The seeds of birch were obtained from the State Key Laboratory of Tree Genetics and Breeding (Northeast Forestry University) and planted in a mixture of vermiculite and soil (v: v= 1:3) in pots. The birch seedlings were cultivated in a thermostatic greenhouse at a temperature of 25 ± 2°C, relative humidity of 65-70%, light intensity of 400 μmol·m^-2s^-1, and a light/dark photoperiod of 16 h/8 h.

The DNA and protein sequences of 40 BpGRASs were searched from the birch genome database (GenBank accession: PRJNA285437). The ExPASy tool (http://www.expasy.org/tools/protparam.html) was used to predict the physicochemical parameters of the putative 40 BpGRAS proteins, such as the molecular weight (MW) and isoelectric point (pI). Based on the birch genome database, the chromosomal locations and duplications of 40 BpGRASs were physically mapped on the 14 chromosomes of birch.

The GRAS proteins of the other 5 plant species were obtained as followed ways: Arabidopsis from TAIR (http://www.arabidopsis.org/), rice from and PlantTFDB v5.0 (http://planttfdb.cbi.pku.edu.cn/), Camellia sinensis (tea) from Wang et al. (2018), Phoenix dactylifera (P. dactylifera) and Theobroma cacao (T. cacao) from Cenci and Rouard (2017). Phylogenetic analysis was performed with 40 BpGRAS proteins, 32 GRAS proteins from Arabidopsis, 38 from rice, 52 from tea plant, 59 from P. dactylifera and 44 from T. cacao using the neighbor-joining (NJ) method in the MEGA X program (Kumar et al., 2018; Zhang et al., 2021). Multiple sequence alignments of the selected 6 GRAS proteins of birch plants and 3 GRAS proteins of different species were performed using ClustalW (Thompson et al., 1997). The MEME/MAST program (http://meme-suite.org) was used for conserved protein motif analysis with a maximum of 20 motifs. Putative promoter sequences of 6 selected BpGRASs were obtained using 2 kb of a genomic sequence upstream of the translation start site of the 6 BpGRASs and were extracted from the birch genome database, respectively. Cis-acting elements were analyzed using the website PlantCARE (http://bioinformatics.psb.ugent.be/webtools/plantcare/html).

Two-month-old birch seedlings grown in the soil were watered with a solution of 200 mM NaCl for 3, 6, 12, 24, and 48 h, respectively, and treatment for 0 h was provided as the control. Three seedlings were collected after each treatment process. Total RNA of the root, stem, and leaf tissues of birch was extracted using the Universal Plant Total RNA Extraction Kit (BioTeke, Beijing, China). The extracted RNA was reverse-transcribed into cDNA with oligo (dT) primers in a reaction volume of 10 μL using a PrimeScript RT Reagent Kit (TaKaRa, Beijing, China) as a template for RT-qPCR. Actin (GenBank accession: MK388227) and β-tubulin (GenBank accession: MK388229) were used as reference genes for RT-qPCR analysis (Li et al., 2019). Each 20-μL volume of the reaction mixture included 10 μL of SYBR Green Real-time PCR Master Mix (Toyobo, Osaka, Japan), 2 μL of cDNA template (100 ng), and 0.5 μL of specific primers (10 μM). Amplification was performed by the reaction mixture at 94°C for 30 s, followed by 45 cycles at 94°C for12 s, 58°C for 30 s, 72°C for 45 s, and 82°C for 1 s during plate reading. Real-Time PCR Thermal Cycler-qTOWER³ (Analytik Jena AG, Jena, Germany) was used to perform RT-qPCR. Three replicates were used for each sample and the purity of the PCR products was evaluated using a melting curve. The expression levels were calculated from the cycle threshold using the 2^−ΔΔCt method (Livak and Schmittgen, 2001), and used to generate a heat map using R studio. The primers used are shown in Supplementary Table 1 .

Total RNA of birch plants was extracted using the Universal Plant Total RNA Extraction Kit (BioTeke, Beijing, China). Total RNA was reverse transcribed into cDNA using a PrimeScript RT Reagent Kit (TaKaRa, Beijing, China), which was used as a template for PCR. We designed primers for the cloning of 6 GRAS genes from different tissues of birch plants. All the primers are listed in Supplementary Table 2 . The PCR procedure was as follows: the reaction mixture at 94°C for 3 min, subjected to 30 cycles at 94°C for 30 s, 58°C for 30 s, 72°C for 1 min and 30 s, and 72°C for 7 min. The PCR products were purified and recovered using a Cycle Pure Kit (Omega, Norcross, GA, America). The obtained full-length cDNA of GRAS genes was inserted into the pROKII plasmid, regulated by the CaMV 35S promoter (35S:BpGRAS), and inverted-repeat cDNA sequences of GRAS genes were constructed into the pFGC5941 RNAi vector (pFGC : BpGRAS) for silencing gene expression. The recombinant plasmids exhibiting overexpression (35S:BpGRAS) and inhibited expression (pFGC : BpGRAS) of BpGRASs were transformed into Agrobacterium tumefaciens strain EHA105 via electroporation.

The recombinant plasmids exhibiting an overexpression (35S:BpGRAS) and inhibited expression (pFGC : BpGRAS) of BpGRASs were transferred into 4-week-old birch seedlings via high-efficiency transient transformation by the method of Ji et al. (2014), using the empty vector (pROKII) as a control. Stable transgenic birch lines were obtained using method of Agrobacterium tumefaciens-mediated transformation (Guo et al., 2017) with recombinant plasmids exhibiting an overexpression of BpGRAS34 with the wild-type birch (WT) as the control. Whole transient-transformation plants of overexpression (OE), inhibited-expression (IE) and control plants were treated with 1/2 MS or 1/2 MS containing 150 mM NaCl for 24 h for RT-qPCR and the measurement of physiological indexes. The RNA of whole birch plants of stable transgenic lines was extracted and reverse transcribed into cDNA for RT-qPCR to analyze expression levels, respectively. The primers used were listed in the Supplementary Table 2 . Stable transgenic lines were treated with 1/2 MS or 1/2 MS containing 150 mM NaCl for 24 h to measure physiological indexes. The electrolyte leakage assay was performed and the malondialdehyde (MDA) content was measured in accordance with the method described by Ji et al. (2014) and Wang et al. (2010). The level of ROS was determined using the Plant ROS Elisa Kit (SenBeiJia, Nanjing, China) and hydrogen peroxide (H₂O₂) content was measured with the Hydrogen Peroxide assay kit (Nanjing Jiancheng, Nanjing, China). Superoxide dismutase (SOD) and peroxidase (POD) activities were detected using the protocols described by Asada et al. (1973) and Wang et al. (2010), and proline content was measured using the method described by Bates et al. (1973). Three biological replicates were performed in each experiment.

After treatment with 1/2 MS containing 150 mM NaCl for 2 h, the leaves of birch seedlings were used for biological staining. Cell death was observed via Evans blue staining, using the protocol described by Kim et al. (2003). The H₂O₂ and superoxide O 2 − · contents were determined via diaminobenzidine (DAB) and nitroblue tetrazolium (NBT) staining of the detached leaves based on the methods described by Zhang et al. (2011). After grown in pots with the soil for two months, stable transgenic birch plants and wild-type birch plants were used for phenotype analysis watered with 200 mM NaCl for 10 days. Plants treated with water were served as the control.

Statistical analyses were performed using SPSS (IBM SPSS 22, IBM Corp., Armonk, NY, USA). Data were analyzed using the Student’s t-test. The differences were significant if p< 0.05; this is represented by the * symbol in figures. Three biological replicates were generated for statistical analyses.

A total of 40 BpGRASs were obtained from the birch genome database and identified. Their physicochemical properties were further analyzed using ExPasy ( Supplementary Table 3 ). Most of these proteins had typical GRAS domains containing approximately 350 amino acids (aa), while the GRAS domains of BpGRAS6 and BpGRAS8 were severely truncated and had less than 150 aa. The predicted lengths of the 40 BpGRAS proteins and their MWs (kDa) ranged from 182 aa to 830 aa and 21.07 kDa to 90.86 kDa, respectively. For most BpGRAS proteins, the theoretical pI values ranged from 4.89 to 6.88; four of the BpGRAS proteins were alkalescent, indicating that most BpGRAS proteins were acidulous and may cause variations in BpGRAS protein functions in different environments. The grand average of hydropathy (GRAVY) of all BpGRAS proteins (ranging from -9.773 to -0.102) suggested that all BpGRAS proteins are hydrophilic; these results were similar to the results obtained for GRAS proteins in Prunus mume (Lu et al., 2015). Most predicted instability index values exceeded 40 (ranging from 40.65 to 61.17), indicating that a majority of BpGRAS proteins were unstable, except for BpGRAS35 (34.86) and BpGRAS36 (39.39). Most BpGRASs had no introns, which indicates that the sequences of BpGRASs are conservative at a certain extent.

The identified 40 BpGRASs were further mapped and positioned on 14 birch chromosomes (Chr1 to Chr14) ( Figure 1 ). In general, 40 BpGRASs had uneven distributions on 14 birch chromosomes expect BpGRAS2. The densities of BpGRASs distributed on birch chromosomes were different and uneven among different chromosomal regions. There were no BpGRASs found on the Chr4, 7 and 9. Chr11 contained the most BpGRASs and 14 BpGRASs (35%) were distributed on this chromosome, followed by Chr6 (9, 22.5%) and then both Chr3 and Chr8 (5 each, 12.5%). Only 1 BpGRASs (2.5%) was located on the Chr1, 2, 5, 10, 12, 13 and 14. We speculated that there was no obvious connection and correlation between GRASs’ number and chromosome length according to the previous research (Chen et al., 2019; Liu et al., 2019). A tandem duplication event of genes was defined that a chromosomal region within 200 kb contained 2 or more genes, and plays a vital role for gene family in occurrence further expansion of novel functions (Fan et al., 2021b). Six tandem duplication events were found on the Ch6 and 11 including BpGRAS20/BpGRAS21, BpGRAS21/BpGRAS22, BpGRAS22/BpGRAS23, BpGRAS27/BpGRAS28, BpGRAS38/BpGRAS39, BpGRAS39/BpGRAS40, involving total 9 BpGRASs. All the genes involved in tandem duplication events belonged to the same subfamily. Except BpGRAS38, BpGRAS39 and BpGRAS40 from PAT subfamily, 6 genes in tandem duplication events belonged to LISCL subfamily, indicating that LISCL group played an important role in expansion of GRASs as the largest subfamily (Fan et al., 2021a).

Based on the latest genome assemblies, we found 265 putative GRAS genes: 40 in birch, 32 in Arabidopsis, 38 in rice, 52 in tea plant, 59 in P. dactylifera and 44 in T. cacao, respectively. An unrooted phylogenetic tree was constructed using MEGA X using the NJ method with a bootstrap value of 100 for the identification of the evolutionary relationships among the 40 BpGRAS proteins, 32 GRAS proteins of Arabidopsis, 38 of rice, 52 of tea plant, 59 of P. dactylifera and 44 of T. cacao ( Figure 2 ). Phylogenetic analysis showed that these 265 GRAS proteins could be divided into 17 groups (LISCL, SCL3, RAM1, RAD1, DELLA, SCLA, SCLB, DLT, SCR, SCL4/7, LS, NSP1, NSP2, HAM, SHR, SCL32 and PAT). These findings revealed the basic role of GRAS family proteins in the evolution and development of different plant species and were similar to those of previous reports of some other plant species, including Vitis vinifera, Musa acuminata, Coffea canephora and so on (Cenci and Rouard, 2017). BpGRAS proteins were distributed in 17 subfamilies unevenly, and most of them belonged to LISCL subfamily (9 members). Only 1 BpGRAS protein could be observed in subfamilies RAM1, RAD1, SCLA, DLT, SCLB, SCL4/7, LS and NSP1, respectively. However, no BpGRAS protein belonged to SCL3 group. The phylogenetic tree showed that some BpGRAS proteins were closely related to those of other species (bootstrap support ≥ 80), indicating that these BpGRAS proteins might be orthologous to the GRAS proteins of other plants and have similar functions.

To further explore the sequence features of GRAS TFs in birch plants, we performed a comparative analysis of the conserved motifs between birch and Arabidopsis ( Figure 3 ). The structural details of the GRAS proteins were analyzed via 20 motifs predicted by the MEME program. In general, similar motif compositions could occur among GRAS proteins of the same subfamily, suggesting that GRAS proteins in the same subfamily may have similar functions. Almost all GRAS proteins contained motifs 1, 3, 5, 6 and 8, indicating that these motifs were highly conserved and may play important roles in the GRAS family. Motifs 14 and 18 were only distributed in DELLA subfamily; motifs 13, 19 and 20 were only found in LISCL subfamily; motif 7 was only distributed in SCL4/7, LISCL and PAT subfamilies; motif 16 was absent in DELLA, RAM1, DLT, NSP2, HAM, SCL32, and SCLB subfamilies. Motif 3 was found in all proteins except BpGRAS6; BpGRAS6 contained only motif 12, and BpGRAS8 contained only motifs 1, 3, 8 and 9. Some motifs were distributed only at certain locations in the pattern. For example, motif 5 was always distributed at the end of the pattern, and motif 9 was almost always distributed at the start. The functions of most of these conserved motifs still need to be understood. From the differences in the distribution of these motifs between subfamilies, it can be seen that GRAS proteins of different subfamilies may have different functions; meanwhile, different genes from the same subfamily also exhibited a different distribution of motifs, indicating that the functions of such genes may also be different. In specific GRAS protein subfamilies, similar motifs tended to be clustered together, indicating that there might be functional similarities among those proteins.

Based on the latest birch genome assembly results, 26 BpGRASs were successfully cloned for further gene function studies. Thus, different tissues (from the root, stem, and leaf) were collected after treatment with 200 mM NaCl for 3, 6, 12, 24, or 48 h for RT-qPCR, to analyze the expression patterns of the 26 BpGRASs ( Figures 4 and 5 ). The results showed that all genes were expressed in the root, stem, and leaf tissues at each time point, which indicated that the GRAS genes might play a role in plant growth and development. Most of the BpGRASs (17 genes) were highly induced by salt stress at 6 h and were significantly expressed in the leaf tissues except BpGRAS26 in the stem. BpGRAS13 was significantly induced in the leaf at 6 and 12 h after salt treatment ( Figures 4 and 5 ). Only the expression level of BpGRAS37 peaked under salt stress conditions at 48 h in the leaf ( Figures 4 and 5 ). In the root tissue, BpGRAS20 and BpGRAS36 were expressed significantly at 3 and 48 h, respectively. On the other hand, BpGRAS11 was also significantly expressed at 6 h in the leaf and stem tissues, and BpGRAS30 was highly induced by salt stress at 6 h in the leaf and 24 h in the root. These results indicated that high levels of most of the BpGRASs were induced in the leaves in response to salt stress at 6 h.

Six BpGRASs, i.e., BpGRAS1, BpGRAS16, BpGRAS19 (GenBank accessions: MN117546-MN117548), BpGRAS26, BpGRAS34, and BpGRAS40 (GenBank accessions: MZ062900-MZ062902), were selected for further study because they were significantly upregulated under salt treatment conditions and had better expression patterns as shown in Figure 3 . These 6 BpGRAS proteins, which exhibited a high level of homology to GRAS proteins of Arabidopsis and rice, were selected for multiple sequence alignment analysis. The results indicated that the GRAS proteins of birch and other plant species shared a highly conserved binding domain at the C-terminus ( Supplementary Figure 1 ), and the six selected BpGRAS proteins contained certain GRAS domains that are characteristically found in the GRAS family (Pysh et al., 1999).

The distribution of cis-acting elements in promoters may be responsible for the diversity of functions and expression patterns of different genes. Cis-acting elements were identified from the 2-kb region upstream of the start codon in the promoters of 6 selected GRAS genes (BpGRAS1, BpGRAS16, BpGRAS19, BpGRAS26, BpGRAS34 and BpGRAS40), for further identifying their role in the development of tolerance to salt-shock-induced stress ( Figure 6 ).

Six cis-acting elements were analyzed and found to be involved in response to abiotic stress or phytohormone conduction; these included 6 stress-response elements and 10 phytohormone-related elements. All 6 genes contained 7 to 11 cis-acting elements as shown in Figure 6 . Each of these 6 GRAS genes had at least one element related to the stress response, such as TC-rich repeats, GT-1-box, HSE, LTR, and LTRE, and played a role in generating a stress response. MBS had a drought inducibility-related function because it acted as the binding site of MYB. Meanwhile, 10 phytohormone-related elements of the 6 GRAS genes occurred in most plant hormones; these included the elements that played a role in the abscisic acid-responsiveness (ABRE), auxin-responsiveness (TGA-element, AuxRR-core), gibberellin-responsiveness (TATC-box, GARE-motif), jasmonic acid-responsiveness (CGTCA-motif), and salicylic acid-responsiveness (TCA-element). These results suggested that the 6 BpGRASs might confer tolerance to abiotic stresses such as salt, cold, and drought, and participate in the plant growth and development process.

Six BpGRASs were used for constructing BpGRAS overexpressing and inhibiting recombinant vectors, and transiently transformed birch plants were collected for RT-qPCR via high-efficiency transient transformation (Ji et al., 2014). The results showed that birch plants exhibiting transient overexpression or the knockdown of the 6 BpGRASs had been obtained successfully, with highest expression levels of all the OE plants among variety of transiently transformed plants induced by salt treatment ( Supplementary Figure 2 ).

Physiological determinations of 6 BpGRASs were performed for further identifying whether 6 BpGRASs conferred tolerance to salt stress. Cell death is often measured to detect stress tolerance in plants. Evans blue staining was performed to study cell death after salt stress treatment (Zhang et al., 2011). Under normal growth conditions, three types of plants, i.e., plants exhibiting the overexpression and inhibition of 6 transiently transformed BpGRASs, and control plants (Control) showed a consistent level of staining. Under salt stress, OE plants were stained more lightly than control and IE plants, and the staining intensity of IE plants was the highest ( Figure 7A ). Under salt stress, the electrolyte leakages of IE plants of these 6 BpGRASs were higher than that of control plants, while the OE plants had the lowest electrolyte leakages ( Figure 7B ). An assessment of the MDA contents showed that there were no significant differences in OE, IE and control plants of the three transient transgenic plants under normal growth conditions. However, the MDA content of OE plants was the lowest, compared to the control plants after salt stress treatment ( Figure 7C ). These results showed that overexpression of BpGRASs resulted in minimal levels of cell death, indicating that overexpression of BpGRASs resulted in better salt stress tolerance in birch plants.

To study whether overexpression of BpGRASs could improve the salt tolerance of birch, BpGRAS34 was randomly selected to obtain stable transgenic overexpression plant. We obtained 11 transgenic lines of BpGRAS34 overexpression and detected their expression levels via RT-qPCR. Lines BpGRAS34-5 and BpGRAS34-7 (OE34-5 and OE34-7) were high expressed compared to the other lines ( Supplementary Figure 3 ) and selected for next measurement. Phenotype analysis could intuitively show degree of injury of plants under stress conditions. Under the normal condition, there were not substantially different phenotypes of control plants and OE34-5 and OE34-7 plants, suggesting that BpGRAS34 could not affect growth and phenotype of birch. However, leaves of control plant got wilting while OE34-5 and OE34-7 plants remained alive and greener after salt treatment ( Figure 8A ). Evens blue staining was used to investigate cell death under salt stress. Compared with the control plant, OE lines reduced staining after salt stress, indicating lower cell death ( Figure 8B ). As well as the results of electrolyte leakages and MDA contents, OE34-5 and OE34-7 plants were lower than control plants under salt stress treatment; while there was no significant difference between the control plants and overexpression plants ( Figures 8C, D ). These results suggested that OE34 reduced cell death under salt treatment.

ROS plays an important role in the evaluation of plant stress tolerance (Gechev et al., 2006). NBT and DAB staining were used to determine the level of ROS accumulation via the detection of O 2 − · and H₂O₂ — the two main components of ROS. NBT and DAB staining, H₂O₂ content, SOD and POD activities and ROS content were evaluated to view if OE34 can improve ROS scavenging. No obvious difference in NBT and DAB staining was observed among OE34 and control lines under control conditions. However, compared with control plants under salt treatment condition, the results of histochemical staining of birch tissues using NBT and DAB showed that the levels of both O 2 − · and H₂O₂ in OE plants were lower than those in control plants ( Figure 8B ). The results of these analyses were consistent with results indicating H₂O₂ content and ROS accumulation levels. There was no obvious difference between OE and control lines in the measurement of H₂O₂ content. OE34-5 and OE34-7 had lower H₂O₂ content than control plants after salt treatment ( Figure 8E ), indicating that OE34 reduced H₂O₂ accumulations in birch under salt stress condition. At the same time, transiently transformed plants of 6 BpGRASs also showed better ability of decreasing the H₂O₂ accumulations in the Supplementary Figures 4A, B . Furthermore, SOD and POD are two major ROS scavenging enzymes whose activities have extensively been used as an indicator of stress tolerance in plants (Zang et al., 2015). Under normal growth conditions, the activities of SOD and POD in OE34 plants were not different from those in control plants. However, the activities of SOD and POD in OE34 plants were significantly higher than those in the control plants under salt stress ( Figures 8F, G ). Additionally, OE34-5 and OE34-7 resulted in the lower ROS levels compared to those observed in the control plants ( Figure 8H ). Similarly, these analyses were repeated for transiently transformed plants of 6 BpGRASs and the results were consistent ( Supplementary Figure 4C and Supplementary Figure 5A, B ). These outcomes indicated that a decrease in ROS accumulation was attributable to the overexpression of BpGRASs in birch plants, thereby enhancing SOD and POD activity under salt stress.

In addition, we compared the proline levels in OE34-5, OE34-7 and WT plants, to investigate whether BpGRASs can regulate proline biosynthesis under salt stress conditions ( Figure 8I ). Under normal growth conditions, the proline levels in OE34-5, OE34-7 and control plants were almost the same; however, the proline content in OE34 plants subjected to salt stress treatment was significantly higher than that of the control ( Figure 8I ). When exposed to salt condition, the similar results were found overexpression of 6 BpGRASs in transiently transformed plants, which could increase proline content ( Supplementary Figure 5C ). Thus, the overexpression of BpGRASs can increase the proline content in birch plants under salt stress conditions. The above-mentioned results preliminarily showed that overexpression of BpGRAS34 can improve tolerance to salt by decreasing cell death, enhancing ROS scavenging ability and increasing proline content, further proved that overexpression of BpGRASs can enhance salt tolerance of birch, consistent with cis-acting analysis ( Figure 6 ).