Quantitative interactome analysis with chemical crosslinking and mass spectrometry

Abstract

Structural plasticity and dynamic protein-protein interactions are critical determinants of protein function within living systems. Quantitative chemical crosslinking with mass spectrometry (qXL-MS) is an emerging technology able to provide information on changes in protein conformations and interactions. Importantly, qXL-MS is applicable to complex biological systems, including living cells and tissues, thereby providing insights into proteins within their native environments. Here, we present an overview of recent technological developments and applications involving qXL-MS, including design and synthesis of isotope labeled crosslinkers, development of new LC-MS methodologies, and computational developments enabling interpretation of the data.

Graphical Abstract

1. Introduction

Chemical crosslinking of proteins with mass spectrometry (XL-MS), combined with other structural methods, has become a powerful technique of increasingly utility for study of protein conformations and interactomics. This approach is based on the reactivity of crosslinkers to specific protein sites – usually primary amines, including side chains of lysine residues and protein N-termini, identifying proximal residues and yielding information on protein structures and interactions. Quantitative crosslinking with mass spectrometry (qXL-MS) brings another and yet more complex level of information, providing insights into large-scale changes in protein interactomes with varying biological states, including system perturbations such as drug treatment, age, phenotype, or disease state[1–*3]. More importantly, qXL-MS can be applied directly to complex biological systems, such as living cells, tissues, or organelles, therefore providing information on changes in protein structural dynamics and interactions as they occur within a native environment[4,5].

A variety of strategies have been used for qXL-MS. Label-free qXL-MS[6] includes extracting MS¹ chromatographic peak areas, as well as the MS²-based quantitation employing parallel reaction monitoring (PRM)[7]. Isotopic labeling methods include stable isotope labeling by amino acids in cell culture (SILAC)[8] with XL-MS, and the use of light and heavy isotope-labeled crosslinkers[9], which enables large-scale qXL-MS studies. Isobaric labeling methods apply (i) isobaric reagents (e.g., TMT[10]) to label crosslinked peptides[11], or alternatively (ii) the isobaric quantitative protein interaction reporter (iqPIR) strategy, which includes isotope-encoded isobaric crosslinkers**[12] (Figure 1). Recently, a multiplexed version of the iqPIR has been developed, allowing for crosslinking and multiplexing of up to 6 different samples (Chavez et al., submitted). Moreover, recent developments in the publicly available XLinkDB platform enable the visualization and interpretation of qXL-MS results[13]. Software suitable and applied for qXL-MS include the MS¹-based MassChroQ[14], MaxQuant[15,16], pQuant[17], xTract[5], and Skyline[18]. Here we review recent developments in qXL-MS methodologies and applications described in the following sections: (2) isotopic labeling, (3) label-free, including (3.2) PRM, and (4) isobaric labeling (Figure 1).

Figure 1. Quantitative crosslinking with mass spectrometry (qXL-MS).

Isotope-labeled, label-free, PRM-based, and isobaric-labeled qXL-MS strategies are discussed in sections 2, 3, 3.2, and 4, respectively. iqPIR: isobaric quantitative protein interaction reporter.

2. Isotopic labeling

2.1. Isotopic labeling of cells with SILAC for qXL-MS

Isotopic qXL-MS strategies include the labeling of living cells with SILAC[8] followed by crosslinking. This approach uses the MS¹ information from light and heavy crosslinked peptide pairs to provide relative quantitation of crosslink levels between two samples. One advantage of SILAC over isotope-coded crosslinkers is that SILAC generates wider mass shifts in MS¹ scans, that may provide higher quantitative accuracy. Optimized SILAC isotope combinations for quantitative crosslinking applications have not yet been explored, but could further improve quantitative capabilities.

A SILAC-based qXL-MS illustrated that differences related to acquired chemoresistance to the active agent in the anticancer therapy Irinotecan were detectable at the interactome level, and these differences could in part, help explain activity differences in cells that contributed to the Irinotecan resistance phenotype[1]. A significant aspect of those results was that they revealed for the first time that in vivo crosslinking with mass spectrometry could provide quantitative insight on the interactome inside cells. Following this, Chavez et al. employed SILAC and PIR crosslinking of cancer cells to probe altered conformations and interactions resultant from treatment of cells with multiple Hsp90 inhibitors[2]. These results revealed interactome changes that were both drug concentration-dependent and drug mechanism-specific. These studies revealed drug-dependent changes in Hsp90, indicating compact conformation enrichment from the cellular ensemble which appears specific to N-terminal ATP pocket-targeting Hsp90 inhibitors. Moreover, the study identified changes in Hsp90 interactome with co-chaperones STIP1, Hsp70, CHRD1, and CDC37. This approach has also been applied by the same group to investigate changes in protein conformations and interactions in cancer cells after treatment with the mitotic inhibitor paclitaxel (PTX), using different concentrations of the drug*[3]. The qXL-MS results provided insights into dose-dependent changes in cytoskeleton organization (Figure 2), which caused stabilization of microtubules and arrest of mitosis, consistent with the PTX mechanism of action [19,20].

Figure 2. Quantitative crosslinking reveals PTX stabilized microtubules (MT).

(A) MT structure (PDB 3EDL) displayed as a molecular surface with a ribbon structure inset, illustrating the a-tubulin (maroon) and b-tubulin (gold) subunits. Crosslinked Lys residues are shown as green space-filled residues, with crosslinks displayed as colored lines connecting them (TBA1A K60-K370, orange; TBA4A K60-K370, purple; TBB5/TBB4B K216-TBA1A K326, blue; TBB4B K58-TBA1A K370, green; TBB5 K58-TBA1A K370, magenta). The non-exchangeable GTP binding site (N-site) is indicated by a yellow-highlighted region on a-tubulin. The exchangeable GTP binding site (E-site) is indicated by a cyan-colored region on b-tubulin. The PTX binding site on b-tubulin is red. (B) Protein and crosslink levels measured by SILAC for four tubulin isoforms (TBA1A, dark red dashed line; TBA4A, red dashed line; TBB4B, gray dashed line; TBB5, black dashed line). Crosslinks are colored the same as in (a). (C) Quantified levels of TBB5/TBB4B K216-TBA1A K326 and TBA1A K60-K370 with PTX (50, 100, and 500 nM), CA4 (5 nM), and STLC (5 mM). Error bars represent 95% confidence intervals for n = 6 replicate injections of 2 biological samples. Reproduced with permission from Chavez et al., 2019[3].

Yu et al. combined the Quantitative analysis of Tandem Affinity-purified in vivo crosslinked (X) protein complexes (QTAX)[21] strategy with SILAC and disuccinimidyl sulfoxide (DSSO) crosslinker[22] for a qXL-MS study of the human 26S proteasome, assessing interactomic changes in response to H₂0₂-induced oxidative stress*[23]. Briefly, cells expressing histidine/biotin- (HB-) tagged proteasome were grown in SILAC media, crosslinked in vivo with formaldehyde, and on-bead purified proteasomes were crosslinked in vitro with DSSO. MS³ spectra were used for the identification of 746 crosslinks, and MS¹ signal used for SILAC-based quantitation of 343 crosslinks (~46% of total links). While SILAC-based qXL-MS has enabled initial interactome dynamic measurements, in some cases even inside cells, the challenges associated with MS¹ signal quantitation limit the efficiency of crosslink quantitation for complex systems. For instance, of the 3,323 crosslinked peptides detected in Chavez et al., 2016[2], only 2,582 were quantified (77,7%) and even fewer (559 crosslinks, ~17%) were quantified across all 5 different drug concentrations. This results from issues of co-eluting species that congest MS¹ signals and contaminate extracted ion signals used for quantitation. Filtering of all forward and reverse SILAC-qXL-MS quantitation data for 95% confidence limits eliminates these spurious results, significantly reducing the depth of quantitative data. One strategy that could potentially improve upon the issue of MS¹ signal congestion would be to further separate crosslinked peptides using gas phase fractionation techniques such as high-field asymmetric waveform ion mobility spectrometry (FAIMS)[24].

2.2. Isotopic crosslinkers

Isotope-labeled crosslinkers for qXL-MS include DSS-d0d12*[25], BDP-NHP-d0d8[26], CBDPS*[27], Leiker[28], DMDSSO-d0d10[29], BS3-d0d4[30] and BS3-d0d12[31,32], and DSBU*[33] (Table 1), which include variable numbers of deuterium atoms as the heavy labels. These isotope-encoded crosslinkers are useful for binary comparisons and most of them have been applied for in vitro studies, where extreme complexity in the MS¹ signal is not likely to be such an issue [34–38]. The Rappsilber group applied BS3-d0d4 in solution to investigate conformational changes in the human complement protein C3 in comparison to its activated cleavage product C3b[34]. The same group developed a computational tool for analysis of isotope-coded crosslinking data, called XiQ[39]. BS3-d0d4 was also applied in vitro to study F₁F_O-ATPase isolated from chloroplasts[35], with conformational changes in the catalytic interface of the enzyme related to phosphorylation. Furthermore, Boelt et al. used BS3-d0d4 in solution to investigate conformational changes of calreticulin, a protein part of the endoplasmatic reticulum that regulates Ca²⁺ homeostasis[38]. More recently, BS3-d0d12 was applied to study the role of LEM2 in the reformation of nuclear envelope during cell division[40]. The crosslinking results indicated specific sites of interaction between CHMP7 and the WH domain of LEM2 that was proposed to form a macromolecular O-ring seal at the confluence between membranes, chromatin and the spindle. The qXL-MS results agreed with those from other experiments used in the study, thus providing insights on protein conformation and interactions of isolated complex samples. As with SILAC-based qXL-MS, quantitative measurements with isotopic-labeled crosslinkers also rely on MS¹ information for crosslink quantitation and are subject to similar peak assignment challenges, particularly with complex large scale qXL-MS applications. Additionally, peak assignment can be further complicated by retention time shift between light and heavy deuterium-labeled peptides during LC separation. To date, all applications of isotope-labeled crosslinkers have utilized a binary comparison between a light and heavy pair. Conceptually, it should be possible to expand beyond binary situations, allowing for a greater number of comparisons to be made with the limitation of increasing the MS¹ signal complexity by N as well as diluting the signal of any crosslinked product by 1/N (where N= number of isotopologue crosslinkers).

Table 1.

Isotope-coded crosslinkers for qXL-MS. Dashed line indicates the cleavage site.

Crosslinker	Structure	Feature	Reference
DSS	X indicates H or D isotope for d0-DSS (X = H) or d12-DSS (X = D)	Vendor: Creative Molecules	Khakzad et al.1
		Formula (X = H): C16H20N2O6
		MW (X = H): 368.34
BS3	X indicates H or D isotope for d0-BS3 (X = H) or d4-BS3/d12-BS3 (X = D)	Vendor: ThermoFisher Creative Molecules	Schmidt et al.2 Linden et al.3 Appen et al.4
		Formula: C16H18N2Na2O14S2
		MW: 572.43
DMDSSO	X indicates H or D isotope for d0-DMDSSO (X = H) or d10-DMDSSO (X = D)	MS cleavable	Yu et al.5
		Formula: C14H16N2O9S
		MW: 388.35
DSBU	X indicates H or D isotope for d0-DSBU (X = H) or d12-DSBU (X = D)	MS cleavable	Ihling et al.6
		Formula: C17H22N4O9
		MW: 426.38
CBDPS	X indicates H or D isotope for d0-CBDPS (X = H) or d8-CBDPS (X = D)	MS cleavable	Makepeace et al.7
		Biotin tagged
		Formula: C27H33N9O10S3
		MW: 739.80
Leiker (bAL2)	X indicates H or D isotope for d0-bAL2 (X = H) or d6-bAL2 (X = D)	Photo cleavable	Tan et al.8
		Biotin tagged
		Formula: C43H52N8O13S
		MW: 920.98
PIR BDP-NHP	X indicates H or D isotope for d0-PIR (X = H) or d8-PIR (X = D)	MS cleavable	Zhong et al.9
		Biotin tagged
		Peptide synthesis
		Formula: C64H79N13O22S
		MW: 1414.45
2-plex iqPIR BDP-NHP	* indicates the position can be isotope-coded with C13	MS cleavable	Chavez et al.10
		Biotin tagged
		Peptide synthesis
		Formula: C68H85N15O24S
		MW: 1532.55
6-plex iqPIR BDP-NHP	* indicates the position can be isotope-coded with C13 or N15	MS cleavable	Chavez et al. (submitted)
		Biotin tagged
		Peptide synthesis
		Formula: C68H85N15O24S
		MW: 1546.42

3. Label-free qXL-MS

As with traditional quantitative proteomics, another strategy for qXL-MS is to utilize label-free quantitation (LFQ). Benefits of LFQ include its compatibility with nearly any crosslinking reagent, avoiding costly heavy isotope labeled reagents, and allowing for higher level comparisons. A consequence of LFQ is that each sample needs to be prepared and analyzed separately, requiring increased and reproducible sample preparation and LC-MS instrument time and being subject to experimental variabilities introduced at each step along the process. Müller et al. evaluated the quantitative reproducibility of a MS¹-based LFQ qXL-MS approach using BS3 crosslinked human serum albumin (HSA) samples and found more variability resulted from comparison across different crosslinking reactions than from LC-MS injections[6]. The qXL-MS strategies for label-free and isotope labeled BS3 from the Rappsilber laboratory were summarized in a recent protocol[42]. MS¹-based LFQ of crosslinks and acetylated peptides revealed that acetylation disrupts dimer formation and decreases the activity of the muscle isoform of creatine kinase [43]. MS¹-based LFQ was used to reveal the conformational changes to the peroxisome proliferator-activated receptors gamma (PPARγ) upon binding the antagonist SR11023[44]. To group multiple protein conformations detected in different biological samples, Kurt et al. developed a software that clusters crosslink identifications according to their MS¹-based quantitative profile across multiple samples, the QUIN-XL[45].

3.1. DIA-QCLMS

The majority of qXL-MS studies have used a DDA approach. However, data-independent acquisition (DIA) qXL-MS methods have also been explored. Müller et al. demonstrated a LFQ DIA qXL-MS approach at the MS¹ and MS² levels using samples consisting of a mixture of 7 proteins crosslinked with BS3*[46]. The authors discussed that, despite the ratio compression in MS¹/MS² spectra observed for complex mixtures, the Spectronaut-based (Biognosis) quantitation of crosslinks showed increased accuracy and reproducibility. The Rappsilber laboratory further demonstrated use of the DIA qXL-MS protocol[47] to detect pH-dependent conformational changes in photo-crosslinked HSA and cytochrome c[48].

3.2. Targeted PRM

Targeted quantitation offers multiple benefits to qXL-MS including increased reproducibility, sensitivity, precision and accuracy. Furthermore, targeted analysis is applicable with both LFQ approaches or those using isotope labeling. A limitation of targeted analysis is that it often requires some a priori knowledge of the crosslinked analytes of interest, requiring development of special LC-MS methods devoted to a select number of analytes, making it a relatively low throughput strategy. The first application of parallel reaction monitoring (PRM) qXL-MS was applied in a cross-laboratory study which utilized Skyline[18] for method sharing and data analysis[7]. PRM qXL-MS was also used to help elucidate important conformational changes in heat shock protein 90 (Hsp90) that occur with nucleotide binding, interaction with co-chaperone Aha1, and with a phosphomimetic mutant Y313E [49]. Gutierrez et al. utilized PRM qXL-MS with three different MS-cleavable crosslinkers to detect structural dynamics of the COP9 signalosome*[50]. Mehnert et al. used PRM qXL-MS as part of a multilayered proteomic workflow to detect topological changes associated with cancer mutations on kinase complexes[51]. As with traditional quantitative proteomics, when using DIA and PRM for quantitation of crosslinked peptides it is important to implement good practices to ensure accurate quantitation of the correct analyte[52,53]. These include use of proteotypic peptides, quantifying multiple high mass accuracy matching fragment ions, ideally originating from both peptides, and ensuring accurate retention time matching. Spectral library searching can also be useful to confirm identification of quantified species[54].

4. Isobaric labeling

4.1. TMT labeling for qXL-MS

A qXL-MS multiplexed method based on labeling of crosslinked peptides with isobaric mass tags, i.e., TMT, was developed by Yu et al., called Quantitation of Multiplexed Isobaric-labeled cross (X)-linked peptides (QMIX)[11]. The authors used DSSO to crosslink cytochrome c, followed by proteolytic digestion of separate samples. Digested samples were then labeled with a binary set of TMT reagents (126 or 127 reporter) followed by multiplexing samples, and LC-MS analysis of crosslinked peptides. To eliminate the signal interference at the MS¹ level and compression of measured TMT ratios due to co-isolation of precursor ions[55], the quantitation of crosslinks was done at the MS³ level. Hence, the QMIX approach benefits from advanced mass spectrometers with higher sensitivity MS³ capabilities.

4.2. Isobaric quantitative PIR (iqPIR)

Recently, isobaric quantitative Protein Interaction Reporter (iqPIR) crosslinking technology was developed and demonstrated in our laboratory**[12]. iqPIR crosslinkers are synthesized using ¹³C and ¹⁵N isotope labels, enabling the quantitation of crosslinked peptide pairs using the relative abundance of multiple isotope fragment ions unique to each crosslink in the MS² spectra. This strategy eliminates any retention time shifts – as it happens with deuterated crosslinkers (Section 2.2), and ratio compression of the reporter ion signal by quantifying fragment ions that retain an isotope encoded portion of the tag, analogous to the use of complement reporter ions with TMT[55] or the EASY-tag approach[56] used in traditional quantitative proteomics. Fragmentation of iqPIR crosslinked peptides generates a number of quantifiable ions, including the released intact peptides, as well as b-type and y-type backbone fragment ions containing the crosslinked residue. The iqPIR, as well as other PIR-based crosslinkers[57], contain MS-cleavable bonds, affinity tags, and membrane permeability for the crosslinking of live cells, tissues, and isolated organelles. These features enable isotope-encoded crosslinking of proteins in complex biological systems with one single labeling step, thus eliminating the need for additional chemical labeling reactions and cleanup steps. Moreover, the biotin affinity tag enables crosslink enrichment, which combined with the quantitative features, allows complex cellular applications.

iqPIR was pursued to enable multiplex interactome quantitation but was initially demonstrated with binary comparisons with 2 different isotope-coded crosslinkers**[12]. Recently, a multiplexed strategy was developed with 6 different iqPIR crosslinkers, called 6-plex iqPIR (Chavez et al., submitted). Because the 6-plex iqPIR crosslinkers are isobaric, the precursor ion signal from a crosslinked peptide pair originating from 6 different multiplexed samples overlap in MS¹ but generate unique fragments in MS² (Figure 3), which are further used for crosslink quantitation (Chavez et al., submitted). The 6-plex iqPIR strategy was demonstrated in vivo on MCF-7 cells treated with five different Hsp90 inhibitors, revealing drug class specific interactome dynamics from qXL-MS analysis of a multiplexed sample (Wippel et al., submitted). From 1,756 crosslinks identified in this study, 1,650 were quantified (~94%), and a total of 1,257 crosslinks were quantified across all 6 channels (~71%). These results represent an improvement in crosslinking quantitation if compared with the values discussed above from Chavez et al.[2], where 77,7% of crosslinked peptide pairs were quantified, and 17% quantified across all drug concentrations.

Figure 3. 6-plex iqPIR for multiplexed qXL-MS.

Representation of the 6-plex iqPIR qXL-MS LC-MS pipeline. 1) Crosslinked peptide pairs are separated by liquid chromatography and analyzed with high resolution MS/MS; 2) MS¹ spectrum highlighting the isotope envelope of the precursor ion, embedding all 6 crosslinked peptides; 3) MS² spectrum with highlighted fragment ions from peptides A (b/y ions in green) and B (b/y ions in purple), and isotope envelopes from released reporter ions and backbone peptides A and B, with the peaks colored according to the corresponding channel/Hsp90 inhibitor (shown as an example of environmental conditions here); 4) Relationship check between reporter ion and peptides A and B, comparing the monoisotopic neutral masses with the experimental masses observed (with maximum 10 ppm error allowed)[58]. After Mango, in silico analysis continues with Comet[59] for peptide search of 6-plexed fragment ions, followed by validation of identified crosslinks, and quantitation of 6-plex crosslinks (Chavez et al., in preparation). Reproduced with permission from Wippel et al., submitted.

CONCLUSIONS

A challenge of systems-level biology is to better understand how molecules act and interact to confer function within a crowded cellular environment. XL-MS is an emerging tool that can probe protein conformations and interactions to gain unique insight on molecular mechanisms and biological pathways. Just as seeing how the components of complex machines or instruments work together to function can improve fundamental understanding, visualizing how protein conformations and interactions change within a system can help better understand large-scale protein biological function. Thus, perturbations in the system are frequently necessary to allow comparative analysis among different conditions, where qXL-MS can uniquely reveal protein and interactome dynamics in response to environmental changes. While many successful qXL-MS studies have used MS¹ signal levels for crosslink quantitation or from targeted MS² spectra, isobaric approaches based on MS² spectral quantitation offer increased quantitation efficiency as well as multiplexed quantitation. Recent developments have allowed for the crosslinking, multiplexing, and quantitation of up to 6 different biological samples, with mixing at the protein level. In parallel, computational improvements are needed to allow the visualization of qXL-MS data. Although far from comprehensive, quantitative XL-MS studies and the advancements made so far are providing unique insight on protein, complex and interactome dynamics.

HIGHLIGHTS.

• Quantitative chemical crosslinking provides information on changes in protein interactomics;

• Overview of recent technological developments involving qXL-MS;

• Applications of qXL-MS in biological systems with varying conditions.

ABBREVIATIONS

BS3

bis(sulfosuccinimidyl)suberate

CBDPS

CyanurBiotinDimercaptoPropionylSuccinimide

DSBU

disuccinimidyl dibutryic urea

DSS

disuccinimidyl suberate

iqPIR

isobaric quantitative protein interaction reporter

LC-MS

liquid chromatography – mass spectrometry

MS

mass spectrometry

NHP

N-hyrdoxyphthalimide

PIR

protein interaction reporter

PRM

parallel reaction monitoring

PTX

paclitaxel

SILAC

stable isotope labeling by amino acids in cell culture

TMT

tandem mass tags

qXL-MS

quantitative chemical crosslinking with mass spectrometry