We determined that HPG labeling did not occur in the presence of chloramphenicol, and that the presence of methionine at 1 or 2 μM reduced or eliminated HPG labeling percentages in four separate tests of natural assemblages (Figure S3).

An experiment that tested the competition of HPG with ³⁵S-methionine incorporation suggested the presence of HPG (8 or 18 nM) did not significantly change ³⁵S-methionine incorporation while the presence of cold-methionine (8 or 18 nM) decreased ³⁵S-methionine incorporation by 2.3- and 3.7-fold, respectively (p < 0.05; Figure S4). In a second experiment, presence of HPG at 20 nM resulted in an insignificant reduction of ³⁵S-methionine incorporation (1.3-fold; p = 0.06), but 200 nM and 2 μM HPG elicited significant reductions (2.8-fold; p < 0.05 and 12.5-fold; p < 0.05, respectively; Figure S4).

Additional experiments examined the potential for HPG to competitively inhibit methionine incorporation. Enzyme kinetic curve fit calculations indicated that K_m for ³⁵S-methionine decreased with higher HPG concentrations but there was no effect on V_max. The calculated inhibition constant (K_i) for HPG was 373 nM (numerically equal to K_m for HPG incorporation) while the K_m for methionine incorporation was an order of magnitude lower at 31 nM (Figure 2). The results support the conclusion that HPG competes with methionine for incorporation into cells within an enzyme-substrate binding framework (Copeland, 2005).

We calculated a conversion factor using 3 different approaches and adopted the “top 10% isolate mean” calculation (see Materials and Methods). The individual cell HPG signal intensities of the Alteromonas spp. AltSIO and As1 used for the conversion factors ranged 1.795–5.617 log₁₀ RFU cell⁻¹ (n = 1697). The range in protein for the top 10% cells was 18.2–136.1 fg cell⁻¹ and the resulting conversion factor was 5.358 log₁₀ RFU fg⁻¹ protein; this value was used to convert a single cell HPG signal to a single cell protein production rate (scPP) for all labeled cells in natural communities. The detection limit of the method was determined to be 0.07 fg cell⁻¹ (SE = 6.1 × 10⁻⁴, n = 7136) and was calculated by taking the average value of all control cells from all samples after applying the conversion factor.

Protein mass determined from all cell volumes ranged 8.1–136.1 fg cell⁻¹. The two remaining calculations, while not adopted, provided the following conversion values: (i) “isolates regression,” 5.620 log₁₀ RFU fg⁻¹ protein (R² = 0.41, p > 0.0001) and (ii) “community sum ³⁵S-methionine,” 5.166 log₁₀ RFU fg⁻¹ protein. The “isolates regression” calculation assumed all labeled cells, including low intensity cells, harbored protein where most methionine had been replaced by HPG. This regression fit calculation was sensitive to disproportionate influence by low intensity cells. The “community sum ³⁵S-methionine” factor assumed: (i) the sum crystal area for the ~900 cells we measured could be normalized to all cells per mL, allowing us to correlate crystal area with bulk growth rate and (ii) the bulk HPG community sum intensity could be accurately divided into the bulk ³⁵S-methionine protein synthesis rate.

HPG labeling across 21 seawater samples (i.e., communities) ranged 14.6–100% (Table 2). HPG signal intensities in labeled cells ranged six orders of magnitude (−0.155 to 6.949 log₁₀ RFU cell⁻¹; raw data shown in Figure S5 and Table S2). The labeled cells were rank-ordered by intensity within each community, and the sum HPG signal intensity for all cells in each community was calculated. The percentage of rank-ordered cells that comprised 10% of this “community sum” ranged 0.2–1.6% (Table 3). The percentage of “community sum” comprised by the top 10% of active cells ranged 34.7–77.8% while the bottom 50% of active cells ranged 1.9–13.7% (Table 3). Upon application of the conversion factor to each cell's HPG signal intensity, scPPs ranged 3.1 × 10⁻⁶ to 3.9 fg protein cell⁻¹ h⁻¹ (n = 13645; Table 3 and Figure S6).

For all labeled natural assemblage cells in the 21 Scripps Pier samples (n = 13645), a model II linear regression and correlation analysis of cell volume and log-transformed HPG signal intensity were significantly non-zero and correlated, respectively (p < 0.0001; Spearman R = 0.306; Figure 3). There was also a significantly non-zero model II linear regression and correlation between cell volume and size-normalized cell signal intensity (p < 0.008; Spearman R = 0.063; data not shown).

The scPPs within each community exhibited non-Gaussian distributions, likely caused by a high number of cells with low rates. Neither log-normal nor exponential regressions of the frequency distributions were significant for any of the 21 communities. However, the distributions fit well to power law regressions (Figure 4) with R² > 0.99 in 9 of the 21 seawater samples while the remaining had R² > 0.75. The power law fits displayed similar visual trends in that all had negative slopes and were relatively confined to the same x-y space. However, the scaling factors (variable A) and exponents (variable B) were wide ranging (Figure 4 inset). A consensus regression model fit to the global dataset did not account for all of the variation (extra sum of squares F test, p < 0.0001; Akaike's informative criteria, 99.99% probability). Notably, the innermost and outermost regressions (15-Dec-2010 and 9-Dec-2011) originated from water samples that differed greatly with regard to sampling depth and chlorophyll concentration.

When we applied the standard concentration and incubation protocol to seawater particles (20 nM HPG for 1 h), 80% of DAPI-stained attached bacteria were labeled, with a median scPP of 0.025 fg protein cell⁻¹ h⁻¹ (Figure 5; example images in Figure S7). Extended incubation times (up to 2 d) and higher HPG concentration (2 μM) increased labeling characteristics for particle-associated cells. Median signal intensity in 20 nM HPG increased by 2.8-fold from 1 h to 2 d then remained approximately constant through 8 d (Figure 5). In contrast, median intensity in 2 μM HPG increased by 4.8-fold from 1 h to 2 d, then another 6.6-fold from 2–8 d. Signal intensities for 20 nM HPG treatments were lower than 2 μM HPG treatments at each time point (Figure 5). The same pattern occurred for free-living cells in the same vessels (data not shown). Signal intensities for particle-attached cells followed non-Gaussian distributions (D'Agostino and Pearson omnibus and Shapiro-Wilk normality tests, p < 0.001). Methodologically, omission of filter-transfer-freeze for 8 μm filters had little effect on the particle attached cell measurements because high background fluorescence, as occurs on 0.2 μm filters, was due to small, closely-spaced pores. Background intensities measured from circular regions (1 μm diameter) were only slightly higher on 8 μm polycarbonate filters relative to cover glass following filter-transfer-freeze (1.25-fold in 20 nM samples; 1.09-fold in 2 μM samples).

We hypothesized that limited variation in HPG labeling intensity would characterize marine bacterial monocultures. To address this, isolates AltSIO and As1 were grown in separate experiments to examine patterns and distribution of HPG incorporation in clonal cells. Labeling percentages were 100% for both isolates as determined from aliquots tested each hour from 16 to 22 h. Single-cell signal intensities ranged approximately three orders of magnitude at each timepoint, with the brightest cells of AltSIO and As1, respectively, at 6.176 and 6.279 log₁₀ RFU (Figure S8). By comparison, the most fluorescent single cell measured among all Scripps Pier samples was 5.949 log₁₀ RFU. Population signal intensities for both isolates followed non-Gaussian distributions at all timepoints except T = 18 h (D'Agostino and Pearson omnibus and Shapiro-Wilk normality tests, 0.03 < p < 0.0001).

HPG is specifically incorporated into de novo synthesized protein. Incorporation of HPG into new protein has been previously demonstrated for E. coli (Beatty et al., 2005; Wang et al., 2008). We used chloramphenicol inhibition of protein synthesis to test whether this held true for natural assemblages of marine bacteria. Chloroamphenicol blocks peptidyl-transferase and prevents protein chain elongation. The presence of chloramphenicol eliminated HPG labeling, indicating that HPG is very likely incorporated into newly synthesized protein in natural bacteria. We did not directly address the interaction of HPG and/or methionine at the level of membrane transport and translation, nor did we directly test whether HPG and/or methionine was converted to other compounds or brought into intracellular inclusion bodies rather than become incorporated into proteins.

HPG is a competitive inhibitor of methionine. Methionine addition reduced or eliminated HPG percent labeling in a dose-dependent manner, supporting the basic premise that HPG can be used as a substitute for methionine in protein (Figure S3). Given the non-zero, albeit significantly reduced, HPG labeling percentages in the presence of methionine on 1-Oct-2009, the experiment was repeated three times. The 16-May-2010 and 6-Oct-2010 samples showed no labeling, as expected, while 11-Oct-2010 maintained a low amount of labeling (Figure S3). We speculate that community acclimation to high exogenous methionine concentrations within the particular water masses resulted in preferential uptake of HPG. These results may also be due to naturally-occurring alkynes in the cells of these samples (Udwary et al., 2007). While such cells should have appeared in many samples, it may be that there was a higher abundance of such cells on 1-Oct-2009 and 11-Oct-2010. This result clearly defies expectation and is without a well-defined explanation.

Bulk methionine incorporation was reduced by the presence of HPG (Figure S4). Michaelis-Menten kinetics of HPG-methionine interaction showed competitive inhibition at 1000 nM, but overlap of the 95% confidence intervals for the non-linear uptake kinetic fits in the presence of 0, 30, and 100 nM HPG suggested insignificant competitive inhibition at these concentrations despite exhibiting similar V_max rates alongside increasing K_m concentrations with increasing inhibitor (i.e., HPG; Figure 2). This may partially be explained by: (i) the bacterial community exhibiting ~10-fold lower affinity for incorporation of HPG based on its high K_i and (ii) metabolic variation in the complex assemblage of cells within the triplicate measurements. Additionally, this affinity may explain why 8 or 18 nM HPG did not significantly reduce ³⁵S-methionine incorporation while as little as 10 nM cold methionine was sufficient (Figure S4). Nevertheless, since typical methionine concentrations in surface seawater are ~200 pM (Zubkov et al., 2004), availability of 20 nM HPG should overcome competitive inhibition by methionine.

HPG percent labeling is comparable to ³⁵S-methionine microautoradiography. For parallel seawater incubations, the HPG method (without FTF) and ³⁵S-methionine microautoradiography yielded similar percent labeling (Table 1). Thus, the HPG method performs at least as well as microradiography for quantifying proportions of protein-synthesizing bacteria in natural assemblages. We attempted to directly compare the two methods by co-incubating seawater with ³⁵S-methionine and HPG, transferring co-labeled cells to photographic emulsion on a single slide, then sequentially processing the slide for microautoradiography or click. We found that AlexaFluor-488 azide bound non-specifically to the microradiography emulsion, rendering HPG-labeled cells indistinguishable from background. Application of the click reaction on the filters prior to microautoradiography processing was also unsuccessful as no HPG-labeled cells were observed, possibly due to fluorescence fading during 3 day exposure at 4°C.

Short incubations and high sensitivity. These prerequisites are desirable in order to limit time-dependent changes in protein synthesis rate. We adopted 1 h incubation time as standard protocol because the results of a time course experiment demonstrated that percent cell labeling increased 1.6-fold as incubation time increased from 30 min to 1 h, with no further increase up to 4 h incubation (data not shown). The 1 h incubation time, combined with use of the FTF technique, permitted detection of very low individual cell HPG fluorescence intensities. The scPP estimates were calculated by first acquiring accurate and precise measurements of fluorescence intensities and bacterial cell dimensions using the Nikon NIS Elements software. The HPG intensities were divided by the conversion factor to yield scPP. Using this individualized measurement approach, the lowest detected scPP was 3.1 × 10⁻⁷ fg protein cell⁻¹ h⁻¹.

From data acquired in the particle experiment, we found that increasing HPG concentration from 20 nM to 2 μM yielded mean incorporation rates that were 1.5-fold higher while median incorporation rates increased 3.8-fold. As a standard protocol we chose to use 20 nM HPG to yield conservative estimates of the incorporation rate; even at this conservative HPG concentration bacterial communities on 3 out of 21 sampling dates exhibited 100% labeling.

When we applied the conversion factor to natural assemblages in 21 samples, the resulting scPP ranged ~10⁻⁶ to 3.9 fg protein cell⁻¹ h⁻¹. This broad range is in agreement with Sintes and Herndl (2006) in which values ranged 1–100 amol Leu cell⁻¹ d⁻¹, or equivalent to 0.15 and 15.1 fg protein cell⁻¹ h⁻¹, respectively, assuming 3.6 kg protein per mole of leucine incorporated (Simon and Azam, 1989). Comparatively, our minimum rate detected with HPG is very low. The FTF technique enabled a high signal-to-background ratio for HPG-labeled cells and extended the lower threshold for detecting active cells. This is one potential reason why percent labeling was higher for HPG than for microradiography in parallel samples (Table 1). It may also be influenced by the dynamic range obtained from pixels vs. exposed silver grains. Using fluorescence, the collection of photons from a single cell occurs over hundreds of pixels comprising its dimensions, and not based on the tens of silver grains in its vicinity.

Several different methods for quantifying cell activity have shown that a majority of cells are inactive or “ghosts” or dormant in seawater (Fuhrman and Azam, 1982; Zweifel and Hagstrom, 1995; Lebaron et al., 1998; Schumann et al., 2003; Hamasaki et al., 2004). By contrast, the majority of cells were actively synthesizing protein in all samples we tested, albeit many cells were doing so at a low rate. The large proportion of low activity cells points to the existence of a background microbial community with low metabolic activity potentially primed to respond advantageously. It is known that bacterial communities harbor multiphasic uptake systems (Azam and Hodson, 1981). Low activity cells with high K_m permeases may maintain a minimal metabolic state until favorable conditions arise (Roszak and Colwell, 1987; Lauro et al., 2009). These low-activity populations may thus represent a substantial portion of the “microbial seed bank” suggested to exist throughout the ocean (Gibbons et al., 2013).

A positive linear relationship was observed between cell volume and HPG incorporation (i.e., log₁₀ RFU intensity), consistent with previous studies of marine bacteria (Gasol et al., 1995; Lebaron et al., 2002). Larger volume cells tended to have higher protein synthesis rates (Figure 3). These relationships held when incorporation rates were normalized by volume, i.e., larger cells exhibited higher incorporation even when each cell's rate was divided by its volume. This suggests that carbon and nutrient incorporation is not similar for all cells per cell mass, showing that cell size is a significant parameter affecting anabolic processes derived from ambient material. This lends further support that size is an important factor governing microbially-mediated nutrient dynamics, along with substrate degradation specificity and growth efficiency (el Giorgio and Cole, 1998).

The strong skew in fluorescence intensities within each sample led us to use log-log plots for viewing and comparing the spectrum of scPP across samples after fitting the data distributions to power law regressions. The visual clustering of regressions (and the global fit; Figure 4) indicates there is stability in the distribution of growth rates across individuals in the communities at Scripps Pier as well as in the offshore ecosystem of the San Diego trough. The consensus fit, however, could not explain all of the variation seen in the data, indicating each sampled water mass harbored microbial communities with unique continuums of single cell protein synthesis profiles that: (i) exerted differential influence over elemental cycling and (ii) reflected fluctuating nutrient availability supporting individual cell growth. Indeed, the two most divergent regression lines (15-Dec-2010 and 21-Dec-2011; Figure 4, bold lines) reflected the two most divergent sample types, specifically offshore waters at 60 m (overall lower community activity) and an intense bloom of the dinoflagellate Lingulodinium polydrum (overall higher community activity). Phytoplankton-derived organic carbon availability is among the factors that may explain these two divergent regressions (Mague et al., 1980; Fuhrman et al., 1985; Nelson and Carlson, 2012; Sarmento and Gasol, 2012).

The power law scaling values are useful to consider alongside means and medians when describing non-Gaussian distributions of single cell microbial activity. These values define the allocation of activity within the community (Figure S9). Increasing the scaling factor A (analogous to the y-intercept) results in an overall increase in community protein production rate. Changing the scaling exponent B (analogous to the slope) affects the relative proportion of higher or lower activity cells while also controlling the distribution of community activity. We found no significant relationship between chlorophyll and scPP mean or median, scaling factor, or scaling exponent. However, we did find significant positive correlations between: (i) the scPP means and scaling factors (r = 0.61, p = 0.003) or scaling exponents (r = 0.65, p = 0.002) and (ii) scPP medians and scaling factors (r = 0.93, p < 0.0001) or scaling exponents (r = 0.95, p < 0.0001).

The power law distributions are in general agreement with previous work examining single-cell incorporation of various substrates in the Arctic Ocean using microautoradiography (Nikrad et al., 2012). In one of the few examples of this distribution, the authors noted a negative slope of percent active cells vs. silver grain area (data were plotted on log-linear axes), similar to the decreasing trend between percent cells and the activity bins shown in our HPG study using log-log axes. One implication is that “cell-specific rate,” i.e., bulk community rate divided by the total abundance of cells, does not accurately reflect activity distribution at the single cell level. This calculation assumes Gaussian rate distribution within the natural community (Smith and del Giorgio, 2003; and references therein) but this assumption is clearly not valid for HPG-based protein synthesis rates given their fit to a power law regression.

We applied the method to particles because these microenvironments are ubiquitous in coastal seawater and may regulate bacterial community activity at larger scales. The standard protocol for free-living bacteria (20 nM HPG, 1 h incubation; Figure 1) also worked for particle-attached bacteria, and additional treatments in the particle incubations gave insight into the nature of protein production on particles. The experiment included parallel treatments of 2 μM HPG since we recognized a priori that high methionine concentrations could originate from particles, e.g., due to bacteria-mediated proteolysis, which could reduce HPG incorporation and thereby underestimate growth rates for bacteria closely-associated with particles (Smith et al., 1992; Azam and Long, 2001; Kiørboe and Jackson, 2001). This assumption may have been accurate since signal intensities were higher in 2 μM than in 20 nM HPG (Figure 5). Also, the dramatic increase in HPG signal intensities over 8 day for particle-associated cells suggests that methionine concentrations in the particles were initially high and decreased over time. Possible explanations include chemical diffusion away from the particles or bacterial incorporation and conversion into biomass. As a result, the removal of methionine relieved competition with HPG, which led to increased incorporation of HPG. An alternative explanation for low HPG signal intensities at 1 and 48 h is the occurrence of antagonistic interactions within the attached consortia that constrained cell growth (Long and Azam, 2001a).

The observations are consistent with previous work on marine particles that demonstrated particle-attached bacteria are active. One study used fluorescence microscopy to detect incorporation of the thymidine analog BrdU (Hamasaki et al., 2004). BrdU percent labeling was higher on particles than among free-living cells (56 vs. 35%) but signal intensities were not quantified for particle-attached bacteria. A few studies have used microautoradiography and showed that attached bacteria were active and exhibited higher percent labeling than free-living bacteria (Kirchman and Mitchell, 1982; Paerl, 1984). Specifically, size-fractionated samples demonstrated that particle-associated cells accounted for disproportionately high incorporation relative to free-living cells; the tested substrates included ¹⁴C-glucose and ¹⁴C-glutamate (Kirchman and Mitchell, 1982), or an ³H-amino acid mixture and ³³PO₄ (Paerl, 1984).

Our results suggest that percent labeling on particles is influenced by incubation times and concentrations of both a compound and its analog in the surrounding environment. Single cell protein production rates on a particle can vary widely which may reflect microscale heterogeneity of resources on the particle itself, e.g., protein-depleted microscale patches may have enhanced HPG incorporation. The HPG method may enable cells to act as bioindicators for the “nutritional status” of a particle, simultaneously providing information on particle lability and single cell growth rate.

When we applied the conversion factor to cultures of Alteromonas AltSIO and As1, protein synthesis rates varied by greater than 3 orders of magnitude. Here cells were grown ~21 h in non-enriched seawater in contrast to >40 h growth used in the conversion factor experiments. The large variance was unexpected but indicates that growth rate heterogeneity can occur among clonal cells undergoing logarithmic growth (Figure S8). The pattern is consistent with the adaptive strategy of bacterial aging in E. coli (Rang et al., 2011) in that heterogeneity may indicate repeated accumulation of oxidative damage to mother cells (lower intensity cells) that serve to rejuvenate daughter cells (higher intensity cells). If growth heterogeneity in AltSIO and As1 reflects similar behavior in natural marine communities, then individuals of specific taxa may exhibit a broad range of growth rates that depend on the overall health of their predecessors. This may partially explain the existence of dividing cells containing one HPG-labeled and one non-labeled cell (Figure 7). Such asymmetrical growth coupled with the prevalence of very slow growing cells could have important consequences for the rank-abundance of microbial taxa within the rare biosphere (Sogin et al., 2006).

Chloramphenicol inhibition of protein synthesis was used to determine the specificity of HPG incorporation. Samples were incubated with 50 μg mL⁻¹ chloramphenicol + 20 nM HPG for 1 and 6 h, processed, and compared to unamended controls.

E. coli uses methionyl-tRNA synthetase to incorporate HPG into protein in place of methionine (Kiick et al., 2002; Wang et al., 2008). We conducted four incorporation kinetics experiments to test that natural marine microbial communities also incorporated HPG in place of methionine.

Methionine inhibition of HPG incorporation. We incubated seawater samples with 20 nM HPG plus 0, 1, or 2 μM methionine for 0.5, 1, 4, and 6 h. Samples were click-processed (as above, without filter-transfer-freeze), imaged, and analyzed to quantify methionine inhibition of HPG incorporation.

HPG inhibition of ³⁵S-methionine incorporation. Scripps Pier seawater was incubated for 1 h with 2 nM ³⁵S-methionine alone and with 8 nM, 18 nM, 20 nM, 200 nM, 2 μM HPG, or 8 nM HPG + 10 nM cold methionine. ³⁵S incorporation was quantified by the centrifugation method (triplicate samples with duplicate controls; Smith and Azam, 1992).

HPG inhibition of methionine incorporation: Competitive or non-competitive? We evaluated uptake kinetics to test whether HPG is a competitive, non-competitive, or mixed inhibitor of methionine (Michaelis and Menten, 1913). We used natural coastal seawater bacterial assemblages collected from the Scripps Pier, and analyzed the concentration-dependence of ³⁵S-methionine incorporation (K_m) at 5 concentrations (2, 5, 10, 30, and 100 nM). Deviation of ³⁵S-methionine incorporation kinetics due to the presence of 0, 30, 100, and 1000 nM HPG was analyzed to determine the K_i for HPG, which is numerically equivalent to HPG K_m (Experimental details in Table S1).

³⁵S-methionine microautoradiography. Parallel incubations were conducted to compare microautoradiography and click chemistry for detection of the proportion of protein synthesizing cells from natural assemblages. Slides were prepared for microautoradiographic analysis from samples incubated with ³⁵S-methionine in the presence (200 nM and 2 μM) or absence of HPG. Simultaneously, samples were incubated with HPG and click-processed for visualization: (i) directly on filters and (ii) on glass slides after applying the filter-transfer-freeze technique. The objective was to compare the proportion of positively labeled cells by the three approaches, as well as to quantify the potential inhibitory effect of the presence of HPG on cellular protein synthesis.

Seawater was incubated for 1 h in the dark unless otherwise noted. Following incubation with 2 nM ³⁵S-methionine + 18 nM methionine, samples were formaldehyde fixed (2% final concentration) for >15 min. Killed controls received a mixture of radiolabeled amino acid and formaldehyde. Samples were filtered onto 0.2 μm pore size polycarbonate filters backed with 0.45 μm pore size mixed cellulose ester filters (Isopore membrane filters and MF membrane filters, Millipore, Billerica, MA). Filters were rinsed with 0.2 μm filtered 1× PBS or Milli-Q water, dried on absorbent tissue, and stored at −20°C.

Filters were cut into 1/8 to 1/4 pieces. Preparations for microautoradiography were performed under darkroom conditions using a Kodak GBX-2 safelight with a 15 W bulb placed ~5 m away from the working area. This extremely faint light ensured low background, but necessitated the use of night vision monocular goggles (D-112MG; Night Optics USA, Inc., Huntington Beach, CA). We modified an emulsion coating and cell transferring protocol based on those from previous publications (Fuhrman and Azam, 1982; Teira et al., 2004; Longnecker et al., 2010). Ammersham LM-1 emulsion was melted at 43°C for 15 min to 1 h. Slides were dipped in emulsion for 5 s, allowed to drip for 5 s, wiped to remove emulsion on the back of the slide, and immediately placed onto a flat sheet of aluminum foil on ice. After 5–10 min, a filter slice was placed cell side down onto the emulsion. When this was completed for all filters, each slide was placed in a slide box with desiccant. The slide box was sealed with black tape, wrapped in aluminum foil, placed in a cardboard box, kept at RT for 1 h, and then transferred to 4°C for 72 h exposure. We did not test longer exposure times than 72 h, but note that this duration is longer than published studies examining both %-active and silver-grain sizes using ³H-labeled compounds (Cottrell and Kirchman, 2003; Sintes and Herndl, 2006; Longnecker et al., 2010). Since ³⁵S emits nearly 10x more energy than ³H, our results likely reflect the upper bound of %-active cells as revealed using ³⁵S-methionine.

Slides were developed with Kodak D19. In 50 mL tubes, 1.6 g of developer was mixed with 40 mL water and dissolved at 43°C for 15 min. Nine grams of fixer was mixed with 40 mL water at room temperature until fully dissolved. Four separate slide mailers (Fisher Scientific, Pittsburgh, PA, USA) were used as reagent containers and filled with 20 mL of developer, developer wash, fixer, and fixer, respectively. Under darkroom conditions, the slides were removed from the darkened box, and the filters were gently removed from each slide using forceps, thereby transferring the cells from the filters (now embedded in exposed emulsion).

Slides were submerged in developer for 4 min, washed for 10 s, then submerged in fixer for 5 min, washed again for 5 min, and then allowed to dry on absorbent paper. Once the slide was dry, 50–100 μL of 1 μg mL⁻¹ DAPI was added to the location of the cells and incubated for 10 min. The slide was then washed in autoclaved filtered water and dried again. A coverslip (#1.5 thickness) was mounted to the slide with 15–20 μL of antifade mounting medium.

Microscopy visualization was performed as described above with omission of the green channel. Transmitted light was used to observe the presence or absence of exposed emulsion grains. Percent labeling was calculated as the number of DAPI-labeled cells located within exposed emulsion divided by the total number of DAPI-labeled cells in each field of view for both live and control samples. The data were arcsine transformed, averaged, blank subtracted, and then back transformed. Autofluorescent cyanobacteria and/or eukaryotic cells were excluded from analyses.

We tested 3 approaches to calculate a valid factor converting sum intensity (SI) to fg protein, called “top 10% isolate mean,” “isolates regression,” and “community sum ³⁵S-methionine.”

Alteromonas species As1 and AltSIO (Pedler et al., 2014) were used to quantify the “top 10% isolate mean.” These strains were acclimated to and maintained in GF/F-filtered autoclaved seawater (FASW) to simulate natural seawater conditions. Exponentially growing cells of each isolate were inoculated into triplicate flasks containing 100 ml FASW at a starting concentration of 10² cells ml⁻¹. Flasks were incubated at room temperature for 16 h, at which point concentrations were ~10³ cells mL⁻¹, corresponding to growth by 4 doublings (confirmed via DAPI counts). HPG (20 nM) was added to one flask of each isolate and the other remained unamended as a control. Assuming each cell had ≤ 4.0 × 10⁶ methionine residues (an upper limit), 20 nM HPG would be an ample supply for up to ~10⁹ cells and was not limited in the incubation volumes. After 24 and 48 h additional incubation (T40 and T64 h total growth time), 20 mL sub-samples were formaldehyde fixed (2%) and filtered. Each sample was click processed, imaged, counted and measured. Images for the green channel were acquired with 50 ms exposure due to bright labeling of cells. Cell volume (V) and protein content (P) was calculated as above. After fluorescence background subtraction, sum intensities for the cells (of both isolates) were divided by volume-derived protein content (P) to obtain a conversion factor of fluorescence in units of sum intensity per fg protein. This value was multiplied by 20 to scale the 50 ms exposure to images acquired from the standard 1 s exposure used in natural assemblages. We concluded this intensity quantification was valid after measuring the intensity linearity of calibrated fluorescent beads. The average intensity:protein value of the top 10% most intensely labeled cells at T40 and T64 was calculated as a conversion factor. Here it was assumed that only the top 10% most intensely labeled cells were “saturated” with HPG.

The “isolates regression” approach used the cell signal intensities (from the experiment described above) and plotted these against the cell volume-derived protein content (P). A linear regression was fit to the data, and the inverse of the slope in terms of log₁₀ RFU fg protein⁻¹ was the calculated conversion factor. Here it was assumed that all cells, both faintly and brightly labeled, were saturated with HPG.

Lastly, the “community sum ³⁵S-methionine” approach was calculated from two experiments that quantified parallel HPG labeling and ³⁵S-methionine bulk incorporation. On two dates, 1 h incubations of Scripps Pier seawater with 2 nM ³⁵S-methionine + 18 nM methionine were performed alongside separate 1 h incubations with 20 nM HPG. Bulk protein synthesis was calculated from ³⁵S-methionine incorporation assuming zero isotope dilution and 2.2 mol%. The bulk assemblage HPG community sum (i.e., of ~ 1 × 10⁶ cells mL⁻¹) was extrapolated from directly measured cells. First, the HPG labeling percentage (calculated via microscopy) was multiplied by the DAPI-based cell density to provide the number of HPG⁺ cells in the bulk assemblage. This was divided by the number of microscopy-based HPG⁺ cells to yield a proportional value that was multiplied by the community sum to yield the bulk assemblage community sum. It was then divided by the ³⁵S-methionine bulk protein synthesis rate to provide a conversion factor in units of log₁₀ RFU fg protein⁻¹.

The “top 10% isolates mean” factor was applied to individual cell HPG intensities acquired from natural samples and other experiments to obtain single-cell protein production (scPP; fg cell⁻¹ h⁻¹).

Seawater was collected from the Ellen Browning Scripps Memorial Pier (32° 52.02′ N, 117° 15.43′ W; “Scripps Pier”). Measurement of environmental parameters including temperature, salinity, and chlorophyll fluorescence were conducted as part of the Southern California Coastal Ocean Observing System (SCCOOS; www.sccoos.org). In total, 21 samples were collected and processed.

Offshore samples were collected at 3, 30, and 60 m depth from the San Diego trough (32° 38.022′ N, 117° 34.034′ W) via Niskin bottles attached to a CTD rosette (Sea-Bird Electronics, Bellevue, WA).

Comparison of HPG-based activity profiles among all the sampling dates was performed by plotting the frequency distributions of cell-specific protein production on log₁₀-log₁₀ axes. The percent cell labeling vs. fg cell⁻¹ h⁻¹ within 0.05 fg bins were fit to power-law functions for each sample, and the resulting equations were used to generate a linear regression by reorganizing the power law equation (y = Ax^B) into logarithmic form (log y = B log x + log A). This enabled B, the scaling exponent, to be used as the regression slope alongside A, the scaling factor, as the y-intercept.

Duplicate 40 mL seawater samples were treated as follows: (i) 20 nM HPG, (ii) 20 nM HPG + 2% formaldehyde, (iii) 2 μM HPG, (iv) 2 μM HPG + 2% formaldehyde, and (v) no addition. Each was incubated for 1 h, 48 h, and 8 day. At each time point, 30 mL was removed, formaldehyde-fixed, and filtered onto 8 μm white polycarbonate filters (30 mL). Filters were click-processed, but filter-transfer-freeze was not applied to 8 μm filters (to minimize potential damage to particles and/or alteration of bacterial associations). Ten particles (>10 μm in diameter and containing at least 20 cells) were imaged on the 8 μm filters. Particle-attached cells were only measured and quantified if they were in focus, a characteristic of most particles observed in this study. Particles with high proportions of unfocused cells were not visualized, and those with few unfocused cells were visualized, but the cells were not analyzed.

Alteromonas spp. strains As1 and AltSIO were inoculated at ~10³ cells mL⁻¹ into GF/F filtered and autoclaved seawater. Through preliminary growth experiments in the same growth medium, it was determined that sampling would begin 16 h after inoculation to target the onset of exponential growth phase. One hour incubations amended with 20 nM HPG were carried out every hour from 16 to 22 h (duplicate with one control). Samples were fixed, filtered, stored, and then click- and filter-transfer-freeze processed as above. Images were acquired at 200 ms to maintain unsaturated pixels and signal intensities measured for each cell were multiplied by 5 for congruence with all samples.