Peptide Fingerprinting Using Single-Molecule Fluorescence

Aberrant post-translational modifications (PTMs) on arginine residues, such as methylation and citrullination, are closely linked to oncogenic processes. Detection of these PTMs is challenging with current technology and can be hampered by low abundance and mass differences that are difficult to resolve by mass spectrometry. To overcome these challenges, we developed methods for PTM detection using Quantum-Si's Platinum single-molecule protein sequencing platform. Proteins are sequenced on Quantum-Si's Platinum instrument using our library prep protocol. Briefly, proteins are digested into peptide fragments, conjugated to linkers, then immobilized at the bottom of nanoscale reaction chambers on a semiconductor chip, resulting in exposed N-termini for sequencing. During sequencing, the immobilized peptides are exposed to a solution of dye-labeled N-terminal amino acid (NAA) recognizers that reversibly bind to their cognate NAAs with distinct kinetic properties. Aminopeptidases sequentially remove NAAs to expose subsequent amino acids for recognition. Fluorescence lifetime, intensity, and kinetic data are collected in real time and analyzed to determine primary sequence and PTM content. The data output consists of distinct pulsing regions called recognition segments (RSs). Each RS corresponds to a period of time between aminopeptidase cleavage events, during which an NAA recognizer binds on/off to its exposed target NAA. Chemical modification to a NAA or to a nearby downstream amino acid can modulate recognizer affinity, resulting in a characteristic change in the average pulse duration (PD) during an RS. To demonstrate the detection and differentiation of arginine PTMs, we applied our platform to distinguish between asymmetric dimethylarginine (ADMA) and symmetric dimethylarginine (SDMA) and between citrulline and native arginine residues. Sequencing traces revealed that arginine and ADMA bind the arginine recognizer (PS621) with similar PD, whereas SDMA exhibited no binding. These results indicate that SDMA, in contrast to ADMA, has reduced binding affinity with PS621, providing a clear kinetic difference between these isomeric arginine PTMs. Moreover, citrulline and arginine side chains also exhibited distinguishable kinetic signatures. Citrullination eliminated the N-terminal arginine recognized by PS621 and resulted in a large increase in the median PD of preceding amino acid residues. Taken together, single-molecule protein sequencing offers an alternative approach to detection of arginine PTMs that is not based on m/z, but rather on the kinetic signature of binding between recognizers and N-terminal amino acids. The ability to directly detect arginine PTMs offers potential for biomedical research. We envisage applications using Quantum-Si's platform to key areas of cancer research including biomarker development and drug discovery. Citation Format: Kenneth Skinner, David Kamber, Brian Reed. Detection of arginine posttranslational modifications by single-molecule protein sequencing on the Quantum-Si platform. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5301.

The vast complexity of the proteome currently overwhelms any single analytical technology in capturing the full spectrum of proteoform diversity. In this study, we evaluated the complementarity of two cutting-edge proteomic technologies—single-molecule protein sequencing and individual ion mass spectrometry—for analyzing recombinant human IL-6 (rhIL-6) at the amino acid, peptide, and intact proteoform levels. For single-molecule protein sequencing, we employed the recently released Platinum instrument. Next-generation protein sequencing (NGPS) on Platinum utilizes cycles of N-terminal amino acid recognizer binding and aminopeptidase cleavage to enable parallelized sequencing of single peptide molecules. We found that NGPS produces single amino acid coverage of multiple key regions of IL-6, including two peptides within helices A and C, which harbor residues that reportedly impact IL-6 function. For top-down proteoform evaluation, we used individual ion mass spectrometry (I2MS), a highly parallelized Orbitrap-based charge detection MS platform. Single ion detection of gas-phase fragmentation products (I2MS2) gives significant sequence coverage in key regions of IL-6, including two regions within helices B and D that are involved in IL-6 signaling. Together, these complementary technologies delivered a combined 52% sequence coverage, offering a more complete view of IL-6 structural and functional diversity than either technology alone. This study highlights the complementarity of these protein detection methods to cover protein segments relevant to biological interactions more comprehensively.Graphical Supplementary InformationThe online version contains supplementary material available at 10.1007/s00216-025-06120-7.

ProtSeq: Toward high-throughput, single-molecule protein sequencing via amino acid conversion into DNA barcodes

Thioredoxins are small (about 12 kDa) proteins which reduce disulphide bridges on target enzymes [6]. Protein or nucleic acid sequences have been established for different thioredoxins from vertebrates [7, 8, 15, 16], algae [3, 4, 12] and bacteria [6, 10]. The active site shows a highly conserved sequence -C-G-P-C-. Recently a larger consensus thioredoxin signature (AT)-x-W-C(AG)-(PH)-C has been defined in the PROSITE database [ 1 ]. It matches all thioredoxins but no other protein sequences. In higher plants a spinach f-type cDNA [9], an m-type cDNA [ 14] and an m-type protein [ 11 ] have been sequenced. The f-type has a signal peptide typical of chloroplast targeting. Both enzymes are encoded in the nucleus, but are localized and active in the chloroplast in their mature form. M-type activates NADP-malate dehydrogenase wild f-type activates enzymes of the photosynthetic carbon cycle. In addition, cytosolic thioredoxin activities have been characterized in spinach. The proteins have been purified but not yet sequenced [5]. We have isolated a complete or nearly complete clone from a cDNA library from Nicotiana tabacum cells cultivated in vitro. This 699 nucleotide insert hybridizes on northern blots with a 0.9 kb mRNA. It contains an ORF of 378 bp (Fig. 1). The 3' untranslated region is 242bp long, including 16bp of poly(A) ÷ tail. Two AT-rich regions could serve as polyadenylation signals. The ORF codes for a protein of 126 amino acids. At positions 43-49 it contains a sequence which perfectly matches the thioredoxin signature. In addition, the protein sequence shows 40-50 ~o homologies with Chlamydomonas Chl assumed to be a cytoplasmic protein [4] and vertebrate thioredoxins using FASTA [13]. In contrast, the homology with m and f spinach chloroplast thioredoxins is only 30~o. Figure2 presents a multiple alignment [2] with some of the published thioredoxin sequences sorted by decreasing homology with the tobacco sequence. Chlamydomonas [3, 4], rabbit [7], Anacystis [ 12], Corynebacterium [ 10] and m-type spinach 2 [ 11] sequences were obtained by protein sequencing. Human [16], rat [15], chicken [8], Escherichia coli [6], f-type [9] and m-type 1 [14] spinach sequences are derived from cDNA sequencing, showing the presence of a large signal peptide in the spinach thioredoxins. The tobacco sequence matches the mature protein sequences and has no signal peptide, suggesting that the protein encoded by the clone we have isolated is cytoplasmic.

Adsorptive Biphasic Column Technology for Protein Sequence Analysis and Protein Chemical Modification

Protein sequencing techniques

ABSTRACTGrowth in the pharmaceutical industry has led to an increasing demand for rapid characterization of therapeutic monoclonal antibodies. The current methods for antibody sequence confirmation (e.g., N-terminal Edman sequencing and traditional peptide mapping methods) are not sufficient; thus, we developed a fast method for sequencing recombinant monoclonal antibodies using a novel digestion-on-emitter technology. Using this method, a monoclonal antibody can be denatured, reduced, digested, and sequenced in less than an hour. High throughput and satisfactory protein sequence coverage were achieved by using a non-specific protease from Aspergillus saitoi, protease XIII, to digest the denatured and reduced monoclonal antibody on an electrospray emitter, while electrospray high voltage was applied to the digestion mixture through the emitter. Tandem mass spectrometry data was acquired over the course of enzyme digestion, generating similar information compared to standard peptide mapping experiments in much less time. We demonstrated that this fast protein sequencing method provided sufficient sequence information for bovine serum albumin and two commercially available monoclonal antibodies, mouse IgG1 MOPC21 and humanized IgG1 NISTmAb. For two monoclonal antibodies, we obtained sequence coverage of 90.5–95.1% for the heavy chains and 98.6–99.1% for the light chains. We found that on-emitter digestion by protease XIII generated peptides of various lengths during the digestion process, which was critical for achieving sufficient sequence coverage. Moreover, we discovered that the enzyme-to-substrate ratio was an important parameter that affects protein sequence coverage. Due to its highly automatable and efficient design, our method offers a major advantage over N-terminal Edman sequencing and traditional peptide mapping methods in the identification of protein sequence, and is capable of meeting an ever-increasing demand for monoclonal antibody sequence confirmation in the biopharmaceutical industry.

The Age of the Proteome

At the MPSA 1988 in Berlin we presented a general strategy for the combined use of mass spectrometric molecular weight information and automatic Edman degradation for sequence determination of proteins (Roepstorff et al., 1989). Among the mass spectrometric methods available then, i.e. plasma desorption mass spectrometry (PDMS) and fast atom bombardment mass spectrometry (FABMS), we had chosen PDMS because this method in our experience yielded better sensitivity and higher practical mass range than FABMS. The mass accuracy although not as good as for FABMS was in most cases adequate. PDMS had further advantages over FABMS in simplicity in sample preparation, instrument operation and data interpretation. This has been of utmost importance for our routine use of mass spectrometry in protein sequencing. In the past years we have sequenced several proteins using this strategy (Andersen et al., in press; Andreassen et al., submitted; Hojrup et al., 1986; 1991; Klarskov et al., 1989; Mikkelsen et al., 1989; Parello et al., in preparation; Talbo et al., 1990).

Requirements for protein sequence analysis have dramatically changed in the last decade. Before the development of DNA sequencing methods, complete protein sequences could only be determined by combining extended N-terminal sequence information obtained from purified proteins with the sequences of overlapping cleavage fragments. Micromoles of purified protein and a tenacious protein chemist were needed to fill in the last gaps in the sequence. To minimize the number of overlaps required, the primary focus for improvement of protein sequencing methods was to devise instrumentation and methodologies permitting longer sequencing runs (Frank, 1982).

Traditional "bottom-up" proteomic approaches use proteolytic digestion, LC-MS/MS, and database searching to elucidate peptide identities and their parent proteins. Protein sequences absent from the database cannot be identified, and even if present in the database, complete sequence coverage is rarely achieved even for the most abundant proteins in the sample. Thus, sequencing of unknown proteins such as antibodies or constituents of metaproteomes remains a challenging problem. To date, there is no available method for full-length protein sequencing, independent of a reference database, in high throughput. Here, we present Database-independent Protein Sequencing, a method for unambiguous, rapid, database-independent, full-length protein sequencing. The method is a novel combination of non-enzymatic, semi-random cleavage of the protein, LC-MS/MS analysis, peptide de novo sequencing, extraction of peptide tags, and their assembly into a consensus sequence using an algorithm named "Peptide Tag Assembler." As proof-of-concept, the method was applied to samples of three known proteins representing three size classes and to a previously un-sequenced, clinically relevant monoclonal antibody. Excluding leucine/isoleucine and glutamic acid/deamidated glutamine ambiguities, end-to-end full-length de novo sequencing was achieved with 99-100% accuracy for all benchmarking proteins and the antibody light chain. Accuracy of the sequenced antibody heavy chain, including the entire variable region, was also 100%, but there was a 23-residue gap in the constant region sequence.

Thioacetylation method of protein sequencing: Derivatization of 2-methyl-5(4 H)-thiazolones for high-performance liquid chromatographic detection

The N-terminal amino acid sequence of a 42.5 kDa subunit of the NADH: ubiquinone oxidoreductase (complex I) from potato has been determined by direct protein sequencing. The sequence was found to be homologous to that of the nuclear-encoded 49 kDa complex I subunit of bovine and Neurospora mitochondria and to the sequence deduced from the mitochondrial nad7 gene identified in the mitochondrial (mt) DNA of tryp anosomes and the moss Marchantia. An oligonucleotide probe derived from the potato N-terminal protein sequence hybridized only to the plant mtDNA. Immunoprecipitation of in-organello 35S-labelled potato and wheat mitochondrial translation products with an antibody directed against the Neurospora 49 kDa complex I subunit indicates that at least in these plants the NAD7 protein is synthesized within the organelle. Comparisons of genomic, cDNA and protein sequences of the 5' coding region reveal three codons that are changed by RNA-editing and confirm translation of the edited transcripts in plant mitochondria. The NAD7 protein appears to undergo post-translational processing since the N-terminal methionine residue is absent from the mature mitochondrial protein.

Three overlapping cDNA clones encoding methylmalonate-semialdehyde dehydrogenase (MMSDH; 2-methyl-3-oxopropanoate:NAD+ oxidoreductase (CoA-propanoylating); EC 1.2.1.27) have been isolated by screening a rat liver lambda gt 11 library with nondegenerate oligonucleotide probes synthesized according to polymerase chain reaction-amplified portions coding for the N-terminal amino acid sequence of rat liver MMSDH. The three clones cover a total of 1942 base pairs of cDNA, with an open reading frame of 1569 base pairs. The authenticity of the composite cDNA was confirmed by a perfect match of 43 amino acids known from protein sequencing. The composite cDNA predicts a 503 amino acid mature protein with M(r) = 55,330, consistent with previous estimates. Polymerase chain reaction was used to obtain the sequence of the 32 amino acids corresponding to the mitochondrial entry peptide. Northern blot analysis of total RNA from several rat tissues showed a single mRNA band of 3.8 kilobases. Relative mRNA levels were: kidney greater than liver greater than heart greater than muscle greater than brain, which differed somewhat from relative MMSDH protein levels determined by Western blot analysis: liver = kidney greater than heart greater than muscle greater than brain. A 1423-base pair cDNA clone encoding human MMSDH was isolated from a human liver lambda gt 11 library. The human MMSDH cDNA contains an open reading frame of 1293 base pairs that encodes the protein from Leu-74 to the C terminus. Human and rat MMSDH share 89.6 and 97.7% identity in nucleotide and protein sequence, respectively. MMSDH clearly belongs to a superfamily of aldehyde dehydrogenases and is closely related to betaine aldehyde dehydrogenase, 2-hydroxymuconic semialdehyde dehydrogenase, and class 1 and 2 aldehyde dehydrogenases.

Classifying protein sequences from biological data has lot of importance in the field of pharmacology. The application of machine learning to find the sequence of amino acids has recently received popularity from various researchers. This chapter proposes a protein sequence classification technique using 1D Convolutional Neural Network (CNN). We also have discussed how NLP algorithms can be used for protein sequencing. We have achieved an accuracy of 85% with the proposed 1D CNN and further improved to 92% after increasing the filter size.KeywordsProtein sequenceCNNMachine learningCNN—1DNLP