ZmCCD10a Encodes a Distinct Type of Carotenoid Cleavage Dioxygenase and Enhances Plant Tolerance to Low Phosphate1

Yanting Zhong,a,2 Xiaoying Pan,a,b,2 Ruifeng Wang,a Jiuliang Xu,a Jingyu Guo,a Tingxue Yang,a Jianyu Zhao,c Faisal Nadeem,a Xiaoting Liu,a Hongyan Shan,d Yanjun Xu,e and Xuexian Lia,f,3 Department of Plant Nutrition, College of Resources and Environmental Sciences, China Agricultural University, MOE, Beijing 100193, China Guangdong Provincial Key Laboratory of Crop Genetic and Improvement, Crop Research Institute, Guangdong Academy of Agricultural Sciences, Guangzhou 510640, China Department of Vegetable Sciences, China Agricultural University, Beijing 100193, China State Key Laboratory of Systematic and Evolutionary Botany, Institute of Botany, The Chinese Academy of Sciences, Beijing 100093, China Department of Applied Chemistry, China Agricultural University, Beijing 100193, China Center for Crop Functional Genomics and Molecular Breeding, China Agricultural University, Beijing 100193, China

A particularly interesting feature of a number of CCD genes is their active involvement in phosphate (Pi) stress responses. Upon Pi limitation, enhanced MAX4 expression boosts strigolactone production, reducing tiller proliferation and stimulating AMF-mediated Pi acquisition (Umehara et al., 2008;Czarnecki et al., 2013). ZmCCD7 transcription is greatly up-regulated, especially in crown roots of Pi-limited seedlings (Pan et al., 2016). PHOSPHATE STARVATION RESPONSE REGULATOR 1 (PHR1), a MYB family transcription factor, is a crucial regulator of plant responses to Pi limitation via binding to the P1BS cis-element (GNATATNC) of target genes (Chiou and Lin, 2011). Heterologous expression of ZmPHR1 in Arabidopsis increases the shoot biomass and inorganic Pi content (Wang et al., 2013). Auxin response factors ARF7 and ARF19 function upstream of PHR1 to modulate its transcription in Arabidopsis roots (Huang et al., 2018). PHOSPHATE1 (PHO1) mediates Pi loading into the xylem and root-to-shoot translocation (Poirier et al., 1991). Regulated by PHR1 and miRNA399 at the transcription level, PHO2 degrades PHO1 and Pi transporters to down-scale Pi absorption and translocation; pho2 mutant leaves show Pi toxicity due to Pi overaccumulation (Delhaize and Randall, 1995;Bari et al., 2006;Liu et al., 2012). The family of PHOSPHATE TRANSPORTER 1 (PHT1) plays crucial roles in Pi uptake from the soil and translocation within the plant (Nussaume et al., 2011). Knockout of AtPHT1;1 significantly reduces Pi uptake, and AtPHT1;5 regulates Pi translocation from source to sink tissues to maintain Pi homeostasis (Shin et al., 2004;Nagarajan et al., 2011).
Maize is an evolutionarily, genetically, and economically important crop promoting global food and feed security (Kellogg, 1997;Coe, 2001;Kumar and Jhariya, 2013). Ancient polyploidization and mobilization of highly abundant transposable elements has led to a 2.5-gigabase maize genome (Lai et al., 2004;Schnable et al., 2009). Maize grains contain highly concentrated phosphorus (P) to nourish embryo and seedling development (Nadeem et al., 2011); however, Pi limitation frequently occurs due to the low bioavailabilty of soil Pi (Shen et al., 2011). It is particularly necessary to explore evolutionary novelties of CCDs in maize and their potential relationship with P nutrition. Here, we constructed a unified CCD phylogenetic tree and found that CCD10 represents a previously unanchored basal clade adding to CCD1, CCD7, CCD8 clades in the entire gene family. ZmCCD10a encoded a new CCD in terms of cleavage sites. Down-regulation of ZmCCD10a expression made plants susceptible to Pi deficiency, whereas its heterologous expression in Arabidopsis enhanced plant tolerance to Pi limitation.

ZmCCD10 was a Distinct Member of the CCD Family
To analyze CCD genes in maize, 14 CCDs were retrieved from the maize genome and classified into four basal clades according to systemic phylogenetic analysis: CCD1, CCD7, CCD8, and CCD10 (Fig. 1). ZmCCD1, two ZmCCD4s, and seven ZmNCEDs were members of the CCD1 clade, which existed in bacteria, fungi, and plants, and underwent two rounds of duplication along with plant adaptation to land, giving rise to NCED and CCD4 subclades. ZmCCD7 belonged to a second basal clade persisting in bacteria, archaea, and plants. ZmCCD8 appeared as a single-copy gene in the CCD8 clade spreading across five fundamental kingdoms. CCD10 was named as CCD-like or more recently ZAS in different contexts (Wei et al., 2016;Wang et al., 2017;Chen et al., 2018;Wang et al., 2019); here, we renamed it as CCD10 according to three criteria: (1) reflecting its fundamental functions in carotenoid cleavage; (2) following CCD1, CCD4, CCD7, CCD8, NCED2, NCED3, NCED5, NCED6, and NCED9; and (3) accommodating wide functional divergence across species beyond mediation of zaxinone synthesis in rice. Thus, we constructed a unified CCD phylogenetic tree with four basal clades (Fig. 1). ZmCCD10a and ZmCCD10b represented a previously unanchored basal clade containing only plant species, most likely due to their loss in bacteria, archaea, and fungi. Notably, basal clade genes CCD1, CCD7, CCD8, and CCD10 contained 1 to 14, 7 and 8, 4 to 10, and 9 to 15 exons in land plants, respectively, indicating large variations in gene structure across four CCD clades (Supplemental Fig. S1). Together, our analysis suggested that ZmCCD10 encodes a distinct CCD in a basal clade.
To trace CCD10 footprints over plant evolution in angiosperms, we constructed a phylogenetic tree with 24 representative species from 10 orders, revealing that CCD10 was highly dynamic and underwent recurrent gain and loss events during genome expanding and restructuring (Fig. 2). Its copy number dramatically varied from 0 (i.e. in the Brassicaceae family) to 6 in the examined species ( Fig. 2; Supplemental Table S1). In dicots, CCD10 duplicated independently before Fabales and Malvales occurrence, and CCD10 doubled before monocot emergence. Different from maintenance of CCD10 duplicates in Poales, only one copy was retained in Elaeis guineensis in subclade a or in Musa acuminate in subclade b. The a subclade further duplicated before Poales split and gave rise to the c subclade. Particularly in the Poaceae family (Figs. 2 and 3), CCD10a remained a single copy in maize and rice and gained an extra copy in Brachypodium distachyon and sorghum (Sorghum bicolor). CCD10b persisted in maize and sorghum although it was eliminated in rice and B. distachyon. CCD10c was lost in maize and sorghum while maintaining two paralogs in rice and four in B. distachyon, indicative of a unique evolutionary path.  . Phylogenetic analysis of the CCD10 gene family in angiosperms. The pink, green, and blue rectangles represent the CCD10a, CCD10b, and CCD10c subclades, respectively. Red solid circles indicate typical duplication events. ZmCCD10a is in red and bold. CCD10s of amborellales serve as the outgroup. Numbers on branches indicate bootstrap values above 50.
In maize, CCD10a was previously reported as ZmCCD8b, a CCD8 closely related name requiring further functional justification (Vallabhaneni et al., 2010). To verify our evolutionary classification, ZmCCD10a and ZmCCD8 were constitutively expressed in max4 (Auldridge et al., 2006a). Heterologous expression of ZmCCD8 functionally complemented the mutant and fully suppressed its over-branching, whereas ZmCCD10a transgenic plants still had as many (6-8, n 5 20) bolts as the max4 mutant (Fig. 4A). Dramatically different phenotypes of these two transgenic lines provided functional evidence to favorably position ZmCCD10a into a separate clade rather than the ZmCCD8 clade in this context.
ZmCCD10a was also quite different from ZAS, its ortholog in rice (Wang et al., 2019); the latter lost 50 amino acids in a conserved region on the C-terminal (Fig. 4B). Within this conserved region, Gly-541 in ZmCCD10a was highly conserved across all angiosperms, corresponding to a Glu residue in lycophytes, liverworts, and charophytes ( Fig. 4B; Supplemental Fig.  S2). The Glu/Gly residue is involved in binding of the Fe 21 cofactor to condition enzymatic functions of CCDs (Takahashi et al., 2005;Kloer and Schulz, 2006). . CCD10 collinearity across four representative cereal species. Collinearity in the CCD10a (A), CCD10b (C), and CCD10c (E) regions across maize, sorghum, rice, and B. distachyon. Dot plots were drawn based on CoGe (https://genomevolution.org/ coge/SynMap.pl) BlastP analysis and putative homologous pairs were plotted as purple dots corresponding to their genomic positions. Red lines cross at the CCD10 locus. Local collinearity of CCD10a (B), CCD10b (D), and CCD10c (F) across maize, sorghum, rice, and B. distachyon are shown. Solid red arrows represent CCD10 genes. Open arrows in the same color represent homologous genes across species, except for solid gray arrows and triangles to indicate nonhomologous genes or elements.

Expression of ZmCCD10a was Upregulated by Low Phosphate
To investigate the fundamental function of ZmCCD10a, we characterized its expression pattern by reverse transcription quantitative PCR (RT-qPCR). At the silking stage, transcript levels of CCD10a were highest in the silk tissue, less abundant in the ear and root, very low in the husk, shank, and leaf blade, and hardly detectable in the tassel, node, and internode (Fig. 5A). Interestingly, CCD10a expression was dramatically stimulated by ;73-fold in the root under low phosphate (LP) and ;6-fold under low nitrogen, but remained at a similar level compared with the control under low potassium (Fig. 5B). Time course analysis revealed that CCD10a expression in the root was nearly 10-fold on day 5, increased to 82-fold on day 10, and decreased to ;20to ;30-fold from day 15 to 25 under LP compared with the control (Fig. 5C). We also quantified CCD10a expression in the root in another maize inbred line Kansas Yellow Saline and four teosinte lines (Z. mays ssp. parviglumis, Z. mays ssp. huehuetenangensis, Zea luxurians, and Zea diploperennis) on day 10 under LP and found ;55-, 11-, 23-, 16-, and 34fold upregulation, respectively (Supplemental Fig. S3, A-E), indicating conserved upregulation of CCD10a expression in the root by LP despite intensive artificial selection and dramatic habitat changes over domestication.
We next performed in situ hybridization to examine tissue expression patterns of ZmCCD10a within the root. As shown in Figure 5D, no signal was detected in the root tip of the negative control. ZmCCD10a signals were mostly present in the pericycle under the Pi sufficient condition. Under LP, strong ZmCCD10a signals were observed in cell division and elongation zones in addition to pericycle signals. When transiently Figure 5. ZmCCD10a expression and localization. A, Relative expression of ZmCCD10a in different maize organs/tissues. The gene expression level in the root was normalized to one for comparison with that in the other organs/tissues. B, Relative abundance of ZmCCD10a transcripts under LN (low nitrogen), LP, LK (low potassium), 100 mM NaCl, and PEG (5% [w/v] PEG6000) 10 d after the treatment compared with the control (Ctrl). C, Stimulation of ZmCCD10a expression during the 25-d LP treatment. Error bars represent the SE of four biological replicates. Asterisks indicate significant difference as determined by oneway ANOVA (**P , 0.01). D, ZmCCD10a expression in the primary root tip in maize by in situ hybridization. Negative control (NC) was hybridized with the sense probe. ZmCCD10a antisense probes were applied to the longitudinal and cross section of the maize root. P, Pericycle; M, meristem. Scale bars 5 500 mm. E, Subcellular localization of GFP-tagged ZmCCD10a. GFP protein (plasmid University of California [pUC]-GFP) and GFP-tagged ZmCCD10a (ZmCCD10a-GFP) were transiently expressed in maize mesophyll protoplasts. GFP signals in green, auto-fluorescent signals of chloroplasts in red, and merged signals in yellow. Scale bars 5 5 mm. expressed in maize leaf protoplasts, GFP-only signals diffused within the entire cell, whereas GFP-tagged CCD10a mirrored auto-fluorescent signals of chloroplasts (Fig. 5E), indicating that, like CCD7 and CCD8 (Auldridge et al., 2006a), CCD10a localized to plastids.
We performed headspace solid-phase microextraction coupled with gas chromatography (GC)-mass spectrometry analysis to qualitatively investigate products of ZmCCD10a-mediated carotenoid metabolism in E. coli and found geranylacetone in the E. coli culture expressing phytoene, 6-methyl-5-hepten-2-one (MHO) in the culture expressing lycopene, MHO and a-ionone in the culture expressing d-carotene, a-ionone in the culture expressing «-carotene, and b-ionone in the culture expressing b-carotene ( Fig. 6H; Supplemental Figs. S4, A-F, and S5, A-F). No volatile was identified in the culture expressing zeaxanthin, probably due to low concentrations of resulting products. Thus, ZmCCD10a was able to cleave the above carotenoids at 5, 6 (59, 6') and 9, 10 (9', 10') positions.

Heterologous Expression of CCD10a in Arabidopsis Conferred LP Tolerance
ZmCCD10a was constitutively expressed in Arabidopsis driven by the Ubiqutin promoter (Desfeux et al., 2000), and three independent transgenic lines were selected for further analysis (Fig. 7, A and B). With sufficient Pi, the total carotenoid levels in siliques of the transgenic plants significantly decreased compared with those of wild-type plants, including neoxanthin, violaxanthin, antheraxanthin, lutein, zeaxanthin, and b-carotene (Table 1); no such significant change occurred in leaves and roots (Table 1).
CCD10a was dramatically upregulated under Pi limitation in maize (Fig. 5B); here, we took advantage of CCD10a transgenic lines to analyze biological consequences of CCD10a upregulation in planta. Both CCD10a transgenic plants and wild-type plants showed robust growth under nutrient-sufficient conditions (Fig. 7A); however, CCD10a transgenic plants had significantly greater shoot and root dry weights and similar root/ shoot ratios compared with the wild-type plants (Fig. 7, C-E). Transgenic plants remained normal even 10 d after Pi limitation (Fig. 7A); by sharp contrast, a typical Pi-deficient phenotype of reddish-purple discoloration occurred in leaves of wild-type plants (Jiang et al., 2007). The anthocyanin concentration in leaves of wild-type plants (16.9 mg g 21 fresh weight) was approximately two times that of the transgenic plants under LP (Fig. 7F). Importantly, when compared with substantial decreases in biomass accumulation in the shoot of wild-type plants under Pi limitation, transgenic plants had relatively lower reductions (Fig. 7D). As a result, the root/shoot ratio of transgenic plants was significantly lower than that of wild-type plants under Pi limitation (Fig. 7E). Therefore, ZmCCD10a transgenic plants were less stressed by LP.
We speculated that ZmCCD10a expression affected the uptake, accumulation, or allocation of Pi in transgenic plants. Under Pi-sufficient conditions, no significant difference in the P concentration was observed between wild-type and transgenic plants (Fig. 7, G and H). However, under Pi limitation, the P concentration in the root of transgenic plants was 24.48% lower than that of wild-type plants (Fig. 7G), and the P concentration in the shoot of the transgenic plants was 30.82% higher than that of wild-type plants (Fig. 7H). We further analyzed related gene expression in the root. CCD10a heterologous expression significantly stimulated expression of AtPHT1;1 and AtPHT1;5 under Pi limitation (Fig. 7, I and J). Pi limitation also caused up-regulation of AtPHR1 and AtPHO1 and down-regulation of AtPHO2 in the transgenic plants ( Fig. 7, K-M). Thus, ZmCCD10a transgenic plants were more LP resistant by allocating more P into the shoot. Signaling-and stress-related genes were also involved in such regulation as indicated by their significant enrichment and overall up-regulation in the transcriptome profile (Supplemental Fig. S6, A-C).
Different from downregulation of PHO2 expression in Arabidopsis under LP (Lin et al., 2008), we found its upregulation at a late time point ( Fig. 7M; Supplemental Fig. S7A), implying a potentially dynamic functional mode of PHO2 in the Arabidopsis response to Pi limitation. As positive regulators of PHR1 (Huang et al., 2018), transcript levels of ARF19 in the root of CCD10a overexpression lines under LP were significantly higher than those in the wild-type root, whereas ARF7 expression in the root did not respond to LP (Supplemental Fig. S7, B and C).
Pi limitation also affects carotenoid accumulation. The total carotenoid concentration in leaves of transgenic plants was significantly higher than that of wildtype plants under Pi limitation, with a particular increase in lutein accumulation and others remaining unchanged (Table 1). However, when compared with  (Table 1). We detected no significant difference in carotenoid accumulation in siliques between wild-type and transgenic plants under Pi limitation (Table 1).
ZmCCD10a silencing made maize seedlings more sensitive to LP To investigate in vivo functions of ZmCCD10a, ZmCCD10a knockdown lines were generated by virusmediated gene silencing (Benavente et al., 2012;Wang et al., 2016). Most positive plants showed 15% to 85% reductions in CCD10a expression compared with the vector control under the Pi-sufficient condition, whereas ZmCCD10b expression remained the same (Fig. 8, A and B). Seedlings with 68% to 78% reductions in CCD10a expression were then used for phenotypic characterization. When compared with the control seedlings, ccd10a seedlings had similar biomass accumulation in the shoot under Pi-sufficient conditions; however, the shoot dry weight of ccd10a knockdown lines significantly decreased by 35.25% under LP (Fig. 8D). Notably, the P concentration decreased by 23.79% in the shoot by contrast with an increase of 43.54% in the root of ccd10a lines under Pi limitation (Fig. 8, F and G), suggesting that more P retention in the root negatively affected shoot growth. Interestingly, accumulation of neoxanthin, violaxanthin, lutein, zeaxanthin, and total carotenoids significantly decreased in leaves of ccd10a seedlings with sufficient Pi compared with that in control leaves. In addition to reduction in accumulation of the above carotenoids, b-carotene accumulation also significantly decreased in the leaf of ccd10a transgenic lines under LP (Table 1). However, the ccd10a root accumulated significantly more antheraxanthin, lutein, zeaxanthin, b-carotene, and total carotenoids than the control root under Pi limitation (Table 1). We also tried to qualify aromatic volatiles derived from CCD10a functioning in the maize root with the authentic standard and bacterial culture as positive controls. When compared with the obvious peak of the standard, we observed the fairly low peak from the bacteria culture and were unable to detect meaningful peaks from root samples, likely due to extremely low levels of aromatic metabolites in the root (Supplemental Fig. S8).
One incompatible result derived from CCD10a overexpression and knockdown was that both transgenic lines promoted biomass accumulation in the root (Figs. 7C and 8C). Monocots and dicots have quite different developmental patterns (Atkinson et al., 2014). A given gene may have primarily conserved functions across monocots and dicots (Kim et al., 2006); however, it may confer divergent phenotypes on some other perspectives (Guan et al., 2012;Huang et al., 2017). The root dry weight is programmed by a wide array of endogenous developmental and exogenous environmental cues (Zhang and Forde, 1998;Remans et al., 2006;Di Mambro et al., 2017). Inconsistent effects of CCD10a expression on the root dry weight in Arabidopsis and maize may be attributed to accumulative evolutionary divergence of these two species or other yet unknown mechanisms to be further investigated in the future.

CCD10 Represents a Previously Unanchored Basal Clade of CCDs Within the Unified Four-Clade CCD Family Tree
CCDs play a wide array of fundamental roles in plant development, secondary metabolism, stress biology,  and interspecies interaction (Schwartz et al., 1997;Akiyama et al., 2005;Auldridge et al., 2006b;Guan et al., 2012;Liu et al., 2015). How CCDs originated and evolved over species diversification and sophistication has been under intensive investigation. Four CCD and five NCED clades are most frequently placed on the phylogenetic trees as parallel clades (Cui et al., 2012;Helgeland et al., 2014;Wei et al., 2016;Wang et al., 2017), and inclusion of CCD-like and ZAS clades (Wei et al., 2016;Wang et al., 2019) makes a more complicated and yet unresolved CCD tree. However, the primary evolutionary path of CCDs remains unknown. More and more functional characterization of CCDs also calls for a better understanding of CCD evolution history. We found that all CCDs and NCEDs formed a unified phylogenetic tree with CCD1, CCD7, CCD8, and CCD10 as four basal clades (Fig. 1). First, ZmCCD10 represented an unexpectedly basal clade within the large CCD family spreading across bacterial, archaeal, fungal, plant, and animal kingdoms (Fig. 1). The CCD10 clade accommodated CCD-like and ZAS genes (Wei et al., 2016;Wang et al., 2019). Second, CCD4 and NCEDs were not independent basal clades, but subclades of the large CCD1 clade (Fig. 1). CCD2 belonged to the CCD1 clade (Frusciante et al., 2014). Thus, our work resolved the evolutionary puzzle of the large CCD gene family, providing a fundamental frame for in-depth analysis of evolution patterns of each subclade and informative guidance for further functional investigation of CCDs.
As a plant-specific clade, CCD10s (9 to 15 exons) mostly have more exons than CCD7s (7 and 8 exons) and CCD8s (4 to 10 exons) and display less variation in exon number than CCD1s (1 to 14 exons) in land plants ( Fig. 1; Supplemental Fig. S1), indicating unique gene structure of CCD10s. Duplication is an essential driving force of gene and species evolution by providing genetic materials for selection (Lynch and Conery, 2000;Bowers et al., 2003); selection and adaptation may cause gene losses on the other hand (Lai et al., 2004;Koskiniemi et al., 2012;Albalat and Cañestro, 2016). Copy number of CCD10 neither followed a general ascending path from algae to land plants, nor was proportional to genome size across species (Fig. 2; Supplemental Table S1). CCD10 expands from one copy in lycophyte Selaginella moellendorffii to three copies in basal angiosperm Amborella trichopoda and shrinks back to one copy in monocots M. acuminate or E. guineensis ( Fig. 2; Supplemental Table S1). Two rounds of CCD10 duplication predates Poales origination (Fig. 2), continues with asymmetric evolution in the Poaceae family, or further expands in B. distachyon (6 CCD10s) in contrast to two copies in maize (Fig. 3). Together, CCD10 is a previously unanchored basal clade of the CCD family illustrating highly dynamic evolutionary features.

ZmCCD10a Is a Distinct Broad-Spectrum CCD
Maize, a genetically and evolutionarily important crop (Doebley et al., 2006;Schnable et al., 2009), contains two CCD10 genes, with ZmCCD10a as the orthologous gene across examined cereal crops (Fig. 2). ZmCCD10a shows higher levels of expression in the silk, ear, and root at the reproductive stage, with concentrated expression in pericycle in the root (Fig. 5, A  and D). GFP-tagged ZmCCD10a is localized to plastids-the organelle containing carotenoids (Fig. 5E). When coexpressed in E. coli, ZmCCD10a significantly cleaves differentially structured carotenoid substrates including phytoene, lycopene, d-carotene, «-carotene, and b-carotene at the 5, 6 (59, 6') and 9, 10 (9', 10') positions and generates C 8 (MHO) and C 13 apocarotenoids (geranylacetone, a-ionone, and b-ionone; Importantly, ZmCCD10a knockdown leads to overaccumulation of lutein, antheraxanthin, zeaxanthin, b-carotene, and total carotenoids in the root under LP in comparison with wild-type seedlings; consistently, heterologous expression of ZmCCD10a in Arabidopsis significantly reduces accumulation of lutein and total carotenoids in the LP root (Table 1), suggesting that ZmCCD10a is a functional CCD in planta. ZmCCD10a expression in Arabidopsis also reduces accumulation of neoxanthin, violaxanthin, antheraxathin, lutein, zeaxanthin, b-carotene, and total carotenoids in siliques (Table 1). Similarly, knockout of related CCDs (CCD1, CCD4, or NCED2) significantly promote accumulation of total and other individual carotenoid in siliques in Arabidopsis (Gonzalez-Jorge et al., 2013). In addition, leaves, when compared with nongreen tissues (fruits and roots), have very different metabolic and accumulation patterns of carotenoids (Shewmaker et al., 1999;Ye et al., 2000;Lätari et al., 2015). Under LP, accumulation of lutein and total carotenoids increases in leaves of Arabidopsis with ZmCCD10a expression (Table 1); ZmCCD10a knockdown reduces accumulation of neoxanthin, violaxanthin, lutein, zeaxanthin, b-carotene, and total carotenoids in maize leaves. Such positive association of carotenoid accumulation in leaves with ZmCCD10a expression in two contrasting biological contexts needs to be further investigated (Table 1).
Together, our in vivo and in vitro results suggested that ZmCCD10 functions in a highly divergent manner in   terms of substrate preference and cleavage positions compared with its counterpart ZAS in rice (Wang et al., 2019). Our conclusion was supported by another line of evidence that OsZAS lost a conserved 50-amino acid domain at its C terminus ( Fig. 4B; Supplemental Fig. S2; Takahashi et al., 2005;Kloer and Schulz, 2006); such domain loss may result in substantial functional divergence from ZmCCD10 and other CCDs (Wang et al., 2019).

ZmCCD10a Is a Positive Regulator of Plant Tolerance to LP
Dramatic and persistent upregulation of CCD10a expression in the maize and teosinte root in response to LP strongly suggests association of CCD10 with Pi nutrition (Fig. 5B; Supplemental Fig. S3, A-E). Heterologous expression of ZmCCD10a in Arabidopsis promotes P accumulation in the shoot, resulting in higher shoot dry weight under Pi limitation (Fig. 7, D and H). PHT1;1 and PHT1;5 mediate Pi acquisition and translocation from source to sink organs, respectively (Shin et al., 2004;Nagarajan et al., 2011). Enhanced expression of PHR1, PHO1, PHT1;1, and AtPHT1;5 and reduced expression of PHO2 in roots of ZmCCD10aexpressing plants favors Pi uptake and translocation toward the LP shoot (Fig. 7, H-M). In opposition to preferential P retention in the shoot of LP-stressed Arabidopsis transgenic lines, ZmCCD10a silencing causes more P retention in the root and less P allocation to the shoot under LP, which hampers shoot growth and biomass accumulation (Figs. 7H and 8, D-G). ZmPHT1;6 knockout dramatically reduces aboveground biomass in the field (Willmann et al., 2013). Lower PHT1;6 expression is in agreement with decreases in shoot biomass of ZmCCD10a silencing plants (Fig. 8, D and I). Elevation of ZmPHT1;3 and ZmPHT1;13 expression in the ZmCCD10a silencing lines may be a feedback outcome of more severe P shortage in the shoot, which probably facilitates Pi absorption or translocation in the root (Fig. 8, F, H, and J; Liu et al., 2016). Together with up-regulation of PHO2 expression and downregulation of ZmPHR1;1, ZmPHR1;2, and PHO1 expression tend to reduce P allocation to the shoot under LP (Fig. 8, G and K-N). Therefore, CCD10a modulates relative Pi allocation between the root and shoot depending on its transcript abundance; enhanced transcription, probably under regulation of ZmPHR1;1 and ZmPHR1;2 (Fig. 9D), favors Pi flow to the shoot to support aboveground functioning under LP.
As a putative loop linker, ZmCCD10a overexpression and silencing leads to contrasting expression of PHR1, PHO2, PHO1, and related Pi transporters. How does ZmCCD10a transact on the PHR1 pathway? In Arabidopsis, ARF19 positively regulates PHR1 expression specifically in the root (Huang et al., 2018). Here, transcript levels of ARF19 in Arabidopsis roots with heterologous CCD10a expression significantly increase under LP compared with those in wild type, whereas transcript abundance of ZmARF27 (ARF19's ortholog) in the CCD10a knockdown root is significantly reduced upon Pi limitation (Supplemental Fig. S7, C-E). Thus, one possibility is that CCD10a may modulate PHR1 expression in an ARF19-dependent manner. Alternatively, ZmCCD10a triggers the LP signaling cascade via potential protein partners or yet to be validated metabolites, which is supported by surprising enrichment and upregulation of signaling related genes in Arabidopsis roots with heterologous ZmCCD10a expression (Supplemental Fig. S6, A-C). Together, ZmCCD10a functions as a positive regulator of plant tolerance to Pi limitation, and the presence of CCD10a in cereal ancestors probably facilitates their adaptation to Pi limited habitats. Finally, given that only mycorrhizal plants contain CCD10 (Wang et al., 2019) and that C 13 apocarotenoids are strongly associated with colonization and functional maintenance of mycorrhizas (Walter et al., 2000;Floss et al., 2008a;Floss et al., 2008b), elevated expression of CCD10a under LP accelerates carotenoid cleavage to boost C 13 apocarotenoid provision, potentially favoring Pi acquisition via the mycorrhiza fungal pathway.

Plant Materials and Growth Conditions
Hydroponic culture of maize (Zea mays; inbred lines B73 and Kansas Yellow Saline) and teosinte seedlings (Z. mays ssp. parviglumis; Z. mays ssp. huehuetenangensis; Figure 9. Regulation of ZmCCD10a by ZmPHR1s. A, The P1BS element in the ZmCCD10a promoter. B, ZmPHR1;1 and ZmPHR1;2 binding to ZmCCD10a p:LacZ, rather than ZmCCD10a m:LacZ with a mutated P1BS element in the yeast one-hybrid assay. C, Binding of PHR1;1 and PHR1;2 to the biotin-labeled P1BS-containing oligonucleotide derived from the ZmCCD10a promoter in the EMSA assay. D, Transcriptional enhancement of ZmCCD10a by transactivation of PHR1;1 (62K-PHR1;1) and PHR1;2 (62K-PHR1;2) compared to the vector control (V/C), as indicated by the relative LUC/ REN ratio. Error bars represent the SE of six biological replicates. Asterisks indicate significant difference as determined by one-way ANOVA (*P , 0.05).
Zea luxurians; Zea diploperennis) followed our well-established protocol (Pan et al., 2016). The full nutrient solution comprised of 2 mM NH 4 NO 3 , 2 mM CaCl 2 , 0.75 mM K 2 SO 4 , 0.1 mM KCl, 0.65 mM MgSO 4 , 0.25 mM KH 2 PO 4 , 0.2 mM Fe-EDTA, 1 3 10 23 mM ZnSO 4 , 1 3 10 23 mM MnSO 4 , 1 3 10 24 mM CuSO 4 , 5 3 10 26 mM (NH 4 ) 6 Mo 7 O 24 , and 1 3 10 23 mM H 3 BO 3 . For the low nitrogen treatment, the NH 4 NO 3 concentration was reduced from 2 mM to 40 mM; 2.5 mM KH 2 PO 4 and 0.1 mM KCl were adopted for the LP and low potassium treatment, respectively, whereas other nutrient concentrations remained unchanged or balanced. To simulate salt or drought stress, 100 mM NaCl or 5% (w/v) polyethylene glycol (PEG) 6000 was applied to the full nutrient solution. Every treatment had three technical and three biological replicates, and the nutrient solution (pH 5 6.0) was changed every 2 d. Plants were grown in a controlled growth chamber having continuous aeration, 14/10-h light/dark photoperiod, 28°C/22°C day/night temperature, and 60% relative humidity. Seedlings were harvested on 5, 10, 15, 20, and 25 d after treatments. Field experiments with four biological replicates were conducted at the Shangzhuang Experimental Station, Beijing (Liao et al., 2012;Pan et al., 2015;Pan et al., 2016). Maize samples at silking were harvested as the root (five randomly selected crown roots), node (next to the ear), internode (immediately next to the ear), leaf blade of the ear leaf, shank, ear, husk, silk, and tassel. All samples were immediately frozen in liquid nitrogen and kept in the 2 80°C freezer for further analysis.
Arabidopsis (Columbia-0; wild-type and transgenic lines) was grown in hydroponic solution (pH 5 5.8) containing 0.5 mM Ca(NO 3 ) 2 , 2 mM K 2 SO 4 , 2 mM RT-qPCR, Gene Cloning, In Situ Hybridization, and Subcellular Localization RNA isolation, RT, real time qPCR with gene specific primers (Supplemental Table S2), and data analysis were carried out following standard protocols (Livak and Schmittgen, 2001;Pan et al., 2016). ZmUbi (for maize) and AtTub4 (for Arabidopsis) were used as the internal control. Each treatment for all RT-qPCR in this study had three biological replicates, and each biological replicate had three technical replicates.
The full CCD10a coding sequence was amplified with addition of restriction enzyme/cleavage sites (SalI and SpeI), cloned into the pMD19-T vector (Takara), and sequenced as previously described (Pan et al., 2016). The PCR fragments were listed in the Supplemental Table S2.
The root tips of maize seedlings were fixed and hybridized following previous protocols (Zhang et al., 2013). Probes were generated through PCR amplification using gene-specific primers (Supplemental Table S2). Images were captured by an Olympus BX53 microscope.
The CCD10a coding sequence lacking a termination codon was inserted into the pUC-GFP vector for GFP-fused protein expression driven by the 35S promoter. Maize protoplasts were cultured, isolated according to Yoo et al. (2007), and transformed with the 35S::CCD10a-GFP construct using the PEG-mediated method (Zuo et al., 2015). Briefly, 0.2 mL of suspended protoplasts (0.5 3 10 6 cells mL 21 ) with 20% (w/v) PEG4000 (Sigma) were transformed with 20 mg plasmids and incubated at 23°C for 15 min. Transformed protoplasts were cultured at 25°C for 16 h in the dark and observed under a Zeiss confocal microscope.

Analysis of CCD10a Activity
CCD10a was cloned into pDEST15 (Invitrogen) from CCD10a-pDONR221 by recombination. Plasmids containing individual carotenoid biosynthetic genes for phytoene, lycopene, d-carotene, «-carotene, b-carotene, and zeaxanthin were solely transformed (as the negative control) or cotransformed with pDEST15 harboring CCD10a into the Escherichia coli strain BL21-AI (Invitrogen) for L-arabinose induced expression; ZmCCD10a expression was analyzed by the standard SDS-PAGE method. Three individual colonies per carotenoid strain were used to separately inoculate 5 mL Luria-Bertani medium containing 100 mg mL 21 ampicillin and 34 mg mL 21 chloramphenicol (Sigma). The 5-mL cultures were grown overnight at 37°C. Four milliliters of overnight culture was used to inoculate 200 mL of Luria-Bertani medium containing 100 mg mL 21ampicillin and 34 mg mL 21 chloramphenicol (Sigma) in a 500-mL flask.
For HPLC analysis, cultures were grown with vigorous shaking for 9 to 10 h at 30°C to an absorbance of 0.9 to 1 at 600 nm, then divided into two equal parts in two 250-mL flasks. One flask was induced with a final concentration of 0.1% (w/v) L-arabinose (Sigma), and another flask was added equal volume of water as control. The cultures were grown for additional 16 h at 25°C. Cells were pelleted at 5000 g for 5 min. Lycopene, phytoene, d-carotene, and «-carotene were extracted from cell pellets by vortexing in 12 mL acetone. b-Carotene was extracted from cell pellets by vortexing in 4-mL ethyl alcohol followed by extraction in 8-mL ethyl acetate. To collect zeaxanthin, cell pellets were extracted in 12 mL of 95% (v/v) ethanol. All solvents were purchased from Fischer Scientific, and all steps were carried out on ice and shielded from strong light. After filtration, 20 mL extract was injected onto the HPLC column. The Waters HPLC was equipped with a photodiode array detector and Agilent SB-C 18 (150 mm 3 4.6 mm; 5 mm; Waters). For lycopene, d-carotene, «-carotene, b-carotene, and zeaxanthin, mobile phases consisted of acetonitrile/water (90:10 [v/v]; A) and ethyl acetate (B). The linear gradient elution increased to 100% B from 0 in 15 min with a flow rate of 1 mL min 21 , and the detection wavelength was 475 nm. For phytoene, the mobile phase was ethanol (1 mL min 21 ) at room temperature and the detection wavelength was 286 nm. Standard phytoene, lycopene, b-carotene, and zeaxanthin were from Aladdin, and d-carotene and «-carotene were from CaroteNature.
Volatile compounds were determined by headspace solid-phase microextraction coupled with GC-mass spectrometry . Cultures were grown for 5 to 6 h at 30°C to an absorbance of 0.6 to 0.8 at 600 nm, divided, and induced at 16°C for an additional 16 h. Then 5 mL culture and 1 g NaCl were placed into a 15-mL sample vial tightly capped with a polytetrafluoroethylene-silicon septum and heated at 40°C for 30 min on a heating platform with agitation at 400 rpm. The solid-phase microextraction (85 mm polyacrylate, 100 mm polydimethylsiloxane, and 50/30 mm Divinylbenzene/Carboxen/Polydimethyl-siloxane [DVB/CAR/polydimethylsiloxane]; Supelco), preconditioned according to manufacturers' instructions, was then inserted into the headspace, where the extraction lasted for 30 min with heating at 40°C and agitation by a magnetic stirrer. The fiber was subsequently desorbed in the GC injector for 25 min. The Agilent 6890N GC was equipped with an Agilent 5975B mass spectrometer. The column was a HP-INNOWAX column (60 m 3 0.25 mm internal diameter, 0.25 mm; J&W Scientific). Helium was used as carrier gas with a flow rate of 1 mL min 21 . The GC oven was held at 40°C for 1 min, increased up to 220°C at a rate of 3°C min 21 and held at 220°C for 5 min. The mass spectrometer operated in the electron impact mode at 70 eV, and samples were scanned in the range of mass-to-charge ratio 50 to 500. Identification was based on retention indices of reference standards and mass spectra matching in the National Institute of Standards and Technology 08 library.

Constructs and Generation of Transgenic Plants
The CCD10a coding region was cloned into SalI and SpeI sites of the pSuper1300 1 vector (Yang et al., 2010). The construct was transformed into Agrobacterium tumefaciens strain GV3101 for Arabidopsis transformation (Desfeux et al., 2000). Positive transformants were selected on Murashige and Skoog plates with 35 mg mL 21 hygromycin. The resulting positive lines (T1 generation) were transferred to soil (perlite:vermiculite, 1:1 [v/v]) in a growth chamber as described above for seed proliferation and homozygote screen on the selective growth media. Three independent T3 homozygous lines were selected for further analyses according to the expression level of ZmCCD10a.
To generate gene silencing lines, a 238-bp fragment from the CCD10a coding sequence (C 1185 to C 1422 ) was cloned into the pCMV201-2b N81 vector, transformed into the A. tumefaciens strain C58C1, coinfiltrated with C58C1 cultures harboring pCMV101 or pCMV301 (with a ratio of 1:1:1) into the leaves 5 and of 6-week Nicotiana benthamiana seedlings, and incubated for 3 d . CCD10a silencing constructs were introduced into excessive amount of maize seeds (0.15-0.16 g grain 21 ; 100 seeds per strain per treatment) by vascular puncture after virus extraction from infected leaves, with empty pCMV201-2b N81 as the negative control . Maize seeds were then germinated and grown in control and LP nutrient solutions for 10 d. The cotyledon tip (2 cm 3 0.5 cm) was cut off to screen for CCD10a silencing seedlings using the virus specific primers; 2-cm long root tips were sampled and frozen in liquid nitrogen to analyze gene expression. The remaining tissue of individual plants was used for other phenotypic analysis with four biological replicates. The whole experiment was repeated three times. Dry mass was determined after oven drying at 120°C for 30 min and 65°C for 72 h using a BSA124S analytical balance (Sartorius). The samples were then ground into fine powder for P analysis (Soon and Kalra, 1995). Primers are listed in Supplemental Table S2.

RNA Isolation, Library Construction, and Transcriptome Sequencing
Total RNA was extracted from the root of Col-0 and ZmCCD10aoverexpressing plants grown in the full nutrient solution for 4 weeks, followed by library construction, RNA sequencing, and data analysis according to detailed protocols described previously (Pan et al., 2015). A representative transgenic line (no. 15) was selected for sequencing with three biological replicates.

Elemental Assay
The appropriate amount of the ground plant material was digested by a modified Kjeldahl procedure using H 2 SO 4 1 H 2 O 2 as a catalyst. Digests were analyzed for P concentrations by automated colorimetry using the molybdovanadate method (Soon and Kalra, 1995).

Isolation and Quantification of Carotenoids in the Plant Tissue
Fresh plant tissues (10 to 20 mg for the leaf and 400-500 mg for the root) were ground into powder and suspended in the 500 mL extraction buffer (acetone: ethyl acetate, 3:2 [v/v]) containing 0.1% (w/v) butylated hydroxytoluene. After 10 min of shaking, 100 mL HPLC-grade water was added followed by 10 min centrifugation at 2500g. The 20 mL top organic phase was used for the HPLC injection. Then 500 mL of extraction buffer (methanol: chloroform, 2:1 [v/v]) with butylated hydroxytoluene and 100 mL of HPLC-grade water/200 mL of chloroform was used for extraction of silique carotenoids. Carotenoids were identified by comparison of the retention time with that of the authentic standard. Standard curves were prepared using commercial standards, with antheraxanthin, neoxanthin, and violaxanthin from CaroteNature and lutein, zeaxanthin, and b-carotene from Aladdin.

EMSA
Coding sequences of ZmPHR1;1 and ZmPHR1;2 were cloned into the PET-30a vector, transformed into the BL21 strain, cultured for 5 to 6 h at 30°C until an absorbance of 0.6 to 0.8 at 600 nm, and then induced with 1% (w/v) isopropylthio-b-galactoside for 16 h at 16°C. The recombinant protein was purified with Ni-NTA-Agarose. EMSA was carried out following the instructions of the LightShift Chemiluminescent EMSA Kit (Thermo Scientific). The ;40-bp probe containing the P1BS element of ZmCCD10a was synthesized and labeled with biotin, with unlabeled biotin as the negative control. The primers and oligonucleotides were listed in Supplemental Table S2.

Transient Activation Assay in N. benthamiana Leaves
Coding sequences of PHR1;1 and PHR1;2 were cloned into pGreen 62-SK as effectors, and the 2054-bp promoter of CCD10a was fused to pGreen II 0800-LUC as the reporter. Both vectors were transformed into A. tumefaciens GV3101 (pSOUP 19). After construct verification, the culture mixture (9 effector:1 reporter) was infiltrated into N. benthamiana leaves and incubated for 3 d, with the empty effector vector as the negative control. Activities of firefly luciferase (LUC) and Renilla luciferase (REN) in leaves were measured with the Dual Luciferase Assay System (Promega) on a GLO-MAX 20/20 Luminometer (Promega). The ratio of LUC to REN was calculated with six biological replicates, and the value derived from the vector control was normalized to one. Primers are listed in Supplemental Table S2.

Statistical Analysis
Significant differences were analyzed using the one-way ANOVA in SAS (v8; SAS Institute Inc.). Means of different treatments were compared based on the LSD at a 0.05 or 0.01 level of probability.

Accession Numbers
Accession numbers of all genes or proteins analyzed in this study are listed in Supplemental Datasets S1 and S2 and Supplemental Table S2.

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. The structure of CCD family genes in land plants.
Supplemental Figure S2. Partial alignment of CCD10 protein sequences.
Supplemental Figure S3. CCD10a expression in maize and teosinte lines.
Supplemental Figure S4. Coomassie Blue-stained SDS-PAGE gels showing His-tagged ZmCCD10a expression in E. coli.
Supplemental Figure S5. GC-mass spectrometry analysis of products derived from carotenoid cleavage by ZmCCD10a.
Supplemental Figure S7. Gene expression analysis in wild-type Arabidopsis and ZmCCD10a-overexpressing and silencing seedlings.
Supplemental Figure S8. Aromatic metabolite analysis in roots of ZmCCD10a silencing lines by GC-MS.
Supplemental Table S1. Variations in copy number of CCD10 over evolution.
Supplemental Table S2. Primers used for gene cloning, expression, and vector construction.
Supplemental Dataset S1. Gene identifiers and related information for phylogenetic analysis of the CCD gene family.
Supplemental Dataset S2. Gene identifiers and related information for phylogenetic analysis of the CCD10 gene family in angiosperms.