Using HPLC, the major pigments from the floral tissues of anthocyanin species (A+) species were found to comprise delphinidin and cyanidin, whereas no detectable anthocyanins were found in the non-anthocyanin species (A-) flowers ( Figure 3 ). To further characterize the mechanism responsible for the non-pigmentation of flowers in A- species, E. sagittatum was used as a model for the enzymatic function.

To determine whether changes in gene expression might be involved in the non-pigmentation phenotype of A- species, the transcript levels of putative anthocyanin biosynthetic enzymes were examined in petal tissue ( Figure 4 ; Supplementary Figure 1A ). Expression of CHS1 not CHS2 and CHS3 was found to be significantly lower in E. wushanese (A-) than in other species. Similarly, down-regulation of CHS2 was observed in E. frachetii (A-), suggesting that loss of anthocyanin may result from low levels of expression of different copies of CHS in E. wushanese and E. franchetii. For CHI and F3H, we found no significant correlation between expression level and the loss of anthocyanins in spur tissues of all A- species. Among the structural genes, ANS was the only ABP locus where all A- species had significantly lower expression levels than that of A+ species. This suggests that the lack of pigmentation production in all A- species could be caused primarily by lower ANS expression. The expression level of DFR was significantly lower in A- species than in A+ species, except for E. lishihchenii. It has been reported that substrate competition between FLS and DFR creates a metabolic flux of the flavonoid biosynthetic pathway in Arabidopsis. In this study, low expression of DFR in the A- species E. franchetii and E. wushanese was correlated with increased accumulation of FLS expression. On the other hand, up-regulation of DFR was positively correlated with FLS expression in E. lishihchenii but E. sagittatum showed no correlation with DFR. In summary, these results suggest that the loss of anthocyanin in E. frachetii, E. wushanese and E. sagittatum may be primarily related to alterations at the ANS locus, affecting gene expression.

To further analyze the loss of anthocyanin in sepals ( Figure 5 ; Supplementary Figure 1B ), gene expression was analyzed across five Epimedium species using the same primers. Expression of DFR was lowest in the three A- species and was correlated with ANS expression, suggesting the expression of DFR and ANS could be regulated by a common transcription factor. CHS1 transcripts were not detected in the sepals of E. wushanese, which also had the lowest CHS1 expression in petals. These observations suggest that negative regulation of the DFR and ANS genes together was also correlated with the lowest CHS1 in sepals and petals in E. wushanese.

To study the catalytic activity of E. sagittatum ABP gene products, 35S::EsF3’H and 35S::EsDFR genes were individually transferred into their respective Arabidopsis mutants; transparent testa 7 (tt7) lacking flavonoid 3’-hydroxylase, and transparent testa3 (tt3) lacking dihydroflavonol reductase under the control of the cauliflower mosaic virus 35S promoter (Peer et al., 2001). Transgenic and mutant control seedlings were grown under nitrogen stress to determine if the Epimedium genes could rescue the Arabidopsis anthocyanin-null mutant phenotypes. Accumulation of anthocyanins was observed in transgenic seedlings ectopically expressing EsF3’H and EsDFR ( Figure 6 ). However, the tt7 and tt3 mutant controls did not exhibit anthocyanin accumulation in cotyledons. Thus the E. sagittatum genes showed catalytic activity in Arabidopsis. Given the lack of an Arabidopsis mutant for F3’5’H, in order to determine whether EsF3’5’H can function in vivo, we overexpressed 35S::EsF3’5’H in wild-type Arabidopsis. Under normal conditions on 1/2 MS medium, the seedlings overexpressing EsF3’5’H showed comparable anthocyanin production to wild-type controls ( Figure 6 ).

Eight Epimedium species (E. acuminatum, E. franchetii, E. leptorrhizum, E. epstenii, E. sagittatum, E. lishihchenii, E. wushanense, and E. zhushanense) grown in the specialized Epimedium nurseries of Wuhan Botanical Garden, Chinese Academy of Sciences, Wuhan, China ( Figure 2 ). All plants were transplanted from wild populations and growing under the same environmental conditions. Floral tissues including petaloid sepals and spur-like petals were collected in the spring of 2011. The eight species were separated into two groups corresponding to anthocyanin (A+) (E. acuminatum, E. leptorrhizum, E. epstenii and E. zhushanense) and non-anthocyanin (A-) (E. franchetii, E. sagittatum, E. lishihchenii, E. wushanense) based on visual observation of the floral tissues. The samples were weighed, packaged in aluminum foil, flash-frozen in liquid nitrogen, and then stored at -80°C.

The profiles of anthocyanidins from the samples of A- species and A+ species were analyzed using HPLC. The precursors of anthocyanin pigments were extracted from 100 mg of fresh corolla tissue. For each sample, 20 μL of supernatant was injected into a Shimadzu LC-20 AT liquid chromatograph (Shimadzu Corporation, Japan) and a 250×4.6 mm reverse phase C18 column (Sigma-Aldrich, USA) at a flow rate of 1 ml min^-1. The organic solvent was composed of acetonitrile and 0.1% trifluoroacetic acid, and the polar solvent was 0.1% trifluoroacetic acid in HPLC-grade water. The anthocyanin was measured at 550 nm. The chemical compounds cyanidin, delphinidin, malvidin, pelargonidin, peonidin, and petunidin (Poypehenols Laboratories, Norway), were used as anthocyanidin standards.

In total, 12 genes from E. sagittatum involved in the ABP were cloned following RT-PCR amplification using degenerate primers or specific primers based on our previous investigation(Zeng et al., 2010; Huang et al., 2013a; Huang et al., 2013b; Huang et al., 2015). These genes were CHS1, CHS2, CHS3, CHI1, CHI2, F3H1, F3H2, F3’H, F3’5’H, FLS, DFR, ANS. In this study, all pairs of primer from E. sagittatum were transferred to other Epimedium species.

Total RNA was extracted from inner sepals and petals at anthesis, at which time the biosynthesis of anthocyanin is completed. First-strand cDNA was synthesized using PrimeScript RT reagent Kit (Takara, Japan) following the manufacturer’s instructions. In each 20 μL qRT-PCR reaction, 50 ng of cDNA was amplified using SYBR^® Premix Ex TaqTM II (Takara, Japan) and 100 mM of primers in an ABI7500 Real-Time PCR machine (ABI, USA) as per the manual. Actin was amplified as the control gene. The samples from three tissues were used and three technical replicates were performed for each sample. Data were analyzed by ABI7500 software. In this study, all pairs of primer (CHS1, CHS2, CHS3, CHI1, CHI2, F3H1, F3H2, F3’H, F3’5’H, FLS, DFR, and ANS) from E. sagittatum were transferred in other Epimedium species. All primers used in this manuscript are listed in the supplementary database.

For functional analyses, the E. sagittatum (A-) genes EsF3’H and EsDFR were overexpressed in their respective Arabidopsis thaliana (ecotype Landsberg) mutants, each lacking anthocyanins at the seedling stage. EsF3’5’H, The coding regions of EsF3’H, EsF3’5’H, and EsDFR were cloned into pMD19-T (Takara, Japan). The SalI and SacI digested fragment of each gene was purified and ligated into the pMV plasmid (derived from pBI121) behind the cauliflower 35S promoter. The plasmids were then transformed into Agrobacterium strain EHA105. Arabidopsis wild-type and mutants (tt3 and tt7) were transformed by the floral dip infiltration method(Zhang et al., 2006). Transformants were selected on 1/2 Murashige and Skoog medium supplemented with 50 μg/mL kanamycin. Resistant seedlings were then transferred into the soil to harvest seeds. T1 seedlings were screened on 1/2 MS medium minus nitrogen for observation of anthocyanin accumulation.