The sweet potato genomic information was derived from I. batatas “Taizhong 6” genomic data provided by the Ipomoea genome hub (https://ipomoea-genome.org/). The MADS-box family HMM model files SRF-TF (PF00319) and K-box (PF01486) were downloaded from the Pfam database (https://pfam.xfam.org/). The whole genome MADS-box family genes of sweet potato were identified by HMMER software (Krejci et al., 2016), and the MADS-box genes were identified by the SMART website (http://smart.embl-heidel-berg.de/) again. The genes containing the MADS-box domain were named from IbMADS1 to IbMADS95, and the three genes matching the published sequences of IbMADS1 (Ku et al., 2008), IbMADS4 (Kim et al., 2002), and IbMADS79 (Kim et al., 2005) were still named as IbMADS1, IbMADS4, and IbMADS79, respectively. IbMBP1, IbMBP4, and IbMBP8 (Dong et al., 2018) were named as IbMADS47, IbMADS38, and IbMADS51, respectively, in this study. According to the sweet potato genome gff3 annotation information, GSDS online website was used (http://gsds.gao-lab.org/). The MADS-box gene structure of sweet potato was analyzed visually. TBtools (Chen et al., 2020) was used to plot the innovative prediction results visually. TBtools (Chen et al., 2020) was used for structure visualization. According to the annotation information of the sweet potato genome, the MADS-box genes structure of sweet potato was visualized using GSDS online website tool (http://gsds.gao-lab.org/).

The Arabidopsis MADS-box protein sequence was obtained from the Plant TFDB website (http://planttfdb.gao-lab.org/index.php) (Jin et al., 2017). I. trifida MADS-box proteins were obtained according to the research of Zhu et al. (Zhu et al., 2020). The MADS-box genes of sweet potato were divided into two categories, type I and type II, based on the phylogenetic analysis of the MADS gene with Arabidopsis. ClustalW was used to compare the two types of MADS-box proteins; a neighbor-joining (NJ) tree was constructed using bootstrap analysis with 1,000 replicates, the amino acid p-distance model, missing data treatment option set at partial deletion, and a site coverage cutoff of 60%; and MEGA-X (Kumar et al., 2018) was used to construct the NJ phylogenetic tree (Saitou and Nei, 1987). The parameters for constructing the evolutionary tree were as follows: Poisson model, uniform rates, Sam (homogeneous), and pairwise deletion. According to the classification method of Sheng et al. (2019), the MADS-box gene subfamily of sweet potato was categorized based on the MADS-box gene subfamily of Arabidopsis, using Evolview online tools (https://www.evolgenius.info/evolview/) to beautify the tree.

According to the annotation information of the sweet potato genome downloaded from the Ipomoea genome hub, the positions of 95 sweet potato MADS-box genes on chromosomes were obtained, and the chromosome mapping was performed using Map Chart (Voorrips, 2002) software. Origin 8.5 (https://www.originlab.com) was used to draw the distribution cake map of sweet potato MADS-box genes.

The sweet potato MADS-box protein sequence was submitted to MEME online tool (http://meme-suite.org/) (Bailey et al., 2009) for the motif prediction. The discovery number of Motif was set to be 8, the width of motif was 6–50, and other default settings were used. The MADS-box motif domain of sweet potato was visualized by TBtools.

According to the annotation information of the sweet potato genome and sweet potato genome sequence, referring to Sumiya (2021) and Zhang et al. (2021), the 2-kb upstream sequence of sweet potato MADS-box genes was submitted to PlantCare website (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) (Lescot et al., 2002) for cis-acting element prediction. The PlantCare analysis results were simplified, and TBtools was used for visualization.

According to the annotation information and the whole genome protein sequence of sweet potato, MCScanX (Wang et al., 2012) software was used to analyze the MADS-box collinearity of sweet potato, and Circos (Krzywinski et al., 2009) software was used for visualization.

The tuberous root (greater than 10 mm in diameter), pencil root (2–5 mm in diameter), fibrous root (diameter of less than 1.5 mm), flower (an incompletely opened bud), fruit (the peel is green and the fruit form full), and stem tip (1–2 rachises of the rolled leaf at the stem apex, excluding the rachis where the expanded leaf is located) of “Jishu 26” (I. batatas cv. Ji 26) cultivated in an open field for 100 days were selected. After rapid freezing in liquid nitrogen, with dry ice sent to Biomarker Technologies Company (Beijing, China, https://www.biocloud.net) for total RNA extraction, library construction, and transcriptome sequencing, raw data were uploaded to National Center for Biotechnology Information (NCBI) project PRJNA744414; the expression of sweet potato MADS-box genes was analyzed based on genome information and local blast; the FPKM value (Florea et al., 2013) was calculated to reflect the gene expression; and the calculation formula is as follows: FPKM = {cDNA Fragments\over{Mapped Fragments(Millions) * TranscriptLength(kb)}}. Pheatmap package and ggplot2 package (Ito and Murphy, 2013) of R Statistics (https://www.r-project.org) were used for data processing and visualization.

The tuberous root is the main harvest organ of sweet potato. According to the expression profile of the MADS-box gene in different parts, several genes highly expressed in tuberous root were selected for qRT-PCR expression verification. The double-stranded cDNA of fibrous root (F.R, sampling details were the same as 1.7), tuberous root (T.R, sampling details were the same at 1.7), leaf (L, expanded leaf 1–2 at the apex of stem), stem (S, 1–2 rachises of the expanded leaf at the stem apex, excluding the rachis where the rolled leaf is located), stem tip (S.T, sampling details were the same at 1.7), fruit (F.T, sampling details were the same at 1.7), and other tissues cultivated in open field for 100 days “Jishu 26” were used as templates; β-actin was used as a reference gene, using the CFX96™ quantitative PCR instrument to perform the fluorescent quantitative PCR. The reaction program was as follows: predenaturation at 95°C for 3 min, 95°C for 5 s, 60°C for 30 s, 40 cycles; and 95°C for 15 s, 60°C for 6 s, 95°C for 15 s. The reaction system was as follows: 2× universal SYBR green fast qPCR mix (ABclonal Technology, Wuhan, China) for 5 μl, ddH₂O for 3 μl, cDNA for 1 μl, forward primer for 0.5 μl, and reverse primer for 0.5 μl. Primers were designed using Primer premier 6 (Singh et al., 1998) based on the sequence of the amplified CDs of the MADS-box genes, and the primer sequences are listed in Supplementary Table S1. Three biological replicates and three technical replicates were set up, and relative gene expression was calculated by the 2^−ΔΔCt method (Livak and Schmitten, 2001). Data processing and analysis were performed using SPSS software, and column graphs were drawn using origin 8.5.

A total of 95 MADS-box genes were identified from the whole genome of sweet potato and named according to IbMADS1–IbMADS95 (Supplementary Table S2). According to the gene structure, the sweet potato MADS-box genes were divided into two categories: type I and type II (Figure 1). The different types of MADS-box gene structures of sweet potato have apparent differentiation. Most type I genes are less than 2 kb in length and have one to two long fragment exons. On the other hand, the type II gene is usually over 3 kb and contains more than six exons, and most exonic fragments are short. Smart protein structure prediction found that the sweet potato MADS-box gene contains other domains in addition to the MADS domain. Most type I genes contained variable numbers of low-complexity region (LCR) structures and coiled-coil region (CCR) structures. The 21 type II genes contained a K-box domain; the K-box domain was commonly found to be associated with SRF-type TFs, also the signature domain of type II MADS-box genes. And the type II genes without a K-box domain usually belong to the MIKC* subfamily.

The NJ phylogenetic tree of MADS-box proteins from I. batatas, I. trifida, and Arabidopsis were constructed by MEGA-X. Based on the subfamily classification of Arabidopsis MADS proteins, sweet potato type I MADS proteins were divided into three subfamilies: Mα, Mβ, and Mγ (Figure 2A). In comparison, type II MADS proteins were divided into 12 subfamilies (Figure 2B). The type I MADS proteins of sweet potato and Arabidopsis were cross-distributed in each subfamily. Among them, both Mα subfamily and Mβ subfamily had 27 proteins each, and Mγ subfamily was less abundant with 16 proteins, which was similar to the distribution of the Arabidopsis protein subfamily. However, the type I proteins of I. trifida were mainly concentrated in Mα subfamily, with only one gene each in the Mβ and Mγ subfamilies.

The type II gene of I. batatas is phylogenetically closer to that of I. trifida, as indicated by gene number and subfamily distribution. The type II gene of I. batatas did not appear in the FLC subfamily or TT16 subfamily, and other subfamilies were distributed, with one to six proteins varying in number. The proteins for I. trifida are mainly distributed in subfamilies such as SEP, PI, and SVP and show deletions in several subfamilies. Twenty-two type II proteins contain K-box region, while three sweet potato type II proteins have no K-box region, one of them belongs to the AP1 subfamily, and the other two belong to the MIKC* subfamily.

Chromosome localization of 95 MADS-box genes in sweet potato was performed (Figure 3). It was found that except LG3 chromosome, the distribution of MADS-box genes in sweet potato varied in quantity in the other 14 chromosomes. More than 10 MADS-box genes were distributed on LG7, LG8, LG11, and LG15 chromosomes. LG1, LG5, and LG12 chromosomes contain fewer MADS-box genes, only one to three. Mα subfamily is mainly distributed on chromosomes LG7, LG8, and LG15, with more than five genes. In contrast, the Mβ subfamily (blue) is more fragmented, with one to four genes distributed in all chromosomes except LG16 and LG12. Seven MADS-box genes on LG6 were Mγ subfamily, and the rest were sporadically distributed on LG4, LG9, LG10, LG12, and LG14. The MIKC subfamily genes are widely distributed, except for LG1, LG3, LG6, LG9, and LG12, which are more numerous on chromosome LG11, with seven genes belonging to the MIKC subfamily. Gene cluster exists for LG6, LG7, LG8, LG14, and LG15, especially in the interval 10413181 to 10505372 of LG15 containing nine MADS-box genes. Gene clusters were present on chromosomes LG6, LG7, LG8, LG14, and LG15. In particular, the interval 10413181 to 10505372 of LG15 contains nine MADS-box gene presence distributed centrally.

Conserve motif analysis of sweet potato MADS-protein sequence was performed by online MEME tool. There are some differences in the number of MADS-protein motif in sweet potato. Each protein contains two to six motifs (Figure 4). All genes have motif 1 domain encoding 28 amino acids, and more than 90 genes contain motif 2 and motif 5. The 51 amino acids encoded by these three motifs are the core motifs of MADS-box genes, which are related to the SRF-box region. In addition, the motif 3 domain also plays an important role, and more than 50 proteins contain highly conserved motif 3 domains encoding 29 amino acids, including 22 type II MADS proteins.

The 2-kb upstream CDS sequences were extracted from the promoter regions of 95 MADS-box genes of sweet potato. The cis-acting elements in the promoter region of sweet potato MADS-box were predicted by PlantCARE online tool. All sweet potato MADS-box genes contain three to 12 light-responsive elements, including Box 4, G-box, GT1-motif, and TCT-motif. In addition, the promoter regions of sweet potato MADS-box genes are mainly more than 200 MeJA-responsive elements, including CGTCA-motif and TGACG-motif. And ABA-responsive elements (ABRE), auxin-responsive elements (TGA-element and AuxRR-core), and about 100 MYB response elements are involved in drought inducibility and light-responsiveness. In addition, there are a small number of regulatory elements such as circadian control elements, meristem expression elements, salicylic acid-responsive elements, low-temperature-responsive elements, gibberellin-responsive elements, defense, and stress-responsive elements. They indicated that MADS-box genes could respond to various environmental changes and coordinate the average growth and development of sweet potato (Figure 5).

Based on the collinearity analysis by MCScanX, a large number of genes were found to be collinear relational (light pink and light green) between and within chromosomes in sweet potato. There are 19 sweet potato MADS-box genes with collinearity, of which 15 genes belonged to inter-chromosome segmental duplications and four genes belonged to tandem duplications. Among them, IbMADS4, IbMADS14, IbMADS32, and IbMADS56 genes share a reciprocal segmental duplication relationship, with IbMADS51 showing a segmental duplication relationship with IbMADS56 and IbMADS4 but not with IbMADS14 or IbMADS32. There are tandem duplications between IbMADS58 and IbMADS59, and between IbMADS68 and IbMADS69. And there are segmental duplications in the other 10 intergenic pairs. The most MADS-box genes with collinear relationships were on chromosome LG11 with five, and the remaining chromosomes had fewer than two or no MADS-box genes with collinear relationships (Figure 6).

To investigate the expression patterns of MADS-box genes in sweet potato, the 75 MADS-box gene expression profiles were detected in six different tissues by transcriptome sequencing. The results showed that sweet potato MADS-box genes were expressed and differentiated in different tissues (Figure 7). Cluster analysis clustered the sweet potato MADS-box gene expression profiles into seven classes, with class I, II, and III genes mainly expressed in different types of roots. Among them, seven genes in class II were highly and specifically expressed in the tuberous root. The class IV genes were expressed in the stem tip and fibrous root. The class V genes were highly and specifically expressed in the stem tip. The class VI genes were predominantly expressed in flower, and a small proportion of these genes were also somewhat expressed in fruits. The class VII genes are highly and specifically expressed in fruit. Overall, 22 MADS-box genes were highly expressed in flowers, 19 were highly expressed in stem tips, 12 were highly expressed in fruits, and the remaining genes were differentially expressed in different types of roots.

qRT-PCR was used to further validate the expression profiles of the 10 genes, which were more highly expressed in roots than other tissues, according to transcriptome sequencing of MADS-box genes in different tissues (Figure 8). Ten genes showed tissue-specific expression at different levels in all tissues; and IbMADS1, IbMADS18, IbMADS19, IbMADS79, and IbMADS90 were highly expressed in the fibrous root or tuberous root. Seven genes were more highly expressed in the leaf, only IbMADS15 was highly expressed in the stem, and the remaining genes were relatively poorly expressed in the stem. Ten genes were all expressed at lower levels in the stem tip. IbMADS18, IbMADS31, and IbMADS83 were significantly expressed in the fruit. Some genes were highly expressed only in a single tissue, and others were not significantly expressed, such as IbAMDS1, which was significantly expressed only in the fibrous root; IbMADS17 and IbMADS20, which were highly expressed only in the leaf; and IbMADS31 and IbMADS83, which were significantly expressed only in the fruit. The five remaining genes were more highly expressed in two to three tissues.