Prochlorococcus Cells Rely on Microbial Interactions Rather than on Chlorotic Resting Stages To Survive Long-Term Nutrient Starvation

The ability of microorganisms to withstand long periods of nutrient starvation is key to their survival and success under highly fluctuating conditions that are common in nature. Therefore, one would expect this trait to be prevalent among organisms in the nutrient-poor open ocean. Here, we show that this is not the case for Prochlorococcus, a globally abundant and ecologically important marine cyanobacterium. Instead, Prochlorococcus relies on co-occurring heterotrophic bacteria to survive extended phases of nutrient and light starvation. Our results highlight the power of microbial interactions to drive major biogeochemical cycles in the ocean and elsewhere with consequences at the global scale.

different organisms, e.g., due to variations in maximum growth rate or the ability to utilize the specific substrate tested. Second, cells might be at different physiological states, e.g., exponentially growing, starved, or dying, and thus exhibiting different levels of metabolic activity (2,3). Third, cells show stochastic fluctuations in their activity due to noise in gene expression or regulatory networks (4). Finally, some organisms respond to environmental stress by producing resting stages or spores. Such cells often exhibit very low (or undetectable) metabolic activity and yet are viable, namely, able to return to an active state and reproduce when environmental conditions return to favorable (5). The presence of such resting stages, together with a fluctuating activity at the single-cell level and the genetic variability found within natural populations, is suggested to promote the survival of the population as a whole (2,6).
Understanding the factors affecting the metabolic activity (vitality) of phytoplankton is of special interest. These microbial primary producers perform about one-half of the photosynthesis on Earth, providing energy through carbon fixation at the base of the aquatic ecosystem (7). As phytoplankton grow, they take up elements such as nitrogen (N) and phosphorus (P) from the environment, potentially leading to low nutrient concentrations that may constrain the growth of both the phytoplankton themselves and co-occurring organisms (8,9). Phytoplankton viability, including their ability to survive under conditions of nutrient stress, has been extensively studied, especially for organisms that produce massive blooms that emerge and decline rapidly (for reviews, see references 10, 11, and 12). For example, some bloom-forming cyanobacteria, such as Aphanizomenon species, produce morphologically distinct spores that show very little photosynthetic activity and yet remain viable in the sediment for long periods of time, providing the inoculum for the next growth season (13). In laboratory cultures of Synechococcus elegantus PCC 7942 and Synechocystis PCC 6803, two unicellular freshwater cyanobacteria, nitrogen starvation results in a programmed process where cells enter a resting stage, enabling them to survive prolonged periods of stress (14,15). As part of this process, cells degrade their photosynthetic apparatus in a controlled manner, resulting in a loss of chlorophyll autofluorescence and culture bleaching (a process termed chlorosis). However, the observation that chlorotic cells are viable resting stages is not universal. Chlorotic cultures of Microcystis aeruginosa PCC 7806 were shown to contain a small population of nonchlorotic cells with high chlorophyll autofluorescence (described throughout this study as "high-fl"). Only these high-fl cells were suggested to revive after the re-addition of a nitrogen source, whereas the low-fl cells are presumably dead (16). Chlorotic cells were also observed in eukaryotic phytoplankton. However, it is not yet clear to what extent such cells remain viable, since it may depend on the specific organism and stress conditions (11,17,18).
Prochlorococcus is a pico-cyanobacterium that is extremely abundant in the oligotrophic oceans, performing an estimated ϳ8.5% of global ocean photosynthesis (19). The carbon fixed by Prochlorococcus, which is estimated to produce up to 75% of the daily photosynthetic carbon in the surface layer of the Pacific subtropical gyre (20), can then be utilized by co-occurring heterotrophic bacteria. Prochlorococcus cells in the oceans exhibit extremely high genetic diversity (21), and some of this diversity has been linked with their ability to grow under conditions of extreme nutrient limitation (22,23). It has therefore been suggested that this genetic diversity enables Prochlorococcus as a group to thrive across a wide variety of oceanic conditions (24). While the physiological and transcriptional responses of multiple Prochlorococcus lineages to short-term nutrient starvation have been extensively studied (22,(25)(26)(27)(28)(29), little is known about their ability to survive more than a few days under such conditions. A study on the response of Prochlorococcus strains to a different type of stress, extended darkness (i.e., C starvation), has shown that these organisms can survive light starvation only for a limited time (30). In these experiments, low-fl cell populations reminiscent of chlorotic cells in other cyanobacteria appeared after the light-starved cultures were reexposed to light (30). Therefore, phenotypic evidence exists that Prochlorococcus can undergo a chlorosis-like process, and yet whether these chlorotic cells are active, and whether they are resting stages that can resume growth when conditions are favorable, is currently unknown. Our experiments were therefore designed to answer the following questions. (i) Do Prochlorococcus respond to long-term nutrient starvation by producing chlorotic cells? (ii) If so, are such cells metabolically active (vital), and are they able to reproduce and grow when stress conditions end (viable)? To address these questions, we used fluorescence-activated cell sorting (FACS) to obtain distinct chlorotic subpopulations from axenic and unialgal laboratory cultures of Prochlorococcus which were preincubated with isotopically labeled tracers for photosynthesis (H 13 CO 3 ) and nutrient uptake ( 15 NH 4 ϩ ), and we visualized their activity using nanoscale secondary ion mass spectrometry (NanoSIMS). This method enabled us to measure photosynthesis and N uptake at single cell resolution by quantifying the change in isotopic ratios (31,32). Our results show that while Prochlorococcus do undergo a chlorosis-like process, with some of the chlorotic cells still photosynthesizing and taking up NH 4 ϩ , the chlorotic cells are unable to regrow and thus do not represent resting stages. Instead, coculture with heterotrophic bacteria enables Prochlorococcus to survive long-term stress even without producing resting stages.

RESULTS AND DISCUSSION
Emergence of chlorotic subpopulations in Prochlorococcus cultures. As Prochlorococcus batch cultures reach stationary stage and start declining in abundance, the green color of the cultures disappears, and subpopulations of cells emerge with lower chlorophyll autofluorescence that can be identified by flow cytometry (Fig. 1A and B). This phenomenon is observed in strains from all major cultured ecotypes, as well as in a marine Synechococcus, strain WH8102 (Fig. 1C). In the high-light adapted strain Prochlorococcus MIT9312, lower chlorophyll populations emerged in batch cultures that reached stationary stage due to N or P starvation imposed by the medium composition. However, the timing of the subpopulation emergence and the forward light scatter and chlorophyll autofluorescence (analyzed by flow cytometry) were different under the two nutrient stresses (see Fig. S1A and B in the supplemental material) (33). Cells with Interactions and Starvation in Prochlorococcus ® lower chlorophyll autofluorescence also appeared in populations of another strain, the low-light-adapted MIT9313, when these cultures were inhibited in a coculture with high cell densities of the heterotrophic bacterium Alteromonas sp. strain HOT1A3 (see Fig. S1C and D) (34). Thus, the emergence of populations of cells with lower chlorophyll autofluorescence under a variety of stress conditions is a pervasive phenomenon across marine pico-cyanobacteria. We focused our experiments aiming to better characterize this phenomenon on Prochlorococcus sp. strain MIT9313, since the response to stress in this strain has been extensively studied (22,26,27,(34)(35)(36). In addition, in this strain, three clearly separate subpopulations can be observed when cultured in Pro99 media, facilitating the sorting and subsequent NanoSIMS analyses (Fig. 1B, referred to here as high-, mid-, and low-fl populations).
Assessing the metabolic activity of sorted chlorotic subpopulations. We next sought to determine whether the high-, mid-, and low-fl populations differ in their vitality, measured here as their photosynthesis and nutrient uptake rates (incorporation of H 13 CO 3 and 15 NH 4 ϩ , respectively). The uptake ratio of labeled versus unlabeled nutrients were then used to calculate the metabolic activity of the sorted cells (Table 1). As shown in Fig. 2 and Table 1, the mean uptake of both H 13 CO 3 and 15 NH 4 ϩ was highest in the high-fl population, followed by the mid-and low-fl populations, with the latter population indistinguishable from the control, i.e., glutaraldehyde-killed cells (66 to 114 cells were measured [ Table 1]; see also Tables S1B and S2). We repeated the entire workflow in an independent experiment, and the results are very similar (see Fig. S2A and B; Table 1, n ϭ 88 to 208 cells). These results are reminiscent of observations in several eukaryotic phytoplankton (18). The mean uptake rates for glutaraldehyde killed cells (control) were 0.06 Ϯ 0.15 fg cell -1 day -1 for C and 0.18 Ϯ 0.02 fg cell -1 day -1 for N and most likely depict the absorption of the label by nonspecific binding or diffusion.
Within each of the populations, cell-cell heterogeneity was observed in both 13 C and 15 N uptake ( Fig. 2; see also Fig. S2A and B). Within all of the populations (including the high-fl), some cells were inactive, and this could not be explained by the limited purity of the FACS-sorting procedure (see Table S1 and Text S1 in the supplemental material). The coefficients of variation in C and N uptake rates were within the range shown for other organisms, or higher (see Table S2) (32,37). Similar levels of heterogeneity (primarily in N uptake) were also seen in cells grown under a 12:12 light/dark cycle, where the Prochlorococcus cell cycle follows a diel rhythm, suggesting that this heterogeneity is not due to different stages of the cell cycle or the diel cycle (38) (see Table S2 and Fig. S2C to F in the supplemental material). Cell-cell heterogeneity was also observed in cells from an exponentially growing, nutrient-replete culture (see  Fig. 2 (see also Fig. S2A and B), with the uptake rates and number of cells noted for each experiment separately. c The results for old MIT9313 cultures under light/dark (L/D) refer to the experiment presented in Fig. S2C to F in the supplemental material. The experiment was sampled after a longer period due to slower growth under a light-dark cycle compared to constant light (Fig. S2C). d V, uptake rate. Means and standard deviations were calculated from the uptake rates determined for single cells in each experiment.  Table S2), suggesting that this heterogeneity is not exclusively limited to ageing or stressed cells. This is in accordance with studies assessing the vitality of Prochlorococcus cells using various dyes, which consistently show that a significant fraction of the cells in laboratory cultures are inactive or potentially dead (39,40).
In addition to differing in their chlorophyll autofluorescence and metabolic activity, the high-, mid-, and low-fl cell populations also differ by their forward and side light scatter properties, which are related to cell size and (in larger cells) morphological complexity (see Fig. S4A and B). In agreement with these observations, cells sorted from the high-fl population and observed by scanning electron microscopy (SEM) were 20 to 30% larger than those from the mid-and low-fl populations (see Fig. S4C to E). Interactions and Starvation in Prochlorococcus ® Evaluating the viability of subpopulations. We next sought to determine whether the low-fl cells are viable resting stages. We tested this indirectly by determining the ability of Prochlorococcus MIT9313 cells, cultured in Pro99 media, to grow upon transfer to new growth media at different times during exponential growth and upon culture decline. As shown in Fig. 3A, only cells from cultures where the high-fl cells were dominant could grow when transferred to new growth media. No growth was observed upon transfer of cells from stationary or declining cultures where no high-fl cells were observed. Intriguingly, the presence of high-fl cells was not enough to ensure culture growth (e.g., day 34 in Fig. 3A). This is consistent with a previous study showing that cells belonging to a different Prochlorococcus strain, MED4, that were incubated for 3 days in the dark were unable to resume growth after return to light despite showing no clear difference in the chlorophyll autofluorescence (30). The probability of growth after transfer did not depend on the number of transferred cells (41), with as many as 2.5 ϫ 10 7 cells/ml failing to grow after transfer during culture decline (cells at ϳ1/10 of this density grew after being transferred during exponential stage). Thus, nonchlorotic cells (defined as being within the range of chlorophyll autofluorescence exhibited by exponentially growing cells) are not necessarily viable.
One problem with performing experiments in Pro99, which is commonly used to culture Prochlorococcus, is that the conditions causing cells to reach stationary stage are not always clear (e.g., (42). We therefore repeated these experiments under conditions where entry into stationary phase is induced by N or P starvation ( Fig. 3B; see also Fig. S5) (43). When entry into stationary stage was induced by N or P starvation, chlorotic cells appeared much faster, and the cultures became nonviable much earlier, i.e., essentially immediately after the cessation of exponential growth. Similar results were obtained with a different strain of Prochlorococcus, MIT9312 (see Fig. S5A and B). However, a marine Synechococcus strain (WH8102) behaved differently, surviving N starvation much longer and being able to regrow in nutrient-replete media long after the culture started declining and when essentially all cells were chlorotic (see Fig. S5D). This is reminiscent of the ability of (presumably axenic) cultures of two freshwater cyanobacteria, Synechococcus elegantus PCC 7942 and Synechocystis PCC 6803, to revive after extended N starvation (14,15).
The inability of axenic Prochlorococcus strains to survive long-term nutrient starvation was surprising, and we therefore hypothesized that their survival would be enhanced by interactions with co-occurring heterotrophic bacteria. Indeed, when cocultured with a heterotrophic bacterium, Alteromonas HOT1A3 (34,44), Prochlorococcus strains representing all major cultured ecotypes were able to regrow after 60 days of N and P stress, whereas all axenic strains failed to do so ( Fig. 4; see also Fig. S6A and B in the supplemental material). Interestingly, strain MIT9313, which was initially inhibited by this Alteromonas strain (Fig. 4A) (34,45), was also able to survive long-term starvation in coculture, suggesting that fundamentally different interactions occur during exponential growth compared to long-term, presumably nutrient-limited growth. These results are consistent with the ability of heterotrophic bacteria to extend the survival time of different Prochlorococcus strains under conditions of constant darkness (albeit for only several days [30]) and with the ability of different heterotrophic bacteria to support the long-term viability of batch cultures of Synechococcus WH7803 (46).
It has previously been suggested that detoxification of waste products may be one mechanism whereby heterotrophic bacteria support the long-term survival of pico-cyanobacteria (46). To test this, we transferred Prochlorococcus cultures from nutrient-replete media into sterile seawater with no added nutrients (see Fig. S6B). Under these conditions, Prochlorococcus reach much lower densities than in laboratory media (ϳ10-fold lower fluorescence, particularly in coculture, compare Fig. S6B and, e.g., Fig. 4), and thus it is unlikely that they are inhibited by their own waste products. Although coculture with Alteromonas HOT1A3 reduced the growth of all Prochlorococcus and Synechococcus strains under these conditions, potentially Interactions and Starvation in Prochlorococcus ® due to competition for scarce nutrients, only the cocultures were able to survive and grow when transferred into fresh media after 40 days (right panel in Fig. S6B). We therefore hypothesize that the ability of the heterotrophs to support Prochlorococcus survival in extended coculture is due to nutrient remineralization rather than to the detoxification of potential waste products, although a potential role for detoxification cannot be ruled out.

Stress survival in pico-cyanobacteria: why is Prochlorococcus different?
In the present study, we demonstrate that phenotypic heterogeneity between clonal Prochlorococcus cells occurs at multiple "scales." In exponentially growing axenic laboratory cultures of two strains, MIT9313 and MED4, C and N uptake rates differ significantly    Table S2 in the supplemental material). This variation is independent of genetic variability. In addition, as axenic cultures become stressed, a larger phenotypic change occurs as cells lose their chlorophyll autofluorescence and become chlorotic. Under these experimental conditions, most cells are inactive (primarily in the low-fl population, as measured in strain MIT9313), although we cannot rule out that even low-fl cells still retain a residual level of activity not detectable by the NanoSIMS. Nevertheless, some cells from the chlorotic populations retain at least part of their photosynthetic capacity and indeed can fix carbon and take up NH 4 . Yet, in our experiments, they do not regrow when conditions become more favorable. In Synechococcus elegantus PCC 7942, chlorotic cultures retain approximately 0.01% of their photosynthetic activity, as well as a residual level of protein translation, although it remains unclear whether this is a process shared by all cells in the culture or whether this activity is only due to a small subset of more active cells (14). The clear difference between the ability of axenic Synechococcus elegantus PCC 7942 and Synechocystis PCC 6803 to survive long-term N starvation, as well as the inability of axenic Prochlorococcus cultures to do so, suggests an inherent difference in the physiology and genomic functional capacity between these unicellular cyanobacteria.
Entry into chlorosis in Synechocystis is a regulated process that involves the organized degradation of the phycobilisomes in parallel with an increase in the storage products glycogen and polyhydroxybutyrate (PHB) (15). The photosynthesis apparatus of Prochlorococcus is different from that of other cyanobacteria, using unique chlorophyll a 2 /b 2 binding proteins rather than phycobilisomes (47). Indeed, Prochlorococcus lack orthologs of the nblA gene required for phycobilisome degradation during chlorosis (15). Synechococcus WH8102, which was able to survive N starvation much longer in axenic culture than the tested Prochlorococcus strains (Fig. S5D), has a divergent nblA-like gene (48). In addition, while Prochlorococcus likely use glycogen as a C storage pool (49), they lack the phaA, phaB, phaC, and phaE genes required for PHB biosynthesis which are induced in Synechocystis PCC 6803 under chlorosis (although these genes are not required for revival from chlorosis [15]). Taken together, these differences suggest that Prochlorococcus lack the genetic toolkit employed by Synechocystis PCC 6803 and Synechococcus elegantus PCC 7942 to enter into a resting stage. Thus, chlorotic cells in Prochlorococcus are not resting stages.
If Prochlorococcus are indeed incapable of producing resting stages in response to nutrient or light starvation, what are the evolutionary drivers of this phenotype, and what are the consequences for the dynamics of Prochlorococcus populations in the ocean? While the open oligotrophic ocean is often considered a relatively stable environment, nutrient concentrations do fluctuate (8), and phytoplankton (including Prochlorococcus) inhabiting these waters show multiple signs of nutrient stress (50). Many of the microbes that live in such environments comprising a large fraction of the surface ocean have small, highly streamlined genomes (51), and this has been suggested to be an adaptation to low nutrient concentrations (24,51,52). Therefore, it is possible that the lack of resting stages is a result of this genome streamlining: the genomes of Synechococcus elegantus PCC 7942 and Synechocystis PCC 6803 are ϳ3.2 and ϳ4 Mbp with their plasmids, respectively, compared to ϳ1.4 to 2.5 Mbp for Prochlorococcus strains.
Surviving nutrient stress "with a little help from my friends." The ability of Prochlorococcus to thrive under conditions of extreme nutrient limitation is often explained by their small cell size (increasing their biomass-specific diffusion), their generally low nutrient requirements, and their specific metabolic strategies to minimize the per-cell elemental quotas (53)(54)(55). However, these mechanisms appear not to work in axenic laboratory cultures, and thus we propose that interactions with co-occurring microorganisms enable Prochlorococcus to survive when nutrient-saving mechanisms are not sufficient, as suggested for its close relative Synechococcus (46). This may take the form of recycling of inorganic nutrients by the heterotrophic bacteria, as well as possibly by the production of organic compounds that contain elements such as N or P. Indeed, Prochlorococcus can compete with heterotrophic bacteria for amino acids (56,57). Importantly, the utilization of organic compounds (mixotrophy) may provide Prochlorococcus also with carbon, sulfur, or energy sources and may potentially help them survive also light starvation (30,(58)(59)(60)(61)(62).
Regardless of the specific forms of dissolved organic matter being utilized by the cells, and on the exact mechanism enabling the cells to survive long-term nutrient starvation in coculture, the lack of any observed mechanism for the production of resting stages by Prochlorococcus may be considered another manifestation of the "black queen hypothesis." This hypothesis states that microorganisms "outsource" essential survival mechanisms such as detoxification of reactive oxygen species to the surrounding microbial community (63). These forms of microbial interactions likely affect the distribution and activity of Prochlorococcus on a global scale. The increased survival of Prochlorococcus under harsh conditions, supported by its associated heterotrophic bacteria, may enable it to remain active at the single cell level even during long periods of unfavorable conditions (64,65). Thus, the tight interactions between Prochlorococcus and its bacterial "supporters" likely affects photosynthesis and carbon cycling at the base of the aquatic food web, with potentially profound implications for overall oceanic productivity and carbon cycling.

MATERIALS AND METHODS
Prochlorococcus growth and stable isotope incubations. Axenic Prochlorococcus strains were grown in Pro99 media under constant cold while light (27 mol photons m Ϫ2 s Ϫ1 ) at 22°C. Axenicity of the strains was tested routinely, as well as before all major experiments, using test media (ProMM [41]). In addition, no evidence for contaminating heterotrophic cells was observed using flow cytometry or SEM or when axenic Prochlorococcus cultures were used as negative controls for 16S amplicon sequencing. Bulk chlorophyll fluorescence (excitation, 440 nm; emission, 680 nm) was measured almost daily using a fluorescence spectrophotometer (Cary Eclipse; Varian). In parallel, samples for flow cytometry were taken for cell numbers. When three distinct subpopulations appeared in the flow cytometry, the cultures were labeled with 1 mM Sodium bicarbonate-13 C and 1 mM ammonium-15 N chloride (Sigma-Aldrich) for 18 to 24 h. The optimal incubation time is based on preliminary isotope labeling experiments with Prochlorococcus MED4, showing that uptake is identified already after 3 h and is linear until 24 h under our growth conditions (see Fig. S7). Incubations were stopped by fixing 2 ml of the culture with 2ϫ EM-grade glutaraldehyde (2.5% final concentration; Sigma) and subsequent storing at 4°C until the sorting analysis. Nonlabeled cells that were killed before labeling (by adding 2.5% glutaraldehyde) were used as a negative control.
Cell sorting and filtration. Sorting of subpopulation was carried out using a BD FACSAria III sorter (BD Biosciences) at the Life Sciences and Engineering Infrastructure Center, Technion, Israel. Each sample was sorted for three subpopulations: nonchlorotic (high-fl), semichlorotic (mid-fl), and chlorotic (low-fl) ( Fig. 2A). The sorting gates for each subpopulation were determined from the population observed in forward scatter (FSC; a proxy for cell size) and autofluorescence (PerCP, chlorophyll autofluorescence). After sorting, the sorted subpopulation was gently filtered on 13-mm-diameter polycarbonate filters (GTTP, 0.2-m pore size; Millipore, MA), washed twice with sterile seawater, and air dried. The filters were cut into two parts. One half was stored at 4°C until NanoSIMS analyses, and the other half is used for SEM.
NanoSIMS and data analysis. The samples were coated with a layer of ϳ30-nm gold with a Cressington 108 auto sputter coater (Watford, United Kingdom). Random spots were used for NanoSIMS analyses. SIMS imaging was performed using a NanoSIMS 50L instrument (Cameca, Paris, France) at the Leibniz-Institute for Baltic Sea Research Warnemünde (IOW). A 133 Cs ϩ primary ion beam was used to erode and ionize atoms of the sample. Images of secondary electrons-12 C -, 13 C -, 12 C 14 N, and 12 C 15 N -were recorded simultaneously using mass detectors equipped with electron multipliers (Hamamatsu). The mass resolving power was adjusted to be sufficient to suppress interferences at all masses allowing, e.g., the separation of 13 Cfrom interfering ions such as 12 C 1 H -. Prior to the analysis, sample areas of 50 ϫ 50 m were sputtered for 2 min with 600 pA to erode the gold, clean the surface, and reach the steady state of secondary ion formation. The primary ion beam current during the analysis was 1 pA; the scanning parameters were 512 ϫ 512 pixels for areas of 30 ϫ 30 to 48 ϫ 48 m, with a dwell time of 250 s per pixel. A total of 60 planes were analyzed.
Analyses of NanoSIMS measurements. All NanoSIMS measurements were analyzed with the Matlab based program look@nanosims (66). Briefly, the 60 measured planes were checked for inconsistencies and all usable planes accumulated, regions of interest (ROIs) (i.e., Prochlorococcus cells and filter regions without organic material for background measurements) defined based on 12 C 14 N mass pictures, and both 13 C/ 12 C and 12 C 15 N/ 12 C 14 N ratios calculated from the ion signals for each ROI.
Uptake rate calculation. Uptake rate was estimated using the following equation, based on that of Legendre and Gosselin (67), as follows: Roth-Rosenberg et al.

®
July/August 2020 Volume 11 Issue 4 e01846-20 mbio.asm.org 10 Where %P t * is the concentration (atom %) of the heavy isotope in the particulate matter at the end of the incubation, %D i * is the concentration of the dissolved tracer added to the incubation (and assumed not to change over the short incubation time), and %P 0 * and %D 0 * are the natural heavy isotope concentrations in the particulate and dissolved matter, respectively. We estimated Q, the cell quota (in fg cell Ϫ1 ) of C or N, based on measurements of the biomass of MED4 and MIT9313 (66 and 158 fg cell Ϫ1 , respectively [68]) and assuming that C comprises 50% and N comprises 7.5% of the cell biomass. For heavy isotopes concentration in the particulate and dissolved phases before incubation we used the natural values for isotopic ratios of 13 C and 15 N (1.12 and 0.37%, respectively). For the experiment shown in Fig. S7 in the supplemental material, we measured the NH 4 ϩ concentration in the media and added the 15 N tracer to 50% final concentration. Since all other experiments were performed in declining cultures, we assumed that the NH 4 ϩ was depleted from the media, and thus %D t * was defined as 90%, based on previous measurements of NH 4 ϩ concentrations in old cultures. We used a value of 50% for the initial percentage of 13 C, based on dissolved inorganic carbon (DIC) measurements (43). For the terminal concentrations of 15 N and 13 C in the particulate phase (%P t * ), we used the values of 13 C/ 12 C and 15 N/ 14 N that were obtained from the NanoSIMS analysis of the cells. 13 C/ 12 C and 15 N/ 14 N ratios below the natural values resulted in negative uptake values and were treated as zero uptake.
Means and standard deviations (SD) of C and N uptake rates were calculated from the uptake rate values of individual cells ( Table 1). The uptake rate values were not corrected for negative control (killed cells), which are presented for comparison in Table 1. Since 13 C/ 12 C and 15 N/ 14 N values of individual cells were not normally distributed, for significance analysis we used nonparametric tests (Mann-Whitney and Kruskal-Wallis tests) performed using the real statistics resource pack software (release 5.4).
Scanning electron microscopy. Immediately after filtration, filters dehydrated in an ethanol series of 30, 50, 70, 80, 90, and 100% (vol/vol) ethanol (dilutions were in deionized water) for 10 min each. Samples were then dried, mounted on stubs with carbon tape, and coated with 5-nm gold. Cells were obtained on a Zeiss Sigma SEM by using a SE2 detector (2 to 2.5 kV, WD ϭ 8 mm).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.02 MB.