Utility of a PCR-based method for rapid and specific detection of toxigenic Microcystis spp. in farm ponds

Abstract

Microcystis is a widespread freshwater cyanobacterium that can produce microcystin, a potent hepatotoxin harmful to animals and humans. Therefore, it is crucial to monitor for the presence of toxigenic Microcystis spp. to provide early warning of potential microcystin contamination. Microscopy, which has been used traditionally to identify Microcystis spp., cannot differentiate toxigenic from non-toxigenic Microcystis. We developed a PCR-based method to detect toxigenic Microcystis spp. based on detection of the microcystin synthetase C (mcyC) gene and 16S rRNA gene. Specificity was validated against toxic and nontoxic M. aeruginosa strains, as well as 4 intergeneric freshwater cyanobacterial strains. Analytical sensitivity was as low as 747 fg/µL genomic DNA (or 3 cells/µL) for toxic M. aeruginosa. Furthermore, we tested 60 water samples from 4 farm ponds providing drinking water to swine facilities in the midwestern United States using this method. Although all water samples were positive for Microcystis spp. (i.e., 16S rRNA gene), toxigenic Microcystis spp. were detected in only 34 samples (57%). Seventeen water samples contained microcystin (0.1–9.1 μg/L) determined with liquid chromatography–mass spectrometry, of which 14 samples (82%) were positive for mcyC. A significant correlation was found between the presence of toxigenic Microcystis spp. and microcystin in water samples (p = 0.0004). Our PCR method can be a low-cost molecular tool for rapid and specific identification of toxigenic Microcystis spp. in farm ponds, improving detection of microcystin contamination, and ensuring water safety for farm animals.

Keywords

microcystin PCR

Introduction

Cyanobacteria, also known as blue-green algae, are a group of diverse photoautotrophic prokaryotes living in different aquatic environments (e.g., fresh, brackish, and saline water), in soil, and even on rocks.^{18,22,24,35,60} Numerous cyanobacterial species, such as Microcystis, Anabaena, and Nodularia, can grow explosively under warm and nutrient-rich conditions to form noticeable blooms in freshwater bodies, such as ponds, lakes, and slow-moving rivers, with different appearances, such as scums, mats, and foams.^19,23,31 Such blooms are called harmful algal blooms (HABs) and are detrimental to aquatic environments in which they occur. Two detrimental effects caused by HABs are depletion of dissolved oxygen as a result of cyanobacterial cell decomposition after death, and the release of cyanotoxins. Cyanotoxins are a threat to the health of humans or animals that ingest contaminated water.^18,66 Cyanobacterial blooms are predicted to become more widespread, intense, and of longer duration under anticipated climate change scenarios.^8,45

Microcystin is a potent hepatotoxic cyanotoxin with a low median lethal dose that can cause adverse health problems in animals and humans by damaging their hepatocytes after ingestion of contaminated water and causing liver failure and death.^13,21,51,58 Over 80 microcystin isoforms have been discovered based on their common peptides and heptapeptide structure. All are synthesized through enzymatic catalysis by microcystin synthetases, which consist of several polyketide synthases and non-ribosomal peptide synthetases and are encoded by clustered microcystin synthetase (mcy) genes. In the case of Microcystis aeruginosa, which is one of the well-known microcystin producers and is a well-characterized freshwater cyanobacterial species,^9,56,57 the bacterium has 10 clustered mcy genes in its genome, designated A–J. The mcy gene cluster encompasses 2 operons, mcyABC and mcyDEFGHIJ, transcribed in opposite directions, terminating at mcyC and mcyJ at the 3′-ends, respectively.⁶¹

Farm ponds are a common drinking water source for livestock, and tend to be nutrient-rich, which can be a favorable environment for cyanobacterial growth leading to HABs.⁴⁸ Therefore, a microcystin-contaminated water supply can be a health risk to livestock if not properly managed. Although HAB monitoring is a common component of state and federal water quality monitoring programs concerning public and aquatic ecosystem health, such mandatory programs do not exist for water bodies serving animal production.^49,50 Furthermore, blue-green algae toxicoses are not diseases reportable to veterinary authorities. It would be helpful to the livestock industry to develop more inexpensive methods to monitor HABs, especially toxicity in farm ponds because animal deaths from cyanotoxin toxicoses are common.^{4,10,17,37,52}

Microscopic examination is a traditional approach to identify the presence of Microcystis spp. in water.^23,58 Microscopic examination is, however, laborious and time-consuming, and requires taxonomic knowledge and experience to determine accurate identification. Furthermore, many Microcystis spp. have both toxic and nontoxic strains that share identical morphologic traits.^42,43 Liquid chromatography–mass spectrometry (LC-MS) analysis has been employed in microcystin identification and quantification.^13,58,61 However, LC-MS requires sophisticated instruments and complex analytical processes, which may limit its implementation, and may cause some laboratories to use a more convenient and less expensive antigen ELISA instead of LC-MS, although accuracy and/or low sensitivity can be issues with the ELISA method.

Molecular techniques have been developed that overcome the limitations of microscopy. These techniques include the development of nucleic acid–based methods to detect toxigenic M. aeruginosa, such as random amplified polymorphic DNA, amplified fragment length polymorphism, and PCR.^{7,38,40,41,62,64} Given that the clustered mcy genes are one of the most important genomic distinctions of toxic strains from nontoxic strains,^30,71 many genes (e.g., mcyA, B, D, E, and G) have been selected and shown by molecular assays to be specific genetic markers for toxigenic Microcystis spp.^{6,32,54,62,64,72} Analysis of genetic markers of mcyC has not been commonly done despite its 3,876-bp size and conserved regions for specific primer designing.⁵⁸ On the other hand, the translated enzyme McyC (i.e., protein) plays an indispensable role in the completion of microcystin synthesis because it contributes to functions such as condensation, aminoacyl adenylation, and cyclization on the linear precursor. Moreover, transcription of mcyC also indicates the accomplishment of transcription of mcyABC operon given its 3′-terminal locus.⁶¹ Therefore, the mcyC gene may be an ideal candidate for detection of toxigenic Microcystis spp. and downstream gene expression studies.

We developed a PCR-based method targeting the mcyC and 16s rRNA genes of Microcystis spp. to specifically detect toxigenic Microcystis spp. in a rapid and differential manner and evaluated its performance on pond waters providing drinking water to swine facilities in the midwestern United States. We also evaluated whether the nutrient level of pond water (i.e., eutrophication) is linked to microcystin content and cyanobacterial biomass.

Materials and methods

Cyanobacterial strains and culturing

Each of 7 cyanobacterial strains (Table 1) was cultured in a transparent polystyrene tube containing 5 mL of BG-11 medium (UTEX, Austin, TX) at 20°C under the light intensity of ~ 40 μmol of photons/m² s (equivalent to 3,000 lux) following a light/dark regime of 12 h/12 h per day.^69,70 Cells were collected when they grew to log phases, which empirically occurs ~ 1 wk after inoculation.

Table 1.

Cyanobacterial strains used in development of a PCR-based method for detection of toxigenic Microcystis spp. in farm pond water.

Strain	Source	Microcystin production (μg/L)*
Microcystis aeruginosa LB2385	UTEX	149
M. aeruginosa LB2388	UTEX	146
M. aeruginosa LB2386	UTEX	0†
Nodularia spumigena B2091	UTEX	0
Oscillatoria sp. 29135	ATCC	0
Anabaena sp. 29211	ATCC	0
Cylindrospermum licheniforme 29412	ATCC	0

ATCC = American Type Culture Collection; UTEX = Culture Collection of Algae at the University of Texas at Austin.

Microcystin production was tested by liquid chromatography–mass spectrometry.

†

Zero was used for “not detectable” to be consistent with other numerical values presented in the table.

Preparation of cyanobacterial genomic DNA

All cyanobacterial cultures were centrifuged at 11,000 × g for 10 min at ambient temperature. The pelleted cells were dispersed with 180 μLof buffer ATL plus 20 μLof proteinase K solution (DNeasy blood & tissue kit; Qiagen, Germantown, MD). The suspension was vortexed vigorously and incubated at 55°C overnight for thorough cell lysis. DNA extraction was then carried out as per the kit’s protocol for gram-negative bacteria with the minor modification of doubling the volume of buffer ATL, proteinase K solution, and buffer AL. Each genomic DNA (gDNA) extract was finally eluted with 200 μLof autoclaved distilled and deionized water (ddH₂O), and gDNA concentration was measured spectrophotometrically (NanoPhotometer classic; Implen, Westlake Village, CA).

Design, validation, and selection of specific primers for toxigenic Microcystis spp

Twenty primer pairs (Table 2) were devised according to the mcy cluster (genomic loci: 3,480,000–3,545,000 bp) of M. aeruginosa NIES-843 (GenBank accession NC010296) based on its alignment with homologous sequences from the non-redundant/nucleotide (nr/nt) and genome databases by the Basic Local Alignment Search Tools (BLAST, http://blast.ncbi.nlm.nih.gov/Blast.cgi) in search of conserved regions. To identify the genus of Microcystis, an extra primer pair (16S_F683: 5′-CGAAAGCGTGCTACTGGGCTGTAT-3′; 16S_R1217: 5′-TCGCTGGCTCTCGCGAGTTC-3′) was designed based on the 16S rRNA gene (16S hereafter) sequence of M. aeruginosa strain SPC777 (GenBank accession EF121241) aligned against 18 homologous sequences from other cyanobacterial genera. Primer specificities were analyzed in silico using the online bioinformatics tool Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). These primers were then synthesized (Integrated DNA Technologies, Coralville, IA).

Table 2.

Validation of specificity and analytic sensitivity of 20 mcy primer pairs used in the development of a PCR-based method for detection of toxigenic Microcystis spp. in farm pond water.

	Primer*	Sequence (5′–3′)†	Specificity						Analytical sensitivity (fg/μL)
			Microcystis aeruginosa		Nodularia spumigena	Oscillatoria sp.	Anabaena sp.	Cylindrospermum licheniforme	7.47 × 10⁵	7.47 × 10⁴	7.47 × 10³	7.47 × 10²	7.47 × 10¹
			Toxic	Nontoxic	Nodularia spumigena	Oscillatoria sp.	Anabaena sp.	Cylindrospermum licheniforme	7.47 × 10⁵	7.47 × 10⁴	7.47 × 10³	7.47 × 10²	7.47 × 10¹
1	mcyB_F3039	TWCYYTAGTGAGTAAAATTCAGGAAAA	+	–	–	–	–	–	+	+	+	–	–
	mcyB_R3594	RCCATCACARATAATATGATGAATTTC
2	mcyC_F580	GCATGGCAAAATMAGTKGTTAGAAAG	+	NS	–	NS	–	NS	NT	NT	NT	NT	NT
	mcyC_R1298	TGACCGYYCATTCTRGMGAT
3	mcyC_F1411	GAGRTATCCACGCGCTAAACC	+	NS	–	–	NS	–	NT	NT	NT	NT	NT
	mcyC_R2043	AATTTGACCTCTCCAGTTAAAMCAGA
4	mcyC_F2170	CATATAATCCAAGCGCTCAGGG	+	NS	–	–	–	NS	NT	NT	NT	NT	NT
	mcyC_R2607	CCTGAAACYATCGCCAGAATGA
5	mcyC_F3181	CCGAAAGAACGTCCTTATGTG	+	–	–	–	–	–	+	+	–	–	–
	mcyC_R3311	TCATCGCTAATATACCGGAAAG
6	mcyC_F3181	CCGAAAGAACGTCCTTATGTG	+	–	–	–	–	–	+	+	+	+	–
	mcyC_R3294	CGTTAKATACTCYGCTAAATTGGC
7	mcyE_F8905	GATAAYTTYTTTGAAATYGGYGG	+	NS	–	–	–	NS	NT	NT	NT	NT	NT
	mcyE_R9186	ATAARCACTKGGCATTCCRTA
8	mcyF_F250	GTCATTGGCTGTTGTACGGCTC	+	–	–	–	–	–	+	+	+	+	–
	mcyF_R585	ACACCCAGAAATATAGGAGCGACTG
9	mcyG_F852	GGCMGAAGAACTTAAGCACGYC	+	–	–	–	–	–	+	–	–	–	–
	mcyG_R1119	CCGGATAGTTGCTCCMGGAATCGGC
10	mcyG_F3793	GATGACYCCTTAAACCAAACCGM	+, NS	NS	–	–	NS	NS	NT	NT	NT	NT	NT
	mcyG_R4705	GTGGAACAGTTGCAGATGGGG
11	mcyG_F4160	GTAACCATTTYACTAATKCYGGG	+	NS	NS	NS	NS	NS	NT	NT	NT	NT	NT
	mcyG_R4705	GTGGAACAGTTGCAGATGGGG
12	mcyG_F5518	GGYGCSGGAACYGGAGCWAC	+	NS	NS	NS	NS	NS	NT	NT	NT	NT	NT
	mcyG_R5893	CGCTAAAWCKCCACCAMCCTTG
13	mcyH_F181	GCCTCTAATGCAGTTAATAGCTTCG	+	–	–	–	–	–	+	+	–	–	–
	mcyH_R452	CTTTGATCYGGATTATCCACATCG
14	mcyH_F571	GCCGCCTTGATTTTAGTGGC	+	NS	–	–	–	–	NT	NT	NT	NT	NT
	mcyH_R903	CCCCCGAATATAGGCYGGAGC
15	mcyI_F71	TGTATGATGAAGCGGGGSAA	+	NS	–	NS	NS	NS	NT	NT	NT	NT	NT
	mcyI_R391	GSAGCTTTTTGGCAAGGGCT
16	mcyI_F432	TCGCAATCAKATSCAACCGA	+	–	–	–	NS	NS	NT	NT	NT	NT	NT
	mcyI_R824	TCAAAGACATCKAGGGCGCA
17	mcyJ_F72	CCTTAATCCRGYTCGAGAAATGGCG	+	NS	–	–	–	–	NT	NT	NT	NT	NT
	mcyJ_R439	TTATCCTGTCCGATAACCCTGC
18	mcyJ_F385	CAARTAGAAATTGCTCAAGAAMGAGT	+	NS	–	NS	NS	NS	NT	NT	NT	NT	NT
	mcyJ_R871	CAAAYGCCATAAACCAMCCCC
19	mcyJ_F441	ACTTCAAGTCGGCTCGGCTA	+	NS	NS	NS	NS	NS	NT	NT	NT	NT	NT
	mcyJ_R621	CCCGCAGCCAGAAGTTAATT
20	mcyJ_F442	CTTCAAGTMGGCTCGGCTACTC	+	NS	–	–	–	NS	NT	NT	NT	NT	NT
	mcyJ_R892	CAAACGCCATAAACCAACCCCG

NS = nonspecific amplification; NT = not tested; + = specific amplification; – = no amplification.

Primers in bold are pairs highly specific for toxigenic Microcystis spp.

†

Degenerate bases are: W = A or T; Y = C or T; R = A or G; M = A or C; K = G or T; S = C or G.

PCR

DNA amplification was done in a 25-μL reaction volume (ExTaq PCR kit; Takara Bio, Mountain View, CA) consisting of 1 μLof template gDNA, 2.5 μLof 5× ExTaq buffer, 2 μLof total dNTP (2.5 mM), 1 μLof each primer (10 mM), 0.125 μLof ExTaq polymerase (5 U/μL), and 17.375 μLof ddH₂O. PCR was performed (ABI 2720 thermal cycler; Applied Biosystems, Foster City, CA), and the thermal cycling conditions were: 94°C for 5 min for a hot-start to inhibit nonspecific annealing and primer dimer formation; 35 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s; 72°C for 5 min to ensure the complete amplicon elongation; and 4°C thereafter. All PCR products were analyzed by capillary electrophoresis (QIAxcel system; Qiagen).

Evaluation of PCR analytical performance

Genomic DNA of all 7 cyanobacterial strains was used to validate the specificity of primers in PCR. Furthermore, 10 mL of M. aeruginosa LB2385 (6 × 10⁶ cells/mL) culture were collected and suspended in 180 μLof buffer ATL plus 20 μLof proteinase K solution, and total gDNA was extracted and dissolved in 50 µL of ddH₂O. The DNA concentration was adjusted to 74.7 ng/µL. Then, serial 10-fold gDNA dilutions were prepared in 20 µL of ddH₂O each. Finally, gDNA dilutions containing 7.47 × 10⁵, 7.47 × 10⁴, 7.47 × 10³, 7.47 × 10², and 7.47 × 10¹ fg DNA/µL, which would be equivalent to M. aeruginosa LB2385 cultures containing 3 × 10³, 3 × 10², 3 × 10¹, 3 × 10⁰, and 3 × 10^‒1 cells/μL, respectively, were used to estimate the analytical sensitivity (i.e., limit of detection) of the PCR as per protocols in previous sections.

PCR products were sequenced to confirm the amplification specificity and lack of cross-contamination between samples and controls. As per manufacturers’ recommended protocols, PCR products were purified (QIAquick PCR purification kit; Qiagen), or they were ligated into a plasmid vector (pGEM-T vector system; Promega, Madison, WI) and cloned into DH5α Escherichia coli competent cells that grew on lysogeny broth (LB) agar plates for blue-white screening followed by re-culturing of positive colonies in liquid LB media. Purified PCR products or extracted recombinant plasmids (ZR plasmid miniprep, classic kit; Zymo Research, Irvine, CA) were then submitted to the Iowa State University DNA Facility (Ames, IA) for sequencing.

Collection and processing of water samples

Water was collected from 4 farm ponds (designated as 1–4) located in the midwestern United States. These ponds were the main drinking water sources for nearby swine facilities. Surface water was sampled from 2 shore sites in the pond and assigned as “surface bloom” (SB) or “open water” (OW). SB samples were collected where blooms or aggregated cyanobacteria (i.e., algae) were visible; OW samples were collected where water was clear. Sampling was done biweekly from June to October in 2015, except on June 25 when only SB water was sampled because of a miscommunication. In total, 60 water samples (32 SB and 28 OW) were obtained. Each sample was aliquoted into 3 portions on site. Lugol solution (2.5 mL; LabChem, Zelienople, PA) was added to 500 mL of water in an amber plastic bottle for microscopic phytoplankton identification.⁶³ The other 2 portions of 100 mL of water each were stored in a clear plastic bottle. All samples were shipped immediately on ice to our laboratory and frozen upon arrival until analyses were conducted.

For molecular testing, water samples were thawed and filtered through a 100-µm opening screen (Nitex; Sefar, Buffalo, NY) to remove large particles. Filtrates were then centrifuged at 11,000 × g for 10 min at ambient temperature. Genomic DNA was extracted from the resultant pellets in the previously mentioned manner. Final gDNA solutions were serially diluted in a 10-fold gradient up to 10^‒4 in order to minimize potential PCR inhibitory substances.

Microscopic phytoplankton identification

Phytoplankton samples were sub-sampled and examined under a bright–field microscope at 400× magnification. Samples were analyzed using a nanoplankton counting cell similar to a Palmer–Maloney counting cell, following methods described previously.^1,2 Phytoplankton were identified to the genus level. Cell numbers were counted based on natural taxonomic units, and linear dimensions were measured on the first 50 natural taxonomic units. Simple geometric model formulas were used to calculate the biovolume (µm³/L) of plankton divisions and the total phytoplankton community.^16,25 Although biomass has been estimated from biovolume assuming a density of 1.1 g/cm³, wet biomass was calculated from the biovolume using a density of 1 g/cm³ for ease of conversion as reported previously.^15,27

Microcystin and nutrient measurement

A liquid chromatograph–mass spectrometry (LC-MS) method was employed to detect microcystin in water samples.⁶⁶ For testing, 20 mL of each water sample was freeze–thawed 5 times, sonicated for 10 min, and loaded into a silica-based bonded phase cartridge (SepPak C18 cartridge; Waters, Milford, MA) that was prewashed with 5 mL of absolute methanol and 10 mL of ddH₂O. The prewashed sample in the cartridge was then washed with 5 mL of 20% methanol aqueous solution. Samples were then eluted with 7 mL of absolute methanol, dried under N₂, and re-dissolved in 400 μLof absolute methanol. Finally, microcystin analysis was performed on a LC-MS instrument (Varian 310 LC/MS; Agilent Technologies, Santa Clara, CA) with a high-performance liquid chromatography (HPLC) pump and auto-sampler (ProStar 201 HPLC pump; ProStar 410 auto-sampler; Agilent Technologies). Microcystins were separated on a 150 × 2 mm HPLC column (5 C18-A; Polaris Industries, Medina, MN). The mobile phase consisted of A (0.1% formic acid) and B (10 mM ammonium acetate plus 0.1% formic acid in methanol) with a flow rate of 0.2 mL/min. The separation was done using a gradient starting at 50% B to 95% B in 10 min and then re-equilibrated back to 50% B in 6 min. Mass spectral analysis was done in electrospray ionization with a detector voltage of 1,500, needle voltage of 3,500, drying gas temperature of 325°C, and nebulizer gas pressure of 50 psi. Four common microcystin isoforms were detected as follows: RR-, molecular weight (MW)+520.1, capillary voltage (CV)-108; LA-, MW+910.5, CV-80; LR-, MW+995.6, CV-132; YR-, MW+1045.6, CV-76. Retention times were 8, 10.6, 9.3, and 8.8 min, respectively. A 100 μg/L mixed standard consisting of the 4 isoforms (RR, LA, LR, and YR) was prepared by combining 100 μL of each microcystin standard at 1 ng/μL and diluting to 1 mL with methanol. Microcystin LA, LR, YR, and RR were quantified simultaneously against standard curves established by injecting 20, 10, 4, and 2 μL of the standard mixture to obtain 100, 50, 20, and 10 μg/L of each isoform. The detection limit of the LC-MS was 0.1 μg/L.

Nitrates and nitrites (N), and phosphates (P), were analyzed on the 28 OW samples following persulfate digestion of unfiltered water samples using protocols established in a commercial nutrition analysis laboratory. Based on values from controls, the N level was determined using the second-derivative spectroscopy method¹²; the P level was estimated using the ascorbic acid method.³⁴ Furthermore, N:P ratios were calculated as molar ratios.

Statistical analysis

Logistic regression was used to evaluate the predictive value of Microcystis biomass for existence of mcyC (i.e., toxigenicity).¹¹ The presence and absence of mcyC were manually set as 1 and 0 (i.e., binary assignment), respectively, and were then plotted against corresponding biomasses in a coordinate. This regression was done to test the null hypothesis that biomasses between samples are not significantly different regardless of presence or absence of mcyC (i.e., biomass has no predictive value for toxigenicity or toxicity).

Total microcystin (RR, LA, LR, and YR) contents were plotted against corresponding Microcystis biomasses, and the linear regression was performed by Pearson correlation coefficient (Pearson r).²⁰ It was postulated that the 2 parameters would be positively correlated because intuitively, microcystin production should be more intense when Microcystis is increasing in biomass and vice versa. Likewise, potential correlation of microcystin content and abundances of Microcystis and all cyanobacterial taxa to N, P, or N:P were also investigated by linear regression, assuming that nutrient is a key stimulator in cyanobacterial growth and toxin production. In order to investigate factors likely explaining variance in microcystin content and biomass among ponds, SB and OW data in each pond were aggregated per date to calculate pond averages (except 6 of 25 samples) for all variables, resulting in a new 32-sample dataset.

A McNemar test was performed to determine whether the presence of microcystin (via LC-MS) could be predicted by the presence of mcyC (via PCR).³⁶ The null hypothesis was that such prediction was invalid. The presence and absence of either microcystin or mcyC were numerically set as 1 and 0 (i.e., binary), respectively, and individual samples (n = 60) were compared regarding their assay results between LC-MS and PCR.

Results

Analytical performance of PCR assays

Microcystin production from M. aeruginosa LB2385 and M. aeruginosa LB2388 was confirmed by LC-MS; no detectable amounts of microcystin were determined for the remaining 5 strains (Table 1). Cyanobacterial gDNA was extracted and diluted to ~ 1 ng/µL for validation of PCR specificity. Although all mcy primers were specific to toxic Microcystis spp. by in silico analysis, 6 of the 20 primer pairs were determined to be specific for toxic M. aeruginosa LB2385 and LB2388 among all reference cyanobacterial stains (Table 2). No amplification from M. aeruginosa LB2386 elucidated the absence of mcy cluster in nontoxic strains. In comparison, all 3 M. aeruginosa strains were tested positive by 16S PCR but not the other 4 intergeneric strains (Fig. 1A).

Figure 1.

Electrophoretic results of 7 cyanobacterial strains by PCR using primer pairs: A. 16S_F683/R1217; B. mcyC_A; C. mcyC_B. Lanes 1–8 represent Microcystis aeruginosa LB2385, M. aeruginosa LB2388, M. aeruginosa LB2386, Nodularia spumigena B2091, Oscillatoria sp. 29135, Anabaena sp. 29211, Cylindrospermum licheniforme 29412, and negative control, respectively.

Two primer pairs (mcyF_F250/R585 and mcyC_F3181/R3294) showed the highest analytical sensitivity (i.e., limit of detection) among the 6 specific primer pairs (Table 2), which reached as low as 747 fg gDNA/µL in a reaction or 3 cells/μL in water. Because of the potential advantages of mcyC, the mcyC_F3181/R3294 pair was eventually selected for subsequent experiments and designated mcyC_A. Given that subspecies sequence variations were observed at genomic loci where mcyC_A binds, the mcyC_F3181/R3311 pair was selected as a secondary primer pair for point mutation investigation and designated mcyC_B (Fig. 1B, 1C).

Microcystis detection by 16S PCR and biomass calculation

All 60 water samples were positive for Microcystis sp. by 16S PCR. Microscopically, however, Microcystis sp. was not detected in 4 samples (Fig. 2). In general, Microcystis biomasses were at a relatively low level and differed temporally across ponds. Overall, ponds 2 and 4 had the lowest and highest average biomasses, respectively. Intra- and inter-pond Microcystis accumulation peaked on different dates.

Figure 2.

Temporal changes of Microcystis and cyanobacterial biomasses (mg/L), microcystin (μg/L), and nutrient concentrations (μg/L), and presence and absence of mcyC in field water samples (SB vs. OW) from 4 farm ponds (1–4). Solid and clear bars represent the microcystin concentrations in SB and OW by liquid chromatography–mass spectrometry, respectively. Square, rhombus, triangle, and circle represent Microcystis biomass, cyanobacterial biomass, N concentration, and P concentration, respectively, and no symbol is shown for samples measured 0. Plus (+) and minus (–) signs on top of each panel indicate presence and absence of mcyC by PCR, respectively; X marks refer to the 4 lost samples.

Detection of toxigenic Microcystis spp. by mcyC PCR

In contrast to the 16S PCR result, only 34 of 60 water samples (57%) were positive for mcyC (i.e., toxigenic Microcystis spp.) by mcyC_A, including the 18 positive samples (30%) by mcyC_B (Fig. 3). All PCR products were shown to be derived from mcyC of Microcystis spp. according to their sequence similarities (>99% in BLAST). All ponds had toxigenic Microcystis spp. at the beginning of survey and more than once during the study, and water samples from pond 1 were positive more frequently for mcyC (n = 12 of 15) than those from the other 3 ponds (n = 6, 7, or 8 of 15; Fig. 2). Neither SB nor OW had consecutive and consistent PCR-positive outcomes throughout the entire sampling period in any pond, displaying non-preferential distributions of toxigenic Microcystis spp. between the 2 collection sites at each pond.

Figure 3.

Venn diagram of 37 water samples positive for microcystin by liquid chromatography–mass spectrometry (LC-MS) and/or mcyC by PCR with mcyC_A and mcyC_B. Numbers in overlapped and open areas in circles are actual sample numbers positive for the intended target(s) by corresponding assays. The percentages in parentheses are the proportions of each type of positive result in the 37 positive samples.

Microcystin and nutrient contents

Seventeen of the 60 water samples (28%) were determined to contain microcystin (mainly microcystin-LR) by LC-MS (0.1–9.1 μg/L; Fig. 3), some of which were beyond the limit of 1 μg/L for human drinking water safety recommended by the World Health Organization.⁶⁷ The highest concentrations appeared in SB in October for all ponds except pond 2 (Fig. 2).

Fourteen (82%) and 8 (47%) of the 17 microcystin-positive samples were positive for mcyC (i.e., toxigenic Microcystis spp.) by PCR using mcyC_A and mcyC_B, respectively (Fig. 3). All 8 positive samples by mcyC_B were included in the 14 positive samples by mcyC_A. The remaining 3 microcystin-positive water samples that were mcyC-negative by PCR were still negative when retested by PCR using a mcyF-specific primer set (mcyF_F250/R585), the other primer pair with high analytical specificity and sensitivity (Table 2). Interestingly, cyanobacterial genera capable of producing microcystin other than Microcystis (e.g., Anabaena and Aphanizomenon) were observed microscopically in these 3 samples.

The levels of N and P in the water samples were 1,160–2,560 μg/L and 58–335 μg/L, respectively. Only 4 of 28 OW samples were eutrophic as indicated by N >1,880 μg/L, whereas 21 samples were P eutrophic (>84 μg/L).²⁹ Additionally, the N:P ratios of 23 samples were less than the Redfield ratio (16:1).⁵³ These results suggest that most samples were N-stressed, which is unfavorable for nitrogen-rich microcystin production.

Pairwise relationship between biomass, mcyC, microcystin, and nutrient

No significant difference was observed in Microcystis biomass between mcyC-positive and -negative samples given that p values in logistic regressions for mcyC_A and mcyC_B were both 1, respectively (Fig. 4), suggesting that Microcystis biomass is not a useful predictor of the existence of toxigenic Microcystis spp. Furthermore, Microcystis biomass explained little variance in microcystin concentration (R² = 0.184, p < 0.01; Fig. 5).

Figure 4.

Logistic regression plots of Microcystis biomass (mg/L) against existence of mcyC by PCR using mcyC_A and mcyC_B for 60 water samples. Values 1.0 and 0.0 on the ordinate refer to the probabilities of presence and absence of mcyC, respectively. Regression equations are shown beside the curves.

Figure 5.

Linear regression plot of Microcystis biomass (mg/L) against corresponding microcystin concentration (μg/L) for 60 water samples by Pearson r.

No significant linear correlation was found between the nutrient parameters (i.e., N, P, and N:P) and response variables (i.e., cyanobacterial biomass, Microcystis biomass, and microcystin concentration) across the 4 ponds by either original data (n = 60 individual water samples) or averaged data (n = 32; combining both SB and OW data for each pond at each sampling time), or in each individual pond (n = 15 in each analysis; R² < 0.5). Nevertheless, when data were further aggregated to calculate pond averages per site (SB or OW) spanning the whole sampling stretch and compared across the 4 ponds (n = 4 in each analysis), a positive correlation was observed only between microcystin and N (R² = 0.940, p < 0.05) in SB, between biomass and N (R² = 0.861, p < 0.05) in OW, and between biomass and P (R² = 0.948, p < 0.05) in SB (Fig. 6).

Figure 6.

Linear correlations between microcystin concen-tration, cyanobacterial biomass, and nutrients in 4 ponds. Data were aggregated to calculate pond average (•) per site during the whole sampling period and were analyzed by Pearson r. A positive correlation was observed only between microcystin and nitrate and nitrite (N) in surface bloom (SB) samples, between biomass and N in open water (OW) samples, and between biomass and phosphate (P) in SB samples.

Based on a McNemar test, the presence of microcystin was not independent of existence of mcyC determined by mcyC_A (p = 0.0004), whereas strong independency was found regarding mcyC_B (p = 0.819), demonstrating the significant connection between the 2 variables using the former primer pair. Therefore, PCR testing using mcyC_A could provide a more reliable prediction of the presence of microcystin in water samples.

Discussion

As reported in the literature and observed in our cultures, it is very difficult, if not impossible, to morphologically differentiate toxic and nontoxic strains of Microcystis spp.^42,43 This difficulty has been a major problem in the development of a microscopic method for early detection of toxic Microcystis spp. and other toxic cyanobacteria. Microcystins can be produced by many Microcystis species, such as M. aeruginosa, M. botrys, and M. viridis.^41,46 Their mcy genes share almost identical sequences, which makes it even more difficult, if not impossible, to specifically differentiate M. aeruginosa from other toxigenic Microcystis spp. It has been postulated that the taxonomic classification within Microcystis should be restructured, and it has been suggested that many species of Microcystis should be unified into a single M. aeruginosa, and that some should be regarded as a morphologic variant of this unified species.⁴⁴ We could speculate that most toxigenic Microcystis spp. should be phylotypic congeners of M. aeruginosa that are mistakenly classified into various species, which could explain why their mcy genes share highly similar sequences. This agrees with what was observed in the alignment of downloaded mcy sequences and in sequencing PCR products of mcyC from water samples, in which all DNA sequences were identical or similar to those from M. aeruginosa. Our data suggest that our newly developed PCR method likely detects all toxigenic Microcystis spp.

Among the 20 primer pairs targeting different mcy genes, mcyC_A was 1 of only 2 pairs outweighing the rest of the pairs regarding analytical performance (i.e., specificity and sensitivity). Even though a conventional gel-based PCR was employed, the estimated sensitivity using mcyC_A (i.e., 747 fg DNA or 3 cells/µL) is comparable with those via quantitative PCR previously published on the same topic, further demonstrating excellent analytical sensitivity of our assay.^6,54,64,72 Our study also demonstrated the importance of primer design to overcome sequence variability. The reference M. aeruginosa strain NIES-843 and our stock strain LB2385 were isolated from Japan and Canada, respectively.^28,42 Inter-strain mutation was noticed in their mcyC (e.g., 2 and 4 nucleotide mutations were found in the reverse primers mcyC_R3294 and mcyC_R3311, respectively). Then, the 2 mutational nucleotides were replaced with 2 degenerate bases in mcyC_R3294 comprising mcyC_A, in contrast to the non-degenerate mcyC_R3311 composing mcyC_B; these 2 primer pairs were intentionally used to investigate the influence of point mutations on PCR detection of the target. As suspected, such a mutation in the primer-binding site interfered with assay sensitivity given that PCR testing using mcyC_A detected a significantly higher proportion of mcyC-positive samples (57%) than one using mcyC_B (30%), which detected just a subset of samples by mcyC_A. Therefore, it is important to pay attention to mutations when designing primers and to circumvent the issue by using degenerate nucleotides.

Although no Microcystis sp. was observed microscopically in 4 water samples, the existence of Microcystis sp. was confirmed by 16S PCR, demonstrating the usefulness of molecular techniques to identify low-biomass or morphologically unidentifiable cyanobacteria (i.e., cellular degradation). Despite such an advantage, 3 water samples were found to contain microcystin without finding mcyC by PCR. Cyanobacterial genera other than Microcystis, such as Anabaena, Aphanizomenon, and Planktothrix, can produce microcystin but have unique mcy clusters regarding genetic arrays and sequences.^9,56,57 Those microcystin producers were found in the 3 microcystin-positive water samples and logically could have produced the detected microcystin. However, intergeneric specific primers are not designed yet, which remains an important task for developing toxigenicity tests in the future. It is noteworthy that microcystin levels in the 4 farm ponds were sporadically higher than the lethal dose reported in swine, emphasizing the importance of monitoring for microcystin and toxic cyanobacteria in farm ponds.²³

Negative mcyC PCR results may also be attributable to a trace amount of toxigenic Microcystis spp. and/or potent PCR inhibition. Even though dilution of DNA extracts was able to reduce PCR inhibition, such a procedure could also diminish template concentration below the detection level of the assay. This possibility necessitates better sample processing to remove PCR inhibitory substances to the greatest extent prior to PCR tests.⁶⁸ Because cyanobacterial cells are covered with a thick and rigid peptidoglycan cell wall that increases toughness, sample pretreatment should be improved for more effective cellular disruption to gain a higher DNA yield.²⁶ Bead beating is a method to consider given that it has been proven to effectively disrupt marine phytoplankton cells.⁶⁹ Additionally, accompanying exogenous substances that are often found in aquatic environments and act as PCR inhibitors (e.g., humic substances, urea) should be removed.⁵ A reasonable means of removal is to apply a strong disaggregating chemical (e.g., chloroform) that excludes some interferents.⁶⁹ Introduction of these extra steps could help to increase DNA yield and purity from cyanobacterial samples, which would improve the accuracy of cyanobacterial detection, including detection of toxigenic taxa.

Twenty water samples contained mcyC yet lacked microcystin. Given that PCR test results were statistically correlated with LC-MS data, these false-positive PCR results may be attributed to the perceived higher sensitivity of nucleic acid–based molecular assays against their analytical chemistry rivals. Nonetheless, non-expression of mcy genes cannot be ruled out especially at the transcriptional level. If this is the case, detection of genes (DNA) can be a good indicator for toxigenicity but not occurrent toxicity. Initiation of gene expression may require certain environmental stimulations such as rich nitrogen sources, iron limitation, or intense illumination.³⁹ It is speculated that the detection of gene transcripts (messenger RNA) may have a better connection with toxin production given that transcription is the primary step in gene expression. It remains to be investigated whether messenger RNA detection outperforms DNA detection with respect to precise revelation of toxicity, and what intra- and extra-cellular factors influence toxin production.

It is well known that nitrogen and phosphorus are indispensable elements for phytoplankton survival in the forms of N and P, respectively, and phytoplankton can bloom explosively as a result of excessive nutrient availability such as occurs during eutrophication.³ In limnology, N, P, and N:P are 3 critical parameters to indicate a waterbody’s trophic status and are usually closely linked to microalgal growth.⁵³ Significant correlations did not exist between any of these 3 parameters and cyanobacterial biomass or microcystin content on the whole, but discernible correlations were observed when pond averages per site spanning the entire sampling timeframe were compared, which supports findings from large empirical studies of temperate lakes.^14,65 Similar to previous reports,⁶⁰ N was also a significant predictor for microcystin in our study. Positive relationships between nitrogen-rich cyanotoxins and N are well documented and hypothesized to result from the large N demands for synthesizing cyanotoxins. Interestingly, N:P failed to predict cyanobacterial biomass, which contradicts the discovery that cyanobacteria could dominate under N-limitation because some genera have the ability to fix atmospheric nitrogen.⁵⁹ Average N:P ratios suggested common N-deficient pond conditions because 23 of the 28 nutrient data indicated N-deficiency (i.e., N:P < 16). Moreover, no congruent temporal trends were found within individual farm ponds with respect to nutrients versus microcystin content, existence of mcyC, or cyanobacterial biomass. The drastic temporal variabilities are in agreement with a previous study that aquatic N or P could not be simply associated with cyanobacterial growth or cyanotoxin production because free N and P can easily be influenced by various biological and abiotic activities, even if cyanobacteria dominate phytoplankton communities or cyanotoxins are detectable.^33,47,55 These findings suggest that multivariate approaches would be needed to account for complex interactions among predictive factors that create necessary conditions to stimulate cyanobacterial growth and cyanotoxin production.

Footnotes

Acknowledgements

We thank Amie Curtis, Dwayne Schrunk, Ju Ji, and Yikang Chai for their excellent assistance in sample collection and transportation, storage, and statistical analysis. We also thank Daniel Kendall who performed microscopic analyses of phytoplankton samples.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The study was supported in part by funding from the Innovative Swine Industry Enhancement Grant Program by Iowa Attorney General’s Office, Iowa State University (ISU) Health Research Initiative, and ISU Veterinary Diagnostic Laboratory Research Support Fund.

ORCID iD

Jian Yuan

References

Acker

. Preparation of algal samples for analysis using Palmer-Maloney cells. In: Charles

, et al., eds. Protocols for the Analysis of Algal Samples Collected as Part of the US Geological Survey National Water-Quality Assessment Program. Philadelphia, PA: Academy of Natural Sciences, 2002:55–60.

Amand

ALS

. Biological examination. In: Clesceri

, et al., eds. Standard Methods for the Examination of Water and Wastewater. 20th ed. Washington, DC: American Public Health Association, 1998.

Anderson

DM.

Red tides. Sci Am 1994;271:52–58.

Backer

, et al. Canine cyanotoxin poisonings in the United States (1920s–2012): review of suspected and confirmed cases from three data sources. Toxins 2013;5:1597–1628.

Bessetti

An introduction to PCR inhibitors. J Microbiol Methods 2007;28:159–167.

Beversdorf

, et al. Microcystin mcyA and mcyE gene abundances are not appropriate indicators of microcystin concentrations in lakes. PLoS One 2015;10:e0125353.

Burns

, et al. Molecular detection of genes responsible for cyanobacterial toxin production in the genera Microcystis, Nodularia, and Cylindrospermopsis. In: Spencer

JFT

Ragout de Spencer

, eds. Public Health Microbiology: Methods and Protocols. Vol. 268. Totowa, NJ: Humana Press, 2004:213–222.

Carey

, et al. Eco-physiological adaptations that favour freshwater cyanobacteria in a changing climate. Water Res 2012;46:1394–1407.

Christiansen

, et al. Microcystin biosynthesis in Planktothrix: genes, evolution, and manipulation. J Bacteriol 2003;185:564–572.

10.

Classen

, et al. Microcystin toxicosis in nursery pigs. J Swine Health Prod 2017;25:198–205.

11.

Cox

DR.

The regression analysis of binary sequences. J R Stat Soc B 1958;20:215–242.

12.

Crumpton

, et al. Nitrate and organic N analyses with second-derivative spectroscopy. Limnol Oceanogr 1992;37:907–913.

13.

Delaney

, et al. Toxicity of microcystin-LR, isolated from Microcystis aeruginosa, against various insect species. Toxicon 1995;33:771–778.

14.

Downing

, et al. Predicting cyanobacteria dominance in lakes. Can J Fish Aquat Sci 2001;58:1905–1908.

15.

Filstrup

, et al. Cyanobacteria dominance influences resource use efficiency and community turnover in phytoplankton and zooplankton communities. Ecol Lett 2014;17:464–474.

16.

Findenegg

GR.

Relations between carbonic anhydrase activity and uptake of HCO₃- and Cl- in photosynthesis by Scenedesmus obliquus. Planta 1974;116:123–131.

17.

Fitzgerald

, et al. Toxicosis due to microcystin hepatotoxins in three Holstein heifers. J Vet Diagn Invest 1993;5:651–653.

18.

Francis

Poisonous Australian lake. Nature 1878;18:11–12.

19.

Galat

, et al. Large-scale patterns of Nodularia spumigena blooms in Pyramid Lake, Nevada, determined from Landsat imagery: 1972–1986. Hydrobiologia 1990;197:147–164.

20.

Galton

Regression towards mediocrity in hereditary stature. J R Anthropol Inst G 1886;15:246–263.

21.

Giannuzzi

, et al. An acute case of intoxication with cyanobacteria and cyanotoxins in recreational water in Salto Grande Dam, Argentina. Mar Drugs 2011;9:2164–2175.

22.

Glover

, et al. Photosynthetic characteristics of coccoid marine cyanobacteria. Arch Microbiol 1981;129:42–46.

23.

, et al. Microcystis bloom formation in the lower Nakdong River, South Korea: importance of hydrodynamics and nutrient loading. Mar Freshw Res 1999;50:89–94.

24.

Herath

The algal bloom problem in Australian waterways an economic appraisal. Aust J Agric Resour Econ 1995;63:77–86.

25.

Hillebrand

, et al. Biovolume calculation for pelagic and benthic microalgae. J Phycol 1999;35:403–424.

26.

Hoiczyk

, et al. Cyanobacterial cell walls: news from an unusual prokaryotic envelope. J Bacteriol 2000;182:1191–1199.

27.

Holmes

, et al. Collection, fixation, and enumeration of phytoplankton standing stock. In: National Research Council (U.S.), ed. Recommended Procedures for Measuring the Productivity of Plankton Standing Stock and Related Oceanic Properties. Washington, DC: National Academy of Sciences, 1969:17–46.

28.

Hughes

, et al. Toxicity of a unialgal culture of Microcystis aeruginosa. Can J Microbiol 1958;4:225–236.

29.

Jorgensen

. Water quality: the impact of eutrophication. In: Williams

, ed. Lakes and Reservoirs. Vol. 3. Shiga, Japan: United Nations Environment Program–International Environment Technology Center, and the International Lake Environment Committee Foundation, 2001.

30.

Kaneko

, et al. Complete genomic structure of the bloom-forming toxic cyanobacterium Microcystis aeruginosa NIES-843. DNA Res 2007;14:247–256.

31.

Kellar

, et al. Physiological adaptations in response to environmental stress during an N₂-fixing Anabaena bloom. Appl Environ Microbiol 1980;40:587–595.

32.

Kurmayer

, et al. The abundance of microcystin-producing genotypes correlates positively with colony size in Microcystis sp. and determines its microcystin net production in Lake Wannsee. Appl Environ Microbiol 2003;69:787–795.

33.

Kuypers

, et al. The microbial nitrogen-cycling network. Nat Rev Microbiol 2018;16:263–276.

34.

Lipps

. Inorganic nonmetallic constituents. In: Clesceri

, et al., eds. Standard Methods for the Examination of Water and Wastewater. 20th ed. Washington, DC: American Public Health Association, 1998.

35.

Lopes

, et al. Planktonic and benthic cyanobacteria of European brackish waters: a perspective on estuaries and brackish seas. Eur J Phycol 2011;46:292–304.

36.

McNemar

Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika 1947;12:153–157.

37.

Mez

, et al. Identification of a microcystin in benthic cyanobacteria linked to cattle deaths on alpine pastures in Switzerland. Eur J Phycol 1997;32:111–117.

38.

Neilan

, et al. Genetic diversity and phylogeny of toxic cyanobacteria determined by DNA polymorphisms within the phycocyanin locus. Appl Environ Microbiol 1995;61:3875–3883.

39.

Neilan

, et al. Environmental conditions that influence toxin biosynthesis in cyanobacteria. Environ Microbiol 2013;15:1239–1253.

40.

Nishihara

, et al. Random amplified polymorphic DNA (RAPD) analyses for discriminating genotypes of Microcystis cyanobacteria. Biosci Biotech Bioch 1997;61:1067–1072.

41.

Ordorika

, et al. Distribution of microcystin-producing and non-microcystin-producing Microcystis sp. in European freshwater bodies: detection of microcystins and microcystin genes in individual colonies. Syst Appl Microbiol 2004;27:592–602.

42.

Otsuka

, et al. Morphological variability of colonies of Microcystis morphospecies in culture. J Gen Appl Microbiol 2000;46:39–50.

43.

Otsuka

, et al. Phylogenetic relationships between toxic and non-toxic strains of the genus Microcystis based on 16S to 23S internal transcribed spacer sequence. FEMS Microbiol Lett 1999;172:15–21.

44.

Otsuka

, et al. A proposal for the unification of five species of the cyanobacterial genus Microcystis Kutzing ex Lemmermann 1907 under the rules of the Bacteriological Code. Int J Syst Evol Microbiol 2001;51:873–879.

45.

Paerl

, et al. Climate change: a catalyst for global expansion of harmful cyanobacterial blooms. Environ Microbiol Rep 2009;1:27–37.

46.

Park

, et al. Temporal variabilities of the concentrations of intra- and extracellular microcystin and toxic Microcystis species in a hypertrophic lake, Lake Suwa, Japan (1991–1994). Environ Toxic Water 1998;13:61–72.

47.

Paytan

, et al. The oceanic phosphorus cycle. Chem Rev 2007;107:563–576.

48.

Peltzer

, et al. Effects of agricultural pond eutrophication on survival and health status of Scinax nasicus tadpoles. Ecotoxicol Environ Saf 2008;70:185–197.

49.

Poikane

, et al. A hitchhiker’s guide to European lake ecological assessment and intercalibration. Ecol Indic 2015;52:533–544.

50.

Pollard

, et al. The promise and potential of continental-scale limnology using the US environmental protection agency’s national lakes assessment. Limnol Oceanogr Bull 2018;27:36–41.

51.

Pouria

, et al. Fatal microcystin intoxication in haemodialysis unit in Caruaru, Brazil. Lancet 1998;352:21–26.

52.

Puschner

, et al. Blue-green algae toxicosis in cattle. J Am Vet Med Assoc 1998;213:1605–1607, 1571.

53.

Redfield

AC.

On the proportions of organic derivations in seawater and their relation to the composition of plankton (reprint). Benchmark Papers Ecol 1934;1:176–192.

54.

Rinta-Kanto

, et al. Quantification of toxic Microcystis spp. during the 2003 and 2004 blooms in western Lake Erie using quantitative real-time PCR. Environ Sci Technol 2005;39:4198–4205.

55.

Rinta-Kanto

, et al. Lake Erie Microcystis: relationship between microcystin production, dynamics of genotypes and environmental parameters in a large lake. Harmful Algae 2009;8:665–673.

56.

Rouhiainen

, et al. Genes coding for hepatotoxic heptapeptides (microcystins) in the cyanobacterium Anabaena strain 90. Appl Environ Microbiol 2004;70:686–692.

57.

Saker

, et al. Detection of microcystin synthetase genes in health food supplements containing the freshwater cyanobacterium Aphanizomenon flos-aquae. Toxicon 2005;46:555–562.

58.

Shirai

, et al. Toxicity and toxins of natural blooms and isolated strains of Microcystis spp. (Cyanobacteria) and improved procedure for purification of cultures. Appl Environ Microbiol 1991;57:1241–1245.

59.

Smith

VH.

Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 1983;221:669–671.

60.

Thomas

, et al. Spatial and temporal distribution of cyanobacterial soil crusts in the Kalahari: implications for soil surface properties. Geomorphology 2007;85:17–29.

61.

Tillett

, et al. Structural organization of microcystin biosynthesis in Microcystis aeruginosa PCC7806: an integrated peptide–polyketide synthetase system. Chem Biol 2000;7:753–764.

62.

Tillett

, et al. Detection of toxigenicity by a probe for the microcystin synthetase A gene (mcyA) of the cyanobacterial genus Microcystis: comparison of toxicities with 16S rRNA and phycocyanin operon (phycocyanin intergenic spacer) phylogenies. Appl Environ Microbiol 2001;67:2810–2818.

63.

Utkilen

, et al. Fieldwork: site inspection and sampling. In: Chorus

Bartram

, eds. Toxic Cyanobacteria in Water: A Guide to Their Public Health Consequences, Monitoring and Management. London, UK: CRC Press, 1999:329–345.

64.

Vaitomaa

, et al. Quantitative real-time PCR for determination of microcystin synthetase E copy numbers for Microcystis and Anabaena in lakes. Appl Environ Microbiol 2003;69:7289–7297.

65.

Watson

, et al. Patterns in phytoplankton taxonomic composition across temperate lakes of differing nutrient status. Limnol Oceanogr 1997;42:487–495.

66.

Woodhouse

, et al. Cyanotoxins. In: Sharma

, et al., eds. Cyanobacteria: An Economic Perspective. Oxford, UK: Wiley, 2014:257–268.

67.

World Health Organization (WHO). Guidelines for Drinking-water Quality: Health Criteria and Other Supporting Information (Addendum to Vol. 2). Geneva, Switzerland: WHO, 1998.

68.

Yilmaz

, et al. Improved methods for the isolation of cyanobacterial DNA from environmental samples. J Phycol 2009;45:517–521.

69.

Yuan

, et al. An improved DNA extraction method for efficient and quantitative recovery of phytoplankton diversity in natural assemblages. PLoS One 2015;10:e0133060.

70.

Yuan

, et al. Development of a rapid detection and quantification method of Karenia mikimotoi by real-time quantitative PCR. Harmful Algae 2012;17:83–91.

71.

Zhang

J-Y

, et al. Complete genome sequence and genomic characterization of Microcystis panniformis FACHB 1757 by third-generation sequencing. Stand Genomic Sci 2016;11:11.

72.

Zhang

, et al. Analysis of cylindrospermopsin- and microcystin-producing genotypes and cyanotoxin concentrations in the Macau storage reservoir. Hydrobiologia 2014;741:51–68.