Phylogeny and Toxin Profile of Freshwater Pufferfish (Genus Pao) Collected from 2 Different Regions in Cambodia

The species classification of Cambodian freshwater pufferfish is incomplete and confusing, and scientific information on their toxicity and toxin profile is limited. In the present study, to accumulate information on the phylogeny and toxin profile of freshwater pufferfish, and to contribute to food safety in Cambodia, we conducted simultaneous genetic-based phylogenetic and toxin analyses using freshwater pufferfish individuals collected from Phnom Penh and Kratie (designated PNH and KTI, respectively). Phylogenetic analysis of partial sequences of three mitochondrial genes (cytochrome b, 16S rRNA, and cytochrome c oxidase subunit I) determined for each fish revealed that PNH and KTI are different species in the genus Pao (designated Pao sp. A and Pao sp. B, respectively). A partial sequence of the nuclear tributyltin-binding protein type 2 (TBT-bp2) gene differentiated the species at the amino acid level. Instrumental analysis of the toxin profile revealed that both Pao sp. A and Pao sp. B possess saxitoxins (STXs), comprising STX as the main component. In Pao sp. A, the toxin concentration in each tissue was extremely high, far exceeding the regulatory limit for STXs set by the Codex Committee, whereas in Pao sp. B, only the skin contained high toxin concentrations. The difference in the STX accumulation ability between the two species with different TBT-bp2 sequences suggests that TBT-bp2 is involved in STX accumulation in freshwater pufferfish.


Introduction
Pufferfish of the family Tetraodontidae contain tetrodotoxin (TTX) and/or saxitoxins (STXs), but the toxin ratio differs depending on the genus or species. Marine pufferfish of the genus Takifugu inhabiting

Phylogeny
A total of 16 freshwater pufferfish were sampled from small lakes near Phnom Penh (PNH) and Kratie (KTI) (Figures 1 and 2, and Table A1). Partial sequences of the three mitochondrial genes, CYTB, 16S rRNA, and COI, were determined for each sample and phylogenetically analyzed for taxonomic identification. Phylogenetic analysis of 870-bp CYTB sequences showed that the PNH and KTI sequences each formed a tight cluster together with Pao cochinchinensis (AB741980.1) or Pao baileyi (AB741978.1) (Figure 3). Similar results were obtained from analyses of concatenated datasets of CYTB+16S rRNA (555 bp) and CYTB+COI (621 bp) ( Figures A1 and A2). Considering the genetic variation of P. suvattii in Figure 3, each cluster seems to represent a single species, though the intraspecific variation in the genus Pao might be unexpectedly large (see Pao leiurus and P. cochinchinensis in Figures 3, A1, and A2). Hereafter, we thus refer to the samples from PNH and KTI as Pao sp. A and Pao sp. B, respectively.

Phylogeny
A total of 16 freshwater pufferfish were sampled from small lakes near Phnom Penh (PNH) and Kratie (KTI) (Figures 1 and 2, and Table A1). Partial sequences of the three mitochondrial genes, CYTB, 16S rRNA, and COI, were determined for each sample and phylogenetically analyzed for taxonomic identification. Phylogenetic analysis of 870-bp CYTB sequences showed that the PNH and KTI sequences each formed a tight cluster together with Pao cochinchinensis (AB741980.1) or Pao baileyi (AB741978.1) (Figure 3). Similar results were obtained from analyses of concatenated datasets of CYTB+16S rRNA (555 bp) and CYTB+COI (621 bp) ( Figures A1 and A2). Considering the genetic variation of P. suvattii in Figure 3, each cluster seems to represent a single species, though the intraspecific variation in the genus Pao might be unexpectedly large (see Pao leiurus and P. cochinchinensis in Figures 3, A1, and A2). Hereafter, we thus refer to the samples from PNH and KTI as Pao sp. A and Pao sp. B, respectively.   A 492 to 778 bp region of the nuclear rhodopsin gene was directly sequenced from the polymerase chain reaction (PCR) product of each sample. As no double peaks were observed in the Sanger sequence electropherograms, the sequenced regions were considered to be homogeneous in all the samples. Furthermore, there was no sequence variation between any pair of samples. Compared with sequences of other Pao species in the database, all the mutations found were synonymous, and the amino acid sequence was identical among the Pao species examined.  A 492 to 778 bp region of the nuclear rhodopsin gene was directly sequenced from the polymerase chain reaction (PCR) product of each sample. As no double peaks were observed in the Sanger sequence electropherograms, the sequenced regions were considered to be homogeneous in all the samples. Furthermore, there was no sequence variation between any pair of samples. Compared with sequences of other Pao species in the database, all the mutations found were synonymous, and the amino acid sequence was identical among the Pao species examined.
A 298 to 536 bp region of the nuclear TBT-bp2 gene was determined for all the samples except KTI-3 by direct sequencing of the PCR products. We observed double peaks at several nucleotide positions in the Sanger sequence electropherograms of almost all the samples, indicating that the TBT-bp2 gene had multiple alleles (Table A2). Allele types were apparently similar between samples of Pao sp. A and Pao sp. B. Compared with the allele types found in Pao sp. A and Pao sp. B, there was no sample in which the allele types could be represented as a hybridization of Pao sp. A and Pao sp.
B. An ML phylogenetic analysis clearly separated the samples into PNH and KTI, namely Pao sp. A and Pao sp. B, as in the case of the mitochondrial genes ( Figure 4). Amino acid sequences also differed between Pao sp. A and Pao sp. B. Note that although we tried to obtain longer sequences, PCR amplifications were unsuccessful, probably due to the presence of microsatellite-like sequences in neighboring introns.  A 298 to 536 bp region of the nuclear TBT-bp2 gene was determined for all the samples except KTI-3 by direct sequencing of the PCR products. We observed double peaks at several nucleotide positions in the Sanger sequence electropherograms of almost all the samples, indicating that the TBT-bp2 gene had multiple alleles (Table A2). Allele types were apparently similar between samples of Pao sp. A and Pao sp. B. Compared with the allele types found in Pao sp. A and Pao sp. B, there was no sample in which the allele types could be represented as a hybridization of Pao sp. A and Pao sp. B. An ML phylogenetic analysis clearly separated the samples into PNH and KTI, namely Pao sp. A and Pao sp. B, as in the case of the mitochondrial genes ( Figure 4). Amino acid sequences also differed between Pao sp. A and Pao sp. B. Note that although we tried to obtain longer sequences, PCR amplifications were unsuccessful, probably due to the presence of microsatellite-like sequences in neighboring introns.

Toxin Profile
Toxins were extracted from the skin, muscle, liver, and gonads of each freshwater pufferfish. Analysis of the extracts by high-performance liquid chromatography with post-column fluorescence derivatization (HPLC-FLD) for STXs [12,33] revealed that the tissues of each individual, except for one or two tissues of KTI-1 and KTI-2, contained STXs comprising STX as the main component and decarbamoylSTX (dcSTX) as a minor component (typical chromatograms are shown in Figure A3). NeoSTX was also contained in some tissues as a minor component, but no other known STX components, such as C toxins and gonyautoxins (GTXs), were detected. When the same extracts were analyzed for TTX by liquid chromatography tandem mass spectrometry (LC-MS/MS) [12,34], no TTX was detected at all.

Toxin Profile
Toxins were extracted from the skin, muscle, liver, and gonads of each freshwater pufferfish. Analysis of the extracts by high-performance liquid chromatography with post-column fluorescence derivatization (HPLC-FLD) for STXs [12,33] revealed that the tissues of each individual, except for one or two tissues of KTI-1 and KTI-2, contained STXs comprising STX as the main component and decarbamoylSTX (dcSTX) as a minor component (typical chromatograms are shown in Figure A3). NeoSTX was also contained in some tissues as a minor component, but no other known STX components, such as C toxins and gonyautoxins (GTXs), were detected. When the same extracts were analyzed for TTX by liquid chromatography tandem mass spectrometry (LC-MS/MS) [12,34], no TTX was detected at all.
The toxin concentration (converted value to mg STX·diHCl eq/kg) in each tissue from the pufferfish is shown in Figure 5 and Table A1. The toxin concentrations of Pao sp. A were generally high and far above the regulatory limit for STX (0.8 mg STX·diHCl eq/kg) in all tissues. The mean toxin concentration in the ovary (58.7 mg STX·diHCl eq/kg) was the highest, followed by the skin (40.8 mg STX·diHCl eq/kg). The mean toxin concentration in the liver (23.1 mg STX·diHCl eq/kg) was lower than that in the skin, although in some individuals, the toxin concentration in the liver exceeded that in the skin. The mean toxin concentration in the muscle (22.3 mg STX·diHCl eq/kg) was the lowest, but still almost 28-fold higher than the regulatory limit. In contrast, the toxin concentrations of Pao sp. B were generally much lower than those of Pao sp. A. The mean toxin concentration in the skin (6.0 mg STX·diHCl eq/kg) was the highest, followed by the ovary (1.3 mg STX·diHCl eq/kg). The mean toxin concentrations in the testis, liver, and muscle were below the regulatory limit (0.5-0.6 mg STX·diHCl eq/kg), but some individuals had concentrations that slightly exceeded the regulatory limit.  The toxin concentration (converted value to mg STX·diHCl eq/kg) in each tissue from the pufferfish is shown in Figure 5 and Table A1. The toxin concentrations of Pao sp. A were generally high and far above the regulatory limit for STX (0.8 mg STX·diHCl eq/kg) in all tissues. The mean toxin concentration in the ovary (58.7 mg STX·diHCl eq/kg) was the highest, followed by the skin (40.8 mg STX·diHCl eq/kg). The mean toxin concentration in the liver (23.1 mg STX·diHCl eq/kg) was lower than that in the skin, although in some individuals, the toxin concentration in the liver exceeded that in the skin. The mean toxin concentration in the muscle (22.3 mg STX·diHCl eq/kg) was the lowest, but still almost 28-fold higher than the regulatory limit. In contrast, the toxin concentrations of Pao sp. B were generally much lower than those of Pao sp. A. The mean toxin concentration in the skin (6.0 mg STX·diHCl eq/kg) was the highest, followed by the ovary (1.3 mg STX·diHCl eq/kg). The mean toxin concentrations in the testis, liver, and muscle were below the regulatory limit (0.5-0.6 mg STX·diHCl eq/kg), but some individuals had concentrations that slightly exceeded the regulatory limit.
The total amount of toxin per individual is shown in Figure 6. As described above, the toxin concentration in each tissue was much higher in Pao sp. A than in Pao sp. B, but due to the larger individual size of Pao sp. B over Pao sp. A (Table A1), the total toxin amount of KTI-3-5 (172-417 µg STX·diHCL eq/individual) was generally comparable to that in Pao sp. A individuals (122-463 µg STX·diHCL eq/individual) other than PNH-1, in which the toxin concentration in the skin was exceptionally low. The toxin distribution profile in the body, however, differed between Pao sp. A and Pao sp. B. In Pao sp. A, in addition to the skin, the amount of toxin in the muscle and ovary was high, whereas in Pao sp. B, most of the toxin amount in the body was accounted for by the skin, irrespective of the sex.
The toxin composition in each tissue per species is shown in Figure 7. In Pao sp. B, the toxin concentrations in the muscle, liver, and gonads were very low, and neoSTX concentrations were below the limit of quantification. Other than this, however, there was no considerable difference between Pao sp. A and Pao sp. B. In each tissue, most of the toxins were accounted for by STX. In the whole body, both Pao sp. A and Pao sp. B contained 6% to 7% neoSTX and 3% to 4% dcSTX as the minor components, in addition to the main component STX (~90%). The total amount of toxin per individual is shown in Figure 6. As described above, the toxin concentration in each tissue was much higher in Pao sp. A than in Pao sp. B, but due to the larger individual size of Pao sp. B over Pao sp. A (Table A1), the total toxin amount of KTI-3-5 (172-417 µg STX·diHCL eq/individual) was generally comparable to that in Pao sp. A individuals (122-463 µg STX·diHCL eq/individual) other than PNH-1, in which the toxin concentration in the skin was exceptionally low. The toxin distribution profile in the body, however, differed between Pao sp. A and Pao sp. B. In Pao sp. A, in addition to the skin, the amount of toxin in the muscle and ovary was high, whereas in Pao sp. B, most of the toxin amount in the body was accounted for by the skin, irrespective of the sex.   The total amount of toxin per individual is shown in Figure 6. As described above, the toxin concentration in each tissue was much higher in Pao sp. A than in Pao sp. B, but due to the larger individual size of Pao sp. B over Pao sp. A (Table A1), the total toxin amount of KTI-3-5 (172-417 µg STX·diHCL eq/individual) was generally comparable to that in Pao sp. A individuals (122-463 µg STX·diHCL eq/individual) other than PNH-1, in which the toxin concentration in the skin was exceptionally low. The toxin distribution profile in the body, however, differed between Pao sp. A and Pao sp. B. In Pao sp. A, in addition to the skin, the amount of toxin in the muscle and ovary was high, whereas in Pao sp. B, most of the toxin amount in the body was accounted for by the skin, irrespective of the sex.  concentrations in the muscle, liver, and gonads were very low, and neoSTX concentrations were below the limit of quantification. Other than this, however, there was no considerable difference between Pao sp. A and Pao sp. B. In each tissue, most of the toxins were accounted for by STX. In the whole body, both Pao sp. A and Pao sp. B contained 6% to 7% neoSTX and 3% to 4% dcSTX as the minor components, in addition to the main component STX (~90%).

Discussion
The present study revealed that freshwater pufferfish collected from two regions of Cambodia, Phnom Penh and Kratie, are different species belonging to the genus Pao (Pao sp. A and Pao sp. B, respectively), and both possess STXs, with Pao sp. A having high toxicity and Pao sp. B having low toxicity.
The CYTB analysis suggests that Pao sp. A is the same species as P. cochinchinensis (AB741980.1), whereas CYTB sequences of other P. cochinchinensis specimens (AP011925.1, JQ681922.1) are distantly related to Pao sp. A (Figure 3), which makes it difficult to identify species with DNA barcoding methods. According to Igarashi et al. [35], at least one of the CYTB sequences is likely an error resulting from the misidentification of specimens. Likewise, Pao sp. B is closely related to P. baileyi as well as to P. abei and P. leiurus (KF667490.1, GU057266.1) (Figure 3), and could be the same species as one of these. The phylogenetic positions of P. leiurus, however, largely differ among specimens ( Figure 3). As discussed in Igarashi et al. [35], these situations are most likely due to similarities in the morphologic characteristics, geographic distribution, and ecology of the Pao species. Our mitochondrial gene analyses indicated the presence of actual and/or potential genetic boundaries of species (i.e., species-level clusters in the phylogenetic trees), suggesting that an accumulation of genetic information could contribute to resolve the taxonomic status and relationships among Pao species. On the other hand, if crossbreeding occurred in the Pao species, an introgression of the mitochondrial genome might have caused such discordance between morphologic species identification and mitochondrial DNA phylogeny. Crossbreeding is frequently observed in the marine Takifugu pufferfish species (e.g., Takahashi et al., 2017 [36]), and an introgression of the mitochondrial genome from the freshwater Rhinogobius goby species was reported [37]. As demonstrated by the analysis of the TBT-bp2 gene, this possibility should be carefully considered in parallel to taxonomic reexamination, though no evidence was found for the presence of crossbreeding between Pao sp. A and Pao sp. B.

Discussion
The present study revealed that freshwater pufferfish collected from two regions of Cambodia, Phnom Penh and Kratie, are different species belonging to the genus Pao (Pao sp. A and Pao sp. B, respectively), and both possess STXs, with Pao sp. A having high toxicity and Pao sp. B having low toxicity.
The CYTB analysis suggests that Pao sp. A is the same species as P. cochinchinensis (AB741980.1), whereas CYTB sequences of other P. cochinchinensis specimens (AP011925.1, JQ681922.1) are distantly related to Pao sp. A (Figure 3), which makes it difficult to identify species with DNA barcoding methods. According to Igarashi et al. [35], at least one of the CYTB sequences is likely an error resulting from the misidentification of specimens. Likewise, Pao sp. B is closely related to P. baileyi as well as to P. abei and P. leiurus (KF667490.1, GU057266.1) (Figure 3), and could be the same species as one of these. The phylogenetic positions of P. leiurus, however, largely differ among specimens ( Figure 3). As discussed in Igarashi et al. [35], these situations are most likely due to similarities in the morphologic characteristics, geographic distribution, and ecology of the Pao species. Our mitochondrial gene analyses indicated the presence of actual and/or potential genetic boundaries of species (i.e., species-level clusters in the phylogenetic trees), suggesting that an accumulation of genetic information could contribute to resolve the taxonomic status and relationships among Pao species. On the other hand, if crossbreeding occurred in the Pao species, an introgression of the mitochondrial genome might have caused such discordance between morphologic species identification and mitochondrial DNA phylogeny. Crossbreeding is frequently observed in the marine Takifugu pufferfish species (e.g., Takahashi et al., 2017 [36]), and an introgression of the mitochondrial genome from the freshwater Rhinogobius goby species was reported [37]. As demonstrated by the analysis of the TBT-bp2 gene, this possibility should be carefully considered in parallel to taxonomic reexamination, though no evidence was found for the presence of crossbreeding between Pao sp. A and Pao sp. B.
The toxin concentration in each tissue differed greatly between Pao sp. A and Pao sp. B: it was generally much higher in Pao sp. A than in Pao sp. B. Ngy et al. [7] examined the toxicity of two Cambodian freshwater pufferfish species, Pao turgidus and Pao sp., during the rainy and dry seasons, and reported that only the skin and ovary of P. turgidus were toxic in both seasons. The highest toxicity scores of the skin and ovary were 37 mouse unit (MU)/g (5.5 mg STX·diHCl eq/kg) and 27 MU/g (4.0 mg STX·diHCl eq/kg), respectively, which indicate low levels of toxicity, similar to Pao sp. B. Ngy et al. [7] did not conduct a genetic analysis, and while there is some uncertainty in species identification, it is assumed that several species of low-toxicity freshwater pufferfish inhabit Cambodia. The Bangladeshi L. cutcutia (highest toxicity score of the most toxic tissue, the skin, 20 MU/g (3.0 mg STX·diHCl eq/kg)) [6] could also be considered a low-toxicity species, similar to Pao sp. B. According to Kungsuwan et al. [4], the Thai freshwater pufferfish P. leiurus and P. suvattii are both highly toxic, with the highest mean toxicity scores of the skin, muscle, liver, and ovary ranging from 61 to 109 MU/g (9.1-16 mg STX·diHCL eq/kg) in P. leiurus and 42 to 200 MU/g (6.3-30 mg STX·diHCL eq/kg) in P. suvattii. The toxin concentration of Pao sp. A is comparable to or even higher than that of P. leiurus/P. suvattii, and this species is one of the most highly toxic species among freshwater pufferfish.
The total toxin amount per fish was generally similar between Pao sp. A and Pao sp. B, with the highest score exceeding 400 µg STX·diHCL eq/individual in both species. As the minimum lethal dose for humans is approximately 400-1000 µg STX·diHCL eq [38,39], ingestion of one-two whole bodies of these freshwater pufferfish can cause death by poisoning. Therefore, both species should be considered extremely dangerous to eat. In Pao sp. B, however, the majority of the harbored toxin was distributed in the skin, and the toxin concentrations of the other tissues were below or only slightly above the regulatory limit. Therefore, at least with the individuals examined in this study, removing the skin may decrease the possibility of poisoning. In addition to individual differences, regional and seasonal variations are observed in pufferfish toxicity [1,40]. Careful accumulation of data on the phylogeny and toxin profile of freshwater pufferfish, with such considerations in mind, may help to establish a taxonomic system linked with toxicity profiles and the presentation of species that can be consumed by removing toxic parts and other treatments.
In both Pao sp. A and Pao sp. B, the main toxin was STXs comprising STX as the main component, and neoSTX and dcSTX as the minor components, and TTX was not detected at all. The toxin of P. turgidus from Cambodia and Pao freshwater pufferfish from Thailand is similarly composed of STX-dominated STXs and does not contain TTX [4,5,7]. There are some exceptional cases, such as the case of the Bangladeshi L. cutcutia in which a unique STX derivative (carbamoyl-N-methylSTX) is the main component [6,41], but even in the case of marine pufferfish that simultaneously possess STXs and TTX, the main component of STXs is often STX [9,11,[13][14][15]. The origins of STXs in Floridian Sphoeroides and Philippine Arothron marine pufferfish are considered to be STX-producing dinoflagellates Pyrodinium bahamense and P. bahamense var. compressum, respectively, based on similarities in the toxin composition [8,9]. Although the origin of STXs in freshwater pufferfish is presumed to be STX-producing cyanobacteria [6], little is known about the toxicity of cyanobacteria in the freshwater environments of Cambodia, and the origin of STXs in Cambodian freshwater pufferfish remains to be elucidated.
The STX distribution profiles clearly differed between Pao sp. A and Pao sp. B. In contrast to TTX, of which the absorption, transportation, and accumulation in marine pufferfish have been studied with respect to candidate molecules involved in these processes (i.e., PSTBP), such molecules for STXs remain unknown regardless of the marine or freshwater pufferfish species. PSTBP genes in marine pufferfish obviously evolved as a result of the fusion of two TBT-bp2 genes at the sixth and first exons, and the sequences remain quite similar to TBT-bp2 genes on the same genome (Hashiguchi et al. 2015 [32]; Yamada et al., unpublished). TBT-bp2 was originally identified as a protein that binds the environmental toxin TBT in the fish body [31]. These findings together suggest that the TBT-bp2 gene could be one of the molecules involved in the kinetics of STX in the pufferfish body. As expected from the observed difference in the STX distribution profiles between Pao sp. A and Pao sp. B, we found different amino acid sequences of TBT-bp2 between Pao sp. A and Pao sp. B, which could result in interspecific differences in the structure and function of TBT-bp2. Given the identical amino acid sequence of the rhodopsin gene in Pao species, including Pao sp. A and Pao sp. B, the TBT-bp2 gene might be one of the genes that determines the physiologic and ecologic characteristics of the host. A more comprehensive analysis of the expression, distribution, and function of PSTBP/TBT-bp2 isoforms coupled with TTX/STX distribution profiles could provide a better understanding of the molecular mechanisms and evolutionary processes of TTX/STX accumulation in Tetraodontidae.

Pufferfish Specimens
In November 2019, 11 (PNH-1-11) and 5 (KTI-1-5) freshwater pufferfish were sampled from a small lake connected by a channel to the Mekong River near Phnom Penh and Kratie, Cambodia, respectively (Figures 1 and 2, and Table A1). The specimens were transported alive to the laboratory at the University of Kratie, where PNH-1-3 were dissected to obtain the skin (including the fins), muscle, liver, and gonads. Each tissue was placed in a plastic bag, frozen, and transported to the laboratory at Nagasaki University. The other specimens were frozen as whole bodies and transported to the laboratory at Nagasaki University, where they were similarly dissected. Each tissue was stored at −80 • C until DNA extraction or toxin quantification.

Phylogenetic Reconstructions
The sequences determined were checked and trimmed manually. In the case of the 2 nuclear genes, the Sanger sequence electropherograms were carefully examined for clear double peaks with comparable height, which reflects allelic heterogeneity at the locus. These sites were coded with IUPAC ambiguity codes in the sequence alignments. For the mitochondrial genes, a couple of previous studies [35,45] published CYTB sequences of closely related species with 16S rRNA or COI sequences. Phylogenetic reconstructions using Bayesian inference (BI) and maximum likelihood (ML) methods were conducted for CYTB and the 2 concatenated datasets, CYTB+16S rRNA, and CYTB+COI. Best fit models of nucleotide substitution for BI analyses by MrBayes 3.2.7 [46] were inferred using Kakusan 4 [47], and the GTR+I+G model was used for ML analyses by FastTree 2 [48]. In the BI analyses, 4 runs of 5 million generations were performed with 4 chains each, trees were sampled at 1000-generation intervals, and the first 25% of the trees were discarded as burn-in. In the ML analyses, bootstrap values were calculated using 1000 trees generated by SEQBOOT in the PHYLIP Package 3.698 [49], and the consensus tree was obtained by CompareToBootstrap.pl in FastTree 2 [50]. For TBT-bp 2, as the MrBayes analysis did not provide ambiguity codes, only ML analysis was conducted as described above, but with the JC+CAT model.

Conflicts of Interest:
The authors declare no conflict of interest.

KTI-1, 2, 4, 5
Pao abei (AB741960.    Sample  7  23  24  33  138  169  193  225  228  231  232  286  367  429  432  448  499  505  524   PNH-1  A  T  G  C  C  T  C  C  T  G  C  G  A  G  C  C  T  T  A  T G . C G G 1 Y = C or T, W = A or T, R = A or G, M = A or C, S = C or G, K = G or T, nd = not determined.  T G . C G G 1 Y = C or T, W = A or T, R = A or G, M = A or C, S = C or G, K = G or T, nd = not determined. Figure A3. HPLC-FLD chromatograms of the ovary extract from PNH-4 (left), skin extract from KTI-3 (middle), and of the saxitoxin (STX), decarbamoylSTX (dcSTX), and neoSTX standards (right).