Rapid Mapping of Protein Interactions Using Tag‐Transfer Photocrosslinkers

Abstract Analysing protein complexes by chemical crosslinking‐mass spectrometry (XL‐MS) is limited by the side‐chain reactivities and sizes of available crosslinkers, their slow reaction rates, and difficulties in crosslink enrichment, especially for rare, transient or dynamic complexes. Here we describe two new XL reagents that incorporate a methanethiosulfonate (MTS) group to label a reactive cysteine introduced into the bait protein, and a residue‐unbiased diazirine‐based photoactivatable XL group to trap its interacting partner(s). Reductive removal of the bait transfers a thiol‐containing fragment of the crosslinking reagent onto the target that can be alkylated and located by MS sequencing and exploited for enrichment, enabling the detection of low abundance crosslinks. Using these reagents and a bespoke UV LED irradiation platform, we show that maximum crosslinking yield is achieved within 10 seconds. The utility of this “tag and transfer” approach is demonstrated using a well‐defined peptide/protein regulatory interaction (BID80‐102/MCL‐1), and the dynamic interaction interface of a chaperone/substrate complex (Skp/OmpA).

Abstract: Analysing protein complexes by chemical crosslinking-mass spectrometry (XL-MS) is limited by the sidechain reactivities and sizes of available crosslinkers,their slow reaction rates,a nd difficulties in crosslink enrichment, especially for rare,t ransient or dynamic complexes.H ere we describe two new XL reagents that incorporate amethanethiosulfonate (MTS) group to label ar eactive cysteine introduced into the bait protein, and ar esidue-unbiased diazirine-based photoactivatable XL group to trap its interacting partner(s). Reductive removal of the bait transfers at hiol-containing fragment of the crosslinking reagent onto the target that can be alkylated and located by MS sequencing and exploited for enrichment, enabling the detection of lowa bundance crosslinks.Using these reagents and abespoke UV LED irradiation platform, we show that maximum crosslinking yield is achieved within 10 seconds.T he utility of this "tag and transfer" approach is demonstrated using aw ell-defined peptide/protein regulatory interaction (BID 80-102 /MCL-1), and the dynamic interaction interface of ac haperone/substrate complex (Skp/OmpA).
Chemicalcrosslinking-massspectrometry(XL-MS)isapowerful component of the structural biology toolkit. [1] XL-MS methods have been enhanced by new data analysis software, [2] cleavable crosslinkers, [3] strategies for crosslink enrichment, [4] footprinting reagents, [5] and structural modelling approaches. [6] However,several challenges remain. Forexample,many XL-MS reagents have limited chemical reactivities (e.g. succinimide ester-based reagents) [7] restricting the residues for which information can be obtained. Secondly,t he low-abundance of crosslinked products often necessitates enrichment prior to MS. [8] Finally,c rosslinked peptide identification is difficult due to spectral complexity,p oor fragmentation efficiency,a nd the increased search space associated with the large numbers of peptide combinations. [2] Bulky crosslinking reagents may also perturb native interactions or create new aberrant interactions. [9] Furthermore,m any biological processes,f or example,p rotein folding,b inding and conformational changes,o ccur on short timescales.C onsequently,c rosslinks detected for ad ynamic/non-equilibrium system report the average of multiple states or are dominated by the longest-lived state of all those populated. Photocrosslinkers such as diazirines can address this challenge and have been used previously to tag Cys residues in peptides. [10] Diazirine-generated carbenes react with proteins in ns, [11] yet long irradiation times (minutes to hours [10b, 12] )a re often required to generate acceptable crosslink yields due to the use of low intensity lamps.
Here we exploit a" tag and transfer" approach [13] to develop two new XL-MS reagents and aw orkflow that enables crosslink identification for both well-defined and dynamic protein-protein interactions (PPIs). These heterobifunctional reagents comprise am ethanethiosulfonate (MTS) group for specific attachment onto as ingle Cys residue introduced into the "bait" protein, creating ac leavable disulfide bond within the linker arm, and adiazirine that crosslinks to a" target" protein ( Figure 1a,b,F igure S1,S2). Commercially available MTS-benzophenone-biotin tags have been used in photoinduced-crosslinking (PI-XL), but these bulky tags (752 Da) may perturb PPI interfaces. [14] In the reagents described here,r eduction of the disulfide bondcontaining linker arm between the crosslinked proteins leaves at hiol tag on the target (87 or 204 Da in size) (Figure 1b, Figure S3). Them odified residue can then be alkylated (e.g. with iodoacetamide [IAA]) and localised using MS protocols for identifying post-translational modifications (PTMs), [2] or enriched before MS.W ea lso describe the construction of a3 65 nm UV LED irradiation platform (Figure 1c,F igure S4a,b) which enables crosslinking reactions to be completed in just 10 seconds with marginal heating.W eexemplify this methodology by mapping two PPIs,each with adifferent binding mode:acomplex between BID 80-102 and MCL-1( K D = 50 AE 20 nm), [15] where the tight binding affinity is mediated by several key residues [15a, 16] and the dynamic chaperone/substrate complex Skp/OmpA (K D = 22 AE 16 nm), [17] in which the binding interface involves many rapidly interconverting interactions. [17,18] MTS-diazirine and MTS-trifluoromethyl phenyl diazirine (MTS-TFMD) (Figure 1a,F igure S1) were chosen as photoactivatable groups due to the superior performance of diazirines in comparative PI-XL studies, [12c, 19] their small size,and rapid, indiscriminate reactivity ( Figure S2). [9,11] Both diazirine-and TFMD-containing crosslinkers were synthesised as the photochemistry of TFMD leads to higher crosslinking efficiency,b ut it is bulkier. [20] In the bait, au nique solvent exposed Cys for crosslinker conjugation is introduced by mutagenesis or synthesis.I np roteins/peptides which lack Cys,o rc ontain buried Cys,t his is straightforward. When asolvent exposed Cys is already present in the bait this can be exploited, or substituted with Ala or Ser,a nd an ew Cys introduced in al ocation of interest. Knowing the location of the Cys on one partner in the PPI reduces the MS/MS search space from n tot 2 to n target (where n tot is the number of crosslinkable residues in the bait and target-for diazirines, this is every residue in the proteins-and n target is the number of residues in the target). Here,weintroduce aunique Cys by mutagenesis into the E. coli outer membrane protein OmpA, while for the BID 80-102 peptide,w hich comprises the binding domain of the pro-apoptotic protein BID,Cys was introduced via solid-phase peptide synthesis.I mportantly,t he reactivity of the photoactivated diazirine places no restrictions on the amino acids in the target protein that can be detected once ac omplex is formed.
Theh uman apoptotic regulatory pair BID/MCL-1 is at arget for cancer drug discovery. [22] TheB H3 binding domain of BID (BID 80-102 )a dopts ah elix on binding to asurface groove on MCL-1 (Figure 3a). [15b,16] Five single Cys variants of BID 80-102 (Figure 3a)w ere synthesised (see Supporting Information), each with au nique Cys residue in ad ifferent position. Subsequently,t he peptides were tagged with MTS-diazirine or MTS-TFMD.T he modified peptides bound MCL-1 with high affinity,a lthough those with the tagged residues in the centre of BID 80-102 (I86C and V93C) had areduced EC 50 (Figure 3b,T able S1). Crosslinking of the BID 80-102 peptides with MCL-1 was achieved by UV LED irradiation. Crosslinked complexes were detected by SDS-PAGE (Figure 3c). Thea bsolute crosslinking efficiencies varied from 9-40 %f or MTS-diazirine,w ith reduced efficiency for lower affinity variants (Table S1). Absolute crosslinking efficiencies were higher for the MTS-TFMD tagged peptides (26-53 %) (Table S1), as expected given the greater yield of TFMD-derived carbenes. [20] Gel bands of the crosslinked complexes were excised and the linker arm cleaved by reduction of the disulfide (Figure 1b)releasing MCL-1 bearing the transferred thiol-containing tag at the interaction sites. Thefree thiol in each tag was capped by alkylation with IAA and the protein digested in-gel with trypsin (see Supporting Information). Peptides modified with the transferred tag were identified by LC-MS/MS.This allowed BID 80-102 to be mapped  Lines are exponential fits to the data. Relative intensity of crosslinked product [%] was calculated as the intensity of the crosslinked product band on an SDS-PAGE gel at time t divided by the intensity at t final , assuming the maximal yield is achieved when the graph plateaus (see example SDS-PAGE of the XL reaction (bottom) and Figure S4c). b) Sample heating at irradiation times required to reach 15, 55 and 100 %maximal crosslinked product for the Hg-Xe and UV LED lamps. n.d. = not determined.
into the MCL-1 binding groove unambiguously,inagreement with the known "bind-and-fold" interaction [23] and NMR/Xray structures of the complex (Figure 3d,T ables S2 and S3, Figures S6-S8). [15b, 16] Both crosslinkers gave similar results ( Figure S6), demonstrating that the bulkier TFMD-diazirine can be used even when sidechain interdigitation drives association. Crosslinked positions on MCL-1 were quantified by MS/MS from each Cys residue introduced into BID 80-102 labelled with MTS-diazirine or MTS-TFMD.T his revealed that crosslinking yields were greater when the Ca-Ca Euclidian distance of the residues in the complex was < 10 ,b ut crosslinks could be detected for distances up to 15 ( Figures S9,S10).
We next tested the ability of our workflow and tag transfer reagents to study ap rotein complex stabilised by transient interactions.C haperone-client binding often involves ad ynamic interaction of chaperones with an unfolded/ partially folded client protein to aid folding or prevent aggregation. [24] These interfaces are challenging to map using crosslinking since the interactions can be dynamic and diffuse. Here,w eu sed the interaction between the periplasmic chaperone Skp from E. coli,a nd a b-barrel outer membrane protein substrate,O mpA, as am odel for this type of PPI. [25] Skp,ahomotrimer,h as aj ellyfish-like structure resulting in ac age in which substrates are sequestered (Figure 4a). [25b,26] Skp binds its OMP substrates with nm affinity, [17] similar to that of the well-defined BID 80-102 /MCL-1 complex. [15] However,S kp-bound OmpA is in a" fluid-globule" state that forms many weak and transient interactions in the Skp cage. [17,18] Tw os ingle Cys variants of OmpA were generated, W7C and T144C,l ocated in the b-barrel and as urface exposed loop in folded OmpA, respectively. [18] MTS-diazirine or MTS-TFMD conjugation of both variants was efficient (Figure S11), and did not prevent folding ( Figure S12). TheS kp/OmpA complex was assembled by dilution of urea-denatured OmpA into aS kp-containing solution, and PI-XL was then performed. Thecrosslinked Skp/OmpA complex was detected by non-reducing SDS-PAGE as as ingle band with an apparent mass of approximately 70 kDa (Figure 4b). Four non-overlapping positions from OmpA-(T144C) to residues within Skps "cage" (Figure 4d (top);F igure S13) could be identified by ingel digestion of the crosslinked proteins followed by LC-MS/MS ( Figure 5, Method 1). This result was surprising since it is inconsistent with previous studies which have shown that OmpA tumbles dynamically on as ub-ms timescale within Skp. [18] We reasoned that the lack of modified sites resulted from the relatively low abundance of these peptides in the Skp-OmpA complex, rather than reflecting as pecific interaction surface involving these four sites.G iven that each Skp trimer binds as ingle OmpA, and each OmpA contains only as ingle crosslinker,then only one modified peptide would result from each 70 kDa Skp/OmpA complex.
To test this conclusion we developed aprotocol to enrich chemically modified Skp,r emoving background peptides arising from OmpA and unmodified Skp ( Figure 5). We used thiopropyl Sepharose 6B beads to purify Skp monomers containing the thiol fragment obtained from the tag-transfer reaction and the Cys-containing OmpA bait. Thet hiolcontaining proteins were then eluted from the resin, separated by SDS-PAGEa nd trypsinised in-gel (Method 2). Alternatively,trypsin digestion was performed on-bead [27] and unbound peptides removed by washing the resin prior to elution (Method 3). Using Method 2 the number of modified peptides detected increased > 2-fold (from 6t o1 4), whilst Method 3 yielded afurther 2.5-fold increase in the number of modified peptides identified (from 14 to 35) (Figure 4d, Figure S13), aca. 6-fold increase in detection over Method 1. No significant sidechain bias was observed for either crosslinker ( Figure S14), consistent with the reactivity of the diazirine-derived carbenes. [9, 19b] Similar crosslinking sites were observed for OmpA(W7C) and OmpA(T144C) using both crosslinkers (Figures S13,S14). Using these enrichment protocols modifications on Skp were identified all around the internal cavity (Figure 4c), consistent with OmpA tumbling randomly in Skp. [18] Theability to enrich crosslinked peptides  Table S1). (c) SDS-PAGE of BID 80-102 labelled with MTS-diazirine (top) or MTS-TFMD (bottom) crosslinked to MCL-1. d) Residues of MCL-1 (magenta on ag rey ribbon) that crosslinked to at least one BID 80-102 peptide labelled with MTS-diazirine(left) or MTS-TFMD (right). BID 80-102 is shown in cyan and residues substituted as Cysa re coloured as in (a) (PDB ID:2KBW [16] ). See also Figures S6-S8, Tables S2, S3.
using our tag-transfer protocol, combined with the promiscuous reactivity of diazirines,e xemplifies the power of the technique to monitor even the most dynamic of protein interfaces (Figure 4c &d ).
In summary,w eh ave demonstrated that Cys-containing variants of ab ait protein conjugated with MTS-diazirine or MTS-TFMD-based tag-transfer crosslinkers can map PPIs in both well-defined and dynamic interfaces.S ince the location of the crosslink in the bait protein is known and only the transferred tag is considered in downstream analysis,t he background from bait peptides is removed and the search space for crosslinked products is reduced. These features are particularly important as target protein size increases.T he workflow described is simple to implement, only requiring the appropriate crosslinking reagents,L C-MS/MS and proteomics software for mapping PTMs.E nrichment and digestion of modified peptides enables low protein concentrations to be used, minimising the possibility of aggregation or other aberrant interactions.A dditionally,t he custom UV LED platform enables PI-XL on a10second timescale,not possible with arc-based lamps,a nd only previously achieved using pulsed lasers in solution or in the gas phase. [5,10a] Thelow cost (% $300) and simplicity of our UV LED system makes these timescales accessible to any researcher. Our workflow thus opens the door to time-resolved XL on the second timescale without the need for expensive lasers and enables the study of conformational changes within dynamic protein complexes versus time. . low reaction efficiencies mean that uncrosslinked material will remain). Enrichment can be performed using one of three methods (see the text). MS analysis of the peptides is then performed, and the data searched to identify the peptides/residues modified with the crosslinking reagent (with the free thiol capped by reaction with IAA).  Table S4). One monomer in the Skp trimer is highlighted. In cand d, red represents residue-level informationa nd blue represents crosslinks localisedw ithin a2or more residue region. d) Crosslinked residues identified by the enrichment methods shown in Figure 5f or OmpA(T144C)[MTS-diazirine]. The lower plot shows the total number of unique sites identified by the three enrichment strategies shown in Figure 5. Representative mass spectra of modified Skp peptides are shown in Figures S15,S16 and alist of crosslinked peptides is in Tables S4,S5.