Introduction

The world’s population is expected to reach over 9 billion by 2050, as the latest FAO forecast reports (OECD/FAO 2020), determining a 70% increase in the global food demand, especially on cereals (York 2016; Chen et al. 2017; FAO 2017). To satisfy this demand, the development of improved cultivars along with the adoption of new and optimized management practices will be essential. As reported by Beres et al. (2020), the accomplishment of this goal will necessitate a significant yield boost for crops such as wheat, which rate of gain needs to be improved by 30–40% (FAO 2017; Cassman et al. 2020). Wheat is one of the most globally grown crop, supplying about 20% of the human daily calories and protein intake (CRP-WHEAT 2016). Most of the cultivated wheats belong either to the hexaploid Triticum aestivum L. or to the tetraploid T. durum species, the latter representing only the 5% of the total wheat production. Durum wheat is an élite crop in dry lands, such as the Mediterranean basin, with a yearly production average of 40 million tonnes (MT) on a planting area of 17 million hectares globally (IGC 2020). Durum wheat is mostly grown under rain-fed conditions, so its productivity is severely influenced by rainfall as well as biotic (pests and diseases) and abiotic (drought, cold, salinity) stresses. So far, breeding programs aiming at durum wheat yield improvement, will also have to deal with the variable effects induced by the climate changes we are witnessing (Reynolds et al. 2020). The increasing atmospheric CO2 is affecting global temperatures, water availability and weather events, all having strong impacts on plant productivity and carbon assimilation, as well as on agriculture and food production.

So far, bigger efforts will be required to face and deal with these issues and guarantee sustainable crop production and provide climate-smart agriculture, especially from a genetic perspective. Indeed, the identification of candidate genes having a major role in the control of central metabolic pathways is of crucial importance in underpinning and revealing the genomic region actively involved in those traits’ control.

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, EC 4.1.1.39) is responsible for the photosynthetic assimilation of CO2 into organic compounds, the rate limiting step in photosynthesis (Spreitzer et al. 2002). As reported by Perdomo et al. (2021), Rubisco is a quite complex enzyme, consisting of eight large subunits, encoded by a single plastidial gene (RbcL), and eight small subunits which are instead encoded by a nuclear multigene family (RbcS), (Schmidt et al. 1986; Roy 1989; Carmo-Silva et al. 2015; Bracher et al. 2021).

Despite being the most abundant protein in plants (and on Earth) (Raven 2013), Rubisco has been reported as one of the most inefficient enzymes ever, both because of its very low catalytic turnover rate and the predisposition to be inhibited by sugar-phosphate derivatives which latches the active (Portis 1995; Carmo-Silva et al. 2015; Bracher 2021). Enhancing Rubisco functionality has significant repercussions concerning the improvement of plant productivity and resource use efficiency (Parry et al. 2007; Whitney et al. 2011). It has been demonstrated that its activity depends on interaction with the Rubisco activase (Rca), a chloroplast ATPases associated with diverse cellular aActivities (AAA + protein family) encoded by a nuclear gene (Bhat 2017).

Rca plays a significant role in photosynthesis, as catalyzes the Rubisco activation in the photosynthetic light-independent Calvin-Benson-Basham cycle. Rca utilizes the energy derived by ATP hydrolysis to release the sugar-phosphates from Rubisco active sites, thus reestablishing its catalytic competence (Bhat et al. 2017; Shivhare et al. 2017).

Bread wheat (T. aestivum) contains two Rca genes, as most of flowering plant species, including grasses. Carmo-Silva et al. (2015) reported the presence of two Rca sequences, in tandem, on chromosome group 4 in bread wheat (specifically, long arm for A genome and short arm for B and D genomes). TaRca1 has a simpler structure, consisting of only two exons, and originates a single short mature protein named TaRca1-β. TaRca2 is a longer gene, with a total of six exons; as alternative splicing has been demonstrated to occur at the end of exon 5, two isoforms might be obtained: a shorter one, TaRca2-β, and a complete one, TaRca2-α, 37 aa longer than the previous. The two β isoforms different just for 0.5 kDa (TaRca1-β and TaRca2-β being 42.7 and 42.2 kDa, respectively), while TaRca2-α isoform has a predicted molecular weight of 46.0 kDa (Carmo-Silva et al. 2015).

Little is known about Rca detailed mechanism(s) in plants, since only recently plant Rubiscos have been recombinantly expressed (Ng et al. 2020).

Some specific conditions might affect its efficiency, such as abiotic or biotic stresses, especially light, temperature and pathogen infection. Some studies have been carried out to determine the expression of Rca specific isoforms during the 24 h diel cycle (Perdomo et al. 2021), and heat-stress condition (Degen et al. 2020). Nevertheless, several proteomic studies have highlighted a differential expression of Rca enzymes after pathogen infection. Interestingly, Fusarium graminearum seemed to determine a change in Rca protein abundance after infection both in wheat and barley (Zhou et al. 2006; Geddes et al. 2008). The recently published bread and durum wheat genome sequences (Appels et al. 2018; Maccaferri et al. 2019), might be considerably valuable in disclosing the genetic complexity of those species. By exploiting these and other publicly available data, we focused on the isolation and characterization of Rca genes in durum wheat, as well as on the depiction of their phylogeny with other main crops and model species. Also, a RIL map** population previously developed to study Fusarium graminearum resistance (Giancaspro et al. 2016) was used to genetically and physically map the genes, and in silico analysis was performed in order to identify differential expressions under different environmental condition. A deeper knowledge of Rca genes structures as well as their physical and genetic map** might be of greater importance in planning future breeding programs based on the most recent and precise genetic techniques, such as the genome editing approaches, to improve both crop yield and/or pathogen resistance, such as to Fusarium.

Materials and methods

Ribulose-1,5-bisphosphate carboxylase-oxygenase activase (Rca) genes isolation and characterization in durum wheat

Bread wheat Rca gene sequences (accession numbers LM992844 (TaRca1) and LM992846 (TaRca2) (Carmo-Silva et al. 2015) were used as queries and launched against the publicly available durum wheat cv Svevo genome browser (https://d-gbrowse.interomics.eu/). The retrieved gene sequences were further analyzed for gene structure detection and used to primer pairs design for further investigation in different wheat genotypes.

Phylogenetic analysis of Rca genes

Rca1 and Rca2 orthologous genes for T. aestivum, T. durum, Triticum turgidum L. subsp. dicoccoides, Triticum urartu, Aegilops tauschii, Hordeum vulgare, Brachypodium distachyon, Oryza sativa, Sorghum bicolor and Zea mays were retrieved from the EnsemblePlant database (http://plants.ensembl.org/) by blasting the two genes of T. aestivum against each of the considered species’ genome.

For ease of phylogenetic tree reading, abbreviations of common names and genus names will be used as follows. Each plant species considered in this paper was indicated with a two-letter prefix (followed by each gene symbol): Ta for T. aestivum, Tt for T. durum, Tdic for T. dicoccoides, Tu for T. Urartu, Ae for Ae. tauschii, Hv for H. vulgare, Bd for B. distachyon, Os for O. sativa, Sb for S. bicolor and Zm for Z. mays.

All retrieved genes’ cDNAs were aligned by using the ClustalW method via MegaX software (Kumar et al. 2018). Phylogenetic analysis was carried out on 28 CDS sequences using the Maximum Likelihood based on the Tamura-Nei model (Tamura et al. 1993).

The tree was drawn to scale, with branch lengths measured in the number of substitutions per site. The tree was generated with MegaX as well and modified with the FigTree program (http://tree.bio.ed.ac.uk/software/figtree/).

Plant material and polymorphism detection through ecotilling analysis

Two wheat genotypes, 02-5B-318 and Saragolla, parents of a map** population consisting of a set of 135 recombinant inbred lines (RILs) segregant for FHB resistance (Giancaspro et al. 2016), were used to eventually identify polymorphisms within Rca genes sequence to genetically and physically map them. 02-5B-318, a breeding line derived from the resistant Chinese cv. Sumai-3, is a FHB-resistant bread wheat accession, while durum wheat cv. Saragolla is FHB-susceptible. Primer combination for were designed. By using Oligo Explorer (http://www.genelink.com/tools/gl-oe.asp) and Primer3 (http://frodo.wi.mit.edu/primer3/) software, a set of genome specific primer pairs were designed based on the retrieved Svevo Rca1 and Rca2 gene sequences and used to amplify target DNA from both 02 and Saragolla parental lines (Table 1). The RIL population was also used for map** yield and Protein content QTLs (Giancaspro et al. 2019).

Table 1 Primer sequences used to amplify Rca genes from 02-5B-318 and Saragolla parental lines
Full size table

DNA amplifications were carried out in 20 μL reaction mixtures, each containing 50 ng DNA template, 200 M of dNTP, each primer in 0.5 µM concentrations, 1 × buffer, 0.02 U/µL Taq polymerase. The following PCR protocol in a BIORAD thermo cyclers was used: initial denaturation at 95 °C for 3 min, followed by 35 cycles of 94 °C for 30’’, 60.5 °C/63 °C for 30’’, 72 °C for 1 min with a final extension at 72 °C for 10 min (Nigro et al. 2014).

In order to discover and map eventual polymorphism between the genomic sequences of the two varieties previously described, Single Nucleotide Polymorphisms (SNPs) were detected using the Surveyor nuclease kit (Transgenomic, Inc.), following manufacture’s instruction. The heteroduplex formation, CELI digestion and gel analysis were carried out as reported by Nigro et al. (2014, 2017).

To confirm the polymorphisms within genome specific genes, the heteroduplex hybridization digestion pattern was compared to the ones obtained in each parental lines. Furthermore, PCR products giving a digestion pattern after CEL1 treatment were re-amplified and sequenced by 96 capillary 3100 Avant (Life Technologies) (Nigro et al. 2014, 2017).

Development of Rca1 specific markers

The Rca1 sequences of the two wheat genotypes 02-5B-318 and Saragolla were aligned using ClustalΩ (https://www.ebi.ac.uk/Tools/msa/clustalo/) from EBI website to identify polymorphisms. The polymorphic markers were mapped in the ‘Saragolla’ and’02-5B-318’ map** population. The observed segregation ratio for the marker was tested by Chi-square analysis for deviation from the expected 1:1 ratio. The linkage analysis was performed by JoinMap v. 4.0 (Ooijen et al. 2006) and the Kosambi map** function was used to calculate map distances (1943).

Rca genes in-silico expression analysis

In silico expression analysis and the identification of upregulated gene models was carried out using the RNAseq data available at http://www.wheat-expression.com/ using gene models from ‘Chinese spring’. All experimental conditions were considered. Additionally, the homologous genes from ‘Svevo’ were picked in durum wheat.

Results

Isolation of genomic sequences of Rca genes in wheat

The bread wheat TaRca1 and TaRca2 sequences were used as queries by blast analysis in the Svevo portal (https://d-gbrowse.interomics.eu/gb2/gbrowse/Svevo/), and several different durum wheat sequences annotated as the Ribulose-1,5-bisphosphate carboxylase-oxygenase activase, were retrieved. Specifically, 79 different splicing isoforms were identified on 4A chromosome, while 74 different ones were identified for 4B homoeologous.

Indeed, differently from bread wheat and other species, the two genes were not separately and individually annotated, but reported in tandem as different splicing form of the same gene: TRITD4Av1G139700 on minus strand of 4A chromosome, and TRITD4Bv1G060980 on plus strand of 4B. By sequence analysis and comparison of CDS and predicted aminoacidic sequence, it was possible to determine the most similar and more likely to be the durum wheat Rca1 and Rca2 sequences for both A and B genome: Rca1-4A: TRITD4Av1G139700.79 (chr4A:442,229,641.0.442,232,675); Rca2-4A: TRITD4Av1G139700.4; Rca1-4B: TRITD4Bv1G060980.1; Rca2-4B: TRITD4Bv1G060980.64. Both Rca1 homoeologous genes comprise 2 exons, a complete CDS of 1299 bp and a predicted protein of 432 aa.

Rca2 genes have 6 exons, a CDS of 1404 bp and a protein of 467 aa. It should be reminded that in this case, an alternative splicing at exon 5 might induce a shorter isoform (Fig. 1).

Fig. 1
figure 1

Durum wheat Rca1 and Rca2 sequences for A and B genome and physical position on reference Svevo genome

Full size image

Phylogenetic analysis

Orthologues Rca1 and Rca2 gene sequences were retrieved for ten species from the EnsemblePlant database (http://plants.ensembl.org/), including T. aestivum, T. durum, T. dicoccoides, T. urartu, Ae. tauschii, H. vulgare, B. distachyon, O. sativa, S. bicolor and Z. mays. The above reported species were chosen as either sequenced wheat genome progenitor and related species, or model plants which genome sequences have been fully sequenced and annotated. Furthermore, two C4 species were chosen in order to compare the evolutive distance of Rca1 genes with C3 ones.

The identified CDS were checked for sequence structure and similarity. A total of 28 gene sequences were retained to build a phylogenetic tree comprising the ten considered species (Table 2), which were firstly aligned by using the ClustalW method via MegaX software (sequences and alignment are reported in Supplementary file 1).

Table 2 Ensembl entries of the Rca1 and Rca2 genes retrieved in T. aestivum, T. durum, T. dicoccoides, T. urartu, Ae. tauschii, H. vulgare, B. distachyon, O. sativa, S. bicolor and Z. mays. (EnsemblPlants website: http://plants.ensembl.org/)
Full size table

The evolutionary history was inferred by using the Maximum Likelihood method based on the Tamura-Nei model (Tamura et al. 1993). The tree with the highest log likelihood (− 8623.3812) is shown (Fig. 2). Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Codon positions included were 1st + 2nd + 3rd + Noncoding. There were a total of 1774 positions in the final dataset. Evolutionary analyses were conducted in MEGAX (Kumar et al. 2018).

Fig. 2
figure 2

Molecular Phylogenetic analysis by Maximum Likelihood method of Rca1 and Rca2 genes retrieved from T. aestivum, T. durum, T. dicoccoides, T. urartu, Ae. tauschii, H. vulgare, B. distachyon, O. sativa, S. bicolor and Z. mays

Full size image

As shown by the tree in Fig. 2, two main clades, and a third smaller one, were generated. The first one, grouped 11 Rca1 sequences (reported in pink) belonging to the 7 of the analyzed species; as expected, sequences belonging to A genome from Triticum species clustered altogether with the Urartu one, and the same situation was observed for orthologous Rca1 belonging to B and D genome, the latter clustering with the Rca1 of Ae. tauschii as well. H. vulgare and B. distachyon had the more dissimilar sequences, as expected. The same situation was observed for the second cluster, which grouped 12 Rca2 sequences. Triticum orthologous belonging to the same genome clustered together, as previously observed for Rca1 CDS. Interestingly, a third smaller cluster formed, which grouped both Rca1 and Rca2 from Sorghum and Maize, the only C4 species included in the analysis. A divergent sequence was also found out, a Rca2 CDS sequence of rice.

Genetic and physical map** of wheat Rca genes sequences

ECOTILLING approach requires a treatment of the amplified DNA with CELI endonuclease, or any of a number of single strand endonucleases, after heteroduplex formation between the lines to be investigated. Surveyor nuclease cleaves with high specificity at the 3′ side of any mismatch site in both DNA strands, including all base substitutions and insertion/deletions up to at least 12 nucleotides. The treatment of all amplicons for each Rca gene allowed the identification of a mismatch in Rca1 sequence, and specifically, a SNP was identified within the second exon of Rca1-4A gene. Precisely, the T/C SNP identified between the two parental line, 02-5B-318 (C) and Saragolla (T), was mapped in the RIL map** population, and was localized at 123.9 cM (Fig. 3). Analysis of SNP in the predicted mature protein showed that the polymorphism resulted in amino acid substitution in position 260, a leucine to phenylalanine switch L-F (C/T). Unfortunately, no polymorphism was detected within Rca2 gene sequence. The projected SNP of Rca1 in the Svevo genome mapped at 37.7 cM, at physical position 442,230,162 bp. The metaQTL analysis conducted by Maccaferri et al. (2019), identified 14 different QTLs underlying the Rca genes region, most of which related to yield traits, but two of them found to be involved for Fusarium graminearium and leaf rust resistance. On these bases a new QTL analysis for Fusarium resistance was contacted using the same phenotypic data and genetic map of Giancaspro et al. (2016) adding data from Rca1. QTL analysis conformed the presences of a QTL for FHB resistance coincident with Rca gene with a LOD of 3.

Fig. 3
figure 3

Graphic representation of chromosome 4A and Rca-1 gene location

Full size image

Expression profile of Rca genes in wheat

Using the genome browser for ‘Svevo’ (https://d-gbrowse.interomics.eu/gb2/gbrowse/Svevo/) reference genomes and the RNAseq data available at http://www.wheat-expression.com/ (Borrill et al. 2016), we carried out an in silico gene expression analysis to identify in which tissue and phenological stage Rca1 gene transcripts were more abundant.

In addition, the analysis was conducted to correlate the gene expression with biotic stress conditions and detected where the expression was higher during plant development.

The Rca1-4A gene expression was detected in leaves (including flag leaf), followed by roots and spikes. Considering the developmental stages and types of leaf, the higher level of Rca1-4A gene expression was reported at seedling, three leaf and reproductive stages and during grain filling.

According to stress response, the wheat Rca1-4A gene showed to be particularly expressed during Zymoseptoria tritici, Stripe rust, Powdery mildew and Fusarium infections (Fig. 4), conforming what obtained with QTL analysis for FHB resistance.

Fig. 4
figure 4

In silico expression analysis of durum Rca1_4A gene in leaves/shoots under stress disease, in particular in response to Zymoseptoria tritici, Stripe rust pathogen, Powdery mildew and Fusarium infections

Full size image

Expression analysis of Rca1-4A under abiotic stress included: drought stress, heat stress, combined drought and heat stresses, water stress, chitin addition and PEG 6000 treatment to simulate drought and cold stress.

The homeologus gene Rca1-4B was highly expressed in leaves, while lower levels were detected in roots and spikes. Comparing the expression of the Rca1-4A and Rca1-4B genes under stress conditions, the latter one showed a higher expression level under the Powdery mildew infection. Overall, the expression data reported for the Rca genes located on chromosome 4B appeared to be more abundant compared to the ones on the 4A homoeologues.

Discussion

The rise in atmospheric temperatures has affected the length of cereal growing season in large areas of Europe. This, together with water scarcity, are two main constraining factors for crop productivity, as reported by Perdomo et al. (2015, 2016).

Over the past 150 years, the concentration of carbon dioxide in the atmosphere increased by 32%, going from 280 to 370 parts for million in volume. As known, CO2 concentration directly affects photosynthesis, a process extremely susceptible to both drought and heat stress (Chaves et al. 2009; Carmo-Silva et al. 2012; Vile et al. 2012; Mathur et al. 2014; Singh et al. 2014).

Rubisco, the key enzyme of CO2 carboxylation in the Calvin–Benson cycle, represent 50% or more of total proteins in leaves. Its deactivation represents a major element in decreased CO2 assimilation rate at high temperatures (Parry et al. 2013; Bracher et al. 2017). Rubisco activase (Rca) facilitates the dissociation of inhibitory sugar phosphates from the active site of Rubisco (Spreitzer et al. 2002), affecting the efficiency of photosynthetic activity.

Due to its pivotal role in photosynthetic processes, the enhancement of Rubisco activation by Rca may be a potential strategy for improving a photosynthesis-driven increase in crop yield.

In this paper, we focused on the gene structure characterization of Rca genes in durum wheat, and their map** on a RIL population previously developed and used to map QTL involved in Fusarium graminearum resistance.

Previous proteomic studies have indeed highlighted a differential expression of Rca enzymes after Fusarium graminearum infection. Interestingly, the pathogen seemed to determine a change in Rca protein abundance after infection both in wheat and barley (Zhou et al. 2006). Zhou et al. (2006), identified a down-regulation of Rubisco activase (2,sevenfold change detected) along with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and degradation of Rubisco, suggesting that photosynthesis was disrupted or at least decreased after F. graminearum infection in wheat spikes. Recently, a study of Kosovà et al. (2017) analyzed the effect of an artificial infection with Fusarium culmorum and application of deoxynivalenol (DON) on barley spikes at flowering by proteomic analysis (2D-DIGE technique combined with LC–MS/MS). They found out a decrease in photosynthesis-related proteins in Fusarium- and DON-treated plants with respect to control indicating an adverse effect of stress on photosynthetic apparatus. Rubisco activase small isoform A were detected only in Fusarium- and DON-treated plants while they were ab sent in control plants of both analyzed genotypes. According to Geddes et al. (2008), enhanced level of Rubisco activase isoform1was found in Fusarium-treated susceptible and intermediate-resistant barley cultivars, while no change was found in resistant ones, suggesting indicating that they Rca were latterly induced as a consequence of both pathogen infection and mycotoxin application.

We mapped Rca gene on 4A chromosome in 02-5b-318 × Saragolla RIL population, and the SNP was projected on the reference Svevo genome (Macaferri et al. 2019). MetaQTL analysis previously performed highlighted a very interesting situation. Indeed, the region on Rca1-4A gene was surrounded by 14 different QTL. Despite 12 of them were involved in yield and yield-related traits control, two of them were a FHB resistance QTL and a LR resistance one (Prat et al. 2014; Aoun et al. 2016). This confirm our data that identified a QTL coincident with the Rca1 gene location on 4A chromosome. The same investigation carried out on the 4B homoeologous genes region, showed similar result, with 35 different QTL underlying the region, and specifically two involved in FHB resistance QTL and a LR resistance (Marone et al. 2009; Ruan et al. 2012), suggesting that this region and its homoeologous are actively involved in both yield and pathogen resistance.

Furthermore, the SNP (C/T) identified between the two parental lines determined an amino acidic change. Specifically, the amino acid substitution is located at position 260, at the beginning of the α helical subdomain of the AAA + module, and determines a leucine to phenylalanine switch L-F. Previous studies have reported single amino acid substitution in Rca proteins, and in some cases, they have been shown having a solid effect on the enzyme thermostability.

Scafaro et al. (2019) found that two mutant wheat Rca2b proteins showed 5–7 °C increases in thermo stability. Those mutants were found in warm adapted species and presented eight and 11 residue substitutions, respectively, and both including the mutation M159I. Also, Scafaro et al. (2019) reported that a single lysine residue in the C-terminal extension of Rca was responsible for ADP sensitivity decrease.

Degen et al. (2020) demonstrated that a single amino acid residue substitution (methionine/ isoleucine at position 159) in the wheat Rca2b isoform (Scafaro et al. 2016) extended the temperature optimum while maintaining the efficiency of Rubisco activation by Rca. It was supposed that this substitution altered the regulatory properties of Rca which was more thermostable, resulting in a more efficient activation of Rubisco. The occurrence of an isoleucine at position 159 in Rca2 might indicate an adaptation to warm environments, and some authors hypothesized that the ancestral monocots Rcas had a methionine residue at this position.

Further investigation will be carried out to determine its involvement in FHB resistance and whether also the single substitution we reported here might have similar effects on the Rca activity, as a thermal and regulatory switch that can be exploited to improve the efficiency and climate resilience of wheat carbon fixation, besides with.

Ultimately, the importance of targeting Rca in future breeding programs is also enhanced considering its pivotal role in photosynthetic processes, which will be adapting to always more environmental changes. The strong correlation and interaction between C and N metabolisms have been widely studied, and recent reports pointed their attention on how Rubisco and Rca might change their expression whether different N supplement were given to plants. Tetard-Jones et al. (2013), compare the effect of contrasting components (fertilization and crop protection regime) of organic and conventional crop** systems, and found out an up-regulation of both GS2 and Rca when wheat plants were supplemented with mineral fertilizers, as also reported by other authors (Fortunato et al. 2019; Lacolla et al. 2019). Another study from Yousuf et al. (2017) analyzed the proteomic pattern in two wheat cultivars with contrasting nitrogen efficiencies when both high temperature and low nitrogen stresses were combined. The authors found the Rca significantly increased in abundance under stressful conditions, suggesting its potential to regulate Rubisco activity.

Rca has been widely demonstrated to be a valuable candidate gene to enhance Rubisco activity and CO2 assimilation under mutable environmental conditions, especially water and nutrient availability which are predicted to be occurring even more frequently due to climate change scenarios. So far, develo** new traditional as well as innovative breeding programs, such as with specific genome editing experiment to induce new and useful amino acid substitutions, might be a promising strategy for the improvement of yield capacity and sustainability of crops under global warming.