A common phytoene synthase mutation underlies white petal varieties of the California poppy

The California poppy (Eschscholzia californica) is renowned for its brilliant golden-orange flowers, though white petal variants have been described. By whole-transcriptome sequencing, we have discovered in multiple white petal varieties a single deletion leading to altered splicing and C-terminal truncation of phytoene synthase (PSY), a key enzyme in carotenoid biosynthesis. Our findings underscore the diverse roles of phytoene synthase in shaping horticultural traits, and resolve a longstanding mystery of the regaled golden poppy.

Since white poppy petals are deficient in carotenoid pigments 11 , we focused on genes of the carotenoid biosynthetic pathway. Comparing expression of carotenoid biosynthesis genes between white and orange poppy varieties, only phytoene synthase (PSY) showed significantly altered expression, with an average 2.5-fold reduced transcript levels in white varieties (P = 0.003, two-sided Student's t-test) (Fig. 1a, right). While this finding focused attention on PSY, the modest reduction in white varieties was unlikely to account for a near absence of carotenoid pigment.
Comparing the aligned PSY transcript reads between white and orange petal poppy varieties, all four white varieties (but none of the three orange varieties) exhibited an apparent 5 bp gap within the PSY transcript (Fig. 2a,b and Supplementary Fig. 1). The gap occurred within the coding region, at the site of an inferred exon-exon junction (by comparison to the tomato reference genome). To define the alteration at the genome level, we designed PCR primers to amplify and sequence across the exon-exon junction from genomic DNA (isolated from poppy seeds). In the white petal varieties, the resultant PCR product was consistently smaller (Fig. 2c). Sequence alignment revealed a 76 bp deletion within the PSY intron, which extended through the 3′ splice acceptor site and 5 bp of the downstream exon (explaining the apparent 5 bp alignment gap from the RNAseq reads) (Figs 2d, 3). By comparing the white-petal PSY genome sequence and assembled transcript contig, loss of the splice acceptor site led to usage of a cryptic splice acceptor site within the intron, resulting in a coding frameshift with early translational termination and predicted C-terminal truncation of the PSY protein ( Fig. 2d and Fig. 3). Early translational termination is associated with nonsense-mediated mRNA decay 13 , consistent with our observed reduced PSY transcript levels. Notably, the C-terminal truncation abolishes a highly-conserved putative enzyme active site (DXXXD motif) in PSY (Fig. 2b,d) 14 .
That the PSY deletion is present in all four white petal varieties (minimally 8 alleles if diploid), but in none of the three orange petal varieties (minimally 6 alleles), demonstrates strong segregation with the white petal phenotype (P = 0.0003, two-sided Fisher's exact test). Together, the genetic and inferred biochemical (predicted loss of active site) data provide strong evidence that the 76 bp deletion underlies the white petal trait. Moreover, that all four white petal poppy varieties (notwithstanding subtle differences in petal hues) harbor precisely the same deletion suggests that they were likely all derived from a single white-petal poppy origin.
In addition to studying commercial varieties, we also sought to examine white-petal specimens among wild poppy plants. To this end, we undertook expeditions to find and collect white petal specimens among California poppy fields across Santa Clara, San Mateo, and Solano counties. Of three specimens collected, all harbored precisely the same PSY mutation (Fig. 2e), suggesting that they likely represent commercial seed contaminants among orange petal varieties that were seeded rather than wild plants. Interestingly, in one field we noted orange and white poppies together with an uncommon yellow-orange petal variant. PCR analysis revealed that the yellow-orange poppy carried both the wildtype and deletion PSY allele, suggesting an F1 hybrid between previously seeded orange and white petal varieties (Fig. 2e).
Carotenoids are flower petal pigments, but they also provide essential roles as accessory pigments and antioxidants in chloroplasts for photosynthesis 4 . Thus, given the PSY null mutation identified from flower buds, the existence of other PSY encoding genes seemed likely. To investigate that possibility, we carried out RNAseq from green leaf material from orange and white petal poppy varieties. Aligning the reads, only a small fraction of the www.nature.com/scientificreports www.nature.com/scientificreports/ PSY reads from the white-petal leaf specimen exhibited the deletion (Fig. 4a). A distinct set of single nucleotide polymorphisms (SNPs) present only in the leaf RNA segregated with the non-deletion reads, allowing us to design haplotype-specific PCR primers to amplify across the exon-exon junction. Notably, PCR of genomic DNA using the non-deletion haplotype-specific primers revealed two larger PCR products (Fig. 4b), where sequencing disclosed two different intron sequences (Fig. 4c). This finding indicates the presence of two additional PSY genes (which we have designated PSY1B and PSY1C), expressed in poppy leaves.
Based on the relative frequencies of SNPs in the PSY transcripts from petal and leaf tissue (Supplementary Fig. 2   Close-up view highlighting the 5 bp alignment gap in the white-petal varieties. Note, the alignment gap overlaps with a putative enzyme active site (DXXXD motif, indicated below). (c) PCR across the exon-exon junction (site of alignment gap) results in a shorter PCR product (red arrow), indicative of genomic DNA deletion in the white-petal varieties. Image of full-length gel is available in Supplementary Fig. 4. (d) Illustration summarizing the PSY gene, mRNA and protein products inferred from sequence alignment of the PCR products (PCR primers indicated; see Fig. 3 for annotated cDNA sequences). In all four white-petal varieties, a 76 bp intronic deletion (red hatched box) ablates the 3′ splice acceptor. Usage of a cryptic splice acceptor leads to a coding frameshift with early translational termination. The resulting C-terminal truncation destroys one of two highlyconserved putative enzyme active sites (pink rectangles); brown lines indicate PSY conserved motifs. (e) PCR analysis (and subsequent sequencing) reveals the identical 76 bp intronic deletion in three white-petal poppy plants discovered among ostensibly wild California poppy fields from three geographically distinct locales. Note, a yellow-orange petal poppy plant discovered in Field 1 (also pictured in inset, white arrow) carries both the PSY wildtype and deletion allele, suggesting that it represents an F1 hybrid cross between an orange and white petal variety. Image of full-length gel is available in Supplementary Fig. 4. is more abundantly expressed (accounting for 97% of leaf PSY transcripts in orange petal varieties, reduced to 78% in white-petal varieties) compared to PSY1B/1C. California poppy PSY1A exhibits high (99%) homology to PSY1B/1C at the nucleotide sequence level, and 100% identity at the amino acid sequence level, suggesting relatedness by recent gene duplication. However, we note limitations of our analysis, including variable read coverages (particular at the ends of the PSY genes), and the challenges of phasing short reads and assigning SNPs to individual genes. A definitive analysis will require cloning the individual PSY cDNAs and genome loci.
The finding of multiple PSY gene paralogs in plants, first detailed in the tomato 15 , is now common. Like for the California poppy, some such PSY paralogs are expressed primarily in green (photosynthetic) tissues, while others drive carotenoid accumulation in flowers, fruits, or roots. For example, in the tomato (Solanum lycopersicum), PSY1 is predominantly expressed in the petals and ripening fruit, while PSY2 is predominant in leaves 16 . In the loquat (Eriobotrya japonica), PSY1 is expressed in the fruit peel, PSY2A in the ripening fruit flesh, and PSY2B in leaves 17 . And in the carrot (Daucus carota), PSY1 and PSY2 are expressed in the root, while PSY1 is also expressed in leaves 18 . A comparative analysis of PSY protein sequences among eudicots reveals California poppy PSY to be most closely related to PSY from the recently sequenced opium poppy (Papaver somniferum) genome 19 , and overall more closely related to the so-called Eudicot PSY1 clade ( Fig. 5 and Supplementary Fig. 3) 20 .
In summary, by transcriptome sequencing of California poppy flower buds, we have identified a frameshifting deletion in phytoene synthase that is common to multiple commercial white petal varieties. All have distinct white hues (likely due to different genetic backgrounds), but nonetheless appear to have been bred from the same common originator. Importantly, the white-petal trait in 15 different natural and commercial California poppy variants was previously shown to map to a single genetic locus 12 . That study included the Alba and Ivory Castle varieties also analyzed here. Thus, we can infer that PSY1A mutations underlie all previously studied white-petal California poppy variants. Whether other variants share the same 76 bp frameshifting deletion mutation remains to be determined.
PSY variants/mutations have previously been associated with agriculturally important traits, e.g., color variation in tomatoes, peppers, cassavas, and loquats 17,21-23 . We have now also connected PSY mutations to ornamental horticulture. Our discovery resolves a decades old mystery of the molecular underpinnings of white-petal California poppies, and adds to the cultural legacy of the California golden poppy.

Methods plant materials.
Commercial California poppy varieties were purchased as seeds from Eden Brothers (Ivory Castle, White Linen, Alba, Orange, and Golden West), Vermont Wildflowers (White), and Cornucopia (California Golden). Seeds were germinated in individual pots, and subsequent developing flower buds collected and frozen on dry ice. In some cases, poppy leaf material was also collected. Mature flowers from the same plants were examined and photographed to verify the advertised varieties. We also collected ostensibly wild California poppy flower samples from fields across three San Francisco Bay Area counties. For RNA isolation, plant material (flower buds with calyx caps removed, or leaves) was pulverized in liquid nitrogen using a mortar and pestle, and  Castle flower, and Ivory Castle leaf. Note that the 5 bp alignment gap present in Ivory Castle flower is observed in only a minority of reads from Ivory Castle leaf. Note also in Ivory Castle leaf the presence of 3 SNPs flanking the gap that segregate with the non-deletion reads, permitting design of haplotype-specific (leaf vs. flower expressed) PCR primers (shown below). The 3 SNPs are also present in Orange leaf, but masked by the higher (flower) PSY expression. (b) Haplotype-specific PCR of genomic DNA across the exon-exon junction (site of alignment gap) using leaf-expression specific PCR primers results in two longer PCR products (blue arrows), distinct from the shorter flower-expression specific PCR products, and indicative of additional PSY genes (annotated PSY1B and PSY1C) expressed in poppy leaves. Image of full-length gel is available in Supplementary  Fig. 4. (c) Partial genome sequences for PSY1A (top), PSY1B (middle), and PSY1C (bottom), as determined by exon-spanning PCR. Intron sequences are in lower case text.