Antenna proton sensitivity determines photosynthetic light harvesting strategy

The extent of proton sensitivity in antenna proteins determines the photosynthetic light harvesting strategy in plants and algae.

In higher plants, several processes contribute to excess light energy dissipation, but only the pH-dependent one, the so-called energy-dependent quenching mechanism (nonphotochemical quenching; NPQ) is considered photoprotective (Demmig-Adams and Adams, 1992;Ruban, 2013). Proof of a strict connection between NPQ and pH is that reverse ATPase activity can stimulate NPQ even in the dark (Gilmore and Yamamoto, 1992). Besides controlling xanthophyll cycle activity, in several phototrophs pH exerts a direct control on NPQ. This is thought to act via a regulation of antennas (e.g. Dekker and Boekema, 2005;Horton et al., 2008;Peers et al., 2009;Grossman et al., 2010). Indeed, a very similar thermal dissipation process to that in vivo can be induced in vitro in purified antennas by lowering pH and detergent concentration (Ruban et al., 1994a). Starting from this evidence, it was proposed that antenna aggregation is at the basis of the NPQ process (Horton et al, 1996), and subsequent findings employing liposomes started to clarify how pH and ions together with lipids and lipid to antenna ratios control the 'aggregation state' of antennas (Moya et al., 2001;Kirchhoff et al., 2008;Akhtar et al., 2015;Kaňa and Govindjee, 2016;Natali et al., 2016;Crisafi and Pandit, 2017). In higher plants, antennas are the site of energy dissipation, whilst xanthophylls and the PsbS protein seem to be simply controllers of the process Walters et al., 1994;Li et al., 2000;Betterle et al., 2009). Evidence that npq1, a mutant lacking zeaxanthin, and npq4, a mutant without PsbS, could both perform NPQ indicated their dispensability, thus placing antennas and pH as the only key elements of the process (Niyogi, 1999;Johnson et al., 2009;. Nevertheless, xanthophylls play an important role modulating the kinetics of NPQ activation and dissipation (Johnson et al., 2010). Pre-conditioning of leaves with light exposure, for instance, makes NPQ fast and persistent because of the conversion of violaxanthin into zeaxanthin (Ruban and Horton, 1999). Zeaxanthin, a highly hydrophobic pigment, in turn, makes antennas more dehydrated and therefore sensitive to pH and prone to quench compared with violaxanthin-enriched antennas. Interestingly, this idea was put forward not only for higher plant antennas (Ruban et al., 1994a), but also for antennas from distant organisms such as diatoms (Gundermann and Büchel, 2008), brown algae (Ocampo-Alvarez et al., 2013) and alveolates .
The state-of-the-art model of NPQ for plants claims that, under high light, lumen acidification induces antenna protonation, which in turn triggers protein conformational changes, aggregation and energy dissipation. However, it seems that preaggregation in vivo can affect efficiency of antenna protonation and vice versa (e.g. Petrou et al., 2014). Optical changes induced by aggregation can be visualized spectroscopically (Lokstein et al., 2002), specifically as an increase in the fluorescence yield of red-shifted emission from antennas at low temperatures (Ruban et al., 1991;Bassi and Dainese, 1992;Miloslavina et al., 2008;Belgio et al., 2012). Based on dicyclohexylcarbodiimide (DCCD) binding and mutagenesis work (Ruban et al., 1998;Belgio et al., 2013;Ballottari et al., 2016), it was concluded that sensors for low pH are negatively charged residues located in a lumen-exposed antenna protein loop and in the C-terminus. Once protonated, those residues become neutral, thus making the whole protein more hydrophobic and easier to aggregate and quench. Although in vitro fluorescence quenching as a function of pH has been observed for various types of antennas (Gundermann and Büchel, 2012;Kaňa et al., 2012;Schaller-Laudel et al., 2015), identification of putative protonable residues so far concerned mainly antennas from the green lineage (Ruban et al., 1998;Li et al., 2004;Liguori et al., 2013;Belgio et al., 2013;Ballottari et al., 2016).
Despite the progress in our understanding of NPQ in higher plants, this subject has been less explored in algae. The alveolate Chromera velia represents an interesting system in this context, as it shows efficient non-photochemical quenching (Kotabová et al., 2011;Quigg et al., 2012;Mann et al., 2014) with similarities on the one side to higher plants, and on the other to brown algae and diatoms (see below). Isolated from stony corals from Sidney harbor, this facultative symbiont is globally distributed in the marine environment at depths not exceeding 5 m (Obornik and Lukes, 2013). The phylogenic origin of the alga is complex. C. velia is an alveolate, and therefore closely related to dinoflagellates and other algae in the SAR clade (such as diatoms and brown algae), but all phylogenic analyses have invariably demonstrated its genuine relationship to apicomplexan parasites (Oborník et al., 2016). In any case, C. velia is considered a 'red-clade' alga, i.e. an alga whose chloroplast was obtained by secondary endosymbiosis from a red algal ancestor (Kotabová et al., 2011;Sobotka et al., 2017). In C. velia, NPQ is connected to the xanthophyll cycle (Kotabová et al., 2011) as in brown algae (Ocampo-Alvarez et al., 2013); however, differing from them (Garcia-Mendoza et al., 2011) but similar to diatoms (Ruban et al., 2004;Lavaud and Kroth, 2006;Grouneva et al., 2008), its activation is extremely fast (almost monophasic) and pH-dependent (see Belgio et al., 2018).
In the present paper, we investigated the reasons for the characteristic high rate of quenching displayed by the alga. We compared the NPQ of C. velia with that of a higher plant (a well-known system) and showed that the mechanism of heat dissipation and in particular NPQ activation is different in the two evolutionarily distant phototrophs. Our data indicated that the Chromera light harvesting complex (CLH) is more sensitive to protons than the higher plant antenna (light harvesting complex II; LHCII). We propose that protonation of the antenna is the basis of the 'constitutively' fast NPQ found in C. velia and, as previously suggested for diatoms (Lavaud and Kroth, 2006;Lavaud and Lepetit, 2013), ΔpH by itself is important for NPQ activation. This conclusion might also explain the unusual high light acclimation strategy recently reported for C. velia, consisting of a decrease in reaction centers whilst still maintaining a full antenna content (Belgio et al., 2018).

Plant material
Chromera velia (strain RM12) was grown in artificial sea water with additional f/2 nutrients (Guillard and Ryther, 1962). Cells were cultivated in glass tubes at 28 °C, in a continuous light regime of 200 µmol m −2 s −1 while aerated with air.

Isolation of C. velia and plant light harvesting complexes
C. velia cells were broken and solubilized as described in  and then loaded on a fresh, continuous 5-15% sucrose density gradient prepared using a home-made gradient maker in buffer containing 25 mM HEPES pH 7.8 and 0.04% n-dodecyl β-D-maltoside (β-DM). The ultracentrifugation was performed at 140 000 g at 4 °C for 20 h (with rotor SW28, for 40 ml tubes, of an L8-M ultracentrifuge; Beckmann, USA). The resulting band no. 2 contained a strong double band at 18 and 19 kDa, previously identified as 'fucoxanthin chlorophyll a/c binding protein (FCP)-like antenna' (Tichy et al., 2013). The band analysis by Pan et al., (2012) and Tichy et al. (2013) placed this antenna protein within the main FCP-like group of light-harvesting complexes and so it was named Chromera light harvesting complex (CLH).
After separation by sucrose gradient, the antenna protein was desalted using a PD10 column (GE Healthcare) in a buffer containing 20 mM HEPES (pH 7.6) and 0.01% (w/v) β-DM. Spinach LHCIIb was isolated as previously described (Ruban et al., 1994b) and then purified, desalted and eluted in the same buffer as CLH. In both cases, antennas were isolated from samples dark-adapted for 30-45 min.

Non-photochemical fluorescence quenching in native cells and isolated chloroplasts
Chlorophyll fluorescence was measured using a double modulation fluorometer FL-3000 (Photon System Instruments, Czech Republic). A multiple turnover saturating flash was applied to measure the maximum quantum yield of the photochemistry of photosystem II (F v /F m ) according to (F m −F 0 )/F m , where the difference between the maximum (F m ) and minimum (F 0 ) fluorescence is used to calculate the variable fluorescence (F v ) (van Kooten and Snel, 1990). Cells were then illuminated with an orange actinic light (625 nm, 500 µmol photons m −2 s −1 ), during which periodic saturating flashes were applied. NPQ was calculated as (F m −F m ′)/F m or F m ′, where F m ′ is the maximum fluorescence measured in the presence of actinic light. Non-photochemical quenching of fluorescence was measured in whole cells of C. velia (chlorophyll concentration 0.7 µg ml −1 ) and isolated spinach chloroplasts (chlorophyll concentration 1.4 µg ml −1 ). NPQ formation rates (NPQ as a function of time) in different xanthophyll cycle de-epoxidation states (DEPSs) were determined from the measured fluorescence traces as described in the 'Data analysis and model fitting' section.
Where indicated ( Fig. 2; Supplementary Fig. S1 at JXB online), the effect of an uncoupler on the fluorescence quenching was examined by adding NH 4 Cl (final concentration of 15 mM) at different time points of the measuring protocol.

In vitro fluorescence quenching of antennas
Isolated antennas (OD 676 =1 cm −1 ), solubilized in 0.01% DM, were diluted 20 times, while constantly stirring, in a room temperature buffer containing 10 mM sodium citrate and 10 mM Tris-HCl and adjusted with small drops of HCl to give the desired pH (for further details see Ruban et al., 1994a;Belgio et al., 2013). Chlorophyll fluorescence was continuously monitored using an FL 3000 fluorometer (PSI, Czech Republic, blue excitation at 464 nm, 184 µmol m −2 s −1 ). The pK a values for quenching kinetics were calculated as described in the 'Data analysis and model fitting' section.

Absorption measurement
Absorption spectra were recorded with a Unicam UV 500 spectrometer (Thermo Spectronic, UK).

Pigment extraction and HPLC analysis
Cells or chloroplasts were collected on GF/F filters (Whatman, UK) and soaked in 100% methanol (overnight at −20 °C) and disrupted using a mechanical tissue grinder. Filter and cell debris were removed by centrifugation (12 000 g, 15 min) and the supernatant used for absorbance measurements at 652, 665, and 730 nm. Chlorophyll concentration was determined according to Porra et al. (1989). HPLC was carried out on an Agilent 1200 chromatography system equipped with a diode array detector. Pigments were separated on a Luna Phenomenex C8 (2) column (particle size, 3 µm; pore size, 100 Å; dimensions, 100 × 4.6 mm), by applying a 0.028 M ammonium acetate-MeOH gradient (20/80) as described in (Kotabová et al., 2011) and the eluted pigments were quantified at 440 nm. The de-epoxidation state of the xanthophyll cycle pigments (DEPS) was calculated as: (zeaxanthin+0.5 antheraxanthin)/(violaxanthin+antheraxan thin+zeaxanthin) (Johnson et al., 2009;Kotabová et al., 2011;Oborník et al., 2011). For purified antennas, the same procedure was applied simply skipping the first step of filtration through GF/F filter.

Zeaxanthin enrichment
Plant chloroplasts and C. velia cells with a DEPS of 10% were obtained from dim-light-adapted samples (30 min). Enrichment in zeaxanthin was achieved as described previously (Ruban et al., 1994b;Belgio et al., 2014) by pre-conditioning leaves with 350 µmol photons m −2 s −1 under 98% N 2 for 20-40 min for 20% and 40% DEPS, respectively. For C. velia, 10 min illumination with 500 µmol photons m −2 s −1 was sufficient to obtain 40% DEPS, in agreement with what has been previously published (Kotabová et al., 2011). DEPS was assessed by immediate incubation in methanol followed by HPLC analysis (see 'Pigment extraction and HPLC analysis' section).

In silico studies
For in silico studies, the LHCIIb structure resolved at 2.5 Å resolution (PDB code: 2BHW; Standfuss et al., 2005) was employed. The structure of the CLH polypeptide (CveliaI_19753.t1 taken from Tichy et al. (2013)) was predicted using Phyre 2 (http://www.sbg.bio.ic.ac.uk/phyre2/html/ page.cgi?id=index) and YASARA software (http://www.yasara.org). The protonation states of protein ionizable groups were computed in both cases using the H++ program (http://biophysics.cs.vt.edu), an automated system that calculates pK values of ionizable groups in macromolecules and adds missing hydrogen atoms according to the specified pH of the environment. Results shown for LHCIIb are relative to chain A, but results for chains B and C were very similar, in agreement with (Xiao et al., 2012). As recommended for typical physiological conditions and deeply buried residues, the external dielectric value was set to 80, the internal dielectric value to 4, salinity to 0.15 and pH to 7.5.

Data analysis and model fitting
NPQ formation rates (NPQ in function of time; see Fig. 1) were determined using a well-established methodology valid for both algae and vascular plants (see Serôdio and Lavaud (2011) concerning the applicability of the Hill equation to NPQ in algae). Briefly, average data from three to six independent measurements of C. velia cells and spinach chloroplasts in different DEPSs were fitted using the sigmoidal Hill equation threeparameter implementation in SigmaPlot 12.5 (Systat Software, Inc., San Jose, CA, USA). The standard error of the estimate was between 0.02 and 0.08, meaning that ~95% of the data fell within 2% of the fitted line; moreover R 2 values were above 0.97, thus confirming the appropriateness of the approach.
In order to determine the quenching pK a of antennas, we used a method previously established for various antennas including mutants (Ruban et al., 1994a;Ruban and Horton, 1999;Belgio et al., 2013;Zaks et al., 2013). Briefly, the relationship between quenching kinetics and pH (see Fig. 4) and relative parameters (Table 1) were obtained from experiments like the one shown in Fig. 3 as follows. Quenching kinetics were calculated at each pH point by fitting the measured traces ( Fig. 3) with the three-parameter hyperbolic decay function: y=(y 0 +ab)/(b+x) where 1/b represents the rate of the process. Then the data points from Fig. 4 were fitted by the sigmoidal Hill equation y= [ax b ]/[c b +x b ] in order to obtain Hill coefficients (b), pK a values (c) and quenching kinetics at pH 4.97 (see also Petrou et al., 2014). The standard error of the estimate was again low (below 0.1) and R 2 above 0.90, confirming the validity of the approach.
The lack of an evident kinetic effect of xanthophylls in C. velia can also be seen from the shape of the NPQ formation curve. Whilst in spinach the increase in zeaxanthin (zea) concentration from 10 to 40% reduced curve sigmoidicity from 3.2 to 1.8 ( Fig. 1B; Table S1), in C. velia no evident change could be seen (Fig. 1A) and the Hill coefficients were not significantly different in the two conditions (3.0 ± 0.4 versus 2.5 ± 0.3; see Supplementary Table S1). The increased deepoxidation (from 10% to 40%; Fig. 1A), therefore, did not seem to affect NPQ kinetics as strongly as in spinach, but it stimulated the total NPQ (NPQ max ; see Table S1). This is in agreement with a previous report (Kotabová et al., 2011) showing NPQ enhancement by zeaxanthin in C. velia.
In C. velia, NPQ was induced almost instantaneously with the turning on of the actinic light, and we therefore used NH 4 Cl to investigate the possibility that lumen acidification was the basis of fast NPQ. As with spinach, in C. velia NH 4 Cl reversed fluorescence quenching independent of its addition time during irradiation ( Fig. 2A), proving a strict link between protons and NPQ in C. velia. However, the kinetics of NPQ relaxation at different time points were very different from each other and from those of spinach ( Supplementary  Fig. S1A). Whilst in spinach NPQ relaxed almost immediately after NH 4 Cl injection, with 70% fluorescence recovery within 10 s, it took at least 500 s to achieve a similar recovery in C. velia (Fig. 2B). Interestingly, in C. velia, the later NH 4 Cl was added, the faster NPQ relaxed (Fig. 2B). This is in strict contrast to higher plants ( Supplementary Fig. S1B), where faster relaxation kinetics were observed at the beginning of NPQ formation (see e.g. Fig. 3A in (Ruban et al., 2004), suggesting a different sensitivity of NPQ to lumen acidification. The connection between NPQ and protons was further investigated in vitro using isolated antennas.
'Fluorescence quenching titration' is an efficient way to systematically study the pH dependency of quenching in vitro, by injecting purified antennas into buffers of increasingly acidic pH (Ruban et al., 1994b;Wentworth et al., 2000;Kaňa et al., 2012;Belgio et al., 2013). This method was employed  Ruban and Horton 1999). Circles, mean averages from at least three independent replicates; solid lines, fittings; dashed lines, 95% confidence intervals. The error bar shows typical standard deviation.
to assess the hypothesis that faster NPQ activation (Fig. 1) related to antenna protonation, rather than to zeaxanthin content. Therefore LHCII and CLH complexes were isolated from dark-adapted material and the absence of zeaxanthin was confirmed by HPLC analysis (see Supplementary Table S1 at JXB online).
Upon injection, CLH displayed a progressive quenching proportional to the acidity of the buffer (Fig. 3A). Sample integrity was constantly monitored by absorption spectroscopy (Fig. 3, inset) and by reversibility of the quenching after detergent addition (data not shown). Besides the general similarity of the process, pointing to a fundamentally conserved quenching mechanism, the differences between the two types of sample are notable. At each pH value, fluorescence quenching was consistently higher in CLH compared with LHCII, with the biggest difference found around pH 6.0. From the traces in Fig. 3, a titration curve of quenching kinetics as a function of pH was constructed (Fig. 4). It shows that, to attain the same rate of fluorescence quenching, a lower pH is required in LHCII compared with CLH. In particular, almost 50% of maximum quenching rate was observed at pH 5.5 in CLH, whilst a pH of 5.0 was necessary to get the same quenching rate in spinach. Similarly, CLH showed almost the maximum quenching rate (90 ± 5%) at pH 4.97, whereas for LHCII it was only 50%. This was reflected in a shift by 0.5 pH unit to higher values in the calculated quenching pK a of CLH compared with LHCII, i.e. from 5.5 ± 0.1 to 5.0 ± 0.1 (Table 1). The pK a value for LHCII was in good agreement with that previously reported (see e.g. pK a =4.9 in Petrou et al. (2014)). The Hill coefficient for CLH was not significantly different from that of LHCII (7.2 ± 1.9 and 7.5 ± 1.1 for LHCII and CLH, respectively; Table 1) and in both cases they were higher than those for in vivo quenching (see Fig. 1), consistent with the absence of zeaxanthin (see Supplementary Table S1 and Discussion). In summary, the shift in quenching pK a confirmed a higher proton sensitivity of CLH compared with LHCII, independent of xanthophylls.
In order to address possible reasons behind the higher pH sensitivity found in CLH, a comparative in silico analysis was conducted using the amino acid sequences of CLH and LHCII ( Supplementary Fig. S2). A schematic overview of the two proteins is presented in Supplementary Fig. S3. We have explored in particular the protein lumenal loop to identify residues that are protonable within the physiological range. The protein structure predicted for CLH is presented in Fig. 5. We found 24 negatively charged amino acidic residues in total (i.e. aspartic and glutamic acids) in CLH, four of which are located in the luminal loop  and one in the C-terminus .
The estimated in situ pK a values were calculated and compared with LHCII (2.5 Å resolution structure from Standfuss et al. (2005)) and are presented in Table 2. Results for LHCII are in good agreement with a previous report (Xiao et al., 2012), where two residues in particular (Glu-107 and Asp-215) were indicated as putative pH sensors for NPQ as their quenching pK a values are within the thylakoid physiological range (3.9-7.5). The same analysis applied to CLH revealed the presence of three plausible protonable residues: Asp-107, Asp-119 and Glu-205 (see Table 2, Fig. 5 right). Furthermore, their pK a values were shifted to higher pH values compared with LHCII, confirming that the lumenal loop is more sensitive to protonation in CLH (see Asp-107, Asp-119 and Glu-205 and their pK a in Table 2).
An overall comparison between LHCII and CLH protein structures (Table 3) indicated that, despite a similar number of total protonable residues (~11.4-11.5% in both cases), CLH displayed a lower protein charge than LHCII at pH 7.6, that is −6 versus −24, respectively. This means that LHCII tends to be more charged than CLH and a stronger protein-protein repulsion is expected for LHCIIs at pH 7.6 (see Discussion). In agreement with this, the CLH isoelectric point was ~0.4 higher than LHCII, implying that ~30 times fewer protons are required to neutralize negatively charged residues compared with LHCII. In summary, the in silico results supported the experimental data well and provided theoretical explanations for the faster, more efficient quenching found for CLH.

Discussion
In the present paper, we investigated reasons for fast NPQ activation in C. velia. In higher plants, the kinetics of NPQ induction are influenced by xanthophyll composition (Fig. 1B). Demmig-Adams (1990) was the first to provide evidence for a connection between the xanthophyll cycle and NPQ. She showed that the conversion of violaxanthin into zeaxanthin, stimulated under light by lumen acidification, strongly enhanced NPQ. Later it was noticed (Ruban and Horton, 1999) that the NPQ of zeaxanthin-enriched samples was much faster, as zeaxanthin changed the NPQ dependency (cooperativity) as a function of ΔpH, from sigmoidal (violaxanthin) to hyperbolic (zeaxanthin) (see also Horton et al., 2000;Johnson et al., 2009). Here, we confirmed with a control sample (spinach) that the transition into the quenched state is slower for leaves enriched in violaxanthin compared with zeaxanthin (Fig. 1B) as the Hill coefficient Table 1. pH versus quenching titration curve fitting parameters in CLH and LHCII

Sample
Hill coefficient Estimated pK a Quenching kinetics at pH 4.97 LHCII 7.2 ± 1.9 5.0 ± 0.1 0.50 ± 0.01 CLH 7.5 ± 1.1 5.5 ± 0.1 0.90 ± 0.05 Titration parameters were determined by fitting measured traces like those represented in Fig. 3 as described in the 'Data analysis and model fitting' section of 'Materials and methods'. Standard fitting errors were provided by SigmaPlot software. (For more details, see Ruban et al. 1994a;Ruban and Horton 1999;Petrou et al. 2014.) decreased in the presence of zeaxanthin. In C. velia however, we found a different behavior. Although NPQ was greater in zeaxanthin-enriched samples, confirming the first observations (Kotabová et al., 2011), its rate was insensitive to xanthophyll composition (Fig. 1A, C), indicating that the reason for the fast NPQ in C. velia resided elsewhere. The NH 4 Cl-infiltration experiment (Fig. 2) following the procedure of Ruban et al. (2004) and Lavaud et al. (2002), suggested that fast NPQ related to lumen acidification and protons. Incomplete diffusion of the uncoupler was in fact ruled out by previous evidence of efficient NH 4 Cl penetration in C. velia cells (see Fig. 4b in Belgio et al., 2018). The titration of NPQ as a function of pH confirmed that CLH was significantly more sensitive to protons than LHCII (Figs 2, 3). Most importantly, the rate of quenching was increased, especially between pH 5 and 6.5. Within this pH range a very rapid quenching formed almost instantly upon injection of a CLH sample (Fig. 3A). This resulted in a shift in quenching pK a of CLH to higher pH values than LHCII (Table 1). Therefore, to attain 50% of maximum quenching kinetics, a 0.5 unit lower pH value (corresponding to 3.16 times more protons) was required for LHCII than CLH, indicating that the latter had an increased sensitivity to acidification. The high level of structural similarity between CLH and LHC protein families (Pan et al., 2012;Tichy et al., 2013; see also Supplementary Fig. S3) prompted an in silico comparison between lumenal loop residues. The analysis identified five protonable (i.e. negative) residues in the lumenal loop of CLH (Table 2). Three of them were predicted to be protonable within the physiological range (assuming a chloroplast lumen pH ranging between 3.9 and 7.5 (Peltier et al., 2002). Compared with the corresponding residues in LHCII, the quenching pK a values of these three residues are significantly higher in CLH (Table 2), which is consistent with the shift in the quenching pK a found from experimental data (Fig. 4). We propose that during lumen acidification and in vitro quenching (Fig. 3), these residues shift from negative to neutral (i.e. become protonated) and this occurs earlier (i.e. at higher pH values) in CLH than in LHCII. This mechanism would explain the faster quenching (Fig. 3) and the shift to higher values of CLH quenching pK a found experimentally (Fig. 4). Interestingly, the pK a of Lys-211 (which in standard conditions has a pK a of 10.67) was found to be reduced to 7.4 in CLH (Table 2). If confirmed by further studies, this means that this lumenal loop residue can also be protonated during lumen acidification and therefore might play a role in NPQ activation in C. velia. The comparison between LHCII and CLH ( Table 3) indicated also that the algal antenna is less charged at physiological pH than LHCII. As protein clustering proved to be crucial for efficient quenching of LHCII (Betterle et al., 2009;Belgio et al., 2012;Petrou et al., 2014), an overall less charged protein like CLH could be more prone to aggregate and therefore quench more easily, due to a minor protein-protein electrostatic repulsion. The predicted isoelectric point is consistent with this. For CLH, in fact, a ~0.4 lower pI was found (Table 3), corresponding to 30-40 less protons required for charge shielding. Considering the in vitro and in silico results together, we suggest that the increased NPQ formation kinetics relate to inbuilt antenna properties, in terms of  The putative residues that can be protonated within the physiological pH range (3.9-7.5) are in shown bold. Residues in CLH with a higher pK a than LHCII have been marked with an asterisk. Set values in the simulation were: for internal dielectric, 4; external dielectric, 80; and salinity, 0.15; in agreement with Xiao et al. (2012). Predicted sequences and protein structures are shown in Supplementary Figs S2 and S3, respectively.
both a higher number of lumen-exposed protonable residues and an overall increased protein hydrophobicity. We hypothesize that this applies also to other similar antennas, such as diatom FCPs. In fact, although CLH binds only chlorophyll a and xanthophylls (see Kotabová et al., 2011), due to its structural properties, this protein was classified as 'FCP-like', i.e. closely related to antennas from dinoflagellates, brown algae, and diatoms (Lepetit et al., 2010;Pan et al., 2012;Tichy et al., 2013). Moreover diatoms can also be characterized by fast NPQ activation (Ruban et al., 2004;Lavaud and Kroth, 2006;Grouneva et al., 2008). It was experimentally shown for brown algae (Nitschke et al., 2012), alveolates (Belgio et al., 2018), and other microalgae (Goss and Jakob, 2010) that the habitat and particularly the light conditions affect NPQ capabilities of algae from the SAR clade. As a coral symbiont, C. velia is expected to be mainly exposed to rather 'moderate' light intensities.
However, as this organism can be also found 'free-living' outside the coral, at depths of 3-5 m, light intensities of up to 1000 μmol m −2 s −1 are normally experienced on a sunny day (Behrenfeld et al., 1998;Oborník et al., 2011). We can speculate that, due to fast quenching of antennas, in C. velia there was no selective pressure towards proteins capable of enhancing NPQ rate such as PsbS or Lhcsr (Pan et al., 2012). These proteins in fact have a role as NPQ enhancers in vascular plants and green microalgae, respectively (Goss and Lepetit 2015). Spinach and C. velia seem therefore to have evolved very different 'antenna behaviors' in relation to different acclimation strategies. They can be summarized as follows (Fig. 6): • Chromera velia carries antenna proteins that are 'natural quenchers'; PsbS, a strong NPQ enhancer (Li et al., 2000), is absent (Pan et al., 2011) and the thylakoid membrane is highly enriched in violaxanthin, an 'anti-quenching' pigment . As a consequence its NPQ kinetics are characterized by fast formation/slow relaxation. • Higher plant antenna proteins, here represented by spinach LHCII, are 'natural light harvesters', so PsbS is required for effective but in particular fast quenching  and little or no free violaxanthin is present the membranes (Dall'Osto et al., 2010;Xu et al., 2015). The NPQ kinetics are characterized by slow formation/fast relaxation. Fig. 6. Scheme showing the different light harvesting strategies of C. velia and higher plants. The C. velia thylakoid membrane carries CLH proteins that are 'natural quenchers' with three protonable lumen-facing residues, D107, D119, and E205 (indicated by small protrusions). The membrane is highly enriched in unbound, 'anti-quenching', violaxanthin pigments, and PsbS protein is absent. The higher plant thylakoid membrane supports the LHCII protein, a 'natural light harvester' with two protonable lumen-facing residues. PsbS protein is required for effective quenching, and the amount of unbound violaxanthin in the membrane is negligible. The scheme does not represent real stoichiometries/proportions. For more details, see main text. In this scenario, 'free' violaxanthin plays a role of quenching inhibitor, particularly important for C. velia and less crucial for LHCII. A similar role of violaxanthin was previously suggested for some brown algae (Ocampo-Alvarez et al., 2013). It explains the abundance of violaxanthin in algae like C. velia, where the violaxanthin to Chl a ratio is ~0.36 (mol mol −1 ), ~8 times higher than in plants (see e.g. Kotabová et al., 2011), which is supported by work showing quenching modulation by 'free', i.e. not firmly bound to protein, xanthophylls (Ruban et al., 1994a;Lepetit et al., 2010;Mann et al., 2014;Xu et al., 2015;. Finally, the model presented (Fig. 6) provides an explanation also for the unusual acclimation strategy observed in C. velia: whilst plants (carrying 'natural harvester' antennas) protect themselves from high light by reducing their antenna size (see Kouřil et al., 2013), in C. velia (characterized by 'natural quencher' antenna proteins) the antenna size is unaffected even after days of exposure to high light (Belgio et al., 2018). This evidence, at first puzzling, seems now more logical in view of the results presented here.

Conclusions
In conclusion, we have shown a similar quenching mechanism in antennas from a higher plant compared with those from an alveolate. In both cases the trigger is low pH and the likely sensors are protonable lumenal residues. However, the actual sensitivity to lowering pH is different for the two proteins as CLH is more sensitive to protons than LHCII. We propose that this is due to subtle differences in the amino acid composition of the protein lumenal loop. As a result, CLH switches into a dissipative quenched state more easily than LHCII and therefore the higher plant antenna protein can be considered a 'natural light harvester' whilst the CLH protein is a 'natural quencher'.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. NH 4 Cl induces fast NPQ relaxation in spinach chloroplasts. Fig. S2. Sequence of LHCIIb and CLH used in the present study. Fig. S3. Schematic overview of LHCII (left) and CLH (right) antenna protein structures used in the present study. Table S1. Pigment composition of antennas isolated from C. velia and spinach.