The eight Synechococcus spp. strains used in this study were retrieved from the Roscoff Culture Collection (RCC¹), including representative strains of the three known CRD1 ESTUs (CRD1A—C) and one or two of each of the four dominant clades in the field (clades I—IV) used as controls (Table 1 and Supplementary Figure 1). Cells were grown in 50 mL flasks (Sarstedt, Germany) in PCR-S11 culture medium (Rippka et al., 2000) supplemented with 1 mM sodium nitrate. Cultures were acclimated for at least 2 weeks in temperature-controlled chambers across a range of temperatures dependent on the thermal tolerance of each strain and under a continuous light of 75 μmol photons m^–2 s^–1 (hereafter μE m^–2 s^–1) provided by a white-blue-green LED system (Alpheus, France). For each experiment, cultures were grown in triplicate, inoculated at an initial cell density of ∼3 × 10⁶ cells mL^–1, and samples were harvested daily to measure growth rate and fluorescence parameters as described below.

In order to compare the capacity of strains to repair the D1 subunit of photosystem II (PSII; see “Measurement of PSII Repair Rate” section), cultures grown in 250 ml flasks at 75 μE m^–2 s^–1 were acclimated at 18, 22, and 25°C, temperatures at which all strains were able to grow, were subjected to high light stress (375 μE m^–2 s^–1). Exponentially growing cultures were sampled at T0 and after 15, 30, 60, and 90 min of stress, before shifting cultures back to the initial light conditions and then sampling again after 15, 30, 60 min, and 24 h of recovery (R). While D1 repair measurements were performed at all-time points, cell concentrations were measured by flow cytometry only at T0, T30 min, T90 min, R30 min, and R24 h and liposoluble pigment content was determined only at T0.

Culture aliquots (200 μl) sampled for flow cytometry were fixed using 0.25% (v/v) glutaraldehyde (grade II, Sigma Aldrich, United States) and stored at −80°C until analysis (Marie et al., 1999). Cell concentrations were estimated using a Guava easyCyte flow cytometer (Luminex Corporation, United States) and maximum growth rates (μ_max) were calculated as the slope of the linear regression of ln (cell density) vs. time during the exponential growth phase. Synechococcus cells were identified based on their red (695 nm) and orange (583 nm) fluorescence, proxies for their chlorophyll a and phycoerythrin content, respectively. Fluorescence, forward scatter and side scatter values were normalized to that of standard 0.95 μm beads using Guavasoft software (Luminex Corporation, United States).

The maximum PSII quantum yield (F_V/F_M) was estimated using a Pulse Amplitude Modulation fluorimeter (Phyto-PAM II, Walz, Germany) during the exponential growth phase after 10 min dark acclimation followed by addition of 100 μM of the PSII blocker 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU, Sigma-Aldrich, United States; Campbell et al., 1998).

The PSII quantum yield was calculated as:

where F₀ is basal fluorescence, F_M maximal fluorescence level and F_V variable fluorescence (Campbell et al., 1998; Six et al., 2007).

Fluorescence excitation (with emission set at 580 nm) and emission (with excitation set at 530 nm) spectra were generated using a LS-50B spectrofluorometer (Perkin-Elmer, United States) as described in Six et al. (2004). The fluorescence excitation ratio (Exc_{495:550 nm}) was used as a proxy for the PUB:PEB ratio. Phycobilisome (PBS) rod length and the degree of coupling of the PBS to PSII reaction center chlorophylls was then assessed using fluorescence emission spectra by calculating the phycoerythrin (PE, F_max = 565–575 nm) to phycocyanin (PC, F_max = 645–655 nm) ratio as well as the PC to PBS terminal acceptor (TA; F_max = 680 nm) ratio, respectively (Pittera et al., 2017).

Triplicate cultures were harvested during the exponential phase when F_V/F_M was maximum for each temperature condition. Cultures (50 mL) were subjected to centrifugation in the presence of 0.01% (v/v) pluronic acid (Sigma-Aldrich, Germany) at 4°C, 14,000 × g for 7 min, using an Eppendorf 5804R (Eppendorf, France). Pellets were resuspended and transferred to 1.5 ml Eppendorf tubes and centrifuged at 4°C, 17,000 × g for 2 min using an Eppendorf 5417R centrifuge (Eppendorf, France). Once the supernatant was removed samples were stored at −80°C until further analysis. Pigment content was subsequently assessed using calibrated high-performance liquid chromatography (HPLC 1100 Series System, Hewlett Packard, St Palo Alto, CA), as previously described (Six et al., 2005).

Each culture acclimated to 75 μE m^–2 s^–1 and 18, 22 or 25°C was split into two new 50 mL flasks (Sarstedt Germany) with one used as a control and the other flask supplemented with lincomycin (0.5 mg mL^–1 final concentration, Sigma-Aldrich, United States) in order to inhibit protein synthesis and thus D1 repair (Guyet et al., 2020). Both sub-cultures were then subjected to light stress by exposing cultures to 375 μE m^–2 s^–1 continuous light (at the same temperature), and F_V/F_M measured at different time points as described above. The PSII repair rate for each strain at each temperature was determined from the coefficient differences between the exponential curves fitted over the 90 min time course of F_V/F_M measurements for control and +lincomycin samples. This light stress experiment was replicated on four independent cultures.

The realized niches of CRD1 and clades I-IV ESTUs were determined using petB reads extracted from metagenomic data from the Tara Oceans and Tara Polar circle expeditions, the Ocean Sampling Day (OSD; June 21st 2014) campaign, and petB metabarcodes from (i) various oceanographic cruises (CEFAS, BOUM, Micropolar, RRS Discovery cruise 368 and several in the northwestern Pacific Ocean as detailed in Xia et al., 2017), (ii) two individual sampling sites in the Mediterranean Sea (Boussole, Point B) as well as (iii) a bi-monthly sampling at the long-term observatory site SOMLIT (“Service d’Observation en Milieu Littoral”)-Astan located 2.8 miles off Roscoff between July 2009 and December 2011 (Supplementary Table 1).

petB metagenomic recruitment using the Tara and OSD datasets was performed as described previously (Farrant et al., 2016). Synechococcus petB sequences from both metagenomes and metabarcodes were used to define operational taxonomic units (OTUs) at 97% identity using Mothur v1.34.4 (Schloss et al., 2009) that were then taxonomically assigned using a petB reference database (Farrant et al., 2016). OTUs encompassing more than 3% of the total Synechococcus reads for a given sample were grouped into ESTUs and used to determine the whole temperature range occupied by each of the five major Synechococcus ESTUs.

The Cyanorak v2.1 information system (Garczarek et al., 2021)² was used to compare the phyletic pattern i.e., the presence/absence pattern of each cluster of likely orthologous genes (CLOG) in each strain, for CRD1 strains and their clades I-IV counterparts for a number of selected genes potentially involved in adaptation to low temperature based on previous literature (see section “Results”).

In order to determine the temperature optima and boundary limits of the different CRD1 strains and to compare them to those of typical cold and warm Synechococcus thermotypes, representative strains of each of the three CRD1 ESTUs and strains of clades I, II, III, and IV were grown over a range of temperatures from 6 to 36°C. The growth responses of all strains to temperature followed a typical asymmetric bell-shaped curve over the selected temperature range (Figure 1), with a progressive rise in growth rate (μ) until T_opt (the temperature associated with maximum μ: μ_max) was reached, and a sharp decline above T_opt. BIOS-U3-1 (CRD1-B) was able to grow between 12 and 29°C with a T_opt at 25°C (μ_max = 0.78 ± 0.02 d^–1), a growth pattern most similar to that of the clade IV strain BL107, while the clade I strain MVIR-18-1 was able to grow at much lower temperatures, down to 8°C but could not grow above 25°C (Figure 1A). MITS9220 (CRD1-A) and BIOS-E4-1 (CRD1-C) displayed thermal growth characteristics more similar to the clade II (A15-62 and M16.1) and III (WH8102) strains, representatives of warm thermotypes (Figure 1B). While most strains in this category displayed a minimal growth temperature of 16°C, large variations between strains were observed at the highest thermal boundary limit (T_max). Maximum growth temperature was obtained for M16.1 (II; T_max: 34°C), then A15-62 (II) and WH8102 (III; both with T_max at 32°C), MITS9220 (CRD1-A; T_max: 31°C) and finally for BIOS-E4-1 (CRD1-C; T_max: 30°C). The latter strain also displayed the highest minimal growth temperature (T_min: 18°C) and thus possesses the narrowest temperature range for growth of all the strains studied (12°C vs. 15–18°C). It is also worth noting that CRD1 strains display a lower maximum growth rate and more generally lower growth rates at most temperatures than their clade I, II, III, and IV counterparts.

We then compared the fundamental thermal niches of all studied strains, i.e., the whole temperature range over which they can grow in a laboratory setting in the absence of biotic interactions (e.g., competition or predation), with environmental realized niches (sensu Pearman et al., 2008) of the corresponding ESTUs. For this, we determined the distribution limits of each of these ESTUs along the Tara Oceans and Tara Polar circle transects, 203 samples from OSD2014 and additional oceanographic cruises and individual sampling sites, altogether encompassing 413 samples worldwide covering a wide range of temperature conditions (Figure 1, Supplementary Figure 2, and Supplementary Table 1). This made it possible to have much finer estimates of the limits of the thermal niches of the different ESTUs than in the study performed by Farrant et al. (2016), in particular for the cold adapted ESTUs, which were poorly represented in the initial Tara Oceans dataset (Supplementary Figure 2).

This analysis showed that the CRD1B ESTU displayed a reduced thermal tolerance range in the environment (14–24.5°C) compared to the BIOS-U3-1 strain in culture (12–29°C), while the typical cold thermotypes colonized larger thermal niches in situ than their representative strains (Figure 1A). Environmental realized niches indeed ranged from 2.5 to 24°C for ESTU IA (compared to 8–25°C for MVIR-18-1) and from 8.5 to 25°C for ESTU IVA (compared to 12–29°C for BL107). Interestingly, the median temperature of the CRD1B ESTU is 3°C higher than that observed for ESTUs IA and IVA.

As concerns warm thermotypes, CRD1C displayed a fairly narrow thermal tolerance range in situ (22–29.5°C), which, similar to the cold thermotype CRD1B, was even narrower than for its representative strain BIOS-E4-1 (18–30°C; Figure 1B). Comparatively, the CRD1A ESTU was detected across a wider temperature range (14–30.5°C) than the other two CRD1 ESTUs and also slightly larger than the corresponding strain in culture (MITS9220, 16–31°C). Still, the most extended temperature range was observed for ESTU IIA and IIIA (12–32°C) that reached significantly lower temperature limits than the corresponding clade II (16 to 32–34°C) and III (16–32°C) strains. Of note, although both IIA and IIIA ESTUs displayed a similar temperature range, the median temperature of ESTU IIA (25°C) was about 3°C higher than that of ESTU IIIA (22°C) and the maximum median temperature was surprisingly observed for the CRD1C ESTU (26.5°C). In this context, it is also worth mentioning that although clade II strains are clearly both warm thermotypes, M16.1 displays a significantly higher temperature limit for growth than A15-62 and more generally than all other strains. This suggests that ESTU IIA may encompass two distinct ESTUs, but such a high temperature niche (>32°C) where they could be discriminated is exceptional and not available in our dataset (Supplementary Table 1).

In order to assess whether the cold, temperate thermotype BIOS-U3-1 (CRD1B) exhibits similar adaptation mechanisms to those previously described for the cold-adapted clades I and/or IV, we examined a number of clusters of likely orthologous genes (CLOGs) from all Synechococcus genomes belonging to clades I-IV and CRD1 present in the Cyanorak v2.1 information system (Garczarek et al., 2021). First, we looked for the occurrence of two amino-acid substitutions in phycocyanin α- (RpcA) and β-subunits (RpcB), which were shown to differ between cold- (Gly in clades I and IV for RpcA43; Ser in RpcB42) and warm-thermotypes (Ala in clades II and III for RpcA43; Asp in RpcB42), these substitutions being potentially responsible for the differential thermotolerance of this phycobiliprotein between thermotypes (Pittera et al., 2017). In all three CRD1 strains, both sites displayed the warm-type residue (Supplementary Figures 3A,B), suggesting that in contrast to typical cold and warm thermotypes, the molecular flexibility of this phycobiliprotein does not differ between CRD1 thermotypes. We then looked at fatty acid desaturases that are essential for regulating membrane fluidity and thus the activity of integral membrane proteins, including photosynthetic complexes (Mikami and Murata, 2003; Pittera et al., 2018; Breton et al., 2020). All three CRD1 strains surprisingly possess in addition to the core Δ9-desaturase gene desC3, a second Δ9-desaturase, desC4 (Supplementary Table 2), previously thought to be specific to cold-adapted strains as well as the Δ12-desaturase desA3 found in both cold-adapted clades I and IV as well as in clade III, a warm thermotype subjected to much stronger seasonal variability than its (sub) tropical clade II counterparts (Pittera et al., 2018). Furthermore, BIOS-U3-1 also possesses desA2, thought to be specific to warm environments, while this gene is in contrast absent from the other two CRD1 warm-adapted strains. Thus, CRD1 strains exhibit a different desaturase gene set and potentially display a larger capacity to regulate membrane fluidity than typical cold- or warm-adapted thermotypes. Finally, while all clades I, III and IV genomes possess the ocp operon, involved in the protection of PSII against photoinactivation through the dissipation of excess light energy (Kirilovsky, 2007) and which was recently shown in marine Synechococcus to play a key role at low temperature (Six et al., 2021), none of the three CRD1 genomes possess this operon.

PSII quantum yield (F_V/F_M), used as a proxy of photosynthetic activity, was measured for each strain over their whole temperature growth range. Most strains displayed a decrease in this parameter at both low and high boundary limits of their growth temperature range and this effect was particularly striking for BIOS-U3-1, reaching values down to 0.11 at 14°C and 0.32 at 28°C (Figure 2A). Besides MVIR-18-1 that exhibited a quite constant F_V/F_M over its whole temperature range, the decrease in F_V/F_M at high temperature was stronger for cold than warm thermotypes that are able to maintain a quite high F_V/F_M in the warmest growth conditions (Figure 2B). Finally, as for growth rate, CRD1 strains exhibited lower F_V/F_M at all temperatures than clade I to IV strains.

The Exc_495:550nm fluorescence excitation ratio, used as a proxy for PUB:PEB ratio, was consistent with the pigment type of each strain (Humily et al., 2013; Table 1 and Supplementary Figure 4A). This ratio remained pretty constant over the whole temperature range for all strains except for the chromatic acclimator BL107 (pigment type 3dA), for which a sharp increase was observed at its maximal growth temperature (28°C) to reach a value (1.35) intermediate between that typically observed in green light (or white light; 0.6–0.7) and blue light (1.6–1.7). This suggests that the chromatic acclimation process could be affected by growth temperature, at least in this strain. The phycobilisome (PBS) rod lengths and the degree of coupling of PBS to PSII reaction center chlorophylls, as estimated from PE:PC and PE:TA ratios, respectively, showed fairly limited variations over the temperature range, indicating that the phycobiliprotein composition of PBS is quite stable over the growth temperature range of each strain (Supplementary Figures 4B,C). One notable exception was a rise in both ratios for strain A15-62 at its minimal growth temperature, likely attributable to the partial decoupling of individual phycobiliproteins and of the whole PBS from PSII, a phenomenon typically observed under stressful conditions (Six et al., 2007; Guyet et al., 2020). It is also worth noting that MITS9220 and to some extent BIOS-E4-1, exhibited a significantly higher PE:PC ratio than the five other strains, potentially indicating a different phycobiliprotein composition and/or length of PBS rods.

In terms of liposoluble pigments, the β-carotene/chlorophyll a (β-car/Chl a) ratio tended to increase with temperature in BIOS-E4-1 and MITS9220, as observed for the other warm thermotypes, whilst this ratio was more stable in the cold thermotypes BIOS-U3-1 and BL107, and seemed to slightly increase in the lower part of the thermal range for the clade I strain MVIR-18-1 (Figure 3). For all strains, these ratios result from a concomitant increase with temperature of Chl a and β-car content per cell (Supplementary Figure 5), indicating an enhancement of the surface of thylakoids per cell at higher temperatures that was particularly marked for BIOS-E4-1 and A15-62, whilst this variation was fairly limited in the other two CRD1 strains. As these two pigments are present in different proportions in PSI and II (Umena et al., 2011; Xu and Wang, 2017), the higher β-car/Chl a ratio measured in clades I and IV strains also suggests that they may have a higher PSII:PSI ratio than all other strains, including BIOS-U3-1, and that this ratio might be more strongly affected by temperature in warm than cold thermotypes.

As concerns the zeaxanthin/chlorophyll a (Zea/Chl a) ratio, although an increase in this ratio was measured at low temperature for all strains, the amplitude was globally larger for cold than for warm thermotypes, with BIOS-U3-1 behaving very similarly to the clade IV strain BL107 that exhibits the largest variation in this ratio (Figure 3). Changes in this ratio likely originate partially from the decrease in Chl a content in response to cold, a strategy typically used by cells to regulate light utilization under slow growth conditions (Inoue et al., 2001). However, several strains also displayed an increase in their Zea content per cell at low temperature, a response particularly striking in BIOS-U3-1 and A15-62, but that also seems to occur in M16.1 and in the two other CRD1 strains BIOS-E4-1 and MITS9220 (Supplementary Figure 5). Thus, although Zea has been hypothesized to be involved in the photoprotection of cold-adapted strains by dissipating excess light energy under low temperature conditions (Kana et al., 1988; Breton et al., 2020), this process seems to be present in both cold and warm-adapted CRD1 strains and in most warm thermotypes as well. In this context, it is also worth noting that the two clade II strains, A15-62 and M16.1, displayed fairly distinct temperature-induced variations in their Zea:Chl a ratios and individual pigment contents, possibly linked to their different isolation temperatures (see section “Discussion” below).

The ability of the different strains to repair PSII in response to light stress (375 μE m^–2 s^–1) was determined in cultures acclimated to 18, 22, and 25°C by measuring changes in F_V:F_M over time after adding the protein synthesis inhibitor lincomycin, or not (Supplementary Figure 6). While a decrease in F_V:F_M ratio during the 90 min light stress period was observed in both cultures supplemented with lincomycin and controls, this ratio only re-increased back up to initial F_V:F_M values, after shifting cultures back to standard light conditions (75 μE m^–2 s^–1), in the control group in most strains and temperature conditions. Thus, all studied strains were able to recover from this light stress, as long as the D1 repair cycle was not inactivated by inhibition of protein synthesis. Yet, a fast decrease in F_V:F_M was observed for all three CRD1 cultures supplemented with lincomycin, while the +/− lincomycin curves overlapped during the first 15–30 min of light stress in most other strains and conditions. This suggests that the initial decrease in F_V:F_M in clades I–IV strains was not due to D1 damage but rather to dissipation of light energy as heat through non-photochemical quenching (Campbell et al., 1998), whilst the damage and hence repair of D1 proteins only occurred later on.

The PSII repair rate (R_PSII), as calculated from the time course of F_V:F_M with and without lincomycin, increased with temperature in most strains, except for BIOS-U3-1 that displayed its highest rate at 22°C (Figure 4). Strikingly, all three CRD1 strains displayed significantly higher R_PSII than clade I-IV strains at all three tested temperatures, a difference ranging from 3- to nearly 40-fold at the lowest common temperature (18°C). Furthermore, CRD1 strains displayed fairly limited variation in R_PSII with temperature (ranging from 1.33 to 1.87-fold) compared to the other strains, the strongest increase in R_PSII being observed for the clade I strain MVIR-18-1 (21.5-fold) and the clade III strain WH8102 (5.5-fold). This indicates that CRD1 strains exhibit a constitutively high level of PSII repair compared to the other strains whatever the growth temperature and only trigger a moderate increase in R_PSII in response to temperature variations.