## Introduction

Ocean warming with climate change is occurring at an unprecedented rate as a result of increasing loads of atmospheric CO2 and other greenhouse gases1. Earth system models predict that ocean warming will result in a reduction in marine primary productivity (up to 20%) throughout the twenty-first century and driven by rising temperatures that exceed the limits of thermal tolerance2,3 and nutrient limitation in a warmer, more stratified ocean4. Oceanic primary production is accomplished by photosynthetic microorganisms5. Their short generation times and high population densities make evolutionary responses to climate change possible6. There is growing evidence of adaptation by phytoplankton through evolutionary responses to global change drivers, such as elevated CO27,8,9,10, pollutants11, and temperature12,13. For example, previous studies reported fast adaptation of phytoplankton to warming after 80–450 generations10,12. Latitudinal patterns in thermal growth responses suggest that phytoplankton are adapted to environmental temperature, with tropical species showing optimum growth at the high in situ oceanic temperatures14. Rising temperatures leading to poleward shifts in the thermal niches of phytoplankton has been predicted by the end of this century along with a sharp decline in tropical phytoplankton diversity in the absence of evolutionary responses to warming14. Thus evolutionary responses of phytoplankton to warming will be essential to avoid a reduction in the biodiversity of the tropical ocean where replacement by temperate species is not expected. Diatomic algae are a relevant group contributing largely to oceans primary production and biogeochemical processes. Diatoms dominate primary producers in cold waters, and although are not dominant in tropical waters still play there a major role in the biological pump and other biogeochemical processes15,16. Diatoms have lower activation energies (i.e. less sensitive to increasing temperature) than other phytoplankton groups17, but its adaptation to warming will be crucial to maintain an efficient biological pump in the tropical ocean16.

In order to test adaptation to warming temperatures of warm-adapted tropical organisms, we conducted a long-term experiment on four diatomic phytoplankton species, Chaetoceros tenuissimus, Chaetoceros sp., Thalassiosira sp. and Synedra sp., isolated from the warm surface waters of the Red Sea. The four species of diatoms were maintained in the laboratory for about six months (~200–600 generations) at 26 °C (ambient temperature control) and at 30 °C (experimental warming conditions). 26 °C represents the mean Red Sea surface temperature (SST) for the 1982–2015 period (Fig. S1). For the projected warming conditions at 30 °C (mean Red Sea temperature +4 °C), we followed the high-emission scenario (RCP 8.5) projected for the turn of the next century (2100) by IPCC in 2014, that will represent a global mean warming ranging between 2.6 and 4.8 °C. The choice of ambient control temperature is problematic when considering interannual and seasonal variability22 as the sea surface temperature (SST) of Red Sea range from 21 to 32 °C23. However, phytoplankton species can adapt their realized niches to track average increases in water temperature regardless of annual and seasonal fluctuations, so the mean SST should be a reliable experimental control value24.

We investigated the thermal adaptation strategies of the four diatoms by combining experimental evolution with measurements of fundamental physiology (Fig. S2). The four species showed a fast adaptation to warmer temperature after 200–600 generations. Two of the species increased their optimal growth temperature (Topt) and maximum growth rate. The other two diatoms did not increase Topt and growth, but shifted from specialist to generalist increasing their maximum critical thermal limit. However, none of the species was able to improve both and did not experience a “win-win” adaptation. The results suggested that despite adaptation was fast the trade-offs associated constrained growth or photophysiological performance.

## Results

After the six-month temperature selection period, Chaetoceros sp. and Thalassiosira sp. produced a similar number of generations of 439 and 440 generations and 413 and 393 generations for the experimental ambient and warming conditions, respectively. Due to its higher specific growth rate, C. tenuissimus produced nearly 600 generations (ambient: 556 generations, warming: 570 generations), whereas Synedra sp. produced only 200 generations (ambient: 188 generations, warming: 194 generations) due to its lower growth rate.

We used the mixed-effects model (Table 1) to understand the selection effect of time and temperature on the four species. Increasing temperature significantly affected the growth of all four species of diatoms. Selection time (F = 16.45, p < 0.001) and selection temperature (F = 42.61, p < 0.001) as well as the interaction of selection time and selection temperature (F = 189.34, p < 0.001) significantly affected the growth of Chaetoceros sp. (Fig. 1a, Table S1). Increased temperature increased the growth of Chaetoceros sp. at the beginning of the selection period, while this positive effect was reversed toward the end of selection period (Fig. 1a). Among Thalassiosira sp., the growth of warming-adapted cells was significantly lower than that of ambient-adapted cells (F = 142.56, p < 0.001) (Fig. 1b, Table S1). Among C. tenuissimus, increasing temperature increased the growth (F = 543.58, p < 0.001), and this positive effect was amplified by selection time (F = 39.00, p < 0.001) (Fig. 1c). Similarly, the growth of Synedra sp. was significantly affected by selection temperature (F = 29.85, p < 0.001) (Fig. 1d), with a positive effect on growth throughout the selection period (F = 2.92, p = 0.0889) (Fig. 1d). While two of the phytoplankton species (C. tenuissimus and Synedra sp.) benefitted from the temperature increase, the other two (Chaetoceros sp. and Thalassiosira sp.) did not.

## Discussion

Diatoms become the predominant phytoplankton in the Red Sea when nutrient availability increases28,29. Diatoms proportion increase in the North during winter, in the cyclonic eddies and largely in the South, the warmest area in the Red Sea where temperature ranges between 26 and 32 °C and nutrient availability increases through the influx of Gulf of Aden Intermediate Water28,29,30. Their adaptation capacities to warming, on one hand, may help them to maintain their presence and distribution in the tropical ocean, maintaining the biogeochemical processes sustained by diatoms16,31,32. On the other hand, the trade offs associated with the thermal adaptations may decrease their competitive fitness, and could reshape the communities.

A recent comparative analysis coupled with an eco-evolutionary species distribution model demonstrated that the geographic variation in thermal niches of phytoplankton closely matched the temperature regime of their natal environment14. This study suggests that rising temperatures will cause poleward shifts in the thermal niches of phytoplankton and a sharp decline in trophic phytoplankton diversity in the absence of evolutionary adaptation. Our long-term experiments on diatoms from a tropical sea suggested that ~200–600 generations (6 months) were sufficient to evoke adaptation to ocean warming. Our observations that the Red Sea diatoms could shift their optimal growth temperatures or critical thermal limits provide an empirical basis for parameterizing thermal evolution in eco-evolutionary models of phytoplankton dynamics. Our study suggests that rapid thermal adaptation may help to reduce the sharp decline in tropical phytoplankton diversity as projected in the absence of evolutionary responses.

Several recent studies have shown that adaptation to warming becomes evident after 80–100 generations12,13. However, experiments investigating evolutionary responses to environmental change in phytoplankton have thus far focused on phytoplankton from natal environments in the temperate and sub-tropical oceans. Our study builds on this knowledge by investigating adaptive responses in organisms isolated from one of the warmest seas in the world23. Our findings demonstrate that photosynthetic microorganisms living in oceanic extreme temperatures can adapt to future warming conditions. The tropical diatoms tested showed contrasting adaptation capacities with different strengths and weaknesses. Specialization to increased optimal temperatures optimizes responses against competitors at higher water temperatures, but the thermal niche allows a generalist species to persist in a changing thermal environment. However, these adaptation strategies involve costs for some species, associated with trade-offs, including reduction in the maximum growth or the photosynthetic capacity, that should be considered for projections (Fig. S7). Since the oceans are now undergoing multiple environmental changes including ocean acidification, deoxygenation, increased light exposure and decreased nutrient availability within the upper mixing layers33, the adaptation associated with such trade-offs may alter the resistance of a phytoplankton species to other stressors. At the same time, the thermal adaptation of phytoplankton under natural conditions might also be amplified or retarded by other environmental changes and it might also depend on co-evolutionary interactions with other species that remain difficult to predict.

Fast organisms’ adaptation will be required in fast changing environments. The Red Sea is warming at 0.17 ± 0.07 °C decade−1, exceeding the global rate23. Our results confirmed fast adaptation capacity to warming (i.e. within months) for Red Sea diatoms. Irwin et al.24 found that phytoplankton species in the CARIACO station in a 15-years time series adapted to a change of 1 °C over a decade24. Our study was conducted at fixed ambient and warming scenarios ignoring the temperature fluctuations in the natural environment and its influence on the rate of adaptation. Future studies with fluctuating treatment mimicking temperature variation in the region would contribute to help identify more accurate adaptation rates as well as examine thermal adaptation strategies and trade-offs.

In conclusion, our results suggest organisms from the warmest sea on earth can adapt fast to ongoing warming, on top of already experiencing thermal extremes, with trade-offs. Experiments examining evolutionary responses to increasing temperature in the ocean, like those performed in this study, provide important insights into responses of organisms to climate change. They also help to identify potential evolutionary winners and losers in the context of climate change. The thermal adaptation strategies reported in the present study and associated constrains to performance, may improve our knowledge on the ecological evolutionary responses to warming.

## Methods

### Species isolation and culture conditions

Four diatom species, Chaetoceros sp., Thalassiosira sp., Chaetoceros tenuissimus and Synedra sp., were isolated from coastal Red Sea waters near the Al Fahal Reef (22.2528°N, 38.9612°E). Surface water samples were collected and then were passed through a 45-μm filter. Approximately ten clonal cultures were established by single-cell isolation under a microscope. All cultures were maintained as batch cultures in filtered seawater that was taken from the same location and enriched with f/4 medium and silicate. The cultures incubated at 24 °C in a precise temperature-controlled incubator (Percival, United States). The cultures grew with a light: dark cycle of 12 h: 12 h under 50 μmol photons m−2 s−1.

After pure cultures of each species were established, mono-specific cultures of Chaetoceros sp., Thalassiosira sp., Chaetoceros tenuissimus and Synedra sp. were grown in 200 mL Erlenmeyer flasks at 26 ± 0.1 (experimental ambient, termed as ambient hereafter) and 30 ± 0.1 °C (experimental warming, termed as warming hereafter). The design of the warming temperature that ignoring the regional temperature variation (both annual and seasonal) in the Red Sea rather serves as a temperature increase by average as projected by high-emission scenario (RCP 8.5, IPCC 2014) for the turn of next century, to explore the extent the adaptive capacity. The cultures grew under 400 μmol photons m−2 s−1 with a light: dark cycle of 12 h: 12 h. Four independent replicated cultures (n = 4) were run semi-continuously for about 6 months under ambient and warming conditions by renewing the medium every 3 days for Chaetoceros sp., Thalassiosira sp. and C. tenuissimus and every 7 days for Synedra sp. due to their lower growth rate. This long-term nature of the experiments allowed the diatoms to adapt to the experimental temperature environment. The initial cell concentration was set at 1000 cells mL−1, and the medium was partially renewed every 3 or 7 days to restore the cell density to the initial level (i.e., growth batch cycle). Nutrients were not limiting as the cell abundances achieved at the end of the batch cycles were far from those expected at the stationary phase. Cell abundance was quantified every 3 days for Chaetoceros sp., Thalassiosira sp. and C. tenuissimus and every 7 days for Synedra sp., by examining the samples under an optical microscope (LEICA DMI 3000B-Germany) by hemocytometer. The specific growth rate, μ (d−1), during each batch growth cycle was calculated as

$${\rm{\mu }}=\frac{ln({C}_{1}-{C}_{0})}{({t}_{1}-{t}_{0})},$$
(1)

where C1 and C0 are the cell densities at times t1 and t0 (t1 − t0 = 3 or 7 days), respectively. The number of generations per transfer (g) is equivalent to the number of doubles and was calculated as follows:

$$g=\frac{{t}_{1}-{t}_{0}}{ln(2)/{\rm{\mu }}},$$
(2)

where t1 − t0 is the time interval of the transfer (d), ln (2)/μ is the doubling time (d) and μ is the specific growth rate (d−1).

### Thermal growth curves

The thermal growth responses for all four diatoms adapted to ambient and warming conditions for about 6 months were determined at ten (Chaetoceros sp., Thalassiosira sp. and Synedra sp.) or twelve (C. tenuissimus, depending on its high thermal capacity) assay temperatures. At the end of the long-term selection period, cultures of ambient and warming conditions were inoculated into 200 mL flasks at an initial cell density of 1000 cells mL−1, and then incubated at 18, 20, 22, 24, 26, 28, 30, 32, 34 and 36 °C (Chaetoceros sp., Thalassiosira sp. and Synedra sp.) and 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38 and 40 °C for species C. tenuissimus. To accommodate the different growth rates, batch cycles lasted 3 days for Chaetoceros sp. (~6–8 generations), Thalassiosira sp. (~5–7 generations) and C. tenuissimus (~9–11 generations) and 7 days for Synedra sp. (~7–8 generations) of the different temperatures. At the end of each growth cycle, the cell densities were determined by counting cell abundances under an optical microscope, and specific growth rates (μ) for each assay temperature were calculated from cell abundances versus time.

### Photosynthetic responses to temperature

Photosynthetic responses were measured at the end of each growth cycle using Pulse-Amplitude-Modulation (PAM) fluorometry (Phyto-PAM, Walz, Germany) at the ambient and warming temperatures in a reciprocal assay in which warming-adapted and ambient-adapted populations were compared against the respective non-adapted (i.e., 30 °C and 26 °C, for warming and ambient, respectively) and vice versa (Fig. S2). All the measurements were made in the middle of the light phase.

The maximum quantum yield (Fv/Fm) of photosystem II (PSII) was measured on samples that were adapted to the dark for 15 min and then determined by a saturated pulse (5000 μmol photos m−2 s−1). For the rapid light curve (RLC) measurements, we determined the relative electron transport rate (rETR) at 12 different light levels (1, 16, 32, 64, 164, 264, 364, 564, 764, 1064, 1364 and 1664 μmol photons m−2 s−1), each lasting for 20 s. The rETR (an arbitrary unit) was calculated as

$$rETR={\rm{\Phi }}\mathrm{PSII}\times 0.5\times \mathrm{PAR},$$
(3)

where ΦPSII is the photochemical quantum yield of PSII in light, PAR is the actinic light intensity (μmol photons m−2 s−1), and the factor 0.5 accounts for approximately 50% of all the absorbed energy allocated to PSII. RLC was fitted with the model proposed in ref.34. The photosynthetic parameters maximum electron transport rate (ETRmax), light usage efficiency (α) and saturated light intensity (Ik) were derived from the fitted curves (Supplementary Methods).

### Statistical analyses

We used the linear mixed-effects model to quantify trajectories of specific growth rates under experimental ambient and warming temperatures. For the analysis, μ was considered the dependent variable, time (in days) and treatment temperature were the fixed effects, while slopes and intercepts were treated as random effects at the level of replicates nested within the selection temperature35. The significance of the parameters was assessed using likelihood ratio tests, comparing models with common slopes and intercepts for each selection temperature (Table 1). Comparisons of models were done using Akaike’s information criterion (AIC) and likelihood ratio tests on models fitted with restricted maximum likelihood (REML) for comparison of random effects, and maximum likelihood (ML) for comparison of fixed effects36.

The thermal reaction norms of the four diatoms adapted to the two treatment temperatures were assessed by applying the equation described in refs14,37:

$$f(T)=a{e}^{bT}[1-{(\frac{T-z}{w/2})}^{2}],$$
(4)

where specific growth rate, f, depends on temperature, T, and is defined as a function of the parameters z, w, a and b. w is the temperature niche width (the range of temperatures over which the growth rate is positive), while the other three (z, a, b) possess no explicit biological meaning but interact to influence the rate of increase in growth with temperature, the maximum growth rate and the optimum temperature for growth. Specifically, z determines the location of the maximum of the quadratic portion of this function. a and b are the Epply curve coefficient and Epply curve exponent, respectively. We estimated the critical thermal minimum (CTmin, the lowest temperature at which the growth of phytoplankton is zero), the critical thermal maximum (CTmax, the highest temperature at which the growth of phytoplankton is zero)38, the maximum growth rate (μmax), the optimal temperature for growth (Topt), identified as the temperature at which the growth rate is maximal, and thermal breadth (expressed as B80, 80% performance of the maximum growth rate breadth) by numerically maximizing the equation after estimating the parameter values for each replicate. We note that the estimated μmax in ambient- and warming-adapted C. tenuissimus showed followed opposite trends with that of measured μmax. The measured μmax values are presented in Fig. 2i. For the other three species, μmax is given in the values estimated by equation (4). Student’s t-test was used to test the differences in CTmin, CTmax, Topt, μmax and B80 between two temperature-adapted populations.