A Key Marine Diazotroph in a Changing Ocean: The Interacting Effects of Temperature, CO2 and Light on the Growth of Trichodesmium erythraeum IMS101

Trichodesmium is a globally important marine diazotroph that accounts for approximately 60 − 80% of marine biological N2 fixation and as such plays a key role in marine N and C cycles. We undertook a comprehensive assessment of how the growth rate of Trichodesmium erythraeum IMS101 was directly affected by the combined interactions of temperature, pCO2 and light intensity. Our key findings were: low pCO2 affected the lower temperature tolerance limit (Tmin) but had no effect on the optimum temperature (Topt) at which growth was maximal or the maximum temperature tolerance limit (Tmax); low pCO2 had a greater effect on the thermal niche width than low-light; the effect of pCO2 on growth rate was more pronounced at suboptimal temperatures than at supraoptimal temperatures; temperature and light had a stronger effect on the photosynthetic efficiency (Fv/Fm) than did CO2; and at Topt, the maximum growth rate increased with increasing CO2, but the initial slope of the growth-irradiance curve was not affected by CO2. In the context of environmental change, our results suggest that the (i) nutrient replete growth rate of Trichodesmium IMS101 would have been severely limited by low pCO2 at the last glacial maximum (LGM), (ii) future increases in pCO2 will increase growth rates in areas where temperature ranges between Tmin to Topt, but will have negligible effect at temperatures between Topt and Tmax, (iii) areal increase of warm surface waters (> 18°C) has allowed the geographic range to increase significantly from the LGM to present and that the range will continue to expand to higher latitudes with continued warming, but (iv) continued global warming may exclude Trichodesmium spp. from some tropical regions by 2100 where temperature exceeds Topt.


Introduction
The ocean is a major sink for anthropogenic emissions [1], of which the capacity to store CO 2 is strongly affected by biological processes [2]. As atmospheric CO 2 increases, the dissolved inorganic carbon (DIC) concentrations in the oceans increases, the pH declines, and the inorganic speciation changes [3]. Such changes are expected to have diverse effects for marine  Trichodesmium has used a single independent variable (e.g. CO 2 whilst keeping temperature and light constant or temperature whilst keeping CO 2 and light constant), of which there are typically 3 to 8 treatments and all cultured for short time periods and often with several undefined growth conditions (Table 1).
To investigate the integrated effects of key physical/chemical variables (temperature, pCO 2 and light) that will be altered by climate change on the growth rate of Trichodesmium, we performed a systematic, multivariable experiment, where Trichodesmium IMS101 was cultured over long durations (> 9 months), at multiple treatments (n = 174) with controlled and defined growth conditions, thus ensuring that balanced growth and complete physiological acclimation was achieved (S1 Table). The aims were to assess the response of Trichodesmium IMS101 growth to temperature, CO 2 and light intensity. We tested the hypotheses that (i) the thermal niche width of Trichodesmium IMS101 is reduced (i.e. T min increased and/or T max decreased) under suboptimal light or pCO 2 , (ii) the optimal temperature (T opt ) for growth is unaffected by light or pCO 2 , (iii) the maximum photochemical efficiency of PSII (F v /F m ) is affected by suboptimal and supraoptimal temperature, light and CO 2 , and (iv) when light is limiting, suboptimal pCO 2 further reduces growth rate.

Experimental setup
Cultures were grown at low volumes (5 mL) in 12 mL glass test tubes. Test tubes were acid washed and autoclaved prior to culturing, and each dilution was made into a new tube to avoid the build-up of contaminants. Growth rates were quantified from changes in fluorescence (F o ) measured daily (between 09:00 to 10:30) on dark-adapted cultures (20 minutes) using a FRRfII Fastact Fluorometer (Chelsea Technologies Group Ltd, UK). The FRRfII parameters were optimised prior to the experiment to ensure a saturating fluorescence curve was achieved for both low (post-dilution) and high (pre-dilution) cell density cultures.
Cultures were kept at the lower section of the exponential growth phase (S1 Fig) and optically thin to avoid nutrient limitation, self-shading and minimise CO 2 drift [32]. Tubes were gently inverted twice a day to minimise trichomes aggregating at the meniscus. Subject to the temperature and CO 2 , high light cultures were usually diluted every fourth to fifth day, while low light cultures every tenth to twelfth day. When F o declined at an extreme growth condition (e.g. high temperature), three attempts were made to re-grow that treatment, using culture from the closest growth condition.

Culture medium and carbonate chemistry
A single batch (25 L) of filter-sterilised (0.25 μm pore) YBCII media [44] was made and stored in acid-washed, autoclaved Duran bottles (no headspace). The inorganic carbon chemistry of each bottle was determined via CO2SYS [45]; using a 15 mL sample for TCO 2 analysis (Shimadzu TOC-V Analyser & ASI-V Autosampler), and a 10 mL sample for pH (Thermo Scientific Orion Ross Ultra pH Electrode EW-05718-75, UK). The pH probes were rinsed and calibrated with fresh (< 2 weeks) artificial seawater buffers (TRIS and AMP) prior to use [46].
Once a culture reached a pre-determined F o value, it was diluted (0.5 mL culture to 4.5 mL media) with filter-sterilised (0.2 μm pore) YBCII media that had been adjusted to a target pH (and thus target CO 2 ) to return the culture to a starting F o value. To obtain a targeted CO 2 concentration in the YBCII media used to dilute the semi-continuous cultures, the medium was bubbled with a CO 2 -air mixture to a targeted pH (precision of ± 0.002). Once the media reached the desired CO 2 concentration (± 1%) it was immediately distributed into the test tubes, already containing the culture (S1 File). Test tubes were sealed via a PTFE lined screw cap and PTFE tape on the test tube threads, ensuring a gas-tight seal and preventing exchange of CO 2 with the atmosphere. Prior to screwing the cap on, the gaseous headspace was flushed with a filtered (0.2 μm pore) standard gas mixture of a target CO 2 concentration (BOC Industrial Gases, UK). All carbon chemistry calculations were made in CO2SYS [45], using the 1 st and 2 nd equilibrium constants (K1 and K2) for carbonic acid [47], the dissociation constant for KSO 4 [48], the boric acid constant (KB) [49], and the total pH scale. The CO2SYS program calculates CO 2 concentrations as μatm; however, as the CO 2 concentrations reported here are to zero decimal places the equivalent units of parts per million are used (ppm).
Prior to every dilution, 3.5 mL of culture was collected in a 5 mL plastic cryogenic vial (Sigma-Aldrich V5257-250EA) and was used to measure the post-culturing pH. Assuming alkalinity remained constant throughout the entire growth phase [50], a post-culturing CO 2 was calculated using the post-culturing pH and initial alkalinity.

Temperature gradient
To measure the effect of temperature on growth, a custom-made water-jacketed aluminium temperature block was used to house test tube cultures. The temperature ranged from 18 to 33˚C, and the temperature steps along the gradient were at~0.5˚C increments. The temperature (± 0.1˚C) of each tube was measured using a temperature probe (Sper Scientific 840038, Arizona USA), and varied < 0.2˚C over a diel period for any given culture along the gradient. A clear Perspex sheet was secured beneath of the block to hold the tubes in place and allow illumination from below. Light was provided by four fan-cooled LED strips (The Optoelectronic Manufacturing Corporation Ltd. 3ft T5 Daylight, UK), and was adjusted using neutral density filters, with half of the test tubes illuminated at low light (40 μmol photons m -2 s -1 ± 0.4), and the remaining tubes at high light (400 μmol photons m -2 s -1 ± 4), all at a 12:12 L:D cycle. The light was measured for each test tube using a light meter (Li-Cor Li-250A, Nebraska USA), and was checked weekly throughout the experiment. For each of the low and high light treatments, and for each temperature treatment, Trichodesmium IMS101 was cultured at three targeted CO 2 concentrations (180, 380 and 720 ppm), giving a total of 120 treatments.

Light gradient
To measure the light response of growth a second temperature block was setup and run concurrently. All test tube cultures were maintained at 26˚C. Using neutral density filters, a light gradient was setup ranging from 10 to 1400 μmol photons m -2 s -1 , at a 12:12 L:D cycle. The light steps along the gradient were distributed in order to resolve the light-limited and saturated sections of the growth curve. For each light treatment, Trichodesmium IMS101 was cultured at three targeted CO 2 concentrations (180, 380 and 720 ppm), giving a total of 54 treatments.
The measured pCO 2 in the cultures just prior to dilution into fresh medium were between 20% and 30% lower than the target pCO 2 concentrations of 180, 380 and 720 ppm, and were similar across all temperatures and light intensities ( Table 2).

Chlorophyll fluorescence
In addition to the minimum fluorescence (F o ), the FastPro software (Chelsea Technologies Group Ltd, UK) generated dark-adapted values of maximum fluorescence (F m ). The maximum photochemical efficiency of PSII in the dark-adapted state (F v /F m ) was calculated using the following: Data processing to obtain balanced growth rates Acclimation took~12 to 16 weeks with up to 20 dilutions required to allow cultures to achieve balanced growth at the extreme temperature and light limits. For each dilution, a growth rate was calculated from linear regression of ln(F o ) versus time. Balanced growth was assumed to have been achieved once growth rates from a minimum of three successive growth curves had stabilised (S2 File, S2 Fig). A script written in the open source statistical software R [51] was used to process and analyse the growth rate data for each treatment [52] (S4 File). This objective approach improved the efficiency of data processing, given the large number of treatments (n = 174) and the duration of the culturing (~9 months) and removed potential bias or subjectivity when determining a growth rate from numerous data points.

Growth rate versus light curve
The growth-light (μ-I) curves were modelled using the following function [53]: where μ max ' is the hypothetical maximum growth rate (d -1 ); α is the initial light-limited slope of the growth-light curve (d -1 (μmol photons m -2 s -1 ) -1 ); β is the parameter that characterises photoinhibition at supraoptimal light intensities (d -1 (μmol m -2 s -1 ) -1 ); I is the light intensity (μmol photons m -2 s -1 ); and I c is the compensation light at which the light-limited growth rate extrapolates to zero (μmol photons m -2 s -1 ). The achieved maximum growth rate (μ max ), light at which growth is maximal (I opt ), lightsaturation parameter (I k ), and light inhibition parameter (I p ) were calculated from the fitted parameters as follows: Growth versus temperature curve The growth-temperature (μ-T) curves were modelled using a newly formulated modified sine function, which returns estimates for T min , T max and the maximum growth rate at T opt : where μ max is the maximum growth rate (d -1 ); T is the temperature of the culture (˚C); T min is the minimum temperature limit for growth (˚C); T max is the maximum temperature limit for growth (˚C); θ is a shape determining parameter, which alters the skewness; and F is a shape determining parameter, which alters the kurtosis. The optimum temperature (T opt ) is calculated from the fitted parameters as follows: The growth-light and growth-temperature curve fits for each CO 2 or light treatment were fitted using a weighted non-linear squares algorithm, where weights were the reciprocals of the standard errors associated with the median growth rates. Standard errors were propagated when parameters (e.g. I k , T opt ) were calculated from curve fit values (S3 Table). Further statistical analysis (Sigmaplot 11.0) was used to assess differences between CO 2 and light treatments. When appropriate the data was log transformed to ensure normality, and a Two or Three-way ANOVA and post hoc Tukey test was applied on all of the growth rate data per light, CO 2 treatment, as opposed to the single calculated curve fitted parameter values.

Results
Growth rates increased with increasing CO 2 at all temperatures (Three-way ANOVA, Tukey post hoc test; P < 0.001). The growth responses were non-symmetrical around the optimum temperature for growth (T opt ranging between 24.7 to 26.9˚C); specifically the effect of pCO 2 on growth was more pronounced at suboptimal than supraoptimal temperatures (Fig 1A and  1B). Growth rates at temperatures below T opt were markedly lower at mid than high pCO 2 , whereas similar rates were observed at both mid and high pCO 2 at temperatures above T opt .
The growth-temperature curves exhibited marked reductions of growth rate (μ) accompanying small changes in temperature near the lower and upper tolerance limits, and smaller rates of change of growth with changes of temperature closer to the optimum (Fig 1A and 1B). With the exception of the high light-low pCO 2 treatment, all other curves exhibited negative skewness (θ > 1) ( Table 3).
The minimum temperature limit for growth (T min ) was affected by both CO 2 and light ( Table 3). Under low light, T min declined from 20.8 to 19.4˚C between low and high CO 2 , whilst at high light the T min declined from 21.3 to 19.7˚C between low and high CO 2 . The maximum temperature limit for growth (T max ) did not significantly vary between most treatments, averaging about 30.7˚C ( Table 3). The temperature niche width (w = Tmax-T min ) was~1.5 to 2˚C smaller at low than at mid and high CO 2 .
There were significant differences in the growth rates between the low and high light treatment (40 versus 400 μmol photons m -2 s -1 ) at all CO 2 treatments (Three-way ANOVA, Tukey post hoc test; P < 0.001). The maximum growth rate (μ max ) at the optimum temperature increased by about 30% (low light) and 70% (high light) as CO 2 increased from low to mid, and by an additional 7% (both low light and high light) from mid to high CO 2 (Table 3).
A more comprehensive assessment of the light-dependence of growth rate was made at the optimum temperature of 26˚C (Fig 1C). The initial slope (α) showed little response to pCO 2 , but the maximum growth rate (μ max ) increased by 65% from about 0.21 d -1 at low CO 2 to 0.33 d -1 at the mid and high CO 2 ( Table 4). The light intensity at which the highest maximum Table 3. The temperature dependent growth curve parameters (± S.E.) for Trichodesmium erythraeum IMS101 generated by fitting a five-parameter function to each of the two light (LL = 40 μmol photons m -2 s -1 ; HL = 400 μmol photons m -2 s -1 ) treatments at the three CO 2 (Low = 180 ppm, Mid = 380 ppm and High = 720 ppm) treatments. Units; μ max (d -1 ), the maximum growth rate; T min (˚C), the minimum temperature limit for growth; T max (˚C), the maximum temperature limit for growth; w (˚C), the thermal niche width for growth; Φ, the peakedness-shape determining factor (higher value = more peaked); θ, the skewness-shape determining factor (< 1 = positive skewness, > 1 = negative skewness). Both shape determining factors, Φ and θ, influence the shape of the curve without modifying μ max , T min or T max .

Parameters
doi:10.1371/journal.pone.0168796.t004 Multiple Stressors on Trichodesmium Growth growth rate occurred (I opt ) also increased with pCO 2 from 290 μmol photons m -2 s -1 at low CO 2 to 434 μmol photons m -2 s -1 at high CO 2 . The light-saturation parameter (I k ) was much lower at low CO 2 in comparison to both mid and high CO 2 treatments (Table 4). The maximum photochemical efficiency of PSII (F v /F m ) was greatest at T opt , and was significantly higher (Three-way ANOVA, Tukey post hoc test; P < 0.001) for low light than high light treatments (Fig 1G and 1H). In addition, F v /F m increased as CO 2 increased from low to mid and high CO 2 (Three-way ANOVA, Tukey post hoc test; P < 0.001) (Fig 1G,H). At low light and for all CO 2 treatments, F v /F m decreased as temperature decreased from T opt to T min , while there was no difference in F v /F m (~0.3) as temperature increased above T opt to T max (Two-way ANOVA, Tukey post hoc test; P < 0.001). Conversely at high light, there was no significant difference of F v /F m at the T min , T opt or T max values between CO 2 treatments (Fig 1H). The maximum F v /F m recorded (~0.35) was at T opt and the lowest light intensity (20 μmol photons m -2 s -1 ) (Fig 1I).

Discussion
Our key findings for Trichodemisum IMS101 are: (i) at T opt , CO 2 affected μ max but not the initial slope (α) of the growth-light curve; (ii) low CO 2 constrained the thermal niche width more than low light; (iii) there was greater divergence in the temperature dependence of growth rate due to differences in CO 2 below T opt than above T opt ; (iv) the maximum photosynthetic efficiency (F v /F m ) was influenced more strongly by varying temperature and light than CO 2 ; and (v) CO 2 affected T min while there was no effect of CO 2 on T max . Here we discuss the physiological mechanisms that may explain these results.

Light and CO 2 dependencies of growth at T opt
The observed increase in μ max with increasing pCO 2 is qualitatively consistent with previous results [32,34], although there are slight quantitative differences in the percentage increase in growth between mid (~380 ppm) and high (~720 ppm) CO 2 treatments.
Previous studies have attributed the low growth rates at low CO 2 to high energy demands required to establish, maintain and operate a carbon concentrating mechanism (CCM), which in turn limits the energy available for N 2 fixation, reducing growth [34,42,43]. Our observations of the effect of CO 2 at light-saturation and optimal temperature support these findings, while the lack of variability in the light-limited initial slopes is inconsistent with this hypothesis because we would expect the initial slope to be lower at low CO 2 if operation of the CCM imposed significant energetic cost. The lack of variation in the initial slopes were perhaps due to low data resolution (too few low-light data points) and the curve fitting process (more heavily weighted to the high-light data points). Consistent with this suggestion is our observation from the thermal response curves that growth rate at low light (40 μmol photons m -2 s -1 ) increased with increasing pCO 2 by about 20% from low to mid CO 2 with a further 10% increase from mid to high CO 2 .
Operation of the CCM is required to maintain high internal CO 2 concentrations within the carboxysome to inhibit photorespiration. Conversely, at high CO 2 concentrations, when the CCM is fully saturated, Trichodesmium spp. can down-regulate CCM activity and up-regulate other cellular processes (e.g. N 2 fixation), which indirectly increases growth [40,42]. Although this is an attractive hypothesis, direct measurements of the magnitude and cost of operating the CCM in Trichodesmium spp. have not been reported. However, the effect on growth rate may be relatively small since the photon requirement for operating the CCM accounts for only 5-15% of the photon requirement for growth (S3 File, S2 Table).
The values of the light-saturation parameters (I k ) were similar to previous observations [21], and exhibited a clear CO 2 response, which was driven largely by the changes in μ max . The photoinhibition parameter (I p ) was higher at high CO 2 than mid CO 2 , indicating an increased capability at alleviating the effects of photoinhibition at high CO 2 , most likely attributed to CCM down-regulation and the up-regulation of photoprotective-related mechanisms. Interestingly, the I p was also higher at low CO 2 than mid CO 2 , which could be attributed to a low CO 2 stress response triggering an up-regulation of photoprotective mechanisms. However, due to the standard error of the modelled slope of photoinhibition (β), the propagated error of the low CO 2 I p value is large, making a difference between the low and mid CO 2 treatment a possible artefact of the curve fit.

Minimum temperature for growth (T min )
The minimum temperature at which Trichodesmium spp. can grow is likely set by the ability to maintain anoxic conditions that are required to prevent nitrogenase inhibition [8]. Unlike heterocystous diazotrophs, Trichodesmium spp. do not possess a glycolipid barrier that prevents or reduces the diffusion of dissolved gases into the cell [8]. To prevent O 2 inhibiting nitrogenase, Trichodesmium spp. uncouple photosynthesis from N 2 fixation both temporally and spatially [54,55], maintains a high mitochondrial respiration rate [8,56] and may also use O 2 scavenging processes (Mehler reaction and hydrogenase activity) to maintain an anoxic state within the diazocytes [34,57,58]. Stal [8] calculated that respiration would not be able to maintain anoxia at T < 17˚C, which is slightly below the T min of 19 to 21˚C that we (Table 3) and others [26] have observed. The higher T min that we observed at low CO 2 may reflect lower availability of substrate for respiration due to lower photosynthesis rate, or it may reflect the higher metabolic demand for operation of a carbon concentrating mechanism, which diverts energy that would otherwise be available to fuel N 2 fixation.

Supra-optimal temperatures for growth (T opt to T max )
The shape of the growth curve from T opt to T max exhibited two distinct sections, where growth steadily decreases to~29˚C, before plummeting to zero at T max . This two-part decline in growth was most pronounced at low light-low CO 2 , which may in part be due to an enhanced CCM activity limiting the efficacy of certain heat stress processes.
The maximum temperature limit for growth (T max ) was similar across all light and CO 2 treatments. The primary targets of thermal damage in vascular plants include the oxygen evolving complex along with the associated cofactors in photosystem II (PSII), the activity of the carboxylating enzyme Rubisco and the ATP generating system [59]. The maintenance of high values of F v /F m at temperatures above T opt under light-limiting conditions indicates that PSII was not damaged by high temperature in these benign low light conditions irrespective of pCO 2 . In contrast, under light-saturated conditions, F v /F m peaked near T opt under all pCO 2 treatments.
Reduced growth rates caused by temperatures exceeding T opt may perhaps be due to the inhibition of Rubisco activity or an increase in the ratio of oxygenation to carboxylation by Rubisco. In comparison to other photosynthetic enzymes, Rubisco has a low turnover rate requiring high concentrations to maintain a sufficient rate of photosynthesis. Rubisco activase has a lower maximum temperature tolerance than Rubisco itself. Therefore, if the temperature exceeds the temperature limit to which an organism is adapted, the activity of Ruisco activase is severely reduced and unable to offset the rate of Rubisco deactivation, reducing or inhibiting photosynthesis [60][61][62]. Previous studies have shown Rubisco activase activity to significantly decrease at~30˚C, with maximum inefficiency occurring at 45˚C [61]. Although the T max values reported here were~31˚C, it is likely that the concentration, substrate affinities and the temperature adaptations of Trichodesmium's Rubisco activase and Rubisco itself would differ from terrestrial plants and other cyanobacteria.
Rubisco in Trichodesmium spp. is characterised by a low affinity for CO 2 relative to ambient CO 2 concentrations [63][64][65]. As temperatures increase, the solubility of CO 2 decreases more rapidly than that of O 2 . As such, increasing temperature favours the oxygenation of Rubisco (photorespiration), which reduces the rate of 3-Phosphoglycerate production and requires significant amounts of energy and reductant to process the NH 3 and potentially enzyme-inhibiting (i.e. 2-P glycolate) by-products. The solubility of CO 2 and O 2 within the carboxysomes is principally governed by temperature and not by light. Despite this there was a low light, low CO 2 integrated effect on T max . A possible explanation for this observation could be that at low CO 2 , the CCM is up-regulated and functioning at maximum efficiency to maintain a sufficient internal CO 2 concentration within the carboxysomes. At low light, photosynthetic rates are reduced, limiting the amount of energy that can be invested into CCM-related proteins. Thus, the combination of low light with low CO 2 may make cells more susceptible to photorespiration, particularly at elevated temperatures where the solubility of CO 2 is reduced.
To recap, we suggest that the observed two-part decline in growth between the T opt and T max values are due to two processes, photorespiration and Rubisco inhibition. The former primarily mediated by the temperature-driven changes to the solubility of O 2 and CO 2 , where the response is compounded by other co-limiting factors (i.e. low CO 2 and light); and the latter solely mediated by the temperature-driven response on enzyme kinetics, and is determined by Trichodesmium's thermal tolerance and is not influence by other abiotic factors.

Potential effects of future climate change on the biogeography of Trichodesmium
Although temperature, CO 2 and light intensity place limits on the potential for Trichodesmium growth, whether this potential is achieved depends on the availability of limiting nutrients (e.g. P, Fe). Trichodesmium spp. are most abundant in regions of the sea that receive high inputs of Fe via deposition of dust transported from arid source regions [66]. This may reflect Trichodesmium's high metalloenzyme inventory, which suggests that iron, instead of phosphorus, may be the key nutrient in constraining Trichodesmium growth and productivity in the present and future oceans [67]. Increases in the supply of Fe to the ocean via increased dust deposition as desertification increases in a warmer climate may also favour Trichodesmium spp. [68][69][70]. However, this advantage may be negated if increases in the dust load occur where elevation of temperature above 26˚C (T opt ) inhibits growth of Trichodesmium.
Climate models indicate that the water column in oligotrophic regions of the ocean will become more stratified as global sea surface temperatures rise due to global warming, decreasing vertical mixing and thereby limiting the supply of new nutrients from the deep ocean [5]. The documented increase of water column transparency (decline of phytoplankton chlorophyll) in low to mid latitude oligotrophic oceans over the past 120 years suggests that phytoplankton abundance has declined. This is presumed to be due to an increase in vertical stratification [71], although this century scale observation does not necessarily appear to explain local changes of chlorophyll and stratification on shorter time scales [72]. Increased water column stratification with lower inorganic N concentrations in the surface mixed layer should give photosynthetic diazotrophs such as Trichodesmium spp. a competitive advantage over other phytoplankton [73]. However, exposure to high light intensities in shallower near surface mixed layers may reduce growth by requiring more energy be used for prevention or repair of damage due to photoxidative stress [74]. Although vertical inorganic P fluxes will also decline as the water columns become more stable, Trichodesmium spp. have several adaptations that allow it to exploit P-limited environments. These include an extracellular hydrolysis of dissolved organic P (DOP) by alkaline phosphatase activity [75], which can provide an additional source of P for growth. Trichodesmium spp. can also upregulate synthesis of proteins associated with high-affinity transport and hydrolysis of phosphonate compounds [76]. To date, this pathway is absent from other sequenced marine cyanobacterial genomes, and thus represents a unique adaption for scavenging and hydrolysing phosphorus compounds from organic sources, and growing in otherwise P-limited regions. In addition, Trichodesmium spp. may be capable of mining phosphate by increasing their density by carbohydrate production facilitating sinking to below the nutricline, where they take up phosphate before using gas vesicles to increase buoyancy enabling a return to the surface of the euphotic zone [77,78]. Finally, the additional N and energy investments required for exoenzymatic breakdown of DOP appears to give N 2 fixers a competitive advantage in oligotrophic regions [79].
The thermal niche of Trichodesmium, which is confined to a relatively narrow temperature range from 19-30˚C, is a key factor that sets the limits of its geographic distribution. It is clear that temperature plays a pivotal role in constraining Trichodesmium's global distribution as i), peak abundance for Trichodesmium sp. occurs in regions that are supra-optimal temperature for growth of IMS101 (Fig 2) and other Trichodesmium isolates (e.g. GBRTRL 1101, KO4-20, Fig 2. Trichodesmium is found most frequently and its biomass is highest where water temperature is supra-optimal for growth of IMS101. The dashed lines represent the minimum (T min ), optimal (T opt ) and maximum (T max ) temperature limits for Trichodesmium IMS101 growth for the mid CO 2 (~380 ppm), high-light treatment. Data from the MARine Ecosystem Data archive (https://doi.pangaea.de/10.1594/PANGAEA.774851). [27,28] and ii), the growth response to temperature is negatively skewed. Growth rates decrease significantly with a 3 to 4˚C increase above the optimal (26˚C). The areal distribution of Trichodesmium spp. is predicted to increase as the 20˚C isotherm slowly shifts pole-wards [80,81]. None-the-less, the negative effects of increased temperature and light at low latitudinal regions where temperature already exceeds T opt may lead to a contraction of the range at low latitudes. How future increases of SST will influence the distribution will depend on the capacity for Trichodesmium spp. to adapt by increasing its upper thermal tolerance limit.

RLI or 2175) that have been investigated
Increases in atmospheric CO 2 are transmitted to the ocean, increasing the dissolved CO 2 , and causing ocean pH and carbonate concentration to decline and bicarbonate concentration to increase. Recent research suggests that ocean acidification has greater potential to increase phytoplankton growth rates in areas of the ocean where temperature is equal to or less than the T opt [82]. In regards to Trichodesmium IMS101, previous research indicates that increasing CO 2 above 380-400 ppm can lead to modest increases in Trichodesmium growth rate [33,34,[38][39][40]42,43], some studies have reported declines in growth at elevated CO 2 [36,37] and others no change [32,35]. Our study suggests that ocean acidification will have a small to negligible effect for growth at the supraoptimal temperatures above 27˚C. Differing growth responses to elevated CO 2 between key phytoplankton types could cause sufficient shifts in competitive fitness to alter community structure [83]. Thus, the effect of ocean acidification, albeit indirectly, may still play a role in constraining Trichodesmium's global distribution. applied to each treatment to generate a median growth rate (Treatment reported here: high CO 2 (~720 ppm), 400 μmol photons m -2 s -1 ). The R code was applied to the full data set of each treatment (A). Criterion 1 identifies and discards data associated with crashed cultures (large circles) or lagged growth (B). Criterion 2 then identifies and discards individual data points (small circles) which, if incorporated, would significantly alter the gradient of a slope (C). Finally, criterion 3 identifies and discards slopes (dashed lines) associated with acclimation (D). The remaining slopes are associated with balanced growth, and were used to calculate a median growth rate (E). (TIF) S1 File. Control of inorganic carbon chemistry in the culture medium.  Table. The range of the experimental treatments used to measure Trichodesmium's light and temperature growth curves. Temperature response growth conditions; low (1 80 ppm), mid (~380 ppm) and high (~720 ppm) CO 2 , 40 μmol photons m -2 s -1 (LL) and 400 μmol photons m -2 s -1 (HL), ranging between 18-31˚C. Light response growth conditions; low (~180 ppm), mid (~380 ppm) and high (~720 ppm) CO 2 , 26˚C, ranging between 20-1400 μmol photons m -2 s -1 ). A circle (O) represents a growing culture; a cross (X) represents a condition where growth did not occur; a dash (-) represents a condition that was not used for culturing. (DOCX) S2 Table. Calculation of the minimum energetic requirement for growth of Trichodesmium. Footnotes to S2 Table. a CO 2 fixation to carbohydrate in the Calvin cycle according to the following stoichiometry. CO 2 + 3 ATP + 2 NADPH ! CH 2 O + H 2 O + 3ADP +3 P i. The photon requirement (9 photons/CO 2 fixed) is from Raven et al. [84]. b Carbon concentrating mechanism where the only energised step is the influx of HCO 3 at one membrane between the medium and Rubisco. Lower value assumes no leakage, whereas the upper value assumes leakage rate equals to the rate of photosynthesis [84]. c The energetic cost of N 2 fixation was calculated assuming complete recycling of H 2 to recover ATP was calculated from the following stoichiometry: N 2 + 6 H + + 6 e -+ 13 ATP ! 2 NH 3 + 13 ADP + 13 P i. d The cost of ammonium assimilation into amino acids is 1 ATP/NH 3 and 1 NADPH (2 reducing equivalents) assimilated via GOGAT. Protein synthesis would require an additional 4 ATP per peptide bond formed. e Based on a typical photosynthetic quotient of 1.2 O 2 evolved per CO 2 fixed for algae growing with ammonium as the inorganic N source. This accounts for the more reduced state of lipids and proteins relative to carbohydrates. f Total cost of synthesising 1 unit of Cbiomass assumes a Redfield C:N ratio of 106C:16N and that protein accounts for all of the cell N. g Photon requirements were calculated based on 1/3 ATP generated per photon absorbed during linear photosynthetic electron transfer from H 2 O to O 2 , with the additional ATP requirement from provided either by LPET from H 2 O to H 2 O (water-water cycle) with 1/3 ATP generated per photon absorbed (higher estimate) or by cyclic photosynthetic electron transfer around photosystem I with 1 ATP generated per photon absorbed. (DOCX) S3 Table. The summary of rules for error propagation. Key; x is the calculated value, σ x is the calculated error of uncertainty; a, b and c are known quantities; σ a , σ b and σ c are errors of uncertainty for a, b and c, respectively; y is a constant with no measure of uncertainty. (DOCX)