Multiple Modes of Cell Death Discovered in a Prokaryotic (Cyanobacterial) Endosymbiont

Programmed cell death (PCD) is a genetically-based cell death mechanism with vital roles in eukaryotes. Although there is limited consensus on similar death mode programs in prokaryotes, emerging evidence suggest that PCD events are operative. Here we present cell death events in a cyanobacterium living endophytically in the fern Azolla microphylla, suggestive of PCD. This symbiosis is characterized by some unique traits such as a synchronized development, a vertical transfer of the cyanobacterium between plant generations, and a highly eroding cyanobacterial genome. A combination of methods was used to identify cell death modes in the cyanobacterium. Light- and electron microscopy analyses showed that the proportion of cells undergoing cell death peaked at 53.6% (average 20%) of the total cell population, depending on the cell type and host developmental stage. Biochemical markers used for early and late programmed cell death events related to apoptosis (Annexin V-EGFP and TUNEL staining assays), together with visualization of cytoskeleton alterations (FITC-phalloidin staining), showed that all cyanobacterial cell categories were affected by cell death. Transmission electron microscopy revealed four modes of cell death: apoptotic-like, autophagic-like, necrotic-like and autolytic-like. Abiotic stresses further enhanced cell death in a dose and time dependent manner. The data also suggest that dynamic changes in the peptidoglycan cell wall layer and in the cytoskeleton distribution patterns may act as markers for the various cell death modes. The presence of a metacaspase homolog (domain p20) further suggests that the death modes are genetically programmed. It is therefore concluded that multiple, likely genetically programmed, cell death modes exist in cyanobacteria, a finding that may be connected with the evolution of cell death in the plant kingdom.


Introduction
Programmed cell death (PCD) is a self-inflicted geneticallybased cell death mechanism in eukaryotic organisms, and genetic and cytological studies have led to the identification of pathways and molecular components that underlie this process [1][2][3]. Emerging evidence suggests that PCD may also be operative in prokaryotes, which were previously considered to be immortal unless killed or eaten by predators. PCD may for instance be involved in developmental life cycles and in optimizing adaptations in natural prokaryotic populations subjected to environmental stresses [4][5][6]. Mechanisms that balance life and death are also known to protect against antibiotics and macrophages during bacterial biofilm formation [7].
Studies of prokaryotic PCD have primarily focused on autolysis and in a limited number of bacteria, such as Echerichia coli [7][8][9]. Homologs of metacaspases, a family of caspase-like proteins originally described in plants, fungi and protozoa [10] have now been identified in numerous prokaryotic genomes and here we use the term metacaspase for these prokaryotic homologs. These genes were recently shown to have a limited distribution (18%) among the 1400 sequenced prokaryotic genomes [11], although they are widely scattered over the bacterial radiation and particularly frequent among phenotypically complex representatives of Alphaproteobacteria, Deltaproteobacteria and Cyanobacteria. Notably, some of these bacteria hold intriguing evolutionary positions related to the origin of mitochondria and chloroplasts and the evolution of eukaryotes.
The globally widespread cyanobacterial phylum exhibits Gramnegative characteristics, introduced oxygenic photosynthesis, and their evolutionary significance in shaping the biosphere is indisputable [12]. Besides being widespread as free-living entities, cyanobacteria entered pigment-free eukaryotes ,2500 MYA in a monophyletic event [13,14], and subsequently evolved into chloroplasts. Through this endosymbiosis, plants conquered global land masses 5-600 MYA ago [12][13][14], and today constitute the largest biomass on Earth. Whereas the identity of the cyanobacterial progenitor of chloroplasts remains unknown, some extant cyanobacteria (genus Nostoc) form intimate relationships with plants in a range of divisions [15]. A driving force behind contemporary cyanobacterial-plant symbioses is the ability of the cyanobacterium to fix atmospheric dinitrogen and to fully support a combined nitrogen-independent life-style of the host plant. In the small pteridophyte (fern) Azolla, cyanobacteria occupy cavities located within each individual plant leaf [16,17]. The unique interaction between the partners is reflected in the loss of cyanobacterial autonomy, a consequence of a massive pseudogenization and an eroding genome [18], the first discovered in a plant symbiosis. Moreover, a small cyanobacterial inoculum is kept during the sexual reproduction of the Azolla plant using a vertical transfer mechanism that is unique amongst plant symbioses [19,20]. To date, PCD-like events have been documented experimentally in some free-living cyanobacteria, notably the unicellular Microcystis and the filamentous genera Anabaena and Trichodesmium [4,21,22]. Indeed, the genomes of the two latter are by now also known to hold particularly many metacaspase genes, although some may have other roles [11].
To deepen our understanding of the role of death events in prokaryotes in general, we selected the cyanobacterium subject to an endosymbiotic lifestyle in an ancient plant (i.e., Azolla) as our model system. We hypothesized that PCD events may underlie some cell developmental features typical of the morphologically complex endosymbiont [16,17] as well as in the strict regulation of the small cyanobacterial population size inside the plant and during the vertical transfer, and hence in symbiotic homeostasis [18,20]. Several approaches were optimized and used to monitor cell death events, such as light-and electron microscopy,, following published guidelines to reduce the risk of artifacts [23,24] as well as biochemical assays and genetic analysis. Our data emphasize the existence of sophisticated cyanobacterial cell death events, and provide structural profiling for cell death as well as important clues for a better understanding of potential PCD processes in prokaryotes. We discuss evolution of PCD in higher life forms, such as plants, in light of our findings.

Results
In order to search for evidence for the presence of programmed cell death (PCD) in a prokaryote a battery of approaches were used (morphological, biochemical, ultrastructural and genetic), each specific in its own way. The biological test material selected was a cyanobacterium living in symbiosis within leaves of the small water-fern Azolla microphylla (from now on Azolla) grown under green-house conditions.

Morphological Criteria
Approximately 25,000 cyanobacterial cells were isolated from leaf cavities of 102 Azolla microphylla sporophytes, representing all developmental stages of the plants (Figure 1, left panel), and from 55 sporocarps, the reproductive generation of ferns ( Figure S1 and S2). Of these cells, 9,043 were examined under bright field and fluorescence microscopy (filter sets for blue and green light and UV light). In total, 17.16% (1,552 cells) of these cells were considered dead or dying based on the following morphological criteria (middle panel, Figure 1A-G): loss of cell membrane integrity; leaking of cellular content; considerable reduction of cell volume; and chlorotic or swollen appearance with a shrunken cytoplasm, retracting from the cell wall. Moreover, only cells that lacked the bright red autofluorescence (non-affected cells) were defined as dead (right panel, Figure 1). The loss of the autofluorescing pigments chlorophyll a and phycobiliproteins from dying cells results in cells being weakly red or green when excited by green light (550 nm; right panel, Figure 1B-D,E) [25] or blue when excited by UV light (330 nm; right panel, Figure 1F).
Based on these criteria, cyanobacterial cell death events were apparent in all leaf cavities, whereas the proportion and cell types affected varied. The frequency increased with leaf age and peaked at 53.6% of the total cell population in leaf no. 30 ( Figure 1H) or older plant parts depending on growth conditions. In recently developed cavities at the plant apex, occupied by cyanobacteria in the small-celled motile hormogonia stage ( Figure 1A), about 10% of the cells were dying, while the majority were actively dividing, colonizing new leaf cavities and starting to differentiate heterocysts ( Figure S1). In younger leaves with few heterocysts, 77.4% of the total number of dead cells was in the vegetative cell stage (leaf no. 10), while in older leaves the proportion of dead heterocysts reached 51.9% (leaf no. 25), and that of akinetes 52.7% (leaf no. 30) of the total number of dead cells ( Figure 1I).
During sexual reproduction of the Azolla plant, motile hormogonia at the apex are attracted to megasporocarp initials ( Figure  S1 and S2) in which they subsequently colonize the inducium chamber. Of the approximately 631649 cyanobacterial cells that initially colonized the smaller female megasporocarp about 21% (135 cells) were sacrificed before its maturation, while the rest of the cyanobacterial inoculum was transferred to the next plant generation. The larger male microsporocarps were initially colonized by hormogonia, but unlike in the female megasporocarp all cyanobacterial cells disappeared during the microsporocarp maturation ( Figure S2).

Biochemical Criteria
In order to examine whether the cell death events observed using morphological criteria were related to PCD, the cyanobacterial population within the Azolla plants were subject to biochemical tests. To this end two assays were used: the Annexin V-green fluorescent protein (Annexin V-EGFP) assay, which registers externalized phosphatidylserine (PS) at the plasma membrane, and is used as an indicator of early apoptosis [23], and the TUNEL assay, which incorporates fluorescein-dUTP into the ends of fragmented DNA to detect DNA cleavage and condensation, and is used as an indicator of late apoptotic events [26]. In both assays, the cyanobacterial cells stained coincided approximately with cells identified as dying/dead cells using the morphological criteria. Annexin V-EGFP fluorescence was apparent in cyanobacterial cells from leaf cavities and developing sporocarps ( Figure 2). The Annexin V-EGFP binding was localized at the cell membranes in all cyanobacterial cell types, i.e., vegetative cells, akinetes and heterocysts (Figure 2A-C). Label was also detected in dying and disrupted akinete initials leaking their cellular content, and the number of heterocysts and akinetes exhibiting signs of apoptosis was higher in older leaves ( Figure 2D-E). Even some pro-akinetes within the developing sporocarps were characterized by cytoplasm condensation and Annexin V-EGFP labeling ( Figure 2F).
The TUNEL assay provides biochemical evidence for the existence of later stages in the PCD pathway. TUNEL fluorescence signals were found in hormogonia, vegetative cells, heterocysts and akinetes, of both leaf cavities and developing sporocarps ( Figure 3). The green positive TUNEL signal was often detected in groups of two to four vegetative cells along the filaments in leaf cavities, and in pro-akinetes within developing megasporocarps ( Figure 3A-B). Based on cell structure, cell volume and TUNEL fluorescence signal intensity, the dying cells were grouped into apoptotic-like and necrotic-like cells ( Figure 3B), the latter characterized by cytoplasmic leakage. Intermixed with non-affected cells, cells with a range of TUNEL-positive signal strengths were identified within the same filament. Based on the intensity and DNA fluorescence distribution, the TUNEL assay  Figure 3C): i) initial phase in DNA fragmentation, weak fluorescence and few scattered patches of fluorescein-dUTP-incorporated DNA (spots); ii) intermediate phase, more pronounced DNA fragmentation, larger and additional fluorescent spots; iii) late phase, advanced DNA fragmentation, a strong fluorescent TUNEL signal; and iv) final phase, fully degraded DNA, characterized by a strong smeared TUNEL signal. Morphological changes were not visible (at the LM level) at steps i) to iii), whereas plasma membrane rupture and leakage of cellular inclusions were significant for cells at step iv). The data were verified by examining cells pre-treated with DNase I which exhibited green fluorescence while green fluorescence was lacking when the TdT enzyme was omitted ( Figure 3D-E).

Cytoskeleton Criteria
Next, a link between the distribution of the cyanobacterial cellular cytoskeleton [27] and cell death modes was examined using a FITC-phalloidin, a stain which specifically binds to F-actin of the cytoskeleton of eukaryotes. A bright green fluorescing actinlike network was apparent in both vegetative cells and heterocysts ( Figure 4 and 5B), at the same time showing a bright red autofluorescence and a strong blue DAPI (DNA) stain, the two latter characteristic of healthy cells. Heterocysts with features of apoptosis, i.e. invaginated but intact cell wall and plasma membrane, still fluoresced green but lacked the well-organized meshed network of actin and exhibited weaker autofluorescence and DAPI staining ( Figure 5E-H). Heterocysts with characteristics of autophagy, such as intact plasma membrane but retracted cytoplasm, still retained actin-like structures as well as fluorescing photosynthetic pigments and DNA ( Figure 5I-L). A disrupted organization of the actin-like network was found in heterocyst which showed necrotic features, such as swollen cells, ruptured plasma membranes and leakage of cellular contents ( Figure 5M-P). Heterocysts exhibiting autolytic characteristics lacked the actinlike structure and most cellular components ( Figure 5Q-T). Together these data suggest that the cyanobacterium possess an actin-like cytoskeleton, which shows a reorganization and depolymerization during cell death which reminds of eukaryotic PCD.

Ultrastructural Criteria
To complement the Annexin V-EGFP and TUNEL assays, which do not discriminate between various PCD cell death modes: apoptotic, autophagic, necrotic and autolytic cell death [28], transmission electron microscopy (TEM) analysis was performed on cyanobacterial cells in Azolla leaf cavities and megasporocarps. A total of 940 ultrathin sections and 3,950 cyanobacterial cells, representing a random distribution of cells along the approximately 80 Azolla plants tested, were analyzed. Among these, 3,219 cyanobacterial cells appeared healthy showing intact cellular features at the ultrastructural level (see Figure S3, S4 and S6A), whereas the remaining ,18.5% (731) were dying or dead ( Figure 6 and S3i-vi). Hence, and not surprising, detailed TEM analyses lead to a somewhat higher percentage of recognized dead cells compared to lower resolution LM analyses (17.2%).
Closer inspection of the dying cells at the ultrastructural level revealed the existence of multiple cell death modes in the cyanobacterium. These included an apoptotic-like cell death mode in vegetative cells ( Figure 6A-C; Figure S4F), characterized by cell shrinkage and thylakoid membrane disintegration at an early stage, while cell walls (membrane/peptidoglycan layer) remained intact and some inclusions, such as carboxysomes and the large cyanophycin granules (N-storage), were retained. The apoptotic-like cell death mode was observed in vegetative cells, heterocysts and akinete initials in developing sporocarps ( Figure S3 and S4). The autophagy-like cell death mode was characterized by an intact cell wall system, but a gradual condensation and degradation of the cytoplasm and high levels of vacuolization ( Figure 6D-F; Figure S4). This cell death mode affected vegetative cells, pro-heterocysts and mature heterocysts ( Figure S3 and S4). In the necrotic-like cell death mode, the cell wall was ruptured at an early stage, resulting in leakage of cellular content. Based on the membrane systems, this cell death mode could in turn be divided into two subtypes: I, characteristic of vegetative cells, the intracellular components were dispersed into the extracellular space, while part of the membrane system remained intact until the late stage ( Figure 6G-I; Figure S4); II, characteristic of the thick-walled akinetes, a delay in the leakage of the degraded cytoplasm was apparent ( Figure 6J-L). The necrotic-like cell death mode was observed in leaf cavities and sporocarps, and in all cell types ( Figure 6G-I,J-L; Figure S3 and S4). Finally, in the autolytic-like cell death mode, the peptidoglycan layer of the vegetative cells lysed at the early stage and the cytoplasm disintegrated gradually and vesicles formed at late stages ( Figure 6M-O; Figure S3 and S4). Autolytic-like cell death was the case in vegetative cells ( Figure 6M-O) and heterocysts ( Figure  S3). These cell death modes occurred independently of plant age and plant organ.
Interesting shifts in the cyanobacterial death modes in relation to the developmental stage/age of the plant was also apparent ( Figure 7). Clearly, apoptotic-and autophagic-like modes operate and dominate in young plant parts (apex to leaf no. 15), while necrotic-and autolytic-like events appear at later developmental stages and dominate in cyanobacteria of older plant leaves.

Stress Enhanced Cell Death
Next, Azolla plants were subjected to selected stresses (darkness, nutrient deprivation and c-radiation). We hypothesized that if PCD is in action in the cyanobacterium, stresses would enhance the process, as shown for other organisms [2]. In total, 1,004 cells from the plant apex (leaf no. 1) and leaf no. 5 were examined by TEM and LM. All stress treatments enhanced distinct morphological alterations and in particular in cyanobacterial cells residing at the plant apex. Dead/dying cells were found at a 3-10 fold higher frequency under stress than under non-stress conditions ( Figure 8A). Dark treatment (5 days) lead to an autophagic-like death mode ( Figure S5), first cytoplasmic condensation and plasma membrane preservation, then cytoplasmic disorganization followed by vacuolization. This suggests that the cyanobacterium may be subject to PCD. It is also apparent that cyanobacteria at the apex depend on their own photosynthesis, while they live heterotrophically in more mature leaf cavities [16]. Annexin V-EGFP ( Figure 8B) and phalloidin-FITC staining (data not shown)   of cyanobacterial cells in Azolla plants subject to nutrient deprivation (N-depletion) verified that a substantial proportion (32.1%) of the cells examined were negatively affected. Using cradiation, at 40 and 80 Gy, enhanced the cyanobacterial cell death frequency substantially, to 83.7% and 65.3% for cyanobacteria at the apex and in leaf no. 5, respectively ( Figure 8A). At the ultrastructural level a clear cytoplasmic condensation and a reduction in cellular volume was apparent indicative of necroticlike and autolytic-like cell death modes ( Figure S6).
Finally, we examined the ultrastructure of dead cells and cell debris in leaves and sporocarps ( Figure S7). When the peptidoglycan layers of the gram negative cell wall were digested, numerous vesicles appeared at the original location of the peptidoglycan layer ( Figure S7E,F). Some of the dying/dead cells were still at this late stage connected to healthy-looking cyanobacterial cells (in both leaf cavities and megasporocarps) and the majority was in close proximity to bacteria. To what extent the cyanobacterium undergoes auto-digestion and/or digestion by adjacent bacteria is however an open question ( Figure S7A-C).

Molecular Criteria
To obtain molecular information related to PCD in the endosymbiotic cyanobacterium of Azolla, the primer pair cas202 and cas203 were designed based on encoding sequences for conserved catalytic subunit p20 of the apoptotic enzyme caspase (in GenBank accession no. CP000117). Two fragments (754 bp and 508 bp) were amplified from genomic DNA isolated from the cyanobacterium of Azolla (Figure 9A), and the sequences were deposited in GenBank with accession numbers FJ541185 and FJ541186, respectively. A BLASTN search showed that FJ541185 and FJ541186 are 99% identical (99% query coverage) to the Cterminal (position 968-1707) and N-terminal (position 221-727) of a ''caspase catalytic subunit p20'' gene (locus tag Aazo_1578) in the sequenced genome of 'Nostoc azollae' 0708, the cyanobiont of A. filiculoides [18]. This gene in 'Nostoc azollae' 0708 was identified as a metacaspase [11,20]. The translated amino acid sequence of FJ541186 (GenBank accession ACM79336) was used as query in a BLASTP search, which identified several homologs in bacteria (including cyanobacteria, proteobacteria and nitrospirae) as well as in plants and fungi (see Table S1 online). The majority of hits are annotated as ''caspase'', ''metacaspase'' or ''caspase catalytic subunit p20'', and all significant hits in completed genomes were previously identified [11]. This indicates that ACM79336 belongs to the superfamily of caspase-like proteases that was previously described as the ''metacaspases'' in plants, fungi and some bacteria [29]. Figure 9B shows a multiple sequence alignment of ACM79336 with representative homologs in cyanobacteria, proteobacteria, plants and fungi (see also Table 1). The caspase domain (Pfam id pf00656) spans the entire length of ACM79336.1 and is also present in the homologous sequences ( Figure 9B). The alignment shows that the catalytic His-Cys dyad common to caspases is replaced with Tyr-Ser in the cyanobacterial sequences. This substitution has been reported previously for cyanobacteria [30] and is the most common type of amino acid substitutions found in bacterial metacaspases [11]. In the protozoan Trypanosoma brucei, a Cys-Ser substitution in the active site is known to inactivate some cysteine proteases [31]. However, the alternative enzymatic roles performed by metacaspases holding this (or other) substitution(s) remains to be established in prokaryotes and in protozoa.
Using primer Hcw3/4, which was based on the encoding sequence of cell wall hydrolase of Anabaena PCC 7120 (GenBank accession no. AF216288), a fragment (506 bp) was amplified and the sequence was deposited in GenBank under accession no. FJ541183. Our BLASTN analysis also showed that FJ541183 is 99% identical (96% query coverage) to an N-acetylmuramoyl-Lalanine amidase/autolysin gene (locus tag Aazo_0861) in the genome of 'Nostoc azollae' 0708. A BLASTP search of the translated amino acid sequence of FJ541183 (GenBank accession ACM79334) identified 42 homologs (Table S2 online). All identified homologs are from cyanobacteria, both unicellular and filamentous/multicellular species.

Discussion
Our data collectively suggest that several programmed cell death events may take place in the life cycle of a prokaryotic cyanobacterium that exhibits advanced developmental and adaptive capacities, as exemplified here in a cyanobacterium adjusting to the highly stringent and stressful conditions offered inside a eukaryotic host. This conclusion is based on the findings that i) the cell death events identified (cell lysis and degradation) operate in a photosynthetic prokaryote intimately associated with a eukaryotic host; ii) four distinct cell death modes are identified, apoptotic-like, autophagic-like, necrotic-like (type I and II, depending on cell type) and autolytic-like, which are similar in several respects to PCD events in eukaryotes [2] (Table 1); iii) the cyanobacterial cell death events may be genetically underpinned as suggested by the presence of autolysin and metacaspase, the latter supported by the rich metacaspase gene pool (with conserved p20 and Cys-His dyads similar to those of eukaryotic caspases) recently found in prokaryotes, and particularly many in developmentally complex cyanobacteria [11,30]; iv) cell death events are not restricted to photosynthetic vegetative cells, but also take place   in heterocysts and resting akinetes, and are potentially therefore involved in differentiation of cells; and that v) these cell death events occur in cyanobacteria irrespective of their location in various plant organs (leaves and sporocarps) and host plant developmental stage (vegetative growth and sexual reproduction), although the relative proportion of the various death modes varied with plant age. Finally, as external stress factors (darkness, nutrient deprivation and ionizing radiation) enhance cell death further supports the existence and potentially, a functional importance, of PCD events in this cyanobacterium. Hence, our data confirm and considerably extend earlier studies on experimentally verified PCD events of some free-living cyanobacteria [4,21,22] and show, for the first time, that autophagic-, autolytic-and necrotic-like PCD events may take place in a cyanobacterium.
The finding that all cyanobacterial cell types (i.e., vegetative cells, hormogonia, heterocysts and akinetes) responded positively to Annexin V-EGFP staining suggest that all can pass early stages of apoptotic-like and autophagic-like PCD, and ahead of any visible phenotypic changes [32,33]. The TUNEL assay further verified the potential existence of later-stage PCD events and progressive DNA fragmentation, yet another hallmark of PCD. This assay suggests that cell death events in cyanobacteria do not occur in a chaotic manner, but represent a regulated process. The distinct phenotypic features identified (cell shrinkage, membrane integrity, cytoplasm preservation, PS exposure and DNA fragmentation) match criteria recently defined for PCD in eukaryotes [24].
Studies have demonstrated that alteration in the actin cytoskeleton is not just a consequence of PCD, but may play a functional role in the initiation and regulation of PCD [34,35]. We here provide the first account of a link between actin dynamics and four cell death modes in a cyanobacterium, based on different F-actin distribution patterns and polymer levels (e.g. actin stabilization, depolymerization and degradation). The data coincide with findings of cytoskeletons in cyanobacteria [27,36] and that actin depolymerizes at early stages of apopotosis while it is preserved largely during middle-later stage autophagy. An intact cytoskeleton network stabilizes cellular conformations, while polymerization of actin destroys the original morphology, resulting in the irregular appearance of the apoptotic-like cyanobacterial cells. Actin alternations may therefore be used as an indicator of potential cyanobacterial PCD modes.
Metacaspases (caspase-like proteases) are known to play a crucial role in PCD in plants, fungi and unicellular eukaryotes [29], and a metacaspase was recently found to be active in cell death events of the cyanobacterium Trichodesmium sp. [22]. Our identification in the cyanobacterium of genes that encode the conserved catalytic subunit (p20) of caspases, combined with the discovery of a gene encoding N-acetylmuramoyl-L-alanine amidase, which catalyzes cell-autolysis, provides support for the existence of PCD pathways in this cyanobacterium. Genomic surveys of 33 cyanobacterial [30] and more recently 1,463 prokaryotic [11] genomes, revealed the existence of 58 and 671 putative metacaspase genes, respectively, including a metacaspase in the recently sequenced Azolla endosymbiont [18]. Genes involved in autophagy or necrosis were not detected in the Azolla endosymbiont genome using primers based on conserved domains of some important eukaryotic proteins, such as Atg4, Atg5, microtubule-associated protein1 light chain3 (LC3) [23], and receptor-interacting serine/threonine-protein kinase1 (RIPK1), proposed to be involved in necrosis [2]. Either the similarity between the eukaryotic and prokaryotic genes is too low, or evolutionary homologs of these genes are absent in prokaryotes. However, autophagy and apoptosis, and even necrosis in some cases, may be regulated by the same factors, and 13 proteins have been identified that have dual roles in autophagy and apoptosis [2,37,38].
Although cyanobacteria as opposed to eukaryotic cells lack membrane-bound organelles, they do contain elaborate membrane systems and subcellular inclusions; for instance photosynthetic thylakoids, cyanophycin and carboxysomes. Still, a comparative evaluation of PCD events in animals, plants and cyanobacteria, based on nomenclature recommendations by Kroemer and co-workers [24], suggests a number of apoptoticlike parameters to be shared between cyanobacteria and eukaryotes as summarized in Table 1. This supports the existence of a cyanobacterial apoptosis process. One difference is that cyanobacteria, unlike eukaryotes, do not produce apoptotic bodies whilst undergoing cell death. Moreover, autophagy-like parameters are shared between animal [38] and cyanobacterial cells, and cyanobacterial autophagy is similar to vacuolar cell death events in plants [39]. We propose that autophagic-like cyanobacterial cell death resembles that induced by macro-autophagosomes, which degrade the cargo of dead animal cells, and vacuolar cell death in    plant cells [36][37][38][39]. Self-consumption, or bacterial consumption, of cyanobacterial remains may also exist in the Azolla symbiosis.
There seem to be few differences between necrosis-like events in eukaryotes and the cyanobacterium. The status of the peptidoglycan layer, a key component of the cell wall of all prokaryotes, may serve as a critical indicator when assessing the living/dying status of a bacterium. Our data show that retention of the peptidoglycan layer until a late stage of cell death is indicative of the apoptotic-like death mode, disappearance at an early stage is indicative of the autolytic-like mode, while maintenance at an early stage and disappearance at a late stage is indicative of the autophagic-like mode. The necrotic-like mode is characterized by an early rupture of the peptidoglycan layer. Examining 150 cyanobacterial cells showed that changes in the peptidoglycan layer positively correlated with the cell death mode in action. The peptidoglycan layer, which is composed of murein, is likely to be degraded by the murein hydrolase identified (Nacetylmuramyl-L-alanine-amidase) [7], which may be contained in the vesicles [17] formed at the site of the cyanobacterial peptidoglycan layer during cell death.
Cyanobacteria living within the Azolla plant are subject to adaptive forces manifested at various organization levels [15][16][17][18][19][20]40,41]. It is therefore likely that cell death restricts the cyanobacterial population size to maintain the symbiotic homeostasis [15,16]. The data also suggest that PCD-like cell death events are operative during the unique vertical transfer of cyanobacteria to the next plant generation via the megasporocarp, again by trimming the population, and as earlier shown, the genome size [18]. PCD may also serve as a defense mechanism for the cyanobacterium against the host plant defense system. The cell death pathways identified here in a contemporary cyanobacterium may have existed early during the evolution of the cyanobacterial phylum. This is interesting, considering the fact that plant chloroplasts arose from a free-living cyanobacterium, and possibly from one with a complex morphology as the Azolla symbiont [42]; and that mitochondria, now functioning as contemporary regulators of metazoan PCD, also arose from prokaryotes [43].
Taken together, our data provide structural and biochemical evidence for the existence of four PCD or PCD-like modes in a prokaryotic cyanobacterium (plant endosymbiont) potentially genetically based in metacaspases, and show features similar to cell death pathways in eukaryotes [24] (Table 1). While the PCDlike events identified here may not represent all prokaryotes, they clearly expand the type of death modes found in non-eukaryotes. Based on our observations, we speculate that that this potentially metacaspase-based prokaryotic form of PCD-like cell death may represent an ancient form involved in the regulation of cell proliferation and cell differentiation, although the process today seems to be limited to 57% [30] of the cyanobacteria and 18% of the prokaryotes [11]. Given that chloroplasts evolved from cyanobacteria, it is also tempting to speculate that chloroplasts (together with mitochondria) may play a role as sensors and regulators in PCD processes in plant cells. Further experimental evidence is now warranted to definitively prove the operation of active cell death programs in cyanobacteria.

Abiotic Stress Induction
The Azolla symbiosis was stressed in the following ways: Azolla plants were transferred to pots (15-cm diameter) containing a medium described by Watanabe et al. [44] and subjected to darkness for 5 days prior to analysis of the cyanobacteria by bright field light microscopy (LM) and transmission electron microscopy (TEM) and to nutrition deprivation by cutting off the root systems and floating the plants on distilled water for 15 days. Cyanobacteria in cavities of leaf no. 13 were then examined using LM combined with Annexin V-EGFP label, which detects cellular PS exposure (see below). Plants were also subjected to gamma (c) radiation by placing fronds in petri dishes transparent to 40 and 80 Gy c-radiation at a dose rate of 1.5 Gy/min generated by a 60 Co source (Radiation Technique Center, Rice Research Institute, Fujian Academy of Agricultural Sciences, Fuzhou, China). After c-ray exposure, plants were sampled and processed for LM and TEM (see details below).

Light Microscopy (LM)
The cyanobacterial filaments from leaf cavities of different developmental stages were picked with a fine needle and gently spread on microscope slides coated with poly-lysine (Sigma, USA). To isolate the cyanobacterial filaments within Azolla sporocarps, a cover slip was placed over the sporocarp, and lightly tapped with a fine needle. The released filaments were smeared uniformly on a slide. The morphology of the cyanobacterial cells was examined using bright field and fluorescence microscopy (Olympus BX60) with filter sets for UV light and blue and green light. Exposure times were 1/15-20 s for bright field, 1/100-110 s for green light (band pass (BP) 545-580 nm), 1.

Transmission Electron Microscopy (TEM)
Azolla fronds or individual leaves and sporocarps of different developmental stages were processed for TEM according to Zheng et al. [17]. The ultrathin sections obtained were examined using transmission electron microscopy (Zeiss EM-906 or JEOL 100CX) set at an accelerating potential of 80 kV.

Annexin V-EGFP Staining
Detection of PS was performed using the Annexin V-EGFP Apoptosis Detection Kit (BioVision, CA, USA). Based on specific features of the cyanobacterial cell types examined, the protocol recommended by the manufacturer was modified as follows: isolated cyanobacterial filaments with heterocysts and proakinetes (both with thick cell wall envelopes) were suspended in 20 ml of 16 binding buffer for 4 min with gentle mixing to prevent filament clumping. Five microliters of Annexin V-EGFP was added and the cyanobacteria were incubated in darkness in a humid chamber at room temperature for 30 min, followed by three washes in distilled water and transfer to microscopy slides, along with a drop of ProLong Antifade Reagent (Molecular Probes, Oregon, USA). Samples were examined by fluorescence microscopy (as above) using the filter set for blue light (BP 460-490 nm).

The TUNEL Assay
To detect DNA fragmentation in cyanobacteria, the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed using the In Situ Cell Death Detection Kit (Roche Diangnostics, CA, USA), according to the manufacturer's instructions, but with the following modifications: cyanobacterial cells from leaf cavities and developing sporocarps were fixed in freshly prepared 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 10 min, followed by two washes in PBS. The fixed cells were made permeable by incubation in lysozyme (20 mg ml 21 with 100 mM EDTA in PBS) at 37uC for 60 min and were subsequently rinsed twice in PBS. Before labeling, the cells were incubated in Proteinase K (20 mg ml 21 in 10 mM Tris/ HCl [pH 8.0], 50 mM EDTA) at 37uC for 20 min, followed by 0.1% Triton X-100 at 4uC for 4 min, to facilitate the entry of TdT into the cells. After two rinses in PBS, the samples were labeled with the TUNEL reaction mixture according to the manufacturer's instructions. Samples were incubated in a humidified (100%) chamber in darkness. As a positive control, permeabilized cyanobacterial cells were incubated with DNase I (1 U ml 21 ) for 10 min at room temperature. A negative control was obtained by omitting TdT from the reaction mixture. DNA fragmentation was examined directly after the TUNEL reaction using fluorescence microscopy (as above) with filter sets for blue light (BP 460-490 nm).

FITC-phalloidin Assay
To visualize the cytoskeleton of the cyanobacterium, cells were isolated from Azolla fronds (greenhouse grown) and placed on glass slide, fixed with 3% formaldehyde and washed with PBS. This was followed by decolorization of the cells in a series of acetone (50%-100% in PBS, each for 2 min) and washing 2 times in PBS. The cells were then incubated in 5 ml FITC-phalloidin (Alexa FluorH 488, Invitrogen, USA) diluted 1:10 in 40 ml buffer (0.9 M NaCl, 20 mM Tris/HCl [pH 7.5] and 0.1% (w/v) SDS), 2,5 ml 1% Triton X-100 and 2.5 ml 3% formamide, and kept at room temperature for at least 30 min. After washing with ddH 2 O two times, the cells were then stained with 50 ml DAPI (2 mg ml 21 DAPI in PBS) at room temperature for 2 h and then washed two times with ddH 2 O. The edges of the cover slip were sealed with nail-polish prior to fluorescence LM-examinations as above.

DNA Extraction
Azolla plants were harvested, the roots removed and the plants rinsed three times with tap water. Cyanobacterial filaments were isolated as described by Ran et al. [18], with the following modifications: the cell suspension was centrifuged first in 40% Percoll (GE Healthcare, Germany) for 5 min at 3606g. The cell pellet was then re-suspended in 80% Percoll for 15 min at 12006g, rendering a ,95% enriched cyanobacterial cell suspension, as evidenced by light microscopy. Next, DNA was extracted from the isolated cyanobacterial cell material using a genomic DNA Extraction Kit (BioFlux, Beijing, China), with the following modifications of the manufacturer's protocol: the sample was diluted in buffer A before all other reagents were added. Proteinase K (20 mg ml 21 , Promega, WI, USA) was included in the solution at step 2, followed by incubation at 37uC for 1 h and at 56uC for 30 min. Extracted DNA was purified with PEG 8000 and stored at 220uC until used.

Sequence Similarity and Alignments
To identify cyanobacterial genes and proteins that are implicated in PCD, PCR products and their translated sequences were used as queries in BLASTN and BLASTP searches, respectively, against the non-redundant (nr) database (Dec 6, 2011) with E-value cut off = 0.01. The multiple sequence alignment for ACM79336 was guided by the seed alignment of the PF00656 domain using Muscle (v3.8.31) [45], with default settings, and gaps were manually adjusted in Jalview (v. 2.7) [46].

Accession Numbers
Sequence data from this article can be found in the GenBank/ EMBL data libraries under accession numbers FJ541183 and FJ541186 and deduced amino acid accession numbers ACM79334 and ACM79336, which correspond to peptidoglycan hydrolase and the p20 subunit of caspase, respectively.