The miR156b–GmSPL2b module mediates male fertility regulation of cytoplasmic male sterility-based restorer line under high-temperature stress in soybean

Overexpression of miR156b decreased male fertility in CMS-based restorer line under HT stress in soybean

Our previous small RNA sequencing data showed that miR156b was a heat-responsive miRNAs in soybean CMS-based F₁ during flowering under HT stress (Ding et al., 2021b). To investigate whether miR156b plays any role in soybean male fertility regulation under HT stress, two soybean transgenic lines overexpressing miR156b were generated in Williams 82 (W82), a restorer line of soybean CMS line NJCMS1A (Figure 1a,b).

Under normal temperature (NT) condition, there was no significant difference in male fertility between miR156b overexpression (miR156bOE) lines and W82 plants (Figure 1c,d). After HT treatment for 4 days, the anther of miR156bOE was dehiscent and the number of viable pollen grains was only about 25%, while the anther dehiscence of control W82 was normal and its number of viable pollen grains was about 40% higher than that of miR156bOE lines (Figure 1e). After HT treatment for 5 days, miR156bOE showed anther indehiscence and had no fertile pollen, but the anthers of W82 were still slightly dehiscent with about 30% fertile pollen (Figure 1c–e). Similar to the result of pollen I₂-KI staining, a transverse section was taken from the unopened flowers (the flowers that will open the next morning). Compared with W82, partial pollen grains of miR156bOE plants developed abnormally after 4 days of HT treatment, forming empty and shrivelled pollen (Figure 1f). After HT treatment for 5 days, all pollen grains of miR156bOE plants were shrivelled and had abnormal morphology, while only partial pollen grains of W82 developed abnormally.

Subsequently, we detected the germination of pollen in vitro under transient HT stress (Figure 2a,b). The germination rates of miR156bOE and W82 in vitro were about 95% under NT condition. However, HT only reduced the pollen viability of W82 by about 10%, while the pollen viability of miR156bOE lines decreased by 35%–40%. In addition, some pollen tubes of the miR156bOE lines were shorter than those of W82 under HT stress.

In order to further understand whether miR156b is involved in the regulation of male fertility of soybean CMS-based F₁ under HT stress, we conducted a cross study between the CMS line NJCMS1A and miR156bOE lines. Similar to miR156bOE, the number of leaves in (NJCMS1A × miR156bOE) F₁ was also increased compared to (NJCMS1A × W82) F₁ (Figures 1a and 2c). Male fertility observation showed that the anthers of (NJCMS1A × miR156bOE) F₁ and (NJCMS1A × W82) F₁ were dehiscent and their pollen fertility rate was about 50% under NT condition (Figure 2d–f). The HT treatment results showed that (NJCMS1A × miR156bOE) F₁ was more sensitive to HT than (NJCMS1A × W82) F₁ and their male parent miR156bOE lines (Figures 1c–e and 2d–f). After HT treatment for 2 days, the pollen fertility rate of (NJCMS1A × miR156bOE) F₁ decreased sharply, only about 12%–26%. After HT treatment for 3 days, (NJCMS1A × miR156bOE) F₁ showed anther indehiscence and almost completely male sterility, while (NJCMS1A × W82) F₁ remained anther dehiscence and its pollen fertility rate was about 30%. All results indicated that miR156b negatively regulated the male fertility of soybean CMS-based restorer line under HT stress.

Overexpression of miR156b inhibited the expression of pollen development and HT response related genes under HT stress in soybean

To further elucidate the regulatory network of miR156b, we performed transcriptome sequencing of miR156bOE-1 under both NT and HT conditions (W82 as the control). RNA sequencing (RNA-seq) data revealed that more than 3500 genes were differentially expressed in flower buds of miR156bOE-1 and W82 under HT stress compared to those under NT condition (Figure S1; Tables S1 and S2). Most differentially expressed genes (DEGs) were down-regulated in miR156bOE-1NT vs miR156bOE-1HT combination, unlike most of the DEGs were up-regulated in W82NT vs W82HT combination (Figure S1, Tables S1 and S2). Similarly, more than 70% of the DEGs (219) were down-regulated in flower bud of miR156bOE-1 compared to W82 under HT stress (Figure 3a, Tables S3 and S4). Then, the DEGs in the W82HT vs miR156bOE-1HT combination were subjected to gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Functional annotation of DEGs revealed that overexpression of miR156b affected multiple molecular function and biological processes related to gene expression regulation, plant growth and stress responses, including transcription regulator activity, antioxidant activity, reproduction and reproductive process (Figure S2). A KEGG pathway analysis showed that the DEGs were mainly related to the pentose and glucuronate interconversions and the protein processing in endoplasmic reticulum (Figure 3b).

In W82HT vs miR156bOE-1HT combination, some DEGs were related to the flavonoid and phenylpropanoid biosynthesis, including FLS1 and peroxidase 5-like (POD5-like) (Figure 3c). Further analysis indicated that most of the DEGs related to carbon metabolism and sugar transport were down-regulated, such as pectinesterase/pectinesterase inhibitor-like (PE/PEI-like), pectinesterase-like (PE-like), pectate lyase-like (PL), polygalacturonase (PG), exopolygalacturonase (Exo-PG), sugar transport protein 10-like (STP10-like) and sucrose transport protein SUC8-like (SUC8-like) (Figure 3c). In addition, genes related to plant hormone signal transduction, pollen development and HT stress, such as auxin-induced protein AUX22 (AUX22), pollen receptor-like kinase 3 (PRK3), pollen-specific leucine-rich repeat extensin-like protein 1 (PLREP), calcium-dependent protein kinase 17-like (CDPK17-like), heat shock factor A6b (HSFA6b), small heat shock protein (sHSP) and heat shock protein 83 (HSP83), underwent a reduction in expression in miR156bOE-1 under HT compared to W82 (Figure 3c,d). Nearly all of the above-mentioned DEGs (except HSP17.9) had ≥1 GTAC motif (SBP-binding element) in their promoters (Figure 3c,d). Most importantly, many of them were highly expressed in flower of soybean (Figure S3), such as FLS1, CDPK17-like, POD5-like, carbon metabolism, sugar transport and pollen development-related DEGs. To verify the RNA-Seq data, 11 of them were selected for quantitative reverse transcriptase PCR (qRT-PCR) analysis, and the coincidence rate between qRT-PCR and RNA-seq was 100% (Figure 3e).

Overexpression of miR156b altered flavonoid accumulation in soybean CMS-based restorer line under HT stress

To further illustrate the mechanism of miR156b in regulating male fertility of soybean CMS-based restorer line under HT stress, a widely targeted metabolome study was performed using the flower buds of miR156bOE-1 and W82 under both NT and HT conditions. A total of 1026 metabolites were detected in the flower buds of miR156bOE-1 and W82, including 11 classes of metabolites, among which flavonoid was the most abundant class metabolite (Figure S4). Approximately 243 metabolites showed differences (¦fold change (FC)¦ ≥ 2, variable importance in project (VIP) ≥ 1) between miR156bOE-1 and W82 in four combinations (W82NT vs miR156bOE-1NT, miR156bOE-1NT vs miR156bOE-1HT, W82NT vs W82HT and W82HT vs miR156bOE-1HT) (Figure 4a, Tables S5–S8). Of these, 18 metabolites were down-regulated and 10 were up-regulated in miR156bOE-1 compared to W82 under HT stress (Figure 4b). Most importantly, one third of them was flavonoid. KEGG data also showed that differential metabolites involved in flavonoid metabolic pathway (isoflavonoid biosynthesis, flavonoid biosynthesis, flavone and flavonol biosynthesis) were enriched in W82HT vs miR156bOE-1HT comparison (Figure S5). In particular, the amount of five (six in total) flavonoids in the flower buds of both miR156bOE-1 and W82 was induced by HT stress (Figure 4c). Interestingly, four of them were down-regulated in miR156bOE-1 under both NT and HT conditions compared to W82. Given that most of the differential flavonoids were down-regulated in miR156bOE-1HT, this indicated that miR156b altered the biosynthesis and accumulation of flavonoid in the flower bud of soybean CMS-based restorer line under HT stress.

We then examined the expression of genes involved in flavonoid metabolic pathway in miR156bOE lines. The transcript levels of two flavonoid metabolism-related genes, including GmFLS1 (Glyma.05G088100) and GmCHI (Glyma.10G29220), were down-regulated in miR156bOE-1HT compared to W82HT (Figure 4d,e). Through promoter analysis, 2 GTAC motifs were found in the promoter region of the GmFLS1, but not in the promoter region of GmCHI (Figure 3c). Therefore, the miR156b–SPL module may directly or indirectly participate in flavonoid metabolism by regulating the expression of GmFLS1 and GmCHI genes, thus regulating the male fertility of soybean CMS-based restorer line under HT stress (Figure 4f).

GmSPL2b knockout increased HT susceptibility in soybean CMS-based restorer line during flowering

miR156 mainly plays its role by negatively regulating its target genes that encoding SPL, which have miR156-binding sites (Lei and Liu, 2016). Our previous studies found that 20 GmSPLs were targeted by miR156b, and miR156b negatively regulated GmSPL2b during flowering in soybean CMS-based F₁ under HT stress (Ding et al., 2020b, 2021b). Furthermore, GmSPL2b was down-regulated in miR156bOE plants under both NT and HT conditions (Figure 5a). To better understand the function of GmSPL2b, a bioinformatics analysis on GmSPL2b was carried out first. GmSPL2b encodes a long peptide containing 433 amino acids, and amino acid sequence alignment revealed that GmSPL2b has a typical domains of SPL, including two Zn-finger-like structures (Zn1 and Zn2) and a nuclear localization signal (NLS) (Figure 5b,c). Phylogenetic analysis showed that GmSPL2b had high sequence identity with SPL2 or SPL12 in Medicago truncatula and Vigna unguiculata, etc (Figure 5d). Subcellular localization analysis revealed that the 35S::GmSPL2b–GFP fusion protein was exclusively localized in the nucleus, which was consistent with its predicted NLS domain (Figure 5b,e). GUS staining of the pGmSPL2b::GUS-transformed Arabidopsis confirmed that GmSPL2b was expressed in anthers of inflorescence during flowering (Figure 5f), suggesting that GmSPL2b plays an important role in regulating pollen development. Furthermore, the transcription activity assay found that the Y2H strain containing BD-GmSPL2b could grow on both SD/−Trp and SD/−Trp/−His/−Ade culture medium (Figure 5g), indicating that GmSPL2b has transcriptional activation activity and can bind with the promoter element of downstream regulatory gene to regulate its transcriptional process.

To further investigate the putative biological function of GmSPL2b in regulating male fertility of soybean CMS-based restorer line under HT stress, we employed CRISPR/Cas9 method to knockout GmSPL2b and obtained gmspl2b mutants with deletion of 14 bp, 10 bp, 5 bp and 7 bp (Figure 6a,b). All mutation types of gmspl2b resulted in a frame-shift with premature termination of GmSPL2b translation and loss of SBP functional domain (Figure 6c). A 10-bp deletion of gmspl2b-1 and a 5-bp deletion of gmspl2b-2 homozygous single mutants were selected for subsequent experiments (Figure 6d). Male fertility of the three materials including W82 was normal under NT condition (Figure 6e–g). The pollen fertility rate of gmspl2b mutants and W82 decreased gradually with the extension of HT treatment time, but the pollen fertility of gmspl2b mutants was lower than that of W82 (Figure 6f,g). However, unlike miR156bOE lines, gmspl2b mutants did not exhibit an anther indehiscence phenotype until after 5 days of HT treatment (Figures 1c and 6e), indicating that GmSPL2b may be only one of the target genes of miR156b involved in the regulation of male fertility in soybean CMS-based restorer line under HT stress. Furthermore, qRT-PCR analysis found that the HSP-related genes, including GmHSP17.6 (Glyma.04G054400) and GmHSP83 (Glyma.16G178800), were also down-regulated in gmspl2b mutants under HT stress (Figure 6h), indicating that they may be directly or indirectly regulated by GmSPL2b.

ROS metabolism in anthers of miR156bOE and gmspl2b was disturbed under HT stress in soybean

ROS is closely related to male fertility of plants under HT stress (Begcy et al., 2019; Djanaguiraman et al., 2018; Zhao et al., 2018). Thus, the ROS staining of anthers was performed in miR156bOE, gmspl2b and W82 under NT and HT conditions (Figure 7a,b). The 5-(and 6)-chroromethyl-2′,7′-dichlorodihydrofluorescein diacetate (CM-H2DCFDA) staining showed that the ROS content in W82 anthers was higher than that of miR156bOE and gmspl2b mutants under NT. Then, the ROS accumulation decreased in W82HT compared to W82NT. However, more ROS accumulation was found in miR156bOEHT compared to either W82HT or miR156OENT. Under HT stress, the ROS content in anthers of gmspl2b mutants increased compared to NT condition, but only slightly increased compared to W82HT, without reaching a significant difference level. This may be because GmSPL2b is only one of the target genes of miR156b involved in male fertility of soybean under HT stress. Intriguingly, our RNA-seq found that two other SPLs, GmSPL4a and GmSPL4b (namely GmSPL12e and GmSPL12d in Ding et al., 2020b), were down-regulated in miR156bOE plants under both NT and HT conditions (Figure S6a,b). Furthermore, they are targeted by miR156b via degradome analysis (Figure S6c,d; Ding et al., 2019b). Similar to GmSPL2b, GmSPL4a/b were also highly expressed in flower of soybean (Figure S6e; Phytozome 13). The difference is that the target site of miR156b is located in the CDS region of GmSPL2b, but in the 3′UTR of GmSPL4a/b (Ding et al., 2020b).

Considering that miR156b was participated in the regulation of flavonoid metabolism (Figures 3c,e and 4), flavonoid was involved in ROS clearance during plant flowering under HT stress (Muhlemann et al., 2018). We analysed the expression level of GmFLS1 in the gmspl2b mutants and found it was also down-regulated in flower buds of gmspl2b mutants compared to W82 (Figure 7c). All these results suggested that miR156b–GmSPL2b may be involved in ROS clearance by mediating the expression of flavonoid metabolism genes, thereby regulating the male fertility of soybean CMS-based restorer line under HT stress.

GmHSFA6b acted as one of the downstream regulatory genes of miR156b–GmSPL2b module

Previous experimental results showed that the male fertility of miR156bOE lines and gmspl2b mutants was highly sensitive to HT, and GmHSFA6b was down-regulated in miR156bOE lines (Figure 3e). Since HSFA family members are important factors that activate the protective mechanism of plant male organs during flowering under HT stress (Fragkostefanakis et al., 2016; Giorno et al., 2010), we decided to investigate whether GmHSFA6b contributes to the regulation of plant male fertility under HT stress. Furthermore, four GTAC motifs (F1–F4) were found in the promoter of GmHSFA6b, and qRT-PCR analysis confirmed that GmHSFA6b was down-regulated in gmspl2b mutants (Figure 8a,b). All these data suggested that GmHSFA6b might act as a downstream regulatory gene of miR156b–GmSPL2b module under HT stress.

Next, the interaction between GmSPL2b and the GmHSFA6b promoter was detected by yeast one hybrid (Y1H) assay. A 42-bp fragment containing three copies of F1–F4 (GTAC) with 5-bp promoter sequence on both sides was cloned into the pAbAi vector and used as bait, respectively, while GmSPL2b was used as prey. However, self-activation detection found that F4 bait had strong self-activation ability (Figure S7). Moreover, Y1H assay found that GmSPL2b can specifically bind F1 and F3 (Figure 8c). We then validated the activation of GmSPL2b to GmHSFA6b gene using a transient luciferase (LUC) complementation imaging assay in tobacco (Nicotiana tabacum) leaves. Luminescence detection revealed that the fluorescence intensity of 35S::GmSPL2b coexpressed with pGmHSFA6b::LUC was significantly stronger than that of the control (Figure 8d). These results suggested that GmSPL2b activated the expression of GmHSFA6b by directly binding to its promoter.

Overexpression of GmHSFA6b conferred HT tolerance in Arabidopsis during flowering

To further confirm the role of GmHSFA6b in the regulation of plant male fertility under HT stress, two independent GmHSFA6bOE transgenic Arabidopsis lines were selected for HT treatment (Figure 9a). Both GmHSFA6bOE transgenic Arabidopsis lines and wild-type (WT) were exposed in an incubator at 42 °C for 4.5 h, and observed the male fertility within next 6 days after HT stress, mainly including anther dehiscence and pollen fertility. On the second day after HT treatment, the GmHSFA6bOE plants showed anther dehiscence with pollen shrinkage slightly (Figure 9b–d). On the sixth day after HT treatment, the GmHSFA6bOE plants showed a little pollen abortion. However, the WT showed pollen deformity (shrinkage greatly and the red colour is lighter) with anther indehiscence and most pollen abortion on the second and sixth day after HT treatment, respectively (Figure 9b–d). The nitro blue tetrazolium (NBT) and CM-H2DCFDA staining showed that the ROS content in inflorescence of GmHSFA6bOE transgenic Arabidopsis lines was lower than that of WT under both NT and HT conditions (Figure 9e,f). All these results revealed that the miR156b–GmSPL2b–GmHSFA6b module might regulate plant male fertility through ROS metabolic pathway under HT stress.

miR156b–GmSPL2b was one of the key module for regulating male fertility of soybean CMS-based restorer line under HT stress

The miR156–SPL module plays an important role in the development regulation of flowering time, plant architecture and nodulation of soybean (Cao et al., 2015; Sun et al., 2019b; Yun et al., 2022). It was also found that miR156 had different responses to HT at different developmental stages of plants, and it was mainly up-regulated by HT stress at seedling stage of wheat (Xin et al., 2010), cabbage (Yu et al., 2012), tomato (Zhou et al., 2016) and maize (He et al., 2019). Meanwhile, HT can reduce the expression of SPL by inducing the expression level of miR156, and then induced the production of HSPs, thus enhancing the HT tolerance of Arabidopsis and alfalfa at seedling stage (Matthews et al., 2019; Stief et al., 2014). Contrary to seedling stage, miR156 was mainly inhibited by HT stress at flowering stage (Li et al., 2015; Liu et al., 2017b; Pan et al., 2017). Pan et al. (2017) conducted small RNA sequencing and qRT-PCR analysis on tomato stamens under normal condition and HT treatment, and found that miR156e and its target gene SPL15 showed negative regulation at the expression level. However, only one material was used in this study. The relationship between miR156e–SPL15 module and male fertility of tomato under HT stress is not clear. Our previous study found that soybean miR156b and its target gene GmSPL2b not only had a negative regulatory relationship in the flower buds of soybean HT-tolerant CMS-based F₁ before and after HT treatment, but also had the same relationship between HT-tolerant CMS-based F₁ and HT-sensitive CMS-based F₁ under HT treatment (Ding et al., 2021b). In addition, the expression level of miR156b in HT-sensitive CMS-based F₁ was higher than that in HT-tolerant CMS-based F₁ under HT stress. Furthermore, our previous study found that the miR156bOE transgenic Arabidopsis showed male sterility after HT treatment (Ding et al., 2021b). However, the mechanism of miR156b–GmSPL2b module regulating the male fertility of soybean CMS-based restorer line under HT stress remains unclear.

In this study, we showed that miR156b negatively regulated male fertility in soybean CMS-based restorer line under HT stress, and GmSPL2b was down-regulated in miR156bOE transgenic soybean under both NT and HT conditions (Figures 1, 2 and 5a). Previously, we showed that GmSPL2b was highly expressed in flower buds of soybean during flowering (Ding et al., 2020b). In addition, Wang et al. (2016) found that the reduced transcription of its AtSPL2 homologue in Arabidopsis was similar to the phenotype of miR156 transgenic Arabidopsis (anther size and the amount of pollen grains per anther decreased). Moreover, knocking out GmSPL2b also reduced the pollen fertility rate of soybean CMS-based restorer line under HT stress (Figure 6e–g). However, there was still a certain gap between gmspl2b mutants and miR156bOE lines under HT stress, both in terms of pollen fertility rate/anther dehiscence or ROS content in anthers under HT treatment (Figures 1c–f, 6e–g and 7a,b). Intriguingly, in addition to GmSPL2b, two other SPLs, GmSPL4a and GmSPL4b, were down-regulated in miR156bOE plants under both NT and HT conditions (Figure S6a,b). Furthermore, they were targeted by miR156b via degradome analysis (Figure S6c,d; Ding et al., 2019b). Thus, GmSPL2b may act as one of the target genes of miR156b regulating male fertility of soybean CMS-based restorer line under HT stress. It cannot be ruled out that GmSPL4a/b may play a synergistic role in the regulation of male fertility of soybean CMS-based restorer line under HT stress, which needs further study.

HSF exists widely in plants and is one of the transcription factors (TFs) that play an important role in response to HT stress during reproductive development (Ding et al., 2020a; Fragkostefanakis et al., 2016; Giorno et al., 2010). In this study, GmHSFA6b was down-regulated in miR156bOE lines and gmspl2b mutants under HT stress (Figures 3e and 8b). Since GmHSFA6b promoter (2 kb in length) contains four GTAC motifs bounded by SPL TF (Figure 8a), it is generally assumed that miR156b–GmSPL2b module may play a role in the regulation of male fertility in soybean CMS-based restorer line under HT stress by regulating GmHSFA6b. Importantly, we proved that GmSPL2b is a transcriptional activating nuclear protein that can directly bind to the promoter of GmHSFA6b (Figures 5g and 8). Furthermore, overexpression of GmHSFA6b in Arabidopsis conferred its HT tolerance during flowering (Figure 9). Hence, this research disclosed a novel mechanism wherein the miR156b–GmSPL2b module regulates male fertility of soybean CMS-based restorer line under HT stress by activating GmHSFA6b. In addition, gene cloning, sequence alignment and bioinformatics analysis showed that the promoter core elements and precursor/CDS sequences of miR156b, GmSPL2b and GmHSFA6b were identical in both HT resistant and sensitive restorer lines, although their promoter sequences were slightly different in the two materials (Figures S8–S10). These results indicated that miR156b regulated the male fertility of soybean restorer line under HT stress by regulating GmSPL2b–GmHSFA6b, and there may be different regulatory factors in the upstream of miR156b, resulting in the different expression trends of miR156b between the two materials. Further study is needed to determine the upstream regulatory factors of miR156b.

The normal development of plant anther and pollen requires the participation of many types of genes, such as anther/pollen development, sugar transport and calcium signalling transduction related genes. The abnormal expression of any one of above genes may lead to male sterility in plants. For example, BcPLL9 and CsSUT1 were found to regulate anther pectin metabolism and sugar transporter in Brassica campestris and cucumber, respectively, and their down-regulation resulted in male semi-sterility or sterility in plants (Jiang et al., 2014; Sun et al., 2019a). Our previous research also found that genes related to anther/pollen development and sugar transport were closely related to male fertility development of soybean CMS-based F₁ under HT stress (Ding et al., 2020b). In Arabidopsis, the double disruption of CPK17/34 disrupts the pollen transmission (Myers et al., 2009). In rice, the RNA interference of OsCPK9 leads to the reduction of pollen fertility rate (Wei et al., 2014). Here, we found that many of their homologous genes in soybean were also down-regulated in miR156bOE lines compared to W82 under HT stress, such as GmPME, GmExo-PG, GmSUC8 and GmCDPK17 (Figures 3c). Furthermore, there was at least one SPL-binding site in the promoter of these DEGs (Figure 3c). However, whether miR156b regulates the expression of these DEG through GmSPL2b remains to be further investigated.

HSPs are a group of chaperone proteins that control the folding/unfolding of substrate proteins and the degradation of misfolded/denatured proteins under HT stress (Singh et al., 2016; Sun et al., 2021b). In this study, 11 HSP20 family members and 1 HSP83 gene were induced by HT in both miR156bOE lines and W82, but all of them were down-regulated in miR156bOE under HT compared to W82 (Figure 3d). The qRT-PCR analysis found that some of them were also down-regulated in gmspl2b mutants under HT compared to W82 (Figure 6h). So we speculated that the decreased pollen fertility rate of miR156bOE and gmspl2b mutants under HT stress might be related to the down-regulation of these HSP genes. Just as previous studies have shown that HSPs play an important role in the regulation of HT tolerance in plants (Yang et al., 2017; Zhai et al., 2016), and overexpression of GmHSP20a could improve the HT resistance of Arabidopsis during flowering (Ding et al., 2020a). Interestingly, there were many GTAC and HSE motifs in the promoters of these HSPs (Figure 3d), which can be bound by SPL and HSF, respectively. However, whether the expression of these HSPs in miR156bOE, gmspl2b mutants and W82 is regulated by GmSPL2b or GmHSFA6b remains to be investigated.

miR156b–GmSPL2b module involved in the plant male fertility regulation under HT stress by regulating ROS metabolism

HT can cause a large amount of ROS produced by mitochondria in reproductive organs, and the abnormal accumulation of ROS is one of the direct causes of male sterility in plants (Begcy et al., 2019; Djanaguiraman et al., 2018; Zhao et al., 2018). After sensing HT signals in the environment, plants can induce gene (including miRNA and its target genes) changes at transcription and post-transcription levels through related signal transduction pathways to activate the ROS scavenging system in vivo, thus effectively eliminating intracellular ROS and reducing damage to male reproductive organs (Chen et al., 2020; Liu et al., 2017a). It is well known that plants have enzymatic ROS scavenging system including antioxidant enzymes such as POD, as well as non-enzymatic ROS scavenging system including antioxidant substances such as flavonoid, which play important roles in plant reproduction and development (Liu et al., 2018; Muhlemann et al., 2018; Zhao et al., 2018). In this study, the miR156bOE presented with higher amount of ROS but with lower content of flavonoid compared to W82 under HT stress (Figures 4 and 7). Interestingly, GmFLS1, a gene involved in flavonoid biosynthesis, was down-regulated in the flower buds of miR156bOE lines and gmspl2b mutants under HT stress (Figures 3e and 7c). Recent studies have shown that transcription factor MYB21 can activate the expression of flavonol synthesis gene FLS1 in anther and pollen of Freesia and Arabidopsis, mediating ROS clearance and male fertility regulation (Shan et al., 2020; Zhang et al., 2021). Meanwhile, some studies have shown that flavonol can reduce ROS accumulation induced by HT stress and enhanced the tolerance of tomato to HT during reproductive development (Muhlemann et al., 2018; Paupière et al., 2017; Rutley et al., 2021). Moreover, there were GTAC motifs in the promoter of GmFLS1 (Figure 3c), and we speculated that miR156b–GmSPL2b may regulate ROS accumulation by regulating the expression of GmFLS1, thus participating in the regulation of male fertility in soybean CMS-based restorer line under HT stress (Figure S11).

Intriguingly, we found that the inflorescence of GmHSFA6bOE transgenic Arabidopsis accumulated lower content of ROS under HT stress (Figure 9e,f), suggesting that GmHSFA6b may be involved in ROS clearance, thus improving the HT tolerance of transgenic Arabidopsis. Wang et al. (2020) found HSFA8a could increase the expression of flavonoid synthesis genes MYB12, ANS and FLS to promote the accumulation of flavonoid, thus effectively eliminating ROS and improving the drought tolerance of apple seedlings. In addition, Feng et al. (2019) found that sHSP positively regulates plant resistance to HT stress during seedling by regulating enzymatic ROS scavenging system. Considering the existence of HSE elements (-nGAAnnTTC-) in promoters of some differentially expressed sHSPs in this study (Figure 3d), they may be regulated by GmHSFA6b. These results suggested that GmHSFA6b may positively regulate male fertility in soybean CMS-based restorer line under HT stress through the flavonoid–ROS pathway (Figure S11). Given that the promoter of GmHSFA6b also has SPL-binding sites (Figure 8a), we concluded that the miR156b–GmSPL2b module may play its role in regulating male fertility in soybean CMS-based restorer line under HT stress by directly activating the expression of GmHSFA6b. Further research will be helpful to dissect the molecular mechanism of miR156b–GmSPL2b–flavonoid–ROS pathway involved in the regulation of male fertility in soybean CMS-based restorer line under HT stress.

In summary, our study found that miR156b–GmSPL2b module plays an important role in regulating male fertility in soybean CMS-based restorer line under HT stress. Based on our findings, we proposed a model that miR156b mediated male fertility of soybean CMS-based restorer line by regulating GmSPL2b. Due to GmSPL2b acted as a transcriptional activator, its degradation will lead to the down-regulation of downstream regulatory genes, including GmHSFA6b, GmFLS1, GmPOD5, carbon metabolism and sugar transport related genes, which interfered the ROS scavenging and anther/pollen development, respectively, and ultimately led to abnormal male fertility in soybean CMS-based restorer line (Figure S11). As the other two target genes of miR156b, GmSPL4a/b may also play the same role as GmSPL2b in this process. Furthermore, we can precisely manipulate the miR156b-binding site in GmSPL2b in the future, or use the related gene resources in miR156b–GmSPL2b pathway as fertility enhancer genes to create HT-tolerant strong restorer lines, so as to cultivate excellent new CMS-based hybrid soybean varieties with stable male fertility.

Plant materials and HT treatment

Soybean CMS line NJCMS1A (Ding et al., 1999, 2002; Gai et al., 1995) and its restore line N4608, YY6 (Ding et al., 2020a), W82, two miR156bOE lines #1 and #2, two GmSPL2b knockout lines #1 and #2, (NJCMS1A × W82) F₁ and (NJCMS1A × miR156bOE) F₁ were grown in an illuminated incubator (RXZ-430D, Ningbojiangnan, Ningbo, China) at 26/20 °C with a 12 h/12 h (light/dark) photoperiod and 70% relative humidity during seedling. The hybridization between the CMS line NJCMS1A and miR156bOE lines/W82 was carried out in the greenhouse of Nanjing Agricultural University. For HT treatment, the flowering plants (R1 stage) of W82 and transgenic lines/F₁ plants were incubated at 40 °C/34 °C with a 12 h/12 h (light/dark) photoperiod and 70% relative humidity in an illuminated incubator and maintained for 3–5 day. All types were grown at 30 °C/24 °C (12 h light/12 h dark) as control. Flower buds of each genotype (miR156bOE, W82, gmspl2b mutant) were collected from three individual plants as three independent biological replicates under NT and HT, respectively, and then frozen in liquid nitrogen immediately and stored at −80 °C for further use.

Nicotiana benthamiana was used for subcellular localization and dual-luciferase (LUC) assay. The Arabidopsis thaliana Columbia (Col-0) ecotype was used for pGmSPL2b::GUS and GmHSFA6b genetic transformation and WT. They were grown under long-day conditions (16 h light/8 h dark) in an illuminated incubator (RXZ-430D, Ningbo Jiangnan, Ningbo, China) at 23 °C as previously described (Ding et al., 2020a). The HT treatment on male fertility of GmHSFA6b transgenic Arabidopsis and WT was performed in an illuminated incubator (RXZ-430D, Ningbo Jiangnan, Ningbo, China) at 42 °C for 4.5 h and then transferred to normal growth condition, according to Kim et al. (2001) described with minor modification. All types were grown at 23 °C as control.

Gene sequence comparison and bioinformatics analysis

The DNA and total RNA from the flower buds of soybean cultivar N4608 and YY6 were isolated using the Plant Genomic DNA Extraction Kit (TIANGEN, Beijing, China) and TRIzol reagent (Invitrogen, Carlsbad, CA, United States) according to the manufacturer's introductions. The miR156b precursor, and promoters (2000 bp) of miR156b precursor, GmSPL2b (Glyma.18G005600) and GmHSFA6b (Glyma.10G029600) were cloned from DNA of soybean cultivar N4608 and YY6. The full-length CDS regions of GmSPL2b and GmHSFA6b were cloned from cDNA of soybean cultivar N4608 and YY6. All primers used for PCR are listed in Table S9. The multiple sequence alignment of the gene nucleotide and amino acid sequences was generated using the DNAMAN. The WebLogo (http://weblogo.berkeley.edu/logo.cgi) tool was used to generate an SBP-box domain logo. The phylogenetic tree of SPL2 was constructed using MEGA (version 5.02; Tamura et al., 2011) by the neighbour-joining method with 1000 bootstrap repetitions. The FPKM values of DEGs between miR156bOE-1 and W82 in soybean root, stem, leaf and flower tissues were obtained from the RNA-seq data in Phytozome v13.0 (https://phytozome.jgi.doe.gov/pz/portal.html#), and the heat map was conducted using MeV 4.9 software. The promoter sequences of DEGs in RNA-seq between miR156bOE-1 and W82 were obtained from the NCBI database (https://www.ncbi.nlm.nih.gov/). The regulatory elements in the promoters of miR156b precursor, GmSPL2b, GmHSFA6b were predicted using PlantCARE on the web (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/).

Vector construction and genetic transformation

The miR156bOE vector was got from our previous study (Ding et al., 2020b). The full-length coding domain sequence (CDS) regions of GmSPL2b and GmHSFA6b were cloned from cDNA of soybean cultivar N4608 and cloned into the pMD-19T vector, then inserted into pCAMBIA3301-26 vector (with BamHI linearized) using a one-step cloning kit (Vazyme, Nanjing, China) to generate the p35S::GmSPL2b and p35S::GmHSFA6b expression vector, respectively. Three vectors (pU6-sgR, pU3-sgR and pGmUbi-Cas9) from Professor Yu's laboratory (National Center for Soybean Improvement, Nanjing Agricultural University, Nanjing, China) was used to knock out GmSPL2b according to an established protocol (Du et al., 2016; Zhang et al., 2020). The target sequences of the sgRNAs were designed by the web-based tool CRISPR-P 2.0 (http://cbi.hzau.edu.cn/cgi-bin/CRISPR2/CRISPR). Mutation detection and primer design were performed using CRISPR-GE software (Zeng et al., 2018). To generate β-glucosidase (GUS) reporter construct of GmSPL2b, the promoter of GmSPL2b (2000 bp) was amplified by PCR using DNA from soybean cultivar N4608 and cloned into the pMD-19T vector, and then replaced the 35S promoter of pCAMBIA3301-GUS using HindIII and NcoI digestion, resulting in a plasmid of pGmSPL2b::GUS.

The miR156bOE and GmSPL2b knockout vectors were transformed into Agrobacterium tumefaciens strain EHA105 via the freeze–thaw method, then transformed into soybean cultivar W82 according to an established protocol (Li et al., 2017). The GmHSFA6OE and pGmSPL2b::GUS vectors were transformed into Arabidopsis according to the Agrobacterium-mediated floral dip method (Clough and Bent, 1998). All primers used for PCR are listed in Table S9.

Subcellular localization and transcriptional activation analysis

The full-length CDS of GmSPL2b minus the stop codon was cloned from its pMD-19 T vector and then inserted into pCAMBIA3301–GFP vector (with BglII linearized) using a one-step cloning kit (Vazyme, Nanjing, China) to generate the p35S::GmSPL2b–GFP expression vector. The pCAMBIA3301–GFP vector was used as the control. The fusion constructs or control plasmids were transformed into tobacco (Nicotiana benthamiana) leaves as described previously (Ding et al., 2020a). GFP fluorescence was visualized under a confocal laser scanning microsystem LSM780 (Carl Zeiss, Jena, Germany) with 488-nm excitation wave length.

Transcriptional activation analysis of GmSPL2b was performed according to previously reported method (Ma et al., 2021). First, the full-length CDS of GmSPL2b was cloned from cDNA of soybean cultivar N4608 and then inserted into pGBKT7 vector (with EcoRI and BamHI linearized) using a one-step cloning kit (Vazyme, Nanjing, China), yielding a BD-GmSPL2b. Second, the recombinant plasmid was transferred to the Y2H strain and cultured on SD/−Trp medium, and then selected on SD/−Trp/−His/−Ade medium for transcriptional activation analysis. The pGBKT7 vector was used as the control and all primers used for PCR are listed in Table S9.

Male fertility investigation

To observe the pollen fertility of soybean and Arabidopsis, the anthers of unopened flowers (the flowers that will open the next morning) in the afternoon were taken and stained with a 1% I₂-KI solution (Nie et al., 2019) and Alexander's staining solution (Ding et al., 2020a), respectively, and then observed under an Olympus CX31 microscope (Tokyo, Japan). For anther morphology observation, anthers from opened flowers of soybean and Arabidopsis were randomly selected in the morning between 8:00 and 10:00 and observed under an Olympus CX31 microscope (Tokyo, Japan). To histologically analyse the anther response of miR156bOE and W82 to HT, the unopened flowers (the flowers that will open the next morning) in the afternoon were fixed, dehydrated, embedded, sectioned, stained according to a previous report (Yang et al., 1998), then observed under an Olympus CX31 microscope (Tokyo, Japan) and photographed with a digital colour camera system (Olympus DP27, Tokyo, Japan).

Pollen germination and HT treatment

For soybean pollen germination investigation, flowers were randomly selected from three plants in each genotype in the morning between 8:00 and 10:00. Flowers were air-dried for 1 h at room temperature, and the fresh pollen was then dusted on the modified pollen germination medium (PGM) in vitro (Djanaguiraman et al., 2019). The PGM was prepared with 15 g sucrose, 0.03 g calcium nitrate, 0.04 g boric acid and 0.6 g agar in 100 mL of deionized water and boiled for 10 min, and then poured into culture dishes (Koti et al., 2004; Salem et al., 2007). The pollen was dusted on the PGM and the plates were then covered and incubated in an illuminated incubator (RXZ-430D, Ningbojiangnan, Ningbo, China) at 30 °C for 0.5 h. Pollen grain was considered germinated when its tube length was greater than the diameter of the pollen grain (Prasad et al., 2006) and observed using an OLYMPUS CX31 microscope (Tokyo, Japan). For HT stress treatment, pollen in PGM was incubated at 40 °C for 0.5 h (Djanaguiraman et al., 2019).

RNA-seq and data analysis

Total RNAs from the flower buds of miR156bOE-1NT, W82NT, miR156bOE-1HT and W82HT (three independent biological replicates for each genotype) were isolated using the TRIzol reagent (Invitrogen, Carlsbad, CA, United States) according to the manufacturer's introduction. Library construction and RNA sequencing were performed by MetWare (Wuhan, China). Clean data were remapped to the soybean Williams 82 reference genome (Wm82.a4.v1) by HISAT (v2.1.0, Kim et al., 2015). DESeq2 (v1.22.1, Love et al., 2014) was used to analyse the differential expression between the two groups, only genes with ¦Log₂FC¦ ≥ 1 and corrected P value ≤0.05 were identified as significant DEGs. DEG enrichment analysis mainly includes GO enrichment analysis and KEGG pathway enrichment.

RNA isolation and qRT-PCR analysis

Total RNAs from miR156bOENT, W82NT, miR156bOEHT, W82HT and gmspl2b mutants in soybean and GmHSFA6bOE lines and WT in Arabidopsis were isolated using the TRIzol reagent (Invitrogen, Carlsbad, CA, United States) according to the manufacturer's introduction. Total RNA from the same soybean samples that constructed the cDNA library was used for the validation of RNA-seq. The first-strand complementary cDNA of miRNA and mRNA was generated using the miRNA 1st-Strand cDNA Synthesis Kit (Vazyme, Nanjing, China) and HiScript Q RT SuperMix for qPCR Kit (+gDNA wiper, Vazyme, Nanjing, China), respectively. The qRT-PCR analysis of miRNA and mRNA and internal control genes setting for soybean and Arabidopsis were carried out according to the method described in our previous studies (Ding et al., 2021a, 2021b). All reactions were run with three independent biological replicates, and the relative gene expression levels were calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001). Primers used for qRT-PCR analysis are listed in Table S9.

Metabolomic analysis

The sample used in the metabolomic analysis was the same as the sample used for RNA-seq. Metabolite extraction was conducted according to the manufacturer's instruction (Wuhan MetWare Biotechnology Co., Ltd., Wuhan, China), and metabolomic analysis was carried out using UPLC-MS/MS (UPLC: Shim-pack UFLC SHIMADZU CBM30A, Kyoto, Japan; MS/MS: Applied Biosystems 4500 Q TRAP, Foster City, CA, USA) according to the method described in our previous study (Ding et al., 2019a). Three independent biological replications were performed. The identified metabolites were annotated using the metware database (http://www.metware.cn). Differential metabolites between groups were determined by ¦Log₂FC¦ ≥ 1 and VIP ≥1.

Y1H assay

The full-length CDS of GmSPL2b was cloned from its pMD-19T vector and inserted into the pGADT7 plasmid to generate the effector vector pGADT7–GmSPL2b. The F1–F4 fragment containing GTAC motif from the GmHSFA6b promoter region was integrated into the pAbAi plasmid to generate a reporter vector pAbAi–pGmHSFA6bF1 to F4, respectively. The assay was conducted according to the previously reported methods (Ding et al., 2020a; Li et al., 2021). The pAbAi–pGmHSFA6bF1 to F4 was first introduced into the Y1H gold yeast (Clontech, Dalian, China) and cultured on SD/−Ura/Aureobasidin A (AbA) medium for self-activating detection. After that, the pGADT7–GmSPL2b and negative control vectors were introduced into the bait-reporter strain and cultured on SD/−Ura/−Leu/AbA for Y1H assay to examine the interaction of GmSPL2b and the GmHSFA6b promoter fragment. Primers used in Y1H assay are listed in Table S9.

LUC assay

The LUC assay was performed basically following the protocols described by Dai et al. (2018) and Wang et al. (2020). The plasmid p35S::GmSPL2b was transformed into Agrobacterium tumefaciens strain GV3101 act as an effector. The promoter sequence of GmHSFA6b was recombined into the pGreenII 0800-LUC plasmid and subsequently co-transformed with the helper plasmid pSoup19 into GV3101 act as the reporter. The pCAMBIA3301-26 plasmid was used as negative control. The reporter strain was mixed with the effectors strain harbouring p35S::GmSPL2b or pCAMBIA3301-26 at a ratio of 1:1 and cotransformed into tobacco (Nicotiana tabacum) leaves (Dai et al., 2018). The dual-luciferase reporter assay system (Promega, E1910, Wisconsin, USA) was used for luminescence detection. Primers used in LUC assay are listed in Table S9.

GUS and NBT staining

GUS staining of inflorescence from the pGmSPL2b::GUS transgenic Arabidopsis material was performed as Jefferson et al. (1987) described. ROS content was observed in the inflorescence of Arabidopsis using NBT staining under NT and HT conditions according to the previously reported method (Kumar et al., 2014). All staining samples were observed under an OLYMPUS SZ61 stereomicroscope (Tokyo, Japan) and photographed with a digital colour camera system (Olympus DP27, Tokyo, Japan).

Fluorescence detection of ROS in soybean and Arabidopsis

The fluorescence detection of ROS in anther of soybean and inflorescence of Arabidopsis was performed using CM-H2DCFDA staining followed the protocols described by Muhlemann et al. (2018) and Zhang et al. (2021). Detached soybean flower buds (the flowers that will open the next morning after removed sepals and petals) and Arabidopsis inflorescence were subjected to HT stress for 1 day and 4.5 h, respectively, and then the samples were harvested and placed in a 2.0 mL centrifuge tube containing CM-H2DCFDA staining solution (10 mmol/L Tris–HCl, 50 mmol/L KCl, pH7.0; 5 μM CM-H2DCFDA and 0.005% (v/v) Triton X-100). Soybean flower buds or Arabidopsis inflorescence under NT condition were used as control. The samples were incubated in the dark (28 °C) for 20 min and then washed twice with fresh solution (10 mmol/L Tris–HCl, 50 mmol/L KCl, pH7.0). The stained soybean anthers (removed from the flower bud using tweezer) and Arabidopsis inflorescence were placed on a microscope slide. Fluorescence was visualized under a stereo fluorescence microscope (LEICA M165 FC, Germany). The quantification of ROS fluorescence intensity in soybean anther was performed using Fiji (Schindelin et al., 2012).

Statistical analysis

Data were shown with the mean ± SD from three independent experiments, and SAS was used for data analysis. One-way ANOVA and Duncan's test and Student's t-test were used to identify significant differences.

Accession numbers

The RNA-seq data generated as part of this study can be found in the NCBI database using accession number PRJNA935333.