XL-maps of the 26S proteasome contain much more residue-to-residue connectivity. In addition, the three SDASO linkers have captured more interactions of the stable and compact 20S CP, but less of the dynamic and flexible 19S RP than DSSO. Because the spacer arm lengths of DSSO and SDASO linkers are similar, variance in PPI coverages is mostly attributed to crosslinkers' reactivity and kinetics (29). Collectively, our results have demonstrated the value of SDASO photocross-linkers in probing PPIs of both simple and complex samples. The extensive SDASO XL-MS data have allowed us not only to obtain comprehensive XL-maps complementary to those of existing cross-linkers but more importantly to better assess the reliability and capability of diazirine cross-linking in probing PPIs. Therefore, this work has established a solid foundation for future applications of photocross-linking in complex XL-MS studies.
DATA AVAILABILITY Raw data have been deposited at the PRIDE Archive proteomics data repository (ID: PXD022690). Annotated spectra for cross-link identifications can be viewed through MS-Viewer (https://msviewer.ucsf.edu/prospector/cgi-bin/ msform.cgi?form=msviewer) using the provided links in the supplemental data.