The Identification of Broomcorn Millet bZIP Transcription Factors, Which Regulate Growth and Development to Enhance Stress Tolerance and Seed Germination

Broomcorn millet (Panicum miliaceum L.) is a water-efficient and highly salt-tolerant plant. In this study, the salt tolerance of 17 local species of broomcorn millet was evaluated through testing based on the analysis of the whitening time and the germination rate of their seeds. Transcriptome sequencing revealed that PmbZIP131, PmbZIP125, PmbZIP33, PmABI5, PmbZIP118, and PmbZIP97 are involved in seed germination under salt stress. Seedling stage expression analysis indicates that PmABI5 expression was induced by treatments of high salt (200 mM NaCl), drought (20% W/V PEG6000), and low temperature (4 °C) in seedlings of the salt-tolerant variety Y9. The overexpression of PmABI5 significantly increases the germination rate and root traits of Arabidopsis thaliana transgenic lines, with root growth and grain traits significantly enhanced compared to the wild type (Nipponbare). BiFC showed that PmABI5 undergoes homologous dimerization in addition to forming a heterodimer with either PmbZIP33 or PmbZIP131. Further yeast one-hybrid experiments showed that PmABI5 and PmbZIP131 regulate the expression of PmNAC1 by binding to the G-box in the promoter. These results indicate that PmABI5 can directly regulate seed germination and seedling growth and indirectly improve the salt tolerance of plants by regulating the expression of the PmNAC1 gene through the formation of heterodimers with PmbZIP131.


Introduction
Cereals are the most important sources of carbohydrate acquisition and are the most widely planted crops in global agriculture. On a worldwide basis, the main crops grown in Asia are maize (Zea mays L.), wheat (Triticum aestivum L.), broomcorn millet (Panicum miliaceum L.), and sorghum (Sorghum bicolor L.); rice (Oryza sativa L.), barley (Hordeum vulgare L.), and foxtail millet (Setaria italica L.) are the most popular crops produced in the Americas, while rye (Secale cereale L.) and rice are heavily cultivated in Europe [1]. In 2018, United States cereal production was 8.69 tons per hectare, while that of China was only 6.08 tons per hectare [2]. In a situation of global climate extremes, land degradation, and decreasing amounts of cultivated land, adverse factors can cause a range of morphological, physiological, and biochemical damage to plants, resulting in serious losses to agricultural production and economic yields [3]. Therefore, improving the tolerance of crops to unsuitable ecological conditions (i.e., abiotic stress and biotic stress) is a persistent goal of plant breeders and researchers [4].
Abiotic stresses represent abiotic environmental conditions that are harmful to plant survival, growth, and development, even leading to injury, damage, and death. The main adverse environmental conditions that plants often encounter are extreme temperatures, drought, salinity, and heavy metal toxicity [5]. Drought stress can occur at any stage of crop Note: CK is seed germination in RO water, NaCl is seed germination in 250 mmol/L. On Day 1 and Day 2, seeds were pre-chilled (4 • C). On the basis of the sequencing scheme set to identify varieties demonstrating tolerance to extremes of the two treatments in three periods, it was found, after preliminary experiments, that there were three nodes in the process of seed germination, with the following periods: 4 • C culture results in total seed inhibition; 25 • C 3 h after germination, expression of early salt-stress-related genes begins; 25 • C when cultivating 6 to 24 h after seed germination, a large number of seed-germination-and salt-stress-related genes are expressed. Considering that the germination rate of Xingmi 1 is too low to allow reaching the sample quantity required for RNA-Seq analyses, Yumi 9 and Yumi 1 were confirmed as the sample varieties.

Transcriptome Sequencing and Assembly
To fully characterize the transcriptome and gene expression levels of P. miliaceum under water (CK) and salt stress (NA) in the three periods, 36 cDNA samples of broomcorn millet seeds were prepared and sequenced using the Illumina HiSeq TM platform. The RNA-Seq raw data are summarized in Table 2, whereby the results are the averages of three replicates. After removing low-quality and adapter-related reads, 317.55 Gb clean reads were obtained, with an average of 8.82 Gb (29.5 million) reads for each sample, and the percentage of Q30 base in each sample was not less than 85.06%. The clean data of each sample were compared with the reference genome. The ratio of mapped reads to the reference genome ranged from 72.93% to 84.57%, and unique mapped reads ranged from 71.51% to 82.51%.

Gene Annotation and Functional Classification
All gene sequences were queried to screen databases such as COG, GO, KEGG, KOG, SwissProt, Pfam, eggNOG, and Nr, after which the genes were annotated with the corresponding information identified. On the basis of the sequencing results (Table 3) There were about 60,331 genes annotated in total, of which 24,955 genes were longer than 300 bp and shorter than 1000 bp, and 32,893 genes were longer than 1000 bp. Gene Ontology (GO) annotation was used to demonstrate the possible functional classifications of broomcorn millet genes. Successfully annotated and classified genes were distributed into 54 functional groups in the categories of cell component (CC), molecular function (MF), and biological process (BP) (Figure 1a). The biological process category covered 22,969 genes and was successfully identified in 20 subgroups, where most annotations were in "metabolic process" (5164, 22.5%), "cellular process" (4369, 19.0%), "singleorganism process" (4070, 17.7%), and "response to stimulus" (2268, 9.9%); 24,226 genes were identified in 17 subgroups for the cell component category, where the top four GO subgroups were "cell part" (6249, 25.8%), "cell" (6237, 25.7%), "organelle" (5550, 22.9%), and "membrane" (2646, 10.9%); and 10,322 genes were annotated in 17 subgroups for the molecular function category, where the top three GO terms were "catalytic activity" (4457, 43.2%), "binding" (4225, 40.9%), and "transporter activity" (532, 5.2%). KEGG annotation showed that most of the genes were in the metabolism category (Figure 1b), and many genes were in the environmental information processing category, which is mainly enriched in signal transduction genes (including those participating in plant hormone signal transduction and the phosphatidylinositol signaling system). According to the KOG annotation, 28,998 genes were grouped into 25 functional classifications (Figure 1c). Around one-fifth of genes were classified as "general function prediction only" (6391, 19.59%), and fewer as "function unknown" (1655, 5.07%). Aside from this, the top three groups were "posttranslational modification, protein turnover, chaperones" (3572, 10.95%), "signal transduction mechanisms" (2939, 9.01%), and "transcription" (1894, 5.81%). The nr database had identified the most annotated genes of the databases used in searching ( Figure 1d). According to the results, P. miliaceum showed the most matches to Setaria italica (41,276, 68.83%), followed by Sorghum bicolor (7199, 12.01%), and Zea mays (5196, 8.67%).

Statistics of Differential Expression Genes (DEGs) in Salt Stress
The twelve groups of differential expression genes (DEGs) are summarized in Table 4, where it can be seen that a large number of DEGs were identified in the groups of Y9_CK3 vs. Y9_NA3 and Y1_CK3 vs. Y1_NA3, including 2860 upregulated and 1456 downregulated genes and 1447 upregulated and 1604 downregulated genes, respectively. In Y9_CK2 vs. Y9_NA2, 1898 DEGs (613 up-and 1285 downregulated) were identified, compared with Y1_CK2 vs. Y1_NA2 (97 up-and 328 downregulated), showing that the response to salt stress in Yumi 9 was earlier. Further research revealed that the number of genes found to be differentially expressed between Yumi 1 and Yumi 9 was relatively stable in the sampling stages 2 and 3, and the numbers of DEGS were 2442 (Y1_CK2 vs. Y9_CK2) and 1968 (Y1_CK3 vs. Y9_CK3), respectively. However, the number of DEGs decreased to 1406 (Y1_NA2 vs. Y9_NA2) and 1338 (Y1_NA3 vs. Y9_NA3) after NaCl treatment.  , and the darker bars to differentially expressed genes (DEGs, blue). The genes were categorized into three main types: "biological process", "cellular component", and "molecular function". (b) KEGG annotation showed that most of the genes were in the metabolism category. The genes were grouped into "cellular process", "genetic information processing", "organismal systems", "metabolism", and "environmental information processing". (c) KOG annotation grouped all genes into 25 functional classifications. The number of genes annotated for this function and the percentage of total genes are shown in square brackets. (d) nr annotation showed that most of the sequenced broomcorn millet genes are highly similar to corresponding genes in Setaria italica and Sorghum bicolor.

Statistics of Differential Expression Genes (DEGs) in Salt Stress
The twelve groups of differential expression genes (DEGs) are summarized in Table  4, where it can be seen that a large number of DEGs were identified in the groups of Y9_CK3 vs. Y9_NA3 and Y1_CK3 vs. Y1_NA3, including 2860 upregulated and 1456 downregulated genes and 1447 upregulated and 1604 downregulated genes, respectively. In Y9_CK2 vs. Y9_NA2, 1898 DEGs (613 up-and 1285 downregulated) were identified, compared with Y1_CK2 vs. Y1_NA2 (97 up-and 328 downregulated), showing that the response to salt stress in Yumi 9 was earlier. Further research revealed that the number of Figure 1. GO, KOG, NR, and KEGG annotation. (a) GO annotation to demonstrate the possible functional classifications of broomcorn millet genes. The lighter bars correspond to all genes (black), and the darker bars to differentially expressed genes (DEGs, blue). The genes were categorized into three main types: "biological process", "cellular component", and "molecular function". (b) KEGG annotation showed that most of the genes were in the metabolism category. The genes were grouped into "cellular process", "genetic information processing", "organismal systems", "metabolism", and "environmental information processing". (c) KOG annotation grouped all genes into 25 functional classifications. The number of genes annotated for this function and the percentage of total genes are shown in square brackets. (d) nr annotation showed that most of the sequenced broomcorn millet genes are highly similar to corresponding genes in Setaria italica and Sorghum bicolor. Moreover, the Venn diagram in Figure 2a displays unique and overlapping sets of DEGs under salt treatment. There were 123 DEGs (I, 28 up-and 95 downregulated, Figure 2b) expressed in all comparisons, demonstrating that these genes were shared and function continuously in response to salt stress. A majority were expressed in the third stage of sampling under salt stress, 1267 DEGs were marked as overlapping (V, 328 up-and 939 downregulated), indicating that these genes showed a relatively late response to salt. There were 235 DEGs (IV, 137 up-and 98 downregulated) in Yumi 9 that specifically responded to salt stress; however, there were only 11 DEGs in Yumi 1. We also identified 699 (V, 275 up-and 424 downregulated), 854 (VII, 265 up-and 589 downregulated) and 2124 (VIII, 1541 up-and 583 downregulated) unique genes differentially expressed only in Y1_CK3 vs. Y1_NA3, Y9_CK2 vs. Y9_NA2 and Y9_CK3 vs. Y9_NA3, respectively.
The Venn diagram in Figure 2c presents unique sets of DEGs in two extreme-tolerant plant varieties, suggesting that these genes may be responsible for the variety differences in terms of salt tolerance. The two groups with the highest numbers of DEGs were treated with RO water; the 1211 DEGs (476 up-and 735 downregulated, Figure 2d) were uniquely expressed in the second stage of sampling (Y1_CK2 vs. Y9_CK2, ii). In the third stage of sampling, they were treated using RO water (Y1_CK2 vs. Y9_CK2, ii), after which 921 DEGs (297 up-and 624 downregulated) were uniquely expressed. However, under salt stress, the number of differentially expressed genes was the highest in the third stage (Y1_NA3 vs. Y9_NA3, vi), where there were 351 upregulated and 204 downregulated DEGs. The number of differentially expressed genes was the lowest in the first stage (Y1_NA1 vs. Y9_NA1, iv), where there were 241 upregulated and 88 downregulated DEGs.

GO Term and KEGG Enrichment Analysis for DEGs
Gene enrichment analysis uses predefined sets of genes and sequences to identify significant biological changes or patterns of gene co-expression, and it can assess the functional relevance of target sets of genes derived from a set of experiments [31]. The most enriched GO terms were classified into 20 functional groups in the following categories: cell component (CC, 4 groups), molecular function (MF, 7 groups), and biological processes (BP, 9 groups) ( Figure 3a). Among them, DEGs were mostly enriched in the oxidation-reduction process (742, belonging to biological process), heme binding (295, belonging to molecular function), and electron carrier activity (280, belonging to molecular function). The KEGG analysis results showed that DEGs were mostly enriched in phenylpropanoid biosynthesis (162, ko00940), plant hormone signal transduction (142, ko04075), carbon metabolism (141, ko01200), starch and sucrose metabolism (127, ko00500), and phenylalanine metabolism (126, ko00360) ( Figure 3b).

Expression Analysis of TFs Involved in Seed Germination under Salt Stress
Seed germination marks the beginning of a new growth cycle of higher plants, and it is affected by complex regulatory mechanisms of internal and environmental signals [32]. In eukaryotes, the synergy between transcription factors and the core transcriptase RNA polymerase initiates gene expression [33]. We comprehensively analyzed the transcription level changes of different expression TFs in seed germination under salt stress ( Figure 4). The heatmap revealed that a large number of bZIP transcription factors (36 PmbZIPs) exhibit differences in expression during this period, which means that they are involved in the germination of seeds under salt stress. In addition, the transcription factors (TFs) of other families mainly include CBF transcription factor (members of the AP2/ERF family), TCP transcription factor, MYB transcription factor, and PLATZ transcription factor (plant AT-rich protein and zinc-binding protein). It is worth noting that EVM0016128 PmABI5), EVM0059197 (PmbZIP118, differential expression, but very low), EVM0045095 (PmbZIP97), and EVM0056648 (PmbZIP89) were differentially expressed. Of these 7 PmbZIP transcription factors, 5 PmbZIPs (PmbZIP131, PmbZIP125,

Expression Analysis of TFs Involved in Seed Germination under Salt Stress
Seed germination marks the beginning of a new growth cycle of higher plants, and it is affected by complex regulatory mechanisms of internal and environmental signals [32]. In eukaryotes, the synergy between transcription factors and the core transcriptase RNA polymerase initiates gene expression [33]. We comprehensively analyzed the transcription level changes of different expression TFs in seed germination under salt stress ( Figure 4). The heatmap revealed that a large number of bZIP transcription factors (36 PmbZIPs) exhibit differences in expression during this period, which means that they are involved in the germination of seeds under salt stress. In addition, the transcription factors (TFs) of other families mainly include CBF transcription factor (members of the AP2/ERF family), TCP transcription factor, MYB transcription factor, and PLATZ transcription factor (plant AT-rich protein and zinc-binding protein). It is worth noting that EVM0016128 (PmbZIP131, PmABF2), EVM0022428 (PmbZIP125), EVM0000147 (PmbZIP33, PmABF4), EVM0038393 (PmbZIP30, PmABI5), EVM0059197 (PmbZIP118, differential expression, but very low), EVM0045095 (PmbZIP97), and EVM0056648 (PmbZIP89) were differentially expressed. Of these 7 PmbZIP transcription factors, 5 PmbZIPs (PmbZIP131, PmbZIP125, PmbZIP33, PmABI5, PmbZIP118) were upregulated by salt treatment and 2 PmbZIPs (PmbZIP89, Pm-bZIP97) were downregulated. PmbZIP33, PmABI5, PmbZIP118) were upregulated by salt treatment and 2 PmbZIPs (PmbZIP89, PmbZIP97) were downregulated.

PmbZIPs Response to Various Stress Conditions in the Plant Seedling Stage
To explore the response of seven PmbZIP transcription factors to different stresses, we analyzed the expression patterns using quantitative real-time reverse transcription (qRT)-PCR. The experimental results showed that PmABI5 could be significantly induced by salt, drought, and low temperature in Y9 millet after treatment for 48 h (salt-tolerant) (Figure 5f). Beyond this, expression of PmABI5 was also rapidly induced by ABA in Y1 millet when treated for 6 h (salt-sensitive), indicating that the two varieties possess different patterns of gene expression corresponding to differences in regulation (Figure 5e). Meanwhile, PmbZIP131 was induced by ABA, low temperature, and high temperature in Y1 (salt-sensitive), while there were no obvious expression changes in Y9 (Figure 5b,e). This result shows that the gene has different expression patterns in the two selected Heatmap of TFs were clustered by columns. Y9, Y1 represent Yumi 9 and Yumi 1, respectively. CK represents control, NA represents salt stress (250 mM NaCl). 1, 2, 3 represent the three periods. Y9_CK1 refers to Yumi 9 germination under RO water in the first period. The treatments and varieties are similarly indicated in the remaining labels. The information on the righthand side of the heatmap corresponds to the gene ID and annotation.

PmbZIPs Response to Various Stress Conditions in the Plant Seedling Stage
To explore the response of seven PmbZIP transcription factors to different stresses, we analyzed the expression patterns using quantitative real-time reverse transcription (qRT)-PCR. The experimental results showed that PmABI5 could be significantly induced by salt, drought, and low temperature in Y9 millet after treatment for 48 h (salt-tolerant) (Figure 5f). Beyond this, expression of PmABI5 was also rapidly induced by ABA in Y1 millet when treated for 6 h (salt-sensitive), indicating that the two varieties possess different patterns of gene expression corresponding to differences in regulation (Figure 5e). Meanwhile, Pm-bZIP131 was induced by ABA, low temperature, and high temperature in Y1 (salt-sensitive), while there were no obvious expression changes in Y9 (Figure 5b,e). This result shows that the gene has different expression patterns in the two selected varieties. PmbZIP33, Pm-bZIP125, and PmbZIP118 exhibited no expression change under most treatments ( Figure S1). Although PmbZIP33 showed significant differences after treatment, the expression level was relatively low (Figure 5a). Two U-subfamily genes (PmbZIP97 and PmbZIP89) were strongly induced by ABA, salt, low temperature, high temperature, and drought stress in Y1 millet ( Figure S1). These results are consistent with the previous germination stage by RNA-Seq, indicating that PmbZIP33, PmbZIP131, and PmABI5 are upregulated under salt stress. However, the fact that PmbZIP89 and PmbZIP97 were upregulated in the seedling stage may be due to inhibition of regulation at a specific time. varieties. PmbZIP33, PmbZIP125, and PmbZIP118 exhibited no expression change under most treatments ( Figure S1). Although PmbZIP33 showed significant differences after treatment, the expression level was relatively low (Figure 5a). Two U-subfamily genes (PmbZIP97 and PmbZIP89) were strongly induced by ABA, salt, low temperature, high temperature, and drought stress in Y1 millet ( Figure S1). These results are consistent with the previous germination stage by RNA-Seq, indicating that PmbZIP33, PmbZIP131, and PmABI5 are upregulated under salt stress. However, the fact that PmbZIP89 and PmbZIP97 were upregulated in the seedling stage may be due to inhibition of regulation at a specific time.

PmbZIPs Subcellular Localization and BiFC Assay
The PmbZIP proteins were fused with enhanced green fluorescent protein (eGFP) at the N-terminus and transiently expressed in N. benthamiana leaf via Agrobacterium-mediated transformation, and the eGFP signal showed exclusive nuclear localization in all cases except for PmABI5 ( Figure 6). Fluorescent signals corresponding to PmABI5:eGFP transiently expressed in N. benthamiana were found in both the nucleus and the endoplasmic reticulum.

PmbZIPs Subcellular Localization and BiFC Assay
The PmbZIP proteins were fused with enhanced green fluorescent protein (eGFP) at the N-terminus and transiently expressed in N. benthamiana leaf via Agrobacterium-mediated transformation, and the eGFP signal showed exclusive nuclear localization in all cases except for PmABI5 ( Figure 6). Fluorescent signals corresponding to PmABI5:eGFP transiently expressed in N. benthamiana were found in both the nucleus and the endoplasmic reticulum. To further confirm the possibility of a direct interaction, we performed bimolecular fluorescence complementation (BiFC) assays. YFP-N-PmABI5 and YFP-C-PmbZIP33 were co-transformed into Nicotiana benthamiana leaves, a strong yellow fluorescent signal was observed in the nucleus (Figure 7). YFP-N-PmABI5 and YFP-C-PmbZIP131, as well as YFP-N-PmABI5 and YFP-C-PmABI5, had the same phenotype. These results indicate that PmABI5 directly interacts with PmbZIP33, PmbZIP131, and PmABI5 in vivo, forming homodimers or heterodimers in order to function.

PmbABI5 Positively Regulates Root Growth in Arabidopsis and Rice
To investigate the possible role of PmABI5 in salt stress during germination and root elongation, we analyzed the germination rate and root growth in Arabidopsis PmABI5 overexpression lines (OE: PmABI5-3-1; OE: PmABI5-4-6; and OE: PmABI5-7-4) cultured under 0 and 100 mM NaCl. Under salt-free conditions, the germination rates of PmABI5 overexpression lines and the wild type were the same ( Figure S2), but root length, root area, and lateral root number were significantly higher than those of the wild type ( Figure  8). Moreover, the germination rate and root development of PmABI5 overexpressing plants were also significantly higher than those of wild type on medium containing 100 mM NaCl ( Figure S2). To further confirm the possibility of a direct interaction, we performed bimolecular fluorescence complementation (BiFC) assays. YFP-N-PmABI5 and YFP-C-PmbZIP33 were co-transformed into Nicotiana benthamiana leaves, a strong yellow fluorescent signal was observed in the nucleus (Figure 7). YFP-N-PmABI5 and YFP-C-PmbZIP131, as well as YFP-N-PmABI5 and YFP-C-PmABI5, had the same phenotype. These results indicate that PmABI5 directly interacts with PmbZIP33, PmbZIP131, and PmABI5 in vivo, forming homodimers or heterodimers in order to function. To further confirm the possibility of a direct interaction, we performed bimolecular fluorescence complementation (BiFC) assays. YFP-N-PmABI5 and YFP-C-PmbZIP33 were co-transformed into Nicotiana benthamiana leaves, a strong yellow fluorescent signal was observed in the nucleus (Figure 7). YFP-N-PmABI5 and YFP-C-PmbZIP131, as well as YFP-N-PmABI5 and YFP-C-PmABI5, had the same phenotype. These results indicate that PmABI5 directly interacts with PmbZIP33, PmbZIP131, and PmABI5 in vivo, forming homodimers or heterodimers in order to function.

PmbABI5 Positively Regulates Root Growth in Arabidopsis and Rice
To investigate the possible role of PmABI5 in salt stress during germination and root elongation, we analyzed the germination rate and root growth in Arabidopsis PmABI5 overexpression lines (OE: PmABI5-3-1; OE: PmABI5-4-6; and OE: PmABI5-7-4) cultured under 0 and 100 mM NaCl. Under salt-free conditions, the germination rates of PmABI5 overexpression lines and the wild type were the same ( Figure S2), but root length, root area, and lateral root number were significantly higher than those of the wild type ( Figure  8). Moreover, the germination rate and root development of PmABI5 overexpressing plants were also significantly higher than those of wild type on medium containing 100 mM NaCl ( Figure S2).

PmbABI5 Positively Regulates Root Growth in Arabidopsis and Rice
To investigate the possible role of PmABI5 in salt stress during germination and root elongation, we analyzed the germination rate and root growth in Arabidopsis PmABI5 overexpression lines (OE: PmABI5-3-1; OE: PmABI5-4-6; and OE: PmABI5-7-4) cultured under 0 and 100 mM NaCl. Under salt-free conditions, the germination rates of PmABI5 overexpression lines and the wild type were the same ( Figure S2), but root length, root area, and lateral root number were significantly higher than those of the wild type ( Figure 8). Moreover, the germination rate and root development of PmABI5 overexpressing plants were also significantly higher than those of wild type on medium containing 100 mM NaCl ( Figure S2). It is worth noting that leaves of the wild type under salt stress had already started to show signs of stress (anthocyanin accumulation), while the overexpression lines still did not exhibit the stress phenotype at 8 days old (Figure 9a). For further confirmation of our hypothesis, the 15-day-old PmABI5-overexpressing lines and wild-type seedlings were transferred to soil (Figure 9b). After a further 4 weeks of growth, PmABI5 overexpression lines had the same number of rosette leaves as the wild type, but with significantly larger leaves (Figure 9c,d). Figure 10e shows 7-day-old seedlings of rice, and PmABI5-overexpressing rice lines have the same phenotype as the corresponding Arabidopsis line in terms of root development. The seed size of the transgenic strain after harvest was also larger than that of Nipponbare ( Figure S3).  It is worth noting that leaves of the wild type under salt stress had already started to show signs of stress (anthocyanin accumulation), while the overexpression lines still did not exhibit the stress phenotype at 8 days old (Figure 9a). For further confirmation of our hypothesis, the 15-day-old PmABI5-overexpressing lines and wild-type seedlings were transferred to soil (Figure 9b). After a further 4 weeks of growth, PmABI5 overexpression lines had the same number of rosette leaves as the wild type, but with significantly larger leaves (Figure 9c,d). Figure 10e shows 7-day-old seedlings of rice, and PmABI5overexpressing rice lines have the same phenotype as the corresponding Arabidopsis line in terms of root development. The seed size of the transgenic strain after harvest was also larger than that of Nipponbare ( Figure S3). It is worth noting that leaves of the wild type under salt stress had already started to show signs of stress (anthocyanin accumulation), while the overexpression lines still did not exhibit the stress phenotype at 8 days old (Figure 9a). For further confirmation of our hypothesis, the 15-day-old PmABI5-overexpressing lines and wild-type seedlings were transferred to soil (Figure 9b). After a further 4 weeks of growth, PmABI5 overexpression lines had the same number of rosette leaves as the wild type, but with significantly larger leaves (Figure 9c,d). Figure 10e shows 7-day-old seedlings of rice, and PmABI5-overexpressing rice lines have the same phenotype as the corresponding Arabidopsis line in terms of root development. The seed size of the transgenic strain after harvest was also larger than that of Nipponbare ( Figure S3).

PmbABI5 Can Regulate PmNAC1 Expression
To delineate the exact sequence of the G-box motifs in the PmNAC1 promoter that are bound by PmABI5, three copies of G-box (CACGTG) and its mutant sequence g-box (CATTTG) were cloned into the pAbAi vector (Figure 10a,b). Then, they were transferred to the yeast Y1H strain after linearization to obtain Gbox-AbAi and gbox-AbAi strains: AD (empty pGADT7 vector), AD-PmbZIP131 (full-length PmbZIP131 fused to GAL4 activation domain), and AD-PmABI5 via transformation of each as newly constructed strains. Our results show that both PmABI5 and PmbZIP131 were able to bind to the wild-type Gbox, but not the mutant (in which CACGTG was mutated to CATTTG) (Figure 10c). Furthermore, the strength of PmbZIP, at 131 interactions, was higher than that of PmABI5.

Discussion
PmABI5 is induced by various stresses and is presumably involved in the regulation of multiple stress responses. Several studies have demonstrated that this family of genes can improve plant tolerance to the corresponding stresses by regulating the expression of key resistance genes [34][35][36]. Several studies have highlighted the involvement of ABI5 in seed germination and post-germinative growth, suggesting that it may interact with JA to mediate these processes in Arabidopsis thaliana and rice [37][38][39]. Moreover, previous research has revealed that ABI5 regulates the expression of PYL11 and PYL12 by directly binding to their promoters, and PYL11 and PYL12 positively modulate ABA-mediated seed germination and seedlings [40]. ABI5 protein can heteromerically interact with HY5 to activate its own expression, regulating the adaptation of young seedlings to environmental stresses [41]. However, the detailed molecular mechanisms underlying ABI5 regulation of seed germination and post-germinative growth in salt-tolerant millet remained elusive. In this study, we further investigated the regulatory role of PmABI5 in modulating ABA and salt responses during seed germination and seedling growth. The results show that PmABI5 can be induced by salt, ABA, heat, and drought, heteromerically interacting with PmbZIP131 or homologously interacting with itself to promote seed germination under salt stress. In addition, overexpressing PmABI5 in Arabidopsis thaliana and rice can significantly increase seed germination rate and root development under saline stress. These results are consistent with those of previous studies. Nonetheless, the ABI5-regulated inhibition of seed germination and early seedling growth, to protect the plant against adverse conditions via regulation of development, has also been reported [42]. However, a previous study showed that seedlings germinated and developed normally in the absence of ABA in spite of ABI5 accumulation [43]. Therefore, it is reasonable to suggest that PmABI5 may improve stress resistance by enhancing seedling growth and passing through the stress-sensitive period as early as possible.

PmbABI5 Can Regulate PmNAC1 Expression
To delineate the exact sequence of the G-box motifs in the PmNAC1 promoter that are bound by PmABI5, three copies of G-box (CACGTG) and its mutant sequence g-box (CATTTG) were cloned into the pAbAi vector (Figure 10a,b). Then, they were transferred to the yeast Y1H strain after linearization to obtain Gbox-AbAi and gbox-AbAi strains: AD (empty pGADT7 vector), AD-PmbZIP131 (full-length PmbZIP131 fused to GAL4 activation domain), and AD-PmABI5 via transformation of each as newly constructed strains. Our results show that both PmABI5 and PmbZIP131 were able to bind to the wild-type G-box, but not the mutant (in which CACGTG was mutated to CATTTG) (Figure 10c). Furthermore, the strength of PmbZIP, at 131 interactions, was higher than that of PmABI5.

Discussion
PmABI5 is induced by various stresses and is presumably involved in the regulation of multiple stress responses. Several studies have demonstrated that this family of genes can improve plant tolerance to the corresponding stresses by regulating the expression of key resistance genes [34][35][36]. Several studies have highlighted the involvement of ABI5 in seed germination and post-germinative growth, suggesting that it may interact with JA to mediate these processes in Arabidopsis thaliana and rice [37][38][39]. Moreover, previous research has revealed that ABI5 regulates the expression of PYL11 and PYL12 by directly binding to their promoters, and PYL11 and PYL12 positively modulate ABA-mediated seed germination and seedlings [40]. ABI5 protein can heteromerically interact with HY5 to activate its own expression, regulating the adaptation of young seedlings to environmental stresses [41]. However, the detailed molecular mechanisms underlying ABI5 regulation of seed germination and post-germinative growth in salt-tolerant millet remained elusive. In this study, we further investigated the regulatory role of PmABI5 in modulating ABA and salt responses during seed germination and seedling growth. The results show that PmABI5 can be induced by salt, ABA, heat, and drought, heteromerically interacting with PmbZIP131 or homologously interacting with itself to promote seed germination under salt stress. In addition, overexpressing PmABI5 in Arabidopsis thaliana and rice can significantly increase seed germination rate and root development under saline stress. These results are consistent with those of previous studies. Nonetheless, the ABI5-regulated inhibition of seed germination and early seedling growth, to protect the plant against adverse conditions via regulation of development, has also been reported [42]. However, a previous study showed that seedlings germinated and developed normally in the absence of ABA in spite of ABI5 accumulation [43]. Therefore, it is reasonable to suggest that PmABI5 may improve stress resistance by enhancing seedling growth and passing through the stress-sensitive period as early as possible.
Plant-specific NAM/ATAF/CUC (NAC) transcription factors (TFs) were initially associated with development [44] but are now being increasingly appreciated for the roles they play in stress responses and senescence [45,46]. In this study, PmABI5 was found to regulate the promoter region of PmNAC1, which in turn results in the regulation of PmNAC1 expression; additionally, the NAC transcription factor family was shown to be involved in salt stress response as well as improving crop stress resistance in many plants. The three subgroup III genes, namely, ANA019, ANAC055, and ANAC072 (RD26), are induced by drought, high salinity, and the hormones abscisic acid (ABA) and jasmonic acid (JA) [47]. The overexpression of the three NAC TFs results in the upregulation of several stress-inducible genes and improved drought tolerance [48], and ANAC019 and ANAC055 function as positive regulators of JA signaling defense responses [49]. Collectively, the molecular, morphological, and physiological evidence clearly indicates that PmABI5 confers higher tolerance under salt stress conditions by regulating effector genes for increasing salt tolerance or root development in crops. Our study also found that ABI5-NAC modules may regulate the expression of abiotic-responsive genes, but further investigation of the downstream regulatory genes is still needed.

Plant Materials and External Characteristics of Seeds
To perform the seed germination experiments, we selected 17 broomcorn millet populations which represent commonly available cultivars in Northwest China (Table 5). An OLYMPUS microscope (Zoom Stereomicroscope SZ61/SZ51) was used to observe the seed color, size, shape, and skin ( Figure 11).  On the basis of the preliminary experiments on germination rates, two varieties (Yumi 1 and Yumi 9) were selected for sequencing. Samples were originally collected from Yulin, Shaanxi, China. The population of Yumi 1 is a waxy broomcorn millet (seed coat color: white) with a glutinous taste due to the amylopectin content of starch in the endosperm. The other population was Yumi 9, which is a non-waxy broomcorn millet (seed On the basis of the preliminary experiments on germination rates, two varieties (Yumi 1 and Yumi 9) were selected for sequencing. Samples were originally collected from Yulin, Shaanxi, China. The population of Yumi 1 is a waxy broomcorn millet (seed coat color: white) with a glutinous taste due to the amylopectin content of starch in the endosperm. The other population was Yumi 9, which is a non-waxy broomcorn millet (seed coat color: yellow).

Salt Stress Treatment and Seed Germination Experiment
In total, 100 seeds were germinated on two pieces of filter paper in Petri dishes (9 cm × 9 cm × 1 cm). Seeds in germination dishes were pre-chilled (4 • C) for 2 d prior to germination, to aid in breaking seed dormancy as per the Association of Official Seed Analysts (AOSA) guidelines, and further cultured at 25 • C in continuous light. Germination conditions were set at 0 mM NaCl (only RO water) as control and 250 mM NaCl treatment.
Under suitable temperature conditions, the seeds first absorb water. Subsequently, the protoplasm inside the seed changes from a gel to a sol state, and enzymatic activity rapidly increases. The nutrients inside the seeds begin to transform and decompose from the endosperm or cotyledons and are transported to the radicle. In addition, germ cells were reconstructed, referred to as a sprout. After radicle elongation, white buds were revealed, which are called dew white. Seeds that reach this period are defined as germinated. The number of seeds that were germinated was counted and recorded daily when the seeds began to grow (i.e., from the beginning of the period of incubation 25 • C in continuous light).

RNA-Seq Sampling and Total Ribonucleic Acid (RNA) Preparation
The samples of Yumi 1 and Yumi 9, both cultured in 0 and 250 mM NaCl, were collected 0 and 3 h upon the emergence of dew white seeds following the onset of continuous light (each sample had three independent biological replicates). Samples were then rapidly frozen with liquid nitrogen and stored at −80 • C. There were 36 samples in total used for RNA-Seq and differential expression analyses.
Total RNA was extracted from seeds using the RNA prep pure polysaccharide polyphenol plant total RNA extraction kit (DP441) (TIANGEN, Beijing, China). RNA degradation and contamination were monitored on 1% agarose gels. RNA purity was checked using the NanoPhotometer ® spectrophotometer (IMPLEN, Westlake Village, CA, USA). RNA concentration was measured using Qubit ® RNA Assay Kit in Qubit ® 2.0 Fluorometer (Life Technologies, CA, USA). RNA integrity was assessed using the RNA Nano 6000 Assay Kit of the Agilent Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA).

Complementary DNA (cDNA) Library Preparation for Transcriptome Sequencing
A total amount of 1.5 µg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using NEBNext ® Ultra™ RNA Library Prep Kit for Illumina ® (NEB, Ipswich, MA, USA) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in NEB-Next First Strand Synthesis Reaction Buffer (5×). First-strand cDNA was synthesized using random hexamer primer and M-MuLV Reverse Transcriptase (RNase H-). Secondstrand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities. After adenylation of 3 ends of DNA fragments, NEB Next Adaptor with hairpin loop structure was ligated to prepare for hybridization. To select cDNA fragments of preferentially 250~300 bp in length, the library fragments were purified with an AMPure XP system (Beckman Coulter, Beverly, CA, USA). Then, 3 µL USER Enzyme (NEB, MA, USA) was used with size-selected, adaptor-ligated cDNA at 37 • C for 15 min followed by 5 min at 95 • C prior to PCR. PCR was then performed using Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) Primer. Finally, PCR products were purified (AMPure XP system) and library quality was assessed on the Agilent Bioanalyzer 2100 system.

Raw Data Assembly and Gene Functional Annotation
The adaptor sequences and low-quality sequence reads were removed from the datasets. Raw sequences were transformed into clean reads after data processing. These clean reads were then mapped to the reference genome sequence [50]. Only reads with a perfect match or one mismatch were further analyzed and annotated on the basis of the reference genome. Tophat2 tools software was used to perform mapping to the reference genome.
Gene function was annotated on the basis of the following databases: nr (NCBI non-redundant protein sequences); nt (NCBI non-redundant nucleotide sequences); Pfam (Protein family); KOG/COG (Clusters of Orthologous Groups of proteins); Swiss-Prot (a manually annotated and reviewed protein sequence database); KO (KEGG Ortholog database); GO (Gene Ontology).

Differential Expression Genes (DEGs) Analysis
Differential expression analysis of two conditions/groups was performed using the DESeq R package (1.10.1). DESeq provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting p-values were adjusted using Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted p-value < 0.05 found by DESeq were denoted as differentially expressed.
Gene Ontology (GO) enrichment analysis of the differentially expressed genes (DEGs) was implemented using the GOseq R package-based Wallenius noncentral hypergeometric distribution [51], which can adjust for gene length bias in DEGs.
KEGG [52] is a database resource for understanding high-level functions and utilities of the biological system, such as the cell, the organism, and the ecosystem, from molecularlevel information, especially large-scale molecular datasets generated by genome sequencing and other high-throughput experimental technologies (http://www.genome.jp/kegg/, accessed on 15 November 2018). We used KOBAS [53] software to test the statistical enrichment of differential expression genes in KEGG pathways.
Heatmap used the R package pheatmap, which uses default parameters. The combination of distance and algorithm is Euclidean distances and complete linkage.

Plant Seedling Stage Conditions and Stress Treatments
Millet seeds after disinfection were sown on Petri plates and incubated at 24 • C (8 h dark/16 h light) in a growth chamber. Seedlings at the two-leaf stage were transferred into vegetative soil matrix culture for four weeks. Millet seedlings were then treated with either 20% PEG6000, 100 µM ABA, 200 mM NaCl, 4 • C cold incubation, or 42 • C heat shock. Leaves were collected at 0, 3, 6, 9, 12, and 24 h after treatment and stored at −80 • C.
The Arabidopsis materials used in this study were in the Columbia (Col-0) background. After germinating in 1 2 Murashige and Skoog (MS) medium, wild-type (WT) plants and transgenic lines were grown in a growth chamber at a temperature of 22 • C, relative humidity of 65%, and with a 16 h light/8 h dark photoperiod.

Quantitative Real-Time PCR
Total RNA (1 µg) from relative samples was used to generate complementary DNA (cDNA) templates using the PrimeScript™ RT reagent Kit with gDNA Eraser (Perfect Real Time) (Takara, RR047Q). Gene-specific primers and TB Green™ Premix Ex Taq™ II (Tli RNaseH Plus) (Takara, RR820Q) were used for QPCR on StepOne System (Applied Biosystems, Darmstadt, Germany) according to the manufacturer's instructions. Genespecific primers were designed by NCBI (https://www.ncbi.nlm.nih.gov/tools/primerblast/index.cgi?LINK_LOC=BlastHome, accessed on 13, November 2019) ( Table 6). PCRs included 2.0 µL of first-strand cDNA (1:10 diluted cDNA), 5 µL of SYBR Mix, and 10 nM of each primer in a final volume of 10 µL. The primers were designed in non-polymorphic regions across accessions. For each gene, we used cDNA samples of three biological replicates per accession as a template for QPCR. Table 6. Primers of qRT-PCR for validation of the selected DEGs.

Gene Name Description Primers
PmbZIP33 Quantification of gene expression was carried out by the relative quantification method (2 −∆∆CT method) implemented in StepOne software v2.3. Since the efficiencies of the several primers were not identical in the three-millet group, we corrected the relative quantification results according to the efficiency of each of the genes in each of the groups by StepOne software v2.3 on the basis of the following equations: Efficiency = 10 (−1/slope) for each primer and RQ = Etarget − (Ct treatment − Ct control)/Endogenous − (Ct treatment − Ct control).
In the BiFC assays, the full-length PmbZIPs (including PmABI5) were fused with the C-terminal of C-YFP via the pCAMBIA2300-VYCE vector and PmbZIPs (including PmABI5) were also fused to the N-terminal of N-YFP using the pCAMBIA2300-VYNE vector, recombinant plasmid pairwise combination, and via Agrobacterium-mediated cotransformation into Nicotiana benthamiana leaves. All the leaves were collected at 48 h after transformation, and the fluorescent signal was detected using confocal microscopy (IX83-FV1200, Olympus, Philadelphia, PA, USA).

Construction of Transgene Vector and Plant Transformation
The full-length PmbZIP1, PmbZIP2, PmbZIP3, PmbZIP4, PmbZIP5, and PmbZIP7 ORFs were cloned into the pCAMBIA2300-eGFP vector to generate the PmbZIPs:eGFP fusion constructs. The parental line of Arabidopsis thaliana L. was Col-0, the overexpressing lines Floral Dip were used for transformation. The parental line of Oryza sativa L. was Nipponbare, and the method used was as previously described [54].
Plant samples were imaged, and the surface area, average root diameter, length, and volume of roots were measured by EPSON V700 Root Analysis System (Seiko, EPSON Crop, Nagno, Japan), and analyzed by WinRhizo (Regent Instruments Inc., Quebec, QC, Canada).

Y1H Assays
Yeast one-hybrid assay, used to identify the promoter binding site, was performed with yeast strain Y1HGold cells (Clontech). Three copies of G-box (CACGTG) and three copies of its mutant sequence g-box (CATTTG) were cloned into the pAbAi vector (Figure 11a,b). pGbox-ABAi and pgbox-ABAi were transformed into the Y1HGold strain after linearization, Gbox-ABAi and gbox-ABAi reporter strains were obtained by homologous recombination. At the same time, fusion of full-length PmABI5 to the C-terminal GAL4 activation domain using the pGADT7 vector. Fusion vector was transformed into two reporter strains and the control comprised the empty pGADT7 vector. All yeasts were tested on SD/−Ura/−Leu and SD/−Ura/−Leu medium with 800 ng/mL AbA.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.