The mulberry (Morus alba L.) and Nicotiana benthamiana seeds were taken from the State Key Laboratory of Resource Insects, Southwest University, Chongqing, China. In a phytotron, mulberry and tobacco seedlings were grown under specific conditions. They received 16 h of light and 8 h of darkness per day, the temperature was maintained at 24 ± 2°C, and the relative humidity was kept at 55%. The mulberries and fruit stalks were taken from the mulberry garden of Southwest University. When we shook the branches vigorously, the mulberries were divided into four categories according to their falling condition and maturity. Including normal mature fruit stalk (MN), mature easy-to-drop fruit stalk (MD), normal young fruit stalk (YN), and easy-to-drop young fruit stalk (YD). At the 2-leaf stage, mulberry seedlings were sprayed with 100 mM ABA and 10 mg/L 2,4-D (auxin hormone regulator). The control group was treated with water. The leaf samples were collected at 0, 1, 3, 6, 12, and 24 h, respectively, and immediately frozen in liquid nitrogen. All samples were subjected to three biological replicates. Mulberry variety “Hongguo No.2” stem was used for lignin staining and paraffin sections.

The M. alba genome database (version ASM1206604v3) was downloaded from the NCBI database (https://www.ncbi.nlm.nih.gov/). After this, the GH9 family (PF00759) domain sequence file was retrieved from the Pfam database (Finn et al., 2008) (http://pfam-legacy.xfam.org/) by searching for the keyword “Glyco_hydro_9”. Using this sequence file, we constructed a local database of white mulberry and searched for candidate members of the GH9 gene family using blastp. The presence of complete GH9 domain was checked using hmmer (https://www.ebi.ac.uk/Tools/hmmer/) and NCBI-CDD (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). Redundant and incomplete sequences were removed manually. Moreover, isoelectric point (PI) and molecular weight (MW) of M. alba GH9 proteins were accessed using the ExPASy database (http://www.expasy.org/tools/). The in silico analysis of subcellular location, signal peptides, and transmembrane domains was conducted using the Plant-mPLoc server (http://www.csbio.sjtu.edu.cn/bioinf/plant-multi/), SignalP 5.0 (https://services.healthtech.dtu.dk/services/SignalP-5.0/), and TMHMM 2.0 (https://services.healthtech.dtu.dk/services/TMHMM-2.0/), respectively. Sequences of GH9 members of poplar (Populus L.) and tomato (Solanum lycopersicum L.) were downloaded from the NCBI (https://www.ncbi.nlm.nih.gov/). Sequences of GH9 members of Arabidopsis were downloaded from the Arabidopsis thaliana database (https://www.arabidopsis.org/index.jsp).

The internal Muscle program of MEGA version 11.0 software (Tamura et al., 2021) was used to perform multiple sequence alignment of the identified M. alba GH9 protein sequences, and a non-root phylogenetic tree with 1,000 bootstrap repeats was constructed using the neighbor-joining (NJ) method. Similarly, another phylogenetic tree was generated among GH9 proteins of M. alba, A. thaliana, Solanum lycopersicum, and Populus tremula for studying phylogenetic relationships. The 2,000-bp upstream promoter sequence of each gene was analyzed using the PlantCARE (Rombauts et al., 1999) (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) website for cis-acting regulatory element analysis. Conserved motifs were analyzed using the Motif Elicitation program (http://meme-suite.org/tools/MEME), searching up to 10, and the number of repetitions was arbitrary (Bailey et al., 2009). For the exon–intron structure analysis of the mulberry GH9 coding sequence, their corresponding DNA sequence was used. Gene conserved structure information was obtained from NCBI-CDD (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi), and then TBtools (Chen et al., 2020) was used to map the gene structure. Gene structure information and chromosome location information (Jiao et al., 2020) were extracted from genome structure annotation files and displayed with TBtools software. Collinearity analysis was performed between mulberry and other species, and the results were plotted using TBtools with MCScanX (Wang et al., 2012).

Total RNA was isolated from all collected samples using Trizol reagent (Invitrogen, USA) according to the manufacturer ‘s protocol. Extracted RNA (1 μg) was reverse transcribed into cDNA using the PrimeScript™ RT kit with gDNA Eraser (TaKaRa, Japan). Primer Premier 5.0 was used to design the specific primers and Primer Blast (https://blast.ncbi.nlm.nih.gov/Blast.cgi) program was used to further verify the specificity of all primers in the NCBI database. qPCR was performed on an ABI Quant studio 3 RealTime system (Applied Biosystems, USA) using SYBR ^® Green RealTime PCR Master Mix (Novoprotein, China) with the following reaction program: denaturation (95°C 30 s), amplification and quantification (95°C 3 s, 60°C 30 s, 40 cycles), and melting curve analysis (60–95°C, heating rate 0.3°C/s). Actin gene was used as internal reference. The relative expression levels of genes were calculated by the 2^−ΔΔCT method. Statistical analyses were performed using Student’s t-test (t-test), with three technical replicates established for each qPCR method. Relevant figures were drawn using TBtools and prism 8.0.

The cDNA sequence of M. alba GH9B6, without a stop codon, was inserted into the 2300-YFP vector using homologous recombination. The resulting construct was then transformed into Agrobacterium tumefaciens GV3101. Two sets of A. tumefaciens were used to infect the leaves of 1-month-old N. benthamiana seedlings: one containing the vector with the fused protein sequences, and the other containing an empty vector (positive control). After 48 h, the infected leaves were observed for fluorescence using a Zen 2 (Olympus) confocal microscope.

To prepare the sample, the base of the fruit stalk of the mulberry was cut off near the bark. A 2-cm section of the fruit stalk tissue was manually cut longitudinally. The resulting slice was then positioned on a glass slide, and phloroglucinol staining solution was applied. After staining for 3 min, 25% concentrated hydrochloric acid was added dropwise for 2 min. Subsequently, the sample was examined and photographed under a ZEISS stereomicroscope. For GUS gene expression, the 1,689-bp upstream promoter sequence of MaGH9B6 was cloned into the PVCT024 vector using homologous recombination. The leaves of 1-month-old N. benthamiana seedlings were infected with Agrobacterium containing the vector with GUS coding sequence under the control of the MaGH9B6 promoter, and the Agrobacterium containing the GUS coding sequence under the 35S promoter (positive control). Wild-type N. benthamiana leaves without infection were used as a negative control. Treatments consisted of 24 h exposition to dark, 24 h exposition to light, spraying of solutions containing 100 mM ABA, 10 mg/L 2,4-D, 100 μM gibberellic acid (GA3), 100 mg/L ethylene, and water (control). After 48 h, take 1 cm² of infected leaves with a blade for staining with 40 mM X-gluc (0.2 mL of 0.5 M EDTA, 0.3 mL of 40 mM X-Gluc, 0.1 mL of 10% Triton X-100, and 1 mL of 1 M sodium hydrogen phosphate buffer) dye at 37°C under vacuum for 24 h and then decolored with alcohol. We selected evenly stained leaves for observation and repeated the procedure several times to get the results. The samples were observed and photographed under a stereomicroscope (ZEISS, German).

Samples of the mulberry stalk (5 mm length) were fixed in 4% paraformaldehyde with 1% glutaraldehyde at 4°C for 24 h. Samples were then dehydrated in an ethanol series and embedded in paraffin prior to cutting 10-μm sections. Sections were stained for morphological observation using 1% (w/v) Safranin O (Amresco, Solon, USA) and 1% (w/v) Fast Green FCF (Merck, Overijse, Belgium) (Zou et al., 2011). Then, they were observed and photographed under a stereomicroscope (ZEISS, German).

With the downloaded Glyco_hydro_9 (PF00759) and the reported GH9 as two queries, the M. alba protein database was established, and nine non-redundant sequences were finally determined by local blast search ( Supplementary Table 1 ). Based on the affinity with other species, GH9 members were named MaGH9A1, MaGH9A2, and MaGH9B1 to MaGH9B7. All GH9 family members were randomly distributed on eight chromosomes and contain 495–1,114 amino acid residues with a molecular weight of 54.52–123.69 kDa. MaGH9B4 had two conserved domains, Chalcone_2 (PF16035.8) and Glyco_hydro_9 (PF00759). Two members (GH9A1 and GH9A2) had transmembrane domains of 75–97 and 77–98 amino acids respectively. However, all members of the GH9B subfamily had signal peptides of 23–34 amino acids except MaGH9B4. This is consistent with the prediction results of previous studies on the GH9 family (Urbanowicz et al., 2007).

A phylogenetic tree was constructed among M. alba GH9 members for evolutionary history analysis. Based on evolutionary relationships, MaGH9 members can be divided into two subgroups: A and B. There are two and seven members in group A and B, respectively.

According to the phylogenetic tree among Arabidopsis, poplar, tomato, and mulberry GH9 members, there were more genes in subgroup B ( Figure 1 ). There were 11, 49, and 7 genes in subgroups A, B, and C, respectively. MaGH9 members were more closely related to the members of P. tremula GH9 genes. However, the number of GH9 family members in Populus was nearly triple that of M. alba. This could be attributed to variations in the evolution of GH9 proteins in the two species, as well as differences in the evolution of their chromosomes.

Gene structure analysis showed that the number of exons in the GH9 family ranges from 4 to 21, and most of the glycoside hydrolase structural domains have four to nine exons. Members of subfamily B exhibit a significant variation in gene structure, which may correspond to their complex and diverse functions. The structural domain of glycoside hydrolase is distributed over four to nine exons in subfamily B. On the other hand, subgroup A glycoside hydrolases mainly consist of five exons that distribute the structural domains ( Figures 2B, C ). Motif analysis showed that all the 10 identified conserved motifs were found only in the glycoside hydrolase domain, and these motifs were highly similar in the A and B subfamilies. This indicates that the glycoside hydrolase domain is very consistent in M. alba ( Figure 2A ). The N-terminal members of subfamily A have transmembrane domains as they are membrane-bound proteins. On the other hand, six members of subfamily B are predicted to have signal peptides, which suggests that they may be secretory proteins.

Several hormone (salicylic acid, methyl jasmonate, gibberellin, ABA, and auxin) and stress-related cis-elements were predicted ( Supplementary Table 2 ). The cis-elements were unevenly distributed in different genes ( Figure 3 ). These results suggest that mulberry GH9 members might have critical roles in hormone stress signaling pathways.

Cellulase has an adsorption effect on lignin and can effectively degrade lignin content (Lan et al., 2020; Zhao et al., 2022). We stained the fruit stalks with phloroglucinol at five stages of mulberry fruit ripening such as green, green to red, red, red to purple, and purple, which were named S1, S2, S3, S4, and S5, respectively ( Figure 4F ). The lignin part in the fruit stalk was stained red, and the lignin-free part was not stained ( Figures 4A–E ). Lignin in the AZ of the carpopodium appeared as a continuous strip in the S1 and S2 stages. From the S3 stage, the lignin content in the AZ of the carpopodium decreased, forming a phenomenon of lignin deposition far away from the AZs. These findings indicate that the variation in lignin content signifies the developmental stages of the AZ, a process intricately linked to plant flower and fruit drop and closely associated with the GH9 family.

The observation of AZ of fruit stalks of mulberry “Hongguo No.2” at S1, S3, and S5 stages by paraffin section showed that there was no AZ in the carpopodium at the S1 stage and a layer of dense cell is present. From S3, the AZ began to appear from the cortical layer and extended to the medulla, and the color of the cells in the AZ began to darken and shrink, forming a scar in the AZ. At S5, the AZ formed a penetrating structure and the cells began to rupture ( Figure 5 ).

We examined the expression patterns of GH9 genes at various fruit developmental stages, including normal mature fruit stalk (MN), mature easy-to-drop fruit stalk (MD), normal young fruit stalk (YN), and easy-to-drop young fruit stalk (YD). We studied the expression patterns of GH9 gene at these stages. The primers used are listed in Supplementary Table 3 . The results displayed significant differences in the expression levels of MaGH9A1, MaGH9B1, MaGH9B5, MaGH9B6, MaGH9B3, and MaGH9B7. Moreover, MaGH9B1, MaGH9B5, MaGH9B6, and MaGH9B3 exhibited significantly higher expression levels in MD stalks compared to those in MN during the same period. Additionally, the expression level of MaGH9A1, MaGH9B6, and MaGH9B7 in YD was significantly higher compared to that in YN ( Supplementary Figure 1 ). Understanding the formation of the AZ is important in exploring the mechanisms behind plant organ abscission. Our investigation into the subcellular localization of MaGH9s, combined with previous research indicating that cellulase genes are predominantly localized to the cell wall and cell membrane, contributes to a comprehensive understanding of these processes (Yan and Wu, 2013; Li et al., 2019b). These results predict the role of this family in the cell wall and related different biological pathways. Based on the results, MaGH9B6 was selected for further study and cloned. It was transiently expressed in the epidermal cells of N. benthamiana leaves, and the fluorescence signal was observed. The results showed that the empty vector-controlled GFP signal was seen in both the nucleus and cytoplasm, and MaGH9B6-controlled GFP signals accumulate in the cell wall ( Figure 6 ). This finding supports the potential function of MaGH9B6 in cell wall degradation.

In silico promoter analysis of all genes predicted their involvement in multiple hormone and stress signaling pathways. To get a better understanding, the 1,689-bp upstream sequence of MaGH9B6 was cloned. X-gluc was used as colorimetric substrate. Transient expression of GUS gene in N. benthamiana leaf epidermal cells was assessed through staining after treatment with water, light, darkness, ABA, 2,4-D, ethylene and gibberellin, respectively. Three replications were performed and the Cauliflower mosaic virus (CaMV) 35S promoter was used as a positive control and those injected with injection buffer were used as a negative control ( Figure 7A ). As compared to the water-treated group, groups following light-treated, gibberellin, ABA, and 2,4-D treatments were all stained in shades of blue. These results are consistent with in silico analysis of the MaGH9B6 promoter ( Figure 7B ). The MaGH9B6 promoter showed more response to ABA and 2,4-D, suggesting its response to these hormones.

Furthermore, in 2-month-old mulberry seedlings following 100 mM ABA and 10 mg/L 2,4-D, the quantitative analysis of mulberry leaves showed that MaGH9B6 expression began to increase significantly from 3 h after ABA stress and continued to increase until 24 h compared to the control. Under 2,4-D stress treatment, MaGH9B6 expression also increased significantly after 3 h, but the expression level decreased significantly at 24 h compared to the control ( Figure 7C ).