Pigment concentrations and cell abundance data used in this study were compiled from measurements carried out during oceanographic campaigns that focused primarily on the tropical and subtropical waters of the Atlantic and Pacific basins (Table 1 and Figure 1). Seawater samples were collected using a conductivity-temperature-depth (CTD) profiler equipped with a rosette of Niskin bottles. Further details on sampling strategies for individual oceanographic datasets can be found in the references provided in Table 1.

Phytoplankton pigments were identified and quantified using High-Performance Liquid Chromatography (HPLC). For each of the datasets, the filtration volumes, organic solvents used for pigment extraction, and corresponding references are summarized in Table 2. Note that DVChl-a and DVChl-b, pigment markers unique to Prochlorococcus, are not available for the Bermuda Atlantic Time-series Study (BATS) as the HPLC method employed did not allow for these pigments to be completely separated from their monovinyl derivatives (Steinberg et al., 2001). A quality check on the HPLC-determined pigment data was conducted using linear regressions (model I) of total accessory pigment concentration on total chlorophyll-a concentration (monovinyl + divinyl Chl-a; TChl-a) (Trees et al., 2000). Outliers with estimated standard deviations greater than 3 in absolute value (Uitz et al., 2006) were removed (final slope = 0.82, R² = 0.90), reducing the pigment database to 2,960 observations, which was 83% of the original data.

Cell counts of the picocyanobacteria Prochlorococcus and Synechococcus were determined by Analytical Flow Cytometry (AFC) on either freshly collected or preserved samples. Synechococcus populations were identified from the orange fluorescence emitted by their characteristic phycoerythrin pigment-protein complex (Moore et al., 1995), and Prochlorococcus populations were distinguished based on their red fluorescence associated with DVChl-a and right-angle light scattering properties (Zubkov et al., 1998). However, Prochlorococcus cells collected from the surface layer of the subtropical gyres typically contain a very low concentration of DVChl-a, which makes it difficult to enumerate the cells accurately using their natural fluorescence signal (Zubkov et al., 1998; Grob et al., 2007a). Therefore, alternative approaches have been developed at different laboratories to estimate the abundance of surface Prochlorococcus populations that are weakly fluorescent as summarized below.

For the BIOSOPE project, Grob et al. (2007a); Grob et al. (2007b) determined the concentration of Prochlorococcus cells with low fluorescence by using the software Cytowin to fit a normal curve to the signal distribution plot. In cases where the fluorescence signal was too dim, the cell number was derived from the pigment DVChl-a assuming an average cell-specific concentration 0.23 fg DVChl-a cell^-1 which was intracellular pigment concentration collected within the surface layer above the 5% level of the incident light and corresponded to the lowest possible values reported in Blanchot and Rodier (1996) and Partensky et al. (1999) for surface, HL Prochlorococcus populations. The BIOSOPE samples with Prochlorococcus cell abundances indirectly estimated from DVChl-a (N = 23) were included in our analyses, and they appeared to agree well with the general trends observed for other datasets.

For the AMT (Zubkov et al., 1998; Hartmann et al., 2014; Lange et al., 2018) and DCM datasets (Veldhuis and Kraay, 2004), Prochlorococcus in surface water samples were DNA-stained with fluorescent dyes, such as TOTO-1 (Zubkov et al., 1998), or SYBR Green (Hartmann et al., 2014), or PicoGreen (Veldhuis and Kraay, 2004), prior to analysis by flow cytometry. The staining technique allows for better visualization of the dim Prochlorococcus cells based on the green fluorescence originating from the DNA-binding dyes in addition to their chlorophyll (red) fluorescence and scattering signals.

Flow cytometric measurements for the HOT program between 1990 and 2005, which corresponds to the period when most of the HOT data used in this study were collected, were carried out using a flow cytometer fitted with a 1-W 488-nm laser as described in Campbell and Vaulot (1993). However, the authors noted that the flow cytometer configuration often failed to detect surface populations of Prochlorococcus. Thus, under these circumstances, a laser that generates an excitation beam in the range 457-515 nm was used instead (Olson et al., 1990) so that the excitation energy more closely matches the absorption bands of the DVChl-a and DVChl-b (Olson et al., 1990). This flow cytometer set-up was also optimized to increase the sensitivity of the instrument to detect surface assemblages of Prochlorococcus at the BATS station (DuRand et al., 2001).

For each of the datasets, the depth of the mixed layer (z_m ) was calculated from vertical profiles of the potential density (σ_θ ) [kg m^-3] obtained from online data repositories. When the online information of σ_θ was not readily available, the σ_θ was calculated from temperature and salinity profiles using the ‘swSigmaTheta’ function in the R statistical package ‘oce’ (Kelly, 2018). The z_m was defined as the depth at which the σ_θ was offset from the near surface (10 m) reference value by a threshold value of 0.125 kg m^-3 (Bouman et al., 2006).

The euphotic depth (z_e ) is often considered as the maximum depth in the water column where phytoplankton photosynthesis takes place, and typically corresponds to the depth at which the downwelling irradiance of photosynthetically-available radiation (PAR) decrease to 1% of its surface value (Kirk, 2011). Assuming that the spectrally non-resolved vertical attenuation coefficient for PAR, K_d (PAR) [m^-1], is constant with depth, the z_e can be estimated from K_d (PAR) using the relation (Kirk, 2011):

K_d (PAR) was converted from the attenuation coefficient at 490 nm, K_d (490) [m^-1], using the relationship reported in Morel et al. (2007) for open-ocean (Case I) waters:

The K_d (490) was calculated from TChl-a based on an empirical relationship derived by Morel and Maritorena (2001):

where K_w (490) = 0.0166 m^-1 is the attenuation coefficient for pure water at 490 nm, and Chl is the average TChl-a [mg Chl-a m^-3] in the mixed layer (Grébert et al., 2018).

Profiles of pigment ratios and intracellular pigment concentrations of picocyanobacteria are presented here as a function of optical depth (the nominal depth divided by the estimated euphotic depth, z/z_e ). The advantages of using the dimensionless z/z_e index rather than nominal depths are twofold. First, it provides an approximation of ambient light conditions at a given depth when in situ vertically-resolved measurements of irradiance were not available (Uitz et al., 2008). Second, as pointed out in Lange et al. (2018), it implicitly carries information about attenuation of light through the water column, which has been shown to be a more effective predictor of long-term changes in phytoplankton pigments and cell abundance in the open ocean than does the instantaneous variation of light intensity expressed in absolute units.

Spatial distribution of pigment composition of phytoplankton populations in surface waters across different trophic domains encountered during the BIOSOPE transect in the South Pacific is illustrated in Figure 2. There is a clear relationship between the shoaling of the mixed layer and increased nutrient supply which likely reinforced changes in the relative contribution of PPPs as a result of the combined effect of taxonomic changes and photoacclimation. When we focus on depth-dependent changes in the ratio of PPPs (sum of zeaxanthin (ZEA), diadinoxanthin (DIAD), diatoxanthin (DIAT), violaxanthin (VIOL), alloxanthin (ALLO), and carotenes (CARO)) to the sum of all pigments (referred to in this study as the PPP index) within the oligotrophic samples, we observed a striking systematic increase from productive, upwelling stations to mesotrophic conditions, and to the high-nutrient, low-chlorophyll (HNLC) regions and the (hyper)oligotrophic central gyres (Figure 3A). These trophic-dependent changes that are evident throughout the euphotic zone are likely the result of corresponding shifts in both the photoacclimatory status and the taxonomic composition of phytoplankton communities.

Given that the BIOSOPE data were all collected within a restricted time period and latitudinal range, variation in incident solar radiation should be minimal and thus differences in the pigment composition of surface phytoplankton communities are most likely primarily caused by shifts in taxonomic structure. Thus, we used the BIOSOPE dataset to explore the relative importance of cyanobacterial- versus eukaryotic-associated pigments in controlling the overall variation in the PPP index covering four major nutrient regimes (eutrophic, mesotrophic, oligotrophic and HNLC). In the HNLC and oligotrophic stations, ZEA, a diagnostic pigment of the picocyanobacteria Synechococcus and Prochlorococcus, accounted for more than half of the total PPP concentration (Figure 2A). On the other hand, DIAD and DIAT, associated with eukaryotic phytoplankton including photosynthetic picoeukaryotes (Dimier et al., 2009), contributed a low percentage (⪅ 30%) to the total PPP budget in these regions (Figure 2B). The relative abundance of ZEA decreased, while that of DIAD + DIAT increased as trophic conditions progressed toward a more productive state, indicating a domain shift from a prokaryote-dominated to a eukaryote-dominated surface phytoplankton community (Karl, 1999; Bouman et al., 2011).

The domain shift in pigment composition was also reflected in a progressive increase in the relative proportion of the photosynthetic pigment (PSP) fucoxanthin (FUCO), which is a pigment marker for diatoms, from oligotrophic open-ocean region to the coastal upwelling zone (Figure 2C). The PSPs chlorophyll-b (Chl-b) and prasinoxanthin (PRAS) along with the PPPs VIOL and neoxanthin (NEOX) were in significant quantities at several upwelling sites (data not shown) and indicated the presence of picoeukaryotic green algae (Ras et al., 2008), such as those in the genus Ostreococcus which has been reported to thrive in various coastal environments (Tragin and Vaulot, 2019). The near ubiquity of 19′-hexanoyloxyfucoxanthin (19′-HEX) throughout the transect (Figure 2E) reflects the cosmopolitan nature of haptophytes (Liu et al., 2009; Swan et al., 2016), and the elevated fraction of this carotenoid observed under oligotrophic conditions and in HNLC waters suggests that some of these cells may fall within the picophytoplankton size fraction (Ras et al., 2008; Liu et al., 2009).

The HNLC stations have surface TChl-a values (0.2-2 mg Chl-a m^-3) similar to those of the mesotrophic sites (i.e., near the Marquises Islands, Figure 1), yet they exhibited a markedly higher abundance of picocyanobacteria, especially Prochlorococcus (average of ~200×10³ cells mL^-1) in surface waters (0-80 m) of the HNLC region compared to ~60 × 10³ cells mL-^-1 at other mesotrophic sites (see also Grob et al., 2007a). This was reflected clearly in a markedly higher relative proportion of ZEA (Figure 2A). The numerical dominance of Prochlorococcus in these low-latitude HNLC waters was associated with highly stable water columns which were likely influenced by the South Pacific Tropical Water (Ras et al., 2008). Molecular studies (Rusch et al., 2010; Malmstrom et al., 2013) have revealed that Prochlorococcus isolated from HNLC regions of the tropical Pacific and Indian Oceans are adapted to iron limitation by reducing their iron requirements through the loss of several genes encoding iron-containing proteins which are mostly involved in the electron transport chain. These HNLC isolates also possess genes that enable them to efficiently acquire iron bound to organic ligands in seawater (Rusch et al., 2010; Malmstrom et al., 2013; Larkrin et al., 2020).

The taxonomic composition of open-ocean phytoplankton communities is governed largely by the supply of nutrients (Cullen et al., 2002). The region in the water column where there is a sharp increase in nitrate concentration is commonly referred to as the nitracline, and its depth, termed the nitracline depth, is widely regarded as an indicator of the supply of the deep-ocean nitrate to the upper mixed layer (Cullen et al., 2002; Cermeño et al., 2008). Along the BIOSOPE transect, the depth of the nitracline was found to increase steadily from the productive upwelling system nearshore to the more oligotrophic open ocean (Figure 2F). In the gyres, the deep nitracline and the corresponding reduced nutrient supply to surface waters resulted in a high abundance of picocyanobacteria indicated by higher values of the PPP index (Figure 3A). As the nitracline shoaled, the surface phytoplankton community structure shifted toward a eukaryote-dominated ecosystem (i.e., higher relative contributions of DIAD + DIAT and FUCO (Figures 2B, C), and lower PPP index values (Figure 3A)) supported by entrainment of new nutrients from depth.

Figure 3 shows that the PPP index decreased vertically coinciding with a reduction in the relative availability of light in the water column as indexed by z/z_e (Uitz et al., 2008). For all of the datasets, the magnitude of the PPP index varied appreciably within the surface layer (i.e. z/z_e < 0.5), and the values appeared to converge to a minimum below the ~1.5z_e depth.

HPLC data from the BATS time series revealed that the PPP index at the sea surface varied seasonally with consistently low and less variable values in winter and spring, and higher and more variable PPP index values observed during the summer-autumn months (Figures 3E–H). The uniformly low PPP index observed during winter (late December to mid-March) (Figure 3E) could be attributed to the combined effect of deep vertical mixing typical of the Sargasso Sea (Steinberg et al., 2001) and a seasonal reduction in incident solar radiation leading to a weakly-stratified water column. Nutrient-replete conditions established previously by deep winter mixing allowed eukaryotic phytoplankton to dominate within the surface layer until early spring (late March to mid-April). The relative similarities in the ambient nutrient and light conditions during winter and spring likely resulted in surface populations dominated by low light-acclimated cells and thus were reflected in the PPP indices for both seasons falling within a relatively similar range (Figures 3E, F). As the upper water column became thermally stratified due to intense solar heating, and the intensity of the wind-driven surface mixing was weakened during the late spring and summer months (late June to mid-September), the mixed layer shoaled such that phytoplankton were confined within the upper mixed layer and thereby continuously exposed to high UV and visible radiation. Photoacclimation of phytoplankton to high-light conditions resulted in an overall increase in the PPP index observed for both eukaryote-dominated and picocyanobacteria-dominated populations (Figure 3G). Picocyanobacterial cells continued to dominate the surface phytoplankton as surface waters became more nutrient-limited during the early autumn (late September to early October) when the sea-surface temperature remained relatively high (> 25°C) and the water-column was strongly stratified. However, as the autumn progressed, surface cooling and a deepening of the mixed layer allowed for the upper water column to be replenished with nutrients and resulted in a phytoplankton community dominated by larger eukaryotic cells acclimated to low-light intensities. This prokaryote-to-eukaryote domain shift occurring intraseasonally between early and late autumn was supported by the contrasting profiles of the PPP index (Figure 3H).

The BIOSOPE dataset illustrates that domain-specific (prokaryotic to eukaryotic) shifts in gross phytoplankton community structure governs horizontal changes in the relative contribution of photoprotective pigments to total pigment biomass in the surface ocean. However, the strong vertical decline in the PPP index was a striking feature exhibited across the various datasets that span a range of spatial and temporal scales (Figure 3). Since picocyanobacteria have been shown to exhibit tremendous variation in their pigment composition in response to both seasonal and vertical changes in light conditions, we examined the photoacclimatory characteristics of this group at the cellular level via the intracellular content of their diagnostic pigments. Our analyses were focused primarily on the picocyanobacteria-dominated waters of the subtropical and tropical oligotrophic gyres where simultaneous measurements of abundance determined by flow cytometry and pigment concentrations measured using the HPLC method were widely available.

In picocyanobacteria, photoacclimatory changes in the magnitude of the PPP index could be attributed to variation in the intracellular concentration of either photoprotective (Kulk et al., 2011; Mella-Flores et al., 2012), or photosynthetic pigments (Kana et al., 1988; Moore et al., 1995), or both. To assess the relative influences that each of these two major pigment groups has on variation in the PPP index, intracellular concentration of ZEA (Figure 4) found in both picocyanobacterial genera Synechococcus and Prochlorococcus and DVChl-a, which is the major photosynthetic pigment for the genus Prochlorococcus (Figure 5) were plotted against the z/z_e index as a proxy for the relative light availability. When calculating the intracellular concentration of ZEA, we partitioned this pigment between the two cyanobacteria groups based on culture observations that have shown that ZEA per cell for Synechococcus is approximately three-times higher than that of Prochlorococcus when grown over a similar range of light intensities (Moore et al., 1995).

The intracellular concentration of ZEA remained relatively uniform from the well-lit surface layer down to the bottom of the euphotic zone (Figure 4), whereas the amount of DVChl-a cell^-1 gradually increased with depth (Figure 5). However, for the BIOSOPE dataset, the calculated DVChl-a cell^-1 was relatively constant within the top 0-0.5z_e layer with most values falling between 0.2-0.3. This is partially the result of the authors estimating Prochlorococcus abundance based on pigment data assuming 0.23 fg DVChl-a cell^-1 for surface samples (Figure 5, red circles) where the natural fluorescence of these cells was too low to detect them using flow cytometry (Grob et al., 2007a). In addition, mesotrophic stations sampled during the BIOSOPE cruise off the Marquesas Islands exhibited intracellular concentrations of DVChl-a that were substantially higher (nearly 1.0 fg DVChl-a cell^-1) and more uniform over the vertical profile than the rest of the transect (Figure 5, blue circles). Vertical density profiles revealed that these stations were more deeply mixed (mixed layer depths of ~70-100 m compared to< 50 m typically observed at other stations), which would reduce the average daily mixed-layer irradiance for phytoplankton communities. If we assume that Prochlorococcus cells were uniformly mixed throughout the mixed layer, the reduction in the average mixed-layer irradiance would result in an increase in the intracellular concentration of their major light-harvesting pigments and therefore explain the relatively high and uniform DVChl-a cell^-1 and low PPP index (Figure 3A, blue circles) measured throughout the water column at these stations.

Uniform ZEA cell^-1 values over a range of z/z_e values (Figure 4) are reflected in the linear relationship between bulk ZEA concentration and the number of picocyanobacterial cells (Pico) depicted in Figure 6 (log₁₀ Pico = 0.90 log₁₀ ZEA + 3.65, R² = 0.61). Upper and lower envelopes shown in Figure 6 correspond, respectively, to the maximum (q_max ) and minimum (q_min ) quantiles (Sathyendranath et al., 2009). Quantile regressions facilitated constraining the variability in values of ZEA cell^-1 by allowing outliers that fell outside the computed maximum and minimum quantiles to be identified and removed. The lower and upper bounds could be viewed as being representatives of extremely high-light (surface) and low-light (bottom of the euphotic zone) conditions, respectively. Note that higher ZEA cell^-1 values could result from the presence of other cyanobacterial populations not accounted for in the flow cytometry counts, in particular the colonial diazotroph Trichodesmium that is known to be abundant in the North Atlantic Subtropical Gyre.

Quantile regressions were also fitted to the plot of DVChl-a concentration against Prochlorococcus cell abundance (Pro) (Figure 7A; log₁₀ Pro = 0.82 log₁₀ DVChl-a + 3.53, R² = 0.40). In this case, the physiological interpretation of the upper and lower envelopes is the opposite to that for the ZEA-cell abundance relationship (Figure 6): the lower boundary (q_min = 0.01) corresponds to the upper limit of DVChl-a cell^-1 for Prochlorococcus cells acclimated to extremely low light, whereas the upper boundary (q_max = 0.99) represents the lower limits of DVChl-a cell^-1 resulting from cells photoacclimated to maximum surface irradiances. The overall variability in the relationship between DVChl-a and cell abundance (Figure 7A) was far greater than that of ZEA and cell abundance (Figure 6). This highlights the significance of the photosynthetic pigment fraction in driving changes in the PPP index in picocyanobacterial-dominated systems.

The carbon-to-DVChl-a ratio of Prochlorococcus (θ_Pro) was estimated based on simultaneous measurements of cell abundance and the concentration of DVChl-a illustrated in Figure 8. Prochlorococcus cell carbon was estimated from abundances (determined by flow cytometry) using two sets of conversion factors. The first conversion factor of 49 fg C cell^-1 (Cailliau et al., 1996) based on the assumption that both HL- and LL-ecotypes have similar carbon contents. In non-axenic cultures of a Prochlorococcus strain, Cailliau and colleagues (1996) derived the cellular carbon contents from flow cytometric cell counts and measurements of particulate organic carbon (POC) that were corrected for the presence of heterotrophic bacteria. Their study found that the values obtained for Prochlorococcus under low and high light were not statistically different, suggesting that this value may be applied to both HL and LL cells.

We also applied conversion factors of 52 and 158 fg C cell^-1 (Casey et al., 2013) to Prochlorococcus populations residing within the euphotic (0-120 m) and sub-euphotic zone (> 120 m), respectively, during the period of water-column stratification which is reflected in the presence of a deep chlorophyll maximum. Casey and colleagues (2013) derived the carbon biomass of Prochlorococcus at the BATS stations based on an empirical relationship between flow cytometric forward light scatter and cellular carbon content. Their study found that Prochlorococcus carbon content increases significantly with depth as cells acclimate to low-light conditions potentially by increasing the number of thylakoid membranes to allow increased synthesis of PSPs (Casey et al., 2013).

Values of the cellular carbon content of surface Prochlorococcus populations obtained by both Cailliau et al. (1996) from cultures and Casey et al. (2013) from field samples are comparable to published estimates from other studies [e.g. 53 fg C cell^-1 reported in DuRand et al. (2001); 54 fg C cell^-1 in Campbell et al. (1994) and Grob et al. (2007b)], although it is important to stress that determination of the carbon content from cell volume has a significant degree of uncertainty since, as pointed out in Bertilsson et al. (2003), the amount of carbon per unit cell volume observed in marine bacteria can vary both between and within species in response to changes in growth conditions. Our choice of conversion factors is therefore based on estimates of Prochlorococcus cellular carbon content derived from direct measurements of POC (Partensky et al., 1999; Bertilsson et al., 2003) and the conversion factors used in this study fall within those obtained from studies of natural picophytoplankton populations.

When samples collected from below the 1.5z_e depth and potential outliers (Figure 8) were removed, our estimates of θ_Pro varied from 9 to 499 g C (g DVChl-a)^-1 for all samples, and θ_Pro values for surface populations (z/z_e < 0.1) ranged from 66-361 g C (g DVChl-a)^-1. To assess whether the extent of the observed variability in θ_Pro was reasonable, these results were compared with θ_Pro measured in various culture experiments and field studies using an approach outlined in Marañón (2005). Given that (1) the intracellular carbon content of Prochlorococcus previously reported in the literature ranged from about 29 to 158 fg C cell^-1 (Casey et al., 2013 and references therein), and (2) the intracellular content of DVChl-a of Prochlorococcus cells subject to extremely high irradiances typical of the subtropical surface ocean (e.g. > 1,000 μmol quanta m^-2 s^-1) can be as low as ~0.2 fg DVChl-a cell^-1 (Blanchot and Rodier, 1996; Grob et al., 2007a) and that of Prochlorococcus growing under very low irradiances (e.g. 7 μmol quanta m^-2 s^-1) is up to 4.0 fg DVChl-a cell^-1 (Partensky et al., 1993; Blanchot et al., 2001), the resulting upper boundary of θ_Pro values should be in the range of 145-790 g C (g DVChl-a)^-1 and the lower boundary in the range of 7-40 g C (g DVChl-a)^-1. In general, our estimated values of θ_Pro fell within these derived upper and lower limits.

Using open-ocean samples collected from the Northwest Atlantic and the Arabian Sea, Sathyendranath et al. (2009) indirectly estimated an upper bound on θ for different groups of phytoplankton in the field based on quantile linear regressions of POC and Chl-a concentration and obtained an average θ_Pro value of 145 g C (g DVChl-a)^-1 for stations classified as Prochlorococcus-dominated. When this average value was superimposed on the scatterplot of Prochlorococcus cell carbon content and DVChl-a concentration (Figure 8), the line of best fit from the Sathyendranath et al. (2009) study lies remarkably closed to the best equation obtained in this study, despite being derived from independent datasets. It should be noted that most of the data used in Sathyendranath et al. (2009) were from samples collected within the top 40 m of the water column, and therefore their estimate likely applies to cells acclimated to mixed-layer light intensities. The estimated value of θ_Pro obtained in the Sathyendranath et al. (2009) study was higher than the average θ_Pro values reported by Blanchot et al. (2001) for surface (0-40 m) Prochlorococcus samples from the oligotrophic (66 g C (g DVChl-a)^-1) and HNLC (93 g C (g DVChl-a)^-1) regions of the equatorial Pacific (Table 3). It has been shown that Prochlorococcus samples from mesotrophic regions contain more carbon biomass relative to chlorophyll pigment than samples associated with oligotrophic environments (Wei et al., 2019) and this may explain the higher θ_Pro values in the Sathyendranath et al. (2009) study.

The vertical patterns in the cellular content of PPPs and PSPs observed in this study are consistent with those obtained by Matsumoto and Ando (2009) who examined vertical profiles of picocyanobacterial pigments in the nutrient-limited western Pacific and found a steady vertical increase in the intracellular Chl-a concentration but no obvious variation in ZEA content. Our findings are also supported by culture experiments. Kana et al. (1988) were amongst the first to observe the constant ZEA cell^-1 in nutrient-replete batch cultures of Synechococcus WH7803 grown under different irradiance levels. Similarly, Moore et al. (1995) found that Synechococcus WH8103 and Prochlorococcus MED4 maintained relatively constant cellular content of ZEA under varying irradiances, although for the Prochlorococcus SS120 strain, there was a slight increase with the increased light level. In contrast, the intracellular concentration of (DV)Chl-a was found to decrease significantly at high light for all picocyanobacterial isolates, with the DVChl-a cell^-1 of Prochlorococcus being up to six times lower than the Chl-a cell^-1 of Synechococcus. A molecular study by Mella-Flores et al. (2012) revealed that Synechococcus WH8103 possess more genes involved in photoprotection and the dissipation of excess energy as heat than Prochlorococcus PCC9511 (very close to the MED4 strain), and that there was a close correlation between expression levels of these genes and diel variations of growth irradiances for Synechococcus which was not always the case for Prochlorococcus. The study also suggested that, rather than intensively inducing a variety of photoprotective mechanisms (including the production of ZEA) as observed in Synechococcus, high light-acclimated Prochlorococcus reduces their energy demand by temporarily down-regulating the synthesis of light-harvesting pigment proteins, ATP formation, and the light-independent reactions of photosynthesis. This is not only to avoid photodamage during a period of high irradiance, but also allows Prochlorococcus cells to generate the energy from photosynthesis that is sufficient to satisfy their minimized energy requirement despite having only low amounts of PSPs to carry out light absorption for photochemistry (Hartmann et al., 2014).

The large dynamic range of DVChl-a cell^-1 values observed in our dataset appears to be indicative of the ability of Prochlorococcus to acclimate to a broad range of light conditions (Partensky et al., 1999). Yet, culture experiments by Moore et al. (1995) on Prochlorococcus strains MED4 and SS120 suggested that plasticity in the intracellular content of DVChl-a in different ecotypes may be restricted to a range of light levels not greater than ~2.5 orders of magnitude (Moore et al., 1995; Partensky et al., 1999). The high dynamic range in the DVChl-a cell^-1 in the field data can be explained by the combined result of both photoacclimation and a vertical shift in dominant (HL and LL) ecotypes, given the limited range of intracellular pigment content observed in cultures.

Figure 7A shows that data points tended to separate into two clusters: one above the line of best fit corresponding to lower cell-specific DVChl-a concentration, and another below the line of best fit which corresponded to samples exhibiting higher DVChl-a content per cell. These clusters represent two distinct populations of Prochlorococcus cells occupying different vertical niches in the water column. The occurrence of a double or bimodal population is a common feature in flow cytometric studies that have examined the intensity of cellular fluorescence in the red due to DVChl-a (Campbell et al., 1994; Blanchot and Rodier, 1996; Thompson et al., 2018). Dim populations have a lower intracellular concentration of DVChl-a and thus lower cellular fluorescence, and tend to be found within the surface mixed layer. On the other hand, bright populations have markedly higher DVChl-a cell^-1 and thus a much stronger fluorescence signal, and typically dominate near the bottom of the euphotic zone. Previous studies (Moore et al., 1998; Moore and Chisholm, 1999) on cultured isolates distinguished the dim and bright Prochlorococcus as HL- and LL-adapted ecotypes, respectively, according to the observation that the dim isolates have lower ratios of DVChl-b to DVChl-a (B/A ratios) than the bright ones over a range of growth irradiances. Molecular studies (Partensky and Garczarek, 2010 and references therein) later confirmed that the two ecotypes differentiated by B/A ratios are, in several ways, genetically distinct.

As illustrated in Figure 7B, ecotype differentiation in Prochlorococcus could account for variation in the intracellular DVChl-a concentration measured in this study. In the absence of genetic information for most of our datasets, B/A ratios were used instead to distinguish ecotypes in Prochlorococcus assemblages, with an increase in B/A ratios indicating a shift from HL ecotype-dominated populations toward populations dominated by LL-adapted strains. Note that only a subset of the HPLC dataset could be used for this analysis since for some datasets, DVChl-b was not separated from monovinyl Chl-b.

Figures 7A, B show a clear separation of HL (lower B/A ratios) and LL ecotypes (higher B/A ratios) based on their DVChl-a cell^-1, highlighting how the two Prochlorococcus ecotypes are readily distinguished from each other on the basis of their intracellular pigment concentration. Note, however, that this rather crude pigment criterion for classifying HL and LL ecotype-dominated populations does not reflect the fact that both ecotypes can co-exist. A study by Thompson et al. (2018) conducted at station ALOHA reported the presence of the LL-adapted strain NATL (with relatively high B/A ratios; Rocap et al., 2002) in the surface Prochlorococcus populations characterized by low intracellular chlorophyll concentration. This was likely to be the result of deep vertical mixing of the water column that brought LL ecotype into the surface layer (Bouman et al., 2006; Thompson et al., 2018). The physiological and genetic characteristics that are intermediate between HL and LL ecotypes (Partensky and Garczarek, 2010) and the ability to be better acclimated to high light relative to other members of the same LL ecotype (Rocap et al., 2002; Malmstrom et al., 2010) allow the Prochlorococcus NATL to occupy a larger depth range than other clades (Partensky and Garczarek, 2010).

To examine the impact of shifts in ecotype diversity on variation in DVChl-a cell^-1 requires that ecotype abundances were measured using molecular methods for each sample. Since this information was not available for many of the datasets, the B/A ratios were used to assist in the identification of certain Prochlorococcus strains. Comparative culture experiments by Partensky et al. (1993) on the isolates MED4 (HLI), SARG (LLI), and NATL (LLI) grown at different irradiance levels (~7.5-133.0 μmol quanta (white light) m^-2 s^-1) showed that the relationship between intracellular concentration of DVChl-a and DVChl-b (Figure 2 in Partensky et al., 1993) for the MED4 strain had a regression slope that was markedly different from those for SARG and NATL1 strains. Also, although B/A ratios for all strains exhibited a general decreasing trend with increasing growth irradiances, the B/A ratio of the MED4 strain was always less than 0.14, while those of the SARG and NATL1 strains were never found to fall below this value. The range of B/A ratios of the MED4 isolate reported in Partensky et al. (1993) is comparable to the results of Moore and colleagues (1995) who observed the range between 0.05 and 0.20 (growth irradiance >10-400 μmol quanta (white light) m^-2 s^-1), and of Rocap et al. (2002) who reported that the MED4 strain acclimating to low light at 18 μmol quanta (white light) m^-2 s^-1) (~1% of typical surface irradiance at subtropical gyres) had a B/A ratio of 0.20, which is the lowest among the 32 Prochlorococcus strains examined.

Based on the culture study of Partensky et al. (1993), Prochlorococcus samples which were likely to be dominated by the MED4 strain (B/A ratio< 0.14) were identified. As shown in Figure 7B (green outline circles), populations dominated by the MED4-like cells exhibited very low intracellular concentrations of DVChl-a, typical of HLI ecotypes, although the values were somewhat variable. Vertical profiles (data not shown) showed that DVChl-a cell^-1 for samples identified as resembling Prochlorococcus HLI populations increased from ca 0.1 near the well-lit surface to up to 1.0 fg DVChl-a cell^-1 below the euphotic depth (z/z_e > 1). This increase was similar to the range previously observed in culture studies (Moore et al., 1995; Moore and Chisholm, 1999) and suggested photoacclimation in this ecotype.

Thus, the results presented in this study show that variability in DVChl-a cell^-1 is likely to be the result of (1) changes in vertical distribution of genetically distinct HL and LL-adapted Prochlorococcus ecotypes, (2) photoacclimation of each ecotype to the local light environment through regulation of the intracellular concentration of the major light-harvesting pigment. The apparent link between genetic diversity and the inherent photophysiological characteristics of Prochlorococcus provides important insight to its ecological success (Johnson and Lin, 2009). Prochlorococcus is able to occupy a broad range of ocean niches from the sea surface to the bottom of the euphotic zone, and as a result, is a significant contributor to global marine primary production (Lange et al., 2018; Visintini et al., 2021). Because of its small and highly-sequenced genome, unique pigmentation, and its rapid identification based on its unique optical signature, Prochlorococcus has been widely used as a model organism for understanding relationships between environmental forcing and microbial processes across vast scales of organization (Partensky and Garczarek, 2010). However, several studies have pointed to differences in photoprotective mechanisms among picocyanobacterial taxa as well as when compared with eukaryotic phytoplankton (e.g. Key et al., 2010; Kulk et al., 2011; Mella-Flores et al., 2012). The question then arises whether Prochlorococcus can serve as an appropriate model for elucidating phytoplankton photoacclimation in general. In other words, are the photoacclimatory characteristics of Prochlorococcus similar to, or different from, those of eukaryotic species, including photosynthetic picoeukaryotes? Since, in many global ocean biogeochemical models, Prochlorococcus is normally classified together with Synechococcus and picoeukaryotes as a ‘pico-autotroph’ functional type (Le Quéré et al., 2005) based on their shared characteristics of being highly effective in nutrient and photon acquisition (Raven, 1998; Le Quéré et al., 2005), it is important to be able to compare photoacclimatory properties of picophytoplankton groups so that it can be justified, from a physiological point of view, whether it is appropriate to combine them into a common biogeochemical functional type.

The carbon-to-Chl-a ratio (θ) of phytoplankton is widely used in phytoplankton growth models to represent the photophysiological status of microalgae and to convert Chl-a concentration to estimates of carbon standing stock (Geider, 1987; Geider et al., 1996; Geider et al., 1997; Sathyendranath et al., 2009; Sathyendranath et al., 2020). Studies (Geider, 1987; MacIntyre et al., 2002) have shown that, for various phytoplankton species grown under constant temperature and nutrient-replete conditions, θ increases linearly with increasing light level. Light-dependent variability in θ is the result of photophysiological mechanisms used by the cell to avoid the harmful overexcitation of photosystems under high light by lowering the cellular content of Chl-a and accumulating more carbon reflecting the increase in the photosynthetic utilization of the absorbed light energy. However, when cultures are maintained under constant light levels, θ decreases significantly with increased temperature, suggesting that high temperature may inhibit phytoplankton growth and thus reduce the demand for energy in the form of fixed carbon (Geider et al., 1997). Variation in θ is also regulated strongly by the availability of nutrients, particularly nitrogen which is a major component of the light-harvesting apparatus (Halsey and Jones, 2015). In addition, since the cellular quota of Chl-a differ widely between phytoplankton taxa (MacIntyre et al., 2002), it can be expected that θ also varies across taxa.

The choices of the carbon conversion factor used to convert phytoplankton cell density to carbon biomass could introduce unconstrained variance in our estimates of θ_Pro. The conversion factors used in this study were either directly measured on cultures (Cailliau et al., 1996) or on oligotrophic populations (Casey et al., 2013). In a different approach, Grob et al. (2007a) independently derived the cellular carbon content of Prochlorococcus from total particle beam attenuation coefficient measurements conducted in the South Pacific Subtropical Gyre and obtained a much lower value of 29 fg C cell^-1 which is in line with that determined by Zubkov et al. (2000) based on cell volume (by size fractionating filtration) on field samples. Later, Zubkov (2014) carried out alternative measurements of cellular biomass using the ¹⁴CO₂-tracer method and acquired a considerably low value of 11 fg C cell^-1 for surface Prochlorococcus populations in the equatorial and temperate South Atlantic. This suggests that there is no universal cell-specific carbon conversion factor (Veldhuis and Kraay, 2004) and shows that such conversion factors vary widely depending on the method as well as the natural and experimental growth conditions of the cells. Attempts have been made, particularly in culture studies on axenic strains (Bertilsson et al., 2003), to refine taxon-specific conversion factors to aid the assessment of the contribution of the phytoplankton standing stock to the ocean carbon pool. Independent information including the genome size and optical properties (Veldhuis and Kraay, 2004) will also help to improve phytoplankton carbon conversion factors for use in marine biogeochemical studies. As the divergence between clusters of the surface and deep populations lessens when applying a higher conversion factor to deeper samples (Figure 8B) and as the explained amount of variance increases at the same time, we feel that further effort is required to measure the carbon content on natural Prochlorococcus populations in general, and within and below the deep chlorophyll maximum, in particular. Differences in the size distribution of surface and deep Prochlorococcus populations suggest that applying conversion factors based on biovolume may help address the problem of converting cell abundances into carbon standing stocks (Casey et al., 2013).

The high and variable values of θ_Pro observed in this study clearly demonstrate that Prochlorococcus has tremendous physiological plasticity to acclimate to the wide range of environmental conditions in which it lives. In general, picocyanobacteria tend to have higher and more variable θ than eukaryotic micro- and nanophytoplankton (i.e. diatoms, dinoflagellates, and nano-prymnesiophytes (Veldhuis and Kraay, 2004; Sathyendranath et al., 2009). Larger eukaryotes, notably diatoms which are known to be successful in fluctuating light environments (more deeply mixed waters) (Claustre et al., 1994), have been shown to be intrinsically less sensitive to photodamage not only owing to their large cell size and corresponding high pigment packaging (Key et al., 2010), but also because they are able to maintain their constitutive resource investments in photoprotection under varying light regimes (Talmy et al., 2013; Halsey and Jones, 2015), which in turn is reflected in a relatively stable θ. In contrast, small picocyanobacteria, which are adapted to more stable water columns, can adjust the proportion of cellular resources allocated to light-harvesting apparatus in response to shifts in irradiance (Talmy et al., 2013), resulting in highly variable intracellular concentrations of photosynthetic pigments (Moore et al., 1995) and thus θ. However, within the picoplankton size range, it remains unclear whether cyanobacteria significantly differ from photosynthetic eukaryotes with respect to variation in θ since studies that directly compare the θ values of the two picophytoplankton groups under similar growth conditions are rare (Table 3).

In comparison to picocyanobacteria, picoeukaryotes have been found to be better adapted to grow in ocean habitats characterized by more extreme physical dynamics (Bouman et al., 2011). Laboratory studies (Kulk et al., 2011; Stawiarski et al., 2018) have demonstrated that photosynthetic picoeukaryotes are able to maintain the high growth rates over a wide range of irradiances and are less affected by photoinhibition at high light. Therefore, it is reasonable to anticipate that our estimates of θ_Pro should be higher and more variable than those of picoeukaryotes. This expectation is consistent with the findings of Veldhuis and Kraay (2004) who examined flow cytometry samples from oligotrophic subtropical waters of the North Atlantic using an independent approach for converting cellular DNA content to phytoplankton carbon biomass. The study showed that θ_Pro decreased substantially from a maximum of 450 g C (g DVChl-a)^-1 at the surface to a minimum of 15 g C (g DVChl-a)^-1 at 150 m, whereas θ of the collective pico- and nanoeukaryote communities also decreased with depth, although to a much lesser extent (3-7 fold variation) with surface values of 30-80 g C (g Chl-a)^-1. The reduced vertical variation in θ observed for photosynthetic eukaryotes was likely due in part to a co-variation of cell pigment content and size as reflected by the increases in both the intracellular concentration of Chl-a and the genome size (proportional to cell size) with depth. A culture experiment by Kulk et al. (2011) also showed that θ_Pro varied from 14 ± 1.27 to 43 ± 1.17 g C (g DVChl-a)^-1 under variable light intensities (50-125 μmol quanta m^-2 s^-1), while the observed range for the picoeukaryote Ostreococcus sp. was only between 39 ± 1.89 and 40 ± 0.99 g C (g DVChl-a)^-1. Determining whether the observed differences in the overall variability in θ between picophytoplankton populations reported here is the result of changes in the inherent difference in photoadaption or an artifact of not accounting for shifts in cell size within natural populations should be a focus of future research as it will allow us to leverage the wealth of global data on their abundance and pigment markers in order to better represent these globally-significant primary producers in the ocean carbon cycle.