Introduction

Intramolecular disulfide bonds naturally occur in various peptide families, including cyclotides [1–3], defensins [4], or animal toxins [5, 6]. Such peptides are increasingly used as a source of inspiration for drug design [7] because of the ability of disulfide bonds to increase peptide stability against proteases [8] and to improve antimicrobial activities [9]. Drugs against chronic constipation [10], neuropathic pain [11], or type-2 diabetes [12] have been developed thanks to disulfide bond-containing peptides.

The characterization of disulfide-bridged peptides is far from being trivial. It includes peptide sequencing, which can readily be performed after reduction and alkylation, leading, however, to the loss of information about cysteine pairing. This pairing can be obtained using nuclear magnetic resonance (NMR) [13, 14] or various protocols involving enzymatic digestions, chromatographic separations, and mass spectrometry (MS) [15, 16]. Nevertheless, these techniques are time- and sample-consuming, and often inefficient when vicinal cysteines are present in the sequences. Ion mobility-mass spectrometry (IM-MS) is proposed here as a tool to overcome the limitations of classic methods because of its capacity to separate ions in the gaseous phase owing to differences in their charge, shapes, and sizes [17, 18]. The collision cross section (CCS) Ω, is correlated to the former three parameters [19] and can be approximated as the rotationally averaged projection of an ion onto a surface. IM-MS has already been shown to be a useful technique to visualize the compaction effects given by the presence of phosphorylation modifications in peptides [20, 21].

In the context of this study, it can be assumed that the presence of disulfide bonds in small- to medium-sized peptides constrain the peptide structures. The effect of the presence of one, two, or three disulfides on the structure of small- and medium-sized peptides (118 peptides, mass <5200 Da) is studied using IM-MS. We propose to quantify, by measuring the CCSs, the effect of compaction of peptide structures (in Å2) caused by the disulfide bond(s). The influence of the cysteine connectivity (Cys1-Cys2/Cys3-Cys4; Cys1-Cys3/Cys2-Cys4; Cys1-Cys4/Cys2-Cys3) on the compaction of peptides containing two disulfide bridges is also investigated.

Materials and Methods

Chemicals

Formic acid (FA) and tris(carboxyethyl)phosphine (TCEP), used as a reducing reagent, were purchased from Sigma-Aldrich (St. Louis, MO, USA) and acetonitrile (ACN) was purchased from Biosolve (Dieuze, France). All the peptides analyzed in this study, named p1 to p79, are described in detail in the Supplementary Information (SI 1). p1, p2, p3 and p4 were purchased from Sigma-Aldrich, and p12 was acquired from Eurogentec (Liège, Belgium). The peptides p70 to p79 were extracted from a Conus textile venom. All the other peptides were chemically synthetized as described in the following (Peptide Synthesis) section .

Peptide Synthesis

The different peptides were synthesized using solid-phase synthesis. The fluorenylmethyloxycarbonyl chloride (Fmoc) strategy was adopted for the synthesis of the linear/reduced form of the peptides on a 25 μmol scale. Briefly, the protected amino acid (Fmoc-aa, 5 equiv.) coupling was performed twice, 5 min in N-methyl-2-pyrrolidinone (NMP; Sigma-Aldrich, St. Quentin en Fallavier) using 1H-benzotriazolium 1-[bis(dimethylamino)methylene]-5chloro-,hexafluorophosphate(1-),3-oxide (HCTU) (5 equiv.) as coupling reagent and N-methylmorpholine (NMM; Sigma-Aldrich, St. Quentin en Fallavier) (10 equiv.) as base, followed by a capping step using Ac2O/NMM (10/10 equiv.), as described elsewhere [22]. For long peptides (length >25 aa), 2 times 10 equiv. of Fmoc-aa, 10 equiv. HCTU, and 20 equiv. NMM were used. The Fmoc protecting group was removed using 20% piperidine in NMP. The peptide was cleaved from the resin as well as its side chains protected with trifluoroacetic acid (TFA)/anisole/thioanisole/water/triisopropylsilane (TIS) 82.5/5/5/5/2.5 v/v/v/v/v solution (8 mL) during 2 h and precipitated in cold ether (final volume 50 mL). The solid was washed three times with 45 mL of cold ether and lyophilized from 10% acetic acid solution.

The formation of the disulfide bonds leading to the folded peptides was performed in a redox system using [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid ] HEPES buffer (0.1 M, pH 7.5) containing 20% acetonitrile and 1 mM cysteine and 0.1 mM cystine. The conditions showed excellent results for toxins containing two disulfide bridges [22] and also for three brigdes (data not published). The folding step was carried out at 4 °C during 48 h at a reduced peptide concentration of 20 μM. The solution was then acidified using 20% TFA solution until pH 1 and the final peptides were concentrated and desalted using solid-phase extraction (SPE) C18 method. Briefly, the peptide is adsorbed on the C18 phase and washed several times with milliQ water (Millipore, Billerica, MA, USA) until neutral pH. The peptide is then eluted from the C18 using H2O/ACN 1/2 v/v solution and lyophilized.

IM-MS Sample Preparation

Samples ranging from p1 to p69 were analyzed at a concentration of 1 μM in H2O/ACN 50/50 (v/v) spiked with 0.1% FA. The CCSs of some peptides [with and without disulfide bond(s)] were measured at different solvent compositions (ranging from 100% ACN to 100% H2O), These measurements are reported in the Supplementary Information (SI 2) and the results will be discussed later in the paper.

In order to increase the number of peptides without disulfides, p1, p2, p3, p4, p5, p59, p60, p61, p62, p63, p64, p65, p66, p67, p68, and p69 were chemically reduced. Five μL of 10 μM peptide solutions were mixed together with 1 μL of TCEP (1 M), 1 μL of formic acid (5%), and 3 μL of water. The mixture was then incubated for 1 h at 56 °C before being purified using a ZipTip C18 (Millipore) according to the manufacturer’s protocol (http://www.millipore.com/userguides/tech1/pr02358). Reduced peptides were eluted by 50 μL of H2O/ACN 50/50 (v/v) spiked with 0.1% FA.

Furthermore, a tryptic digest of alcohol dehydrogenase (ADH) was analyzed. To perform this digestion, the ADH was chemically reduced by dithiothreitol (DTT) during 1 h at 56 °C. Alkylation was then performed by iodoacetamide (IAA) for 30 min at room temperature. The mixture was digested overnight (by trypsin) and was stopped by adding 0.5% TFA.

Ion Mobility-Mass Spectrometry Experiments

All peptides were analyzed using a SYNAPT G2 mass spectrometer (Waters, Manchester, UK) equipped with a traveling-wave ion mobility cell. The following parameters were used for all peptide analyses: capillary voltage = 3 kV, sampling cone = 20 V, extraction cone = 4 V, source temperature = 100 °C, desolvation temperature = 200 °C, desolvation gas flow = 100 L/h, trap/transfer CE = 4 V, trap bias = 35 V, trap wave height = 0.5 V, trap wave velocity = 300 m/s, IMS wave height = 30 V, IMS wave speed = 2000 m/s, transfer wave height = 0.1 V, transfer wave velocity = 508 m/s, IMS pressure = 1.68 mbar, trap and transfer pressure = 0.5 mbar. Data were processed using Waters Masslynx v. 4.1 software and arrival time distribution (ATD) peaks were fitted using PeakFit v. 4.11 by Systat Software. Graphs shown in this study were carried out using Sigmaplot 12.0 by Systat Software and Microsoft Excel.

Collision Cross Section (CCS) Calibration

The CCS calibration protocol described by Ruotolo et al. [23] was followed in order to convert ATDs into CCS. Reported CCS values of bradykinin [24], ubiquitin [25], myoglobin [26], cytochrome c [27], and a tryptic digest of bovine serum albumin BSA [28] were used. The calibration curve and the list of calibrants are provided in the Supplementary Information (SI 3 and SI 4).

Results

In this section, first the ATDs of a large set of peptides will be discussed. In a second time, experimental CCS values (obtained from the ATDs) will be analyzed for each charge state. Then, the effect on the measured CCS attributable to the substitution of a proton by a potassium cation will be investigated. Finally, in order to better visualize the effect of the presence of disulfide bond(s) in peptidic structures, the behavior of three peptides having the same sequence but different cysteine connectivities is discussed.

General Trend of Arrival Time Distributions for Peptides

Figure 1 depicts the effect of the charge state on the drift time of the 118 peptides. These peptides bear zero, one, two, or three disulfide bridges. For every charge state, a quasilinear trend of the ATDs as a function of the mass can be observed. However, when the charge state (z) increases, the data points, whatever the number of disulfide bond(s) in the peptides, become more scattered. This suggests that the higher the charge state, the larger the differences in terms of behaviors. The drift times are correlated through the CCS (Ω) to the ion mobility K, which is proportional to the ratio z/Ω according to Equation 1 (Mason-Schamp [29]):

$$ K=\frac{3}{16}\frac{e}{N}{\left(\frac{2\pi }{\mu {k}_bT}\right)}^{\frac{1}{2}}\frac{z}{\Omega} $$
(1)

where N is the drift gas number density, e is the elementary charge, T is the temperature, k b represents the Boltzmann constant, and μ is the reduced mass of the collision couple.

Figure 1
figure 1

Plot of the drift time as a function of the mass for 118 peptides (see SI 1) at various charge states (from z = 2 to z = 6)

Full size image

The plot of Ω (CCS) as a function of the m/z ratio (see SI 5) also depicts the effect of the charge states and reveals identical observations. As the charge state increases, the CCS values become larger and the deviations from linearity increase. This behavior can be exemplified by the results of the z = 6 charge state, where two trend lines are visible (from 3000 Da). The two trends are associated with two distinct populations: smaller drift times yielded by peptides bearing disulfide bonds and higher drift times measured for peptides without disulfide bonds.

The CCSs of some peptides [with and without disulfide bond(s)] were measured at different solvent compositions, ranging from 100% ACN to 100% H2O), showing that the solvent compositions do not affect the measured CCS values (see SI 2).

Influence of the Charge State

Figure 2 depicts the relation between the CCS (Ω) and the mass (m) for charge states ranging from z = 2 to z = 6. The results clearly show that differences exist between peptides bearing or not bearing disulfide bridges, the amplitude depending on the charge state. For z = 2, it appears that peptides bearing one or two disulfide bond(s) display a similar behavior. It is therefore impossible to discriminate between these peptides. For higher charge states, however, the differences linked to the number of disulfide bonds become visible.

Figure 2
figure 2

Plot of the CCS (Ω, in Ȧ 2) as a function of the mass for the charge states z = 2 to z = 6. Peptides with disulfide bond(s) are depicted in green, whereas peptides without disulfide bond are depicted in red

Full size image

In order to improve data vizualisation and following a data treatment described elsewhere [30], we determine the difference in collision cross sections ΔΩ for each charge state. ΔΩ (Equation 2) designates the difference between the averaged CCSs of the reduced ‘unfolded’ peptides (Ω peptide without disulfide bond ) and the CCSs of the disulfide-bound peptides (Ω peptide with disulfide bond ).

$$ \varDelta \Omega ={\Omega}_{peptide\ with out\ disulfide\ bond}-{\Omega}_{peptide\ with\ disulfide\ bond} $$
(2)

where Ω peptide without disulfide bond is the CCS extracted from a linear fit equation (Equation 3 where A and b are fitting parameters, Ω the CCS and m the mass) applied on data points, charge state by charge state, of peptides without disulfide bonds. The fitted points and the applied fits are depicted in Figure SI 6.

$$ \Omega = Am+b $$
(3)

Table 1 exhibits the fitting parameters A and b obtained for the different charge states. This table reveals that the R 2 values are always higher than 0.97, evidencing a good fit. By applying the different fits on the charge states, a normalization of the data can be performed.

Table 1 List of the fitting parameter A and b values obtained by fitting the CCS values of the peptides without disulfide bonds for each charge state, using Equation 3 (Ω = Am + b). R 2 values of the fit correlations are shown as well
Full size table

ΔΩ values were calculated for the disulfide bond-containing peptides in order to evaluate the difference as a descriptor of the number of disulfide bond(s). The plot of the ΔΩ values as a function of the mass for different charge states is shown in Figure 3. A graphical illustration of the ∆Ω calculation (for z = 5) is depicted in Figure SI 7).

Figure 3
figure 3

∆Ω versus m plots for different charge states (from z = 4 to z = 6). Peptides at the same mass having the same number of disulfide bonds differ only by their cysteine connectivity

Full size image

∆Ωs as a Function of the Mass Spread Over a Large Range of Values

For z = 4, peptides with three disulfide bonds have on average a higher ΔΩ value than the peptides bearing only two disulfide bonds, especially in the mass range located between 2500 and 3000 Da. This behavior can be confirmed using the trend lines of peptides and the 95% confidence band (see SI 8 a).

In the SI 8a, it can be seen that a comparison between peptides bearing two and three intramolecular disulfide bonds can only be done in the range between 2200 and 3300 Da. It is also interesting to mention that the ΔΩ tendency as a function of the mass for z = 4 passes through a maximum, the ΔΩs increasing from 1500 to 3500 Da and decreasing again from 3500 to 4000 Da.

For z = 5, a similar situation as the one for z = 4 can be found: peptides bearing one, two, and three disulfide bonds can be distinguished, especially between 2500 and 3500 Da. This behavior is confirmed in the SI 8b. The ΔΩ values first increase when the mass increases from 2000 to 3500 Da and then decrease again from 3500 to 5000 Da. This last tendency, which is also found for z = 4, will be discussed in the “Discussion” section.

For z = 6, peptides having three disulfide bonds all exhibit ΔΩ values around 100 Ȧ2. The lack of points for peptides bearing one or two disulfide bonds prevents further discussion for this charge state.

Effect of the Pairing for Peptides Bearing Two Disulfide Bonds

A peptide possessing two disulfides (four cysteines) can present three different connectivities. Indeed, each cysteine can be linked to one of the three others, yielding different connectivities. The different connectivities provide a particular structure to the peptide, leading to varying compaction effects. To verify this hypothesis, all the peptides bearing two disulfide bonds were annotated according to their connectivities patterns (Figure 4).

Figure 4
figure 4

∆Ω versus mass plots for peptides bearing two disulfide bonds but having different cysteine connectivities for z = 3, z = 4, and z = 5

Full size image

Figure 4 shows that species exhibiting a 1,3/2,4 cysteine connectivity generally display higher ∆Ω values than the two others, this connectivity bringing more compactness to the peptide. This trend can also be observed for the 4+ and 5+ charge states. This behavior is confirmed in the SI 8c and SI 8d, where 95% confidence bands are provided for the 1,3/2,4 connectivity.

For z = 3, the situation is less evident but small ∆Ω differences are observed in the range 1200–1400 Da. Similar to what has been seen above, it appears that the separation is better for higher charge states.

Effect of K+ Adducts on Collision Cross-Sections

Potassium adducts have been detected in the spectra because of contaminations (traces) of the samples. The comparison of protonated or potassiated peptides bearing zero or two disulfide bonds for two different charge states (z = 2 and z = 3) is shown in Figure 5, where the Ω is plotted as a function of the mass.

Figure 5
figure 5

Plots of the CCS as a function of the mass for the charge states z = 2 and z = 3 for peptides with and without disulfide bond. Red trend lines are for the peptides bearing only H+ adducts, whereas black trend lines are for peptides having one K+ adduct

Full size image

Figure 5 highlights that the substitution of a H+ by a K+ ion for a given charge state causes a reduction in CCS of the peptides independent of the presence of disulfide bonds in the peptide structure. This behavior can be correlated to a previous study [31], which demonstrates a sequence-dependent size reduction effect on linear peptides produced by a trypsin digest upon the substitution of protons by potassium cations. It should also be noted that no multiply potassiated species ([Peptide + 2 K+]2+, [Peptide + 3 K+]3+, [Peptide + 1H+ + 2 K+]3+,…) were found in the spectra of peptides bearing disulfide bonds, even when adding 100 μM KCl to the peptide solutions.

Ion Mobility Separation of Peptidic isomers with Three Different Cysteine Connectivities

For further characterization of the structural effects of disulfide connectivities on a peptide, this section will focus on three sequence-identical peptides bearing four cysteines but having three different disulfide bond pairings (p50 for 1,2/3,4 connectivity, p51 for the 1,3/2,4 connectivity, and p52 for the 1,4/2,3 connectivity). The ATDs were measured for each of the three compounds as well as in a mixture of the three isomers. Figure 6 depicts the ion mobility contributions of the 4+ and 5+ charge states.

Figure 6
figure 6

Plots of the ATDs of different peptide adducts for sequence-identical peptides having different cysteine connectivities. The first line represents the ATDs of the mixtures of p50, p51, and p52 for [M + 4H+]4+, [M + 5H+]5+, and [M + 4H+ + K+]5+. The following lines show each the ATDs of the separately-injected p51, p52, and p50 peptides as [M + 4H+]4+, [M + 5H+]5+ and [M + 4H+ + K+]5+ complexes. All injections were performed at 1 μM in a H2O/ACN 50/50 spiked with 0,1% FA. Peptide structures that are presented are for the purpose of illustration. ATDs of the [Peptide + 4H+ + K+]5+ species have lesser intensities

Full size image

The peptide having a 1,3/2,4 connectivity exhibits the smallest drift time, meaning that it is more compact than its two other isomers. The separation is enhanced when going from the 4+ to the 5+ charge state. Regarding the 4+ ATDs, only the 1,2/3,4 connectivity can be clearly distinguished (CCS = 612 Ȧ2) from the two other isomers, which, in turn, show similar ATDs, one compared with the other (CCS = 586 Ȧ2 and 591 Ȧ2).

For the 5+ ATDs, the 1,3/2,4 connectivity shows a CCS of 624 Ȧ2, whereas a CCS of 638 Ȧ2 is associated with the 1,4/2,3 connectivity . The 1,2/3,4 connectivity has a CCS of 646 Ȧ2. The three peptides can hence be differentiated in their 5+ charge state. Figure 5 also shows that the substitution of one proton by a potassium cation enhances the separations. The 1,3/2,4 connectivity species remains unshifted, whereas the CCS of the 1,4/2,3 connectivity species is shifted by 7Ȧ2 (from 638 Ȧ2 to 645 Ȧ2) and the CCS of the 1,2/3,4 connectivity species is shifted by 12Ȧ2 (from 646 Ȧ2 to 658 Ȧ2) .

From a general point of view, it appears that replacing a proton by a potassium cation can be in some cases a useful technique in order to enhance ion mobility separations of disulfide-bridged peptides differing by their connectivity.

Discussion

Our results reveal that for all the analyzed peptides at different charge states (see Figure 3), ∆Ω values increase with the mass until they reach a maximum value, after which the ∆Ωs begin decreasing again. Given the definition of ∆Ω [Equation 2: ∆Ω = Ω peptide without disulfide bond - Ω peptide with disulfide bond(s) where Ω peptide without disulfide bond can be obtained by a fit equation in order to yield an estimation of the Ω peptide without disulfide bond at a given mass], the behavior at small masses could be explained by the fact that the disulfide bonds do not cause a large CCS variation because the constrains are applied on only a small number of amino acids. When the mass increases, the ∆Ω increases to a maximum value. After passing the ∆Ω maximum, the lengths of the sequences are sufficient to screen the Coulomb repulsion, which becomes smaller compared with other intramolecular interactions (hydrogen bonds, hydrophobic interactions between aliphatic amino acids,…). Thus, the peptides bearing no disulfide bonds are able to adopt smaller conformations translating as smaller CCS values and the ∆Ωs decrease.

Our results show that it is possible to differentiate peptides bearing one, two, or three intramolecular disulfide bond(s) based on their CCS values or on the difference of their CCS with the CCS of reference peptides bearing no disulfide bond.

Moreover, distinctions between peptides bearing two disulfide bonds but having differing cysteine connectivities can be made. Indeed, peptides displaying a 1,3/2,4 connectivity have, on average and especially for some charge states (e.g., z = 4), a smaller CCS than the two other connectivities. This observation is most interesting because the disulfide bond assignment is one of the major challenges nowadays in this field of research [7]. The predictive nature of this method has been investigated by performing fits of the different populations and by using 95% confidence bands. Results reveal that an effective comparison/prediction can only be perfomed at given charge states (z = 4 and z = 5) and at given mass ranges.

CCS measurements have also been performed in different solvent compositions. Kinetically trapped structures during ESI are not detected. The invariance of the CCS values when changing solvent compositions shows that our approach can also be used with direct analysis of peptides coming from a gradient mode HPLC/UPLC of a mixture of peptides bearing intramolecular disulfide bond(s) (e.g., animal venom).

The use of cationizing agents other than protons has also been shown to be useful tool to enhance the separation of peptides displaying identical sequences but different cysteine connectivities. However, this effect still has to be verified for multiple peptide sequences to attest its unanimous veracity.

The method developed in this paper could be used on disulfide bond isomers of peptides bearing two intramolecular disulfide bonds that cannot be distinguished using LC-MS [32].

Also, after tryptic digestion of proteins in non-reducing conditions, linear tryptic peptides as well as peptides linked together by one (or more) disulfide bond(s) could be distinguished using the presented method. At the same m/z, for high z, a distinction similar to the one we found between linear and branched polymers [30] will be observed, the bridged peptides being less able to extent their structure. The peptides still connected by disulfide bridges will be easier to detect among all the digestion peptides. The localization of disulfide bonds in proteins will be facilitated.

Conclusions

The purpose of this study was to evaluate IM-MS as a tool to differentiate peptides bearing different numbers of intramolecular disulfide bonds as well as peptides having the same sequence but different cysteine connectivities. This may lead to a rapid verification of correct foldings, which is difficult to obtain by MS/MS.

We analyzed 118 peptide sequences, including reduced and disulfide-bound peptides. We showed that IM-MS is a useful technique to perform structural classifications of peptide folds bearing disulfide bonds. This classification was performed owing to the calculation of CCS differences (∆Ωs) between the disulfide bond-containing peptides and the reduced peptides.

Only small distinctions between peptides bearing one and two disulfide bonds can be made at small charge states (z = 2 and z = 3). Peptides bearing three disulfide bonds are, most of the time, more compact than reduced peptides and the peptides bearing one or two disulfide bonds. This behavior has been confirmed using trend lines and 95% confidence bands.

In the case of sequence-identical model peptides bearing two disulfide bonds, the influence of the cysteine connectivity was investigated. The 1,3/2,4 cysteine connectivity seems to be more compact than the 1,4/2,3 connectivity. This difference in compactness can be found for the charge states z = 3 to z = 5. These results may be used for quick structural classifications of peptides bearing two intramolecular disulfide bonds. It is also noteworthy that the compactness is not influenced by varying solvent compositions (H2O/ACN) from which the ions were produced. The gradient used in HPLC will hence not affect the results.

Finally, the CCSs of potassium adducts may help differentiating structures by enhancing the separations of specific connectivities for sequence-identical peptides.

In terms of prospects, the list of peptides will be extended to improve the confidence in the observed trends. The connectivity identification will be specifically addressed for peptides bearing six cysteines where three disulfide bonds can be formed. The number of possible disulfide isomers is described by the combinatorial formula [33] [Equation 4 where P(n) is the number of possible isomers and n the number of disulfide bonds considered]. The potential cysteine connectivities increase from 3 (for two disulfides) to 15 for three disulfides.

$$ P(n)=\frac{(2n)!}{2^n\ n!} $$
(4)

Peptides bearing four intramolecular disulfide bonds could also be analyzed. Other studies from our group on polymers [34] suggest that other fit equations, different from Equation 3, could also be used in order to calculate the ∆Ωs.

The method developped in this paper could be used to check if the right cysteine pattern was synthetized or found. This last information could be specifically interesting if liquid chromatography is unable to separate these disulfide isomers [32]. This method, used in addition to tryptic digestions in non-denaturing conditions, could also be utilized to facilitate the detection and the identification by MS/MS information of disulfide bond localization in proteins.

Furthermore, preliminary results show that improving the IM resolution by using (e.g., a trapped ion mobility spectrometry TIMS) setup, can push the disulfide bond characterization even further owing to a higher resolution that could allow a better separation between disulfide isomers [35].