Picophytoplankton Niche Partitioning in the Warmest Oligotrophic Sea

Pico-sized Synechococcus, Prochlorococcus, and eukaryotes are the dominant photosynthetic organisms in the vast warm and oligotrophic regions of the ocean. In this paper, we aim to characterize the realized niches of the picophytoplanktonic community inhabiting the Red Sea, the warmest oligotrophic sea, which is considered to be a model for the future ocean. We quantify population abundances and environmental variables over several oceanographic surveys, and use stepwise regression, principal-component analysis (PCA), and compositional-data analysis to identify the realized niches of the three picophytoplanktonic groups. Water temperature varied from 21.4 to 32.4°C within the upper 200-m water column, with the warmest waters being found in the South, where nutrients increased. Synechococcus dominated the biomass, contributing 47.6% to the total picophytoplankton biomass, followed by picoeukaryotes (26.4%) and Prochlorococcus (25.9%), whose proportions contrast significantly with those reported in the subtropical ocean, where Prochlorococcus prevails. There were positive and significant relationships between temperature and the three populations, although these were weak for Prochlorococcus (R2 = 0.08) and stronger and steeper for Synechococcus (R2 = 0.57). The three populations centered their maximum abundances (Lorentzian fits) at similar low nutrient values. Synechococcus were centered close to the surface at ≈77% of surface photosynthetically active radiation (PAR) and ≈30.6°C. The picoeukaryotes were centered at lower light (≈6.4% surface PAR) and warm waters (≈30°C). Prochlorococcus was segregated from the surface waters and centered deep at low light (≈3.2% surface PAR). Light and temperature were the most influential factors determining the community composition, with Synechococcus dominating ∼74% of the picophytoplankton biovolume in the warmest (>30°C) waters. In the warm and mesotrophic southern Red Sea, the moderate abundances of picoeukaryotes and Synechococcus suggest increasing competition with nano and microphytoplankton. Our observations agree with predictions of increasing vertical segregation of picophytoplankton communities with future warming and reveal Synechococcus’s significant capacity to adapt to warming.

The genetic, molecular and physiological differences between organisms determine their fundamental niches, defined as the range of environmental conditions under which they are technically capable of living (Hutchinson, 1957). However, it is the realized niche (which is a subset of the fundamental niche) that describes the range of environmental conditions under which species actually live by considering the effects of different ecological interactions such as interspecific competition or predation. For Synechococcus, Prochlorococcus, and some picoeukaryotes, there is genetic and physiological evidence to outline their fundamental niches, as indicated by strong differences in their nutrient requirements (Moore et al., 2002;Bertilsson et al., 2003;Heldal et al., 2003;Zubkov et al., 2003), tolerance to light and ultraviolet radiation (Moore et al., 1995;Llabrés and Agustí, 2006), and thermal performances (Moore et al., 1995;Agawin et al., 1998;Chen et al., 2014). The realized niches for picophytoplankton vs. other taxonomical or functional types of phytoplankton in the ocean have been examined in several studies (Irwin et al., 2012;Brun et al., 2015). Other studies analyzed the niche partitioning of the three picophytoplankton populations (i.e., Synechococcus, Prochlorococcus, and picoeukaryotes) in both oceans and lakes (Winder, 2009;Bouman et al., 2011;Flombaum et al., 2020). In situ temperature, light, nutrients, and water-column mixing have been identified as significant influences upon picophytoplankton niche partitioning (Ferris and Palenik, 1998;Agustí, 2004;Bouman et al., 2011;Pittera et al., 2014).
At present, one of the major environmental drivers stressing organisms and ecosystems is climate change (e.g., Doney et al., 2002;Bijma et al., 2013;Bindoff et al., 2019). Ocean warming is inducing rapid changes in ecosystems (Burrows et al., 2011), and marine organisms are responding by varying their distributions, biomass, and phenology (Edwards and Richardson, 2004;Gregory et al., 2009;Boyce et al., 2010;Hoegh-Guldberg and Bruno, 2010). Phytoplankton is responding to climate change faster than other primary producers, with shifts described in both their distribution and phenology (Poloczanska et al., 2013). Organisms from polar and tropical areas are the most affected by climate change (Tittensor et al., 2010;Wassmann et al., 2011;Thomas et al., 2012). The increase in seawater temperature in the tropical ocean will represent a new niche for the existing species that will need to accommodate such temperatures at their upper thermal limits, as tropical strains show temperature optima close to or lower than the mean in situ temperature (Thomas et al., 2012). This will force poleward shifts in species' thermal niches and can imply a decline in phytoplankton biomass and diversity, which strongly depends upon adaptation processes (Thomas et al., 2012;Bestion et al., 2020).
Recent studies have addressed predictions for the abundance or distribution of picophytoplankton in the future ocean (Morán et al., 2010;Flombaum et al., 2013;Agustí et al., 2019). However, such predictions require a global-change framework that explicitly considers the cascade of effects and feedbacks that affect the response of biota to warming (Acevedo-Trejos et al., 2014;Duarte, 2014;Agustí et al., 2019). This framework must include changes in many relevant ocean processes, in addition to the direct thermal effect upon the organisms themselves, and therefore challenges the use of present realized niches for future ocean predictions, especially if adaptation processes occur. The relatively unexplored Red Sea is considered to be a model for the future ocean due to its elevated water temperatures and strong environmental gradients (Berumen et al., 2019), and it can therefore provide key information to fill gaps in our knowledge.
The Red Sea is a semi-enclosed, subtropical sea that shows characteristically high seawater temperatures, reaching record values of >33 • C at the surface (Chaidez et al., 2017), higher than the maximum seawater temperatures reported in most of the subtropical and tropical oligotrophic areas inhabited by picophytoplankton (Grob et al., 2007;Bouman et al., 2011;Agustí et al., 2019). It is also highly saline and highly deficient in major nutrients such as SiO 4 , PO 4 , and NO 3 (Weikert, 1987). The Red Sea is also warming at a mean rate of 0.17 ± 0.07 • C decade −1 , higher than the global ocean rate, with the northern Red Sea warming at an even faster rate (0.40-0.45 • C decade −1 , Chaidez et al., 2017). Therefore, the Red Sea offers the opportunity to investigate potential acclimatization and realized adaptation to ocean warming, with a large number of studies addressing its coral reef habitats (Sawall and Al-Sofyani, 2015;Ziegler et al., 2019) and symbionts responses (Hume et al., 2016). It also constitutes a unique opportunity to identify adaptation mechanisms, and physiological and molecular responses to warming (Hume et al., 2016;Giomi et al., 2019;Anton et al., 2020a) including how the elevated temperature may affect biological processes and the interactions between organisms (Hume et al., 2016;Silva et al., 2018;Abdulrahman Ashy and Agustí, 2020;Anton et al., 2020b).
The goal of the present study is to characterize and model the realized niches of the Synechococcus, Prochlorococcus, and picoeukaryote populations inhabiting the Red Sea, the warmest of the oligotrophic seas. Characterizing the realized niches of picophytoplankton in this ecosystem can serve as a gauge for projecting ocean niches in the warming oligotrophic ocean. To do so, we measured picophytoplankton abundances and environmental parameters across their latitudinal and vertical range. To establish the relationships among picophytoplankton and the environment, we analyzed the univariate correlations between cell abundances and environmental parameters, ran a multivariate analysis to define the realized niches of the different picophytoplankton groups, and compared our results with previous studies in oligotrophic waters. We also modeled the realized niches obtained for picophytoplankton in the Red Sea by applying a stepwise statistical method and by compositional-data analysis in three dimensions using barycentric plotting.

Sampling and Biogeochemical Measurements
The hydrography of the Red Sea is dominated by a deepwater current which flows continuously southward out of the main basin through the Strait of Bab al-Mandab, while the surface current is wind-driven and varies seasonally (Manasrah et al., 2019). A seasonal cycle of stratification in spring has been observed, with the maintenance of a shallow thermocline in summer, and a subsequent deepening of the thermocline in winter (Manasrah et al., 2019). Chlorophyll a in the Red Sea also follows a distinct seasonality with maximum concentrations during winter and minimum concentrations during summer (Raitsos et al., 2013).
A total of 29 stations were studied during five different cruises on the Red Sea on board the RV Thuwal (Threats, Dust, and CCF-1 cruises) and RV Al-Azizi (CCF-2 and CCF-3) between October 2016 and March 2018 (Table 1 and Figure 1). At each station, we collected seawater samples using an Idronaut or Sea-Bird CTD-rosette system (depending on the vessel) provided with 12-24 10-L Niskin bottles. We performed sampling at several depths (from 4 to 11 depths per station) between the surface (3-5 m) and 200 m, measuring in situ temperature and salinity. The integrated photosynthetically active radiation (PAR) between 400 and 700 nm was recorded with a PAR sensor installed in a BIC Compact radiometer (Biospherical Instruments, San Diego, CA, United States) or a PUV-2500 Profiling UV Radiometer (Biospherical Instruments, San Diego, CA, United States) deployed at each station at solar noon, thereby avoiding shading from the vessel. The diffuse attenuation coefficient (K d ) was estimated as the slope of the linear regression of the natural logarithm of the instantaneous down-welling irradiance (µmol m −2 s −1 ) vs. depth (m), as described by Overmans and Agustí (2019). The percentage of incident PAR (%PAR) at each specific depth used in this study was estimated from the K d values measured at each station.
At each sampling location and depth, we filtered 300-mL samples into Whatman GF/F filters to measure chlorophyll a (Chl a) concentrations following the fluorometric method described by Parsons et al. (1984). We added 7 mL of 90% acetone solution to each filter and kept them refrigerated in the dark for 24 h to extract the pigment. After incubation, we performed sensitive chlorophyll a measurements by considering its specific fluorescence excitation and emission wavelengths using a calibrated Trilogy Laboratory Fluorometer (Turner Designs, Inc.).
To count the cells of picoeukaryotes and cyanobacteria, we collected approximately 1 mL of seawater at each depth. The samples were analyzed fresh immediately after sampling. We identified and determined cell abundances running between 100 and 200 µL of sample on a CyFlow R space (Sysmex Co.) flow cytometer. The CyFlow R space is a high-performance multi-laser flow cytometer equipped with a 200 mW blue laser beam (488 nm) and complemented with a CyFlow R autoloading station to facilitate sample loading and processing. The Flow cytometry techniques are particularly helpful in the study of phytoplankton based on the different cell sizes and fluorescent properties that the photosynthetic pigments confer upon these organisms (Veldhuis and Kraay, 1993). The three groups targeted here are photosynthetic and present chlorophyll a, which produces red fluorescence. Within both cyanobacteria groups, only Synechococcus contains accessory pigments besides chlorophyll a, including the phycobiliprotein phycoerythrin, which emits orange fluorescence. Picoeukaryotes produce higher scattering and red fluorescence signals due to their higher size with respect to cyanobacteria. Thus, the different light-harvesting pigments and sizes of Synechococcus, Prochlorococcus, and picoeukaryotes ( Supplementary Figure 1) result in distinctive combinations of orange (FL2, 590 ± 50 nm) and red (FL3, 675 ± 20 nm) fluorescence and side-scattering (SSC) signals, allowing their identification and quantification (e.g., Crosbie and Furnas, 2001).
To measure nutrient concentrations (NO 3 and PO 4 ), we collected approximately 15 mL of seawater and kept them frozen until analyzed on land using a Segmented Flow Analyzer (SEAL Analytical) following standard autoanalyzer techniques (Hansen and Koroleff, 1999).

Statistical Analysis
We represented the cellular abundances of each group by depth and latitude using contour plots in R (version 3.4.3, reshape2 and stringr packages). To define the individual effect that the specific set of values of each environmental variable measured on site has on the organisms studied we analyzed the univariate relationships between the environmental variables measured (including temperature, salinity, chlorophyll a, %PAR, phosphate and nitrate concentrations, and cellular abundances) for Synechococcus, Prochlorococcus, and picoeukaryote populations. To do so, we performed regression analysis using GraphPad Prism R 7.0a. We applied linear and Lorentzian regressions and selected the best fit based on the lowest AICc index. The Lorentzian fit equation was given as where y is the cellular abundance in log 10 , x is an environmental variable in log 10 (except for temperature and salinity), amplitude is the height of the center of the distribution in y units, center is the x value at the center of the distribution (where the highest cell abundances are found), and width is a measure of the breadth of the distribution (full width at half maximum) in the same units as x.
In order to assess the relationships between cell abundances and the set of environmental variables that characterize their environment we followed different multivariate methods. We applied a principal-component analysis (PCA) considering cellular abundances, depth, latitude, and environmental parameters (temperature, salinity, %PAR, chlorophyll a, nitrate, and phosphate) using JMP R Pro 12.1.0. This exploratory method allowed us to determine the set of linear combinations that explain most of the variance in cellular abundances and therefore illustrate the realized niches for each group. The cellular abundances and environmental parameters other than temperature and salinity were log-transformed prior to analysis. We also modeled the realized niches for each of the three picophytoplankton groups based on the relationships among environmental parameters and cellular abundances using general Stepwise Regression on JMP R . Of the candidate models produced, we selected the best multivariate model by AICc with forward selection.
The abundances of the three groups in each of the different samples were converted to biovolume by assuming spherical cell shapes with diameters 0.6 µm for Prochlorococcus, 0.9 µm for Synechococcus, and 1.4 µm for picoeukaryotes (as defined in Raven, 1998). These were expressed as proportions of the total picophytoplankton community biovolume, and represented by compositional-data analysis on a ternary plot created in R (Ternary package) to illustrate three-dimensional cohabitation between the three groups.

Environmental Variability and Picophytoplankton Community
We covered a latitudinal transect from 17.09 • N (North Farasan Islands) to 27.60 • N (Duba) (Figure 1 and Table 1), sampling a total of 29 open-ocean stations. Seawater temperatures varied between 21.4 and 32.4 • C, reaching maximum values at the superficial waters in the southernmost latitudes toward the end of summer-fall (Figure 2A). Salinity was high overall, ranging from 37.6 to 40.6 psu, with the lowest values found at depths of less than 100 m in southern stations at all the seasons sampled (including end of summer, fall and winter) ( Figure 2B). The Chl a concentration, representing the total phytoplankton biomass, was low, averaging 0.28 (S.E. = ±0.02) µg L −1 , and remaining below 1 µg L −1 except in two samples collected at deep The mean phosphate concentration was 0.24 ± 0.02 µM, with a median of 0.09 µM and an inter-quartile range from 0.04 to 0.25 µM ( Figure 2D). Nitrate concentrations were low; the mean concentration measured was 3.52 ± 0.32 µM, with median 1.08 µM and an inter-quartile range from 0.47 to 4.73 µM ( Figure 2E). The highest nutrient concentrations were measured toward the south, reaching maximal values of 21.7 µM for NO 3 and 3 µM for PO 4 at approximately 80-m depth (Figures 2D,E).
Synechococcus was more abundant in the southernmost stations, while Prochlorococcus reached its highest concentrations at intermediate latitudes (Figures 2G,H). Moreover, Synechococcus predominated closer to the surface, while Prochlorococcus was more abundant below 40 m (Figures 2G,H). The vertical profiles of averaged cell abundances showed the mean Synechococcus abundances remaining high up to a depth of 50 m, at which point they sharply declined (Supplementary Figure 2). The maximum averaged Prochlorococcus concentrations were located at around 70 m depth (Supplementary Figure 2). The mean picoeukaryotes distribution showed two peaks, one at the surface (∼20 m) and other and at a depth of ∼70 m, similar to that found for Prochlorococcus (Supplementary Figure 2), reflecting the contrasting distributions of picoeukaryotes in the southern and northern stations where they were placed upward or downward, respectively, in the water column ( Figure 2I).
By applying regression techniques, we observed a decrease in Synechococcus abundances with increasing latitude (linear fit, Figure 4A and Table 2), a higher abundance of Prochlorococcus at intermediate latitudes, close to 23 ± 0.5 • N (Lorentzian fit, Figure 4B and Table 2), and no significant trend for picoeukaryote abundance ( Figure 4C and Table 2). The Lorentzian fits defined the maximum cellular abundances for Synechococcus at 6.7 ± 9.3 m depth, for Prochlorococcus at 59 ± 8.1 m, and for picoeukaryotes at 33.9 ± 10.4 m (Lorentzian fit, Figures 4D-F). This fit also suggested that the distributions of Synechococcus, Prochlorococcus, and picoeukaryotes have vertical widths ranging from the surface to depths of 64.5 ± 11.8 m, 150 ± 25 m, and 116 ± 23.8 m, respectively (Lorentzian fit, Figures 4D-F).

Niche Partitioning
Our PCA analysis helped to identify environmental links among the three groups (Figure 5). The first two components explained 73.1% of the overall variability (Figure 5). The principal component 1 (53.5% of the variability) showed a positive correlation of the three picophytoplankton populations, together with temperature, %PAR, and chlorophyll a concentration, as well as a negative correlation with depth, nutrients, and salinity ( Figure 5 and Table 3). The principal component 2 (19.6% of the variability) was positively correlated with latitude and salinity, negatively correlated with phosphate, and slightly correlated with nitrate, separating Prochlorococcus (positive) from Synechococcus (negative). The principal component 1 also indicated that nutrients and temperature had opposite effects upon the picophytoplankton populations. We explored the joint effects of nitrate and temperature upon the picophytoplankton-cell distribution using contour plots (Supplementary Figure 3). Under low temperatures  (<26 • C) and high nutrient concentrations, the abundance of Synechococcus cells was lowest, although Prochlorococcus followed by picoeukaryotes showed high values. The distribution was inverted at the highest temperatures >30 • C, where Synechococcus and picoeukaryotes abundances increased at a wider range of nutrient concentrations, including the maximum abundances observed at the highest nutrient values (Supplementary Figures 3A,C), where Prochlorococcus was almost absent, reaching higher abundances at intermediate nitrate concentrations (Supplementary Figure 3B). We also explored the realized niches by considering cellular abundances and environmental parameters via multiple regression. Using a general stepwise-regression technique, we produced a set of candidate models describing picophytoplankton abundances and selected the best models based on the RMSE values calculated per number of terms and the estimated AICc values for each candidate (Supplementary Table 1). The best model for Synechococcus showed an R 2 value of 0.83 and was composed of five terms: the independent effects of temperature and %PAR and the combined effects of temperature with nitrate and %PAR with salinity and with phosphate ( Table 4 and Supplementary Figure 4). For Prochlorococcus, the R 2 value was lower (0.44) and was also defined by five terms: the independent effects of %PAR and nitrate, and the combined effects of temperature with salinity and %PAR with nitrate and with phosphate. The model for picoeukaryotes had an R 2 value of 0.58 and was defined by four terms: the independent effects of salinity and phosphate and the combined effects of temperature with phosphate and %PAR with nitrate. The different terms in these models showed a good fit when considered separately (p < 0.05, Table 4).
The relative biovolumes of the three groups with respect to the total picophytoplankton biovolume were used to describe partitioning of niches and to identify dominance between the three groups. The average biovolume proportions for the three groups were 47.6 ± 2.5%, 25.9 ± 2.1%, and 26.4 ± 1.3% for Synechococcus, Prochlorococcus, and picoeukaryotes, respectively. In the southern Red Sea, the average biovolume proportion of Synechococcus increased to 60.6 ± 4.2% for Synechococcus, decreased to 15.5 ± 3% for Prochlorococcus, and decreased to 23.9 ± 2.3% for picoeukaryotes. For the rest of the basin (>19 • N), the average proportions still indicated the predominance of Synechococcus, representing 41.1 ± 2.9% of the picophytoplankton biomass, although this increased to 31.2 ± 2.7% for Prochlorococcus and 27.7 ± 1.5% for picoeukaryotes. The proportion of Synechococcus biovolume decreased with salinity at a rate of -25.1 ± 3% per psu and increased with temperature at a rate of 7.6 ± 0.8% per • C, FIGURE 5 | Principal-component analysis (PCA) of the first two components, summarizing the distribution of Synechococcus, Prochlorococcus, and picoeukaryotes with respect to geographical and environmental parameters (latitude, depth, temperature, salinity, chlorophyll a, %PAR, NO 3 , and PO 4 ) in the Red Sea. Principal component 1 explained 53.5% of the variability and showed a positive correlation with picophytoplankton abundances, temperature, %PAR, and chlorophyll a concentration, as well as a negative correlation with depth, nutrient concentration, and salinity. Principal component 2 (19.6% of variability) was positively correlated with latitude, salinity, and Prochlorococcus abundance and negatively with nutrients and Synechococcus abundance.

DISCUSSION
Our study characterized the realized niches of Synechococcus, Prochlorococcus, and picoeukaryote populations inhabiting the Red Sea (i.e., the warmest of the oligotrophic seas), where seawater temperatures often exceed 30 • C, matching future projected temperatures for the subtropical ocean. The results indicated that, in the warmest waters (>30 • C), Synechococcus populations dominated the picophytoplankton community with an average biovolume proportion of 73.6%, while the average proportion of Prochlorococcus and picoeukaryotes decreased to 7.5 and 18.8%, respectively. Seawater temperature and light, followed by nutrient concentrations and salinity, defined the niche partitioning of the picophytoplankton community in the Red Sea, resulting in a strong vertical segregation of the populations.
The picoeukaryotes in the Red Sea belong to several taxa, including members of the classes Mamiellophyceae and Prasinophyceae (Pearman et al., 2017). During their work, TABLE 3 | Loading values for each factor (latitude, depth, environmental parameters, and cellular abundances) and the first two components of the PCA depicted in Figure 5.   Pearman et al. (2017) observed different distribution preferences between picoeukaryotic groups, with the genus Ostreococcus thriving in the north basin, the prasinophyte clade VII in the central basin, and an increased abundance of genera Bathycoccus and Micromonas in the south. This taxonomic diversity may allow niche plasticity for the picoeukaryotic group (Kirkham et al., 2013;Choi et al., 2016). The Synechococcus population is largely dominated by clade II in the Red Sea (Fuller et al., 2003(Fuller et al., , 2005, which is also very abundant in the Indian Ocean (Zwirglmaier et al., 2008). For Prochlorococcus, previous studies identified the dominant HL at the surface and LL in deeper water (Shibl et al., 2014(Shibl et al., , 2016, as observed elsewhere (e.g., Moore et al., 1998); however, few studies have yet been conducted on the taxonomic diversity of picocyanobacteria along the Red Sea (Veldhuis and Kraay, 1993;Acosta et al., 2013). The picophytoplankton abundances found in the Red Sea during this study were similar to those reported in previous studies in the Gulf of Aqaba and the northern Red Sea (Lindell and Post, 1995;Fuller et al., 2005;Stambler, 2005), and along the coast of the Central Red Sea (Al-Otaibi et al., 2020). However, the abundances of the three populations differed strongly from the proportions reported for the oligotrophic ocean: Prochlorococcus is the prevalent picophytoplankton population in the oligotrophic subtropical gyres (Li, 1995;Grob et al., 2007;Bouman et al., 2011;Casey et al., 2013;Agustí et al., 2019), and it was also reported to predominate at more northerly Red Sea latitudes than those studied here, only during summertime (Lindell and Post, 1995). However, we observed that in the Red Sea basin Synechococcus was the most abundant picophytoplankton population, with a moderate average abundance of 8.2 × 10 3 cells mL −1 . Also, the cellular abundances of Prochlorococcus and picoeukaryotes in the Red Sea basin were an average of 19 and 2.6 times lower, respectively, than the mean cellular abundance reported for the global tropical and subtropical oceans , indicating a strong change in picophytoplankton community composition, as well as a reduction in the total picophytoplankton abundance. The prevalence of Synechococcus in the Red Sea basin found in this study is in agreement with qPCR data by Pearman et al. (2017) describing that, with the exception of the northern region in the spring, Synechococcus reads dominated the cyanobacteria in the Red Sea. Therefore, Synechococcus was the prevalent picophytoplankton within the Red Sea area studied, with the highest proportions found at the southern Red Sea.
In terms of biovolume, Synechococcus dominated the picoplankton community in the Red Sea (47.6%), in contrast with a lower proportion of picoeukaryotes (26.4%) and Prochlorococcus (25.9%). For the Pacific Ocean, the contribution of Prochlorococcus varied from 69.3% in the warm oligotrophic area to 58.7% in the equatorial upwelling (Blanchot et al., 2001), with picoeukaryotes accounting for 35% on average and Synechococcus for 7% (Blanchot et al., 2001). In the Central North Atlantic Ocean, Li (1995) found a dominant contribution of picoeukaryotes to total picophytoplankton biomass. The proportions found in the Red Sea are also far from the averages reported for the subtropical Indian Ocean, where Prochlorococcus biomass predominated (Baer et al., 2018), and the contributions of each group to biomass were 69, 19, and FIGURE 6 | Ternary plot showing the biovolume proportions of the three picophytoplankton groups. The triangular regions at the corners represents the biovolume proportions >66% for picoeukaryotes (darkest blue), Synechococcus (darkest yellow), and Prochlorococcus (darkest red). The intermediate-colored areas correspond to biovolume proportions between 33 and 66% (intermediate light colors), and the regions opposite each corner represent 0-33% for the corresponding group. Different symbols represent the three k-means clusters, where circles correspond to cluster A, for the highest average biovolume proportions for Synechococcus (p < 0.05, 80.2 ± 1.6%) and the lowest for picoeukaryotes (p < 0.05, 14.2 ± 1.2%); triangles correspond to cluster B, for the highest average Prochlorococcus-biovolume proportion (p < 0.05, 60.3 ± 1.4%) and the lowest for Synechococcus (p < 0.05, 11.7 ± 1.7%); and crosses correspond to cluster C, for the highest picoeukaryote proportions of the three clusters (p < 0.05, 50.5 ± 1.8%).
12% for Prochlorococcus, Synechococcus, and picoeukaryotes, respectively . In the warm Red Sea between the surface and a depth of 50 m, Synechococcus represented a higher portion of total picophytoplankton biovolume (65.8%), by contrast with the predominance of Prochlorococcus found at similar depth intervals in other oligotrophic sites in the Atlantic and Pacific Oceans (Campbell and Vaulot, 1993;Partensky et al., 1996;Simmons et al., 2016). Therefore, both in terms of cell abundance and biovolume Synechococcus was the prevalent picophytoplankton within the Red Sea area studied, with the highest proportions found at the southern Red Sea surface.

Univariate Relationships of Picophytoplankton to Environmental Parameters in the Red Sea
The three Red Sea picophytoplankton communities were positively correlated with %PAR. However, Synechococcus showed the strongest relationship and dominated the surface waters under high light intensities. The slope of the relationship (0.52) doubled that observed for the global oligotrophic subtropical ocean (slope = 0.28, Agustí et al., 2019), as well as that reported for the oligotrophic Atlantic Ocean (slope = 0.24, Partensky et al., 1996), indicating that high light levels in the Red Sea had a stronger positive effect upon Synechococcus populations. For Prochlorococcus, the correlation was also positive (0.24), but smoother than that for Synechococcus, as observed in the Atlantic Ocean (Partensky et al., 1996, slope = 0.16) and in the global tropical and subtropical ocean , slope = 0.2). Our observations agree with previous studies indicating that Synechococcus is limited by low irradiance and is less sensitive to high light levels than Prochlorococcus (Moore et al., 1995;Mella-Flores et al., 2012). Moreover, the high transparency of the Red Sea allows deep penetration of potentially harmful ultraviolet radiation (Overmans and Agustí, 2019), to which Prochlorococcus shows a high sensitivity (Llabrés and Agustí, 2006;Mella-Flores et al., 2012). Previous studies with natural populations from the Atlantic Ocean (Llabrés and Agustí, 2006;Agustí and Llabrés, 2007) and the Mediterranean Sea (Llabrés et al., 2010) confirmed that Synechococcus could be highly resistant to ultraviolet-B radiation.
The Red Sea waters studied here exceeded the temperatures of other oligotrophic open waters by including values above 30 • C. Red Sea picophytoplankton abundances showed positive correlations with increasing temperature, although with different slopes of 0.29, 0.07, and 0.11 for Synechococcus, Prochlorococcus, and picoeukaryotes, respectively. For Synechococcus and picoeukaryotes in our study, the effect of temperature upon cellular abundance was steeper than that found in tropical and subtropical gyres (slopes = 0.03 and 0.01, respectively; Agustí et al., 2019). For Prochlorococcus, the slope observed here is similar to that reported in other studies from the Atlantic Ocean (slope = 0.03, Jameson et al., 2010) and for the global tropical and subtropical ocean (slope = 0.04, Agustí et al., 2019). Our observations from the Red Sea suggest that Synechococcus would thrive at high temperatures above 30 • C, and that Prochlorococcus would be segregated to colder waters following its thermal niche.
Prochlorococcus covered a wide salinity range of 38.3-40.7 psu, and was present in areas of the northern Red Sea with salinities ≥40 psu, where the Synechococcus and picoeukaryote concentrations were reduced. This association of Prochlorococcus with higher salinity agrees with results from the Mediterranean Sea, where and increasing viability of Prochlorococcus with increasing salinity was reported across the basin (Lasternas et al., 2010). It also agrees with other studies from continental margins in the Atlantic Ocean, where Prochlorococcus was identified as a tracer of saline currents and water masses (Calvo-Díaz et al., 2004;Ribeiro et al., 2016).
The univariate responses of Red Sea picophytoplankton to increased nutrient concentrations obtained here were negative, in contrast with the global oligotrophic ocean, where increased nutrients led to negative correlations with Prochlorococcus, but positive for Synechococcus and picoeukaryotes . For picoeukaryotes, however, the effect of increased nutrients observed here was less negative than that for cyanobacteria. Overall, the relationships obtained here between temperature and nutrient concentrations are similar to those reported by Shibl et al. (2016) for a similar latitudinal sampling along the Red Sea basin. Nutrient concentrations in natural ecosystems are usually negatively correlated with seawater temperature, making it difficult to discriminate the effects of each factor upon the picophytoplankton distribution and its responses to temperature (e.g., Li, 1998;Agawin et al., 2000;Agustí et al., 2019). Despite Red Sea waters being considered highly deficient in major nutrients (Weikert, 1987), the intrusion of warm and nutrient-rich water from the Gulf of Aden at the Southern Red Sea leads to latitudinal gradients in nutrient concentrations along the basin (Churchill et al., 2014). However, we did not find differences in the slopes of NO 3 vs. temperature when comparing waters from the southern and northern Red Sea; rather, the concentration remained relatively high, even at northerly latitudes. However, increased PO 4 concentrations in the south led to a smoother decline with temperature when compared to stations at higher latitudes. In fact, the negative slopes found here between both nitrate and phosphate and temperature were two times larger than those reported for the global tropical and subtropical ocean . This may be due to the smaller temperature range observed in the water column in the Red Sea.
Applying a bell-shaped fit to univariate regressions between cell abundances and an environmental factor can allow us to estimate an optimal value and approximate limits. Amongst other non-linear fits, the Lorentzian model has been proven to be a good fit in previous studies, e.g., to assess the relationship between plankton biomass and diversity and several hydrographic features (Baranyi et al., 2002;Breton et al., 2006;Török et al., 2016;Várbíró et al., 2017). With respect to %PAR, the maximum abundance of Synechococcus was centered at 77% PAR in the Red Sea, while those of Prochlorococcus and picoeukaryotes were centered at lower %PAR values of 3.2 and 6.4%, respectively. Similar results were observed for the Atlantic Ocean (Partensky et al., 1996) and the oligotrophic subtropical ocean  for Prochlorococcus and picoeukaryotes. However, the maximum abundance of Synechococcus was located at approximately 32% PAR in the subtropical oceans , highlighting a possible higher tolerance of Synechococcus Red Sea populations to high light conditions. In agreement with our observations, Veldhuis and Kraay (1993) observed a predominance of Synechococcus in surface waters of the southern region of the Red Sea, with Prochlorococcus peaking closer to the DCM. According to our results, at 1% PAR (≈70 m deep), Synechococcus abundance was strongly reduced, and only Prochlorococcus and picoeukaryotes were able to thrive under such low light conditions. This agrees with previous results from the oligotrophic ocean (Partensky et al., 1996;Agustí et al., 2019).
As expected for picophytoplankton, the three groups showed maxima at low nutrient concentrations, without any significant differences between them. Prochlorococcus and the picoeukaryotes displayed wider niches, reaching waters with higher upper nutrient limits, as they were found deeper in the water column. This was also observed in other studies (Xiao et al., 2018). Overall, the optimal nitrate concentrations for the cyanobacteria studied here (0.42-0.6 µM) were higher than those observed in other areas with similar or lower nitrate concentrations, such as the transition area in NE Pacific Ocean (Simmons et al., 2016), or the Brazilian Bight (Ribeiro et al., 2016). The Red Sea picophytoplankton optima for nitrate were slightly below the threshold of 1 µM, which determines the decline in the dominance of picophytoplankton (Agawin et al., 2000). Synechococcus and picoeukaryote cellular abundances increased at southerly latitudes in the Red Sea, as has been previously observed (Kürten et al., 2014), suggesting that Synechococcus and picoeukaryotes responded to the increased nutrient inputs as described elsewhere (e.g., Mackey et al., 2009). Synechococcus populations increased in warm nutrient-enriched coastal waters (Partensky et al., 1999;Phlips et al., 1999) and areas influenced by upwelling, as reported for the North Eastern African upwelling (Alonso-Laita and  and other areas (Zubkov et al., 1998;Blanchot et al., 2001;Ribeiro et al., 2016). This response to increased nutrient availability has also been reported for picoeukaryote groups (Zubkov et al., 1998;Blanchot et al., 2001;Mouriño-Carballido et al., 2016;Ribeiro et al., 2016;Pearman et al., 2017). However, the responses of picoeukaryotes in the Red Sea to nutrient inputs was not as strong as observed elsewhere, probably due to the notable responses of nano-and microphytoplankton in the Southern Red Sea, which accounted for 43 and 27% of the primary production rate in this area, respectively (Kheireddine et al., 2017). In open-ocean, oligotrophic waters, the size distribution of biomass is heavily skewed toward picophytoplankton (Marañón, 2015). But under increased nutrient concentrations, the contribution of small cells to autotrophic biomass is reduced due to increased loss rates (Agawin et al., 2000) and the pressure of more competitive, fast-growing big cells that can take advantage of nutrient inputs (Marañón et al., 2013). In the main Red Sea basin, picophytoplankton accounted for 30-75% of the phytoplankton biomass (Kheireddine et al., 2017); this proportion decreased to 20-40% in the southern Red Sea (Kheireddine et al., 2017), where the intrusion of nutrient-rich waters from the Gulf of Aden favored the growth of mesotrophic groups such as diatoms (Kheireddine et al., 2017;Pearman et al., 2017), confirming the competition and segregation against smaller phytoplankton cells.

Realized Picophytoplankton Niches in the Warmest Sea
Multiple-regression models and PCA analysis showed strong agreement in indicating that temperature and %PAR were the main factors conditioning the distribution of picophytoplankton, followed by (and negatively correlated with) nutrient concentrations and salinity. The strong influence of PAR availability on niche segregation of picophytoplankton in the Red Sea has been observed elsewhere (Agustí, 2004Flombaum et al., 2020). Temperature was also a key parameter affecting the distribution of the three groups on the Red Sea basin. Synechococcus and picoeukaryotes were advantaged at the highest temperatures measured here (>30 • C), unlike Prochlorococcus. We also observed a strong salinity effect that would mark a succession in niche partitioning, with a negative correlation between Synechococcus and picoeukaryote abundances and salinity (also observed in Kürten et al., 2014), while Prochlorococcus was more abundant where salinity increased.
To summarize, our results suggest that the realized niche for Synechococcus in the Red Sea is centered at high PAR (77%), high temperature (center ≈30.6 • C, lower value ≈26.9 • C) and moderate salinity (center ≈38.2 psu). The Prochlorococcus niche, besides being defined by lower %PAR (center ≈3.2 %PAR), was associated with lower temperature and a larger salinity range of 38.3-40.7. The picoeukaryotic group shared the niche with both cyanobacteria, showing a preference for high temperatures (above 30 • C) with Synechococcus, but favored by low %PAR levels close to those observed for Prochlorococcus (center ≈6.4%), although they were more successful in moderate-salinity waters (center ≈38.2 psu).

The Red Sea as a Warm Ocean Scenario
The present features of the Red Sea have been used as a gauge of future ocean scenarios in the specialized literature (Sawall and Al-Sofyani, 2015;Hume et al., 2016;Berumen et al., 2019;Ziegler et al., 2019;Anton et al., 2020a). Understanding the factors affecting the realized niches of picophytoplankton communities in the warm and oligotrophic Red Sea can provide key information for estimating picoplankton distributions in a future ocean in which small phytoplankton will be of increased importance (Morán et al., 2010). Different representativeconcentration-pathway (RCP) scenarios project increased surface temperatures in the global ocean over the late 21st Century, with expected increases from 0.7 to 3 • C (Collins et al., 2013). Considering that, in some areas of the tropical Pacific and Indian Oceans, the long-term mean surface temperatures already surpass 29 • C (Deser et al., 2010;Signorini et al., 2015), the expected warming would mean more ecosystems subjected to temperatures beyond 30 • C, covering the temperature range considered here, which has not been included in previous oligotrophic analyses. Moreover, climate projections estimate that increased temperatures will lead to stronger stratification, expanding oligotrophic areas in both hemispheres (Polovina et al., 2008;Signorini et al., 2015). The Red Sea is also experiencing a high warming rate of 0.17 ± 0.07 • C decade −1 , while the northern Red Sea is warming between 0.40 and 0.45 • C decade −1 (Chaidez et al., 2017).
As observed here, the Synechococcus distribution was strongly linked to warmer, fresher, and well-illuminated surface waters, reaching higher concentrations toward the nutrient-enriched Southern area of the Red Sea, growing under low to moderate nutrient concentrations and temperatures above 30 • C. In the southernmost area, both Synechococcus and picoeukaryotes could grow in warm-water layers closer to the surface, where they could benefit from the increased nutrient concentrations.
Following our results, Synechococcus predominated at temperatures >30 • C, while the niches of Prochlorococcus and less-abundant picoeukaryotes were centered deeper in the water column under lower-%PAR and -temperature conditions. The segregation of Synechococcus at the warmest surface layer in the Red Sea revealed this genus's large capacity for adapting to ocean warming. Toward deeper waters, the percentages of the Prochlorococcus and picoeukaryote biovolumes increased in detriment to Synechococcus. In the southern region of the Red Sea, Veldhuis and Kraay (1993) observed a predominance of Synechococcus in surface waters, with Prochlorococcus peaking closer to the DCM. The descent of Prochlorococcus and picoeukaryotes to deeper layers agrees with predictions pointing to vertical segregation of the picophytoplankton communities with future warming. Models considering changes in several environmental properties associated with warming predict a decrease of populations at the surface and an increase in subsurface blooms for Prochlorococcus and picoeukaryotes in the tropical and subtropical ocean . Vertical migration of phytoplankton to deeper layers is predicted to occur with increased warming as a consequence of the deepening of thermal isolines, forcing organisms to follow their thermal niche to deeper layers, with populations becoming increasingly compressed at the base of the photic layer (Jorda et al., 2020). Adaptation to warming, however, will allow phytoplankton species to continue growing in the warmer conditions, as our results suggest for Red Sea Synechococcus, leading to a marked vertical segregation between this group and Prochlorococcus and picoeukaryotes, that would remain relegated to deeper layers.
In addition to the temperature conditions, the picophytoplankton communities in the Red Sea are influenced by the strong nutrient and salinity gradients that characterize this basin. The warmest waters of the Red Sea are found in the south, which are enriched in nutrients due to interchange and upwelled waters from the Indian Ocean. The influence of mesotrophic waters in the Southern Red Sea resulted in the displacement of picophytoplankton toward the surface, the exclusion of Prochlorococcus, and the increasing dominance of tropical nano-and microphytoplankton species (Kheireddine et al., 2017), whose diatoms are also adapted to the high seawater temperatures (Jin and Agustí, 2018). The larger capacity of Red Sea nano-and microphytoplankton to respond to nutrients was evidenced by nutrient enrichment mesocosms experiments with oligotrophic waters (Pearman et al., 2016), which showed the fast response of Synechococcus during the first days, soon declining due to a viral infection (Coello-Camba et al., 2020), followed by a large bloom dominated by diatoms.
Our observations help to identify the realized niches of picophytoplankton in the unique Red Sea environment, characterized by their overall high temperatures and latitudinal nutrient and salinity gradients. However, a profound comprehension of these niches and their definition in space would require a more thorough spatial and seasonal coverage, and even complementary transplant experiments to identify the range limits of these picophytoplankton populations (Hargreaves et al., 2014). Besides this, in scenarios like the Red Sea, defined by gradients in ecologically important variables, the quality of the habitat and a species' abundance typically decline toward the edges of its distribution range (Holt, 2003;Hargreaves et al., 2014), without sharp boundaries, and therefore the definition of distribution limits is not straightforward. Also, the distribution of a species is a dynamic feature that can be modified by biotic interactions such as predation or competition (Case et al., 2005), dispersal processes and habitat adaptation (Holt, 2003). This way, future scenarios may be conditioned by the species' ability to adapt to the new conditions, with our results highlighting a good adaptation of Red Sea Synechococcus to surface warming conditions, while less adapted Prochlorococcus and picoeukaryotes thrive deeper.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
SA designed the study. AC-C analyzed the data. Both authors obtained the data and wrote the manuscript.

FUNDING
This research was supported by King Abdullah University of Science and Technology through base line funding BAS/1/1072-01-01 to SA and FCC/1/1973-21-01.

ACKNOWLEDGMENTS
We thank the technical personnel of the Coastal and Marine Resources Core Laboratory (CMOR) and the RV Thuwal and RV Al-Azizi crews for their help during sampling.

SUPPLEMENTARY MATERIAL
The  Figure 6. Circles represent Cluster A, triangles Cluster B, and crosses Cluster C.

Supplementary Table 1 | (A)
Candidate models for each picoplankton group, ordered by number of terms, together with their corresponding quality estimators. Gray shading indicates the models selected in our work. (B) Decrease in RMSE with increasing number of terms in the candidate models for each phytoplankton group.