Trichodesmium – a widespread marine cyanobacterium with unusual nitrogen fixation properties

The last several decades have witnessed dramatic advances in unfolding the diversity and commonality of oceanic diazotrophs and their N2-fixing potential. More recently, substantial progress in diazotrophic cell biology has provided a wealth of information on processes and mechanisms involved. The substantial contribution by the diazotrophic cyanobacterial genus Trichodesmium to the nitrogen influx of the global marine ecosystem is by now undisputable and of paramount ecological importance, while the underlying cellular and molecular regulatory physiology has only recently started to unfold. Here, we explore and summarize current knowledge, related to the optimization of its diazotrophic capacity, from genomics to ecophysiological processes, via, for example, cellular differentiation (diazocytes) and temporal regulations, and suggest cellular research avenues that now ought to be explored.


Introduction
Balancing of nitrogen (N) inputs and exports in global oceans requires substantial biogenic fixation of atmospheric nitrogen (N 2 ). In this context, planktonic colonyforming cyanobacteria of the genus Trichodesmium are recognized as major players. Representatives within the genus have consistently been shown to be stable components of tropical and subtropical segments of the Atlantic, Pacific, and Indian Oceans where they may form enormous surface accumulations ('blooms') visible to the naked eye (see Capone & Carpenter, 1982;Capone et al., 1998;Karl et al., 2002;Tyrrell et al., 2003;Davis & McGillicuddy, 2006;Westberry & Siegel, 2006;Carpenter & Capone, 2008). Trichodesmium contributes to sustaining marine life via active release of key nutrients, for example carbon and nitrogen, and upon death and decay, hence making this fully photoautotrophic genus a vital player in the biogeochemical cycling of basic elements in contemporary oceans (Carpenter & Capone, 2008). The global input via N 2 fixation by Trichodesmium was initially estimated to amount to about 5 Tg N annually by Capone & Carpenter (1982), an estimate that by now has risen to about 60-80 Tg N annually (Capone et al., 1997;Mahaffey et al. 2005;Westberry & Siegel, 2006;Carpenter & Capone, 2008), which makes up a substantial part of the current estimate of global marine N 2 fixation, 100-200 Tg N annually (Karl et al., 2002).
As the N 2 -fixing enzyme, nitrogenase, encoded by the nifHDK genes, is rapidly inactivated by O 2 , diazotrophic cyanobacteria either fix N 2 at night (to avoid photosynthetically evolved oxygen) or differentiate a thick-walled, photosystem II-deficient heterocystous cell type to specifically sustain daytime N 2 fixation (Kumar et al., 2010). Members of the genus Trichodesmium fix N 2 exclusively in the light (Dugdale et al., 1961;Capone et al., 1997), although the genus is affiliated to Section III filamentous cyanobacteria that are unable to form heterocysts and therefore expected to fix N 2 (primarily) during the dark phase (see Bergman et al., 1997). Knowledge has expanded dramatically in regard to the diazotrophic physiology and molecular biology of Trichodesmium, but there are still gaps related to its unique cell biology and overall behavior in an ecophysiological context. We here summarize our current knowledge by highlighting its diazotrophic peculiarities from various perspectives.

Speciation
Trichodesmium erythraeum was named by Ehrenberg in 1830 after observing blooms that discolored the water at the Bay of Tor in the Red Sea (Ehrenberg, 1830). Jules Verne (1869) in'20 000 leagues under the sea' also mentions blooms in this Bay (Box 1). Two other species, T. thiebautii and T. hildebrandtii, were named by Gomont (1892), and Wille (1904) later described another three species, T. contortum, T. tenue, and T. radians. These species were re-examined in 1995 (Janson et al., 1995) using specimens from the Indian Ocean, Caribbean, and Sargasso Seas. Ultrastructural arrangement of gas vesicles and glycogen clusters (carbon storage) were used as primary markers and separated the species into two clades: (i) T. tenue and T. erythraeum and (ii) T. thiebautii, T. hildebrandtii, and T. contortum. In 2001, a close relationship between Trichodesmium spp. (Lundgren et al., 2001) and marine cyanobacterial members of the genus Katagnymene (K. pelagica and K. spiralis; Lemmermann, 1900) was discovered using phylogenetic analysis of nifH gene sequences. Using the more variable fragment of the hetR gene, and a few other genetic markers, as targets revealed that the two Katagnymene species in fact cluster within one of the two Trichodesmium clades (Orcutt et al., 2002;Lundgren et al., 2005). Despite the morphological differences, K. pelagica and K. spiralis were, in addition, found to be the same species (Lundgren et al., 2005). Examining 21 cultivated isolates of Trichodesmium/Katagnymene, using genetic and morphological markers, Hynes et al. (2012) verified the existence of two Trichodesmium clades and suggested that these may inhabit different ecological niches, based on different pigment characteristics.
However, a full revision of the genera Trichodesmium and Katagnymene is now warranted, as the latter also includes several freshwater species (T. iwanoffianum Nygaard, T. lacustre Klebahn, K. accurata Geitler, K. mucigera Compére and K. spirulinoides An; see Komárek & Anagnostidis, 2005), one of which, in addition, represents the 'type strain' of the genus Katagnymene. Sequencing of additional genomes within the Trichodesmuim genus is also needed if we are to fully comprehend the taxonomy and phylogeny of the genus.

Box 1
From Jules Verne, '20,000 leagues under the sea' [translated by Lewis Mercier, (Verne, 1872)]: 'Here it is, M. Aronnax. According to my idea, we must see in this appellation of the Red Sea a translation of the Hebrew word 'Edom'; and if the ancients gave it that name, it was on account of the particular colour of its waters.' 'But up to this time I have seen nothing but transparent waves and without any particular colour.' 'Very likely; but as we advance to the bottom of the gulf, you will see this singular appearance. I remember seeing the Bay of Tor entirely red, like a sea of blood.' 'And you attribute this colour to the presence of a microscopic seaweed?' 'Yes.' 'So, Captain Nemo, it is not the first time you have overrun the Red Sea on board the Nautilus?' 'No, sir.'

An expanding and flexible genome
Trichodesmium erythraeum IMS101 (from now on Trichodesmium IMS101) was one of the first strains isolated into axenic cultures (Prufert-Bebout et al., 1993) and still represents the only sequenced genome within the genus (http://genome.jgi-psf.org/trier/trier.home.html). The genome, which comprises 7.75 Mbp, is among the larger cyanobacterial genomes sequenced to date (Fig. 1a). A recent phylogenetic survey of 58 sequenced cyanobacterial genomes, based on a concatenated alignment of 285 protein orthologs, verified that Trichodesmium IMS101 is affiliated to a lineage composed of other filamentous nonheterocystous species within Oscillatoriales ( Fig. 1b; Larsson et al., 2011): the marine Lyngbya sp. PCC 8106 and two species within Arthrospira (A. platensis and A. maxima; previously denoted Spirulina). These are all ecologically successful and widespread inhabitants of marine waters and alkaline lakes, respectively. Recent analyses based on the 16S rRNA gene sequence give a similar clustering of Trichodesmium IMS101 (Schirrmeister et al., 2011). Larsson et al. (2011) also showed that the capacity to fix N 2 within this Trichodesmium clade was lost in A. platensis and A. maxima, as also verified by Latysheva et al. (2012), while retained in the deeper-branching Trichodesmium and Lyngbya sp. PCC 8106. Interestingly, these four lineages all possess hetR (Larsson et al., 2011), a gene encoding a protease with a key function in cell differentiation (N 2 -fixing heterocysts and resting akinetes; see Kumar et al., 2010), although they all lack these Maximum-likelihood phylogenetic tree based on a concatenated alignment of 285 single-copy orthologs. The tree is a subtree of a larger phylogeny of 58 cyanobacteria (see Larsson et al., 2011). Specific phenotypes for cyanobacteria are shown by the colored boxes next to the tip labels. The clade containing Trichodesmium (order Oscillatoriales) is highlighted with blue branches. Thick and thin branches indicate bootstrap support values (200 replicates) of 100 and between 58 and 84, respectively. Bar, 0.4 expected substitutions per site. (c) Ancestral genome sizes (reconstructed by parsimony) in the phylogeny from b. Organism names are abbreviated (see below for full names). Contemporary genome sizes (Mbp) are shown in the right margin and at specific nodes in the tree. Organism abbreviations are as follows: Acam = Acaryochloris marina MBIC11017, Anav = Anabaena variabilis ATCC29413, Artm = Arthrospira maxima CS328, Artp = Arthrospira platensis str. Paraca, Crow = Crocosphaera watsonii WH8501, Cya0110 = Cyanothece sp. CCY0110, Cya51142 = Cyanothece sp. ATCC51142, Cya7424 = Cyanothece sp. PCC7424, Cya7425 = Cyanothece sp. PCC7425, Cya7822 = Cyanothece sp. PCC7822, Cya8801 = Cyanothece sp. PCC8801, Cya8802 = Cyanothece sp. PCC 8802, Cylr = Cylindrospermopsis raciborskii CS505, Lyns = Lyngbya sp. PCC 8106, Mica = Microcystis aeruginosa NIES 843, Micc = Microcoleus chthonoplastes PCC7420, NoAz = 'Nostoc azollae' 0708, Nods = Nodularia spumigena CCY9414, Nosp = Nostoc punctiforme PCC73102, Noss = Nostoc sp. PCC7120, Rapb = Raphidiopsis brookii D9, Scys6803 = Synechocystis sp. PCC6803, Syn7002 = Synechococcus sp. PCC7002, Thee = Thermosynechococcus elongatus, Trie = Trichodesmium erythraeum IMS101, Ucyn = cyanobacterium UCYN-A. The figures are adapted from Larsson et al. (2011) with the author's permission. developmental capacities. Additionally, comparative genomic analyses show that several other gene orthologs involved in heterocyst differentiation are present in the Trichodesmium IMS101 genome (e.g. hetCF, patB) while others, not unexpectedly, are missing such as those involved in the deposition of the heterocyst outer envelope (e.g. hglCDE, hepB) (Table S1; El-Shehawy et al., 2003;Larsson et al., 2011). However, orthologous genes are not always functionally equivalent. For instance, the sepJ gene of Trichodesmium is missing a vital domain essential for filament integrity under nitrogen deprivation, although it fully complements a sepJ deletion mutant of Nostoc sp. PCC7120 when grown in the presence of combined nitrogen (Mariscal et al., 2011). The nif gene operon of Trichodesmium IMS101 is conserved in a manner typical of some heterocystous cyanobacteria, although Trichodesmium lacks the large DNA insertion element present in the structural nifD gene of several of the heterocystous cyanobacteria, as well as the intergenic region between nifB and nifVZT/cysE (Fig. S1). These findings strengthen an evolutionary relationship between the genus Trichodesmium and the heterocystous clade, whereas distinct differences are also apparent, relationships now worth examining in greater detail.
Another notable feature of the Trichodesmium IMS101 genome is its comparatively low coding capacity (Larsson et al., 2011). With c. 40% noncoding DNA, it holds one of the lowest coding percentages among all to date sequenced genomes of cyanobacteria ( Fig. 1a) and other bacteria (Hou & Lin, 2009). The intergenic sequences within the Trichodesmium IMS101 genome (459-bp median length) are also relatively large for cyanobacteria (14.5-to 231-bp median intergenic length in 39 other finished cyanobacterial genomes). The reason for these large and presumably noncoding intergenic spacers in the Trichodesmium IMS101 genome is unknown. It is, however, interesting to note that among the 58 genomes examined, the genome of Trichodesmium IMS101 is one of a few in which the genome is currently expanding in size, as is also the case for the genomes of the limnic Microcystis aeruginosa NIES 843 and the marine Acaryochloris marina MBIC11017 ( Fig. 1c; Larsson et al., 2011). This suggests that the genome of Trichodesmium IMS101 is in an expanding dynamic state, in contrast to the shrinking genomes of the unicellular marine genera Synechococcus and Prochlorococcus (with genomes < 2 Mbp; Palenik et al., 2006;Kettler et al., 2007), genera which to a large extent share the same tropical/subtropical marine habitat as Trichodesmium. Based on these data, it is suggested that different strategies are used to cope with the various constraints enforced by these oligotrophic oceans (Larsson et al., 2011). Trichodesmium may use a strategy to flexibly adapt by incorporating functions and capacities when needed (e.g. via horizontal gene transfer, HGT) and maintain gene duplications (in-paralogs) affecting c. 10% of all genes in the Trichodesmium IMS101 genome, as a mechanism to promote genome expansion and organismal adaptations (Swingley et al., 2008;Treangen & Rocha, 2011). One example of HGT in Trichodesmium IMS101 is the acquisition of long eukaryotic triglyceride collagen protein fibers that may sustain Trichodesmium colony formation (Layton et al., 2008). Paralogs and horizontally gained genes that do not provide a fitness advantage will undergo inactivation (sequence divergence and loss of function) and eventually be lost. Indeed, 21% of in-paralogs in Trichodesmium IMS101 appear to have been subject to inactivation and are now present only as pseudogenes within the genome (Larsson et al., 2011), thereby contributing to the abundance of noncoding nucleotides (Fig. 1a). Considering that bacterial genomes are subject to a deletion bias (Mira et al., 2001), the large noncoding proportion of the Trichodesmium IMS101 genome which cannot be attributed to remnants of previously functional genes (pseudogenes) is enigmatic. However, it is possible that at least parts of these intergenic regions contain as yet nonannotated genes or small RNAs (Hewson et al., 2009;Shi et al., 2009). It appears that the unicellular marine cyanobacteria with reduced genomes (e.g. Prochlorococcus spp. in Clade II) use an opposite strategy to compete for life-space and survival, that is maintain a large surface-tovolume ratio and fewer genes, a strategy recently characterized as 'cryptic escape' (Yooseph et al., 2010).

The diazocytes -separation in space
A reduced oxygen environment is a prerequisite for effective N 2 fixation activity in Trichodesmium as in all other bacteria. Both transcription of nif genes and biosynthesis of the nitrogenase enzyme complex have been found to be sensitive to oxygen inactivation (Zehr et al., 1997;Staal et al., 2007). Ever since the pioneering studies by Dugdale and co-workers (Dugdale et al., 1961), a compelling research area has therefore been to elucidate how Trichodesmium reconciles oxygenic photosynthesis and oxygenophobic N 2 fixation within its 'heterocyst-free' physiology. Initially, N 2 fixation was proposed to be limited to low-oxygen or anaerobic regions in the center of Trichodesmium colonies (Paerl & Bebaut, 1988). However, later, it became apparent that colony formation was not a prerequisite as 'free' trichomes, the dominant form in Trichodesmium laboratory cultures, are also able to fix N 2 . Attention then switched to the structural differences along the Trichodesmium trichomes that early on were observed using light microscopy, recognized as 'nongranulated' or 'lighter' cell regions in parts of the trichomes (Carpenter & Price, 1976;Bryceson & Fay, 1981;Li & Lee, 1990), in both The diazotrophic physiology of Trichodesmium single trichomes, and colony-associated trichomes. This cellular arrangement occurs in both natural populations and cultures of Trichodesmium IMS101 (Lin et al., 1998;El-Shehawy et al., 2003). On average,~15% of the total cell population of the trichomes may be described as less granulated or lighter (more transparent). These cells are arranged in strings or 'zones' composed of~2-30 cells, but are not always obvious in LM (Fig. 2a). However, when stained (e.g. with Lugol's solution; Fig. 2c), each trichome harbors typically 1-2 such zones per trichome, but up to four zones have been observed in longer trichomes (El-Shehawy et al., 2003). The nongranulated appearance is caused by a diminished number and/or size of subcellular structures such as cyanophycin granules, gas vacuoles, and polyphosphate granules ( Fig. 2c and e), while additional membranes are synthesized (Fredriksson & Bergman, 1997). The 'lighter' cells were in 1991 (Bergman & Carpenter, 1991) proven to be the nitrogenase enzyme containing cells in Trichodesmium (Fig. 2d). The existence of a 'spatial' nitrogenase sequestration mechanism in a nonheterocystous cyanobacterium was thereby proven. The cells were subsequently termed diazocytes: di (two) azo (nitrogen) cyte (cell) (Fredriksson & Bergman, 1997). The exclusive localization of nitrogenase in diazocytes was corroborated by immuno-TEM (sectioned trichomes) and immuno-LM (whole-mount intact trichomes; Fig. 2d) analyses of both cultured and natural populations from the Indian, Pacific, and Atlantic Oceans, using a battery of antibodies (Bergman & Carpenter, 1991;Bergman et al., 1993;Janson et al., 1994;Fredriksson & Bergman, 1997;Berman-Frank et al., 2001b), including also one monoclonal anti-Trichodesmium IMS101-NifH antibody (targeting the smaller Fe protein subunit; Zehr et al., 1990;Bergman et al., 1993). The frequency of the diazocytes is lower at dawn and increased toward noon and is negatively regulated by the presence of combined nitrogen (Fredriksson & Bergman, 1995;Lin et al., 1998;Sandh et al., 2009Sandh et al., , 2011. Indeed, theoretical models suggest that a spatial separation of processes (such as nitrogen fixation and photosynthesis) favors biomass production compared to temporal separation (Rosetti & Bagheri, 2012) as, for instance, in unicellular cyanobacteria . As a few other studies have suggested that the nitrogenase enzyme is present in all cells within the trichomes (Paerl et al., 1989;Ohki, 2008;Orcutt et al., 2009), a variation in cellular localization may exist depending on different environmental conditions or species examined. Using 15 N and Nano-SIMS, Finzi-Hart et al. (2009) observed that the fixed N is rapidly distributed into the majority of cells along Trichodesmium trichomes, although cells in the center showed a lower 15 N label, a pattern that may suggest a zone of diazocytes. A lower 15 N label is also typical for heterocysts analyzed by Nano-SIMS due to a most rapid transfer of fixed nitrogen out of these cells (Popa et al., 2007;Ploug et al., 2010).
It was recently shown that, as for the differentiation of heterocysts, removal of combined nitrogen from the growth medium is the sole and sufficient mean needed to elicit the development of the centrally located nitrogenase containing diazocytes (Fig. 2d) in Trichodesmium IMS101 (Sandh et al., 2012). The fact that the development of diazocytes takes between 8-27 h and that changes in cellular ultrastructure precede the expression of the nitrogenase enzyme (Sandh et al., 2012) strongly argues for a genetically based developmental background. Phylogenetically, the nonheterocystous Trichodesmium clade is a sister group to the heterocystous clade (Fig. 1b) and Trichodesmium shares several genomic and behavioral features with this clade (see above, below and Table S1). For instance, some heterocystous cyanobacteria (Anabaena sp.) first develop strings or subsets of adjacent proheterocysts upon nitrogen deprivation, while only the central cell develops into a mature heterocyst and the other regresses into vegetative cells (Wilcox et al., 1975). The similarity of the patterning of proheterocysts and diazocytes hints that several early regulatory elements may be shared in the nitrogen-regulated pathways of heterocystous genera and Trichodesmium. Diazocytes may during evolution have 'frozen' at this more minimalistic initial stage (Fig. 3), as a full differentiation is disadvantageous for the conditions offered in oceans (Staal et al., 2003;: Stal, 2009). In contrast to heterocysts, the diazocytes are not terminally differentiated cells and retain their ability to divide ( Fig. 2e; Fredriksson & Bergman, 1995), which may also contribute to their more flexible life style and allow the diazocytes (Trichodesmium) to more easily/rapidly adapt to prevailing conditions. To what extent Trichodesmium and heterocystous cyanobacteria share additional regulatory mechanisms that govern nitrogen deprivation signaling and pattern formation is now of great interest to be resolved.

Diazotrophy -separation in time
The timing of the N 2 fixation physiology of Trichodesmium is governed by the circadian clock (Chen et al., 1996(Chen et al., , 1998Dong & Golden, 2008). This regulates the transcription of the nif genes, the daily de novo synthesis of the subunits of the nitrogenase enzyme (NifHDK), a post-translational modification of NifH (Capone et al., 1990;Zehr et al., 1993;Chen et al., 1996), and the supply of appropriate levels of energy and reducing equivalents necessary for N 2 fixation activity (Staal et al., 2007).
Hence, it appears that Trichodesmium spp. not only separates N 2 fixation physically from net oxygen evolution via the development of a special cell type (diazocytes), although this may be the major protective mechanism, but also separates these two incompatible processes temporarily (Berman-Frank et al., 2001b), although in a more subtle way than in other nonheterocystous cyanobacteria . For instance, in contrast to in the latter and in concert with heterocystous cyanobacteria, nitrogenase activity in Trichodesmium operates within the day/light phase of the diel cycle, however, at a period around mid-day when the oxygen production is lowered and oxygen-scavenging mechanisms enhanced (respiration/Mehler reaction). This 'mid-day depression' in photosynthetic oxygen evolution is manifested as a lower quantum yield (~50%) and a low or negative net O 2 evolution (Berman-Frank et al., 2001b). Chlorophyll fluo-rescence kinetic microscopy at the single-cell level has also revealed flexible temporal and spatial switching between high fluorescence states (within a row of cells) and recovery states during subsequent non-N 2 -fixing periods in Trichodesmium IMS101 (Küpper et al., 2004). This rapid sequential switching within type I cells (functional PSII activity and enhanced Mehler reaction; potentially being the diazocytes) would allow diazotrophy even in cells lacking thick cell walls, may be orchestrated by rearrangements of the phycobilisomes between PSI and PSII (Küpper et al., 2009;Andresen et al., 2010), and may constitute the part of Trichodesmium's nitrogenase-protecting mechanism.
The increase in Mehler reaction (Kana, 1993;Milligan et al., 2007) and dark respiratory activity during the The diazotrophic physiology of Trichodesmium diazotrophic periods (Carpenter & Roenneberg, 1995;Kranz et al., 2009) is supported by the enhanced levels of the respiratory enzyme cytochrome c oxidase detected using proteomic analyses on N 2 -fixing cultures of Trichodesmium IMS101 (Sandh et al., 2011) and the enhanced levels of this enzyme specifically in the diazocytes, detected using immunogold localization in natural Trichodesmium populations (Bergman et al., 1993). In addition, Bryceson & Fay (1981) described zones of cells with higher tetrazolium salt depositions within specific areas of the trichomes, which support a reducing environment within the diazocytes. Although the cellular localization is not yet resolved, recent reports on novel carotenoids with strong antioxidant activity in Trichodesmium (Kelman et al., 2009) are also interesting in this context and worth examining further. In addition to a lower net oxygen evolution, the CO 2 fixation is also lowered at mid-day (Berman-Frank et al., 2001b) as recently verified using Nano-SIMS (Finzi-Hart et al., 2009). Earlier 14 C labeling experiments also demonstrated a lower CO 2 incorporation in central parts of the trichomes (Paerl, 1996). The decreased CO 2 fixation impacts the subcellular storage of carbon (glycogen) and the general carbon metabolism. For instance, carbohydrate storages are degrading ( Fig. 2c; El-Shehawy et al., 2003;Sandh et al., 2011), and recent proteomic analysis verifies a shift toward a catabolic carbon metabolism under diazotrophy, that is, a down-regulation of enzymes involved in glycogen storage and an up-regulation of enzymes involved in central carbon metabolism (Sandh et al., 2011). Diazotrophic conditions also led to enhanced levels of proteins involved in the biosynthesis of reducing equivalents (NADPH; through the oxidative pentose phosphate pathway) and the generation of a micro-oxic environment, consistent with increased respiratory activities (Sandh et al., 2011) and oxygen levels at mid-day.
Trichodesmium IMS101 also practices a light/dark (day/ night) separation of other basic cellular processes. While the highly energy-demanding processes, such as CO 2 fixation and diazotrophy, take place in light/day phase, cell division and diazocyte development are more pronounced in the dark/night phase (Chen et al., 1999;Sandh et al., 2009). Temporal separation of similar processes on a diel basis has previously been observed in marine unicellular cyanobacteria (Holtzendorff et al., 2001(Holtzendorff et al., , 2002Stockel et al., 2008;Shi et al., 2010). The circadian clock governs many of these processes and is entrained by the cellular ATP/ADP ratio, which in turn is governed by photosynthesis (Rust et al., 2011). Taken together, current data suggest that besides the development of diazocytes, some more subtle physiological mechanisms may act in concert to optimize diazotrophy in Trichodesmium.

Adaptation to nutrient stress
The nitrogen fixed in Trichodesmium is, as in other cyanobacteria, assimilated via the glutamine synthetaseglutamate synthase (GS-GOGAT) pathway, and as in heterocysts, the GS protein levels are higher in the diazocytes (Carpenter et al., 1992) to prevent feedback inhibition of the nitrogenase activity by the accumulation of the ammonia produced. Likewise, externally administrated sources of nitrogen negatively affect the expression of the nif genes, the synthesis of the nitrogenase enzyme, the nitrogenase activity, and diazocyte abundance in Trichodesmium (Ohki et al., 1991;Lin et al., 1998;El-Shehawy et al., 2003;Holl & Montoya, 2005;Sandh et al., 2011). As in other cyanobacteria (Herrero et al., 2004), when subject to N deprivation, there is a significant upshift in the cellular C : N ratio in Trichodesmium (Kranz et al., 2009), which may be a signal for enhanced transcription by the transcription factor NtcA (Table S1) of N-regulated genes. However, Trichodesmium appears to be flexible in this context, being able to fix N 2 in the presence of low concentrations of dissolved inorganic and organic nitrogen (Holl & Montoya, 2005), and the transcript of ntcA is not exclusively regulated by the availability of, for example, ammonium (Post et al., 2012), which suggests that our knowledge in this area is still limited.
Phosphorus and iron are critical nutrients restricting growth and N 2 fixation in today's oceans (Sanudo-Wilhelmy et al., 2001;Mills et al., 2004;Sohm et al., Fig. 3. A potential evolutionary scenarium for the development of the diazocytic phenotype in Trichodesmium. (1) An ancient nonheterocystous filamentous cyanobacterium, a forerunner of Nostocales and Oscillatoriales (Fig. 1b and c), under nitrogen deprivation conditions develop strings of proheterocysts (as in Anabaena-type spp.; Wilcox et al., 1975). (2) The majority of the proheterocysts revert back into vegetative cells, while one continus the evolution into a thickwalled heterocyst, the dominating nitrogen-fixing phenotype in limnic and terrestrial ecosystems today. (3) In oceans, this proheterocystous phenotype may have been retained and evolved into the strings of diazocytes we see in Trichodesmium spp. today, while the closest relatives ( Fig. 1b) either fix nitrogen in the dark (Lyngbya spp.; e.g. Lundgren et al., 2003) or have lost the capacity to fix nitrogen (Arthrospira spp.; Larsson et al., 2011;Latysheva et al., 2012). 2008; Moore et al., 2009). To overcome P limitations in oligotrophic waters, the Trichodesmium colonies migrate vertically in the water column to scavenge P and other nutrients using a buoyancy-regulating mechanism. This is provided by the pronounced gas vacuoles of Trichodesmium, which can withstand pressures down to depth of about 100-200 m (the highest known; Kromkamp & Walsby, 1992). In the upper euphotic zone, the colonies capture and store carbon and nitrogen (as glycogen and cyanophycin granules; Romans et al., 1994), and with this 'ballast', the colonies sink into deeper waters where P species may be acquired (Romans et al., 1994;Villareal & Carpenter, 2003;White et al., 2006b;Hewson et al., 2009). As the cellular ballast is metabolized in the deeper darker waters, the subsequently lighter colonies return to the euphotic zone to again capture light energy.
Trichodesmium is also known to adjust to periods of low P bioavailability by adopting high cellular N : P ratios (White et al., 2006a) in part by a substitution of its phospholipids by non-P membrane lipids (Van Mooy et al., 2009). Phosphorus uptake is also maximized via the uptake of both inorganic and organic phosphorous species (Stihl et al., 2001;Fu et al., 2005;Dyhrman et al., 2006;Orchard et al., 2009;Beversdorf et al., 2010;White et al., 2010). Enzymes hydrolyzing phosphoesters (alkaline phosphatase; Stihl et al., 2001;Orchard et al., 2009) and phosphonates (e.g. phosphonate hydrolase; Dyhrman et al., 2006) to yield phosphate have been identified in Trichodesmium. However, the globally widespread T. thiebautii lacks one of the alkaline phosphatase-encoding genes, phoA (Orchard et al., 2003). It has been shown that colonial alkaline phosphatase activities in Trichodesmium may rather be further enhanced by quorum-sensing signals (acylated homoserine lactones) released from colony-associated microorganisms (Van Mooy et al., 2012). The various phosphate pools available can be utilized either individually or in combination to sustain growth and N 2 fixation . Any 'luxury' uptake of phosphate is stored in subcellular structures (polyphosphate granules), which are common in natural Trichodesmium populations (Romans et al., 1994). As increased polyphosphate storages have been found in phosphatestarved cells (Orchard et al., 2010), the regulation of these subcellular structures in Trichodesmium is enigmatic. The variability among Trichodesmium species in relation to P-uptake genes (Orchard et al., 2003) also raises the question of niche differentiation and calls for further investigation.
Iron is a pivotal cofactor in a number of cellular processes, such as photosynthesis, N 2 fixation (in nitrogenase), and oxygen scavenging (Kustka et al., 2003;Shi et al., 2007). The high iron content of Trichodesmium cells suggests an efficient uptake and detainment capacity of iron and may even make Trichodesmium colonies ecologically valuable sources of this otherwise poorly soluble element in oligotrophic oceans (Kustka et al., 2003;Whittaker et al., 2011). No obvious genes in the Trichodesmium genome seem to encode for siderophores (Chappell & Webb, 2010), while the genes for transporters of siderophore-bound Fe 3+ and Fe 2+ and for enzymes related to cellular storage of iron are present (Castruita et al., 2006;Chappell & Webb, 2010). However, the involvement of siderophores in enhancing the uptake of low iron concentrations in Trichodesmium colonies has been reported (Achilles et al., 2003). The current hypothesis is that these siderophores are synthesized by associated microorganisms that indirectly facilitate the uptake of iron by Trichodesmium (Achilles et al., 2003). On the other hand, a recent study showed that such siderophore-bound iron is rather consumed by the bacteria than by the Trichodesmium cells (Roe et al., 2011). Another option is that the Trichodesmium colony formation per se facilitates the capture of enough particulate iron from, for example, eolian dust depositions to feed the colonies (Rubin et al., 2011).
Iron depletion is known to lead to a decrease in the frequency of diazocytes (Berman-Frank et al., 2001a;Küpper et al., 2008) and to a down-regulation of N 2 fixation, while photosynthetic capacities are maintained (Shi et al., 2007;Brown et al., 2008;Küpper et al., 2008). Iron limitation may also elicit a switch in the phycobiliproteins being used (Küpper et al., 2008). Other cyanobacteria have also been shown to replace ferredoxin with the ironfree flavodoxin (Sandmann et al., 1990). Monitoring isiB transcription, encoding a flavodoxin, has suggested that Trichodesmium populations in the Atlantic Ocean are rarely, or not at all, iron limited, while those in the Pacific Ocean are (Chappell et al., 2012). However, the expression of the two fld genes, encoding flavodoxins, in Trichodesmium IMS101 is also regulated by N availability and growth stage Chappell & Webb, 2010;Sandh et al., 2011), and their proposed use as iron limitation 'markers' may be questioned. Rather, the enhanced idiA transcription and IdiA levels noted under iron limitation may be a more suitable marker for iron stress in Trichodesmium (Webb et al., 2001;Chappell & Webb, 2010). As a strong up-regulation of Dps, yet another protein related to iron acquisition, was observed on the transfer of Trichodesmium to diazotrophic conditions (N stress), the role of Dps and other proteins related to iron acquisition and metabolism now also needs attention.

Impact on the ecosystem
The Trichodesmium abundance is roughly limited to waters warmer than 20°C, and temperature tolerance for growth The diazotrophic physiology of Trichodesmium and N 2 fixation in cultured strains of Trichodesmium (T. erythraeum IMS101, T. erythraeum GBRTRLI101 and T. tenue H94) ranges from 20 to 34°C, with optimal temperatures being 24-30°C depending on species and other growth conditions (Breitbarth et al., 2007;Chappell & Webb, 2010). However, Trichodesmium-like cyanobacteria (nifH and hetR phylogenies) were recently reported in Arctic waters suggesting wider temperature limits (Díez et al., 2012). Indeed, the very spotty nature of surface blooms of Trichodesmium does not represent the entire population in the ecosystem monitored and illustrates the difficulty in estimating the full global distribution of Trichodesmium via bloom registrations. In spite of this limitation, monitoring such blooms via remote sensing (via the SeaWiFS satellite; Subramaniam et al., 2002) verified that Trichodesmium blooms occur roughly between 20˚N and 20˚S in the eastern Pacific Ocean and that patches may occur even toward 40˚N and 40˚S in the Atlantic and the western Pacific and Indian Oceans (1998 andWestberry & Siegel, 2006). Besides Trichodesmium, numerous unicellular cyanobacteria share the same marine aquatic environment, notably the small-celled non-N 2 -fixing genera Prochlorococcus and Synechococcus (cell diameter~1 lm, genome sizes of 2 Mbp; Partensky et al., 1999;Scanlan et al., 2009), but also several N 2 -fixing unicellular cyanobacteria (Zehr et al., 2001;Montoya et al., 2004;Moisander et al., 2010). Among the latter are representatives of the marine genera Cyanothece, Crocosphaera, and N 2 -fixing cyanobacteria of the 'group A' nifH phylotype (e.g. UCYN-A). These diazotrophic unicellular cyanobacteria may show a broader temperature tolerance (15-30°C) than Trichodesmium, and some have been recovered from waters with detectable nitrate concentrations (Langlois et al., 2005). Observed community shifts from filamentous cyanobacteria in surface waters to unicellular cyanobacteria and/or heterotrophic bacteria in deeper waters (Langlois et al., 2005) may also suggest different ecological niche occupancies. Although estimates of the relative contribution to the total biogenic N 2 fixation in oceans by unicellular cyanobacteria (and heterotrophic bacteria) are increasing (Halm et al., 2012;Sohm et al., 2011a,b;Turk et al., 2011;Zehr & Kudela, 2011), the role of Trichodesmium as a C and N source in the world's oceans is still profound. For instance, Trichodesmium may account for up to 50% of the nifH genes in the North Atlantic Ocean (0°N -42°N and 67°W -13°W; Langlois et al., 2008), and Trichodesmium/Katagnymene represent up to 10 6 nifH genes per liter, while the unicellular cyanobacteria were represented by 10 5 nifH genes per liter and proteobacteria by 10 4 nifH genes per liter (Rijkenberg et al., 2011). Tyrrell et al. (2003) reported an even higher colony abundance of Trichodesmium in the tropical Atlantic Ocean (0-15 o N and 20 o W), and a Trich-odesmium surface bloom covered about 100 000 km 2 in the Arabian Sea (Capone et al., 1998).
For still unknown reasons, large segments of the Trichodesmium population are suddenly trapped at the surface forming easily observed pigmented layers of dying and decomposing cells ('blooms') (Capone et al., 1998). Such decomposing blooms function as gigantic 'fertilizer heaps' releasing large quantities of carbon, nitrogen, and other nutrients for the benefit of nondiazotrophic and heterotrophic biota in the surrounding water bodies. The cause of this destructive 'bloom' phenomenon is unknown, while the involvement of viral infections (Hewson et al., 2004) and/or autocatalyzed cell death processes  have been proposed, but the question is still open. The nitrogen fixed by Trichodesmium may, in addition, enter marine food webs via grazing by tunicates, copepods, and fish (Roman, 1978;Bryceson, 1980;Carpenter, 1983;Oneil & Roman, 1994;Eberl & Carpenter, 2007). Nitrogen isotope ratios in zooplankton in the North Atlantic Ocean strongly indicate that N 2 fixation is a major source of nitrogen for the marine zooplankton community (Montoya et al., 2002). This is at the same time unexpected, as the Trichodesmium toxin production has been inferred as a predator-deterring function (Layton et al., 2008) as well as a cause of death of several eukaryotic organisms, notably the copepod Acartia tonsa (Guo & Tester, 1994), several species of fish (Endean et al., 1993), and pearl oysters (Negri et al., 2004). Trichodesmium is now known to release different secondary metabolites, such as toxins, including the lipophilic chlorinated trichotoxin (Schock et al., 2011), the palytoxin causing clupeotoxism in humans via fish (Kerbrat et al., 2011), and the neurotoxin b-N-methylamino-L-alanine (Cox et al., 2005), the latter also found in diazotrophic bloom-forming cyanobacteria in the Baltic Sea (Jonasson et al., 2010). The ecological function of these toxins is still unknown.
Yet another factor that may play a role besides the mere cell number of an organism is their cell size/cell volume. As illustrated in Fig. 4, the size of Trichodesmium cells (approximate sizes given) makes the cellular volumes of this organism 'gigantic' compared to, for instance, cells of the unicellular genus Prochlorococcus and the diazotrophic genus Croccosphaera. As about 10-20 cells in each Trichodesmium filament are diazocytes (filled with NifH; Fig. 2b) and a large fraction of the nitrogen fixed may be released (as dissolved organic nitrogen or ammonium) from actively growing Trichodesmium populations (Capone et al., 1994;Glibert & Bronk, 1994;Mulholland et al., 2004;Mulholland, 2007), each Trichodesmium cell and filament may be viewed as a highly important source of new nitrogen for all nondiazotrophic small cells (unicellular cyanobacteria and bacteria) when sharing similar N-depleted aquatic marine environments.

Trichodesmium in future scenarios
The global importance of Trichodesmium in oceanic biogeochemistry has triggered numerous studies mimicking future global warming scenarios. As mentioned above, the optimum temperature range for growth and nitrogen fixation of Trichodesmium is 24-30°C (Breitbarth et al., 2007;Chappell & Webb, 2010), but Trichodesmium also survives lower temperatures and darkness (White et al., 2006b;Breitbarth et al., 2007). The ability to live at lower temperatures and in darkness is essential for their vertical migrations (White et al., 2006b) and may explain the occurrence of Trichodesmium in temperate (see LaRoche & Breitbarth, 2005) and potentially in cooler (Díez et al., 2012) waters. Global warming will lead to increased stratification resulting in shallower mixed layers and increased irradiance. High light intensity (up to 1000 lE) stimulates growth, diazocyte abundance, and N 2 fixation in Trichodesmium (Andresen et al., 2010;Kranz et al., 2010;Levitan et al., 2010) and provokes changes in pigment composition (Andresen et al., 2010), and correlates with increased O 2 evolution and CO 2 fixation (Kranz et al., 2010). Higher light intensities also cause faster protein turnover (Andresen et al., 2010) and an increase in RuBisC/O:PSII ratio (Brown et al., 2008). The downregulation in O 2 production via PSII correlates with the earlier noted peaks in N 2 fixation at noon (Berman-Frank et al., 2001b). Increased pCO 2 levels will not only stimulate CO 2 fixation, but also stimulate N 2 fixation (dependent on carbon skeletons for sequestration of the ammonium produced) and growth in Trichodesmium (Hutchins et al., 2007;Levitan et al., 2007Levitan et al., , 2010Ramos et al., 2007;Kranz et al., 2009Kranz et al., , 2010. One mechanism may be energy relocation from the costly carbon-concentrating mechanism (CCM; Badger et al., 2006;Kranz et al., 2011) toward CO 2 and N 2 fixation (Levitan et al., 2007;Kranz et al., 2011). This may in turn increase the release of the newly fixed N into the surrounding water body, thereby enhancing primary production of other organisms (Hutchins et al., 2007). Hence, in a scenario of increased temperatures and CO 2 concentrations in the world's oceans, the abundance of surface blooms of Trichodesmium is expected to increase (Breitbarth et al., 2007;Hutchins et al., 2007;Levitan et al., 2007), unless other unforeseen natural factors provoke the opposite reaction. For instance, element colimitations, rather than single-element limitations, may regulate the growth and N 2 fixation of natural Trichodesmium populations (Mills et al., 2004;Hutchins et al., 2007). Also, Garcia et al. (2011) have shown that the positive effect of higher CO 2 concentrations (on, e.g., N 2 fixation) is primarily seen at lower light intensities, and Rijkenberg et al. (2011) stress that the stimulations expected may be counteracted by decreased supply of nutrients from deeper waters as a consequence of enhanced stratification. In 1968, Fay et al. (1968 proposed that the reducing conditions in the cyanobacterial cell type recognized as 'heterocysts' could be the site for N 2 fixation. In 1973, Fleming & Haselkorn (1973) were the first to isolate nitrogenase from such heterocysts. Hence, heterocysts have during four decades acted as the 'consensus model' for successful light-driven N 2 fixation in cyanobacteria. In 1991, Bergman and Carpenter were the first to show that nitrogenase is localized in subsets or short strings of cells (diazocyte) in Trichodesmium. These are now recognized as a prerequisite for the light-driven N 2 fixation in Trichodesmium. This developmental mechanism is combined The diazotrophic physiology of Trichodesmium with mechanisms temporarily lowering oxygen evolution and orchestrating energy-competing processes in a multifaceted fashion. Hence, Trichodesmium is 'second-tonone' among potent daytime N 2 fixers, as is in particular evidenced by its great global ecological impact. However, many questions still remain. These include identification of genes/proteins that underpin the development of the diazocytes and their regulation, including mechanisms involved in the protection of the oxygen-sensitive nitrogenase. Because freshwater species exist within the (former) genus Katagnymene (see e.g. Komárek & Anagnostidis, 2005), an intriguing question is whether these are capable of developing diazocytes. Sequencing additional genomes within this globally important genus will allow comparative genomic analyses and potentially shed light on, for example, its uniquely low DNA coding proportion and the significance of its apparently expanding genome. Besides genomic analyses, other 'omics' and 'meta-omics' approaches (transcriptomics, proteomics, and metabolomics) need to be introduced if we are to comprehend the unique N 2 -fixing physiology of Trichodesmium at all organization levels. 'Metaomics' may provide more accurate information pertaining to the genetic diversity and the role of Trichodesmium and associated microorganisms than cultures, which often represents a minor part of the total species radiation. Such data may also reveal mutual and potentially life-sustaining interplays between Trichodesmium and the numerous associated microorganisms and open exciting research avenues into microbial evolution and marine microbial interphylum interactions, some potentially of a symbiotic nature. Focus should also be given to proteins that are highly up-regulated under N 2 -fixing conditions (Hewson et al., 2009;Sandh et al., 2011). Another area in need of exploration is the identification of adaptive ecological strategies, and possible niche differentiations, used by natural Trichodesmium populations. Finally, evolutionary aspects related to the uniquely different diazotrophic behavior of the genus Trichodesmium and to its placement in the evolution of affiliated cyanobacteria with unusual diazotrophic behavior are yet other compelling research areas to explore. The diazotrophic physiology of Trichodesmium