Structural analyses of CAMTA genes in riceSeven members of the CAMTA gene family were found in ‘Kitaake’ japonica rice (Supplementary Table S1), located on chromosomes 1, 3, 4, 7, and 10, consistent with previous results7,24, but with different numbers of alternative forms (Fig. 1a, Supplementary Table S2). There were 11, 10, and 18 splice forms of all CAMTA genes in Kitaake (Phytozome database), Nipponbare (MSU and Phytozome databases), and recent Nipponbare data (NCBI database)24, respectively. Five CAMTA1 alternative forms were found in Kitaake, and three forms were found in both the Nipponbare MSU and NCBI databases. CAMTA2 and OsCAMTA3b showed one, two, and three alternative forms in Kitaake, MSU Nipponbare, and NCBI Nipponbare, respectively. Both Kitaake and MSU Nipponbare included only one splice form of CAMTA3, 4, 5, 6, and 7, whereas alternative forms were detected in the NCBI database. There were some differences in nomenclature; accordingly, we named the ‘Kitaake’ rice genes OskCAMTAs based on homology with other species20 to facilitate comparative analyses.These OskCAMTAs encoded proteins with lengths of 878 to 1041 amino acids, molecular masses varying from 99.3 to 115.7 kDa, and predicted isoelectric points of 5.58 to 7.40. The CAMTAs showed a similar exon–intron pattern (Supplementary Figure S1a), indicating conservation of the CAMTA gene family. All CAMTA proteins contain conserved domains, including a nuclear localization signal (NLS), CG-1 DNA-binding domain, transcription factor immunoglobulin domain, ankyrin repeats, IQ motifs, and a calmodulin-binding domain (Supplementary Figure S1b).To investigate the homology between CAMTAs from rice and other species, a phylogenetic tree was constructed using the maximum likelihood method (Fig. 1b). The 62 orthologous CAMTA proteins clustered into six subfamilies (I–VI). OskCAMTAs belonged to three subfamilies: I, II, and VI. Rice CAMTAs were closely related to orthologs from maize (including OskCAMTA1/ZmCAMTA1, OskCAMTA2/ZmCAMTA2, OskCAMTA3/ZmCAMTA3, OskCAMTA4/ZmCAMTA4a/ZmCAMTA4b, OskCAMTA5/ZmCAMTA5, OskCAMTA6/ZmCAMTA6, and OskCAMTA7/ZmCAMTA7a/ZmCAMTA7b). Each CAMTA member within a species was assigned to different subfamilies based on homology. Most OsCAMTAs were highly similar to ZmCAMTAs, which are involved in abiotic or biotic stress and hormone responses20. Therefore, we hypothesized that rice CAMTAs function in the response to environmental stress and hormone crosstalk.Fig. 1Phylogenetic analysis of CAMTA proteins. (a) CAMTAs in Oryza sativa japonica. In total, 39 primary and alternative forms of rice CAMTA amino acid sequences were retrieved from Kitaake, Nipponbare, and previous reports. (b) Comparative analysis of CAMTAs of rice and other plant species. Red circles show rice CAMTAs. Black triangles show maize CAMTAs.Haplotype analysis of CAMTA genesTo investigate natural variation in rice CAMTA genes, single nucleotide polymorphism (SNP) data for CAMTAs and their promoters in 3,024 rice accessions were retrieved from the Rice SNP-Seek database. A total of 527 SNPs were found in all CAMTAs and their promoters. In a haplotype analysis, 446 of 527 SNPs were obtained (Table 1, Supplementary Figure S2). CAMTA1 was subdivided into five haplotypes with 69 SNPs. Eight CAMTA2 and CAMTA3 haplotypes with 30 and 58 SNPs, respectively, were found. CAMTA4 had the most haplotypes (14 haplotypes and 84 SNPs). CAMTA5 and CAMTA6 had 10 haplotypes with 85 and 42 SNPs, respectively. CAMTA7 was subdivided into six haplotypes with 78 SNPs. All CAMTAs also showed a higher transition (56.7–75.4%) than transversion rates (24.6–43.3%) (Fig. 2a). CAMTA1, CAMTA3, CAMTA4, and CAMTA7 had more SNPs in promoter regions (2000 bp upstream) than in other regions (Fig. 2b). The CAMTA7 promoter region contained 75.6% of SNPs in the gene. CAMTA2 and CAMTA5 had similar numbers of SNPs in the promoter and coding regions, whereas CAMTA6 had more SNPs in the coding region. Few SNPs were found in the UTRs of CAMTA genes, and no SNPs were found in the 5’UTR of CAMTA3 or CAMTA6.Table 1 Haplotype analysis and SNP counts.Fig. 2SNPs on CAMTA genes and promoters. (a) DNA substitution types. Transitions are A/G or C/T. Transversions are A/C, A/T, G/C, or G/T. (b) SNP locations. Promoter regions are the regions 2000-bp upstream. (c) SNP positions on promoters and gene structure. Black lines show the SNP positions. Red triangles show a missense SNP.We then analyzed effects of SNPs, with a focus on missense mutations (Fig. 2c, Supplementary Table S3). CAMTA1 had the most missense SNPs. Three out of five CAMTA1 SNPs resulted in a change from nonpolar side chains to polar uncharged side chains and electrically charged groups to nonpolar or polar uncharged groups. CAMTA2, CAMTA3, and CAMTA7 contained one, two, and one missense SNPs, respectively. All resulted in amino acid changes; however, they involved a change to the same side chain group. In CAMTA4, two changes in the nonpolar side chains to electrically charged side chains were observed for missense SNPs. Three CAMTA5 SNPs changed a charged group to a polar uncharged side chain and a polar uncharged group to a nonpolar side chain. The missense SNP in CAMTA6 changed the polar uncharged group to a nonpolar group. To visualize the genetic relationships among subpopulations, we constructed a haplotype network for each CAMTA (Supplementary Figure S3). There were five sub-populations: admixture (admix), aromatic (aro), aus-type (aus), indica (ind), and japonica (jap). The haplotype network revealed distinct separation between the indica and japonica groups in the CAMTA3 and CAMTA6 haplotypes. Moreover, CAMTA1, CAMTA3, CAMTA6, and CAMTA7 haplotypes belonging to indica were genetically distant from japonica haplotypes.Associations between phenotypes and haplotypes were evaluated. We used phenotype data from the Rice SNP-Seek database, which included grain weight and salt injury scores (SIS). The grain weight data consisted of 100-grain weights ranging from 1 to 5. The SIS is typically given on a scale of 1 to 9, depending on plant symptoms under salt treatment. A score of 1 indicates no or few symptoms (tolerant) and a score of 9 indicates severe symptoms (susceptible). These SIS data were obtained at an electrolyte conductivity of 12 dS m− 1 or 100 mM NaCl and scores of 3, 5, 7, and 9 were assigned. This analysis focused on CAMTA1 because it had the fewest haplotype groups and the most missense SNPs (Fig. 3a). Haplotypes 2 and 5 of CAMTA1 showed higher grain weights than those of other haplotypes, and these differences were significant (Fig. 3b). Most of the subpopulations of haplotypes 2 and 5 were japonica. In the salt-responsive phenotype comparison, there was a significant difference between haplotypes 5 and 1/2/3 (Fig. 3c). Haplotypes 1, 2, and 3 tended to show higher salt tolerance than that of haplotype 5. In contrast to haplotypes 2 and 5, haplotypes 1 and 3 subpopulations were typically indica (Fig. 3d). We further analyzed the SNP differences between haplotypes 1 and 5 (which differed in the salt-responsive phenotype). We also investigated linkage disequilibrium (LD) between each SNP site. Variants in the coding region were more closely linked to each other than to those in the promoter region (Fig. 3e).Fig. 3Haplotype analysis of CAMTA1. (a) Haplotype table. ATG position is set to 0. (b) and (c) are phenotype comparisons between haplotypes. (b) Grain weight. (c) Salt injury score at EC12. (d) Haplotype network. (e) LD-blocks.Promoter cis-element analysisWe identified cis-acting regulatory DNA elements in the putative promoter sequences of CAMTA genes from Kitaake and Nipponbare (2000 bp upstream of transcription start sites). A total of 42 cis-elements that regulate transcription by binding to transcription factors within 28 unique transcription factor families were identified (Fig. 4). There were slight differences in cis-element numbers between Kitaake and Nipponbare. These elements are bound by transcription factors involved in the regulation of growth and development, as well as phytohormone and environmental stress responses. All CAMTAs have a large number of cis elements that bind TCP, a plant-specific family of developmental regulators, followed by NF-YB, B3, and AT-Hook. CAMTA6 was found to have the largest number of AP2-binding elements. With the exception of CAMTA1, the promoters of all CAMTA genes were found to contain binding sites for EIN3- and GATA, which are characteristic of ethylene- and light-responsive transcription factors, respectively. Furthermore, with the exception of CAMTA3, all CAMTA promoters contain elements that are bound by LEA5, which responds to environmental stress, and with the exception of CAMTA6, all CAMTA promoters were found to contain cis elements that are recognized by MADF and WRKY. In addition, we found that only CAMTA1 and CAMTA3 possessed cis elements of storekeeper (STK) transcription factors, which are involved in plant development, whereas only CAMTA3 and CAMTA6 have the binding sites for BES1, a brassinosteroid (BR)-responsive transcription factor. Each CAMTA gene possessed a specific combination of these elements, suggesting that they have different functions.Fig. 4Detection of transcription factor binding sites in the 2000-bp sequences upstream of the transcription start sites of CAMTA genes. Numbers indicate the number of cis elements in CAMTA promoters that bind to each transcription factor. The upper panel shows CAMTA promoters in Kitaake from the Phytozome database. The lower panel shows CAMTA promoters in Nipponbare from the MSU and Phytozome databases.
CAMTA gene expressionThe CAMTA gene expression data for the Nipponbare rice genotype were retrieved from the MSU Rice Genome Annotation Project database. All FPKM values were visualized as a heatmap (Supplementary Figure S4). Most OsCAMTAs showed relatively high expression levels in the shoot, pre- or post-emergence inflorescence, embryos at 25 days after pollination (DAP), and pistil compared to the other tissues. All OsCAMTAs, except OsCAMTA3, had relatively lower expression levels in the seeds at 10 DAP, endosperm at 25 DAP, and anther than in the other tissues. In addition, the expression levels of CAMTAs at various time points in the shoots and roots under salt stress were determined using RT-qPCR. The expression levels of most CAMTAs in shoots, roots, or both were altered under salt stress (Fig. 5). In shoots, all CAMTAs except CAMTA6 were significantly up-regulated 3 h after salt treatment (1.5-2.6-fold change). The expression of all CAMTAs except CAMTA5 tended to decrease at 24 h after salt treatment (0.5-0.8-fold change). In the roots, all CAMTAs were up-regulated significantly at 3 h after salt treatment (2.2-7.4-fold change). Overall, all CAMTA genes were significantly up-regulated 3 h after salt treatment in both shoots and roots. These results indicate that all CAMTAs responded rapidly to salt stress at the transcriptional level.Fig. 5Expression of CAMTA genes under normal and salt stress conditions in rice seedlings, as determined using RT-qPCR. EF1α was used as the endogenous control. Values are shown as means ± SD (n ≥ 3). Asterisks represent significant differences between salt stress and control conditions (Student’s t-test, *P < 0.05, **P < 0.01).Co-expression and motif analysesTo identify candidate CAMTA-regulated genes, co-expressed genes were retrieved from the RiceFREND database and analyzed based on reported motifs. Genes containing (C/A)CGCG, (T/G/C) or VCGCGB, and (C/A)CGTGT or MCGTGT motifs in their promoters can be directly regulated by CAMTA. In the rice genome, 70.5% of the genes had at least one of these motifs (Fig. 6a), 42.8% of the genes contained both motifs in their promoters, and 27.7% of the genes had one of the motifs. More genes contained only VCGCGB (22.6%) than only MCGTGT (5.1%). When we identified co-expressed genes of each CAMTA using the RiceFREND database, 2111 non-redundant genes were identified. CAMTA2 and CAMTA6 showed a large number of co-expressed genes (491 and 475 genes, respectively). CAMTA1 had the fewer co-expressed genes (208) (Fig. 6b). Next, we identified overlap between genes with specific motifs and co-expressed genes, revealing 1836 genes. More than half of the co-expressed genes contained both the VCGCGB and MCGTGT motifs in their promoters (Fig. 6c). The expected number of co-expressed genes with specific motifs was calculated based on the proportion of genes with specific motifs in the rice genome and was compared with observed numbers. As evaluated using the chi-squared test, there were significantly more genes in the observed group than in the expected group (Fig. 6d).Based on a weighted Pearson correlation coefficient (PCC) of more than 0.5, 690 genes co-expressed with CAMTA with specific motifs were selected to reconstruct a co-expression network (Fig. 6e). According to the network, 30 genes co-expressed with CAMTA were correlated with at least two CAMTAs. Most of these encoded kinases and zinc finger proteins (Table 2). Only one CAMTA-co-expressed gene was correlated with three CAMTAs, Os05g0312000, which encoded a structural constituent of the ribosome. A GO enrichment analysis of all the genes in the network was performed (Fig. 6f). The positive regulation of the defense response showed the highest fold enrichment in terms of biological process, followed by mRNA process including transcription. Zinc ion binding, a molecular function term, showed the highest significance (false discovery rate enrichment). We also analyzed results for each CAMTA-co-expressed gene set and each CAMTA module (Supplementary Figure S5).Table 2 CAMTA-co-expressed genes for two or more CAMTAs.Fig. 6CAMTA-co-expressed genes with specific motifs. (a) Proportion of genes with specific motifs, VCGCGB or MCGTGT, in the rice genome. (b) Number of CAMTA-coexpressed genes retrieved from the RiceFREND database with mutual rank < 400. (c) Proportion of motif types found in CAMTA-co-expressed genes. (d) Numbers of CAMTA-co-expressed genes with specific motifs. Chi-squared analysis for comparisons of CAMTA-co-expressed genes with specific motifs between observed and expected groups. The number of CAMTA-coexpressed genes with specific motifs was significantly higher in the observed group (orange bar) than in the expected group (blue bar) (**P < 0.01; *P < 0.05). (e) Network of CAMTA-co-expressed genes. (f) GO enrichment analysis of 690 CAMTA-co-expressed genes with specific motifs.According to fold enrichment, the CAMTA1 module was mostly involved in “UDP-glucose transmembrane transporter activity.” The CAMTA2 module was involved in “protein import into peroxisome matrix, docking” and “recycling endosome membrane.” The CAMTA3 module was involved in “positive regulation of defense response to bacterium.” The CAMTA4 module showed the highest enrichment for the term “response to herbivore.” The CAMTA5 module showed the highest enrichment for “proteoglycan metabolic process.” The gene group in the CAMTA6 module was involved in “mRNA processing.” The CAMTA7 module was involved in “S-adenyl-l-methionine transport.” These GO enrichment analyses suggest that each CAMTA regulates a different set of genes.Protein interaction network analysisAn interaction analysis was performed using the STRING database. Thirty interactors were identified with moderate confidence (Fig. 7a, Supplementary Table S4). CaM1 was a common interactor in the first shells of OsCAMTA1 and OsCAMTA5 and in the second shells of OsCAMTA2, OsCAMTA3, OsCAMTA6, and OsCAMTA7. OsCAMTA1 had five direct interactors, including the cytokinin receptor (HK3) and heat shock protein-like protein (OsJ_04594). OsCAMTA5 had the most direct interactors with five predicted functional partners (CaM1, SL1, PRR95, ELP3, and SERR). ELP3 had a subnetwork involved in the regulation of transcription initiation and elongation. Putative protein kinase (OsJ_25984), TAZ zinc finger protein (OsJ_04382), and glucosidase (OS03T0216600-01) were common interactors of OsCAMTA2, OsCAMTA3, OsCAMTA6, and OsCAMTA7. Another OsCAMTA2 interactor was a dephospho-CoA kinase family protein (OS01T0360600-01), with a subnetwork involved in the regulation of intracellular Coenzyme A (CoA) concentration. The OsCAMTA4 network was separated from the other OsCAMTAs and showed four interactors, the rRNA-processing protein, EFG1 domain-containing protein (OS03T0797700-01), peptidase C19, and ubiquitin carboxyl-terminal hydrolase 2 family protein (OS09T0407900-01). However, most edges of this network were determined by text mining and known interactions for homologous proteins in other species. A GO enrichment analysis of all 30 interactors (Fig. 7b) revealed enrichment (with the highest FDR) for the coenzyme A biosynthetic process and dephospho-CoA kinase activity in the biological process and molecular function categories, respectively.Fig. 7Putative CAMTA interactors in Oryza sativa ssp. japonica. (a) Protein association network of CAMTAs. Overlapping genes/proteins between the CAMTA-co-expression network and this network are indicated by yellow circles. (b) GO enrichment analysis of 30 CAMTA interactors.Moreover, we identified three overlapping proteins with the CAMTA-co-expression network analysis in CAMTA2, CAMTA6 and CAMTA7 modules. Os02g0134000 was identified in the CAMTA2-co-expressed gene module and CAMTA2-protein module. Os04g0431200 was in the CAMTA7-co-expressed gene module and in the CAMTA4-protein module. Os06g0639600 or OsJ_22107 was in the CAMTA6-co-expressed gene module and in the CAMTA2- and CAMTA5-protein modules.Phenotypes of camta mutants under salt stressUsing the CRISPR-Cas9 system, we generated knockout mutant lines of OskCAMTA1, OskCAMTA2, and OskCAMTA6. Observations of the phenotypes of camta1, camta2, and camta6 mutant lines and wild-type rice (WT) under normal and salt stress conditions revealed that all camta mutants and the WT were characterized by an increase in salt-induced symptoms, as determined using the standard evaluation of salt injury score (SES) starting on day 2 post-salt treatment (Fig. 8a). However, when assessed on day 4, we found that compared with the WT, the camta1 and camta6 mutants tended to have higher SES stability index (SI) values and all mutants appeared to have higher index values than the WT on day 6 after salt treatment, however only camta1 mutant had statistically significantly higher SES than WT indicating that the camta1 mutant was more sensitive to salt stress than the WT. SPAD values, which provide a measure of leaf greenness, were slightly reduced in all mutant and WT plants subject to salt stress (Fig. 8b). However, we detected differences in the patterns of the SPAD values recorded for the WT and mutants. Whereas lower values were recorded for the WT on day 2, they showed no further decline until day 6. Contrastingly, the SPAD values of camta1 and camta2 mutants continued to decline until day 4 and day 6, respectively. Compared with WT plants, we recorded lower SPAD values for the camta1 mutant on day 4; and lower values for both the camta1 and camta2 mutants on day 6.Fig. 8Phenotypes of camta mutants under salt stress. (a) Standard of evaluation salt injury score (SES). (b) SPAD. (c) Relative water content (RWC). (d) Cell membrane stability (CMS). All values are presented as the means with standard deviations. Four biological replicates were used for all measurements. The statistical analysis was performed using Duncan multiple range test for each day after salt treatment. The asterisk shows significant differences (p < 0.05) compared to WT.We subsequently measured relative water content (RWC) and cell membrane stability (CMS), the value of which was determined based on electrolyte leakage. We accordingly detected reductions in the RWC of all camta mutants and the WT in response to salt stress (Fig. 8c). Compared with that of other plant genotypes, the RWC stability index values obtained for the camta2 mutant appeared to decline to a lower level on day 4, whereas in the case of the camta1 mutant, the recorded RWC SI values continued to decline with prolonged exposure to salt stress, and tended to be lower than those obtained for the WT on day 6. With respect to CMS, we found that all camta mutants and the WT showed similar SI patterns, characterized by gradual declines until day 4 of salt stress (Fig. 8d). Based on these observations, we established that compared with the WT, the phenotypes of the camta1 mutant were negatively affected to greater extent by exposure to salt stress, with the most pronounced effects generally being detected in nearly all parameters determined.
Sequence-based analysis of the rice CAMTA family: haplotype and network analyses
