Widespread Distribution and Identification of Eight Novel Microcystins in Antarctic Cyanobacterial Mats

Abstract

The microcystin (MC) content and cyanobacterial community structure of Antarctic microbial mat samples collected from 40 ponds, lakes, and hydroterrestrial environments were investigated. Samples were collected from Bratina Island and four of the Dry Valleys, Wright, Victoria, Miers, and Marshall. Enzyme-linked immunosorbent assays (ELISAs), liquid chromatography-mass spectrometry (LC-MS), and protein phosphatase 2A (PP-2A) inhibition assays resulted in the identification of low levels (1 to 16 mg/kg [dry weight]) of MCs in all samples. A plot of indicative potencies of MCs (PP-2A inhibition assay/ELISA ratio) versus total MCs (ELISA) showed a general decrease in potency, as total MC levels increased, and a clustering of values from discrete geographic locations. LC-tandem MS analysis on selected samples identified eight novel MC congeners. The low-energy collisional activation spectra were consistent with variants of [d-Asp³] MC-RR and [d-Asp³] MC-LR containing glycine [Gly¹] rather than alanine and combinations of homoarginine [hAr²] or acetyldemethyl 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid (acetyldemethyl ADDA) [ADMAdda⁵] substitutions. Nostoc sp. was identified as a MC producer using PCR amplification of a region of the 16S rRNA gene and the aminotransferase domain of the mcyE gene. Automated ribosomal intergenic spacer analysis (ARISA) was undertaken to enable a comparison of cyanobacterial mat community structure from distant geographical locations. Two-dimensional multidimensional scaling ordination analysis of the ARISA data showed that in general, samples from the same geographic location tended to cluster together. ARISA also enabled the putative identification of the MC-producing Nostoc sp. from multiple samples.

Extreme conditions, including relentless katabatic winds, permanently low temperatures and precipitation, and depauparate carbon supply, ensure that life in the Dry Valleys of Eastern Antarctica is primarily restricted to soil environments (1, 43). However, in selected above surface habitats, e.g., lakes, ponds, and on moist soil, cyanobacteria have thrived, forming thick cohesive mats (50). The cyanobacterial species (predominately members of the orders Nostocales and Oscillatoriales) within these mats are adapted to tolerate harsh physicochemical parameters, including high salinities and UV radiation.

Cyanobacteria worldwide produce a range of natural toxins collectively known as cyanotoxins. The mechanisms of toxicity are very diverse, ranging from hepatotoxicity and neurotoxicity to dermatotoxicity. The most ubiquitous of the cyanotoxins are the hepatotoxic microcystins (MCs). MCs are cyclic peptides, and to date, more than 70 MCs have been isolated and characterized (55). MCs are synthesized nonribosomally by a large peptide synthetase and polyketide synthase enzyme complex (48). An increasing number of species from both planktonic and benthic habitats are known to produce MCs (17, 23, 42). Despite considerable research, the biological and functional roles of MCs are poorly understood. Various hypotheses have been proposed, including defense against grazers (27), gene regulation (10), allelopathic interactions (44), and intraspecific regulation (39). Recently, relatively low concentrations (<15 mg/kg MC-LR [dry weight]) of MCs were identified in cyanobacterial mats from meltwater ponds on McMurdo Ice Shelf in Antarctica (16, 19). The identification of MCs in these mats provides evidence to dispute some of their putative roles, for example, defense against grazers (16). To date, MCs have been identified in Antarctica only from meltwater ponds on Bratina Island, and the extent of their occurrence in other locations in Antarctica was unknown. Additionally, definitive identification of specific MC producers and information on MC variants produced was limited.

In this study, samples from 40 ponds, lakes, and hydroterrestrial environments from four Dry Valleys (Wright, Victoria, Marshall, and Miers) in Eastern Antarctica and Bratina Island were investigated for the presence of MCs. Variations in total MC concentration within samples have been reported when different detection methods were used (26, 28). Therefore, in our study, all samples were analyzed for MCs by at least two of the following methods: liquid chromatography-mass spectrometry (LC-MS), protein phosphatase 2A (PP-2A) inhibition assay, and enzyme-linked immunosorbent assay (ELISA). The genes involved in MC synthesis (mcyA to mcyJ) have been identified and characterized (10, 31, 48), enabling PCR amplification of them to be used as an indication of MC production potential. Sequencing of a region of the mcyE gene and 16S rRNA genes from unicyanobacterial material was used to identify one of the cyanobacterial species responsible for MC production.

The cyanobacterial community structure of each mat was assessed using automated ribosomal intergenic spacer analysis (ARISA). Subsequent multivariant analysis of ARISA profiles allowed the investigation of correlations between community structure, MC production, and geographical locations to be made and enabled the investigation of the influence of water chemistry parameters on cyanobacterial miscellany.

MATERIALS AND METHODS

Samples and sample collection.

Benthic microbial mat material was collected from 13 meltwater ponds on the McMurdo Ice Shelf, located south of Bratina Island (78°00′S, 165°30′E; January 2004), 12 ponds in Wright Valley (77°31′S, 160°45′E; January 2004), six ponds in Victoria Valley (77°22′S, 162°10′E January, 2004), and five locations around Lake Miers (78°6′S, 164°0′E; December 2006). Two samples of hydroterrestrial mats were collected in Miers Valley from moist areas in front of Adams Glacier (MVAG1) and Miers Glacier (MVMG1; December 2006). One sample was collected from the shoreline of Lake Purgatory (S78°03′S 163°51′E; December 2006, included in the Miers Valley samples for all analyses), and three hydroterrestrial mats were collected from the upper reaches of Marshall Valley (78°03′S, 136°55′E; December 2006). Samples were scraped from the sediment surface using a stainless steel spatula (swabbed with alcohol between collections) and placed in sterile 50-ml Falcon tubes.

Water chemistry parameters (Cl⁻, SO₄²⁻, Ca²⁺, and Na⁺ concentrations and pH) were determined for all samples (except Lake Purgatory) as previously described (16, 28). No physicochemical data were collected for hydroterrestrial mats collected in Mier and Marshall valleys.

Isolation of DNA and ARISA fingerprinting and analysis.

Subsamples of the 40 frozen microbial mats were lyophilized. DNA was extracted from approximately 0.1 g of lyophilized material using the MoBio Power Soil kit (Carlsbad, CA) according to the manufacturer's protocol.

ARISA PCRs were carried out using cyanobacterial specific primers as described previously (53). ARISA fragment lengths (AFL) were analyzed by Genetic Profiler V.2 (GE Healthcare, Auckland, New Zealand), and the data were transferred to Microsoft Excel for further processing. All AFL information was transposed to presence/absence data for further analysis. AFL were aligned using an Excel macro. AFL that differed by less than 3 bp were considered identical (53). If multiple AFL fell within this range, then only the AFL with the highest florescence was maintained. AFL less than five times the baseline fluorescence in height were removed, since they could not be fully distinguished from instrument “noise” (14). AFL shorter than 300 bp were removed, as they were considered too short to be true intergenic spacer regions (53).

Nonmetric multidimensional scaling (MDS) based on Bray-Curtis similarities was undertaken using the PRIMER 6 software package (PRIMER-E, Ltd., United Kingdom). This ordination technique ranks the order of similarity of any two communities as an inverse function of the distance between the points representing the communities on the plot (24). Thus, communities with the highest similarity are represented on the plot by points that are plotted closest together. Nonmetric MDS was undertaken with 100 random restarts, and results were plotted in two dimensions. Plots with a stress value of less than 0.20 provide interpretable information (9). Agglomerative, hierarchical clustering of the Bray-Curtis similarities was carried out using the CLUSTER function of PRIMER 6 and plotted onto the two-dimensional MDS at a similarity level of 40%.

Analysis of similarities (ANOSIM) was used to test for significant differences in AFL profiles between samples from Bratina Island and Wright, Victoria, Miers, and Marshall valleys of the Antarctic. ANOSIM produces a sample statistic, R, which is a relative measure of separation of the a priori-defined groups. The R statistic is based both on the difference of mean ranks between groups and within groups. An R value of 1 indicates that community composition is totally different, and an R value of 0 indicates no difference. A Monte Carlo randomization was used to test the statistical significance of R.

To assess which combination of water chemistry variables accounted for the biotic patterns observed, the computer program BEST (8) was used. Only samples for which all water chemistry data was available were used in the analysis.

Microcystin analysis.

Frozen microbial mat samples were lyophilized, and the freeze-dried samples were stored at −18°C. Subsamples (0.2 g) of ground freeze-dried material were placed in 50-ml Falcon tubes, and 15 ml of 70% methanol was added to each tube. The samples were ultrasonicated in a bath (60 min), vortexed, and centrifuged at 20,000 × g at 4°C (10 min). The extraction was repeated, and the supernatants were combined and dried under nitrogen with heating at 35°C. The dried extract was solubilized in 2 ml of 20% methanol in MilliQ water and stored at −18°C (0.07 g of freeze-dried cyanobacterial material per ml). For LC-MS analysis, samples were filtered through a 0.45-μm filter (Minisart RC 4; Sartorius).

The PP2A inhibition assay was carried out in 96-well plates by the method of Mountfort et al. (28). The total 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl-4,6-decadienoic acid (ADDA)-containing MC/nodularin content in the reconstituted extracts was quantified with a competitive indirect ELISA by using the methods of Fischer et al. (11). This method is referred to henceforth as ADDA-ELISA. The 11 samples collected in 2006 were also analyzed using a similar ELISA (AgResearch, Ruakura, New Zealand) that has lower cross-reactivity with free ADDA and nodularin (Lyn Briggs, personal communication). That method is referred to henceforth as ELISA.

The reconstituted extracts for samples collected in 2006 and selected samples from Bratina Island were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS-MS) for 13 MC variants and nodularin (51). MCs were separated by LC (Alliance 2695; Waters Corp., MA) using a 5-μm Luna C₁₈ column (150 by 2 mm) (Phenomenex, CA) with a water-methanol-acetonitrile gradient containing 0.15% formic acid (0.2 ml min⁻¹; 10-μl injection). The Quattro Ultima TSQ mass spectrometer (Waters-Micromass, Manchester, United Kingdom) was operated in the electrospray ionization in positive ion mode with multiple reaction monitoring (MRM) using MS-MS channels set up for MC-RR, didesmethyl MC-RR, demethyl MC-RR, demethyl MC-LR, demethyl MC-YR, didesmethyl MC-LR, desmethyl MC-LR, desmethyl MC-FR, desmethyl MC-WR, desmethyl MC-AR, desmethyl MC-LA, desmethyl MC-LY, desmethyl MC-LW, desmethyl MC-LF, and nodularin. The m/z 135 fragment ion from the protonated molecular cation was selected for each toxin ([MH₂]²⁺ for MC-RR [MC-RR] and variants; MH⁺ for the others). The LC-MS responses were calibrated using mixed standard solutions of MC-RR, MC-LR, MC-YR, and nodularin (Alexis Corporation, Lausen, Switzerland). The MRM response factor for MC-RR was used for quantitation of MC-RR variants, and the MC-LR factor was used for MC-LR variants. Full-scan and fragment ion spectra were also gathered for samples MVAG1 and MVMG1. Full-scan spectra identified molecular species for potential MCs, and parent ion scanning experiments identified the components yielding the ADDA fragment m/z 135 on collisional activation. Daughter ion spectra from the protonated molecular species (collision energy 52 eV for MH⁺ or 30 eV for MH₂²⁺) were gathered for each of the components and examined for MC structural fragment ions.

Identification of a MC producer.

The dominant cyanobacteria species in the MVMG1 sample was determined using an Olympus light microscope (BX51; Olympus, Wellington, New Zealand). Species identification were made were made by referring to the article by Komárek and Anagnostidis (22). Unicyanobacterial material of the dominant species was isolated from the MVMG1 sample using sterile tweezers. Purity was confirmed using microscopic examination, and DNA was extracted as described above. PCR amplification of a cyanobacterial partial 16S rRNA gene segment and a region of the mcyE gene was carried out by the method of Jungblut and Neilan (20). PCR products were purified using a High Pure PCR product purification kit (Roche Diagnostics) and sequenced bidirectionally using the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems). The phylogenetic relatedness of the 16S rRNA and mcyE gene sequence obtained in this study was established using sequences from the NCBI GenBank database. An ARISA profile was obtained from the unicyanobacterial DNA material as described above. This profile was then used for the putative identification of this species in each of the cyanobacterial mat community ARISA profiles.

Nucleotide sequence accession number.

Sequences generated during this work were deposited in the NCBI GenBank database under accession numbers EU359045 and EU359046.

RESULTS

Physical and chemical characteristics of ponds.

The chemical characteristics of pond water overlying mats in the various pond systems in Antarctica are shown in Table 1. Sediments underlying the mats from Bratina Island were black and produce a sulfidogenic odor. In contrast, sediments of the ponds in Wright and Victoria valleys ranged from coarse gravel to sand. The partial chemical analysis of pond water revealed elevated salt levels within some ponds from Bratina Island and Wright and Victoria valleys. However, the salt profiles for the Bratina Island ponds differed from those of ponds in the Wright and Victoria valleys in that sulfate was a major salt constituent, while in the latter, salt mainly comprised of NaCl. The pH of the pond water trended toward being alkaline with the exception of the surface layers of ponds E4 and Ridge (Wright Valley).

TABLE 1.

Chemical analysis of pond water from ponds on Bratina Island and in Wright, Victoria, and Miers valleys in Antarctica^a

Geographical location and pond or study siteb or sample	Concn (g m−3)c				pHc
	Cl−	SO42−	Ca2+	Na+
Bratina Island
Bambi	550	180	30	310	9.6
Brack	1,300	8,200	100	3,800	10.1
Caston	270	310	11	300	9.7
Heart	840	130	11	480	8.8
Moist	ND	ND	ND	ND	ND
No name	430	130	30	210	9.3
P70	1400	230	24	720	9.7
Pancreas	1100	220	29	700	8.9
Retro	550	390	46	340	9.6
Salt	5,100	33,000	200	14,000	9.7
Skua	120	140	8.9	140	9.7
Vent	1200	220	30	720	9.1
Weather station	1400	230	24	740	9.7
Victoria Valley
Basalt	5,300	960	1,100	1,800	8.5
River gauge	4.5	3.2	3.1	1.8	8.1
TP lower	8,800	1,100	1,900	2,800	8.9
TP upper	ND	370	870	1,200	8.3
Upper Victoria	2000	ND	ND	ND	ND
Wright Valley
E3	5.3	12	1.6	6.7	7.9
E4	46	120	33	45	8.6
L01	110	160	39	81	9.3
L15	49	70	23	35	9.4
L16	380	140	28	190	9.2
L26	2,800	580	91	1,300	8.5
L3	8.2	18	6.2	7.4	9.3
L4	11,000	2,000	520	6,000	8.3
L9	750	280	40	460	9
Puddle	430	210	43	250	9.5
Ridge	31	37	6.1	41	7.1
Miers Valley
Lake Miers sites
LMM1	3.9	3	ND	ND	ND
LMM2	3.9	3	ND	ND	ND
LMM3	3.9	3	ND	ND	ND
LMM4	3.9	3	ND	ND	ND
LMM5	2.0	3.9	ND	ND	ND

^aNo data is available for the Marshall Valley, Adams and Miers glacier, and Lake Purgatory sites.

^bUnofficial names of study ponds or study sites.

^cND, not determined.

ARISA.

Analysis of ARISA data for all samples identified a total of 63 distinct ARISA fragment lengths (ALF; i.e., peaks). When AFL were totalled across each of the five locations, the highest diversity was observed in the Bratina Island ponds (Σ = 28, x̄ = 6.2), followed by Wright Valley (Σ = 27, x̄ = 4), Miers Valley (Σ = 26, x̄ = 7.1), Victoria Valley (Σ = 18, x̄ = 5.6), and Marshall Valley (Σ = 9, x̄ = 3).

Multivariate analyses showed that cyanobacterial community structure differed among sampling locations (ANOSIM R = 0.4, P < 0.001). Pair-wise comparisons between each sampling location revealed that Bratina Island and Marshall Valley samples were all significantly different from the samples from other locations, whereas the samples from the Victoria, Miers, and Wright valleys did not vary markedly (Table 2). With the exception of MarV1, MarV3, and Ridge, the two-dimensional MDS ordination analysis separated the samples into two large groups, united at the similarity level of 40%. Within each of these group, samples from the same geographic location tended to clustered together (Fig. 1). One exception to this was the samples from Marshall Valley were all distant from each other.

TABLE 2.

ANOSIM statistics for tests involving a comparison of samples from all five sampling locations

Comparison of samples from different sampling locations	R statistic	P value
Bratina Island and Victoria Valley	0.689	0.001
Bratina Island and Wright Valley	0.508	0.001
Bratina Island and Miers Valley	0.401	0.002
Bratina Island and Marshall Valley	0.788	0.002
Victoria and Wright valleys	−0.106	0.800
Victoria and Miers valleys	0.190	0.066
Victoria and Marshall valleys	0.569	0.018
Wright and Miers valleys	0.177	0.027
Wright and Marshall valleys	0.387	0.011
Miers and Marshall valleys	0.504	0.012

FIG. 1.

Two-dimensional nonmetric multidimensional scaling ordination (stress value of 0.12) based on Bray-Curtis similarities of ARISA fingerprints of cyanobacterial communities from various locations in Eastern Antarctica. Points within a circle cluster at 40% similarity. Symbols: ▵, Bratina Island; •, Marshall Valley; ♦, Miers Valley; □ Wright Valley; ▾, Victoria Valley. Weather Stn, Weather station.

To elucidate potential water chemistry parameters responsible for differences in community structure among the geographic locations, analysis with the BEST computer program was undertaken. Initial pairwise scatter-plots between all combinations of the water chemistry variables suggested that a log(1 + X) transformation of all variables was required (8). The results from the BEST analysis showed that the highest rank correlation (ρ = 0.158, P < 0.008) was due to a combination of pH, Ca, and Na. This value is low in comparison with other examples (9), indicating that this set of environmental variables has weak explanatory power.

MC detection.

Data for microcystins expressed on a μg·kg⁻¹ (dry weight) basis are shown in Tables 3 and 4. MCs were detected by at least one of the detection methods in all samples. With the exception of the higher levels of MC observed for the Adams and Miers glacier samples (MVAG1 and MVMG1), no clear differences could be seen between MC concentrations that could be attributed to geographical location. However, when indicative potencies of MCs (PP-2A inhibition assay/ADDA-ELISA ratio; 27) were plotted against total MCs (ADDA-ELISA; Fig. 2), two trends became evident. (i) Generally, potency decreased as total MC levels increased (this was particularly evident for samples from Miers and Marshall valleys). (ii) Values for sites for discrete locations tended to cluster (this was particularly evident for samples from Bratina Island and Victoria Valley).

TABLE 3.

Determination of microcystins in cyanobacterial mat samples taken at various sites in the vicinity of Bratina Island and in the Wright, Victoria, Miers, and Marshall valley regions

Geographical location, collection date, and pond or study sitea or sample	Total microcystin concnb (μg·kg−1 [dry wt]) by the following method:
	PP-2A inhibition assay	ADDA-ELISAc	ELISAc	LC-MSc
Bratina Island, December 2004
Bambi	12.4	26.1	—	—
Brack	8.8	74.6	—	ND
Casten	1.1	21.2	—	—
Heart	4.5	19.6	—	—
Moist	13.1	12.0	—	—
No name	18.7	31.6	—	ND
P70	5.7	35.6	—	—
Pancreas	18.4	13.7	—	ND
Retro	11.7	15.8	—	—
Salt	30.5	132	—	—
Skua	11.0	38.2	—	—
Vent	9.2	25.8	—	ND
Weather station	2.0	8.0	—	ND
Victoria Valley, December 2004
Basalt	2.5	2.4	—	—
River gauge	1.7	1.0	—	—
TP lower	1.1	1.8	—	—
TP upper	2.6	1.9	—	—
Upper Victoria	11.9	5.9	—	—
Wright Valley, December 2004
E3	26.0	8.6	—	—
E4	5.1	2.8	—	—
L01	13.4	7.9	—	—
LI5	10.8	2.4	—	—
L16	12.3	9.5	—	—
L26	57.7	27.3	—	—
L3	57.6	19.4	—	—
L4	11.2	0.2	—	—
L9	11.7	40.0	—	—
Puddle	0.4	1.3	—	—
Ridge	1.7	0.0	—	—
Miers Valley, December 2006
LMM1	697	360	223	143
LMM2	92.1	8.2	4.3	3.1
LMM3	254	2.1	2.1	2.9
LMM4	116	0.7	0.7	<1
LMM5	408	89.5	64.2	40.1
MVAG1	710	2,960	1,550	1,153
MVMG1	1,510	15,900	7,490	6,609
LPM1	1,020	8.3	8.3	<1
Marshall Valley, December 2006
MarV1	374	187	158	46.6
MarV2	98.3	0.9	0.9	<1
MarV3	2,610	1,070	812	515

^aUnofficial names of study ponds or sites.

^b—, not analyzed; ND, not detected.

^cADDA-containing microcystins.

TABLE 4.

Concentrations of ADDA-containing microcystin congeners in microbial mat samples from Miers and Marshall valleys determined by LC-MS using MRM

MC congener	Concn (μg·kg−1 [dry wt]) of ADDA-containing MC congener in sample:
	LMM1	LMM2	LMM3	LMM4	LMM5	MVAG1	MVMG1	LPM1	MarV1	MarV2	MarV3
Desdimethyl MC-RR (MC-2)	36.4	0.5	<1	<1	9.3	216	568	<1	1.8	<1	89.4
Desmethyl MC-RR (MC-4)	45.0	0.8	<1	<1	8.6	344	1226	<1	2.9	<1	128
MC-RR	<1	<1	<1	<1	<1	10.8	29.7	<1	1.3	<1	4.7
Desdimethyl MC-LR (MC-1)	11.5	<1	<1	<1	5.5	137	1253	<1	6.9	<1	76.2
Desmethyl MC-LR (MC-3)	40.0	1.8	2.9	<1	14.8	371	3226	<1	24.5	<1	200
MC-LR	9.3	<1	<1	<1	1.5	38.6	219	<1	9.2	<1	17.0
MC-FR	<1	<1	<1	<1	1.5	35.9	87.7	<1	<1	<1	<1
Total ADDA-containing MCs	143	3.1	2.9	<1	40.1	1,153	6,609	<1	46.6	<1	515

FIG. 2.

Plot of the PP-2A inhibition assay/ADDA-ELISA ratio versus ADDA-ELISA concentrations for microcystins in cyanobacterial mats taken from different sites in a range of geographical locations in Eastern Antarctica. Symbols: ▵, Bratina Island; •, Marshall Valley; ♦, Miers Valley; +, samples from the base of the glacier in Miers Valley; □, Wright Valley; ▴, Victoria Valley.

LC-MS analysis of selected samples from Bratina Island did not show the presence of MCs. Analysis of samples from the Miers and Marshall valleys by LC-MS identified seven MC congeners. LC-MS peaks for parent MC-RR and MC-LR congeners were present, although the retention times did not exactly match the standards. The most prevalent congeners observed using MRM were a desmethyl MC-LR, a didesmethyl MC-LR, a desmethyl MC-RR, and a desdimethyl MC-RR. More detailed analysis of samples MVMG1 and MVAG1 using parent ion scanning for components yielding the m/z 135 ADDA fragment did not reveal other MC congeners. However, the relative retention time data for the MC-RR and MC-LR congeners found were slightly different from those for the parents or known demethylated analogues of these MCs. Furthermore, full-scan LC-MS revealed the presence of a further four components with MH⁺ 1038 and 1052 (retention region for MC-RR; strong [MH₂]²⁺) and MH⁺ 995 and 1009 (retention region for MC-LR) which did not yield significant m/z 135 on collisional activation. Full daughter ion spectra were obtained for these four components and the four major MCs in the extracts of samples MVAG1 and MVMG1. The spectra were consistent. These samples contained eight novel variants of [d-Asp³] MC-RR and [d-Asp³] MC-LR containing glycine [Gly¹] rather than alanine and combinations of homoarginine [hAr²] or acetyldemethyl ADDA [ADMAdda⁵] substitutions (Fig. 3). MS alone cannot distinguish between the isobaric N-methyl dehydroanaline and dehydrobutyrine, which is another potential substitution (38). The details of these structural assignments will be published elsewhere but are consistent with the analyses of low-energy collisionally activated ion spectra for similar MC variants identified from Nostoc species and Planktothrix agardhii (25, 32, 54). The [Gly¹] substitution is novel but is supported by several peptide fragment ion series (54) and analogous to those observed for [d-Leu¹] MC-LR (34).

FIG. 3.

Structures of microcystins RR and LR and the eight novel variants from Antarctic cyanobacterial mats MVAG1 and MVMG1 (Miers Valley).

The MC content of samples collected in 2006 was determined using four different methods, allowing comparison of results from these methods. There was a strong correlation between the LC-MS results and the two ELISA methods (R² = 0.9997 for ADDA-ELISA and R² = 0.999 for ELISA) and between the two ELISA methods (R² = 0.974). However, the correlation was weak when the PP-2A inhibition assay results were compared to all other methods (R² = 0.168 for LC-MS, R² = 0.160 for ADDA-ELISA, and R² = 0.187 for ELISA).

Confirmation of a MC producer.

On the basis of morphology, the dominant species in sample MVMG1 was identified as a Nostoc sp. with the following features: long and irregularly trichomes surrounded by a diffuse mucilaginous envelope; vegetative cells subspherical, 4 ± 2 μm wide and 2.8 ± 1.2 μm long; heterocytes 5 ± 1 μm wide and 6.4 ± 1.4 μm long. Segments of the 16S rRNA and the mcyE gene were successfully amplified from the purified Nostoc sp. material. The 685-bp 16S rRNA gene sequence (GenBank accession no. EU359045) was submitted to BlastN (2) and matched at greater than 99% sequence homology to Nostoc sp. strain ANT.LH52B.8 (GenBank accession no. AY493593). The 364-bp segment of the mcyE gene (GenBank accession no. EU359046) had high (93%) sequence homology with Nostoc sp. strain 152 (GenBank accession no. AY817163). ARISA analysis from the purified Nostoc sp. material identified two distinct AFL at lengths of 471 and 733 bp. At least one of these peaks was identified in ARISA profiles from samples LMM1, LMM2, MVAG1, MVMG1 (Miers Valley), and MarVM3 (Marshall Valley).

DISCUSSION

MC production.

In this study, we have demonstrated that microcystin production by cyanobacteria in Antarctica is not confined to the meltwater ponds on McMurdo Ice Shelf (16, 18). Using a combination of ELISAs, PP-2A inhibition assay, and LC-MS, MCs were detected in cyanobacterial mats from four distant geographic locations within the Dry Valleys of Eastern Antarctica. Previous studies (16, 18) detected only low levels of MCs, and this was also the case in our study where total MC levels by ELISA were all <16 mg·kg⁻¹ (dry weight). These levels are significantly lower than those reported from planktonic cyanobacterial blooms (7). Interestingly, the highest concentrations of MCs were not detected in the samples from the lake and pond mats but from the hydroterrestrial mats adjacent to the bases of Miers and Adams Glaciers. This is the first report of MCs from hydroterrestrial mats in Antarctica. Jungblut et al. (18) postulated that the low levels of MCs could be due to either the low abundance of the MC producer within the mat community or low levels of biosynthesis by the producer. Nostoc sp. made up a large portion of the biomass of sample MVMG1 (sample with highest MC concentrations); thus, we suggest that the former is the most plausible explanation.

Previous studies have reported only two MC congeners, [d-Asp³] MC-LR and MC-LR in Antarctic cyanobacterial mat samples (16, 18). The quantitative measurements by LC-MS using MRM initially indicated the presence of MC-LR, MC-LR, and MC-FR with higher proportions of four additional congeners, a desmethyl MC-RR, a desdimethyl MC-RR, a desmethyl MC-LR, and a desdimethyl MC-RR (Table 4). However, full-scan LC-MS followed by detailed MS-MS daughter ion analysis revealed that there were eight major MC components which had novel structures based on variants of MC-LR and MC-RR and these included four ADMAdda variants. The latter were not detected by the MRM or parent ion scan experiments because the ADDA fragment at m/z 135 is not significant when the 9-acetoxy substitution is present (Fig. 3) (54). ADMAdda MC analogues have reported in benthic Nostoc strains from Finland (32, 40, 41) and Planktothrix agardhii (planktonic) from Denmark (25). This is the first reporting of such variants from Southern Hemisphere cyanobacteria. Substitution of hAr for Arg was also relatively common in MCs from a Finnish Nostoc sp. The [d-Ala¹] in MCs is generally highly conserved, although substitutions by serine or leucine in MC-LR have been reported (34, 41). Therefore, the novel finding in these Antarctic mats of eight variants of MC-RR and MC-LR all containing [Gly¹] is remarkable.

Hitzfeld et al. (16) and Jungblut et al. (18) hypothesized on potential MC-producing genera; Oscillatoria, Phormidium, and Nostoc were all given as likely candidates. The ability of a Nostoc sp. (in sample MVMG1) to produce MCs was confirmed by PCR amplification and sequencing of a region of the aminotransferase domain of the mcyE gene. Nostoc species have previously been shown to produce MCs (40, 41, 52). Wood et al. (52) detected high levels of MC-RR and a desmethyl MC-RR in benthic Nostoc commune mats collected from a New Zealand lake. The diagnostic ARISA fragment lengths for this species were observed in only five samples from the Miers and Marshall valley regions. Interestingly, four of these samples had the highest total MC concentrations recorded in this study. The absence of the Nostoc sp. AFL from other samples indicates that there are other yet to be identified MC producers within these mat communities.

Several studies have demonstrated variability in MC concentrations when different detection methods were used (6, 28, 52). The ADDA-ELISAs used during the present study measure the total amount of ADDA-containing compounds in the sample, with the second ELISA having lower cross-reactivity with free ADDA and nodularin (Lyn Briggs, personal communication). As both ELISAs used antibodies raised against the ADDA moiety, it is very likely that cross-reactivity to ADMAdda variants was low (25) and that therefore the total MCs in these samples were underestimated. Similarly, the LC-MS (MRM) analyses targeting 13 common ADDA-containing MCs did not determine the ADMAdda variants. This explains the high correlations between results for both ELISAs and the LC-MS (MRM) method. Based on the scanning LC-MS data for samples MVAG1 and MVMG1, it is estimated that including the four major ADMAdda variants would approximately double the total MC concentrations reported in Table 4. The correlations were weak (< R² = 0.19) between the PP-2A inhibition assays and the ELISA or the LC-MS methods, with concentrations by PP-2A inhibition assays being consistently lower for samples containing >20 μg·kg⁻¹ MCs (Table 3). The response of the PP-2A inhibition assay varies depending on the toxicity of MC congeners present in a sample (28), which will explain some of the inconsistencies observed when comparing results obtained via these methods.

Mountfort et al. (28) suggested that for samples containing mixtures of MC congeners, the response ratios (ratio of the amount determined by PP-2A inhibition assay equivalent to MC-LR to the amount determined by ADDA-ELISA) assigns an indicative toxicity to a sample as well as toxin equivalence. In our study when indicative potencies of MCs were plotted against total MCs (as measured by ADDA-ELISA; Fig. 2), samples from discrete locations tended to cluster together. MCs were not detected by LC-MS for the samples from Bratina Island or Victoria and Wright valleys, but it was presumed that the MC congener composition for samples from the same geographic location were similar. The two samples with the lowest PP-2A inhibition assay/ELISA ratio (MVMG1 and MVAG1 from Miers and Marshall valleys) were the samples with the highest total MC levels, and therefore, it is likely that the novel MC-LR and MC-RR variants identified in these samples by LC-MS have significantly lower PP-2A inhibition assay activities than MC-LR [Asp³] variants do. MC-LR [Asp³] variants have been reported to have lower toxicity when administered intraperitoneally and while ADDMAdda variants were toxic when given intraperitoneally (37, 42), somewhat lower PP-2A inhibition assay activities have been reported (25). The effects on toxicity of [Gly¹] or [hAr⁴] substitution have not been determined. None or only low levels of the target 13 ADDA-MCs were detected by LC-MS in the three Miers and Marshall valley samples with the highest PP-2A inhibition assay/ADDA-ELISA ratio (LMM3, LMM4, and MarV2). This potentially indicates the presence of other toxins with high inhibitory potential for PP-2A.

ARISA and MC production.

Morphological surveys (4, 5) and more recently polyphasic approaches using 16S rRNA clone libraries (18, 45, 46, 47) have helped establish an inventory of Antarctic cyanobacteria and allowed investigations into endemism and biogeographical distributions. However, these identification methods are often protracted and therefore not applicable for analysis of large sample numbers. Recently, a sensitive and high-throughput fingerprinting method known as automated rRNA intergenic spacer analysis has been developed (12). This PCR-based method (ARISA) exploits the length heterogeneity of the intergenic spacer region between the 16S and 23S ribosomal genes. Total community DNA is amplified with a fluorescently labeled oligonucleotide, allowing the electrophoretic step to be performed with an automated system in which a laser detects the fluorescent DNA fragments. In this study, ARISA was used to assess cyanobacterial community structure in 40 samples from five distant locations. This enabled us to investigate the following: (i) the influence of community structure on MC production, (ii) biogeographical distribution, and (iii) the effect of selected water chemistry parameters on cyanobacterial community structure.

The nonmetric MDS analysis of the ARISA data showed that community structure appears to have little effect on MC concentration. MVAG1 and MVMG1 plotted close to each other; however, MarV3 which also had a high concentration of MCs, was distant. We postulate that it is the presence and abundance of one or more toxin-producing genotypes, not community structure, that influences the amount of MC in a sample. Numerous studies have shown that the presence of MC genes (i.e., toxic genotypes) correlates with detection of MCs (15, 49).

The study of MC production in extreme environments may help in understanding their functional role. Within these mats, especially the hydroterrestrial mats, there are minimal grazers and few other phytoplankton (16); thus, MC production to prevent grazing or allelopathy seem unlikely. The low levels of MCs found in Antarctic mats so far suggest that minimal biosynthesis is occurring. Although no studies have investigated Antarctic cyanobacterial growth in the field, it is likely that given the extreme cold and dark conditions for many months of the year, growth is minimal. Orr and Jones (33) in a study on cultured Microcystis aeruginosa showed that MC production was limited to the growth phase when the cell concentration was increasing and suggest that MC plays an important (perhaps essential) role in the cellular metabolism of toxigenic strains. If this hypothesis is correct and given the presumably slow growth rate of cyanobacteria in Antarctica, these two factors may explain the low MC levels. MCs are extremely stable and resistant to chemical hydrolysis or oxidation at near neutral pH (42). In the inherently cold and often dark Antarctic environment, it seems likely that these toxins may persist for many months or years. Investigations on Antarctic isolates and on MC gene expression during different phases of the year are planned to further explore this.

ARISA and community structure.

Taton et al. (47) carried out a detailed analysis of cyanobacterial diversity in samples from four different ponds on Bratina Island and identified between 4 and 12 operational taxonomic units (OTUs; based on 16S rRNA gene sequences) per pond. In an analogous study, Jungblut et al. (18) identified 5 to 15 OTUs from three ponds on Bratina Island. A similar diversity was observed in our samples with the number of AFL ranging from 1 to 14. One caveat when interpreting ARISA data is that interoperonic differences in spacer length occur within the genomes of microorganisms (30), such that a single species may contribute more than one peak to an ARISA profile. Previous studies (13, 53) indicate that members of the order Nostocales commonly have two types of intergenic spacer regions (i.e., two AFL), whereas members of the orders Chroococcales and Oscillatoriales have only one. Thus, it is highly likely that the number of AFL is greater than the actual number of OTUs.

Morphological studies (5, 21, 36) suggest that many cyanobacterial species are widespread across the continent of Antarctica. In contrast, recent molecular studies have shown that the communities of four lake mats were distinct, with 71.4% of OTUs found only in one sample (47). This may however be due to artifacts (e.g., produced during DNA extraction, PCR, and cloning) (47) or may reflect the small number of samples used in this study. The results of MDS analyses of our ARISA profiles suggest that cyanobacterial community structure within a geographic location generally does not vary markedly, with most samples showing greater than 40% similarity. Interestingly, there were two clusters on the MDS plot, one containing mainly the Wright and Victoria valley samples and the other primarily consisting of the Bratina Island samples. This result was also shown in the ANOSIM analysis where the Bratina Island samples were significantly different from the samples from Wright and Victoria valleys. Wright and Victoria valleys are adjacent valleys located approximately 150 km north of Miers Valley and Bratina Island. It seems plausible that the close proximity of these valleys has enabled similar cyanobacterial communities to develop. It has been suggested that wind is an important dispersal agent for biomass in Antarctica (3, 35), and this may have also played a role in the similar structures of these communities. The Marshall Valley samples did not cluster close to one another and were usually quite distant from the other samples. These samples were hydroterrestrial mats. Unfortunately, no physicochemical data was collected for these sites, which may have helped in investigating explanations for the community composition differences.

Jungblut et al. (18) suggest that salinity may influence community structure. The ponds and lakes in our study spanned a wide range of salinities. The results of the BEST analysis indicated that differences in the water chemistry parameters (Cl⁻, SO₄²⁻, Ca²⁺, and Na⁺ concentrations and pH) were unlikely to be contributing to the community structure. Thus, rather than physiochemical parameters dictating which cyanobacteria species are present in these ponds, we suggest that some species have adapted to tolerate a wide range of conditions. Taton et al. (47) reached a similar conclusion. Following a comparison of their cyanobacterial OTUs with sequence databases, they suggest that given the ubiquity of several OTUs, cyanobacteria must have the ability to tolerate a range of harsh environmental conditions.