Temporal Patterns and Intra- and Inter-Cellular Variability in Carbon and Nitrogen Assimilation by the Unicellular Cyanobacterium Cyanothece sp. ATCC 51142

Unicellular nitrogen fixing cyanobacteria (UCYN) are abundant members of phytoplankton communities in a wide range of marine environments, including those with rapidly changing nitrogen (N) concentrations. We hypothesized that differences in N availability (N2 vs. combined N) would cause UCYN to shift strategies of intracellular N and C allocation. We used transmission electron microscopy and nanoscale secondary ion mass spectrometry imaging to track assimilation and intracellular allocation of 13C-labeled CO2 and 15N-labeled N2 or NO3 at different periods across a diel cycle in Cyanothece sp. ATCC 51142. We present new ideas on interpreting these imaging data, including the influences of pre-incubation cellular C and N contents and turnover rates of inclusion bodies. Within cultures growing diazotrophically, distinct subpopulations were detected that fixed N2 at night or in the morning. Additional significant within-population heterogeneity was likely caused by differences in the relative amounts of N assimilated into cyanophycin from sources external and internal to the cells. Whether growing on N2 or NO3, cells prioritized cyanophycin synthesis when N assimilation rates were highest. N assimilation in cells growing on NO3 switched from cyanophycin synthesis to protein synthesis, suggesting that once a cyanophycin quota is met, it is bypassed in favor of protein synthesis. Growth on NO3 also revealed that at night, there is a very low level of CO2 assimilation into polysaccharides simultaneous with their catabolism for protein synthesis. This study revealed multiple, detailed mechanisms underlying C and N management in Cyanothece that facilitate its success in dynamic aquatic environments.

Unicellular nitrogen fixing cyanobacteria (UCYN) are abundant members of phytoplankton communities in a wide range of marine environments, including those with rapidly changing nitrogen (N) concentrations. We hypothesized that differences in N availability (N 2 vs. combined N) would cause UCYN to shift strategies of intracellular N and C allocation. We used transmission electron microscopy and nanoscale secondary ion mass spectrometry imaging to track assimilation and intracellular allocation of 13 Clabeled CO 2 and 15 N-labeled N 2 or NO 3 at different periods across a diel cycle in Cyanothece sp. ATCC 51142. We present new ideas on interpreting these imaging data, including the influences of pre-incubation cellular C and N contents and turnover rates of inclusion bodies. Within cultures growing diazotrophically, distinct subpopulations were detected that fixed N 2 at night or in the morning. Additional significant within-population heterogeneity was likely caused by differences in the relative amounts of N assimilated into cyanophycin from sources external and internal to the cells. Whether growing on N 2 or NO 3 , cells prioritized cyanophycin synthesis when N assimilation rates were highest. N assimilation in cells growing on NO 3 switched from cyanophycin synthesis to protein synthesis, suggesting that once a cyanophycin quota is met, it is bypassed in favor

INTRODUCTION
Nitrogen fixing microorganisms (diazotrophs) are critical suppliers of bioavailable forms of nitrogen (N, e.g., ammonium) in natural ecosystems. In the surface ocean where primary production is often limited by N availability, unicellular N 2 -fixing cyanobacteria (UCYN) are now recognized as having key roles in biogeochemical cycles (Zehr et al., 2001;Karl et al., 2002;Montoya et al., 2004;Zehr, 2011;Wilson et al., 2017).
Because of the scarcity of cultured representatives, Cyanothece sp. ATCC 51142 (henceforth Cyanothece 51142, recently reclassified as Crocosphaera subtropica; Mareš et al., 2019) has become an experimental model of UCYN (Reddy et al., 1993;Schneegurt et al., 1994;Colón-López and Sherman, 1998;Sherman et al., 1998;Li et al., 2001). The sequenced Cyanothece 51142 genome and controlled Cyanothece culture studies have provided insights into the genetic controls underlying the temporal segregation of N 2 fixation activity, regulation of photosynthesis Toepel et al., 2008;Welsh et al., 2008), and carbon (C) metabolism (Schneegurt et al., 1994;Colón-López and Sherman, 1998;Aryal et al., 2011;Bernstein et al., 2015). Cyanothece 51142 appear to restrict N 2 fixation to the night time to protect the N 2 fixing enzyme complex, nitrogenase, from inactivation by molecular oxygen produced by photosynthesis during the day time (Reddy et al., 1993;Welsh et al., 2008). Intracellular storage of the newly fixed C and N facilitates temporal separation of activities. Specifically, N fixed during the night is initially stored as cyanophycin until it is used for processes such as protein and nucleic acid synthesis, whereas C fixed during the day is stored as polysaccharides until it is respired the following night to supply reducing equivalents (NADPH) and ATP needed to support N 2 fixation (Schneegurt et al., 1994;Li et al., 2001;Großkopf and LaRoche, 2012;Inomura et al., 2019). C respiration during the late afternoon and night also facilitates nitrogenase activity by depleting molecular oxygen that diffuses into the cell from the environment (Dron et al., 2012;Großkopf and LaRoche, 2012;Červený et al., 2013;Inomura et al., 2019).
In addition to their ability to fix N 2 , Cyanothece (and other UCYN) can assimilate various forms of combined nitrogen (e.g., NH 4 , NO 3 , urea, and amino acids; Mulholland et al., 2001;Holl and Montoya, 2005;Dekaezemacker and Bonnet, 2011;Masuda et al., 2013). Combined nitrogen generally downregulates N 2 fixation because its uptake and assimilation are energetically less costly than the processes supporting nitrogenase activity (Mulholland et al., 2001;Holl and Montoya, 2005;Eichner et al., 2014). This flexibility in nitrogen metabolism is considered to be one reason why Cyanothece thrives in a variety of marine environments with widely ranging nitrogen concentrations (Rippka, 1988;Short and Zehr, 2007;Webb et al., 2009;Bonnet et al., 2011). Cyanothece 51142 appears to efficiently manage C and N storage using multiple copies of genes encoding enzymes in polysaccharide metabolism and conserved gene clusters that coordinate intersecting pathways of C and N metabolism Zhang et al., 2018). For example, the ornithine-ammonia cycle (OAC) may facilitate efficient sequestration and remobilization of N (Zhang et al., 2018). Other N-rich compounds, including arginine, can be catabolized to recycle N within the cell Burnat et al., 2019). CO 2 and N 2 fixation activities are also at least partly regulated according to the cell cycle demands that vary across the day-night cycle (Dron et al., 2013). For example, the N demands of nucleic acid synthesis and the C and energy requirements for new cell synthesis generally restrict cell division to hours when intracellular N reserves and photosynthetic rates are high (Dron et al., 2013;Červený et al., 2013). In contrast to N 2 fixation, combined N assimilation and photosynthetic C assimilation should not require temporal separation, raising questions about how N and C allocations change depending on whether cells are growing on N 2 or combined N.
In this study, we characterized C and N assimilation rates over a diel cycle in Cyanothece 51142 grown under obligate diazotrophic and non-diazotrophic conditions (with added NO 3 ). Our approach combined nanoscale secondary ion mass spectrometry (nanoSIMS) and transmission electron microscopy (TEM) to track the assimilation of 13 C-labeled inorganic C and 15 N-labeled N 2 or NO 3 into individual cells and resolve their assimilation into polysaccharides, cyanophycin, and other inclusion bodies over the day-night cycle. We observed clear differences in N allocation patterns, but also unexpected withinpopulation heterogeneity, including wide variation in labeling of storage inclusions and day-time N 2 fixation. We discuss these observations and highlight how careful evaluation of these complex nanoSIMS data revealed key metabolic mechanisms underlying C and N management in Cyanothece 51142 that facilitate its success in dynamic aquatic environments.

Bioreactor and Semi-Continuous Cultures
Cyanothece 51142 cultures were maintained in 400-or 1,000-mL flat panel photobioreactors (FMT 150,Photon Systems Instruments,Brno,Czech Republic) at 28 • C and 300 µmol photons m −2 s −1 with a 14 h:10 h light:dark cycle (14L:10D), with L0 at 07:30 and D0 at 21:30. The initial and final hours of each light cycle were set to follow a sinusoidal increase or decrease in light intensity, respectively. Triplicate cultures were grown in ASP2 medium (Provasoli et al., 1957;Van Baalen, 1962) either supplemented with 17 mM NO 3 ("NO 3 culture") or prepared with no inorganic N added ("N 2 culture"). Cultures were bubbled with ambient air (400 mL min −1 ) and maintained under turbidostat mode at OD 680 of ∼0.5. Another set of triplicate cultures were grown in ∼300 mL glass tubes in ASP2 medium without NO 3 amendment at 28 • C under 300 µmol photons m −2 s −1 (14L:10D, same light regime as above) under semi-continuous, fed-batch mode ("SC-N 2 culture"). The SC-N 2 cultures were cultivated and maintained to ensure a "back-up" option in case the turbidostat (N 2 ) cultures failed during the experiment. Because both culture conditions were stable throughout the experiment, we opportunistically sampled both cultures. The SC-N 2 culture exhibited a wider range of phenotypic variability compared to the N 2 culture that may be akin to some more dynamic natural environments. Therefore, we chose to include results for both diazotrophic cultures. All cultures maintained constant population sizes for >21 days prior to sampling. Culture and cell characteristics for each treatment are given in Table 1. An analysis of photosynthetic electron transport and the energetic costs of N and C acquisition in the same cultures studied here are given in Rabouille et al. (In Revision).

Stable Isotope Probing Experiments
For stable isotope probing experiments, samples were collected from one culture replicate from each condition. Incubations were performed by sub-sampling cultures into 6 mL gas-tight vials and amending the ASP2 medium with NaH 13 CO 3 (all cultures) and either 15 N 2 (N 2 and SC-N 2 cultures) or Na 15 NO 3 (NO 3 culture). Vials were incubated under light and temperature conditions that were equivalent to culture conditions. Incubation durations were 2 h in the morning, 2.5 h during the day, and 10 h during the night, with shorter incubations in early night (3 h) and late night (5 h). Isotope labeling was calculated from the known amounts of label added to the incubation medium and measured initial concentrations of unlabeled substrate in the bioreactors (Supplementary Table 1). The 15 N 2 enriched stock was prepared by injecting 10 mL of 15 N 2 gas into 43 mL of the ASP2 medium, followed by an equilibration for >24 h. The 15 N-N 2 atom fraction in the incubation medium was calculated assuming that 15 N 2 was fully equilibrated with the stock solution. Since this may lead to an underestimation of N 2 fixation rates (Mohr et al., 2010), we refrain from comparisons of N assimilation rates between N treatments (N 2 vs. NO 3 ). However, comparisons over time and among cells within each treatment are not affected since any potential underestimation would be similar in all 15 N 2 incubations. NO 3 concentration in the incubation medium was estimated by averaging NO 3 concentrations measured in the bioreactor on the respective day of the experiment. Dissolved inorganic carbon (DIC) concentration in the incubation medium was estimated by measuring DIC concentrations in the bioreactor at three time points during the day and interpolating them to the starting time points of our stable isotope incubations. Because the DIC concentration in the cultures varied depending on the time of the day, 13 C-DIC atom fractions varied during our incubations, although the amounts of added NaH 13 CO 3 were the same (Supplementary Table 1).

TEM Analysis
At the end of each isotope-labeling incubation, cells were collected and centrifuged at 2,700 rpm for 10 min at room temperature. One microliter of the pellet was mixed with 1 µL of 20% bovine serum albumin and transferred to a formvarcoated 100 mesh TEM grid. After removing the excess liquid with a filter paper, the grids were frozen in liquid ethane cooled with liquid nitrogen. Freeze-substitution was carried out in a 2% mixture of OsO 4 in 100% acetone (v/v) sequentially at three temperatures: −90 • C (for 96 h), −20 • C (for 24 h), and 4 • C (for 10 h). Temperature was increased at a rate of 5 • C h −1 (from −90 to −20 • C) and 3 • C h −1 (from −20 to 4 • C). After freezesubstitution, the samples were washed three times in acetone and infiltrated sequentially in a 2:1, 1:1, and 1:2 (v/v) mixture of acetone and low-viscosity Spurr resin (EMS) for 1 h in each step. Finally, the samples were incubated overnight in a 100% resin, transferred to embedding molds, and allowed to polymerize. Thin sections (200 nm) were cut with a diamond knife, placed on Cu-indexed TEM grids (rinsed in 30% ethanol), and contrasted for 20 min in saturated ethanolic uranyl acetate (EMS, Hatfield, United States; concentration 13 g/100 mL 50% ethanol; solution filtered before use through a 0.45 µm pore size filter). Images were taken using a JEOL 1010 TEM at 80 kV.

NanoSIMS Analysis
Nanoscale secondary ion mass spectrometry analyses were performed on two types of samples: (i) thin sections that were first imaged by TEM (as described above) and (ii) whole cells collected on polycarbonate filters. For downstream analysis of samples initially imaged by TEM, the lowest primary ion beam current (0.5 pA) was used to achieve the highest lateral resolution afforded by the instrument (∼50 nm). However, because the samples were very thin (∼200 nm), the number of imaged frames was rather low (20-50) before the cell material was sputtered away. This sometimes resulted in a poor signal-to-noise ratio (SNR) in the final secondary ion images, and thus, a low number of cells for which good quality complementary TEM and nanoSIMS images are available. Additional measurements were therefore performed on cells deposited on filters, because the imaging could be done with a stronger beam (2 pA) and over a larger area and many more frames (>200). However, the improved throughput and SNR came at the expense of a lower spatial resolution (see section "Results"). For nanoSIMS analysis of whole cells, the cells were filtered onto polycarbonate filters (2.5 cm diameter, 0.2 µm pore size, Millipore), washed three times, air-dried, and stored at room temperature. Chemical fixation was not performed thus avoiding dilution of the isotope label. Just prior to nanoSIMS analysis, filters were sputtercoated with a 10-nm gold layer, cut into small circular pieces (5 mm diameter) suitable for the nanoSIMS sample holder, and imaged with a Neoscope II JCM-6000 scanning electron microscope (JEOL, Japan) to check sample quality (cell integrity and cell density). Nanoscale secondary ion mass spectrometry measurements were performed with the NanoSIMS 50L instrument (Cameca, France) operated at Utrecht University. Areas of interest were first pre-sputtered with Cs + -ions until secondary ion yields stabilized. Subsequently, the primary Cs + -ion beam was scanned over the sample (areas between 10 µm × 10 µm and 30 µm × 30 µm in size, dwell time of 1 ms pixel −1 ) while detecting secondary ions 12 C − , 13 C − , 16 O − , 12 C 14 N − , 12 C 15 N − , 31 P − , and 32 S − . To increase the overall signal, the same area was imaged multiple times, and the resulting ion count images were aligned and accumulated.

NanoSIMS Data Processing and Quantification of Rates
Nanoscale secondary ion mass spectrometry data were processed with the Look@NanoSIMS software (Polerecky et al., 2012) to quantify 13 C and 15 N atom fractions, denoted as x( 13 C) and x( 15 N) (Coplen, 2011), in regions of interest (ROI's) corresponding to cells or inclusion bodies (polysaccharide granules or cyanophycin inclusions). After drawing ROIs manually, x( 13 C) in the ROI was determined from the total counts of secondary ions 12 C − and 13 C − accumulated over the ROI pixels as x( 13 C) = 13 C − /( 12 C − + 13 C − ). Similarly, x( 15 N) in the ROI was determined from the total counts of 12 C 15 N − and 12 C 14 N − accumulated over the ROI pixels as x( 15 N) = 12 C 14 N − /( 12 C 14 N − + 12 C 15 N − ). The C-and N-specific rates of 13 C and 15 N assimilation into whole cells (k C and k N , respectively) were calculated as: while the C-and N-specific rates of 13 C and 15 N incorporation into polysaccharide granules (p C or p N , respectively) and cyanophycin inclusions (y C and y N , respectively) were calculated as: In Eqs 1-4, x( 13 C) S and x( 15 N) S are atom fractions of the C and N source, respectively, and x( 13 C) ini and x( 15 N) ini are the initial atom fractions of C and N in the ROI, respectively. The isotope labeling of the C and N sources, x( 13 C) S and x( 15 N) S , was assumed to be constant during the incubation and was calculated as described above (Supplementary Table 1). x( 13 C) ini and x( 15 N) ini were determined by averaging data obtained from cells that were not exposed to the labeled substrate (i.e., control cells; x( 13 C) ini = 1.052 × 10 −2 , SD( 13 C) ini = 0.007 × 10 −2 , and x( 15 N) ini = 3.75 × 10 −3 , SD( 15 N) ini = 0.04 × 10 −3 , n = 30). A cell or an inclusion body was considered significantly enriched in 13 C if the 95% confidence interval of its estimated mean 13 C atom fraction did not overlap with that of the control cells, i.e., if x( 13 C) ± 2 × SE( 13 C) did not overlap with x( 13 C) ini ± 2 × SE( 13 C) ini . Here, the standard errors were calculated as SE( 13 C) ini = SD( 13 C) ini / √ n for the control cells, and SE( 13 C) = x( 13 C) × PE( 13 C) for each individual cell or inclusion body, where the relative Poisson error was calculated from the total counts of 12 C − and 13 C − in the cell or inclusion body as PE ( 13 C (Polerecky et al., 2012). The same approach but using the total counts of 12 C 14 N − and 12 C 15 N − was applied to determine significant enrichment in 15 N.
Note that the C-and N-specific rates of 13 C and 15 N assimilation into whole cells, polysaccharide granules, and cyanophycin inclusions have units of per time (i.e., h −1 or day −1 ) and give the rate of 13 C and 15 N assimilation rates normalized to the C and N content of the ROI [i.e., mol C (mol C) −1 h −1 for k C , p C , and y C , and mol N (mol N) −1 h −1 for k N , p N , and y N ]. Evaluation of the variability in assimilation rates among cells and intracellular inclusions required considering how the measured x( 13 C) and x( 15 N) in the ROI were linked to cell growth and metabolism of internal C and N pools (e.g., synthesis and degradation of polysaccharides and cyanophycin inclusions, or recycling of N from existing proteins during cyanophycin synthesis). These considerations are summarized in the Discussion section (Section "Interpreting Isotopic Enrichment Imaging Data"). The assumptions underlying the rate calculations presented above are further explored and discussed in Polerecky et al. (In Revision).

Identification of Intracellular Inclusions
Prominent intracellular inclusion bodies identified in TEM images of Cyanothece 51142 included carboxysomes, cyanophycin inclusions, polyphosphate bodies, polysaccharide granules, and thylakoid membranes (Figure 1). Some, but not all, of these inclusions could be reliably identified in nanoSIMS images when the accumulated secondary ion counts 12 C 14 N − , 32 S − , and 31 P − were combined into RGB overlays. Specifically, carboxysomes, which were identified in TEM images as dark areas with a characteristic hexagonal shape (Figures 1A,C, arrows labeled "c"), had relatively higher CN − and S − but lower P − counts than the surrounding cell material. These differences caused carboxysomes to appear yellow-green in the RGB overlays (Figures 1D,F). Cyanophycin inclusions, which were identified through their oval shape and darker appearance in the TEM images, had markedly higher CN − counts, while the S − and P − counts were not different from the surrounding cell material. The relative enrichment in CN − counts gave cyanophycin inclusions an orange-to-red appearance in the RGB overlays (Figure 1, arrows labeled "cy"). Polyphosphate bodies had higher P − counts and lower CN − and S − counts relative to the surrounding cell material and appeared as bluish spots in the RGB overlays (Figure 1, arrows labeled "p"). Although polysaccharide granules could be identified in TEM images as bright oval shapes ( Figure 1A, arrows labeled "ps"; see also Deschamps et al., 2008), their identification from the nanoSIMS images was not reliable. For example, most polysaccharide granules were associated with localized decreases in CN − counts ( Figure 1D). However, the contrast between the polysaccharide granules and the cell matrix was low, and similar decreases in CN − counts sometimes occurred even when there was no obvious presence of polysaccharide in the TEM images. Neither could variability in S − , P − , C − , or O − counts (C − and O − data not shown) be used to distinguish polysaccharide granules in the nanoSIMS images. Similarly, thylakoid membranes were not identifiable using the nanoSIMS images although their visibility in the TEM images was often good (Figures 1A-C, arrows labeled "t"). Conversely, nucleoids were clearly observed as the violet-colored regions in the RGB overlays due to P counts being markedly higher than the surrounding cell material and CN − and S − counts that were similar to the surrounding cell material (Figures 1D-F, arrows labeled "n"); however, direct nucleoid identification in the TEM images was not possible.

Carbon and Nitrogen Assimilation Rates and Allocation Patterns
Daytime C fixation in Cyanothece 51142 grown under diazotrophic conditions in turbidostat mode ("N 2 culture") was observed in all but one of the 126 cells imaged (Figures 2A,B). The C-specific rates of 13 C assimilation in whole cells, k C , were highest in the morning and declined on average by about 80% in the afternoon ( Table 2). 13 C enrichment was highest in polysaccharide granules and lower and diffusely spread throughout the cell matrix (Figures 3A,B). The C-specific rates of 13 C assimilation in polysaccharide granules, p C , varied (CV≈32%), with 61% of the variance explained by differences among cells and 39% of the variance explained by differences within cells (Supplementary Figure 1A). Moreover, the relative area of the cell sections covered by polysaccharide granules varied among cells (range: 0.08-0.28, CV≈43%) and was significantly positively correlated with the k C values (R = 0.68, p = 0.002; Supplementary Figure 2B).
The majority of cells in the N 2 culture fixed N 2 during the night ( Figure 2C), with the exception of one cell (out of 104 imaged) that fixed N 2 in the morning (green circle in Figure 2A; "asynchronous diazotrophic" cell in Figure 3A). During the early night, cells grouped into two clear subpopulations exhibiting different activities: P1 showed significant N 2 fixation and accounted for 83% of cells, while P2 showed no significant N 2 fixation and accounted for 17% of cells (compare green and red circles in Figure 2C). P1 and P2 showed low but significant C fixation during the early night incubation ( Figure 2C). The k C values in P1 and P2 did not differ at night [ANOVA, , and from a sample collected at 07:00 following a night-time incubation (21:45-07:00) from the NO 3 culture (C,F). Arrows in the images point to sub-cellular structures identified as carboxysomes (c), cyanophycin inclusions (cy), polyphosphate bodies (p), polysaccharide granules (ps), thylakoid membranes (t), and nucleoid (n). NanoSIMS images are shown as RGB overlays of secondary ion counts 12 C 14 N − (red), 32 S − (green), and 31 P − (blue). Note that the hues are not comparable among the images because, for each overlay, the contrast for the three color channels was modified so as to enhance the overall visibility of the intracellular variability. In addition to P − , the polyphosphate bodies had markedly increased O − ion counts (data not shown).  in the early night ( Table 2). Significant C fixation was observed in 5% of cells, and these cells again grouped into subpopulations depending on whether they also fixed N 2 (3%) or not (2%) (compare green and red pluses in Figure 2C). The remaining 95% of cells showed no detectable C nor N 2 fixation during the late night (black pluses in Figure 2C). The C and N 2 fixation patterns observed separately during the day and night were reflected in cells incubated with isotopes over the full night + day cycle (Figures 2D, 3D). The majority of cells (∼87%) fixed both C and N 2 (P3, orange circles in Figure 2D), whereas the remaining 13% of cells had k C values similar to P3 but showed no significant N 2 fixation (P4, red circles in Figure 2D). Average k C and k N values were similar for the majority of cells incubated for the full night + day cycle, but these values were only about 5% of the peak k C and k N values observed during the morning and early night, respectively ( Table 2). These dramatic shifts in metabolism over the day caused k C in the morning to overestimate specific growth rate (µ = 0.22 day −1 ) by 7.7-fold. Theoretically, the average night + day k C value should estimate µ, but was only 0.09 day −1 . Similarly, k N at night overestimated µ by up to 6-fold, and the average night + day k N was only 0.08 day −1 . The discrepancies between µ and the average k C and k N from the night + day (∼20 h) incubations were likely caused by differences in incubation conditions, including gas flow and medium exchange in the turbidostat that were not possible in the isotope labeling incubation. Finally, k C :k N varied widely among cells (range 0.6-3.3, mean = 1.24, SD = 0.78, CV = 63%).
Cyanophycin inclusions showed the greatest 13 C and 15 N enrichment compared to other inclusions within individual cells from the early night incubation (Figures 3C, 4A-C). Similar to the data clustering observed for the whole cells, individual cyanophycin inclusions grouped into two clear subpopulations, one with significant 15 N enrichment (P1) and one with no significant 15 N enrichment (P2) (compare green and red circles in Figure 4B). There was no significant correlation between the C-and N-specific 13 C and 15 N assimilation rates in cyanophycin granules, y C and y N , in P1 (R = 0.10, p = 0.10), and y C :y N were highly variable between cyanophycin granules among cells (range: 0-0.09, mean = 0.015, SD = 0.016, CV = 107%) and within individual cells (Figure 4A). In one cell where we could clearly FIGURE 3 | Images of isotopic enrichment in Cyanothece 51142. Shown are overlays of 13 C (red) and 15 N (green) enrichments measured in cells from three parallel cultures (N 2 , diazotrophic culture grown in turbidostat mode; NO 3 , culture grown on nitrate in turbidostat mode; SC-N 2 , diazotrophic culture grown in semi-continuous batch mode). Cells were incubated during time intervals indicated above the images. Images shown in panels (A,B,E-G) were obtained from thin cell sections analyzed by TEM [panels (A,E,G) correspond to panels (A-C) in Figure 1], whereas panels (C,D,H-J) show images of cells collected on polycarbonate membrane filters. Additional images of filtered cells are shown in Supplementary Figure 2. In each image, the intensity ("brightness") of the red and green color scales linearly with the 13 C and 15 N enrichment, respectively, with black indicating no enrichment. Note, however, that because the scaling for the red and green colors was optimized independently for each image to enhance the visibility of the intracellular heterogeneity, the intensities of the red and green colors are not comparable among the images. In panels (A,C,E-G,I,J), examples of cyanophycin inclusions (cy) and carboxysomes (c) are marked with dashed-line and solid-line arrows, respectively. In panels (A,B,E-G), areas and spots with pronounced 13 C enrichment (red) correspond to polysaccharide granules. Black areas in panels (F,G) correspond to voids (v) due to artifacts associated with the preparation of the thin cell sections. resolve all relevant intracellular structures, we observed low but significant 15 N enrichment in the carboxysomes and a slightly greater 15 N enrichment in the nucleoid in addition to the strong 15 N enrichment in the cyanophycin inclusions ( Figure 4C).

NO 3 Culture
Cyanothece 51142 grown under non-diazotrophic conditions in turbidostat mode (with NO 3 added; "NO 3 culture") showed daily patterns of C fixation that were similar to the N 2 culture. Morning k C was 2.7-fold greater than µ of 0.31 day −1 . In the afternoon, k C decreased on average by 50% and reached about 0.5% of the morning values during the night time (Figures 2E-G and Table 2). Cells in the NO 3 culture always assimilated newly fixed C into polysaccharide granules (Figures 3E-G), whereas in the N 2 culture the highest 13 C enrichment was observed in polysaccharide granules during the day but in cyanophycin inclusions during the night (compare Figures 3A-C).
The daily patterns of N assimilation and intracellular allocation were more complex in the NO 3 culture than in the N 2 culture. Values of k N were highest in the morning (10.3-fold greater than µ) and decreased by about 60% in the afternoon and by 90% during the night (Figures 2E-G and Table 2). In the morning, 15 N was often accumulated in what appear to be newly synthesized cyanophycin inclusions ( Figure 3E and Supplementary Figure 3D, arrows 1) or was added to existing cyanophycin inclusions as manifested by a 15 N-rich "shell" surrounding a 15 N-poor core ( Figure 3E, arrow 2). In many cells, 15 N enrichment was clearly present  Figure 2C) and orange symbols in panel (D) (P6, corresponding to values from the P6 subpopulation shown in Figure 2H). Red data points correspond to cyanophycin inclusions with no significant 15 N enrichment.
in the cell matrix but not in existing cyanophycin inclusions (Supplementary Figure 3D, arrows 3). In the afternoon, 15 N enrichment was distributed relatively homogeneously within the cell matrix, and 15 N enrichment in cyanophycin inclusions was sometimes greater but more often lower than in the cell matrix (Figures 3E,F). In the morning, k C and k N were significantly correlated (R = 0.43, p = 0.045), whereas no significant correlation was observed in the afternoon (R = −0.02, p = 0.90). For both morning and afternoon incubations, k C :k N varied among individual cells from 0.15 to 0.5 (mean = 0.31, SD = 0.08, CV = 26%).
Night-time NO 3 assimilation resulted in 15 N being homogeneously enriched within the cell matrix, but carboxysomes were notably more enriched in 15 N than the cell matrix, and cyanophycin inclusions showed no 15 N enrichment ( Figure 3G). This intracellular N allocation pattern was observed in all cells (Supplementary Figure 3F) despite the large intercellular variability in k N values. In contrast to the N 2 culture, k C and k N in the NO 3 culture were strongly correlated during the night (R = 0.94, p < 10 −4 ; Figure 2G) even though the newly assimilated 13 C and 15 N were allocated into different cell compartments (Figure 3G). Night-time k C :k N values varied only slightly in the NO 3 culture among individual cells (range 0.01-0.03, mean = 0.015, SD = 0.005, CV = 33%).

SC-N 2 Culture
In the majority of Cyanothece 51142 cells grown under diazotrophic conditions in semi-continuous, fed-batch mode ("SC-N 2 culture"), temporal patterns in k C and k N values as well as in C and N allocation were similar to the N 2 culture grown in turbidostat mode. In these cells, the highest k C and k N values were observed during the morning and night, respectively (Figures 2H-J and Table 2). As in the N 2 culture, the highest 13 C and 15 N enrichments were detected in the polysaccharide granules and cyanophycin inclusions, respectively (Figures 3H-J). In the afternoon, there was no measurable N 2 fixation, and k C values decreased by about 80% compared to the morning values ( Figure 2I).
There were, however, two notable differences in the behaviors of the N 2 (turbidostat) and SC-N 2 cultures. Firstly, in the SC-N 2 culture there was a large subpopulation of cells (∼46%) that fixed N 2 during the morning (subpopulation P6 in Figure 2H). In these asynchronous diazotrophic cells, 13 C and 15 N enrichments were concentrated in cyanophycin inclusions (Figure 3H), and k N and k C as well as y C and y N were significantly negatively correlated (R = −0.81, p < 10 −4 , Figure 2H; R = −0.36, p < 10 −5 , Figure 4D, respectively). Additionally, the average k N measured in these morning N 2 -fixing cells was 6.4-fold higher than the average k N during the night time. Consequently, the average k N for the SC-N 2 culture in the morning was about 3.1-fold higher than at night ( Table 2). The second notable difference was that cells from the SC-N 2 culture incubated in the afternoon had some cyanophycin inclusions that were significantly more enriched in 13 C compared to the polysaccharide granules and cell matrix (Figures 3I, 4F).

Intercellular Heterogeneity
In all cultures and incubations, k C and k N were markedly heterogeneous among individual cells (Figure 2). Intercellular heterogeneity in day-time k C values was similar with CV = 16-22% (Figure 5) across the turbidostat cultures (N 2 and NO 3 cultures). Heterogeneity in night-time k N values was also similar between the turbidostat cultures but was about 3-fold greater (CV = 53-57%) than heterogeneity in day-time k C values. Heterogeneity in day-time k N in NO 3 cultures decreased from the morning to the afternoon (CV declined from ∼30 to 15%). Moreover, heterogeneity in k N in the N 2 culture during the night time was higher than in the NO 3 culture during the morning.
The semi-continuous diazotrophic batch culture (SC-N 2 ) showed considerably greater intercellular heterogeneity than the turbidostat cultures (N 2 and NO 3 ), especially for rates of N 2 fixation (Figure 5). In the SC-N 2 culture, the differentiation of cells into subpopulations in the morning was reflected in the high heterogeneity in k C (CV = 85%) and k N (CV = 155%). In the afternoon, heterogeneity in k C in the SC-N 2 culture decreased to a level similar to the N 2 culture (CV = 20%).

Interpreting Isotopic Enrichment Imaging Data
The data yielded from nanoSIMS analyses hold valuable information about metabolic strategies used by cells across time and space. However, knowledge about cells' activities that influence their isotopic composition at the end of an SIP incubation is critical to properly interpret nanoSIMS data. Carbon and nitrogen assimilation in Cyanothece mainly occurred during short and intensive periods either in the few hours after dawn or during the night. These periods of rapid assimilation activities fueled the majority of the cells' C and N needs for growth and were followed by long periods of very low assimilation rates. Our results also show a wide range of Cand N-specific rates of 13 C and 15 N assimilation within and FIGURE 5 | Intercellular variability of C and N assimilation rates in Cyanothece 51142. Shown are the coefficients of variation (CV) of C-specific (A) and N-specific (B) rates rates of 13 C and 15 N assimilation for whole cells, k C and k N , derived from data shown in Figure 2. between whole cells and among inclusions. Here, we critically evaluate the factors that can lead to variations in these measured rates and discuss several important, and to our knowledge previously unrecognized, considerations for using the spatially resolved 13 C and 15 N enrichment data obtained by nanoSIMS to infer rates of substrate assimilation. We first focus on principles of labeling as applied to subcellular structures, such as inclusion bodies, and then discuss these principles in the context of whole cell assimilation rates. A more comprehensive and mathematical analysis of these considerations can be found in Polerecky et al. (In Revision). 13 C or 15 N enrichment depends upon the amount of labeled C or N added to a structure during an incubation relative to the amount of unlabeled C or N present prior to the incubation. Any structure that is newly synthesized during an incubation will have 13 C and 15 N enrichments that match those of the enriched C and N sources. However, C or N that is added to an existing structure during the incubation will cause the average 13 C and 15 N enrichment measured in the structure to be lower than in the C and N sources. The deviation between structure enrichment and source enrichment will decrease with incubation time as a function of the rate of biosynthesis but increase with the initial C and N content of the structure. Consequently, variation in the initial C and N content of structures will lead to apparent differences in 13 C and 15 N enrichments among structures present in an incubation even though the rates of biosynthesis and accumulation of 13 C and 15 N may have been the same. 13 C or 15 N enrichment of a structure will also vary if the C and N used for its biosynthesis are derived from unlabeled sources of C and N, including the turnover of cellular macromolecules, in addition to the labeled sources external to the cell. One way to detect the relative importance of internal macromolecular recycling is to quantify the ratio of C-and N-specific rates of 13 C and 15 N incorporation into a structure (e.g., y C :y N for a cyanophycin granule). Because the C:N ratio of many compounds comprising cell structures is well defined (e.g., cyanophycin has a C:N ratio of two), their biosynthesis will preserve the 13 C: 15 N signature of the C and N sources (except for the minute deviations linked to kinetic isotope fractionation). Consequently, if only the external, labeled pools of C and N are utilized for biosynthesis, the ratio of the C-and N-specific rates of 13 C and 15 N assimilation must be equal to 1. Any departure of this ratio from 1 implies that some of the C or N in the structure originated from a different source (i.e., with a different 13 C: 15 N signature than that of the externally supplied sources).
Similar reasoning is needed when analyzing the isotopic enrichment of whole cells. Average 13 C or 15 N enrichment of a cell depends upon the amount of labeled C or N taken up during an incubation relative to the amount of unlabeled C or N present in the cell prior to the incubation. Consequently, intercellular variability in the amounts of C and N storage compounds will lead to apparent differences in the cellular 13 C and 15 N enrichments among cells even if the rates of 13 C and 15 N assimilation into cells were same. The influence of varying storage compound content can be revealed by analyzing the ratio between the C-and N-specific rates of 13 C and 15 N assimilation into whole cells, k C :k N . For an individual cell, this ratio will be 1 (or very close to 1, if the subtle effects of kinetic isotope fractionation are included) provided the cell is in balanced growth, and the externally supplied labeled pools of C and N were the only sources of C and N assimilated by the cell. Any deviation from 1 indicates that (1) the cell assimilated an additional, unlabeled external source; (2) a storage product was preferentially synthesized over another (e.g., polysaccharides over cyanophycin); or (3) the cell recycled internal, unlabeled stores of C or N.
The foregoing analysis of enrichment sources, biosynthesis, and initial C and N content highlights that k C :k N can reveal the presence, synthesis, or mobilization of intracellular C and N stores. With these ideas in mind, we evaluate the isotope enrichment results for Cyanothece 51142 cultures grown with different N sources across a day-night cycle to understand their C and N assimilation processes and allocation strategies.

Roles of Internal C and N Recycling in Cyanophycin Synthesis
The majority of C used for cyanophycin synthesis at night in cells growing diazotrophically originated from recycling of existing C compounds within the cells. Cyanophycin synthesis involved some 13 C (external C source) assimilation, but the k C :k N and y C :y N values were much lower than 1 (Figures 2C,  4A,B). Thus, the bulk of the CO 2 incorporated into cyanophycin was likely derived from polysaccharide catabolism needed to simultaneously provide energy and reducing power (ATP and NADPH) for N 2 -fixation. Nevertheless, cyanophycin synthesis was detected via their enrichment in 13 C rather than 15 N in about 17% of cells ( Figure 3C). 13 C is assimilated via two CO 2 fixation steps leading to synthesis of the non-ribosomal peptide, cyanophycin, which is comprised of aspartate and arginine . Specifically, CO 2 is incorporated via (i) pyruvate carboxylase or phosphoenolpyruvate carboxykinase yielding oxaloacetate which is transaminated by glutamate to form aspartate and (ii) carbamoyl phosphate synthase together with ornithine transcarbamoylase operating to generate arginine (Zhang et al., 2018).
The differences in y C :y N values between and within N 2fixing cells in the same culture (Figures 3C, 4A,B) were also caused by variations in the relative contributions to cyanophycin synthesis of 15 N assimilated during the incubation and unlabeled N assimilated prior to the incubation. Unlabeled N may originate from efficient recycling of N in polyamines, including degradation of arginine via the recently described AgrE/PutA pathway Lee and Rhee, 2020). We hypothesize that the enrichment patterns observed in the N 2 cultures at night were caused by a variable fraction of cyanophycin-N that originated from active N 2 fixation (P1) or from protein degradation (P2). These findings suggest that diazotrophy demands internal N redistribution at night in all cells (as indicated by the similar 13 C enrichment in cyanophycin inclusions), but distinct subpopulations emerge depending on their rates of N 2 -fixation.
Some cyanophycin inclusions in cells from the SC-N 2 culture showed pronounced 13 C but no 15 N labeling during the afternoon incubation (Figures 3I, 4E,F). We speculate that these cells are part of the asynchronous diazotrophic morning subpopulation that fixed N 2 into cyanophycin (P6, Figure 3H) but that later synthesized cyanophycin using internal (and unlabeled) N sources. This idea is supported by the observation that about half of the population fixed N 2 in the morning and exhibited pronounced 13 C enrichment in cyanophycin granules (Figures 3H,I). These observations suggest that cyanophycin synthesis in Cyanothece can occur throughout the entire light period, with cyanophycin-N derived either from N 2fixation or internal N (e.g., via protein degradation). Our high-resolution imaging shows that cyanophycin effectively collects, stores, and redistributes N to facilitate ongoing protein synthesis and catabolism.

Cyanophycin Biosynthesis Is Prioritized, but N Can Flow Directly to Protein During Growth on NO 3
Whether growing diazotrophically or with NO 3 , Cyanothece prioritized cyanophycin synthesis when the rates of N assimilation were at their highest. When N assimilation rates were lower, most cells growing on NO 3 assimilated N into the cell matrix and carboxysomes, suggesting that N was used directly for protein synthesis without prior storage in cyanophycin. Night-time NO 3 uptake into carboxysomes indicates that this new N was used immediately for the synthesis of RubisCO to maintain its content throughout the diel cycle (Nassoury et al., 2001). How cyanophycin synthesis is prioritized over protein synthesis is not clear, but our data suggest that once the cell has acquired sufficient N storage into cyanophycin, this storage step can be bypassed in favor of direct incorporation into proteins.
Night-time assimilation of N in the NO 3 culture was accompanied by small but detectable assimilation of CO 2 , indicative of pyruvate carboxylase activity during the night. Surprisingly, this new C was directed into polysaccharides ( Figure 3G). Typically, the pyruvate carboxylase reaction is considered important to ensure availability of oxaloacetate for citrate synthase to initiate the TCA cycle that produces NADH and amino acid precursors. However, in NO 3 -grown cells, existing polysaccharides appear to have supplied all of the C for protein synthesis (because no 13 C was incorporated into the carboxysomes coincident with 15 N; see above). These results suggest that the amphibolic nature of the glycolytic/gluconeogenic pathway is directional with respect to the flow of newly fixed C into polysaccharides: newly fixed C flows through gluconeogenesis into polysaccharides simultaneous to glycolytic catabolism of "old C" stored in polysaccharides for use in protein synthesis. Glycolytic and gluconeogenesis pathways are strictly controlled so that they cannot be both highly active at the same time, which would create a futile cycle. The highly sensitive detection of labeled C and N afforded by the stable isotope probing and nanoSIMS technologies combined with TEM allowed us to view these unexpected cell activities that occur at very low levels but that support the careful modulation of C and N storage and re-mobilization in Cyanothece.

Intercellular Heterogeneity in C and N Metabolism
Within-population heterogeneity in 13 C and 15 N enrichments (k C and k N ), such as that observed in N 2 and SC-N 2 treatments, has been reported in previous nanoSIMS-based studies (Foster et al., 2013;Mohr et al., 2013;Masuda et al., 2020), although the causes remain poorly understood (Ackermann, 2015). Intercellular heterogeneity has been attributed to stochastic gene expression or state switching in fast growing bacteria and yeast (Elowitz et al., 2002;Blake et al., 2003;Raj and van Oudenaarden, 2008;Raser and O'Shea, 2013;Sanchez et al., 2013;Damodaran et al., 2015). Cyanothece metabolism is strongly regulated by circadian rhythms, and thus, the intercellular heterogeneity observed in our study is at least partly associated with the regulation of C and N fixation determined by the light period and cell cycle (Caudron and Barral, 2013;Bach and Taucher, 2019).
We find that the metabolism of internal C and N storage compounds is another mechanism contributing to cell-to-cell heterogeneity in isotopic enrichment. Dual-label stable isotope probing combined with sub-cellular resolution imaging enabled us to identify internal recycling of N during cyanophycin synthesis, which led to variation in k N values during night-time N 2 fixation (CV≈55%; Figure 2C) and to the wide ranges of y N (CV≈72%; Figure 4B) and y C :y N (Figures 4A,B) values.
The intercellular heterogeneity in k C values during morning C fixation could be caused by the variable polysaccharide content among cells (see Section "Interpreting Isotopic Enrichment Imaging Data"). The limited cellular volume probed by the nanoSIMS measurement also likely contributes to an apparent population heterogeneity. These alternatives are supported by the variable content and uneven distribution of polysaccharide granules within cells (Figures 1, 3A,B,E-G) and the fact that a large fraction of the variability in k C values (∼50%) was explained by the areal coverage by polysaccharide granules (Supplementary Figure 1B).
Differences in the turnover rates of storage inclusions with different C:N contents (cyanophycin vs. polysaccharides) may underlie the three-fold mismatch between the cell-to-cell variation in k N and k C . Other cyanobacteria and pico-eukaryotes have exhibited similar differences in k N and k C (Berthelot et al., 2019;Masuda et al., 2020). One explanation is that day-time acquisition of C reserves was insufficient to fuel N-fixation and other night-time metabolisms (Dron et al., 2013). However, predawn cells were never completely depleted of polysaccharide granules (data not shown), making it unlikely that C-reserves limited N 2 fixation. The large difference in cellular inclusion content also suggests that their subcellular metabolism influences k N and k C . Cyanophycin comprised ∼3% of cellular N in the N 2 culture, whereas polysaccharides comprised ∼30% of cellular C (Table 1). Moreover, the early night y N values were considerably higher than the morning p C values, suggesting that the turnover rate of cyanophycin is considerably faster than polysaccharide turnover at times of highest N and C assimilation, respectively.
Thus, it appears that during diazotrophy, cells retain large pools of C storage with slow turnover rates and small pools of N storage with high turnover rates to manage their C and N demands. This strategy could result in a greater range of enrichment in cyanophycin inclusions compared to the larger and less dynamic pool of polysaccharides. A high turnover rate of cyanophycin also helps explain why variation in k N values in NO 3 -grown cells was higher in the morning, when the cells assimilated N into cyanophycin, and lower in the afternoon, when the cyanophycin pool was bypassed. Differences in the turnover rates of N-rich proteins might also explain the large variation in k N in NO 3 -grown cells during the night when N assimilation again bypassed cyanophycin.
Differences in the timing of N 2 fixation revealed a surprising amount of within-population cell-to-cell heterogeneity in diazotrophic cultures. While the majority of cells fixed N 2 at night as expected (Mitsui et al., 1986;Gallon, 1992;Tuit et al., 2004;Wilson et al., 2017), subpopulations in both the turbidostat-grown and semicontinuous batch cultures fixed N 2 in the morning. Asynchronous diazotrophy has been suggested to occur when the amount of N 2 fixed at night is insufficient to support growth in the following day (Dron et al., 2013;Rabouille et al., 2014;Rabouille and Claquin, 2016). In Cyanothece 51142, asynchronous diazotrophy coincided with the diel maxima in population-level C fixation (Figure 2), but single-cell analysis revealed that morning N 2 fixation was limited to cells whose C fixation rates, and thus presumably intracellular O 2 concentrations, were low compared to the rest of the population. This behavior may be associated with prolonged deactivation of PSII through the early morning hours (Rabouille and Claquin, 2016). How these activities are regulated is not yet known.
Semi-continuous, fed-batch cultures are exposed to a wider range of nutrient and light concentrations compared to turbidostat cultures. These variations could result in a greater range of cell physiologies within a population. Together with previous reports of N 2 fixation in UCYN occurring during a subjective dark phase under continuous light Pennebaker et al., 2010;Dron et al., 2013), our findings suggest that the timing of N 2 fixation is not only regulated by the circadian rhythm or light/dark cycle but also by the cell's ability to balance N and light energy demands. The greater range of heterogeneity within the SC-N 2 culture compared to the N 2 culture is also consistent with the idea that cell-to-cell metabolic heterogeneity facilitates rapid population adjustment to environmental changes (Ackermann, 2015;Schreiber et al., 2016) such as those present in coastal environments (Rippka, 1988).

CONCLUSION
Dual labeling combined with nanoSIMS imaging enabled a much richer and more complex view of cell activities than previously observed using measurements of bulk activities. Specifically, we observed significant cell-to-cell variation, which we attribute to differences in (1) the degree to which internal storage compounds are used as sources of C and N for cyanophycin synthesis, (2) the turnover rates of different storage pools, (3) the range of environmental conditions experienced by a population over a day-night cycle, and (4) the timing of N 2 fixation. The intercellular heterogeneity potentially reflects adaptive mechanisms that allow Cyanothece to thrive in dynamic environments.
Additional details of C and N metabolism were also elucidated by evaluation of 13 C and 15 N labeling patterns across the day-night cycle. Cyanophycin synthesis is a highly effective N-scavenging pathway that assimilates N from protein degradation as well as external sources (NO 3 or N 2 ). Whether growing on N 2 or NO 3 , cells prioritize cyanophycin synthesis when N assimilation rates are highest. In NO 3 -growing cells, N assimilation switches from cyanophycin synthesis to RubisCO synthesis, suggesting that there is a cyanophycin requirement that, once met, can be bypassed in favor of protein synthesis. In NO 3 -grown cells, night-time CO 2 was assimilated into polysaccharides simultaneous with catabolism of polysaccharides used for protein synthesis, suggesting that one way these cells control C is to maintain a directional flow of new carbon entering the cell: CO 2 → gluconeogenesis → polysaccharides → glycolysis → protein.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author. ILES (Illuminating Lake Ecosystem) project funded through the Leibniz Competition (SAW-2015-IGB-1).

ACKNOWLEDGMENTS
This work results from the 10 th Group for Aquatic Photosynthesis (GAP) workshop held in Třeboň, Czech Republic, in August 2017 and subsequent discussions. With this contribution, we wish to honor the late MG and JK for their collegiality, friendship, and extensive contributions to the field of phytoplankton ecophysiology.