A purple-skinned pepino (S. muricatum) genotype, termed Purple Selection (PS, purple skin and yellow flesh) and a yellow-skinned pepino selection, Yellow Selection (YS, yellow skin and yellow flesh) were propagated by cuttings in a greenhouse at Plant & Food Research, Auckland, New Zealand. After 56 days, rooted cuttings were transferred into 30-L pots and grown in a greenhouse at Plant & Food Research, Pukekohe, New Zealand, at ambient conditions. Ten to thirteen pepino fruit from separate plants were collected at 20, 34, 48 and 73 days after flowering (DAF), and pooled from at least three plants (each stage of fruit development had three biological replicates). Skin tissue was carefully separated from flesh tissue, with any residual flesh tissue removed (pepino has a very strong skin layer) and frozen separately in liquid nitrogen and stored at −80°C.

For functional gene testing, Nicotiana tabacum (tobacco) and pepino plants were grown under glasshouse conditions between 20 to 24°C using natural daylight with extension to 16 h at Plant & Food Research, Auckland, New Zealand.

Pepino extracts were analyzed with a Waters Acquity® Ultra Performance Liquid Chromatography (UPLC) system (Milford, MA, USA) equipped with a photodiode array detector (PDA) and a single quadrupole mass spectrometer (QDa, Waters, MA, USA). The separation of the 5-μL aliquot was performed on a reverse-phase Acquity UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm particle size, Waters, MA, USA). The eluents were 0.1% formic acid in water (A), and 0.1% formic acid in acetonitrile (B). The gradient was as follows: 0 min, 5% B; 5 min, 10% B; 10 min, 20% B; 12 min, 100% B; 14 min, 100% B; 14.5 min, 10% B; 18 min, 10% B. The flow rate was 0.5 mL min⁻¹ and the column temperature was 50°C. Anthocyanins were detected at 520 nm and quantified as petunidin-3-glucoside equivalents using a six-point calibration curve. Furthermore, a validation standard was injected after every tenth injection.

Total RNA was extracted using the Spectrum Plant Total RNA kit® (Sigma-Aldrich). Three biological replicates were used. Library construction, sequencing, mapping to reference genome, and quality control were performed at Australian Genome Research Facility (AGRF; www.agrf.org.au) (Illumina Stranded mRNA with 150 paired end reads and sequencing to 20M depth). The primary sequence data were generated using the Illumina DRAGEN BCL Convert 07.021.624.3.10.8 pipeline. Data yield for the 48 samples from the pipeline ranged from 6.91Gb to 12.7 Gb, with mean yield of 9.3 Gb. Methods at AGRF followed the strict requirements of the International standard ISO17025 for quality control, testing and calibration. The per-base sequence quality for the 48 samples was excellent, with >87% bases above Q30 across all samples. Reads were mapped using the STAR aligner (v2.5.3a) to the Solanum muricatum genome downloaded from http://songlab.bio2db.com/pepino.html. Mapping QC results are presented in Supplementary Figure 1 . Counts were normalized using the Relative Log Expression method in the DESeq2 package (Love et al., 2014). Principal component analysis was performed using common variance on the top 500 most variable genes, to check consistency of the biological replicates. The transcriptome response of the three biological replicates was similar, allowing differences in genotype, tissue type and developmental stage to be examined ( Supplementary Figure 2 ).

Sequence alignment was performed using Geneious Prime® (version 2022.0.1) Muscle alignment with eight iterations and the phylogenetic tree was constructed using the Neighbor-Joining method with 1000 bootstrap replicates.

Total RNA of skin and flesh of the two pepino genotypes, PS and YS, was extracted using the Spectrum Plant Total RNA Kit (Sigma -Aldrich, St. Louis, MO, USA). Removal of genomic DNA and first strand cDNA synthesis was carried out using the QuantiTect Reverse Transcription kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions.

Real-time qPCR DNA amplification and analysis was carried out using the LightCycler 480 Real-Time PCR System (Roche Diagnostics, Switzerland), with LightCycler 480 software version 1.5. The LightCycler 480 SYBR Green I Master Mix (Roche) was used. Each reaction volume was 10 µL; reactions were run with four replicates and a water control was included in each run. qPCR conditions were as follows: 5 min at 95°C, followed by 40 cycles of 5 s at 95°C, 5 s at 60°C, and 10 s at 72°C, followed by 65–95°C melting curve detection. The qPCR efficiency of each gene was obtained by analyzing the raw fluorescence data using the LinRegPCR (2014.x) software (Ruijter et al., 2009). Used primer sequences are listed in Supplementary Table 1 . These were normalized to elongation factor-1 of pepino (Sm06G02378).

Sequences of candidate genes were identified by BLAST (Altschul et al., 1997) from the S. muricatum pepino genome (Song et al., 2022) and PCR-amplified from leaf cDNA using specific primers and Phusion® High Fidelity DNA polymerase (Thermo Fisher Scientific, USA). PCR products of anthocyanin-related regulatory genes (SmMYB113 and SmATV) and genes that encode enzymes of the anthocyanin pathway (SmDFR and SmF3’5’H) were cloned into pENTR-TOPO (ThermoFisher) followed by In-Fusion cloning (Takara Bio) into pHEX2. Clones were confirmed by sequencing (Macrogen).

Transient assays were performed in tobacco as previously reported (Hellens et al., 2005). Approximately 300 μL of Agrobacterium tumefaciens GV3101 culture containing genes of interest were infiltrated into a young N. tabacum leaf. Final color development was evaluated by taking digital photographs four days after the infiltration.

SmMYB113 was transformed into etiolated hypocotyls of pepino (S. muricatum) yellow-skinned selection (YS) following a previously described transformation protocol (Horsch et al., 1985) adapted to use hypocotyls as explants. Briefly, etiolated hypocotyls were cut into 5-mm long pieces. Explants were inoculated with 10 mL Agrobacterium culture for 10 min and blotted dry with sterile filter paper. Inoculated explants were transferred onto co-cultivation medium for 3 days in the dark. After co-cultivation, explants were transferred onto regeneration medium (MS basal salts and vitamins (Duchefa, Haarlem, The Netherlands), 3% sucrose, 0.7% agar (Invitrogen, Waltham, MA, USA), 1 mg/L BAP + 0.1 mg/L NAA and 300 mg/L Timentin) and selection media (MS basal salts and vitamins (Duchefa), 3% sucrose, 0.7% agar (Germantown, 0.1 mg/L NAA and 300 mg/L Timentin) to induce adventitious shoots from callus tissue. Selection was by the inclusion of 150 mg/L kanamycin in the medium. A single shoot was selected from each explant and transferred onto elongation medium which also included 150 mg/L kanamycin. Three to four weeks later shoots started to develop roots. Once roots were established, rooted plants were transferred into potting mix and moved to glasshouse conditions between 20 to 24°C using natural daylight with extension to 16 h.

Two pepino selections that differ in anthocyanin pigmentation of the fruit skin, denoted purple skin (PS) or yellow skin (YS), were chosen for analysis. They produce fruit of comparable size, shape, growth rate and yellow flesh color ( Figures 1A, B ). Anthocyanins were absent from the skin and flesh of the YS genotype, throughout fruit development. Both YS and PS genotypes can produce anthocyanins in leaves and petals, although PS has much darker purple foliage and petals ( Supplementary Figure 3 ). Total anthocyanin concentrations were evaluated throughout the fruit development in the PS genotype ( Figure 1C ). The total amount of anthocyanin was significantly higher at stage 4 than in stages 1 to 3, reaching 1.7 mg g^-1 (dry weight or DW) at stage 4.

Anthocyanin composition was first determined in the skin of the PS genotype by Ultra Performance Liquid Chromatography-High-Resolution Mass Spectrometry (UPLC-HR/MS) and Ultra Performance Liquid Chromatography- diode Array Detector (UPLC-DAD) and compared with available data from the literature. Four predominant anthocyanin compounds were detected in the profiles ( Figure 1D ). Peak 2 represented about 60% of total anthocyanins and was identified as petunidin-3-p-coumaroyl-rutinoside-5-glucoside ( Figure 1D ; Supplementary Figure 4 ). The presence of this anthocyanin has previously been reported in other fruits of the Solanaceae family, such as purple tomatoes (Solanum lycopersicum L.) (Silva Souza et al., 2020), and purple-fleshed potato tubers (Solanum tuberosum L.) (Andre et al., 2007). When anthocyanins are acylated, both cis- and trans- isomers can be found (Saha et al., 2020). Considering the predominance of trans isomers in Solanum species, it is likely that peak 2 is petunidin-3-trans-coumaroyl-rutinoside-5-glucoside (also known as petanin), whereas peak 3 is its cis- stereoisomer.

The other anthocyanin compounds followed the same substitution pattern: -3-acylrutinoside-5-glucoside and exhibited a similar anthocyanidin fragment at m/z 317, indicative of petunidin derivatives. Consequently, peaks 1 and 4 were identified as petunidin-3-caffeoyl-rutinoside-5-glucoside and petunidin-3-feruloyl-rutinoside-5-glucoside, respectively.

There were no significant differences in the proportions of the different anthocyanins at the four developmental stages investigated, suggesting that there is a common regulatory system for all four anthocyanin compounds.

To study the transcripts associated with anthocyanin accumulation in pepino, skin and flesh samples were collected from PS and YS genotypes over the four developmental stages and subjected to RNA-seq. A comparative analysis between purple and yellow tissues at the same developmental stage was performed and differentially expressed genes (DEGs) with the minimum of log₂-fold change in expression (DESeq, adjusted p-value<0.05, mean count 100) between the purple and yellow cultivars were identified. This analysis revealed 1540 up-regulated DEGs and 1467 down-regulated DEGs in skin, and 1524 up-regulated DEGs and 1269 down-regulated DEGs in flesh, when PS was compared with YS. Of these genes, there was an enrichment for candidate genes related to anthocyanin biosynthesis or regulation of this pathway.

Previously, 17 pepino regulatory genes and 11 biosynthetic enzymes have been identified as participating in anthocyanin biosynthesis (Song et al., 2022). Most of these gene models were differentially expressed between YS and PS ( Figure 2 ). In our study, a total of 25 anthocyanin-related DEGs were consistently up-regulated, with four down-regulated in PS skin samples compared with YS ( Figure 2 ). Four genes were DEGs but had variable expression across skin developmental stages. In flesh, 13 and seven DEGs were consistently up- or down-regulated, respectively and when comparing the DEGs in skin and flesh, eight biosynthetic genes were up-regulated (SmuPAL6, SmuPAL7, SmuPAL10, Smu4CL3, Smu4CL4, SmuCHI3, SmuF3’H and SmuUGT). Also, the anthocyanin-related R2R3 MYB regulator SmuMYB113 was up-regulated in both PS tissues, whereas the bHLH SmuTT8 was up-regulated in skin only. In contrast, Smu4CL1 and SmuCHS2 (both anthocyanin biosynthesis-related genes) were down-regulated in both skin and flesh tissues. In addition, five genes were DEGs but had variable expression across flesh developmental stages.

Transcript reads of SmuDFR2 (Sm02G02175) were significantly higher in the purple-skinned genotype than in the yellow-skinned cultivar ( Figure 3 ). Reads of less than 10 per stage were present in flesh of both PS and YS ( Figures 3A, B ). A similar trend was seen with SmuF3’5’H (Sm11G02385), where transcripts were significantly higher in the PS genotype than in the YS cultivar. Reads peaked at stage 4, which correlated with the highest concentrations of anthocyanins ( Figure 1 ). In the flesh, reads for SmuF3’5’H were again less than 5 normalized reads per stage.

An examination of candidate transcriptional regulators of anthocyanin-related enzymes was made. There were four candidate MYB transcription factors in the DEG lists, as well as a bHLH, which is homologous (69% identity) to AtTT8 ( Figure 2A ) and was previously annotated as SmuTT8 (Song et al., 2022) ( Table 1 ). This TT8-like bHLH was highly expressed in PS skin, but not as well expressed in PS flesh. However, the R2R3 MYB SmMYB113 was consistently up-regulated, in both skin and flesh of the purple-skinned selection. Transcript counts were high in all purple fruit tissues ( Figure 3B ). In contrast, reads were absent or very low in YS fruit tissues.

Candidate repressors of anthocyanins, such as R3 MYBs and LOB domain TFs, did not explain the phenotypes, as they were not up-regulated either in YS, or in PS flesh tissues ( Figure 2 ). In addition, a homologue of the tomato anthocyanin repressor SlATV, which we termed SmATV, was identified by blast match in the pepino genome (chr.7, 78.8Mb). Reads mapping to this gene model (previously un-annotated) were low, but present in both skin and flesh ( Figures 3A, B ). Furthermore, a flavonol-related subgroup 7 MYB (SmuMYB12) was expressed in skin of both selections ( Table 1 ), and therefore increased flux of precursor substrates away from anthocyanins does not appear be the cause of the YS phenotype.

Phylogenetic analysis of these MYBs revealed close similarities between SmuMYB113 and StAN1 of potato, and SmuMYB114, and SlANT1 of tomato ( Figure 4 ). In potato, StAN1 is a key regulator of skin coloration (Payyavula et al., 2013). Although differentially expressed, SmMYB114 showed low expression in pepino ( Table 1 ). Therefore, both expression and sequence similarity suggest SmuMYB113 is the causative controller of pepino skin anthocyanin content. However, this does not explain the flesh phenotype, where high levels of SmMYB113 expression are seen, yet no anthocyanins accumulate. In this phylogenetic tree, SmuATV clustered with a group of R3 MYB TFs. Within this group, SmuATV was closely related to PhMYBx and SlMYBATV, both negative regulators of anthocyanin biosynthesis (Kroon, 2004; Albert et al., 2011; Albert et al., 2014b).

Expression of anthocyanin-related genes was verified using RT-qPCR. Accumulation of transcripts of anthocyanin biosynthetic genes SmuDFR2 and SmuF3’5’H was confirmed in purple skin, but these were absent from yellow skin and flesh at all studied stages of fruit development ( Figure 5 ). An increase at 34 DAF for SmuDFR was followed by a decline at later fruit developmental stages, while transcripts of SmuF3’5’H increased at 34 DAF and peaked at 73 DAF.

The expression of SmuMYB113 and SmuATV in flesh and skin was examined using qPCR. For SmuMYB113, the purple-skinned stage 2 of development showed the highest value, followed by a decline. For yellow-skinned (YS) pepino, the values were significantly lower in all stages of development, in both skin and flesh. However, in the flesh of the purple-skinned genotype, high levels of expression were seen at three of the four stages. This confirms patterns seen with RNAseq, and suggests a repressive mechanism exists in the pepino flesh. Expression levels of the potential repressor SmuATV were low in the purple- and yellow-skinned cultivars. However, PS had higher levels of expression generally than YS.

As SmuMYB113 appears to be a key regulatory difference between the purple-skinned and yellow-skinned cultivars, further correlations were made between its expression and other putative target genes within the phenylpropanoid and flavonoid pathways. Early steps in the anthocyanin pathway, such as PAL, C4H, 4CL, CHS, and F3’H, were all well expressed in yellow skin and flesh tissues of both selections ( Table 2 ). This is despite the fact that no anthocyanin accumulated in yellow skin or flesh, or in the flesh of the PS selection. In skin tissues, there was a strong positive correlation between SmuMYB113/anthocyanin and expression of CHI3, F3’5’H, DFR, and the GST, TT19. An exception was the previously annotated UFGT (Sm09G02395) (Song et al., 2022), which showed high expression in all fruit tissues. By blasting the Arabidopsis anthocyanidin 3-O-glucosyltransferase (At5g17050) and grape UFGT (Ford et al., 1998), better matching pepino gene models were retrieved, such as Sm10G01884 ( Table 2 ), which was expressed in a skin-specific manner, especially in PS tissue. We therefore hypothesize that the absence of SmuMYB113 in the yellow-skinned selection of pepino results in a loss of expression of later steps in the anthocyanin pathway. This would also lead to a loss of expression of the bHLH partner, SmTT8, which is known to be part of a hierarchy of regulation in solanaceous plants (Montefiori et al., 2015). However, this does not explain the presence of SmuMYB113 expression in the flesh of the purple-skinned selection, where there is no elevation of expression of these later biosynthetic steps, or indeed of the co-regulator SmTT8.

To functionally test the R2R3 MYB SmuMYB113 and R3 MYB SmuATV, their full-length coding regions were cloned, and placed under the control of the 35S promoter. These sequences were infiltrated using Agrobacterium into tobacco leaves (Espley et al., 2007). Infiltration of SmuMYB113 activated anthocyanin production in tobacco leaves ( Figure 6 ). No anthocyanin production was observed with the negative control GUS construct. The potato StAN1 (NCBI protein accession number: AKA95392) construct (Liu et al., 2016) was infiltrated as a positive control and resulted in strong activation of anthocyanin accumulation ( Figure 6 ). In contrast, co-infiltration of StAN1 with SmuATV significantly reduced anthocyanin accumulation compared with StAN1 alone. Similarly, when SmuATV was co-infiltrated with SmMYB113, there was no anthocyanin accumulation, suggesting strong repression of the R2R3 MYB by this R3 MYB. These findings support the hypothesis that SmuATV acts as a repressor of anthocyanin accumulation.

To characterize the effect of SmMYB113 on pepino anthocyanin biosynthesis, we transformed pepino (YS genotype) with a 35S:SmMYB113 overexpression construct. Three independent lines were produced that showed intense purple pigmentation in leaves and flowers ( Figures 7A, B ). Fruit developed normally and had visibly enhanced purple color ( Figure 7C ). These fruits were tested by qPCR, which showed increased expression of the SmuMYB113 gene in both fruit skin and flesh ( Figure 7D ). A correlation between transgene expression and fruit skin and flesh color intensity was apparent, with transgenic pepino Line 1 showing the highest SmuMYB113 transcript levels and the darkest skin and flesh ( Figures 7C, D ). In contrast, SmuATV transcript was higher in the skin and flesh of a control fruit than in Lines 1 and 3, although increased SmuATV was detected in the skin of Line 2. Transcript abundance of SmuTT8 was also elevated in the transgenic lines, compared with the untransformed YS control. In the flesh of YS wild-type lines, there were very few reads of TT8 in the RNA seq dataset ( Table 1 ). However, when MYB113 was overexpressed as a cDNA transgene, TT8 transcript was induced. Over-expression of SmuMYB113 therefore changes a YS phenotype into a plant with purple-skinned fruit, although the flesh color was still not as intense as expected from a strong promoter. This is further evidence that a strong repressive background is not fully overcome by over-expression of the R2R3 MYB.