Was low CO2 a driving force of C4 evolution: Arabidopsis responses to long-term low CO2 stress

The responses of long-term growth of plants under elevated CO2 have been studied extensively. Comparatively, the responses of plants to subambient CO2 concentrations have not been well studied. This study aims to investigate the responses of the model C3 plant, Arabidopsis thaliana, to low CO2 at the molecular level. Results showed that low CO2 dramatically decreased biomass productivity, together with delayed flowering and increased stomatal density. Furthermore, alteration of thylakoid stacking in both bundle sheath and mesophyll cells, upregulation of PEPC and PEPC-K together with altered expression of a number of regulators known involved in photosynthesis development were observed. These responses to low CO2 are discussed with regard to the fitness of C3 plants under low CO2. This work also briefly discusses the relevance of the data to C4 photosynthesis evolution.


Introduction
The response of plants grown in lower CO 2 concentrations has been much less studied than responses to elevated CO 2 concentrations (Long et al., 2004(Long et al., , 2006Ainsworth and Long, 2005;Gerhart and Ward, 2010). Among these limited studies, some have demonstrated that a large genetic variation in response to low CO 2 exists among Arabidopsis accessions. For example, Sharma et al. (1979) screened 33 Arabidopsis accessions for survival time under limiting CO 2 when grown side by side with C 4 plants (Zea mays L.) in an air-tight chamber where CO 2 concentration was reduced to below the compensation point of C 3 plants and found a 1-2-week difference in the survival time in different accessions and also found substantial genetic segregation among F 2 parents, with extreme differences in survival time near the CO 2 compensation point.
Arabidopsis genotypes from different elevations show significant variation in the response of seed number when grown at low CO 2 (20 Pa) (Ward and Strain, 1997). Ward et al. (2000) performed an artificial selection experiment using Arabidopsis for high seed number over five generations at low CO 2 (20 Pa, or 200 ppm); the selected populations produced 25% more seeds and 35% more biomass on average than control populations which were randomly selected at the fifth generation when grown at low CO 2 . In addition, Ward and Kelly (2004) also observed a high level of genetic variation in survival, reproductive output, and total seed production among the Arabidopsis genotypes when grown at low CO 2 (200 ppm). All these studies suggest that Arabidopsis has adaptive phenotypic plasticity in response to low CO 2 .
In a carbon starvation experiment, 5-week-old Arabidopsis rosettes treated with ambient (350 ppm) CO 2 or compensation point (<50 ppm) CO 2 were collected in the light for 4 h to investigate responses to changing endogenous sugar concentrations in rosettes at the gene expression level using the GeneChip Arabidopsis ATH1 genome array (Bläsing et al., 2005). However, these studies have not addressed the mechanism of long-term responses of plants to low CO 2 .
This study conducted a survey of responses of C 3 plants to long-term low CO 2 treatments at the molecular level. Arabidopsis was chosen as the model system because its genome has been fully sequenced and is still the best annotated plant genome to date (The Arabidopsis Genome Initiative, 2000); the well-annotated Arabidopsis genome facilitates analysis of global gene expression using RNA-Seq technology. This study sequenced the transcriptome of 6-week old Arabidopsis seedlings grown under ambient CO 2 (380 ppm) or low CO 2 (100 ppm). The results are discussed with particular reference to the significance of the altered gene expression to the fitness of C 3 plants under low CO 2 . The relevance of low CO 2 to C 4 evolution is also briefly discussed.

Plant growth and harvest
Arabidopsis thaliana Columbia-0 (Col-0) seeds were imbibed in 0.1% (w/v) agar solution and incubated at 4 °C for 2 d to break dormancy. Imbibed seeds were germinated and grown in Pindstrup soil in a Percival incubator (NC-350HC-LC, Nihonika, Japan) in which CO 2 gas can be accurately and stably controlled in the range of 100-3000 ppm. CO 2 concentrations 100 and 380 ppm were applied in two separate chambers and maintained throughout this study. CO 2 concentrations were monitored and maintained throughout the experiments. Plants were grown under a 8/16 h light/dark cycle (photosynthetic photon flux density 150 μmol m -2 s -1 ) at 21 °C and 70% relative humidity. After 4 weeks, the photoperiod was changed to a 16/8 h light/dark cycle for a further 2 weeks. On day 42, samples were taken during the middle of the light period and mature expanded rosette leaves from 10-15 individual plants were harvested, immediately frozen in liquid nitrogen, and stored at -80 °C until use. The samples were taken from 12 individual pots.

Morphological data collection
Scanning electron microscopy and transmission electron microscopy were used to observe the changes of ultrastructure by low CO 2 . The number of stomata was counted in four fields of view from the fully expanded leaves of no less than eight individual plants for each treatment (Supplmentary Fig. S1 available at JXB online).

RNA preparation and sequencing
Total RNA was prepared with TRIzol (Invitrogen Life Technologies, Shanghai, China), according to the manufacturer's instructions. Following extraction, total RNA was purified using a RNeasy Mini Kit including on-column DNase digestion (Qiagen, Shanghai, China). Purified RNA was checked for integrity and quality using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The cDNA library was constructed for sequencing as described in Illumina TruSeqTM RNA sample preparation version 2 guide (catalog no. RS-930-1021). Sequencing was performed using a Illumina HiSeq 2000 (Illumina, San Diego, USA).

Mapping and quantification of sequence reads
Clean reads were mapped onto the latest A. thaliana Col-0 genome assembly (TAIR 10) or a minimal set of coding sequences of the TAIR 9 genome release (Gowik et al., 2011) using the bowtie version 0.12.7 (Langmead et al., 2009). The best hit of each read with a maximum of three nucleotide mismatches was used (-v 3 --best).
To identify differentially expressed genes, an expression profile matrix was built which integrated the digital gene expression count for each gene in each library, total gene count for each condition were used as background to check if a gene is significantly differentially expressed in low and CO 2 normal conditions by applying the chi-squares test. A FDR-corrected P-value was calculated using where i represents the ascending order of P-values, p(i) represents the ith P-value, C represents a chosen constant, and N represents the size of dataset (Benjamini and Hochberg, 1995). Significantly differentially expressed genes were picked following the criteria P<0.001, FDR<0.025, |log 2 Ratio|≥1.2.

Results
Effects of long-term low CO 2 on biomass growth, stomata density, and chloroplast ultrastructure CO 2 is the major source of carbon for photosynthesis and plays a vital role in plant growth. High CO 2 often increases the growth and reproduction of C 3 annuals, whereas low CO 2 decreases growth (Ward et al., 2000;Ward, 2005). Previous studies showed that minimum CO 2 concentrations between 180 and 200 ppm during the Last Glacial Maximum were already stressful on modern C 3 plants (Dippery et al., 1995;Ward, 2005); therefore, this work set low CO 2 concentration as 100 ppm. Arabidopsis plants grown at 100 ppm for 6 weeks were much smaller than those grown under normal CO 2 (380 ppm) (Fig. 1). In addition, low CO 2 led to a slight delay in flowering time (data not shown). The results showed that low CO 2 (100 ppm) had a dramatic impact on the growth of the C 3 plant Arabidopsis. Stomata control the entry of CO 2 into the leaves of plants for photosynthesis. There is a strong inverse correlation between atmospheric CO 2 and stomatal density (the number of stomata per unit area) . This work examined the stomatal density of abaxial (lower) leaf blade epidermis of Arabidopsis plants grown at either low CO 2 or normal CO 2 for 6 weeks (Supplmentary Fig. S1). As expected, stomatal density was significantly higher (mean±SE 509 ± 59mm -2 ) in plants grown at low CO 2 compared to plants at normal CO 2 (297 ± 54mm -2 ) (Fig. 2).
In plants, photosynthesis occurs exclusively in the chloroplast, and the photosystems (PSI and PSII) exist on the thylakoid membrane inside a chloroplast. PSII is limited to granal thylakoids, while PSI exists exclusively in the thylakoids exposed to the stroma (Albertsson, 1995;Dekker and Boekema, 2005;Sakamoto et al., 2008). The ultrastructure of mature leaves under low CO 2 were examined using transmission electron microscopy, and the size and the arrangement of bundle sheath cells and mesophyll cells was not changed, while Arabidopsis grown under low CO 2 showed decreased stacking in chloroplast grana in both mesophyll and bundle sheath cells under low CO 2 compared to normal CO 2 (Fig. 3).

Some C 4 -cycle genes were upregulated under low CO 2
The mRNA-seq analysis to compare transcriptomes between closely related C 4 and C 3 species within the genus Flaveria and Cleome using Arabidopsis as the reference genome defined a list of enzymes, transporters, and regulatory proteins required for the core C 4 cycle Gowik et al., 2011). It has been reported that Arabidopsis shows the characteristics of C 4 photosynthesis in midveins (Brown et al., 2010), but nothing is known about the plasticity of these characteristics.
In order to check whether C 4 -related characteristics can be regulated by low CO 2 stress, the transcript abundances of putative C 4 -related genes were examined. The transcript encoding the enzyme phosphoenolpyruvate carboxylase (PEPC, At2g42600) showed 2.10-fold higher transcript abundance, followed by PEPC kinase (PEPC-K, At1g08650) with a 1.99-fold increase in abundance (Table 1 and Supplementary  Table S3 available at JXB online). In addition, the transcript abundances for the genes encoding alanine aminotransferase (At1g17290), chloroplast NAD-dependent malate dehydrogenase (At3g47520), pyruvate orthophosphate dikinase regulatory protein (At4g21210), inorganic pyrophosphatase

Photorespiratory genes showed trends of upregulation under low CO 2
Low atmospheric CO 2 concentration would increase photorespiration, so this work also examined the transcript abundances of photorespiration genes. Nearly all genes showed trends of upregulation in plants grown under low CO 2 compared with those under normal CO 2 (Table 2 and Supplementary Table  S4 available at JXB online), except for the gene encoding glycine decarboxylase L-protein (mtLPD1; At1g48030); however, the fold changes were all less than 2. The differential responses of genes involved in the photosynthetic light reactions, Calvin Benson cycle, and ABA and IAA metabolisms were shown in Supplementary Tables S8-11 available at JXB online.
Chloroplast biogenesis-and maintenance-related genes showed differential expression in low CO 2 Given the differential expression of genes involved in chloroplast biogenesis and maintenance between the C 3 and C 4 Flaveria species (Gowik et al., 2011) and the altered chloroplast ultrastructure between low CO 2 and normal CO 2 (Fig. 3), this work examined the transcript abundances of genes involved in chloroplast biogenesis and maintenance under low CO 2 and compared them with previously identified genes differentially expressed between C 3 and C 4 species (Gowik et al., 2011) (Table 3 and Supplementary Table S5 available at JXB online). All the chloroplast biogenesis-and maintenancerelated genes upregulated by low CO 2 shown in Table 3 were also upregulated in C 4 Flaveria species, and five genes downregulated by low CO 2 (At5g52540, At1g52290, At5g20720, At2g32180, and At3g19820) were also downregulated in C 4 Flaveria species (Gowik et al., 2011); however, only At44446, At5g52540, At3g17040, and At1g52290 showed a ratio of expression abundance greater than 2.
Of the genes showing a fold change more than 2, three (At1g44446, At3g17040, and At5g52540) were enriched in C 4 Flaveria species compared to C 3 species. PSII concentrations are well correlated with chlorophyll b synthesis (Bailey et al., 2001), and chlorophyllide a oxygenase (At1g44446) is considered a critical enzyme responsible for chlorophyll b synthesis (Yamasato et al., 2005). HCF107 (At3g17040) is a sequence-specific RNA-binding protein and remodels local RNA structure in a manner that accounts for its ability to enhance translation (Sane et al., 2005;Hammani et al., 2012). The hcf107 mutation in Arabidopsis leads to a defective PSII (Felder et al., 2001). Although many chloroplasttargeted DnaJ proteins have not been characterized, it has been hypothesized that chloroplast-targeted DnaJ proteins participate in protein folding, unfolding, and assembly processes, and some DnaJ proteins are involved in the stabilization of thylakoid membrane complexes such as photosystem II (Chen et al., 2010). Therefore, these three downregulated genes were related to reduced PSII and this is in agreement with the ultrastructural analysis (Fig. 3).

Differentially expressed transcription factors
Ten differentially expressed transcription factors were identified (|log 2 Ratio|≥1.2) ( Table 4). Of these, GOLDEN2-LIKE2 (GLK2, At5g44190), of the GLK family which is involved in chloroplast development (Langdale, 2011), was significantly downregulated under low CO 2 . The GLK2 counterpart GLK1 (At2g20570) was also downregulated in low CO 2 but to a lesser extent.  identified 20 most upregulated genes by GLK1 and GLK2 induction using inducible gene expression combined with transcriptome analysis. The current Table 2

. Transcription abundance of photorespiration genes
The genes in bold represent these that plays a major function in photorespiration and the knockout results in a low CO 2 -sensitive phenotype (Bauwe, 2011). Reads were mapped onto the latest Arabidopsis thaliana Col-0 genome assembly (gene mapping) or a minimal set of coding sequences of the TAIR 9 genome release (core set mapping) using bowtie. Low: low CO 2 , 100 ppm; Nor: normal CO 2 , 380 ppm; rpkm, reads per kilobase per million mapped reads. work assessed the alteration of these 20 primary targets of GLK gene action and found nearly that all of them, except COR15a (At2g42540) were downregulated (Table 5 and  Supplementary Table S6 available at JXB online). COR15a was significantly induced under low CO 2 instead, possibly because COR15a is an indirect, secondary target of GLK2 .

Stress-induced mutagenesis pathway was changed under low CO 2
It has been shown that DNA double-strand break-dependent stress-induced mutagenesis is important to evolution, through producing more mutations under stress in Escherichia coli (Cirz et al., 2005;Shee et al., 2011;Al Mamun et al., 2012). As a severe stress, can low CO 2 induce more mutagenesis in natural populations? This work examined the transcriptional changes in genes encoding products related to human DNA repair proteins and found that genes involved in damage sensing (At5g40450, At2g26980, At4g04720), photoreactivation (At3g15620), homologous recombination (At3g48190), nucleotide excision repair (At2g36490, At3g02060, At5g04560, At1g52500, At3g28030, At5g45400), and DNA polymerases (At4g32700, At1g67500) were upregulated by low CO 2 (Table 6 and Supplementary Table S7 available at JXB online). These results suggest that low CO 2 might induce a similar mechanism of DNA double-strand break-dependent stress-induced mutagenesis to promote evolution.

Discussion
This study, as far as is known for the first time, investigated responses to low CO 2 at the transcriptome level in model plant Arabidopsis. Here, the observed changes of transcriptomics under low CO 2 are briefly discussed, with particular reference to their potential significance for fitness of C 3 plants under low CO 2 and potential linkage to C 4 photosynthesis evolution.

Low CO 2 reduced productivity
Arabidopsis plants grown under low CO 2 had extremely small stature compared with plants grown under normal CO 2 (Fig. 1). This result is in accordance with previous studies on the effect of low CO 2 on plant growth (Ward, 2005). Arabidopsis grown under low CO 2 has about a 7-day delay in flowering time. This has also been observed earlier (Ward and Strain, 1997) and could be interpreted as ORA47 is a regulator of jasmonate biosynthesis (Pauwels and Goossens, 2008) C2C2-GATA At4g26150 CGA1 was regulated by light, nitrogen, cytokinin, and gibberellic acid, and modulated nitrogen assimilation, chloroplast development, and starch production (Bi et al., 2005;Naito et al., 2007;Mara and Irish, 2008;Richter et al., 2010;Hudson et al., 2011); CGA1 play a key role in chloroplast development, growth, and divison in Arabidopsis (Chiang et al.,  SPL3 is involved in regulation of flowering and vegetative phase change (Cardon et al., 1997;Wu and Poethig, 2006;Yamaguchi et al., 2009) C2C2-CO-like At4g27310 BBX28 Zinc finger (B-box type) family protein G2-like At5g44190 GLK2 (Golden2-like 2) GLK2 is required for normal chloroplast development (Fitter et al., 2002); GLK2 together with GLK1 optimize photosynthetic capacity by integrating responses to variable enironmental and endogenous cues AGL42 is involved in the floral transition and RNAi-directed downregulation of AGL24 results in late flowering (Yu et al., 2002) a mechanism to allow for greater accumulation of stored reserves that could be allocated to reproduction, resulting in increased fitness under low CO 2 (Sage and Coleman, 2001;Ward, 2005). Many studies have shown that atmospheric CO 2 concentration negatively regulates stomatal density (Woodward, 1987;Beerling et al., 2001;Franks and Beerling, 2009;Doheny-Adams et al., 2012;Franks et al., 2012). Paleontological research has suggested that the long-term decreases in atmospheric throughout the entire evolutionary history of vascular plants led to the evolution of high densities of small stomata in order to attain the highest g cmax values required to counter CO 2 'starvation' (Franks and Beerling, 2009;Franks et al., 2012). Stomata also exhibit short-term adaptive responses to atmospheric CO 2 over much shorter timescales. For example, A. thaliana Col-0 grown at high CO 2 (720 ppm) had reduced stomata density compared with those grown at ambient CO 2 (360 ppm) (Lake et al., 2001). In the current work, plants grown under low CO 2 developed leaves with higher stomatal density (over 60% increase compared to normal CO 2 ; Fig. 2), suggesting that the plants developed a greater g cmax to counteract the CO 2 limitation of photosynthesis. These results suggest that low CO 2 is a severe stress to C 3 plants and may greatly reduce C 3 plant productivity.

Responses of genes involved in C 4 photosynthesis and photorespiration under low CO 2
In C 4 plants, CO 2 is initially fixed by the enzyme PEPC into a C 4 acid and then transported to the site of Rubisco (Hatch, 1987). The only photosynthetic gene expression patterns common to all independently evolved C 4 lineages are upregulation of PEPC and downregulation of Rubisco in mesophyll cells (Sinha and Kellogg, 1996;Langdale, 2011). Arabidopsis has four genes encoding PEPC, and AtPPC2 (At2g42600) is the only isoform expressed in leaves. Unlike the other three PEPCs, the expression of AtPPC2 is stable and has not been reported to be regulated by any stress (Sánchez et al., 2006;Doubnerová and Ryšlavá, 2011); however, the current work found that AtPPC2 was upregulated by low CO 2 ( Table 1). The regulators of photosynthetic genes are also crucial to maintain C 4 photosynthesis: e.g. plant PEPC activity is further regulated through reversible phosphorylation by PEPC-K (Nimmo, 2003). Transcripts encoding the C 4specific regulatory factors PEPC-K and pyruvate orthophosphate dikinase regulatory protein were upregulated as well (Table 1). However, changes in other C 4 -related genes were less, with fold changes of less than 2.
When grown in low CO 2 , plants would experience relatively high levels of flux through the photorespiratory pathway Table 5. Transcript abundance of GLK-regulated genes Reads were mapped onto the latest Arabidopsis thaliana Col-0 genome assembly (gene mapping) or a minimal set of coding sequences of the TAIR 9 genome release (core set mapping) using bowtie. The most upregulated genes by GLK1 and GLK2 induction identified by  were examined and nearly all of them were downregulated by low CO 2 , except COR15a (At2g42540). Low: low CO 2 , 100 ppm; Nor: normal CO 2 , 380 ppm; rpkm, reads per kilobase per million mapped reads. CAO, chlorophyllide a oxygenase; CHLH, magnesium chelatase; COR15a, COLD-REGULATED 15A; GCN5 related, ornithine N-delta-acetyltransferase; GLK1, Golden2-like 1; GLK2, Golden2-like 2; Lhcb, light harvesting complex subunit; MRU1, mto responding up 1; PORB, NADPH:protochlorophyllide oxidoreductase B. because of the competitive reactions of Rubisco oxygenation. In this study, a trend of upregulation of the photorespiratory genes was observed in plants grown under low CO 2 (Table 2), although most of the genes showed a fold change of less than 2. The recent study of transcriptome analysis using C 3 , C 3 -C 4 intermediate, and C 4 species of Flaveria found that transcript abundances for most genes related to photorespiration in the C 3 -C 4 intermediate species Flaveria ramosissima were even higher than in the C 3 species Flaveria robusta (Gowik et al., 2011), which is indicative of the importance of the photorespiratory pathway during the evolution of C 4 photosynthesis. The different subunits of glycine decarboxylase showed altered expression, although the fold changes of these subunits were about 0.9-1.3. Overall, the data from this study suggest that expression of PEPC and PEPC-K is increased under low CO 2 , which most likely reflects their potential role for refixation of photorespired CO 2 under low CO 2 (Sage et al., 2012). For most of the other C 4 genes, although trends of upregulation were observed, the fold changes were less than 2. Although by using expression level changes of all genes under two conditions as background, this work obtained P-values much less than 0.01 for many C 4 -related genes, it is likely that lack of biological replicates could had potentially led to an overestimation of the reliability of statistical tests and caused problems in identifying significantly changed genes, especially when their fold changes were less than 2. Based on these, this work cannot state that low CO 2 induced upregulation of C 4 genes, except for those genes which showed fold changes over 2 (e.g. PEPC).

Readjustment of balance between light absorption and CO 2 fixation under low CO 2
These data on chloroplast ultrastructure and transcript abundance of genes involved in chloroplast biogenesis and maintenance are consistent with the model for long-term photosynthetic regulation by GLK proteins ). When light is high and atmospheric CO 2 is limiting, the rate of CO 2 fixation is insufficient to use all of the output of the light-harvesting reactions, resulting in an overly reduced photosynthetic electron transport. This triggers a decrease of GLK transcription (GLK1 and GLK2; Table 5). Since GLK transcription factors directly regulate a large suite of genes involved in light-harvesting and thylakoid protein complexes, especially those of PSII , the light-harvesting components in the thylakoid membrane LHCB2.2 (At2g05070), LHCB4.2 (At3g08940), Lhcb3 (At5g54270), Lhcb2.4 (At3g27690), and Lhcb1.4 (At2g34430) were downregulated under low CO 2 . In addition, the downregulation of the chlorophyllide a oxygenase gene led to the decrease of chlorophyll b synthesis. These results were consistent with the fewer and less-stacked grana observed and a higher proportion of nonstacked stromal lamellae, as observed in glk1 glk2 mutants (Fig. 3). Therefore, these observed expression changes in GLK and the genes regulated by GLK can be interpreted as reflecting the altered balance between CO 2 fixation and light absorption.

Evolutionary implications of plants of to low CO 2
Growing evidence suggests that all of the basic elements of C 4 photosynthesis already existed in C 3 plants. For example, all of the enzymes involved in C 4 photosynthesis exist in C 3 plants and play different roles in C 3 plant metabolism (Aubry et al., 2011). Some elements controlling the cell specific expression of C 4 -related enzymes have been found in C 3 plants . Moreover, typical C 3 plants (e.g. tobacco and Arabidopsis) show the characteristics of C 4 photosynthesis in midveins (Hibberd and Quick, 2002;Brown et al., 2010).
Can some features related to C 4 photosynthesis be enhanced under some conditions in a C 3 plant? This work showed that under low atmospheric CO 2 , A. thaliana Col-0 adjusted a series of biological processes, especially the upregulation of PEPC and PEPC-K gene expression, and also the altered expression of some transcription factors related to photosynthesis development, and the downregulation of light-harvesting and thylakoid protein complexes. Although this study also observed that the majority of the other C 4cycle genes were upregulated under low CO 2 in Arabidopsis, their fold changes were less than 2 and therefore no firm statements regarding their changes can be made.
Experiments with more biological replicates and Arabidopsis accessions are still needed to firmly conclude whether low CO 2 can induce upregulation of other C 4 -related genes. Therefore, the results from this paper do not support a scenario where low CO 2 acts as a signal to induce C 4 biochemical features in C 3 plants. It is most likely that the upregulation of PEPC and PEPC-K might be a mechanism that C 3 plants used to refix photorespired and respired CO 2 and also to recapture the released ammonium from photorespiration and hence increase the competitive advantages under low CO 2 conditions.

Supplementary material
Supplementary data are available at JXB online. Supplementary Fig. S1