Photo-crosslinkers boost structural information from crosslinking mass spectrometry

Abstract

Crosslinking mass spectrometry has emerged as a powerful tool in structural biology. This technology utilizes chemical crosslinkers to capture spatial proximities between protein residues to probe the organization, stoichiometry, and flexibility of protein assemblies under near-native conditions. Photo-crosslinking reagents have become increasingly used in crosslinking MS, with chemical properties that offer significant advantages when studying dynamic protein structures. This review explores the fundamentals, applications, and future potential of photo-crosslinkers in crosslinking mass spectrometry.

Introduction

Crosslinking mass spectrometry (Crosslinking MS) is a versatile approach with tremendous potential that bridges the gap between focused, high-resolution structural biology techniques (e.g., X-ray crystallography, cryo-EM, NMR) and proteome-wide protein interaction mapping studies. Crosslinking MS can provide structural insights on heterogeneous and dynamic complexes in near-native conditions and usually requires less protein material than other structural methods (1, 2). For identified crosslinks to be useful as structural information, they should report stable interfaces rather than random, transient encounters within the sample. Popular homobifunctional N-hydroxysuccinimide ester (NHS-ester) crosslinkers like BS³ (bis(sulfosuccinimidyl)suberate) or DSSO (disuccinimidyl sulfoxide) have reactive half-lives of tens of minutes at physiological pH and therefore often produce crosslinks between regions that rarely interact, resulting in noisy data that is difficult to interpret. In contrast, photo-activated crosslinkers (e.g., those with a diazirine as the reactive group (3)) generate reactive intermediates with nanosecond lifetimes, limiting crosslinking mostly to true interfaces [Figure 1A–C] (4). This property, which reduces spurious crosslinks with flexible regions or randomly passing protein complexes we refer to as “high contrast” crosslinking, dramatically improves both data interpretability and structural insights derived from crosslinking MS experiments (4, 5). Here, we discuss the applications of crosslinking MS and how UV-activated crosslinkers are boosting its usefulness for almost all applications.

Figure 1. Photo-crosslinkers provides more structurally relevant data than homobifunctional NHS-ester-based crosslinkers.

A) Comparison of a homo-bifunctional NHS-ester crosslinker (BS³, left) and B) a heterobifunctional NHS-ester-diazirine crosslinker (sulfo-SDA, right). The spacer regions with respective lengths are highlighted with the dashed brackets. NHS-esters (red) are constitutively reactive in native conditions, with preference to lysine residues and protein N-termini. NHS-esters can remain active for tens of minutes before hydrolysis renders them inactive. Diazirines (purple) need to be activated by UV light and remain active for only nanoseconds before they react with nearby protein residues or with bulk solvent. Heterobifunctional crosslinkers, such as sulfo-SDA (shown), typically react first on the NHS-ester side before the diazirine is irradiated with light to maximize the likelihood that the resulting reactive species will find another residue to react with. C) BS³ and sulfo-SDA crosslinks (black) identified from an in vitro crosslinked 70S ribosome sample. The homo-bifunctional NHS-ester crosslinker, BS³, captures many ‘over-length’ crosslinks which are likely formed by chance encounters between ribosomes, not within the same complex. Sulfo-SDA captures only relevant structural information due to the short lifetime of its activated diazirine. Respective histograms are shown below, plotting the number of crosslinks within the given binned Euclidian Cɑ-Cɑ distance from the 3-D structure (given in Å). The maximum theoretically allowed Cɑ-Cɑ distances are 30Å for BS³ and 25Å for sulfo-SDA. D) Chemical structures of a theoretical reaction between SDA (top) or DizSEC (bottom) with a lysine residue (brown) and the NHS-ester, along with an acidic residue (blue) and the UV-activated diazirine, which forms an ester bond (red). Chemical structures of resulting ions from HCD-based fragmentation regimes (black dashed lines) of the DizSEC-reacted peptide (dashed box). “Low-energy” HCD (NCE ~20, beige box) can fragment the ester bond while maintaining the rest of the peptide backbone, as the ester bond is more labile than the peptide bond. Typical HCD energies alone (NCE > 25) can fragment both the ester and peptide bonds, making downstream data analysis challenging. E) Selected MS2 spectrum for a sulfo-SDA-crosslinked peptide spectral match for an APOB intralink (22). Due to various fragmentation possibilities, fragments can come from each peptide alone or combinations of both peptides. In this example, we identify typical HCD-derived fragment ions of both the alpha (purple) and beta (green) peptides alone, but also the alpha peptide with the crosslink remnant stub (“P-S”) and its corresponding beta peptide fragment, which just appears as the peptide alone (“P-0”). This occurs when “low-energy” HCD selectively fragments the ester bond formed from the diazirine reacting with an acidic residue (black dashed line) but does not also fragment the peptides internally. Furthermore, the alpha and beta peptides can fragment at a position that does not affect the covalent crosslink; the selected example shown is highlighted in orange: the two crosslinked peptides plus crosslinker are found as one fragment ion (“b14+P”, orange), while its corresponding fragment counterpart is also identified (y7, light blue).

The chemistry of photo-crosslinking

There are two primary components of crosslinkers: 1) the reactive groups, and 2) the spacer region that connects them. The spacer region dictates the maximum distance between two residues that can be crosslinked [Figure 1A&B] (6). The reactive groups of a crosslinker dictate which parts of the protein structure can be accessed; these chemistries should be reactive in near-native conditions. Ideally, a crosslinker would incorporate broad amino acid reactivity; however, reagents that are specific to a small number of residues are often favored because this reduces the combinations that need to be searched during spectrum matching. The most widely used crosslinkers for crosslinking MS studies use two NHS-esters as their reactive groups (e.g., DSSO (7) and BS³ (8)). These NHS-esters are chosen for their specificity, predominantly reacting with the primary amines of lysine residues and protein N-termini, and, to a lesser extent, with the hydroxyl groups of serine, threonine, and tyrosine residues (9). These residues are often accessible at the surface of proteins and the reaction proceeds at physiological pH. Alternative, lesser-used chemistries include Michael Acceptors (e.g., maleimides and vinyl sulfones) for targeting cysteines (10, 11) and dihydrazines/diazos for targeting aspartate, glutamate, or protein C-termini (12, 13).

An exciting alternative to these constitutively reactive groups is the use of photochemistry, where normally inert groups are activated with photons to produce highly reactive species that have broad reactivity and, importantly, allow for temporal control over the crosslinking reaction. There are three main types of photo-activatable groups used in biological applications: aryl azides (14, 15), benzophenones (14, 16, 17), and diazirines (3, 18, 19), all three of which have been incorporated into commercially available crosslinkers (14, 17, 20).

Diazirine chemistry is the most well-characterized in terms of activity and specificity and remains the top choice in crosslinking MS. Upon UV light exposure (typically at 350–365 nm), the diazirine generates a highly reactive carbene or diazo species capable of inserting into any nearby amino acid to form a covalent bond. Although diazirine reactivity is generally promiscuous, multiple studies have now shown that diazirines will preferentially react with acidic residues to form ester bonds (3, 18) [Figure 1B&D]. A homobifunctional diazirine crosslinker could react with almost any amino acid, but the need to search all potential combinations of residues limits their practicality in crosslinking MS. As a practical compromise, heterobifunctional crosslinkers, like sulfo-SDA (sulfosuccinimidyl 4,4’-azipentanoate), have both an NHS-ester and a diazirine as their reactive groups [Figure 1B]. The more specific NHS-ester reacts first, anchoring the crosslinker to the protein, before the diazirine is activated with UV light. The resulting reactive intermediate persists for a very brief time (nanoseconds) before it reacts with an immediately adjacent protein residue or bulk solvent (typically water) (3, 18). This short reactive lifetime minimizes the crosslinking of rare or non-representative structural states, or simply random encounters, which are often captured by the much longer-lived homobifunctional NHS-ester crosslinkers (4) [Figure 1C].

Some extra considerations are required for the use of diazirines in crosslinking MS: first, the broad reactivity of diazirine-containing crosslinkers, and the tendency for this crosslinker to hydrolyze (forming dead-ends or mono-links), reduces the signal that can be detected in the mass spectrometer (3, 18). This means that offline and/or gas-phase enrichment of the crosslinked peptides is paramount for obtaining optimal results. Second, the ester bonds formed when diazirines react with acidic residues are extremely labile in higher-energy collision-induced dissociation (HCD), which can result in lost fragment information that would evidence the crosslinked residues (20, 21). We mitigate this problem by using stepped-HCD, with one low-energy step to maintain crosslinked peptide fragments without cleavage of the ester bond [Figure 1D&E] (22). Another way to get around this issue is to use an NHS-carbamate instead of an NHS-ester, as in the case of DizSEC [Figure 1D] (20). The NHS-carbamate reacts with lysine to form an MS-cleavable urea group that, when cleaved under HCD conditions, can leave a modification on each peptide. This asymmetric cleavage makes the crosslinking site easier to identify [Figure 1D]. Another heterobifunctional photo-crosslinker, succinimidyl diazirine sulfoxide (SDASO), contains an MS-cleavable sulfoxide moiety that produces a type of this asymmetric bond cleavage (23). Although not all current crosslinking search software are amenable to these specific crosslinkers, as photo-crosslinking MS popularity expands, we believe software capabilities will naturally follow.

Structural applications of photo-crosslinking mass spectrometry

Crosslinking MS is often used to generate structural data on purified complexes in vitro. For most of these applications, we advocate for a shift away from homobifunctional NHS-ester crosslinkers to those with diazirine chemistry for the reasons discussed above. These crosslinkers, in particular sulfo-SDA, have been shown to capture more relevant structural information than their homobifunctional NHS-ester counterparts (e.g., BS³) [Figure 1C] (4, 5). Their broad reactivity often provides clusters of identified crosslinks in 3D space that provide more confidence in localizing structural interfaces [Figure 2A].

Figure 2. Crosslinks can guide structural model building and prediction.

A) Cryo-EM density of the polymerase module from the yeast cleavage and polyadenylation complex with the structural model PDB 7ZGR (24). Crosslinks (red) can help validate an existing structural model, aid construction of a new model, and localize subunits with missing densities. B) Comparison of rpoA (cyan) and rpoC (beige) predicted binding models from both AlphaFold-Multimer (left) and AlphaLink 2 (right) with crosslinks (red). Crosslinks aided prediction in AlphaLink 2 to properly orient the interface of rpoA and rpoC (ipTM = 0.763). AlphaFold failed to produce a good model without incorporation of crosslinking information (ipTM = 0.243).

Crosslinking MS on purified protein complexes is most often used in combination with single particle cryo-EM, complementing cryo-EM’s limitations in resolving the structure of flexible or low-occupancy regions (22, 24–27) [Figure 2A]. Cryo-EM densities are produced by averaging the signal of many particles in vitrified ice. Crosslinks are formed on proteins in solution so can be used as independent experimental distance restraints to validate the fitted structural models (22, 24), to aid in assigning domains to ambiguous, lower-resolution densities (25–27), or even to support de novo structural model building (28, 29). Crosslinks can also indicate where proteins present in the sample are missing from the consensus cryo-EM density, suggesting where to target local refinement approaches or to suggest disruption of structured regions during grid making [Figure 2A]. Crosslinking MS data is also used to complement data from X-ray crystallography, SAXS, and NMR to provide independent validations of proposed structures (30). Crosslinks can also be used to guide protein construct design where the crosslinks suggest contact sites in protein complexes for further structural studies (31).

Crosslinks have been incorporated as sparse experimental distance restraints alongside other structural data in integrative modelling and docking approaches, such as IMP (Integrative modelling platform) (32), HADDOCK (33), and Rosetta (34). These approaches produce a consensus structural model based on all the available data, which can include crosslinking (35–37). The clustering of crosslinks produced with diazirine-containing crosslinkers has significant advantages when used in these approaches. Deep learning-based structure prediction approaches can also benefit from using crosslinks as sparse experimental distance restraints to bias them towards more accurate predictions. In a pioneering approach, Stahl et al. incorporated a diazirine-containing non-canonical amino acid into E. coli proteins and detected crosslinking residues by crosslinking MS (38). The crosslinks were used to bias the pair representation module which predicts residues likely to be in close contact of OpenFold (an open-source variant of AlphaFold), and this led to more accurate structure predictions with their software, AlphaLink (38). Data from longer, soluble crosslinkers, like sulfo-SDA, can also be encoded in the more recent version, AlphaLink 2, which can be used to model protein complexes [Figure 2B] (39). These approaches benefit greatly from the structural specificity of the crosslinks detected from diazirine-containing crosslinkers, particularly those with a short spacer, like SDA.

Interactions between intrinsically disordered proteins (IDPs) and regions (IDRs) with structured domains is an area where photo-crosslinking MS excels at providing structural information. More than a third of the human proteome is intrinsically disordered (40), which highlights the importance of having a structural technique that can tackle this daunting issue. IDPs bind to folded domains with a large range of affinities, making them difficult to study by cryo-EM. We show in Figure 2A how these stretches can be localized onto complexes by crosslinks. NHS-ester-based crosslinkers have proven to be particularly poor at studying these types of samples, as the flexible region will randomly contact many non-specific areas of the proteins (5). On CCR4-NOT, a structurally interactable complex, a diazirine-based photo-crosslinking MS workflow was able to localize the binding sites of IDRs, allowing them to be confirmed by NMR and biophysical assays (41).

Photo-crosslinking dramatically improves clarity of quantitative studies

Proteomics is an effective comparative technology and comparing abundances of crosslinks in different conditions can reveal details of protein complex dynamics and conformational changes (42). As structural states always exist in an equilibrium, quantitative crosslinking MS (qCLMS) experiments are best at comparing conditions that strongly bias the conformation, for example, by ligand/protein binding (43), post-translational modifications (44), or environmental changes (45). Here, again, it has been shown that diazirine-based crosslinking was superior to homobifunctional NHS-ester-based approaches (5). The long reactive half-life of the NHS-esters can mean that all the crosslinks from rare conformations are overrepresented, leading to an averaging away of the interesting structural signatures [Figure 3A]. Diazirines provide a much clearer quantitative signal of the conformational changes (5). Workflows have been developed to do quantitative crosslinking using tandem mass tags (TMT) (46), data-independent acquisition (DIA) (47), or isobaric crosslinkers (48), but the field has so far produced a limited number of qCLMS studies. Moving forward, we suspect that photo-crosslinkers in qCLMS will prove effective at revealing the dynamics of protein conformations and interaction in a variety of applications, dramatically boosting the popularity of the approach.

Figure 3. Photo-crosslinking for comparative and interactome studies.

A) Depiction of a theoretical protein complex that exists in an ensemble of two conformational states. This ensemble is heavily biased to the closed conformation by the addition of a post-translational modification (PTM, orange). Photo-crosslinkers, such as SDA, can more accurately capture these dynamics (middle panel), while homobifunctional, NHS-ester-based crosslinkers, such as BS³, tend to capture the most compact conformational state, due to their long-lasting reactive groups (bottom panel). B) Photo-crosslinkers can be used for both whole-interactome and affinity-purification type experiments. Photo-crosslinkers (purple bar), such as SDA and DizSEC, can crosslink entire cells; however, without specific enrichment, detection of crosslinks is limited to those from the most abundant interactions of the cell (left). Alternatively, photo-crosslinking offers the ability to capture crosslinks on a tagged protein of interest: a bait protein (red) with a tag (cyan) can be crosslinked under native conditions and then purified before crosslinking MS (right). This approach promises to produce much more detailed information on the complex of interest.

In situ photo-crosslinking

While photo-crosslinkers are now regularly used for studying purified protein complexes, the expansive search space and signal dilution compared with homobifunctional crosslinkers have limited their use in more heterogeneous samples. In situ applications of photo-crosslinking MS offer unique opportunities to capture protein-protein interactions (PPIs) and structural information directly in native environments (49). These environments are highly complex, characterized by a vast range of protein abundances and compositional heterogeneity, making crosslinking MS analysis particularly challenging; consequently, these analyses have remained limited to a few specialist laboratories (49). NHS-ester-based crosslinkers have typically been used, but the prolonged reactivity of these reagents often results in non-specific crosslinks between proteins that only transiently pass each other, making the interaction networks difficult to interpret (37, 50, 51).

Photo-crosslinkers, with their short-lived reactive intermediates, should be able to reduce these artefacts. A recent study showed that useful structural information could be produced from crosslinking MS applied to mRNP particles that contained hundreds of proteins, suggesting that careful enrichment of these crosslinks can push past what have been considered the complexity limits for photo-crosslinkers (52). Photo-crosslinking MS for studying an entire cell was first demonstrated by Stephen Fried’s lab, where they used diazirine-based crosslinkers to study the E. coli protein interactome (20). While this study only identified the most abundant interactions, it demonstrated a valuable proof-of-concept [Figure 3B]. Furthermore, the work performed by Stahl et al. 2023 illustrated how photo-Leucine crosslinks can be detected from whole-cell photo-crosslinking (38). NHS-ester-based crosslinkers with enrichable handles, such as Alkyne-A-DSBSO and tBu-PhoX, have been developed to boost the numbers of crosslinks that could be detected in these complicated samples. Only one enrichable, diazirine-containing crosslinker has been reported, but its practical use-cases remains unexplored (53). Analyzing entire cellular interactomes by photo-crosslinking MS will remain difficult for the foreseeable future, but it may be possible to enrich simplified interactomes from crosslinked cells. The Rappsilber lab recently showed that photo-crosslinking can fix transient interactions in situ prior to pulldowns, allowing for the specific enrichment of these labile interactions (54) [Figure 3B]. This work highlights the potential for photo-crosslinkers to be used in increasingly sophisticated applications, such as combining crosslinking MS with proximity labeling or affinity pulldowns, which would boost the information gathered for specific complexes (55).

Conclusions

Diazirine-containing crosslinkers represent a transformative advancement in crosslinking MS, enabling the study of dynamic protein complexes with unparalleled clarity. Their high temporal precision, broad amino acid reactivity, and ability to report on flexible and transient regions makes them powerful tools in structural biology. While challenges such as data complexity and crosslinked peptide enrichment requirements persist, innovations in crosslinker design and analytical techniques promise to overcome these limitations. With ongoing advancements in crosslinking chemistries, sensitivity of mass spectrometers, and computational tools, the capabilities of crosslinking MS continue to expand, making it an increasingly promising technique.