Pyrenoid loss in Chlamydomonas reinhardtii causes limitations in CO2 supply, but not thylakoid operating efficiency

The Chlamydomonas reinhardtii pyrenoid is a Rubisco-containing microcompartment primarily responsible for facilitating carbon accumulation and does not affect thylakoid membrane photosynthetic energetics.


Introduction
The pyrenoid of the model green alga Chlamydomonas reinhardtii has long been implicated in a number of functions, most notably the establishment of a carbon-concentrating mechanism (CCM; Badger et al., 1980;Wang et al., 2015). The pyrenoid may also be needed to maintain structural aspects of the chloroplast, including thylakoid membrane organization and translation of proteins and their assembly into complexes, which was found to be localized to specialized zones at the pyrenoid periphery Zerges, 2007, 2009). Disruption of these pyrenoid-associated processes could potentially impact chloroplast energetics and thus photosynthetic efficiency.
CCMs have arisen multiple times, mostly in aquatic photosynthetic organisms, as a means of increasing carbon fixation under conditions in which CO 2 availability limits turnover of the Calvin-Benson-Bassham cycle (CBBC; Raven et al., 2012;Meyer et al., 2016). Active transport is used to concentrate dissolved inorganic carbon around the primary carboxylase Rubisco, which allows for more rapid ribulose bisphosphate carboxylation and limits photorespiratory activity, leading to an increase in the ratio of carboxylation to oxygenation. For the CCM to function in Chlamydomonas, Rubisco must be confined to a chloroplast microcompartment, the pyrenoid. Assembly of Rubisco into a pyrenoid is controlled by the linker protein EPYC1 (Mackinder et al., 2016), previously designated LCI5 (Lavigne et al., 2001), and two Rubisco small subunit (RBCS) surface α-helices (Meyer et al., 2012). In Chlamydomonas RBCS substitution strains expressing vascular plant RBCS, Rubisco fails to assemble into pyrenoids, rendering these pyrenoid-less (pyr-) cells unable to establish a functional CCM (Genkov et al., 2010;Meyer et al., 2012).
The pyrenoid may also play a role in processes other than the CCM, as suggested by the finding that it is not completely eliminated from wild-type (WT) cells when the CCM is repressed (Borkhsenious et al., 1998). While WT cells redistribute a significant fraction of Rubisco throughout the chloroplast stroma in CCM-repressive conditions, such as at elevated CO 2 or in the dark, at least 50% of cellular Rubisco remains aggregated in the pyrenoid at all times (Borkhsenious et al., 1998;Mitchell et al., 2014). Additionally, several proteins unrelated to the CCM have been localized to pyrenoids, including nitrite reductase (Süss et al., 1995) and nucleic acid processing enzymes (Shukla et al., 2012;Zhan et al., 2015). Eliminating pyrenoid formation may also alter thylakoid membrane structure since distinct thylakoid domains intersect and coalesce within the pyrenoid (Ohad et al., 1967;Goodenough and Levine, 1970;Engel et al., 2015). Finally, the dispersion of Rubisco throughout the stroma might be expected to favour entropically mediated thylakoid stacking (Chow, 1999;Kim et al., 2005). Indeed, initial observations (Genkov et al., 2010;Meyer et al., 2012) suggested that thylakoid membrane hyperstacking occurred in pyr-cells, perhaps reflecting altered photosynthetic energetics in the absence of a functional pyrenoid (Goodenough and Levine, 1969;Chow et al., 2005;Anderson et al., 2008).
The aim of the present study was to understand whether the reduction in growth and photosynthesis, when pyrenoid formation was compromised, was caused directly by the loss of inorganic carbon accumulation capacity, or was also associated with altered thylakoid organization and therefore reduced photosynthetic energetic efficiency. Three pyr-lines, generated by complementing a Chlamydomonas mutant lacking both native RBCS genes with genes encoding vascular plant RBCS (Spinacia, Helianthus, and Arabidopsis;Genkov et al., 2010), were used in the present study. The degree of thylakoid stacking was examined via electron microscopy to determine whether the absence of a pyrenoid altered thylakoid ultrastructure. Furthermore, photosynthetic activity was analysed using a combination of advanced spectroscopic techniques and chlorophyll (Chl) fluorescence measurements. Ultrastructure and whole-cell physiology were studied under contrasting light and CO 2 regimes.
A sole CCM defect limiting the photosynthetic activity of pyr-mutants can be distinguished from additional energetic limitations acting in concert, based on the physiological response of cells to various light and CO 2 regimes. Production of the high energy metabolites NADPH and ATP via the photosynthetic electron transport chain needs to be balanced with their consumption via the CBBC (Allen, 2003;Eberhard et al., 2008;Foyer et al., 2012;Lane, 2014). If CBBC consumption is reduced because the CCM is absent or impaired, it is reasonable to assume that upstream activity of the electron transport chain would be restricted and the cells would experience a higher photon load than in the presence of a CCM. Excess energy would probably be dissipated via non-photochemical quenching (NPQ), which can be monitored by analysing Chl fluorescence (Krause and Weis, 1991;Lazár, 1999;Maxwell and Johnson, 2000;Papageorgiou and Govindjee, 2004). A photosynthetic defect caused by the loss of the CCM should be minimal at low light, when the supply of photon energy becomes more rate limiting than the supply of CO 2 (von Caemmerer, 2013;McGrath and Long, 2014), and any difference between mutants and the WT should be abolished at elevated CO 2 (Fukuzawa et al., 2001;Xiang et al., 2001;Jungnick et al., 2014). In contrast, increased thylakoid stacking or defects in photosynthetic protein complex assembly in the pyrmutants may act to decrease light absorption and thus electron transport chain activity (Falkowski and Raven, 2007), rendering pyr-more susceptible to light limitation. In this case, differences from the WT should be exacerbated at low light, and persist in the presence of elevated CO 2 .
The work presented in this study shows that the pyrenoid can be functionally defined as a CCM component (helping supply a high concentration of CO 2 to aggregated Rubisco). The pyr-mutants were not impaired in structural or energetic features based on rescue of all pyr-phenotypes once a sufficient supply of CO 2 was provided to the cells externally. These findings go hand in hand with novel proteomic work on the same mutants (Mitchell et al., 2017) showing that the effect of pyrenoid absence is limited to the CCM and downstream metabolism, with the large majority of proteins expressed at the same levels in the pyrenoid mutants and WT.

Materials and methods
Strains and culture conditions Pyr-lines were previously generated by expressing the vascular plant RBCS gene, from either Spinacia oleracea, Helianthus annus, or Arabidopsis thaliana, in the ∆∆-RBCS1,2 host Chlamydomonas strain CC-4415 (devoid of the two native RBCS proteins as well as of the state transition regulatory kinase STT7); the control WT cells express the Chlamydomonas native RBCS1 from the same construct in the same ∆∆-RBCS1,2 host (Genkov et al., 2010). An additional pyrenoid-forming line was provided by the HelixAB strain, which makes a chimeric RBCS containing the Chlamydomonas amino acid sequence for the surface α-helices within Spinacia RBCS (Meyer et al., 2012). The resulting holoenzymes assemble into a pyrenoid, but display impaired Rubisco kinetic properties, including a 10-fold reduction in the maximum carboxylation rate.
Cells were grown under continuous illumination at 25 °C in Trisminimal medium (Spreitzer and Mets, 1981). Generally, cultures were maintained under 5% CO 2 and subjected to experimental CO 2 conditions overnight prior to any measurements. Standard conditions are defined here as growth at standard light (SL, 50 µmol photons m −2 s −1 ) and acclimation to air levels of CO 2 (0.04%) for ≥12 h. Experimental treatments included growth at low light (LL, 10 µmol photons m −2 s −1 ) or higher light (HL, 100 µmol photons m −2 s −1 ), and/or with 5% CO 2 supplementation. Cells were harvested for experiments from mid-log phase cultures, which were at ~2 × 10 6 cells ml −1 or 3-4 µg Chl ml −1 ; Chl levels were quantified according to Wellburn (1994).
To establish growth rates (Fig. 5A), three biological replicates for the WT (taken from two independent Chlamydomonas RBCS1 insertion lines) and pyr-(one expression line each using the Spinacia, Helianthus, and Arabidopsis RBCS constructs) were maintained in 50 ml of liquid medium in experimental conditions (SL/5% CO 2 , SL/ air, LL/air) for 6 d in continuous culture. Every 1.5 d, Chl concentrations were measured and cultures were either diluted with fresh medium or supplemented with additional cells from a different culture, in order to attain cultures with mid-log phase growth densities by the next measurement.
Electron microscopy WT and pyr-(Spinacia RBCS strain) samples were fixed for electron microscopy as previously described (Genkov et al., 2010), except that Tris-minimal medium was used instead of PIPES for the first fixation step. Material was imaged using a Field Emission scanning electron microscope (FEI Verios 460L) or a Tecnai G2 80-200 kV transmission electron microscope. Using the latter, 20 cells per experimental condition, each with diameters >4.5 µm, were sampled randomly for quantification of thylakoid stacking. Images were indexed, pooled, and presented in random order for blind analysis. For each image, widths of five representative appressed regions were quantified as the width perpendicular to the orientation of lamellae, using the image processing system Fiji (Schindelin et al., 2012). Statistical analyses of these-and other-results were performed using R version 3.2.4 (The R Foundation, Vienna, Austria).

Chl fluorescence from algal colonies on solid agar medium
To generate algal colonies on solid agar medium, cultures grown in the dark in liquid Tris-acetate phosphate medium (Spreitzer and Mets, 1981) to a density of ~1.5 µg Chl ml −1 were concentrated to 100 µg Chl ml −1 in liquid Tris-minimal medium. Replicates of 20 µl were spotted in random order into individual wells of cell culture plates (24 wells) containing solidified Tris-minimal medium +1.5% (w/v) Bacto-Agar. After 4 d of growth under the experimental conditions (LL or SL, air or 5% CO 2 ), the cells were darkadapted for at least 2 h during transport between Cambridge and Colchester. Experimental gas treatments were sustained over the course of the Chl fluorescence measurements, which were made using a Chl fluorescence imaging system (CFimager, Technologica Ltd, Colchester, UK). In order to capture a range of photosynthetic characteristics rapidly, fluorescence parameters were recorded during an initial 20 min fluorescence induction at 106 µmol photons m −2 s −1 , followed by a 5 min dark period to gauge NPQ relaxation, and finally a light intensity response curve (22,46,106,170,251,356,509, and 679 µmol photons m −2 s −1 ), with each intensity step lasting 2 min. For presenting the information, PSII operating efficiency (ϕ II ) values were transformed to electron transport rates (ETRs) as ETR=ϕ II ×I×a II , where I is the light intensity to which the cells were exposed, and a II is assumed to be 0.84 × 0.5=0.42, accounting for light absorption (0.84) and distribution between the photosystems (0.5), respectively (Baker, 2008).
A suitable form of the mathematical model proposed by Suggett et al. (2003;Equation 1) was fitted to the light response data using Bayesian methods. In its original form, the Suggett et al. (2003) model (Equation 1) describes the change in ϕ II in response to changes in I in terms of the model parameters ϕ m and I k , where ϕ m is the maximum PSII operating efficiency and I k is the light intensity at which photosynthesis becomes light saturated. (1) Given that ϕ m =σ II /a II (Suggett et al., 2011) and that I k is defined as I k =ETR max /σ II (Blackman, 1905), it follows that ϕ m ×I k =ETR max /a II . The relationship can thus be re-expressed in terms of the maximum electron transport rate (ETR max ; Equation 2), a form that can be more easily interpreted in the context of ETR data (Fig. 2).
The version of the model shown in Equation 2 was fitted to ϕ II data by assuming the experimental data points were normally distributed around a deterministic skeleton from the model, allowing for different values of the parameters ETR max and I k for each experimental treatment. The likelihood of the data was then calculated by assuming that all data points were independent, treating the variance of the normal error as an additional nuisance parameter to be estimated (separately for each treatment).
Bayesian methods were used to estimate log-transformed values of model parameters, assuming uninformative improper priors in all cases. The joint posterior of the parameters was estimated using the freely available function MCMCpack (Martin et al., 2011) in R, taking 1 000 000 samples after discarding the first 20 000 as burn in. Proper convergence was assessed by visual inspection of trace plots and by checking that results were unaffected by chain initial conditions. Credible intervals for the mean response of WT and pyr-to light in terms of ETR were calculated-at each value of the independent variable-by considering the distribution formed by taking 10 000 samples from the joint posterior distribution and using them in Equation 2, and then transforming the predicted ϕ II values to ETR as described above.
To test whether model parameters differed between pairs of experimental treatments, posterior probability distributions were estimated for differences in parameters between pairs of experiments (i.e. ΔETR max or ΔI k ) using 100 000 pairs of samples from the posterior distributions. Since this procedure leads to a full probability distribution, it can be used to estimate 95% confidence intervals (CIs) for any difference in parameter values, as well as the probability that the difference in parameters is less than zero (Kruschke, 2011). In interpreting the results, it should be noted that the latter probability would be 0.5 if there were absolutely no difference between the two treatments, meaning that-to the extent it is possible to compare the Bayesian and frequentist inferential frameworks-any value <0.025 is equivalent to a 'significant' difference.

Spectroscopy
For measurements using the Joliot-type spectrophotometer (JTS-10, Bio-Logic Science Instruments SAS, Claix, France), cells grown in liquid medium were suspended in 20 mM HEPES (pH 7.2) containing 10% (w/v) Ficoll at a density of 10-12 µg Chl ml −1 . Where applicable, cultures were split after dark adaptation and half of the culture was supplemented with sodium bicarbonate to a final concentration of 10 mM. Chl fluorescence parameters were estimated using standard procedures (Maxwell and Johnson, 2000) and functional PSI/PSII ratios were estimated from the electrochromic shift (ECS) absorbance (520 nm) using single turnover flashes (Joliot and Delosme, 1974;Bailleul et al., 2010). For a subset of experiments, xenon bulb flashes were used, which were later estimated to have resulted in 1.74 PSI turnovers on the basis of laser flash recordings. Additionally, ECS traces were used to estimate total electron flow rate (TEF; Joliot and Joliot, 2002;Bailleul et al., 2010;Lucker and Kramer, 2013). Fluorescence saturation was measured in samples treated with the PSII inhibitor DCMU (20 µM) to obtain a measure of the functional absorption crosssection of PSII based on the time taken to reach two-thirds of maximal fluorescence (t 2/3 ). In the presence of the PSII inhibitors hydroxylamine (1 mM) and DCMU, P700 oxidation-reduction kinetics were recorded for quantification of total PSI (Hiyama and Ke, 1972;Alric, 2010;Johnson and Alric, 2012), which was used to calculate Chl per PSII. This analysis takes into account the ECS-based PSI/PSII ratio and the total Chl content of the cells, and assumes that each PSI contains 240 Chls (Drop et al., 2014). Alternatively, Chl per PSII was estimated from the Chl a/b ratio according to Drop et al. (2014), assuming a PSI/II ratio of 1.
The same Bayesian approach was used to fit ETR, TEF, and NPQ light response curves as described above for Chl fluorescence data from algal colonies, similarly using the model shown in Equation 2 on ϕ II data on which the ETR curves are based. This relationship was used to derive a modified version for TEF, based on the definition of ETR and assuming ETR≈TEF, as shown in (4) Equation 4 shows the model used for NPQ data (Serôdio and Lavaud, 2011), where NPQ max is the maximum level NPQ that can be attained, I 50 is the light intensity at which half-maximal NPQ is reached, and n is the Hill coefficient that controls the sigmoidicity of the curve.

Results
Thylakoid membrane organization is not affected by absence of the pyrenoid TEM was used to investigate any effect of the pyrenoid on thylakoid ultrastructure (Fig. 1). In WT cells, the pyrenoid is an electron-dense body positioned at the base of the cup-shaped chloroplast and surrounded by a starch sheath (Fig. 1A, C,  E). Pyr-cells exhibit starch granules in the region where the pyrenoid usually assembles (Fig. 1B, D, F). Thylakoids appear as single lamellae or appressed in stacks that often run parallel to the chloroplast envelope and occupy most of the stroma. The extent of stacking, quantified by measuring the width of several appressed regions per cell across 20 randomly chosen cells, was found to be very similar beween the WT and pyr- (Fig. 1G; P=0.912).
WT pyrenoids are traversed by a complex network of modified thylakoid tubules that fuse at the centre of the microcompartment (Ohad et al., 1967;Goodenough and Levine, 1970;Engel et al., 2015). Block face SEM revealed that this network is retained in pyr-mutants (Fig. 1H, see Supplementary  Fig. S1 at JXB online for further images). The tubules appear to meet at the same position as in WT cells, close to where the centre of the pyrenoid would have been. Other than the absence of the pyrenoid matrix, chloroplast ultrastructure thus appears unaltered in pyr-cells.

Differences in ETR between the WT and pyr-are cancelled when CO 2 is not limiting
For an initial investigation into the nature of photosynthetic energetics in the absence of a functional pyrenoid, fluorescence parameters were estimated from WT and pyr-colonies grown on agar in contrasting light and CO 2 regimes and subjected to actinic light intensities of 20-700 µmol photons m −2 s −1 (Fig. 2). Use of a camera-based system ensured that the majority of the signal originates from the topmost layer of cells that have plentiful access to light and air. At ambient CO 2 , the pyr-mutant showed lower ETR through PSII than the WT (Fig. 2A, B). This impact was most pronounced when cells had been grown at SL (50 µmol photons m −2 s −1 ; Fig. 2B) rather than LL (10 µmol photons m −2 s −1 ; Fig. 2A), consistent with limited induction of the CCM in the latter condition as light is more limiting than CO 2 . In contrast, ETR was similar for WT and pyr-cells at 5% CO 2 (Fig. 2C, D).
For comparison with pyr-, a second mutant defective for CO 2 fixation, designated HelixAB, was characterized.
This strain does form a pyrenoid but suffers from impaired Rubisco kinetic properties owing to the chimeric nature of the RBCS in that strain (Meyer et al., 2012). Like pyr-, HelixAB consistently showed significant impairment of TEF and ETR ( Fig. 3A, B) relative to the WT [P(ΔETR max ≤0)=1.8 × 10 -2 ; P(ΔTEF max ≤0)=6 × 10 -5 ] as well as increased NPQ (Fig. 3C). That the phenotype exhibited by pyr-is similar to that of HelixAB, which is limited by the enzymatic activity of Rubisco, suggests that the pyr-defect results from a similarly reduced CBBC turnover as a consequence of being unable to supply saturating CO 2 to Rubisco.
Elevated CO 2 restores pyr-photosynthetic performance via CBBC feedback: To determine the short-term response to elevated CO 2 , saturating levels were generated by the addition of sodium bicarbonate (10 mM final, equivalent to aeration with ≥1% CO 2 ) during spectroscopic measurements of a fluorescence induction time-course at saturating light (350 µmol photons m −2 s −1 ; cf. I k , Figs 2, 3). In the absence of bicarbonate, air-acclimated pyr-cells showed lower ϕ II than the WT (Fig. 4A-C, solid lines, P=1.46 × 10 -5 ), in line with the previous results (Figs 2, 3). In the presence of bicarbonate (dashed lines), pyr-samples exhibited values of ϕ II comparable with those of WT cells (P=0.674) following ≥30 s of illumination (Fig. 4B, C, dashed lines). An initial >5 s lag (Fig. 4A) reappears after each short dark incubation in repeated measurements on the same cells (data not shown), ruling out the idea that the effect is due to a slow entrance of bicarbonate into the cells. Rather, the lag is consistent with activation of the CBBC, which has been well characterized in fluorescence induction experiments (Krause and Weis, 1991;Lazár, 1999;Maxwell and Johnson, 2000;Papageorgiou and Govindjee, 2004). These results thus suggest that the limitation observed in the mutant cells is imposed by slow CBBC turnover under low CO 2 , which is overcome when saturating levels of CO 2 are supplied externally.

Changes in photosystem and accessory pigment composition occur in both the WT and pyr-during acclimation
Levels of Chl, carotenoids, and the photosystems were quantified for the same cultures used for the fluorescence induction time-course measurements (Fig. 4). The WT and pyr-exhibited an equivalent relationship between Chl content and culture density (Fig. 5A, P=0.232). The ratio of carotenoids to Chl (Fig. 5B), quantified according to Wellburn (1994), increased significantly (P=2.85 × 10 -9 ) with growth light intensity, but was also similar between the WT and pyr-(P=0.38). PSI/PSII ratios Fig. 2. Under CCM-inducing conditions, pyr-shows a reduction in CO 2 -limited photosynthesis (ETR max ). Light response curves were collected using a Technologica CFimager from algal colonies grown on agar plates in LL (low light, 10 µmol photons m −2 s −1 ) or SL (standard light, 50 µmol photons m −2 s −1 ) in air or 5% CO 2 , as indicated. The ETR is shown as the mean of six biological replicates ±SE, overlaid with 95% confidence intervals derived from fitting a light response model (Equation 2). Probability density plots of marginal posteriors for the fit parameters ETR max and I k are shown in the figure margins. Fig. 3. High-resolution ECS and Chl fluorescence measurements reveal feedback limitation by Rubisco catalysis. Panels show JTS-10 data of (A) ETR through PSII estimated on the basis of Chl fluorescence data, (B) TEF estimated on the basis of ECS data, and (C) NPQ. Data are the mean ±SE based on ≥3 biological replicates each, overlaid with 95% confidence intervals derived using a Bayesian approach to capture the underlying physiological dynamics in terms of model parameters. ETR max , TEF max , and NPQ max describe the maximum attainable levels of each photosynthetic indicator. I k is the light saturation point, while I 50 is the light half-saturation point. Finally, n is the Hill coefficient which controls the sigmoidicity of the NPQ curve, and can be indicative of allosteric regulation when n>1. Probability density plots of the marginal posteriors for fit parameters are shown in the figure margins. Cells were grown in liquid culture in the absence of aeration at 50 µmol photons m −2 s −1 . Fig. 4. Addition of bicarbonate restores photosynthetic performance in air-acclimated pyr-cells. PSII operating efficiency (ϕ II ) was measured in the JTS-10 after 5, 30, and 120 s at ~350 µmol photons m −2 s −1 (to ensure saturation; see I k ; Figs 2, 3) as detailed above the data panels. Cells were grown in air at a range of light intensities (LL, SL, and HL) as shown on the x-axis. Cultures supplemented with 10 mM sodium bicarbonate directly before measurements are shown by dashed lines. Data are the mean of three biological replicates ±SE. P=2.1 × 10 -14 ). Levels of Chl a relative to Chl b (Fig. 5D) were highest in HL (P=1.5 × 10 -5 ) and, again, similar in the WT and pyr-(P=0.503). Consistently, the allocation of Chl to Fig. 5. Accumulation of pigments undergoes physiological acclimation independent of pyrenoid phenotype. Chl expression on a per cell basis (A) is shown as a scatterplot of individual measurements. Starting with the accumulation of carotenoids relative to Chl (B), data are plotted against growth light intensity and shown as the mean of three biological replicates ±SE. The functional PSI/PSII ratio (C) and the Chl a/b ratio (D), respectively, enable quantification of Chl allocation to PSII from spectroscopic data (E) and Chl extraction data (F). Fluorescence saturation kinetics in the presence of DCMU (t 2/3 ) provide a proxy for functional PSII antenna size (G). Growth rates (H) are shown as percentage change in Chl ml −1 over 24 h as mean ±SE based on four measurements over the course of 6 d of three biological replicates per condition grown in quasi-continuous culture; three asterisks signify statistical significance at the level of P<0.01 estimated through ANOVA.
In contrast, functional association of Chl and PSII (Fig.  5G), measured as PSII absorption cross-section (t 2/3 , an inverse proxy; see the Materials and methods), was reduced in air-acclimated pyr-as compared with the WT (P=2.33 × 10 -3 ) at both SL and HL. No such difference was observed for cells grown at LL (P=0.245) or 5% CO 2 (P>0.999). A similar pattern was observed for growth rates, quantified as percentage change in Chl ml −1 d −1 (Fig. 5H), which differed significantly between air-acclimated WT and pyr-at SL (P=0.009), but not when the light intensity was reduced to LL (P=0.86) or the CO 2 concentration increased to 5% (P=0.94). Thus pyrcells experience the same changes as the WT in their accumulation of photosynthetic pigments during acclimation to the various growth light regimes, despite marked differences in their growth rates when exposed to low levels of CO 2 .

Discussion
Absence of the pyrenoid matrix has no effect on thylakoid ultrastructure In contrast to our initial hypothesis, a quantitative assessment revealed no difference in thylakoid membrane stacking between pyr-and the WT under any of the growth conditions (Fig. 1G). This finding is at odds with the idea that the Rubisco aggregation state could play a role in determining thylakoid stacking via a free energy trade-off (Chow, 1999;Kim et al., 2005). Rather, control of thylakoid stacking appears to lie with lateral heterogeneity of the photosystems: PSII-associated light-harvesting complex II (LHCII) subunits foster cation-modulated attraction between lamellae in flowering plants Anderson et al., 2008), whereas PSI impedes stacking through steric hindrance, which explains the hyperstacking phenotype in PSI-deficient Chlamydomonas mutants (Goodenough and Levine, 1969).
Additionally, the complex network of modified thylakoids normally associated with pyrenoids (Ohad et al., 1967;Engel et al., 2015) was retained in pyr-mutants ( Fig. 1H; Supplementary Fig. S1). This finding is in accord with previous observations that knotted tubules persist in mutants that retain only a very reduced pyrenoid (Ma et al., 2011;Mackinder et al., 2016) as well as in a mutant devoid of Rubisco due to reduced levels of chloroplast ribosomes (Goodenough and Levine, 1970). Since pyr-cells behave as WT except for a defect in the CCM, any CCM-unrelated proteins that localize to the pyrenoid in the WT, such as nitrite reductase or nucleic acid processing enzymes (Süss et al., 1995;Shukla et al., 2012;Zhan et al., 2015), must be functioning normally in the absence of a pyrenoid matrix, and may even be targeted to the same location in pyr-cells. Similarly, there was no significant difference in accumulation of pigments during acclimation to the different growth light regimes between the WT and pyr- (Fig. 5). Synthesis and assembly of photosynthetic protein complexes, which occurs in pyrenoid-associated translation zones in the WT Zerges, 2007, 2009), thus probably also functions normally in pyr-cells [unless a growth arrest (Fig. 5H) is masking a defect confined to low CO 2 ]. Consistently, proteomic work by Mitchell et al. (2017), published in this issue, found the abundances of the large majority of proteins unaffected by pyrenoid phenotype.
Despite the inability of pyr-to aggregate Rubisco into a pyrenoid matrix, pyrenoid-associated non-matrix structures thus seem to form and function as in WT cells. Rubisco aggregation is known to be controlled by two RBCS surface helices (Meyer et al., 2012) and the linker protein EPYC1 (Mackinder et al., 2016), formerly designated LCI5 (Lavigne et al., 2001). It thus seems likely that the vascular plant RBCSs expressed in pyr-lines are simply unable to interact with EPYC1, in the context of an otherwise normal chloroplast. Through an interaction between EPYC1/LCI5 and thylakoid membranes (Turkina et al., 2006), thylakoid tubules may act as an anchor point around which the pyrenoid matrix assembles (Goodenough and Levine, 1970). However, starch accumulates in pyr-at the canonical pyrenoid location without obviously interacting with the knotted thylakoid tubules (Fig. 1B, D, F). Thus thylakoid-independent mechanisms must exist that target chloroplast components to the pyrenoid location.

Pyr-cells are CO 2 limited
When light is plentiful, photosynthesis is generally limited by CO 2 supply, whereas at low light the supply of photons becomes rate limiting (Eberhard et al., 2008;Foyer et al., 2012;von Caemmerer, 2013;McGrath and Long, 2014). That air-acclimated WT and pyr-phenotypes were similar at LL, but not SL (growth, Fig. 5H; ETR, Fig. 2A, B, ϕ II , Fig. 4) was thus a first indication that pyr-cells are simply limited by CO 2 , in line with the established CCM defect (Genkov et al., 2010;Meyer et al., 2012). HelixAB, a strain that forms a pyrenoid but expresses a kinetically impaired chimeric Rubisco (Meyer et al., 2012), shows a similar difference from the WT (Fig. 3), further supporting the notion that the pyr-defect is a consequence of the slower turnover rate of the CBBC.
The capacity to absorb light, but not process all of the incoming energy via the CBBC, means that air-acclimated pyr-cells should have a greater need to dissipate excess energy. Indeed, photoprotective NPQ was found to be higher (Fig. 3C) and the functional PSII absorption cross-section lower (Fig. 5G) in air-acclimated pyr-than in the WT in SL. Such a reduction in cross-section could temporarily reduce light absorption and be under the control of cellular regulation. Increased photoinhibition in pyr-may also contribute to a lower PSII absorption cross-section and increased NPQ. Since the STT7 kinase was deleted alongside RBCS1 and 2 in the genetic host used to generate the strains studied here (Khrebtukova and Spreitzer, 1996), state transitions do not play a role.
The addition of high CO 2 facilitated a complete recovery of pyr-photosynthetic characteristics to those of WT cells, not just after long term-growth at 5% CO 2 (ETR, Fig. 2C, D; growth, Fig. 5H; PSII cross-section, Fig. 5G) but within seconds of illumination following the addition of saturating levels of bicarbonate (Fig. 4B, C). The photosynthetic electron transport chain of air-acclimated pyr-cells is thus just as competent to process incoming light as that of the WT. Therefore, pyrstrains are bona fide CCM mutants (Fukuzawa et al., 2001;Xiang et al., 2001;Jungnick et al., 2014), with photosynthetic impairments the sole consequence of limitations in the supply of CO 2 . That Rubisco must be aggregated for the CCM to function (Genkov et al., 2010;Meyer et al., 2012) implies a role for the pyrenoid matrix in limiting back-diffusion of CO 2 . Through the current work, this microcompartment is now functionally fully defined as a component of the CCM.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Further block face SEM images of intrapyrenoidlike thylakoid tubules in pyr-cells.