Enzymatic study on AtCCD4 and AtCCD7 and their potential to form acyclic regulatory metabolites

Highlight In vitro study shows that AtCCD4 claves all-trans-bicyclic-carotenoids, excludes its direct involvement in generating plastid retrograde signals supposedly derived from cis-desaturation intermediates, and demonstrates that AtCCD7 converts 9-cis-acylic carotenes.

CCD enzymes might also form hitherto unknown regulatory compounds. Apart from ABA and SLs, additional carotenoid-derived regulatory molecules have been postulated and thought to arise through cleavage of desaturation intermediates of the carotenoid biosynthesis pathway ( Fig. 1) (Moise et al., 2014). For instance, the tomato tangerine mutant is impaired in the activity of carotenoid cis-trans isomerase (CRTISO; Fig. 1), which results in 7,9,7',9'-tetra-cis-lycopene (prolycopene) accumulation in fruits, while the r(2997) yellow-flesh does not express the fruit-specific phytoene synthase gene (PSY1), leading to a lack of fruit carotenoids. It was shown that the accumulation of prolycopene and its polycis-configured precursors in the yellow-flesh background is accompanied by a partial restoration of PSY1 transcription. This epistatic behavior of the downstream over the upstream pathway activity is thought to be mediated by carotene cleavage product(s) derived from prolycopene or its less desaturated precursors (Kachanovsky et al., 2012). Moreover, the Fig. 1. Carotenoid biosynthesis and formation of carotenoid-derived signaling molecules. Desaturation intermediates are shaded. The scheme shows established and hypothetical signals generated by carotenoid cleavage dioxygenases (CCDs). The formation of abscisic acid and SLs is initiated by 9-cis-epoxycarotenoid cleavage dioxygenases (NCEDs) and CCD7, respectively. Hypothetical signals are indicated by a question mark. PSY, phytoene synthase; PDS, phytoene desaturase; Z-ISO, ζ-carotene isomerase; CRTISO, carotenoid cis-trans isomerase; LCY-b, lycopene β-cyclase; D27 (DWARF27), all-trans/9-cis-β-carotene isomerase. See text for further explanations.
use of Arabidopsis mutants and carotene desaturation inhibitors pointed to the presence of apocarotenoid(s) that modulates leaf and chloroplast development and which is thought to derive from phytofluene and/or ζ-carotene. It has been suggested that AtCCD4 is involved in the formation of this regulatory signal(s) (Avendaño-Vázquez et al., 2014). Furthermore, potato ccd4 knock-down lines develop tubers with elevated carotenoid content, which also show changed shape and premature sprouting (Campbell et al., 2010). In saffron, CCD4 transcript levels increase under abiotic stress and during senescence (Rubio et al., 2008;Rubio-Moraga et al., 2014).
In the current work, we investigated the cleavage activity of AtCCD4 in vitro, using analytical and kinetic approaches. In addition, we evaluated the potential of this enzyme in generating the signals proposed to derive from cis-configured carotene desaturation intermediates, in comparison with the homolog, SL biosynthetic enzyme AtCCD7. Our data indicate that AtCCD7 rather than AtCCD4 may be a candidate capable of forming these signal molecule(s).
pThio-AtCCD4 was transformed into BL21 (DE3) Escherichia coli cells harboring the plasmid pGro7 (Takara Bio Inc.). Cells were grown at 37 °C in 50 ml of 2 YT growth medium supplemented with chloramphenicol and ampicillin. Protein expression was induced with 0.2% (w/v) arabinose at an OD 600 of 0.5. Cells were grown for 4 h at 28 °C and harvested. Cell pellets were re-suspended in 1 ml of modified LEW buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 1 mg ml −1 lysozyme, 1 mM dithiothreitol; pH 8.0), passed twice through a French pressure cell at 10 000 psi, and centrifuged at 20 000 g for 5 min. Protein was quantified using the Quick Start™ Bradford Protein Assay (Bio-Rad Laboratories) and adjusted to 20 µg µl −1 .
Assays were performed in a total volume of 200 µl. Purified substrates (30 µM in CHCl 3 ) were mixed with 20 µl of ethanolic Triton X-100 (2%, v/v; Sigma), dried using a vacuum evaporator, and dissolved in 100 µl of 2× incubation buffer [2 mM TCEP, 0.4 mM FeSO 4 , 200 mM HEPES-NaOH pH 7.8, and 2 mg ml −1 catalase (Sigma)]. Assays were started by the addition of each 50 µl lysate and H 2 O, and then incubated for 1 h, if not stated otherwise, under shaking (200 rpm) at 30 °C in darkness. For extraction, 2 vols of acetone were added, followed by a short sonication and the addition of 3 vols of light petroleum/diethylether (2:1, v/v). After centrifugation, the epiphase was dried and redissolved in 40 µl of CHCl 3 . A 5 µl aliquot of the extract was subjected to HPLC analysis with system 1, using tocopherol acetate (0.1 mg ml −1 ) as an internal standard.

Dynamic modeling and data processing
The decay of β-carotene (β), β-cryptoxanthin (cry), and zeaxanthin (zea) to β-ionone (β-io) and 3-OH-β-ionone (OH-β-io) via the intermediates β-apo-10-carotenal (β-10) and 3-OH-β-apo-10-carotenal (OH-β-10 is modeled by a set of ordinary differential equations (ODEs) following mass-action kinetics: In total, the model contains 13 free parameters, denoted by θ, which are determined from data. The parameters of interest are the rate constants k i , i = β, cry-OH, cry-β, zea, β-10,OH-β-10. Other parameters are the initial concentrations of the states at time point zero and one conversion factor needed to account for a general change in the enzyme activity between datasets shown in Fig. 3A-C).
The parameters θ are estimated via the maximum likelihood method assuming normally distributed noise. The maximization of the likelihood translates into minimizing the cost function σ where x i denotes the measured value of data point i with uncertainty σ i , and x(t i , θ) gives the model prediction at time t i . The ODEs together with their sensitivity equations are integrated with the lsodes solver (Soetaert et al., 2010). Non-linear derivative-based optimization of the cost function is performed by a trust region optimizer (Nocedal and Wright, 1999). Parameter identifiability and confidence intervals are computed by the profile likelihood method (Raue et al., 2009). All analyses have been performed with the packages cOde/dMod for dynamic modeling in R. The detailed model specification encompassing alle experimental conditions can be found in Supplementary dataset S1 at JXB online. Data pre-processing Each experiment was repeated at least three times; some were repeated four times. Mean values of the experiments were taken as input for the modeling. Measurement uncertainties were estimated by the error model σ 2 where ϕ n i -1 is the probability density of the χ 2 distribution with n i -1 degrees of freedom and the index corresponds to data point i with mean value x i and variance v i determined from n i replicates. Standard errors of the mean were computed from the estimated variance and the number of replicates σ x n i i ( ) .

HPLC analysis and purification
Carotenoids were analyzed using a Shimadzu UFLC XR separation module equipped with an SPD-M20A photodiode array detector (Shimadzu). The column temperature was held at 40 °C. A YMC-pack-C 30 reversed phase column (150 × 3 mm i.d., 5 µm; YMC Europe) was used throughout. The separation system 1 solvent system consisted of A: MeOH/TBME (1:1, v/v) and B: MeOH/ H 2 O/TBME (30:10:1, v/v/v). The flow rate was 0.6 ml min −1 with a gradient from 100% B to 0% B within 20 min and maintenance of the final conditions for 4 min. The separation system 2 solvent system utilized A: MeOH/TBME (1:1, v/v) and B: MeOH/H 2 O/TBME (30:10:1, v/v/v). The flow rate was 0.6 ml min −1 with a gradient from 100% B to 0% B within 24 min and maintenance of the final conditions for 4 min. Separation system 3 employed the solvent system A: MeOH/TBME (4:1, v/v) and B: MeOH/H 2 O/TBME (30:10:1, v/v/v). The flow rate was 0.6 ml min −1 with a gradient from 100% B to 40% B within 20 min, to 0% in 5 min and maintenance of the final conditions for 10 min. Separation system 4 was isocratic with the solvent MeOH/TBME (3:1, v/v) at a flow rate of 2 ml min −1 .

Mass spectrometry
Volatile cleavage products such as β-ionone, 6-methyl-5-hepten-2-one (MHO), and geranylacetone were collected by solid phase micro extraction (SPME; PDMS, 100 µm; Supelco). The fiber was exposed to the assay gas phase for 15 min and desorbed in the injector of the Trace GC coupled to a Trace DSQ II mass spectrometer (Thermo Fisher Scientific). Separation was achieved on a 30 m Zebron ZB-5 column 0.25 mm i.d., 0.25 µm film thickness (Phenomenex). The initial temperature of 50 °C was held constant for 5 min, followed by a ramp of 25 °C min −1 to a final temperature of 280 °C which was maintained for 5 min. The helium carrier gas flow rate was 1 ml min −1 and the injector temperature was set to 220 °C. Electron impact ionization (EI) was used at an ion source potential of 70 eV and a source temperature of 200 °C. Spectra were matched to the NIST (2.0) database using the Excalibur software. Additionally, standards of β-ionone and MHO (Sigma) were used. Non-volatile reaction products were identified by LC-MS using a Dionex UltiMate 3000 UPLC coupled to a Q-Exactive mass spectrometer (Thermo Fisher Scientific). Sample separation was achieved with a Hypersil Gold C 18 UPLC column (150 × 2.1 mm i.d., 1.9 µm) and the solvent system A, 0.05% (v/v) formic acid in H 2 O, and B, 0.05% (v/v) formic acid in acetonitrile. Initial conditions were 70% B for 1 min, followed by a gradient to 100% B within 4 min. The final conditions were maintained for 10 min, all at a flow rate of 0.5 ml min −1 . Ionization of apocarotenoids was achieved with atmospheric pressure chemical ionization (APCI) and analyzed in the positive mode. Nitrogen was used as sheath and auxiliary gas, set to 20 and 10 arbitrary units, respectively. The vaporizer temperature was set to 350 °C and the capillary temperature was 320 °C. The spray voltage was set to 5 kV and the normalized collision energy (NCE) to 35 arbitrary units. For data analysis, the TraceFinder (3.1) software and authentic apocarotenoid standards were used.

AtCCD4 cleaves cyclic carotenoids and apocarotenoids
Carotenoid-accumulating E. coli strains have frequently been used to test carotenoid-cleaving enzymes in vivo (von Lintig and Vogt, 2000;Booker et al., 2004;Prado-Cabrero et al., 2007;Frusciante et al., 2014). However, only a limited number of carotenoid species can be produced in this system, and the stereospecificity of the reactions cannot be determined. We therefore resorted to in vitro assays utilizing an N-terminal thioredoxin-AtCCD4 fusion expressed in BL21 E. coli cells harboring the groEL-ES chaperone system (Alder et al., 2008;Ilg et al., 2009;Bruno et al., 2014). Experiments were carried out with particle-free cell lysates and substratecontaining detergent micelles.
In conclusion, all apocarotenoids regardless of their chain length were cleaved at the C9-C10 double bond.
As shown above, AtCCD4 converted the C 40 carotenoids β-carotene, β-cryptoxanthin, and zeaxanthin into β-apo-10'carotenal and/or 3-OH-β-apo-10'-carotenal. However, this enzyme also cleaved its own products, β-apo-10'-carotenal and/or 3-OH-β-apo-10'-carotenal. To understand the biological function of AtCCD4, it is crucial to determine the preferred substrate, i.e. whether the enzyme favors C 40carotenoids or apocarotenoids. For this purpose, the assessment of standard parameters, K M and V max , may be flawed due to the biphasic system used and to uncertainties concerning the equivalent micellar packing of the different substrate species in an enzyme-accessible form. Therefore, we have resorted to modeling of time course experiments in one homogeneous assay system in which bicyclic carotenoids are converted into monocyclic apocarotenoids that are again subjected to a further cleavage.

Time course and modelling
Dynamic modeling of AtCCD4 activity in time course experiments was carried out to determine rate constants for the primary (cleavage of bicyclic C 40 carotenoids) and secondary cleavage reactions (cleavage of apocarotenoids) depicted in Fig. 3. Based on this, the rank order of bicyclic substrate decay is determined by the number of OH functions present, namely β-carotene (k β ) >β-cryptoxanthin (k cry-OH +k cry-β ) >zeaxanthin (k zea ), the k-ratios being 1.0:1.9:5.7 (Fig. 3A-C). The asymmetric β-cryptoxanthin is preferentially cleaved at the C9-C10 site next to the unhydroxylated ionone (k cry-β =3.5 k cry-OH ; Fig. 3B) corroborating that hydroxylation hinders cleavage.
It is conceivable that the two equivalent half sides of the symmetric β-carotene substrate can be cleaved by AtCCD4 with the same probability in terms of 'half-side substrate recognition'. Instead of the expected k β ≈2k β-10 (reflecting the double concentration of cleavable end-groups), the model supports k β =2.8 k β-10 , indicating discrimination between the uncleaved bicyclic and the pre-cleaved monocyclic substrate. Furthermore, along these lines of thinking, β-iononecontaining substrate half-sides should be equivalent in bicyclic substrates, namely k β ≈k cry-β . However, this is not the case either, so that the binding of the entire carotenoid molecule and the simultaneous recognition of both rings must be assumed. This notion is further supported by the fact that the model is consistent with a completely different behavior of the monocyclic C 27 apocarotenoid substrates. Here, the preference for cleavage at unhydroxylated ionone rings is lost, the rate constants for the cleavage of β-apo-10'-carotenal and Fig. 3. Time course and dynamic modeling of the AtCCD4 reaction. The conversion of three bicyclic carotenoids (shaded in gray) into the corresponding monocyclic C 27 apocarotenoids and the subsequent cleavage of the latter are shown in (A-C). Note that the final products (3-OH)-β-ionone and the C 14dialdehyde were detected but could not be accurately quantified because of volatility, partition behavior, and instability. The symbols denote experimental data points and represent the average (± SEM) of 3-4 replicates. The line through the data points represents the model fit.

CCD4 and/or CCD7 as potential generators of regulatory metabolites
AtCCD4 is thought to cleave the cis-configured phytoene desaturase (PDS) products phytofluene and/or ζ-carotene to generate regulatory molecules involved in leaf development (Avendaño-Vázquez et al., 2014). Furthermore, genetic evidence indicated a role for an undefined CCD in cleaving cis-configured ζ-carotene desaturase (ZDS) desaturation intermediates to generate a feedback signaling compound regulating the transcription of the tomato PSY1 (Kachanovsky et al., 2012). Therefore, we examined four ζ-carotene isomers as potential AtCCD4 substrates. The canonical PDS product 9,15,9'-tri-cis-ζ-carotene as well as 9,9'-di-cis-ζ-carotene were used, the latter assuming a relevant contribution of the C15,C15' double bond-specific ζ-carotene-cis-trans-isomerase (Z-ISO) in planta. However, none of these substrates was cleaved ( Supplementary Fig. S3A-D). Likewise, the enzyme did not convert the canonical PDS intermediate 9,15-di-cisphytofluene ( Supplementary Fig. S3E) or the isomers 15-cisphytofluene and all-trans-phytofluene ( Supplementary Fig.  S3E). Thus, our in vitro data do not support the proposed involvement of CCD4 in leaf development.
However, AtCCD7 cleaved the 9-cis-configured isomer of ζ-carotene (Fig. 4A-D) yielding the product P7 and small amounts of compounds with similar chromatographic and spectral characteristics (marked by asterisks in Fig. 4A). Subsequent LC-MS and GC-MS analyses revealed the formation of tentatively 9-cis-configured ζ-apo-10'-carotenal (P7, C 27 ; Fig. 5A) and the volatile geranylacetone (C 13 ; Fig. 5B). The other minor products detected in this incubation are probably cis/trans-isomers of P7, which can be formed by non-enzymatic isomerization of P7. Incubation with a ζ-carotene sample, which contained all-trans-ζcarotene as the major isomer, led to traces of P7. We also detected a minor peak (P7*) tentatively identified as the all-trans-isomer of ζ-apo-10'-carotenal. It can be assumed that P7* was formed from P7 by non-enzymatic isomerization. The formation of P7 trace amounts from the all-transζ-carotene substrate is most probably due to unavoidable minor cross-contamination with the 9-cis species (indicated by † in Fig. 4C, D). Cleavage of 9-cis-ζ-carotene indicated an activity with 9'-cis-neurosporene that is half identical with ζ-carotene, while the second, more desaturated half corresponds to lycopene. To obtain this substrate, we used recombinant CRTISO that can produce the corresponding 9'-mono-cis isomer from the poly-cis-compounds proneurosporene and prolycopene (Isaacson et al., 2004;Yu et al., 2011). Incubation with AtCCD7 led to a single detectable product, most probably ζ-apo-10'-carotenal (P7), as verified by UV/VIS spectroscopic properties (Fig. 4E). This product suggests a cleavage in the more desaturated half side of neurosporene ( Fig. 4E; Supplementary Fig. S5 XVIII). This half side should be trans-configured, based on CCD7's strict stereospecificity , and on the preference of CRTISO (Yu et al., 2011). Consequently the ζ-apo-10carotenal yielded from 9'-cis-neurosporene is expected to carry a 9'-cis configuration, as shown in Fig. 5C. It needs to be noted that the conversion was much lower as compared with that obtained with 9-cis-ζ-carotene. (Fig. 4A). AtCCD7 showed weak activity with 9-cis-lycopene, yielding product P8 accompanied by traces of P9 (Fig. 4F), while all-translycopene was not converted (Fig. 4G). LC-MS analysis and comparison with an all-trans-apo-10'-lycopenal standard identified both products as apo-10'-lycopenals, of which P8 represents the all-trans-species and P9 the 9-cis species (Fig. 6A, B). As shown in Fig. 4F, unspecific isomerization of 9-cis-to all-trans-lycopene in the presence of detergents and cleared cell lysates was commonly observed (indicated by † in Fig. 4F). In accordance with this observation and due to AtCCD7's specificity for cleavage of the trans-configured C9-C10 double bond, we assume that P8 was formed from the sole cleavage product P9 by non-enzymatic isomerization.

AtCCD4 cleavage site specificity and substrate preference
We have characterized the substrate, stereospecificity, and the cleavage pattern of AtCCD4. The results presented document AtCCD4 as an enzyme strictly cleaving the C9-C10 double bond. This regional specificity of cleavage was observed with bicyclic C 40 carotenes and maintained with standard plant xanthophylls and derived (hydroxy-) apocarotenoids ( Fig. 2A, B; Supplementary Fig. S2). The cleavage site is thus independent of hydroxyl functions in the C3 and C3' positions or on the presence of a second ionone ring. Cleavage site specificity was also unaffected by the chain length in (monocyclic) apocarotenoids (Fig. 2C, D). However, AtCCD4 requires the presence of at least one ionone ring for activity, since we did not observe any conversion of linear carotenes ( Supplementary Fig. S3). Moreover, AtCCD4 is specific for all-trans-configured substrates and did not convert any mono-or poly-cis-configured substrates. This rules out a hypothetically possible direct contribution to ABA and SL biosynthesis (see AtCCD4 and regulatory molecules). Thus, AtCCD4 substrates need to be bicyclic or monocyclic and in the all-trans configuration to be converted.
Dynamic modeling of AtCCD4 time course experiments was carried out using different types of substrates, namely unhydroxylated (β-carotene), monohydroxylated (β-cryptoxanthin), and dihydroxylated (zeaxanthin) carotenoids, and the derived mono-and unhydroxylated apocarotenoids arising as secondary substrates upon primary cleavage. The resulting rate constants demonstrate that (i) unhydroxylated bicyclic C 40 carotenes are preferred over C 27 apocarotenoids; (ii) the presence of one or more OH groups in the C 40 substrate reduces cleavage activity; and (iii) the latter does not apply to C 27 apocarotenoids which, hydroxylated or not, were converted with very similar albeit low rate constants. It is worth noting that AtCCD1, an enzyme also targeting the C9-C10 double bond, prefers apocarotenoids over bicyclic carotenoids (Schmidt et al., 2006;Ilg et al., 2010). This difference is probably due to different biological functions and might be reflected by the cytoplasmatic localization of CCD1 that may contribute to carotenoid degradation in planta (Gonzalez-Jorge et al., 2013; see the Introduction).
Upon incubation with violaxanthin, AtCCD4 showed only a marginal cleavage activity. However, two earlier publications on CCD4 mutants from potato (RNAi) and Arabidopsis (T-DNA insertion) reported on elevated total carotenoid levels, mainly of violaxanthin followed by antheraxanthin, and lutein (Campbell et al., 2010;Gonzalez-Jorge et al., 2013). This apparent contradiction can be explained by different substrate specificity in planta or by end-product accumulation that is caused by reducing the cleavage rate of the β-carotene precursor. Our in vitro data are consistent with the second option and, hence, they are in line with the in vivo data. However, it cannot be excluded that the enzyme's preference is affected by the microenvironment in planta. Fig. 4. HPLC analysis of AtCCD7 activity with cis-configured carotene desaturation intermediates. (A) AtCCD7 cleaved 9-cis-ζ-carotene yielding P7, tentatively identified as 9-cis-ζ-apo-10'-carotenal. The minor peaks (asterisks) represent unspecific ζ-apo-10'-carotenal isomers-confirmed by isobaric masses (see Fig. 5A)-that arise upon sample processing. (B-D) Formation of traces of P7 upon incubation with other ζ-carotene isomers was due to minor cross-contamination with the 9-cis isomer (indicated by †). Product P7* corresponding to all-transζ-apo-10'-carotenal was probably formed from P7 by unspecific isomerization. (E) AtCCD7 showed low cleavage activity with 9'-cis-neurosporene yielding P7, and (F) with 9-cis-lycopene yielding P8 and P9. P8 was identified by LC-MS analysis as all-trans-apo-10'lycopenal, P9 as the putative 9-cis-apo-10'-lycopenal (see Fig. 6). The appearance of the all-trans-lycopene isomer upon incubation (indicated by †) and the formation of P8 are probably due to spontaneous cis-totrans isomerization of 9-cis-lycopene and P8, respectively. (G) AtCCD7 did not convert all-trans-lycopene. UV/VIS spectra are shown as insets. For structures of the substrates, see Supplementary Figure 5. HPLC system 3 (A-D), HPLC system 1 (E), and HPLC system 2 (F, G) were used.
As deduced from the corresponding rate constants, AtCCD4 showed higher preference for C 40 carotenoids than for apocarotenoids, indicating that the enzyme recognizes the entire C 40 substrate and not only half sides. To gain insight into possible underlying structural elements allowing this discrimination, AtCCD4 structure predictions were carried out with I-TASSER (Zhang, 2008). The maize enzyme VP14 (NCED 3 in Arabidopsis), whose crystal structure has been elucidated (Messing et al., 2010), was used as a comparator. AtCCD4 and VP14 share higher amino acid sequence homology (40% sequence identity) and structural similarity (TM score of 0.901) as compared with the other two CCDs for which structural information is available, SynACO (Kloer et al., 2005) and RPE65 (Kiser et al., 2009).
Common CCD structural features (Sui et al., 2013), including a hydrophobic surface patch to mediate association with membranes, are conserved ( Supplementary Fig. S7A, B). Here, a cavity extends (arrowhead in Supplementary Fig.  S7B) towards the active center lined with mainly hydrophobic residues. It is about 40 Å in length and hence well capable of accommodating entire C 40 carotenoid substrates, as suggested with VP14 (Messing et al., 2010). This supports our observations that are in favor of a recognition of the entire substrate molecule by AtCCD4. The cavity features a conserved motif at its back end (Messing et al., 2010) consisting of two charged and three hydrophobic residues (D 499 PMPK 503 in AtCCD4; pink in Fig. 7 and Supplementary Fig. S6) that might act as a 'bumper' for one ionone ring to restrict substrate penetration. These residues thus contribute to the positioning of the C9-C10 site relative to the active center that is formed by the conserved His298, His347, His412, and His590 (orange in Fig. 7 and Supplementary Fig. S6) co-ordinating Fe 2+ on the central axis of the seven-blade β-propeller domain (blue in Supplementary Fig. S6A). Substrate positioning in VP14 is thought also to be aided by the three phenylalanine residues Phe171, Phe411, and Phe589, of which Phe411 is substituted by Ile411 in AtCCD4 (green in Fig. 7 and Supplementary Fig.  S6) (Messing et al., 2010). The position of the second ring appears less defined, at loop structures situated at the tunnel entrance. The structural identity of the aliphatic proportion in Fig. 5. Identification of the 9-cis-ζ-carotene cleavage product. (A) The 9-cis-ζ-carotene cleavage product P7 (see Fig. 4A) was analyzed by LC-MS. AtCCD7 produced a putatively 9-cis-configured ζ-apo-10'-carotenal (7.91) accompanied by variable amounts of unspecific cis-trans-isomers (7.86), as shown by their isobaric masses. UV/VIS spectra of the products are depicted in insets. (B) GC-MS co-elution with the authentic reference (trace extracted at the indicated masses) and spectral comparison with the NIST 2.0 database identified geranylacetone as the volatile second cleavage product. (C) Schematic cleavage pattern of 9-cis-ζ-carotene. m/z denotes calculated isobaric masses. bicyclic substrates would consequently place a high selective significance on the 'bumper' site capable of accommodating unsubstituted β-ionone functions but also C3' hydroxylated functionss albeit with lower effectiveness. The distance to the reactive center of ~15 Å (dotted line, Fig. 7) might consequently determine regional specificity of cleavage.
VP14 is specific for 9-cis-configured epoxidated xanthophyll. This stereospecificity has been attributed to the presence of Val478, Ala478, or Ile478 in NCEDs, while CCDs carry phenylalanine or methionine in this position. With phenylalanine in this position, this would define AtCCD4 as a trans-specific CCD, a condition that is met by our data.

AtCCD4 and regulatory molecules
For decades, evidence for retrograde regulation of carotenoid biosynthesis has been put forward, for instance in tomato (Giuliano et al., 1993), maize (Bai et al., 2009), daffodil (Al-Babili et al., 1999, potato (Diretto et al., 2010), and carrot (Arango et al., 2014). The underlying mechanisms are far from being understood. However, it can be assumed that less lipophilic apocarotenoids that can be transported outside of the plastid play the role of signaling molecules, as known for ABA and SLs (Nambara and Marion-Poll, 2005;Raghavendra et al., 2010;Al-Babili and Bouwmeester, 2015). As outlined (see the Introduction), data by Kachanovsky et al. (2012) and Avendaño-Vazquez et al. (2014) point to poly-cis-configured carotene desaturation intermediates as signal precursors, and the latter publication suggested a participation of AtCCD4 in the generation of such signals.
However, our in vitro studies show that AtCCD4 is incapable of cleaving any carotene desaturation intermediates. AtCCD4 cleaved neither the cis-ζ-carotene isomers 9,15,9'-tri-cis-, 9,9'-di-cis-, and 9-cis-ζ-carotene nor all-transζ-carotene ( Supplementary Fig. S3). Moreover, it did not convert any other linear carotene, such as phytofluene, neurosporene, or lycopene, irrespective of their stereo-configuration ( Supplementary Fig. S3E-H). Our data are consistent with the report of Huang et al. (2009), which showed that   7. Predicted substrate cavity of AtCCD4. The substrate cavity is highlighted in gray and the conserved histidine residues co-ordinating Fe 2+ are in orange. The conserved DPMPK motif on the back of the substrate cavity thought to restrict substrate penetration is shown in magenta. The 'caging' phenylalanine residues are highlighted in green. The distance from the rear of the cavity to the active center is ~15 Å, approximately the size of a β-ionone moiety, thus positioning the C9-C10 double bond for cleavage. For further explanations, see text.
AtCCD4 does not cleave ζ-carotene or lycopene in carotenoid-accumulating E. coli cells.
Our data do not support a contribution of AtCCD4 to the biosynthesis of the two known carotenoid-derived hormones ABA and SLs. In the case of SLs, this assumption is based on the stereospecificity of this enzyme that did not cleave 9-cis-βcarotene ( Supplementary Fig. 2F), which would lead to 9-cisapo-10'-carotenal, the SL biosynthesis intermediate formed by AtCCD7. Similarly, we did not observe any conversion of 9-cis-violaxanthin ( Supplementary Fig. 2G), indicating that AtCCD4 neither contributes to nor directly interferes with ABA biosynthesis. This corroborates the previous observation that ABA levels in ccd4 knock-down potatoes remain unaffected (Campbell et al., 2010).
The enzyme produces all-trans-β-apo-10'-carotenal and -3-OH-β-apo-10'-carotenal from bicyclic carotenoids. Given that AtCCD4 mediates the synthesis of a signaling molecule, it might be speculated that this molecule is a derivative of these apocarotenoids. Indeed, all-trans-β-apo-10'-carotenal in vitro is a precursor of β-apo-13-carotenone (d'orenone) that is formed by CCD8 enzymes (Alder et al., 2008). D'orenone has been shown to exert a regulatory function upon external application, affecting root hair growth in Arabidopsis (Schlicht et al., 2008). Moreover, this compound triggers indole-3-acetic acid (IAA) synthesis in the ectomycorrhizal basidiomycete Tricholoma vaccinum and promotes lateral root growth in the host tree spruce (Wagner et al., 2016). Taken together, our data do not support the hypothesis that AtCCD4 mediates the formation of signaling molecules from linear intermediates of the carotenoid biosynthesis pathway. However, it cannot be excluded that this enzyme plays a role in leaf morphogenesis or other developmental processes, for instance as a structural component of a complex regulating such processes. This assumption is based on the report of Naested et al. (2004) that shows the association of AtCCD4 with the protein VAR3 in a complex required for plastid development.

A possible role for AtCCD7 in forming linear retrograde signals
Intriguingly, both known carotenoid-derived hormones, ABA and SLs, originate from 9-cis-configured carotenoids (Al-Babili and Bouwmeester, 2015). The 9-cis configuration is also present in the acyclic carotenes that are thought to be the precursor of new signaling molecule(s). All desaturation intermediates have a 9-cis configuration (as in phytofluene) or are cis-configured at both the C9-C10 and C9'-C10' double bond ( Fig. 1; for structures, see Supplementary Fig. S5). Moreover, it can be assumed that plants also contain other 9-cis-configured carotenes, such as 9'-cis-neurosporene and 9-cis-lycopene, which do not act as pathway intermediates. 9'-cis-Neurosporene and 9-cis-lycopene can be enzymatically formed in vitro by the carotene isomerase CRTISO (Isaacson et al., 2004;Yu et al., 2011), a property that has been exploited in this study to generate these substrates.
Based on the reported specificity of AtCCD7 for 9-cisconfigured bicyclic carotenoids , we tested the capability of this enzyme in cleaving desaturation intermediates and 9-cis-configured acyclic carotenes. The enzyme cleaved 9-cis-ζ-carotene, but did not show activity with any of the other ζ-carotene isomers tested. Thus, besides the strict specificity for 9-cis-configured carotenoids, AtCCD7 does not require the presence of an ionone ring structure for activity, in contrast to AtCCD4. Though to a much smaller extent, AtCCD7 also cleaved 9'-cis-neurosporene.

Supplementary data
Supplementary data are available at JXB online. Figure S1. GC-MS analysis of AtCCD4 volatile products. Figure S2. HPLC analysis of AtCCD4 activity with different C 40 xanthophylls and carotenes. Figure S3. HPLC analysis of AtCCD4 activity with carotene desaturation intermediates. Figure S4. HPLC analysis of AtCCD7 activity with carotene desaturation intermediates. Figure S5. Structures of substrates cleaved by AtCCD4 and AtCCD7. Figure S6. Sequence alignment of VP14, AtCCD4, and AtNCED3. Figure S7. AtCCD4 3D model prediction. Supplementary dataset S1. Modeling of AtCCD4 activity.