Cyclotide host-defense tailored for species and environments in violets from the Canary Islands

Cyclotides are cyclic peptides produced by plants. Due to their insecticidal properties, they are thought to be involved in host defense. Violets produce complex mixtures of cyclotides, that are characteristic for each species and variable in different environments. Herein, we utilized mass spectrometry (LC–MS, MALDI-MS), transcriptomics and biological assays to investigate the diversity, differences in cyclotide expression based on species and different environment, and antimicrobial activity of cyclotides found in violets from the Canary Islands. A wide range of different habitats can be found on these islands, from subtropical forests to dry volcano peaks at high altitudes. The islands are inhabited by the endemic Viola palmensis, V. cheiranthifolia, V. anagae and the common V. odorata. The number of cyclotides produced by a given species varied in plants from different environments. The highest diversity was noted in V. anagae which resides in subtropical forest and the lowest in V. cheiranthifolia from the Teide volcano. Transcriptome sequencing and LC–MS were used to identify 23 cyclotide sequences from V. anagae. Cyclotide extracts exhibited antifungal activities with the lowest minimal inhibitory concentrations noted for V. anagae (15.62 μg/ml against Fusarium culmorum). The analysis of the relative abundance of 30 selected cyclotides revealed patterns characteristic to both species and populations, which can be the result of genetic variability or environmental conditions in different habitats. The current study exemplifies how plants tailor their host defense peptides for various habitats, and the usefulness of cyclotides as markers for chemosystematics.


Results
Cyclotide diversity in violets from the Canary Islands. The diversity of cyclotides produced by four violets (V. odorata, V. anagae, V. palmensis and V. cheiranthifolia) was explored using MALDI-MS. The highest number of different peptides were found in V. anagae and V. odorata-41 and 37, respectively. Of these, more than 60% were unique to the single species and not found in the other violets tested, judged by their monoisotopic molecular mass (Table 1). A substantially lower number of cyclotides could be identified in V. palmensis (25 individual cyclotides) and V. cheiranthifolia (26) with eight and six masses observed exclusively in those species, respectively. V. palmensis and V. cheiranthifolia had fifteen peptides in common and only 11 masses were found in both V. anagae and V. odorata (Table 1). All monoisotopic masses of cyclotides found in each species are shown in Supplement 1.
Cyclotide sequences in V. anagae-de novo MS/MS sequencing and transcriptome analysis. A cyclotide with monoisotopic molecular mass 3010.24 was isolated from V. anagae, enabling MS/MS Table 1. Total number and number of unique monoisotopic molecular masses (individual cyclotides) found in the mass spectra of extracts from different violets.

Species
Number of cyclotides Unique to species V. anagae 41 25 V. odorata 37 27 V. palmensis 25 8 V. cheiranthifolia 26 6 Scientific Reports | (2021) 11 1126.57, 15 y-ion and five b-ions could be assigned in the MS/MS fragmentation pattern (Fig. 1A). Tryptic cleavage resulted in two fragments with [M+2H] 2+ 929.9 and [M+2H] 2+ 769.87, the MS/MS fragmentation of those allowed to identify the rest of the peptide's amino acid sequence (Fig. 1B,C). The peptide was named vian 2 and its complete sequence is shown in Table 2. Database searches in the CyBase cyclotide database (http:// www. cybase. org. au) and protein BLAST in Uniprot database verified the novelty of the sequence. V. anagae transcriptome analysis yielded 22 cyclotide sequences of which 19 were new and three were known ( Table 2). The new peptides were named using an established nomenclature 40 . Additionally, 22 precursor peptides were characterized; the nucleotide sequences of the transcripts together with translated open reading frames containing cyclotides are shown in Supplement 2. Cyclotide extracts display antifungal activity. V. anagae and V. palmensis appeared to contain more cyclotides and fewer other compounds compared to V. cheiranthifolia. Higher peaks in the region of the chromatogram where the peptides were expected were observed in HPLC trace (Fig. 2). An extract of V. anagae showed the highest activity against three out of the five tested fungal species: M. fragariae, B. cinerea, F. culmorum ( Table 3). The cyclotide extract was most effective against F. culmorum with MIC of 15.62 µg ml −1 . The activity was weaker by one (MIC: 31.25 µg ml −1 ) and four dilution factors (MIC: 125 µg ml −1 ) respectively against B. cinerea and M. fragariae. None of the tested extracts were active against the Aspergillus strains (Table 3). V. palmensis extract showed some activity against F. culmorum (500 µg ml −1 ), whereas the extract of V. cheiranthifolia was not active against any strain at the tested concentrations.   Fig. 3. The profiles of V. cheiranthifolia and V. palmensis show similarities, which is also notable in the LC-UV traces in Fig. 2. In contrast, V. anagae and V. odorata exhibited more dissimilar cyclotide patterns (Fig. 3). The principal component analysis (PCA) was based on the relative abundance of 30 selected cyclotide masses. As shown in Fig. 4A, samples grouped by species in the first two components account for the majority of variation in the data (~ 70%). In addition, species belonging to different sections of genus Viola (Viola or Melanium) formed separate groups. V. odorata and V. anagae (section Viola) were clearly differentiated from the other two tested species along the first component (PC1), whereas V. anagae differed from the other samples along the PC2 axis. The species from section Melanium, V. cheiranthifolia and V. palmensis, showed small differences in their cyclotide production patterns and occupy the same space in the PCA (Fig. 4A). However, the separation of these species became apparent in PCA analysis including only different populations of V. cheiranthifolia and V. palmensis (Fig. 4B). One exception was noted from this trend-the Guajara population, isolated from other populations in PCA space, and clustered with V. palmensis (Fig. 4B).
The PCA for V. cheiranthifolia plants from different sites confirmed the separation of the Guajara population in terms of the cyclotide production patterns (Fig. 5A). PCA prepared for V. palmensis populations did not show any significant differences between them. Minor levels of variability, resulting in separation of specific populations in PCA, was observed in case of V. anagae and V. odorata (Fig. 5B-D). Additionally, PCA for V. odorata showed some degree of variation between populations from the three sampled islands and among the ones from La Palma (Fig. 5C,D). Statistical analyses of differences between populations showed significance (p values < 0.05) in 15 of the 30 analyzed cyclotides for V. anagae and V. cheiranthifolia. The differences between V. odorata populations were significant for 24 out of 30 analyzed cyclotides (Fig. 6). Additional PCA analyses are presented in Supplement 3. Table 2. Sequence alignment of the novel, not described previously (vian 1-20) and known (varv A, cyO2, viba 12) cyclotides found in V. anagae. Novelty or identity of the sequences has been established by comparing them to sequences deposited in Cybase (http:// www. cybase. org. au) and Uniprot (https:// www. unipr ot. org/). a Sequenced by MSMS on the peptide level. X in the sequence stands for either I or L. b Masses not found in mass spectra-undetermined N-terminal cleavage position from the precursor peptide. c The peptide tail sequence recognized by the cyclization enzyme is present in the precursor protein.

Discussion
The current work revealed a variety of novel cyclotides in violets endemic to the Canary Islands, V. anagae, V. cheiranthifolia and V. palmensis. None of these species have been previously investigated for their peptide constituents. These three endemic violets are rare and inhabit small defined areas, and both the number of populations and individual plants are limited 30,35 . All species were analyzed by MALDI-MS, antifungal assays, and HPLC fingerprinting, but only V. anagae was chosen for MSMS de novo peptide sequencing and transcriptome    www.nature.com/scientificreports/ sequencing due to better availability of plant material from a larger population accessible all year, independent of season. It was also the violet from which we expected the highest number of different cyclotides based on preliminary analyses of mass spectra. The peptide vian 2 (monoisotopic molecular mass of 3009.06 Da) was isolated and sequenced with MSMS. This peptide has a proline in loop five and therefore belongs to the Möbius subfamily 1 . When compared to known cyclotide sequences deposited in Cybase (http:// www. cybase. org. au/; Mulvenna et al. 41 ), the peptide loops one, three and four were similar in length and sequence to previously known cyclotide sequences, whereas differences occurred in loop two. Analysis of the transcriptome resulted in three known and 19 new predicted cyclotide sequences. The number of sequences found in the present study is similar to previous work where transcriptome sequencing was used in cyclotide exploration 15 . Some of the masses calculated for these peptides were not found in the mass spectra of extracts from V. anagae. Such differences have been reported before, i.e. peptides expressed on the transcript level are not detected in the peptidome or vice versa 15,42 . Some peptides may be expressed in low abundances, or not at all, and are hence not detected by MS. Some peptides, like vian 2, can be found in high abundance on the peptide level, but not at all in the transcriptome. The explanation of these discrepancies is still unknown, but it is clear that some cyclotides are stored for longer times in the plant 43 . In the current study, a part of the explanation may also be that the sample for transcriptome analysis was not collected in the same time of the year as the rest of the samples due to technical reasons-the season (different environmental and physiological conditions) may have influenced the cyclotide expression. Nevertheless, this shows that the complete repertoire of cyclotides produced by a particular species goes beyond what can be detected at one occasion or using a single www.nature.com/scientificreports/ method. One solution to get the complete picture of all possible cyclotides in a given species would be whole genome sequencing. However, the function of these genes would still have to be monitored on the peptide level.
In addition to differences in cyclotide expression profiles, as judged from MS and HPLC, the cyclotide extracts from V. anagae, V. cheiranthifolia and V. palmensis exhibited different levels of antifungal activities. The lowest MICs against all fungal species tested was shown for the extract of V. anagae, which also contained the highest abundance of cyclotides in relation to other compounds. This may be associated with its cyclotide composition being similar to V. odorata, especially the presence of some masses corresponding to cycloviolacins that have been shown to have antifungal activity 20 . The higher antifungal potency of the V. anagae cyclotide extract may be interpreted as an indication of the natural target(s) of the suite of cyclotides in this particular species, as a result of its environment. It has been suggested that plant species facing a larger variety of pathogens and herbivores display a higher diversity of peptides 14 . Our results may support this hypothesis: smaller numbers of different peptides were observed in V. cheiranthifolia and V. palmensis which grow in very dry, cold conditions at high altitudes 30,35 . On the other hand, V. odorata and especially V. anagae, which reside in the Canary Islands subtropical forests 38 display a larger variety of peptides. It would be ideal to test the antifungal activities of cyclotide extracts prepared from different populations of a species as well. Therefore, it would be possible to show, for a given species, if the environment determines not only the cyclotide composition but also the antifungal effect of the plant extract. However, most of the investigated species are either protected or the populations were small, so that collection of larger quantities of plant material, required to produce enough extract for the assays, was not possible. Therefore, in order to perform the assays, the extracts from plants belonging to different populations of a given species had to be pooled.
In the current study, we used ultrahigh resolution FTICR MALDI-MS imaging for cyclotide abundance. MALDI-MS had been successfully used to measure peptide concentration before, the method developed by Colgrave et al. 44 , allowed precise absolute quantification of a single cyclotide in serum. We measured the relative abundance of peptides instead of performing absolute quantification. Spectra of the whole spots of extracts were collected as an image with 200 µm spatial resolution, from which individual cyclotides were analyzed. This approach provides automation of the analysis and more data points per extract spot, which in turn results in www.nature.com/scientificreports/ more robust average intensities. The method is highly efficient, enabling the analysis of many samples in a short period of time (i.e. hundreds of samples in a matter of few hours) and across multiple peptides. The current study exemplifies the applicability of cyclotides as markers in chemosystematics because different species could be distinguished on the basis of their peptide production patterns in the PCA. This analysis was built on multiple specimens from each population, each collected and analyzed individually to take into account the variability between individual plants. The analysis was qualitative and semiquantitative-different cyclotides were distinguished by their molecular mass but also by their relative quantities (signal intensities) for each peptide. The applied method allowed to select and investigate 30 different cyclotides in the mass spectra of all species and populations. The peptide pattern diversification was notable in case of the V. cheiranthifolia Guajara population, which was clearly distinct from the others. This result supports the genetic differentiation of this population shown by Rodríguez-Rodríguez et al. 30 . Recently, on the basis of morphological features, this population was proposed to be separated into a new species-V. guaxarensis 29 . Interestingly, all sampled individuals of the Guajara population grouped with V. palmensis in the PCA, not with other V. cheiranthifolia populations. The two species, V. palmensis and V. cheiranthifolia, are closely related to each other 30,35 . According to Rodríguez-Rodríguez et al. 30 , the Guajara population is the oldest and presumably other V. cheiranthifolia are successors. The results of the present study lend support to that hypothesis. On the contrary, V. palmensis showed homogeneity in terms of cyclotide patterns among studied populations. These findings may also corroborate a previous study, which showed low genetical variability between V. palmensis populations (Batista and Sosa 35 ).
The above demonstrates the importance of both peptide identity and abundance (i.e. quantity) for chemosystematic applications. In fact, differences in abundance of only two peptides have been shown to be enough to distinguish the species V. odorata and V. uliginosa. They produce very similar set of cyclotides 15,31 but differ in quantities of cycloviolacins cyO2 and cyO13. Both peptides are produced by both violets, but in V. uliginosa cyO2 is a minor peptide and cyO13 one of the major and vice versa in V. odorata 15,31 . The species could be distinguished on this basis regardless of the growth conditions. CyO13 was shown to be the main cycloviolacin in V. uliginosa even under treatment of different plant growth regulators in in vitro culture 15 . To expand this type of analysis as done in the present study (in particular in combination with the robustness and speed of the MALDI-MS assay) may prove to be a very useful tool for chemosystematic investigations.
Analyses of multiple samples from individual plants and populations also showed variations in cyclotide patterns. This raises the question: is this variability in cyclotide expression within species based on genetic variations or environmental factors? Both arguments can be made. To the best of our knowledge, no literature about the genetics, ecology or physiology of V. anagae is available. However, considering the differences in cyclotide production between populations seen in the current analysis, some genetical variation can be hypothesized. It seems less likely that the differences reflect the environment influences as the populations are located in close proximity. V. anagae appears to propagate asexually; during several visits to the sites at different times of the year (December-late May), chasmogamous flowers were observed but never seed capsules (BS pers. obs.). In this case, there would be no gene flow between populations, genetical variability could be preserved, despite proximity of the locations.
In case of V. odorata, the differences in cyclotide production patterns between populations, shown in the current study, may to some degree reflect the environmental influences. The samples were obtained from populations at distant locations residing from very different habitats. Little is known about the genetic variation between the V. odorata populations residing at the archipelago. However, some level of homogeneity can be hypothesized. In Marcussen 2006 it was shown, on the basis of selected molecular markers, that V. odorata from Tenerife was genetically similar to the continental populations 45 . V. odorata is a common species, not endemic to the Islands. It can be presumed that it arrived to the islands relatively recently, making differentiation and speciation unlikely. Therefore, it can be presumed that the genetical variability is low between populations residing at the islands. Specific cyclotides are produced and accumulated in plant organs and tissues to target specific types of pathogens or insect herbivores 20,23 . Therefore, the minor differences in suits of cyclotides produced by V. odorata shown in the current study may reflect the response of the plant to conditions/types of stress present in particular habitat. However, more data, e.g. from transplant experiments would be required to rule out how the genetical and environmental factors contribute to the observed inter-populations differences.

Conclusion
The present study showed the diversity of cyclotides produced by violets from the Canary Islands. Using this as an example, it is demonstrated how the production patterns may reflect inter-species and inter-population variations and that cyclotide analysis can be utilized in chemotaxonomy. The present study supports the hypothesis that specific suits of peptides evolved for the needs of different species and their environments. From each location at least five individuals (above ground parts) were sampled, desiccated on silica gel and stored at room temperature. One additional population of V. anagae was sampled in January 2020 for transcriptome sequencing. A whole plant was collected and living sample was shipped to Uppsala (Sweden). Fresh tissues (fragments of a leaf, petiole and root) were transferred to RNAprotect Tissue Reagent (Qiagen, Düsseldorf, Germany) immediately after arrival and stored at − 20 °C. The collection sites are outlined in Fig. 7, and

Preparation of extracts and MALDI-MS analysis.
Dried leaves from all specimens were placed in a separate 2 ml Eppendorf tubes containing a metal bead and powdered using a TissueLyser (Qiagen, Germantown, MD), 1 min at 25 Hz. Subsequently, 10 mg of each powdered sample was transferred to a separate 2 ml Eppendorf tube with a metal bead and extracted in a TissueLyser for 2.5 min at 25 Hz, with 1 ml 30% acetonitrile (ACN), 0.1% trifluoroacetic acid (TFA) in MiliQ water. Tubes were centrifuged and supernatant was collected.
To match the concentrations with linear response range of matrix-assisted laser desorption/ionization-mass spectrometry (MALDI-MS), all the samples were diluted sixfold. The diluted extracts (0.5 µl of each sample) were spotted on a metal plate and air dried. The plates were sprayed with a MALDI matrix solution (2,5-dihydroxybenzoic acid, 35 mg/ml in 50% ACN, 0.2% TFA) using an automatic matrix sprayer (TM-Sprayer; HTX Technologies, Chapel Hill, NC) at a flow rate of 70 µl/min with a spray nozzle velocity of 1100 mm/min, track spacing of 2 mm (criss-cross pattern), nitrogen pressure of 6 psi, and a spray nozzle temperature of 90 °C for 8 passes. All the spots were analyzed using a MALDI Fourier-transform ion cyclotron resonance (FTICR) (solariX 7T-2ω, Bruker Daltonics, Bremen, Germany) mass spectrometer equipped with a Smartbeam II 2 kHz laser operated in positive ionization mode. The spot areas were imaged at 250 µm lateral resolution. The relative quantitative analysis for monoisotopic cyclotide ions-the mean intensity per pixel per spot (approx. 70 pixels per spot)-was performed in msIQuant software 46 . Spectra were normalized against the root mean square (RMS) of all data points. Monoisotopic ions, not overlapping with others in the average spectra were selected. Additionally, serial dilution of extracts from different species were also spotted on a MALDI-MS target plate to assess the range of dilutions in which the relation of concentration to intensity is linear for a given ion. Ions showing a linear relation for any of the species and being in the linear range in the used extract dilutions were included in the analysis. Average mass spectra were generated in FlexImaging, version 4.0 (Bruker Daltonics). The cyclotides produced by a particular species were distinguished by their monoisotopic molecular mass in between 2.8 and 3.8 kDa.
Principal component analysis (PCA) was performed for the data collected with MALDI-MS. The intensities of selected monoisotopic + 1 ions corresponding to cyclotides were used for the PCA and displayed according to differences between species or locations. For each cyclotide, a Kruskal-Wallis rank sum test followed by a Nemenyi post-hoc test was performed to test for differences in cyclotide variance in the different species 47 . All statistical analyses were conducted in R. Significance testing was done using the package PMCMRplus and plots were produced with the package ggplot2 [48][49][50] . UPGMA clustering analysis of the cyclotides based on Euclidean distances was visualised with a heatmap prepared in R (https:// CRAN.R-proje ct. org/ packa ge= pheat map) 51 .
V. anagae transcriptome sequencing and cyclotide mining. RNA extraction and transcriptome sequencing were performed by Macrogen Inc. (Seoul, South Korea). The sample was shipped on ice, preserved in RNAprotect Tissue Reagent (Qiagen). Libraries were prepared using TruSeq stranded mRNA kit (Illumina, San Diego, USA), followed by Illumina Novaseq 2 × 100 bp paired-end sequencing. The transcriptome was de novo assembled using Trinity 52 . The assembled transcriptome was mined for similar sequences to the cyclotides

LC-MSMS sequencing.
A sample of 650 mg of V. anagae ground plant material was prepared by extracting twice with 10 ml of 30% ACN with 0.05% TFA and once with 10 ml of 60% ACN in 0.05% TFA. The extract was first purified on an ÄKTA basic HPLC system (Amersham Biosciences, Uppsala, Sweden) with a SNAP Ultra C 18 flash column, 20 µm, 100 Å, 12 g (Biotage) on a 20 min gradient from 10 to 70% ACN in 0.05% TFA with a flowrate of 10 ml/min. Fractions were analyzed by electrospray (ESI) MS in positive ionization mode (Finnigan LCQ ion trap, Thermo Electron Co., Waltham, MA, USA). The cyclotide containing fractions were lyophilized and subsequently re-dissolved in 10% ACN (0.05% TFA). Cyclotides were isolated using RP-HPLC using a Shimadzu LC20 system (Shimadzu Scientific Instruments, Kyoto, Japan) equipped with a Jupiter 5 µm C 18 300 Å, 250 × 10 mm column (Phenomenex). Reduction, alkylation and cleavage of the peptides to be sequenced was performed as described before 15,56 and the primary structure of the peptide fragments was analyzed using UPLC-QToF nanospray MS (Waters nanoAcquity, QToF Micro; 75 µm × 250 mm 1.7 µm BEH130 C18). MS scan window was set to 200-2500 m/z and for MS/MS to 50-2000 m/z. Antifungal assays. Cyclotide extracts of each Viola species were prepared by grinding 350-450 mg of dry plant material. Subsequently, the material was extracted twice with 10 ml of 30% ACN (0.05% TFA) in water and once with 10 ml of 60% ACN in 0.05% TFA. The extracts were combined, lyophilized, and resuspended in 2.5 ml Milli-Q water, and desalted using size exclusion chromatography (PD-10, GE Healthcare). The cyclotide extract composition was assessed by UV-RP-HPLC using a Shimadzu LC20 system described above and a 40 min gradient from 5 to 55% of ACN (0.05% TFA). The resulting extracts were then lyophilized and weighted. The enriched cyclotide extracts from V. anagae, V. palmensis and V. cheiranthifolia were tested against five fungal plant pathogens: Aspergillus fumigatus Fresenius, Aspergillus pseudonomius Varga, Mycosphaerella fragariae (Tulasne) Lindau, Botrytis cinerea Pers.; Fr. and Fusarium culmorum Sm. Sacc. B. cinerea, F. culmorum and M. fragariae were received from the Institute of Plant Protection-National Research Institute, Poznań, Poland (accession no. 2235, 2169, 1089 respectively), A. pseudonomius from Westerdijk institute, Netherlands (CBS 119388) and A. fumigatus strain J7, as previously described 57 . Stock suspensions of the fungal strains were prepared using established procedures 20,57 . All cultures were maintained at room temperature (RT) and under aseptic conditions. Standard microdilution assays in flat-bottomed 96-well Nunc A/S plates (Thermo Scientific) were performed 58,59 . The assays were performed in 24 g l −1 potato dextrose broth (PDB, Sigma-Aldrich St. Louis, MO, United States) for B. cinerea, F. culmorum and M. fragariae; and RPMI 1640 medium for both Aspergillus species. The cyclotide enriched plant extracts were dissolved in the respective media at starting concentrations between 0.4 and 1 mg l −1 . The minimal inhibitory concentration (MIC) was assessed by measuring the optical density (OD, absorbance at 600 nm) at the timepoint of inoculation (0 h) and again after 48 h (for M. fragariae, B. cinerea and F. culmorum) or 72 h (for A. fumigatus and A. pseudonomius), using a Varioskan Flash Multimode Reader (Thermo Scientific). The OD 48h-0h values lower than 0.1 indicated growth inhibition at this concentration. All antifungal assays were performed in three replicates.