Insights from transcriptome profiling on the non-photosynthetic and stomatal signaling response of maize carbonic anhydrase mutants to low CO₂

Abstract

Background

Carbonic anhydrase (CA) catalyzes the hydration of CO₂ in the first biochemical step of C₄ photosynthesis, and has been considered a potentially rate-limiting step when CO₂ availability within a leaf is low. Previous work in Zea mays (maize) with a double knockout of the two highest-expressed β-CA genes, CA1 and CA2, reduced total leaf CA activity to less than 3% of wild-type. Surprisingly, this did not limit photosynthesis in maize at ambient or higher CO₂concentrations. However, the ca1ca2 mutants exhibited reduced rates of photosynthesis at sub-ambient CO₂, and accumulated less biomass when grown under sub-ambient CO₂ (9.2 Pa). To further clarify the importance of CA for C₄ photosynthesis, we assessed gene expression changes in wild-type, ca1 and ca1ca2 mutants in response to changes in pCO₂ from 920 to 9.2 Pa.

Results

Leaf samples from each genotype were collected for RNA-seq analysis at high CO₂ and at two time points after the low CO₂ transition, in order to identify early and longer-term responses to CO₂ deprivation. Despite the existence of multiple isoforms of CA, no other CA genes were upregulated in CA mutants. Although photosynthetic genes were downregulated in response to low CO₂, differential expression was not observed between genotypes. However, multiple indicators of carbon starvation were present in the mutants, including amino acid synthesis, carbohydrate metabolism, and sugar signaling. In particular, multiple genes previously implicated in low carbon stress such as asparagine synthetase, amino acid transporters, trehalose-6-phosphate synthase, as well as many transcription factors, were strongly upregulated. Furthermore, genes in the CO₂ stomatal signaling pathway were differentially expressed in the CA mutants under low CO₂.

Conclusions

Using a transcriptomic approach, we showed that carbonic anhydrase mutants do not compensate for the lack of CA activity by upregulating other CA or photosynthetic genes, but rather experienced extreme carbon stress when grown under low CO₂. Our results also support a role for CA in the CO₂ stomatal signaling pathway. This study provides insight into the importance of CA for C₄ photosynthesis and its role in stomatal signaling.

Electronic supplementary material

The online version of this article (10.1186/s12864-019-5522-7) contains supplementary material, which is available to authorized users.

Background

Carbonic anhydrase (CA) is a ubiquitous enzyme across all kingdoms of life, catalyzing the reversible hydration of CO₂ into bicarbonate [1]. In plants, the β-subtype of CA is often present in large quantities and has long been implicated in photosynthesis [2]. The role of β-CA in C₃ plants remains somewhat ambiguous, with reports that CA makes up a significant portion of leaf protein, but β-CA knockdowns show little photosynthetic phenotype [3, 4]. In C₃ plants, β-CA is primarily found in the chloroplast and the prevailing hypothesis is that it facilitates diffusion of CO₂ across the chloroplast envelope [1]. In contrast, β-CA is primarily localized to the mesophyll cytosol in C₄ plants, where it produces bicarbonate for phosphoenolpyruvate carboxylase (PEPC) in the first committed step of the CO₂ concentrating mechanism.

Various modeling studies and mutant analyses have been employed to assess the role of β-CA for C₄ photosynthesis. For example, modeling the apparent in vivo CA activity extrapolated from in vitro enzyme assays suggested that CA may be near-limiting for C₄ photosynthesis [5]. Subsequent mutant analyses in the C₄ dicot Flaveria bidentis found that CA was in excess for photosynthesis, as CO₂ assimilation was unaffected until CA levels reached < 20% of wild-type [6, 7]. However, C₄ monocots such as maize had previously been shown to have naturally low levels of CA activity [8], suggesting that CA may be limiting in these plants. To address this question, Studer et al. [9] knocked out the two highest-expressed β-CA genes in maize (Ca1 and Ca2). Surprisingly, a 97% reduction of wild-type CA activity resulted in plants capable of maintaining wild-type rates of photosynthesis at ambient and above partial pressures of CO₂ (pCO₂), with significant reductions in photosynthesis occurring only under sub-ambient pCO₂. RNAi knockdowns of CA to ~ 10% of wild-type in Setaria viridis showed similar results [10]. Taken together, it appears that CA activity is in excess for what is needed to drive photosynthesis in C₄ plants under ambient pCO₂. However, these results do not provide a satisfactory explanation for the overabundance of CA in C₄ leaves, and also do not indicate if CA mutants are able to compensate through alternate means of enhancing CO₂ diffusion and bicarbonate availability.

In addition to photosynthesis, CA has been shown to serve diverse roles in plant metabolism. Isoforms of α and γ CA subtypes identified in plants are primarily expressed in non-photosynthetic tissue. The function of α-CA in higher plants is not known [11]. Similarly, the role of γ-CA is not well understood, although mutant analyses in Arabidopsis thaliana have implicated γ-CAs in the assembly of Complex I in the mitochondrial electron transport chain and suggested roles for γ-CAs in light-dependent development and photorespiration [12–14]. Although β-CAs are primarily considered for their photosynthetic role, isoforms of β-CAs in C₃ plants have also been implicated in lipid biosynthesis [15], stomatal responses to CO₂ [16, 17], recapture of photorespired CO₂ [14, 18], and amino acid biosynthesis [19]. Furthermore, recent work in C₄ plants has suggested a link between β-CA, nitrogen metabolites, and photosynthesis [20], and a role for β-CA in stomatal movement [21]. Therefore, β-CAs may serve in a variety of roles in C₄ plants.

To further evaluate the role of β-CA in the C₄ plant maize, RNA-seq was performed on ca1 and ca1ca2 mutants to test the transcript-level effect of CA limitation under high and low pCO₂. We hypothesized that CA mutants would upregulate other CA isoforms, photosynthetic genes, or genes influencing CO₂ diffusion into the leaf in order to compensate for the CA limitation. This experiment also provided an opportunity to evaluate the potential role of β-CA in non-photosynthetic roles in a C₄ plant. Overall, our results provide new insights into the role of β-CA in maize and mechanisms of low CO₂ adaptation in CA-limited plants.

Results

RNA-seq was performed on wild-type and ca1 and ca1ca2 mutant alleles (described previously in [9]) in order to evaluate the effect of CA limitation on transcriptional responses to low CO₂. Plants were initially grown at 920 Pa (10,000 ppm) CO₂ as previously described [9] to minimize growth differences between genotypes. To verify that genotypes were similar under the high CO₂ growth conditions, leaf samples were taken for RNA-seq after 12 days at 920 Pa. After the CO₂ levels were reduced to 9.2 Pa (100 ppm), two additional time points were sampled to capture rapid and longer-term response differences between genotypes to reduced CO₂ availability.

Principal component analysis of log₂-transformed reads showed that the CO₂ treatment had the strongest effect on gene expression (PC1: 37%), whereas the genotype corresponded to the second principal component (11% of variance) (Fig. 1a). All genotypes were tightly clustered at high CO₂ (“High”), consistent with previous findings that high CO₂ rescues the CA mutant phenotype [9]. Under high CO₂ conditions, ca1ca2 mutant plants had 170 significantly differentially expressed genes, whereas the ca1 mutant plants had 42 differentially expressed genes. Of these, 19 differentially expressed genes were shared between the genotypes. Gene ontology (GO) term enrichment found no significant GO term enrichment in ca1 mutants at high CO₂, but the ca1ca2 mutants were enriched in genes involved in regulation of gene expression (GO:0010468).

Fig. 1.

Principal component analysis of log-transformed reads (a) and Venn Diagram indicating statistically significant differentially expressed genes (b) in both low CO₂ time points in ca1 and ca1ca2 mutant plants relative to wild-type. Differential expression was determined by an FDR cutoff of 0.05 using the quasi-likelihood negative binomial generalized log-linear model in edgeR

The ca1ca2 mutants were more strongly affected by the first day at low CO₂ than ca1 mutant plants, as evidenced by their separation in the PCA as well as the number of differentially expressed genes (Fig. 1b). Again, the differentially expressed genes of ca1 mutants were not significantly enriched in any GO terms; however, ca1ca2 mutants were enriched in many stress responses, signaling, carbohydrate metabolism, and cellulose biosynthesis genes. After two days at low CO₂ (“Low2”), both ca1 and ca1ca2 mutants were clearly distinct from wild-type in the PCA and many genes were differentially expressed (Fig. 1b). In both mutant genotypes, the differentially expressed genes at Low2 were enriched in GO categories including carbohydrate metabolism, cellulose biosynthesis, trehalose biosynthesis, stress responses, epidermis development, and organic acid metabolism (Additional file 1).

Five genes were differentially expressed in the ca1 and ca1ca2 mutants across all conditions. However, these five genes were unresponsive to CO₂, and located on the same chromosome as ca1 and ca2 (chromosome 3). Differential expression of these genes was determined to be an artifact of backcrossing the W22-based mutant into a B73 background. Due to linkage, homozygous mutant plants have W22 alleles of genes adjacent to the Ca locus and the wild-type plants have B73 alleles. The differential expression of these five genes in a W22 and B73 background is consistent with the expression measured by [22].

Transcripts accumulated for both ca1 and ca2 in the mutant genotypes as expected due to the location of the Ds element in ca2 and the footprint allele in ca1 [9], although in most cases ca1 and ca2 expression was statistically reduced in the mutants (Table 1). However, previous work showed significant reductions in CA activity in ca1 and ca1ca2 mutants [9] despite transcription of the majority of the gene. As previously described in Studer et al. [9], Ca3 was incorrectly annotated as part of Ca2 in the AGPv3 assembly. However, Ca3 is lowly expressed in leaf tissue, and visualization of the gene region in IGV shows no up-regulation of the gene in ca1 or ca1ca2 mutants (Additional file 2). The apparent down-regulation of Ca3 in mutant plants is most likely artefactual and due to misalignment of Ca1 and Ca2 sequences, as Ca3 has very high sequence homology to these other CA genes. Since fewer Ca1 and Ca2 transcripts are present in mutant plants, there are fewer reads aligning to Ca3 as well. Although there are three additional genes encoding β-CAs in maize leaf tissue, none of these were up-regulated in the mutants (Table 1). Consistent with previous reports showing no change in activity of the key photosynthetic enzymes PEPC or Rubisco in ca1 or ca1ca2 mutant plants compared to wild-type [9], there was no evidence of compensation in the expression of C₃ or C₄ cycle components in either mutant genotype, regardless of the pCO₂ during the sampling. In fact, when significant differences were observed, these genes tended to have lower expression compared to wild-type at low CO₂ (Additional file 3). Similarly, genes encoding photorespiratory enzymes were not differentially expressed between mutants and wild-type plants, although these genes tended to be downregulated at low pCO₂ in all genotypes_.

Table 1.

Log-fold change and significance values of β-carbonic anhydrases found in leaf tissue in ca1 and ca1ca2 mutant plants at each CO₂ condition. Significantly differentially expressed genes were defined as FDR < 0.05

		ca1 log-fold change (FDR)			ca1ca2 log-fold change (FDR)
Gene ID	Predicted subcellular localization	High	Low1	Low2	High	Low1	Low2
GRMZM2G121878 (Ca1)	Chloroplastic/cytosolic	−0.86 (0.98)	−1.1 (0.18)	−1.9 (0.0007)	− 1.66 (0.0098)	−1.25 (0.019)	− 1.32 (0.010)
GRMZM2G348512 (Ca2)	Cytosolic	−0.84 (0.14)	−0.75 (0.20)	−1.0 (0.0025)	−1.97 (0.0003)	− 1.95 (0.0002)	−1.28 (0.0021)
GRMZM2G094165	Chloroplastic	0.49 (1)	0.76 (0.19)	−0.13 (0.82)	0.20 (1)	0.37 (0.50)	0.79 (0.084)
GRMZM2G414528	Mitochondrial	−0.06 (1)	0.05 (1)	0.15 (0.80)	−0.27 (1)	0.072 (0.91)	0.34 (0.43)
GRMZM2G145101	Mitochondrial	−.12 (1)	−0.06 (1)	0.18 (0.75)	−0.042 (1)	0.22 (0.65)	−0.25 (0.59)

Aquaporins have been implicated in mediating CO₂ diffusion across membranes in various plant species [23]. However, no aquaporin genes were consistently upregulated in mutant plants at low pCO₂. In ca1ca2 mutants, ZmPIP1;2 (AC209208.3_FG002) was upregulated and ZmPIP2;5 (GRMZM2G178693) was downregulated. In contrast, only ZmPIP1;6 (GRMZM2G136032) was significantly differentially expressed in ca1 mutants and was downregulated at low pCO₂. Interestingly, the expression patterns of the various maize aquaporin genes in response to CO₂ varied, with some increasing at low CO₂ and others decreasing (Additional file 4). Expression of ZmPIP1;2 increased steadily in response to CO₂, with expression in the ca1ca2 mutants surpassing wild-type at the second low CO₂ time point. In contrast, ZmPIP2;5 exhibited an expression pattern similar to the photosynthetic genes – relatively constant between the high CO₂ and first low CO₂ time point, but sharply reduced after two days at low CO₂. Localization data for the mesophyll and bundle sheath provides additional insight into the potential function of these aquaporin genes. ZmPIP1;2 and ZmPIP1;6 are found in both cell types, whereas ZmPIP2;5 is primarily localized to the bundle sheath [24].

Carbonic anhydrase has previously been shown to function in the stomata to initiate stomatal responses to CO₂. The CO₂-sensing pathway of stomata has been characterized in Arabidopsis thaliana, but it is not well known whether monocots or C₄ species use a similar signaling pathway. Similar to what was previously shown in Arabidopsis [16], ca1 and ca1ca2 mutants have altered stomatal conductance and responsiveness [21]. Therefore, we hypothesized that downstream components of the CO₂ signaling pathway are likely conserved in maize. Homologs of the genes in the CO₂ pathway from Arabidopsis were identified and used to evaluate the downstream effects of CA knockouts on stomatal signaling genes. Genes downstream of CA in the stomatal CO₂ signaling pathway altered their expression in response to changing CO₂ as expected. In addition, several genes involved in the CO₂ sensing mechanism of the stomata were differentially expressed between ca mutants and wild-type plants (Fig. 2). Other components of the CO₂ signaling pathway, including RHC1 and HT1, responded as expected to drive stomatal opening (Fig. 2), but differential expression was not observed between genotypes.

Fig. 2.

Expression patterns for genes involved in the CO₂ signaling pathway of the stomata. Wild-type data are presented as an average of both wild-type genotypes. Data are presented as the average of log-transformed counts-per-million +/− standard error. Green area represents a maize guard cell. The bicarbonate pool produced by CA-medicated CO₂ hydration is sensed by RHC1. When bicarbonate levels are high, RHC1 interacts with and binds HT1. This prevents HT1 from inhibiting OST1 (indicated with dashed line). With HT1 sequestered to the plasma membrane, OST1 is able to activate SLAC1, which results in stomatal closure. As pictured here, RHC1 and SLAC1 are both expected to localize to the plasma membrane [28]

Among the genes differentially expressed in both genotypes at low CO₂ were several groups characteristic of carbon starvation responses. Four genes likely encoding trehalose phosphate synthase (TPS) were expressed significantly higher in both ca1 and ca1ca2 mutants at low CO₂. Interestingly, these TPS genes were expressed at similar levels at high CO₂ in all genotypes, and were generally downregulated in response to low CO₂ in wild-type plants. However, the mutant plants either maintained or increased expression levels of several TPS genes in response to low CO₂ (Additional file 5).

Multiple genes relating to starch and sucrose synthesis were also differentially expressed in ca mutants under low pCO₂. In general, genes involved in the breakdown and mobilization of sucrose, including SWEET transporters and sucrose synthase, were upregulated. The sucrose synthase breakdown pathway, driven by shrunken1 in maize, has lower energetic costs and tends to be favored under energy limited conditions [25]. Two genes encoding glucose-6-phosphate translocators were significantly downregulated in the mutants. Furthermore, genes involved in starch synthesis were likewise downregulated. Overall these results are consistent with decreased carbon supply and decreased investment into the starch and sucrose pathways in the mutant plants relative to wild-type under low CO₂ (Additional file 4).

In addition to reprogramming of carbohydrate metabolism, carbon starvation is often characterized by specific changes in nitrogen metabolism. In this experiment, asparagine synthetase, aspartate aminotransferase, and glutamine synthetase were strongly upregulated under low CO₂ in the mutants. Furthermore, several putative amino acid transporters were also upregulated at low CO₂ (Fig. 3; Additional file 6). Interestingly, orthologs of several transcription factors associated with low energy signaling in Arabidopsis, including regulation of asparagine synthetase, were also upregulated in the mutants at low CO₂. This indicates a likely parallel role for these genes in maize (Fig 3; Additional file 6).

Fig. 3.

Top differentially expressed genes involved in synthesis of high N:C amino acids and remobilization. Colors indicate log-fold change relative to wild-type. Significant differences are marked with an asterisk (*). Gene IDs are given in Additional file 6

Another metabolic process affected by low CO₂ in the mutants, consistent with a carbon starvation response, was cell wall synthesis and growth. Four isoforms of cellulose synthase, three isoforms of β-galactosidase, and three isoforms of endoglucanase were significantly downregulated in ca1 and ca1ca2 mutants compared to wild-type plants at low CO₂ (Additional file 4). Interestingly, several genes encoding homologs of expansin-like genes from Arabidopsis were upregulated at low CO₂ in ca1ca2 mutants.

Differentially expressed transcription factors were identified using publicly available data from GRASSIUS [26]. A total of 324 transcription factors were differentially expressed in at least one genotype/condition. At high CO₂, only two transcription factors were identified as differentially expressed in ca1 mutants; these genes were likewise differentially expressed in ca1ca2 mutants. Similarly, three transcription factors were differentially expressed in ca1 mutants at the first low CO₂ time point, which were all shared with ca1ca2 mutants. However, ca1ca2 mutants also differentially expressed many more transcription factors under high CO₂ and the first low CO₂ time point that were not shared with ca1 mutants: 33 and 115, respectively. After two days at low CO₂, ca1 mutants had 111 differentially expressed transcription factors, ca1ca2 mutants had 165, and 48 of these were shared. Many transcription factor families were represented in these differentially expressed genes (Fig. 4). Of the shared genes with Arabidopsis homologs of known function, several were involved in low energy signaling, regulation of asparagine synthetase, and stomatal regulation (Additional files 6 and 7).

Fig. 4.

Transcription factors that were differentially expressed in ca1ca2 mutants at FDR < 0.01, log-fold change > 2 or  < -2 , in at least one CO₂ condition. Genes were clustered according to expression pattern. Colored bars to the right indicate transcription family membership. Families represented by fewer than 5 transcription factors were grouped into “Other”. The families for these genes can be found in Additional file 7

Twelve genes that were found to be either differentially expressed or responsive to CO₂ in the RNA-seq analysis were validated with qRT-PCR. The log₂-fold change between mutant and wild-type was calculated for each condition and compared to the log₂-fold change calculated using edgeR in the RNA-seq analysis. Similar results were observed using both methods (Additional file 8). The log₂-fold change values generated by qPCR and RNA-seq were significantly correlated (p < 2e^− 16, R² = 0.82; Additional file 9). All genes which were differentially expressed in one or more condition showed similar patterns in the qRT-PCR data. Additionally, we included one gene that was not differentially expressed in the RNA-seq data (PEPC); this result was also confirmed in the qRT-PCR results. Furthermore, patterns of CO₂responsiveness were consistent between the qRT-PCR and RNA-seq data (Additional file 10). Thus, the observed gene expression trends are supported by two orthogonal methods providing strong evidence for the observed results.

Discussion

CA mutants at high CO₂

Consistent with previous observations that ca mutants grow normally under high CO₂, both ca1 and ca1ca2 mutants clustered with wild-type plants in the principal component analysis and there were only a few genes differentially expressed between genotypes at high CO₂. As noted by the GO enrichment analysis, the ca1ca2 mutants differentially express many transcription factors at high CO₂; however, these genes are uncharacterized in maize and have unknown functions. Only 19 genes were differentially expressed in both genotypes at high CO₂, indicating that the mutations had a minor effect on gene expression at high CO₂. Importantly, no photosynthetic genes aside from ca1 and ca2 were affected at high CO₂.

Stomatal signaling

One of the goals of this study was to identify both the short-term and longer-term responses to low CO₂ in ca mutants. At the first low CO₂ time point, expression of many genes was affected in the ca mutants; however, this effect was relatively small in the ca1 mutants compared to the ca1ca2 mutants. This result is surprising considering that ca1 contributes to approximately 90% of bulk leaf CA activity, while ca2 makes up only 7% of bulk leaf CA activity. The results of Studer et al. [9] and the significant differences in the temporal response of ca1 and ca1ca2 genotypes to low CO₂ suggest at least a partial subfunctionalization of the two CA genes. Indeed, despite their small reduction in CA activity, ca2 mutants have altered stomatal conductance [21]. The role of ca2 in stomatal signaling is further supported by gene expression changes in ca1ca2 mutants at the first low CO₂ time point. In Arabidopsis, CA functions in stomatal signaling by modulating the bicarbonate concentration in guard cells in response to CO₂. Under high CO₂, the corresponding high bicarbonate concentration in the guard cells is sensed by a MATE transporter, RESISTANT to HIGH CO₂ 1 (RHC1), which then binds and sequesters the kinase HIGH LEAF TEMPERATURE 1 (HT1) to the plasma membrane [27, 28]. The bound HT1 releases the kinase OPEN STOMATA1 (OST1) to trigger stomatal closure through activation of the S-type anion channel SLOW ANION CHANNEL-ASSOCIATED1 (SLAC1) [29, 30]. Since CA functions in the initial step of the CO₂ signaling pathway through modulating bicarbonate concentrations, any effect of the ca1 and ca1ca2 mutants on CO₂ signaling should be seen in differential expression of these downstream CO₂ signaling genes. Indeed, homologs of the downstream CO₂ signaling genes in our study were responsive to CO₂ as expected and differentially expressed in the ca1ca2 but not ca1 mutants. The maize homolog of SLAC1 (GRMZM2G106921) is strongly downregulated in ca1ca2 mutants compared to wild-type at the first low CO₂ time point, although it returns to wild-type levels after two days at low CO₂. The strong downregulation in ca1ca2 mutants relative to wild-type indicates that ca1ca2 plants adjust expression of CO₂-sensitive stomata genes to relieve SLAC1-driven stomatal closure while adjusting for the CO₂ signaling bottleneck due to low CA activity. The ca1 mutants showed a similar, but non-significant, expression pattern for SLAC1. In addition to SLAC1, two isoforms of OST1 were downregulated in response to CO₂, with more pronounced downregulation in the mutants compared to wild-type. One isoform of OST1(GRMZM2G066867) was downregulated at the first day of low CO₂, whereas the second isoform (GRMZM2G081915) was downregulated at the second time point. Since OST1 functions in activating SLAC1 and therefore inducing stomatal closure, downregulation of OST1 would be expected to enhance stomatal opening in the mutants. Interestingly, this effect is seen at the first low CO₂ time point. We hypothesize that the expression level is altered in response to changing CO₂, but after acclimating to the condition over several days, the plants achieve a steady-state expression level similar to wild-type. This hypothesis would explain why altered expression levels are not observed when the plants are grown at high CO₂, and are largely resolved after two days at low CO₂. Although these results support a role for CA in the stomatal CO₂ signaling pathway, there is still debate in the literature about whether the CO₂ signal comes exclusively from the guard cells or whether it is also influenced by a mesophyll signal such as C_i, sucrose, malate, or other metabolites [31]. It is unclear whether the changes in gene expression of stomatal genes may be affected by a mesophyll signal reflecting reduced photosynthetic rate in the mutants under low CO₂. Additional study on these mutants may provide insight into the contribution of mesophyll and guard cell signals for stomatal responses to CO₂.

Photosynthetic metabolism

After two days at 9.2 Pa CO₂, changes in photosynthetic metabolism were evident across all genotypes, but interestingly, relatively few differences were observed between the genotypes. One of our hypotheses was that CA mutant plants upregulated other isoforms of CA or enzymes in the C₄ pathway to compensate for low CA activity. However, our RNA-seq analysis showed that none of the other CA isoforms found in leaves were upregulated in CA mutants relative to wild-type, indicating that the plants were unable to compensate for the mutation. This result is in contrast to previous work on the carbon concentrating mechanism of Chlamydomonas reinhardtii, which identified multiple isoforms of CA that were induced by low CO₂ [32]. Interestingly, however, one of the predicted mitochondrial β-CAs (GRMZM2G414528) strongly increases in expression in response to CO₂, independent of genotype. This result supports the proposed role for mitochondrial CAs in respiratory/photorespiratory CO₂ recycling [14]. Although photorespiratory genes generally decreased at low CO₂, many respiratory genes increased in expression at low CO₂ (Additional file 11). Similar results have been observed for mitochondrial CA in Chlamydomonas reinhardtii when transitioning cells from high to low CO₂, suggesting an anaplerotic role for mitochondrial CA [33, 34].

Photosynthetic metabolism is significantly changed by differences in the availability of atmospheric CO₂, but surprisingly few photosynthetic genes were differentially expressed between genotypes. However, genes involved in light reactions, photorespiration, as well as the C₃ and C₄ cycles were all downregulated at low CO₂, regardless of genotype. In general, the ca mutants tended to downregulate these genes to a larger extent than wild-type plants; although for most genes this trend was not significant after two days at low CO₂. Interestingly, the only core C₃ cycle gene found to be upregulated in the experiment was Rubisco activase (GRMZM2G162282) in ca1 mutants. A similar trend was observed in ca1ca2 mutants, although the difference was not significant (FDR = 0.087). In all genotypes, the small subunit of Rubisco was downregulated in response to low CO₂. This indicates that under low carbon availability, plants may alter the activation state of Rubisco, while reducing the total protein content to conserve resources. Downregulation of light reaction genes is expected to account for the light stress associated with low CO₂ availability. The downregulation of photorespiratory genes along with the C₄ and C₃ cycles is indicative of the precise balancing of these cycles such that photorespiration is minimized even under these low CO₂ conditions.

Aquaporins

Aquaporins have been shown to mediate CO₂ diffusion in several plant species [23]. In particular, CO₂-transporting aquaporins have been implicated in mesophyll conductance [35–37], which is defined as the movement of CO₂ from the intercellular air spaces to the site of initial carboxylation in the mesophyll. In C₃ plants, CA is thought to enhance mesophyll conductance directly by facilitating diffusion of CO₂ through the chloroplast stroma [1]. Alternatively, in C₄ plants, mesophyll conductance is defined as the movement of CO₂ to the sites of carboxylation by PEPC, which also requires the CA-catalyzed CO₂ hydration to bicarbonate. Previous work has demonstrated that high CA activity is important to offset low mesophyll conductance, particularly under CO₂-limiting conditions [38, 39]. Therefore, aquaporins may play an important role in mediating CO₂ diffusion upstream of CA and we hypothesized that CA mutants may increase expression of CO₂-transporting aquaporins to offset low CA activity for mesophyll conductance.

Currently, the aquaporin isoforms in maize that are responsible for CO₂ movement have not been characterized. In our study, three aquaporin genes (AC209208.3_FG002, GRMZM2G178693, GRMZM2G136032) were differentially expressed at low CO₂ in one of the two mutant genotypes. However, these genes did not have a consistent response to CO₂, indicating that their physiological roles likely differ. Distinct physiological roles are also supported by differences in mesophyll/bundle sheath localization. This may be reflective of some aquaporin genes being involved in CO₂ movement, whereas others may be responding to changes in water availability due to stomatal opening at low CO₂. Since no aquaporin was consistently upregulated in the mutant plants, our results do not support an increase in mesophyll conductance via CO₂-transporting aquaporins as a mechanism to enhance CO₂ availability in CA mutants.

Carbon starvation responses

In spite of few differences between CA mutant plants and wild-type in genes enhancing CO₂ uptake or photosynthesis, the gene expression data presents a clear pattern that the CA mutants were experiencing low carbon stress. This is consistent with previous work showing a reduction in photosynthesis under sub-ambient CO₂ [9]. Carbon starvation responses are often characterized by changes in carbohydrate metabolism and allocation of photoassimilate into organic acids [40, 41]. Various studies have implicated trehalose-6-phosphate synthase (TPS) and associated metabolites including trehalose-6-phosphate (T6P) as a critical signal for sugar levels that potentially interacts with various other upstream and downstream pathways [42–45]. In this study, wild-type plants decreased expression of multiple isoforms of TPS at low CO₂. This is expected, as previous studies have shown low levels of T6P in carbon-starved Arabidopsis seedlings [46]. In contrast, both ca1 and ca1ca2 mutants either maintain relatively constant TPS expression across all CO₂ conditions, or even upregulate TPS at low CO₂. This pattern is somewhat paradoxical given other evidence suggesting carbon starvation in CA mutant plants; however, we hypothesize that CA mutant plants utilize TPS to reprogram carbohydrate and organic acid metabolism in response to the extreme carbon starvation they experience at low CO₂.

Many studies have linked TPS to reprogramming of carbohydrate metabolism, particularly in the context of carbon starvation. Early work on T6P and sugar signaling in Arabidopsis led to the conclusion that T6P is critical for proper use of carbohydrates and growth [43]. Schluepmann et al. [43] discussed their findings in the context of similar work in yeast, and hypothesized that T6P may have evolved as a way to regulate glycolysis to handle sudden and large changes in carbohydrate availability, contributing to organisms’ versatility in handling diverse environmental conditions. Since then, a clear link between T6P, sucrose, and organic acid metabolism has emerged [45]. Specifically, T6P appears to act both as a sugar signal as well as a negative feedback regulator of sucrose; high levels of sucrose lead to an increase of T6P, which in turn stimulates consumption of sucrose [47]. Furthermore, recent work has also shown that induction of TPS is associated with altered partitioning of photoassimilate, promoting carbon flux into organic acids [45]. The mechanism linking TPS to organic acids is less well understood, although there is some evidence that T6P may activate nitrate reductase, affecting nitrogen uptake, or PEPC, affecting respiratory flux [45]. Yet additional studies have suggested a role for T6P in stomatal opening [44]. Our results suggest that ca1 and ca1ca2 mutants do not regulate TPS as expected for carbon starvation. Instead, the high levels of TPS, expected to confer high levels of T6P, may indicate a reprogramming of primary metabolism in mutant plants, stimulating sucrose consumption for growth and altering photoassimilate partitioning into organic acids with high N:C to conserve carbon resources.

Investment in high N:C organic acids are a classic symptom of carbon starvation, as observed in several metabolite studies [40, 41]. These organic acids typically include asparagine and arginine, although results have been inconsistent between studies. In our study, two genes encoding asparagine synthetase were strongly upregulated, as well as aspartate aminotransferase which is an essential step in synthesizing asparagine. In addition, asparaginase (an enzyme which degrades asparagine) was downregulated and homologs of transcription factors characterized in Arabidopsis as regulating asparagine synthetase were upregulated. Furthermore, the isoform of aspartate aminotransferase upregulated in this experiment is not expected to have a role in C₄ photosynthesis; in contrast, the C₄ isoform was strongly downregulated at low CO₂. The upregulation of multiple genes with putative functions in amino acid transport also suggests that plants were actively engaging in remobilization of nitrogen in response to low CO₂. Overall, these findings are consistent with an investment in amino acids with a high N:C ratio for remobilization of nitrogen. Although recent work in Arabidopsis has suggested a role for CA in amino acid metabolism [19], we did not observe significant differences in the biosynthesis of other amino acids. Therefore, our results are most consistent with a remobilization of nitrogen in energy efficient amino acids, rather than a direct role for CA in amino acid synthesis.

Conclusions

RNA-seq analysis of ca1 and ca1ca2 mutant plants in response to low CO₂ found that despite experiencing reductions in photosynthetic capacity under low CO₂ due to CA limitation, CA mutant plants do not compensate by upregulating other CA genes or genes involved in CO₂ uptake or assimilation. However, CA mutant plants had altered expression of genes in the CO₂ stomatal signaling pathway consistent with previous work demonstrating abnormal stomatal responses in CA mutants. These findings support a similar CO₂ signaling pathway in maize as was previously described in the C₃ dicot Arabidopsis thaliana. CA mutant plants also exhibited typical signs of carbon starvation, including sugar signaling and amino acid metabolism. This work identified previously uncharacterized transcription factors which may function to regulate metabolic responses to carbon starvation. Future work on the other isoforms of CA will provide additional insight into the role of CA in photosynthesis and stomatal signaling in maize.

Methods

Plant growth and sampling

Maize lines carrying the mutant alleles ca1-m1::Ds and ca1-d1ca2-m1::Ds (generated and provided by A. Studer; previously described in [9]) in a W22 background were crossed to the reference line B73. These F1 plants were then backcrossed to B73 for two additional generations prior to selfing. This resulted in BC2S1 kernels segregating wild-type and ca1 and ca1ca2 mutant alleles. Kernels were germinated and grown in deep-well plug trays (Hummert #11–8606-1), in a controlled-environment growth chamber (Biochambers; TPC-19) at a photosynthetic photon flux density of 500 μmol m^− 2 s^− 1 at plant height, approximately 50% relative humidity, air temperature of 31 °C/22 °C day/night with a 16-h day. On the 7th day after planting (DAP), homozygous and wild-type seedlings were identified using the PCR assay described in Studer et al. [9]. The pCO₂ in the chamber was initially 920 Pa (10,000 ppm) during the photoperiod. On the 12th DAP, the pCO₂ in the chamber was reduced to 9.2 Pa (100 ppm) in the final hours of the photoperiod and maintained during the photoperiod for the remainder of the experiment.

Leaf samples for RNA were taken three hours after the beginning of the photoperiod on days 12, 13 and 15. Each genotype/time-point combination was sampled in triplicate, with each of these three biological replicates containing a pool of leaves from four individual plants. Leaf samples were immediately frozen in liquid nitrogen and stored at − 80 °C until extraction.

RNA extraction and library preparation

Leaf tissue was ground in a Fluid Management Harbil 5G-HD paint shaker. Total RNA from the pooled samples of each line was isolated using TriPure reagent according to manufacturer’s instructions (Invitrogen), and resuspended in 1x RNA Secure solution (Ambion). Library preparation was performed according to the strand-specific protocol developed by Wang et al. [48], resulting in three libraries per genotype/time-point combination for a total of 36 libraries. Libraries were sequenced on two lanes of an Illumina HiSeq2000 (Cornell University), generating 50-bp single-end reads. All libraries were sequenced on both lanes to account for lane effects.

RNA-seq analysis

Raw Illumina reads from RNA-seq libraries were de-multiplexed using standard Casava settings. Reads were trimmed using TrimGalore [49] with a stringency of 2 and default settings. Trimmed reads were aligned to the maize reference genome B73 AGPv3 (Ensembl) using Bowtie v2.2.5.0 [50] and TopHat2 v2.1.0 [51]. Minimum and maximum intron length were 5 and 60,000 bp, respectively; all other parameters were set to default values. Raw read counts were generated using htseq-count in union mode [52]. After verifying no significant effect of sequencing lane by principal component analysis, the data from each lane were merged in subsequent steps. Lowly expressed genes were filtered using edgeR [53], keeping genes with counts-per-million > 1 in at least three samples. Normalization factors for library sizes were calculated using standard methods. The quasi-likelihood negative binomial generalized log-linear model in edgeR (glmQLFit) was used to fit a model with two sets of mutants and two wild-types as treatments, while downstream a priori contrasts were conducted to statistically test differences in expression between groups of interest. Gene ontology enrichment was evaluated using the R package GOstats [54]. RNA-seq data have been deposited in the NCBI SRA (SRP133928).

Sequencing of biological triplicates for all genotypes at each time point yielded on average 30.8 ± 1.0 million reads per genotype:treatment combination. After quality control and adapter trimming, an average of 95.5 ± 0.1% reads mapped to the B73 reference genome. Differential expression was determined at a false-discovery rate of 0.05.

qRT-PCR

Quantitative real-time PCR was performed on 12 genes which were either differentially expressed or showed a CO₂ response in the RNA-seq data. These included SLAC1, three aquaporins, aspartate aminotransferase, two asparagine synthetases, PEPC, TPS, and three transcription factors. Gene IDs and primer sequences are provided in Additional file 12. Primers for each gene were designed using Roche’s Universal ProbeLibrary Assay Design center. The original RNA samples were treated using DNaseI (NEB M0303) according to manufacturer’s instructions followed by an ethanol precipitation, and quantified with a Qubit fluorometer (Invitrogen). cDNA synthesis was performed using the qScript cDNA synthesis kit (Quantabio) according to manufacturer’s instructions with 1 μg of DNase-treated RNA. The reference gene for normalization (GRMZM2G046402) was selected using the leaf gradient RNA-seq data from Wang et al. [55] because it did not vary with leaf development. Expression of this gene from RNA-seq was consistent across all samples, supporting its use as a reference gene.

qPCR was performed on a QuantStudio 7 Real-Time PCR System (Applied Biosystems) using qScript cDNA SuperMix (Quantabio) and standard cycling conditions. Standard curves were generated from a 7-point serial dilution for each primer set to determine assay efficiency and analyzed using the Applied Biosystems Standard Curve Analysis Module. Estimated efficiency was between 90 and 110% for all primer sets. Expression levels were estimated using the relative standard curve method. Reactions were performed in triplicate, and technical replicates were averaged prior to normalizing by the reference gene. For one gene, GRMZM2G078472, several samples did not amplify during the 40-cycle qPCR run. This is expected given the low expression levels observed in the RNA-seq data. In these cases, a Cq value of 40 was imputed to allow for calculations of log₂-fold change.

Abbreviations

CA

Carbonic anhydrase

GO

gene ontology

HT1

HIGH LEAF TEMPERATURE 1

OST1

OPEN STOMATA 1

PEPC

Phosphoenolpyruvate Carboxylase

RHC1

RESISTANT to HIGH CO₂ 1

Rubisco

Ribulose-1,5-bisphosphate carboxylase/oxygenase

SLAC1

SLOW ANION CHANNEL-ASSOCIATED 1

T6P

Trehalose-6-phosphate

TPS

Trehalose-6-phosphate synthase

Authors’ contributions

AK, AS, and AC conceived and designed the experiment. AK and AS performed the experiment, and AK, AS, and OC analyzed the data and interpreted the results. AK wrote the article with contributions of all the authors. All authors have read and approved the manuscript.

Ethics approval and consent to participate

Not applicable

Consent for publication

Not applicable

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.