The Behavior of Mixtures of Paralytic Shellfish Toxins in Competitive

PSTs block voltage-gated sodium channels (Na channel) and elicit neurotoxicity. Animals, including ... Katie O'Neill , Ian F. Musgrave , Andrew Humpag...
0 downloads 0 Views 162KB Size
Chem. Res. Toxicol. 2006, 19, 661-667

661

The Behavior of Mixtures of Paralytic Shellfish Toxins in Competitive Binding Assays Lyndon E. Llewellyn* Australian Institute of Marine Science, PMB 3, TownsVille MC, Queensland, 4810, Australia ReceiVed October 4, 2005

Organisms that contain paralytic shellfish toxins (PSTs) may contain many members of this toxin family. PSTs block voltage-gated sodium channels (Na channel) and elicit neurotoxicity. Animals, including humans, may encounter PST mixtures via consumption of tainted seafood, contaminated water, or the microalgae that produce the toxins. PST binding by the Na channel as well as other proteins such as antibodies and saxiphilin have been used to develop biomolecular assays for PSTs. An equation that predicts the combined effects of binary and ternary PST mixtures has been experimentally validated for two unrelated STX-binding proteins, the rat brain Na channel and a saxiphilin from the xanthid crab Liomera tristis. It was found that the most potent toxin or toxins in any mixture profoundly affect the cumulative potency of the mixture, overwhelming weaker toxins with the transition from strong to weak toxicity and changing in a curvilinear manner. Less active PSTs must be several orders of magnitude more concentrated than stronger toxins for the mixture to reflect their potency. This behavior is important in understanding how toxin mixtures may act at the Na channel receptor via which PSTs exert their neurotoxicity and that the presence of weaker toxins does not dilute the effect of stronger toxins in a linear fashion. This strong dominance of a mixture by the most potent toxins also has implications for measurement of toxic test samples and for standards that may contain low levels of highly potent bioactive impurities. This equation has been extended to mixtures of PSTs containing more than three toxins and may be applicable to other natural contaminants and any competitive binding assays used to detect their presence and measure their concentration. Introduction Saxitoxin (STX)1 and its many analogues (1), collectively referred to here as the paralytic shellfish toxins (PSTs), can contaminate mollusks, crustacea, and fish (2-4). Ingestion of these PST-tainted animals by humans can cause paralytic shellfish poisoning (PSP), which may be fatal. PST neurotoxicity in people and animals that encounter the toxins in the environment is exerted through a blockade of a family of transmembrane voltage-gated sodium channels (Na channels) prohibiting sodium ion movement across cell membranes of excitable tissue (5). PST-contaminated animals and microalgal toxin producers may carry a multitude of STX derivatives with individual toxin concentrations extending from trace amounts to complete dominance (6-11). In the natural situation then, mixtures of PSTs are normal. Public safety-motivated testing of seafood for PST toxicity is usually achieved by injecting seafood extracts into the peritoneum of mice; the time until death is the measure of the total toxicity (12). Bioassay alternatives to this live animal testing include cytotoxicity assays (13-17), Na channel-based tissue sensors (18-20), ELISAs (21), and binding of radiolabeled STX by Na channels and saxiphilins (22-24). Common to all of these approaches is that the assay measure is a direct result of toxin binding, be it by the Na channel receptor or hydrophilic proteins such as antibodies or saxiphilin. * To whom correspondence should be addressed. Tel: +61-7-4753 4449. Fax: +61-7-4772 5852. E-mail: [email protected]. 1 Abbreviations: dcSTX, decarbamoylsaxitoxin; GTX5, gonyautoxin 5; IC50, concentration that inhibits radioligand binding by 50%; Na channel, voltage-gated sodium channel; neoSTX, neosaxitoxin; PSP, paralytic shellfish poisoning; PSTs, paralytic shellfish toxins; STX, saxitoxin; [3H]STX, tritiated saxitoxin.

Mixtures of related chemicals and how they impact upon biomolecule-dependent assays have been little studied in molecular pharmacology and are absent for PSTs. Evaluations of HPLC predictions of the bioassay results have been proxies for studies of the pharmacological and toxicological behavior of PST mixtures. HPLC methods separate PSTs and relate chromatography peak area to PST standards; a cumulative toxicity value is calculated as follows:

total toxicity )

∑[C(i) T(i) D(CF)]

(1)

where C(i) is the concentration of individual PST in µM, T(i) equals the toxicity of individual toxins in mouse units per µmol, D is the dilution factor of the tested extract, and CF is the conversion factor for each PST in µg STX per mouse unit (25). This technique requires standards for each of the analyzed PSTs, a reliable specific toxicity value for each PST, and that toxins are above the detection limit. With this latter point, detection limits for PSTs can differ substantially (25, 26). Agreement between the HPLC analyses and the mouse lethality bioassay is high (25-27), suggesting that PST mixtures injected into mice act in a simple additive manner reflecting eq 1. A model is described here that explains the behavior of PST mixtures in binding assays, and experimental data are reported to validate the model. The model predicts the amount of inhibition by toxin mixtures in competitive binding assays from which an IC50 (concentration that inhibits radioligand binding by 50%) can be calculated. Simple % inhibition values as well as the IC50 have been used for PST quantitation in assays related to those used here (10, 22, 28) as well as in ELISAs (21) and

10.1021/tx050277i CCC: $33.50 Published 2006 by the American Chemical Society Published on Web 04/22/2006

662 Chem. Res. Toxicol., Vol. 19, No. 5, 2006

other PST bioassays (13, 14). This equation may be generalized to accommodate an infinite number of PST analogues.

Llewellyn fraction binding protein occupied ) 1+

Experimental Procedures Overall Experimental Design. Two STX-binding proteins were selected, which have measurably different sensitivities to a series of PSTs. This enabled quantitation of the change in IC50 of binary and ternary toxin mixtures when the binding proteins are exposed to different ratios of toxins. These proteins behave as a single isoform in competitive binding assays (29), and the PSTs used were high quality pure toxins not contaminated with other PSTs. Presently, pure and certified individual PSTs are available from the Canadian National Research Council Certified Reference Material Program and include STX, neosaxitoxin (neoSTX), decarbamoyl-neoSTX, decarbamoylsaxitoxin (dcSTX), and gonyautoxin 5 (GTX5). Toxin purity was confirmed by the Certificate of Analysis provided for each toxin lot where other PSTs were undetectable except for a possible trace amount of STX in the neoSTX solution. The purity was measured by HPLC analysis followed by postcolumn oxidation and fluorescence detection, capillary electrophoresis coupled with UV absorbance detection, and LC in combination with chemiluminescence nitrogen detection. The stated uncertainties in the certified concentrations of each of the toxins ranged from 3.1 to 4.6%. Competition curves were conducted with toxin mixtures titrated with the ratio between the toxins held constant, and IC50 values were derived for comparison to the values predicted by equations described below. Reagents. Tritiated saxitoxin ([3H]STX) was from Amersham Pharmacia Biotech (Buckinghamshire, United Kingdom). STX dihydrochloride, neoSTX, dcSTX, and GTX5 were purchased from the Institute for Marine Biosciences, Canadian National Research Council (Halifax, Canada). All toxin dilutions were in 0.01 M acetic acid. Biological buffers and general chemicals were from Sigma (Sydney, Australia). Water was deionized (≈18 MΩ) with a Millipore system (Sydney, Australia). Saxiphilin from the xanthid crab Liomera tristis and rat brain membranes were prepared as previously described (30) and stored at -80 °C between uses. Assay Protocols. The L. tristis saxiphilin assay was conducted in high protein binding 96 well filtration plates using a protocol to be described in detail elsewhere (Robertson and Llewellyn, unpublished results) using a final protein concentration of 1.5 µg protein/mL. Rat brain Na channel binding of [3H]STX was measured using a microtiter plate method (22, 31) and 14 µg protein rat brain membrane/mL. In both assays, the final assay volume was 150 µL, the concentration of [3H]STX was 1.2 nM, and the background radioactivity was defined in samples containing 5 µM STX. The fraction of toxin binding proteins occupied was calculated after subtraction of background radioactivity and normalized to unity using control samples containing no competitor. Assay mixtures were prepared by sequentially adding stock solutions of reaction components (i.e., buffer, choline chloride or NaCl, toxin 1, toxins 2 and 3 as required, [3H]STX, and water) with the reaction initiated by adding the rat brain Na channels or L. tristis saxiphilin to achieve final assay volume. Stock solutions of the different toxins were varied to attain the desired ratio of toxins. All samples were tested in duplicate. Inhibition (%) and IC50 Prediction of Toxin Mixtures. 1. Binary Combinations. The behavior of two ligands with different affinities for a single population of binding proteins is illustrated below: k+2

Receptor.Toxin2 y\ z Toxin2 + Receptor + k -2

k+1

Toxin1 y\ z Receptor.Toxin1 k -1

The work of Tuck et al. (32) and Jones et al. (33) can be combined to predict the fraction of toxin-binding protein occupied for any combination of two toxins:

1

([

[toxin 1] toxin 1 IC50

])

] [ n1

+

n2

[toxin 2] toxin 2 IC50

(2)

[Toxin 1] and [toxin 2] are the concentrations of the two different toxins in nM, and the IC50 values for toxins 1 and 2 are those experimentally determined in control experiments using the toxins in isolation. n1 and n2 are the slope coefficients for the toxins in isolation. By definition, these should be one because they are from a single toxin acting upon a single binding protein in a reversible binding reaction (34). Equation 2 can then be used to generate the expected inhibition curve from a titration of a binary toxin mixture having a constant ratio. The IC50 of this theoretical curve is then calculated using a standard competition curve: fraction binding protein occupied )

1 1 + 10(logIC50/log[toxin])

(3)

2. Ternary Combinations. Adding a third toxin to a binary mixture with an affinity distinguishable from its partners is illustrated below:

The combined effect of the ternary mixture can be calculated after extending the denominator of eq 2 as follows: fraction binding protein occupied ) 1 [toxin 1] n1 [toxin 2] 1+ + toxin 1 IC50 toxin 2 IC50

([

] [

] [ n2

+

])

[toxin 3] toxin 3 IC50

n3

(4)

The new terms for toxin 3 are the same as for toxins 1 and 2 except that they are derived for toxin 3 in isolation. Equation 4 is used to generate the expected competition curve from a titration of three toxins kept at a constant ratio, and the IC50 is calculated using eq 3. Data Analysis and Modeling. IC50 values and their 95% confidence limits from experimental data were derived using the single site curve equation of Graphpad Prism (Ver 4.0, Graphpad Software, San Diego, CA), with fitting constrained to a slope of 1 and a baseline of 0 (i.e., eq 3 above). When more than one toxin was present, the value for toxin concentration used on the ordinate axis was the cumulative concentration of all toxins present. Modeling using eqs 2-4 in combination was achieved using an Excel (Microsoft, Seattle WA) spreadsheet incorporating a curvefitting routine using sum-of-squares minimization (35).

Results As elaborated in the Discussion, historical knowledge of the affinity of the Na channel and various STX-binding proteins for a variety of PSTs was used to identify two unrelated PSTbinding proteins along with a suite of toxins to test the predictions generated by eqs 2 and 4. The rat brain Na channel and a saxiphilin from the xanthid crab L. tristis possess markedly different affinities for STX itself using similar conditions to those used here, those being KD values of 0.8-1 and 0.06-0.12 nM, respectively (30, 36). STX was chosen as the primary toxin in the mixtures because its tritiated form was the radioligand in the assay. The second toxin chosen was the sulfated PST GTX5, which is weakly bound by both of these proteins but not so weak that excessive amounts of toxin would

Toxin Mixtures in Binding Assays

Chem. Res. Toxicol., Vol. 19, No. 5, 2006 663

Table 1. Inhibition Curve Parameters for Each Individual Toxin in Both the Rat Brain Na Channel and L. tristis Saxiphilin Assaysa binding protein rat brain Na channel toxin

IC50

STX 3.0 neoSTX NDb dcSTX 10.2 GTX5 76.2

L. tristis saxiphilin

95% confidence r2 of fit limits 1.5-6.0 NDb 7.4-14.0 51.9-165

IC50

0.94 0.9 NDb 8.9 0.98 NDb 0.92 199

95% confidence r2 of fit limits 0.8-1.0 6.8-11.6 NDb 147-269

0.99 0.99 NDb 0.96

a In all cases, curve fitting was constrained so that the slope equalled unity. b ND, not determined because this particular toxin was not used for this toxin binding protein (see Results section).

be needed to fully displace the [3H]STX. The toxin with an affinity likely to be in the middle of these two affinities was different for rat brain Na channels and L. tristis saxiphilin. neoSTX has an affinity for L. tristis saxiphilin between that of STX and GTX5 (30) whereas neoSTX was equipotent with STX in displacing [3H]STX from rat brain Na channels (37) and it would be difficult to distinguish from STX itself. dcSTX, however, has an affinity between that of the rat brain Na channel for STX and GTX5 (37). The affinity of L. tristis saxiphilin for dcSTX was significantly weaker than GTX5, and large amounts of toxin would be needed to obtain full inhibition curves against [3H]STX. The above proved true as can be seen from the IC50 values for the toxins in isolation (Table 1). With the rat brain Na channel assay, dcSTX and GTX5 were, respectively, 3.4- and 25-fold weaker than STX in competing with 1.2 nM [3H]STX. With L. tristis saxiphilin, neoSTX and GTX5 were 10- and 221fold less active than STX in displacing 1.2 nM [3H]STX. In all inhibition curves, both here and in later curves with binary and ternary mixtures, the poorest quality curve fit to experimental data was an r2 of 0.92 with several curves attaining an r2 of 0.99. Predicted IC50 values of binary mixtures from eq 2 are depicted in Figure 1A, showing the change in IC50 of a mixture where one toxin has an IC50 of 1 nM and the second toxin has an IC50 ranging from 10- to 100-fold less than that of toxin 1. There is a sigmoidal transition between the IC50 values of the two toxins when the abscissa axis has been log-transformed. Depending on the potency of the second toxin, a 100-1000fold excess of the weaker over the more potent toxin is required for the mixture’s IC50 to attain that of the weaker toxin. The predicted IC50 values generated by using eq 4 for ternary mixtures are depicted in Figure 1B, showing the change in IC50 of a mixture when one toxin has an IC50 of 1 nM and the second and third toxins have IC50 values of 0.01 and 100 nM. The figure also shows the change in IC50 over 6 orders of magnitude with regards to the concentration ratio of the second and third toxins relative to toxin 1. Again, when the abscissa is log-transformed, a sigmoidal surface encompasses the transitions from the IC50 of one toxin to the other two toxins. Table 2 depicts the IC50 values ((95% confidence limits) generated experimentally using binary mixtures of GTX5 and STX in both binding assays employed here. The experimentally determined IC50 values were virtually identical to the predicted IC50 values from eq 2. Table 2 also lists the IC50 values generated using mixtures of GTX5, dcSTX, and STX with the rat brain Na channel radioreceptor assay and for mixtures of GTX5, neoSTX, and STX with L. tristis saxiphilin. Experimentally determined IC50 values were again virtually identical to the predicted IC50 values from eq 4. Examples of the

Figure 1. Predicted IC50 values of binary (A) and ternary (B) toxin mixtures. In panel A, a toxin with an IC50 of 1 nM is mixed with toxins having the depicted IC50 in ratios as per the X-axis. In panel B, a toxin with an IC50 of 1 nM is mixed with a second and third toxin having IC50 values of 0.01 and 100 nM, respectively.

inhibition curves obtained with ternary mixtures are provided in Figure 2. When all of the predicted and actual IC50 values for binary and ternary mixtures are pooled (Table 2), the regression lines yielded were y ) 0.90x + 1.0 (r2 ) 0.99, n ) 20) and y ) 0.96x - 0.6 (r2 ) 0.97, n ) 20) for L. tristis saxiphilin and rat brain Na channels, respectively (Graphpad Prism, Version 4.0, Graphpad Software).

Discussion Exposure to environmental or food contaminants rarely involves a single bioactive compound. A case in point are the PSTs, which usually contaminate seafood or water as a suite of related chemicals. How these toxin mixtures behave pharmacologically must be understood to confidently interpret both bioassay results and subsequent toxicological effects triggered by receptor binding. To gain this understanding, the most controlled experimental situation first needs to be created. Here, this was attained by using high quality, certified toxins of known affinity for well-understood toxin binding proteins, namely, the rat brain Na channel and a saxiphilin (30, 38). With regards to the rat brain Na channel, Usup and colleagues (37) are one of the few to measure the affinity of a large series of different PSTs in rat brain Na channel radioreceptor assays. Of the known

664 Chem. Res. Toxicol., Vol. 19, No. 5, 2006

Llewellyn

Table 2. Agreement between Predicted and Actual IC50 Values ( 95% Confidence Limits (CL) Using L. tristis Saxiphilin and Rat Brain Na Channelsa ratio neoSTX or dcSTX to STX 0

1

10

100

1000

0

1

10

100

1000

95% CL GTX5 to STX

IC50 predicted experimental

IC50 lower

IC50 upper

1 10 100 1000 1 10 100 1000 1 10 100 1000 1 10 100 1000 1 10 100 1000

L. tristis saxiphilin 1.8 2.2 9.5 10.8 62.6 67.1 163.0 154.4 2.4 2.7 9.4 9.6 59.1 54.7 160.2 135.5 5.4 5.6 9.2 8.3 40.5 38.9 139.2 121.2 8.3 7.2 9.0 6.4 15.6 13.9 63.4 64.5 8.8 8.3 8.9 6.8 9.7 8.4 16.9 14.1

1.7 7.5 60.3 107.4 1.9 7.3 43.2 98.1 4.6 5.8 25.0 93.9 5.9 5.3 10.1 34.7 5.6 5.7 6.3 10.7

2.9 15.5 74.8 221.9 3.9 12.7 69.4 187.1 6.8 11.9 60.4 156.3 8.6 7.8 19.2 119.6 12.1 8.1 11.2 18.4

1 10 100 1000 1 10 100 1000 1 10 100 1000 1 10 100 1000 1 10 100 1000

rat brain Na channel 5.8 5.3 23.7 19.9 61.4 56.4 74.4 74.5 6.7 5.5 21.4 15.6 58.6 46.9 73.9 78.9 9.0 7.6 14.7 13.4 42.8 45.7 70.2 61.2 10.0 9.6 10.8 8.8 17.9 15.8 48.1 43.4 10.2 8.8 10.3 12.3 11.1 11.1 18.3 22.1

3.7 13.7 35.0 52.9 3.5 10.9 36.0 44.8 4.8 6.9 32.5 38.2 6.8 6.2 11.8 27.8 2.8 8.8 8.7 16.7

7.5 28.8 90.8 105.0 8.8 22.5 61.2 138.8 12.1 26.3 64.2 98.0 13.6 12.4 21.2 67.7 27.9 17.1 14.0 29.3

a Note that for L. tristis saxiphilin, the second toxin was neoSTX whereas it was dcSTX for the rat brain Na channel.

saxiphilins (30, 38), the isoform from the xanthid crab L. tristis possessed an affinity series for different PSTs that encompassed some of the currently available certified reference PSTs. Marrying this information together, the toxins chosen to explore the effects of mixing PSTs on the rat brain Na channel receptor assay were STX, neoSTX, and GTX5, whereas STX, dcSTX, and GTX5 were chosen for use with L. tristis saxiphilin. These two proteins also possess STX affinities almost an order of magnitude different (30, 36), further testing the ability of eqs 2 and 4 to predict the behavior of a mixture under different conditions. When modeling the behavior of a binary mixture, several phenomena are predicted. When concentrations of two toxins are equal, the mixture’s IC50 is forecast to shift less than 2-fold from the affinity of the most potent toxin even when one of the mixture partners possesses an IC50 100-fold weaker than the more active partner (Figure 1A). This predicted dominance of a binary mixture’s IC50 by the most potent toxin is emphasized further by eq 2’s forecast that 1000 times as much of a 100fold less potent toxin would be needed for the mixture’s IC50 to attain the IC50 of the less potent toxin, a prediction confirmed

by experimental data (Table 2). These experimental data confirm the above predictions that the most potent toxin will strongly dominate the behavior of a binary mixture. In Table 2, it can be seen that the experimentally determined IC50 of a mixture, where the concentration of GTX5 is 1000-fold higher than that of STX, is just over 100 nM and far from the control IC50 for GTX5 in isolation (199 nM, Table 1). The strong control over the potency of a mixture by the most active toxin raises the spectre that highly toxic samples may not be identified by analytical methods that separate individual toxins. Highly potent PSTs may be present in some samples below their analytical detection limits, but the mixture may be highly potent, reflecting the dominant activity of high affinity toxin or toxins. For example, HPLC coupled with fluorescence monitoring detects and quantifies individual PSTs (26, 27). Derived toxicity values are obtained by conversion of individual toxin amounts to STX equivalents using factors derived from mouse lethality studies with each toxin. This technique detects distinct toxins with different sensitivities (26). In the abovedescribed scenario, namely, a highly potent PST being present below the method’s detection limit, the bioassay would indicate that the toxicity is measurably greater than what the HPLC approach indicates. This has occurred (6, 10, 39), and differences between HPLC and various bioassays, biomolecular assays in particular, may not result from experimental error or biological variation but from the phenomenon observed here. To illustrate this point further, the four PSTs used in this study have different detection limits by HPLC fluorescent detection. For example, in one application of this method, the toxins used in this study had detection limits of 25, 41, 52, and 110 fmol for STX, dcSTX, neoSTX, and GTX5, respectively (26). The conditions used in a separate study more readily detected GTX5 than neoSTX, and neoSTX had a detection limit almost 5-fold less sensitive than STX (40). To highlight the above point and using the PSTs used here as an example, it is possible that neoSTX, one of the most potent PSTs in several bioassays (26, 41), may be present in an extract at a concentration beyond the capability of an analytical method sensitivity and yet may strongly influence the apparent total toxicity of a mix of PSTs that includes neoSTX. In this case, the analytical method would underestimate true toxicity. Consideration must therefore be given to analytical standards, which may contain trace impurities either from their initial purification or from degradation during storage. If this impurity is more potent than the standard, specific activity determinations of the standard will reflect that of the minor, yet highly potent, toxic component. Bearing in mind the very close agreement between the experimental and the predicted IC50 values for toxin mixtures that eqs 2 and 4 generate for binary and ternary mixtures, a general equation for predicting the fraction of toxin-binding protein occupied by mixtures of more than three toxins can be proposed

fraction occupied ) 1+

[

([

1

] [ ] [

[toxin 1] toxin 1 IC50

[toxin 3] toxin 3 IC50

n3

n1

] ])

[toxin 2] + toxin 2 IC50

+ ... +

[toxin x] toxin x IC50

n2

(5)

+

nz

The additional terms are the concentrations, IC50 values, and slope coefficients for each new toxin added to the mixture. To predict the IC50 of very complex mixtures, it is necessary to have the IC50 values for all of the toxins in the mixture to insert

Toxin Mixtures in Binding Assays

Chem. Res. Toxicol., Vol. 19, No. 5, 2006 665

Figure 2. Examples of the inhibition of equilibrium levels of [3H]STX bound by L. tristis saxiphilin (A) and rat brain Na channel (B) by the ternary toxin mixtures depicted on the graphs. Examples have been selected that span the breadth of obtained potencies but are separate enough to enable visual distinction of the inhibition curves and data points. Data points are the means ( SE of duplicate measurements. Curves are drawn using eq 3, inserting the IC50 values listed in Table 2 for the respective toxin mixture. Table 3. Predicted Change in the IC50 of a Mixture of Toxins with an Increasing Number of Toxins of an Equal Concentration and of Decreased Potency individual toxin IC50 values

proportion of each toxin in mixture (%)

predicted IC50

1 1, 2 1, 2, 5 1, 2, 5, 10 1, 2, 5, 10, 20 1, 2, 5, 10, 20, 50 1, 2, 5, 10, 20, 50, 100 1, 2, 5, 10, 20, 50, 100, 200

100 50 33.3 25 20 16.7 14.3 12.5

1 1.3 1.8 2.2 2.7 3.2 3.8 4.2

into eq 5. To overcome this technical hurdle and to gain some insight into how very complex mixtures of more than three toxins should behave, we can consider a theoretical situation. Table 3 lists the change in IC50 as toxins of decreasing potency are added to a mixture at a concentration equal to all other toxins. It can be seen that the IC50 of the most complex mixture

is only 4-fold less than the most potent toxin (4.2 nM) even though >75% of the toxins possess IC50 values weaker than the cumulative IC50. Taking this most complex situation further, that is, the eight toxin mixture, increasing the weakest toxin’s concentration to 10 times greater than each of the other seven toxins so that it comprises almost 60% of the total toxin profile, the mixture’s IC50 only changes to 8.8 nM, far from the dominant toxin’s IC50 of 200 nM. Increasing its concentration further to 100-fold more than each of the other toxins (i.e., approximately 94% of the total toxin profile) changes the mixture’s IC50 to 45 nM and by a 1000-fold (i.e., slightly more than 99% of the total toxin profile) increases the cumulative IC50 to 146 nM, values quite distinct from the weakest toxin’s IC50 of 200 nM. A toxin mixture’s cumulative IC50 can therefore be very resilient to significant changes in the concentration of the weaker toxins. PSTs are not the only seafood or environmental toxicant for which the behavior of mixtures of toxins that behave in the

666 Chem. Res. Toxicol., Vol. 19, No. 5, 2006

same manner is a consideration. Enzyme, ELISA, and receptor binding assays are becoming common and sometimes standard techniques, for monitoring seafood and drinking water toxins (42). This model may also explain the behavior of protein phosphatase inhibition assay (43, 44), which has been used to quantitated microcystins as well as okadaic acid and its analogues (45). Brevetoxins, which cause neurotoxic shellfish poisoning, bind to site 5 on the Na channel and have been detected using a radioreceptor assay (46). Domoic acid and its analogues antagonize the ionotropic glutamate receptor and have been measured with a radioreceptor binding assay (47). It remains to be seen whether this model can be extended to other biomolecule-dependent PST assays such as ELISAs or cytotoxicity assays when used to measure toxins that may occur in chemical mixtures and act identically. In the longer term, the relationship in eq 4 might further explain the behavior of toxin mixtures in whole animal bioassays such as the mouse lethality test, which is widely used for PST monitoring (23) and other toxicological events. A necessary event in inducing toxicity is for a toxicant to bind to a molecular target, and understanding the behavior of toxin mixtures at the receptor level is critical to understanding all toxicological events whether it be a human consuming tainted seafood (2-4), agricultural livestock drinking PST-contaminated water (48), or mice being injected for PST monitoring (23). Apart from the toxin mixture in the ingested material, a mixture may ensue from the ingested or administered toxin being metabolically converted to an analogue with significantly different potency. Metabolic modification of PSTs during intoxication has been observed in human poisonings where the toxin profile of the tainted seafood is transformed by the victim (3, 49). Harking back to the above discussion with regards to HPLC analysis of PSTs, this model may also be used to identify new PSTs or other toxins that affect PST binding when HPLC analysis and predicted bioassay measurement markedly disagree. The PSTs are bound by a number of naturally occurring proteins including the Na channel, their pharmacological receptor, the transferrin-like saxiphilins, and a blood protein from pufferfish. In nature, these proteins usually encounter a mixture of PSTs, which then act on the Na channel to induce toxicity or may be ligated by hydrophilic proteins that circulate in their blood, possibly preventing them from exerting their toxicity. Rather than only understanding how these toxins act in isolation then, it is critical to fully understand the natural situation in which these toxins usually find themselves, that is, of being in a mixture. Acknowledgment. I am grateful to Andrew Negri and Jason Doyle of the Australian Institute of Marine Science and Cedric Robillot of Cleveland Biosensors Pty. Ltd. for their critical comments.

References (1) Oshima, Y. (1995) Postcolumn derivatization HPLC methods for paralytic shellfish poisons. In Manual on Harmful Marine Microalgae (Hallegraeff, G., Anderson, D., and Cembella, A., Eds.) pp 81-94, UNESCO, Paris. (2) Anonymous (2002) Neurologic illness associated with eating Florida pufferfish, 2002. MMWR Morb. Mortal Weekly Rep. 51, 321-323. (3) Llewellyn, L. E., Dodd, M. J., Robertson, A., Ericson, G., de Koning, C., and Negri, A. P. (2002) Post-mortem analysis of samples from a human victim of a fatal poisoning caused by the xanthid crab, Zosimus aeneus. Toxicon 40, 1463-1469. (4) Lehane, L. (2001) Paralytic shellfish poisoning: A potential public health problem. Med. J. Aust. 175, 29-31. (5) Catterall, W. A., Goldin, A. L., and Waxman, S. G. (2003) International Union of Pharmacology. XXXIX. Compendium of voltage-gated ion channels: sodium channels. Pharmacol. ReV. 55, 575-578.

Llewellyn (6) Llewellyn, L. E., Negri, A. P., Doyle, J., Baker, P. D., Beltran, E. C., and Neilan, B. A. (2001) Radioreceptor assays for sensitive detection and quantitation of saxitoxin and its analogues from strains of the freshwater cyanobacterium, Anabaena circinalis. EnViron. Sci. Technol. 35, 1445-1451. (7) Chang, F. H., Anderson, D. M., Kulis, D. M., and Till, D. G. (1997) Toxin production of Alexandrium minutum (Dinophyceae) from the Bay of Plenty, New Zealand. Toxicon 35, 393-409. (8) Negri, A. P., Jones, G. J., Blackburn, S. I., Oshima, Y., and Onodera, H. (1997) Effect of culture and bloom development and of sample storage on paralytic shellfish poisons in the cyanobacterium Anabaena circinalis. J. Phycol. 33, 26-35. (9) Oshima, Y., Hasegawa, M., Yasumoto, T., Hallegraeff, G., and Blackburn, S. (1987) Dinoflagellate Gymnodinium catenatum as the source of paralytic shellfish toxins in Tasmanian shellfish. Toxicon 25, 1105-1111. (10) Llewellyn, L. E., Doyle, J., Jellett, J., Barrett, R., Alison, C., Bentz, C., and Quilliam, M. (2001) Measurement of paralytic shellfish toxins in molluscan extracts: Comparison of the microtitre plate saxiphilin and sodium channel radioreceptor assays with mouse bioassay, HPLC analysis and a commercially available cell culture assay. Food Addit. Contam. 18, 970-980. (11) Karunasagar, I., Oshima, Y., and Yasumoto, T. (1990) A toxin profile for shellfish involved in an outbreak of paralytic shellfish poisoning in India. Toxicon 28, 868-870. (12) Van Egmond, H., Speijers, G., and Van den Top, H. (1992) Current situation on worldwide regulations for marine phycotoxins. J. Nat. Toxins 1, 67-85. (13) Manger, R. L., Leja, L. S., Lee, S. Y., Hungerford, J. M., Kirkpatrick, M. A., Yasumoto, T., and Wekell, M. M. (2003) Detection of paralytic shellfish poison by rapid cell bioassay: Antagonism of voltage-gated sodium channel active toxins in vitro. J. AOAC Int. 86, 540-543. (14) Jellett, J. F., Marks, L. J., Stewart, J. E., Dorey, M. L., Watson-Wright, W., and Lawrence, J. F. (1992) Paralytic shellfish poison (saxitoxin family) bioassays: Automated endpoint determination and standardization of the in vitro tissue culture bioassay, and comparison with the standard mouse bioassay. Toxicon 30, 1143-1156. (15) Fairey, E. R., Edmunds, J. S., and Ramsdell, J. S. (1997) A cell-based assay for brevetoxins, saxitoxins, and ciguatoxins using a stably expressed c-fos-luciferase reporter gene. Anal. Biochem. 251, 129132. (16) Gallacher, S., and Birkbeck, T. H. (1992) A tissue culture assay for direct detection of sodium channel blocking toxins in bacterial culture supernates. FEMS Microbiol. Lett. 71, 101-107. (17) Kogure, K., Tamplin, M. L., Simidu, U., and Colwell, R. R. (1988) A tissue culture assay for tetrodotoxin, saxitoxin and related toxins. Toxicon 26, 191-197. (18) Lee, H. O., Cheun, B. S., Yoo, J. S., Watanabe, E., and Han, M. S. (2000) Application of a channel biosensor for toxicity measurements in cultured Alexandrium tamarense. J. Nat. Toxins 9, 341348. (19) Cheun, B. S., Loughran, M., Hayashi, T., Nagashima, Y., and Watanabe, E. (1998) Use of a channel biosensor for the assay of paralytic shellfish toxins. Toxicon 36, 1371-1381. (20) Cheun, B. S., Takagi, S., Hayashi, T., Nagashima, Y., and Watanabe, E. (1998) Determination of Na channel blockers in paralytic shellfish toxins and pufferfish toxins with a tissue biosensor. J. Nat. Toxins 7, 109-120. (21) Usleber, E., Dietrich, R., Burk, C., Schneider, E., and Martlbauer, E. (2001) Immunoassay methods for paralytic shellfish poisoning toxins. J. AOAC Int. 84, 1649-1656. (22) Doucette, G. J., Logan, M. M., Ramsdell, J. S., and Van Dolah, F. M. (1997) Development and preliminary validation of a microtiter platebased receptor binding assay for paralytic shellfish poisoning toxins. Toxicon 35, 625-636. (23) Llewellyn, L. E., and Doyle, J. (1998) A high-throughput, microtiter plate assay for paralytic shellfish poisons using the saxitoxin-specific receptor, saxiphilin. Anal. Biochem 261, 51-56. (24) Krishnan, G., Morabito, M. A., and Moczydlowski, E. (2001) Expression and characterization of Flag-epitope- and hexahistidinetagged derivatives of saxiphilin for use in detection and assay of saxitoxin. Toxicon 39, 291-301. (25) Sullivan, J. J., Wekell, M. W., and Kentala, L. L. (1985) Application of HPLC for the determination of PSP toxins in shellfish. J. Food Sci. 50, 26-29. (26) Oshima, Y. (1995) Postcolumn derivatization liquid-chromatographic method for paralytic shellfish toxins. J. AOAC Int. 78, 528-532. (27) Lawrence, J. F., Niedzwiadek, B., and Menard, C. (2004) Quantitative determination of paralytic shellfish poisoning toxins in shellfish using prechromatographic oxidation and liquid chromatography with fluorescence detection: interlaboratory study. J. AOAC Int. 87, 83-100.

Toxin Mixtures in Binding Assays (28) Negri, A., and Llewellyn, L. (1998) Comparative analyses by HPLC and the sodium channel and saxiphilin 3H-saxitoxin receptor assays for paralytic shellfish toxins in crustaceans and molluscs from tropical North West Australia. Toxicon 36, 283-298. (29) Dill, K., Lin, M., Poteras, C., Fraser, C., Hafeman, D. G., Owicki, J. C., and Olson, J. D. (1994) Antibody-antigen binding constants determined in solution-phase with the threshold membrane-capture system: Binding constants for anti-fluorescein, anti-saxitoxin, and antiricin antibodies. Anal. Biochem. 217, 128-138. (30) Llewellyn, L. E. (1997) Haemolymph protein in xanthid crabs: Its selective binding of saxitoxin and possible role in toxin bioaccumulation. Mar. Biol. 128, 599-606. (31) Llewellyn, L., Negri, A., and Quilliam, M. (2004) High affinity for the rat brain sodium channel of newly discovered hydroxybenzoate saxitoxin analogues from the dinoflagellate Gymnodinium catenatum. Toxicon 43, 101-104. (32) Tuk, B., Van Oostenbruggen, M. F., Herben, V. M. M., Mandema, J. W., and Danhof, M. (1999) Characterization of the pharmacodynamic interaction between parent drug and active metabolite in vivo: Midazolam and R-OH-midazolam. J. Pharmacol. Exp. Ther. 289, 1067-1074. (33) Jones, G., Wortberg, M., Kreissig, S. B., Bunch, D. S., Gee, S. J., Hammock, B. D., and Rocke, D. M. (1994) Extension of the fourparameter logistic model for ELISA to multianalye analysis. J. Immunol. Methods 177, 1-7. (34) Fersht, A. (1999) Structure and Mechanism in Protein Science: A Guide to Enzyme Catalysis and Protein Folding, W. H. Freeman and Company, New York. (35) Bowen, W. P., and Jerman, J. C. (1995) Nonlinear regression using spreadsheets. Trends Pharmacol. Sci. 16, 413-417. (36) Weigele, J. B., and Barchi, R. L. (1978) Analysis of saxitoxin binding in isolated rat synaptosomes using a rapid filtration assay. FEBS Lett. 91, 310-314. (37) Usup, G., Leaw, C. P., Cheah, M. Y., Ahmad, A., and Ng, B. K. (2004) Analysis of paralytic shellfish poisoning toxin congeners by a sodium channel receptor binding assay. Toxicon 44, 37-43. (38) Llewellyn, L. E., Bell, P. M., and Moczydlowski, E. G. (1997) Phylogenetic survey of soluble saxitoxin-binding activity in pursuit of the function and molecular evolution of saxiphilin, a relative of transferrin. Proc. R. Soc. London., Ser. B 264, 891-902.

Chem. Res. Toxicol., Vol. 19, No. 5, 2006 667 (39) Powell, C. L., and Doucette, G. J. (1999) A receptor binding assay for paralytic shellfish poisoning toxins: Recent advances and applications. Nat. Toxins 7, 393-400. (40) Sullivan, J. J., Wekell, M. W., and Kentala, L. L. (1985) Application of HPLC for the determination of PSP toxins in shellfish. J. Food Sci. 50, 26-29. (41) Jellett, J. F., Stewart, J. E., and Laycock, M. V. (1995) Toxicological evaluation of saxitoxin, neosaxitoxin, gonyautoxin-II, gonyautoxin-II plus III and decarbamoylsaxitoxin with the mouse neuroblastoma cell bioassay. Toxicol. in Vitro 9, 57-65. (42) Hungerford, J. M. (2005) Committee on Natural Toxins and Food Allergens. Marine and freshwater toxins. J. AOAC Int. 88, 299-313. (43) Oliveira, A. C., Magalhaes, V. F., Soares, R. M., and Azevedo, S. M. (2005) Influence of drinking water composition on quantitation and biological activity of dissolved microcystin (cyanotoxin). EnViron. Toxicol. 20, 126-130. (44) Della Loggia, R., Sosa, S., and Tubaro, A. (1999) Methodological improvement of the protein phosphatase inhibition assay for the detection of okadaic acid in mussels. Nat. Toxins 7, 387-391. (45) Robillot, C., and Hennion, M. C. (2004) Issues arising when interpreting the results of the protein phosphatase 2A inhibition assay for the monitoring of microcystins. Anal. Chim. Acta 512, 339-346. (46) Whitney, P. L., Delgado, J. A., and Baden, D. G. (1997) Complex behavior of marine animal tissue extracts in the competitive binding assay of brevetoxins with rat brain synaptosomes. Nat. Toxins 5, 193200. (47) Van Dolah, F. M., Finley, E. L., Haynes, B. L., Doucette, G. J., Moeller, P. D., and Ramsdell, J. S. (1994) Development of rapid and sensitive high throughput pharmacologic assays for marine phycotoxins. Nat. Toxins 2, 189-196. (48) Negri, A. P., Jones, G. J., and Hindmarsh, M. (1995) Sheep mortality associated with paralytic shellfish poisons from the cyanobacterium Anabaena circinalis. Toxicon 33, 1321-1329. (49) Gessner, B. D., Bell, P., Doucette, G. J., Moczydlowski, E., Poli, M. A., Van Dolah, F., and Hall, S. (1997) Hypertension and identification of toxin in human urine and serum following a cluster of musselassociated paralytic shellfish poisoning outbreaks. Toxicon 35, 711-722.

TX050277I