Background

As an important role in the process of histone methylation, the SET domain gene family is widely distributed in eukaryotes. The name of SET was taken from the three regulatory factors containing the SET domain, including Su (var) 3–9 [1], Enhancer of zeste [2], and Triohoarx [3], which were first found in Drosophila melanogaster. SET domain gene (SDG) family could offer the activity of histone methyltransferase in the plant, which regulates plant histone lysine methylation. The SDG proteins are a specific catalytic site of histone lysine methyltransferase and are highly conserved in evolution [4].

The SDG family plays significant roles in the regulation of various biological processes, such as plant growth and development [5, 6]. The first reported SDGs in plants were the CLF (CURLYLEAF) and MEA (MEDEA) genes in Arabidopsis [7, 8], which could stably repress the floral homeotic gene and encoded a protein with homology to the product of the polycomb-group gene, enhancer of zeste, ensuring the stable inheritance of expression patterns through cell division and regulating the control of cell proliferation. In Arabidopsis, TRITHORAX1 (ATX1, named SDG27 in this study) has been shown to regulate floral organ development. AtATX1 directly binds the active FLC locus before flowering, and this interaction is released upon the transition to flowering [9]. Meanwhile, AtSDG40 affects flowering time by epigenetically modulated FLC expression. The decreased levels of H3K4me3 and increased H3K27me3 in FLC chromatin led to earlier flowering in AtSDG40 mutants [10]. In rice, OsSDG701 encoded an H3K4-methyltransferase that enhanced the expression of rice Hd3a and RFT1 florigen, and promoted rice flowering under either long-day or short-day conditions [11].

In addition, SDG proteins have also been found to participate in response to external cues, including both abiotic and biotic stresses [12]. The chromatin modifier AtATX1 participated in drought signal transduction in Arabidopsis [13]. At the same time, AtATX1 affected the genes regulating the ABA pathway by up-regulating the expression of NCED3 to improve drought stress tolerance [14]. Low temperature significantly affected the expression of DcSUVH5a, DcATXR5a, and DcSUVR14a, and the expression of DcASHR3, DcSUVR3, DcATXR4, DcATXR5b, and DcSDG49 genes were affected by drought stress in orchids [15]. In Arabidopsis, the histone methyltransferase AtSDG8 defends against fungal pathogens in plants by regulating a subset of genes in the jasmonic acid (JA) and ethylene signaling pathways [16]. These findings indicate that the SDGs could alter the histone methylation signal and affect the growth and adaptation of plants.

The identification and functional analysis of SDG gene family members has been deployed in many plant species, including Arabidopsis thaliana [17], Oryza sativa [18], Zea mays [19], Gossypium raimondii [20], Brassica napus [21], and Triticum aestivum [22] with 49, 43, 43, 52, 122, and 166 members, respectively. However, the study on SDG genes in upland cotton is still limited. In this study, a comprehensive identification of SDG proteins was conducted in four Gossypium species. This involved analyzing various aspects such as the analysis of the phylogenetic tree, chromosomal distribution, evolutionary, gene structure, domain organization, and cis-elements. In addition, the expression of SDG genes during plant growth and development and the response to various abiotic stresses were investigated. We further performed the qRT-PCR assays to verify the GhSDGs in response to abiotic stresses and hormonal treatments. In addition, the WGNCA analysis showed that GhSDG51 co-expressed with many genes that are associated with salt stress. The further VIGS assays showed that the silencing of GhSDG51 could reduce the salt stress tolerance in cotton plants compared with the control. Taken together, this research will provide valuable insights for future investigations on the SDG protein in cotton.

Results

Identification of SDG family members

By scanning the SET domain, a total of 40, 40, 75, and 74 SDG genes were identified in G. arboreum, G. raimondii, G. barbadense, and G. hirsutum, respectively. These genes were subsequently named as GaSDG1–40, GrSDG1–40, GbSDG1–75, and GhSDG1–74 according to their chromosomal location. All of the members in the SDG family contained one SET domain, along with several other types of domains such as SAD, AWS, PWWP, MYDN finger, etc. (Fig. S1). The physical and chemical properties of the members of the SDG family members, including protein length, protein molecular weight (MWs), isoelectric point (pIs), and protein hydrophilicity and hydrophobicity were analyzed (Table S1). The distribution of values of the above physical and chemical characteristics in allotetraploid cotton and their diploid progenitors were approximately equal to each other (Fig. S2).

Phylogenetic analysis and structure of SDG family members

Combined with the identified SDG member in Arabidopsis thaliana and Oryza sativa, a total of 297 proteins were chosen to generate a phylogenetic tree (Fig. 1). As clearly shown in the phylogenetic tree, the SDG proteins can be divided into eight distinct subfamilies, namely subfamily I ~ VII. Each subfamily contains multiple SDG members from Gossypium species and at least one AtSDG member, indicating that there is no Gossypium-specific subfamily was found. Among the above subfamilies, the subfamily III contained the most members (30.97%), whereas the IV only contained seven members (2.35%). The member of each subfamily from G. hirsutum and G. barbadense was almost twice that of those from the diploid cotton.

Fig. 1
figure 1

Phylogenetic tree of SDG proteins from six plant species. A phylogenetic tree was constructed using FastTree based on 297 SDG proteins obtained from Arabidopsis thaliana, Oryza sativa, Gossypium hirsutum, Gossypium barbadense, Gossypium arboreum, and Gossypium raimondii. The SDG proteins of the six species are denoted by distinct symbols and colors, while the eight subfamilies are distinguished by varying background colors

Full size image

To explore the structure of SDG members, a total of 15 motifs were identified (Fig. 2). These motifs performed different distributed patterns within each subgroup. The motif 3 (HIGLFAKRDIPAGEELTYDYG, identified as a part of SET domain) was detected in all of the subfamilies, and the motif 10 (WTTERCAVCRWVEDWDYNKIJICNRCQIAVHQECYGARNVRD, identified as a part of PHD-finger domain) and motif 13 (WVHVTCAWFRPEVSFADDEKMEPALGILRIPSDSFVKICVICKQIHGSCT, identified as a part of PHD-zinc-finger like domain) belonged exclusively to subfamily V (Table S2). The motif 11 was only detected in subfamily III and VII. On the other hand, the gene structure (exon-intron structure) of the GhSDGs exhibited significant variations across different subfamilies. All of the members in subfamily I displayed the mono-exon structure, while the members in subfamily V consisted of 21 ~ 24 exons. These results suggest that the gene structure and protein architecture of GhSDGs are conserved within each specific subfamily, implying that they may possess functions, and are consistent with the classification of the phylogenetic analysis.

Fig. 2
figure 2

Conserved motifs and exon–intron structure analysis of GhSDG genes. A Conserved motifs of GhSDG proteins, the different motifs represented by boxes with different colors; B Exon–intron structures of GhSDG genes, the exons and introns are represented by green boxes and black lines, respectively. The scale at the bottom is used to infer the length of proteins and genes

Full size image

Gene duplication and selection pressure analysis of SDG members

To investigate the expansion pattern of SDG genes, we conducted a further analysis to identify the duplication pattern. Multiple types of duplication events were detected in cotton’s SDG genes. The expansion of the SDG genes in Gossypium species was primarily driven by the dispersed duplication event, with the subsequent whole genome duplication (WGD) event also playing a significant role (Fig. 3 and Table S3). These results provide an explanation for the observation that the number of SDG genes in diploid Gossypium species is only 1.18 times higher than in T.cacao, despite Gossypium species undergoing an additional round of WGD event compared to T.cacao (Fig. 3). A comprehensive search was conducted in an additional 35 plant species, resulting in the identification of a total of 1689 SDG genes across 39 species (Fig. 3). We found that SDGs were widespread in monocotyledonous and dicotyledonous plants. Among these species, the SDG genes in most species were mainly expanded by dispersed duplication, the proportion of this duplication varying from 46.5% ~ 92% (Fig. 3). However, the WGD and tandem duplication events were the major force in expanding the SDG genes in some cases, the 84.3 and 76% of SDG in Glycine max and Citrus sinensis were expanded by WGD and tandem duplication events, respectively (Fig. 3). These results indicated that the expansion of SDG genes in the majority of species is likely to be driven by dispersed events. However, in a few instances, the expansion of SDG genes may also be expanded by WGD and tandem duplication events.

Fig. 3
figure 3

The phylogeny of 39 plants used in this study and the duplication events analysis of identified SDG proteins. The right side shows the amount and the duplication types of SDG protein detected in each plant. The order of tree branches depicted on the left side was retrieved from the Timetree database (http://www.timetree.org/)

Full size image

All the Ka/Ks ratios of gene pairs of SDG genes in Gossypium species, resulting from various duplication events, were found to be lower than 1. This observation suggests the SDG genes might experience strong purifying selection pressure during evolution and suggests that the protein functions may be conserved after the expansion (Table S4).

The allopolyploidization event has been found to have a significant impact on the expansion of SDG genes in cotton. The number of members in the tetraploid cotton species was nearly double that of the diploid cotton species, whereas some SDG members might have been lost during the evolution. The SDG genes of the four Gossypium species exhibited an uneven distribution across nearly all of the chromosomes (Fig. 4 and S3; Table S1). For G. arboreum, the SDG genes were unevenly distributed on each chromosome, while the SDG genes of G. raimondii were distributed on 12 chromosomes except D02. In both the G. hirsutum and G. barbadense, SDG genes were absent on A01, A04, and D01. The distribution of SDG genes in the G. arboreum and At subgenome of tetraploid cotton species exhibited a high level of consistency. However, the distribution of G. raimondii showed slight differences compared to the Dt subgenome of tetraploid cotton species (Fig. 4 and S3). We also performed a collinearity analysis of the SDG genes between the Gossypium species and Arabidopsis. The detected collinearity gene pairs account for less than half of the SDG genes, indicating that these genes may have been conserved from their common ancestor and the rest of the genes might be duplicated in different evolutionary events (Fig. S4).

Fig. 4
figure 4

Chromosome distribution and collinearity of duplicated gene pairs of SDG genes. Chromosomes of the Gossypium arboreum, Gossypium raimondii, and the At and Dt subgenomes of Gossypium hirsutum are represented by differently colored boxes. Collinear gene pairs between GhSDG genes and the SDG genes of the diploid species are represented by blue and orange lines

Full size image

Cis-elements analysis of GhSDGs

The cis-element related to the stress response, light response, growth and development were detected in the promoter region (2000 bp upstream of the start codon) of GhSDGs. The results displayed that the MYC, MYB, and STRE elements were enriched in the promoter of GhSDGs, indicating that these GhSDGs are likely to play a crucial role in the response to environmental stress (Fig. S5). Therefore, we speculated that GhSDGs might play a vital role in regulating the defense against the external environment.

The expression pattern of GhSDGs

According to the analysis of the RNA-seq data, the GhSDGs exhibited distinct expression patterns across various tissues. The GhSDG34, 20, 8, and 49 were expressed relatively highly in all tissues, whereas the GhSDG43, 14, 35, and 60 were rarely expressed in all tissues (Fig. S6). GhSDG13 exhibits a specific and highly expressed pattern in both the leaf and root, while GhSDG72 shows exclusive high expression specifically in the root. These results indicate that GhSDGs might play different roles in the growth and development in upland cotton.

The biological function of a specific gene is closely related to the spatiotemporal expression of the transcript. Based on the results of RNA-seq data (Fig. S7), we chose 6 genes to performed qRT–PCR analysis to investigate their response under the NaCl treatments. The results showed that the under the salt treatment, the expression of GhSDGs were down-regulated after the salt treatment at 2 h and 4 h, while their expression rebound at the 6 h ~ 24 h. The expression of GhSDG20 reached the peak at 24 h, while the expression of GhSDG15 and GhSDG67 continuous decreasing after 6 h. The expression of GhSDG51, GhSDG57, GhSDG57, and GhSDG67 decreased at 12 h but increased at 24 h (Fig. 5). These results indicating that GhSDGs could respond to the salt stress.

Fig. 5
figure 5

The relative expression of GhSDGs under the salt treatment. GhActin7 was used as an internal reference. The primers used are listed in Table S5. Error bars represent the standard deviation of three independent biological replicates

Full size image

WGCNA network analysis predict GhSDG51 related to salt stress

We have additionally discovered that GhSDG51 acted as a hub gene within a weighted gene co-expression network analysis (WGCNA) network that is associated with salt stress. In this network, GhSDG51 exhibited co-expression with several transcription factors, including GhSRG1, GhYPTM2, GhPHOT2, GhCYP76C3, GhRAN3 and GhTCP7 (Fig. 6). We further found that these genes could respond to salt stress by qRT-PCR assays. The expression levels of GhYPTM2, GhPHOT2, GhCYP76C3 and GhRAN3 showed an overall upward trend after salt stress. The expression of GhSRG1 increased at 2 h, 4 h, and 6 h after salt treatment, and the expression of GhTCP7 increased at 12 h and 24 h after salt treatment. Therefore, it was hypothesized that GhSDG51 might play a role in the response to salt stress.

Fig. 6
figure 6

Co-expression network and qRT-PCR analysis. A The co-expression network of GhSDG51 in salt stress. The information of genes displayed in the network were listed in Table S6. The raw RNA-seq data (seedlings under the salt treatment for 0, 3, 6, 9, 12, and 24 h) was retrieved from a previous study (accession number: PRJNA490626). B The relative expression of the genes that co-expressed with GhSDG51. GhActin7 was used as an internal reference. The primers used are listed in Table S5. Error bars represent the standard deviation of three independent biological replicates

Full size image

Silence GhSDG51 reduces tolerance to salt stress

To further explore the function of GhSDG51 under salt stress, we silenced the GhSDG51gene in cotton by VIGS assay. After approximately 2 weeks following the injection, the TRV: GhCLA1 plants exhibited an albino phenotype (Fig. 7B), indicating that the VIGS experiment was successful. Furthermore, the qRT-PCR experiment revealed a significant decrease in the expression level of GhSDG51 in TRV: GhSDG51 plants compared to TRV: 00 plants (Fig. 7C), indicating that this gene expression was successfully silenced. After the salt treatment with 400 mM NaCl for 2 days, it was observed that the GhSDG51-silenced plants exhibited noticeable wilting in comparison to the control group (Fig. 7A). To further investigate the impacts of plant physiological and biochemical characteristics, the content of proline (Pro) and malondialdehyde (MDA) were measured in TRV: GhSDG51 and TRV: 00 plants. Before treatment, the content of MDA and proline in TRV: GhSDG51 and TRV: 00 plants exhibited similarity. After the treatment, the content of MDA was found to be significantly higher in TRV: GhSDG51 plants compared to TRV: 00 plants. Conversely, the content of proline in TRV: GhSDG51 was significantly lower than that in TRV: 00 plants (Fig. 7D, E). These results indicated that the salinity tolerance of cotton plants was reduced following the silencing of GhSDG51. Moreover, it is proposed that GhSDG51 may play a crucial role in positively regulating the response of cotton to salt stress.

Fig. 7
figure 7

Silencing GhSDG51 reduced tolerance to salt stress. A Phenotype of empty control (TRV: 00) and VIGS plants (TRV: GhSGD51) before and under 400 mM NaCl treatment (TRV: GhSGD51); B Positive control (TRV: CLA1); C The expression level of GhSDG51 in blank control and VIGS plants; D The malonaldehyde (MDA) contents of blank control and VIGS plants; E The proline (Pro) contents of blank control and VIGS plants. The data measurement was tested with a one-way analysis of variance (ANOVA) using the student’s t-test. ** and * indicate statistical significance at the 0.01 and 0.05 probability levels, respectively

Full size image

Discussion

Cotton, a significant economic crop known for its inherent abiotic resistance, is extensively cultivated worldwide. Abiotic stresses, such as salinity stress, have a significant impact on the growth of cotton and have led to limitations in its cultivation worldwide. The SET domain-containing proteins (SDGs) are known histone lysine methyltransferases and participate in several processes, including responding to both biotic and abiotic processes [23, 24]. Here, we conducted a comprehensive genome-wide investigation of SDGs in G.hirsutum and explored their potential in salt stress.

This study identified a total of 229 putative SDG genes from G. arboreum, G. raimondii, G. barbadense, and G. hirsutum. In both allotetraploid cotton species G. hirsutum and G. barbadense, the number of identified SDG genes was nearly equal to the combined number of SDG genes in diploid cotton varieties G. arboreum and G. raimondii. This trend was also observed within each subfamily. Further analysis of gene structure and protein sequences revealed that SDG gene members within each subfamily exhibited similar gene structures, sequence lengths, and motif distribution pattern. Therefore, our findings potentially support previous assertion that the allotetraploid Gossypium species are evolved and speciated by the hybridization of the diploid progenitors (Table S1, Fig. 1).

We further found that the dispersed duplication event is the main force in the expansion of the SDG genes in many plants, including the Gossypium species (52.5% ~ 58.67%), which is different from the duplication pattern on the whole-genome level (Table S3). In addition, compared with the diploid Gossypium species, most SDG genes in allotetraploid species were well conserved in the allotetraploidization events, a similar result was also found in the Brassica species in this study (Fig. 4, Fig. 3). Functional differentiation might occur in the duplicated genes, including partial or complete loss of the previous functions, gaining new functions, or maintaining the original functions [25]. And the selection pressure analysis showed that the SDG genes are under the purified selection during the multiple duplication events (Table S4). These results suggest that members of the SDG family may exhibit relatively conserved functions during the evolution.

Promoters contain cis-elements that are crucial for the specific expression and function of a gene during growth and development. Some critical stress and hormone-responsive cis-elements for the plant were found in the promoter region of almost all SDG genes (Fig. S5). Therefore, promoter analysis indicated that the SDG genes in cotton played roles in various stress responses. By performing the RT-PCR, many SDG genes were found to respond to the salt treatment, including GhSDG51. The WGCNA analysis further showed that GhSDG51 might act as a hub gene in a co-expression network related to salt stress. This network includes GhRAN3, GhYPTM2, GhSRG1, GhPHOT2, GhTCP7, and GhCYP76C3, which could participate in the process of plant cell activity, signal transduction, and responding to multiple abiotic stresses [26,27,28,29,30,31]. Next, the qRT-PCR assays indicated the genes that co-expressed with GhSDG51 were also up-regulated in salt stress (Fig. 6).

Based on the above results, we speculate that GhSDG51 is a candidate gene that may play a role in salt stress. Due to factors such as the long time and low efficiency of the cotton transgene, we performed VIGS assay to silence of GhSDG51 to study its functions because of the short experimental period, simple operation method, low cost, and high efficiency of this method. After salt treatment, the TRV: GhSDG51 plants wilted more leaves compared to control plants (Fig. 7A). Additionally, after salt treatment, the TRV: GhSDG51 plants possessed higher MDA, whereas the content of proline was lower than that in control plants (Fig. 7D, E). The proline regulates osmotic potential and scavenging ROS, the content of proline is positively correlated with the abiotic stress resistance of plants, including salt, drought, and high temperature stress [32, 33]. While MDA is the main product of membrane lipid peroxidation, and the content of MDA indirectly reflects the damage degree to plant cells [34]. Therefore, we conclude that GhSDG51 could play a positive role in salt stress. However, the precise mechanism of the function of GhSDG51 requires further study.

Conclusion

In this study, 229 cotton SDG genes were identified in G. arboreum, G. raimondii, G. hirsutum, and G. barbadense, and they could be divided into eight subfamilies. The dispersed duplication and allopolyploidization events played a vital role in the expansion of GhSDGs. The analysis of cis-elements and the qRT-PCR experiments demonstrated that GhSDGs might respond to abiotic stresses. The WGCNA analysis further showed GhSDG51 plays as a hub gene in a co-expression network. Finally, by performing the VIGS assay and the measurement of physiological traits, the GhSDG51 might play a positive role in salt tolerance. This study laid a foundation for further understanding of SDG proteins in cotton, and provided functional candidate genes related to salt stress.

Material and methods

Identification and characterization of SDG proteins

The HMMER software (version 3.3.2) was performed to screen the SDG protein sequences containing SET domain (PF00856) with default parameters in 39 selected plant species, the detailed information on these species is listed in Table S7 [35]. To further confirm that the candidate SDGs contain the SET domain, we employed the InterProScan program (version 5.51RC1–84.0) with the SMART database (SM00317) and PFAM database (PF00856) [36]. The protein properties of the SDGs were calculated by the ProtParam module in Biopython [37].

Phylogenetic analysis of SDG proteins

The SDG protein sequences retrieved from the A. thaliana, O. sativa, and four Gossypium species were aligned by performing the MAFFT software (version 7.310) [38]. The gaps in the alignment result were further removed by BMGE software [39]. The phylogenetic tree was constructed by the aligned protein sequences by FastTree (version 2.1.11) [40]. The visualization of the phylogenetic tree was generated by evolview (version 3.0) [41].

Gene structure, and motif analysis of GhSDGs

The information on gene structure (exon/intron) was retrieved from the genomic annotation (gtf/gff files). The SDG protein sequences were submitted to the MEME website to detect the motifs [42]. The PlantCARE website was used to detect the cis-elements in the promoter (2000 bp upstream genomic DNA sequences) of SDG genes [43]. The results were visualized in TBtools (version 0.1098765) software [44].

Analysis of gene collinearity and duplication events

The duplication event was identified by the duplicate_gene_classifier module in MCScanX [45]. The collinearity of the SGD genes was calculated by MCScanX and Dupgen_finder [45, 46]. The gene collinearity and the chromosome location of SDGs were visualized in Circos software (version 0.69–9). The ratio of nonsynonymous rate and synonymous rate (Ka/Ks) was calculated by ParaAT pipeline and Kaks_calculator (version 2.0) [47, 48].

Expression proflings and WGCNA analysis

The transcriptome data from a previous study were used in this study (Accessions: PRJNA490626) [49]. The analysis pipeline for generating FPKM (Fragments Per Kilobase of transcript per Million mapped reads) value and the analysis of WGCNA were described in our previous study [50].

RNA extraction and qRT-PCR analysis

The seeds of upland cotton (TM-1) were cultured in seedling blocks at 24 / 20 °C for 16 / 8 h. Three-week-old seedlings were treated selected with 200 mM sodium chloride. Three biologically replicated of true leaves were collected at 0, 2, 4, 6, 12, and 24 h and frozen in liquid nitrogen at − 80 °C for further analysis. Total RNA was isolated using the RNA prep Pure Plant Kit (TIARGEN, Beijing, China) and treated with DNase-I to remove genomic DNA. The quality and purity of the RNA were measured by a NanoDrop 2000 spectrophotometer (Thermo Scientific, Massachusetts, USA), and 1 μg of RNA was used to synthesize first-strand cDNA with the Transcriptor First Strand cDNA Synthesis Kit and oligo-dT primers at 42 °C for 60 min, followed by 72 °C for 10 min. Real-time PCR was performed in a 7500 Fast Real-Time PCR system (Applied Biosystems, State of California, USA) using SYBR Green Master Mix. Three biological replicates were performed for each cDNA sample, and each reaction was prepared in a total volume of 20 μL containing 10 μL of SYBR Green PCR Master Mix, 0.4 μL of each primer, 2 μL (200 ng) of diluted cDNA template, and 7.2 μL of nuclease-free H2O. The qRT-PCR was performed using the PCR System from Applied Biosystems. The primers for the qRT-PCR are listed in Table S5. Results were analyzed with the 2−∆∆Ct method [51].

VIGS of GhSDG51 in upland cotton

A specific fragment of GhSDG51 (301–600) was cloned and used to construct a pTRV2:GhSDG51 vector. Then, the recombinant vector plasmid was transformed into Agrobacterium tumefaciens LBA4404. The Agrobacterium tumefaciens carrying pYL156 (negative control, TRV: 00), pYL156-GhSDG51 (TRV: GhSDG51), pYL156-CLA1 (positive control, TRV: CLA1), and pYL192 (accessory vector) were cultured in 28 °C, 180 rpm shaker. When the OD600 of the bacterial liquid reached 1.8, the bacteria were resuspended to OD600 = 1.5 by using the buffer containing 10 mM MgCl2, 10 mM MES, and 200 μM acetosyringone. After being kept in the dark at 25 °C for 3 h, the Agrobacterium resuspended solution containing pTRV2: GhSDG51, pTRV2: 00, and pTRV2: GhCLA1 was mixed with pYL192 in a 1:1 ratio [52]. The mixture was then injected into the cotyledons of TM-1 seedlings. Finally, the injected cotton plants were placed in a dark incubator and moved to the culture room for regular cultivation after 24 h. After growing for about 2 weeks, when the leaves of positive plants exhibited an albino phenotype, the silencing efficiency was subsequently determined through qRT-PCR experiments. The TRV: GhSDG51 and TRV: 00 were irrigated with 400 mM NaCl to observe the phenotype. Cotton leaves were used to determine the content of MDA content and proline activity by using the malondialdehyde assay kit (Jiancheng, Nanjing, China) and proline (Pro) assay kit (Jiancheng, Nanjing, China), respectively.