What is the role and nature of programmed cell death in phytoplankton ecology?

Abstract

Cell death is a fundamental process in all metazoan organisms. In contrast, the ecological role of cell death in phytoplankton has been sorely neglected: the causes and biochemistry of cell death, and the quantitative significance of cell death in the ecology of phytoplankton populations and in broader biogeochemical cycles, are not well understood. Metazoan cell death is much better described, due to its accepted roles in the regulation of multicellular life. In metazoan cells, an influential paradigm suggests that there are two morphological outcomes of cell death, one caused by an ‘active’ pathway within the cell (so-called ‘programmed cell death’; PCD), and the other from a ‘passive’ externally-driven process (necrosis). Here, we examine the development of this paradigm, and associated concepts, in plant, animal, and microbial life, and discuss the role of cell death amongst the diverse taxa of the phytoplankton. Several recent studies suggest PCD operates in cyanobacteria, chlorophytes, and dinoflagellates. A better understanding of phytoplankton cell death will potentially provide insight into bloom development, intercellular signalling and population regulation. Understanding the role of PCD in phytoplankton life-history will likely come through examination of metabolic differentiation within phytoplankton populations, of which at present there are only isolated reports. Although bacterial metabolic differentiation (e.g. in the formation of biofilms) is well accepted, metabolic differentiation and group selection amongst microalgae are poorly understood, and are ideas which merit greater research effort. If a process similar to metazoan PCD is widespread amongst unicellular algae, then a rethinking of the ecological relationships between and within phytoplankton populations will be necessary. We highlight the semantic difficulties present in this relatively new field of study and make recommendations for future study.

Keywords:

• apoptosis
• cell lysis
• microalgae
• necrosis
• phytoplankton mortality
• programmed cell death (PCD)

Phytoplankton cell death: a neglected ecological process

Marine and freshwater phytoplankton are responsible for up to half of global primary production and play critical roles in global carbon cycling (Geider et al., 2001). Although a great deal is known about factors controlling phytoplankton growth, and their physiology during growth and division, there is relatively little understanding of the factors affecting mortality. Traditionally, phytoplankton population losses have been considered to result from grazing and sedimentation, with cells treated essentially as ‘immortal’ (Kirchman, 1999) but, paradoxically, for both marine and freshwater species, there is almost universal recognition that phytoplankton cultures often ‘crash’ in the laboratory. Such events are rarely explored, and the bases for this type of mortality have remained a mystery (Fogg & Thake, 1987). Mass mortality of microalgal populations in nature has also been observed (see below), but questions remain about the frequency and importance of such events.

The issue of phytoplankton losses due to causes other than grazing and sinking received considerable attention in the freshwater literature of the 1970s and 1980s, driven by difficulties in balancing phytoplankton growth with grazing and sinking (Kalff & Knoechel, 1978; Reynolds, 1984). Many of these studies involved species-specific approaches (Lehman & Sandgren, 1985) and whole-lake budgets (Jewson et al., 1981; Crumpton & Wetzel, 1982), which are difficult to accomplish in marine systems. In a wide range of freshwater ecosystems, terms were coined for unexplained losses, including “cell mortality” (Jassby & Goldman, 1974), “physiological death” (Pollingher, 1988), “physiological or internal losses” (v. “ecological or external losses”) (Behrendt & Nixdorf, 1993), “algal autolysis” (Coveney et al., 1977) “non-predatory mortality” (Scavia & Fahnenstiel, 1987), or simply “lysis” (Tilzer, 1984). Although production and grazing/sedimentation do balance for some taxa (Jewson et al., 1981; Forsberg 1985) – this is particularly true for the diatoms, where losses are generally well explained by sinking and grazing (Horn & Horn, 2000) – it is certainly not true for other taxa, including dinoflagellates such as Peridinium spp. (Pollingher, 1988; Sanderson & Frost, 1996). Whole-system approaches are difficult, and attempts to model freshwater carbon budgets have been problematic because of the large errors associated with production, grazing and sedimentation terms. For example, in a C-budget for an offshore Lake Michigan station, an average of 32% of annual production was unaccounted for by grazing and sinking (Scavia & Fahnenstiel, 1987), but this might not have been significant when the errors associated with production, grazing and sinking were propagated. Given the inherent variability of aquatic environments and the limitations of large-scale sampling, it is difficult to see how the high variance in such estimates could be effectively reduced. Perhaps for this reason, the past two decades have seen less work on loss terms.

In marine waters, although phytoplankton mortality terms such as “natural death” have been included in models for some time (Steele, 1974; Fasham et al., 1990), and despite recognition that closure of carbon budgets in most marine ecosystems is problematic because of “enigmatic losses” (Walsh, 1983), loss terms have only more recently received attention (Kirchman, 1999). This awareness has generally been brought about by application of new methods rather than detailed analysis of population dynamics (cf. Noji et al., 1986). For example, use of esterase activity assays has provided indirect evidence of mass lysis in phytoplankton communities (Brussaard et al., 1995; Agustí et al., 1998), whilst evidence from other physiological studies indicate that a considerable portion of cells in nature may not be viable (Veldhuis et al., 2001; Agustí 2004). There are relatively few cases where newer cell-based techniques to examine mortality have been applied to freshwater ecosystems (an exception is Vardi et al., 1999).

The potential importance of phytoplankton cell mortality is thus considerable, both at the ecosystem level, where it affects the balance of energy available to higher trophic levels via detrital foodwebs and the biogeochemical cycling of carbon, and also at the population and community levels, where it is critical to understanding bloom dynamics and species successions.

Phytoplankton cell death: concepts and quantification

Phytoplankton mortality is a poorly-defined concept. Reynolds (1984) summarized “loss terms” in a growth equation where net change in population is expressed:

where N is the biomass, k′ is growth and k_L is the sum of loss rates, which are in turn specified as:

where k_w is hydraulic washout (perhaps equivalent to horizontal advection in a marine setting), k_s is sedimentation, k_d is death and k_g is grazing.

k_d is particularly difficult to define, and its relative importance between systems is unknown. Biological components of k_d include pathogens, many of which have only recently been assessed. For example, viruses (Brussaard, 2004; Suttle, 2005) and bacteria (Cole, 1982; Imai et al., 1993) are recognized as important causes of phytoplankton mortality. Additional complexity is brought about by interactions between loss factors; the magnitude of mortality induced by nutrient limitation can be modified by bacteria (Brussaard & Riegman, 1998), for example. Parasites of freshwater and marine phytoplankton, which can cause death are known to include flagellates (Kuhn, 1998), and chytrids (Canter, 1979; Holfeld, 1998). The role of allelopathy between microalgae is attracting greater attention as interactions on the microscale come under greater focus. For example, field studies on the mutual allelopathy between the cyanobacterium Microcystis and the dinoflagellate Peridinium show that lysis and death of both species can be caused, and allelopathy may therefore be highly significant in structuring microalgal communities (Vardi et al., 2002). Laboratory studies on allelopathy between phytoplankton species have shown sophisticated mechanisms, which may act across trophic levels (Estep & MacIntyre, 1989; Casotti et al., 2005). The physical/chemical factors involved in k_d include nutrient availability and light. The linkages between the biotic and abiotic factors of the environment, and the potential of these factors to cause cell mortality, deserve greater attention in phytoplankton ecology. It is conceivable that certain environmental conditions cause the activation of catabolic processes within cells that can be adaptive, but which also have the potential to contribute to cell death processes under particular circumstances, making the interpretation of the causes of cell death difficult. In addition, the established concepts of cell death, developed to explain metazoan cell death, rely on relatively simple causal relationships. The situation for unicellular phytoplankton cells, suspended in a highly variable environment, interacting with competitors, parasites and conspecifics, is likely to be significantly more complex. Nevertheless, the paradigms that have been developed in the metazoan field are being imported into phycology, probably because of the novelty of cell death study in phycology.

Metazoan cell death: paradigms

In metazoan cell biology, the study of cell death is an enormous endeavour driven in large part by a morphological paradigm (Kerr et al., 1972), which suggests that the morphological changes evident in dead cells are diagnostic for the cell either having been killed by conditions to which it cannot adapt (necrosis), or the cell integrating signals from itself or its environment which produce a ‘programmed’ response. The ‘programmed’ response consists of new gene expression within the cell, which terminates the cell's life, thereby producing a distinctive corpse morphology. In other words, PCD can lead to apoptosis (see Table 1, which introduces simple definitions for the most common terms used in cell death studies). It should be acknowledged that the philosophical basis for distinguishing apoptosis from necrosis is debated (Malmusi & Ackerman, 1999; Ratel et al., 2001; Kanduc et al., 2002), since the specificity of molecular markers (e.g. TUNEL labelling, see below) that are claimed to be diagnostic for apoptosis may not be as specific as claimed (Labat-Moleur et al., 1998; Jerome et al., 2000). What is not questioned however, is that throughout all domains of life, cellular pathways exist that can lead to a biochemical cascade resulting in cell death (Aravind et al., 2001; Ameisen, 2002; Bidle & Falkowski, 2004). The impact of intercellular signalling and environmental conditions on these pathways is now the focus of laboratory investigations into cell death. Although our understanding of the molecular basis of these pathways in phytoplankton has improved (Bidle & Falkowski, 2004), the role of cell death in phytoplankton ecology has not been well addressed. Seeking an evolutionary explanation for the existence of these pathways amongst unicellular algae is a compelling reason for the study of cell death in microalgae. With this in mind, it is worth first discussing the evolutionary arguments for the development of PCD in metazoans.

Table 1 . Simple definitions of terms used to describe the causes of cell death. The term used should reflect the evidence demonstrated; terms like ‘PCD’ should only be used for cells where specific molecular and morphological markers have been demonstrated.

Term	Definition	Comment
Natural cell death	Death that results from an exhaustion of division potential (age), energy limitation (starvation), or intolerable physical conditions (e.g. desiccation).	To avoid exhaustion of division potential, most phytoplankton probably undergo rejuvenation through sex. Amongst phytoplanktonic eukaryotes, sexual cycles are incompletely characterized.
Programmed cell death (PCD)	Cell death (colourfully called ‘cell suicide’) resulting from gene expression within the moribund cell. PCD is mediated by a set of enzymes called caspases and results in a set series of biochemical events as the cell dies, which produce a distinctive dead cell morphology (e.g. apoptosis).	PCD is essential for the successful construction of complex body plans (differentiation) and for homeostasis. PCD is thought to be activated by intercellular signalling, or by environmental stress (cf. natural cell death), causing caspase activation and cell death. PCD can be a cause (though not the ultimate cause) of cell death once triggered.
Apoptosis	A set of morphological characters seen in dying/dead metazoan cells during/after the activation of PCD. These include DNA (and nuclear organelle) fragmentation, compartmentalization of cell contents into apoptotic bodies, cell shrinkage, membrane ‘blebbing’, loss of membrane integrity, and phosphatidylserine inversion.	Apoptotic cells are thought to be produced during the death of phytoplankton and yeast cells (see text) and therefore PCD is thought to occur in these taxa. Apoptosis cannot be a cause of cell death; apoptosis is a consequence of PCD.
Lysis	Rupture of the cell membrane, dissolution of cell contents.	Recognizable as cell loss/disappearance. Quantification of phytoplankton cell lysis has been attempted with the dissolved esterase method. Lysis can be caused by pathogens and parasites. Viruses may subvert cellular machinery to produce ‘PCD-like’ processes prior to lysis.
Necrosis	Cell death that does not involve gene expression. In metazoan cells, necrosis involves loss of membrane integrity, cell swelling and then lysis.	Necrosis results when cells are killed in the absence of any metabolic self-involvement.
Paraptosis	A form of PCD, triggered by a molecular signal, which results in extensive vacuolization and mitochondrial swelling in the absence of caspase activation.	Paraptosis produces an alternative morphological outcome to apoptosis.
Senescence	The process of growing old; difficult to recognize in unicellular organisms. In phycology, ‘senescence’ has been used to refer to the appearance of cells after exponential growth, i.e. when many cells are probably dead due to natural cell death.	In microalgal work, the use of ‘senescence’ is problematic: the appearance of cells that have died is used to explain the process that caused cell death. However, an outcome of a process cannot simultaneously be the cause of the process: the cause of death in these instances is not senescence. ‘Senescence’ in multicellular plants involves PCD.

‘Gain of function’ as the evolutionary basis for metazoan PCD

The concept of PCD was developed by plant biologists in the 1920s to explain plant cell death during fungal infection (Jones, 2001). A more general role for PCD was conceived by Leopold (1961) whereby PCD regulates ontogenetic changes (such as the terminal differentiation of xylem cells) and, with the demonstration of the purely nuclear basis of PCD (Yoshida, 1962), PCD became well established in plant physiology as the governing mechanism of multicellularity (Jones, 2001). For plant physiologists, the recognition of a ubiquitous role for PCD in plant development derives from an appreciation of the link between metazoan ontogeny and events at the single cell level. The balance between cell proliferation and cell death in the shaping of the organism is the essence of metazoan PCD. In plant physiology, there was an early awareness that most of the cell death that multicellular organisms undergo is programmed, and that the term ‘senescence’ could be used for the ‘regular and normal’ orchestrated molecular dismantling (i.e. PCD) of the cell for recycling and reallocation (Jones, 2001). In this scheme, PCD facilitates the sculpting of the organism during ontogeny in a ‘developmental programme’. Plant development uses PCD in order to create ‘functional cell corpses’ (Jones, 2001), which are indispensible to plant structure and function, such as in the aforementioned terminal differentiation of xylem cells, or corpses which are simply surplus to requirements. The shedding of leaves in autumn by deciduous trees is an often quoted example of PCD in action, and provides the inspiration for the term apoptosis (from the Greek for ‘falling away’).

The idea that PCD in plants minimizes the effects of pathogens has developed into the concept of the ‘hypersensitive response’ (Lam et al., 2001), whereby plants trigger localized mass cell death in order to limit the spread of a fungal or bacterial infection. Interestingly, mortality and lysis of phytoplankton cells has been hypothesized to serve the same purpose amongst freshwater phytoplankton populations (Canter, 1979). Possession of hypersensitivity may benefit groups of genetically similar cells by limiting the spread of infection through the population. For phytoplanktonic organisms, this idea is controversial to some, since it implies the existence of traits in individuals which benefit the group (i.e. the species) at the expense of individual fitness (see next section).

The study of cell death in animal cells, during development, homeostasis, and cancer, started later than in plants, but has come to dominate the literature. The concepts of cell death, which arose in plant work (see review of Jones, 2001), have been somewhat overshadowed by the pathological work of Kerr et al. (1972), who proposed the paradigm of apoptosis v. necrosis (see above, and Table 1). An important later development in PCD studies was the idea that PCD could be the ‘default’ fate of cells, with cells relying on the detection of signals (‘social factors’) from neighbouring, equally differentiated, cells in order to prevent the activation of PCD (Raff, 1992); in this way, cells will die if accidentally carried to the wrong position in the body. Many biomedical scientists understand apoptosis to be synonymous with PCD. Strictly speaking, since PCD can occur without resulting in apoptosis (certainly in plants, and most likely in other taxa), apoptosis is better thought of as simply one morphotype that results from PCD (Jones, 2001; see Table 1). As discussed below, it is important to be aware that the term ‘PCD’ has limitations, and carries connotations. It is, however, in widespread use. The original description of apoptosis in rat cells (Kerr et al., 1972) described a relatively simple sequence of morphological changes in dying cells: cells shrink, compartmentalize their organelles and lastly, undergo lysosomal fusion by neighbouring cells. A number of molecular markers have since been added to this scheme:

1. Activation of caspases. Caspases are a family of proteases that have the unusual characteristic of cleaving peptide chains after an aspartate residue. Caspases are responsible for the morphological changes displayed by apoptotic cells (Cohen, 1997). Caspase activity per se is not unique to cell death; caspases are likely also to be involved in general catabolic (‘housekeeping’) pathways (Vaux & Korsmeyer, 1999). However, the greater expression of caspases during cell death, as well as their dire consequences, reveal a clear role in the biochemical mediation of PCD.

2. Inversion of an aminophospholipid (phosphatidylserine), which is usually found on the internal side of the plasma membrane, to the outside of this membrane. Externalization of phoshatidylserine is thought to represent a process of cell self-labelling for subsequent phagosomal disposal in metazoans (Martin et al., 1995). Thus, although there is no obvious reason to suspect such a process in unicells, phosphatidylserine inversion has been detected in yeast and chlorophytes (see below).

3. Low molecular weight DNA fragmentation, detectable by the comet assay and later by TUNEL/ISEL assays, is thought to be an early marker of apoptosis and thus PCD (Patel et al., 1995) in animal cells. The specificity of this marker to apoptosis is questioned (Graslkraupp et al., 1995; Jerome et al., 2000; Kanduc et al., 2002).

The role of PCD in cancer and disease has been the driving force for the biomedical study of PCD. On the simplest level, cancers represent the failure of PCD mechanisms; cancers are uncontrolled cellular proliferations in tissues that require balanced cell growth and loss for normal functioning. The original animal-based morphological description of apoptosis portrayed the process as a ‘tidying-up’ of cells for nutrient re-assimilation by neighbouring cells. Apoptosis thus contrasted strongly with the previously described, and better known form of animal-cell death, necrosis, caused by physical factors such as heat or mechanical damage. Kerr et al. (1972) proposed that the earlier terms ‘physiological cell death’ and ‘autolysis’ could be replaced by apoptosis and necrosis, respectively. This has occurred in the biomedical sciences, although in other disciplines the terms ‘autolysis’ and ‘physiological death’, are still found. For example, in biological oceanography ‘autolysis’ is a loss term used in the modelling of phytoplankton production (Steele, 1974; Fasham et al., 1990). Poorly-defined terms such as ‘autolysis’ reflect the system-level interest of oceanography, as well as a paucity of data on phytoplankton cell death. However, since the fate of aquatic production will be strongly influenced by the form of phytoplankton cell death (Kirchman, 1999), understanding the nature of cell death should be of great interest to aquatic microbial ecologists.

Group selection and PCD in phytoplankton: altruism?

Much of the novelty in this research area comes from the realization that unicellular life could be able to organize itself into co-operating groups; and that individual cell death may be part of a population regulation originating amongst the algae themselves, in response to environmental cues. The idea of co-operation amongst groups of unicells is old news to bacteriologists and some microbial ecologists, but may be new, and controversial, to classically-trained ecologists. We define group selection here simply as the selection for traits that are beneficial to a population at the expense of an individual possessing the trait. There are two arguments for supposing that group selection may have produced adaptations (induction of cell death) that benefit groups of conspecific cells over the individual cell. Firstly, it may be that the ‘group’ in some phytoplanktonic populations has sufficient kinship such that selection at the level of individual cells, and selection at the level of groups of conspecific cells, has become blurred. Phytoplankton populations can be composed largely of asexually-dividing cells, which, under conditions of rapid growth and low mixing, can consist mainly of clonal cells, and could therefore experience selection for group traits. Early allozyme work by Gallagher (1980) showed that blooms of the diatom Skeletonema costatum had ‘distinct and prevalent’ forms, which supported this idea. However, seemingly in contrast, recent microsatellite analysis of blooms of the diatom Ditylum brightwellii indicate that genetic diversity remains high during blooms, with many ‘different’ clonal lines making up the population (Rynearson & Armbrust, 2005). This observation, that the genetic structure of phytoplankton blooms is ‘heterogenous’ (Medlin et al., 2000), has led to the conclusion that bloom conditions will not favour group selection (Thornton, 2002).

But, are clonal differences more important than conspecific affinity in terms of group adaptation to changing environmental conditions? The level of genetic discrimination now possible with modern molecular methods is high, and may cause us to assume physiological heterogeneity at the subspecies level unnecessarily, at the expense of examining population level heterogeneity. There is, after all, no reason to think that differences in intraspecies phylogenetic markers (e.g. ITS regions) translate to differences in physiological responses (Lowe et al., 2005): the two do not match. The second argument for group selection in phytoplankton comes from multilevel selection theory, which proposes that natural selection operates on a nested hierarchy of units (genes, individuals, social groups, species and multispecies communities; Sloan Wilson, 1997): adaptations do not only evolve via selection on individuals. Adaptation only at the individual level would create highly dysfunctional groups (Sloan Wilson, 1997), and would have acted against the evolution of microalgal differentiation and colony formation. Experimental and theoretical approaches have convincingly shown that genetically based interactions among individuals contribute to a response to group selection (Goodnight & Stevens, 1997). In the case of phytoplankton, the ability to sense, and respond to, the presence and physiological status of conspecifics will have been key to the evolution of cell death processes with population-level benefit.

Using the term ‘PCD’ – a health warning

In Table 1, we define PCD as the cell death resulting from specific gene expression within a moribund cell. This definition is slightly unsatisfactory since it is only the observation of elevated caspase activity (and other PCD ‘markers’) that permits us to say that the cell is about to die, and that its cellular apparatus is participating in its own death. Thus, circular reasoning is applied – the proposition (that cells are undergoing PCD) is proved via an assumption that PCD ‘markers’ are produced in cells undergoing PCD. This is less unsatisfactory than suggesting that the morphological characters of a dead cell which died due to PCD (i.e. apoptosis) indicate that the cell was killed by PCD; a dead cell cannot be killed by anything. These problems of interpretation are related to difficulties in defining ‘the moment of death’ in single cells, and have led to substantial confusion in interpreting observations of cell death between cell types (Jones, 2001). There is also a danger of ‘fitting’ observations to a morphological paradigm unjustly, even when there is no particular reason to expect that cell death processes will be the same in widely different cell types. In some ways this last problem is exacerbated by a preoccupation in getting assays (such as TUNEL-labelling) to give positive results.

Secondly, there are significant semantic ramifications in exporting the concept of PCD from multicellular to unicellular life (Ratel et al., 2001). If a cell dies at a specific point in multicellular development (using, for example, time and position cues) then the cell acts as part of a greater system, and death occurs in order to fulfill a functional role. This is the established concept of PCD. Ratel et al. (2001) have argued that the death of unicells, outside of a multicellular grouping, is better thought of as a ‘cell death programme’ since, in cell cultures, the cell is the system, and cell death is the result of a biochemical change whose aim is, therefore, to stop existing. Since PCD, by the original usage, is caused by biochemical changes whose aim is to maintain (multicellular) existence, then most experiments on cell cultures cannot be thought of as experiments into PCD, but simply experiments into the ‘cell death programme’ that cells possess (Ratel et al., 2001). Therefore, such experiments may have limited use in illuminating PCD (but are highly achievable). Whilst this is a lesser problem for unicellular microalgae, it does highlight the great potential for semantic confusion when exporting concepts across disciplines, and teaches us how using the term ‘cell death programme’ is safer than bringing in the concept of a ‘programme’ which is extrinsic to the cell. However, it is true that the term ‘PCD’ is now well established and has permeated across disciplines, and that the term appears to be here to stay. We recommend that the term ‘PCD’ should be applied only after the demonstration of caspase activity, and the dependence of this caspase activity on de novo protein synthesis. If this is not possible, then the simpler term ‘cell death’ should be preferred.

PCD in bacteria, yeast and eukaryotic protists

The ‘traditional’ view that microbial life should have no need for PCD pathways was based on the notion that microbial populations are composed of competing, selfish individuals not subject to any selective pressure for characters that increase group fitness. This view has been strongly challenged by developments in group selection theory as well as empirical demonstrations showing that microbes such as bacteria and yeast can be metabolically heterogenous and kin-linked. Colony-level bacterial processes, such as the lysis of mother cells during sporulation and lysis of vegetative cells in myxobacterial fruiting body formation, involve PCD (Lewis, 2000). In bacteria, so-called ‘autolytic’ processes represent PCD because autolysins – compounds that mediate ‘autolysis’ – are manufactured by the cells themselves (Lewis, 2000). The concept of PCD, developed to explain metazoan cell phenomena, has been bridged to the unicellular world through the observation of such kin-linked groups and colonies, for which the PCD of individual cells may give nutritional benefit for conspecifics, or some other advantage. Thus, the evolutionary stability of a self-mediated lytic process in bacteria (Lewis, 2000) and also yeast (Frohlich & Madeo, 2000) has been explained, to some people controversially, in terms of altruism. Yeast displays apoptotic markers (DNA fragmentation and phoshatidylserine inversion) during cell death induced by oxidative stress (Madeo et al., 1997 1999). Therefore, yeast cell death, under certain conditions, is thought to be PCD. Similarly, observations of DNA fragmentation have been taken as evidence of PCD in the unicellular parasite Leishmania (Lee et al., 2002), although, in Leishmania cell death, DNA fragmentation is not dependent on caspase activity (Zangger et al., 2002), and is therefore different from metazoan PCD. The slime mould Dictyostelium discoideum, a single-celled eukaryote, also shows multiple PCD ‘markers’ (Arnoult et al., 2001; Tatischeff et al., 2001) during differentiation between single-celled to group life-cycle stages. PCD in D. discoideum is thus thought to have a functional role in the regulation of a primitive multicellularity. In all these cases, the functional basis and evolutionary stability of PCD relies on a view of microbial life as having the potential to form consortia of co-operating colonies and cells which have a more complex pattern of group development than previously realized.

Compared with metazoan cells, our knowledge of PCD in unicellular organisms is patchy, and variations on the metazoan paradigm (e.g. failures to detect caspase activities in yeast and Leishmania, and lack of DNA fragmentation in D. discoideum: Olie et al., 1998), are still interpreted as PCD. One should note however, that genes encoding for proteins known as ‘paracaspases’ (bacteria) and ‘metacaspases’ (plants and fungi) have been recognized from sequence work, and that all these enzymes share common active sites (Uren et al., 2000). Key elements of cell death pathways – apoptotic protein domains – are thus highly conserved across taxa (Aravind, 2001) and some cell death workers propose a unicellular origin of PCD with a later adoption by multicellular life for functional advantage (Ameisen, 2002). Such an origin for PCD, coupled with observations of an ecological foundation for ‘functional’ PCD in microbes (through differentiation), argues against the initially counterintuitive existence of PCD in unicellular organisms.

Phytoplanktonic cyanobacteria: differentiation, metabolism and PCD

As occurs in dinoflagellate populations (see below), rapid and seemingly spontaneous cyanobacterial population crashes have been noted in nature and in culture (Eloff et al., 1976; Vonshak et al., 1996). The causes of these crashes appear to be related to oxidative stress and nutrient limitation, which combine to cause rapid mass cell death. A PCD pathway (involving caspase activity and DNA fragmentation) has recently been described in Trichodesmium in response to nutrient limitation, high light and oxidative stress (Berman-Frank et al., 2004). PCD as a differentiating mechanism is an attractive idea for explaining colony development in cyanobacteria: Trichodesmium may use PCD as a way of generating small groups of dispersal cells (hormogonia) that are released from parent colonies after death and during unfavourable conditions (Berman-Frank et al., 2004). Dispersal is also thought to be the adaptive basis for PCD in Calothrix, in which colony fragmentation is achieved through the PCD of specific cells (Adamec, 2005). Colony differentiation involving mortality has also been noted in Anabaena variabilis. A. variabilis produces heterocysts that are viable for only 2–3 days, independent of media and growth conditions, and that show a loss of cell contents during the loss of viability (Reddy et al., 1987). Heterocysts of A. variabilis, therefore, appear to undergo a process of terminal differentiation similar to that shown by higher plants. This case, similar to that of the colonial green alga Volvox (see below), suggests that some cyanobacteria can differentiate through mortality, leading to the corollary that control of single cell mortality may be an emergent property of cell groups, and under genetic control. Like Trichodesmium, another species of Anabaena (A. flos-aquae) displays ‘markers’ of PCD such as DNA fragmentation, increased protease activities and plasma membrane permeability (Ning et al., 2002). Furthermore, A. flos-aquae shows a population death rate that is a function of nutrient-controlled growth rate, and is under circadian, genetic control and therefore programmed in nature (Lee & Rhee, 1999; see also Brussard et al., 1997). Data on the ‘background’ mortality that may be present in growing phytoplankton populations is patchy; it is a common assumption that, if a (usually laboratory) population is growing, then all the cells are in good health. This assumption is confounded by the considerable metabolic heterogeneity of both natural and laboratory microalgal populations, as revealed through physiological assays (Veldhuis et al., 2001; Brussaard et al., 2001; Agustí 2004). Added to this, kinetic analysis shows that populations of symmetrically dividing cells can simultaneously contain dividing, non-dividing and dying cells whilst showing exponential growth, provided that the death rate is constant (Sheldrake, 1974). Thus, considerable metabolic heterogeneity is possible in phytoplankton populations, although little is known about it, and what does exist is likely to have been overlooked up to now.

A competing idea on the cause of cyanobacterial cell death is based on the dependence of bacterial mortality on growth phase (Aldsworth et al., 1999; Mason et al., 1986). This concept, perhaps more familiar to phycologists, suggests that, when growth is arrested by some form of sublethal stress (e.g. nutrient limitation in culture), growth becomes uncoupled from metabolism, leading to cell death through an oxidative burst (Aldsworth et al., 1999). This idea has been used to explain cell death in cultures of phytoplanktonic Synechococcus (Sakamoto et al., 1998) and Synechocystis (Suginaka et al., 1999); metabolic imbalance (unbalanced growth) causes cell death, via unavoidable generation of oxidative stress. That the uncoupling of growth and metabolism causes unavoidable oxidative stress is an attractively simple explanation for cell death, which does not need to encompass an evolutionary explanation of PCD – cells are simply killed due to ‘design constraints’. However, uncoupling of growth and metabolism cannot explain the mechanism of cyanobacterial PCD during differentiation. Differentiation clearly occurs in cyanobacteria and, since metazoan PCD is regulated through cell signalling, the role of signalling in microalgal PCD needs greater attention, and will benefit from studies into the signalling molecules produced by microalgal cells in culture, and also in biofilms. Signalling within bacterial biofilms can lead to PCD via quorum-sensing (Oleskin et al., 2000), or through the induction of prophages and autotoxins (Webb et al., 2003). Quorum sensing (density-dependent gene expression) in bacteria also provides a plausible mechanism for ontogenetic differentiation. Studies of bacterial populations suggest hitherto unrealized complexity in the way microbial populations are regulated. Whether such complexity can be shown amongst microalgal populations is an open and exciting question, which could considerably complicate existing notions of the factors that regulate food-web structure. For example, intercellular signalling (which is allelopathic in nature) between cyanobacteria and dinoflagellates may be sufficiently sophisticated to harness cell death pathways (Vardi et al., 2002); and the exciting discovery of prokaryote–eukaryote signalling (Joint et al., 2002) shows that there is much to be learned about signalling within microbial communities.

PCD in chlorophytes: ‘functional’ and ‘non-functional’ PCD

Work on chlorophytes has revealed two instances of PCD, with two evolutionary interpretations. Older literature suggests that chlorophyte cell death can occur when a type of multicellularity (i.e. differentiation) is an objective of coordinated cell division. In the colonial alga Volvox, somatic cells die after a set number of divisions in a process of ‘programmed senescence’ (Pommerville & Kochert, 1981; Kirk, 1998). ‘Senescent’ Volvox cells shrink, lose chlorophyll, accumulate lipid and then lyse. However, up to the point of lysis (during ‘senescence’), the nucleus is unchanged and new protein expression occurs (Hagen & Kochert, 1980; Pommerville & Kochert, 1981). On a morphological basis, these changes appear similar to what happens in ‘old’ (i.e. nutrient-limited) algal cultures (Freudenthal, 1962; Prezelin, 1982; Fogg & Thake, 1987). The similarities between cell death during Volvox differentiation, which involves nuclear persistence and protein expression, and the cell death that is routinely observed in old cultures, might suggest a common pathway, involving self-mediation via PCD. In this sense, algal senescence (see Table 1) could be functionally equivalent to multicellular plant senescence (which is controlled via PCD) with protease expression and cellular dissolution occurring in order to fulfil a functional role. One possible functional role of such a pathway in unicellular algae (and yeast) is to recycle the organic remains of cells for use by conspecifics (Gomez et al., 1974; Frohlich & Madeo, 2000). This idea is controversial and requires a convincing demonstration of conspecific nutritional (or other) benefit resulting from the cell death of some proportion of the cells (Franklin & Berges, 2004). That cell death could provide nutritional benefit for remaining conspecifics is an idea that is easier to accept for yeast, because yeast cell division can be clearly asymmetric (e.g. in budding yeast; Laun et al., 2001). Asymmetric division is also possible in dinoflagellates (Silva & Faust, 1995), although whether the production and death of different types of cells gives nutritional benefit to remaining cells is unknown, and the utility of such a strategy is obscure. It is also difficult to see how nutritional benefit could be restricted to conspecifics in the planktonic environment. Another problem is that asymmetric division in dinoflagellates occurs during stressful conditions (Reguera & Gonzalez-Gil, 2001), when sexual cycles may also be induced, leading to difficulties in interpreting the causes of changed cell morphology. ‘Small cells’ are thought to be capable of acting as gametes in some dinoflagellates (Silva & Faust, 1995), but the evidence for this is not always conclusive (Reguera & Gonzalez-Gil, 2001).

Returning to chlorophytes, a competing viewpoint on the origin of unicellular PCD has been suggested by detailed work on the chlorophyte Dunaliella tertiolecta. D. tertiolecta, when placed in darkness, undergoes mass cell death, with the apoptotic morphology of dead cells suggesting the existence of a PCD-like process. Insertion of viral lysogenic genes into the D. tertiolecta genome at some point in evolutionary history may explain the existence of this pathway (Berges & Falkowski, 1998; Segovia et al., 2003). Under environmentally stressful conditions, when silencing or repression systems fail (Baulcombe, 2002), these lysogenic genes may be activated in a maladaptive fashion. In this scheme the PCD pathway in unicells is an ‘accident of history’ and offers no contemporary functional advantage. In the case of D. tertiolecta, the ‘metabolic imbalance’ hypothesis on the cause of cyanobacterial cell death (see above) could also be invoked; thus giving a good example of the difficulty of ascribing single causes to observations of microalgal mortality. The lysogenic gene hypothesis (Berges & Falkowski, 1998; Segovia et al., 2003) is unique in neatly encompassing the latter possibility in an evolutionarily coherent fashion. Whether this hypothesis can be applied to other instances of microalgal cell death remains to be seen.

Population crashes and cell death in dinoflagellates

Attempts to ‘balance’ dinoflagellate populations by quantifying cell production and cell loss (as grazing and sedimentation) have not succeeded (Pollingher, 1988; Sanderson & Frost, 1996); unexplained population ‘crashes’ are a feature of many dinoflagellate bloom populations (Heiskanen, 1993; Heiskanen & Kononen, 1994; Lindahl, 1983; Usup & Azanza, 1988). In the freshwater dinoflagellate Peridinium gatunense, PCD is thought to be induced by CO₂ limitation because, during such stress, cell death can be blocked by cysteine protease (caspase) inhibitors, and dying cells show positive DNA fragmentation (Vardi et al., 1999). The symbiotic dinoflagellate Symbiodinium sp. is also thought to undergo PCD (displaying an apoptotic morphology) due to heat stress (Dunn et al., 2002). However, in the dinoflagellate Amphidinium carterae, cell death in darkness, and during culture senescence, results in similar morphological changes without definitive apoptotic characteristics (Franklin & Berges, 2004). Mortality in A. carterae shows greatest morphological similarity with paraptosis (see Table 1) – an alternative outcome of PCD (Sperandio et al., 2000), which was originally observed in mammalian nerve tissue. Interesting work on cell death in the coccolithophore Emiliania huxleyi also reveals a paraptotic morphology resulting from cell death (Bidle et al., 2005). Further molecular and cytological characterization of cells in nature and in culture is needed to identify PCD and to test its ecological importance in a wide range of species. Given the often rapid disappearance of dinoflagellate blooms, PCD in response to environmental stress or energy limitation could potentially explain the enigmatic population crashes seen in nature. Noji et al. (1986) considered that dinoflagellate (Ceratium spp.) bloom declines in the Baltic are caused by the exhaustion of division potential, coupled with an inability to enter a sexual cycle due to the unfavourable environmental conditions that exist at the end of blooms. Recent work highlighting the importance of environmental stress in the sexual induction of Volvox (Nedelcu et al., 2004) suggests that sex and death result from differing magnitudes of environmental stress, which produce a common intracellular consequence (oxidative stress) – ‘mild’ oxidative stress triggers the induction of sex, whereas ‘severe’ oxidative stress kills the cell. In metazoan cells, oxidative stress is considered to be a key ‘trigger’ of PCD (Jabs et al., 1996), and the role of oxidative stress is attracting increasing attention in investigations into microalgal cell death.

In dinoflagellates, cell death can also be caused by direct interspecific contact (Costas et al., 1993; Uchida et al., 1995), and the importance of this interaction in nature is poorly known. Allelopathic interactions within and between taxa may be sophisticated enough to harness PCD mechanisms, but first we need to be able to recognize the extent and nature of PCD in the divergent taxa of the phytoplankton.

Conclusions and suggestions for further work

The investigation of phytoplankton PCD is highly novel. It is, at present, difficult to make generalizations about how PCD operates in phytoplankton, and whether it has greater similarities with animal, plant or fungal models. Whole genome work in Thalassiosira (Armbrust et al., 2004) and Chlamydomonas (Shrager et al., 2003) indicates that microalgal cell death shares certain similarities with higher plants, such as enzyme homology with plant/fungal metacaspases instead of animal caspases, and a lack of animal cell PCD regulators (e.g. Bcl-2 family, p53). In addition, the role of the mitochondrion is unclear: in animal cells the mitochondrion appears to be central but, in plants and protists, its role is far from clear. In Fig. 1 we attempt to summarize the ecological and cellular factors involved in microalgal cell death, along with the current evolutionary explanations for phytoplankton PCD. Much remains to be done. We propose that one of the most important tasks is to assess the quantitative significance of PCD as a loss process, and how PCD might affect species successions and the conversion of carbon (c.f. Kirchman, 1999). In order to do this, we need robust indicators. Some, like DNA fragmentation, will be difficult to apply in the field because of the need for high biomasses and the manipulation of samples. Other techniques, such as cell-specific labelling using Annexin-V and caspase substrates and inhibitors, show more promise when coupled with flow cytometry (Lederman, 2004). Caspases are a good target for developing new approaches as they are reasonably specific to cell death and relatively easy to measure. Freshwater phytoplankton populations are likely to be more tractable to investigation due to their smaller scales, fewer species, high biomasses, easier access and more regular blooms.

Fig. 1. Summary of the causes, consequences and evolutionary implications of cell death processes in unicellular phytoplankton. Simple cause and effect relationships are probably complicated by metabolic heterogeneity amongst unicellular populations, the influence of undocumented life-history processes (sexual cycles, asymmetric division) and intercellular signalling.

Recently, it has been shown that the dinoflagellate Alexandrium ostenfeldii can use signalling to encyst temporarily and thus avoid parasite infection, implying that intercellular communication evoked by biotic interactions can be an important factor in bloom development (Toth et al., 2004). Can the fate of conspecific cells in other phytoplankton species be modulated by water-borne signals, which act on the group level? If a large proportion of the surrounding cells share the same genome, then signal transmission amongst unicellular organisms can, evolutionarily, be compared with systemic signalling in a multicellular plant (Toth et al., 2004), and it is thus plausible that group-selected traits could have evolved. Group level traits could be revealed through examination of metabolic heterogeneity in single populations, and the response of populations to environmental stress. The best way to do this in the laboratory will be with continuous culture techniques (Brussard et al., 1997; Lee & Rhee, 1999). Entraining monospecies cultures to constant conditions and examining mortality rates, and the biochemical characteristics of dead cells, will shed light on intraspecies population regulation and the nature of cell death. Metabolic heterogeneity can be assessed with live/dead stains, stains for respiratory activity (e.g. CTC activity; Berman et al., 2001) and hydrolytic enzyme activities. Flow cytometry, coupled with cell sorting of heterogeneous subpopulations, is likely to be key in these types of experiments, and counterstaining techniques will be useful in understanding the physiological state of individual cells.

The role of extracellular signalling between phytoplankton needs examination through experiments designed to isolate and characterize the biologically active compounds exuded by phytoplankton cells, and their relationship with the induction of cell death programmes. A good example of the complexity in trophic interactions which can be seen with such an approach is found in Casotti et al. (2005). Understanding of surface receptors and signalling pathways will increase as genomic data is generated, and these data should prove useful in predicting the role of signalling. This field has great potential for generating major new insights and, given the revolution in understanding in bacteriology gained from observations of quorum sensing and cell death processes, microbial ecology in its entirety could be set to become much more complex.

Acknowledgements

We thank the reviewers for their positive and constructive comments on ways to improve the manuscript.