Temperature regulates Synechococcus population dynamics seasonally and across the continental shelf

Hourly, year‐round flow cytometry has made it possible to relate seasonal environmental variability to the population dynamics of the smallest, most abundant phytoplankton on the Northeast US Shelf. To evaluate whether the insights from these data extend to Synechococcus farther from shore, we analyze flow cytometry measurements made continuously from the underway systems on 21 cruises traveling between the Martha's Vineyard Coastal Observatory (MVCO) and the continental shelf break. We describe how seasonal patterns in Synechococcus, which have been documented in detail at MVCO, occur across the region with subtle variation. We find that the underlying relationship between temperature and division rate is consistent across the shelf and can explain much of the observed spatial variability in concentration. Connecting individual cell properties to annual and regional patterns in environmental conditions, these results demonstrate the value of autonomous monitoring and create an improved picture of picophytoplankton dynamics within an economically important ecosystem.

The distribution of Synechococcus, a globally important phytoplankter, varies on scales from hours to decades and from microliters to ocean basins.As a result, the drivers of their population dynamics are difficult to identify.The temporal aspect of this challenge is addressed by autonomous sampling at long-term observatories.On the Northeast US Shelf (NES), for example, a FlowCytobot at Martha's Vineyard Coastal Observatory (MVCO) has been taking measurements of individual phytoplankton at 20-min intervals for 20 yr (Olson et al. 2003).This time series has been a valuable resource for studying picoplankton ecology (Sosik et al. 2003;Hunter-Cevera et al. 2014, 2016a;Fowler et al. 2020).In particular, it has revealed strong relationships between environmental drivers (temperature and light) and Synechococcus vital rates, which may help predict the ecosystem's response to future change.
When considering such predictions, however, two questions arise.First, is the community that has been studied at MVCO representative of the broader region?And second, do the relationships between environmental conditions and Synechococcus observed over seasonal cycles hold for other types of environmental variability, such as spatial gradients?To answer these questions, one must consider the spatial variability in Synechococcus dynamics and drivers.We analyzed flow cytometry data collected continuously from 21 research cruises between 2018 and 2022 transiting between MVCO and the edge of the continental shelf (Fig. 1; Supporting Information Table S1) to create an improved picture of Synechococcus dynamics in time and space.This analysis extends our knowledge of Synechococcus ecology across an environment that is both heterogeneous and constantly in motion.
Arguably, the most fundamental metric of picoplankton ecology is cell abundance.It is the extraordinary abundance of Synechococcus that earns these small cells their ecological importance.At MVCO, Synechococcus concentrations are low in winter, then grow exponentially during an annual spring bloom, increasing by two to three orders of magnitude (Hunter-Cevera et al. 2019).Relative to larger phytoplankton, Synechococcus concentrations and ranges are expected to increase as a result of ongoing climate change (Flombaum et al. 2013;Henson et al. 2021).The dominance of small phytoplankton is also predicted to reduce the efficiency of energy transfer to large zooplankton (Legendre and Le Fèvre 1991;Schmidt et al. 1998Schmidt et al. , 2020)).When incorporated into regional fisheries models, expected increases in picophytoplankton can decrease catch potential by up to 20% (Cheung et al. 2011).It is therefore critical that we understand the factors that control Synechococcus concentration in economically important regions like the NES.
Broadly speaking, the concentration of Synechococcus is determined by the balance of adding and removing individuals.Reproduction occurs through division, and mortality is predominately from grazing and viral lysis (Evans et al. 2003;Calbet and Landry 2004).Because cell division and loss happen simultaneously (Landry and Hassett 1982), it is impossible to attribute changes in population size to one or the other without independent vital rate estimates.Most methods for measuring vital rates (e.g., dilution series) require that cells be experimentally manipulated through incubations or staining (Landry and Hassett 1982;Binder and Chisholm 1995).In contrast, diel changes in cell optical properties can be measured in situ and translated into division rates that are consistent with the more labor-intensive methods (DuRand 1995;Sosik et al. 2003;Ribalet et al. 2015).Application of this method to picoplankton populations at MVCO has allowed for an unprecedented view of Synechococcus dynamics, producing thousands of daily division rate measurements across a diversity of natural conditions (Hunter-Cevera et al. 2019).Looking across years, Hunter-Cevera et al. (2014) concluded that Synechococcus population growth at MVCO is lightlimited in the fall, temperature-limited in the spring and that the timing of the spring bloom varies interannually with differences in temperature (Hunter-Cevera et al. 2016a).
In addition to large temporal changes in temperature and light conditions, the NES exhibits spatial environmental heterogeneity.The region's hydrography sets up a fortunate natural experiment to compare seasonal and spatial variability in temperature.The waters at MVCO originate from the Arctic and terrestrial sources and flow southwest along the continental shelf (Chapman and Beardsley 1989).They are colder and less saline than the Slope Sea and Gulf Stream waters offshore (Csanady and Hamilton 1988).The front between shelf waters and slope waters creates a spatial gradient that encompasses the range of temperatures observed at MVCO over the seasonal cycle (Fig. 1A).By sampling across the shelf, we have an opportunity to compare in situ dynamics of Synechococcus at nearby locations experiencing different conditions at the same time of year.In this way, we will demonstrate that the relationships observed at MVCO extend across the region, that they lead to spatial variability in Synechococcus annual cycles and, ultimately, inform predictions of coastal Synechococcus dynamics in the future.Our work builds on the growing body of literature connecting individual cell traits to large-scale marine ecology (see, e.g., Barton et al. 2013;Coles et al. 2017;Dutkiewicz et al. 2020) and creates a dynamic picture of Synechococcus communities across the NES.

Methods
We used automated flow cytometry to count and size individual picophytoplankton cells in 61,113 samples from 21 NES Long-Term Ecological Research (LTER) cruises between 2018 and 2022 (Supporting Information Table S1).The NES-LTER focal transect extends 150 km along the 70 53 0 W longitude line (Fig. 1A), with water depth increasing gradually from 20 to 200 m, and then to over 1600 m past the shelf break.
We grouped samples geographically based on bottom depth intervals ("Inner Shelf," < 50 m; "Mid Shelf," 50-100 m; "Outer Shelf," 100-500 m; and "Upper Slope," > 500 m; Fig. 1A), and into seasons following Hunter-Cevera et al. ( 2019) (winter, 01 January to 31 March; spring, 01 April to 14 June; summer, 15 June to 14 September; and fall, 15 September to 31 December).Each vessel recorded environmental and meteorological data while underway, including surface seawater temperature, salinity, and incident shortwave solar radiation.Vertical profiles of temperature were periodically collected, from which we calculated the mixed layer depth as the shallowest depth at least 0.2 C below the surface temperature or with a difference from surface sigma-t of at least 0.03.Nutrients were measured in water collected from casts as described in Marrec et al. (2021).
Measurements of particle fluorescence and scattering were taken with an Attune NxT flow cytometer (Thermo Fisher Scientific), modified to automatically sample 0.4 mL of seawater every 2 min from each vessel's underway science seawater flow system.Synechococcus were identified and enumerated on the basis of their scattering, phycoerythrin, and chlorophyll fluorescence signals.Resulting data are publicly available online (Stevens et al. 2023).
Daily division rates were estimated by fitting a matrix model to 24 h of hourly binned cell-size distributions, as described by Hunter-Cevera et al. ( 2014) and modified by Fowler et al. (2020).The model assumes that cell growth is a non-decreasing function of light intensity and that larger cells are more likely to divide than smaller cells in a given 10-min time step.On six early cruises, the flow cytometer was not set up for sufficiently sensitive cell-size measurements to apply this method to Synechococcus.For this reason, our division rate analysis includes 31,810 samples from the remaining 15 cruises (Supporting Information Table S1).Individual Synechococcus cells are binned by their volume and the model parameters are fit, 1 d at a time, to the cell-size and light time series.Because we sampled from a moving vessel, our observations may include multiple Synechococcus communities within a single day.Similar to Ribalet et al. (2015), we therefore fit the model to a sliding 24-h window of inputs in order to examine the sensitivity of our estimates to any transitions in cell-size distribution that arise from the movement of the ship (Supporting Information Figs.S2, S3).We consider each daily division rate to be a summary of the activity of Synechococcus populations through which the ship transited during that day.
We compared Synechococcus dynamics measured on the cruises to those previously described at MVCO (see Hunter-Cevera et al. 2019 for details on observatory data).Because daily division rates at MVCO are estimated from 24-h windows starting at dawn, we focused our analysis of the cruise data on the 31 daily division rates calculated from windows starting at dawn.We fit exponential regressions to compare the temperature-division rate relationships across locations.These curves are defined by the equation where μ is division rate, T is temperature, and a and b are parameters controlling the curve's shape.We also fit this curve using an upper quantile regression to describe the envelope of observed values and compare to Eppley (1972) (Supporting Information Fig. S4).

Results
Synechococcus concentration across the shelf follows a seasonal cycle (Fig. 1B-E; Supporting Information Fig. S5), with mean values low in winter (6700 cells mL À1 ) and high in summer (36,000 cells mL À1 ).All values measured on the cruises were within the range of concentrations that have been observed at MVCO (max.270,000 cells mL À1 ).Synechococcus concentrations increased with seawater temperature across seasons and spatial gradients (Supporting information Figs.S6-S8).The relationship between temperature and Synechococcus concentration is strikingly similar to that which has been observed at MVCO (Hunter-Cevera et al. 2019; Fig. 2).
On all six winter cruises, concentrations increased with distance from shore (Supporting Information Fig. S8).This gradient became steeper in spring, due to higher offshore concentrations.Notably, the spring cruise with the largest range of concentrations was also the latest in the season (RB1904; between 20 and 139,000 cells mL À1 ; 12-25 May 2019).In summer, the spatial pattern was reversed.Though Synechococcus was abundant everywhere, concentrations were lower on the Upper Slope than they were at MVCO.The five fall cruises in our time series did not exhibit any spatial gradient in Synechococcus concentration (Supporting Information Fig. S8).
Division rates on cruises varied from 0.02 to 1.2 d À1 , with the lowest values measured during spring cruises (mean 0.15 d À1 ) and the highest during summer (mean 0.85 d À1 ; Fig. 3A).The mean division rate offshore in winter was higher than 97% of winter division rates measures at MVCO, with a maximum value of 0.49 d À1 .Mean spring division rate offshore was higher than only 13% of all spring values at MVCO.Division rates in the summer and fall were not systematically different between MVCO and the cruise data (Supporting Information Fig. S9) and exhibit a similarly positive relationship with daily incident radiation (Supporting Information Fig. S10).
Division rate increased with seawater temperature (p ¼ 0:002; Fig. 3E).All division rates measured offshore were within the range of values previously seen at MVCO at the same temperatures (Supporting Information Fig. S11).The best-fit temperature response curves from cruise and MVCO data have similar growth constants (Table 1) and overlapping prediction bounds (Supporting Information Fig. S4).

Patterns in cell concentration
Synechococcus populations across the NES exhibit an annual cycle in concentration similar to the cycle at MVCO (Fig. 1B-E).An argument for establishing MVCO two decades ago was that its location is open to advection and therefore representative of the broader region.For Synechococcus, our results validate this argument.The seasonal dynamics that are so striking at MVCO also occur across the NES.Moreover, the relationship between Synechococcus concentration and temperature is consistent across the NES, regardless of whether the variability arises over time or space (Fig. 2).This consistent community response strengthens and extends the inferences we can draw from MVCO, supporting the prediction that Synechococcus will respond similarly to other sources of temperature variability, including warming due to climate change.
There are slight differences in the annual cycle in Synechococcus concentration (Fig. 1B-E), which are consistent with the spatial environmental gradient.Cruise data from the Inner Shelf consistently include both the lowest concentrations in winter and the highest in summer (Fig. 1A).The interquartile range in concentration decreases with distance from shore across the shelf (Supporting Information Fig. S12), despite increased sampling effort toward the shelf break (9200, 16,000, and 20,000 samples on the Inner, Mid, and Outer Shelf).Although the true annual peaks in concentration may have occurred between cruises, our analysis supports the conclusion that the seasonal cycle in Synechococcus concentration has a larger amplitude nearshore than offshore.Seawater temperature seems to be a main driver of this difference in amplitude.The relationship between temperature and Synechococcus concentration that we have measured is similar to that compiled from global observations (Flombaum et al. 2013).In particular, the slope of the relationship is steep at lower temperatures and saturates at higher temperatures (Fig. 2B).Over the course of a year, temperature varies roughly between 0 C and 20 C at MVCO and between 5 C and 25 C on the Upper Slope.As we might expect from the global analysis, we measured a wider range of concentrations at the colder location.
We also see evidence of a phenological shift across space.The spring increase in Synechococcus concentration occurs earlier offshore than nearshore.In May 2019, we measured over 10 5 Synechococcus cells mL À1 on the Upper Slope while the Inner Shelf still exhibited winter-level concentrations near 10 3 cells mL À1 .Again, the true peaks in concentration likely occurred between cruises.Additional observations in late spring would be useful for filling in details of the offshore seasonality and may change our conclusions.Nonetheless, we repeatedly saw higher concentrations of Synechococcus offshore in the winter and spring, and lower concentrations offshore in the summer, indicating that both the annual spring increase and annual decline begin earlier offshore than nearshore.

Drivers and controls
An earlier spring bloom offshore is consistent with the hypothesis that Synechococcus at MVCO are temperaturelimited in the winter (Hunter-Cevera et al. 2014, 2019).Our division rate measurements provide further support for this theory.Higher temperatures offshore in the winter allow for higher division rates than are observed at MVCO at the same time of year (Fig. 3A).Cell division offshore is able to outpace loss early in the year, leading to higher concentrations of Synechococcus offshore than nearshore in the winter and spring.
We do not, however, see offshore Synechococcus dividing at their maximal rates throughout the year.For a given temperature, division rates measured from cruise samples fall at or below the maximum rates reported at MVCO (Fig. 3E).In spring, the offshore division rates we measured are particularly low relative to at MVCO.This may be an artifact of our small number of spring cruises, or offshore populations may be limited by a factor other than temperature.In summer 2018, dilution series experiments similarly measured a decrease in growth rate across the shelf for the < 10 μm fraction from about 1.5 to 0.9 d À1 (Marrec et al. 2021).
Lower division rates may reflect increased competition offshore, since Prochlorococcus are seen to dominate the cyanobacteria community beyond the continental shelf (Olson et al. 1990).We might then expect offshore nutrients to be drawn down through competition, but nutrient concentrations in the spring are relatively high compared to summer values (Supporting Information Fig. S13).The low spring division rates may nonetheless indicate co-limitation between temperature and nutrients.
Alternatively, offshore populations might be light-limited as a result of vertical mixing.The bottom depth at MVCO is only 15 m, while the mixed layer depth offshore varies seasonally from 2 to 100 m (Fig. 3C).Incident light data suggest that Synechococcus at MVCO are light-limited through the fall (Hunter-Cevera et al. 2019; Supporting Information Fig. S10), yet fall division rates offshore are comparable to those at MVCO despite significant mixing.Thus, our observations do not indicate that deeper mixed layers are directly associated with lower division rates.
We note that a difference in division rate could reflect a difference in taxonomy rather than physiological response to environmental conditions.The Synechococcus community at MVCO is dominated year-round by a single clade (Hunter-Cevera et al. 2016b), which we might expect to grow in conditions across the shelf.Coastal microbial community composition can vary across short distances, however, with both temperature and distance from shore (Wang et al. 2019;Doré et al. 2022).It may be that Synechococcus strains with different reproductive strategies are abundant in the offshore region despite dividing less rapidly.
High division rates offshore early in the year and a more gradual increase to summer rates can explain the earlier spring increase in concentration offshore (Supporting Information Fig. S14).It is important to emphasize that our division rate measurements are independent of cell concentration.Only by measuring both quantities can we say with some confidence that the cross-shelf patterns arise from bottom-up processes.A study in the Northeast Atlantic described a cross-shelf pattern in summer picoplankton concentration (Schmidt et al. 2020).Similar to our summer data, Synechococcus concentration decreased with distance from shore.The authors attributed the pattern to a cross-shelf difference in grazing pressure, although the study did not include measurements of vital rates.Although grazing processes could drive the pattern, lower division rates offshore would also explain it.Our results and those of Marrec et al. (2021) are consistent with this bottom-up explanation.As an alternative example, a recent study of Prochlorococcus found that trophic interactions define the high-latitude limit of their global range, even though it was long believed that their growth at high latitudes was limited by temperature (Follett et al. 2022).Only by measuring rate information can we unambiguously identify drivers of change in cell concentration over time or space.

Limitations
Long-term high-resolution observations have provided us with valuable insights into picoplankton ecology, but these data have also taught us to be cautious.At MVCO, we see extreme short-term variability in vital rates, particularly in summer (e.g., Figure 3D).This variability could be due to local weather, micro-diversity, rapid adaptation, or differences in cells' environments at the microscale.Only through comparison across many years have we gained confidence in the repeatability of the seasonal and diel patterns we observe.The uncertainty around any single measurement is particularly relevant when considering cruise data, as presented here, which by nature are collected over short windows of time, with only a few time points at any specific location.In addition, this analysis only includes observations of Synechococcus populations in the surface waters and so cannot account for depth as an important dimension of spatial environmental variability.In order to have confidence in any claims about spatial variability in Synechococcus dynamics, there is a need for high-resolution and enduring observation.The NES-LTER work we present here is a substantial step toward bridging the relevant spatial and temporal scales.

Conclusions
We have described the seasonal and spatial patterns in Synechococcus concentration from 21 research cruises and calculated in situ estimates of division rate from 31 d.We find that the seasonal population dynamics of the genus are generally consistent across the NES.More importantly, we find that the underlying relationships between vital rates and the environmental variables which drive this seasonality can predict differences in concentration that result from spatial environmental variability.The observations of Synechococcus made at a single nearshore location provide insight into their dynamics across the shelf.In particular, the relationships among temperature, concentration, and division rate are consistent between populations monitored over time at MVCO and across space and time in the cruise data.Synechococcus in warmer waters farther from shore are dividing more rapidly in winter, supporting the prediction of Hunter-Cevera et al. (2016a) that Synechococcus at MVCO are temperature-limited in early spring and will undergo earlier spring blooms as waters warm due to climate change.In concordance with predictions made from global models (Flombaum et al. 2013), our work suggests that increases in temperature will favor Synechococcus growth across the NES.Ongoing NES-LTER monitoring will allow for more comprehensive evaluation of our predictions.For now, our results make a compelling case for the validity of the inferences we can draw from the time series at the coastal observatory and provide a powerful demonstration of the relationships between environmental conditions and population dynamics across space and time.

Fig. 1 .
Fig. 1. (A) Map of the study region with colored points overlaid at the location of every underway flow cytometry sample in our data set (n = 61,113).Circle color indicates the average water temperature of samples binned latitudinally, demonstrating the overall spatial gradient.Land is shaded in gray and the yellow star indicates MVCO.Bathymetry lines drawn in black at 0, 50, 100, and 500 m delineate our four subregions.(B-E) Seasonal progression of Synechococcus concentration in four subregions.In each panel, gray circles indicate daily averages from MVCO time series (Sosik and Olson 2020) and black lines indicate the average at MVCO for each day of year between 2003 and 2018.Triangles are concentrations of Synechococcus from each of the cruise samples, colored according to time of year.A version of (B) to (E) with concentrations on a log scale is available in the Supporting Information (Fig. S1).

Fig. 2 .
Fig. 2. Relationship between seawater temperature and Synechococcus concentration at MVCO (A) and from cruise samples (B).The color of each point indicates time of year.MVCO values are daily averages; cruise data are for each individual sample (2-min intervals).The cruise data reflect the combination of seasonal and spatial variability shown in Fig. 1.

Fig. 3 .
Fig. 3. Drivers of Synechococcus reproduction over the seasonal cycle at MVCO (circles) and measured from cruise data (triangles).Colored points highlight time of year, while light gray points provide comparison to MVCO measurements, which are reported in Hunter-Cevera et al. (2019).(A) Division rate estimates for days starting at dawn at MVCO and from underway cruise data.(B) Average daily temperatures at MVCO and from cruise samples.(C) Mixed layer depth calculated from vertical temperature profiles for cruises; light gray box indicates depth of water column at MVCO.The relationship between division rate and temperature at MVCO (D) and from cruise data (E).Exponential curves are fit to division rates at MVCO (dashed) and from cruise data (solid).Vertical bars on cruise division rate estimates in (A) and (E) indicate the standard deviation of estimates from days starting within a 6-h window of dawn.Horizontal bars in (E) indicate the standard deviation in temperature exhibited within each dawn-to-dawn day plotted.

Table 1 .
Parameter values for Eq. 1 fit to temperature and division rate data from MVCO and cruises.Note that the confidence intervals (CI) are largely influenced by the number of days in each data set: 3185 and 31 for the observatory and cruises respectively.