In the present study, uniform and well-growing ‘Tianbaojiao’ banana plantlets with six- to seven-leaf stage were exposed to 4 °C cold treatment in a growth chamber (GXZ-280C, Ningbo Jiangnan Instrument Factory, Ningbo, China) at 28 °C, and to a photoperiod of 16/8 h (day/night, 1500 ± 200 lx). After cold treatment (4 °C), the first fully expanded top leaves were harvested at time points 0, 4, 8, 12, 18, and 24 h. The first leaf (expanded leaf) and different tissue parts (leaf, pseudostem, and root) of the control plants at 28 °C were snap frozen in liquid nitrogen and stored in an ultra-low temperature refrigerator at −80 °C and used for total RNA extraction, qRT-PCR expression analysis, and RLM-RACE test materials.

The genomic and protein data of M. acuminata var. DH-Pahang ( Martin et al., 2016), M. balbisiana var. DH PKW ( Wang et al., 2019), and M. itinerans were downloaded from the Banana Genome Database (http://banana-genome-hub.southgreen.fr). Hidden Markov model files of the GRAS structural domains were downloaded from the Pfam (Mistry et al., 2020) database (http://pfam.xfam.org/), and these three genomic data of bananas were searched (E-value ≤ 1e-10) using the HmmSearch program in the Hmmer 3.0 software to obtain preliminary candidate proteins after redundancy removal. The structural domains of the banana MaGRAS, MbGRAS, and MiGRAS candidate proteins were detected by NCBI-CDD (Lu et al., 2019) (https://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml) tool and SMART (Letunic et al., 2020) (http://smart.embl-heidelberg.de/), and the sequences without GRAS structural domains were removed, resulting in 73, 59, and 58 members, respectively. After NCBI-Blastn/Blastp comparison, and with reference to the Arabidopsis GRAS nomenclature and the position of the chromosome (chr) where each member is located, each member was named sequentially. Segmental duplications and tandem duplications genes of the GRAS family in bananas were analyzed using TBtools (Marques-Bonet et al., 2009). Segmental duplications are usually defined as DNA fragments with highly similar sequences greater than 1 Kb (Chen et al., 2020). Tandem duplications are defined as homologous genes located within 100 kb and separated by 10 or fewer non-homologous spacer genes (Hanada et al., 2008; Zhang et al., 2020). Based on information on the physical location of family members on chr, chr positioning and covariance analysis were mapped using TBtools, and values of non-synonymous substitutions (Ka) and synonymous substitutions (Ks) were calculated for GRAS replicated genes in bananas. The gene duplication occurrence time (T) was calculated as T = Ks/2λ, where the evolutionary rate (λ) of Musa was 4.5 × 10⁻⁹ (Lescot et al., 2008).

A total of 33 Arabidopsis and 53 rice GRAS protein sequences were downloaded from the Tair (https://www.arabidopsis.org/) database and transcription factor database (http://planttfdb.cbi.pku.edu.cn/) as comparison sequences. Using the Neighbor-joining method of MEGA 6.06, phylogenetic trees with a total of 265 GRAS amino acid sequences were constructed for three species of banana, M. acuminata, M. balbisiana, and M. itinerans, Arabidopsis, and rice, with the running parameters of Poisson model, complete deletion, and bootstrap value set to 1,000 replicate tests. Other parameter values defaulted.

Molecular weight, theoretical isoelectric point, instability coefficient, and signal peptide of banana GRAS protein were predicted using ExPASy’s (Wiederschain, 2006) (https://web.expasy.org/protparam/) and SignaIP 4.1 Server (Petersen et al., 2011) (http://www.cbs.dtu.dk/services/SignalP/). The conserved motifs of banana GRAS protein sequences were analyzed through the MEME (Bailey and Gribskov, 1998) (Multiple Expectation Maximization for Motif Elicitation) online website. The software TBtools was used to map the motifs, phylogenetic trees, gene structures, and protein structural domains on the obtained data.

Sequences 2000 bp upstream of the transcription start site of the MaGRAS, MbGRAS, and MiGRAS genes were used for cis-acting element prediction using the PlantCare (Lescot et al., 2002) (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) online website. Their promoter transcription factor binding sites were predicted using the PlantTFDB (Jin et al., 2016) online website (http://planttfdb.cbi.pku.edu.cn/) with parameters set to p-value ≤ 1e-6. The data obtained was compared, analyzed, and finally plotted by TBtools software.

The plant miRNA from the miRbase database and the miRNA mature body sequence identified from the banana transcriptome were used as the database, and the psRNATtarget (Dai and Zhao, 2011) online analysis software (http://plantgrn.noble.org/v1_psRNATarget/) was used for the prediction analysis of miRNAs regulating the MaGRAS gene family of banana. The expected value was set to 0–3, and the others were the default parameter values.

Total RNA from the roots, pseudostem, and leaves of ‘Tianbaojiao’ transplants with satisfactory integrity and purity were mixed, and reverse transcribed by RLM-RACE using the GeneRacerTM Kit (Invitrogen, USA) to obtain cleavage site verification cDNA. Cleavage primers of target genes were designed using DNAMAN 6.0 (Lynnon Biosoft), with primers shown in Table 1 . Two rounds of nested PCR amplification were performed when combined with the universal primers GeneRacer™ 5 ′ Primer, GeneRacer™ 5 ′ Nested Primer, and the specific primers. The primers were synthesized by Fuzhou Shangya Biotechnology Co.

The PCR reaction system was 25 µL, Dream TaqTM Green PCR Master Mix (2X) 12.5 μL, ddH2O 9.5 μL, template cDNA 1 µL, and 1 µL each of the upstream and downstream primers. The amplification procedure was 94 °C predenaturation for 3 min, 94 °C denaturation for 30s, 55 to 60 °C annealing for 30s, 72 °C extension for 30s, 34 cycles, and the last 72 °C extension for 10 min. The amplified products were detected by 1% agarose electrophoresis, and the target bands were recovered by Gel/PCR Extraction Kit. After being ligated into the pMD 18-T vector and transformed into DH5α sensors for bacteriophage PCR, several single colonies were picked for sequencing after TA cloning. The sequencing of the samples was entrusted to Shanghai Boshang Biotechnology Co.

According to the results predicted by psRNATarget prediction, the key members of the MaGRAS gene family were further selected, and qPCR-specific primers were designed on both sides of the cleavage site. The key members of the MaGRAS gene family were analyzed for their specific expression in different tissue sites of bananas and under low-temperature stress. Primer design was performed using DNAMAN 6.0 with a target fragment size of 80–160 bp, and the primer sequence and annealing temperature are shown in Table 2 .

We took samples of roots, false stems, and leaves of transplanting seedlings under 28 °C control (CK) and 4 °C cold treatment for 4 h, 8 h, 12 h, 18 h, and 24 h, and then extracted the total RNA of each sample according to the operating instructions of RNAprep Pure Plant Kit (TIANGEN, China). The concentration, purity, and integrity of the extracted RNA were determined by 1.0% non-denaturing agarose gel electrophoresis and a UV spectrophotometer. RNA processed at different tissue and various time points was reverse-transcribed to cDNA as a template for qPCR using PrimerScriptTM RT Reagent Kit (Perfect Real Time) reverse transcription kit (Takara, Japan), according to the manufacturer’s instructions. CAC was used as the internal reference gene (Chen et al., 2011), and U6 was used as the internal reference gene for miRNA. Roche LightCycler 480 fluorescence quantitative PCR instrument was used to amplify the miRNA. Three biological replicates were performed for each sample, and the mean value was taken. The relative expression level of the target gene was calculated by 2^−ΔΔCT, and Microsoft Excel 2016 and GraphPad Prism version 8.0.2 for Windows (GraphPad Software, San Diego, California USA, www.graphpad.com) were used for calculation analysis and mapping. IBM SPSS Statistics for Windows, version 26.0 (IBMCorp., Armonk, N.Y., USA) was used for significance analysis of the expression level of each gene member.

NCBI-CDD and SMART yielded 73 MaGRAS, 59 MbGRAS, and 58 MiGRAS family members. Coding sequences (CDS) alignment analysis using NCBI-BLAST revealed that some MaGRAS genes contained multiple transcripts; Ma03_g02280 and Ma08_g09130 had three transcripts each, and Ma07_g05950 and Ma11_g18690 had two transcripts each. The resulting members were named sequentially according to the genome database annotation, coding sequences chromosomal mapping information, and the NCBI-Blastp/Blastn alignment results, and with reference to the Arabidopsis GRAS nomenclature method.

ExPASy’s online tool ProtParam was used to analyze the basic physicochemical properties of the proteins encoded by the banana GRAS family ( Supplementary Table 1 ). The number of amino acids in each member of the banana MaGRAS family ranged between 401 and 805 amino acids (aa), and protein molecular weight was between 44902.63–90217.59Da. The theoretical isoelectric points were located between 4.79 and 8.92, and the instability coefficient was between 42.12–565. The number of amino acids for each member of the MbGRAS family was between 316–991 aa, and the protein molecular weight was 33149.59–108846.87Da. The theoretical isoelectric points were found between 4.43 and 8.87, and the instability coefficient was between 40.77–66.99. The number of amino acids in each member of the MiGRAS family was between 338–888 aa, and protein molecular weight was between 37943.74–98811.03Da. The theoretical isoelectric points were between 4.85 and 9.14, and the coefficient of instability was between 42.03–67.58. The theoretical average values of other electric points of MaGRAS, MbGRAS, and MiGRAS were 6.04, 5.91, and 5.91, respectively, and the instability coefficient was greater than 40, indicating that most of them were weakly acidic and unstable proteins.

Chromosomal localization results of the GRAS gene family in bananas showed that the 73 MaGRAS members were unequally distributed on 12 chrs, with chrUn_random and chr 2 having the least distribution, both having only one member. Chr 4 had the most distribution with 12 members, followed by chr 6, 8, and 10 with 10, 9, and 8 members, respectively ( Figure 1A ). A total of 59 MbGRAS members were distributed on 11 chr. Chr 2 had the least distribution with only one member, chr 4 had the most distribution with 13 members, followed by chr 6, 8, and 10 with 7, 7, and 8 members, respectively ( Figure 1B ). M. itinerans assembled only on the scaffold level, whereas MiGRAS was localized on 51 scaffolds(S). All had only one member except for S333, S658, S1345, S1584, and S3004, which had two members each, and S784, which had three members ( Figure 1C ).

To investigate the gene duplication events in the GRAS family of bananas, co-linear analysis was performed on the GRAS genes of banana, rice, and mimosa ( Figures 2A–E ). There were 43 segmental duplication pairs and 1 tandem duplication pair (Ma06_g25650 and Ma06_g25670) in MaGRAS, distributed on all chrs except chrUn_random. There were 28 segmental duplication gene pairs in MbGRAS with no tandem duplication, and at least 2 segmental duplication genes were present on each chr. There were 16 segmental duplication gene pairs and 1 tandem duplication gene pair (Mi_g027597 and Mi_g027598) in MiGRAS, unevenly distributed over 20 scaffolds. GRAS covariance analysis of the three genomes of bananas revealed 121 homologous gene pairs between MaGRAS and MbGRAS, 95 homologous gene pairs between MaGRAS and MiGRAS, and 70 homologous gene pairs between MbGRAS and MiGRAS. In addition, many GRAS genes were found to be relatively conserved among bananas, and 56 of the 73 MaGRAS genes were homologous in both MbGRAS and MiGRAS. Moreover, banana MaGRAS was also homologous in rice and mimosa, with co-linearity between 12 AtGRAS and 9 MaGRAS and between 19 OsGRAS and 24 MaGRAS.

The Ka and Ks values of the orthologous gene pairs of the banana GRAS family were calculated and analyzed whether the family was under natural selection pressure and traced its duplication time during evolution ( Supplementary Table 2 ). The Ka/Ks values of all fragment repetitive gene pairs within the banana GRAS family were less than one, indicating that the banana GRAS genes were subject to purifying selection during evolution. Prediction of the time of GRAS gene duplication events in bananas showed that the time of duplication events in the MaGRAS, MbGRAS, and MiGRAS families was 39.3333–93.4333 million years ago, 41.6936–268.5561 million years ago, and 50.3393–360.9675 million years ago, respectively. It suggests that the MiGRAS gene may be the earliest to start the duplication event, followed by the MbGRAS gene and, finally, the MaGRAS gene.

Analysis of the exon–intron structure of the banana GRAS gene family revealed that MaGRAS has 1–3 exons and 0–4 introns ( Figure 3 ). Six members contain three exons, nine members contain two exons, and all other members have only one exon. Most members do not contain introns, two members contain four introns, and three members have three introns. MbGRAS has 1–11 exons and 0–10 introns, only one member contains 11 exons and 10 introns. Most members have only one exon and no introns, and about a quarter of the members contain 2–4 exons and introns. MiGRAS has 1–9 exons and 0–8 introns, about half of the members have only one exon and no introns, and about one-sixth of the members have more than 3 exons or more than 2 introns.

Predictive analysis of the functional structural domains of banana GRAS proteins using TBtools software showed that the conserved structural domains of GRAS were incomplete for four members (MbSCL4-3, MbSCL4-5, MbSCL4-6, and MbSCL28-2). Whereas the remaining members had a complete GRAS superfamily or GRAS conserved domains. In addition to the common GRAS domain, some members also contained other domains. MaSCR-2a had a PHA03307 superfamily functional domain, MiSCL3-3 had a PRK14698 superfamily and a PHA03247 superfamily functional domain, whereas MaSLR1-2, MaSLR1-3, MaSLN1, MaSLR1-1, MbSLN1, MbSLR1-2, MbSLR1-1, MiGAI-1, and MiSLN1 all had a DELLA functional domain. That might be related to their diversity of functions.

The 20 conserved motifs of the banana GRAS proteins were analyzed using the MEME online website and it was found that each MaGRAS contained motif 1, motif 2, motif 4, and motif 6. Except for a few MaGRAS proteins with partial deletions, the rest contained motif 5, motif 7, motif 8, motif 9, and motif 11. Motif 14 and motif 20 were only present in the LISCL subfamily, motif 15 and motif 16 were only present in the AtPAT1 subfamily, and motif 18 was only present in the AtSCR subfamily. Most of the MbGRAS proteins contained motif 1, motif 2, motif 3, motif 4, motif 5, motif 6, motif 7, motif 8, motif 10, motif 11, motif 15, and motif 16. Motif 14 and motif 17 were present only in the AtSCL4/7 subfamily, whereas motif 20 was only present in the AtPAT1 subfamily. Most proteins of the MiGRAS contained the conserved motif, motif 1, and motif 13, besides motif 15 in the LISCL subfamily, motif 16 and motif 20 in the HAM subfamily, and motif 18 in the AtPAT subfamily. The c-terminus sequence of the GRAS proteins was highly conserved and consisted of five motifs in sequence, namely, HLR I, VHIID, LHR II, PFYRE, and SAW 5, present in motifs 1–9, motif 11, and motif 17 of MaGRAS, motifs 1–8, motif 10, and motif 15 of MbGRAS, and motifs 1–7, motif 9, motif 10, and motif 13 of MiGRAS. The majority of motifs of GRAS protein appeared at the C-terminus, suggesting that the family proteins are highly conserved at the C-terminus. The N-terminus motifs, on the other hand, were more discrete, indicating that there may be dissimilarities in the function of the banana GRAS family.

To analyze the phylogenetic relationships between GRAS proteins from rice and Arabidopsis and GRAS proteins from MaGRAS, MbGRAS, and MiGRAS, 265 GRAS protein sequences from bananas, Arabidopsis, and rice were used to construct phylogenetic trees using the Neighbor-joining method of the software MEGA 6.06. The typical GRAS domain minimum length is about 350 amino acids, and some members of the GRAS family in bananas, rice, and Arabidopsis have a GRAS domain of less than 350 amino acids and are not reliable, so these members were excluded when constructing the phylogenetic tree ( Figure 4 ). The GRAS family of bananas can be divided into 10 subfamilies: SCR, SHR, DELLA, PAT1, HAM, LISCL, LAS, SCL3, SCL4/7, and DLT. The number of members of each subfamily varies, and the most numerous subfamily is PAT1. Each subfamily contains MaGRAS, MbGRAS, MiGRAS, AtGRAS, and OsGRAS genes. OsGRAS4, OsGRAS19, and OsGRAS43 have previously been reported as rice-specific proteins (Liu and Widmer, 2014), but the results of this evolutionary tree indicate that there are multiple banana GRAS proteins clustered within a subfamily with each of these two rice genes. This suggests that OsGRAS4, OsGRAS19, and OsGRAS43 are homologous to these banana GRAS proteins.

The miRNA-regulating MaGRAS family members were predicted using psRNATarget online prediction software. The analysis showed that when the E value was 0–3, 14 members were shown to be targets of 38 miRNAs. Among them, MaSCL27-1, MaSCL27-2, MaSCL27-3, MaSCL27-4, MaSCL6-1, MaSCL6-2, MaSCL22, MaSCL15-1, MaSCL15-2, and MaSCL15-3 were shown to be targets of 22 microRNAs of the miR171 family. MaSCL15-3 was also targeted by five miR408. MaCIGR2-1 can be targeted by 3 miR172, and MaNSP2 was shown to be a target of 10 microRNAs of the miR171 family ( Supplementary Table 3 ).

Further study found that miR171 acted in a manner similar to siRNA, but somewhat different from other miRNAs, with almost all miRNAs acting in a completely complementary pairing manner. Therefore, when the E value was 0, MaSCL27-1, MaSCL27-2, MaSCL27-3, MaSCL6-1, MaSCL6-2, and Mascl22, which are regulated by both mac-miR171a-3p.1 and mac-miR171a-3p.2, and MaSCL27-4, which is also regulated by mac-miR171a.2, mac-miR171c.1, and mac-miR171b.2, were selected as candidate members for subsequent expression analysis ( Table 3 ).

The RLM-RACE method has been used for showing cleavage site verification ( Figure 7 ). MaSCL6-1, MaSCL6-2, MaSCL22, MaSCL27-1, MaSCL27-2, and MaSCL27-3 were all cleaved by both mac-miR171a-3p.1 and mac-miR171a-3p.2, all of which produced single cleavage sites. In addition, MaSCL27-4 was simultaneously cleaved by miR171a.2, miR171b.2, and miR171c.1, generating two cleavage sites. The shear sites of mac-miR171a-3p.1 for MaSCL6-1, MaSCL6-2, MaSCL22, MaSCL27-1, and MaSCL27-2 were located between nucleotides at positions 8 and 9 of the binding region. The MaSCL27-3 cleavage site was between the nucleotides at positions 11 and 12 of the binding region. The shear sites for MaSCL22, MaSCL27-1, and MaSCL27-2 were between the nucleotides at positions 7 and 8 in the binding region. However, the MaSCL27-3, MaSCL6-1, and MaSCL6-2 cleavage sites were between the nucleotides at positions 10 and 11 of the binding region. It indicates that the same miRNA can mediate different members of the same target gene family with the same shear site and can be involved in regulating the expression of target genes simultaneously. In addition, MaSCL27-4 can be cleaved by mac-miR171a.2, mac-miR171b.2, and mac-miR171c.1 simultaneously, producing two cleavage sites at positions 7–8/11–12, 8–9/11–12, and 8–9/11–12 of the binding region, respectively. Interestingly, these miR171 sequences complementarily recognized with target genes can be fully paired and almost all of the shearing sites were located at the GC where the miRNA binds to the target genes.

The expression of seven key MaGRAS members (MaSCL27-1, MaSCL27-2, MaSCL27-3, MaSCL6-1, MaSCL6-2, MaSCL22, and MaSCL27-4) targeted by miR171 was analyzed in different tissue parts of ‘Tianbaojiao’ transplants using CAC as an internal reference gene ( Figure 8 ). The seven key members of the MaGRAS gene family were expressed in roots, leaves, and pseudostems, but the expression was lower in roots than in leaves and pseudostems. Except for MaSCL27-1 and MaSCL27-4, the expression of the other five genes was significantly higher in leaves than in pseudostems, reaching a highly significant difference (P<0.01), and their expression patterns were more similar in each tissue site. Overall, MaGRAS genes showed low levels of expression in both roots and high levels in leaves and pseudostems, with most members reaching the highest expression in leaves.

The results of the relative expression analysis of different MaGRAS members under low-temperature stress are shown in Figure 9 . The expressions of all seven MaGRAS members were downregulated at 8 h of low-temperature treatment in ‘Tianbaojiao’ transplants compared with the 28 °C control. In addition, the expression patterns of the seven MaGRAS during low-temperature stress compared with the 28 °C control level can be divided into two categories. In the first category, GRAS expression levels were positively regulated by intermittent upregulation with increasing low-temperature stress. MaSCL27-4, MaSCL6-2, and MaSCL22 were upregulated or significantly upregulated at 4 h, 12 h, and 24 h, and MaSCL27-1 was significantly upregulated from 12 h to 18 h. The second category was the continuous negative regulation of GRAS expression level with the increasing degree of low-temperature stress. The expression of MaSCL27-3 and MaSCL6-1 was always lower than that of the control without low-temperature treatment, and the expression was lowest at 24 h and 4 h, respectively, while the expression of MaSCL27-2 was continuously lower than that of the control after 4 h and dropped to the lowest at 12 h.

The expression patterns of MaSCL27-1, MaSCL27-2, MaSCL27-3, MaSCL27-4, MaSCL6-1, MaSCL6-2, and MaSCL22, and mac-miR171a-3p.1, mac-miR171a-3p.2, mac-miR171a.2, mac-miR171c.1, and mac-miR171b.2 were analyzed under low-temperature stress at 4 °C, using U6 as an internal reference gene ( Figure 9 ). Compared with CK, the expression of mac-miR171a-3p.1 and mac-miR171a-3p.2d significantly increased at 8h and 18h, respectively, and significantly decreased at 12h and 24h, respectively. The expression of mac-miR171a.2, mac-miR171c.1, and mac-miR171b.2 increased significantly at 18h and decreased significantly at 8h, 12h, and 24h, respectively. MaSCL27-4 was largely negatively correlated with miR171a.2, miR171b.2, miR171c.1, and MaSCL22 and MaSCL6-1 were also largely negatively correlated with miR171a-3p.1 and miR171a-3p.2. In addition, miR171a-3p.1 and miR171a-3p.2 inhibited their targets (MaSCL6-2, MaSCL27-1, MaSCL27-2, and MaSCL27-3) during the pre-low-temperature stress period. To some extent, miRNAs are confirmed to negatively regulate the expression of target genes at the post-transcriptional level.

The GRAS gene family regulates plant growth and development, including GA signaling, root and axillary shoot development, and response to abiotic stress. They have been identified in a variety of plants, but have been reported relatively rarely, in bananas, especially as genome-wide identifications. Based on genomic data of M. acuminata, M. balbisiana, and M. itinerans, 73 MaGRAS, 59 MbGRAS, and 58 MiGRAS genes, respectively, were identified by bioinformatics methods. Four of the MaGRAS genes had more than one transcript (Ma03_g02280, Ma07_g05950, Ma08_g09130, and Ma11_g18690), with significant amplification of membership compared to various species such as Arabidopsis (33), water lily (38), and tomato (53). In addition, the GRAS gene number in M. acuminata banana was the highest, which may be the largest by this genome. Chromosome localization information showed that MaGRAS and MbGRAS genes were unevenly distributed on all the banana chrs, and only one GRAS gene was found on chr2, while most members were distributed on chr4 and formed short clusters of tandem repeats. Most MiGRAS genes were distributed on all scaffolds, but a few scaffolds had multiple GRAS genes.

Gene structure analysis showed that 60%, 67.8%, and 53.4% of MaGRAS, MbGRAS, and MiGRAS genes, respectively, had no intron. GRAS genes without intron also occupied a large proportion of other species. For example, 82.2%, 67.6%, 55%, 81.3%, and 77.4% of Prunus mume ( Lu et al., 2015), Arabidopsis, rice (Liu and Widmer, 2014), lychee (Chen et al., 2021), and tomato (Huang et al., 2015), respectively, indicated that GRAS family genes were highly conformed in different species. Studies have shown that early eukaryotes had very abundant introns. With the evolution of organisms, most species have fewer introns and only a few species have increased introns. However, most of the introns in the extant species with a high number of introns in their genomes were inherited from their ancestors, with few new additions (Iwamoto et al., 1998). It was speculated that the GRAS gene of bananas had intron loss during its evolution. The genes with more introns in bananas mainly exist in the PAT and SCR subfamilies in MaGRAS, PAT and SCR subfamilies in MbGRAS, and SCR and LISCL subfamilies in MiGRAS, suggesting these subfamilies may have had important functions in banana evolution. Functional differences among plant gene family members are regulated to some extent by cis-acting elements in their promoter region (Baranwal et al., 2016). Our study found that there are a large number of light responses, phytohormone responses, and stress responses as well as elements related to growth and development in the promoter region of the GRAS gene family in bananas. These results indicate that the GRAS gene is involved in various stress and hormone responses during banana growth and development, and plays an important role in regulating banana growth at different developmental stages. Many studies have shown that transcription factors play a role in the growth and development of bananas and in banana stress-endurance, for example, bZIP plays an important role in the early development and ripening of bananas (Hu et al., 2016), and C2H2 regulates the ripening of bananas by regulating ethylene synthesis, and regulates cold stress of bananas by inhibiting transcription of ICE1 (Llave et al., 2002; Han and Fu, 2019). There are many kinds of transcription factors in the GRAS gene of bananas, and there are 23 kinds of TF in MaGRAS, MbGRAS, and MiGRAS, indicating that the GRAS expression of bananas is regulated by many kinds of transcription factors, which then regulate the growth and development of bananas.

There are some differences in GRAS family phylogenetic relationships among different plants. Studies have shown that the GRAS families of Arabidopsis and rice can be attributed to 8 and 14 subfamilies (Tian et al., 2004; Lee et al., 2008), while poplar can be attributed to 13 subfamilies (Liu and Widmer, 2014). The clustering analysis of the GRAS family members of banana, Arabidopsis, and rice in the phylogenetic tree showed that the GRAS family in bananas can be classified into 10 subfamilies, indicating that the GRAS family may have acquired new functions during evolution. Combining phylogenetic tree and miRNA prediction analysis, we noted that seven MaGRAS genes regulated by miR171 belong to the HAM subfamily and have high homology with AtHAM1, AtHAM2, AtHAM3, AtSCL15, and AtSCL26 in Arabidopsis. Previous studies have shown that AtHAM1 (AtSCL27), AtHAM2 (AtSCL22), and AtHAM3 (AtSCL6) are involved in the development of apical meristem and axillary meristem of Arabidopsis ( Engstrom et al., 2011), and similar effects have been found in petunia (Stuurman et al., 2002). The results of qRT-PCR showed that the expression levels of these seven members were different in different tissues. The expression levels of seven members in pseudostems and leaves were higher than that in roots, and the expression levels of five members in leaves were the highest. It is suggested that the seven MaGRAS genes belonging to the HAM branch may play an important role in regulating the growth and development of bananas amid environmental stress, and that, except for MaSCL22, the other six members of these seven members contain one to two low-temperature response cis-acting elements, suggesting that the HAM subfamily may play a role in the process of low-temperature response in bananas.

Whole genome duplication is considered to be the main cause of plant gene family amplification. In the process of evolution, plants form genes with similar functions or structures through gene duplication events (Cannon et al., 2004), and series and fragment duplication genes in plant genomes play an important role in coping with environmental stress (Hanada et al., 2008; D’hont et al., 2012). Studies have found that GRAS gene family tandem repeat events are relatively common, with about 2 to 6 tandem repeat genes in each cluster, which occurred twice (4 genes), 12 times (40 genes), 7 times (15 genes), and 4 times (8 genes) in Arabidopsis, poplar, rice, and pepper, respectively (Liu and Widmer, 2014; Zhang et al., 2017). In contrast, the occurrence frequency of tandem repeats in the banana GRAS gene family was less, with only two tandem repeats in MaGRAS and MiGRAS (4 genes), and only one tandem repeats in MiGRAS (2 genes), representing short clusters. It is speculated that these genes may be the product of the banana’s long evolutionary process and that they play an important role in its reproduction. It has been described that bananas may have undergone three genome-wide duplication events during their evolution, namely, α, β (approximately 75 million years ago), and γ (approximately 100 million years ago) (Zhu et al., 2013). The duplications of the MaGRAS family occurred 39.3333–93.4333 million years ago, corresponding to the α or β event. In MbGRAS and MiGRAS, the duplication events of MbSCL9-1 and MiSCL9-1 occurred approximately 268.5561 and 360.9675 million years ago, respectively, corresponding to γ events, while the duplication events of other genes occurred later than 100 million years ago, corresponding to α or β events. It is speculated that whole-genome duplication, gene tandem, and segmental duplication all promoted the GRAS gene family amplification. The duplication event of SCL9-1 occurred earliest in MaGRAS, MbGRAS, and MiGRAS, suggesting that this gene may be originally present in the GRAS family of bananas, rather than produced by subsequent gene duplication. In addition, selection pressure analysis revealed that all banana GRAS duplication genes underwent purifying selection, suggesting that the family may be functionally conserved.

According to promoter identification, nearly half of the GRAS genes in bananas have a low-temperature response element, which is not active under normal conditions but can be activated when plants are subjected to low-temperature stress, and plays an important role in plant cold resistance response. In addition, six members of the banana GRAS family were found to be regulated by miR171. Plant miRNAs are widely involved in the regulation network of response to low-temperature stress (Zhang et al., 2009; Ye, 2013), and miR171 is a class of low-temperature responsive miRNAs, which are upregulated in Arabidopsis and downregulated in rice and Populus trichocarpa in response to low-temperature stress (Liu et al., 2008; Lv et al., 2010). MiR171 in Arabidopsis affects flower and root development, GA response, light signal transduction, lateral organ polarity, meristem formation sieve tube development and response to adversity, by regulating SCL6-II(SCL27), SCL6-III(SCL22), and SCL6-IV(SCL6) of the GRAS family (Llave et al., 2002; Wang et al., 2010). The miR171 family of plants is highly conservative, so it is possible that miR171 in bananas may be involved in regulating the hypothermia response mechanism. The analysis found that MaGRAS were targeted by multiple members of the miR171 family, suggesting that they may play a role in the banana cold-resistance response. In addition, GRAS is involved in the response of various plants to stress. Ye Jiejun (Ye, 2013) found that GRAS gene expression in citrus plants was significantly upregulated under low-temperature stress, and the increase of GRAS gene expression was positively correlated with plant resistance to low temperature. Guo Peng et al (Guo et al., 2013). used Northern hybridization to study the expression characteristics of ZmSCL7 in maize under 4 °C stress, and the results showed that ZmSCL7 was significantly induced by low temperature. LaSCL18 can actively respond to low-temperature stress, and its downregulated expression trend appeared after 12 h of low-temperature treatment but was always higher than the control level, indicating that this gene may be related to the low-temperature tolerance of LaSCL18 (Li LL et al., 2017). By analyzing GRAS gene expression in ‘Tianbaojiao’ transplants during low-temperature stress, it was found that the MaGRAS gene expression pattern during low-temperature stress was not a simple linear change relationship, but a response of multi-directional regulation. MaSCL27-2, MaSCL27-3, and MaSCL6-1 showed negative regulation in response to low temperature, while the other four members showed intermittent positive regulation. In particular, the expression level of MaSCL27-4 reached nearly eight times that of the control after 24 h of low-temperature treatment, and promoter prediction analysis found that MaSCL27-4 contained two low-temperature response cis-acting elements, suggesting that MaSCL27-4 may play a major role in the process of low-temperature response in bananas. Therefore, the GRAS family of bananas may participate in the cold-resistance response of bananas and play an important role in the cold-resistance process.

In general, miRNAs mainly exist in the form of single chains, which are not completely complementary to the binding region of target mRNA in action, resulting in mismatch, inhibition of translation or degradation of mRNA, and regulation of gene expression at a post-transcriptional level without affecting the stability of mRNA. The online prediction results of miRNA showed that miR171 and its target GRAS binding region were completely complementary and paired, which was very similar to the way and function of siRNA degrading target mRNA by means of RNAi. This was consistent with the results of Llave et al., which showed that miR171 and GRAS internal sequences of Arabidopsis were completely complementary. The RLM-RACE verification results showed a clear targeting relationship between miR171 and MaGRAS, which was basically consistent with the software prediction results and research reports. Further analysis of the regulation mode of miR171 and its target genes under low-temperature stress showed that MaSCL22, MaSCL6-1, and MaSCL27-4 were negatively correlated with miR171. These miRNAs regulate the post-transcriptional level expression of the three MaGRAS members in response to low-temperature stress by cleaving target mRNA. The response patterns of the other four members were not completely negatively correlated with miR171, but significantly inhibited each other in the early stage, while there was a certain degree of synergistic and complementary inhibition in the later stage, which fully indicated the complexity of post-transcriptional regulation of miRNA and mRNA in response to low-temperature stress in plants. The regulatory functions of miRNA and the regulatory modes of target genes are diverse.