Streptomyces sp. 11-1-2 produces a diverse array of specialized metabolites

To further explore the specialized metabolites that are responsible for the observed bioactivities, the organic culture extracts were subjected to untargeted LC-MS/MS, and the resulting data were analyzed using ion identity molecular networking in the GNPS environment along with SIRIUS⁴² and BUDDY⁴³ for molecular formula prediction and annotation of compounds. The molecular networking analysis revealed that the number of unique features was higher in the MeOH extracts than in the EtOAc extracts, and for both EtOAc and MeOH, the SA extracts had more unique features than the other media (Supplementary Fig. ^S3 online). The molecular networks for the EtOAc and MeOH culture extracts are shown in Fig. 4 and Supplementary Figure ^S4 online, respectively. The subnetworks for some predicted metabolites are numbered and highlighted.

Open in a separate window

Figure 4

Ion identity molecular networks of Streptomyces sp. 11-1-2 metabolites. The metabolites were extracted from M4, OBA, PMA, SA, SFM and YMS plate cultures using ethyl acetate and were analyzed using untargeted LC-MS/MS in both positive and negative ionization mode. Each node in the networks represents one fragmentation spectrum, and nodes are linked if the cosine score is>0.7 and there are at least 6 matched fragment ions. Networks containing selected specialized metabolites are highlighted and numbered based on the main metabolite: A, musacin D; B, nigericin; C, meridamycins; D, elaiophylin; E, guanidylfungin A; F, echoside C; G, galbonolide B; H, geldanamycin.

Echosides are para-therphenyl compounds produced by Streptomyces spp., and some possess DNA topoisomerase inhibition activity and weak antibacterial activity⁴⁴. In addition, echosides share structural similarities with terfestatin A, which has been shown to exhibit auxin signaling inhibitory activity⁴⁵. The presence of three members of the echoside family was predicted in the EtOAc extracts from the different plate cultures, whereas none were found in the corresponding MeOH extracts (Supplementary Figs. ^S5−^S7 online). Echoside C (1; m/z 453.1191, [M-H]⁻) was predicted in all the EtOAc media extracts, with the relative level being greatest in the PMA extract (Supplementary Fig. ^S5 online). However, the presence of this molecule could not be confirmed using targeted reverse-phase high performance liquid chromatography (RP-HPLC) (data not shown), suggesting that the amount produced in all the test media is very low. Echosides D and E (2 and 3; m/z 494.0914 [M-H]⁻ and m/z 524.1019 [M-H]⁻, respectively) were also predicted in the same network (Supplementary Figs. ^S5−^S7 online). Other features in the echoside network include two matches to the GNPS library for the plant metabolites baicalin (4) and chrysin-7-O-glucoronide (5), and two more features for which the molecular formula were predicted by SIRIUS and BUDDY, but the structures remain unresolved. The production of plant metabolites, specifically flavonoids, by Streptomyces strains has been previously reported^46–48, but the mechanism by which 4 and 5 are biosynthesized in 11-1-2 remains unclear.

Previous analysis of the 11-1-2 genome using antiSMASH version 6.0 revealed the presence of a region (#30) with 100% similarity to the reported BGC for echosides A-E²⁶. This same region is now designated region 32 using the updated version of antiSMASH (version 7.1.0) (Supplementary Table ^S1 online). The echoside BGC was described as containing a NRPS-like polyporic acid synthase (encoded by echA) that is essential for biosynthesis of these metabolites, along with a number of genes with putative regulatory and tailoring functions²⁷. BiG-SCAPE analysis showed that the echoside BGC and closely related BGCs can be subdivided into three gene cluster families (GCFs); moreover, the BGC network suggests that echosides and similar metabolites might be widespread among Streptomyces (Supplementary Fig. ^S5 online). Interestingly, a comparison between the BGC predicted at region 32 and the terfestatin/echoside BGC from Streptomyces sp. RM-5-8 showed important differences both upstream and downstream of the proposed core genes of the BGC (Supplementary Fig. ^S5 online). These differences might account for the production of different echoside analogues, as previously proposed⁴⁹.

Elaiophylin is a macrolide with diverse bioactivities, including antimicrobial activity against some Gram-positive bacteria, and anticancer activity⁵⁰. Elaiophylin is often co-produced with geldanamycin and nigericin, and several analogues like efomycins and halichoblelides^50–55 have been identified from different Streptomyces strains of diverse origins. The production of elaiophylins and efomycins is likely performed by the same BGC, and some discrete differences may account for the production of the different analogues, as previously suggested⁵². The antiSMASH 7.1.0 analysis of the 11-1-2 genome revealed that region 46 has 100% similarity to the efomycins K/L BGC (Supplementary Table ^S1 online), which is also responsible for the production of elaiophylins⁵².

The untargeted LC-MS/MS analysis revealed the presence of a feature in negative ion mode that is consistent with elaiophylin (6; m/z 1023.5900, [M-H]⁻ and 1069.5951, [M+FA]⁻) (Fig. 5 and Supplementary Figs. ^S8 and ^S9 online). Targeted detection by RP-HPLC (Supplementary Fig. ^S10 online) confirmed the presence of this compound in all the extracts except for the M4 and SFM extracts, and the levels were significantly higher in the PMA extracts than in the other extracts, with the PMA EtOAc extract having the greatest levels (Table ​(Table11 and Fig. 5). A number of related features were also detected, most notably, 2-methylelaiophylin (7; m/z 1037.6052, [M-H]⁻ and 1083.6106, [M+FA]⁻), and efomycin G (8; m/z 1009.5739, [M-H]⁻ and 1055.5799, [M+FA]⁻) (Supplementary Figs. ^S8 and ^S9 online), which possess antimicrobial activity against Gram-positive bacteria^50,56. Intriguingly, other compounds predicted in these networks do not match the known elaiophylin analogues, suggesting the possible presence of novel elaiophylin derivatives being produced by 11-1-2.

Open in a separate window

Figure 5

(a) Ion identity molecular network (negative ion mode) for elaiophylin metabolites in the Streptomyces sp. 11-1-2 culture extracts. Each node in the network represents one fragmentation spectrum, and the predicted molecular formula and structure of each is shown. Nodes are linked if the cosine score is>0.7 and there are at least 6 matched fragment ions, and the thickness of the line increases at higher cosine values. The size of the node is relative to the total abundance of each compound in the analyzed culture extracts. The pie charts represent the relative abundance of each compound in the six different culture extracts: Red=M4, Yellow=OBA, Green=PMA, Teal=SA, Blue=SFM, Violet=YMS. (b) Alignment of the elaiophylin BGC and the efomycin K/L BGC available in the MIBiG database and the predicted elaiophylin BGC at region 46 in Streptomyces sp. 11-1-2. Genes coloured the same belong to the same functional group, and homologues are linked by shaded areas that indicate the % amino acid identity of the corresponding protein products.

Table 1

Nigericin, geldanamycin and elaiophylin quantification in different organic culture extracts.

Extraction solvent	Culture medium	Nigericin (mM)	Geldanamycin (mM)	Elaiophylin (mM)
Ethyl acetate	M4	6.93±0.55	0.19±0.02	N.D.2
	OBA	0.82±0.28	0.22±0.02	0.02±0.01
	PMA	9.44±2.26	0.51±0.04	1.21±0.11
	SA	1.58±0.17	0.01±0.00	0.02±0.01
	SFM	4.53±0.37	0.15±0.02	N.D.2
	YMS	8.48±0.66	0.15±0.01	0.02±0.00
Methanol	M4	2.93±0.50	0.03±0.04	N.D.2
	OBA	0.98±0.43	0.15±0.09	0.03±0.01
	PMA	3.79±2.32	0.18±0.04	0.27±0.04
	SA	1.00±0.23	N.D.2	0.01±0.00
	SFM	1.42±0.33	0.05±0.02	N.D.2
	YMS	4.10±0.89	0.06±0.01	0.01±0.00

Open in a separate window

¹The values correspond to the average of three replicates±one standard deviation.

²Not detected.

A prominent feature found in the LC-MS/MS dataset was predicted to be musacin D (9, m/z 155.0706 [M+H]⁺), and the relative level of this feature was greatest in the PMA and YMS EtOAc (Fig. 6 and Supplementary Fig. ^S11 online). Musacin D is among several musacins first detected in Streptomyces griseoviridis, and all were initially reported as having no significant bioactivity^57,58. Later, it was found that the plant pathogen Nigrospora sacchari produces a phytotoxic compound with an identical structure to musacin D except for a different stereochemical configuration of the hydroxy group at the carbon five position⁵⁹. With the current information, it is not possible to conclude if the feature predicted as musacin D has a role in the phytotoxicity of the extracts, or if its identity truly corresponds to musacin D or the phytotoxin from N. sacchari. There is also currently no information regarding the genes necessary for the biosynthesis of musacins, and thus it was not possible to determine if 11-1-2 contains the genes necessary to produce these compounds. Purification and characterization of the feature will be necessary to establish its significance for the pathogenic phenotype of 11-1-2.

Open in a separate window

Figure 6

Ion identity molecular network (positive ion mode) for musacin D in the Streptomyces sp. 11-1-2 culture extracts. The features of the network are as described in the legend for Fig. 5.

Niphimycins are macrolides with structural resemblance to azalomycins and guanidylfungins, all of which have antifungal activity^60–63. The analysis of the different extracts identified a niphimycin-like compound with a match to the GNPS library in the majority of the extracts (10; m/z 1130.7292, [M+H]⁺ and 1128.7141, [M-H]⁻) (Supplementary Figs. ^S12−^S14 online). This compound is likely to be guanidylfungin A, a molecule with antimicrobial activity against fungi and Gram-positive bacteria and found to be co-produced with guanidylfungin B⁶⁴. Guanidylfungin B was also predicted to be present in some of the extracts (11; m/z 1116.7121, [M+H]⁺) (Supplementary Figs. ^S12−^S14 online). The presence of guanidylfungin A in the extracts may account for some of the observed antifungal activity against the S. cerevisiae indicator strain (Fig. 1). However, as this compound was detected in the OBA EtOAc extract that had no observable antifungal activity in our bioassays, further investigation of this compound and its production by 11-1-2 is required. To date, there is no BGC assigned specifically for the production of guanidylfungins in Streptomyces. Given the structural similarities between niphimycins and guanidylfungins, the cluster in region 50 (87% similarity to the niphimycins C-E cluster; Supplementary Table ^S1 online) of the 11-1-2 genome is most likely responsible for production of this compound (Supplementary Fig. ^S12 online).

A feature predicted to be galbonolide B (12) was also detected in the extracts (m/z 365.2315, [M+H]⁺) (Supplementary Figs. ^S15 and ^S16 online). Galbonolide B is a 14-membered macrolide with antifungal activity⁶⁵. Region 5 in the 11-1-2 genome is predicted by antiSMASH 7.1.0 to contain a BGC with 20% similarity to the known rustmicin BGC (Supplementary Table ^S1 online). Rustmicin is also known as galbonolide A, and the related compound neorustmicin is also known as galbonolide B; both molecules are 14-membered macrolides with antifungal activity⁶⁵. The rustmicin BGC from Streptomyces galbus contains 30 genes and is separated into two operons. Genes galABCDE are sufficient for the production of neorustmicin/galbonolide B⁶⁶, while genes galGHIJK are required for production of rustmicin/galbonolide A⁶⁵. The role of the remaining genes is unclear, but it is speculated that they are tailoring enzymes^65,67. The galABCDE core is present in region 5 of 11-1-2, suggesting that this strain can at least produce galbonolide B, while the flanking genes could serve as post-polyketide synthase tailoring enzymes yielding novel galbonolides (Supplementary Fig. ^S15 online).

Meridamycin is a 27-membered macrolide that was first isolated from a strain of Streptomyces hygroscopicus⁶⁸. Meridamycin binds to the human FKBP12 protein more efficiently than the closely related specialized metabolites FK506 or rapamycin, thus inhibiting the target of rapamycin (TOR) complex involved in immune responses⁶⁸. The FKBP12 protein is also present in plants, where it participates in different development pathways associated with the TOR complex^69,70. The use of rapamycin shows limited effects on seedling development^71,72, but the use of ATP-competitive TOR inhibitors reduces Arabidopsis root growth⁷². It is not clear how meridamycin would affect plant development, but its higher binding efficiency compared to rapamycin makes it a metabolite of interest in the context of the plant pathogen 11-1-2. Two features predicted to be meridamycin (13; m/z 820.5215, [M-H]⁻ and 822.5331, [M+H]⁺) and its analogue meridamycin A (14; m/z 838.5325, [M-H]⁻ and 862.5268, [M+Na]⁺) (Supplementary Figs. ^S17−^S19 online) were detected in the extracts, and region 11 of the 11-1-2 genome partially matches the reported meridamycin BGC (Supplementary Table ^S1 online). Other meridamycins analogues (meridamycin B, C, D) have been described in Streptomyces sp. LZ35 and are likely to be produced by the same BGC as meridamycin⁷³. The meridamycin BGC in Streptomyces sp. NRRL 30748 and Streptomyces sp. DSM 4137 (syn: Streptomyces malaysiensis DSM 4137) is classified as a hybrid non ribosomal peptide-polyketide synthase (NRP-PKS) cluster^74,75, and the 11-1-2 cluster shows some differences compared to these reported clusters (Supplementary Fig. ^S17 online). Firstly, the 11-1-2 region and the meridamycin BGC show two PKS genes, merC and merD, while the cluster sequenced in Streptomyces sp. DSM 4137 contains a single PKS. Secondly, a group of four genes encoding two putative transcriptional regulators (LysR- and TetR-family), one transporter (major facilitator superfamily, MFS) and a DNA polymerase III subunit alpha is located upstream of the PKS genes in 11-1-2, whereas they are situated downstream of the PKS genes in S. malaysiensis and are absent from the Streptomyces sp. NRRL 30748 cluster. Thirdly, three genes located downstream from the PKS genes in 11-1-2 are not present in the other clusters. Two of the genes encode a predicted two-component sensor histidine kinase and a LuxR transcriptional regulator, while the third gene encodes a protein of unknown function. It remains unclear how these genomic differences might affect the production of meridamycins in 11-1-2.

Previously, we reported that 11-1-2 produces geldanamycin and nigericin along with putative intermediates and other predicted related compounds²⁶. Molecular networks for these compounds were again identified in the current study, though the geldanamycin network had fewer features compared to our previous analysis²⁶. Notably, a feature from the nigericin network (15; m/z 784.5188, [M+NH₄]⁺) was predicted in the EtOAc and MeOH extracts from all media, but it was more prominent in both the PMA and SA organic culture extracts (Supplementary Fig. ^S20 online). This feature was not previously detected in the cultures of 11-1-2, and it is unclear if this compound exhibits similar biological activity as nigericin. Targeted analysis revealed that the nigericin production levels were significantly greater than that of geldanamycin in all the media tested (Table ​(Table1).1). Both compounds were detected in both the EtOAc and MeOH extracts, with the EtOAc extracts containing higher amounts of each. Of the media tested, PMA supported the highest amounts of both compounds, while SA contained the least amount of each (Table ​(Table11).

Organic extraction of Streptomyces sp. 11-1-2 cultures

Culture extracts were prepared from the 11-1-2 plate cultures as follows. The agar in each plate was cut in half, and each half was further cut into small pieces using a sterile pipettor tip. The pieces were transferred to two clean 50 mL plastic conical tube and the agar pieces were stored at −80 °C until ready for extraction. The agar pieces were first thawed at room temperature before addition of the organic solvent. ACS grade EtOAc or MeOH (20 mL) was added to each vessel, and the contents were mixed and left to incubate at room temperature overnight. Following incubation, the extracts were filtered using Whatman® #1 filter paper (GE Healthcare Life Sciences) and transferred into clean conical tubes. The remaining agar pieces were rinsed with 10 mL of fresh solvent, which was subsequently filtered and combined with the corresponding extract. The solvent was removed from each extract by rotary evaporation using an IKA® RV 10 instrument (IKA Works, USA). Methanol extracts were also subjected to freeze-drying for 3 days using a Labconco Freezone 12 Freeze Dryer (Labconco Corp., MO, USA) to remove the remaining aqueous liquid following rotary evaporation. The dried MeOH extracts were resuspended in 1 mL of 50% LC-grade MeOH, while the dried EtOAc extracts were redissolved in 1 mL of 70% v/v LC-grade MeOH. The extracts were then stored at −80 °C. Non-inoculated plates of each medium were used as controls in all cases. The control plates were prepared, incubated, and subjected to organic extraction alongside the inoculated media.

LC-MS/MS analysis

Untargeted metabolomics analysis of the culture extracts was conducted at the BioZone Mass Spectrometry Facility in the University of Toronto using a Thermo Scientific Ultimate 3000 UHPLC coupled to Thermo Scientific Q-Exactive equipped with a HESI-II probe. Metabolite separation was carried out using a Thermo Scientific Hypersil Gold C18 column (2.1×50 mm, 1.9 µm, 175 Å, equipped with guard column) maintained at 40 °C and a mobile phase gradient of water/acetonitrile with 0.1% v/v formic acid. Mass spectra were recorded using fast polarity switching. Then, the raw LC-MS/MS data files were converted into mzXML format and separated based on their polarity using MSConvert for further analysis. Both the raw and the converted files are available in the Mass Spectrometry Interactive Virtual Environment (MassIVE) data repository (massive.ucsd.edu) under the accession number MSV000091858.

Molecular networking

The raw data files were pre-processed using MSConvert⁸¹ with the recommended parameters (https://ccms-ucsd.github.io/GNPSDocumentation/fileconversion/). The files were converted into mzXML files with a 32-bit binary encoding precision and no zlib compression. The Peak Picking filter was selected with the vendor algorithm for MS-Levels 1-2.

The spectral data obtained were analyzed using Feature-Based Molecular Networking (FBMN)^20,21 complemented with Ion Identity Molecular Networking²². For this, the mzXML files were imported and analyzed using MZmine (version 2.37.corr17.7)⁸². The parameters used for the analysis are detailed in Supplementary Table ^S2 online. The peak area of each ion in the feature quantification table was adjusted by subtracting the area from the corresponding control extract. The spectral summary files (.mgf files), edited feature quantification tables (.csv files) and supplementary edge files (.csv files) were then processed using the FBMN workflow²¹ within the GNPS web platform (https://gnps.ucsd.edu). The parameters of the FBMN analysis are detailed in Supplementary Table ^S3 online. The networks generated were visualized using Cytoscape⁸³. To further characterize the results in each network, the spectra of compounds without matches to the GNPS reference libraries were analyzed using MetFrag⁸⁴, SIRIUS, including the CSI:fingerID option^42,85 and BUDDY with no chemical database restriction⁴³.

Targeted detection of Streptomyces metabolites

Detection of geldanamycin in culture extracts was performed by RP-HPLC using an Agilent 1260 Infinity Quaternary LC system (Agilent Technologies Canada Inc., Mississauga, ON) as described previously²⁶. A standard curve was generated using known amounts of a pure geldanamycin standard (Cayman Chemicals, USA) and was used for metabolite quantification.

Nigericin was detected using a modified version of a previously published protocol²⁶. Organic culture extracts were analyzed using an Agilent 1260 Infinity LC-6230 TOF LC-MS system. Extracts (5 µL) from triplicate cultures were loaded onto a Zorbax SB C-18 column (4.6×150 mm, 5 µm particle size; Agilent Technologies Canada Inc.) held at 22 °C. The column was equilibrated in 20% ammonium acetate buffer (20 mM):80% MeOH, and compounds were eluted using a linear gradient to 100% MeOH over 22.5 min at a constant flow rate of 1 mL/min. Data acquisition was performed using Agilent MassHunter version B.08.00 (Agilent Technologies Canada Inc.) and MestReNova version 14.1.2 (Mestrelab Research S.L.) was used for data analyses. Quantification of nigericin was achieved by generating a standard curve using known amounts of a pure nigericin sodium salt standard (Cayman Chemicals, USA).

Detection of elaiophylin in culture extracts was performed by RP-HPLC using an Agilent 1260 Infinity Quaternary LC system. Extracts (5 µL) prepared from triplicate cultures were loaded onto a Poroshell 120 EC-C18 column (4.6×50 mm, 2.7 µm particle size; Agilent Technologies Canada Inc.) held at 40 °C. The initial mobile phase consisted of 70% H₂O:30% MeOH, and this was held constant for 0.5 min followed by a linear gradient to 100% MeOH over 15 min. The mobile phase was maintained at this concentration for 1 min, and was then returned to 70% H₂O:30% MeOH over 1 min. The flow rate was held constant at 1 ml/min. Elaiophylin was monitored using a detection wavelength of 254 nm, and the ChemStation software version B.04.03 (Agilent Technologies Canada Inc.) was used for data acquisition. A standard curve was generated using known amounts of a pure elaiophylin standard (Cayman Chemicals, USA) and was used for metabolite quantification.

Detection of echoside C in culture extracts was performed by RP-HPLC using an Agilent 1260 Infinity Quaternary LC system. Extracts (5 µL) prepared from triplicate cultures were loaded onto a Poroshell 120 EC-C18 column (4.6×50 mm, 2.7 µm particle size) held at 40 °C. Metabolites were eluted using a linear gradient of acetonitrile (ACN) and H₂O with 0.1% v/v formic acid in both phases. The initial mobile phase consisted of 95% H₂O:5% ACN, and this was held constant for 0.5 min before changing to 5% H₂O:95% ACN over a period of 10 min. The mobile phase was maintained at this concentration for 1 min, and was then returned to 95% H₂O:5% ACN over 1 min. The flow rate was held constant at 1 mL/min. Echoside C was monitored using a detection wavelength of 260 nm, and the ChemStation software version B.04.03 was used for data acquisition. A known amount of a pure echoside C standard (Chemspace LLC, USA) was used as a reference.

Plant bioassays

The phytotoxic activity of agar cores, organic culture extracts and pure compounds was evaluated using a potato tuber slice bioassay as described before^26,86. The potato tubers used in the assays were purchased from local vendors. The phytotoxicity of extracts and pure compounds was also evaluated using a radish seedling assay as previously described²⁶. Following incubation, the root and shoot length of each seedling was measured, and the number of lateral roots per centimeter of total root length was determined using a dissecting microscope³⁴. The radish seeds (cv Cherry Belle) used in the assays were purchased from McKenzie Seeds Canada. All experiments/protocols involving plant materials were performed with relevant institutional, national, and international guidelines and legislation.

Antimicrobial bioassays

The antimicrobial activity of 11-1-2 agar cores and culture extracts were determined using different microbial indicator organisms. Cultures of B. subtilis, S. epidermidis and P. syringae pv. tomato DC3000 were prepared by inoculating a single colony into 10 mL of TSB, while S. cerevisiae cultures were prepared by inoculating a single colony into 10 mL of YPD broth. The cultures were incubated overnight at 28 °C (B. subtilis, P. syringae pv. tomato, S. cerevisiae) and 37 °C (S. epidermidis), and then 2 mL of each was added to 200 mL of molten TSA (for bacterial cultures) or YPD agar (for yeast cultures). The cells were mixed with the melted agar by swirling, and the agar was then poured into sterile Corning® Square BioAssay dishes (245 mm×245 mm) and allowed to solidify. Agar cores from 11-1-2 cultures plates, or 6 mm Whatman® filter disks (GE Healthcare Life Sciences) with 20 µL of 11-1-2 culture extract, were placed equidistantly onto the bioassay plates, after which the plated were incubated at the appropriate temperature for 24 h. The diameter of the zone of inhibition around the agar core or paper disk was recorded after the incubation period.

Bioinformatics analyses

To identify specialized metabolite biosynthetic gene clusters in 11-1-2, the genome sequence (NCBI accession NZ_CP022545) was analyzed using antiSMASH 7.1.0⁸⁷ with all the extra features available. The 11-1-2 genome was also analyzed and compared using the BiG-SCAPE workflow against the MIBiG database with the default parameters (https://git.wur.nl/medema-group/BiG-SCAPE)⁸⁸. For the large-scale network and phylogenetic analysis of specific BGCs, 2136 Streptomycetales genomes (as of November 2022) were downloaded from the National Center for Biotechnology Information (NCBI) and were processed using the command-line version of antiSMASH 5.1.2 with the bacterial setting and otherwise default parameters. Sequence similarity networks and phylogenetic relationships for the BGCs were generated using the BiG-SCAPE workflow with the default parameters (https://git.wur.nl/medema-group/BiG-SCAPE)⁸⁸. Network files were visualized using Cytoscape version 3.8.2⁸³, and the BGCs present within each network were retrieved and compared using Clinker with default parameters and Clustermap.js. was used to visualize the BGC alignment results⁸⁹. Further visualization of selected BGCs was performed using the CAGECAT suite⁹⁰ and Gene Graphics⁹¹.

The authors would like to thank Dr. Ibrahim M. Abu-Reidah of the Memorial University Centre for Chemical Analysis, Research and Training (C-CART) for their assistance with the nigericin LC-MS analysis. This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant to DB (RGPIN-2018-05155). GD-C was supported in part by a Memorial University School of Graduate Studies Fellowship.