Keywords

1 Introduction

Proteins play pivotal roles in all biological processes. As the structure of a protein dictates its function, investigating the 3D-structure of a protein and clarifying its interactions with other proteins is one of the most important tasks to elucidate biological processes. While the 3D-structural analysis of proteins is commonly achieved by nuclear magnetic resonance (NMR) spectroscopy, X-ray crystallography, and cryo-electron microscopy (cryo-EM), protein-protein interactions might be identified by co-immunoprecipitation or Förster resonance energy transfer (FRET) (Operana and Tukey 2007).

To date, NMR and X-ray crystallography are still the dominating techniques to determine high-resolution protein structures as is indicated by the large number of structures available in the PDB (∼120,000 structures obtained by X-ray crystallography versus ∼12,000 structures obtained by NMR spectroscopy). Limitations of both high-resolution techniques, however, persist in the investigation of very large and transient protein complexes as well as membrane proteins. Cryo-EM overcomes some of these limitations as structural analysis can be performed at rather low protein concentrations (less than 1 μM) and highly complex protein assemblies can be targeted (Li et al. 2013).

Cross-linking/MS is an approach that complements the high-resolution 3D-structural techniques and has emerged as promising tool for the structural investigation of proteins and protein complexes in the last years (Young et al. 2000). Especially the combination of cryo-EM with cross-linking-MS has proven beneficial to provide insights into large protein assemblies (Greber et al. 2014; Weisz et al. 2017; Benda et al. 2014) that cannot be obtained by X-ray crystallography or NMR spectroscopy.

Cross-linking/MS relies on introducing covalent connections between functional groups of amino acid side chains by a chemical reagent. This cross-linker possesses a defined length and connects only these amino acids that are in the appropriate distance to be cross-linked. Usually, the analysis of the cross-linked amino acids is performed in a classical proteomics “bottom-up” approach where the cross-linked protein(s) are enzymatically digested and the peptide mixtures are analyzed by liquid chromatography tandem mass spectrometry (LC/MS/MS). This highly sensitive method allows examining as low as femto- to attomole amounts of proteins. The distance constraints that are derived from the cross-linked amino acids are subsequently employed for de-novo or homology modeling approaches (Leitner et al. 2016; Rappsilber 2011; Sinz 2014; Walzthoeni et al. 2013; Politis et al. 2014).

Importantly, the cross-linking/MS approach is not only applicable to the 3D-structural analysis of purified proteins, but it also allows elucidating protein-protein interaction networks (Häupl et al. 2016; Schweppe et al. 2017). Protein-protein interaction studies are often based on an affinity enrichment of a tagged bait protein to a specific matrix (Puig et al. 2001; Gavin et al. 2002), together with its interaction partners. Here, the washing procedure applied to remove non-interacting proteins is a crucial step, which however harbors the risk of losing transiently or weakly bound protein interaction partners. Due to the covalent fixation of proteins in the cross-linking/MS approach, the loss of weakly bound proteins during the washing procedure is circumvented.

In this chapter, we give an introduction into the principles of cross-linking/MS and present examples for successful applications of this approach to derive 3D-structural information of proteins and to identify protein interaction networks.

2 The Cross-Linking/MS Strategy

Once the 3D-structure of a protein or protein complex is covalently fixed by a cross-linking reagent in-vitro or in-vivo, the identification of the cross-linked amino acids will ultimately give insights into the spatial organization of the protein system under investigation. After the cross-linking reaction, the reaction mixture is analyzed by one-dimensional gel electrophoresis (SDS-PAGE) to visualize the result of the cross-linking reaction and to eventually optimize the reaction conditions (Sinz 2006; Rappsilber 2011) (Fig. 8.1). As mentioned above, the analysis of the cross-linked amino acids is commonly achieved by a “bottom-up” approach, including enzymatic digestion and LC/electrospray ionization (ESI)-MS/MS analysis of the resulting peptide mixture. Proteolysis is realized either by in-gel or in-solution digestion. Applying the in-gel approach, the band containing the protein or protein complex of interest is excised and digested within the gel. Alternatively, proteolytic cleavage can be carried out directly in-solution without previous separation of the proteins. The resulting peptide mixture is highly complex as it not only contains non-cross-linked, i.e., linear, and cross-linked peptides of the target proteins, but also peptides of possible contaminants or the protease used for digestion. In subsequent LC/MS/MS analysis, mass spectra and fragment ion mass spectra are recorded, followed by the identification of cross-linked peptides by specific software tools, such as xQuest (Rinner et al. 2008), pLink (Yang et al. 2012), StavroX (Götze et al. 2012a) or Kojak (Hoopmann et al. 2015), that automatically match MS/MS spectra of cross-linked peptides to cross-linking candidates. The cross-links deliver (i) distance information for 3D-structural computational modeling and (ii) insights into the identity of protein interaction partners and protein-protein interaction sites.

Fig. 8.1
figure 1

Cross-linking/MS workflow. A protein or protein complex is stabilized by introducing a covalent bond with a cross-linking reagent. Separation of the cross-linked protein(s) by SDS-PAGE is followed by enzymatic in-gel or in-solution digestion, resulting in a peptide mixture containing cross-linked and non-cross-linked (linear) peptides. Applying the peptide mixture to MS analysis enables the identification of cross-linked peptides by customized software tools that match MS/MS spectra to potential cross-linking candidates. The cross-links identified provide distance information for modeling protein 3D-structures or for identifying protein interaction partners

Full size image

3 Experimental Design of the Cross-Linking/MS Workflow

3.1 Cross-Linker Design and Reactivity

The commonly used cross-linkers comprise two reactive head groups that are separated by a spacer with a defined length (Sinz 2006). The spacer determines the distance between the amino acids to be covalently connected and as such serves as “molecular ruler” within a protein or protein complex. Cross-linking reagents are categorized into homobifunctional cross-linkers, comprising identical head groups, or heterobifunctional cross-linkers with non-identical reactive sites (Table 8.1). The most frequently employed reactive groups are N-hydroxysuccinimidyl (NHS) esters targeting primary amines in lysines and protein N-termini. For NHS esters, an additional reactivity for hydroxy groups in serines, threonines, and tyrosines has been observed (Mädler et al. 2009; Kalkhof and Sinz 2008). Heterobifunctional linkers often contain NHS esters as one of the reactive groups (Hermanson 1996). The second reactive site can be a maleimide, targeting cysteine residues, or a photo-reactive group, such as diazirines or benzophenones. Photo-reactive moieties react in a non-specific manner, potentially connecting all 20 amino acids that are in spatial proximity. For diazirines, a preference for acidic amino acids was observed (Ziemianowicz et al. 2017; Jumper et al. 2012; Iacobucci et al. 2018), while benzophenones target mostly methionines (Wittelsberger et al. 2006). Cross-linking of acidic amino acids is still a challenging task at experimental conditions that do not interfere with the native protein structure. Hydrazines react with aspartic and glutamic acid residues as well as with the C-terminus of proteins upon activation with 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) (Novak and Kruppa 2008) or 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMTMM) (Leitner et al. 2014a). The activated acidic amino acid might then also react with primary amines of lysines and the N-terminus, forming a direct connection between a carboxylic acid and an amine group (Schwarz et al. 2016).

Table 8.1 Functional groups in cross-linking reagents
Full size table

In addition to the cross-linking reagents that are externally introduced (Table 8.1) photo-reactive, unnatural amino acids are available that are directly incorporated into proteins (Suchanek et al. 2005; Piotrowski et al. 2015). These photo-reactive amino acid analogues contain photo-reactive groups, such as benzophenones or diazirines, and they are incorporated into proteins during the translation process in living cells. In general, two strategies are applied, where the photo-reactive amino acids are incorporated into the protein(s) either in a site-specific or a non-directed fashion. The site-specific incorporation of the photo-reactive amino acid makes use of the amber stop codon that is placed at the desired position in the DNA. The amber stop codon encodes for the photo-reactive amino acid, e.g. para-benzoylphenylalanin (Bpa) (Ryu and Schultz 2006; Schwarz et al. 2016). To incorporate the photo-reactive amino acid, a specific transfer RNA (tRNA) is needed, which is encoded by an additional plasmid to be transformed or transfected into the cell. The tRNA binds the photo-reactive amino acid and incorporates it into the protein at the amber stop codon position. The major advantage of this approach is that the cross-linking reaction will specifically take place at the desired position within the Bpa-labeled protein. On the other hand, the non-directed incorporation of the photo-reactive amino acid exploits the translation machinery of the cell to incorporate photo-reactive amino acids. As such, photo-methionine (photo-Met) or photo-leucine (photo-Leu) can be incorporated into proteins by the respective tRNAs for methionine and leucine (Suchanek et al. 2005; Piotrowski et al. 2015; Lössl et al. 2014; Häupl et al. 2017; Iacobucci et al. 2013). Efficient incorporation of photo-Met into proteins has been shown for different cell types (E. coli, HEK 293 and HeLa cells) with incorporation rates of 30–35% (Piotrowski et al. 2015). A detailed protocol of a cross-linking approach using the complementary cross-linking principles of BS2G and photo-reactive amino acids is provided in (Lössl and Sinz 2016).

3.2 Identification of Cross-Linked Peptides

As shown in the cross-linking/MS workflow, cross-linked peptides are generated by enzymatic cleavage of the cross-linked proteins by a specific protease (Fig. 8.1). The most prominent protease is trypsin that cleaves proteins C-terminally to basic amino acids (lysine and arginine residues). Cross-linking/MS however differs from the usual proteomics workflow as the use of one single protease is in some cases not sufficient. As two peptides are covalently connected, high molecular weight products are generated exceeding the optimal range of peptide MS detection. Applying a protease additionally to trypsin, such as GluC (cleaving C-terminally to glutamate and aspartate residues), will decrease the molecular weight of cross-linked peptides (Piotrowski et al. 2015). Another approach to generate peptides with lower molecular weight is to conduct proteolysis by an unspecific protease, such as proteinase K (Petrotchenko et al. 2012).

In general, cross-linked peptides are categorized into three different classes as type 0, type 1, and type 2 cross-links (Table 8.2) (Schilling et al. 2003). Type 0 (“dead-end” or “mono-link”) describes a peptide, in which one amino acid is modified by a cross-linker reagent. Here, only one reactive group of the cross-linker has reacted with an amino acid, while the other one has been hydrolyzed or has reacted with the reagent that was used for quenching the cross-linking reaction. “Dead-end” cross-links can deliver insights into the solvent-accessible surface of specific amino acids and as such, give information on the overall topology of the protein under investigation. Type 1 (intrapeptide or “loop-link”) describes the connection of two neighboring amino acids within one peptide. Only limited information on the protein’s tertiary structure is provided by these cross-links. Type 2 (interpeptide) cross-links connect two peptides originating from one protein or interacting proteins. This class represents the most valuable cross-linked products that yield information on the structural proximity of specific amino acid residues and allow deducing 3D-structural information. According to the systematic nomenclature provided by Schilling et al., the higher molecular weight peptide is termed “α-peptide” whereas the peptide with the lower molecular weight is referred to as “ß-peptide” (Schilling et al. 2003).

Table 8.2 Nomenclature of cross-linked peptides
Full size table

The number of software tools for identifying cross-linked peptides from MS data is steadily increasing and not easy to review. Table 8.3 provides an overview comprising several of the so far developed software applications. The general workflow of these software tools includes an in-silico digestion of proteins. Subsequently, potential cross-link candidates are automatically compared to the recorded MS/MS spectra and matching cross-link candidates are reported. In addition to the software tools available for cross-link identification, an increasing number of software applications is available to further examine the cross-links, such as xVis (Grimm et al. 2015) or Xlink-DB (Zheng et al. 2013). These applications provide a visualization of the identified cross-links either in a schematic fashion or by mapping them into published PDB structures. Also, there are software tools available for quantifying cross-linked peptides, e.g. xTract (Walzthoeni et al. 2015) or XiQ (Fischer et al. 2013).

Table 8.3 Selected software tools for identifying cross-linked peptides
Full size table

3.3 Facilitated Analysis of Cross-Linked Products

The identification of cross-linked products is still challenging due to their low abundance in the peptide mixtures generated after enzymatic digestion (Fig. 8.1). Specifically, we see two main difficulties: (i) Cross-linked peptides might be missed during MS analysis, (ii) false-positive identifications of cross-links might occur due to their great variability. To overcome these problems, sample complexity can be reduced or cross-linked products can be enriched. Alternatively, cross-linkers carry an isotope signature (usually by introducing deuterium atoms) or create specific fragment ion patterns during MS/MS experiments for automated data analysis.

Strong cation exchange (SCX) and size-exclusion chromatography (SEC) are the methods of choice to enrich cross-linked peptides (Leitner et al. 2010, 2014b; Schmidt and Sinz 2017). Also, an affinity enrichment of cross-linked products is based on a biotin tag that specifically binds to avidin. Biotin is either incorporated in the cross-linking reagent (Tang and Bruce 2010) or is introduced after the cross-linking reaction by click chemistry (Nury et al. 2015). In order to utilize SCX enrichment of cross-linked peptides, enzymatic cleavage has to be performed with a protease cleaving C-terminally to basic amino acids, such as trypsin. Consequently, every peptide carries a positive charge at the C-terminus that can be utilized for SCX enrichment. As cross-linked products are composed of two peptides, they accommodate a higher number of positive charges than linear peptides. Thus, the cross-linked peptides can be enriched by SCX (Leitner et al. 2010; Fritzsche et al. 2012; Schmidt and Sinz 2017; Tinnefeld et al. 2017). As cross-linked peptides usually possess higher molecular weights than non-cross-linked peptides, they can also be enrichment by SEC (Herzog et al. 2012; Rampler et al. 2015). A major advantage of SEC is its applicability for all peptide mixtures, independent of the protease used for digestion, but on the other hand, low-molecular weight cross-linked peptides might get lost during SEC enrichment.

For an unambiguous identification of cross-linked products, cross-linking reagents with unique characteristics have been designed. The first class of cross-linkers contains isotope labels, in most cases deuterium atoms, such as bis(sulfosuccinimidyl)suberate (BS3) D 0/D 4 (Fig. 8.2a) (Müller et al. 2001; Schmidt et al. 2005). The deuterated and non-deuterated version of the cross-linker are mixed in a 1:1 ratio and added to the protein solution. Hence, every cross-linked product is visible in mass spectra as a specific doublet of signals. Both species generate nearly identical MS/MS spectra, but differ in the fragment ions containing the deuterated or non-deuterated cross-linker. Acquiring MS/MS spectra from both isotope species helps in unambiguously identifying cross-linked peptides as two spectra with the specific mass shift are only obtained for cross-linked products, but not for linear peptides.

Fig. 8.2
figure 2

Strategies to facilitate the analysis of cross-linked peptides. Structures of respective cross-linkers are presented in the upper panel. Specific cleavage sites of the cross-linkers upon collisional activation inside the mass spectrometer are indicated. The middle panel (MS) displays how the cross-linked peptides appear in the mass spectrum, while the lower panel (MS/MS) shows characteristics of the cross-linked products in fragment ion mass spectra. (a) For the isotope-labeled cross-linker BS3(D 0/D 4), two signals are detected in the mass spectrum for one cross-linked peptide pair differing by the isotope label (4 amu). Subsequently, two MS/MS spectra for one cross-linked peptide pair are recorded, black – BS3D 0; bold – BS3D 4. (b) For the MS-cleavable cross-linker DSBU, characteristic fragment ion signatures of the linker (two doublets) are visible in MS/MS spectra. (c) PIR cross-linkers exhibit characteristic fragment ions (dashed line) and peptides modified with cross-linker fragments (bold lines) in MS/MS spectra

Full size image

A highly attractive approach that is currently gaining more and more importance is employing cross-linkers with an MS-cleavable moiety (Sinz 2017). The cross-linker is cleaved during collisional activation in the gas phase inside the mass spectrometer resulting in specific fragment ions that contain parts of the cross-linker. In Fig. 8.2b, the MS-cleavable cross-linker disuccinimidyl dibutyric urea (DSBU, formerly BuUrBu) is presented. DSBU comprises two NHS esters as reactive sites and a cleavable urea moiety (Müller et al. 2010). After fragmentation of the linker, two doublet signals are visible for each interpeptide (type 2) cross-linked product in the MS/MS spectrum. These doublets result from cleavage of one of the two NH–CO bonds of the urea moiety. Thus, two pairs of asymmetric fragments are generated, exhibiting a specific mass difference of 25.979 amu that allows an unambiguous identification of a cross-linked product.

There are other MS-cleavable cross-linkers available that generate characteristic fragment ions, such as a class of reagents termed “protein interaction reporters (PIR)” (Fig. 8.2c) (Tang and Bruce 2010). PIR cross-linkers comprise two cleavage sites releasing a specific part of the cross-linker after collisional activation inside the mass spectrometer. This specific fragment ion as well as peptides containing the remaining fragment of the cross-linker are present in every MS/MS spectrum of a cross-linked product and allow its unambiguous assignment.

4 Structural Investigation of Purified Proteins and Large Protein Assemblies

As outlined above, the distance constraints derived from the cross-linked amino acids serve as basis for a computational modeling of purified proteins or protein complexes. The distance constraints are provided as Cα-Cα or Cβ-Cβ distances and defined by the length of the side chains of the cross-linked residues plus the cross-linker spacer length (Merkley et al. 2014; Hofmann et al. 2015). In principle, two strategies are employed to implement these constraints into the modeling process: (i) modeling of protein structures using the cross-linking distances as input, (ii) filtering the theoretical models by the experimentally determined cross-linking distances. A number of different software tools are available for modeling protein structures, such as ROSETTA (Kaufmann et al. 2010), I-TASSER (Zhang 2009), PEP-FOLD (Maupetit et al. 2009) or Abalone (http://www.biomolecular-modeling.com/Abalone/index.html). The optimal cross-linker spacer lengths for protein modeling have been evaluated by combining and analyzing simulated and experimentally observed cross-linking constraints for various proteins (Hofmann et al. 2015). A specific equation was developed to predict the ideal spacer length in correlation to the size of the targeted protein. As a result, the optimal spacer length of a cross-linker to study the 35-kDa human phosphatase activator protein (PDB-ID: 2IXM) was determined to be 12.5 Å (Hofmann et al. 2015).

The investigation of a purified protein aims at elucidating its tertiary structure and the assembly state of eventually present oligomeric forms. Furthermore, newly developed cross-linking reagents are usually evaluated using small proteins, such as myoglobin, human serum albumin or bovine serum albumin (Brodie et al. 2017; Belsom et al. 2016, 2017; Giese et al. 2016a; Iacobucci et al. 2017). Small proteins, such as myoglobin and the FK506 binding protein (FKBP), have also been used for studying the impact of cross-linking applied to structural modeling with a discrete molecular dynamics (DMD) simulation based on cross-linking constraints (Brodie et al. 2017). Five short-range cross-linkers with various functional groups targeting different amino acids were employed to derive distance constraints of both proteins. The cross-linking data obtained were subsequently used as input for the DMD. Clustering of the generated models identified three clusters for FKBP and two for myoglobin. Representative model structures of each cluster were similar to the known PDB structures (Fig. 8.3a), underlining the strength of the cross-linking/MS approach to obtain native-like conformations.

Fig. 8.3
figure 3

Structural investigation of single proteins and a large protein complex. (a) Comparison of cross-link-based modeled structures to available PDB structures of FK506 binding protein (FKBP) and myoglobin. The best scored structures of the largest clusters are superimposed with the PDB structures (dark grey). Figure is adapted with permission from (Brodie et al. 2017). (b) Structure of Psb28 docked to the RC47 subcomplex of the photosystem II of the cyanobacterium Synechocystis sp. (Figure is adapted with permission from Weisz et al. 2017)

Full size image

Applying cross-linking to large protein complexes illustrates that the size of the complexes of interest is unrestricted for cross-linking/MS approaches, while NMR or X-ray crystallography are limited in protein size by sample preparation and data acquisition. Cross-linking/MS is able to deliver structural information on small protein assemblies, such as nidogen-1/laminin γ1 (Lössl et al. 2014) or chaperone Hsc70/α-synuclein complexes (Nury et al. 2015), but also on large protein systems, such as the mammalian mitochondrial ribosome (Greber et al. 2014) or the ribosome post-recycling complex (Kiosze-Becker et al. 2016).

To simplify the study of protein-protein interactions, representative peptides can be employed that harbor known interaction sites, as predicted by preceding biochemical studies or computational approaches. As an example, Munc13 peptides containing the respective calmodulin (CaM) binding site were synthesized to investigate presynaptic CaM/Munc13 complexes (Dimova et al. 2009; Lipstein et al. 2012). For this, the unnatural photo-reactive amino acid Bpa was incorporated during peptide synthesis into Munc13 peptides and the peptides were applied for photo-cross-linking experiments. Additionally, the heterobifunctional cross-linker N-succinimidyl-p-benzoyldihydrocinnamate (SBC), containing a NHS ester and a benzophenone group, as well as BS3 and bis(sulfosuccinimidyl)-2,2,4,4-glutarate (BS2G) were applied to obtain complementary cross-linking data on the interaction between CaM and Munc13 peptides. The cross-linking constraints were then subjected to computational modeling of the CaM/Munc13 peptide complexes using the PatchDock and ROSETTADock software applications. The resulting structures revealed all Munc13 isoforms to bind similarly to CaM, indicating a common CaM-binding motif of all four Munc13 isoforms (Dimova et al. 2009; Lipstein et al. 2012).

Photosystem II intermediate complexes from cyanobacterium Synechocystis sp present an impressive example for applying cross-linking/MS to large protein assemblies (Weisz et al. 2017). To study the interaction of the cytoplasmic photosystem II protein Psb28 with the membrane bound photosystem II component, complete photosystem II was purified from Synechocystis sp. Structural analysis was carried out with the homobifunctional, amine-reactive cross-linking reagent BS3 (D 0/D 12). The cross-linking data revealed an interaction of Psb28 with cytochrome b559, PsbE, and PsbF subunits of the photosystem II included in the RC47 subcomplex. Subsequent docking of Psb28 to the RC47 subcomplex of the photosystem II, which is known to interact with Psb28, was performed with the DOT 2.0 docking software. The resulting structures were validated by comparing the cross-linking data with the generated models. The final model of the protein complex is displayed in Fig. 8.3b, showing that Psb28 is interacting with PsbE and PsbF of the RC47 subcomplex as confirmed by the cross-linking data.

Currently, the majority of cross-linking studies is performed in-vitro as these studies allow a specific targeting of the proteins of interest. Deriving structural information on protein and protein complexes in-vivo is still a daunting task due to the enormous complexity of cellular samples. In many studies, cross-linking experiments are combined with immunoblotting. Examples include studies of the COX-2/mPGES complex and TS3-regulating proteins LcrG and LcrV (Henderson and Nilles 2017) as well as investigating the assembly state of α-synuclein (Corbille et al. 2016). The analysis of α-synuclein assemblies in living cells was performed by disuccinimidylglutarate (DSG) and dithiobis(succinimidyl)propionate (DSP). Both cross-linkers are NHS esters targeting amine groups in proteins, but they differ in spacer length. Additionally, the cross-linker DSP contains a disulfide bond that can be cleaved under reducing conditions (Corbille et al. 2016). Cross-linking was induced by adding the cross-linkers directly to the cell suspension, followed by disruption of the cells. Analysis of the complexes was then performed by immunoblotting using an anti-α-synuclein antibody revealing the presence of α-synuclein dimers and pentamers in the cells. This result was verified by the cleavable DSP cross-linker as the pentamer disappeared after reduction of the disulfide bond in DSP.

5 Identification of Protein-Protein Interaction Networks

The identification of protein-protein interaction networks is the key to understanding biological processes and cross-linking/MS can make here major contributions by identifying interacting proteins as well as defining their interaction sites. Often, the result of the cross-linking reaction is monitored via SDS-PAGE or immunoblotting, giving insights only into these interaction partners for which antibodies are available (Hetu et al. 2008; Maadi et al. 2017; Henderson and Nilles 2017). The combination of cross-linking with MS will give more detailed insights, giving a more comprehensive picture on protein interaction networks.

In-vitro MS based procedures usually include an affinity-based identification of protein interaction partners. The starting point for these studies is the immobilization of a bait protein on a matrix, followed by incubation with cell lysates or cellular fractions. Several washing steps are performed to remove non-interacting proteins, followed by enzymatic digestion to identify protein binding partners (Fig. 8.4a, upper panel). Unfortunately, the washing procedure can prevent the detection of weakly or transiently bound protein-protein interaction. Applying a cross-linking reagent for a covalent fixation of interacting proteins prior to the washing procedure prevents losing potential protein binding partners (Fig. 8.4a, lower panel). Afterwards, the enriched interacting proteins are enzymatically digested and analyzed by LC/MS/MS. Cross-linked peptides as well as non-cross-linked peptides are used to identify the interaction partners. Non-cross-linked peptides identify the binding proteins, while cross-linked peptides additionally yield information on the protein interfaces.

Fig. 8.4
figure 4

Elucidation of protein-protein interaction networks. (a) The immobilized bait protein is incubated with a cell lysate, followed by a washing procedure to remove the non-binding proteins. LC/MS/MS analysis identifies the interacting proteins. Upper panel: common affinity-based strategy, lower panel: cross-linking-based strategy. (b) Identified interaction partners of protein kinase D2 applying BS2G to capture interacting proteins by the strategy presented in a) lower panel. Figure is adapted with permission from (Häupl et al. 2016). (c) Workflow for the identification of protein-protein interaction network from murine mitochondria. Circles display proteins, lines indicate cross-links. The color depth of each dot is proportional to the frequency of the respective protein within the eleven samples. (Figure is adapted with permission from Schweppe et al. 2017)

Full size image

A combined affinity purification cross-linking/MS strategy was applied to identify partners of protein kinase D2 (PKD2) from Golgi preparations and whole cell lysates. For these studies, the external cross-linker BS2G (D 0/D 4) as well as the unnatural amino acids, photo-Leu and photo-Met, incorporated into proteins during translation in HeLa cells, were applied (Häupl et al. 2016, 2017). To investigate PKD2 interaction partners, glutathione-S-tranferase (GST)-tagged PKD2 was immobilized on GSH sepharose beads and incubated with cell lysate or a Golgi preparation. PKD2-interacting proteins were covalently bound by adding the amine-reactive cross-linker BS2G (D 0/D 4) to the reaction mixture or by inducing photo-cross-linking of the unnatural amino acids by UV-A irradiation. LC/MS/MS analysis allowed identifying the covalently fixed PKD2 interaction partners. The results obtained by the BS2G cross-linking are illustrated in Fig. 8.4b (Häupl et al. 2016). With the photo-reactive amino acids, similar PKD2 interaction partners were identified, but the complementary reactivity and shorter spacer length of the photo-reactive amino revealed additional interacting proteins (Häupl et al. 2017). A similar approach, exclusively based on BS2G (D 0/D 4), was employed for investigating protein interaction partners of tissue-type plasminogen activator (t-PA), an established tumor marker in various cancers (Bosse et al. 2016). Proteins secreted by erlotinib-sensitive (PC9) and erlotinib-resistant (PC9ER) non-small cell lung cancer (NSCLC) cells were investigated, indicating differences of t-PA interacting proteins between erlotinib-sensitive and -resistant cells.

To enable protein interaction partner studies in-vivo, cross-linking reagents are added to cell cultures or cell suspensions, which are crossing the cell membrane to react with target proteins within the cell (Weisbrod et al. 2013; de Jong et al. 2017). In case of photo-reactive amino acid incorporation, the reactive groups enabling cross-linking reactions are already incorporated during cell growth. Desired cross-linking is afterwards induced by exposure of the cells to UV-A light (Yang et al. 2016a). Subsequent proteolysis of whole cells or cell lysates results in enormously complex peptide mixtures that hamper a thorough identification of cross-linked peptides. Consequently, an enrichment of cross-linked peptides is mandatory before performing LC/MS/MS analyses. Affinity strategies can be applied if the protein of interest contains a tag for the specific isolation of the desired protein complexes (Walker-Gray et al. 2017). Alternatively, the biotin label is contained in the cross-linker. As described above, the biotin label can either be incorporated in the cross-linker itself (Tang and Bruce 2010; Tan et al. 2016; Yang et al. 2016b) or it is added after to the cross-linking reaction by a click-reaction (Nury et al. 2015). The latter approach is based on orthogonal chemistry strategies developed for proteomic analyses (Speers and Cravatt 2005; Weerapana et al. 2007).

Abovementioned PIR cross-linkers (Fig. 8.2c) have been used for in-vivo studies identifying protein-protein interactions in mitochondria – cell organelles that are comprised of more than 1000 proteins (Schweppe et al. 2017). Cross-linking experiments were conducted on active mitochondria isolated from mouse heart and allowed the identification of protein interaction partners as well as the 3D-structural investigation of the respective protein complexes. The PIR cross-linker used for this study was membrane-permeable, and comprised two NHS esters as reactive head groups as well as a biotin group for an enrichment of cross-links (Fig. 8.4c). After the cross-linking reaction, mitochondria were disrupted and the proteins were enzymatically digested. Cross-linked peptides were first fractionated by SCX and further enriched by affinity chromatography. Subsequent LC/MS/MS analysis by identified 327 proteins and 2427 cross-linked peptides, which additionally allowed gaining insights into the 3D-structures of selected protein complexes.

6 Conclusion

Cross-linking/MS has matured as a valuable technique in structural biology that complements existing techniques, such as X-ray crystallography, NMR spectroscopy, and cryo-EM. The major applications of the cross-linking/MS approach are to derive 3D-structural information of purified proteins and protein complexes, providing distance information for computational modeling studies, and to elucidate protein-protein interaction networks from cell lysates or even in intact cells. Although cross-linking of proteins in-vivo is still a challenging task, innovative approaches have been developed and are continuously being improved. A comprehensive analysis of protein-protein interaction networks in the cellular environment has become feasible.