Three model organisms were used in this study: Escherichia coli K-12 MG1655, Synechococcus sp. WH7803, and Prochlorococcus sp. MED4. Axenic cultures were grown in batch mode on specified minimal media and harvested during late log phase growth (Supplementary Table 1). Harvesting in this growth phase was necessary for sensitivity in the subsequent nutrient and enzymatic analysis. The limit of detection for phosphate analyses was 25 nm and 0.005 μmol/hr for APase activity. E. coli growth was monitored by OD₆₀₀, Synechococcus WH7803 by phycoerythrin fluorescence (ex. 498nm, em. 573nm) and Prochlorococcus MED4 by chlorophyll a fluorescence (ex. 430nm, em. 664nm). E. coli was chosen due to its rapid growth and our ability to quickly test methodological approaches before extending them to more environmentally relevant organisms like Prochlorococcus and Synechococcus. All growth measurements were performed in a 96-well plate reader (Tecan Infinite 200 Pro).

Cyanobacterial cultures were grown in P-replete ASW and PRO 99 medium (Lindell et al., 1998; Moore et al., 2007) with orthophosphate as the sole phosphorus source (Supplementary Table 1). E. coli MG1655 was grown in M9 minimal medium with either orthophosphate (0.24 mM) or glycerophosphate (0.24 mM), as the sole phosphorus source. SRP-deplete conditions were achieved through E. coli growth in glycerophosphate-only M9 media to induce APase expression. To induce P-starvation in cyanobacterial cultures, cultures were grown in P-replete medium, harvested by centrifugation, and resuspended in medium with no added P. P starvation was assessed by assaying APase activity in whole cells, as described below. By this assay, P starvation was apparent in Prochlorococcus MED4 within 24–48 h, whereas Synechococcus WH7803 required up to 120 h, consistent with previous reports (Moore et al., 2005).

Phosphorus starvation was confirmed prior to lysis by assaying APase activity (see below). Just before lysis, cells were harvested by centrifugation at 8000 × g for 10 min and washed three times with 50mM Tris (pH 8) to remove residual phosphate from the growth medium. Triplicate cell culture aliquots (2ml; cell density > 10⁹cells/ml) were then lysed by incubation with egg-white lysozyme (Fisher) at a final concentration of 5mg/ml for 1 h at 37°C in 50mM Tris buffer (pH 8.0). Lysates were immediately assayed for APase activity and for soluble reactive phosphorus (SRP) and measured in triplicate. Samples for total dissolved phosphorus were frozen at −20°C until analysis. Cell counts were determined via flow cytometry with SYBR Gold staining. Effective lysis was confirmed by visual inspection using an epifluorescence microscopy. In addition, DNA extracted from cell lysates was quantified using a NanoDrop Spectrophotometer and compared against estimates of intracellular DNA/cell as a test of DNA yield and lysis efficiency > 90%.

Incubations were comprised of pelleted and rinsed cells lysed in a Tris buffer solution as mentioned above. Lysed cells and cellular debris were left suspended in a 2.0 ml microcentrifuge tube at 25°C for several days in the dark and monitored at 24–48 hr intervals. The suspension was centrifuged and the supernatant of this lysate solution was removed for nutrient and enzymatic assays. Some incubations were amended with DOP (final concentration 1mM; compounds detailed below) to compare rates of hydrolysis against unamended lysate incubations. Additional incubations were completed in solutions filtered through a 0.22 μm syringe filter to probe APase activity the particulate and dissolved (<0.2 lysate μm) fractions. A tabular summary of the different incubations, treatments, and the analyses performed is included (Supplementary Table 4).

APase activity was assayed using the fluorogenic substrate 6,8- Difluoro-4-methylumbelliferyl phosphate (DiFMUP) (Duhamel et al., 2010). Hydrolysis of DiFMUP produces the fluorescent product 6,8-Difluoro-7-hydroxy-4-methylcoumarin (DiFMU). DiFMU standards (0–4 μM) were prepared in sterile 50mM Tris pH = 8.0 and their fluorescence measured using a temperature controlled Tecan Infinite 200 Pro plate reader, with excitation at 358 nm and emission at 450 nm. Aliquots of intact cultures, 200 μl (to monitor P starvation during growth), or cell lysates (200 μl unfiltered lysate following lysozyme treatment) were incubated with 10 μM DiFMUP and fluorescence monitored over an assay period of 40 min. The rate of DiFMU production was calculated by the change in fluorescence over time. This concentration of substrate was determined to be saturating by varying DiFMUP concentration (0–100 μM) during enzyme assays and comparing subsequent rates. Thus, enzymatic activities presented here are interpreted as maximum activities.

Purified APase (Sigma) was used to test the enzyme activity and enzymatic lifetime when the enzyme is isolated from cell and cellular debris while being incubated in the presence of degrading proteins like Proteinase K. Purified calf alkaline phosphatase (Sigma Aldrich CAT: P6774) was assayed and contrasted against alkaline phosphatase activity from fresh E. coli lysate, with proteinase K added to both solutions. Proteinase K was used to test the ability to degrade alkaline phosphatase in lysate incubations given its broad specificity to degrade proteins. Alkaline phosphatase activity was measured via [SRP] as a proxy for hydrolysis via addition of glycerophosphate. The activation of proteinase K in solution was facilitated by increasing temperature to 55°C in a Tris solution, described above.

Soluble reactive phosphorus was measured spectrophotometrically using the phosphomolybdate blue method of Murphy and Riley (1962) adapted for this study by matrix matching of standards and samples. Potassium phosphate (KH₂PO4) was used as an inorganic phosphate reagent, following 60°C oven drying for adsorbed water removal and Fisher egg-white lysozyme (CAT: 89833) was used as the lysing agent. Standard and sample solution matrices were the same compositionally, with same final concentration of lysozyme and analyzed in 50mM Tris, and diluted with DI water if necessary, to remain within the linear portion of the [SRP] (0–15 μM) standard curve. The dilution also reduced a spectrophotometric interference at 883nm related to the lysozyme.

Total dissolved phosphorus (TDP) was measured using published methods, where [TDP] = [SRP] + [DOP] (Monaghan and Ruttenberg, 1999). Samples were acid hydrolyzed and subsequently ashed, prior to final measurement as TDP. Standards (5’ Adenosine monophosphate, Na-Glycerophosphate, and sodium-pyrophosphate) and samples were treated with HCl (final concentration 0.1M HCl) and 50(w/v)% Mg(NO₃)₂ in combusted and acid cleaned glassware at 80°C for 24hrs and dried for at least 24hrs at 120°C. Following drying, samples were ashed in a muffle furnace at 500°C for 2hrs. Ashed material was dissolved in 3ml of 0.75M HCl and the pH adjusted to 2.0 with 1.0M HCl prior to final measurement as SRP, as described above. Hydrolysis and recovery experiments with the above-mentioned standards yield > 98% hydrolysis and TDP yield. Kimball glass scintillation vials capped with Teflon-lined caps proved most effective at limiting phosphorus contamination, while maintaining structural integrity, and were re-used following acid washing, DI rinsing, and liner replacement.

Model DOP compounds (1 mM) or pyrophosphate (0.5 mM) were added to pure-culture lysates to track the hydrolysis and breakdown of model compounds via SRP production. Concentration was determined to be saturating via the aforementioned procedure with DiFMUP. Hydrolysis rates of model DOP compounds were compared against measured rates of APase activity to explore whether additional phosphatases might be acting to hydrolyze DOP. Model DOP compounds included sodium glycerophosphate (CAT#: G9422; Fisher Scientific), sodium pyrophosphate decahydrate (CAT#: S390–500; Fisher Scientific), RNA isolated from yeast (CAT#: R5636; Sigma Aldrich), and yeast extracted 5’adenosine monophosphate (CAT: 01930; Sigma Aldrich). All compounds and chemicals were of ACS grade or higher.

Z-tests were completed on each time point of the lysate incubations. The mean of triplicate measurements was compared under the different phosphorus conditions to assess the statistical significance of the nutrient conditions through time. Z-scores were then used in the generation of a p-value as a test of the significance.

Immediately following lysis, P was quantified as total dissolved P (TDP), soluble reactive P (SRP), and dissolved organic P (DOP) (Table 1). The quantity of P released varied dramatically across taxa, with E. coli releasing three-fold more SRP per cell than Synechococcus, and 100-fold more than Prochlorococcus, when grown in P-replete conditions (Table 1). This SRP release scales with the total cellular quota of P, which ranges from nearly 1 fmol/cell for E. coli to about 200 amol/cell for Synechococcus and 32 amol/cell for Prochlorococcus in P-replete conditions (Supplementary Table 2). Similarly, the immediate SRP release in E. coli was 107 pmol/cell under p-replete conditions and 83 pmol/cell under p-deplete conditions.

Cellular P content is known to be reduced in P-limited picocyanobacteria (Bertilsson et al., 2003; Van Mooy et al., 2008). Consistent with this, P limitation caused a dramatic reduction in initial SRP release from Synechococcus WH7803 (Figure 1A and Table 1). The immediate release of SRP from P-replete cultures was 33 amol/cell compared to 10 amol/cell from P-deplete cultures. This reduction in SRP release largely accounted for the lower TDP release from P-deplete Synechococcus cells (40 amol/cell) compared to P-replete cells (68 amol/cell). SRP constituted a smaller fraction of TDP released from P-deplete cells (25%) compared to P-replete cells (49%).

Prochlorococcus MED4 cultures released far lower amounts of TDP upon lysis, consistent with their smaller cell size and lower cellular P quotas (Supplementary Table 2). We measured SRP release of 0.9 and 0.7 amol/cell for P-replete and P-deplete MED4 cultures, respectively (Table 1). This SRP release accounted for nearly all of the TDP released from MED4. TDP analysis on Prochlorococcus MED4 lysates appears more variable than other incubations, likely due to the precision of the assay at nanomolar P concentrations.

Following lysis, regeneration of P continued for several days. TDP increased over time (Figures 1A,C), due to solubilization and enzymatic digestion of particulate cell debris. Likewise, SRP generally increased for several days as well (Figure 1). This accumulation of P as SRP was most pronounced in E. coli. For example, with the P-replete E. coli cultures, the initial post-lysis release of 105 amol SRP/cell increased to 350 amol/cell by day 5 (Figure 1E). Unlike the picocyanobacterial strains, E. coli exhibited similar SRP accumulation trajectories for both P-deplete and P-replete cells over the 96 hr period.

Picocyanobacteria showed lower per-cell increases in SRP and TDP over time, consistent with their lower cellular P quotas. In lysates from P-replete Synechococcus, DOP declined while SRP increased (Figure 1), consistent with a transformation from the former pool to the latter, such that by 96 h 75% of the TDP was SRP and only 25% was DOP. By contrast, DOP continued to make up over 70% of the TDP from P-deplete Synechococcus, suggesting that, unlike in the P-replete case, this DOP was not readily hydrolysable. Continued solubilization of particulate P released additional TDP over the 4 day incubation.

Prochlorococcus release of SRP and TDP was not as statistically distinct when comparing P-replete and P-deplete conditions (p < 0.1). Prochlorococcus incubations exhibited a more irregular pattern of SRP accumulation over time (Figures 1C,D and Table 1). In lysates from P-replete Prochlorococcus, there was a significant increase in TDP (transformed from the particulate phase) over time, mostly in the form of DOP; this DOP was not readily hydrolzyed to SRP, hence the fraction of SRP declined over time (Figures 1C,D). The fraction of cellular P regenerated in dissolved form during the 4 day incubation reached about 15% for P-deplete and 3–5% for P-replete Prochlorococcus cultures.

Enzymatic activity following lysis was sustained over the 96hr period incubation with rates of activity linked to nutrient status. In all three taxa, APase activity, measured via DiFMUP hydrolysis, was significantly higher (p < 0.01) in lysates from P-deplete cells compared to P-replete cells (Figure 2), consistent with the known upregulation of APase under P-limited conditions (Torriani-Gorini et al., 1994; Donald et al., 1997; Moore et al., 2005). This activity persisted for at least 96 h following cell lysis (Figure 2). The continued activity of APase provides an important mechanism to transform the dissolved and particulate nutrient pools.

Alkaline phosphatase activity in lysates, as measured with the substrate DiFMUP, was distinct from rates of DOP hydrolysis and SRP production in these same lysates. Although APase activity was low and near the limit of detection in lysates from P-replete cells of all three taxa (Figure 2), SRP continued to be regenerated (Figure 1). Synechococcus lysates showed six-fold higher APase activity from P-deplete cells than from P-replete cells. Yet SRP did not accumulate significantly over time in lysate incubations of P-deplete cultures; most TDP remained as DOP (Figures 1B, 2B). In contrast, nearly as much SRP was regenerated post-lysis in the P-replete Synechococcus lysate, via hydrolysis of DOP, as had been released during the initial lysis step, despite very low APase activity. This finding of SRP regeneration with limited APase activity in the lysate is further explored below.

We posited freshly lysed cells and recently generated debris might house APase activity given the membrane-bound nature of the enzyme. Consistent with this prediction, APase activity in the dissolved (<0.22 μm) lysate fraction was 10-fold lower than unfiltered E. coli lysate (1.2 ± 0.11 μmol L^–1 hr^–1 vs. 12 ± 0.17 μmol L^–1 hr^–1).

Lysate and purified enzyme showed contrasting results for the fate of enzyme activity when incubated with Proteinase K (Figure 3). Purified alkaline phosphatase activity was diminished immediately following temperature activation of proteinase K at 55°C, while APase activity in E. coli lysate continued uninhibited by proteinase K activation. However, Proteinase K in solution prevented the direct measurement of APase activity via a fluorophore. E. coli lystate unaffected by Proteinase K addition furthers supports our suggestion of the membrane localization of the enzyme.

Model DOP compounds were used as analogs to freshly lysed organic matter. We spiked E. coli cell lysates with model DOP compounds and monitored SRP accumulation over time. Lysates from both P-replete and P-deplete E. coli exhibited the ability to hydrolyze the range of DOP compounds we tested (Figure 4). However, the fraction of the DOP spike hydrolyzed varied across compounds and with cell physiology. Addition of 5’-AMP, for example, led to nearly complete phosphorus release as SRP in lysates from both P-replete and P-deplete cells after 96 h, while only 10–15% of P added as RNA was fully hydrolyzed to SRP (Figure 4). Glycerophosphate (GYP) hydrolysis (measured by an increase in SRP) was nearly 100% in lysates from P-deplete cells but only 50% in lysates from P-replete E. coli cells (Figure 4, top panels); despite the fact that the resultant 1 mM [SRP] was enough to inhibit significantly APase activity (experiments not shown).

The efficiency of SRP regeneration from P-deplete E. coli cells was higher than from P-replete cells. This is consistent with the 10 to 40-fold higher APase activity in P-deplete cultures (Figure 4). In these P-deplete lysate incubations, the rates of GYP hydrolysis were over two orders of magnitude higher than those of DiFMUP for the APase assay. In the P-replete lysate incubations, GYP hydrolysis rates were still high even though APase activity was minor. 5’-AMP incubations similarly showed much more rapid hydrolysis in P-deplete culture lysates, though the hydrolysis of the spike essentially went to completion in both P-deplete and P-replete lysates within 2–3 days. Extensive hydrolysis of GYP (P-deplete) and 5’-AMP (both P-deplete and P-replete) produced APase inhibiting concentrations of SRP. RNA associated P was surprisingly stable, with only about 10% of the spike fully hydrolyzed to liberate phosphate. It is possible that the ribonucleases and auto-hydrolysis of RNA were inhibited in the lysate solution or that polynucleotides were partially stabilized in the lysate solution. Pyrophosphate stability was only tested in P-replete lysate. It rapidly hydrolyzed (<24 hrs) to yield essentially all of its P as SRP.

Similar experiments with picocyanobacterial cell lysates were completed in the presence of GYP. Lysates from P-deplete Prochlorococcus MED4 cells showed a rapid increase in SRP after GYP addition, with a spike of 1mM GYP resulting in over 700 μM SRP after 96 h (Figure 5). Lysates from P-replete Prochlorococcus, on the other hand, showed no corresponding increase in SRP, consistent with their much lower APase activity (Figure 5). APase assays and GYP-spike incubations produced results consistent with one another. Initial APase activity was higher per cell in the P-deplete Prochlorococcus lysate than for Synechococcus and P-replete E. coli (Figure 2), whereas P-replete Prochlorococcus had undetectable APase activity. The P-deplete Prochlorococcus APase activity declined steadily over the 4-day incubations, from an initial rate of roughly 1.3 amol cell⁻¹ hr⁻¹ to 0.7 amol cell⁻¹ hr⁻¹ by the end. Such declines were not observed in Synechococcus nor E. coli. GYP hydrolysis rates were rapid and appeared to accelerate over the course of the incubations of P-deplete Prochlorococcus lysate. Whereas GYP hydrolysis was negligible in lysate from P-replete cultures. Hydrolysis rates of GYP (Figure 5) were within a factor of two of the APase activities (Figure 2) for Prochlorococcus, while E. coli and Synechococcus. showed hydrolysis rates more distinct from GYP rates, five to ten-fold greater than APase activities.

Alkaline phosphatase assays and DOP-spike incubations on the Synechococcus lysate yield more complex results. Synechococcus lysates from P-replete cells showed greater SRP accumulation than lysates from P-deplete cells in the presence of GYP (Figure 5), despite higher APase activity in P-deplete cells. APase activity was eight times higher in P-deplete lysate than P-replete (Figure 2). Yet hydrolysis of the naturally produced DOP that was released and incubated in the unspiked lysate (Figure 1) and hydrolysis of GYP in the GYP-spiked lysate (Figure 5) were much more rapid in the P-replete lysate than P-deplete. Synechococcus APase is expected to hydrolyze GYP, so the roughly order of magnitude slower rate of GYP hydrolysis than DiFMUP hydrolysis in the P-deplete lysate conflicts with our expectation. We surmise that there are other enzymes that play a stronger role in hydrolyzing GYP than APase. This is also indicated by the significant hydrolysis of GYP by P-replete E. coli lysate (Figure 5) in which assayed APase activity was low (Figure 2).

Cellular lysis is an important route for nutrient regeneration in aquatic microbial ecosystems, especially in oligotrophic regions that are heavily reliant on internal cycling of nutrients to sustain productivity. Our cell lysis and lysate incubation experiments show that both direct release of cytosolic SRP and the sustained catalytic activity of phosphohydrolase enzymes, lead to the hydrolysis of DOP compounds to liberate SRP and are important for returning nutrients to bioavailable forms. This was true for cultures of E. coli, Synechococcus WH7803, and Prochlorococcus MED4, though the patterns of P release were different among strains and were sensitive to the nutrient status of the cells prior to lysis.

Soluble reactive phosphorus release from E. coli, Synechococcus WH7803, and Prochlorococcus MED4 following lysis included a rapid burst of presumably cytosolic SRP. There could also be a contribution of SRP from DOP compounds that were very unstable and hydrolyzed on the time scale of the lysis reaction and time zero sampling (roughly an hour-long process). We do not consider this potential flux separately from the pulsed release of cytosolic SRP in the following discussion. Although initial SRP release was greater from P-replete cells, the efficiency of P-release was greater from the lower P-content, P-deplete cells. In both P-deplete and P-replete cultures, the initial burst of SRP release was generally followed by significant sustained release of SRP for 2–4 days (Figure 3), with most of the release happening in the first 2 days.

Published nutrient quota measurements provide useful context to contrast per cell P inventory for strains and regeneration trajectories determined above. Cyanobacterial strains have documented flexibility in cellular stoichiometry (Bertilsson et al., 2003; Krumhardt et al., 2013; Martin et al., 2014), in part owing to the ability of some strains to produce polyphosphates.

Polyphosphate storage is well documented in E. coli and Synechococcus strains and in other surface ocean microbial communities (Rao et al., 1998; Diaz and Ingall, 2010; Martin et al., 2018). Although Prochlorococcus genomes contain genes related to polyphosphate breakdown and/or storage (e.g., ppk, ppx) polyphosphate storage is not as well documented in this cyanobacterium (Heldal et al., 2003; Martiny et al., 2006; Temperton et al., 2011). Previous laboratory Synechococcus WH7803 cultures found P quota to be strain specific with polyphosphate storage resulting in at least 2-fold increases in cellular P-quota (Heldal et al., 2003). Our immediate SRP release from P-replete and P-deplete Synechococcus cultures constitutes 17% and 5% of cellular P, respectively, when compared against documented polyphosphate storing laboratory Synechococcus WH7803 (Scanlan et al., 1993; Supplementary Table 2). The immediate SRP release as a fraction of cellular P increases to 33% and 15% of Synechococcus WH7803 cultures when comparing against P inventories excluding polyphosphates. Normalizations of phosphorus release to other cellular characters, like biovolume, were considered, however, data are limited on field populations (Supplementary Table 3). Compared to lower-limit published P cell quotas, our results indicate roughly 40% of the cellular P in Synechococcus P-deplete cells may be immediately released upon lysis into the TDP pool, compared to approximately 26% of the cellular P in P-replete cells (Supplementary Table 2). Similarly, Prochlorococcus releases > 70% of P as DOP immediately following lysis. Relative to published cellular P quotas for MED4, at most 10% of cellular P was released immediately by lysis into the TDP pool (Bertilsson et al., 2003; Krumhardt et al., 2013; Martiny et al., 2016; Supplementary Table 2). These findings support our hypothesis that DOP is a major phosphorus reservoir that ultimately supplies a significant regenerative flux post-lysis, while the initial pulse of SRP is combination of intracellular P and solubilized intracellular components (i.e., polyphosphate).

Alkaline phosphatase was our enzyme of focus due to its putative link to nutrient stress status, its broad specificity for phosphorus-rich substrates, and our ability to quantify activity with a straightforward fluorophore. APase activity persisted for all three strains over the 4 day duration of the lysate incubations with only slight declines, largely unchanged from the initial activity levels on harvested live cells. Glycerophosphate (GYP), 5’ AMP, and pyrophosphate were readily hydrolyzed even in lysate from P-replete cultures for which APase activity was minimal. RNA hydrolysis was relatively inefficient compared to the other compounds tested. Moreover, APase activity differences in Proteinase K incubations supports our hypothesis that 1) APase in the membrane-bound fraction of the unfiltered lysate was shielded from proteinase K degradation, analogous to debris-bound enzymes in natural aquatic environments; and/or 2) other membrane-specific phosphohydrolases are equally capable of DOP hydrolysis, but shielded from proteinase K as well. The overall set of observations indicates that phosphohydrolase enzymes in addition to APase remained active in the lysates. Most APase activity was linked to cellular particulate debris, and APase activity remained associated almost entirely with cell membranes.

Discrepancy in the GYP-derived rates of hydrolysis and those from DiFMUP suggest that additional phosphatases that do not hydrolyze DiFMUP could be important for releasing SRP in lysates from P-replete cells. In field studies, this implies that APase assays might underestimate phosphohydrolytic capacity and response to nutrient stress. Therefore, other phosphatases remain important in the hydrolysis of DOP compounds with a range of specificities and affinities for substrate (Button et al., 2004). By extension, the enzymes responsible for DiFMUP hydrolysis do not necessarily act proportionately on all DOP compounds present in cell lysates. Clearly, taxa, the compounds strains generate, and their physiological state have an impact on the nutrient release and the ultimate path of nutrient regeneration.