The conserved plant materials C. morifolium ‘Zhongshanzigui’, L. paludosum, and their F₁ hybrids (C. morifolium × L. paludosum) were collected from the Chrysanthemum Germplasm Resource Preserving Center, Nanjing Agricultural University, China. All the plants were propagated from cutting and grown in a greenhouse under natural light. The seedling, flower bud differentiation, flowering, and flowering decay stages of all plant materials were observed and observations were recorded for at least two years. Plant height and leaf area, weak stem thickness, and survival rate (60 days after transplanting) were also collected. Significance testing was performed using SPSS 20.0 (http://www-01.ibm.com/software/cn/analytics/spss/ ).

Total RNA was isolated from young, fully expanded leaves of L. paludosum using the Total RNA Isolation System (Takara Bio) following the manufacturer’s instructions. From 1μL total RNA, first-strand cDNA was synthesized using random primers and SuperScript III Reverse Transcriptase (Invitrogen). PCR amplification was performed using degenerate primers (FT-F: AYACIYTIGTIATGGTIGAYCC; FT-R: CCISWYTCICKYTGRCARTT). The PCR products were cloned using the PMD19 TA cloning kit (Takara, Japan) for confirmation by DNA sequencing. RACE PCR was then used to obtain the full-length cDNA. For the 3′ RACE reaction, the first-strand cDNA was synthesized using an oligo (dT) primer incorporating the sequence of the adaptor primer, followed by nested PCR using gene-specific primer pairs and the adaptor primer. The nested PCR adaptor primer (Abridged Anchor Primer, AAP) and the Abridged Universal Amplification Primer (AUAP) provided with the 5′ RACE System kit v2.0 (Invitrogen) were used to reamplify 5′ RACE and the internal gene-specific primer pairs to isolate full-length cDNAs of LpFTL1. A neighbor-joining tree was generated using MEGA 5 software, applying the Poisson model with gamma-distributed rates and 1000 bootstrap replicates.

Owing to the high conservation of the FTL genes (amino acid sequence identity values> 90%), LpFTL2 and LpFTL3 primer sequences were designed according to CmFTL2 and CmFTL3 (Oda et al., 2012a). The PCR products were cloned using the PMD19 TA cloning kit (Takara) and sequenced. Finally, internal gene-specific primers were designed to distinguish CmFTL2, CmFTL3, LpFTL2, and LpFTL3. The primer sequences and PCR conditions are listed in Table S1 .

A total of five day fully expanded fourth and fifth leaves harvested from three plants were collected at the seedling stage (March 5-10), flower bud differentiation (April 5-10), and flower decay (June 5-10) of L. paludosum; the seedling stage (May 5-10), flower bud differentiation (August 5-10) and flower decay (November 5–10) of C. morifolium ‘Zhongshanzigui’; the seedling stage (March 5-10), flower bud differentiation (April 5-10 and August 5-10) and flower decay (November 1-5) of C. morifolium × L. paludosum.

RNA was extracted using the Total RNA Isolation System (Takara) following the manufacturer’s instructions. Before reverse transcription, total RNA was treated with RNase-free DNase I (Takara) at 37°C for 30 min to avoid genomic DNA contamination. For each sample, first-strand cDNA was synthesized using random primers and SuperScript III Reverse Transcriptase (Invitrogen) according to the manufacturer’s instructions. Quantitative real-time PCR was performed using a Mastercycler ep realplex 2 S system (Eppendorf, Germany). The cDNA was amplified using SYBR^® Premix Ex TaqTM II (Takara, Japan). PCR conditions were 95°C for 2 min, followed by 40 cycles of 95°C for 30 s, 55-58°C for 30 s, and 72°C for 30 s. Expression levels of genes were calculated relative to EF1α genes using the comparative quantification analysis. The relative transcript abundance was calculated using the 2−△△Ct method. The results are presented as the mean ± SD of three replicates. The primer sequences are listed in Table S1.

The restriction sites of Sal I and Not I were added to the ends of LpFTL1 and LpFTL3, respectively. After double restriction digestion, T4 ligase was used to bind to the pENTR1A entry vector. The pENTR1A-LpFTL1 and pENTR1A-LpFTL3 entry vectors were then recombined with the pMDC43 vector using LR recombination technology to obtain pMDC43-LpFTL1 and pMDC43-LpFTL3 expression vectors.

A. thaliana (Col-0) was transformed using the floral dip method. The seeds of the T1 generation were screened and cultured on 1/2 MS + 25 mg/L Hyg plate medium, the T3 generation was obtained by self-pollination, and the zygosity of the transgene was identified using an RT-PCR assay. Then, the T3 generation seeds and the wild type (Col-0) were cultured for 3–4 weeks in a short-day light incubator (23°C/8 h light, 18°C/16 h dark, 75%-80% relative humidity) and transferred to long-day light with the conditions of 23°C/16 h light, 18°C/8 h dark. The transgenic process was repeated twice (codes C and B) to get sufficient transgenic lines. After bolting, statistical analysis was performed every two days. Photographs were taken when 60% of the transgenic lines bloomed. The data were statistically analyzed using Excel and SPSS software.

The full flowering periods of C. morifolium and L. paludosum occurred from late October to early November and late April to May, respectively. The plants of C. morifolium × L. paludosum bloomed from late April to May and late October to early November under natural conditions without any treatment ( Figure S1 ). This means that the hybrids of C. morifolium × L. paludosum flowered twice a year, from spring to early summer and autumn to winter which shown a superimposition of parental flowering patterns. It is emphasized that the state of C. morifolium ( Figure 1A ) and L. paludosum ( Figure 1B ) was natural while only a few flowers can bloom normally in spring for two years but are accompanied by a large number of poorly developed flower buds that cannot open in the later stage ( Figures 1C, D ). In addition to flowering traits, the ornamental traits of the hybrid were weaker than those of their parents. The main performance criteria consisted of decreased plant height and leaf area, slow growth, weak stem thickness, and low survival rate, which showed strong hybrid weakness ( Figure 1E ).

Three FT-like genes have been successfully cloned. The full-length sequences of LpFTL1, LpFTL2, and LpFTL3 cDNA contained a total of 522 nucleotides. LpFTL1, LpFTL2, and LpFTL3 are predicted to encode 174 polypeptide residues. The predicted LpFTL1, LpFTL2, and LpFTL3 protein had calculated molecular mass of 19.76 kDa, 19.79 kDa, and 19.62 kDa, and theoretical pI of 9.14, 8.07, and 7.44, respectively.

The sequences of several other FT orthologs were used to construct a phylogenetic tree (maximum likelihood method) for the FT group genes ( Figure 2A ). The phylogenetic tree was grouped into three major clades: FT, BFT, and TFL1. As expected, LpFTL1, LpFTL2, and LpFTL3 were classified in the FT clade. It was more closely related to CmFTL3/CsFTL3 than to the others. As shown in Figure 2B , these three sequences encoded residues for Y85 (shown in red box) and Q140 (shown in black box), which are required for the formation of an external loop, and the conserved segments (indicated by red and green lines)

qRT-PCR was used to examine the expression patterns of FTLs. The LpFTL1 and LpFTL3 transcripts mainly accumulated during the flower bud differentiation (April), were 15.5 and 22.1 times of the seedling stage (March), respectively, and showed a sudden decrease at the flower decay stage (June) ( Figure 3A ). The expression levels of CmFTL1, CmFTL2, and CmFTL3 in ‘Zhongshanzigui’ are shown in Figure 3B . Three CmFTLs were expressed at low levels in the seedling (May) and flower decay (November) stages. The CmFTL3 was strongly transcribed with a 29.2 value at the key flowering gene than the leaf collected at seedling and flower bud differentiation stages (August).

Although hybridization may affect gene expression to a certain extent, the expression of all FTLs tended to be silent during the seedling stage (March). During flower bud differentiation (April and August), the potential key flowering genes, LpFTL1/LpFTL3 and CmFTL3 transcripts, significantly increased in April and August, respectively, showing an overlay of expression modes in C. morifolium × L. paludosum ( Figure 3C ).

In the present study, a total of 26 and 30 independent T₃ Arabidopsis lines constitutively expressing LpFTL1 and LpFTL3 were produced, and two representative Ox lines of each gene were selected from the T3 generation seeds of transgenic A. thaliana to observe for expression of phenotypic effect in LpFTL1 ( Figure 4A ) and LpFTL3 ( Figure 4B ) for Arabidopsis. Among these, FTL1-C4, FTL1-B2 Ox, FTL3-C16, and FTL3-B9 in multiple independent transgenic Arabidopsis calls showed earlier bolting and flowering than WT (Col-0) plants. The most intuitive reason is the timing of transgenic lines showing bolting or even flowering; WT Arabidopsis was still in vegetative growth and the phyllotaxy at bolting remained the same. While the number of rosette leaves was not different at flowering stage (WT VS. transgenic lines), ranging from 12 to 15 ( Figure 4C ). This result shows that LpFTL1 andLpFTL3 are involved in controlling flowering in Arabidopsis.

Statistical analysis showed that the average flowering days from the sowing of LpFTL1 ox were 37.2 d and 38.9 d and those from the sowing of LpFTL3 ox were 37.4 d and 37.4 d, while that of control WT Arabidopsis was 41.50 d; the difference was more than 2.5 d, above the significance level of 0.05. The mean flowering days of the transgenic plants were significantly different from those of the control WT Arabidopsis ( Figure 5 ).

Hybridization is an important mechanism of genetic variation and functional evolution (Richards, 2003; Kearney, 2005). New combinations of favorable genes come from different parents, potentially resulting in heterosis. In addition to the parent phenotypes, nascent hybrids often show novel traits that do not exist in the contributing parents, such as increasing levels of heat, cold, drought tolerance, pest resistance, flowering time change, and organs allowing their offspring to thrive in harsh environments (McClintock, 1983; Paun et al., 2006; Birchler, 2015; Botet and Keurentjes, 2020). In Arabidopsis allotetraploids, the expression of A. thaliana and A. arenosa FLC loci is additive, leading to a late-flowering phenotype (Wang et al., 2006). The flowering time directly affects plant reproduction and adaptation. In our study, the natural flowering of C. morifolium was in autumn, while L. paludosum was in the spring. Intergeneric hybrids are powerful, and the only way for Chrysanthemums to put forth florescence. Although hybrids between different genera are difficult to obtain, particularly between two species with relatively distant classifications, we obtained hybrids between C. morifolium and L. paludosum. After two years of observation, we ensured that all hybrids exhibited continuous flowering phenotype, and their flower bud differentiation in April and August, which led to flowering under natural conditions without any treatment. These results were consistent with our breeding objectives.

In the analysis of FTLs gene expression in hybrids, the results clearly showed that the FTLs gene is involved in the opening of C. morifolium × L. paludosum. In the flowering process, of L. paludosum and C. morifolium hybrids, it can be tentatively concluded that the first spring flowering is mainly affected by the LpFTL1 and LpFTL3 genes from L. paludosum, whereas the second flowering in autumn is determined by the CmFTLs gene from C. morifolium. Therefore, we speculate that the introduction of LpFTLs also promotes early flowering. From the results, it can be concluded that the LpFTLs gene from L. paludosum regulates flowering and the flowering period of A. thaliana, successfully converting LpFTL1 and LpFTL3 and was 3–5 days earlier than that of WT. Some studies on the verification of the transgenic function of A. thaliana cloned from C. morifolium showed that function was super-expressed by significantly promoting the early flowering of A. thaliana (Oda et al., 2012a; Higuchi and Hisamatsu, 2015; Bernardini et al., 2022). These results are in accordance with our study and prove that the FTL gene prevalent in a variety of plants plays a significant role in regulating flowering and this gene acting on reproductive growth is also similar (Storchova et al., 2015).

After the FTL protein was identified as an anthocyanin, a large number of researchers have isolated, identified, and functionally studied FTL homologous genes from different varieties of crops such as rice, wheat, fruit (apples), vegetables (tomatoes, potatoes), and ornamental crops (orchids, orange stalks, and chrysanthemums), which have a large number of early or late flower expression (Trankner et al., 2011; Li et al., 2014; Higuchi and Hisamatsu, 2015). Among the many varieties, the rational use of interspecific hybridization, intergeneral hybridization, and transgenics will create a broader possibility of flowering regulation (Chen et al., 2014; Qi et al., 2018; Mitrofanova et al., 2020). Most ornamental chrysanthemums are short-day plants, and if the LpFTLs gene is overexpressed in a chrysanthemum, it is likely to change its original flowering characteristics, such as the phenotype of continuous flowering in offspring. However, the exogenous FTL gene and the endogenous FTL gene may have a mutual influence. In an experiment to transfer the cotton FTL gene to tobacco, data suggested that adequate levels of transgenic cotton FTL might interfere with the balance of endogenous FTL inducers and inhibitors, leading to phenotypic changes in transgenic tobacco (Faure et al., 2007). Therefore, to apply the FTL gene to the breeding of continuous flowering, it is also necessary to study species as target plants in a more detailed and clearer model.

Elucidating the molecular basis of the hybrid phenotype will contribute to our understanding of evolution and enhance crop breeding. In the present study, all hybrids exhibited a continuous-flowering phenotype. As chrysanthemum is produced by vegetative propagation, it can be cloned using cuttings, which could keep early flowering characteristics. However, success is temporary, and the growth characteristics show serious hybrid weakness compared to their parents. Therefore, in our later study, it is reasonable to expect a backcross between the hybrid and paternal ‘Zhongshanzigui’ for better ornamental traits.

A prevailing model explains how hybrid phenotypes in which deleterious parental alleles are compensated in hybrid progeny by cumulative enhancement of gene expression (Chen et al., 2014; Tikhenko et al., 2015). Although this matches with our aim to get continuous flowering as an improvement of ornamental traits, it was found in our research center that this is caused by compound expression leading to shorter vegetative growth, especially after the appearance of the first flowers in less than two months. This directly leads to the weakness of heterosis in hybrid offspring. Compared with other crops, continuous flowering is a fascinating goal for ornamental plants; however, strong growth and higher biomass are often ignored. Nevertheless, an intuitive model of continuous flowering observed in our present study appears to be accompanied by hybrid weakness under the balance of vegetative and reproductive growth. Therefore, in future chrysanthemum breeding programs, a suitable balance point must be confirmed to ensure the target flowering time is under normal growth ( Figure 6 ).

A previous study in maize documented cases of positive associations between targeted gene expression and superior phenotypes in hybrids, which is inclusive of the occurrence of hybrid weakness because the actual results can be condition-dependent (Romagnoli et al., 1990). Meanwhile, some slow growth may be due to the haploid level of each parental genome (Sun et al, 2017b). Although the model discussed above appears to be a simple genetic model, the molecular basis of hybrid weakness is usually complex and often involves incremental effects and interactions of genes at multiple loci, most of which have not been characterized, particularly in ornamental plants (Tikhenko et al., 2015). Thus, breeding for continuous flowering either through introgression or transgenics should simultaneously target multiple loci to achieve a desired level of equilibrium (Alcazar et al., 2009). This suggests the need for further identification and molecular characterization of the weakness of hybrid loci. Such findings could facilitate the development of ornamental plants to efficiently utilize hybridization.