Microcystin in Missouri reservoirs

Abstract

During summers (May–Aug) 2004–2006, 177 Missouri reservoirs were sampled monthly at open pelagic locations to assess regional patterns in microcystin concentration, frequency of occurrence over successive summer seasons and relations with environmental factors. Microcystin was detected in 58% of Missouri reservoirs and 23% of samples (n = 1402). Total microcystin concentrations, measured by enzyme-linked immunosorbent assay, ranged from ≤ 0.1 to 21 μg/L. Concentrations ≥ 1 μg/L were detected in 10% of reservoirs and exceeded the human health concern limit of 20 μg/L once in a single sample. Microcystin occurred throughout summer, with maximum concentrations in individual reservoirs observed in each month. Occurrence was consistent across years, with about one-half of Missouri reservoirs having detectable microcystin each summer. Eleven reservoirs with microcystin maxima ≥1 μg/L were sampled multiple seasons; of these, 8 had detectable microcystin each summer, which indicates that short-term surveys can identify water bodies with the greatest potential for toxin production. Eutrophic reservoirs in northern Missouri had the greatest microcystin occurrence and concentrations. Reservoirs with detectable microcystin had significantly (p < 0.01) greater nutrient and chlorophyll values and significantly shallower Secchi depths than reservoirs without detection. All correlations, however, had r-values ≤ 0.35, and bivariate plots indicated nonlinear trends. Cylindrospermopsin was measured by enzyme-linked immunosorbent assay in 36 reservoirs once in late summer 2006; 14% had small detectable levels (total concentrations < 1 μg/L). This is the first report of cylindrospermopsin in Missouri.

Key words:

• cyanobacteria
• cylindrospermopsin
• microcystin
• Missouri reservoirs
• toxin

Microcystins (MC) are a class of cyanobacterial hepatotoxins common in lakes and reservoirs worldwide (Chorus and Bartram 1999, Huisman et al. 2005). Regional studies indicate MC occurs frequently in the Midwestern United States and may reach concentrations of human health concern (McDermott et al. 1995, Dodds 1996, Graham et al. 2004, Hedman et al. 2008, Lindon and Heiskary 2008). Whereas several studies in the United States have assessed regional patterns in MC occurrence (McDermott et al. 1995, Dodds 1996, Zimba and Grimm 2003, Graham et al. 2004, Boyer 2007, Williams et al. 2007, Hedman et al. 2008, Lindon and Heiskary 2008), few have evaluated frequency of occurrence during successive summer seasons. Likewise, few regional studies have developed empirical relations between environmental factors and MC concentration. Knowledge of MC occurrence during summer and the general environmental conditions associated with elevated concentrations will improve understanding of factors favoring toxic blooms.

Most Missouri reservoirs are located in the agricultural Osage and Glacial Plains, the Ozark Highlands, which are covered by forest and pasture, or the ecotonal Ozark Border section (Jones et al. 2008a). Reservoir nutrients are affected by hydraulic flushing and depth but exhibit an increase with cropland and a decrease with forest cover (Jones et al. 2004, 2008b). An initial survey of particulate MC (net collections > 64 μm) in Missouri reservoirs indicated MC occurrence was more common in the eutrophic Osage and Glacial Plains reservoirs than those in the Ozark Highlands (Graham et al. 2004). A subsequent study of MC in the size fractions of natural plankton communities indicated particulate MC underestimates total toxin concentration because of algal loss in net collections (Graham and Jones 2007). This study reports total MC concentrations in whole water samples collected in conjunction with a statewide inventory of 177 Missouri reservoirs during summers 2004–2006 (Jones et al. 2008a). Data on the occurrence and concentration of MC are analyzed to describe regional patterns, detection frequency across three summer seasons and relations with reservoir nutrients and transparency. Also, cylindrospermopsin, another cyanobacterial hepatotoxin (Chorus and Bartram 1999), was measured in 36 reservoirs once in late summer 2006 to gather preliminary data on its occurrence; algal taxa capable of producing this toxin (Anabaena, Aphanizomenon and Cylindrospermopsis) can dominate summer algal assemblages in Missouri reservoirs (Graham et al. 2006a, Jones et al. 2008a). This information can facilitate human health protection in lakes and reservoirs used for recreational activities and drinking-water supplies.

Methods

Sample collection

Missouri reservoirs (n = 177) were sampled May through August 2004–2006: 94 were sampled one summer season; 45 were sampled 2 summers; and 38 were sampled all 3 summers. Most reservoirs were sampled monthly (n = 3 or 4 per summer), with individual reservoirs sampled on 3–26 occasions during this study. The reservoirs encompassed the full range of landscape, morphological and limnological conditions typical of Missouri reservoirs (Table 1; Jones et al. 2004, 2008a, 2008b).

Table 1 Provincial medians (med) and ranges of landscape, morphological, and limnological variables.

	Ozark Highlands			Ozark Border			Osage Plains			Glacial Plains
Variable	n	Med	Range	n	Med	Range	n	Med	Range	n	med	Range
Landscape
Watershed area (Ha)	29	5,309	155–1,544,940	25	502	50–38,632	20	2,611	253–4,631,671	103	677	25–1,503,965
% Urban cover	29	0	0–6	25	0	0–47	20	2	0–77	103	2	0–32
% Forest cover	29	66	28–94	25	40	0–82	20	10	0–31	103	12	0–84
% Crop cover	29	2	0–11	25	10	0–52	20	24	7–59	103	28	0–70
% Grass cover	29	24	1–59	25	34	0–74	20	46	9–73	103	39	1–78
Morphological
Mean depth (m)	29	3.7	1.8–12.3	25	3.4	1.4–9.8	20	3.2	1.0–9.6	103	2.7	1.1–8.5
Reservoir area (Ha)	29	96	6–51,334	25	46	7–379	20	109	14–53,814	103	40	5–18,524
Reservoir volume (Ha m)	29	1,900	80–2,020,000	25	796	60–13,782	20	1,400	57–1,202,700	103	450	10–500,040
Flushing rate (Yrs)	29	1.75	0.23–87	25	0.66	0.22–42	20	1.5	0.43–6.0	103	0.73	0.03–7.0
Limnological
Temperature (°C)	245	26.6	16.3–33.5	127	27.6	16.8–33.5	140	26.3	19.5–32.3	890	26.2	14.2–34.3
Dissolved oxygen (mg/L)	245	6.0	2.7–11.5	127	6.0	1.5–11.6	140	5.3	1.8–9.7	890	6.9	2.1–15.3
Secchi (m)	245	2.1	0.4–7.9	127	1.1	0.2–4.8	140	0.8	0.09–2.5	890	1.0	0.01–7.1
Total nitrogen (mg/L)	245	380	40–1,050	127	640	270–2,230	140	870	380–1,810	890	840	240–3,120
Total phosphorus (μg/L)	245	14	3–67	127	36	5–257	140	49	20–237	890	41	6–346
TN:TP	245	25.6	5.7–180	127	18.7	5.0–78.3	140	17.4	4.8–35.6	890	19.8	4.5–66.2
Chlorophyll (μg/L)	245	6	1–70	127	15	2–265	140	27	4–185	890	18	1–342
Chl:TP	245	0.4	0.08–1.16	127	0.4	0.02–2.0	140	0.5	0.03–1.4	890	0.4	0.01–2.2
Nonvolatile solids (mg/L)	245	1.1	0.07–9.6	127	1.9	0–14.3	140	3.6	0.6–22.0	890	3.0	0–42.8
Volatile solids (mg/L)	245	1.4	0.3–9.8	127	2.6	0.3–21.6	140	4.3	1.0–17.0	890	3.5	0.2–37.6
Total solids (mg/L)	245	2.7	0.6–14.6	127	4.7	0.9–28.8	140	8.4	2.1–38.0	890	7.1	0.3–51.2
Conductivity (μS)	245	168	25–368	127	118	16–489	140	193	67–401	890	186	24–613
Turbidity (NTU)	245	2.1	0.5–19.4	127	3.9	0.8–36.3	140	6.3	1.9–80.2	890	5.2	0.5–126
Dissolved organic carbon (mg/L)	245	3.5	1.0–7.0	127	5.5	3.0–11.9	140	5.4	3.3–7.8	890	6.6	3.9–12.4
Note: For landscape and morphological variables n indicates the number reservoirs; for limnological variables n indicates the number of measurements.

Temperature, dissolved oxygen and Secchi transparency were measured at open pelagic locations in each reservoir, typically near the dam. Water samples were composited from the surface layer (0.25–0.5 m) and analyzed for total MC, conductivity (μS), turbidity (NTU), chlorophyll (Chl; uncorrected for degradation products), total phosphorus (TP), total nitrogen (TN), non-volatile suspended solids (NVSS), volatile suspended solids (VSS), total suspended solids (TSS), and dissolved organic carbon (DOC). Whole-water samples underwent 3 freeze-thaw cycles after Graham and Jones (2007) to lyse cyanobacterial cells and allow determination of total MC. In 2004, Envirogard® enzyme-linked immunosorbent assay (ELISA) kits (detection limit 0.1 μg/L; kit is cross-reactive with microcystin-LR, -YR, and -RR and nodularin) were used to measure MC. During 2005–2006, Abraxis® ELISA kits were used (detection limit 0.1 μg/L; -adda specific). Both kits provided similar results, and among-reservoir variability was greater than among-kit variability (J. Graham, unpublished data). Total cylindrospermopsin (CYL) was measured in 36 Missouri reservoirs once in late summer 2006 using Abraxis® ELISA kits (detection limit 0.04 μg/L). DOC was determined colorimetrically (Technicon Method No. 451–76W and Ontario Ministry of the Environment document JC 7501, 1972). Details of other analytical methods and land-use data are presented elsewhere (Knowlton and Jones 1995, Graham et al. 2004, Graham and Jones 2007, Jones et al. 2008a, 2008b).

Statistical analyses

Reservoir trophic status was determined using geometric means (n = 3–18 per summer), which were averaged across all seasonal means (n = 1–3), and criteria for Missouri reservoirs (Jones et al. 2008a). Summary statistics were calculated using all available data; where MC was not detected, values of 0.05 μg/L (one-half the method detection limit) were substituted (Sokal and Rohlf 1995). All data were grouped into 3 MC categories based on concentration: no detectable MC; MC between 0.1–1 μg/L; and MC > 1 μg/L. Detections of large MC concentrations were too few to allow creation of additional MC categories above 1 μg/L. Significant differences in environmental variables associated with each MC category were determined using one-way analysis of variance (ANOVA) followed by Tukey's pairwise comparisons (Sokal and Rohlf 1995). Relations between total MC concentration and environmental variables were developed using Pearson correlation analysis (Sokal and Rohlf, 1995). All data were log₁₀ transformed (land-use data were logit transformed) for ANOVA and correlation analyses (Sokal and Rohlf 1995). Statistical analyses were performed using SAS® (9.1), and significance was set at p < 0.05 for all analyses.

Bivariate plots of the relations between total MC concentration and environmental variables showed trends that were not linearized through transformation. Microcystin was undetectable across the range of all variables, and the shape of any given relation was defined by the upper limits created by MC maxima. Nonlinear interval-maxima regression (IMR) (Blackburn et al. 1992, Scharf et al. 1998) was used to define the shape of these upper limits after Graham et al. (2004). Each independent variable was divided into equal increments, resulting in 10–16 intervals. The maximum MC concentration and the associated environmental variable value were obtained from each interval and used in nonlinear regression analysis (Blackburn et al. 1992, Scharf et al. 1998). Microcystin detection frequency in each interval was calculated by dividing the number of detections in the interval by the total number of samples (n = 1402). We performed IMR using SigmaPlot® (10.0) and set significance at p < 0.05.

Results

Descriptive limnology

Missouri reservoirs encompass a broad range of trophic conditions (Jones et al. 2008a; Table 1). Most (79%) reservoirs sampled in 2004–2006 were mesotrophic or eutrophic, with 7% oligotrophic and 14% hypereutrophic. Based on TN:TP ratios (4–180), potential phosphorus limitation (TN:TP > 17) or co-limitation (17 < TN:TP > 10) by phosphorus and nitrogen (Forsberg and Ryding 1980) was common (93% of samples). Most (88%) samples had Chl:TP ratios substantially < 1, suggesting bloom conditions were uncommon (White et al. 1985; Table 1).

Regional patterns in limnology matched those described by Jones et al. (2008a), with the greatest nutrient and Chl values and the shallowest Secchi depths occurring in Osage and Glacial Plains reservoirs (Table 1). Median nutrient and Chl values in the Plains were 2–3 times greater than in the Ozark Highlands, and the Ozark Border was intermediate; median Secchi depth in the Ozark Highlands was 2 times greater than observed in other physiographic sections, including the Ozark Border (Table 1).

Microcystin occurrence and concentration

Overall, 58% of Missouri reservoirs (n = 177) and 23% of samples (n = 1402) had detectable MC (Fig. 1 and Fig. 2). Total MC concentrations ranged from 0.1 to 21 μg/L but were usually < 1.0 μg/L (Table 2; Fig. 1 and Fig. 2). Few Missouri reservoirs (10%) had MC concentrations exceeding the World Health Organization (WHO) finished drinking-water guideline of 1 μg/L (Table 2; Fig. 1 and Fig. 2), and the WHO recreational guideline of 20 μg/L (Chorus and Bartram 1999) was exceeded once during our study. Of the 36 reservoirs sampled for CYL during late summer 2006, 14% had low detectable levels (0.12–0.81 μg/L).

Table 2 Regional means, medians, and maxima of total microcystin values. Summary statistics are based on samples collected in each region.

			Total microcystin (μg/L)
Region	Reservoir n	% Detection	n	% Detection	Mean	Median	Maximum
Ozark Highlands	29	52	245	11	< 0.1	< 0.1	1.2
Ozark Border	25	44	127	20	0.15	< 0.1	3.1
Osage Plains	20	80	140	34	0.19	< 0.1	4.9
Glacial Plains	103	58	890	25	0.32	< 0.1	21
Note: Reservoir n indicates the number of reservoirs in each region. n indicates the number of samples collected in each region.

Figure 1 Physiographic location of reservoirs sampled in Missouri and total microcystin occurrence and concentration.

Figure 2 Occurrence and concentration of total microcystin in Missouri reservoirs. The World Health Organization (WHO) microcystin guidelines for finished drinking water and full-body contact recreation are 1 and 20 μg/L, respectively. (A) Occurrence and concentration among samples (n = 1402), (B) occurrence and concentration among reservoirs (n = 177).

Occurrence and concentration of MC were greatest in eutrophic Osage and Glacial Plains reservoirs (Table 2; Fig. 1). It occurred most frequently in Osage Plains reservoirs (80% of reservoirs, 34% of samples), although the greatest concentrations were collected from Glacial Plains reservoirs (mean = 0.32 μg/L, max = 21 μg/L; Table 2). Despite general regional patterns, MC values > 1 μg/L were observed throughout Missouri, with the exception of the southern Ozark Highlands (Table 2; Fig. 1).

Microcystin occurred in Missouri reservoirs throughout summer with mean, median and maximum concentrations being similar across summer months during the 3 summer seasons sampled for this study; however, MC was detected about twice as often in June through August compared to May (Table 3). Occurrence and concentration of MC were similar across years. Between 42 and 52% of reservoirs and 20–26% of samples had detectable MC each summer (Table 4). Mean and median MC concentrations were similar among summers, though 2006 maxima were an order of magnitude less than the others (Table 4). Of the 83 reservoirs sampled during 2 or 3 summers, 60% had detectable MC in more than one summer. Within this group, 11 reservoirs had MC maxima ≥ 1 μg/L and 8 (73%) had detectable MC each summer of sample collection, indicating frequent MC occurrence in some reservoirs.

Table 3 Monthly means, medians, and maxima of total microcystin values. Summary statistics are based on samples collected during each month.

			Total microcystin (μg/L)
Month	Reservoir n	% Detection	n	% Detection	Mean	Median	Maximum
May	153	18	281	12	0.14	< 0.1	13
June	176	40	464	24	0.33	< 0.1	21
July	177	35	464	20	0.28	< 0.1	11
August	148	35	381	22	0.18	< 0.1	4.6
Note: Reservoir n indicates the number of reservoirs sampled each month. n indicates the number of samples collected during each month.

Table 4 Annual means, medians, and maxima of total microcystin values. Summary statistics are based on samples collected during each year.

			Total microcystin (μg/L)
Year	Reservoir n	% Detection	n	% Detection	Mean	Median	Maximum
2004	76	42	514	23	0.34	< 0.1	21
2005	95	52	380	26	0.25	< 0.1	11
2006	127	46	508	20	0.16	< 0.1	4.9
Note: Reservoir n indicates the number of reservoirs sampled each year. n indicates the number of samples collected during each year.

Relation with environmental variables

We detected MC in reservoirs of all trophic status but occurrence and concentration increased across the trophic gradient: 33% of oligotrophic (n = 12), 45% of mesotrophic (n = 40), 60% of eutrophic (n = 100), and 80% (n = 25) of hypereutrophic reservoirs had detectable MC. Maximum concentrations in eutrophic and hypereutrophic reservoirs were 2 orders of magnitude greater than in oligotrophic reservoirs and one order of magnitude greater than in mesotrophic reservoirs (Table 5).

Table 5 Trophic class means, medians, and maxima of total microcystin values. Summary statistics are based on samples collected in each trophic class.

			Total microcystin (μg/L)
Trophic class	Reservoir n	% Detection	n	% Detection	Mean	Median	Maximum
Oligotrophic	12	33	94	5	< 0.1	< 0.1	0.33
Mesotrophic	40	45	374	10	< 0.1	< 0.1	3.1
Eutrophic	100	60	776	27	0.30	< 0.1	21
Hypereutrophic	25	80	158	42	0.51	< 0.1	10
Note: Reservoir n indicates the number of reservoirs sampled in each trophic class. n indicates the number of samples collected in each trophic class.

As indicated by occurrence and concentration across the trophic gradient (Table 5), nutrient status, Chl content and transparency differed among the 3 MC concentration categories (Table 6). Mean nutrient and Chl concentrations were greatest in the MC > 1 μg/L category and smallest in the nondetect category; the MC 0.1–1 μg/L category was intermediate. Total N and TP significantly increased by ∼ 2-fold across the MC categories, and Chl by ∼ 4-fold (Table 6). Secchi transparency showed the opposite pattern. Mean values decreased significantly by ∼ 2-fold across the MC categories (Table 6). Mean TN:TP ratios were significantly greater in the nondetect and MC 0.1–1 μg/L categories (19–20) than the > 1 μg/L category (15), although the difference is not as marked as observed for other variables (Table 6). Several landscape and morphological variables also differed among the 3 MC concentration categories, likely reflecting effects on reservoir nutrient values (Jones et al. 2004, 2008b).

Table 6 Comparison of environmental variables among three microcystin concentration categories: no detectable microcystin (nd), total microcystin concentrations between 0.1–1 μg/L (0.1–1), and total microcystin concentrations greater than 1 (> 1). Significant differences between means were determined using one-way ANOVA and Tukey's pairwise comparisons.

Variable	n	Mean	Range	F-value	p-value
Landscape
% Forest cover
nd	1,071	22 ^a	0–94	24.13	< 0.01
0.1–1	270	12 ^b	0–92
> 1	49	8 ^b	1–56
% Crop cover
nd	1,071	10 ^a	0–78	35.27	< 0.01
0.1–1	270	23 ^b	0–78
> 1	49	28 ^b	17–78
% Grass cover
nd	1,071	35 ^a	1–78	7.26	< 0.01
0.1–1	270	39 ^b	2–78
> 1	49	44 ^b	17–78
Morphological
Mean depth (m)
nd	1,070	3.4 ^a	1.1–12.3	5.73	< 0.01
0.1–1	266	3.2 ^a	1.1–12.3
> 1	46	2.6 ^b	1.6–4.2
Limnological
Secchi (m)
nd	1,080	1.2 ^a	0.01–7.9	26.93	< 0.01
0.1–1	271	0.9 ^b	0.2–6.8
> 1	49	0.7 ^c	0.2–3.5
Total nitrogen (μg/L)
nd	1,082	646 ^a	40–3070	77.05	< 0.01
0.1–1	271	891 ^b	140–3120
> 1	49	1170 ^c	450–2230
Total phosphorus (μg/L)
nd	1,082	32 ^a	3–251	51.33	< 0.01
0.1–1	271	47 ^b	6–346
> 1	49	76 ^c	14–237
TN:TP
nd	1,082	20 ^a	4.5–180	9.48	< 0.01
0.1–1	271	19 ^a	5.5–67
> 1	49	15 ^b	5.6–44
Chlorophyll (μg/L)
nd	1,082	11 ^a	1–342	94.34	< 0.01
0.1–1	271	26 ^b	1–306
> 1	49	46 ^c	3–140
Chlorophyll: Total phosphorus
nd	1,082	0.4 ^a	0.01–2.2	70.99	< 0.01
0.1–1	271	0.6 ^b	0.06–1.6
> 1	49	0.6 ^b	0.2–1.4
Volatile suspended solids (mg/L)
nd	1,073	2.6 ^a	0.2–25	84.86	< 0.01
0.1–1	267	4.4 ^b	0.2–38
> 1	49	7.6 ^c	0.6–19
Total suspended solids (mg/L)
nd	1,072	5.4 ^a	0.3–51	37.35	< 0.01
0.1–1	267	7.2 ^b	0.7–51
> 1	49	11.5 ^c	1.0–39
Turbidity (NTU)
nd	1081	4.0 ^a	0.5–126	56.53	< 0.01
0.1–1	271	5.7 ^b	0.7–44
> 1	49	11.5 ^c	1–48
Dissolved organic carbon (mg/L)
nd	1082	5.4 ^a	1.0–11.9	44.40	< 0.01
0.1–1	271	6.4 ^b	1.4–12.4
> 1	49	7.6 ^c	3.8–11.4

Despite increases in occurrence and concentration with elevated nutrients and reduced transparency (Table 5 and 6), correlations between MC and trophic state variables were weak (TN: r = 0.31, TP: r = 0.26, Secchi: r = −0.19; all n = 1402 and all p < 0.01). Correlations of MC concentration with other measured variables in Table 1 also were weak (all r ≤ 0.35, data not shown). Similar to Graham et al. (2004), MC-TN and MC-TP IMR maxima in Missouri reservoirs were characterized by unimodal curves (r² = 0.87 and 0.75, respectively; both p < 0.01), and MC-Secchi IMR maxima were characterized by exponential decline (r² = 0.93, p < 0.01; Fig. 3). Frequency distributions of MC occurrence along these environmental gradients showed similar patterns. Along the TN gradient the greatest MC values (> 10 μg/L) occurred between 600 and 1900 μg/L; for TP, MC peaked between 30 and 170 μg/L, ranges indicative of eu-hypereutrophic conditions in Missouri reservoirs (Jones et al. 2008a; Fig. 3). Microcystin > 10 μg/L occurred at Secchi depths < 2 m, encompassing meso-hypereutrophic conditions (Fig. 3). The most frequent detection of MC occurred within narrower environmental ranges. We detected MC most frequently at TN between 600 and 1,200 μg/L (eutrophic conditions), TP between 15 and 70 μg/L (meso-eutrophic conditions) and Secchi depths between 0.5 and 1.0 m (eutrophic conditions), a result likely due in part to the large number of samples collected within these ranges (Fig. 3).

Figure 3 Total microcystin (MC), total nitrogen (TN), total phosphorus (TP) and Secchi bivariate relations. Data shown as grouped into intervals for interval-maxima regression (IMR). Curves were estimated using IMR. Black points indicate the data used for IMR analysis (n = 8–16) and gray points indicate all other data. Bar graphs represent frequency distributions of MC detection within each interval used for IMR. Hash marks on the x-axes indicate oligotrophic/mesotrophic, mesotrophic/eutrophic and eutrophic/hypereutrophic cutpoints for Missouri reservoirs as defined by Jones et al. 2008a. (A) MC–TN relation (r² = 0.87, p < 0.01, values for line fitted to black points only), (B) MC–TP relation (r² = 0.75, p < 0.01), (C) MC–Secchi relation (r² = 0.93, p < 0.01). All n = 1402.

Discussion

Cyanobacteria frequently dominate summer algal biomass in Missouri reservoirs, and Anabaena is the most commonly occurring potential MC producer (Jones et al. 2008a). Microcystin was common in Missouri reservoirs, but few concentrations were large enough to cause human health concerns (Fig. 2). Occurrence and concentrations were similar across years, with approximately 50% of reservoirs sampled having detectable MC each summer (Table 4). Widespread occurrence with relatively few large values is characteristic of regional MC studies worldwide (Chorus and Bartram 1999, Graham et al. 2004, Boyer 2007, Williams et al. 2007, Hedman et al. 2008), a consequence of extreme temporal and spatial variability in MC concentrations (Lindholm 1991, Kotak et al. 1995, Welker et al. 2003). Among reservoirs with MC maxima ≥1μg/L, 73% had detectable MC each summer, indicating seasonal inventory studies can identify reservoirs with potential for toxin production.

Microcystin maxima in Missouri reservoirs (21 μg/L) were orders of magnitude less than cyanobacterial blooms in Wisconsin (7,600 μg/L) and Minnesota (8,400), although such large values are not necessarily typical of Midwestern blooms (Hedman et al. 2008, Lindon and Heiskary 2008). The monthly collections from open-water pelagic locations likely missed concentrated surface accumulations of cyanobacteria; therefore, concentrations large enough to cause human-health concerns during recreational activities are likely under-represented (Chorus and Bartram 1999, Graham et al. 2008, Lindon and Heiskary 2008). Toxic cyanobacterial blooms have been documented in the Midwestern United States since the late 1800s (Yoo et al. 1995), and anecdotal reports of animal illness and death after exposure to cyanobacterial blooms are common throughout the region, including Missouri (Graham 2006).

About one-half of reservoirs in this study had detectable MC, similar to the earlier findings of Graham et al. (2004) in Missouri. Graham et al. (2004) sampled particulate MC using a 64-μm mesh plankton net, which may enhance detections but under-represent total concentrations by loss of algal material through the net (Graham and Jones 2007). Maximum MC concentrations were ∼3 μg/L in the Graham et al. (2004) study, compared to 21 μg/L in this dataset. These MC data come from whole-water samples, which better represent the entire cyanobacterial community and more accurately estimate MC concentration (Graham and Jones 2007).

Metalimnetic populations of potential MC producing cyanobacteria occasionally bloom at depth, particularly in mesotrophic systems (Lindholm 1991, Graham et al. 2008). Collection of surface samples will miss detection of these subsurface blooms and may result in underestimates of toxin occurrence. In Missouri reservoirs, there is typically little difference in Chl values at the surface and at the Secchi depth, indicating subsurface Chl peaks are uncommon (Jones et al. 2008a); therefore, subsurface cyanobacterial blooms likely do not represent a large proportion of microcystin occurrences in Missouri reservoirs.

This is the first report of CYL in Missouri. Potential CYL producers, including Anabaena, Aphanizomenon and Cylindrospermopsis, can dominate summer algal assemblages in Missouri reservoirs (Graham et al. 2006a, Jones et al. 2008a), but the toxin was detected in few samples and always at concentrations < 1 μg/L. In most of the United States, CYL occurs relatively infrequently (< 10% of samples) and when detected usually at concentrations < 1 μg/L (Boyer 2007, Hedman et al. 2008), although in Florida CYL has occurred in up to 30% of samples (Williams et al. 2007) and at concentrations as large as 200 μg/L (Burns 2008).

In temperate climates, toxic cyanobacterial blooms historically have occurred most frequently during late summer–early fall (Chorus and Bartram 1999, Huisman et al. 2005); however, in this study, there were no distinct patterns during summer, and MC maxima were observed anytime between May and August (Table 3). Similarly, MC concentrations of concern for recreational exposure were observed from May to September in 12 eutrophic Minnesota lakes (Lindon and Heiskary, 2008). Microcystin maxima may have occurred during months other than May–August in Missouri reservoirs. Previous studies in Missouri have documented year-round MC occurrence, and peak concentrations occasionally occur during winter (Graham et al. 2006a, 2006b). Thus, MC is a potential recreational concern throughout summer and a potential drinking water concern year-round in Midwestern reservoirs and lakes.

Missouri reservoirs with the greatest MC values had increased nutrients and a reduced light environment (Table 6; Fig. 3), characteristics conducive to cyanobacterial dominance (Chorus and Bartram 1999). General patterns in MC with respect to trophic state match these patterns, with peak occurrence and concentrations in eutrophic to hypereutrophic reservoirs (Table 5; Fig. 3). Even so, MC occurred in reservoirs of all trophic status and large nutrient values and shallow Secchi depths were not consistently related to elevated MC values (Table 5 and Table 6; Fig. 3). Because of co-occurrence of toxic and nontoxic cyanobacterial strains (Vézie et al. 1998) and the range of factors that may drive cyanobacterial community dynamics (Reynolds 1998), regional relations between environmental variables and MC occurrence and concentration are invariably complex (Kotak et al. 2000, Chorus 2001, Graham et al. 2004).

Nonlinear relations that describe MC maxima along gradients of nutrients and light (Fig. 3) were similar to those described by Graham et al. (2004) for particulate MC in the Midwest; however, TN and TP ranges where MC maxima were encountered in Missouri are shifted to peak at lower nutrient levels and are narrower than those defined for the region (Fig. 3; Graham et al. 2004). Nonlinear interval-maxima regression is limited by the range of MC values and environmental conditions encountered (Blackburn et al. 1992, Scharf et al. 1998, Graham et al. 2004). Concentrations of MC in Missouri were orders of magnitude greater, and TN and TP values orders of magnitude less, than in the midcontinent region sampled by Graham et al. (2004). The Midwestern relations are likely more representative of global conditions and are similar to relations described in Germany (Chorus 2001) and Alberta, Canada (Kotak et al. 2000). The Missouri relations are local, and may be used to identify conditions under which elevated MC concentrations most frequently occur.

This study demonstrates that MC is common in Missouri reservoirs; most values are low, but concentrations of concern were detected. Though timing of peak concentrations were unique to individual reservoirs, statewide occurrence was relatively consistent among months and years, indicating that including MC as part of routine water-quality monitoring programs may be sufficient to identify reservoirs with frequent MC occurrence; these reservoirs mostly are in the midrange of the trophic continuum. These data enhance understanding of when and where elevated MC concentrations may occur and can lead to more effective identification of human health risks and lake and reservoir management strategies.

Acknowledgments

We thank Steven Heiskary, Minnesota Pollution Control Agency, and Keith Loftin, U.S. Geological Survey, for reviews that provided valuable input and improved the quality of this paper, and D. Obrecht, C. Radcliffe and K. Veum for field and laboratory assistance. Funding and support for this project was provided by Missouri Department of Natural Resources, Food and Agriculture Policy Research Institute, Missouri Agricultural Experiment Station, and the U.S. Geological Survey. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Notes

^a ,

^b ,

^c Indicate significant differences in mean values for each variable. Table 1 variables without significant differences (p > 0.05) are not shown.