Single-Organelle Quantification Reveals Stoichiometric and Structural Variability of Carboxysomes Dependent on the Environment

The carboxysome is a complex, proteinaceous organelle that plays essential roles in carbon assimilation in cyanobacteria and chemoautotrophs. It comprises hundreds of protein homologs that self-assemble in space to form an icosahedral structure. Despite its significance in enhancing CO2 fixation and potentials in bioengineering applications, the formation of carboxysomes and their structural composition, stoichiometry and adaptation to cope with environmental changes remain unclear. Here we use live-cell single-molecule fluorescence microscopy, coupled with confocal and electron microscopy, to decipher the absolute protein stoichiometry and organizational variability of single β-carboxysomes in the model cyanobacterium Synechococcus elongatus PCC7942. We determine the physiological abundance of individual building blocks within the icosahedral carboxysome. We further find that the protein stoichiometry, diameter, localization and mobility patterns of carboxysomes in cells depend sensitively on the microenvironmental levels of CO2 and light intensity during cell growth, revealing cellular strategies of dynamic regulation. These findings, also applicable to other bacterial microcompartments and macromolecular self-assembling systems, advance our knowledge of the principles that mediate carboxysome formation and structural modulation. It will empower rational design and construction of entire functional metabolic factories in heterologous organisms, for example crop plants, to boost photosynthesis and agricultural productivity. One Sentence Summary Determination of absolute protein stoichiometry reveals the organizational variability of carboxysomes in response to microenvironmental changes The authors responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Author (www.plantcell.org) is: Lu-Ning Liu (luning.liu@liverpool.ac.uk).


INTRODUCTION
Organelle formation and compartmentalization within eukaryotic and prokaryotic cells provide the structural foundation for segmentation and modulation of metabolic reactions in space and time. Bacterial microcompartments (BMCs) are self-assembling organelles widespread among bacterial phyla (Axen et al.,β014). By physically sequestering specific enzymes key for metabolic processes from the cytosol, these organelles play important roles in CO β fixation, pathogenesis, and microbial ecology (Yeates et al.,β010;Bobik et al.,β015). According to their physiological roles, three types of BMCs have been characterized: the carboxysomes for CO β fixation, the PDU microcompartments for 1,β−propanediol utilization, and the EUT microcompartments for ethanolamine utilization.
The common features of various BMCs are that they are ensembles composed of purely protein constituents and comprise an icosahedral single-layer shell that encases the catalytic enzyme core. This proteinaceous shell, structurally resembling virus capsids, is selfassembled from several thousand polypeptides of multiple protein paralogs that form hexagons, pentagons and trimers (Kerfeld and Erbilgin,β015;Sutter et al.,β016;Faulkner et al.,β017). The highly-ordered shell architecture functions as a physical barrier that concentrates and protects enzymes, as well as selectively gating the passage of substrates and products of enzymatic reactions (Yeates et al.,β010;Bobik et al.,β015).
Carboxysomes serve as the key CO β -fixing machinery in all cyanobacteria and some chemoautotrophs. The primary carboxylating enzymes, ribulose-1,5-bisphosphate carboxylase oxygenase (Rubisco) (Rae et al.,β01γ), are encapsulated by the carboxysome shell that facilitates the diffusion of HCO γ and probably reduces CO β leakage into the cytosol Rubisco (comprising the large and small subunits RbcL and RbcS) and -carbonic anhydrase ( -CA, encoded by the ccaA gene). The colocalized -CA dehydrates HCO γ to CO β and creates a CO β -rich environment in the carboxysome lumen to favor the carboxylation of Rubisco. In addition, CcmM and CcmN function as "linker" proteins to promote Rubisco packing and shell-interior association (Kinney et al.,β01β). CcmM in the -carboxysome appears as two isoforms, a γ5-kDa truncated CcmMγ5 and a full-length 58-kDa CcmM58 (Long et al.,β007;Long et al.,β010;Long et al.,β011). CcmMγ5 contains three Rubisco Understanding the physiological composition and assembly principles of carboxysome building blocks is of key importance not solely to unravel the underlying molecular mechanisms of carboxysome formation and biological functions, but also for heterologously engineering and modulating functional CO β -fixing organelles to supercharge photosynthetic carbon fixation in synthetic biology applications. Previous estimations of the carboxysome protein stoichiometry from either the whole cell lysates or the isolated forms using immunoblot and mass spectrometry illustrated the relative abundance of carboxysome proteins (Long et al.,β005;Long et al.,β011;Rae et al.,β01β;Faulkner et al.,β017). Moreover, it was revealed that carboxysome biosynthesis in Syn794β is highly dependent upon environmental conditions during cell growth, such as light intensity (Sun et al.,β016) and CO β availability (McKay et al.,199γ;Harano et al.,β00γ;Woodger et al.,β00γ;Whitehead et al.,β014). The exact stoichiometry of all building components in the functional carboxysome and how carboxysomes manipulate their compositions, organizations and functions to cope with environmental changes have remained elusive.
Here, we construct a series of Syn794β mutants with individual components of carboxysomes functionally tagged with the bright and fast-maturing enhanced yellow fluorescent protein (YFP) and report the in vivo characterization of protein stoichiometry of carboxysomes at the single-organelle level, using real-time single-molecule fluorescence microscopy, confocal and 4 electron microscopy combined with a suite of biochemical and genetic assays. Quantification of the protein stoichiometry of -carboxysomes in Syn794β grown under different conditions demonstrates the organizational flexibility of -carboxysomes, and their ability to modulate functions towards local alterations of CO β levels and light intensity during cell growth, as well as the regulation of the spatial localization and mobility of -carboxysomes in the cell. This study provides fundamental insight into the formation and structural plasticity of carboxysomes and their dynamic organization towards environmental changes, which could be extended to other BMCs and macromolecular systems. A deeper understanding of the protein composition and structure of carboxysomes will inform strategies for rational design and engineering of functional and adjustable metabolic modules towards biotechnological applications.

Protein stoichiometry of functional carboxysomes at the single-organelle level
We constructed ten Syn794β strains expressing individual -carboxysome proteins (CcmKγ,CcmK4,CcmKβ,CcmL,CcmM,CcmN,RbcL,RbcS,CcaA,RbcX) fused with YFP at their Ctermini individually (Supplemental Figure 1) Figure 1D), indicative of the structural heterogeneity of -carboxysomes. The modal average stoichiometry of each protein subunit per carboxysome was defined by the measured peak from each distribution of the raw stoichiometric data ( Figure 1D, Supplemental Figure γ), after subtracting the background fluorescence distribution, primarily from chlorophylls, which was determined from the WT cells (Supplemental Figure 4).
In the -carboxysome synthesized in cells grown under Air/ML, Rubisco enzymes are the predominant components, as indicated by the RbcS content (Table 1). CcmM is the second most abundant element; there are over 700 copies of CcmM molecules per -carboxysome.
In addition, the CcmK4 content is greater than that of CcmKγ by a factor of γ. Our study represents a direct characterization of protein stoichiometry at the level of single functional carboxysomes in their native cellular environment.
As a control, we fused RbcL with mYPet, a monomeric-optimized variant of YFP. The RbcL-YFP and RbcL-mYPet cells show no significant difference in the subcellular distribution of carboxysomes as well as cell doubling times and carbon fixation (Supplemental Figure 5), demonstrating that there are no measurable artefacts due to putative effects of dimerization of the YFP tag.
We also examined the relative abundance of individual carboxysome proteins in the YFP-

Stoichiometry of carboxysome proteins exhibit a dependence on the microenvironment conditions of live cells
Our previous study showed that the content and spatial positioning of -carboxysomes in Syn794β are dependent upon light intensity during cell growth, revealing the physiological regulation of carboxysome biosynthesis (Sun et al., β016). Whether the stoichiometry of different components in the carboxysome structure changes in response to fluctuations in environmental conditions is unknown. Here we addressed this question by taking advantage of the far greater throughput of confocal microscopy compared to Slimfield, whilst still using the single-molecule precise Slimfield data as a calibration to convert the intensity of detected foci from confocal images into estimates for absolute numbers of stoichiometry. We achieved this by identifying the peak value of the foci intensity distribution from each given cell strain obtained from confocal imaging with the peak value of the measured Slimfield foci stoichiometry distribution for the equivalent cell strain under Air/ML. This approach allows us to generate a conversion factor which we then applied to subsequent confocal data acquired under lower light (LL), higher light (HL) and ML with the air supplemented by γ% CO β , and to estimate relative changes in the stoichiometry of carboxysome building components using large numbers of cells, without the need to obtain separate Slimfield datasets for each condition ( Table 1.
Interestingly, we find that the variation of CcmL abundance per carboxysome rises with increasing light illumination and CO β availability (

Variation of carboxysome diameter represents a strategy for manipulating carboxysome activity to adapt to environmental conditions
The change in the protein content per carboxysome signifies the variation of -carboxysome size and organization among different cell growth conditions. Indeed, electron microscopy (EM) of Syn794β WT cells substantiates the variable structures of -carboxysomes in response to the changing environment ( Figures 4A and 4B). The average diameter ofcarboxysomes is 19β ± 41 nm (n = γγ) in Air/ML, 144 ± β4 nm (n = β5) in γ% CO β , 151 ± ββ nm (n = β7) in LL, and β08 ± β8 nm (n = 51) in HL ( Figure 4B, Supplemental Table 5, Supplemental Figure 9). These results reveal that both the CO β level and light intensity can 1γ result in alternations of carboxysome size ( Figure 4B). Larger -carboxysomes can accommodate more Rubisco enzymes (estimated on the basis of RbcS content) ( Figure 4C).
An exception is the carboxysomes under LL, which are around 5% larger than the carboxysomes under γ% CO β but comprises only 67% of Rubisco per carboxysome under CO β ( Figure 4C, Supplemental Table 5). EM images reveal that the lumen of -carboxysomes synthesized under LL often contain regions with low protein density ( Figure 4A, arrows; Supplemental Figure 9), 59% for LL (16 out of β7 carboxysomes) compared with 9% for Air/ML (γ out of γγ), 1β% for CO β /ML (γ out of β5) and 8% for HL (4 out of 51), which likely accounts for the reduced and uneven Rubisco loading within the -carboxysome.
We also find that CO β -fixing activity per carboxysome increases as the -carboxysome structure enlarges, which is correlated to strong light intensity during cell growth ( Figure 4D), demonstrating the correlation between -carboxysome structure and function in vivo.
Moreover, under HL the CO β -fixation activity per Rubisco of the -carboxysome declines as the carboxysome size and Rubisco density in the carboxysome lumen increase ( Figure 4E, Supplemental Table 5). This may suggest that Rubisco density and local Rubisco packing are important for determining CO β -fixation activity of individual Rubisco (Supplemental Table 5).

Patterns of spatial localization and diffusion of -carboxysomes in live cells change dynamically depending upon light intensity during growth
The patterns of -carboxysome localization within the cyanobacterial cells appears to be Exceptionally, fluorescence tagging on CcmP and CcmO does not show normal carboxysome assembly and localization compared to other YFP-tagged strains (Supplemental Figure 11).
In this work, therefore, we did not include estimation of the protein abundance of CcmP and CcmO, as well as RbcL and CcmKβ that cannot be fully tagged with YFP. capping CcmL will create large space within the shell, as a possible mechanism of modulating shell permeability, or will be compensated for by incorporation of other shell proteins, for example the additional CcmP trimers that are speculated to be responsible for permeability, remains to be further investigated. Our results also suggest that carboxysomes could possess a flexible molecular architecture, resonating with the observation of structural "breathing" of virus capsids which has been reported to be key to cope with temperature change (Roivainen et al., 199γ;Li et al., 1994 While the abundance of most of the structural components varies, the ratio of CcmK4 and CcmKγ is relatively unaffected (ranging from γ.6 to 4.1, Supplemental Table 5) under the tested growth conditions, implying their spatial colocalization within the carboxysome shell ββ ( Figure 6A). This is reminiscent of the recent observation that CcmKγ Cultures were constantly diluted with fresh medium to maintain exponential growth phase for the following imaging and biochemical analysis. Escherichia coli strains used in this work, DH5a and BWβ511γ, were grown aerobically at γ0 or γ7°C in Luria-Broth medium. Medium supplements were used, where appropriate, at the following final concentrations: ampicillin 100 mg•mL -1 , chloramphenicol 10 mg•mL -1 , apramycin 50 mg•mL -1 , and arabinose 100 mM.
All YFP-fusion mutants were generated following the REDIRECT protocol (Supplemental  Table   6). Primers used in this work were listed in Supplemental Table 7. The same strategy was also applied for the mYPet mutant. For these mutant strains, BG-11 medium was supplemented with apramycin at 50 g•mL -1 .

Cell doubling time and growth curve measurement
Cultures were inoculated at OD 750 of 0.05-0.1 with fresh BG-11. Growth of cells was monitored at OD 750 using a spectrophotometer (Jenway 6γ00 spectrophotometer, Jenway, UK) every β4 hours. Doubling times were calculated using exponential phase of growth from day 1 to day 4. Four biological replicates from different culture flasks were recorded. Data are presented as mean ± standard deviation (SD). For each experiment, at least three biological replicates from different culture flasks were analyzed.

Slimfield microscopy and data analysis
Live cells were applied at the small volume onto the BG-11 agarose pad at 0.β5 mm thickness to maintain physiological growth, air dried to remove excessive medium and then For high-copy-number strains, intensity of carboxysomes was very high compared to the chlorophyll but for CcmL (typically ~βx, compare Supplemental Figure γ with Supplemental Figure 4A) the fluorescence intensity per carboxysome was comparable (although generally brighter) to small regions of bright chlorophyll, detected as foci by our software, as confirmed by looking at the parental strain with no YFP present. To correct for this chlorophyll content, we tracked parental WT Syn794β cells as YFP-labelled cells to calculate the apparent chlorophyll stoichiometry distribution (Supplemental Figure 4A). The CcmL distribution was then corrected by subtracting the apparent chlorophyll distribution. To investigate putative periodic features in the stoichiometry distribution, we used the raw uncorrected values to minimize dephasing artefacts ( Figure 4C) using a kernel width of 0.5 molecules (equivalent to the error in determining the characteristic intensity). The peak values in other strains were far from the chlorophyll peak and so unaffected by this correction.

Confocal microscopy imaging and data analysis
Preparation of Syn794β cells for confocal microscopy was performed as described earlier ( Confocal microscopic images were processed using FIJI Trackmate plugins (Tinevez et al., β017) to retrieve peak intensities of carboxysomes based on the Find Maxima detection algorithm. Noise tolerance was determined by background intensities in empty regions.
Imaging for different treatments in the same strain was performed under the same imaging β6 settings. For strains with visible cytosolic signals, the cytosolic background intensity was determined by the average peak intensities in non-carboxysome regions over the central line of the cell and was subtracted to obtain peak intensities. Raw data were processed by Origin Lab and MATLAB (Mathworks) for profile extraction and statistical analysis and the goodnessof-fit parameter for Violin plot visualization. Violin plots were generated by R to illustrate the fluorescence intensity distribution of individual building proteins per carboxysome fitted by kernel smooth fitting. The representative values and deviations of signal intensities were represented by Peak value ± half width at half maximum (HWHM) measured from kernel density fitted profiles, respectively. The significance of differences between treatments was evaluated by Mann-Whitney U-tests pair-wisely (Supplemental Table 4). Standard errors of sampling were determined through randomized grouping of intensity entries, with each group containing a minimum of 70-100 entries. Errors were controlled below 5% to have accurate estimation from the distributions. The relative protein abundance of carboxysomes was estimated by confocal imaging under Air/ML, CO β /ML, LL, and HL was normalized by the definite copy number of each strain under Air/ML determined by Slimfield imaging.

Live-cell time-lapse confocal imaging and data analysis
A β mm-thick BG-11 agar mat was prepared in stacked sandwiches to accommodate drops of diluted Syn794β cells. Cells were incubated on the BG-11 agar mat on the microscope for 1-β hours before imaging. The continuous light illumination was provided at the intensity relatively equal to HL, ML, or LL that were used for cell growth, in order to maintain cell physiology. The same illumination was applied to the cells during time-lapse imaging with a hand-made module that switched off the light during laser scanning (less than 5 s per minute intervals).
The interval time was set to 60 s to guarantee sufficient light illumination between imaging.
The laser power was set to the minimum (1%) to reduce the bleaching for signals during longterm tracking. Images were initially corrected for horizontal drifting by Descriptor-based series registration (βd/γd+T) plugin, and then were processed by the Trackmate plugin in FIJI for particle tracking. Retrieved track data was analyzed using bespoke MATLAB (Mathworks) scripts for MSD. Diffusion coefficient calculations and data visualization were modified as previously described (Ewers et al.,β005;Sbalzarini and Koumoutsakos,β005). Diffusion coefficients were calculated by fitting the first 6 points of the MSD vs. curves. As the MSD vs. curves indicated potentially non-Brownian diffusion at higher values, we described the diffusion coefficients as "apparent diffusion coefficients". Tracking and diffusion coefficient β7 determination were tested by computational simulations (Supplemental Movie β). Bespoke Matlab code was written to generate simulated image stacks of carboxysomes diffusing inside cells. Images were simulated by integrating a model γD point spread function over a γD model for the cell structure (Wollman and Leake β015). This model comprises an inner cytosol surrounded by thylakoid membranes (indicated by chlorophyll fluorescence) and γ carboxysomes with a diameter of β00 nm. Each component's intensity was adjusted to match real images before representative Poisson noise was applied. Carboxysomes were simulated undergoing Brownian motion with a diffusion coefficient of 1.γ x 10 -5 µm β s -1 over 40 image frames. Trackmate tracking and diffusion coefficient calculation yielded a mean diffusion coefficient of 1.γβ ± 0.0β x 10 -5 µm β s -1 , giving a 1.5% error.

Immunoblot analysis
Immunoblot examination was carried out following the procedure described previously (Sun et al.,β016). 150 µg of cell lysate, measured by Pierce Coomassie (Bradford) Protein Assay Kit (Thermo Fisher Scientific), was loaded on 10% (v/v) denaturing SDS-PAGE gels. Immunoblot analysis was performed using the primary mouse monoclonal anti-GFP (Invitrogen, γγ-β600), capable of recognizing series of GFP variants including YFP, the rabbit polyclonal anti-RbcL (Agrisera, AS0γ 0γ7), the horseradish peroxidase-conjugated goat anti-mouse IgG secondary antibody (Promega, W40β1) and a Goat anti-Rabbit IgG (H&L), HRP conjugated (Agrisera AS10 1461). Anti-CcmKβ antibody was kindly provided by the Kerfeld lab (Michigan State University, US) (Cai et al.,β016). Protein quantification from immunoblot data was carried out using FIJI. Our nominal assumption that the ratios of YFP-tagged to total RbcL or CcmKβ in carboxysomes are similar to those in cell lysates.

In vivo carbon fixation assay
In vivo carbon fixation assay was carried out to determine carbon fixation of Syn794β WT and mutant cells, as described in the previous work (Sun et al.,β016). For each WT and mutant, at least three biological replicates from different culture flasks were assayed. Significance was assessed by two-tailed Student's t-tests.

Electron microscopy and carboxysome size measurement
β8 Electron microscopy was carried out as described previously (Liu et al.,β008;Sun et al.,β016). Carboxysome diameter was measured as described previously (Faulkner et al.,β017) and was analyzed using Origin.

Accession Numbers
Accession numbers of genes in this article are provided in Supplemental Table 6.  Supplemental Table 4. Evaluation and quality control of quantitative microscopy.  γ0

Competing interests
The authors declare no conflict of interest. (D) Distribution of YFP copy number detected for individual carboxysomes in corresponding mutants, rendered as kernel density estimates using standard kernel width. Heterogeneity of contents was observed, and a "preferable" copy number, represented by kernel density peak values could be determined. Statistics of copy numbers (Peak value ± HWHM) are listed in Table 1 for ML conditions. The corresponding Slimfield images and histogram for complete strain sets are shown in Supplemental Figure γ.     intensity reduces: β.76 ± β.8γ x 10 -5 µm β s -1 for HL (mean ± SD, n = 105), 1.48 ± 1.0γ x 10 -5 µm β s -1 for ML (n = 84), and 0.β8 ± 0.19 x 10 -5 µm β s -1 for LL (n = γγ6). p = γ.05 x 10 -5 between HL and ML; p = β.77 x 10 -5 between ML and LL, two-tailed Student's t-test).  Table 1). *Rubisco content was estimated from RbcS stoichiometry based on the RbcL 8 S 8 Rubisco structure. The majority of shell facets shown in light blue is tiled by the major shell γβ protein CcmKβ. The total abundance of CcmM58 and CcmMγ5 was estimated. The components RbcL, CcmKβ, CcmO and CcmP were not directly determined in this work and thus are not shown in this model.

(B)
The carboxysome diameter is variable in response to changes in the CO β level and light intensity.