Fragment correlation mass spectrometry: Determining the structures of biopolymers in a complex mixture without isolating individual components

Abstract

Fragment correlation mass spectrometry correlates ion pairs generated from the same fragmentation pathway, achieved by covariance mapping of tandem mass spectra generated with an unmodified linear ion trap without preseparation. We enable the identification of different precursors at different charge states in a complex mixture from a large isolation window, empowering an analytical approach for data-independent acquisition. The method resolves and matches isobaric fragments, internal ions, and disulfide bond fragments. We suggest that this method represents a major advance for analyzing structures of biopolymers in mixtures.

Consider a complex biopolymer mixture in which the goal is to determine the molecular structure of each biopolymer by tandem mass spectrometry (1, 2). Whether employing bottom–up or top–down approaches (3), the first step is to separate individual components. If the sample is limited in amount or if the constituents have posttranslational modifications (4), this task is often challenging, especially when compounds coelute in the separation step. Methods based on Fourier-transform ion cyclotron resonance (5) and linear ion trap (6) have been developed to resolve mixtures in two dimensions (2-D): mass-to-charge (m/z: Th) of precursor and product ions. However, when mixtures contain isobaric or multiply charged precursors, these ions cannot be separated by such a 2-D method. We describe here the extension of an approach (7–9) involving the coincident detection of fragment ion pairs generated from the same pathway by correlating their signal intensities.

Recently, data-independent acquisition (DIA) has emerged as a powerful technology for proteomics by cofragmenting all precursor ions in a large isolation window, instead of selecting the most abundant ones for further fragmentation as in data-dependent acquisition (DDA) (10). However, complex fragment ion spectra increase the challenges of data analysis. We demonstrate fragment correlation mass spectrometry for determining the structures of seven peptides, coisolated at different charge states and cofragmented using collision-induced dissociation (CID). This approach does not require mixture components to be preseparated; assigns identical internal ions specifically to their different structures by their different companion ions; and is not hindered by isobaric precursor or fragment peaks.

We exploit covariance mapping techniques to correlate fragment ions generated from the same pathway of the same precursor, whose signals fluctuate in synchrony (11). Mapping of simple covariance is effective to study fragmentation of small molecules (12), but for analysis of biomolecules it becomes overwhelmed by spurious correlations arising from signal fluctuations because of, e.g., ionization source instability (13). Previously, a partial-covariance method was used for DDA of both peptides (7) and proteins (8) including a database search engine (9). In this work, we generalize this technique using contingent covariance, which can additionally suppress undesired correlations produced by nonlinear coupling to fluctuating experimental parameters.

Results

We report a universal method for fragment correlation mass spectrometry using a large isolation window in an unmodified linear ion trap without prior separation. This method exploits intrinsic fluctuations of fragment signals resulting from the stochastic nature of the fragmentation process and inherent to any tandem mass spectra (MS²) measurement. We mixed seven peptides, several of which are structural analogs that are challenging to resolve, including therapeutic cyclic peptides (14) containing disulfides (15). We cofragmented all peptides using a 50-Th isolation window and continuously collected MS² spectra. We calculated contingent covariance between ion intensities, followed by peak centroiding and jackknife resampling to further evaluate the covariance signal-to-noise (S/N) (further details in SI Appendix). Correlation peaks with robust signals, matched to peptide precursor structures, are shown in Fig. 1A as a correlation plot. We show an example in the orange box of Fig. 1B: [b₆]⁺ and [y₃]⁺ signal intensities from desmopressin are correlated (orange), while they are not correlated with [y₃]⁺ of oxytocin (green). Signals of the fragment ion pairs fluctuate synchronously, generating a positive covariance between their intensities. We reconstitute spectra in Fig. 1B as a fingerprint for each precursor structure, the so-called “extracted covariance tandem mass spectrum.” This is achieved by projecting the symmetric correlation plot onto one m/z axis with the added signals using the sum of correlation peak intensities at a certain m/z location.

Fig. 1.

Fragment correlation mass spectrometry. (A) a correlation plot generated from positive contingent covariance between signal intensities in tandem mass spectra. Each data point in the correlation plot indicates a pair of correlated fragments generated by the same pathway from their color-coded precursor. (B) Extracted covariance tandem mass (MS²) spectra were generated by projecting the symmetric correlation plot along the axes for each precursor classification. An example in the orange box indicates synchronizing measurements of fragment pairs in orange. (C) Coverage maps show terminal ions assigned to specific structures (in black slashes) with additional sequence cleavages from correlated isobaric ions, internal ions, and/or disulfides fragments (pink).

We plotted in Fig. 2 the correlated peaks in the covariance maps prior to jackknife resampling. The method can directly measure disulfide bond fragments (octreotide in purple in Fig. 2A as an example), resolve isobaric fragment ions, and match internal ions (Fig. 2 B and C). By comparing covariance peaks (colored peaks on the covariance maps) with the averaged tandem mass spectra peaks (in black traces at the back of the maps), we find that the present method can not only correlate fragments but also uncover more peaks from their nearby chemical noise.

Fig. 2.

Covariance maps generated by calculating contingent covariance matrix from fluctuating intensities of each fragment pair from MS² scans, with averaged mass spectra shown on the back. (A) The generated covariance map at a large m/z range, showing four peaks with the highest covariance S/N ratios after jackknife resampling assigned to four different fragment pairs from four different precursors in the mixture (with pathways in orange and green used in Fig. 1B); (B) isobaric fragments resolved by correlating with companion ions that have different m/z values; (C) internal ions assigned by correlating with terminal ions generated from a pathway that leads to three fragments including an undetected neutral terminal fragment.

Discussion

In the mixture of seven peptides, the structure of each analyte of interest is successfully elucidated in the presence of matrix effects from the other seven peptides. In Fig. 1A, among all 91 peaks in the MS² spectrum of the mixture, 36% of fragment ions are isobaric with at least one ion of a different fragment structure. As shown in the coverage maps, fragment correlation mass spectrometry significantly improves sequence coverage of each compound in the mixture compared to standard tandem mass analysis using terminal ions alone. This happens especially when the structures are difficult to analyze, such as cyclic peptides with disulfide bond connectivity and structures with identical b ions (e.g., dynorphin with its analog).

Note that the internal ion, [bi_(4-5)-H₂O]⁺, is assigned unambiguously to oxytocin in green in Fig. 2B. Although the same internal ion exists in all three compounds in the mixture, the other two compounds have different companion ions correlated with that internal ion (see desmopressin in orange and oxytocin analog in brown).

Our method selectively amplifies previously missing structural signals over neighboring signals generated by chemical or electrical noise. When the fragmentation pathway yields only two fragment ions, then the sum of the two fragment masses equals the precursor mass (Fig. 2B). When it yields three fragments of which one is neutral, it generates two correlated fragment ions, as shown in Fig. 2C. In this case, mass of the neutral loss was deduced by subtracting the sum of the fragment ion pair masses from the mass of the precursor that was matched. As demonstrated here, our method can be immediately applied to resolve mixed samples and improve sequence coverage in MS² measurements of multiply charged precursors. Future developments should leverage high scan rate instruments, such as time-of-flight instrumentation, for tandem mass spectrometry to reduce the required collection time for each correlation map to several seconds. This will enable our method to be coupled to a liquid chromatography (LC-MS²) workflow.

This study empowers DIA by resolving and matching isobaric fragments, internal ions, and disulfide bond fragments without resorting to a modified instrument, high resolution, prior separation, or derivatization.

Materials and Methods

A solution (10 to 100 µM) of seven different peptides was prepared without the need for sample workup. Aliquots of 10 to 30 µL were nanoelectrosprayed into a linear ion trap mass spectrometer in which MS² spectra were obtained in positive mode using CID with a 50-Th isolation width centered at m/z 520. Samples were sprayed continuously until 10,000 MS² scans were collected in less than 28 min in each trial for calculation of a covariance matrix. See SI Appendix for more details.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Datasets generated in this study are available via the Stanford Digital Repository (https://purl.stanford.edu/gs029dv3773) (16). The codes to implement the contingent covariance calculation are available via https://github.com/TaranDriver/FCMS (17).