Marine diatoms are major oceanic primary producers (Field et al., 1998; Armbrust, 2009) with distinctive biooptical (Key et al., 2010) and metabolic (Armbrust, 2004; Allen et al., 2011; Hopkinson et al., 2011) characteristics, including tight metabolic coupling between their chloroplasts and mitochondria (Prihoda et al., 2012; Bailleul et al., 2015). Like all oxygenic photoautotrophs, diatoms use Photosystem II (PSII) to photooxidize water and generate the reductant that supports their biological productivity. PSII, however, suffers inactivation under illumination (Vass, 2011; Vinyard et al., 2013a,b), at a rate inversely proportional to diatom cell size (Key et al., 2010). This inactivation must be countered through an expensive protein repair cycle (Edelman and Mattoo, 2008; Nixon et al., 2010; Raven, 2011; Komenda et al., 2012) which shows a light-dependency in cyanobacteria and in green plants (Edelman and Mattoo, 2008; Komenda et al., 2012). Under low to moderate illumination the rate of PSII inactivation is a nearly linear function of light level (Tyystjärvi and Aro, 1996; Campbell and Tyystjärvi, 2012) even though at least two mechanisms (Oguchi et al., 2011; Vass, 2011; Tyystjärvi, 2013; Zavafer et al., 2015) drive photoinactivation. As in cyanobacteria (Nixon et al., 2010; Komenda et al., 2012) in marine centric diatoms this repair cycle is mediated by FtsH protease complexes (Campbell et al., 2013), but diatoms can accumulate significant pools of PSII repair cycle intermediates, even under moderate light conditions (Wu et al., 2011, 2012). These sub-pools of PSII components interact with complex regulation of excitation dissipation (Lavaud et al., 2007, 2016; Lavaud and Lepetit, 2013), within a physical context of triply stacked thylakoids (Gibbs, 1962; Drum, 1963; Lepetit et al., 2012) that are distinct from the thylakoid arrangements (Rast et al., 2015) in green (Baroli and Melis, 1996; Edelman and Mattoo, 2008) or cyanobacterial (Singh et al., 2008; Komenda et al., 2012) model species. PSII repair thus represents a significant (Li et al., 2015), light-dependent burden on protein metabolism in diatoms, like other photoautotrophs. Under variable light the instantaneous rate of PSII inactivation can outrun the repair rate, at the expense of drawdown of pools of reserve subunits (Wu et al., 2011, 2012; Lavaud et al., 2016) or pools of active PSII (PSII_active) (Behrenfeld et al., 1998).

Given recent findings of ATP shuttling between mitochondria and chloroplasts (Prihoda et al., 2012; Bailleul et al., 2015) to drive CO₂ fixation in diatoms, we hypothesized that dark repair of PSII, supported by respiration, could confer an advantage upon marine diatoms, which thrive in fluctuating light environments (MacIntyre et al., 2000; Lavaud et al., 2007). We thus measured pools of PSII_active, and the degradation and synthesis rates of key PSII proteins across diel cycles (Chisholm and Costello, 1980) and growth light levels. We find that diatoms maintain an active cycling of PSII subunits during dark periods, and may indeed rely upon dark repair of PSII to balance cycles of inactivation and replacement of PSII.

Across a 16:8 L:D cycle under low growth light (75 μmol photons m⁻² s⁻¹) the small diatom T. pseudonana showed nearly steady contents of PSII_active (Figure 1A) and the core PSII protein PsbA (Figure 1B) in a ~1:1 ratio showing that PSII_active accounts for almost all PsbA protein. This indicates that only small pools of inactive PSII repair cycle intermediates were present under this growth condition.

Under a 16:8 L:D cycle at higher growth light (300 μmol photons m⁻² s⁻¹) the content of PSII_active (Figure 1C) was significantly lower [3-factor ANOVA, df = 1, F = 53.669, p = 3.66 × 10⁻¹⁰, with Tukey Honest Significant Difference post-tests (RStudio)], even though the quantum yield of PSII (Supplemental Figures 4A,C) was only marginally lower. This highlights an important distinction between the pool size for PSII_active and the fluorescence based quantum yield for PSII. At midday PSII_active and PsbA protein content converged at ~58 fmol μg protein⁻¹, before diverging late in the period of high light as PSII_active declined and PsbA protein accumulated (Figure 1D), because PsbA synthesis outran PsbA degradation. During the subsequent dark period PSII_active dropped significantly [3-factor ANOVA, df = 1, F = 14.418, p = 0.000315 with Tukey Honest Significant Difference post-tests (RStudio)] to only ~30% of the pool of PsbA protein, with a large pool of reserve protein that is not part of a PSII_active (compare Figures 1C,D, dark period). The content of PSII_active then increased from the late night measurement to measurements at dawn (Figure 1C), even when PsbA protein synthesis was blocked by the chloroplast protein translation inhibitor lincomycin (Supplemental Figures 1C,D). Thus, in the dark the cells contained a reserve of PSII protein available for assembly into PSII_active around dawn, even when new protein synthesis was blocked. PsbA protein content then dropped again in the morning as PSII_active and PsbA protein contents converged (Figure 1D). Analogous experiments with T. pseudonana under 8:16 and 4:20 diel cycles (Supplemental Figure 1) showed similar responses. The parallel set of experiments on the larger diatom T. punctigera (Supplemental Figure 2) showed generally dampened diel responses and a lower fraction of the PsbA pool in the PSII_active pool under low growth light (Supplemental Figures 2A,B,E,F).

The content of the key FtsH protease that mediates removal of PsbA (Nixon et al., 2005; Campbell et al., 2013) was steady through the diel cycle in both T. pseudonana and T. punctigera, while T. pseudonana showed a significant up-regulation of FtsH protein under the higher growth light [2-factor ANOVA, df = 1, F = 4.757, p = 0.032 with Tukey Honest Statistical Difference post-tests (RStudio)] (Supplemental Figures 7A–D).

We estimated the rate constants for inactivation of PSII, k_pi and for the countering activation, k_{psii activation} of new units of PSII_active (Figures 2A,B), by tracking [PSII_activefluor] over 7200 s time intervals, with or without blockage of chloroplast protein synthesis by lincomycin, for T. pseudonana (Supplemental Figure 1) and for T. punctigera (Supplemental Figure 2). We repeatedly estimated these rate constants over 7200 s periods during darkness (Figures 2A,B closed bars), moderate low growth light (Figures 2A,B, shaded bars) or higher growth light (Figures 2A,B, open bars). 3-way ANOVA found no significant effects of culture photoperiod length, nor time within the photoperiod, upon the rate constants, but there were significant effects of dark vs. light during the measurement period, and of the growth light level. Therefore, to increase statistical power for the estimation of rate constants Figure 2 pools data from multiple determinations on cultures grown under 16:8 (Supplemental Figures 1, 2), 8:16 (Supplemental Figure 1) and 4:20 (Supplemental Figures 1, 2) L:D diel cycles. The smaller T. pseudonana shows higher rate constants for PSII inactivation under higher light growth [3-factor ANOVA, df = 1, F = 534.448, P <2e-16, with Tukey Honest Significant Difference post-tests (RStudio)] consistent with the higher intrinsic susceptibility of smaller diatoms to photoinactivation of PSII_active (Key et al., 2010; Wu et al., 2011, 2012). T. pseudonana also shows a surprising rate constant for inactivation of PSII_active even in the dark (Figure 2A, dark bars), since measurable loss of PSII_active was observed relative to total protein pools during dark periods in this species. The counteracting rate for activation of new units of PSII_active, k_{psii activation}, did not show a statistically significant response to culture photoperiod length, nor to time within the photoperiod, but did show a significant increase under higher growth light [3-factor ANOVA, df = 1, F = 59.91, P = 0.0015,with Tukey Honest Significant Difference post-tests (RStudio)] compared to moderate light and during the photoperiod vs. darkness [3-factor ANOVA, df = 1, F = 19.56, P = 0.0115,with Tukey Honest Significant Difference post-tests (RStudio)].

The rate constants for removal of PsbA protein, k_{PsbA degrade}, again measured in the presence of lincomycin to block counteracting synthesis of PsbA (Supplemental Figures 1, 2), did not vary significantly across determinations in darkness, low or higher light for T. pseudonana (Figure 2C) nor for T. punctigera (Figure 2D). Under low light k_{PsbA degrade} was larger than k_pi in both T. pseudonana (Welch Two Sample t-test, t = −4.529, df = 43.254, p = 4.612 × 10⁻⁵) (Figures 2A,C) and in the larger T. punctigera (Welch Two Sample t-test, t = −6.586, df = 38.453, p = 8.5 × 10⁻⁸) (Figures 2B,D). Furthermore, the k_{PsbA degrade} was larger than the protein specific growth rates presented for each combination of species and growth light, for comparison (Figures 2C,D).

We estimated rate constants for synthesis of PsbA protein, k_{PsbA synth}, by comparing changes in PsbA content in the absence and in the presence of the chloroplast translation inhibitor lincomycin for T. pseudonana (Supplemental Figure 1) and for T. punctigera (Supplemental Figure 2) over 7200 s periods repeated through the diel cycle. k_{PsbA synth} showed significant PsbA synthesis during the dark period, particularly after the higher daytime growth light level. In parallel with our analyses of degradation and synthesis of PsbA we also analyzed degradation and synthesis of the PsbD core protein of PSII (Supplemental Figure 5). As expected in these diatoms (Wu et al., 2011, 2012) turnover of PsbD was of comparable magnitude to the turnover of PsbA, although the kinetics of the turnover of the two proteins were not tightly correlated on a sample by sample basis.

To analyze the cumulative diel progression of the PSII repair cycle we plotted cumulative inactivation of PSII (Figure 3), cumulative removal of PsbA protein (Figure 3) and cumulative synthesis of PsbA protein (Figure 3) over a 16:8 L:D diel cycle starting at 08:00 with low growth light (Figures 3A,B) or high growth light (Figures 3C,D) for T. pseudonana (Figures 3A,C) and for T. punctigera (Figures 3B,D). These trajectories are based upon the pools of PSII_active and PsbA over the diel cycle (Figure 1; Supplemental Figures 1, 2), along with the rate constants estimated for loss or synthesis of the pools estimated for the illumination levels and species (Figure 2). Two-way ANOVA showed that growth light level (Df = 1, F = 9.048, Pr = 0.011) and species (Df = 1, F = 12.841, Pr = 0.004) both showed significant influences on cumulative PSII inactivation. In contrast growth light or species did not show significant effects upon cumulative PsbA degradation, nor upon cumulative PsbA synthesis. In T. pseudonana under low light (Figure 3A) less than 100 fmol PSII_active μg protein⁻¹ suffered inactivation up to the end of the 16 h photoperiod at 24:00 (Figure 3A, closed circles). Over the same 16 h of low light illumination ~200 fmol PsbA μg protein⁻¹ was cleared (Figure 3A, open squares), while ~270 fmol PsbA μg protein⁻¹ was synthesized. Therefore, the cells were degrading and replacing protein subunits from PSII units that had not suffered inactivation. Furthermore, during the subsequent dark period PSII inactivation stopped but PsbA protein turnover continued, so that over the 24 h cycle ~350 fmol PsbA μg protein⁻¹ was degraded and replaced through synthesis, while only ~100 fmol PSII μg protein⁻¹ was inactivated (Welch Two-Sample t-test df = 13.805, p = 0.0107).

In T. pseudonana under high light illumination (Figure 3C) inactivation approximately balanced PsbA clearance and replacement over 24 h, with PsbA turnover continuing during the dark hours to catch up to photoinactivation incurred during the photoperiod and no significant difference between diel PSII photoinactivation and diel PsbA degradation.

For T. punctigera the lower k_pi than T. pseudonana [3-factor ANOVA, df = 1, F = 21.235, P = 2.52 × 10⁻⁵,with Tukey Honest Significant Difference post-tests (RStudio)] (Figure 2B) and lower content of PSII_active (Supplemental Figure 2A) meant much less cumulative inactivation of PSII_active over the 16 h high light illumination up to 24:00 (Figure 3D). In principle this could have lowered the metabolic repair costs for T. punctigera because fewer PSII inactivations mean lower requirements for PSII protein turnover (Li et al., 2015). In fact cumulative diel PsbA degradation (Figures 3B,D) and diel PsbA synthesis (Figures 3B,D) in T. punctigera again significantly exceeded cumulative diel inactivation under both low (Welch Two-Sample t-test df = 17.441, p = 7.76 × 10⁻⁵) and high growth light (Welch Two-Sample t-test df = 17.983, p = 5.8 × 10⁻⁴), showing that the larger T. punctigera was also turning over PsbA proteins even for PSII_active that had not suffered inactivation.

Cumulative diel PsbA protein turnover in these strains is already light saturated at a growth light of 75 μmol photons m⁻² s⁻¹, consistent with findings in other organisms (Edelman and Mattoo, 2008) that PsbA protein turnover saturates under low light. For T. pseudonana growing under low light much of the daily clearance of PsbA took place during the dark period following illumination (Figure 3A, open circles and squares). For T. pseudonana the metabolic costs of PSII protein turnover were actually similar under low and under higher light, because even though photoinactivations were slow under low light (Figure 2A), the cells under low light contained more PSII_active (Figure 1A), so the total diel burden of PSII turnover was similar under low and high light. For T. punctigera cumulative PSII protein turnover exceeded cumulative PSII inactivation, under both low and high growth light (Figures 3B,D). In contrast we found no evidence for statistically significant rates of turnover of the PsaC subunit of Photosystem I over the same measurement periods (data not presented), suggestin that in these cells turnover of Photosystem I proteins is negligible compared to PSII protein turnover, at least up to moderately high growth lights.

Labeling experiments, reviewed in Edelman and Mattoo (2008) showed that in model plants and in cyanobacteria (Nixon et al., 2005) PSII protein turnover is largely light dependent and reaches rate saturation under low light. In our measures of PSII protein turnover in marine centric diatoms we find that significant PsbA and PsbD degradation, and some synthesis, continue during darkness. Indeed dark turnover represents a large fraction of total diel cycling of PSII proteins in the diatoms tested. In the small diatom Thalassiosira pseudonana the burden of PSII cycling upon chloroplastic protein metabolism was actually similar under low light, with slow photoinactivation per PSII but large pools of PSII, and under higher growth-saturating light with faster photoinactivation per PSII but smaller pools of PSII. In the larger diatom T. punctigera cumulative diel turnover of PSII proteins consistently outran the cumulative inactivation of PSII_active. Thus, in diatoms PSII inactivation (Komenda et al., 2007) is not a necessary precursor step to protein degradation and the chloroplasts in these large diatoms are degrading subunits taken from PSII that have not been inactivated. The diatoms, like other chlorophyll a/c organisms, have a triply layered thylakoid organization (Gibbs, 1962; Drum, 1963; Lepetit et al., 2012) that may place kinetic limitations upon access of the FtsH protease (Komenda et al., 2012) to PSII, in contrast to the single layer thylakoid arrangement in cyanobacteria, or the grana/stroma thylakoid organization of higher plants. The 1–2 ATP per peptide bond to drive chloroplastic protein degradation (Nixon et al., 2010; Raven, 2011; Komenda et al., 2012; Campbell et al., 2013) in the dark must derive from the tight coupling of chloroplastic and mitochondrial ATP and amino acid metabolism in diatoms (Allen et al., 2011; Prihoda et al., 2012; Bailleul et al., 2015). This dark PsbA turnover capacity is particularly evident in the small, coastal strain T. pseudonana in this study. We suggest that diatoms have evolved an ongoing PSII protein turnover in the dark as an adaptation to match kinetic limitations on their instantaneous PSII protein turnover, with fluctuating instantaneous photoinactivation rates in variable light environments. In another study (Li et al., 2016) we show that for the larger T. punctigera dark periods are actually necessary for the species to achieve maximal growth rates. The diatoms may thus offset the burden of PSII repair upon ATP, protein and amino acid metabolism to dark periods. This allows transient burn down of reserve pools of PSII_active during intervals of high illumination (Behrenfeld et al., 1998), with a subsequent clean up in the succeeding dark period.

GL conducted all growth and culturing experiments, harvested all samples for subsequent analyses, conducted all photophysiological measures, did some of the PsbA quantitations, generated draft figures and curve fits, and contributed to the text. AW conducted quantitations of FtsH and some of the PsbA quantitations, generated draft figures, contributed to the text and performed the statistical analyses with DC. ND conducted quantitations of PsaA and PsaC, managed the reagents for protein quantitations and contributed to the text. AC supervised the protein quantitation work and contributed to the text. DC supervised the culturing and photophysiology work, contributed to figure drafting and curve fitting, wrote the draft of the text and performed the statistical analyses with AW. All authors edited and revised the manuscript.