Cyanobacteria and their secondary metabolites in three freshwater reservoirs in the United Kingdom

Bloom-forming cyanobacteria occur globally in aquatic environments. They produce diverse bioactive metabolites, some of which are known to be toxic. The most studied cyanobacterial toxins are microcystins, anatoxin, and cylindrospermopsin, yet more than 2000 bioactive metabolites have been identified to date. Data on the occurrence of cyanopeptides other than microcystins in surface waters are sparse. We used a high-performance liquid chromatography–high-resolution tandem mass spectrometry/tandem mass spectrometry (HPLC–HRMS/MS) method to analyse cyanotoxin and cyanopeptide profiles in raw drinking water collected from three freshwater reservoirs in the United Kingdom. A total of 8 cyanopeptides were identified and quantified using reference standards. A further 20 cyanopeptides were identified based on a suspect-screening procedure, with class-equivalent quantification. Samples from Ingbirchworth reservoir showed the highest total cyanopeptide concentrations, reaching 5.8, 61, and 0.8 µg/L in August, September, and October, respectively. Several classes of cyanopeptides were identified with anabaenopeptins, cyanopeptolins, and microcystins dominating in September with 37%, 36%, and 26%, respectively. Samples from Tophill Low reservoir reached 2.4 µg/L in September, but remained below 0.2 µg/L in other months. Samples from Embsay reservoir did not exceed 0.1 µg/L. At Ingbirchworth and Tophill Low, the maximum chlorophyll-a concentrations of 37 µg/L and 22 µg/L, respectively, and cyanobacterial count of 6 × 104 cells/mL were observed at, or a few days after, peak cyanopeptide concentrations. These values exceed the World Health Organization’s guideline levels for relatively low probability of adverse health effects, which are defined as 10 µg/L chlorophyll-a and 2 × 104 cells/mL. This data is the first to present concentrations of anabaenopeptins, cyanopeptolins, aeruginosins, and microginins, along with microcystins, in U.K. reservoirs. A better understanding of those cyanopeptides that are abundant in drinking water reservoirs can inform future monitoring and studies on abatement efficiency during water treatment.


Background
Cyanobacteria, also known as blue-green algae, are prokaryotes that can thrive under diverse environmental conditions and occur globally in aquatic and terrestrial habitats [1]. Adaptations that facilitate their survival and provide them with competitive advantages include the presence of a wide range of pigments (e.g., chlorophyll-a, phycocyanin, phycoerythrin Filatova et al. Environ Sci Eur (2021) 33:29 absorbing across a wide light spectrum), buoyancy, nitrogen fixation, production of dormant cells, and formation of colonies or filaments [2]. These advantages allow cyanobacteria to form high-density communities termed, blooms [3]. One major issue associated with bloom-forming cyanobacteria is the production of bioactive secondary metabolites, some of which are identified as toxins. These cyanotoxins include microcystins, cylindrospermopsin, anatoxins, and saxitoxins. Microcystins are heptapeptide molecules that contain a characteristic Adda moiety [(2S,3S,4E,6E,8S,9S)-3amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid], or derivatives thereof, that is linked to their hepatotoxic effects [4]. Microcystins are produced by multiple cyanobacteria genera, including Microcystis, Dolichospermum (previously Anabaena), Planktothrix, Nostoc, Oscillatoria, and Anabaenopsis. While more than 300 microcystin variants have been identified to date [5], a few microcystins are routinely included in the analysis of surface waters. Anatoxin-a is the most widely studied of the known anatoxins and it has a potent neurotoxic activity. Anatoxin-a binds to nicotinic acetylcholine receptors in peripheral nerve cells, which can be lethal by respiratory arrest [6]. Anabaena is the main producing genus of anatoxin-a besides Planktothrix, Oscillatoria, Microcystis, Aphanizomenon, Cylindrospermum, and Phormidium [7]. Cylindrospermopsin is a cyclic guanidine alkaloid that blocks protein synthesis, which can affect liver, kidney, thymus, and heart functions [4,8]. Cylindrospermopsin is known to be produced by Cylindrospermopsis, Aphanizomenon, Anabaena, and Lyngbya genera [9]. Due to the toxicological risks posed by various cyanobacterial metabolites, guidelines values have been introduced by several countries (such as the EU, USA, Canada, Brazil, Australia, South Africa, China, and Japan) to protect the public from exposure to cyanotoxins [10]. The World Health Organization (WHO) has set a guideline threshold value of 2 × 10 4 cyanobacterial cells/mL for recreational waters that would pose a relatively low probability of adverse health effects, which may correspond to a concentration of 20 µg/L microcystins for a Microcystis bloom [4]. WHO has set a guideline value of 1 µg/L for total microcystins-LR (MC-LR) in drinking water [11]. However, recently, an update of the WHO guideline has been finalized, and this value was modified. Now, not only MC-LR is considered, but also a total microcystins' content. The value for lifetime is 1 µg/L and for short-term events 12 µg/L in drinking water [12]. Additionally, threshold values for cylindrospermopsin, anatoxin-a, and saxitoxins are now also included, those are 3, 30, and 3 µg/L in drinking water [13][14][15]. Microcystins and anatoxin-a have been regularly reported in European waters with concentrations up to 100 and 14.4 µg/L, respectively, during seasonal studies in the recent years (e.g., Italy, Poland, Czech Republic, Germany, Russia, France, and Netherlands [16] and references herein). Many freshwater bodies in the United Kingdom (U.K.) frequently experience cyanobacterial blooms and high microcystin concentrations [17,18]. One study reported that 53% of 117 samples from cyanobacterial blooms reached microcystin concentrations above 0.2 µg/L and 13% of samples exceeded the WHO alert level of 20 µg/L [19]. Anatoxin-a has been detected in several lakes in Ireland reaching a maximum of 112 µg/L [20].
Beyond these regulated cyanotoxins, more than 2000 secondary metabolites from cyanobacteria have been structurally identified to date [5]. They are co-produced with known cyanotoxins by bloom-forming cyanobacteria [21][22][23][24][25]. In addition to microcystins, these include cyanopeptides with characteristic moieties such as cyclic cyanopeptolins [β-lactone ring, Ahp moiety (3-amino-6-hydroxy-2-piperidone)] and anabaenopeptins (ureido bond connecting the primary amine of lysine with the primary amine of the neighboring amino acid) as well as linear cyanopeptides such as aeruginosins [Choi moiety (2-carboxy-6-hydroxyoctahydroindole)] and microginins [Ahda moiety (3-amino-2-hydroxy-decanoic or octanoic acid)]. Compounds from these classes have shown acute toxicity in planktonic grazers (LC 50 < 10 mg/L) and to inhibit various enzymes (IC 50 < 30 µg/L often < 8 µg/L) [26]. While their mode of action and toxic potency is not sufficiently studied yet, their occurrence has been reported in several surface waters [27][28][29][30][31][32]. With the exception of some microcystin variants, no analytical reference standard materials exist for these other cyanopeptides. Only few studies used available bioreagents to perform absolute quantification of up to two anabaenopeptins, three cyanopeptolins, and one microginin in U.S. and Canadian lakes [27,28,32]. Globally, there is a knowledge gap regarding the entry of these cyanopeptides into drinking water treatment plants, their removal during treatment, along with cyanobacterial cells and known toxins. A recent study reported similar concentrations of cyanopeptolins and anabaenopeptins in raw drinking water compared to well-known microcystins [28]. Awareness of those cyanopeptides that are abundant in drinking water reservoirs can facilitate the prioritization of additional compounds that should be included in monitoring along with cyanotoxins and other indicator parameters such as chlorophyll-a and cell abundance. These measurements aim to improve the early identification of bloom formation and design of water management strategies. A better understanding of those cyanopeptides that are abundant in drinking water reservoirs would also aid the prioritization of compounds for which their removal efficiency during water treatment should be assessed.
In this study, we analysed toxins in raw water from three reservoirs that supply the drinking water systems in the U.K. between August and October 2019. We used a combination of target-and suspect-screening approaches based on high-performance liquid chromatographyhigh-resolution tandem mass spectrometry (HPLC-HRMS/MS). We quantified the co-occurrence of toxins and cyanopeptides across reservoirs and sampling times and compared these profiles to chlorophyll-a concentrations and cyanobacterial cell counts. To the best of our knowledge, this is the first time that anabaenopeptins, cyanopeptolins, aeruginosins, and microginins have been quantified in U.K. waters along with microcystins and anatoxin-a.

Study sites, sample collection, and storage
The water reservoirs sampled were Tophill Low (53°54′59.2416" N, 000°22′20.4024" W), Ingbirchworth (53°32′59.3808" N, 001°40′40.9800" W), and Embsay (53°59′13.9344" N, 002°00′11.1492" W). Tophill Low consists of 2 water reservoirs D and O (due to their shape), their volumes, maximal depth, and surface area are 900 million L (ML) and 773 ML, 3.8 m and 6.1 m, 0.239 km 2 and 0.141 km 2 , respectively. Both reservoirs provide 45 ML/d drinking water each. Ingbirchworth is the biggest reservoir with a volume of 1370 ML, maximal depth of 18.5 m, and a surface area of 0.235 km 2 . The Embsay reservoir has a volume of 797 ML, maximal depth of 15 m, and a surface area of 0.11 km 2 . Both Ingbrichworth and Embsay provide 20 ML/d of drinking water each. While these waters are used for recreational activities, including bird watching, game bird shooting, and sailing, their primary purpose is to serve as drinking water sources.
Samples were collected from three water treatment facilities connected to these reservoirs operated by Yorkshire Water. At Tophill Low, river water from the River Hull is pumped into two storage reservoirs, operating in series, and water is abstracted for treatment after about 30 days retention time via sub-surface draw-offs at around 4 m depth. Ingbirchworth reservoir impounds water from the Blackwater Dike, which enters the treatment plant under gravity via a draw-off tower at a depth of around 14 m. Embsay reservoir receives water from Lowburn Gill and Moor Beck and enters the treatment plant by gravity via a draw-off tower at a depth of around 12 m. Samples were collected on August 13th, September 3rd, and October 10th 2019, at the inlets to the respective water treatment facilities; in each case travel time of the raw water storage at the reservoir to the treatment plant is only a few minutes, so that the samples represent water from the raw water storage. Samples were collected in green polyethylene terephthalate bottles and transported to the laboratory in coolers, in the dark. For biological analyses, 1 L water samples were kept at 4-8 °C in the dark. Taxonomic analysis by microscopy was performed on the day following sampling and chlorophyll-a concentrations were assessed by filtration and solvent extraction followed by spectrophotometric measurement (see details in Additional file 1: Text S1, S2). Analysis of chlorophyll-a, total ammonium, nitrate, and total phosphate was done spectrophotometrically (see details in Additional file 1: Text S3-S5, and results in Additional file 1: Tables S2, S3 for Ingbirchworth and Tophill Low reservoirs). For cyanotoxin and cyanopeptide analysis, 1 L water samples were frozen and shipped to Swiss Federal Institute of Aquatic Science and Technology (Eawag), where they were stored frozen until sample preparation.

Sample preparation for chemical analysis
Cyanopeptides were extracted and pre-concentrated according to a recently validated method [33]. Briefly, 300 mL of each freshwater sample was sonicated in an ultrasonic bath (30 min, 200 W, 60 Hz) to lyse cells and release intracellular cyanopeptides. Next, samples were centrifuged at 3219.84 × g for 7 min. A 250 mL supernatant aliquot was collected and subjected to two sequential solid-phase extraction (SPE) procedures based, respectively, on Oasis HLB (500 mg, 6 cc, Waters Corporation, Milford, MA, USA) and Supelclean ™ ENVI-Carb ™ (500 mg, 6 cc, Supelco, Sigma-Aldrich, St. Louis, MO, USA) cartridges. The HLB cartridges were conditioned with methanol and equilibrated with water (10 mL each). Then, the supernatant was loaded at 1 mL/ min, and the elution of the cartridge was accomplished using 20 mL of methanol at 50 °C. The resulting extract was collected, basified up to 0.1% ammonia, and then transferred to a Supelclean ™ ENVI-Carb ™ cartridge, pre-conditioned with methanol, and equilibrated with water containing 0.1% ammonium hydroxide (10 mL each). After loading at 1 mL/min under vacuum, elution was carried out by back-flushing the cartridge with 20 mL of heated methanol (50 °C) containing 0.5% formic acid. Thus, the loading was sequential, and the extraction step was performed separately for each cartridge. Both Oasis HLB and Supelclean ™ ENVI-Carb ™ extracts were combined, concentrated in a Turbovap (Biotage) under a gentle stream of gaseous nitrogen at 25 °C, and before reaching dryness 500 µL methanol:water (1:9, v/v) was added and solutions were stored at -20 °C until analysis. All samples were analysed in triplicate (extraction step), except for Ingbirchworth October samples, which were analysed in duplicate.

Analysis of cyanopeptides
Cyanopeptides extracted from water samples were analysed using high-performance liquid chromatography coupled with high-resolution tandem mass spectrometry. Samples were injected into the HPLC system (Ulti-mate3000, Dionex, Thermo Scientific) using a CTC PAL autosampler fitted with a 20 µL stainless sample loop; the injection volume was 20 µL. Chromatographic separations were performed over a Kinetex C 18 column [2.1 × 100 mm (I.D. x L), 2.6 µm particle diameter] fitted with a SecurityGuard C 18 guard cartridge and in-line filter (aluminum frit, 0.7 µm pore diameter) at 40 °C. Mobile phase comprised 0.1% v/v formic acid in nanopure water (solvent A) (18.2 MΩ-cm-resistivity) and methanol (solvent B), which were used to generate the following binary gradient elution profile: 0/20/50/70/100/100/20/20% B at 0/1.5/5.5/21.4/21.5/26/26.1/30 min at a flow rate of 0.255 mL/min. HPLC eluates were introduced into a high-resolution tandem mass spectrometer (Exactive-Plus ™ Orbitrap, Thermo Scientific) via an electrospray ionization (ESI) probe operated with the following conditions: + 3.5 kV spray voltage, 325 °C capillary temperature, 35 arbitrary units (AU) sheath gas, 17 AU auxiliary gas, 1 AU spare gas, and 275 °C probe heater temperature. A data-dependent top-N MS 2 acquisition procedure was applied for analyte detection. Here, full-scan mass spectrometry data were acquired in profile mode from 450 to 1350 m/z at 70,000 FWHM (full-width at half maximum, at 200 m/z) resolution with automated gain control (AGC) of 5e5, 100 ms maximum ion injection time, 1 microscan, and 60% S-lens RF setting. Data-dependent MS 2 scans were triggered for the top-3 most-intense ions (with intensity > 2e4) from the preceding full scan, using the following acquisition parameters: profile acquisition mode, 17,500 FWHM resolution, 1 m/z isolation window (0 m/z offset), 1 microscan, 5e4 AGC target, 70 ms maximum ion injection time, 5 s dynamic exclusion, 'True' for 'pick others' , and stepped normalized collision energies of 15%, 30%, and 45%. The scan range for MS 2 events was dynamically adjusted (by the instrument) based on the target ion's m/z value. The suspect-screening included 1219 cyanopeptides in total with 160 microcystins, 177 cyanopeptolins, 73 anabaenopeptins, 65 cyclamides, 78 microginins, 79 aeruginosins, and 587 other compounds, accounting for structural isomers and the mass window of 450-1350 m/z. External calibration was used to support cyanopeptide quantification. Here, matrix-matched calibrants were prepared in the range 5-125 µg/L for each sampled water reservoir, using water derived therefrom as calibrant matrix. For Aeruginosin 98B, external calibration in the mobile phase was used at 0.5-500 µg/L. Each calibrant was prepared using reference standards of 10 microcystins and nodularin as well as 6 bioreagents of additional cyanopeptides. The dominant precursor ions and the limits of detection and quantification in nanopure water and lake matrices are listed in Additional file 1: Table S1. For nanopure water, the LODs were between 0.02 and 1.01 µg/L.

Analysis of anatoxin-a and cylindrospermopsin
Analysis of anatoxin-a and cylindrospermopsin was performed by HPLC (Dionex UltiMate3000 RS pump, Thermo Scientific) with an autosampler (CTC Analytics). Chromatographic separation was carried out on a reversed-phase C 18 Atlantis T3 ® column (3 mm × 150 mm, 3 µm particle size) fitted with precolumn (VanGuard ® Cartridge, Waters) and in-line filter (BGB ® ). The mobile phase consisted of nanopure water (solvent A) and acetonitrile (solvent B) both acidified with formic acid (0.05%). Binary gradient elution was carried out at a constant flow rate of 300 μL min −1 with the following gradient profile: 0/5/55/75/95/95/5% B at 0/1/10/20.5/24/26/26.1 min; and column was re-equilibrated for 3.4 min under the initial conditions. The injection volume was 20 μL. Detection of analytes was carried out by HRMS/MS (Q-Exactive-Plus ™ Orbitrap, Thermo Scientific) with an ESI probe in positive mode operated at the following conditions: 4 kV spray voltage, 325 °C capillary temperature, 35 arbitrary units (AU) sheath gas, 17 AU auxiliary gas, 1 AU spare gas, 275 °C probe heater temperature. Full-scan mass spectrometry data were acquired from 90 to 1100 m/z with a nominal resolving power of 70,000 FWHM, AGC target of 1e6, maximal injection time of 100 ms with 1 ppm mass accuracy. Data-dependent MS 2 mode was acquired at a resolving power of 17,000 FWHM, AGC target of 1e5, and maximal injection time of 50 ms. Data-dependent MS/MS acquisition was optimized for target compounds, and the applied collision energies for anatoxin-a and cylindrospermopsin were 10 eV and 25 eV, respectively. External calibration in the mobile phase was included in the analysis at 0.05-50 µg/L (details in Additional file 1: Table S1).

Identification and quantification
Compound Discoverer version 3.1.0.305 was used to process HPLC-MS/MS data files. Feature detection (extracting mass-chromatographic peaks), grouping (grouping of features with correlated retention time profiles), deconvolution (assigning adduct and isotope annotations to grouped features), and compound annotation were effected using a customized non-targeted workflow based on CyanoMetDB (v01), a database of known cyanopeptides [5]. Compound annotations were assigned based on one or more of the following conditions: match occurred between a) an experimental MS2 spectrum and one or more mzCloud library spectrum/spectra; b) an experimental MS2 spectrum and mzVault library spectrum/spectra, c) base ion (i.e. [M + H] +1 or [M + 2H] +2 ) mass and one or more entries in the mass list, i.e., metabolites listed in CyanoMetDB or; d) predicted elemental composition and one or more formulae in the mass list. Next, the features (i.e., mass-chromatographic peaks of dominant adducts + H, + Na, + NH 4 ) associated with each 'compound' annotation were inspected to ensure Gaussian-like peak integration and an isotopic 'S-fit' greater than 50%. Where no peak was identified, e.g., in the blank samples, the 'Fill gaps' workflow node was activated, retaining filled peak areas associated with 'filled by redetected peak' and 'filled by matching ion' flags. We used the confidence level scheme for mass spectrometry outlined by Schymanski et al. [34]. Herein, only those compounds that could be identified by one of the following criteria were reported: a cyanopeptide was defined as a tentative candidate (Level 3) based on exact mass (< 5 ppm mass error), accurate isotopic pattern, and evidence from fragmentation data (which was evaluated manually); a cyanopeptide was defined as a probable structure (Level 2b) based on indicative fragmentation information supporting the connectivity of the building blocks of the peptide; and a cyanopeptide was defined as a confirmed structure (Level 1) when these parameters were in agreement with available reference standards or bioreagents.
For all reference standards and bioreagents, linear regression models of the calibration curves were determined. The limits of detection and quantification (LOQ, LOD) ranged in the high ng/L to low µg/L range, and were calculated from the regression models as three or ten times, respectively, the standard deviation of the response, divided by the slope parameter (Additional file 1: Table S1). For those compounds for which a reference standard or bioreagent was available, the concentrations were quantified using the data derived from matrix-matched external calibration. Since standard reference materials were not available for most cyanopeptides, we used the class-equivalent approach for quantification according to previous work [22]. In this class-equivalent approach, quantification of cyanopeptides is achieved using the regression models of the structurally most similar bioreagent or reference standard assigned for each compound (Additional file 1: Table S2). For those compounds for which no class equivalent could be assigned, the calibration parameters of MC-LR were used for quantification. Concentrations were only reported when the peak area was above the limits of quantification (LOQs) of the respective regression model. Recoveries from the sample preparation procedure were previously assessed for anatoxin-a, cylindrospermopsin, and several microcystins in the range of 50-87% [33]. The analysis sequence spanned 3 days, with a nanopure water-based calibration curve analysed midway through and a matrix-matched calibration curve analysed at the end of the sequence. Blank injections comprised nanopure water and were introduced after each set of triplicate for a given sample and 2-3 blanks after calibration curves. No retention time drift was observed across the sequence and matrices.

Total metabolite concentrations
Water from three reservoirs was analysed for the presence of toxins and cyanopeptides. The analysed samples represent water-soluble concentrations after partial liberation of intracellular compounds (cell lysis during sonication) but no active extraction of the biomass by additional organic solvents. Overall, 8 cyanopeptides were identified by the comparison to their reference standards that were used for absolute quantification (Level 1, confirmed structures, Additional file 1: Figures S1-S8). A further 20 cyanobacterial metabolites were revealed through suspect-screening, with levels of confidence ranging from 'tentatively identified' to 'probable structures' based on the interpretation of the fragmentation spectra (Level 2-3, Additional file 1: Table S2, Figures S9-S11). Figure 1 summarises the total concentration and number of identified toxins and cyanopeptides recorded for each individual reservoir and month. Overall, samples from Ingbirchworth reservoir had the highest concentration of cyanobacterial secondary metabolites, reaching 5.2 ± 0.4 µg/L in August, 61.2 ± 1.6 µg/L in September, and 1.1 ± 0.1 µg/L in October. Overall, 16 out of 21 metabolites showed the same concentration pattern, with lower concentrations occurring in August and October, and the highest concentration in September in Ingbirchworth samples. Among the compounds that did not follow this pattern was the target analyte anatoxin-a with overall lower concentrations in the ng/L range (26 ng/L in August, 12 ng/L in September, and 6 ng/L in October). Samples from Tophill Low reservoir also had the highest concentration of cyanobacterial metabolites during September, reaching 2.4 ± 0.1 µg/L. In August and October, concentrations were below 0.2 µg/L for this reservoir. Only traces of single cyanobacterial metabolites were tentatively identified in samples from the Embsay reservoir with comparable concentrations across the sampling months (0.10 ± 0.02 µg/L).

Diversity of metabolites
Compounds belonging to different cyanopeptide classes were identified, including microcystins, anabaenopeptins, aeruginosins, cyanopeptolins, microginins, and anatoxina. Figure 2 shows the concentrations of individual metabolites recorded in Ingbirchworth reservoir samples in September, when the highest concentrations and greatest number of cyanobacterial metabolites were observed. We were able to quantify 59% of the total metabolite concentration through targeted analysis of microcystins and anabaenopeptins using available reference standards or bioreagents. Anabaenopeptins, cyanopeptolins, and microcystins were the three dominating classes representing 37%, 36%, and 26% of total number of identified compounds, respectively. Among the anabaenopeptins, the dominating compounds were anabaenopeptin B (12 ± 2 µg/L), anabaenopeptin A (9.3 ± 0.7 µg/L), and oscillamide Y (1.5 ± 0.6 µg/L), all of which were quantified with bioreagents. The cyanopeptolin anabaenopeptilide 202A (identification confidence level 3) was the most abundant compound with 22 ± 2 µg/L; however, this compound was only quantified indirectly by the classequivalent approach, because no bioreagent was available at the time of analysis. Among the detected microcystins, the most abundant were MC-RR, an MC-RR variant 1024 (group of isomeric microcystins, details see Additional file 1: Table S2), MC-LF or MC-FL, [Dha 7 ]MC-LR, and MC-LR. The maximum MC-LR concentration recorded in Ingbirchworth reservoir was 1.8 ± 0.2 µg/L, while the maximum total microcystin concentration summing all variants was 16 ± 1 µg/L. Both of these values are below the water quality guideline value of 20 µg/L suggested by the WHO for a moderate health alert in recreational waters [4]. Notably, individual anabaenopeptins occurred at slightly higher concentrations than the maximal total microcystins concentration. The risks posed by anabaenopeptins remains unclear. They are not classified as toxins, yet do have inhibitory effects on enzymes. Like some microcystin congeners, anabaenopeptins have been shown to inhibit protein phosphatases, albeit with lower potency [35]. Anabaenopeptins were also identified as potent inhibitors of carboxypeptidases and the concentrations recorded here (1.5-12 µg/L individual and > 22 µg/L total), exceed the IC 50 values for anabaenopeptin B (1 µg/L) by tenfold (IC 50 of 371 µg/L for anabaenopeptin A) [36]. The only difference between these two anabaenopeptins is the C-terminal amino acid outside of the cyclic structure, where anabaenopeptin has an arginine while anabaenopeptin A has a tyrosine moiety. The dominating cyanopeptolin, anabaenopeptilide 202A, can be produced by genera Anabaena, and while toxicological studies have not been reported, other cyanopeptolins are known to inhibit proteases involved in metabolism and blood coagulation [26,37]. Figure 3 shows the concentration of individual metabolites for the September sample from Tophill Low reservoir, when, analogous to Ingbirchworth, the highest concentrations and greatest number of cyanobacterial metabolites were observed. Anabaenopeptins accounted  [38]. Aeruginosins also inhibit human serine proteases involved in blood coagulation at IC 50 values ranging from 4 to 93 µg/L [39,40]. No microcystins were detected in samples from Tophill Low reservoir.
Cylindrospermopsin was not identified in any reservoir sample.

Seasonal trends of chlorophyll-a and cyanobacterial cell counts
Cyanobacteria are known to be present in the reservoirs analysed herein, which is supported by the identification of cyanotoxins and other cyanobacterial metabolites discussed above. Ingbirchworth reservoir has experienced periphyton blooms in recent years, which may be related to increased farming activity in the catchment area. Authorities monitor water quality and cyanobacteria in these reservoirs to manage possible taste and odour development and to prevent issues related to cyanotoxins. Figure 4 summarises the chlorophyll-a concentrations and cyanobacterial cell counts recorded at Ingbirchworth and Tophill Low, between July and Both values exceed the WHO guideline level for relatively low probability of adverse health effects, which is currently defined as 10 µg/L chlorophyll-a [4]. For Ingbirchworth and Tophill Low reservoirs, this was the case most of the year, for 9 and 6 months in 2019, respectively (Additional file 1: Tables S3, S4). While a general trend of chlorophyll-a concentration with cyanobacterial abundance is expected [41][42][43], an absolute relationship across water bodies cannot be inferred. Based on the samples evaluated in this study, the peak cyanopeptide concentration at Ingbirchworth was almost 30-fold higher compared to Tophill Low, while the maximum chlorophyll-a concentrations were more comparable. The cyanobacterial cell counts peaked in both reservoirs about 2 weeks after the peak cyanopeptide concentration, reaching 6 × 10 4 cells/mL with being Anabaenopsis the most abundant genus. Similar to the chlorophyll-a concentrations, the relationship of absolute cyanopeptide concentrations with cyanobacterial cell count differs between reservoirs. Compared to other bloom-forming cyanobacteria, such as Microcystis, Dolichospermum/ Anabaena, and Oscillatoria/Planktothrix, less information is available about secondary metabolites produced by Anabaenopsis, though it has been reported that they produce microcystins [44,45]. At Tophill Low, Anabaena (1.1 × 10 4 cells/mL) and Oscillatoria (0.4 × 10 4 cells/ mL) also contributed to the cyanobacterial abundance in September. At Ingbirchworth, meanwhile, only a low co-abundance of Microcystis (0.09 × 10 4 cells/mL) and Snowella (0.05 × 10 4 cells/mL) was detected alongside the dominant Anabaenopsis (5.8 × 10 4 cells/mL). Despite Anabaenopsis being the major cyanobacterial genus at both Inbirchworth and Tophill Low, the cyanopeptide profiles of these two reservoirs were not identical. Both had high concentrations of anabaenopeptin A, anabaenopeptin B, oscillamide Y, and aeruginosin 822, though only Ingbirchworth samples were rich in the cyanopeptolin anabaenopeptilide 202A as well as various microcystins. Tophill Low samples contained no detectable microcystins, though did contain additional anabaenopeptins and aeruginosins. This discrepancy suggests that other cyanobacterial species may have contributed to the observed metabolite profiles, and/or that the production dynamics of metabolites from Anabaenopsis differed between proportions of major metabolite classes is present in the inset pie chart. Compounds marked with an asterisk (*) were substances quantified with reference standards or bioreagents; all other compounds were identified by suspect-screening and quantified as class-equivalents. Error bars represent the standard deviation of triplicate samples reservoirs. No cyanobacteria were identified in Embsay reservoir, which agrees with only traces of cyanotoxins detected herein. We also did not observe a correlation of total ammonium, nitrate, total phosphate, and temperature with cyanobacterial abundance or toxin concentrations in Ingbirchworth or Tophill Low reservoirs (data in Additional file 1: Tables S3, S4).

Conclusions
In this study, 28 cyanopeptides were identified in three freshwater reservoirs in the U.K, including microcystins, anabaenopeptins, aeruginosins, cyanopeptolins, microginins, and anatoxin-a. Of these, 8 were directly quantified using reference standards or bioreagents, while the remaining 20 cyanopeptides were identified by suspect-screening and quantified using a class-equivalents approach. Total concentrations of microcystins did not exceed WHO guideline level of relatively low probability of adverse health effects in recreational water. However, levels of chlorophyll-a and cyanobacterial cell count were higher than the guideline values. These two parameters had a similar pattern as cyanobacterial metabolites (peaking in September).

Fig. 4
Chlorophyll-a concentrations and cyanobacteria cell counts (secondary y-axis) measured in a Ingbirchworth and b Tophill Low reservoir samples. Total cyanopeptide concentrations are plotted for comparison (additional y-axis on the left). Shaded areas highlight the biological data closest to the cyanopeptide data sampled on August 13th, September 3rd, and October 10th 2019