Application of fractional mass for the identification of peptide-oligonucleotide cross-links by mass spectrometry

Amino Acid Composition Determination From the Fractional Mass of Peptides

A peptide's fractional mass is directly associated with its elemental composition and thus amino acid composition. Here it is demonstrated that a peptide's fractional mass alone can be a useful identifier or indicator of that composition for small to mid-sized peptides (5-7 amino acids) and can significantly reduce the number of viable amino acid compositions for larger peptides (> 8 residues) to include or exclude certain possibilities. Separate consideration of the integer portion of the peptide's mass helps to reduce the number of possibilities where many duplicate fractional mass values are found. Adoption of this fractional mass strategy should aid approaches that are presently employed for peptide identification, including in the use of mass map data to search protein databases for proteomics applications.

Ad hoc learning of peptide fragmentation from mass spectra enables an interpretable detection of phosphorylated and cross-linked peptides

Mass spectrometry-based proteomics provides a holistic snapshot of the entire protein set of living cells on a molecular level. Currently, only a few deep learning approaches exist that involve peptide fragmentation spectra, which represent partial sequence information of proteins. Commonly, these approaches lack the ability to characterize less studied or even unknown patterns in spectra because of their use of explicit domain knowledge. Here, to elevate unrestricted learning from spectra, we introduce ‘ad hoc learning of fragmentation’ (AHLF), a deep learning model that is end-to-end trained on 19.2 million spectra from several phosphoproteomic datasets. AHLF is interpretable, and we show that peak-level feature importance values and pairwise interactions between peaks are in line with corresponding peptide fragments. We demonstrate our approach by detecting post-translational modifications, specifically protein phosphorylation based on only the fragmentation spectrum without a database search. AHLF increases the area under the receiver operating characteristic curve (AUC) by an average of 9.4% on recent phosphoproteomic data compared with the current state of the art on this task. Furthermore, use of AHLF in rescoring search results increases the number of phosphopeptide identifications by a margin of up to 15.1% at a constant false discovery rate. To show the broad applicability of AHLF, we use transfer learning to also detect cross-linked peptides, as used in protein structure analysis, with an AUC of up to 94%.

Constellation: An Open-Source Web Application for Unsupervised Systematic Trend Detection in High-Resolution Mass Spectrometry Data

The increasing popularity of high-resolution mass spectrometry has led to many custom software solutions to process, interpret, and reveal information from high-resolution mass spectra. Although there are numerous software packages for peak-picking, calibration, and formula-finding, there are additional layers of information available when it comes to detecting repeated motifs from polymers or molecules with repeating structures or products of chemical or biochemical transformations that exhibit systematic, serial chemical changes of mass. Constellation is an open-source, Python-based web application that allows the user first to expand their high-resolution mass data into the mass defect space, after which a trend finding algorithm is used for supervised or unsupervised detection of repeating motifs. Many adjustable parameters allow the user to tailor their trend-search to target particular chemical moieties or repeating units, or search for all potential motifs within certain limits. The algorithm has a built-in optimization routine to provide a good starting point for the main trend finding parameters before user customization. Visualization tools allow interrogation of the data and any trends/patterns to a highly specific degree and save publication-quality images directly from the interface. As Constellation is deployed as a web application, it is easily used by anyone with a web browser; no software download or high-powered computer is required, as computations are performed on a remote high-powered data server run by our group.

Mass Spectrometry of Nucleic Acid Noncovalent Complexes

Nucleic acids have been among the first targets for antitumor drugs and antibiotics. With the unveiling of new biological roles in regulation of gene expression, specific DNA and RNA structures have become very attractive targets, especially when the corresponding proteins are undruggable. Biophysical assays to assess target structure as well as ligand binding stoichiometry, affinity, specificity, and binding modes are part of the drug development process. Mass spectrometry offers unique advantages as a biophysical method owing to its ability to distinguish each stoichiometry present in a mixture. In addition, advanced mass spectrometry approaches (reactive probing, fragmentation techniques, ion mobility spectrometry, ion spectroscopy) provide more detailed information on the complexes. Here, we review the fundamentals of mass spectrometry and all its particularities when studying noncovalent nucleic acid structures, and then review what has been learned thanks to mass spectrometry on nucleic acid structures, self-assemblies (e.g., duplexes or G-quadruplexes), and their complexes with ligands.

The Kendrick analysis for polymer mass spectrometry

The mass spectrum of a polymer often displays repetitive patterns with peak series spaced by the repeating unit(s) of the polymeric backbones, sometimes complexified with different adducts, chain terminations, or charge states. Exploring the complex mass spectral data or filtering the unwanted signal is tedious whether performed manually or automatically. In contrast, the now 60-year-old Kendrick (mass defect) analysis, when adapted to polymer ions, produces visual two-dimensional maps with intuitive alignments of the repetitive patterns and favourable deconvolution of features overlaid in the one-dimensional mass spectrum. This special feature article reports on an up-to-date and theoretically sound use of Kendrick plots as a data processing tool. The approach requires no prior knowledge of the sample but offers promising dynamic capabilities for visualizing, filtering, and sometimes assigning congested mass spectra. Examples of applications of the approach to polymers are discussed throughout the text, but the same tools can be readily extended to other applications, including the analysis of polymers present as pollutants/contaminants, and to other analytes incorporating a repetitive moiety, for example, oils or lipids. In each of these instances, data processing can benefit from the application of an updated and interactive Kendrick analysis.

Photodissociative Cross-Linking of Diazirine-Tagged Peptides with DNA Dinucleotides in the Gas Phase

Non-covalent complexes of DNA dinucleotides dAA, dAT, dGG, dGC, and dCG with diazirine-tagged Cys-Ala-Gln-Lys peptides were generated as singly charged ions in the gas phase. Laser photodissociation at 355 nm of the diazirine ring in the gas-phase complexes created carbene intermediates that underwent covalent cross-linking to the dinucleotides. The dinucleotides differed in the cross-linking yields, ranging from 27 to 36% for dAA and dAT up to 90-98% for dGG, dGC, and dCG. Collision-induced dissociation tandem mass spectrometry (CID-MS³) of the cross-linked conjugates revealed that fragmentation occurred chiefly in the dinucleotide moieties, resulting in a loss of a nucleobase and backbone cleavages. The CID-MS³ spectra further revealed that cross-links were primarily formed in the 3'-nucleotides for the dAT, dGC, and dCG combinations. Gas-phase and solution structures of dGG complexes with S-tagged CAQK were investigated by Born-Oppenheimer molecular dynamics (BOMD) and density functional theory calculations. The low free-energy complexes had zwitterionic structures in which the peptide was protonated at the N-terminus and in the Lys residue whereas the carboxyl or dGG phosphate were deprotonated, corresponding to the respective (Cys⁺, Lys⁺, COO^-)⁺ and (Cys⁺, Lys⁺, phosphate^-)⁺ protomeric types. Both types preferred structures in which the peptide N-terminal cysteine carrying the S-photo-tag was aligned with the 3'-guanine moiety. BOMD trajectories at 310 K were analyzed for close contacts of the incipient peptide carbene with the positions in dGG that pointed to frequent contacts with the N-1, NH₂, and N-7 atoms of 3'-guanine, in agreement with the cross-linking results. Carbene insertion to the guanine N-1-H and NH₂ bonds was calculated by density functional and Møller-Plesset perturbational theory to be 350-380 kJ mol^-1 exothermic. Based on calculations, we proposed a mechanism for the carbene reaction with guanine starting with an exothermic attack at N-7 to form a dipolar intermediate that can close an aziridine ring in another exothermic reaction, forming a stable covalent cross link. Graphical Abstract.

Mass Defect from Nuclear Physics to Mass Spectral Analysis

Mass defect is associated with the binding energy of the nucleus. It is a fundamental property of the nucleus and the principle behind nuclear energy. Mass defect has also entered into the mass spectrometry terminology with the availability of high resolution mass spectrometry and has found application in mass spectral analysis. In this application, isobaric masses are differentiated and identified by their mass defect. What is the relationship between nuclear mass defect and mass defect used in mass spectral analysis, and are they the same? Graphical Abstract ᅟ.

Application of the half decimal place rule to increase the peptide identification rate

RATIONALE

Many MS2 spectra in bottom-up proteomics experiments remain unassigned. To improve proteome coverage, we applied the half decimal place rule (HDPR) to remove non-peptidic molecules. The HDPR considers the ratio of the digits after the decimal point to the full molecular mass and results in a relatively small permitted mass window for most peptides.

METHODS

First, the HDPR mass filter was calculated for the human and other proteomes. Subsequently, the HDPR was applied to three technical replicates of an in-solution tryptic digest of HeLa cells which were analysed by liquid chromatography/mass spectrometry (LC/MS) using a quadrupole-orbitrap mass spectrometer (Q Exactive). In addition, the same sample was analysed three times with a fixed exclusion list. The exclusion list was based on only choosing doubly charged ions for fragmentation.

RESULTS

The peptide spectrum match (PSM) rate increased by 2-4% applying HDPR filters from 0.1-0.25 Da and 75-150 ppm, respectively. Excluding all MS2 events by applying an HDPR filter of doubly charged ions, we were able to improve PSMs by 0.9% and the PSM rate by 2.5%.

CONCLUSIONS

An algorithm to filter precursors based on the HDPR was established to improve the targeting of the acquisition of MS2 spectra in data-dependent acquisition (DDA) experiments. According to our data, a total gain of PSMs of 1-5% might be achievable if the HPDR filter would already be applied during MS data acquisition. Copyright © 2016 John Wiley & Sons, Ltd.

Click chemistry generated model DNA-peptide heteroconjugates as tools for mass spectrometry

UV cross-linking of nucleic acids to proteins in combination with mass spectrometry is a powerful technique to identify proteins, peptides, and the amino acids involved in intermolecular interactions within nucleic acid-protein complexes. However, the mass spectrometric identification of cross-linked nucleic acid-protein heteroconjugates in complex mixtures and MS/MS characterization of the specific sites of cross-linking is extremely challenging. As a tool for the optimization of sample preparation, ionization, fragmentation, and detection by mass spectrometry, novel synthetic DNA-peptide heteroconjugates were generated to act as mimics of UV cross-linked heteroconjugates. Click chemistry was employed to cross-link peptides to DNA oligonucleotides. These heteroconjugates were fully characterized by high resolution FTICR mass spectrometry and by collision-induced dissociation (CID) following nuclease P1 digestion of the DNA moiety to a single nucleotide monophosphate. This allowed the exact site of the cross-linking within the peptide to be unambiguously assigned. These synthetic DNA-peptide heteroconjugates have the potential to be of use for a variety of applications that involve DNA-peptide heteroconjugates.

Photo-cross-linking and high-resolution mass spectrometry for assignment of RNA-binding sites in RNA-binding proteins

RNA–protein complexes play pivotal roles in many central biological processes. While methods based on next-generation sequencing have profoundly advanced our ability to identify the specific RNAs bound by a particular protein, there is a dire need for precise and systematic ways to identify RNA interaction sites on proteins. We have developed an integrated experimental and computational workflow combining photo-induced cross-linking, high-resolution mass spectrometry, and automated analysis of the resulting mass spectra for the identification of cross-linked peptides and exact amino acids with their cross-linked RNA oligonucleotide moiety of such RNA-binding proteins. The generic workflow can be applied to any RNA–protein complex of interest. Application to human and yeast mRNA–protein complexes in vitro and in vivo demonstrates the powerful utility of the approach by identification of 257 cross-linking sites on 124 distinct RNA-binding proteins. The software pipeline developed for this purpose is available as open-source software as part of the OpenMS project.

Selective detection of peptide-oligonucleotide heteroconjugates utilizing capillary HPLC-ICPMS

A method for the selective detection and quantification of peptide:oligonucleotide heteroconjugates, such as those generated by protein:nucleic acid cross-links, using capillary reversed-phase high performance liquid chromatography (cap-RPHPLC) coupled with inductively coupled plasma mass spectrometry detection (ICPMS) is described. The selective detection of phosphorus as (31)P(+), the only natural isotope, in peptide-oligonucleotide heteroconjugates is enabled by the elemental detection capabilities of the ICPMS. Mobile phase conditions that allow separation of heteroconjugates while maintaining ICPMS compatibility were investigated. We found that trifluoroacetic acid (TFA) mobile phases, used in conventional peptide separations, and hexafluoroisopropanol/triethylamine (HFIP/TEA) mobile phases, used in conventional oligonucleotide separations, both are compatible with ICPMS and enable heteroconjugate separation. The TFA-based separations yielded limits of detection (LOD) of ~40 ppb phosphorus, which is nearly seven times lower than the LOD for HFIP/TEA-based separations. Using the TFA mobile phase, 1-2 pmol of a model heteroconjugate were routinely separated and detected by this optimized capLC-ICPMS method.

Investigation of protein-RNA interactions by mass spectrometry--Techniques and applications

Protein-RNA complexes play many important roles in diverse cellular functions. They are involved in a wide variety of different processes in growth and differentiation at the various stages of the cell cycle. As their function and catalytic activity are directly coupled to the structural arrangement of their components--proteins and ribonucleic acids--the investigation of protein-RNA interactions is of great functional and structural importance. Here we discuss the most prominent examples of protein-RNA complexes and describe some frequently used purification strategies. We present various techniques and applications of mass spectrometry to study protein-RNA complexes. We discuss the analysis of intact complexes as well as proteomics-based and crosslinking-based approaches in which proteins are cleaved into smaller peptides. This article is part of a Special Section entitled: Understanding genome regulation and genetic diversity by mass spectrometry.

Improving mass defect filters for human proteins

The mass defect of a substance can be used in mass spectral analysis to identify peaks as likely belonging to a compound class, such as peptides, if the mass defect is within the known range for that compound class. For peptides, a range of possible mass defects was calculated previously, using a set of theoretical peptides, where all possible amino acid combinations were considered (Mann , M. Abstract from the 43(rd) Annual Conference on Mass Spectrometry and Allied Topics; Conference Proceedings , 1995). We compare that range of theoretical peptide mass defects to new values obtained from in silico tryptic digests of proteins that are abundant in human serum and human seminal fluid. The range of mass defect values encompassing 95% of peptides for the human protein data sets was found to be up to 50% smaller than the previously reported mass defect range for the theoretical peptides. The smaller range established for human tryptic peptides can be used to improve peptide mass defect filters by excluding more species that are not likely to be peptides, thus improving filter selectivity for peptides during proteomic data analysis.

Method for the identification of lipid classes based on referenced Kendrick mass analysis

A rapid method for the determination of lipid classes with high sensitivity is described. The referenced Kendrick mass defect (RKMD) and RKMD plots are novel adaptations of the Kendrick mass defect analysis that allows for the rapid identification of members of a homologous series in addition to identifying the lipid class. Assignment of lipid classes by the RKMD method is accomplished by conversion of the lipid masses to the Kendrick mass scale and then referencing the converted masses to each lipid class. Referencing of the masses to a given lipid class is achieved by first subtracting the heteroatom and lipid backbone contributions to the mass defect, leaving behind the contribution to the mass by the fatty acid constituents. The final step in the referencing makes use of spacing differences in mass defects between members of the same Kendrick class to identify members of the lipid class being referenced. The end result of this is that a lipid belonging to the class being referenced will have an integer RKMD with the value of the integer being the degrees of unsaturation in the lipid. The RKMD method was able to successfully identify the lipids in an idealized data set consisting of 160 lipids drawn from the glyceride and phosphoglyceride classes. As a real world example the lipid extract from bovine milk was analyzed using both accurate mass measurements and the RKMD method.

Sequence analysis of peptide:oligonucleotide heteroconjugates by electron capture dissociation and electron transfer dissociation

Mass spectrometry analysis of protein-nucleic acid cross-links is challenging due to the dramatically different chemical properties of the two components. Identifying specific sites of attachment between proteins and nucleic acids requires methods that enable sequencing of both the peptide and oligonucleotide component of the heteroconjugate cross-link. While collision-induced dissociation (CID) has previously been used for sequencing such heteroconjugates, CID generates fragmentation along the phosphodiester backbone of the oligonucleotide preferentially. The result is a reduction in peptide fragmentation within the heteroconjugate. In this work, we have examined the effectiveness of electron capture dissociation (ECD) and electron-transfer dissociation (ETD) for sequencing heteroconjugates. Both methods were found to yield preferential fragmentation of the peptide component of a peptide:oligonucleotide heteroconjugate, with minimal differences in sequence coverage between these two electron-induced dissociation methods. Sequence coverage was found to increase with increasing charge state of the heteroconjugate, but decreases with increasing size of the oligonucleotide component. To overcome potential intermolecular interactions between the two components of the heteroconjugate, supplemental activation with ETD was explored. The addition of a supplemental activation step was found to increase peptide sequence coverage over ETD alone, suggesting that electrostatic interactions between the peptide and oligonucleotide components are one limiting factor in sequence coverage by these two approaches. These results show that ECD/ETD methods can be used for the tandem mass spectrometry sequencing of peptide:oligonucleotide heteroconjugates, and these methods are complementary to existing CID methods already used for sequencing of protein-nucleic acid cross-links.

Non-linear classification for on-the-fly fractional mass filtering and targeted precursor fragmentation in mass spectrometry experiments

MOTIVATION

Mass spectrometry (MS) has become the method of choice for protein/peptide sequence and modification analysis. The technology employs a two-step approach: ionized peptide precursor masses are detected, selected for fragmentation, and the fragment mass spectra are collected for computational analysis. Current precursor selection schemes are based on data- or information-dependent acquisition (DDA/IDA), where fragmentation mass candidates are selected by intensity and are subsequently included in a dynamic exclusion list to avoid constant refragmentation of highly abundant species. DDA/IDA methods do not exploit valuable information that is contained in the fractional mass of high-accuracy precursor mass measurements delivered by current instrumentation.

RESULTS

We extend previous contributions that suggest that fractional mass information allows targeted fragmentation of analytes of interest. We introduce a non-linear Random Forest classification and a discrete mapping approach, which can be trained to discriminate among arbitrary fractional mass patterns for an arbitrary number of classes of analytes. These methods can be used to increase fragmentation efficiency for specific subsets of analytes or to select suitable fragmentation technologies on-the-fly. We show that theoretical generalization error estimates transfer into practical application, and that their quality depends on the accuracy of prior distribution estimate of the analyte classes. The methods are applied to two real-world proteomics datasets.

AVAILABILITY

All software used in this study is available from http://software.steenlab.org/fmf

CONTACT

hanno.steen@childrens.harvard.edu

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Shifting unoccupied spectral space in mass spectrum of peptide fragment ions

Ions near the high-end border of a mass defect distribution plot for native peptide fragment ions have potential as signature markers that are based on mass-to-charge ratio determination. The specificity of these marker ions, including phosphoryl ions, can be improved by removing interfering isobaric ions from the border region on the distribution plot. These interfering ions are rich in Asp and Glu content. The masses of amino acid residues and peptides are rescaled from the IUPAC scale (12C = 12 u as the mass reference) to the averagine scale (averagine mass = 111 u* as the mass reference with zero mass defect; u*: the mass unit on the averagine scale), using a scaling factor of 0.999493894. It is theoretically predicted that esterification of Asp and Glu side-chain carboxylates with n-butanol can achieve a sufficient retreat of the high-end border on a mass defect distribution plot based on the use of mass spectrometers with better-than-medium resolution. Theoretical calculations and laboratory experiments are performed to examine effects of various esterifications on the averagine-scale mass defect distribution of peptide fragment ions and on the specificity of two positive phosphoryl ions: the phosphotyrosine immonium ion and a cyclophosphoramidate ion.