Exploring the diversity of cysteine-rich natural product peptides via MS/MS fingerprint ions

Abstract

Natural product extracts present inherently complex matrices in which the identification of novel bioactive peptide species is challenged by low abundance masses and significant structural and sequence diversity. Additionally, discovery efforts often result in the re-identification of known compounds, where modifications derived in vivo or during sample handling may obscure true sequence identity. Herein, we identify mass spectral (MS²) “fingerprint” ions characteristic of cyclotides, a diverse and biologically active family of botanical cysteine-rich peptides, based on regions of high sequence homology. We couple mass shift analysis with MS² spectral fingerprint ions cross referenced with CyBase - a cyclotide database - to discern unique mass species in Viola communis extracts from mass species that are likely already characterized and those with common modifications. The approach is extended to a related class of cysteine-rich peptide, the trypsin inhibitors, using the characterized botanical species Lagenaria siceraria. Coupling the observation of highly abundant MS² ions with mass shift analysis, we identify a new set of small, highly disulfide bound cysteine-rich L. siceraria peptides.

Graphic synopsis:

We identify mass spectral (MS²) “fingerprint” ions characteristic of cyclotides, a diverse and biologically active family of botanical cysteine-rich peptides, based on regions of high sequence homology. Mass shift analysis is coupled with MS² spectral fingerprint ions to discern unique mass species in Viola extracts from mass species that are likely already characterized and those with common modifications.

Introduction

A global rise in antimicrobial resistance necessitates the identification and characterization of novel antimicrobial compounds to supplement an increasingly dry drug pipeline. Natural product peptides are a source of antimicrobial therapeutics with unique chemistries; however, highly complex natural product matrices, inherent sequence variability, and low abundance challenge the identification of new antimicrobial peptide (AMP) sequences. Antimicrobial peptide secondary/tertiary structure, charge, post-translational modification, and molecular weight are non-uniform, ranging from 2–4 residue, negatively-charged non-ribosomally synthesized, linear bacterial lipopeptides to >40 residue, basic, and highly disulfide-bound plant defensins. Additionally, the expression and abundance of AMPs are dynamic, influenced by environmental factors (e.g. season, stress) and circadian rhythms. Thus, despite the >3000 AMPs reported among databases (dbaasp.org,¹ http://aps.unmc.edu/AP/²), the chemical suites of many sources remain obscured by extract complexity, diversity, and periodicity, necessitating new approaches to explore the true AMP sequence diversity in nature.

Mass spectrometry (MS) drives primary AMP discovery, where modern instruments coupled with ultra-high performance liquid chromatography (UHPLC) can identify and resolve low abundance peptidyl species in complex mixtures with unprecedented limits of detection, resolving power, and mass accuracy. Advances in multistage MS (MSⁿ) allow for the determination of extract constitution at the MS¹ level as well as the identity of constituents via tandem MS. Fragmentation methods (e.g. collision-induced dissociation [CID], electron-transfer dissociation [ETD]) and orthogonal chemical derivatization strategies enable the characterization of diverse and difficult-to-sequence mass species without prior isolation. Even so, the hundreds to thousands of compounds present in natural product extracts as detected via LC-MS complicate comprehensive AMP identification/sequence elucidation, and additional strategies are needed to explore the depths of natural product peptide sequence diversity. Conserved features among AMP families can inform MS-based methods towards the exploration of novel sequences in natural product extracts.

Cysteine-rich peptides (CRPs) represent the majority of characterized botanical natural product peptides³ and are produced extensively in plants to mediate various elements of physiology, reproduction, growth/development, and defense against pathogens.^4–8 Disulfide bonding among cysteine residues scaffolds these highly stable and enzymatically-resistant molecules; CRPs are typically classified by cysteine number and pattern along the backbone.³ Chemical derivatization strategies can be exploited to detect the presence of CRPs in complex matrices by mass shift analysis, where the analysis of extracts before and after the reduction and alkylation of Cys residues produces an easily observable characteristic mass shift. Mass shift analysis is often the first step towards the identification of CRPs in botanical extracts,^9,10 however this alone can lead to under- or over-estimation of the number of unique CRPs given the high complexity of botanical matrices and the presence of low abundance/modified (e.g. oxidized) CRPs. Additionally, low quality MS/MS spectra typical of untargeted, data-dependent experiments may preclude full sequence characterization.

Mass spectral fingerprinting is a method of profiling MS ions as chemical signatures that are characteristic of classes of molecules, organisms, or other target group, and can be derived from in silico database-based prediction or from experimental data. Currently, MS fingerprinting is primarily performed at the MS¹ level, without or complementary to MS/MS data. Database-based protein identification with peptide mass fingerprinting often employs in-vitro enzyme digestion prior to MS, where a unique combination of peptides is used as a chemical “fingerprint” of a given protein when compared to those predicted from the database. With this approach, increasing sample complexity generates increasing peptide interference and may challenge accurate protein identification/reduce confidence in protein assignments.¹¹ Spectral patterns acquired via MALDI-TOF and processed with advanced statistical algorithms have proven invaluable as a tool to discriminate bacterial^12–19 and fungal^20–22 species, including drug resistant strains, though the specific molecular origin of diagnostic ions is often unknown in such chemometric approaches. MS fingerprinting has been exploited extensively in the field of natural products to determine chemical composition as it relates to the geographical origin^23,24 and purity/adulteration^25–29 of a variety of natural substances. Additionally, it has been applied to the identification of CRPs at the MS¹ level to reveal misidentification/mislabeling of herbal medicines.³⁰ MS/MS data is routinely collected during LC-MS experiments and can provide information about the type/characteristics of components in a complex matrix without complete sequence elucidation. Common and abundant CRP MS/MS fingerprint ions can be leveraged to prioritize the identification of novel peptidyl species and reduce the re-characterization of known CRPs.

Knottins are low molecular weight (~3 kDa), functionally diverse CRPs characterized by three conserved disulfide bonds and a canonical ‘knotted’ cysteine connectivity of C1-C4, C2-C5, and C3-C6.³¹ Peptides conforming to the knottin structural class are further subdivided into families based on sequence homology: cyclotides are head-to-tail cyclized knottins found largely in the Violaceae, Rubiaceae, and Cucurbitaceae botanical families. Cyclotides possess diverse, innate bioactivities with >500 sequences known, and thousands more are thought to await discovery. Three cyclotide subfamilies exist (bracelets, Möbius, and trypsin inhibitors) where regions between the six Cys residues (loops 1–6) facilitate sequence comparison among cyclotide species (Figure 1). Largely homologous, the bracelet and Möbius cyclotides are differentiated by the absence or presence of a cis-Trp/Pro or cis-Tyr/Pro bond in loop five, respectively, in addition to some regions of sequence homology specific to each subfamily. Despite sharing a cyclic knot structure, trypsin inhibitor sequences are more dissimilar to the aforementioned subfamilies and offer a comparatively limited range of bioactivities.³²

Figure 1.

Fingerprint ions derived from the Glu-C digestion and b/y fragmentation of the prototypical cyclotides cyO2 (top, bracelet) and kalata B2 (bottom, Möbius). Fragment ion color-coding aligns to the sequence colors; below the fragment ion m/z is the chance of finding that ion in the MS² of a non-cyclotide (% occurrence), followed by the chance of finding that ion in combination with those preceding it towards the nearest terminus (% combined) in a non-cyclotide. Loop regions between cysteine residues are numbered above and below sequences in grey and allow for facile sequence comparison among cyclotide sequences; cysteine connectivites, conserved among all cyclotides, are represented by black bars above the cyO2 sequence. Structures of cyO2 (PDB 2KNM) and kalata B2 (PDB 1PT4) modeled in PyMol (Schrödinger) are shown to the left of the corresponding sequences, where the Glu cleavage site for linearization is represented by a blue star.

The number and identity of CRP species, such as cyclotides, can dramatically vary among natural product extracts, where the comprehensive characterization of the suite of CRP masses in a single botanical species can challenge even the most sensitive and accurate analytical methods. Cysteine derivatization and mass shift analysis are commonly used to identify CRP constituents, and cyclic CRPs may require in vitro enzymatic digestion (e.g. endoproteinase Glu-C) to cleave the cyclic backbone and facilitate MS² sequence characterization. Regardless, complete sequence characterization is often challenged by low abundance cyclotide species in complex matrices, numerous sample preparation steps, and incomplete MS/MS fragmentation.

Regions of sequence homology among CRPs can be exploited to guide the identification of novel mass species and reduce the time-intensive sequence characterization of known CRPs, where shared sequence motifs are reflected in MS/MS spectra as “fingerprint” ions. Herein, theoretical MS/MS fingerprint ions most representative of the bracelet and Möbius cyclotide subfamilies are identified: bracelet/Möbius-indicative fingerprint ions coupled with the knowledge of sequence characteristics specific to each subfamily may aid sequencing efforts of novel mass species. Additionally, the Möbius subfamily represents only ~1/3 of characterized cyclotides; as such, Möbius-specific fingerprint ions may be utilized to target masses belonging to this under-characterized subfamily. Proof-of-principle is achieved by coupling a mass shift analysis, where CRPs are identified by a +348.16 Da mass shift consistent with the reduction and alkylation of three disulfide bonds, with a “fingerprint” analysis of the reduced, alkylated, and enzymatically-digested botanical extracts from the cyclotide-producing species Viola communis. Common m/z fingerprints observed with these data are discussed and sequence information derived from fingerprint analysis is cross-referenced with CyBase, the database of cyclic proteins,³³ to identify sequences of interest. False discovery rate is determined via in silico fragmentation of the Swissprot database and enumeration of the instances in which these diagnostic fingerprint ions would be produced by a non-cyclotide mass. Over 500 species of Viola can be found in nature, each with a diverse and complex suite of expressed cyclotides; environmental variation may be reflected in the cyclotide expression in even a single plant. As such, wild-grown Viola aerial material from two independent harvests was chosen to demonstrate the utility of MS/MS fingerprint ions, focusing on the bracelet and Möbius cyclotides abundant in Viola spp. Without full sequence characterization or targeted runs, and only low quality MS/MS spectra resulting from poor fragmentation and low abundance fragment ions, we demonstrate that a rapid assessment of MS/MS fingerprint ions can reveal novel cyclotide masses, modified cyclotides, and masses that are likely known cyclotide sequences. We extend our analysis to CRPs in the seeds of the well-characterized Cucurbitaceae species, Lagenaria siceraria (bottle gourd). Highly abundant MS/MS fingerprint ions are identified in the known linear trypsin inhibitors LLTI-II and III.³⁴ Mass shift analysis is then coupled with the identification of fingerprint ions in unknown mass species to identify and subsequently sequence three new L. siceraria CRPs, LSCRP-I, II, and III.

Thus, we demonstrate that fingerprint ions in the low m/z range can indicate whether a mass is (a) likely a cyclotide, (b) belongs to a cyclotide subfamily, (c) a potentially new cyclotide motif type that warrants full sequence characterization, (d) an oxidized or over-alkylated cyclotide, aiding in dereplication – the identification of known compounds to prevent their re-characterization, and thus prioritize novel mass species, (e) can inform what types of cyclotides are changing in abundance among sample types (e.g. tissue samples, seasonal variation), and (f) can differentiate multiple unique peptide species close or identical in mass that may have been overlooked otherwise based on mass alone.

Experimental

Fragment Database Generation.

Fragmentation databases were generated by reading .fasta sequences of cyclotides (CyBase, the database of cyclic proteins [http://www.cybase.org.au], 433 sequences, accessed January 5, 2020, removing sequences containing no Glu residues and cleaving at the first N-terminal Glu after the loop six ‘NG’ ligation site in sequences containing >1 Glu) into an in-house developed Python program that simulates Glu-C digestion and b/y-ion fragmentation. A Swissprot fragmentation database was generated by reading .fasta sequences from Swissprot (558,712 sequences, accessed January 5, 2020) into an in-house developed C++ program (accessible at github.com/hickslab, Repositories: FragmentSearch) that simulates Glu-C digestion, b/y-ion fragmentation, and compares the resulting set of in silico predicted fragments with a user-specified experimentally-derived list of fingerprint fragment masses within a mass tolerance, defaulting to 0.02 Daltons. For false discovery rate (FDR) determination, all cyclotide species (208 sequences) were removed from the Swissprot database prior to in silico manipulation. Monoisotopic fragment masses were calculated, where Cys has a fixed modification of +57.0214 Da. Digestion with Glu-C was simulated by splitting each sequence at the C-terminus of each glutamic acid residue and the resulting theoretical b/y-ions of the linear sequence were predicted.

Plant Material and Extraction.

Viola communis Pollard aerial material was harvested from North Carolina residential areas (35°55′25.6″N, 78°56′14.5″W) in the months of March and December 2019. Voucher specimens of V. communis are available at the University of North Carolina at Chapel Hill Herbarium (NCU), NCU accession numbers 672919 and 672920 (Parsley, s.n.), respectively. Lagenaria siceraria seeds were purchased from Strictly Medicinal (Eugene, OR). Viola aerial material was ground under liquid nitrogen to a fine powder, and L. siceraria seed material was ground with a seed mill. Peptides were extracted with 10% acetic acid buffer (1:3 w/v aerial tissue, 1:10 w/v seed material). Peptidyl constituents were selected for by filtering extracts through a 30 kDa MWCO filter (Millipore) and dialyzing with a 0.1–1 kDa cutoff membrane (Spectrum Labs) into 5 mM ammonium formate pH 2.7. Remaining small molecule components were removed by strong cation exchange (PolySulfoethyl A column, 100 mm × 4.6 mm, 3 μm particles, PolyLC) using a linear salt gradient (mobile phase A: 5 mM ammonium formate, 20% acetonitrile, pH 2.7; mobile phase B: 500 mM ammonium formate, 20% acetonitrile, pH 3.0). Peptide-containing fractions were pooled, salts were removed via solid-phase extraction (SPE) cartridges (Waters Sep-Pak, 500 mg), and elutions were dried in a vacuum centrifuge (Labconco).

Mass Shift Analysis and Proteolysis.

Material was reduced (10 mM dithiothreitol, 45 °C, 850 rpm, 30 min) and alkylated (100 mM iodoacetamide, 25 °C, 850 rpm, 15 min). Reduced and alkylated material was incubated 1:200 enzyme:substrate with endoproteinase Glu-C enzyme (Sigma) in 100 mM ammonium bicarbonate, pH 7.8, at 37 °C for 3 h. Pierce SPE C₁₈ spin columns were used to remove salts (Thermo Scientific) prior to LC-MS analysis.

LC-MS/MS Analysis.

One μg of acidified reduced/alkylated or reduced/alkylated/Glu-C digested sample was injected onto a nano-LC-ESI-MS/MS platform comprised of a NanoAcquity (Waters, Milford, MA) coupled to a TripleTOF5600 MS (AB Sciex, Framingham, MA). A Symmetry C₁₈ trap column (100 Å, 5 μm, 180 μm × 20 mm, Waters) and HSS T3C₁₈ analytical column (100 Å, 1.8 μm, 75 μm × 250 mm, Waters) were used for nUPLC separation of peptides, with a flow rate of 0.3 μL/min and a 30 minute linear ramp of 5%–50% B (mobile phase A, 1% formic acid in water; mobile phase B, 1% formic acid in acetonitrile). The TripleTOF5600 MS was operated in positive-ion, high-sensitivity mode with the MS survey spectrum using a mass range of 350–1600 Da in 250 ms and IDA where the first 20 MS² spectra were collected with a mass range of 100–1800 Da each in 87 ms. Targeted CID MS/MS data was acquired for the three L. siceraria CRP masses, 2605.93 Da (869.65 m/z, +3), 2705.02 Da (902.68 m/z, +3), 2762.01 Da (921.68 m/z, +3), using the reduced/alkylated samples and a collision energy (CE) of 40 and a CE spread of 5. All V. communis and L. siceraria mass spectrometry data has been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository³⁵ with the data set identifier PXD019501and DOI 10.6019/PXD019501. Mass shift analysis was performed by identifying 348.16 Da mass shifts with an in-house developed Python script using “Peptide ion data” deisotoped peak lists generated in Progenesis QI for Proteomics (Nonlinear Dynamics, v.2.0) with a retention time filter of 14–45 min and a maximum charge of +10.

Results

Theoretical Fingerprint Ions

Cyclotide sequences (CyBase.au.org) were Glu-C cleaved and fragmented in silico to produce theoretical b/y MS/MS ions. Fingerprint ions of cyclotides and cyclotide subfamilies were assessed for abundance in the theoretical database, initially focusing on the more common bracelet and Möbius species (Table 1). Möbius cyclotides are defined primarily by an aromatic-Pro motif in loop five; total Möbius cyclotides, cis-Trp/Pro and cis-Tyr/Pro Möbius sequences were separately examined. Necessitated by design, all cyclotides contain a 147.08 Da fragment ion corresponding to a C-terminal Glu residue. Notably, ~87% of all known cyclotide species contain the C-terminal residues ‘CGE,’ providing a 148.06, 205.08, and 365.11 m/z fingerprint handle for masses likely belonging to the cyclotide family. The Swissprot database³⁶ was then used to estimate the FDR for these cyclotide fingerprint ions based on the frequencies of identifying sequences/fragment ions characteristic of cyclotides in non-cyclotide masses. The likelihood of finding the individual cyclotide-indicative fragment ions 148.06, 205.08, 365.11 m/z in non-cyclotide masses was calculated to be 97%, 0.46%, and 0.02%, respectively, where the likelihood of identifying the 148.06 and 205.08 m/z fragment ions in the same spectrum would be 0.45% and of identifying all three fragment ions would be only 0.005% (Figure 1). This indicates that the ‘CGE’ motif is unlikely to be found in non-cyclotide mass species; therefore these fragment ions in combination can be used as a basis to assign novel putative cyclotides from LC-MS/MS data.

Table 1.

Top 15 most abundant theoretical b/y MS/MS fingerprint ions in, from left to right, all cyclotides, the bracelet subfamily, the Möbius subfamily, the cis-Trp/Pro Möbius, and the cis-Tyr/Pro Möbius. Masses are shown in the +1 charge state (m/z). The column representing the percent of cyclotides containing a given ion is color-coded, from high (orange) to low (yellow) percent.

All Cyclotides (423)					Bracelets (294)					All Möbius (129)
m/z	No.	Seq.	Ion Type	%	m/z	No.	Seq.	Ion Type	%	m/z	No.	Seq.	Ion Type	%
148.06	423	E	y	100	148.06	294	E	y	100	148.06	129	E	y	100
205.08	373	GE	y	88	205.08	246	GE	y	84	205.08	127	GE	y	98
365.11	368	CGE	y	87	365.11	243	CGE	y	83	365.11	125	CGE	y	97
88.04	242	S	b	57	88.04	226	S	b	77	102.05	110	T	b	85
248.07	242	SC	b	57	248.07	226	SC	b	77	262.09	110	TC	b	85
347.14	207	SCV	b	49	347.14	205	SCV	b	70	409.15	54	TCF	b	42
575.25	194	IPCGE*	y	46	462.17	180	PCGE	y	61	561.23	48	PVCGE	y	37
462.17	182	PCGE	y	43	575.25	153	IPCGE	y	52	464.18	48	VCGE	y	37
102.05	163	T	b	39	494.21	78	SCVF	b	27	674.32	47	LPVCGE	y	36
262.09	163	TC	b	39	607.29	78	SCVFI	b	27	731.34	45	GLPVCGE	y	35
533.22	87	SCVW	b	21	632.27	77	GIPCGE	y	26	575.25	41	PICGE*	y	32
632.27	80	GIPCGE	y	19	704.34	76	SCVFIP	b	26	478.20	41	ICGE	y	32
494.21	78	SCVF	b	18	864.37	76	SCVFIPC	b	26	523.20	36	TCFGG	b	28
607.29	78	SCVFI	b	18	533.22	71	SCVW	b	24	845.38	35	NGLPVCGE	y	27
704.34	76	SCVFIP	b	18	646.30	70	SCVWI	b	24	1001.48	34	RNGLPVCGE	y	26
WP Möbius (93)					YP Möbius (36)
m/z	No.	Seq.	Ion Type	%	m/z	No.	Seq.	Ion Type	%
148.06	93	E	y	100	148.06	36	E	y	100
205.08	91	GE	y	98	205.08	36	GE	y	100
365.11	90	CGE	y	97	365.11	35	CGE	y	97
102.05	82	T	b	88	102.05	28	T	b	78
262.09	82	TC	b	88	262.09	28	TC	b	78
464.18	44	VCGE	y	47	409.15	23	TCF	b	64
674.32	43	LPVCGE*	y	46	575.25	14	PICGE	y	39
561.23	42	PVCGE	y	45	478.20	14	ICGE	y	39
731.34	41	GLPTCGE	y	44	523.20	11	TCFGG*	b	31
1001.48	31	RNGLPVCGE	y	33	537.25	11	TCFK	b	31
845.38	31	NGLPVCGE	y	33	688.33	9	LPICGE	y	25
409.15	31	TCF	b	33	466.18	7	TCFG	b	19
575.25	27	PICGE	y	29	510.20	7	TCFT	b	19
478.20	27	ICGE	y	29	567.22	7	TCFTG	b	19
361.15	26	TCV	b	28	88.04	7	S	b	19

^*Includes permutation of some residues within the sequence.

More specific molecular information can be attained from other highly abundant MS/MS fingerprint ions. In addition to the presence/absence of the cis-Trp/Pro or cis-Tyr/Pro motif in loop five, bracelet and Möbius cyclotides offer regions of sequence homology within subfamilies. As such, the b-ions 248.07, 347.14, and y-ion 462.17 m/z are seen primarily in bracelets, where b-ions 102.05, 262.09, 409.15 m/z and y-ions 464.18 and 561.23 m/z are seen largely in Möbius cyclotides (Figure 1, Table 1, summarized in Table 2). Differences between the cis-Trp/Pro and cis-Tyr/Pro Möbius sequences are seen primarily in the cis-Trp/Pro y-ion series: 464.18, 674.32, 561.23, and 731.34 m/z (Table 1). The percent of bracelet and Möbius cyclotides in CyBase that contain a given MS/MS fingerprint ion can be calculated (Table 2). Additionally, the percent in which a given MS/MS fingerprint ion belongs to a bracelet or Möbius sequence can be calculated (Table 2), and indicates the specificity of a given sequence to a subfamily (e.g. if a sequence motif is identified, does it suggest a bracelet or Möbius identity, aiding in sequence characterization). Bracelet sequences outnumber Möbius sequences ~2:1, consistent with calculated Bracelet:Möbius ID values of ~67:33. Though this does not preclude the identification of these fragment ions in the alternative subfamily, these ions can be used to predict cyclotide subfamily type in addition to contributing additional evidence that a mass belongs to the cyclotide family. A conserved aromatic residue in bracelet subfamily loop two (N-terminal sequence ‘SCVX(Aromatic)’) can provide additional sequence information, where instances of Phe, Tyr, and Trp incorporation into this b4 position are roughly equal (Table 2) and can aid in the discrimination of sequence identities. In addition to these theoretical cyclotide, bracelet, and Möbius MS/MS fingerprint ions, experimentally observed fragment ions (Table 2) can be used to quickly differentiate cyclotides close in mass, identify common modifications (e.g. oxidation, alkylation), and identify likely new/known sequences in tandem with database searching towards a more accurate representation of sequence diversity in a plant extract, and are discussed in detail through examples below.

Table 2.

Diagnostic b/y MS/MS fingerprint ions indicative of (A) all cyclotides, (B) bracelets, (C) Möbius, and (D) notable and often highly abundant bracelet ions that provide valuable information towards the identification of a mass species.

(A)	All cyclotides						(B)	Bracelets
m/z	b/y	Seq	% Brac	% Möb	Brac ID	Möb ID	m/z	b/y	Seq	% Brac	% Möb	Brac ID	Möb ID
148.06	y	E	100	100	70	30	248.07	b	SC	77	12	93	7
205.08	y	EG	84	98	66	34	347.14	b	SCV	70	2	99	1
365.11	y	EGC	83	97	66	34	462.17	y	EGCP	61	2	99	1
(C)	Mobius						(D)	Informative Bracelet Peaks
m/z	b/y	Seq	% Brac	% Möb	Brac ID	Möb ID	m/z	b/y	Seq	% Brac	% Möb	Brac ID	Möb ID
102.05	b	T	18	85	33	67	494.20	b	SCVF	27	0	100	0
262.09	b	TC	18	85	33	67	510.20	b	SCVY†	20	9	85	15
409.15	b	TCF	4	42	18	82	533.21	b	SCVW	24	0	100	0
464.18	y	EGCV	1	37	6	94	565.21	b	SCVW(diOX)*	-	-	-	-
561.23	y	EGCVP	5	37	25	75	590.23	b	SCVW(Alk)*	-	-	-	-

^*m/z species generated from common, often unintentional, modifications (di-oxidation, alkylation) of the aromatic bracelet peak; values for percent bracelet, percent Möbius, bracelet identification, and Möbius identification are not applicable.

^†All %Bracelet and Bracelet ID values derived from the shown sequence; however, %Möbius and Möbius ID values were calculated from a same mass (510.20 m/z) but different sequence: the bracelet and Möbius values here are not derived from the same amino acid sequence, but both produce the 510.20 m/z ion, necessitating that the 510.20 m/z value be assigned to a bracelet ‘SCVY’ motif only when observed in combination with 248.07 and 347.14 m/z b-ions.

Experimental Fingerprint Ions

Wild grown Viola communis material collected during two independent harvests was reduced, alkylated, and Glu-C digested for mass shift and fingerprint ion analysis. Total ion chromatograms (TICs) of the intact V. communis samples suggest that the extract components are of roughly the same composition; however, differences in the TICs support the analysis of both extracts to provide additional mass species for fingerprint analysis (Figure 2A).

Figure 2.

(A) Total ion chromatograms (TICs) of V. communis harvests. (B-H) MS² spectra of V. communis reduced, alkylated, and Glu-C digested material. Common fingerprint ions are labeled; the bracelet aromatic peak often differentiates masses and is bolded in green in B-G.

Approximately 200 V. communis masses produced a characteristic +348.16 Da mass shift upon reduction and alkylation and were within the cyclotide mass range of 2500–3700 Da. Sequence elucidation of all 200 mass species would take considerable time, additional targeted experiments to improve MS/MS fragmentation, and would likely result in the identification of known cyclotides from other botanical species as well as duplicate sequences present in the same plant species with modifications (e.g. oxidation) that obscure sequence identity. Thus, the use of cyclotide-indicative fingerprint ions identified herein can guide sequencing prioritization.

Novel cyclotide species.

Mass 3508.57 Da (878.15 m/z, +4, and the neutral, intact mass (NI), calculated by subtracting 348.16 Da for cysteine modification and 18.01 Da for Glu-C digestion, = 3142.40 Da) produced very few MS/MS fragment ions that could be used towards full sequence elucidation, however, cyclotide y-ion fragments 148.06, 205.08, and 365.11 m/z corresponding to the C-terminal sequence ‘CGE’ are observed (Figure 2B). Additionally, the common bracelet specific b-ions 248.06, 347.13, and y-ion 462.16 m/z are present, corresponding to the N-terminal and C-terminal sequences ‘SCV’ and ‘PCGE,’ respectively (Table 1). Additionally, a 510.20 m/z fragment is present and corresponds to a Phe in the aromatic b4 position, indicating an N-terminal sequence ‘SCVF.’ The neutral, intact mass 3142.40 Da is within 1 Da of four entries in CyBase: mden F, viphi F, mra5, and kalata B18, identified in the botanical species Melicytus dentatus, Viola philippica, Melicytus ramiflorus, and Oldenlandia affinis, respectively. Comparing these known sequences with the experimentally-derived MS/MS fingerprint ions, mden F and viphi F contain N-terminal ‘TCF’ and ‘SCVF’ sequences, respectively, while mra5 and kalata B18 share the N-terminal ‘SCVY’ motif with 3142.42 Da, but both contain a C-terminal ‘PCAE’ motif. Thus, based on the known sequences, all of these peptides can be ruled out strictly from the four N- and four C-terminal residues/predicted fingerprint ions, indicating that 3142.42 Da may be a novel mass and should be pursued further for sequence characterization.

Dereplication.

Natural product discovery is plagued by the re-identification of known compounds, and dereplication strategies are key to the discovery of novel mass species. Like the putatively novel mass above, mass 3518.61 Da (880.66 m/z, +4, and NI = 3152.45 Da), did not produce enough fragmentation for full sequence determination, however, the characteristic cyclotide fingerprint ions 205.08 and 365.11 m/z were observed, in addition to the bracelet b-ions 248.07, 347.13, and the y-ion 462.16 m/z, indicating that this mass is a cyclotide in the bracelet subfamily (Figure 2C). The MS/MS spectrum also contains the b4 aromatic peak 533.31 m/z, indicating that this mass contains the N-terminal sequence ‘SCVW’ (Table 2). The neutral, intact mass 3152.45 Da searched on CyBase is within 22 ppm of the known cytotoxic bracelet cyclotide, vitri A (3152.38 Da), identified in the botanical species Viola tricolor, Viola biflora, and Psychotria leptothyrsa. The observed MS/MS fingerprint ions are consistent with this cyclotide, with a ‘SCVW’ N-terminal motif and a ‘PCGE’ C-terminal motif. Though there could still be internal same mass amino acid substitutions making 3152.45 Da a unique sequence, based on the presence of vitri A among numerous botanical species, mass 3152.45 Da should be deprioritized for sequence characterization when targeting novel mass species.

Modifications.

Mass 3550.61 Da (888.66 m/z, +4, and NI = 3184.45 Da) has the characteristic cyclotide fingerprint y-ions 148.06, 205.08, and 365.11 m/z and the bracelet b-ions 248.07, 347.13 and y-ion 462.16 m/z; however, we see a shift of the aromatic b-ion peak 533.20 m/z to 565.20 m/z (Figure 2D), indicating N- and C-terminal sequences ‘SCVW(+32)’ and ‘PGCE,’ respectively. This mass shift is commonly observed experimentally, where bracelet-type cyclotides are modified with a doubly oxidized loop two Trp residue (+32 Da),³⁷ and may result from in vivo or experimental conditions (e.g. acidic environments). The true intact, neutral mass of this cyclotide is thus calculated by subtracting the 32 Da contributed by di-oxidation, and is consistent with a doubly oxidized vitri A. Without this observation, 3184.45 Da may be misidentified as a new cyclotide species or reassigned incorrectly: the neutral doubly-oxidized mass is identical to two CyBase cyclotides, Mobo B and Kalata B16. Mobo B can be eliminated from consideration immediately, as the N-terminal sequence motif is ‘TCAK’ and would not produce the 248.07 and 347.14 m/z ions indicative of the common bracelet N-terminal motif ‘SCV.’ Kalata B16 has the N-terminal 248.07 and 347.13 m/z ions, however, these are followed by a Tyr residue, representing the motif ‘SCVY,’ and is inconsistent with the 565 m/z ion, as well as the −32 Da, b–ion 533.20 m/z peak derived from the vitri A ‘SCVW’ motif. Additionally, kalata B16 has a C-terminal motif ‘PCAE,’ and would not contain the y–ions 205.08, 365.11, and 462.16 m/z.

Notably, the solvent-exposed loop two bracelet tryptophan residue is particularly susceptible to modifications during sample handling. It has been observed that common in vitro alkylation with iodoacetamide (+57.02 Da) can result in the over-alkylation of the loop two tryptophan residue, shifting this diagnostic fragment ion from 533.20 m/z to 590.22 m/z (data not shown). The true reduced and alkylated mass is thus calculated as 57.02 Da less, and there should not be a correlated mass in the intact sample MS¹ spectra (as this modification occurs upon reduction and alkylation). As with tryptophan oxidation, an assessment of the MS² spectrum may quickly identify this common modification and demonstrates the use of fingerprint ions in dereplication strategies essential to MS-based AMP analysis of natural product extracts.

Expanding chemical diversity.

Cyclotides with different sequences can have the same mass (e.g. cycloviolacins O2 and O9), and without MS² fingerprint ions the chemical diversity of cyclotide species in a given sample may be underestimated. In the V. communis December reduced, alkylated, and Glu-C digested spectra, common cyclotide fingerprint ions in the MS² spectra of the masses 3565.65 (892.42 m/z, +4, and NI = 3199.49 Da), 3565.61 (892.41 m/z, +4, and NI = 3199.45 Da), and 3564.57 Da (892.15 m/z, +4, and NI = 3198.41 Da) (Figures 2E, 2F, and 2G, respectively) are present. Upon closer examination, several characteristic fragment ions indicate that these masses are comprised of at least three different peptidyl species. All share the same base characteristic cyclotide y-ion 365.11 m/z and bracelet specific b-ions 248.07 and 347.14 m/z and y–ion 462.17 m/z, however, the b-ion loop two aromatic residue is clearly shifted to 494.21, 510.20, and 565.20 m/z corresponding to a Phe, Tyr, and doubly-oxidized Trp in 3199.49, 3199.45, and 3198.41 Da, respectively. Standard cyclotide-identifying fingerprint ions coupled with the shifting of the tell-tale aromatic peak has allowed the quick and facile discrimination of three unique sequences close in mass, which may have been overlooked as a single mass otherwise. Masses 3199.49 and 3199.45 Da are within 1 Da of three masses on CyBase, and mass 3166.41 Da (3198.41 Da with 32 Da subtracted to account for di-oxidation) is within 1 Da of four masses on CyBase. Using the information gathered here, these masses can be subsequently analyzed for possible novelty/redundancy with known sequences as described above.

Linear species.

Masses exhibiting a 348.16 Da mass shift but failing to appear at the predicted reduced, alkylated, and Glu-C digested m/z are either not visible after the multiple sample preparation steps from low abundance, have >1 or 0 Glu residues, the off-target Glu-C cleavage at Asp residues was significant (Glu-C cleaves ~100× slower at Asp residues), or the original mass was a putative linear cyclotide variant. The neutral intact mass 3004.26 Da is present in both the March and December harvests, however, despite the initial high abundance in both samples, this mass is not present at the anticipated reduced/alkylated/Glu-C digested m/z. The reduced and alkylated V. communis samples, however, contain the expected +348.16 Da mass in the MS¹ (3352.41 Da) and unlike cyclic cyclotide variants, exhibited fragmentation in the MS². Upon initial observation, no cyclotide fingerprint ions were present and indicates this is either a non-cyclotide mass – and would not have any cyclotide fingerprint ions – or a linear cyclotide variant, and thus would not be cleaved at the loop one Glu residue (in vivo cyclotide ligation occurs in loop six, in which would be the natural termini of a linear variant). The intact neutral mass 3004.26 Da is within 10 ppm of the cyclotide violacin A, a linear cyclotide with a Möbius cis-Trp/Pro motif in loop five, found in the botanical species Viola odorata and Psychotria leptothyrsa. In silico prediction of violacin A aligns with several prominent fragment ions in the reduced and alkylated MS² spectrum (data not shown). Returning to the reduced, alkylated, and Glu-C digested spectrum, the predicted digested violacin A mass is present in the MS¹ spectrum, where now the linear mass is cleaved into two discrete peptides via an internal Glu residue. The MS² from the larger sequence (2648.15 Da) resulting from the Glu-C digest of the linear violacin A cyclotide contains the expected Möbius fragment b-ions 262.09 and 409.15 m/z (Figure 2H).

Trypsin inhibitor fingerprint analysis

Mass spectral fingerprinting is not limited to bracelet and Möbius subfamilies, and can be translated to the facile identification and prioritization of trypsin inhibitor CRPs. Three highly homologous Lagenaria siceraria seed trypsin inhibitors have been characterized previously, LLTI-I, II, and III, and conform to the three disulfide connectivity of the cysteine-knot motif, albeit with a linear backbone.³⁴ Mass shift analysis of L. siceraria material revealed ~20 masses demonstrating a +348.16 Da mass shift upon derivatization. Of these, masses 3666.53 Da (917.64 m/z, +4, NI = 3318.37 Da) and 3795.61 Da (949.91 m/z, +4, NI = 3447.43 Da) were consistent with the L. siceraria trypsin inhibitors LLTI-II and LLTI-III, respectively (Figure 3). MS² spectra of 3666.53 and 3795.61 Da reveal distinct and shared y-ion fingerprint ions (y2: 236.07, y5: 543.22, y8: 945.39, y9: 1044.46, y10: 1204.49, y11: 1319.51, y12: 1390.56, and y13: 1503.64 m/z) consistent with known LLTI sequences (Figures 3, 4A, 4B). Exploration of the reduced and alkylated L. siceraria MS data resulted in the identification of an additional, unknown mass 3852.61 Da (964.16 m/z, +4, NI = 3504.45 Da), differing from the mass of LLTI-III by 57.02 Da and demonstrating the characteristic LLTI-II and III fingerprint ions. A third trypsin inhibitor, LLTI-I, has been characterized in L. siceraria and differs in sequence from LLTI-III only by a modified N-terminal Glu (to a pyroglutamic acid residue, −18.01 Da) and the addition of a C-terminal Gly residue (where the mass of Gly is 57.02 Da). Relying on this mass difference, it may be assumed that 3504.45 Da is a LLTI-I sequence variant with an unmodified N-terminal Glu residue, differing only by a single Gly, and may be prioritized for sequence characterization to investigate this possibility. However, closer examination of the MS² spectrum (Figure 4C) reveals the same fingerprint y–ion series as in LLTI-II and III, and is incompatible with the LLTI-I sequence derived from an extra C-terminal Gly (which would offset all y-ions by the mass of Gly). With this information, it is likely that 3852.61 Da is an over-alkylated LLTI-III sequence. Though the site of alkylation cannot be immediately localized with the observed MS/MS fingerprints, as with the bracelet aromatic peak, this demonstrates the use of fingerprint ions in other CRPs to prioritize or deprioritize mass species of interest.

Figure 3.

Sequences and neutral, disulfide-intact (NI) molecular weights (Da) of Lagenaria siceraria known trypsin inhibitors LLTI-I, II, and III (top) and previously uncharacterized Lagenaria siceraria cysteine-rich peptides LSCRP-I, II, and III (bottom). Bolded, blue Glu residue (E) at the LLTI-I N-terminus is modified to a pyroglutamic acid. Regions where sequences diverge are highlighted in blue.

Figure 4.

MS² spectra of L. siceraria known trypsin inhibitors (A-C) and novel cysteine-rich sequences (D-F); fingerprint ions common to each type of CRP are labeled.

New sequence types of novel CRPs may be identified through mass shift analysis complemented with MS² fingerprint ion identification, where masses that belong to the same class of CRPs can be grouped based on similar, highly abundant MS² ions prior to sequence characterization. The three L. siceraria masses 2605.93 Da (869.65 m/z, +3, NI = 2257.76 Da), 2705.02 Da (902.68 m/z, +3, NI = 2356.86 Da), 2762.01 Da (921.68 m/z, +3, NI = 2413.85 Da) were identified with mass shift analysis to likely contain three disulfides and are not consistent with any characterized L. siceraria CRPs. MS² spectra of the reduced and alkylated mass species contained similar, highly abundant fragment ions (Figure 4D–F), indicating that these masses belong to the same sequence family of CRPs. Complete sequence elucidation via tandem MS revealed three novel, linear sequences, confirmed to contain six Cys residues, differing each by only one amino acid residue, and are named LSCRP-I, II, and III (Figure 3). The initially observed abundant fragment ions 300.17, 460.20, 717.28, 903.36, 1066.41, 1226.45, 1341.48, and 1501.52 m/z present among the LSCRP spectra correlate to a conserved b-ion series and can be used as a handle to identify additional LSCRP species, either in L. siceraria or perhaps other Cucurbitaceae species. The biological function and activities, if any, of these peptides is unknown; however, the cysteine-rich nature of these small, homologous peptides dictates that follow up studies concerning biological activity should be pursued. Here, a standard mass shift analysis experiment coupled with on the fly MS² assessment has aided the facile identification of a small set of novel CRPs in L. siceraria.

Conclusion

Stable and commonly bioactive CRPs are typically discovered via mass shift analysis. However, the accurate assessment of unique CRP species in a given botanical species is often challenged by same mass species, post-translational modifications or modifications derived from sample handling, and incomplete MS² fragmentation. Mass spectral fingerprint ions can be leveraged to gain additional information about a mass species prior to full sequence characterization and with only poor quality MS² spectra. Orthogonal methods, e.g. transcriptomics, can be utilized to aid the accurate sequence characterization of targeted masses demonstrating MS/MS fingerprint ions. Herein we identify sets of mass spectral fingerprint ions characteristic of the CRP cyclotide family, which may indicate a mass belongs to a specific cyclotide subfamily, and “tell-tale” ions that are of importance when discriminating putative cyclotide species, including common oxidation and over-alkylation ions observed experimentally. Cyclotide-containing V. communis material is used as proof-of-principle, where experimental cyclotide fingerprint ions are explored. Fingerprint ions derived from a third type of CRP, the trypsin inhibitors, are assessed in the gourd L. siceraria. Combining mass shift analysis with the identification of prominent MS² fingerprint ions is then used to identify three novel CRPs. We demonstrate that abundant mass spectral fingerprint ions can be used to quickly discern masses of interest in complex matrices and masses that are already characterized, aiding prioritization of the most promising novel mass species in a natural product sample for characterization.