The tomato (Solanum lycopersicum) varieties M82 and Rutgers served as a reference “wild-type.” The tangerine t³⁰⁰² and yellow flesh r²⁹⁹⁷ mutants and the wild species S. pimpinellifolium accession LA1589 were obtained from the Tomato Genetics Resource Center (University of California, Davis, CA). Mutants t³⁴⁰⁶, r³⁷⁵⁶, and z²⁸⁰³ were isolated from the variety M82 following mutagenesis and screening (Menda et al., 2004). Introgression line IL3-2, which carries a single chromosomal segment from S. pennellii (LA716) in the genetic background M82 (Eshed and Zamir, 1995), was obtained from Prof. Zamir (The Hebrew University of Jerusalem, Israel). Virus-induced gene silencing (VIGS) experiments were carried out in a transgenic Moneymaker line overexpressing the Delila (Del) and Rosea1 (Ros1) Myb transcription factors from Antirrhinum majus under control of the fruit-specific promoter E8 (Butelli et al., 2008; Orzaez et al., 2009). Plants were grown in the greenhouse as previously described (Neuman et al., 2014).

DNA was extracted from approximately 15 mg of young leaves as previously described (Eshed and Zamir, 1995). Homozygous F2 plants from crosses between the different carotenoid mutants were identified by visual screening and confirmed through DNA genotyping. The tangerine alleles t³⁴⁰⁶ and t³⁰⁰², and the zeta mutant z²⁰⁸³ were identified by the virescence appearance of their shoots and the tawny flower color typical of tangerine. The genotyping of the homozygous yellow flesh mutants IL3-2 (r^sp), r²⁹⁹⁷, r³⁷⁵⁶, and Delila + Rosea (DR) was confirmed by polymerase chain reactions (PCR) amplification using the following primers: For PSY1 in IL3-2, 5′-AATACTTTTAGGGTCAAACAATTAA-3′ (forward) and 5′-AAAAATTGACCCACATTGAAAAA-3′ (reverse). Due to a deletion of 672 bp in the PSY1 from S. pennellii and its introgression line IL3-2, the PCR amplification yielded a 1065 bp fragment in the wild-type tomato S. lycopersicum and 393 bp in the IL3-2 introgression line. For PSY1 in r²⁹⁹⁷, the primers 5′-CGGGAGTCATTAGCATAGTTCC-3′ (forward) and 5′- CGAGGCATAGGAATTTGGTG-3′ (reverse) were used for PCR amplification. A 5 kb insertion in the PSY1 from r²⁹⁹⁷ generates a fragment of 5,326 bp in r²⁹⁹⁷ and 447 bp in the wild-type. Amplification of such a large fragment by PCR was possible using the KAPA HiFi HotStart ReadyMix kit #KK2601 (Roche). For PSY1 in r³⁷⁵⁶, 5′-CGAGGCATAGGAATTTGGTG-3′ (forward) and 5′-ACCTATCTAAGGCTGCCGGGGTAATA-3′ (reverse). The PCR product was sequenced to identify the presence of a transition mutation changing codon 151 from Trp to an early stop codon (Kachanovsky et al., 2012).

The presence of the transgenic DR genes was confirmed in seedlings using PCR amplification with the following primers: 5′AAGGCTTCTGATACGGACAAG3′ (forward) and 5′TCTTACGGCTTCCATCACTTC3′ (reverse).

Total RNA was extracted from 200 mg fruit tissue with TRI Reagent RNA isolation reagent (Sigma-Aldrich), according to the manufacturer’s protocol. For cDNA preparation, reverse transcription was done with the iScript™ gDNA Clear cDNA Synthesis Kit #172–5,035 (Bio-Rad). Rapid amplification of 5′ cDNA end (5′-RACE) analyses to determine the 5′ end of the transcripts of PSY1 were carried out with SMART^® RACE 5′/3′ Kit from Clontech Laboratories Inc. (Mountain view, Ca, United States) according to the manufacturer’s protocol. The reverse primer was 5′-TCCATACGCATTCCTTCAATC-3′ (exon #7).

To measure transcript levels of PSY1 and PSY2, the cDNA was amplified using ProFlex PCR System (Applied Biosystems by Thermo Fisher Scientific) in a quantitative PCR protocol using the Applied Biosystems^™ Fast SYBR^™ Green Master Mix on a StepOnePlus^™ Real-Time PCR System (Applied Biosystems). Cycling conditions were 95°C for 20 s, followed by 40 cycles of 95°C for 3 s, 60°C for 30 s and fluorescence acquisition at 60°C. For each gene, the relative mRNA level was determined in three biological replicates. The primers used for the RT-PCR amplifications were 5′-AACTTGTTGATGGCCCAAAC-3′ (forward) and 5′-CTGTATCGGACAAAGCACCA-3′ (reverse), for PSY1 (Solyc03g031860); 5′-AGTTCTGCTAGTAGATGGCC-3′ (forward) and 5′-GGGCACTAGAGATCTTGCAT-3′ (reverse) for PSY2 (Solyc02g081330); and 5′-AACAGTTGGCCTAATGAGTGTGC-3′ for PSY3 (Solyc01g005940) The ACTIN gene (Solyc11g005330) served as a control for normalization, using the primers 5′-TTGCTGACCGTATGAGCAAG-3′ (forward) and 5′-GGACAATGGATGGACCAGAC-3′ (reverse) that differentiate between genomic DNA and cDNA sequences.

To investigate the enzymatic activity of the different PSY1 variants from r²⁹⁹⁷, the complete cDNA of this gene was obtained from RNA isolated from pulp of fresh fruit of r²⁹⁹⁷ followed by RT-PCR using the primers 5′-AGCCACTAGTTGGCTTGTTGAGTGAAGCATAT-3′ (forward) and 5′-GTCGCTCGAGCTCATGCTTTATCTTTGAAGAGAGG-3′ (reverse). The PCR products were cloned into the plasmid pBluescript SK⁺ between the SpeI and XhoI restriction sites, and the plasmids were transfected into Escherichia coli cells of the strain XL1Blue. Sequencing of different bacterial colonies obtained from this cloning revealed two variants that can produce possible functional open reading frames of PSY1. One of these variants resulted from skipping over exon #4 of PSY1 while the other from skipping over exon #2 and exon #4 of this gene. Thus, these two new plasmids were termed pPSY1V1 and pPSY1V2, respectively. As a control, the full-length transcript was amplified from wild-type tomato fruit and cloned into the plasmid pBluescript SK⁺ between the SpeI and XhoI sites, generating the pfull_PSY1 vector. The inserts were sequenced to identify possible PCR-derived mutations.

The plasmid pACCRT-EIB carries the Pantoea agglomerans genes Idi, crtE, crtI, and crtB that produce lycopene when expressed in E. coli (Cunningham and Gantt 2007). The crtB gene, which encodes phytoene synthase, was knocked out pACCRT-EIB through site-directed mutagenesis using the protocol of QuikChange^® Site-Directed Mutagenesis Kit of Stratagene (La Jolla, CA) with the KAPA HiFi HotStart ReadyMix enzyme with the primers 5′-TCAGGAAGTGGCTATGCTCATGATATCGCCCC-3′ (forward) and 5′-GGGGCGATATCATGAGCATAGCCACTTCCTGA-3′ (reverse). The new plasmid was called pACCRT-EI. Plasmids pPSY1V1, pPSY1V2, and the full-length PSY1 cDNA plasmid pfull_PSY1, were co-transfected with pACCRT-EI to E. coli strain XL1-Blue grown on Luria-Bertani (LB) medium containing the antibiotics ampicillin and chloramphenicol. To enhance the expression of these genes, 24 mg/l of Isopropyl 1-thio-β-D-galactopyranoside (IPTG) was added to the LB medium. E. coli cells were grown overnight at 37°C on LB agar plates followed by 5 days at room temperature for pigment accumulation.

Fresh samples of fruit were collected from three biological replicates. Fruit pigments were extracted from 200 to 250 mg of fresh pericarp tissue at the “breaker” and “ripe” (breaker plus 7 days) stages. The tissue was ground in 1 ml of 1:1 water chloroform mixture. The chloroform phase was separated by centrifugation and dried under a stream of N₂. The dried carotenoid extracts were dissolved in 300 μl acetone. Carotenoids were separated by high-performance liquid chromatography analysis as previously described (Karniel et al., 2020).

A VIGS experiment was established as previously described (Fantini et al., 2013). The pTRV2_DR plasmid was kindly supplied by Dr. Giovanni Giuliano, ENEA, Italy. A PSY2-specific silencing sequence was amplified with the primers 5′-GGGGACAAGTTTGT ACAAAAAAGCAGGCTGACGTTGCCCATTGCTTATGC-3′ (forward) and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTACTCAAATGAAGTCAATTATC-3′ and cloned in the pTRV2_DR using the Gateway BP Clonase II enzyme mix and Gateway LR Clonase II enzyme according to the manufacturer’s protocol.

Previous studies suggested that the locus yellow flesh r²⁹⁹⁷ is genetically linked to the PSY1 gene in chromosome #3. However, the molecular basis of the mutation has remained unknown (Kachanovsky et al., 2012). To find the r²⁹⁹⁷ mutation, a mapping population of 310 F2 plants was created from a cross between the homozygous r²⁹⁹⁷ tomato line LA2997 in the Rutgers variety genetic background and the red-fruited wild species S. pimpinellifolium. The mapping relied initially on two polymorphic markers between S. pimpinellifolium and S. lycopersicum, named INDEL1 and INDEL2, found in the genomic sequence spanning ~1600 kb around the PSY1 gene in chromosome #3 (Supplementary Table S1; Supplementary Figure S1). Sixty recombinant plants between these markers were phenotyped for fruit color and were further screened with additional fourteen PCR markers (Supplementary Table S1). Due to the recessive nature of the mutation, heterozygous red-fruited plants were self-pollinated, and fruit color was checked in the F3 generation (Supplementary Figure S1A). The association between fruit phenotypes and the DNA marker results enabled us to localize the r²⁹⁹⁷ mutation within 24 kb between markers PSY1-3′ and 8,587 that spanned the whole PSY1 gene (Supplementary Figure S1B). Using PCR amplification with primers located in this region, we have amplified an insertion of 4,867 bp in exon #4 of the PSY1 gene in yellow flesh allele r²⁹⁹⁷. The nucleotide sequence of the insertion (Supplementary Figure S2) is identical to a known Ty1-copia-like retrotransposon of the Rider family (GenBank: EU195798.2; Cheng et al., 2009), which contains 397 nucleotide long terminal direct repeats (LTR) at both borders, an open reading frame encoding a Ty1-copia-type polyprotein, and a primer-binding site (PBS) and polypurine tract site (PPT). Interestingly, the 5′ end of the retrotransposon sequence is identical to the previously published sequence of GTOM5, an abortive mRNA transcript of PSY1 from the mutant yellow flesh r in the tomato variety Ailsa Craig (Fray and Grierson, 1993a).

A rapid amplification of 5′ cDNA ends (5′-RACE) analysis was carried out to determine the transcription initiation site of PSY1 in the wild-type tomato variety. The results showed that 81 percent of the transcripts started in the first exon and the others in exon #4 (Supplementary Figure S3). In the fruits of mutant r²⁹⁹⁷, all the PSY1 transcripts were initiated in exon #1 (Figure 2). The primary one was a fusion of exon #4 with the retrotransposon sequence (transcript I in Figure 2). This fused transcript creates an early stop codon, similar to the one described in GTOM5 from yellow flesh r in the Ailsa Craig variety (Fray and Grierson, 1993a). Therefore, we conclude that r²⁹⁹⁷ in LA2997 and r in Ailsa Craig are essentially the same yellow flesh allele derived from the ancestral yellow-fruited tomato line. In addition to the most abundant abortive PSY1 transcript, fruits of r²⁹⁹⁷ produce additional transcripts resulting from alternative splicing events (Figure 2). These transcript variants lack the retrotransposon-containing exon #4 and include exons 5–9, which encode the C-terminus of the PSY1 protein where the catalytic domain is located (Cao et al., 2019). The functionality of the alternatively spliced variants was tested in E. coli cells. For this purpose, the cDNA of variants II and III (Figure 2), which contain the coding sequence of exons 5–9, and the normal cDNA sequence of exons 4–9 were cloned in the plasmid pBluescript SK⁺. These plasmids were transfected into E. coli cells carrying plasmid pACCRT-EI, which encodes the bacterial genes CrtE and CrtI for geranylgeranyl diphosphate synthase and phytoene desaturase, respectively. In the presence of phytoene, these enzymes can synthesize lycopene. As expected, the E. coli cells that expressed the full-length PSY1 were red due to lycopene accumulation. In contrast, PSY1 variants lacking exon #4 are not translated into enzymatically active proteins, as the alternatively spliced PSY1 variants II and III gave rise to colorless bacteria that did not produce carotenoids (Supplementary Figure S4). These results show that PSY1 variants lacking exon #4 are enzymatically inactive, as is evident in fruits of the mutant yellow flesh r²⁹⁹⁷.

The epistasis phenomenon found in the double mutant yellow flesh r²⁹⁹⁷/tangerine (Kachanovsky et al., 2012) was further investigated in various allele combinations. Fruits of the introgression line IL3-2, which carries the PSY1 gene from S. pennellii in the variety M82, are yellow and lack carotenoids due to a substantial reduction in PSY1 expression compared to its isogenic cultivated line M82 (Liu et al., 2003; Joung et al., 2009). Therefore, the PSY1 gene from S. pennellii in the genetic background of the cultivated tomato can be considered as a yellow flesh allele, r^sp. The IL3-2 line was crossed with the isogenic tangerine t³⁴⁰⁶ and zeta z²⁸⁰³ mutants, which are impaired in the genes CRTISO and ZISO, respectively, to produce in F2 generation double mutants IL3-2 r^sp/t³⁴⁰⁶ and IL3-2 r^sp/z²⁸⁰³. The genotypes of these mutants were confirmed with DNA markers. In agreement with previous observations of the yellow flesh allele r²⁹⁹⁷, fruit of the double mutant IL3-2 r^sp/t³⁴⁰⁶ accumulated carotenes, though at a lower level than the wild type, mainly in cis-configurations (Table 1). These results confirm that the epistasis of tangerine over yellow flesh is a general phenomenon and is not confined to a specific r allele. Another yellow flesh allele, r³⁷⁵⁶, which was isolated by mutagenesis in the tomato variety M82, contains an early stop codon in exon #4 of the PSY1 gene (Kachanovsky et al., 2012). Fruits of the double mutant r³⁷⁵⁶_/t³⁰⁰² were orange due to the accumulation of low levels of carotenes (Table 1 and Supplementary Figure S5). These results indicate that tangerine is also epistatic over the null yellow flesh allele r³⁷⁵⁶. As previously reported (Kachanovsky et al., 2012), the epistasis phenomenon is linked to the CRTISO impairment but not to the ZISO function loss because fruits of the double mutant IL3-2 r^sp/z²⁸⁰³ were yellow and contained only trace amounts of carotenoids (Table 1).

Previous results have shown that the mutation tangerine leads to a substantial increase in the transcription of PSY1 in r²⁹⁹⁷ fruit compared to the single mutant r²⁹⁹⁷ (Kachanovsky et al., 2012). To analyze the effect of tangerine on the expression of PSY1, we established several new qRT-PCR protocols with primers for exon #6 to measure the mRNA level of PSY1. The results confirmed that the transcript levels of PSY1 in the double mutant tangerine/r²⁹⁹⁷ appeared to be higher than in r²⁹⁹⁷, however, to a lower extent than previously reported (Figure 3A). The increase of PSY1 transcript in IL3-2(r^sp)/t³⁴⁰⁶ compared with IL3-2(r^sp) was relatively modest, but the data were not significant in this case (Figure 3B). Expression of PSY2 in fruits of the yellow flesh/tangerine double mutants was lower than in the yellow flesh alleles r²⁹⁹⁷, r³⁷⁵⁶, and IL3-2 (r^sp; Figure 3C). In compliance with the PSY3 expression pattern in tomato,¹ the root-specific SlPSY3 (Solyc01g005940) could not be detected with qRT-PCR in fruits of any of these lines.

Since tangerine obstructs carotenoid biosynthesis at the isomerization of cis-lycopene, we investigated the effects on phytoene synthesis of other pathway inhibitors. Norflurazon (Zorial) is a specific inhibitor of the plant enzyme phytoene desaturase (PDS; Sandmann et al., 1989). To test the effect of norflurazon, its commercial formula Zorial was injected into mature green fruits of the yellow flesh mutant r²⁹⁹⁷. After ripening, diffused red sectors appeared in the treated fruit due to the accumulation of carotenoids (Figure 4A). A similar experiment was done in ex-planta fruit tissues. Disks of pericarp taken from mature green fruits of yellow flesh r²⁹⁹⁷ and the wild-type variety Rutgers were incubated in vitro and treated with Zorial. The Rutgers fruit disks appeared yellow to orange, as opposed to red in the untreated disks. The colors of the mutants’ disks were not changed compared to untreated control. However, carotenoid composition in the disks indicated that PDS inhibition induced phytoene synthesis in both yellow flesh fruit mutants ex-planta (Figure 4B).

The effect of inhibition of carotenoid biosynthesis downstream to lycopene was investigated in yellow flesh mutants treated with the lycopene β-cyclase inhibitor 2-(4-Chlorophenylthio)-triethylamine hydrochloride (CPTA; Sandmann et al., 1989; Cunningham et al., 1993). CPTA was injected to mature green fruits of mutants r²⁹⁹⁷, r^3756, and IL3-2 (r^sp). In all mutants, the lycopene cyclase inhibition altered the fruit color in distinct sectors, which appeared red due to lycopene synthesis (Figure 5A, Table 2). However, the carotenoid composition varied among the different yellow flesh mutants. While in r²⁹⁹⁷, CPTA induced the accumulation of cis-carotenes, which led to red color appearance, in r³⁷⁵⁶ and IL3-2 (r^sp) it caused the accumulation of mainly lycopene (Table 2). The expression of PSY1 did not significantly change in the CPTA-treated sectors compared to the non-treated sectors. However, the expression level of PSY2 was doubled (Figures 5B,C).

To investigate the potential involvement of PSY2 in the restoration of phytoene production in the double mutants of tangerine and yellow flesh, expression of the gene PSY2 was inhibited by a transient VIGS. To this end, we created a tomato line that combined the mutations yellow flesh r²⁹⁹⁷ and tangerine t³⁰⁰² with the transgenic line overexpressing the transcription factors Delila (Del) and Rosea1 (Ros1; DR), as visual reporters for silencing in tomato fruit (Butelli et al., 2008; Orzaez et al., 2009). The quadruple mutant line was named DR/r²⁹⁹⁷/t³⁰⁰². A TRV-based silencing vector was constructed to silence both PSY2 and the transgenes Del and Ros1. The Si-RNA sequence was designed to specifically silence PSY2, but not PSY1 or PSY3 (Materials and Methods). The silencing TRV vectors were injected to green fruits of the lines DR/r²⁹⁹⁷/t³⁰⁰² and DR as a control. The silenced sectors were identified in ripe fruit by eliminating anthocyanins. In the DR/r²⁹⁹⁷/t³⁰⁰² fruit they appeared yellow and in DR they were red (Figure 6A). Carotenoid analysis indicated that non-silenced tissues of DR/r²⁹⁹⁷/t³⁰⁰² accumulated cis-carotenoids typical to the r²⁹⁹⁷/t³⁰⁰² double mutant, whereas carotenoid synthesis in the PSY2-silenced sectors was nearly abolished (Figure 6B; Supplementary Table S2). Similar silencing of PSY2 in the Del/Ros1 Moneymaker line did not change the carotenoid composition (Figure 6; Supplementary Table S2). Quantification of mRNA confirmed specific and efficient silencing of PSY2 in the silenced sectors, while the expression of PSY1 was unchanged (Figure 7). The elimination of carotenoid synthesis by PSY2 silencing in the DR/r²⁹⁹⁷/t³⁰⁰² line confirms that carotenoid biosynthesis recovery in the double mutant r²⁹⁹⁷/t³⁰⁰² depends on PSY2 activity.

Several yellow flesh mutants in the locus r have been identified in tomatoes (Tomato Genetics Resource Cente).² The allele r²⁹⁹⁷ (LA2997), which is described as a spontaneous mutation in the variety Rutgers, is considered the oldest allele (Chetelat, 2002) that was originated from the yellow variety brought to Europe in the 16th Century (McCue, 1952; Bergougnoux, 2014) and genetically characterized in the early 20th century (Price and Drinkard, 1909; Gilbert, 1912; Jenkins, 1948). We have confirmed that the mutation in r²⁹⁹⁷ is caused by an insertion of a copia-type retrotransposon Rider in the coding region of PSY1. The sequence of the insertion in exon #4 (Supplementary Figure S2) indicates a full-length Rider retrotransposon (Jiang et al., 2012). Previously, we missed identifying the mutation in r²⁹⁹⁷ while using genomic DNA sequencing based on PCR amplification (Kachanovsky et al., 2012). The existence of a complete retrotransposon bordered by identical long terminal repeats (LTR) often forms a stable stem-loop secondary structure. Most Taq-polymerases used in PCR amplifications skip stem-loop structures due to the replication slippage mechanism observed in the presence of direct repeats in the DNA (Viguera et al., 2001). Unlike previous analyses, here we used an engineered Taq polymerase, KAPA2G Fast, with higher processivity and speed under conditions that enabled the amplification of the full-length retrotransposon. LTR retrotransposons comprise more than 60% of the tomato genome. Most of them exist for millions of years in the exact chromosomal location (Tam et al., 2007; Cheng et al., 2009; Du et al., 2010; Jiang et al., 2012). Therefore, it is reasonable to assume that the retrotransposon mutation in r²⁹⁹⁷ is ancient and that it has prevailed throughout the tomato’s history since it was brought to Europe from America. The sequence of the retrotransposon LTR was previously detected in the GTOM5 mRNA from a yellow flesh mutation in the variety Ailsa Craig (NCBI Reference Sequence: X67143.1; Fray and Grierson, 1993a). Fray and Grierson (1993a) have predicted that this sequence belonged to a transposable element based on Southern blot hybridizations. According to the sequence of PSY1 in yellow flesh r²⁹⁹⁷ (Supplementary Figure S2), it is evident that this is the same r allele described before in the Ailsa Craig variety (Fray and Grierson, 1993a). The phenotype of yellow flesh r²⁹⁹⁷ has been ascribed to the lack of PSY1 transcription (Kachanovsky et al., 2012). This notion was supported by the fact that functional PSY1 genes introgressed to S. lycopersicum from green-fruited wild tomato species exhibit yellow flesh phenotypes, as demonstrated by the allele r^sp (Table 1). The existence of a 4.8 kb Rider retrotransposon in the first coding exon of PSY1 explains the drastic reduction of mRNA measured by RT-PCR analysis based on the amplification of cDNA sequences downstream of exon #4.

The epistasis of the mutation tangerine over yellow flesh r²⁹⁹⁷ was attributed to increased transcription of PSY1 in the double mutant r²⁹⁹⁷/t³⁰⁰² (Kachanovsky et al., 2012). Loss-of-function mutations in the gene CRTISO were found to increase PSY1 expression and enhance total carotenoids in fruits of tangerine tomato (Isaacson et al., 2002; Isaacson, 2005) and the yofi in melon (Cucumis melo; Galpaz et al., 2013). In both cases, the accumulation of cis-carotene intermediates could play a role in regulating gene expression (reviewed in: Cazzonelli et al., 2020; Escobar-Tovar et al., 2020). Measuring mRNA of PSY1 in fruits of r²⁹⁹⁷/t³⁴⁰⁶ by RT-PCR amplification of exon #6 sequence showed a slight elevation compared to yellow flesh r²⁹⁹⁷ (Figure 3). However, this small increase in Psy1 transcript cannot explain the restoration of phytoene synthase activity in the double mutant. This result differs from previous measurements in strain r²⁹⁹⁷/t³⁰⁰² (Kachanovsky et al., 2012). The explanation for the discrepancy can be related to the different genetic backgrounds of t³⁰⁰² (Rutgers) and t³⁴⁰⁶ (M82). Moreover, the Rider retrotransposon in exon #4 is inserted after codon 107 of the PSY1 coding sequence. Consequently, the primary transcript of PSY1 in r²⁹⁹⁷ consists of exons 1–3 and part of exon #4 fused to the LTR sequence of the retrotransposon (Figure 2). It thus potentially encodes a chimeric non-functional polypeptide of 178 amino acids. Other rare transcript variants of PSY1 in r²⁹⁹⁷ skip exon #4, which encompasses the retrotransposon (Figure 2). However, a truncated PSY1 translated from exons 5–9 was inactive in the E. coli complementation assay (Supplementary Figure S4). These results indicate that the tomato mutant r²⁹⁹⁷ lacks a functional PSY1 enzyme and, therefore, the gene expression level is irrelevant. PSY1-specific antibodies are not available, and consequently, it was not possible to obtain quantitative data on the aberrant PSY1 protein level. The phenomenon of tangerine epistasis has also been observed with allele r³⁷⁵⁶ that carries a loss-of-function mutation in PSY1 (Table 1). This result challenges previous conclusions derived from the double mutant r³⁷⁵⁶/t³⁰⁰² on the epistasis of tangerine over yellow flesh (Kachanovsky et al., 2012).

The results of the experiments described here (summarized in Figure 8) support an alternative option for recovering phytoene synthesis in the yellow flesh/tangerine double mutants by activating the leaf-specific phytoene synthase, PSY2. PSY1 is mainly expressed in the fruit, where its transcripts increase dramatically at the “breaker” ripening stage from practically undetectable levels at the mature green fruit. By contrast, PSY2 transcript level is quite similar in green and ripening fruit (Fraser et al., 1999, 2007; Fernandez-Pozo et al., 2017; Shinozaki et al., 2018; Tomato Expression Atlas).³ Moreover, in yellow flesh fruit, the PSY2 protein has been detected immunologically, and its enzymatic activity of phytoene synthesis has been demonstrated in vitro (Fraser et al., 1999). Nevertheless, PSY2 is not active in the fruits when PSY1 is impaired in yellow flesh mutants (Table 1; Fray and Grierson, 1993a; Fraser et al., 1999; Kachanovsky et al., 2012; Fantini et al., 2013; Chen et al., 2019; Zhao et al., 2020). The minuscule amount of carotenoids, mainly lutein and β-carotene, found in the fruits of yellow flesh, tangerine and zeta mutants, as well as in the norflurazon-treated wild-type fruits, are more likely residues from the chloroplasts prior to their transition to chromoplasts rather than a low basal activity of PSY2. An upregulation of PSY2 transcription in the yellow flesh/tangerine double mutants was ruled out (Figure 3). The unequivocal evidence that PSY2 sustains phytoene synthesis in r²⁹⁹⁷/t³⁰⁰² fruits is the outcome of PSY2 silencing in these fruits. As seen in Figure 6, the lack of PSY2 abolished carotenoid biosynthesis in the double mutant. Why then PSY2 is not active in yellow flesh fruits, and what activates it in the genetic background of tangerine? It has been hypothesized that PSY2, which usually operates in chloroplasts, is not active in fruits due to failure to interact with other enzymes in the carotenogenic pathway in chromoplasts (Fraser et al., 1999). In such a case, tangerine could alter the carotenoid metabolon in chromoplasts in a way that enables PSY2 to function. One possibility is the localization of the enzyme in the plastids. The location of the biosynthesis enzymes, including PSY, is associated with the accumulation and sequestration of carotenoids (reviewed in: Li et al., 2016). A change in the PSY2 compartment within the plastids could enable its accessibility to the carotenoid metabolon in the chromoplasts. It was reported that PSY isozymes from various plants differ in their plastid sub-organellar localization, such as specific membranes or plastoglobuli (Shumskaya et al., 2012; Ampomah-Dwamena et al., 2015). The mutation tangerine eliminates carotenoids downstream to lycopene and causes an abnormal accumulation of cis-carotenes, mainly tetra-cis-lycopene (prolycopene). These changes in carotenoid constituent could modify the plastidial membrane organization, as seen in the Arabidopsis ccr mutants, which alter the membranous structures in etioplasts (Park et al., 2002). In this context, it must be emphasized that the effect of the tangerine mutation is already manifested in mature green fruit where significant amounts of cis-carotenoids accumulate in chloroplasts prior to their transition to chromoplasts (Isaacson et al., 2002).

Similar phenomena are likely to occur when carotenoid biosynthesis in fruits chromoplasts is interrupted by inhibitors. Fruits or pericarp tissues at the mature green stage treated with norflurazon accumulate upon ripening excessive amounts of phytoene (Figure 4; Filler-Hayut, 2012). The buildup of phytoene, which was begun in the chloroplasts, possibly affected the assembly of membranes during the transition to chromoplasts in a manner that enabled PSY2 functioning. The difference in the carotenoid composition following treatment with norflurazon between ex-planta pericarp and whole fruit of yellow flesh (Figure 4) can be attributed to the diffusion of the inhibitor in the fruits during ripening that dilutes it to a low concentration that no longer blocks PDS. Moreover, the borders of the fully inhibited fruit sectors are diffused and sometimes unclear.

Similar to norflurazon, the lycopene cyclase inhibitor CPTA also induced carotenoid biosynthesis in fruits of yellow flesh r³⁷⁵⁶ (Figure 5). Treatment with CPTA increased protein level of phytoene synthase in Narcissus pseudonarcissus flowers (Al-Babili et al., 1999) and caused transcriptional changes in Citrus sinensis (Lu et al., 2019). A slight increase of PSY2 expression was measured in CPTA-treated fruit sectors while PSY1 transcript levels were unchanged (Figure 5). However, since allele r³⁷⁵⁶ carries a null mutation in PSY1, the induction of lycopene synthesis must have occurred due to activation of the PSY2 enzyme. In green tissues, CPTA eliminates cyclized carotenoids and leads to an accumulation of lycopene and other intermediate cis-carotenes (Fedtke et al., 2001; La Rocca et al., 2007). These changes in yellow flesh fruits treated with CPTA at the green stages could alter sub-organellar structures in the chloroplasts that linger during their transition to chromoplasts and influence PSY2 activity in ripening fruits. Activation of phytoene synthesis in fruits of yellow flesh mutants as a result of lycopene cyclase inhibition by CPTA has been recently reported by Gupta et al. (2022). The authors discovered that CPTA did not influence the expression levels of carotenoid biosynthesis genes and that ectopic accumulation of lycopene in chloroplasts was associated with the transition from chloroplasts to chromoplasts and the activation of PSY2 (Gupta et al., 2022). A comparable but slightly different case exists in pepper (Capsicum annuum) where PSY1 is the key enzyme responsible for fruit color and PSY2 functions in leaves (Berry et al., 2019; Dyachenko et al., 2020; Wei et al., 2020). However, it was demonstrated that PSY2 activity contributes to synthesizing a basal level of carotenoids in the fruit when PSY1 is not functional (Jang et al., 2020).

It has been established in Arabidopsis, melon, and sweet potato that the DnaJ chaperon protein ORANGE (Or) stabilizes PSY and regulates its activity by direct interaction between the two proteins (Zhou et al., 2015). Differential interaction of Or with paralogous PSY enzymes was reported in saffron (Ahrazem et al., 2020). The activation of PSY2 prompted by changes in the carotenoid composition may be enabled by a unique interaction of Or with PSY2 that either stabilizes the enzyme or interferes with its import to chromoplasts (Yuan et al., 2021).

Interestingly, disruption of carotenoid biosynthesis at ζ-carotene in the mutant zeta (z²⁰⁸³) did not activate phytoene synthesis in yellow flesh fruits despite a significant accumulation of phytoene (Table 1; Kachanovsky et al., 2012). This phenomenon illustrates that only accumulation of a specific carotenoid intermediates are leading to activation of PSY2.

In conclusion, our results demonstrate that although PSY1 is the sole enzyme that produces phytoene in tomato fruit, under certain circumstances, PSY2 can be activated to sustain carotenoid biosynthesis. The results support the hypothesis on the existence of a carotenoid biosynthesis metabolon with distinct features in chloroplasts and chromoplasts. Furthermore, although transcriptional regulation is the primary mechanism determining carotenoid biosynthesis in tomato fruit, additional post-transcriptional mechanisms also play a role in this process. Additional future experiments that will determine the protein levels of PSY1 and PSY1 and their sub-organellar localization are required to elucidate the mechanism underlying the activation of PSY2 in fruit chromoplasts.