- Email: [email protected]
HPLC-PDA detection of cylindrospermopsin—opportunities and limits
Water Research 36 (2002) 4659–4663
Technical note
HPLC-PDA detection of cylindrospermopsin—opportunities and limits Martin Welkera,b,*, Heike Bickela, Jutta Fastnera,b b
a Federal Environmental Agency, Corrensplatz 1, 14195 Berlin, Germany Institute of Chemistry, Technische Universitat . Berlin, Franklinstr. 29, 10587 Berlin, Germany
Received 25 February 2002; received in revised form 3 April 2002; accepted 18 April 2002
Abstract The cyanobacterial hepatotoxic alkaloid cylindrospermopsin (CYL) is of increased concern to public health due to the spreading of its main producer, Cylindrospermopsis raciborskii, around the globe. Here we present results of an evaluation of the possibility to analyse environmental samples for their content of CYL based on HPLC with photo diode array detection as an alternative to costly LC-MS approaches. A gradient from 0% to 50% aqueous methanol (+0.05% trifluoroacetic acid) in 20 min proved to be highly reproducible with respect to peak height, peak area, and retention time of purified CYL. Good linearity of peak area response was found for 1–300 ng CYL on column. For a good performance the duration of equilibration prior to individual runs was crucial. Extraction from cell material (culture and bloom) was efficiently done with pure water in one extraction step and CYL contents determined matched well with results previously obtained by LC-MS. When different seston matrices were added to cultured cells to mimic realistic environmental samples, however, peaks eluting close to CYL in chromatograms restrained the performance. The data presented show a limitation of HPLC-PDA analysis for trace amounts of CYL in environmental samples but also underline the potential of an inexpensive and fast analysis for various purposes. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Cylindrospermopsis raciborskii; Cylindrospermopsin; HPLC; Photo diode array detection; Cyanotoxin
1. Introduction Cyanobacteria have become notorious for the diverse toxic and irritant compounds they are able to produce. Some of these toxins are well studied with respect to their biosynthesis, their effect on an array of organisms, their environmental fate, and possibilities of respective water treatment. For the hepatotoxic peptides microcystins a wealth of knowledge has been gathered accordingly and a variety of analytical methods has been developed for their detection during the last two *Corresponding author. Institute of Chemistry, Technische Universit.at Berlin, Franklinstr. 29, 10587 Berlin, Germany. Tel.:+49-303142-3974; fax:+49-303142-4783. E-mail address: [email protected] (M. Welker).
decades. More recently, another toxin, the hepatotoxic alkaloid cylindrospermopsin (CYL) (Fig. 1), has drawn increased attention of toxicologists and health authorities because the main producer, Cylindrospermopsis raciborskii (Nostocales) is expanding its geographical extension with a considerable pace. This potentially toxic species has recently spread from its supposed origin in Eastern Africa and Australia throughout most of the tropic and subtropic regions of the world [1,2] and most recently northward in temperate zones as far as Northern Germany [3]. The toxin produced by C. raciborskii is an alkaloid highly soluble in water due to its zwitterionic character. Like in microcystins, CYL expresses its toxicity mainly in the vertebrate liver with, however, different toxicological mechanisms [4]. Death of cattle has been repeatedly attributed to the occurrence of CYL [5].
0043-1354/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 3 - 1 3 5 4 ( 0 2 ) 0 0 1 9 4 - X
M. Welker et al. / Water Research 36 (2002) 4659–4663
4660
OH -O S 3
O
NH
N
HN
NH
Voyager-DEt PRO Biospectrometryt workstation (Applied Biosystems, USA) with 2,5-dihydroxy benzoic acid as matrix.
3. Results and discussion
+ 3.1. Elution conditions
NH
O
Fig. 1. Molecular structure of CYL. In deoxy-CYL the oxygen atom printed in bold is lacking.
The production of CYL is not restricted to C. raciborskii. The same toxin has been isolated and identified in Umezakia natans [6], Aphanizomenon ovalisporum [7], Raphidiopsis curvata [8] and may be found in further species. A variant of the toxin, deoxyCYL has been found and confirmed in several strains [9,8]. LC-MS protocols have been established as a standard method for the identification and quantification of CYL [10]. This method works well and the aim of the present study was not to improve the analytical method but to test weather a less expensive alternative analytical protocol is possible: In most laboratories in countries that encounter toxic blooms of C. raciborskii (and other CYL producers) a LC-MS approach is not available for monitoring CYL concentration in field samples.
2. Methods and material CYL standard was kindly provided by Geoff Eaglesham and Peta Senogles (Queensland Health Scientific Service) as well as dried material of a laboratory culture of C. raciborskii AWT 205. Further culture material came from strains isolated from German lakes and another strain and bloom material was kindly provided by Martin Saker (Univ. Porto). The HPLC analyses were carried out on a system from Waters Inc., Milford, USA, consisting of a 600 controlling unit, a 996 photodiode array detector, a 717 auto sampler and an in-line degasser. The column was a LiChroSpher RP 18 cartridge from Merck with a precolumn of the same sorbent material. Detection was by UV-scanning from 200 to 350 nm at a rate of 1 spectrum s1. CYL has an UV-absorption spectrum with a maximum at l ¼ 262 nm. Eluents were water, methanol, acetonitrile, and mixtures thereof. Eluents were applied either as supplied by the manufacturer or acidified with 0.05% v/v trifluoroacetic acid (TFA). HPLC fractions were collected manually and analysed by matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS) on a
The application of several gradients and eluent pairs clearly revealed that all gradients with acetonitrile, either acidified or not, and those with pure water and methanol did not result in a retention of CYL on the column used (data not shown). The toxin peak invariably eluted together with or shortly after the solvent front. Only gradients of acidified water and methanol showed any retention of the toxin and therefore the following will be concentrated solely on respective gradients. Satisfactory retention was achieved with different proportions of water and MeOH when 0.05% TFA (v/v) was added. All chromatograms discussed in the following refer to chromatograms extracted from PDA data at a wavelength of l ¼ 262 nm. In a series of linear gradients starting with 100% H2O+0.05%TFA to a final eluent condition with 5– 50% aqueous MeOH+0.05% TFA after 20 min gave retention times of pure CYL in relation to the slope of the gradient (Fig. 2). For shallow gradients the retention time was longest and shortened successively with higher final MeOH proportions and could be manipulated according to a saturation function fitted to the data. Relative deviation in retention times were low for all gradients with a maximum coefficient of variance (CVRT,max) o2.0% for particular gradients. Longer retention times, however, resulted in increasingly broader peaks or decreases in peak height, respectively. Peak heights, too, were reproducible with CVh,maxo3.0%. In contrast, peak area was only weakly influenced by variations in eluent gradients with slightly lower values for steep gradients that, however, could be the result of the actual integration settings in the Millenium 32 software. Most important, peak area was reproducible, too, with CVA,maxo1.1% for all gradients. For the performance of the tested gradients equilibration of the column prior to individual runs with 100% water+0.05% TFA was crucial. Equilibration times shorter than 8 min resulted in considerable shifts of retention time and equilibration times o5 min prevented retention at all for gradients with 30–50% MeOH+0.05% TFA as final condition. Therefore, equilibration time was set to 15 min before each injection. All applied gradients thus had comparably good performances. A further increase of MeOH proportion in final gradient conditions did increase peak sharpness only slightly but with the considerable cost of extended
M. Welker et al. / Water Research 36 (2002) 4659–4663
quality was impaired by background noise. A calibration showed good linearity ðR2 ¼ 0:9999Þ for a load range from 1 to 300 ng o.c.
10
3.2. Extraction of CYL from cell material (a) 5
0
Retention time [min]
16
14 (b) 12
10 85 80
(c)
75 70 0
10
20
30
40
50
Final eluent condition in % MeOH Fig. 2. Peak height (a), retention time (b), and peak area (c) of 25 ng CYL on column dependent on elution conditions. Gradients of MeOH and water (+0.05% v/v TFA) were from 0% to X % MeOH in 20 min as indicated by the categories on the abscissa. Data represent mean7SD of three injections.
equilibration time (data not shown). For further experiments we chose the gradient from 0% to 50% MeOH+0.05% TFA in 20 min for it gave sharp peaks with a ratio of height to half-maximum width of h:w >50 for 40 ng CYL on column (o.c.) and allowed a high throughput due to short retention time of CYL. Reproducibility was tested with five runs of 10 and 100 ng CYL o.c., respectively. For 10 ng o.c. coefficients of variation were CVh=0.6%, CVA=1.0%, and CVRT=0.1% while for 100 ng o.c. respective values were CVh=0.3%, CVA=0.1%, and CVRT=0.1%. The next step was to assess the detection limit under the selected gradient condition. An amount of 5 ng o.c. gave a good peak with a absorbance spectrum equal to that of 40 ng o.c. (match angle o51 in the Millenium 32 spectral match calculation [11]). An amount of 1 ng o.c. still resulted in a significant peak but the spectrum
Freeze-dried cell material was kindly provided by Peta Sinogles from cultures of C. raciborskii AWT 205. Extraction of the toxin was done in reaction tubes with varying solvents. A weighted sample of freeze dried material (7–14 mg in a reaction tube) was suspended in 1 mL of respective solvents (see below). The suspension was sonicated for 15 min in a water bath, shaken for 1 h at room temperature, and sonicated again for a single extraction cycle. Suspensions were then centrifuged and the supernatant was collected and either directly injected to the HPLC or stored at 41C until processed further. As solvent we tested 100% MeOH, 5% acetic acid, MeOH:5% acetic acid=1:3, and 100% water. Only 100% water gave satisfactory results with all other solvent yielding maximally 30% of CYL when compared to the aqueous extract on a w/w base after one extraction cycle. For extractions with acetonitrile and MeOH it is not clear, whether the low recovery in HPLC is actually due to bad extraction performance or due to a partial break-through caused by organic solvents in the sample (though a CYL peak at the expected retention time was recorded). However, extraction with 100% water proved to be very efficient and was completed with a single extraction step (Fig. 3, P > 0:99; ANOVA). To avoid clogging of the column by insoluble compounds a precipitation step was introduced. After extraction and centrifugation TFA was added to the clear supernatant at a concentration of 0.1% v/v. The sample was shaken
2
Toxin content [µg CYL (mg DW) -1]
Peak height [mAU]
15
Peak area [mAU*sec]
4661
1
0 1
2
3
+ 0.1%
Number of extraction steps TFA Fig. 3. Toxin content of a C. raciborskii sample calculated after each extraction step (light columns). Dark column: addition of TFA to the supernatant aquaeous extract followed by a precipitation/centrifugation step. Mean of 5 extractions 7SD.
M. Welker et al. / Water Research 36 (2002) 4659–4663
for 1 h and allowed to stand for 3 h at room temperature before repeated centrifugation. The addition of TFA resulted in the precipitation of dissolved cell components (likely proteinic compounds) but had no influence on the determined CYL content (Fig. 3). A bloom sample collected in an aquaculture pond in Townsville (Australia) and kindly provided by Martin Saker, Univ. Porto, Portugal, was extracted and analysed accordingly. In the chromatogram two peaks were found with an absorbance spectrum resembling of CYL: one with the retention time of CYL and the other one with a retention time short to 15 min (Fig. 4). Mass spectral analysis revealed that the first peak was a compound with a mass of m=z ¼ 416:11 Da [M+H]+compared to 416.12 Da as theoretical mass thus confirming CYL. The second peak was a compound with a mass of m=z ¼ 400:12 Da [M+H]+which corresponds to the de-oxy variant of CYL. The CYL content determined by our extraction and detection method was 7.0 mg CYL (mg dry weight)1 (DW) matching well to the content determined by LC-MS of 6.8 mg CYL (mg DW)1 (Martin Saker, pers. comm.). In isolates of C. raciborskii from German lakes we could not detect CYL and the finding was confirmed by LCMS analysis (Eaglesham and Fastner, unpublished data). We thence assumed that the method developed so far was suitable for detection and quantification of CYL in bloom samples dominated by C. raciborskii or in culture experiments. In mid-European waters C. raciborskii was seldom found to be the dominant cyanobacterium in any water body up to now (except for one mass development reported in France [12]). Relative biomasses only rarely exceeded 10% and thus the method had to be tested for its robustness with more realistic samples. Different seston samples were used to be mixed with dried cells of AWT 205 with a share of C. raciborskii spanning from 8% to 35% w/w. The seston samples covered a broad
0.5
a
AU
0.4 0.3 0.2
300
CYL detected [ng on column]
4662
250
200
MS1 MS2 LS WS ScS
150
( 100 100
150
) 200
250
300
CYL calculated [ng on column] Fig. 5. Detected vs. calculated amount of CYL in extracts of 8– 36% C. raciborskii culture material added to seston samples from Berlin lakes. MS: Muggelsee, . LS: Langer See, WS: Wannsee, ScS: Schlachtensee. Dotted line: 100% equivalent; full line: linear regression calculated excluding extracts from ‘Wannsee’.
range of possible natural plankton communities with main component being Planktothrix, Aphanizomenon/ Microcystis, diatoms, or rotifers. From previously analysed samples of the cultured cells we calculated the theoretical CYL content of the composed samples (ranging from 1.7 to 7.5 mg CYL (mg DW)1). For each seston matrix three replicates were extracted and analysed. The amount of CYL measured was then compared to the amount theoretically expected (Fig. 5). In all samples except those with seston from lake Wannsee (WS) CYL could be detected according to retention time and UV-spectra. In some samples, however, unknown compounds nearly coeluted and in the resulting chromatograms CYL had to be quantified from double peaks. In comparison to expected values the measured ones were satisfactory close to equality for most samples. The complete failure of detection of CYL in ‘Wannsee’ samples, however, clearly shows the limits of the method developed so far. Complex matrices are the main impediment when natural samples are analysed and particular compounds of interest are to be quantified.
b
0.1
4. Conclusions
0.0 0
5
10
15
20
Retention time [min] Fig. 4. HPLC-PDA chromatogram (l ¼ 262 nm) of a cyanobacterial bloom sample dominated by C. raciborskii collected in an aquaculture pond in Townsville (Australia). Elution with 0– 50% MeOH (+0.05% TFA) in 20 min (a) CYL, (b) deoxyCYL.
Our results demonstrate that HPLC-PDA analysis of CYL is possible in principle: retention time is reproducible on a reversed phase column, linearity is given in a reasonable range, the detection limit is sufficiently low for most purposes, and the required chemicals were mainly inexpensive water and MeOH. Nonetheless, the
M. Welker et al. / Water Research 36 (2002) 4659–4663
application of the protocol to environmental samples proved to be hampered by the fact that the extraction with pure water, though very efficient for CYL, gave a considerable matrix background and occasionally covered CYL completely in chromatograms. Therefore, future efforts will be directed to develop efficient cleanup steps to remove at least part of the matrix background and to reliably measure concentrations of CYL from environmental samples with low (o0.2%) CYL contents.
Acknowledgements We thank Geoff Eaglesham and Peta Sinogles, Queensland Health Scientific Service, for providing purified CYL and dried cell material and Martin Saker, University of Porto, for a toxic bloom sample of Cylindrospermopsis raciborskii. This study was funded by the Deutsche Forschungsgemeinschaft (Grant CH 113/1-1).
[4]
[5]
[6]
[7]
[8]
[9]
References [1] Padis!ak J. Cylindrospermopsis raciborskii (Woloszynska) Seenayya et Stubba Raju, an expanding, highly adaptive cyanobacterium: worldwide distribution and review of its ecology. Arch Hydrobiol Suppl 1997;107:563–93. [2] deSouza RCR, Carvalho MC, Truzzi AC. Cylindrospermopsis raciborskii (Wolosz.) Seenaya and Subba Raju (Cyanophyceae) dominance and a contribution to the knowledge of Rio Pequeno arm, Billings Reservoir, Brazil. Environ Toxicol Water Qual 1998;13:73–81. [3] Mischke U, Wiedner C, Nixdorf B. Distribution of the neophyte Cylindrospermopsis raciborskii (Wolosz.) Seenayya and Subba Raju (Cyanobacteria) in Brandenburg/
[10]
[11]
[12]
4663
Germany. Deutsche Ges. Limnol, Proceedings 2000; 715–9 (in German). Shaw GR, Seawright AA, Moore MR, Lam PKS. Cylindrospermopsin, a cyanobacterial alkaloid: evaluation of its toxicologic activity. Therapeutic Drug Monitoring 2000;22:89–92. Saker ML, Thomas AD, Norton JH. Cattle mortality attributed to the toxic cyanobacterium Cylindrospermopsis raciborskii in an outback region of north Queensland. Environ Toxicol 1999;14:179–82. Harada K-I, Ohtani I, Iwamoto K, Suzuki M, Watanabe MF, Watanabe M, Terao K. Isolation of cylindrospermopsin from a cyanobacterium Umezakia natans and its screening method. Toxicon 1994;32:73–84. Banker R, Carmeli S, Hadas O, Teltsch B, Porat R, Sukenik A. Identification of cylindrospermopsin in Aphanizomenon ovalisporum (Cyanophyceae) isolated from lake Kinneret. J Phycol 1997;33:613–6. Li R, Carmichael WW, Brittain S, Eaglesham GK, Shaw GR, Li Y, Watanabe MM. First report of the cyanotoxins cylindrospermopsin and desoxycylindrospermopsin from Raphidiopsis curvata (cyanobacteria). J Phycol 2001; 37:1121–6. Norris RL, Eaglesham GK, Pierens G, Shaw GR, Smith MJ, Chiswell RK, Seawright AA, Moore MR. Deoxycylindrospermopsin, an analog of cylindrospermopsin from Cylindrospermopsis raciborskii. Environ Toxicol 1999;14:163–5. Eaglesham GK, Norris RL, Shaw GR, Smith MJ, Chiswell RK, Davis BC, Neville GR, Seawright AA, Moore MR. Use of HPLC-MS/MS to monitor cylindrospermopsin, a blue–green algal toxin, for public health purposes. Environ Toxicol 1999;14:151–4. Sievert H-JP, Drouen ACJH. Spectral matching and peak purity. In: Huber L, George SA, editors. Diode array detection in HPLC. Chromatographic science series, Vol. 62. New York: Dekker, 1993. p. 51–126. Cout!e A, Leitao M, Martin M. Premiere observation du genre Cylindrospermopsis (Cyanophyceae, Nostocales) en France. Cryptogamie Algol 1997;18:57–70.
Technical note
HPLC-PDA detection of cylindrospermopsin—opportunities and limits Martin Welkera,b,*, Heike Bickela, Jutta Fastnera,b b
a Federal Environmental Agency, Corrensplatz 1, 14195 Berlin, Germany Institute of Chemistry, Technische Universitat . Berlin, Franklinstr. 29, 10587 Berlin, Germany
Received 25 February 2002; received in revised form 3 April 2002; accepted 18 April 2002
Abstract The cyanobacterial hepatotoxic alkaloid cylindrospermopsin (CYL) is of increased concern to public health due to the spreading of its main producer, Cylindrospermopsis raciborskii, around the globe. Here we present results of an evaluation of the possibility to analyse environmental samples for their content of CYL based on HPLC with photo diode array detection as an alternative to costly LC-MS approaches. A gradient from 0% to 50% aqueous methanol (+0.05% trifluoroacetic acid) in 20 min proved to be highly reproducible with respect to peak height, peak area, and retention time of purified CYL. Good linearity of peak area response was found for 1–300 ng CYL on column. For a good performance the duration of equilibration prior to individual runs was crucial. Extraction from cell material (culture and bloom) was efficiently done with pure water in one extraction step and CYL contents determined matched well with results previously obtained by LC-MS. When different seston matrices were added to cultured cells to mimic realistic environmental samples, however, peaks eluting close to CYL in chromatograms restrained the performance. The data presented show a limitation of HPLC-PDA analysis for trace amounts of CYL in environmental samples but also underline the potential of an inexpensive and fast analysis for various purposes. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Cylindrospermopsis raciborskii; Cylindrospermopsin; HPLC; Photo diode array detection; Cyanotoxin
1. Introduction Cyanobacteria have become notorious for the diverse toxic and irritant compounds they are able to produce. Some of these toxins are well studied with respect to their biosynthesis, their effect on an array of organisms, their environmental fate, and possibilities of respective water treatment. For the hepatotoxic peptides microcystins a wealth of knowledge has been gathered accordingly and a variety of analytical methods has been developed for their detection during the last two *Corresponding author. Institute of Chemistry, Technische Universit.at Berlin, Franklinstr. 29, 10587 Berlin, Germany. Tel.:+49-303142-3974; fax:+49-303142-4783. E-mail address: [email protected] (M. Welker).
decades. More recently, another toxin, the hepatotoxic alkaloid cylindrospermopsin (CYL) (Fig. 1), has drawn increased attention of toxicologists and health authorities because the main producer, Cylindrospermopsis raciborskii (Nostocales) is expanding its geographical extension with a considerable pace. This potentially toxic species has recently spread from its supposed origin in Eastern Africa and Australia throughout most of the tropic and subtropic regions of the world [1,2] and most recently northward in temperate zones as far as Northern Germany [3]. The toxin produced by C. raciborskii is an alkaloid highly soluble in water due to its zwitterionic character. Like in microcystins, CYL expresses its toxicity mainly in the vertebrate liver with, however, different toxicological mechanisms [4]. Death of cattle has been repeatedly attributed to the occurrence of CYL [5].
0043-1354/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 3 - 1 3 5 4 ( 0 2 ) 0 0 1 9 4 - X
M. Welker et al. / Water Research 36 (2002) 4659–4663
4660
OH -O S 3
O
NH
N
HN
NH
Voyager-DEt PRO Biospectrometryt workstation (Applied Biosystems, USA) with 2,5-dihydroxy benzoic acid as matrix.
3. Results and discussion
+ 3.1. Elution conditions
NH
O
Fig. 1. Molecular structure of CYL. In deoxy-CYL the oxygen atom printed in bold is lacking.
The production of CYL is not restricted to C. raciborskii. The same toxin has been isolated and identified in Umezakia natans [6], Aphanizomenon ovalisporum [7], Raphidiopsis curvata [8] and may be found in further species. A variant of the toxin, deoxyCYL has been found and confirmed in several strains [9,8]. LC-MS protocols have been established as a standard method for the identification and quantification of CYL [10]. This method works well and the aim of the present study was not to improve the analytical method but to test weather a less expensive alternative analytical protocol is possible: In most laboratories in countries that encounter toxic blooms of C. raciborskii (and other CYL producers) a LC-MS approach is not available for monitoring CYL concentration in field samples.
2. Methods and material CYL standard was kindly provided by Geoff Eaglesham and Peta Senogles (Queensland Health Scientific Service) as well as dried material of a laboratory culture of C. raciborskii AWT 205. Further culture material came from strains isolated from German lakes and another strain and bloom material was kindly provided by Martin Saker (Univ. Porto). The HPLC analyses were carried out on a system from Waters Inc., Milford, USA, consisting of a 600 controlling unit, a 996 photodiode array detector, a 717 auto sampler and an in-line degasser. The column was a LiChroSpher RP 18 cartridge from Merck with a precolumn of the same sorbent material. Detection was by UV-scanning from 200 to 350 nm at a rate of 1 spectrum s1. CYL has an UV-absorption spectrum with a maximum at l ¼ 262 nm. Eluents were water, methanol, acetonitrile, and mixtures thereof. Eluents were applied either as supplied by the manufacturer or acidified with 0.05% v/v trifluoroacetic acid (TFA). HPLC fractions were collected manually and analysed by matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS) on a
The application of several gradients and eluent pairs clearly revealed that all gradients with acetonitrile, either acidified or not, and those with pure water and methanol did not result in a retention of CYL on the column used (data not shown). The toxin peak invariably eluted together with or shortly after the solvent front. Only gradients of acidified water and methanol showed any retention of the toxin and therefore the following will be concentrated solely on respective gradients. Satisfactory retention was achieved with different proportions of water and MeOH when 0.05% TFA (v/v) was added. All chromatograms discussed in the following refer to chromatograms extracted from PDA data at a wavelength of l ¼ 262 nm. In a series of linear gradients starting with 100% H2O+0.05%TFA to a final eluent condition with 5– 50% aqueous MeOH+0.05% TFA after 20 min gave retention times of pure CYL in relation to the slope of the gradient (Fig. 2). For shallow gradients the retention time was longest and shortened successively with higher final MeOH proportions and could be manipulated according to a saturation function fitted to the data. Relative deviation in retention times were low for all gradients with a maximum coefficient of variance (CVRT,max) o2.0% for particular gradients. Longer retention times, however, resulted in increasingly broader peaks or decreases in peak height, respectively. Peak heights, too, were reproducible with CVh,maxo3.0%. In contrast, peak area was only weakly influenced by variations in eluent gradients with slightly lower values for steep gradients that, however, could be the result of the actual integration settings in the Millenium 32 software. Most important, peak area was reproducible, too, with CVA,maxo1.1% for all gradients. For the performance of the tested gradients equilibration of the column prior to individual runs with 100% water+0.05% TFA was crucial. Equilibration times shorter than 8 min resulted in considerable shifts of retention time and equilibration times o5 min prevented retention at all for gradients with 30–50% MeOH+0.05% TFA as final condition. Therefore, equilibration time was set to 15 min before each injection. All applied gradients thus had comparably good performances. A further increase of MeOH proportion in final gradient conditions did increase peak sharpness only slightly but with the considerable cost of extended
M. Welker et al. / Water Research 36 (2002) 4659–4663
quality was impaired by background noise. A calibration showed good linearity ðR2 ¼ 0:9999Þ for a load range from 1 to 300 ng o.c.
10
3.2. Extraction of CYL from cell material (a) 5
0
Retention time [min]
16
14 (b) 12
10 85 80
(c)
75 70 0
10
20
30
40
50
Final eluent condition in % MeOH Fig. 2. Peak height (a), retention time (b), and peak area (c) of 25 ng CYL on column dependent on elution conditions. Gradients of MeOH and water (+0.05% v/v TFA) were from 0% to X % MeOH in 20 min as indicated by the categories on the abscissa. Data represent mean7SD of three injections.
equilibration time (data not shown). For further experiments we chose the gradient from 0% to 50% MeOH+0.05% TFA in 20 min for it gave sharp peaks with a ratio of height to half-maximum width of h:w >50 for 40 ng CYL on column (o.c.) and allowed a high throughput due to short retention time of CYL. Reproducibility was tested with five runs of 10 and 100 ng CYL o.c., respectively. For 10 ng o.c. coefficients of variation were CVh=0.6%, CVA=1.0%, and CVRT=0.1% while for 100 ng o.c. respective values were CVh=0.3%, CVA=0.1%, and CVRT=0.1%. The next step was to assess the detection limit under the selected gradient condition. An amount of 5 ng o.c. gave a good peak with a absorbance spectrum equal to that of 40 ng o.c. (match angle o51 in the Millenium 32 spectral match calculation [11]). An amount of 1 ng o.c. still resulted in a significant peak but the spectrum
Freeze-dried cell material was kindly provided by Peta Sinogles from cultures of C. raciborskii AWT 205. Extraction of the toxin was done in reaction tubes with varying solvents. A weighted sample of freeze dried material (7–14 mg in a reaction tube) was suspended in 1 mL of respective solvents (see below). The suspension was sonicated for 15 min in a water bath, shaken for 1 h at room temperature, and sonicated again for a single extraction cycle. Suspensions were then centrifuged and the supernatant was collected and either directly injected to the HPLC or stored at 41C until processed further. As solvent we tested 100% MeOH, 5% acetic acid, MeOH:5% acetic acid=1:3, and 100% water. Only 100% water gave satisfactory results with all other solvent yielding maximally 30% of CYL when compared to the aqueous extract on a w/w base after one extraction cycle. For extractions with acetonitrile and MeOH it is not clear, whether the low recovery in HPLC is actually due to bad extraction performance or due to a partial break-through caused by organic solvents in the sample (though a CYL peak at the expected retention time was recorded). However, extraction with 100% water proved to be very efficient and was completed with a single extraction step (Fig. 3, P > 0:99; ANOVA). To avoid clogging of the column by insoluble compounds a precipitation step was introduced. After extraction and centrifugation TFA was added to the clear supernatant at a concentration of 0.1% v/v. The sample was shaken
2
Toxin content [µg CYL (mg DW) -1]
Peak height [mAU]
15
Peak area [mAU*sec]
4661
1
0 1
2
3
+ 0.1%
Number of extraction steps TFA Fig. 3. Toxin content of a C. raciborskii sample calculated after each extraction step (light columns). Dark column: addition of TFA to the supernatant aquaeous extract followed by a precipitation/centrifugation step. Mean of 5 extractions 7SD.
M. Welker et al. / Water Research 36 (2002) 4659–4663
for 1 h and allowed to stand for 3 h at room temperature before repeated centrifugation. The addition of TFA resulted in the precipitation of dissolved cell components (likely proteinic compounds) but had no influence on the determined CYL content (Fig. 3). A bloom sample collected in an aquaculture pond in Townsville (Australia) and kindly provided by Martin Saker, Univ. Porto, Portugal, was extracted and analysed accordingly. In the chromatogram two peaks were found with an absorbance spectrum resembling of CYL: one with the retention time of CYL and the other one with a retention time short to 15 min (Fig. 4). Mass spectral analysis revealed that the first peak was a compound with a mass of m=z ¼ 416:11 Da [M+H]+compared to 416.12 Da as theoretical mass thus confirming CYL. The second peak was a compound with a mass of m=z ¼ 400:12 Da [M+H]+which corresponds to the de-oxy variant of CYL. The CYL content determined by our extraction and detection method was 7.0 mg CYL (mg dry weight)1 (DW) matching well to the content determined by LC-MS of 6.8 mg CYL (mg DW)1 (Martin Saker, pers. comm.). In isolates of C. raciborskii from German lakes we could not detect CYL and the finding was confirmed by LCMS analysis (Eaglesham and Fastner, unpublished data). We thence assumed that the method developed so far was suitable for detection and quantification of CYL in bloom samples dominated by C. raciborskii or in culture experiments. In mid-European waters C. raciborskii was seldom found to be the dominant cyanobacterium in any water body up to now (except for one mass development reported in France [12]). Relative biomasses only rarely exceeded 10% and thus the method had to be tested for its robustness with more realistic samples. Different seston samples were used to be mixed with dried cells of AWT 205 with a share of C. raciborskii spanning from 8% to 35% w/w. The seston samples covered a broad
0.5
a
AU
0.4 0.3 0.2
300
CYL detected [ng on column]
4662
250
200
MS1 MS2 LS WS ScS
150
( 100 100
150
) 200
250
300
CYL calculated [ng on column] Fig. 5. Detected vs. calculated amount of CYL in extracts of 8– 36% C. raciborskii culture material added to seston samples from Berlin lakes. MS: Muggelsee, . LS: Langer See, WS: Wannsee, ScS: Schlachtensee. Dotted line: 100% equivalent; full line: linear regression calculated excluding extracts from ‘Wannsee’.
range of possible natural plankton communities with main component being Planktothrix, Aphanizomenon/ Microcystis, diatoms, or rotifers. From previously analysed samples of the cultured cells we calculated the theoretical CYL content of the composed samples (ranging from 1.7 to 7.5 mg CYL (mg DW)1). For each seston matrix three replicates were extracted and analysed. The amount of CYL measured was then compared to the amount theoretically expected (Fig. 5). In all samples except those with seston from lake Wannsee (WS) CYL could be detected according to retention time and UV-spectra. In some samples, however, unknown compounds nearly coeluted and in the resulting chromatograms CYL had to be quantified from double peaks. In comparison to expected values the measured ones were satisfactory close to equality for most samples. The complete failure of detection of CYL in ‘Wannsee’ samples, however, clearly shows the limits of the method developed so far. Complex matrices are the main impediment when natural samples are analysed and particular compounds of interest are to be quantified.
b
0.1
4. Conclusions
0.0 0
5
10
15
20
Retention time [min] Fig. 4. HPLC-PDA chromatogram (l ¼ 262 nm) of a cyanobacterial bloom sample dominated by C. raciborskii collected in an aquaculture pond in Townsville (Australia). Elution with 0– 50% MeOH (+0.05% TFA) in 20 min (a) CYL, (b) deoxyCYL.
Our results demonstrate that HPLC-PDA analysis of CYL is possible in principle: retention time is reproducible on a reversed phase column, linearity is given in a reasonable range, the detection limit is sufficiently low for most purposes, and the required chemicals were mainly inexpensive water and MeOH. Nonetheless, the
M. Welker et al. / Water Research 36 (2002) 4659–4663
application of the protocol to environmental samples proved to be hampered by the fact that the extraction with pure water, though very efficient for CYL, gave a considerable matrix background and occasionally covered CYL completely in chromatograms. Therefore, future efforts will be directed to develop efficient cleanup steps to remove at least part of the matrix background and to reliably measure concentrations of CYL from environmental samples with low (o0.2%) CYL contents.
Acknowledgements We thank Geoff Eaglesham and Peta Sinogles, Queensland Health Scientific Service, for providing purified CYL and dried cell material and Martin Saker, University of Porto, for a toxic bloom sample of Cylindrospermopsis raciborskii. This study was funded by the Deutsche Forschungsgemeinschaft (Grant CH 113/1-1).
[4]
[5]
[6]
[7]
[8]
[9]
References [1] Padis!ak J. Cylindrospermopsis raciborskii (Woloszynska) Seenayya et Stubba Raju, an expanding, highly adaptive cyanobacterium: worldwide distribution and review of its ecology. Arch Hydrobiol Suppl 1997;107:563–93. [2] deSouza RCR, Carvalho MC, Truzzi AC. Cylindrospermopsis raciborskii (Wolosz.) Seenaya and Subba Raju (Cyanophyceae) dominance and a contribution to the knowledge of Rio Pequeno arm, Billings Reservoir, Brazil. Environ Toxicol Water Qual 1998;13:73–81. [3] Mischke U, Wiedner C, Nixdorf B. Distribution of the neophyte Cylindrospermopsis raciborskii (Wolosz.) Seenayya and Subba Raju (Cyanobacteria) in Brandenburg/
[10]
[11]
[12]
4663
Germany. Deutsche Ges. Limnol, Proceedings 2000; 715–9 (in German). Shaw GR, Seawright AA, Moore MR, Lam PKS. Cylindrospermopsin, a cyanobacterial alkaloid: evaluation of its toxicologic activity. Therapeutic Drug Monitoring 2000;22:89–92. Saker ML, Thomas AD, Norton JH. Cattle mortality attributed to the toxic cyanobacterium Cylindrospermopsis raciborskii in an outback region of north Queensland. Environ Toxicol 1999;14:179–82. Harada K-I, Ohtani I, Iwamoto K, Suzuki M, Watanabe MF, Watanabe M, Terao K. Isolation of cylindrospermopsin from a cyanobacterium Umezakia natans and its screening method. Toxicon 1994;32:73–84. Banker R, Carmeli S, Hadas O, Teltsch B, Porat R, Sukenik A. Identification of cylindrospermopsin in Aphanizomenon ovalisporum (Cyanophyceae) isolated from lake Kinneret. J Phycol 1997;33:613–6. Li R, Carmichael WW, Brittain S, Eaglesham GK, Shaw GR, Li Y, Watanabe MM. First report of the cyanotoxins cylindrospermopsin and desoxycylindrospermopsin from Raphidiopsis curvata (cyanobacteria). J Phycol 2001; 37:1121–6. Norris RL, Eaglesham GK, Pierens G, Shaw GR, Smith MJ, Chiswell RK, Seawright AA, Moore MR. Deoxycylindrospermopsin, an analog of cylindrospermopsin from Cylindrospermopsis raciborskii. Environ Toxicol 1999;14:163–5. Eaglesham GK, Norris RL, Shaw GR, Smith MJ, Chiswell RK, Davis BC, Neville GR, Seawright AA, Moore MR. Use of HPLC-MS/MS to monitor cylindrospermopsin, a blue–green algal toxin, for public health purposes. Environ Toxicol 1999;14:151–4. Sievert H-JP, Drouen ACJH. Spectral matching and peak purity. In: Huber L, George SA, editors. Diode array detection in HPLC. Chromatographic science series, Vol. 62. New York: Dekker, 1993. p. 51–126. Cout!e A, Leitao M, Martin M. Premiere observation du genre Cylindrospermopsis (Cyanophyceae, Nostocales) en France. Cryptogamie Algol 1997;18:57–70.