“Zero-length” cross-linking in solid state as an approach for analysis of protein–protein interactions

Abstract

We have developed a new approach for the analysis of interacting interfaces in protein complexes and protein quaternary structure based on cross-linking in the solid state. Protein complexes are freeze-dried under vacuum, and cross-links are introduced in the solid phase by dehydrating the protein in a nonaqueous solvent creating peptide bonds between amino and carboxyl groups of the interacting peptides. Cross-linked proteins are digested into peptides with trypsin in both H₂¹⁶O and H₂¹⁸O and then readily distinguished in mass spectra by characteristic 8 atomic mass unit (amu) shifts reflecting incorporation of two ¹⁸O atoms into each C terminus of proteolytic peptides. Computer analysis of mass spectrometry (MS) and MS/MS data is used to identify the cross-linked peptides. We demonstrated specificity and reproducibility of our method by cross-linking homo-oligomeric protein complexes of glutathione-S-transferase (GST) from Schistosoma japonicum alone or in a mixture of many other proteins. Identified cross-links were predominantly of amide origin, but six esters and thioesters were also found. The cross-linked peptides were validated against the GST monomer and dimer X-ray structures and by experimental (MS/MS) analyses. Some of the identified cross-links matched interacting peptides in the native 3D structure of GST, indicating that the structure of GST and its oligomeric complex remained primarily intact after freeze-drying. The pattern of oligomeric GST obtained in solid state was the same as that obtained in solution by Ru (II) Bpy₃²⁺ catalyzed, oxidative “zero-length” cross-linking, confirming that it is feasible to use our strategy for analyzing the molecular interfaces of interacting proteins or peptides.

One of the most recognized approaches for uncovering protein function in cellular machinery is to map the protein–protein interactions in protein complexes, the molecular “machines” of the cell (Figeys 2002; Dziembowski and Seraphin 2004). Current pull-down approaches identify protein interactors but do not provide clues about the structures of protein complexes. Detailed knowledge about interacting protein interfaces are of outstanding importance for understanding protein functions in the cell (Sinz 2003).

Analytical techniques to study the topology of protein complexes using chemical cross-linking in combination with mass spectrometry (MS) have recently been developed (Sinz 2003). The ultimate goal of an intermolecular cross-linking experiment is to gain knowledge about how proteins physically contact each other (Back et al. 2003). The intra-molecular cross-links derived from these experiments also provide information, in the form of distance constraints, about the protein fold, the protein tertiary structure (Young et al. 2000), and, in the context of protein–protein interactions, the protein quaternary structure.

A variety of chemical cross-linking approaches were recently reviewed (Kluger and Alagic 2004). Most approaches use bifunctional or trifunctional linkers combined with MS to determine cross-linked peptides (Sinz 2003). Cross-linking using chemical linkers, however, complicates the interpretation of obtained mass spectra. The reagents used frequently exhibit low solubility and stability in water and are prone to nonspecific chemical reactions. The cross-linkers introduce reagent-specific spatial constraints for cross-linking due to their length and conformation so, on occasion, only limited contact points can be identified.

“Zero-length” cross-linking with carbodiimides creates an amido bond between the amino and the carboxyl amino acid residues within interacting protein surfaces (Grabarek and Gergely 1990). This approach represents an alternative to the above methods because it allows direct evaluation of contact interactions between peptides. However, reactive esters, the key intermediates in this cross-linking, are too unstable in water to make the yield of this cross-linking reaction high enough for reliable analysis. A new approach, zero-length photo cross-linking of tyrosine residues, which is mediated by Ru(II)(bpy)₃²⁺ and ammonium persulfate, was shown recently to be a potential method for mapping protein interactions in multiprotein complexes (Fancy and Kodadek 1999; Denison and Kodadek 2004). However, this method requires the tedious isolation of cross-linked peptides and the presence of ammonium persulfate in solution, which could generate oxidative artifacts (e.g., cross-links via disulfide formation). Clearly, a generic method for determining protein interactions is needed that would allow the analysis of “intact” protein complexes without the current obstacles of the adverse effects from solvents or linkers and the tedious and inefficient isolation of cross-linked proteins and/or peptides.

Strategy

Our strategy (Scheme 1) represents a generic method for analyzing interacting protein interfaces that combines multiple recent findings. The first finding is the observation of Simons et al. (2002), who demonstrated the cross-linking of proteins in freeze-dried protein samples in the absence of any chemical reagent. They showed that heat and vacuum can lead to the dehydration of proteins and the formation of inter- and intramolecular amide cross-links between amine and carboxyl groups. The second finding is a method for identifying cross-linked proteins in protein complexes that uses the replacement of two C-terminal oxygens with ¹⁸O catalyzed by trypsin (Back et al. 2002). If the replacement is complete, two C-terminal ¹⁸O oxygens are introduced into each carboxyl group of a peptide, providing a 4-amu (atomic mass unit) shift in the MS spectra for a single peptide and an 8-amu shift for cross-linked peptides. The third method incorporated into our strategy is the computational analysis of the MS and MS/MS spectrum of cross-linked peptides using software developed by Malin Young at Sandia National Laboratories (http://roswell.ca.sandia.gov/~mmyoung/index.html). The analysis uses two programs. The first program, Automated Assignment Program (ASAP) (Young et al. 2000), analyzes the MS spectra of cross-linked proteins and identifies cross-linked peptides. The second program, MS2Assign, analyzes the MS/MS spectra of cross-linked peptides (Schilling et al. 2003).

Scheme 1.

Strategy for direct identification of cross-linked peptides in intact protein complexes by using solid-state cross-linking combined with MS analysis.

Our strategy is based on the premise that the intact state of a protein complex of interest can be preserved by freeze-drying followed by specific cross-linking at the contact interfaces of proteins via dehydration. We improved the yield and the specificity of the dehydration-mediated cross-linking by introducing carbodiimide in acetone instead of heat and vacuum. The cross-linked peptides are labeled with ¹⁸O and directly analyzed by capillary LC (liquid chromatography)-MS without fractionation or separation prior to analysis. The cross-linked peptides are selected and then identified using the ASAP program. Finally, the identity of the cross-linked peptides is confirmed by MS/MS analysis of the 8-amu-shifted peptides followed by analysis using the MS2Assign program.

In this work we provide experimental support for the premise that protein complexes remain substantially intact after freeze-drying. We show that the dehydration of protein complexes in freeze-dried samples leads to cross-linking and that the formation of amide bonds between the interacting molecules can be catalyzed by carbodiimide in an organic solvent. Finally, we show that this cross-linking is specific and demonstrate the proof of principle by identifying 27 intramolecular and two intermolecular cross-links in homo-oligomeric glutathione-S-transferase (GST), some of them corresponding to the native structure of the protein.

Materials and methods

Materials and reagents

GST was purchased from Sigma Chemical Co. The GST sequence was taken from SwissProt, accession no. P08515. The concentration of the stock solution was determined by a Coomassie Plus Protein Assay (Pierce Biotechnology). All reagents were prepared immediately prior to use. The water used in all components of the procedure was MilliQ-type water. All chemicals used, including ammonium bicarbonate, acetonitrile, methanol, iodoacetamide, ditiothreitol, glutathione-reduced (GSH), N-ethyl-N′-(3-dimethylaminopropyl) carbodiimide (EDC), and 1-chlo-ro-2,4-dinitrobenzene (CDNB) were analytical grade purchased from Sigma. Trypsin was purchased from Pro-mega, 4%–12% Bis-Tris gels; sample buffer, nitrocellulose, and molecular weight markers were from Invitrogen. The isotopically enriched water used for the digestion procedures was 95% ¹⁸O (Spectra Stable Isotopes).

Cross-linking and digestion

Protein samples were dialyzed against 50 mM ammonium bicarbonate, flash-frozen in liquid nitrogen, and lyophilized for 2 h. Cross-linking reactions were initiated by dehydration in acetone, heat and vacuum cycles, and various concentrations of EDC. EDC was incubated with dry sample in acetone for 1 h at room temperature. The reactions were stopped by addition of 100 μL of 50 mM ammonium bicarbonate (pH 8.0). The solution was freeze-dried again under vacuum for additional 2 h. For gel analysis, the protein samples were dissolved in Laemmli reducing buffer (Invitrogen) to a final protein concentration of 1μg/μL, boiled for 5 min, and analyzed by 4%–12% Bis-Tris gels SDS-PAGE. Proteins in gels were visualized with GelCode blue stain (Pierce Biotechnology) and/or In Vision His-Tag (Invitrogen) fluorescence gel stain specific for hexahistidine affinity tags.

Cross-linked GST was subjected to complete denaturation by using procedures reviewed by Back et al. (2002) prior to proteolytic digestion. Cross-linked protein was incubated for 1.5 h at 60°C in 10 mM DTT and 8 M urea and subsequently alkylated with 50 mM iodoacetamide in the dark for 1 h at 37°C. After eightfold dilution with 50 mM ammonium bicarbonate (pH 8.3), the samples were digested with sequencing grade trypsin in a ratio of 1:50 (w/w) at 37°C for 24 h. The digested samples were then cleaned with PepClean C₁₈ columns (Pierce), dried in vacuum, and stored at −80°C. The digestion procedures were performed either in H₂¹⁶O or in H₂¹⁸O.

GST activity

GST activity using CDNB and GSH substrates was determined by monitoring changes in absorbance at 340 nm (Habig et al. 1974) by using an Agilent 8453E UV-Visible spectroscopy system (Agilent). Assays were conducted at room temperature in 0.1 M potassium phosphate (pH 6.5). The concentrations of GSH and CDNB were 1 mM and 0.5 mM, respectively. All assays were linear functions of protein concentration and of time for at least 3.5 min.

LC-MS analysis

LC-MS analysis was performed by using a PNNL-developed high-efficiency capillary LC system in conjunction with a Finnigan model LTQ FT ion trap-FTICR MS (ThermoQuest) with electrospray ionization (ESI). A high-performance liquid chromatography (HPLC) column was coupled to the mass spectrometer by using an in-house manufactured interface. No sheath gas or makeup liquid was used. The heated capillary temperature and spray voltage were 200°C and 2.2 kV, respectively. The capillary LC system used a pair of 100-mL Isco model 100DM syringe pumps and a series D controller (Isco Inc.); an in-house manufactured stir-bar style mobile phase mixer (2.5 mL volume); two four-port, two-position Valco valves for mobile phase and capillary column selection; and one six-port, two-position Valco valve equipped with a 10 μL sample loop for manual injections. The mixer and valves were mounted on an in-house manufactured rack assembly that was custom-fit to a PAL autosampler (Leap Technologies) for unattended routine analysis. Reversed-phase, capillary HPLC columns were manufactured in-house by slurry packing 5 μm Jupiter C18 stationary phase resin (Phenomenex) into a 60-cm length of 360-μm outside diameter × 150-μm inside diameter fused-silica, capillary tubing (Poly-micro Technologies) incorporating a 2-μm retaining screen in a 1/16-inch capillary-bore union (Valco Intruments). The mobile phase consisted of 0.2% acetic acid and 0.05% trifluoroacetic acid (TFA) in water (A) and 0.1% TFA in 90% acetonitrile and 10% water (B). The mobile phase was degassed with an in-line Alltech vacuum degasser. The LC system was equilibrated at 5000 psi with 100% mobile phase A for initial starting conditions. The mobile phase selection valve was switched from position A to position B at 20 min after injection, creating an exponential gradient as mobile phase B displaced A in the mixer. An ~5-cm length of 360 inside diameter fused-silica tubing packed with 5 μm C18 was used to split ~25 μL/min of flow before the injection valve. The split flow controls gradient speed under conditions of constant-pressure operation. Flow through the capillary LC column was ~1.8 μL/min when equilibrated to 100% mobile phase A. MS and MS/MS mass spectra were acquired by using Xcalib software (ThermoQuest)

Identification of cross-linked peptides

Distinguishing the 8-amu-shifted, cross-linked peptides in MS spectra was performed by using in-house software developed for analyzing differentially labeled peptides (Masselon et al. 2005). The 8-amu-shifted, cross-linked peptides were then identified by using ASAP, the latest version of ASAP (Links), and the MS2Assign programs, developed by Malin Young at Sandia National Laboratories (http://roswell.ca.sandia.gov/~mmyoung/index.html). These programs required the protein sequence, the MS peak list, the charge state, the mass error limit, the proteolytic enzyme, and the chemical modification, if any. For each m/z value, ASAP searches the library for masses within the error limit, in this case ± 100 ppm. MS2Assign was applied to predict the possible fragments of each cross-linked peptide and its fragmentation ions.

Structural analysis

The monomer (Protein Data Bank [PDB] ID 1GNE) and dimer (PDB ID 1Y6E) structures of GST from Schistosoma japonicum were used for comparison of cross-links to X-ray-determined structural models. Distances between cross-linked amino acids were measured by using the visual molecular dynamics (VMD) program (Humphrey et al. 1996) and the built-in Tcl/Tk scripting commands.

Results

Acetone and carbodiimides facilitate zero-length cross-linking of freeze-dried protein complexes by dehydration

To demonstrate the principle of our approach, cross-linking of freeze-dried GST (Schistosoma japonicum) and urease (Jack beans) was performed by suspending lyophilized purified proteins in acetone, then drying the suspension in a vacuum enhanced by heat; this cycle of suspension in acetone followed by drying in a heated vacuum was repeated multiple times (Fig. 1). Indeed, in agreement with the experiment of Simons et al. (2002), proteins that form oligomers in their native state were covalently cross-linked by heating in a vacuum. The investigators, however, used water between the cycles and heated the proteins to temperatures up to 85°C for 24 h. This method could be detrimental to the protein complex and native protein structures and could lead to nonspecific cross-linking.

Figure 1.

Acetone/heat/vacuum (AHD) cycles induce cross-linking of GST and urease. Proteins (10 μg) were freeze-dried in 10 mM ammonium bicarbonate (pH 7.5) and then suspended in 100 μL acetone, and the suspension was dried at 60°C in a vacuum of 4.6 mtorr for 30 min. This AHD cycle was then repeated up to three times. The proteins were analyzed by 4%–12% gradient SDS-PAGE.

We discovered that using acetone, instead of water between the cycles, prevented dissociation of the protein complexes, thus significantly increasing the cross-linking yield. Also, the cross-linking time was shortened to 30 min at 60°C. We obtained similar results with other homooligomeric protein complexes such as egg lysozyme, oxidoreductase SO3120 from Shewanella oneidensis, and bovine serum albumin (data not shown). Other organic solvents such as dioxane, methanol, or ethanol were also used but resulted in lower cross-linking yields.

We further increased the yield of cross-linked GST by adding EDC, which facilitated the zero-length cross-linking by dehydration in the solid state. Figure 2 shows the SDS-PAGE separation of GST cross-linked by different concentrations of EDC in acetone. A distinctive dimer (50.74 kDa) above the GST monomer (25.37 kDa) appeared on the SDS-PAGE in the presence of EDC (Fig. 2A). We observed a similar effect by using cyclohexyl carbodiimide (data not shown). Higher concentrations of carbodiimide evenly decreased the solubility of all cross-linked products. We hypothesize that the decrease in solubility was most likely due to the increased intramolecular cross-linking. Last, but not least, the carbodiimide-mediated, solid-state cross-linking in organic solvent resulted in defined products that are better distinguished in gel compared with the cross-linking products prepared by the dehydration using heat and vacuum.

Figure 2.

Cross-linking of GST with EDC in acetone. (A) SDS-PAGE of cross-linked GST. (Lane 1) Molecular weight marker; (lane 2) un-cross-linked GST; (lane 3) freeze-dried GST in acetone; (lanes 4–6) freeze-dried GST preincubated with 0.1, 1, 5 mM EDC in acetone for 30 min at room temperature. (B) Enzyme activity of GST as a function of cross-linking method.

Freeze-drying itself showed only a marginal effect on the GST activity, indicating that the native protein structure was not substantially affected, the structural changes were reversible (Fig. 2B), or both. The GST activity was indirectly proportional to the degree of zero-length cross-linking. This suggests that intramolecular cross-links occur in addition to intermolecular cross-links, and that some of the cross-links may occur between essential amino acids of the active site.

Zero-length cross-linking in solid state is specific

Specificity of the zero-length cross-linking in the solid state is fundamental to its use as the preferred cross-linking method for studying the structure of protein complexes. Therefore, we mixed the purified homo-oligomeric GST protein complexes with a 100-fold excess of many other unrelated proteins (Escherichia coli lysate) and cross-linked the mixture after freeze-drying. The cross-linked products were analyzed via Western blot probing for GST (Fig. 3). The presence of the crude lysate did not affect the cross-linking pattern of the purified GST complex, indicating that no nonspecific proteins from the crude E. coli lysate cross-linked to GST in its solid state. The somewhat lower efficiency of GST cross-linking in the presence of the crude cell lysate compared with the purified protein can be attributed to the higher consumption of EDC in the presence of the large excess of non-specific proteins. We obtained similar results with other homooligomers (e.g., oxidoreductase from Shewanella oneidensis) (data not shown). These results showed that the zero-length cross-linking of GST by dehydration in solid state is specific. Only proteins that specifically interact in solution cross-linked in the solid state.

Figure 3.

GST is specifically cross-linked in its solid state even in the presence of a large excess of other proteins. GST from Schistosoma japonicum was freeze-dried in the absence or the presence of a crude Escherichia coli lysate (1:100 [w/w]), cross-linked by dehydration with EDC in acetone, and analyzed by SDS-PAGE followed by Western blot analysis using rabbit anti-GST polyclonal antibodies.

Efficiency of ¹⁸O labeling

Cross-linked peptides were distinguished by an 8-amu shift (vs. a 4-amu shift from non-cross-linked peptides) obtained after the ¹⁸O isotope exchange of the C-terminal carboxyl groups catalyzed by trypsin as described in the Materials and Methods section. Doubly charged cross-linked peptides will produce a 4-amu shift in the spectra. Figure 4 shows an example of a MS spectrum of a doubly charged cross-linked peptide with peaks separated by 4 amu. Peaks caused by non-cross-linked peptides exhibited only their normal isotopic distribution. A ¹⁶O/¹⁸O ratio of nearly 1:1 of the shifted peptide compared with the control indicated that ~100% ¹⁸O labeling was achieved (the labeled and the unlabeled samples were mixed in a 1:1 ratio before the MS measurement). The use of H₂¹⁸O in the digestion process represents a general strategy for introducing stable isotope labels into cross-linked peptides.

Figure 4.

Mass spectra of a doubly charged cross-linked peptide YLKSSK-YEEHLYERDEGDK showing ¹⁶O/¹⁸O ratio. As a result of the incorporation of four ¹⁸O atoms into the peptide C termini, 4-amu shifts were recorded for doubly charged masses at m/z of 1195.1⁺².

MS analysis

Two separate experiments were performed by using the ThermoFinnigan LTQ FT MS. First, a conventional LC-MS experiment was used to construct a selection list of precursor ions of interest for the subsequent MS/MS experiment. A two-dimensional representation of an LC-MS analysis is shown in Figure 5A. The selection list was created by using in-house-developed algorithms and software tools. Second, to extract and produce reliable measurements for the peptides detected Unique Mass Class (UMC) species (Masselon et al. 2005) were first distinguished. A UMC was composed of multiple measurements of the same ion species observed through multiple contiguous spectra (i.e., a species eluting from the LC separation); one entity attributed with mass defined as the median mass of its member’s masses, abundance, defined as the sum of member’s abundance, charge state, and elution range. Once established, coeluting UMCs were compared with the expected mass difference of 8 amu that is attributed to cross-linked peptide pairs observed in both ¹⁶O and ¹⁸O isotopic versions (see example in Fig. 5B). Based on this analysis, 172 of 8-amu difference pairs potentially attributable to cross-linking and 1269 of 4-amu difference pairs potentially attributable to single peptides were found. The 172 of 8-amu difference UMC pairs and the most abundant ions from both UMCs were chosen for targeted LC-MS/MS analysis.

Figure 5.

(A) Two-dimensional representation of the LC-MS experiment used to construct the selected list for targeted MSMS experiment. (B) Illustration of UMC pair with 8-Da mass difference. Size of the spot is proportional to the ion intensity.

Identification of cross-linked peptides using ASAP

To identify the cross-linked peptides from the selection list, the peptides were analyzed with the ASAP program (Young et al. 2000), initially applying a highly conservative mass cutoff of 100 ppm. The input to ASAP required the GST sequence, the amino acid modification table, and a text file containing a list of m/z (monoisotopic) values for the 8-amu-shift peaks. Knowing the chemistry of EDC-catalyzed dehydration, we assumed that the most likely cross-links occur between the carboxyl groups of Asp or Glu and the amino group of Lys forming a peptide bond between the interacting amino acid residues. Also, we observed that nearly all free amines (N-terminal amino group and K) were carbamylated, which was most likely due to heating of the sample to halt residual trypsin activity before LC-MS measurements. We performed calculations for two proposed single cross-linking types (Schilling et al. 2003): (1) interpeptide cross-linking products type 2, in which spatially close amino acid residues are cross-linked between two peptides, and (2) interpeptide cross-linking products type 2,1, the same cross-link as above with an intrapeptide cross-link within one peptide. The peptide chain of a type 2 cross-linked peptide is donated by an α-chain, and the other peptide chain is donated by a β-chain. We did not consider multiple cross-links within one peptide or between the peptides.

A summary of the cross-linked peptide identifications obtained by the application of the ASAP program on the 8-amu-shifted peptides is shown in Table 1. Twenty-four cross-linked peptides of the above types of cross-links and 29 cross-links were identified. For example, the peaks at m/z 703.9⁺³, 1208.1⁺², 816.8⁺³, and 833.9⁺² were assigned to the type 2,1 cross-linking. The new version of the ASAP program, Links, allowed us to identify intermolecular cross-links Lys77–Glu87 and Lys39–Glu36 (peaks at m/z 802.1⁺³ and 774.1⁺³) between two molecules of GST.

Table 1.

Automatic assignment of MS cross-linked peptide spectra using ASAP

Exp. mass	Error (ppm)	Charge	Amino acid no.	Sequence	Cross-linked sites	Distance intramolecular (Å)b	Distance intermolecular (Å)c	Cross-linking type
447.8	41.2	2	87–88, 175–179	ER-LVCFK	Glu87–Cys177	23.6	41.4	2
573.8a	42.3	2	9–10, 35–41	IK-DEGDKWR	Lys10–Glu36	21.2	44.7	2
606.4	80.7	1	87–88, 194–196	ER-SSK	Glu87–Ser194	29.2	51.5	2
703.9a	15.8	3	35–41, 78–88	DEGDKWR-HNMLGGCPKER	Asp38–Lys86,	50.5	31.9	2,1
				DEGDKWR	Glu36–Lys39	3.5	58.4
709.4	63.7	3	191–196, 119–130	YLKSSK-VDFLSKLPEMLK	Lys196–Asp120	39.7	32.7	2
771.4	42.3	3	119–124, 64–77	VDFLSK-LTQSMAIIRYIADK	Asp120–Thr65	32.4	20.0	Intermolecular
774.1a	65.8	3	35–43, 35–43+1 carb	DEGDKWRNK-DEGDKWRNK	Lys39–Glu36	3.5	59.0	2
776.7	53.0	3	64–72, 108–118	LTQSMAIIR-IAYSKDFETLK	Ser67–Glu115	26.4	30.1	2
777.1	68.0	3	191–196, 64–77	YLKSSK-LTQSMAIIRYIADK	Ser195–Asp76	30.4	38.7	2
797.7	10.3	3	175–181, 89–102	LVCFKKR-AEISMLEGAVLDIR	Lys179–Glu90	23.4	34.6	2
802.1a	85.6	3	73–77, 73–88	YIADK-YIADKHNMLGGCPKER	Lys77–Glu87,	18.2	16.2	Intermolecular
					Asp76–Lys77	7.48	19.8
816.8a	41.5	3	113–118, 182–196	DFETLK-IEAIPQIDKYLKSSK	Glu115–Lys193,	31.8	53.7	2,1
				IEAIPQIDKYLKSSK	Asp189–Lys190	10.8	60.4
817.1	65.3	3	42–44, 89–107	NKK-AEISMLEGAVLDIRYGVSR	Lys43–Asp100	34.2	21.9	2
820.2	59.0	4	119–130, 103–118	VDFLSKLPEMLK-YGVSRIAYSKDFETLK	Asp120–Ser111	12.2	14.6	2
831.1	34.2	3	119–124, 87–102	VDFLSK-ERAEISMLEGAVLDIR	Ser123–Glu90	25.7	41.5	2
833.9a	75.6	2	42–44, 108–118	NKK-IAYSKDFETLK	Lys44–Asp113,	37.8	26.1	2,1
				IAYSKDFETLK	Glu115–Lys112	10.7	23.7
837.9a	64.8	2	125–130, 1–8	LPEMLK-SPILGYWK	Glu127–Lys8	37.3	30.0	2
863.4	98.4	2	180–181, 113–124	KR-DFETLKVDFLSK	Lys180–Glu115	25.5	49.0	2
906.4a	81.6	3	18–26, 181–193	LLLEYLEEK-RIEAIPQIDKYLK	Glu21–Lys190	7.5	52.6	2
1038.9	41.8	3	73–86, 181–193	YIADKHNMLGGCPK-RIEAIPQIDKYLK	Cys84–Glu183	31.3	35.8	2
1039.2	38.8	3	11–17, 87–107	GLVQPTR-ERAEISMLEGAVLDIRYGVSR	Thr16–Glu95	9.6	14.1	2
1092.9a	43.3	3	182–190, 18–34	IEAIPQIDK-LLLEYLEEKYEEHLYER	Lys190–Glu21	7.5	52.6	2
1195.1a	14.2	2	191–196, 27–39	YLKSSK-YEEHLYERDEGDK	Lys196–Glu28	9.53	53.9	2
1208.1a	16.1	2	35–41, 73–86	DEGDKWR-YIADKHNMLGGCPK	Glu36–Lys77,	24.9	47.1	2,1
					Asp76–Lys86	14.2	19.5

^a The MS/MS data of these peptides were measured, and they confirmed the ASAP assignments.

^b Distance between cross-linking candidate amino acid residues within the monomer calculated using the structural data from PDB ID 1GNE.

^c Distance between to be cross-linking candidate amino acid residues between two monomers in the dimer calculated using the structural data from PDB ID 1Y6E.

We also investigated the possibility of forming esters or thioesters from the hydroxyls of Ser and Thr or the thiols of Cys under our cross-linking conditions. However, only nine cross-linked peptides were identified by using these criteria, suggesting that this type of cross-link is less frequent than the one between the amino groups and carboxyls. For example, type 2 cross-linked peptides were assigned to the peaks at m/z 447.8⁺², 606.4⁺, 831.1⁺³, and 1038.9⁺³.

Confirmation of identity of cross-linked peptides and identification of cross-linking type

MS/MS spectra were analyzed to confirm the identity of some of the cross-linked GST peptides proposed by the ASAP analysis and to evaluate the type of cross-linking. Due to the low concentrations and fragment ion intensities of many cross-linked peptides, we obtained interpretable MS/MS data for only a limited set of 11 8-amu-shifted peptides (Table 1). Fragmentation data were analyzed to either accept or reject predicted structures of a cross-linked peptide by using the MS2Assign program (Schilling et al. 2003). For example, the MS/MS spectrum of an interpeptide cross-linking from GST shown in Figure 6 can be assigned to two peptides: YLKSSK¹⁹⁶ and YE²⁸EHLYERDEGDK. Cleavage products could be assigned to one or the other of the two peptide chains, primarily type a-, b-, and y-ions. The fragment ions corresponding to these two peptides are, therefore, designated with either the α or the β subscript to indicate the originating peptide. For the cross-linked peptide YEEHLYERDEGDK, four y-ions are present at m/z 563.2 (y_5β), 719.3 (y_6β), 1011.3 (y_8β), and 1502.7 (y_12β − 17). A set of a- and b-ions are also observed for the same peptide a_11β, a_8β − 18, b_12β, and b_11β + 18. Moreover, fragment ions originating from cleavages involving both peptide chains, e.g., (b_6αb_8β)⁺ − 17 and (y_3αy_13β)⁺ were observed.

Figure 6.

MS/MS spectra of the cross-linked peptides YLKSSK¹⁹⁶ and YE²⁸EHLYERDEGDK at m/z 1195.1⁺². Fragments from peptide YLKSSK are labeled with an α subscript; those from peptide YEEHLYERDEGDK, with a β subscript. Most peaks were generated from MS2Assign, as indicated in the figure.

Figure 7 shows the types of cross-linked products. In the spectrum shown in Figure 7, the peak at m/z 833.9⁺² corresponds to the type 2,1 cross-link of the peptide IAYSKD¹¹³FETLK with NKK⁴⁴. The peptide IAYS KDFETLK contains an internal cross-link between residues Lys112 and Glu115. Two a-ions are present at m/z 890.4 (a₈ − 18) and 981.4 (a₉ − 28), and one b-ion is present at m/z 918.5 (b₈ − 18). Also, inter-cross-linked peptides NKK⁴⁴ (α-chain) and IAYSKD¹¹³ FETLK (β-chain), generated a pattern similar to the previous example and consisted of a series of dominant y-, a-, and b-ions.

Figure 7.

MS/MS spectra of 2,1 type cross-linking. One interpeptide cross-link occurred between the α-chain of the peptide NKK ⁴⁴ and the β-chain of the peptide IAYSKD¹¹³FETLK. An intrapeptide cross-link was formed between K12 and E115 of the peptide IAYSK¹¹²DFE¹¹⁵TLK, which is labeled with asterisks.

Intermolecular cross-links between two molecules of GST were identified by using the Links program and are shown in Figure 8. A cross-link was identified between Lys77 of the peptide YIADK⁷⁷ (α-chain) and Glu88 of the peptide YIADKHNMLGGCPKE⁸⁷R (β-chain). The MS/MS spectrum revealed some fragments, for example, b_8β − 17 at m/z 956.5 and y_10β − 28 at m/z 1076.5, as well as a_5α − 28 at m/z 535.4, that correspond to the sequences YIADKHNMLGGCPKER and YIADK, respectively.

Figure 8.

MS/MS spectra of the intermolecular cross-link. The interpeptide cross-links was found between the sequence YIADKHNMLGGCPKER of one molecule of GST containing the intrapeptide cross-link in the peptide between D76 and K77 (labeled with asterisks) and the peptide YIADK of another molecule GST.

MS/MS spectra were not obtained for some cross-linked peptides due to their low intensity. The confidently assigned peptides are listed in Table 1 (those without the stars) and their sequence confirmed that predicted from the ASAP program.

Validation of cross-links using three-dimensional structure of native GST

We validated our predictions of cross-linked GST peptides by visualizing them in the three-dimensional (3D) oligomeric structure of a GST monomer (Lim et al. 1994) and the GST dimer (Rufer et al. 2005). Distances between the hydrogen donors (lysine amino groups, serine hydroxyl groups, cysteine sulfhydryl groups) and hydroxyl groups of related aspartate or glutamate amino acid residues for cross-linked peptides were calculated for both the monomer and the dimer (Table 1) crystal structures.

Given the flexibility of the long side chains between which zero-length cross-links can form, we used a distance of 10 Å as a cut-off for determining whether or not the cross-link is consistent with the monomer or the dimer. Columns 7 (Distance Intramolecular) and 8 (Distance Intermolecular) report the average of the distance between cross-linked candidate residues within each monomer and between the same residues in the dimer, respectively. Among the 24 cross-linked peptides listed in Table 1, nine (m/z values 703.9⁺³, 774.1⁺³, 802.1⁺³, 816.8⁺³, 833.9⁺², 906.4⁺³, 1039.2⁺³, 1092.9⁺³, 1195.1⁺²) were consistent with distances measured within the monomers. The peptide YIADK-YIADKHNMLGGCPKER with the m/z of 802.1 of those can also form an intermolecular cross-link as indicated. The sequences of all peptides except for the m/z 1039.2 species were also confirmed by MS/MS followed by MS2Assign analysis.

We found numerous peptide assignments with distances >10 Å for both intramolecular and intermolecular cross-links. One explanation of those is that there could be another orientation of the GST monomer to either another GST monomer or the GST dimer, or both. Indeed, our experimental data indicate that the next size after the dimer is a trimer (Fig. 2A), which formation would require a new monomer orientation relative to GST monomer/dimer. Determining these possible orientations by using molecular modeling tools such as distance geometry or constrained conformational searches is a computationally intensive process beyond the scope of this article, but it is important to note that this method provides constraints that could be used for modeling protein quaternary structure.

Discussion

To circumvent the limitations of current protein complex identification methods and cross-linking methods used for structural evaluation of proteins and protein complexes, we developed a new strategy for zero-length cross-linking of intact protein complexes in the solid state. The core of our strategy is to perform the cross-linking reaction in solid state by using the natural chemical reactivity of amino acid residues that are in contact due to their interaction at a protein–protein interface. Instead of the tedious purification of cross-linked peptides, we identified cross-linking products directly in the MS spectra by using trypsin-catalyzed ¹⁸O replacement in C-terminal peptide carboxyls using the ASAP program and/or by subjecting peptides to MS/MS analysis.

The formation of covalent amide cross-links between interacting ammonium and carboxylate groups of protein amino acid residues by dehydration with heat and vacuum has recently been described by Simons et al. (2002). However, the Simons et al. approach used prolonged heating for up to 24 h at temperatures as high as 85°C, which could cause structural changes to the protein complex and could cause thermal modifications. Under such conditions, proteins can cross-link nonspecifically, as the investigators showed for the cross-linking two unrelated proteins—the bovine pancreatic RNase A and the egg lysozyme.

We improved the cross-linking method in solid state by introducing an organic solvent and a dehydration reagent—the carbodiimide. Our conditions facilitated much faster cross-linking (~30 min) in the solid state, resulting in high yields (up to 50%) at room temperature while maintaining conditions that are more amenable to preserving intact protein complexes. Cross-linking in solid state mediated by carbodiimide not only increased the yield of cross-linking compared with using heat and vacuum but also provided well-resolved products in gel. The same products were observed for purified protein and in the presence of a large excess of many other proteins, suggesting a high specificity of our cross-linking approach.

Protein cross-linking adds to the complexity of the peptide mixture being analyzed. Therefore, we looked for a method that would simplify MS analysis by focusing it on cross-linked peptides in order to make our cross-linking useful and, eventually, amenable to study the structure of not only purified protein complexes but also proteins of interest in crude protein mixtures. Most of the available methods focus on modifications to the cross-linking reagent that can be distinguished in the MS spectra (Bennett et al. 2000; Tang et al. 2005). The ¹⁸O labeling method of C-terminal carboxyls of (Back et al. 2002) looked the most suitable for our zero-length cross-linking. Although the ¹⁸O exchange is driven by enzyme kinetics, the reaction conditions could be decoupled from protein digestion (Yao et al. 2003) and adjusted so that the exchange is complete. Indeed, we were able to find conditions under which nearly complete isotope exchange occurred, thus facilitating automated identification of cross-linked peptides.

Even with help of the ¹⁸O exchange, the analysis of the MS spectra was still quite complex and challenging. Out of more than, e.g., 1269 peaks from an LC-MS analysis, there were 172 of 8-amu-shifted peaks that originated from the cross-linked peptides in the mass spectra. The ASAP program (Young et al. 2000) was quite helpful in assigning the correct cross-linked peptides.

We intentionally selected a simple protein, GST, for which the 3D structure has been solved and which specifically associates in higher-order oligomers (Lim et al. 1994) to prove the principle of our strategy. The same protein was used to develop photo-inducible zero-length cross-linking catalyzed by Ru(II)bpy₃²⁺ and ammonium persulfate in solution (Fancy and Kodadek 1999). This cross-linking approach preferentially targets protein tyrosine residues, creating stable covalent bonds between them. Despite the different origin of our zero-length cross-linking method, we observed a similar pattern and yield of cross-linked GST in solid state to that described by Fancy and Kodadek (1999) in solution. While Fancy and Kodadek did not measure the enzyme activity before and after cross-linking, we observed a decrease of GST activity proportional to the degree of cross-linking. This loss of activity is most likely due to the modification of Asp100, which is one of the essential amino acids of the GST active site (Lim et al. 1994) and is involved in the cross-link NK⁴³K-AEISMLEGAVLD¹⁰⁰IRYGVSR.

Some of the GST amino acid residues were cross-linked ambiguously to multiple corresponding partners (e.g., Glu36, or Asp76), suggesting that there might be some flexibility in the 3D structure. The same amino acid residues were shown to be simultaneously involved in multiple intramolecular or intermolecular interactions. Both the Glu36 and the Asp76 are involved in hydrophilic interactions between the GST molecules (Lim et al. 1994), which could explain why 100% cross-linking efficiency of GST could not be achieved.

Although proteins are known to markedly change their conformations due to lyophilization (Klibanov and Schefiliti 2004), reversibility of these changes suggests that the key peptide–peptide interactions holding the backbone of the active protein structure or protein complexes remain intact. Our results support this suggestion—nine of the 24 identified cross-linked peptide pairs fit those that interact via their carboxyls and amides/hydroxyls in the native structure of GST as shown in its 3D structure. The few unexplained cross-links could be caused by conformational flexibility, or they may belong to intermolecular cross-links that were not identified by our method.

Conclusions

In this report, we show the first successful cross-linking experiments on homo-oligomeric protein complexes using the method of “zero-length” protein cross-linking by dehydration in solid state with an organic solvent and dehydration reagents (e.g., carbodiimides). Solid state cross-linking pattern is shown to be specific to a particular protein complex in the presence of many other proteins. Our key contribution to the current knowledge and technologies about protein complex identification and characterization is in performing cross-linking in solid state, wherein most of the interactions between proteins or peptides can be preserved at the point of freeze-drying. Our strategy does not count on any further purification of cross-linked peptides or protein complexes—just their direct identification by using LC-MS, LC-MS/MS, and computer analysis of the spectra. We anticipate that the analysis will be much more difficult for larger protein complexes and will be limited to proteins and protein complexes with known composition and sequences. If there is enough material to get high-quality MS spectra, then the protein size should not be a limit unless cross-linking of a bigger protein will produce a water-insoluble product that would be impossible to cleave into the peptides by an endopeptidase like trypsin. This could be circumvented by using chemical cleavage methods (e.g., CNBr, N-bromosuccinimide) for solubilizing insoluble cross-linked products.

To overcome the limitations of cross-linked peptide identification using theoretical and experimental mass comparison, thus making our strategy useful for probing the structure of unknown protein complexes, a method based on MS/MS de novo sequencing of cross-linked peptides would have to be developed. This could lead to developing approaches for identification of protein interaction networks in whole proteome.

On the other hand, our method could be used as a new approach for probing the 3D structure of freeze-dried proteins and protein complexes, lending itself as a new tool for understanding freeze-dried state of proteins that is difficult to explore otherwise. This couldhave applications in designing stable protein formulations in protein drug research.

Article and publication are at http://www.proteinscience.org/cgi/doi/10.1110/ps.051685706.