Commonly referred to as “saw dust” on the surface of the sea (10), Trichodesmium grows in filaments (referred to as trichomes) that can have 100–200 cells (Fig. 1B). In the field, Trichodesmium may be found in its colony form, as either puff-shaped or tuft-shaped colonies (Fig. 1A) (11). Colonies typically have 100–200 trichomes. In some systems, free trichomes may predominate (12). Trichodesmium are high-light adapted, and have gas vesicles that keep them near the surface of the ocean (Fig. 1C) (13). They often form thick blooms visible in satellite imagery (14). Trichodesmium grown in culture usually occurs as free trichomes of ≈80–100 cells per trichome during exponential phase, and aggregates during stationary phase.

Diazotrophy is a significant challenge for unicellular microorganisms, because the O₂ produced from CO₂ fixation is inhibitory to nitrogenase, the key enzyme in N₂ fixation. Therefore, diazotrophs have developed different behavioral, biochemical, and physical strategies to protect nitrogenase from the O₂ evolved during photosynthesis. Certain cyanobacteria, such as Gloeothece spp., temporally segregate the processes over a diel cycle by fixing CO₂ during the day and fixing N₂ at night (15). Others, such as Anabaena spp., have terminally differentiated cells, termed heterocysts, with thickened cell walls, and reduced PS II and rubisco activity (16). These cells serve to spatially segregate the 2 processes, with N₂ fixation occurring in the heterocysts, whereas oxygenic photosynthesis and CO₂ fixation occurs in vegetative cells. Trichodesmium is unique in that it is a nonheterocystous cyanobacteria that fixes both CO₂ and N₂ concurrently during the day (1). There is considerable current debate among researchers as to how these processes co-occur. Current theories, and sometimes conflicting data, suggest increased O₂ consumption within cells protects nitrogenase [e.g., increased hydrogenase (17), Mehler (18, 19), and superoxide dismutase activity (20)], in addition to both spatial (21–23) and temporal segregation of CO₂ and N₂ fixation (19, 24, 25). Much of the current debate revolves around the metabolic potential of individual cells in a trichome, and how those capabilities may differ among cells in a trichome or within individual cells over time (26).

NanoSIMS analysis showed substantial subcellular spatial variability in ¹⁵N and ¹³C enrichment along Trichodesmium trichomes and with depth through individual cells (Figs. 2, ,3,3, and and4);4); ¹³C/¹²C and ¹⁵N/¹⁴N ratio images generated from sectioned trichomes, along with correlated TEM maps, provide direct evidence of subcellular uptake localization within cells ≈8 h, and the redistribution of that enrichment after 24-h incubation (Fig. 3). In both sectioned and whole cells, we observed discrete hotspots enriched in ¹⁵N and ¹³C at 4 h, with increased density at 8 h (Figs. 2 I and J, and and33 B and C). These features are likely cyanophycin granules, a type of N-rich storage vacuole comprised of an asparagine and aspartate polymer found in many cyanobacteria (28–30), including Trichodesmium (21, 31). NanoSIMS analyses of Trichodesmium sections from 8 h indicate that the ¹³C:¹⁵N ratio of the cyanophycin granules is consistent with a mixture of asparagine (2:1) and aspartate (4:1), providing chemical evidence to identify these granules as cyanophycin. Also, these ¹⁵N hotspots correspond directly to vacuole-like structures identified as cyanophycin granules in corresponding TEM images (Fig. 3 A and B). The hotspots were no longer evident after 24-h incubation, at which point cells were uniformly enriched with newly fixed ¹⁵N (Figs. 2K and and33E). These results demonstrate the dynamic nature of these subcellular features, where it appears that newly fixed ¹⁵N is stored during the daytime (when N₂ fixation is active), and is subsequently used at night for cellular biosynthesis (when N₂ fixation has ceased). As might be expected, elevated levels of ¹³C are also associated with these cyanophycin granules (Fig. 3C).

NanoSIMS analysis also allowed us to examine subcellular isotope enrichment in the vertical dimension of cells. Based on analyses of multiple cells examined from multiple trichomes and time points, we determined that cyanophycin granules can comprise up to 6.4% (± 0.7%) of cell area/volume, and are evenly distributed along trichomes. In depth profile analyses of whole cells, as the NanoSIMS analyzed deeper and deeper layers, individual cyanophycin granules became apparent, and then disappeared as the ion beam sputtered through them (Fig. 4), suggesting that the granules are randomly distributed throughout the cell volume. It has been hypothesized that certain nondifferentiated cells, called “diazocytes”, along a trichome contain nitrogenase and are the sole sites of N₂ fixation (21, 22, 32). Fredriksson and Bergman (21) found that only 24% of these diazocyte cells contained cyanophycin granules, compared with 61% of non-N₂-fixing cells. We did not find evidence for this theory in our samples. When cells were examined in TEM thin sections (0-, 8-, and 24-h incubations), and with NanoSIMS depth profiling (4- and 8-h incubations), cyanophycin granules were observed at similar densities in all cells.

Previous studies have examined the transient nature of cyanophycin granules in single-celled and heterocystous diazotrophs. Sherman et al. (30) used immunocytochemical analysis to demonstrate the localization of cyanophycin at the polar plugs of mature heterocysts in Anabaena sp. PCC7120. Mackerras et al. (33) used cultures of Anabaena cylindrica (a heterocystous cyanobacteria) and Synechocystis 6308 (a unicellular cyanobacteria that temporally segregates CO₂ and N₂ fixation) grown in media with limited ammonium concentration ([NH₄⁺]). They demonstrate that synthesis and subsequent degradation of cyanophycin granules in these 2 organisms as [NH₄⁺] in the media is drawn down, indicating that the 2 cyanobacteria produce cyanophycin granules in N-replete conditions and use the stores during N-deplete conditions. Our study demonstrates the dynamic nature of the cyanophycin pool in Trichodesmium grown under diazotrophic conditions. Transient accumulation of cyanophycin indicates an uncoupling of CO₂ and N₂ fixation with cell growth. Interestingly, recent modeling of Trichodesmium metabolism (34) suggested that storage pools of C and N are dynamic over the diel cycle, accumulating during the day and consumed during the dark period. Our results provide direct evidence for this prediction.

In more than half of the trichomes analyzed from the 8- and 24-h incubations, we observed a marked reduction in ¹³C and ¹⁵N enrichment in cells located near the middle of trichomes (Fig. 5); this pattern may be related to programmed cell death (PCD) and/or trichome splitting. In Trichodesmium, as cells divide and the trichome increases in length, it is hypothesized that cells at the center of the trichome begin to have decreased metabolic activity and CO₂ and N₂ fixation. Eventually, the trichome splits at this point, yielding 2 shorter trichomes. Although PCD has been described for Trichodesmium (24, 35), we cannot confirm whether zones of reduced ¹³C and ¹⁵N uptake we observed are undergoing PCD, or whether the reduction in activity is due to other factors. However, secondary electron images we obtained before SIMS analysis provide visual evidence of thinner cells at these points (Fig. S1), suggesting that the trichome is stressed and about to break.

We used NanoSIMS analysis of whole cells to calculate uptake rates of CO₂ and N₂ fixation at the single cell level. These data were averaged for 40–120 contiguous cells located in multiple trichomes from each time point (Fig. 6), and were comparable to rates calculated from samples harvested in parallel, and bulk-analyzed by isotope ratio mass spectrometry (Fig. S2). Results from both approaches indicate that CO₂ fixation rates were highest in the morning (≈10 AM), whereas N₂ fixation rates were highest in the afternoon (≈12 PM; Fig. 6 and Fig. S2). These results provide evidence for a temporal decoupling of CO₂ and N₂ fixation in Trichodesmium, and corroborate earlier findings where Berman-Frank et al. (24) used fast repetition rate fluorescence (FRRF), O₂ production, ¹⁴CO₂ uptake, and acetylene reduction, to show maximal CO₂ fixation rates in the morning (≈11 AM) and afternoon (≈3 PM). Maximal N₂ fixation rates were measured at midday (≈12 PM), when CO₂ fixation was down-regulated. Küpper et al. (25), using fluorescence kinetic microscopy, provided further evidence for temporal segregation within individual cells, and suggest that there is a finely regulated shift between photosynthetic activity states throughout the course of the day that contributes to the protection of nitrogenase from O₂ evolved during photosynthesis. Together, our findings and these aforementioned studies build on the theory that increased O₂-consumptive activities within individual cells (17–20) enable them to temporally segregate the incompatible processes of CO₂ and N₂ fixation.

A controversial and evolving area of debate in the field of Trichodesmium physiology has been the suggestion that CO₂ and N₂ fixation within Trichodesmium are spatially segregated between different cells. The earliest microautoradiography studies suggested that spatial segregation happens at the level of the colony, with anoxic microzones occurring near the center of aggregates, where N₂ fixation was localized and photosynthesis down-regulated (36, 37). This theory was later questioned, when it was demonstrated that individual trichomes from disrupted colonies still fix both N₂ and CO₂ (38). More recently, conflicting immunochemical studies have examined the distribution of nitrogenase proteins along trichome cells. Some studies (23, 39) indicate that nitrogenase is dispersed throughout all cells in colonies, whereas other research suggests that only ≈15% of cells along a trichome contain nitrogenase (21, 22), and that these cells occur in clusters along the trichome (21, 32). These clusters, referred to as diazocytes, suggest a spatial segregation of N₂ fixation and CO₂ fixation. It should be noted that cells that contain nitrogenase are not necessarily photosynthetically inactive (15).

In our study, after 2 h of exposure to labeled isotopes, newly fixed ¹³C and ¹⁵N was apparent and relatively evenly distributed in all cells along trichomes (Fig. 2). Although depleted regions did develop by 8 and 24 h, uptake of both C and N was simultaneously lower, and these regions corresponded to morphologically anomalous “thinning” spots thought to be prebreakage points (see discussion above). However, in our experiment, the lack of clear spatial differences in uptake patterns could be attributed to rapid redistribution of recently fixed C and N. Wolk et al. (40), using autoradiography with ¹³N-N₂ gas, demonstrated in the cyanobacteria Anabaena variabilis that N₂ fixing cells are able to fix and redistribute newly fixed N to neighboring cells in <1.5 min. More recently, in experiments with the heterocystous cyanobacteria Anabaena oscillariodes, extant heterocysts were depleted in ¹⁵N relative to vegetative cells (41), despite the fact that the heterocysts are the indisputable sites of N₂ fixation. This phenomenon may be due to the loss of diffusible ions and small molecules in the cytosol that are not fixed in place by cell fixatives used before NanoSIMS analysis (e.g., gluteraldehyde) (42). If a similar mechanism occurred in Trichodesmium cells, it could account for soluble metabolites formed in the early time points being undetectable. Although our data seems to suggest otherwise, it is possible that CO₂ and N₂ fixation did not occur concurrently within all cells.

For NanoSIMS microanalysis, multiple whole Trichodesmium trichomes were filtered, washed with Milli-Q H₂O, and dried onto a silica chip. SIMS was performed at Lawrence Livermore National Laboratory by using a Cameca NanoSIMS 50 instrument, according to previously described methods (41). Briefly, a ≈2pA Cs⁺ primary beam focused to ≈150 nm was scanned across a 256 × 256 pixel 10–15 μm² raster to generate secondary ions in ≈20 serial quantitative secondary ion images. Electron multiplier detectors collected ¹²C⁻, ¹³C⁻, ¹²C¹⁴N⁻, ¹²C¹⁵N⁻, and ³¹P⁻ ions, and secondary electrons were simultaneously imaged. The NanoSIMS was tuned for ≈7,000 mass resolving power to resolve isobaric interferences. Samples were presputtered to at least 100 nm to achieve sputtering equilibrium; depth of measurement analysis ranged from 200 to 350 nm. For each time point, 4–5 trichomes were analyzed. We include data in Figs. 2 and and55 where we were able to analyze at least 40–120 contiguous cells. Data were processed as previously described (41) by using custom software. Briefly, each cell was defined as a region of interest using a 30% ¹²C threshold; individual cells are easily identified in NanoSIMS secondary electron, ¹²C and ¹⁴N images. The isotopic composition for each ROI was calculated, corrected with reference standards, and converted to percentage uptake (Figs. 2 and and5)5) (41). Reference standards used included finely ground bovine liver (NIST SRM 1577b) (41) and a Bacillus subtilis spore preparation (44). Measurement precision, σ_(internal), was 0.4–1.4% (2σ) for individual ¹³C/¹²C and ¹⁵N/¹⁵N measurements, and replicate analyses of the standard yielded an analytical precision, s_(std), of 2.1% (2σ) for an individual measurement. Repeated measurements with depth on selected cells and on multiple filaments within a time point were used to ensure measurement accuracy. Rates of N and C fixation (Fig. 6) are the change C or N percentage uptake hour⁻¹ relative to the initial measurement. TEM analysis (Tecnai G2) was conducted on Trichodesmium filaments from the 0-, 8-, and 24-h time points, dehydrated in a series of ethanol washes, embedded in LR White epoxy, and microtomed to 150 nm. Correlated NanoSIMS analysis was then performed on the filaments imaged by TEM. Percentage area of cyanophycin granules were measured relative to whole-cell area in 65 cells from 6 distinct trichome thin-sections taken from the 0-, 8-, and 24-h time points. NanoSIMS analyses of Anabaena filaments (Fig. S3) are described in Popa et al. (41).