Multiplexed cross-linking with isobaric quantitative protein interaction reporter technology.

Abstract

Chemical cross-linking with mass spectrometry (XL-MS) has emerged as a useful technique for interrogating protein structures and interactions. When combined with quantitative proteomics strategies, protein conformational and interaction dynamics can be probed. Quantitative XL-MS has been demonstrated with the use of stable isotopes incorporated metabolically, or into the cross-linker molecules. Isotope labeled cross-linkers have primarily utilized deuterium and rely on MS1 based quantitation of precursor ion extracted ion chromatograms. Recently the development and application of isobaric quantitative protein interaction reporter (iqPIR) cross-linkers were reported, which utilize ¹³C and ¹⁵N isotope labels. Quantitation is accomplished using relative fragment ion isotope abundances in tandem mass spectra. Here we describe the synthesis and initial evaluation of a multiplexed set of iqPIR molecules, allowing for up to six cross-linked samples to be quantified simultaneously. To analyze data for such cross-linkers, the two-channel mode of iqPIR quantitative analysis was adapted to accommodate any number of channels with defined ion isotope peak mass offsets. The summed ion peak intensities in the overlapping channel isotope envelopes are apportioned among the channels to minimize the difference with respect to the predicted ion isotope envelopes. The result is accurate and reproducible relative quantitation enabling direct comparison among six differentially labeled cross-linked samples. The approach described here is generally extensible for the iqPIR strategy, accommodating future iqPIR reagent design, and enables large-scale in vivo quantitative XL-MS investigation of the interactome

For Table of Contents Only

INTRODUCTION

Chemical cross-linking with mass spectrometry (XL-MS) is a technique that is becoming increasingly useful for the study of protein structures and interactions. XL-MS is considered highly complementary to traditional structural biology techniques such as X-ray crystallography or cryogenic electron microscopy (cryo-EM). Although the information from XL-MS is lower resolution, typically on the order of tens of angstroms, it offers several unique benefits including being applicable to proteins and multicomponent protein complexes of unlimited size¹ as well as biological samples of extreme complexity including intact living cells and tissues².

When combined with quantitative proteomics strategies, quantitative XL-MS (qXL-MS) can provide information on changes to protein conformations and interactions. The majority of qXL-MS studies performed to date have utilized stable isotopes to generate unique mass labels into samples to be compared, allowing for relative quantitation by MS. This has been demonstrated by metabolic incorporation of isotopes into proteins using SILAC³, to differentiate drug resistant and sensitive phenotypes⁴ or evaluate the concentration dependent effects of drug treatment on cells on the interactome^5,6. One challenge for MS1 based quantitation of light and heavy isotope cross-link pairs is that the signal for each cross-link is split into two channels, complicating the already dense MS1 spectra. A number of studies have incorporated stable heavy isotopes directly into the cross-linker molecule, thereby generating isotopically light and heavy versions of each cross-linked peptide pair which can be differentiated and quantified in MS1 spectra. To date, nearly all examples of isotope encoded light and heavy cross-linkers employed in qXL-MS studies have utilized deuterium as the heavy isotope^7–9. While deuterium is useful for generating isotopologue pairs, differences between the C-H and C-D bonds result in an observed retention time (RT) shift during reversed-phase LC fractionation¹⁰. This RT shift further complicates the challenging task of precisely extracting MS1 chromatograms for light and heavy analytes, serving as a potential source of error for quantitative analysis.

Protein interaction reporter technology is a modular concept to cross-linker design, relying on solid phase peptide synthesis (SPPS) chemistry and the wide availability of FMOC-protected amino acids to facilitate synthesis of peptide-based chemical cross-linkers with useful features, including MS labile bonds, affinity tags and isotope labels.

We recently reported the development of iqPIR cross-linkers, allowing for binary qXL-MS analyses¹¹. The iqPIR cross-linkers utilize ¹³C and ¹⁵N isotopes incorporated into the cross-linker molecule at specific locations, allowing for cross-linkers to have the same molecular mass, but generate distinct, isotope encoded fragment ions upon tandem MS analysis that can be used for relative quantitation based on fragment ion spectra. Within iqPIR tandem mass spectra quantifiable fragment ions include the released intact peptide ions formed upon cleavage of the cross-linker MS labile bonds, iqPIR reporter ions, as well as any y-type and b-type peptide backbone ions containing the cross-linked amino acid residues which carry an isotope encoded stump mass modification. These unique features circumvent the issues mentioned above with MS1 based quantitative methods, improving the quality of qXL-MS measurements. In this study we describe the development, synthesis, and initial evaluation of a set of iqPIR cross-linkers allowing for multiplexed analysis of up to six biological samples in a single LC-MS acquisition. We present a computational approach for quantification whereby the summed fragment ion peak intensities in the overlapping channel isotope envelopes are apportioned among the channels to minimize the difference with respect to the predicted ion isotope envelopes, resulting in accurate relative quantitation among six differentially labeled cross-linked samples.

The quantitative performance of the 6-plex iqPIR strategy is evaluated using defined mixtures of purified proteins as well as a more complex multiplexed sample of six HeLa biological replicates. An accompanying manuscript (Wippel et al. 2021, submitted), demonstrates the application of the 6-plex iqPIR reagents on breast cancer cells treated with different heat-shock protein 90 (Hsp90) inhibitors.

EXPERIMENTAL SECTION

Materials

The following Fmoc protected amino acids were obtained from CEM (Matthews, NC): Gly-N-Fmoc, L-Asp-N-Fmoc, beta—o-tert-butyl ester, L-Lys-alpha-N-Fmoc, epsilon-N-Fmoc, L-Pro-N-Fmoc. Fmoc-L-Lys(biotin)-OH was obtained from P3 BioSystems (Louisville, KY).

The following heavy isotope labeled Fmoc protected amino acids were obtained from Cambridge Isotope Laboratories (Andover, MA): L-Asp-N-Fmoc, beta—o-tert-butyl ester (¹³C₄, ¹⁵N), L-Lys-alpha-N-Fmoc, epsilon-N-Fmoc (¹³C6, ¹⁵N₂), L-Lys-alpha-N-Fmoc, epsilon-N-Fmoc (¹³C₆), Gly-N-Fmoc (¹³C₂), L-Pro-N-Fmoc (¹³C₅,¹⁵N). Succinic anhydride, Succinic anhydride-1,4-¹³C₂ and Succinic anhydride-¹³C₄ were obtained from Sigma Aldrich (St. Louis, MO).

Acetonitrile (ACN), dimethylformamide (DMF) dimethyl sulfoxide (DMSO), dichloromethane (DCM), pyridine, trifluoroacetic acid (TFA), formic acid (FA), were all obtained from Sigma Aldrich.

The following standard proteins were all obtained from Sigma Aldrich: alcohol dehydrogenase (ADH1_YEAST) from S. cerevisiae, albumin (ALBU_BOVIN) from B. taurus, cytochrome C (CYC_BOVIN) from B. taurus, hemoglobin (HMG_HUMAN) from H. sapiens, myoglobin (MYG_HORSE) from E. ferus.

See Supplementary Experimental Section for details of the synthesis of iqPIR reagents, protein cross-linking and multiplexed sample preparation for LC-MS, HeLa cell culture and in vivo cross-linking, LC-MS/MS analysis of multiplexed cross-linked peptide samples, cross-link identification data analysis, and calculation of cross-link relative concentrations among samples.

RESULTS AND DISCUSSION

Design and synthesis of 6-plex iqPIR cross-linkers

PIR cross-linkers are peptide-based molecules with engineered features including affinity tags, selectively cleavable bonds, and isotope labels. The 6-plex iqPIR cross-linkers were designed based on the biotin containing Asp-Pro cleavable bond (BDP) scaffold and were synthesized using SPPS methodology. Features of these cross-linkers include a biotin tag, Asp-Pro MS cleavable bonds, amine reactive NHP esters and selective incorporation of stable heavy isotopes. The amino acid building blocks were coupled in order of Gly, Gly, biotin-Lys, Lys, Pro₂, Asp₂, and succinate₂ to grow the peptide from the C-terminus (amide) to the N-terminus which is capped with succinate groups converted to NHP esters to produce the molecule illustrated in Figure 1A. The iqPIR molecules are designed to have the same molecular mass yet generate unique fragment ions upon cleavage of the Asp-Pro bonds. This is accomplished by incorporating heavy isotope labels (¹³C and ¹⁵N) into specific locations within the reporter (cyan outlined portion) and stump (magenta outlined) portions of the molecules (Figure 1A). In this case each of the six cross-linkers contains a total of sixteen ¹³C and two ¹⁵N at unique positions within the molecular structure (Figure 1A & B). The six cross-linkers all have the same monoisotopic molecular mass of 1545.609006 Da as demonstrated by the ESI-MS1 spectra of the intact iqPIR reagents (Figure 1C and Supplementary Figures S1–S6). Upon fragmentation of a precursor ion containing the six iqPIR reagents, the unique reporter ions are generated at 808, 812, 816, 818, 822 and 826 m/z (Figure 1D). The reporter ions are spaced by 4 m/z corresponding to the mass of four ¹³C with the exception of the 816 and 818 reporters which differ by 2 m/z corresponding to the mass of two ¹⁵N. As a result of having overlapping ion peak envelopes, more than a single isobaric cross-linker contributes intensity to many of the observed isotope peaks.

Figure 1 – Molecular structure and design of 6-plex iqPIR cross-linking reagents.

A) Molecular structure of an iqPIR cross-linker illustrating the reporter region outlined in cyan and the stump regions outlined in magenta. Potential atomic locations for heavy isotopes, either ¹³C or ¹⁵N, are indicated by colored circles, coded to match coloring in panel B. B) Table indicating how heavy isotopes were distributed within the cross-linker structure to achieve an isobaric set of 6 distinct iqPIR cross-linker molecules. Numbers in the boxes correspond to ¹³C atoms and/or ¹⁵N atoms (denominator). The nominal reporter ion masses for each of the 6 iqPIR regents are listed to the left of the table. C) Representative MS1 spectrum indicating the intact m/z at 1546.619 for the pseudo-molecular ion for the 6-plex iqPIR reagents. D) MS2 spectrum highlighting the distinct reporter ions generated upon MS fragmentation of a precursor ion containing equal amounts of each of the 6-plex iqPIR molecules. Note coloring in D is to distinguish the reporter ions by highlighting the cross-linker contributing the greatest amount of intensity to the indicated isotope peaks. Due to their overlapping isotope peak envelopes, more than a single isobaric cross-linker contributes intensity to many of the ion isotope peaks. The colors are not related to coloring in panels A or B.

Evaluation of 6-plex iqPIR with multiplexed protein samples

We evaluated the quantitative performance of the 6-plex iqPIR on a set of five purified proteins, including alcohol dehydrogenase (ADH1_YEAST) from S. cerevisiae, albumin (ALBU_BOVIN) from B. taurus, cytochrome C (CYC_BOVIN) from B. taurus, hemoglobin (HMG_HUMAN) from H. sapiens, and myoglobin (MYG_HORSE) from E. ferus. Each protein was cross-linked independently with each of the 6-plex iqPIR reagents as described in the Supplementary Experimental Section. Cross-linked proteins were then mixed at varying ratios to generate two multiplexed protein samples, according to Table 1, prior to tryptic digestion and LC-MS analysis of the resulting peptide samples. The relative ratios of cross-linked peptides spanned the range from 1:1 to 1:12 and included missing channels as well. Example MS2 spectra from an iqPIR cross-linked peptide pair originating from ADH1_YEAST are displayed in Figure 2A. The positions of the cross-linked residues K207 and K226 are indicated on PDB structure 4w6z in Figure 2B. The spectra are displayed as butterfly plots with the MS2 spectrum from multiplexed protein sample 1 on the top and the MS2 spectrum from multiplexed protein sample 2 inverted on the bottom. Major fragment ions including the released peptides A (EK²²⁶DIVGAVLK) and B (VLGIDGGEGK²⁰⁷EELFR), reporter ions as well as b-type and y-type backbone fragment ions from peptides A and B are annotated. Zoomed insets show details of the isotope patterns for 6-plex iqPIR encoded fragment ions. Peptide backbone fragment ions that contain the cross-linked residue, also contain the isotope encoded stump region from the iqPIR reagents and display isotope patterns that can be used for relative quantification. The signal from fragment ions that do not contain the isotope labeled stump region from the iqPIR cross-linkers collapse together into their natural isotope patterns as demonstrated for the y6 ion from peptide A. Signal from the reporter ions is inverted along the x-axis (m/z) relative to the released peptide ions, as the heaviest mass reporter corresponds to the lightest mass peptide and so on. Although the reporter ion signal contains a larger mass spacing between adjacent iqPIR channels (2 or 4 Da) compared with peptide based fragment ions (2 or 1 Da), they are generally not as suitable for quantitation since they suffer from ratio compression resulting from co-isolation of contaminating precursor ions, a phenomenon observed previously for the binary iqPIR¹¹ as well as in traditional isobaric labeling approaches such as TMT¹². Fragment ion spectra from iqPIR cross-linked peptides are rich in quantitative information as every fragment ion containing a cross-linked residue can potentially be used for quantitation. For example, the top (black) and bottom (magenta) individual spectra in Figure 2A contain a total of 22 and 23 quantifiable signals, respectively. Total observed intensity of each ion is apportioned among the six channels to render its relative concentrations.

Table 1 –

6-plex iqPIR cross-linked protein mixtures

6-plex iqPIR multiplexed protein samples with 5 standard proteins							Total (μl)
Multiplexed sample 1	808	812	816	818	822	826
(all 1μg/μL)	μl 1	μl 2	μl 3	μl 4	μl 5	μl 6
MYG	5	15	2.5	2.5	20	0	45
ALBU	20	5	10	5	15	5	60
HBA	2.5	5	15	20	2.5	15	60
CYC	10	20	0	10	10	10	60
ADH1	0	15	2.5	10	30	2.5	60
Multiplexed sample 2	808	812	816	818	822	826
(all 1μg/μL)
MYG	10	5	20	10	5	10	60
ALBU	2.5	0	2.5	30	2	10	47
HBA	30	2	10	0	2	16	60
CYC	0	10	10	20	0	20	60
ADH1	10	10	10	10	10	10	60

Figure 2 – Example 6-plex iqPIR cross-linked peptide pair.

A) MS2 spectra of the cross-linked peptide pair consisting of peptide A = EK²²⁶DIVGAVLK cross-linked with peptide B = VLGIDGGEGK²⁰⁷EELFR from ADH1_YEAST. The spectra are displayed as a butterfly plot with the black spectrum on the top resulting from multiplexed sample 1 (0:6:1:4:12:1 for iqPIR channels 808:812:816:818:822:826) as defined in Table 1 and the magenta inverted spectrum on the bottom resulting from multiplexed sample 2 (equal mixture across all iqPIR channels). Zoomed insets on the top illustrate the details of fragment ion isotope patterns and are color coded according to the iqPIR channels, highlighting the cross-linker contributing the greatest amount of intensity to the indicated isotope peaks in the general case of an ion containing equal amounts of each of the 6-plex iqPIR molecules. Due to their overlapping isotope peak envelopes, more than a single isobaric cross-linker contributes intensity to many of the ion isotope peaks. The y6 ion from peptide A does not contain a cross-linked residue and therefore contains no quantitative value. Fragment ions for intact released peptides A and B as well as y11 from peptide B contain the isotope encoded iqPIR stump mass and display a signature isotope pattern with quantitative information. Likewise the reporter ions also contain quantitative information but are generally not utilized due to compression effects resultant from co-isolation of contaminating precursor ions as noted previously¹¹. B) Structural model for ADH1 (PDB:4w6z) highlighting the position of the cross-linked Lys residues K207 and K226 displayed as green space filled residues. C) Contributions of channels to overlapping stump-containing fragment ion isotope envelope peaks O₀ through O₁₃. Predicted relative isotope peak intensities for the ion in a single channel, based on its chemical composition, are indicated by ξ₀ through ξ₄ for the monoisotopic through fourth offset peak. Illustrated are equal amounts of ion in all channels, such as the case of ADH1_YEAST stump-containing fragment ions in multiplexed protein sample 2.

Method to deconvolute observed ion isotope peak envelopes among contributing cross-linkers

Because ion isotope peaks in more than a single channel often coincide, determining the relative concentration of an ion among the six samples requires deconvoluting the observed overlapping isotope envelopes among the contributing channels, as depicted in Figure 2C for a stump-containing fragment ion of multiplexed protein sample 2 ADH1_YEAST that is equally abundant in the six samples. To achieve this, we adapted the two-channel mode of iqPIR quantitative analysis to accommodate any number of channels with defined ion isotope peak mass offsets. This is done by first determining the theoretical predicted relative intensities of the monoisotopic and first 4 offset peaks, ξ₀ through ξ₄, for each ion based on its chemical composition¹³. The total observed intensity is equal to

$$TotalIntensity=\sum _{\mathrm{i}=0}^{13}{O}_i$$

summed over the 14 combined envelope peaks, O₀ through O₁₃. We then apportion the total ion intensity among all contributing channels to minimize the envelope intensity difference, the overall difference between observed isotope peak relative intensities and those based on the theoretical ion values and channel apportionments:

$$EnvIntensDiff={\sum _{i=0}^{13}(O_i-\sum _{ch=1}^6{Apportionment}_{ch}\sum _{j=0}^4{\xi }_j{\delta }_{i,{Offset}_{ch}+j})}^2$$

where Apportionment_ch is the fraction of total intensity attributed to channel ch, i.e. its relative concentration, ξ_j is the predicted relative intensity of ion offset peak j, Offset_ch is the mass offset in Da between the channel ch monoisotopic peak and that of the 826 Channel (0, 2, 4, 5, 7, and 9 for the six channels, respectively), and Kronecker delta ${\delta }_{i,{Offset}_{ch}+j}$ is equal to 1 when ion offset peak j of channel ch contributes to peak i and 0 otherwise. The sum of all channel apportionments, $\sum _{ch=1}^6{Apportionment}_{ch}$, equals 1. Apportionment of observed ion intensity among the six channels is achieved by estimating an initial apportionment and repeatedly imposing random perturbations to that apportionment, computing its envelope intensity difference, and keeping new apportionments only that reduce the difference over the current best apportionment. Once apportionments are made for an ion, ratios can be calculated for any specified numerator (ch 1) and denominator (ch 2) channel as the ratio of their apportionments:

$${Ratio}_{\raisebox{1ex}{$ch1$}\!\left/ \!\raisebox{-1ex}{$ch2$}\right.}=\raisebox{1ex}{${Apportionment}_{ch1}$}\!\left/ \!\raisebox{-1ex}{${Apportionment}_{ch2}$}\right.$$

Ratios of quantified ions of released peptides and their stump-containing y and b fragments in spectra in which a cross-linked peptide pair is identified are combined together to calculate an average cross-link abundance ratio in the two channel samples. In total 60 non-redundant cross-linked peptide pairs were identified and quantified. Each MS2 spectrum confidently assigned to a cross-linked peptide pair contained on average 15 quantifiable iqPIR encoded isotope clusters, providing multiple opportunities for quantitative comparisons and the ability to generate robust statistics from within a single spectrum Supplementary Figures S7–8.

Relative quantitation of cross-links corresponding to each standard protein was calculated based on ion apportionments of the released peptides and their stump-containing y and b fragments and averaged to yield relative cross-linked protein concentrations among the six channel samples. Figure 3 shows good agreement between the expected and calculated protein relative concentrations among the six channels in the two multiplexed protein samples. Even in cases in which no standard protein was present in a sample, for example ADH1_YEAST in the 808 Channel and CYC_BOVIN in the 816 Channel of multiplexed protein sample 1, the calculated concentrations were close to, or equal to, zero. It should be noted that the actual relative protein concentrations may vary from the expected values shown due to technical variability in steps of sample preparation, including cross-linking, pipetting, and injection volumes in LC, and thus could potentially match even more closely to the calculated concentrations. Calculated relative concentrations of proteins in the 818 and 816 Channels, whose stump-containing fragment ion monoisotopic masses are offset by only 1 Da, are still close to their expected values. The error bars in Figure 3, indicating the standard deviation of calculated channel relative concentrations among contributing cross-links spanning each standard protein of the multiplexed samples, demonstrate the reproducibility of the method. Furthermore, because the cross-linked standard proteins have a wide variety of different relative concentrations in the six channels (Table 1), these results show that the quantitation method is robust to the diversity of cross-link relative concentrations expected in multiplexed biological samples.

Figure 3 –

Expected and calculated relative cross-linked protein concentrations among the six channels (826 through 808) in multiplexed protein samples 1 and 2. Error bars on the calculated values indicate one standard deviation. No HBA_HUMAN cross-links were identified in multiplexed protein sample 1.

An unanticipated protein, TNNI3_BOVIN, was identified and quantified in both multiplexed protein samples. It is likely a contaminant in one of the constituent highly purified standard proteins. Although this protein is bovine, its sequence is highly homologous to that of its human and horse orthologues so could have been present in any standard protein other than ADH1_YEAST. One can see in Figure 3 that the calculated channel apportionments of TNNI3_BOVIN closely match those of CYC_BOVIN in both multiplexed protein samples, suggesting that as its source. In order to assess the significance of similarity between expected and computed apportionments (relative protein concentrations in the six channels) of the cross-linked proteins, a difference (apportionment deviation) between the two is computed as the vector distance where the channels are considered to be independent dimensions:

$$ApportionmentDeviation=\sqrt{\sum _{ch}{({Apportionment}_{ch}-{ActualApportionment}_{ch})}^2}$$

A null distribution of apportionment deviations was then estimated for each multiplexed sample protein by computing the deviations of the protein’s expected apportionment with respect to a set of 1000 random apportionments with channel relative values set with equal probability to an integer value between 0 and 30. With the null deviation distribution a z-score can then be computed for the observed deviation based on the calculated protein channel apportionments. One can see in Supplementary Table S3 that for all multiplexed sample proteins, the deviations of calculated apportionments have very low z-scores consistent with a significantly small value not likely to be due to chance alone. We investigated the apportionment deviation values for each of the standard proteins for the presence of outlier values based on the interquartile range (IQR) test (outlier values defined as those below Q1 − 1.5*IQR or above Q3 + 1.5*IQR) and found no evidence of outliers Supplementary Figure S9. Interestingly, the calculated apportionments of TNNI3_BOVIN in both multiplexed protein samples had deviations with a significantly low z-score with respect to the null deviation distribution of only CYC_BOVIN, supporting its origin in that standard protein. Table 2 shows the search results for the combined MS2 spectra of all five standard proteins that were assigned with high probability to TNNI3_BOVIN. One can see that all of the spectra are from the CYC_BOVIN standard protein, confirming that as its source. That calculated apportionments can be used to identify the source of TNN13_BOVIN demonstrates their great potential for clustering together cross-links with similar relative concentrations in the six samples.

Table 2 –

Search results identifying TNNI3_BOVIN released cross-linked peptides

iprobability	probability	spectrum	ions	peptide	protein	calc_neutral_pep_mass
0.999999	1	101117_cytochromeC_BDP_2hr_1.03831.03831.2	18/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.04152.04152.2	13/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.04179.04179.2	19/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.04580.04580.2	18/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP 2hr_1.05071.05071.1	11/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05073.05073.2	16/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05105.05105.1	13/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05107.05107.2	19/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05153.05153.2	18/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05236.05236.2	14/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05310.05310.2	17/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05375.05375.2	13/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05698.05698.1	12/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05700.05700.2	18/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05728.05728.1	11/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.05730.05730.2	19/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	l	101117_cytochromeC_BDP_2hr_1.05803.05803.2	16/22	R.K[325.13] NIDALSGMEGR. K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	l	l0lll7_cytochromeC_BDP_2hr_l.06027.06027.2	16/22	R.K[325.13] NI DALSGMEGR. K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999999	1	101117_cytochromeC_BDP_2hr_1.06600.06600.2	17/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.999212	0.9994	101117_cytochromeC_BDP_2hr_1.03828.03828.2	12/16	K.KEDTEK[325.13]ENR.E	sp|P08057|TNNI3_BOVIN	1344.57g4
0.998125	0.995	101117_cytochromeC_BDP_2hr_1.05801.05801.1	13/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.996735	0.99l3	101117_cytochromeC_BDP_2hr_1.05234.05234.1	11/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.994441	0.9956	101117_cytochromeC_BDP_2hr_1.01768.01768.2	12/16	K.K[325.13]EDTEKENR.E	sp|P08057|TNNI3_BOVIN	1344.5794
0.99312	0.9817	101117_cytochromeC_BDP_2hr_1.03829.03829.1	12/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.99263	0.9804	101117_cytochromeC_BDP_2hr_1.05373.05373.1	10/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.992414	0.9951	101117_cytochromeC_BDP_2hr_1.01767.01767.2	10/12	R.AHLK[325.13]QVK.K	sp|P08057|TNNI3_BOVIN	1019.54
0.989485	0.9921	101117_cytochromeC_BDP_2hr_1.01766.01766.1	10/16	K.K[325.13]EDTEKENR.E	sp|P08057|TNNI3_BOVIN	1344.5794
0.967101	0.9737	101117_cytochromeC_BDP_2hr_1.03990.03990.2	11/16	K.K[325.13]EDTEKENR.E	sp|P08057|TNNI3_BOVIN	1344.5794
0.958631	0.892	101117_cytochromeC_BDP_2hr_1.06598.06598.1	10/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.942515	0.9537	101117_cytochromeC_BDP_2hr_1.04149.04149.1	8/12	R.AHLK[325.13]QVK.K	sp|P08057|TNNI3_BOVIN	1019.54
0.903968	0.759	101117_cytochromeC_BDP_2hr_1.04578.04578.1	12/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.893492	0.9127	101117_cytochromeC_BDP_2hr_1.01765.01765.1	8/12	R.AHLK[325.13]QVK.K	sp|P08057|TNNI3_BOVIN	1019.54
0.831163	0.6043	101117_cytochromeC_BDP_2hr_1.04150.04150.1	5/22	R.K[325.13]NIDALSGMEGR.K	sp|P08057|TNNI3_BOVIN	1486.6722
0.806749	0.8633	101117_cytochromeC_BDP_2hr_1.04151.04151.2	9/12	R.AHLK[325.13]QVK.K	sp|P08057|TNNI3_BOVIN	1019.54
0.73927	0.7811	101117_cytochromeC_BDP_2hr_1.03830.03830.2	13/16	K.KEDTEK[325.13]ENR.E	sp|P08057|TNNI3_BOVIN	1344.5794

The six channels allow for 30 different pairwise abundance ratios with a designated numerator and denominator channel, with half the ratios being the inverse of the other half. Ratios allow one to designate a reference channel, for example a control sample, to which one or more others, for example drug-treated samples, are compared (See accompanying manuscript Wippel et al. 2021, submitted). A pairwise channel ion ratio is calculated as the ratio of apportionments of the ion in the two channels. For each two-channel ratio, a log2ratio mean and standard deviation are computed for a cross-linked peptide pair based on quantified ion ratios of released peptides and their stump-containing y and b fragments in MS2 spectra in which the cross-link was identified. As previously demonstrated¹¹, dead-end monolinks can be quantified, complementing data from intra-protein links, to provide an estimate of relative protein abundance levels. For multiplexed protein samples 1 and 2 a total of 177 protein level log2ratios were calculated utilizing dead-end mono links and intra-protein cross-links. Figure 4 shows a scatter plot of expected and calculated cross-link (A) and protein (B) log2ratios where the expected ratios are equal to the intended relative protein concentrations in the two channels (Table 1). In order to eliminate redundancy, each channel was included only once at random in a single ratio, as either a numerator or denominator, for each cross-link or protein (See Supplementary Experimental Section). One can see overall good agreement. A small number of calculated cross-link ratios differ widely from the expected ratios. For example, a cross-link in multiplexed protein sample 1 spanning ADH1_YEAST had a calculated log2ratio of 5 in the 826 Channel relative to the 818 Channel, differing widely from the expected log2ratio of 2.09. That is likely because the expected apportionment of ADH1_YEAST in the numerator channel is only 0.04, a small relative intensity that may be lost as noise. Calculated log2ratios corresponding to very large or small expected ratios are often smaller in magnitude due to limitations of detecting zero apportionments. Other differences may be due to the actual ratios being different from the expected values due to variability of sample preparation, as discussed above. 88% of calculated cross-link log2ratios have 95% confidence values within 1, and half within 0.3. This demonstrates reproducibility of iqPIR quantitation conferred by multiple contributing ions. Having the six samples quantified together by qXL-MS in a single multiplexed sample enables cross-links to be quantified in many pairwise channel ratios. For example, half of all quantified cross-links were assigned a log2ratio in 12 or more of the 15 distinct two-channel comparisons.

Figure 4 – Accuracy of calculated multiplexed protein sample cross-link and cross-linked protein pairwise channel ratios.

Calculated versus expected cross-link (A) and protein (B) log2ratios, with the ideal equivalent values indicated as dashed lines. Expected zero and infinite ratios are displayed with log2 values of −5 and 5, the minimum and maximum values of calculated log2ratios, respectively. Error bars indicate one standard deviation. P-values reflecting significance of calculated cross-link (C) and protein (D) ratios differing from those expected for equimolar samples. The p-value corresponding to a Bonferroni corrected 0.01 maximum is indicated by a dashed line (see Supplementary Experimental Section).

P-values computed for the calculated pairwise channel ratios indicating how likely that ratio could be generated by chance from equimolar samples are shown for cross-links (Figure 4C) and proteins (Figure 4D). One can see that significantly low p-values are obtained for the majority of cross-link log2ratios less than −1 or greater than 1, and for all cross-linked protein log2ratios not equal to zero. Calculated log2ratios corresponding to very large or small expected ratios, even with reduced magnitudes, are nevertheless easily detectable as non-equimolar. Interestingly, several log2ratios close to 0 also have low p-values demonstrating consistency of quantitation among their contributing ions.

In order to test the reproducibility of 6-plex iqPIR quantitation on a complex sample, HeLa cells were grown in tissue culture and separately treated with each isobaric cross-linker. These six biological replicate samples were then combined together in a single multiplexed sample and subjected to qXL-MS. This models an experiment such as a time course treatment of a drug in which a minority of cross-links are expected to change in abundance in one sample with respect to another. Whereas the first five channel samples were added at equimolar concentrations, the 808 Channel sample was added at three times that amount in order to simulate large variability of cross-link abundance that may occur during sample preparation. Cross-links were quantified to determine their relative concentrations in the six biological replicate samples. The median cross-link apportionments in the six channels were equal to 0.14, 0.11, 0.12, 0.11, 014, and 0.38, close to their expected relative concentrations of 0.13 for the first 5 channels and 0.38 for the 808 Channel.

Cross-link log2ratios were then calculated for all numerator and denominator channels. Normalization at the cross-link log2ratio level was conducted by subtracting from each log2ratio the median log2ratio among all cross-links. Normalization centers the log2ratio distribution at zero and is justified, as in this case, when most cross-links are not expected to change in abundance in the numerator and denominator channels. Supplementary Figure S10 shows that both normalized cross-link and protein log2ratio values for all two-channel comparisons were, as expected, centered at zero, and largely contained within ±0.75 and ±0.5, respectively. Interestingly, the distributions of the pairwise channel ratios involving the 808 Channel, whose sample was mixed at 3 times the molar amount of other channel samples, were similar to those of other ratios. Normalization was thus able to compensate for the 3-fold difference in amount of applied material to the 808 Channel sample, enabling high quality quantitative comparison with the other channel samples. Reproducibility of the multiplexed iqPIR quantitation is demonstrated by the majority of calculated pairwise channel ratios among the six distinct biological replicates having values close to 1:1. Over 90% (7584/8250) of the log2ratios consistently measured across the six biological replicates with a 95% confidence interval of less than 1, occur within the range of −0.5 to 0.5.

Figure 5 shows the computed p-values for the cross-link and protein level log2ratios indicating their likelihood given a 1:1 ratio. Again, in order to eliminate redundancy, each channel was included only once at random in a single ratio, as either a numerator or denominator, for each cross-link or protein-pair. In contrast to the multiplexed standard protein samples, only small numbers of the 1,052 HeLa cross-links (3%) and 692 protein-pairs (0.6%) have ratios with significantly low p-values, the vast majority of those for log2ratios between −1 and 1. These ratios likely reflect true small differences in the biological replicate samples treated with the different isobaric cross-linkers that are detectable due to the consistency of their contributing ion ratios, and are a testament to the sensitivity of 6-plex iqPIR quantitation. Interestingly, half of quantified cross-links were assigned a log2ratio in 14 or 15 distinct pairwise channel comparisons. This demonstrates the benefit of quantifying the six samples together in a single multiplexed qXL-MS analysis rather than with multiple dual-channel labeling and quantitation that often makes it difficult to quantify the majority of cross-links with all desired pairwise ratios and requires more mass spectrometry and analysis time.

Figure 5 – Significance of calculated multiplexed HeLa sample cross-link and cross-linked protein pairwise channel ratios.

Volcano plots indicating the likelihood that calculated cross-link (A) and protein (B) log2ratios are due to chance from a 1:1 sample comparison. The dashed line indicates the Bonferroni corrected 0.01 maximum threshold below which the p-values are significant.

The presence of isotope labeled fragment ions in tandem mass spectra of 6-plex iqPIR cross-linked peptides results in more complex fragmentation patterns that could potentially affect the ability to identify cross-linked peptides. To evaluate this possibility, we compared the mean XLinkProphet probabilities for 466 cross-linked peptides labeled with 6-plex iqPIR or a non-isotope labeled PIR cross-linker. As shown in Supplementary Figure S11, overall there is a minimal difference in probabilities, with a slight, but insignificant bias (median difference = 4.6E-7, with 95% of all identifications having a difference less than 0.15 introduced by the presence of the isotope peaks. The isotope labeled fragment ions also provide a unique signature for fragment ions containing the cross-linked residues, which could be utilized in future identification efforts, a feature currently being explored for implementation into the Comet search engine¹⁴. Furthermore, despite any minimal impact to identification, the 6-plex iqPIR approach affords the major benefit of high quantitative consistency, as evidenced by greater than 94% of cross-links being quantified across six biological samples of cross-linked HeLa and MCF-7 (Wippel et al. 2021, submitted) samples.

Cross-link ratios for 15 distinct channel pairs of the multiplexed protein and HeLa samples were uploaded to the public cross-link database, XLinkDB^15,16 where they can be viewed in tables 6-plex_iqPIR_5prot_mix_QE_Bruce and 6-plex_iqPIR_HeLa_QE_Bruce, respectively. A tutorial video explaining how to view quantitative iqPIR data within XLinkDB can be seen in Supplementary Video S1. Figure 6A shows part of the heatmap for selected pairwise channel ratios of multiplexed protein samples 1 and 2. Graduated colors of red and green indicate increased and decreased amount of the cross-link in the numerator versus denominator channel samples, respectively. Clicking on a log2ratio shows a summary of ratios of ion types contributing to the cross-link quantitation (Figure 6B). When clicked, it then in turn displays for each quantified ion the observed 14 spectrum peaks of its combined isotope envelope, depicted in black, and to the right of each peak, the combined contributions from the six channels, each depicted in a separate color, based on the calculated relative concentration of the ion among the six channels (Figure 6C). The envelope intensity difference, reflecting the difference, is indicated. This information allows the researcher to confirm the calculated quantitation of all contributing ions. Examples of annotated MS2 spectra with calculated ion apportionments are available in Supplementary Figures S12–14.

Figure 6 – Viewing 6-plex iqPIR quantitation in XLinkDB.

A. Dataset table with heatmap of multiplexed protein sample cross-link pairwise channel log2ratios. B. Log2ratios of ions of both released peptides and their stump-containing y and b fragments contributing to quantitation of multiplexed protein sample 2 ADH1_YEAST cross-link SIGGEVFIDFTK²²⁴EK-VLGIDGGEGK²⁰⁷EELFR in the 818 versus 816 Channels. C. Apportionments of total observed intensities of the stump-containing SIGGEVFIDFTK²²⁴EK y12 fragment and peptide VLGIDGGEGK²⁰⁷EELFR ions among six overlapping channel isotope envelop

CONCLUSIONS

New technological developments in qXL-MS will help researchers illuminate the complex world of protein conformational and interaction dynamics. In the current work we have described the synthesis and application of multiplexed iqPIR cross-linking reagents, allowing for quantitative comparison of up to six samples in a single LC-MS run. The incorporation of ¹³C and ¹⁵N isotopes at selective locations within the iqPIR cross-linkers produces isobaric cross-linked peptides. Upon isolation, fragmentation, and tandem MS analysis, these isobaric peptides display unique isotope patterns which can be used for relative quantitation across iqPIR channels. Importantly, each fragment ion containing the isotope encoded portion of the cross-linker can be used for quantitation, allowing for multiple quantitative measurements and associated statistics to be calculated from a single spectrum.

To enable automated analysis of 6-plex iqPIR data, we present a method for quantifying peptides cross-linked with the six-channel iqPIR cross-linkers. For each cross-linked peptide pair identified in an MS2 spectrum, ions of both released peptides and their stump-containing y and b fragments are apportioned among the six overlapping channel isotope envelopes to minimize the difference with respect to expected peak intensities based on the predicted ion isotope envelopes. Using samples with standard proteins cross-linked with the six isobaric cross-linkers and mixed together at known relative concentrations, we demonstrate that the calculated qXL-MS quantitation is accurate and reproducible over a wide range of relative protein cross-link concentrations. A similar multiplexed analysis of six complex biological replicate HeLa samples resulted in calculated equimolar normalized cross-link concentrations, with pairwise channel log2ratios close to zero. A great benefit of analyzing six samples together for relative quantitation is further demonstrated by obtaining all 15 unique pairwise channel ratios for most quantified cross-links. This method is thus amenable to the quantitative analysis of time courses of environmental perturbations and treatment of cells with increasing doses of drugs. Our algorithm, furthermore, is extensible to quantitation using any number of isobaric cross-linkers with defined stump and reporter masses. XLinkDB was adapted with new features to store and display multiplex iqPIR data, paving the way for its use in large-scale qXL-MS studies such as evaluating drug treatment of breast cancer cells, described in an accompanying manuscript (Wippel et al. 2021, submitted).