Manipulation of ZDS in tomato exposes carotenoid‐ and ABA‐specific effects on fruit development and ripening

Summary Spontaneous mutations in fruit‐specific carotenoid biosynthetic genes of tomato (Solanum lycopersicum) have led to improved understanding of ripening‐associated carotenogenesis. Here, we confirm that ZDS is encoded by a single gene in tomato transcriptionally regulated by ripening transcription factors RIN, NOR and ethylene. Manipulation of ZDS was achieved through transgenic repression and heterologous over‐expression in tomato. CaMV 35S‐driven RNAi repression inhibited carotenoid biosynthesis in all aerial tissues examined resulting in elevated levels of ζ‐carotene isomers and upstream carotenoids, while downstream all trans‐lycopene and subsequent photoprotective carotenes and xanthophylls were diminished. Consequently, immature fruit displayed photo‐bleaching consistent with reduced levels of the photoprotective carotenes and developmental phenotypes related to a reduction in the carotenoid‐derived phytohormone abscisic acid (ABA). ZDS‐repressed ripe fruit was devoid of the characteristic red carotenoid, all trans‐lycopene and displayed brilliant yellow pigmentation due to elevated 9,9′ di‐cis‐ζ‐carotene. Over‐expression of the Arabidopsis thaliana ZDS (AtZDS) gene bypassed endogenous co‐suppression and revealed ZDS as an additional bottleneck in ripening‐associated carotenogenesis of tomato. Quantitation of carotenoids in addition to multiple ripening parameters in ZDS‐altered lines and ABA‐deficient fruit‐specific carotenoid mutants was used to separate phenotypic consequences of ABA from other effects of ZDS manipulation and reveal a unique and dynamic ζ‐carotene isomer profile in ripe fruit.


Introduction
Carotenoids play pivotal roles throughout plant development, among the most observable being coloration of ripe fruit and flowers for attraction of seed dispersing frugivores and pollinators. Carotenoids are additionally requisite to maintain proper function of the photosynthetic apparatus. Photoprotective carotenes (i.e. b-carotene) and xanthophylls (i.e. lutein, zeaxanthin, violaxanthin and neoxanthin) are localized in photosystems I and II where they absorb excess light energy and limit damage from adjacent excited chlorophylls through quenching of reactive singlet oxygen (Amunts et al., 2007;Jahns and Holzwarth, 2012). Numerous aspects of plant development, dormancy and stress responses are dependent on the synthesis of the carotenoid-derived phytohormone abscisic acid (ABA) (Cutler et al., 2010;Hauser et al., 2011;Mauch-Mani and Mauch, 2005;Vishwakarma et al., 2017;Zhang et al., 2006). Recent evidence indicates additional roles of carotenoids as precursors for important regulatory functions via strigolactones (reviewed in Waters et al., 2017) in addition to signalling functions with other organisms through volatile carotenoid metabolites (reviewed in Hou et al., 2016;McQuinn et al., 2015).
Utilization of available natural tomato (Solanum lycopersicum) carotenoid biosynthetic mutants in combination with studies in other model species (e.g. Arabidopsis thaliana and Zea mays) elucidated a nearly comprehensive carotenoid biosynthetic pathway ( Figure S1). Moreover, duplication of genes in a majority of biosynthetic steps in tomato illuminated chloroplast-and chromoplast-specific pathways clarifying ripening-associated carotenogenesis (Galpaz et al., 2006;Sato et al., 2012; Figure S1 and Table S1). Natural mutations in genes involved in chromoplast-specific carotenogenesis (e.g. r, PSY1; tangerine, CRTISO; Beta, CYC-B/CRTL-B2) have allowed researchers to functionally define many steps of the pathway relevant to fruit ripening (Fray and Grierson, 1993;Isaacson et al., 2002;Ronen et al., 2000) and led to some of the colour variation in tomato (yellow and orange fruit varieties) appreciated by consumers. However, the lack of gene duplication within the poly-cis-transformation of 15-cisphytoene to all trans-lycopene limits the assessment of ripening roles for these genes.
In the poly-cis-transformation of 15 cis-phytoene to all translycopene, 15-cis-phytoene undergoes four desaturation steps carried out by two desaturases (phytoene desaturase/PDS and zeta-carotene desaturase/ZDS) in conjunction with two intermediary isomeric conversions carried out by two isomerases (zetacarotene isomerase/ZISO and carotenoid isomerase/CRTISO; Figure S1; Beltr an et al., Brausemann et al., 2017;Chen et al., 2010b;Dong et al., 2007;Isaacson et al., 2002;Qin et al., 2007). Unlike ZISO and CRTISO which carry out reactions that can alternatively be catalysed by light in photosynthetic tissues (Isaacson et al., 2002;Li et al., 2007, respectively), mutations in the single copy genes PDS and ZDS manifest in dramatically reduced plant fitness reflecting highly unstable photosynthetic tissues (Dong et al., 2007;Qin et al., 2007). This is likely the reason that, until recently, PDS and ZDS remained uncharacterized in terms of tomato fruit ripening (McQuinn et al., 2017). Initial characterization of PDS and ZDS was limited to their transient repression via virus-induced gene silencing (VIGS) due to their indispensable nature in photosynthetic tissues (Fantini et al., 2013). More recently, we further characterized PDS via heterologous over-expression of the Arabidopsis PDS in tomato and investigated its impact on the carotenoid accumulation and gene expression throughout the aerial tissues of the tomato plant (McQuinn et al., 2017).
Genes involved in synthesis of 15 cis-phytoene and its poly-cistransformation to all trans-lycopene are induced transcriptionally at the onset of fruit ripening giving the tomato its characteristic red colour. It is well established that the initiation of physiological changes associated with tomato ripening, including carotenoid accumulation, is induced by a climacteric burst of ethylene, and is dependent on an elaborate combination of epigenetic and transcriptional dynamics influenced to varying degrees by ripening-specific transcriptional regulators RIN (a MADS-box transcription factor), NAC-NOR (a NAC domain transcription factor) and CNR (a Squamosa promoter binding protein; Eriksson et al., 2004;Gallusci et al., 2016;Giovannoni et al., 2017;Giovannoni et al., 2004;Ito et al., 2017;L€ u et al., 2018;Manning et al., 2006;Vrebalov et al., 2002).
As the main rate-limiting step of carotenogenesis, phytoene synthase (PSY) has garnered much attention regarding its transcriptional and post-transcriptional regulation across various plant species (i.e. Arabidopsis, Zhou et al., 2015;Melon, Chayut et al., 2017). This holds true in tomato, where PSY1 has been well studied and shown to be transcriptionally regulated by ethylene perception and signalling (Gapper et al., 2013), and a direct target of RIN-mediated transcriptional regulation (Martel et al., 2011), which remains dependent on demethylation of its promoter prior to ripening (Zhong et al., 2013). Further, numerous strategies targeting PSY1 have been deployed to enhance carotenoid accumulation in tomato (reviewed in Fraser et al., 2009). Recently, researchers enhance carotenoid accumulation in tomato upon ectopic expression of the Arabidopsis ORANGE protein, which has dual roles in post-transcriptionally regulating PSY as well as inducing chromoplast biogenesis (Yazdani et al., 2019). While such strategies have proven successful, new bottlenecks emerge in subsequent desaturase steps carried out by PDS and ZDS, thereby limiting their effectiveness. These secondary bottlenecks present additional targets for manipulation to enhance carotenoid content, one of which, PDS, has been successfully exploited to increase downstream health-promoting carotenoids demonstrating its potential in carotenoid biotechnology strategies (McQuinn et al., 2017).
Regarding the enzymes carrying out the poly-cis-transformation of 15-cis-phytoene to all trans-lycopene, very little is known regarding their regulation, other than their induction by ethylene perception and signalling upon initiation of tomato fruit ripening (Alba et al., 2005). Further, while ZISO and CRTISO have been characterized via natural mutations (Gonda et al., 2019;Isaacson et al., 2002, respectively) and now PDS through transgenic approaches (Fantini et al., 2013;McQuinn et al., 2017), ZDS remains the least characterized step in tomato fruit ripening-associated carotenogenesis. Therefore, transgenic manipulation of ZDS and exploration into its regulation is essential and may aid future strategies aimed at enhancing carotenoid content beyond levels achieved via PSY exploitation.
Given the limited knowledge of ZDS in terms of its regulation and phenotypic consequences on fruit development and ripening, we explored the consequences of transgenic manipulation of ZDS throughout tomato fruit development, investigating its function via RNAi-guided repression and over-expression using a heterologous gene from Arabidopsis (AtZDS). The heterologous overexpression of AtZDS was implemented to reduce potential cosuppression of the endogenous tomato ZDS gene, a strategy that has proven useful for the manipulation of PDS in tomato (McQuinn et al., 2017). Herein, we identify ZDS as a heavily regulated gene in carotenogenesis by multiple ripening regulatory factors. Additionally, we highlight the consequences of ZDS manipulation on fruit ripening carotenogenesis and its potential for the enhancement of carotenoid content in tomato fruit. Lastly, we confirm diminished ABA and carotenoid accumulation throughout tomato fruit development reduces fruit growth and expansion, delays the initiation of fruit ripening and alters fruit quality (i.e. primary metabolites and brix o content).

Results
ZDS is a single copy gene in tomato strongly regulated by ripening transcription factors and ethylene In tomato, multiple steps of carotenoid biosynthesis are represented by multigene families (e.g. Solanum lycopersicum PSY1; SlPSY2; SlPSY3, and SlLCY-B/ CRTL-B1; SlBCYC/ CRTL-B2; Table S1), the origins of which can be traced in some cases to a whole-genome triplication event Tomato Genomics Consortium (TGC, 2012). These genes have been further categorized according to chloroplast and chromoplast specificity (Table S1; Galpaz et al., 2006). Chromoplast-or fruit ripening-specific genes have been functionally characterized in tomato through the study of natural mutations that have minor effects, if any, on non-fruit photosynthetic tissues. Given that a tomato ZDS mutant has not yet been reported, and consistent with the many deleterious phenotypes of Arabidopsis and Zea mays ZDS mutants, it is likely that ZDS is represented by a single copy gene in tomato and many other widely studied plant species. Queries of available protein sequences using tomato and Arabidopsis ZDS (SlZDS and AtZDS, respectively) as the query in multiple sequence/genome databases (NCBI, http://www.ncbi.nlm.nih.gov; TAIR, www.arabidopsis.org; Solgenomics, Solgenomics.net) indicates this is the case for tomato and many other plant genomes. It is noteworthy that genome duplication in apple and orange may have led to an increase from one to two or more paralogous copies of ZDS, respectively, in these lineages (Chen et al., 2010a;Velasco et al., 2010).
Analysis of ZDS transcript accumulation in above ground tissues of tomato via qRT-PCR demonstrates ZDS is expressed in all tissues analysed, consistent with available microarray and RNAseq data (TGC, 2012;Fernandez-Pozo et al., 2017;Shinozaki et al., 2018;Zouine et al., 2017;ted.bti.cornell.edu). ZDS was most highly expressed in the leaves, ripening fruit and flowers, in increasing order (Figure 1a). 9,9 0 di-cis-f-carotene is converted to all trans-lycopene by ZDS in conjunction with carotenoid isomerase (CRTISO), whose expression pattern for the most part paralleled ZDS (Figure 1a). ZDS expression analysis through a comprehensive fruit developmental time course reveals ZDS Because ZDS expression is induced upon ripening initiation, ZDS transcript levels were quantified in tomato ripening transcription factor (ripening inhibitor, rin, and non-ripening, nor) and ethylene insensitive (Green ripe, Gr, and Never ripe, Nr) mutants at the 7 days post-breaker (7DPB) stage (Barry et al., 2005;Giovannoni et al., 2004;Hackett et al., 2000;Vrebalov et al., 2002). For accurate fruit staging, wild-type and mutant flowers were all tagged at anthesis and ripening mutant fruits were collected at the same number of days after pollination as that of 7DPB wild-type fruit where 'breaker' was determined by the first sign of colour change. ZDS transcripts are reduced in all four mutants with the greatest reduction observed in nor fruit. These results highlight NOR, a NAC domain transcription factor, as influential in the regulation of ZDS, and remain consistent with the RIN transcription factor and ethylene all as ripening-related ZDS regulatory components (Figure 1c).

ZDS-RNAi lines demonstrate inhibited carotenogenesis in flowers and fruits
ZDS expression was suppressed via RNA interference driven by the constitutive CaMV 35S promoter in the pHELLSGATE2 vector containing cDNA targeting the gene-specific 3'UTR of ZDS. Agrobacterium tumefaciens-mediated transformation of wildtype tomato (S. lycopersicum cv Ailsa Craig) yielded eight independent transgenic ZDS-RNAi lines, three of which (ZDS.2, ZDS.5 and ZDS.7) were selected for further characterization based on insert number, and limited deleterious effects on photosynthetic tissues ensuring flower and fruit set ( Figure S2). A number of chlorotic RNAi lines were removed from the analysis due to their inability to propagate. Quantitative PCR confirmed lines ZDS.5 and ZDS.7 as single insertion lines, whereas ZDS.2 contains two transgene insertions. Plants heterozygous for both inserts (in the case of ZDS.2) and for the single insert in lines ZDS.5 and ZDS.7 were used for subsequent characterization as homozygous plants were severely dwarfed and incapable of proliferation and propagation ( Figure S3). RNAi resulted in significant repression of ZDS in young leaves, anthesis flowers and ripe fruits of the three independent RNAi lines as compared to wild-type and all three lines displayed phenotypes consistent with altered carotenogenesis (Figure 2). The strongest repression of ZDS was observed in anthesis flowers (~8-fold decrease) and in ripe fruit (~16-fold decrease), while expression was repressed approximately twofold in leaves of repression lines (Figure 2a).
Leaves of two-month-old plants began to photobleach when removed from low light conditions and placed in natural light ( Figure S2). Interestingly, this susceptibility to photobleaching was temporary, as newly formed leaves became visually indistinguishable when compared to wild-type throughout the remainder of the growing cycle. Similar results were reported by Dong et al. (2007) for the weaker Arabidopsis ZDS-mutant allele of spontaneous cell death 1 (spc1-1) when grown in short-day conditions. These observations are consistent with photoprotective carotenes and xanthophylls accumulating to wild-type levels in the leaves of the ZDS-RNAi lines (Table 1). Further, given the accumulation of b-carotene and xanthophylls remains unchanged in leaves, it is likely that carotenoid-derived hormones, strigolactone and ABA are not substantially impacted in those tissues as well.
In contrast, photobleaching is prevalent throughout fruit development of ZDS-RNAi lines, with the double-insert ZDS.2 line being the most severe ( Figure 2b). Accumulation of all translycopene, the characteristic red pigment of ripe tomato fruit, is inhibited throughout ripening of ZDS-RNAi tomato fruit as observed in fruit at the full-ripe stage (i.e. 7DPB; Figure 2b). Reduced pigmentation was also observed in ZDS-RNAi anthesis flowers compared with wild-type (Table 1 and Figure S2). Additional phenotypes were observed consistent with reduced synthesis of the carotenoid-derived hormone ABA in seeds ( Figure S4).
Carotenoid profiles in ZDS-RNAi and wild-type control tissues were assessed using high-performance liquid chromatography (HPLC). Wild-type leaves, flowers and ripe fruit accumulated undetectable or low levels of linear carotenoids upstream of ZDS, while all trans-lycopene and/or downstream cyclic carotenoids were abundant (Table 1). CaMV 35S-driven repression of ZDS in leaves, flowers and ripe fruit of the three transgenic lines blocked the pathway to varying degrees, resulting in the accumulation of upstream carotenoids phytoene and phytofluene, while fcarotene isomers only accumulated in anthesis flowers and ripe fruit (Table 1). Downstream photoprotective carotenes (i.e. bcarotene) and xanthophylls still accumulate to near wild-type levels (Table 1). It is possible that the modest reduction of ZDS mRNA in the leaves as compared to fruit accounts for this observation. In contrast, anthesis flowers and ripe fruit showed a statistically significant decrease in the synthesis of downstream carotenoids (lycopene and b-carotene), and total xanthophylls compared to wild-type (Table 1), likely due to the greater inhibition of ZDS mRNA accumulation in these tissues.

AtZDS over-expression reveals an additional bottleneck in fruit ripening-associated carotenogenesis
To further investigate the relevance of changes in ZDS mRNA accumulation in tomato fruit, heterologous over-expression of Arabidopsis thaliana ZDS (AtZDS) in tomato driven by the CaMV 35S promoter was employed using the pART27 vector system (Gleave, 1992). As noted above, we deployed heterologous overexpression rather than use of the tomato ZDS gene to reduce potential for co-suppression of the endogenous gene, a strategy proven useful for manipulation of tomato PDS (McQuinn et al., 2017). Nine independent transgenic lines were identified and one, AtZDS.OE.8, demonstrated phenotypes similar to ZDS-RNAi lines, suggesting ZDS co-suppression ( Figure S5). Of the remaining eight lines, four (AtZDS.OE.1A, AtZDS.OE.4, AtZDS.OE.7, AtZDS.OE.9), each containing a single insert and presenting consistent phenotypic variation, were carried to the T2 generation.
The effectiveness of AtZDS.OE was determined via qRT-PCR of the AtZDS transgene in young leaves, anthesis flowers and 7DPB fruit from each independent transgenic line ( Figure 3). CaMV 35S-driven expression of AtZDS was successful with all tissues analysed displaying substantial transgene mRNA accumulation. The native Solanum lycopersicum ZDS, SlZDS, is slightly reduced in some of the tissues analysed in the AtZDS.OE lines, possibly reflecting a response to the transgene ( Figure 3a). Nevertheless, total ZDS mRNA was greatly increased in all tested tissues and lines. Young leaves showed the lowest relative increase in AtZDS transcript levels ranging from approximately 0.5-fold to 4-fold higher when compared to wild-type ( Figure 3a). In chromoplast rich tissues (mature petals and ripe fruits), AtZDS transcript levels were elevated to greater than 10 on a log 2 scale when compared to wild-type ( Figure 3a). Despite successful over-expression in multiple tissues, visual phenotypes were limited to the ripe fruit at 7DPB (Figure 3b).
Carotenoid profiles of young leaves, flowers and ripe fruits from AtZDS.OE lines were investigated via HPLC analysis and compared to carotenoid content in wild-type control plants. Carotenoid levels remained unchanged in young leaves and flowers in the AtZDS.OE lines in agreement with the lack of visual differences ( Table 2), suggesting ZDS activity is not limiting. A modest increase in total lycopene content was observed in the 7DPB fruit of the AtZDS.OE lines with no significant change in total f-carotene amount ( Table 2).

Manipulation of ZDS exposes a dynamic f-carotene isomer profile in ripe tomato fruit
While total f-carotene remained unchanged in the AtZDS.OE lines, the f-carotene isomer profile was altered greatly in fruit either repressing the endogenous SlZDS or over-expressing AtZDS ( Figure 4). In ZDS-RNAi fruit, f-carotene isomers 9,15,9 0 tri-cis-fcarotene, 9,9 0 di-cis-f-carotene, 9 cis-f-carotene and all trans-fcarotene were elevated in a manner consistent with that observed in fruit with ZDS repressed by VIGS (Fantini et al., 2013;Figure 4a). In AtZDS.OE fruit, the ZDS substrates, 9,9 0 dicis-f-carotene and 9 cis-f-carotene were depleted and undetectable while all trans-f-carotene was significantly increased (Figure 4b). This change in the f-carotene isomer profile is consistent with the lack of reduction in the total f-carotene content of AtZDS.OE lines (Table 2).

Reduced carotenoid content and ABA in ZDS-repressed fruit impacts tomato development
In contrast to vegetative tissues of ZDS-RNAi plants, developing fruit displayed clear phenotypes evident of ZDS repression and the resulting inhibition of carotenogenesis (Figures 2b and 5). The effects of ZDS repression on carotenoids and ABA in developing fruit were determined via HPLC and LC-MS, respectively, at three stages in early fruit development (10, 15 and 25 DPA). The ABAdeficient tomato mutant, notabilis (not), defective in 9-cisepoxycarotenoid dioxygenase (SlNCED1) function and minimally characterized with regard to fruit development was included as an additional control (Burbidge et al., 1999).
Levels of ZDS mRNA were reduced approximately sixfold in ZDS.2 and ZDS.7 developing fruit ( Figure S6a). As expected, levels of acyclic carotenoids upstream of ZDS were elevated in ZDSrepressed fruit and undetectable in both wild-type and the ABAdeficient not mutant ( Figure S6b). Levels of photoprotective bcarotene and xanthophylls were reduced greatly in the ZDS-RNAi lines compared to wild-type and not ( Figure 5a). Furthermore, when comparing wild-type to not and the ZDS-RNAi lines, it appears that carotenoid biosynthesis is limited in wild-type fruit as the cells transition to their predominant expansion phase (1 cm -15 DPA), while the ABA-deficient mutant not, ZDS.2 and ZDS.7 maintain carotenoid concentration despite increasing cell size ( Figure 5a). As suggested by the carotenoid profile in ZDS-RNAideveloping fruit, ZDS.2 and ZDS.7 both have at least 75% reduced ABA accumulation during early stages of fruit development compared to wild-type and similar to not (Rodrigo et al., 2003;Figure 5b).
Reduced levels of ABA in early stages of tomato fruit development result in increased FtsZ-regulated chloroplast division, thereby elevating plastid number (Galpaz et al., 2008). FtsZ transcript levels were measured in ZDS.2-and ZDS.7-developing fruit via qRT-PCR and compared to wild-type and not fruit. In not fruit, a similar increase in FtsZ transcripts in early stages of fruit development was observed, consistent with other ABA-deficient mutants in tomato (i.e. high-pigment 3, flacca and sitiens; Figure 5c; Galpaz et al., 2008). ZDS repression and resulting ABA deficiency led to the up-regulation of FtsZ expression in ZDS.2 and ZDS.7 with FtsZ transcript levels twofold higher than the other ABA-deficient mutants (Figure 5c).
In concert with increased chloroplast division, chlorophyll biosynthetic genes were induced in not, ZDS.2 and ZDS.7 compared to wild-type. Chlorophyll a oxygenase (CAO) expression was elevated significantly in not and ZDS-RNAi lines relative to wild-type ( Figure  S7a). Expression of  protochlorophyllide oxidoreductase (POR-B1 and POR-B2), which catalyses the reduction of protochlorophyllide to chlorophyllide, showed a >2-fold increase in expression in all ABAdeficient lines with the largest and most prolonged in the ZDS-RNAi lines ( Figure S7b). Despite the elevation of chlorophyll biosynthetic gene expression in the ZDS-RNAi lines, ZDS.2 and ZDS.7 fruit did not accumulate chlorophyll at levels comparable to wild-type during development, consistent with reduced photoprotective carotenes and xanthophylls (Figure 2c and 5d) and similar to the effect of reduced ZDS on leaves early in development ( Figure S2). We further explored the effects of reduced photosynthetic pigments on attributes of fruit quality (i.e. primary metabolism and brix o ). A comprehensive assessment of primary metabolism in mature green (MG) and ripe (7DPB) fruit of ZDS-RNAi lines was performed via GC-MS (Figure 6a). Total soluble solids are negatively influenced in the MG fruit with strong decreases in major brix o components (e.g. fructose, glucose, citric acid; Figure 6a). However, as the fruit ripens the brix o components approach wild-type levels, though remain lower (Figure 6a). In addition, both stages showed elevated amino acid content consistent with abiotic stress (i.e. photooxidative stress; Obata and Fernie, 2012; Figure 6a). Interestingly, when comparing total brix o levels of 7DPB fruit of ZDS-RNAi lines with same stage fruit of fruit-specific carotenoid mutants (i.e. r and tangerine), it is apparent that carotenoids are linked to fruit soluble sugar content during ripening (Figure 6b). It was observed that ripe fruit of r and tangerine, which lack any visible defect in photoprotection during early fruit development, displays a similar and significant reduction in total brix o levels (Figure 6b).
In addition to influencing chloroplasts and chlorophyll levels during fruit development, reduced ABA levels appear to also impact fruit size and the number of days between pollination and the initiation of ripening (Figure 6c-f). Post-breaker fruit were weighed and ZDS-RNAi fruits showed a significant reduction in mass as did the ABA-deficient not mutant (Figure 6c). In AtZDS.OE fruit, a significant increase in fruit size was observed in only one of the three lines compared to wild-type ( Figure 6d and Table S2). Ripening initiation was delayed in ZDS-RNAi lines, ZDS.2 and ZDS.7, and the ABAdeficient mutant, not, by approximately 6 days (Figure 6e), consistent with prior observations that exogenous ABA can promote ripening (Zhang et al., 2006). It is noteworthy that this significant ripening impairment has not previously been attributed to tomato ABA-deficient mutants, presumably because they eventually ripen and the phenotype has not been carefully monitored. Consistently, significant reduction of two days in time to ripening initiation was observed in two of three AtZDS.OE lines analysed (Figure 6f and Table S2). Minimal changes in other fruit developmental phenotypes observed in AtZDS.OE lines may be indicative of ZDS not being ratelimiting in photosynthetic tissues, consistent with no detectable levels of ZDS substrates recorded in wild-type leaves and developing fruit.
ABA and ethylene have been demonstrated to have an antagonistic relationship during plant development including leaf epinasty and adventitious root growth as observed in the ABAdeficient tomato not mutation, and during fruit ripening in the ABA-deficient tomato hp3 mutant (Galpaz et al., 2008;Thompson et al., 2004). During fruit ripening of ZDS-RNAi lines, ripening fruit displayed elevated or prolonged ethylene production (Figure 7a). However, no complementary effect was observed in AtZDS.OE ripening fruit where ethylene levels remain comparable to wild-type (Figure 7b).

Discussion
Current understanding of ripening-associated carotenoid biosynthesis has been largely established through the cloning and functional characterization of genes underlying natural fruitspecific carotenoid mutants in tomato (Fray and Grierson, 1993;Galpaz et al., 2006;Isaacson et al., 2002;Ronen et al., 2000). Both desaturase genes (PDS and ZDS) remain undefined by tomato mutations, likely due to their single copy gene status and anticipated deleterious effects on photosynthesis impeding survival and consistent with Arabidopsis and Zea mays ZDS mutants (Dong et al., 2007;Matthews et al., 2003). We implemented transgenic repression of tomato ZDS and heterologous overexpression of AtZDS to provide insight into how the manipulation of ZDS and carotenogenesis may alter tomato fruit development and quality.

The single copy ZDS is up-regulated at ripening initiation by NAC-NOR, yet represents a second bottleneck in ripening-associated carotenogenesis
While most steps of the carotenoid biosynthetic pathway are associated with small gene families including members expressed specifically in chloroplast-or chromoplast-rich tissues, we confirm that ZDS is a single copy gene in tomato crucial for carotenoid biosynthesis throughout plant development. Ripe fruit accumulate elevated levels of carotenoids as attractants for seed dispersing frugivores and are largely dependent upon transcriptional induction of ZDS (Figure 1a).  Figure 1c). ZDS transcripts were significantly reduced in ripe fruit of Nr and Gr, both dominant mutations in an ethylene receptor and a receptor-associated protein, respectively, conferring ethylene insensitivity suggesting ethylene signalling is critical for ZDS transcriptional regulation via a potential interaction with ethylene response transcription factors or ERF's (Adams-Phillips, Giovannoni,2004 andGapper et al., 2013). Similar and stronger repression of ZDS was observed in 7DPB fruit of rin and nor, respectively (Figure 1c). At first glance, this repression of ZDS may reflect an indirect regulation through RIN's and NAC-NOR's influence on ethylene (Adams-Phillips et al. 2004;Ito et al., 2017;Wang et al., 2019). On the contrary, upon mining available databases of RIN and NAC-NOR ChIPSeq data (ted.bti.cornell.edu and www.epigenome.cuhk.edu.hk/encode.html), both transcription factors bind to the promoter region immediately upstream of the transcriptional start site, suggesting a role in the direct regulation of ZDS (Vrebalov et al., 2002, Wang et al., 2019and reviewed in Giovannoni et al., 2017. Herein, we demonstrate that ZDS represents an additional carotenoid biosynthetic gene heavily regulated at the onset of and throughout tomato fruit ripening by ripening transcription factors and ethylene signalling either separately or in combination. Consistent with ZDS expression in the wild-type cultivar (Figure 1a), repression of ZDS via the CaMV 35S promoter inhibits carotenoid biosynthesis downstream of 9,9 0 di-cis-fcarotene significantly in flowers and fruits (Table 1). Further, while the modest repression of ZDS in young leaves allows small quantities of the upstream acyclic carotenoids to accumulate, no significant changes are observed in downstream photoprotective carotenes and xanthophylls and phytohormone (i.e. ABA and strigolactone) precursors after acclimation to normal light conditions (Table 1). That said, a temporary susceptibility to highlight was observed upon transition from low light to normal light   conditions, which was overcome quickly as the plant acclimated despite constant repression of ZDS expression ( Figure S2 and Table 1), similar to the effects of the weak ZDS allele encoded by the Arabidopsis spc1 mutation (Dong et al., 2007). We surmise that it may be possible that weak repression results in low levels of downstream cyclic carotenoids, which then undergo nonenzymatic oxidative cleavage by singlet oxygen due to increased susceptibility to photoinhibition. The resulting apocarotenoid signal (i.e. b-cyclocitral) may provide a means of acclimation to the increased light stress as previously reported (Ramel et al., 2012).
As the focus of this study was to investigate the impacts of ZDS manipulation on ripening-associated carotenogenesis and tomato fruit development, ZDS-RNAi lines with stronger repression and albino phenotypes (i.e. ZDS.1 and ZDS.8) were excluded from this study ( Figure S2) to facilitate viability through fruit set. Indeed, this phenotypic severity suggests greater ZDS reduction is highly detrimental to plant development and survival. Further assessment of these more severe genotypes may provide insight into the role of the previously described higher plant f-carotenederived apocarotenoid signal (Avendano-Vasquez et al., 2014). Consistent with a single ZDS gene, additional phenotypes were observed in the weaker repression lines with impaired synthesis of the carotenoid-derived hormone ABA in seeds causing precocious seed germination like that observed in the maize ZDS mutant, vp9 (Matthews et al., 2003; Figure S4).
Over-expression of AtZDS in tomato generates a modest increase in downstream all trans-lycopene of the ripening fruit (Table 2). Concurrently, there is a total depletion of the ZDS substrate 9,9 0 di-cis-f-carotene, which otherwise persists at low levels in wild-type tomato fruit (Figure 4b). The steps catalysed by ZDS thus represent a minor limitation to maximal potential activity of the wild-type biosynthetic pathway. ZDS may be an effective target for genetic manipulation when its substrate is elevated, such as in the case of the recently described tomato AtPDS overexpression lines (McQuinn et al., 2017).

Carotenoids and ABA influence fruit development and ripening
Abscisic acid is one of two carotenoid-derived hormones (the other being strigolactones), characterized to date. ABA is synthesized throughout the plant aiding in protection against abiotic stresses (i.e. drought, salt and cold) in addition to participating in the regulation of plant development (e.g. seed germination and fruit ripening; reviewed in Nambara and Marion-Poll, 2005). As fruit of multiple plant species develop and mature, ABA undergoes dynamic shifts in its synthesis and catabolism. The first is often during early development and fruit expansion followed by a sharp decrease before the onset of ripening. In tomato and other species such as persimmon, there is an induction of ABA prior to ripening initiation and at least in tomato, capable of inducing the ripening hormone ethylene (Leng et al., 2014). In other fruit, ABA is increased in concert with later maturation, ripening and/or senescence (Cotton: Davis and Addicott, 1972;Persimmons: Zhao et al., 2012;Peach: Soto et al., 2013;Avocado: Chernys and Zeevaart, 2000). Strawberry fruit are highly dependent on ABA for complete ripening (Jia et al., 2011).
While more recent studies indicate that ABA is important in tomato ripening based on its timing of accumulation and ability to initiate ripening when applied exogenously (Mou et al., 2016;Zhang et al., 2009), its precise role in tomato ripening remains unclear. Tomato ABA-deficient mutants provide an opportunity to explore the significance of ABA on ripening. Galpaz et al. (2008) demonstrated ABA's negative regulation of tomato fruit chloroplast division at the transition from cell division to cell expansion through characterization of the ABA-deficient mutant high-pigment 3 (hp3). Reduced ABA in developing fruit of our ZDS-RNAi lines resulted in increased FtsZ expression mimicking the hp3 effect and supporting increased plastid division and associated elevated plastid number as resulting from loss of ABA early in fruit development (Galpaz et al., 2008; Figure 5b and c). The chlorophyll biosynthetic genes, POR B and CAO, are elevated consistent with increased plastid numbers, yet decreased chlorophylls were observed, in contrast to the ABA-deficient not mutant ( Figure S7 and Figure 5d). Reduced chlorophyll in ZDS repression lines may reflect a reduction in photoprotective carotenes and xanthophylls ( Figure 5a). Interestingly, changes in chlorophyll biosynthetic gene expression differ substantially from that observed in the Arabidopsis spontaneous cell death1 (spc1) mutants, suggesting that developing fruit may possess a different retrograde signalling network in response to dysfunctional chloroplasts or chlorophyll deficiency.
ABA content is strongly linked to mature green fruit size, the stage preceding ripening initiation. Galpaz et al. (2008) studied the tomato ABA-deficient mutants, flacca, sitiens and hp3, and observed mature fruit weight was approximately 75% of control fruit for all three mutants. 30%-40% reductions in fruit weight were also observed for transgenic tomatoes with repressed NCED1 expression . The 30%-50% reduction in fruit weight of the ZDS-RNAi lines is consistent with these observations and their resulting decrease in ABA (Figure 6c). Taken together, these results demonstrate that ABA plays an important role in fruit size determination and levels of the hormone are influenced by changes in upstream carotenoid synthesis enzymes including ZDS. Climacteric fruit such as tomato produce a characteristic ethylene burst at the onset of ripening, which is necessary for maturation to proceed (Gapper et al., 2013;Giovannoni et al., 2017). The synthesis and perception of ABA during fruit development has been proposed as an early ripening regulator upstream of ethylene (Zhang et al., 2009). Inhibition of ABA biosynthesis in developing tomato fruit via a carotenoid synthesis inhibitor (NDGA) delayed initiation of ripening for 3 days (Zhang et al., 2009). The ABA-deficient pinalate mutant in Citrus sinensis displays a similar effect in that de-greening is prolonged (Rodrigo et al., 2003). The 7-day delay of ripening initiation in ZDS-RNAi tomato fruit further supports a strong relationship between ABA and fruit ripening (Figure 6e). While the ABA-deficient mutant notabilis was previously shown to inhibit fruit expansion, we show here that it is also ripening delayed (Figure 6e), while fruit from two out of the three AtZDS over-expression lines tested ripen on average 2 days earlier than wild-type fruit (Figure 6f and Table S2). Further, altered plastid levels are known to associate with fruit sugar/TSS content (Nguyen et al., 2014) but the effect of altered carotenoid flux on ABA and downstream signalling molecules may indicate a more direct role of carotenoid metabolism on ripening control. Indeed, other carotenoid impaired mutants such as lutescent2 (l2) that fail to accumulate carotenoids at the onset of ripening due to impaired plastid accumulation are significantly ripening delayed (Barry et al., 2012) further supporting a role of carotenoids in coordination of ripening process and beyond their effects on ripe fruit pigmentation and associated appearance and nutrient quality.
In short, we confirm that ABA influences fruit development and ripening at many junctures. Initially as a mediator of fleshy fruit expansion and later as a component of the complex set of genetic and hormonal interactions that contribute to ripening induction. As the fruit matures, ABA influences ethylene synthesis via its well-documented antagonistic relationship (Cheng et al., 2009). In ZDS-RNAi repression lines, ethylene synthesis is maintained at a higher level for a longer period consistent with observations from the ABA-deficient tomato hp3 mutant (Galpaz et al., 2008;Figure 7a). Together, these results further suggest that in contrast to observations in leaves of Arabidopsis ZDS mutants (chloroplast biogenesis 5 (clb5) and spc1), the overaccumulation of upstream carotenoids (i.e. phytofluene and f-carotene) does not produce any identifiable alterations in tomato fruit development that differ from the expected phenotypes attributed to ABA and carotenoid deficiency.

Plant materials and growth conditions
Wild-type (Solanum lycopersicum cv Ailsa Craig, LA2838A) and homozygous notabilis (LA3614); Never ripe (Barry et al., 2005); Green ripe (LA2435); ripening-inhibitor (LA3754); and nonripening (LA3770) mutant seed were obtained from the Tomato Genetics Resource Center, UC Davis (http://tgrc.ucdavis.edu/). Plants were grown in greenhouses at the Guterman Bioclimate Laboratory and Greenhouse Complex, Cornell University, Ithaca, NY. All plants were grown under low light conditions consisting of 12-h day/night and under a mesh shade cloth for the first month after which plants were transplanted, removed from shade and grown under long day conditions (16-hr days and 8-h nights). ZDS-RNAi transgenic lines were carried on to the T1 generation where heterozygotes were selected for further analysis. Analysis of the AtZDS over-expression lines was carried out to the T2 generation.
Young leaf tissue was collected off the 4th, 5th and 6th leaves from the meristem of two-month-old plants. Whole petals and anthers were harvested from flowers at anthesis/pollination (the day a flower fully opened). Developing fruit was staged relative to pollination by tagging flowers at anthesis and collecting fruit at specific days post-anthesis (i.e. 10, 15, 25 and 30DPA/MG). Ripening stages were determined as days post-breaker where breaker is the initiation of overt ripeninginitial colour change at the blossom end of the fruit. Zygosity and copy number of transgenic lines were determined in the T1 generation and confirmed in the T2 generation.

RNAi and over-expression constructs for plant transformation
The ZDS-RNAi construct was made in the pHELLSGATE 2 vector (provided by Peter Waterhouse, University of Sydney, Australia). The ZDS cDNA sequence used in construct development spans 285bp of the 3 0 UTR starting from the most 5 0 bp of the stop codon. PCR amplification of the RNAi fragments was accomplished using the FastStart High-fidelity PCR system (Cat No. 04-738-292-001, Roche Applied Sciences, IN) using the EST clone TUS-69-K8 with primers ZDSRNAi-for and ZDSRNAi-rev containing recombination sequence specified in the Gateway BP Clonase kit (Cat No. 11789-020, Invitrogen, CA; Table S3). The ZDS-RNAi fragment was Gel purified (Cat No. 28706, Qiagen, MD) and cloned into pHELLSGATE 2 according to BP Clonase kit instructions. The resulting construct was sequence verified and transformed into S. lycopersicum cv Ailsa Craig by Agrobacterium tumafaciens (strain LBA-4404) as described in Van Eck et al. (2006).
The AtZDS over-expression (AtZDS.OE) construct was generated as described previously (Gleave, 1992;McQuinn et al., 2017). RNA from Arabidopsis thaliana (accession Columbia-0) leaf tissue was converted to cDNA via iScript TM cDNA synthesis kit CA). Resulting cDNA was used to amplify the full length AtZDS ORF via FastStart High-fidelity PCR system Roche Applied Sciences,IN) with AtZDS-OE-KpnI.for and AtZDS-OE-XbaI.rev primers (Table S3). The resulting AtZDS.OE construct was sequence verified and transformed into S. lycopersicum cv. Ailsa Craig by Agrobacterium tumafaciens (strain LBA-4404) as described in Van Eck et al. (2006).

DNA isolation and zygosity/copy number analysis
Genomic DNA was isolated from fresh meristematic leaf tissue as previously described Barry et al. (2005). Verification of insertion events in the ZDS-RNAi T0 plants was carried out via PCR using primers specific for the 35S promoter in the pHELLSGATE 2 vector (Table S3). Insertion of AtZDS was validated using internal primers within the 35S promoter and the AtZDS ORF (Table S3). Zygosity and copy number was determined in the T1 generation via

RNA isolation and quantitative RT-PCR analysis
Total RNA was isolated using a modified protocol from the RNeasy Minikit (Cat. No 74106, Qiagen Sciences, MD) as described in McQuinn et al. (2017). Quantitative real-time PCR was performed using the Power SYBR â Green RNA-to-C T TM 1-Step Kit (Cat No. 4309169, Applied Biosystems, NJ) in a 5 µL reaction volume (2.5 µL 2X Master Mix; 1 µM forward and reverse primers; 1 µL of total RNA; 0.46 µL DEPC-treated water; 0.04 µL RT enzyme mix). All tissue samples were represented by a minimum of 3 biological replicates, each with triplicate technical replicas. Gene-specific primers were checked for efficiency using wild-type or reference RNA (for primer sequence, see Table S3). To be able to apply the standard curve method described in User Bulletin #2 (Applied Biosystems, 1997), a standard curve was included on each plate for the specific gene being analysed using wild-type or reference RNA (serial dilutions: 50 ng; 5 ng; 0.5 ng; 0.05 ng; 0.005 ng) in triplicate. For each gene analysis, templatefree and negative-RT controls were included. Real-time PCRs were carried out using an ABI PRISM TM 7900HT Sequence Detection System (Applied Biosystems, CA) under the following reaction conditions: reverse transcription at 48°C for 30 min; enzyme activation at 95°C for 10 min; followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. A dissociation curve was added at the end of the run for verification of primer specificity.
ABI PRISM TM SDS version 2.3 software (Applied Biosystems, California) was used to determine gene-specific threshold cycles (C T ) using the endogenous reference (18S rRNA) for every sample. C T 's were extracted, and the standard curve method (Applied Biosystems, 1997) was applied to calculate relative mRNA levels in comparison to the wild-type control or the reference sample (equal volume and concentration of RNA from each tissue combined).

Carotenoid and chlorophyll extraction and analysis
Carotenoids and chlorophylls were extracted from 200 mg of leaf tissue and frozen tomato pericarp, while carotenoid and xanthophyll contents of anthesis flowers were extracted from replicates of petals and anthers of a single flower. Tissue prep, pigment extraction and analysis were carried out using a modified protocols from Alba et al. (2005) as described previously (McQuinn et al., 2017). All solvents used were HPLC grade.

Brix o (TSS) content quantification
Brix o content was measured from pericarp juice and pulp separately in freshly harvested ripe fruit with a digital refractometer (Artisan TM HR200 Digital Refractometer; APT Instruments; IL).

Ethylene measurements
Ethylene was measured from fruits by sealing the fruit in airtight jars for 3 h at 22°C, after which a 1ml aliquot of headspace gas for each fruit was injected on to an Agilent 6850 Network GC System equipped with a flame ionizing detector (Agilent technologies; CA). Samples were compared to standards of known concentration and normalized for fruit mass.

Primary Metabolite analysis using GC-TOF-MS
Metabolite analysis of primary metabolites by GC-MS was performed as described by Lisec et al. (2006) and optimized for tomato fruit pericarp as described in Rohrmann et al. (2011). Chromatograms and mass spectra were evaluated with Chroma TOF 1.0 (Leco) and TagFinder 4.0 software (Luedemann et al., 2008). Further supporting information regarding the metabolite analysis is available in Tables S4 and S5.

Statistical analysis
Statistical analysis and comparison of mean values (i.e. carotenoid, xanthophyll and chlorophyll amounts, ABA, brix o , fruit size, days to ripening and ethylene) from each mutant genotype to the control genotype were performed using the Student's t-test.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. The carotenoid biosynthetic pathway noting S. lycopersicum mutations. Figure S2. Additional phenotypes in ZDS-RNAi lines. Figure S3. ZDS repression construct homozygosity results in severe deleterious phenotypes. Figure S4. Evidence of reduced ABA accumulation in ZDS-RNAi seeds. Figure S5.  Figure S7. Low ABA induced chlorophyll biosynthesis during early stages of fruit development in ZDS-RNAi lines. Table S1. Carotenoid gene families with chloroplast and chromoplast specificity.