7.0 HAPTOPHYTE BLOOMS: CHRYSOCHROMULINA, PRYMNESIUM, PHAEOCYSTIS
The Haptophyta include bloom-forming, flagellate species belonging to the genera Emiliania, Phaeocystis, Prymnesium and Chrysochromulina. Coccolithophore blooms, particularly those of Emiliania huxleyi, often discolour seawater chalky-white, but are not known to be harmful. Recurrent blooms of toxic Chrysochromulina spp. causing mass mortality of marine life and severe economic losses to aquaculture in Scandinavian waters have attracted recent attention; these bloom events are considered in Section 7.1. Ichthyotoxic blooms of the globally distributed flagellate Prymnesium parvum occur primarily in brackish water habitats, but recent bloom events in more saline fjords, embayments and at fish farm sites have kindled renewed interest in Prymnesium blooms [see Section 7.2]. The ubiquitous gelatinous, colony-forming Phaeocystis spp. produce dense blooms that can have harmful or nuisance impacts; these are considered in Section 7.3.
7.1. Chrysochromulina blooms in Scandinavian waters
7.1.1 Chrysochromulina taxonomy and autecology
Chrysochromulina is a species-rich, nanoplanktonic genus found globally in coastal and oceanic waters, with 55 known species, 38 of which occur in Scandinavian waters (Edvardsen, 2002; Eikrem and Edvardsen, 1999). Identification of Chrysochromulina species is based on species-specific scales that adorn their cells visible only by electron microscope. Molecular evidence suggests that some Chrysochromulina and Prymnesium spp. are closely related (Edvardsen and Paasche, 1998); taxonomic revisions are expected.
The definitive morphological trait of the haptophytes is a haptonema located between their two flagella; its length can be many times that of the cell. The haptonema, retractable and long and coiled in Chrysochromulina, and short and uncoiled in Prymnesium, has multiple functions (Kawachi and Inoye, 1995). In Chrysochromulina hirta, the haptonema coils spontaneously when physically or chemically stimulated, particularly during avoidance reactions, which suggest a sensory response to obstacles. Some Chrysochromulina species produce both amoeboid and non-motile cells in culture, but not resting cysts, unlike Prymnesium parvum which attaches to macroalgae by its haptonema (Edvardsen and Paasche, 1998; Johnsen and Lein, 1989). Should similar docking behavior occur in Chrysochromulina, it might provide an "overwintering" mechanism during unfavorable growth conditions. Chrysochromulina species are photosynthetic, but many species supplement this by mixotrophic feeding on small particles. The haptonema is important in prey capture, as shown for Chrysochromulina spinifera and Chrysochromulina hirta (Kawachi and Inoye, 1995). Chrysochromulina polylepis feeding on small flagellates ingested 1 to 2.5% of its body C per day (Legrand et al., 1996).
Estep and MacIntyre (1989) posited that Chrysochromulina polylepis was allelopathic against competing species after finding dead cells of the latter devoid of cellular contents, but intact during a bloom of C. polylepis. They suggested that this mode of antagonism was a novel, enzymic mechanism that they termed dasmotrophy. Chrysochromulina polylepis subsequently was confirmed to be allelopathic against other species in experiments (Myklestad et al., 1995). Docking of C. polylepis by its haptonema onto the attacked cells, followed by release of toxin (enzyme) would facilitate dasmotrophy.
Chrysochromulina spp., in culture, exhibit strong phototaxic behaviour. Chrysochromulina acantha and Chrysochromulina simplex have two swimming modes: a slow mode (ca. 5 µm s -1) and a rapid mode (up to 260 µm s -1) (Gregson et al., 1993). Swimming speeds of C. polylepis ranged from 65 to 138 µm s -1 (Dahl et al., 1989). Chrysochromulina spp. should be able to undertake nutrient-gathering migrations down into the nutrient reservoir below the euphotic zone for use in photosynthesis during their return migrations back up into the euphotic zone, a well known behaviour of flagellates (see Smayda, 1997).
The diverse haptonema- and swimming-based behavior of Chrysochromulina spp. are important autecological features contributing to their bloom dynamics. Assessments of harmful species dynamics invariably ignore the role of such behaviour; investigators usually focus on abiotic growth factors, grazing, toxin content, and treat the cells primarily as passive particles. Behaviourisms are strategies evolved by species to maximize growth and survival. Aquaculturists in developing management and mitigation procedures must take into account the cellular physiology and behaviour, and any unique aspects of the population behaviour of harmful species that threaten their enterprise.
Chrysochromulina species usually occur in low to modest abundance (10 3 to 10 5 cells L -1), but are capable of explosive growth leading to intense, virtually monospecific blooms (Edvardsen and Paasche, 1998). Some species are eurythermal and euryhaline, such as ichthyotoxic C. polylepis which has a relatively high growth rate (µ = 1.3 d -1) (Edvardsen and Paasche, 1998; Granéli et al., 1993). It grows fairly rapidly between 9° to 22°C, and has an optimal salinity between 20-30 psu, but will grow at 3 psu. Some species are cryophilic: Chrysochromulina birgeri blooms under ice in the Baltic Sea (Hallförs and Niemi, 1974).
The versatile autecology, cellular and bloom behaviour of the species-rich chrysochromulinids, and their threat to aquaculture, have become of interest because of their recent and unprecedented harmful blooms in Scandinavian waters.
7.1.2 The 1988 Chrysochromulina polylepis bloom in southern Scandinavian waters
7.1.2a Scale of mortality
The 1988 toxic bloom of C. polylepis extended over most of the Skagerrak and the entire Kattegat, an area of about 75,000 km 2, and lasted four weeks, from May to early June (Granéli et al., 1993). This bloom has been reviewed and reinterpreted frequently as newer data and insights have become available (Dahl et al., 1989; Lindahl and Dahl, 1990; Maestrini and Granéli, 1991; Granéli et al., 1993; Edvardsen and Paasche, 1998; Gjøsaeter et al., 2000). This continuous interest reflects the social and political concern triggered by this exceptional bloom of an obscure species, and equally exceptional mortality of the natural biota that resulted.
The magnitude and diversity of mortality during the bloom are unparalleled in European coastal waters. All trophic levels experienced mortality, from phytoplankton to zooplankton, benthic macroalgae and fauna, and demersal and pelagic fish. The benthic community was particularly devastated: piles of rotting echinoderms (starfish, sea urchins) were found among dead ascidians, molluscs (snails, whelks, bivalves), sponges and crabs (Lindahl and Dahl, 1990; Gjøsaeter et al., 2000). Among the macroalgae, red algal species exclusively were affected. Intertidal populations of mussels, barnacles ( Semibalanus balanoides) and macroalgae, other than rhodophytes, were relatively unharmed. Fish stocks were severely impacted, although selectively. Cod ( Gadus morhua) larvae suffered great mortality, unlike older fish that may have migrated into deeper waters during the bloom. The 1988 year classes (O-group) of cod, whiting ( Merlangius merlagus) and pollock ( Pollachius pollachius) were strongly reduced, unlike saithe ( Pollachius virens) and eel ( Anguilla anguilla) (Gjøsaeter et al., 2000). In experiments, cod exposed to 12 million cells L -1 of C. polylepis for 4.5-hrs began to swim upside down (Lindahl and Dahl, 1990). Zooplankton populations in situ were decimated and inhibited during experimental exposure to C. polylepis (Lindahl and Dahl, 1990) . Exposure to 200,000 cells L -1 was lethal to an Acartia sp. The copepods Centropages hamatus and Temora longicornis suffered reduced egg production and grazing, and death, the specific effect dependent upon the C. polylepis population density. Fertilization and early developmental stages of Mytilus edulis and the ascidian Ciona intestinalis were negatively affected in exposure experiments (Granmo et al., 1988).
Norwegian fish farms were also impacted, inspiring a mitigation effort that successfully moved 120 fish pens into brackish waters and into inner fjord segments where the bloom either did not occur, or was not harmful (Lindahl and Dahl, 1990). Sea trout kept at <14 psu survived large C. polylepis population densities (20 million L -1), while rainbow trout died after 1.5 hrs exposure to only 0.2 million cells L -1. Despite mitigation efforts, 900 tonnes of farmed salmon and sea trout valued at U.S. $9 million were killed (Eikrem and Throndsen, 1993; Gjøsaeter et al., 2000).
7.1.2b Bloom pattern, dynamics and causation
The C. polylepis bloom was first detected in early May prompted by the anomalous behavior and mortality of farmed fish in Gullmar Fjord, Sweden (Lindahl and Dahl, 1990). The bloom, entrained within coastal currents, spread northward from the Swedish fish farm sites and then westward along the southern Norwegian coast. A wind shift in mid-May reversed the direction of flow and seeded the central Skagerrak with a bloom population. Westward advection of C. polylepis resumed following another wind shift, and in late May the bloom dissipated off the southwest coast of Norway in relatively warm water (8°-12°C). In the central Skagerrak, the bloom ended in early June. During the four week bloom period, there was widespread mortality of sub-tidal and sub-littoral benthic communities, nearshore and coastal plankton, and fish communities along the entire route of bloom dispersal .
The toxic dinoflagellate K. mikimotoi co-occurred with C. polylepis during early bloom stages (Lindahl and Dahl, 1990), but the bloom that developed and spread thereafter was virtually a monospecific population of C. polylepis. Maximum population densities were very high, ranging from 80 to 100 million cells L -1 (Lindahl and Dahl, 1990; Kaas et al., 1991; Edvarsen and Paasche, 1998). Chlorophyll (biomass) concentrations (40-80 mg m -2) were low, however - only 10-20% of spring bloom levels (Dahl et al., 1989; Edvardsen and Paasche, 1998). The fact that the spring diatom biomass levels are usually 5- to 10-times greater than the biomass attained during the 1988 Chrysochromulina bloom is significant. This classifies the bloom as a high population density, low biomass bloom ( HDLB), a classification consistent with the nanoplanktonic size of C. polylepis: it has a cell length of 8-12 µm and a width of 6-8 µm (Eikrem and Throndsen, 1993). HDLB species are usually harmful or a nuisance not because of toxicity, but because their high population density, small cell size and/or nutritional inadequacy often lead to grazer starvation, either directly or through clogging of feeding appendages (Smayda, 1997). This contrasts with the virulence of PSP species, which is primarily a function of their cellular toxin levels. Ingestion of only a few PSP-containing cells can be toxic to grazers (see Smayda, 1992). The abundant, ungrazed populations during HDLB outbreaks lead to another, more widespread detrimental impact. When the accumulated, ungrazed biomass becomes nutrient-limited, it sinks to the bottom sediments and decomposes, which can lead to anoxia and widespread trophic mortality. While the negative impacts of HDLB species are usually more related to their population density (numerical abundance) than to biomass, the ability of C. polylepis to produce toxins distinguishes it from many other HDLB species. However, similar features in Chrysochromulina leadbeateri and Prymnesium parvum, for example, suggest that HDLB haptophyte species, commonly, may be chemically toxic as well.
The high bloom abundance of C. polylepis is consistent with a fairly rapid growth rate, and is not primarily the result of physical accumulation of a relatively slow growing, ungrazed population. A primary production rate of 2.6 g C m -2 d -1 measured during initial bloom stages yields a daily in situ turnover rate of 1.2 d -1 when pro-rated to the standing stock (Lindahl and Dahl, 1990). This is consistent with the maximum daily growth rate (µ = 1.3 d -1) found for C. polylepis in culture (Edvardsen and Paasche, 1998). Near the termination of the bloom, the production rate decreased to 0.5 g C m -2 d -1, corresponding to a population generation time of six days (Lindahl and Dahl, 1990). Chrysochromulina polylepis may also have benefited from an allelochemic repression of competing species. Ceratium (= dinoflagellate) species found outside the bloom waters were pigmented and viable, whereas ceratians within bloom waters were dead, their cells intact but devoid of cellular content (= empty). Ceratian mortality was estimated to have reached 80% (Johnsen and Lømsland, 1990). Myklestad et al. (1995) showed experimentally that C. polylepis can allelochemically inhibit other species. Thus, the allelochemical and multiple anti-grazing capacities, together with high growth rates of C. polylepis faciitated its 1988 bloom.
Controversy exists over the bloom trigger. Aksnes et al., (1995), based on modelling, concluded that anthropogenic nutrients delivered into the Skagerrak from the German Bight by the Jutland Current triggered the bloom. Edvardsen and Paasche (1998) have contended that nutrient stimulation is not supported by the evidence. Based on primary production rates, the cellular composition of C. polylepis and in situ nutrient levels, they concluded that nutrients fluxing into the euphotic zone from deep water supplied the calculated N and P demands during the bloom (Dahl et al., 1989; Lindahl and Dahl, 1990). Aure et al. (1989) have pointed out that large influxes of nutrient rich water into this region from the southern North Sea are recurrent, annual winter/spring events, unlike toxic blooms (Edvardsen and Paasche, 1998). The conclusion reached by Granéli et al. (1993) remains viable: " large scale eutrophication is not a prime requisite for harmful blooms of Chrysochromulina".
There is stronger evidence that the C. polylepis bloom was initiated and driven by a combination of hydrographic and meteorological conditions that corresponded to the classical rainfall-runoff-high irradiance event, a well-known trigger of red tides and harmful blooms [see Section 12.1]. The difference, in this case, was the very large scale of the runoff event. Pre-bloom, winter/spring weather in northern European was mild and wet. Precipitation exceeded the mean rainfall for that period by 2.5-fold. In response, the influx of low saline water into the Kattegat from the Baltic was elevated, as was delivery of nutrient-rich (riverine derived) runoff into the German Bight flowing northward into the Skagerrak. The euphotic zone and pycnocline in Swedish coastal waters were shallower than usual due to the large outflow from the Baltic Sea (Lindahl and Dahl, 1989). The calm, sunny period that prevailed the first two weeks in May, and which rapidly warmed the surface watermass layer from 6° to 12°C, combined with the hydrographic conditions to trigger the bloom. The hydrographic, inorganic nutrient levels and irradiance conditions meet the classical critical depth requirements for a spring bloom event, but the factors selecting specifically for C. polylepis remain unknown (Granéli et al., 1993). Chrysochromulina polylepis has a high selenium requirement (Dahl et al., 1989), prompting speculation that elevated micro-nutrient levels delivered in runoff played an important bloom-role, but selenium was not measured and bioassays testing the potential role of cobalt were inconclusive (Segatto and Granéli, 1995).
Scandinavian investigators, generally, agree that runoff was an important factor in the 1988 bloom , a view strengthened by an 8-year time series (1989 to 1996) for a southern Norwegian coastal site (Dahl et al., 1998). Thrice weekly sampling revealed that C. polylepis was seasonally present (May-June) and achieved modest, non-toxic blooms ranging in abundance from 0.5 to 1 million cells L -1. In 1994 and 1995, when the lowest surface salinities (15-17 psu) and highest N:P ratios for the time series were recorded, bloom abundance increased to 5.5 and 2.5 million cells L -1, respectively. The accompanying reduction in salinity and elevated N:P ratios, due to increased NO 3 concentrations, are symptomatic of local, freshwater runoff, and led Dahl et al. (1998) to conclude that runoff promotes Chrysochromulina blooms. It is unresolved whether elevated N:P ratios during the 1988 bloom selected for C. polylepis (Dahl et al., 1989; Granéli et al., 1993; Gjøsaeter et al., 2000). Mesocosm experiments with natural communities showed the opposite effect: non-toxic Chrysochromulina pringsheimii was favored at low N:P ratios, i.e. during nitrogen limitation rather than under P-limiting conditions, i.e. high N:P ratios (Schöllhorn and Granéli, 1993).
In summary, the unusual climatic conditions and high runoff that preceeded the 1988 C. polylepis bloom are anomalies of interest because they impact several habitat variables that regulate blooms. The most viable, general explanation of the bloom is the conclusion reached by Granéli et al. (1993): it " seems increasingly likely that unusual climatic and hydrographic conditions rather than long-term environmental changes were decisive in preparing the stage for the 1988 bloom". This conclusion applies only to the initial stage of the bloom: the sources of the seed population, whether the bloom developed synchronously throughout the region, and why it became monospecific and grazer-free are unknown.
7.1.2c Phosphorus limitation and toxicity
Phosphate limitation appeared to terminate the bloom (Dahl et al., 1989; Granéli et al., 1993; Edvardsen and Paasche, 1998). Phosphorus had three effects on C. polylepis: it regulated cellular growth rate, population abundance and, when limiting, induced cellular toxicity. Chrysochromulina polylepis produces two haemolytic compounds (1-acyl-3-digalactosyglycerol and octadecapentaenoic acid) (Yasumoto et al., 1990). Cell extracts of C. polylepis and C. leadbeateri tested positively for haemolytic activity (Meldahl et al., 1995). Haemolysins are also produced by the closely related species P. parvum (Igarishi et al., 1995) and Prymnesium patelliferum (Meldahl et al., 1995). The broad spectrum of mortality observed during the 1988 bloom, ranging from plankton to benthic invertebrates to fish, is consistent with the presence and mode of action of hemolysins (prymnesins). These toxins attack membranes, unlike dinoflagellate neurotoxins such as saxitoxin and brevetoxin which specifically target vertebrate neurosystems. Since fish gills have large surface areas and are highly permeable, they are important entry sites for prymnesins.
In culture, P-starved clones from the C. polylepis bloom became haemolytic and also produced feeding repressants having lethal activity against tintinnids (Carlsson et al., 1990; Edvardsen et al., 1990; Edvardsen, 1993). P-limited cells were 60-times more toxic to Artemia salina nauplii than P-sufficient cells. Demonstration of P-limited induction of toxicity in cultures and the high N:P ratios during early bloom stages have led to the generally accepted view that the bloom was (ultimately) a P-limited event, and that the widespread mortality observed resulted from P-induced toxicity. However, P-limitation does not completely explain this toxicity, which was stronger in situ than in laboratory cultures (Granéli et al., 1993). If toxicity in C. polylepis is primarily a senescent population and P-limited trait, the continuous growth and increase of the population over the 4-week bloom period suggest that toxicity and mortality should have been restricted to late bloom stages, rather than persist as the bloom spread and grew within its 75,000 km 2 block, and irrespective of the growth phase (see Figure 1b in Gjøsaeter et al., 2000).
The variability of toxin production within, and among Chysochromulina spp. complicates analysis of what determines whether a given Chysochromulina bloom will become toxic or not, or whether the local species are toxic or not. In addition to the effect of P-limitation on toxigenesis, there is considerable clonal (strain) variability in toxicity (Edvardsen, 1993) influenced by the population density (Jebram, 1980) and whether there is mixotrophic feeding on P-rich bacterial cells (Legrand et al., 1996), which can quench toxin production. Toxicity tests revealed that the potency of strains of C. polylepis isolated during different years varied greatly; the 1988 strain was most toxic (Dahl et al., 1998). Several of the 14 Chysochromulina spp. found to be non-toxic in culture, even when P-limited (Edvardsen, 1993; Edvardsen and Paasche, 1998), were present during a bloom in Danish coastal waters of five Chrysochromulina species, during which there was mortality of both farmed and wild fish (Knipschildt, 1992). Three of the 14 species were toxic to the bryozoan Electra pilosa when "old cultures" were provided as food (Jebram, 1980). These diverse influences on toxin production obscure the contribution of individual Chrysochromulina species to mortality when present in mixed communities that include other toxic species. An unidentified Chrysochromulina sp. was present during a massive bloom in the Skagerrak dominated by the ichthyotoxic raphidophytes Chattonella marina and Heterosigma akashiwo, and during which 1,100 tonnes of farmed Atlantic salmon were killed (Naustvoll et al., 2002). During a non-toxic bloom in Dutch coastal waters, an unidentified Chrysochromulina sp. co-occurred with Chattonella marina (Vrieling et al., 1995).
The diverse evidence and behaviour of C. polylepis and related species suggest that the 1988 bloom was not only exceptional in the regional scope and vastness of its induced mortality, but it was also an exceptional event both for Chrysochromulina spp. and the haptophyte group. The controvery over what stimulated the 1988 bloom underlies an ongoing social and political concern and debate triggered by this exceptional bloom. This debate is whether this (and other) antagonistic blooms are due to natural causes or anthropogenic effects, in this case eutrophication. This type of debate has become commonplace, even among scientists, and is at the center of the ASP - fish farm stimulation debate in Scotland. The great merit of the 1988 C. polylepis event is that it shows the need for, and the success in treating harmful algal blooms as ecological and oceanographic events rather than simply as stimulus (= nutrients, etc.) - response (= blooms, fish kills, etc.) events. The altered views with regard to initiation of the 1988 bloom that have resulted from newer research insights show also the merit of revisiting bloom events as newer data become available in seeking to explain their causes. Through such quantification, ultimate causes are more likely to be reached than through reliance on statistical correlations or anecdotal associations. Extrapolations to other, related blooms and species should be guarded, however, based on two other Chrysochromulina blooms in Scandinavian waters.
7.1.3 The 1991 Chrysochromulina leadbeateri bloom in Northern Norway
In mid May 1991, an extraordinary bloom of Chrysochromulina leadbeater , which lasted five weeks, developed in the Ofotfjord-Vestford area in northern Norway, during which 600 tonnes of farmed salmon were killed along a >200 km belt (Aune et al., 1992; Eikrem and Throndsen, 1993; Granéli et al., 1993; Rey and Aure, 1991). This was the first known bloom of C. leadbeateri in western Norwegian coastal waters since first recorded in 1970, and the first report of its toxicity. Similar to C. polylepis, it is nanoplanktonic (6-8 µm Ø), grows rapidly (µ = 1.4 d -1), and appears to be both eurythermal and euryhaline given its recorded distribution in coastal and oceanic waters of both hemispheres (Eikrem and Throndsen, 1993; Edvardsen, 1995; Edvardsen and Paasche, 1998).
This bloom had some features in common with the 1988 C. polylepis bloom: it developed after collapse of the spring diatom bloom during a period when the surface layer, stabilized by freshwater runoff, became calm and sunny weather prevailed (see Edvardsen and Paasche, 1998). There were major difference in nutrient conditions and bloom features during the two Chrysochromulina blooms. Anthropogenic nutrification was not a factor in the pristine, northern area and the N:P ratio was normal, unlike the high ratio indicative of P-limitation during the 1988 C. polylepis bloom. The bloom was not monospecific, other Chrysochromulina spp. were present, and there was no apparent mortality of the natural biota other than a localized high mortality of sea-urchins (Johannesen et al., 1991).
During the 1990-1991 winter, more than 70% of the total stock of Norwegian spring-spawning herring (1.5 x 10 6 t) overwintered in the Ofotfjord-Vestford bloom area (Legrand et al., 2001). The metabolism of this population, in combination with natural mortality and decomposition, was believed to have altered chemical conditions: to have decreased oxygen levels, to have released organic substances, and to have increased inorganic N concentrations. Legrand et al. (2001) have suggested that these herring induced changes may have favored the C. leadbeateri bloom and its toxicity, and proposed the following mechanism of stimulation. Growth promoting polyamines (cadaverine, putrescine) are present in high concentrations in herring tissue and released from dead fish through bacterial action. This release has two effects: the polyamines serve as a bloom trigger and as co-factors to form a "haemolytic-polyamine complex" that heightens the toxicity of C. leadbeateri. The origin and synthesis of this toxin - phytostimulation complex are very unclear. Strains isolated into culture from the bloom did not test positively for toxicity (Edvardsen, 1995; Legrand et al., 2001) unlike water samples collected then (Edvardsen, 1995).
Johnsen et al. (1999) previously suggested that three phases characterized the C. leadbeateri bloom: eutrophication from herring and mixotrophy ? growth-enhancing polyamines ? synergistic effects of polyamines and ichthyotoxins on toxicity development in C. leadbeateri. While this bloom may have been responsive to such complex nutritional and toxin chemistry, the factors regulating and prolonging this bloom and its toxicity are unknown. Comparisons of the 1988 and 1991 blooms suggest that C. polylepis and C. leadbeateri diverge in their bloom requirements; that multiple environmental scenarios can promote their blooms and toxicity, and that a common Chrysochromulina bloom model is not to be expected. The blooms show also that toxic Chrysochromulina blooms are natural events that can develop both in nutrient rich and unpolluted habitats; that toxicity may be enhanced by high N:P ratios (P-limitation) or by some other mechanism; that the bloom strain may be non-toxic; and that there may be a chemical factor associated with the physical benefit that runoff brings to blooms.
7.1.4 The 1992 Chrysochromulina spp. bloom in Danish waters
The third bloom of Chrysochromulina to develop in Scandinavian waters in the five year period since 1988 occurred in 1992, covering an area of 10,000 km 2 in the southern Kattegat, primarily in the Lillebaelt region of Denmark (Hansen et al., 1995). The bloom developed after the diatom spring bloom, as in 1988 and 1991, and lasted about one month, from mid-April to mid-May. About 50 tonnes of farmed rainbow trout ( Oncorhynchus mykiss) were killed; fish farms lost 30 to 80% of their stocks. As during the 1988 and 1991 Chrysochromulina blooms, there was a pronounced pycnocline, high N:P ratios (30:1); the surface waters (14-16 psu) were brackish, with temperature ranging from 7.5° to 8.3°C. The limited data on pre-bloom habitat and meteorological conditions preclude evaluation of potential bloom triggering mechanisms.
The salient features of the 1992 bloom were the diverse composition of the Chrysochromulina bloom species and their mortality modes. The maximum abundance of about 50 million cells L -1 corresponded to a carbon biomass level of 600 mg C m -3, yielding a chlorophyll biomass level of 12 mg m -3 based on a C:Chl ratio of 50:1. Thus, the 1992 bloom was a HDLB event. Four Chrysochromulina species dominated (90%) the biomass: C. brevifilum, C. ericina, C. hirta and C. spinifera. This composition differed significantly from the Chrysochromulina blooms in 1988 and 1991 which were monospecific, unlike the 1992 event and which differed also in the absence of C. polylepis and C. leadbeateri. The four Chrysochromulina species association with the 1992 toxic bloom was also unique. Excluding the possible inhibition of a bryozoan when fed senescent cells of C. brevifilum (Jebram, 1980), these four species have not tested positively for toxicity (Edvardsen, 1995). Their previous blooms were benign (Dahl et al., 1998), and copepods incubated in bloom water during the 1992 bloom suffered no mortality (Hansen et al., 1995). The cause of farmed fish mortality is unclear; there was not a correlation between Chrysochromulina biomass and the loss of caged fish. The high N:P ratios (30:1) were in the range that favored toxicity of C. polylepis and C. leadbeateri during their blooms. Fish died primarily during early bloom stages; they were not reported during the bloom maximum. This was similar to events during the 1991 C. leadbeateri bloom in northern Norway (Johannesen et al., 1991). Hansen et al. (1995) have suggested that the 1992 Chrysochromulina bloom (without reference to specific species) was differentially toxic, i.e. " killed the weakest part of the caged fish population, while the remaining population acclimatized to the toxic bloom ". In support of this, they pointed out that populations of rainbow trout ( Oncorhynchus mykiss) exposed to concentrations of the heavy metals Cu and Zn. responded in similar fashion.
Planktonic ciliates were virtually absent during the bloom despite the presence of nanophytoflagellates, on which they feed. This scarcity does not appear to be due to haemolysin production by the Chrysochromulina spp. (Hansen et al., 1995). Three of the species, C. ericina, C. hirta and C. spinifera, whose cellular diameters range from 4-10 µm, are adorned with long spines (10 to 36 µm in length) that increase their cellular 'functional diameter' to between 25 and 76 µm. The latter size range exceeds the capacity of the ciliates to ingest these chrysochromulinids and probably led to their food limitation and inhibited their population growth. The evidence suggests that low grazing pressure probably promoted the 1992 Chrysochromulina bloom (Hansen et al., 1995). Experiments suggested that in spite of high grazer biomass, the daily loss to grazing was only 2-3% of the phytoplankton biomass.
The unpredictability, species diversity, and variable scope, scale and range of harmful affects, which range from benign occurrence to severe mortality, and which characterize the Chrysochromulina blooms in Scandinavian waters present an even greater concern: prediction of when, where, and what species will bloom; the bloom triggering event, and which combination of potential multifactors will affect bloom intensity and duration at a potential bloom site. Answers to these queries are beyond present capability. The extent to which fish farms themselves promote or auto-regulate Chrysochromulina blooms is also unknown. The issue of fish farm stimulation of HABS is evaluated in Section 10.
7.2 Prymnesium blooms and fish kills
Blooms of Prymnesium parvum have caused fish kills in European waters since the 1890s (Moestrup, 1994), occurring most frequently in low salinity lakes and in brackish lagoons with high nutrient concentrations (Edvardsen and Paasche, 1998). Prymnesium can also form non-toxic blooms (Haase, 1994). In UK waters, P. parvum blooms are recurrent in the brackish (1-6 psu) Norfolk Broads where they can reach 800 million cells L -1 (Holdway et al., 1978). In Hickling Broad, the largest lake (120 ha) in this system, significant habitat changes which began in 1969 culminated in a major fish kill during a P. parvum bloom (Anonymous, 1991). Intensive drainage of the catchment area increased salinity favouring marked annual increases in the populations of migratory, roosting gulls, and whose wastes fertilized this brackish lake. These habitat changes are thought to have increasingly favoured Prymnesium growth, with consequent kills of fish and their mysid ( Neomysis integer)prey (Anonymous, 1991) .
Equally prodigious blooms of P. parvum have occurred elsewhere (see Table 2 in Edvardsen and Paasche, 1998), and are prominent in the coastal lagoons of the Baltic Sea (Leppänen et al., 1995; Lindholm, 1997; Moestrup, 1994). In Danish waters, a bloom of 655 million cells L -1 was accompanied by fish and bivalve mortality, revealing P. parvum can be toxic against invertebrates also (Otterstrøm and Steemann Nielsen, 1939). That event was also the first connection established between Prymnesium blooms and fish kills. And, a bloom of 126 million cells L -1 caused total mortality in a Danish fish pond (Moestrup, 1994). In a hypereutrophic brackish lake in Morocco, a P. parvum bloom (322 mg chl m -3) caused extensive mortality of fish, shrimp, shellfish and other invertebrates (Sabour et al., 2000). Winter blooms of Prymnesium cf. patelliferum in a brackish Italian lake caused zooplankton and fish kills (Mattioli and Simoni, 1999. Prymnesium parvum first appeared in Israel fish ponds in 1947, and rapidly spread in these inter-linked brackish impoundments to cause extensive mortality of farmed carp and other fish species (Reich and Aschner, 1947). Prymnesium has since become endemic, developing dense blooms of up to 10 billion (10 10) cells L -1 throughout most of the year (Shilo, 1967). Prymnesium parvum was characterized as having become the " most serious natural obstacle to fish breeding in Israel " (Shilo and Shilo, 1955). Continuance of fish-farming has been made possible by application of ammonium sulfate to the fish ponds as a bloom control agent, the active ion being ammonium which lyses Prymnesium cells (Shilo and Shilo, 1953, 1955).
While P. parvum is an HDLB species, similar to Chrysochromulina spp. (see Table 2 in Edvardsen and Paasche, 1998), it differs in being both toxic and morphologically antagonistic to grazers. Prymnesium spp. are nanoplanktonic, similar to Chrysochromulina spp.; P. parvum cells have a length from 8-14 µm and a width of 6-8 µm (Eikrem and Throndsen, 1993). Despite its brackish water preference, P. parvum is euryhaline, exhibiting a range from 5 to 45 psu, with an optimal salinity for growth at 10-20 psu (Edvardsen and Paasche, 1998). It is also eurythermal, achieving blooms over a temperature range from 5°-30°C, with very high growth rates in culture at 26°C (Edvardsen and Paasche, 1998). This tolerance results results a very wide range of habitat conditions at which P. parvum may bloom, and facilitates its global, cosmopolitan distribution. Prymnesium patelliferum, morphologically identical to P. parvum except for differences in scale morphology discernible only by electron microscopy, is equally euryhaline and eurythermal (Edvardsen and Paasche, 1998; Larsen et al., 1993). Prymnesium calathiferum, reported from New Zealand where its bloom caused mortality of fish and shellfish, is truly marine (Chang, 1985).
Prymnesium species and their blooms are of interest to HAB dynamics in Scottish coastal waters because of their unusual series of ichthyotoxic blooms in southwestern Norwegian waters, in the Ryfylke Fjord region, during the 3-year period from 1989 to 1991 (Aure and Rey, 1990; Kaartvedt et al., 1991; Eikrem and Throndsen, 1993). The 1989 bloom was claimed to be the first known instance of a P. parvum bloom in a large marine (estuarine) system (Aure and Rey, 1990). In retrospect, the 5-year period beween 1988 to 1992 might be termed the Haptophyte pentade. During this 5-year period, P. parvum and P. patelliferum bloomed each year between 1989 and 1991 following the 1988 and 1991 blooms of C. polylepis and C. leadbeateri, respectively, and prior to a multispecific bloom of C. brevifilum, C. ericina, C. hirta and C. spinifera in 1992 in the southern Kattegat (Hansen et al., 1995), and discussed in Section 7.1.4. The Prymnesium blooms killed 1,250 tonnes of farmed salmon. Summed with the mortality during the 1988 and 1991 Chrysochromulina blooms, this resulted in a 5-year dieoff of farmed Atlantic salmon equalling 2,650 tonnes valued at about U.S. $15 million (Eikrem and Throndsen, 1993).
The Prymnesium blooms in the Ryfylke Fjord region were summer (July-August) blooms that originated in waters of =10 psu salinity (Granéli et al., 1993) and about 18°C. The inner Ryfylke fjord system consists of three fjord branches into which there is freshwater inflow regulated by discharge from a hydroelectric power plant (Aure and Rey, 1990). This discharge influences recipient waters in several ways stimulatory to Prymnesium blooms. The high freshwater input and sluggish exchange with offshore waters create a marked brackish surface layer (4 to 5 psu) characterized by a high N:P ratio (47:1 to 64:1). This ratio reflects the high NO 3 and low PO 4 content of the discharged water (Aure and Rey, 1990; Kaartvedt et al., 1991). The volume of discharge varies seasonally and interannually, partly as a function of snow melt volume. The freshwater inflow admixes with deeper waters and induces upwelling which pumps nutrients upwards into the euphotic zone. The discharge of freshwater, which is regulated, and the induced upwelling (stated to be intense) are thought to be significant to bloom initiation. During the 1989 and 1990 Prymnesium blooms, the time of initiation closely depended on when freshwater discharge began. Beyond this initial effect, subsequent bloom dynamics and toxicity were influenced by phosphorus availability. Calculations based on mass balance considerations indicated that during the 1990 bloom only 20% of the required PO 4 for growth was supplied by freshwater discharge and induced upwelling (Aure and Rey, 1992). The remaining phosphorus required had to be supplied from in situ remineralization. During the 1989 bloom, alkaline phosphatase activity was high in cells collected from the brackish layer (Kaartvedt et al., 1991), indicative of P-limited growth and the capacity of the cells to assimilate dissolved organic phosphorus.
Chemotaxic attraction of the Prymnesium cells to fish pens allowed Prymnesium to exploit the phosphorus excreted by penned-fish. The directed migration of the cells to the fish pens presumably occurred by following the phosphorus waste chemocline. Populations in the waters surrounding fish-pens were 10 to 40 times greater than in open waters (Kaartvedt et al., 1992). Prymnesium cells also exhibited a high affinity for attachment onto surfaces, accumulating onto the fish cages and fouling fish nets. This behaviour facilitated their access to, and uptake of phosphorus. An indication of the magnitude of this response is evident from the population density of P. parvum cells found attached to the green macroalga, Cladophora spp. Shaking of the algal fronds released up to 400 million Prymnesium cells L -1. These attractions and attachments of Prymnesium to fixed sources of phosphorus would appear to be a rather unique nutrient gathering stategy.
Based on estimated rates pf phosphorus excretion at the fish farm sites and the phosphorus requirements of Prymnesium (based on C. polyepis), the daily crop of new cells produced in an inner, brackish arm of the fjord system would correspond to 10 9 to 10 10 cells m -2 of sea surface (Kaartvedt et al., 1992). This calculation shows that the caged fish excreted quantiative levels of nutrients helping to sustain the Prymnesium bloom. The more significant result is that Prymnesium blooms, similar to the toxic Chrysochromulina blooms, seemingly flourish in P-limited habitats. While P availability may limit population levels, the high N:P ratios favor their increased toxicity. Shilo (1967) has shown that P-limitation increased P. parvum toxicity by 10- to 20-fold. Presumably, toxicity has survival value in helping Prymnesium to cope with P-limiting conditions, but the specific benefit, if any, is unclear.
The Prymnesium blooms that developed during 1989-1991 did not achieve great abundance. The maximum recorded population in open fjord waters in each year was =5 million cells L -1 (Aure and Rey, 1992; Granéli et al., 1993), a population density considerably below that usually associated with fish mortality (see Table 2 in Edvardsen and Paasche, 1998). A P. patelliferum population attained 370 million cells L -1 and a chlorophyll biomass of 250 mg m -3 during an August 1991 bloom (Johnsen et al., 1998). During the 1992 bloom, a Chrysochromulina ericina bloom (5 million cells L -1) also developed (Aure and Rey, 1992).
Prymnesium parvum was first recorded from Norwegian waters in 1965 (Eikrem and Throndsen, 1993). Similar to the Chrysochromulina spp. blooms, there was a multidecadal gap between this initial detection and the first reported bloom event. The fact that their novel blooms clustered in the 4-year period, from 1988-1991, suggests that a Haptophyte niche may have opened during that period, and has since constricted. This does not seem to be associated with a climatic anomaly or meteorological pattern. The freshwater delivery into the inner brackish areas is from unpolluted rivers and hydroelectric dams that have an excess of nitrate relative to phosphate (Granéli et al., 1993).
Toxin production by Prymnesium is a stronger trait than found in Chrysochromulina spp. Prymnesium toxins have very potent haemolytic activity, with two toxins having been isolated, prymnesin-1 and prymnesin-2 (Igarishi et al., 1995; Meldahl et al., 1995). Toxicity appears to increase with P-limitation, as in C. polylepis, also during N-deficiency (Johansson and Granéli, 1999; Granéli and Johansson, 2001), and possibly at high N:P ratios (see Granéli et al., 1993). In addition to being ichthyotoxic, Prymnesium may have allelochemical activity, inhibiting competing phytoplankton species, such as the dinoflagellate Prorocentrum minimum and the cryptomonad flagellate Rhodomonas cf. baltica (Granéli and Johansson, 2001). Prymnesium appears to have a high capacity for phagotrophy. It has been argued that this , particularly predation on bacteria, is an important source of phosphorus when limiting (Legrand et al., 2001). The relationship between toxin content and mixotrophic feeding is unknown. Legrand et al. (2001) review the environmental regulation of toxicity in Prymnesium spp.; toxin synthesis appears to be under multifactorial control.
Prymnesium spp. have a complex haploid-diploid life cycle, including the formation of a resting stage, and two stages of P. parvum identifiable as the forms: f. parvum and f. patelliferum (Edvardsen, 2002; Larsen and Medlin, 1997; Larsen and Edvardsen, 1999). Reference to P. patelliferum in the literature is actually P. parvum f. patelliferum. Prymnesium parvum also has an unusual sessile stage that attaches to benthic algae and other objects forming dense populations (Johnsen and Lein, 1989). The occurrence of two morphological forms, its planktonic and sessile vegetative stages, and a resting stage is evidence that P. parvum is adapted to, and requires multiple niches in its bloom dynamics. This considerably complicates efforts to quantify the regulation of its blooms, since events at any stage of its life cycle, favorable or unfavorable, can affect its subsequent bloom pattern. There is considerable information on Prymnesium cellular ecophysiology and toxicity, but minimal information on its population dynamics and regulation.
7.3 Phaeocystis blooms
This section summarizes some of the evidence that the haptophyte genus Phaeocystis can exhibit nuisance and toxic behaviour. Phaeocystis is found in Scottish coastal waters, but its blooms, should they have occurred, have not attracted much attention (see review in Tett and Edwards, 2002), unlike in contiguous UK waters (Jones and Haq, 1963; Morris, 1971; Owens et al, 1989). Among the haptophytes, harmful blooms of Phaeocystis probably pose a relatively insignificant threat to fish farms unlike potential blooms of Prymnesium and Chrysochomulina. Shellfish culture may be more vulnerable given the experience in Dutch coastal waters. Moestrup (1994) concluded that the effects of adverse Phaeocystis blooms are usually reversible nuisances rather than toxic impairments. It was long believed that there was only one species, Phaeocystis pouchetii. Recent molecular and morphological findings indicate that at least four species have been identified as P. pouchetii, including Phaeocystis globosa, a common bloom species in European waters (Baumann et al., 1994; Medlin et al., 1994). The extensive literature on the biology and ecology of Phaeocystis has been reviewed by Davidson and Marchant (1992) and Lancelot et al. (1998). Phaeocystis, where used here, will refer here to Phaeocystis pouchetii and Phaeocystis globosa.
Phaeocystis has a complex life cycle that includes a unicellular, motile stage and gelatinous colonial stage that can reach a size of several cm. The large size, gelatinous matrix and (possibly) chemical exudates from the colonial stage often deter grazing, although there is increasing evidence that Phaeocystis can be grazed by protozoans, copepods (see, however, Bautista et al., 1992) and fish, including mackerel ( Scomber scomber) and flounder ( Pleuronectes flesus) (Weisse et al., 1994). When grazer control fails, an anoxia can develop. Widespread mortality of benthic invertebrates in the Irish Sea was induced by anoxia that developed after sedimentation of a Phaeocystis bloom (Rogers and Lockwood, 1990). An anoxic event in the western Oosterschelde following the collapse of a P. globosa bloom caused the dieoff of 10,000 t of Mytilus edulis valued at Euro 20 million (Peperzak, 2002).
Significant levels of foam ( Meerschaum) can be produced during Phaeocystis blooms, the consequence of turbulent churning of the copious amounts of mucilaginous material secreted by Phaeocystis (Lancelot et al., 1987; Thingstad and Billen, 1994). When blown onto beaches, the foam deposits as spindrift in layers that can build up to 1 m in thickness. The unsightly conditions affects beach tourism. Malodorous seas resulting from the secretion of dimethylsulfide ( DMS) sometimes occur during the collapse of P. pouchetii blooms (Liss et al., 1994). The volatile gas liberated is linked to the synthesis of acrylic acid, which is antibiotic against bacteria (Sieburth, 1960). Savage (1930) suggested that migrating herring deflected their migratory route to avoid Phaeocystis blooms and associated 'stinking water'. This assumed fish avoidance has led to the widespread view that Phaeocystis is toxic, but the evidence for this is meagre: a fish kill in Norway (see Moestrup, 1994) and the report that a 1997 bloom of P. pouchetii in Hong Kong waters killed " thousands of caged cultured fish in the area [when] during the bloom a thick layer of foam" developed (Songhui and Hodgkiss, 1999). This foam is probably similar to that which accumulates on beaches in Holland. Phaeocystis has been reported to kill larval fish in bioassay experiments, including herring ( Clupea harengus) and cod ( Gadus morhua) larvae, and to depress feeding in farmed Atlantic salmon (Sieburth, 1960; Eilertsen and Raa, 1995; Aanesen et al.,1998). However, the filtrate from the Phaeocystis strain reported to kill cod larvae (Aanesen et al.,1998) functioned more as an anaesthetic than haemolytic toxin when bioassayed in a rarely used fly response test (Stabell et al., 1999).
Phaeocystis blooms are common in North European waters (Lancelot et al., 1998). A multidecadal increase in the dominance and annual bloom duration of Phaeocystis globosa has occurred in the Dutch Wadden Sea since the 1970s, where it has become a dominant bloom species (Cadée, 1986; Cadée and Hegeman, 1986; 2002). This success has been linked to long-term nutrient enrichment of those waters (Lancelot et al., 1987), and where the spawning and clearance rates of cultured Mytilus edulis can be inhibited during Phaeocystis blooms (Pieters et al., 1980; Smaal and Twisk, 1997). This cause of this impairment is unresolved. The latter authors partly attributed this to the large size of the Phaeocystis colonies, while other studies show the colonial stage does not affect Mytilus edulis growth rates (Petri and Vareschi, 1997). This equivocal effect on Mytilus is similar to the contrasting experimental results obtained in studies of zooplankton grazing on Phaeocystis (Bautista et al., 1992; Weiss et al., 1994).
Notwithstanding the diverse nuisance and toxic impacts reported for Phaeocystis, its blooms are usually without the apparent negative impacts, particularly fish kills, reported for the related haptophyte species C. polylepis and P. parvum f. parvum and f. patelliferum.
7.4 Relevance of Haptophyte blooms to Scottish aquaculture
The toxic Chrysochromulina (particularly) and Prymnesium blooms in European coastal waters have been reviewed in great detail because of their relevance to fish farming operations in Scotland. The two extraordinary, toxic Chrysochromulina blooms that occurred in Norwegian coastal waters - C. polylepis in 1988 and Chrysochromulina leadbeateri in 1991 - are of regional concern because of their ubiquity and unexpected toxicity (and for Chrysochromulina generally). Although previously recorded from northern European waters, neither C. polylepis nor C. leadbeateri was known to have bloomed before or to be toxic. In fact, the formal taxonomic description of C. polylepis, first identified from the English Channel in 1955 (Manton and Parke, 1962), states that it is not toxic to fish based on culture experiments. Edvardsen (2002) recently concluded, however, that " all species of Chrysochromulina are potentially toxic", and Gjøsaeter et al. (2002) state that " future blooms of C. polylepis are quite likely". It is notable that harmful Chrysochromulina blooms have been reported only from North Atlantic coastal waters - in Canada, Denmark, Finland, Norway and Sweden (see Table 3 in Edvardsen and Paasche, 1998) - despite their world-wide distribution, including the presence of C. polylepis in New Zealand waters (Chang, 1995). This apparent regional (boreal) preference, and the ecology and impacts of Chrysochromulina blooms in Scandinavian waters are highly relevant to fish farms in Scotland for several reasons: habitat similarities, the occurrence of Chrysochromulina spp. and P. parvum in Scottish waters (Table 5; Hannah and Boney, 1983), the blooming of an unidentified Chrysochromulina sp. in Loch Striven (Tett, 1980), the occurrence of ichthyotoxic C. polylepis (Manton and Parke, 1962) and Prymnesium patelliferum (Green et al., 1982) elsewhere in U.K. coastal waters, and report that toxic C. polylepis was recorded 19 times in the English Channel between 1955-1959 in all months excluding November and December (Manton and Parke, 1962). Given these distributional, bloom and autecological features of Chrysochromulina spp. and the habitat conditions found in Scottish coastal waters and sea lochs, the presence of a diverse Chrysochromulina community is expected, with local fish farms potentially vulnerable to toxic Chrysochromulina and Prymnesium blooms. Ichthyotoxic haptophyte blooms, similar to raphidophyte blooms, differ significantly from the blooms of toxic (harmful) ASP, DSP and PSP species. Blooms of the latter compromise seafood safety and shellfish marketing, but there usually is not an accompanying loss of the aquacultural stock, and harvesting of the shellfish can resume upon their depuration. In contrast, ichthyotoxic flagellate blooms cause fish farm mortalities and non-recoverable financial hardship. Ichthyotoxic flagellate blooms are less predictable, of greater intensity, duration and expanse than the blooms of ASP, DSP and PSP species affecting seafood (shellfish) safety. Although ichthyotoxic haptophyte blooms have not been reported from Scottish waters, I believe that these waters are vulnerable to their blooms. Accordingly, the summaries of Chrysochromulina and Prymnesium blooms in northern European coastal waters have been prepared in anticipation of their future occurrences in Scottish coastal waters.