S ynergistic interactions between carotene ring hydroxylases drive lutein formation in plant carotenoid biosynthesis

Plant carotenoids play essential roles in photosynthesis, photoprotection, and as precursors to apocarotenoids. The plastid-localized carotenoid biosynthetic pathway is mediated by well-defined nuclear-encoded enzymes. However, there is a major gap in understanding the nature of protein interactions and pathway complexes needed to mediate carotenogenesis. In this study, we focused on carotene ring hydroxylation which is performed by two structurally distinct classes of enzymes, the P450 CYP97A and CYP97C hydroxylases and the nonheme di-iron HYD enzymes. The CYP97A and HYD enzymes both function in hydroxylation of β -rings in carotenes, but we show that they are not functionally interchangeable. Formation of lutein, which involves hydroxylation of both β - and ε -rings, was shown to require co-expression of CYP97A and CYP97C enzymes. These enzymes were also demonstrated to interact in vivo and in vitro as determined using bimolecular fluorescence complementation and a pull-down assay, respectively. We discuss the role of specific hydroxylase enzyme interactions in promoting pathway flux and preventing formation of pathway dead-ends. These findings will facilitate efforts to manipulate carotenoid content and composition for improving plant adaptation to climate change and/or for enhancing nutritionally important carotenoids in food crops.


INTRODUCTION
Carotenoids are a large class of isoprenoid pigments synthesized by all photosynthetic organisms, as well as some bacteria, fungi, and aphids (Cuttriss, 2011). In plants, carotenoids serve essential roles in photosynthesis and photoprotection (Jahns and Holzwarth, 2012), and are precursors to apocarotenoids that function in stress and developmental responses (Walter et al., 2010). Plant-derived carotenoids also provide nutritional benefits to humans (von Lintig, 2010;Wurtzel et al., 2012).
The plastid-localized carotenoid biosynthetic pathway is mediated by well-defined nuclear-encoded enzymes. The product of the first committed biosynthetic step, phytoene, is enzymatically converted into all-trans lycopene, the major branch point precursor for downstream carotenoids. (FIG. 1).The linear lycopene is enzymatically converted to carotenes by formation of an ε-ring or β-ring at each end of lycopene. The rings differ only in the position of a double bond. Hydroxylation of the carotene rings is mediated by ring-specific hydroxylases and leads to xanthophylls such as lutein and zeaxanthin.
Though the individual biosynthetic enzymes are known, there is a gap in fundamental understanding of complexes and protein interactions involved in mediating carotenogenesis.
The pathway likely functions as a multienzyme complex(es) to facilitate metabolite channeling as predicted by the absence of pathway intermediates and the presence of complexes containing carotenoid biosynthetic enzymes (Maudinas et al., 1977;Camara et al., 1982;Kreuz et al., 1982;Al-Babili et al., 1996;Bonk et al., 1997;Lopez et al., 2008). In the present study, we examined an intriguing portion of the pathway, where hydroxylation of rings in carotenes is catalyzed by two structurally distinct enzymes, P450 heme (CYP97) and non-heme diiron (HYD) enzymes (Sun et al., 1996;Tian andDellaPenna, 2001, 2004;Kim and DellaPenna, 2006;Quinlan et al., 2007). Hydroxylation of the two β-ionone rings in β-carotene leads to zeaxanthin, while hydroxylation of the one β-ring and one ε-ring in α-carotene leads to lutein.
Hydroxylation of the β-rings in the carotenes is potentially mediated by either the P450-type CYP97A or diiron HYD β-ring hydroxylase enyzmes. Hydroxylation of the ε-ring of αcarotene is performed by another P450 enzyme, CYP97C. Theoretically, a single β-ring hydroxylase should suffice for hydroxylation of the β-ring in both α-carotene and β-carotene. It is unknown why two different β-ring hydroxylases have been maintained throughout evolution; it is possible that their respective activities are not entirely interchangeable. We hypothesized that hydroxylation of each of the carotene rings does not happen independently, but that a synergistic interaction occurs between partner enzymes (CYP97A and CYP97C) to facilitate carotene hydroxylation of α-carotene. To provide support for this hypothesis, we investigated whether carotene hydroxylase enzyme co-expression influenced biosynthesis of enzyme products. We also determined which enzyme partners showed evidence of physical interaction.

Functional complementation in E. coli to test for CYP97 and HYD substrate specificities
A widely used functional complementation method employed in our previous studies demonstrated activity of the P450 and HYD carotene ring hydroxylases from grasses (Quinlan et al., 2007;Vallabhaneni et al., 2009). Compared to β-ring hydroxylation by HYD, CYP97A4 and CYP97C2 were less effective in hydroxylating carotene rings in E. coli accumulating βcarotene or ε-ε-carotene, their respective β-ring or ε-ring substrates. We considered two reasons for the low activity of the CYPs. The first possibility was that the enzymes were not presented with their optimal substrate, α-carotene, which contains both βand ε-rings. The second possibility was that perhaps the CYP97 enzymes did not function optimally as individual enzymes but required co-expression and interaction which would allow for efficient hydroxylation of a mixed-ring compound, such as α-carotene. Biochemical phenotypes of plant knockouts support the hypothesis that CYP97 enzymes act sequentially (first CYP97A and then CYP97C) to hydroxylate α-carotene (Kim and DellaPenna, 2006).
We first tested the effectiveness of α-carotene as a substrate, which can only be produced by engineering bacteria to synthesize both α-carotene and β-carotene. We expressed rice CYP97A4 and CYP97C2 (Quinlan et al., 2007) and maize HYD4 (Vallabhaneni et al., intermediate, β-cryptoxanthin as well as the end product, zeaxanthin. This was the case for HYD4; cells expressing this enzyme accumulated ~30% zeaxanthin. By contrast, cells expressing CYP97A4 mainly accumulated the monohydroxylated intermediate β-cryptoxanthin (17% total carotenoids) while only 3% zeaxanthin was generated (FIG. 2, Table 1). Similar results were observed when cells were engineered to accumulate β-carotene only ( Table 2). It was also expected that these β-ring hydroxylases would hydroxylate α-carotene to form zeinoxanthin, and indeed this product was detected in cells expressing either CYP97A or HYD4, although the HYD4 enzyme was twice as active as CYP97A. In addition, we expected that cells transformed with the ε-ring hydroxylase CYP97C2 would accumulate the monohydroxylated product α-cryptoxanthin. However, this compound was barely detected (~0.7% total carotenoids). These results show that HYD4 was most effective in producing a dihydroxylated carotene, in this case zeaxanthin, which was produced from β-carotene. The above results only partially confirmed the hypothesis that P450 carotene hydroxylases (CYP97A and CYP97C) prefer α-carotene over β-carotene as a substrate. CYP97A appeared to function as a monohydroxylase for either β-carotene or α-carotene, but CYP97C was marginally functional, regardless of the substrate. These experiments also showed that CYP97C could not efficiently hydroxylate carotene β-rings. Such a finding was inconsistent with the proposal that CYP97C could hydroxylate both rings of α-carotene to explain formation of lutein in mutants lacking other known β-ring hydroxylases (Kim et al., 2009). Therefore, we next tested our second hypothesis that CYP97A and CYP97C must be co-expressed and physically interact to convert α-carotene to lutein.
When P450 hydroxylases were co-expressed in the presence of α-carotene and βcarotene, their combined activity was dramatically increased as evidenced by formation of lutein (29% of total carotenoids), representing hydroxylation of the ε-ring in α-carotene by CYP97C2 and the β-ring by CYP97A4 (FIG. 3, Table 3). This level of di-hydroxylated product was comparable to that found for zeaxanthin formation by HYD4 (FIG. 2, Table 1). In contrast, the co-expression of the β-ring hydroxylase HYD4 with the ε-ring hydroxylase, CYP97C2, did not lead to significant levels of hydroxylated carotenes. Perhaps there was a synergistic interaction occurring between P450 enzymes that did not occur between HYD4 and CYP97C2, since creating a monohydroxylated substrate by HYD4 was insufficient for CYP97C2 to efficiently hydroxylate the remaining ε-ring. Our results showed that the CYP97 enzymes must be co-expressed in order for α-carotene to be fully hydroxylated to form lutein, and that the nonheme di-iron β-ring hydroxylase (HYD) was not functionally equivalent to the P450 β-ring hydroxylase CYP97A.
The requirement for co-expression suggested that the CYP97 enzymes might interact with each other to produce the di-hydroxylated carotenes. We predicted that the interacting enzymes should have similar patterns of plastid localization. Moreover, we expected to detect physical interactions in planta between CYP97A and CYP97C but not between CYP97C and HYD enzymes. To test these predictions we carried out the following localization experiments.

Plastid localization of carotene hydroxylases based on chloroplast import studies
Recent proteomic methods utilizing LC-MS/MS showed CYP97A and CYP97C localized to the Arabidopsis chloroplast envelope (Joyard et al., 2009;Ferro et al., 2010).
However, no data were available for location of HYD enzymes. Using the online prediction server TMHMM (Krogh A et al., 2001), HYD4 was predicted to have four transmembrane helices which would be expected to confer an integral membrane localization. The CYP97 structures were not predicted to have transmembrane helices. To test whether the hydroxylases were integrally or peripherally associated with membranes, we conducted in vitro chloroplast import assays (FIG. 4). Radioactively labeled protein precursors were imported into isolated chloroplasts, and then chloroplasts were fractionated into membrane and soluble fractions.
CYP97A4 and CYP97C2 proteins were found in the membrane fraction and dissociated from it upon alkaline treatment, indicating that these proteins were peripherally associated. In addition, a significant amount of the CYP97A4 protein was found in the soluble fraction, which also suggested that the peripheral association of the protein is quite weak, allowing the protein to dissociate into a soluble fraction during the fractionation procedure. In contrast, HYD4, found in the membrane fraction as well, proved to be an integral protein as evidenced by resistance to alkaline treatment. These data are consistent with our structural predictions and earlier studies of a citrus HYD (Inoue et al., 2006).

Testing plastid-localized interactions of partner hydroxylases
The functional complementation experiments in E. coli suggested that a synergistic interaction between CYP97A and CYP97C facilitates lutein formation from α-carotene.
Enzyme interactions were tested in planta using the approach of bimolecular fluorescence complementation (BiFC) (Citovsky et al., 2006) by transient expression in isolated maize protoplasts. Protoplasts maintain their tissue specificity and reflect in vivo conditions (Faraco et al., 2011;Denecke et al., 2012) and are therefore valuable for examining enzyme interactions.
In addition, transient expression is an advantageous approach for monitoring localization of low abundance carotenoid biosynthetic enzymes that evade detection in proteomic studies. In BiFC, putative interacting proteins are fused respectively to non-fluorescent N-terminal (nYFP) and C-terminal (cYFP) halves of the yellow fluorescent protein (YFP); interacting proteins bring together the non-fluorescent fragments, thereby restoring the yellow fluorescence. Constructs encoding fusion proteins were created for CYP97A4, CYP97C2, or HYD4, such that each was fused at their C-termini to N-or C-terminal halves of YFP. The resulting constructs were transiently co-expressed in maize protoplasts and examined using confocal microscopy (FIG.   5A). CYP97A4 and CYP97C2 were found to interact with each other, as shown by restored YFP fluorescence. The interaction of CYP97A4 and CYP97C2 was additionally confirmed by an in vitro pull-down assay (FIG. 6). We also detected HYD4+HYD4 interaction, which suggested HYD4 formed a homodimer. We did not detect homodimers for CYP97A4 or CYP97C2 or heterodimers for CYP97A4 + HYD4 or CYP97C2 + HYD4 (FIG. S2).
Interaction results for all tested protein combinations are summarized in Table 4. For comparison, we also individually expressed the enzymes as GFP fusions to confirm plastid localization in our protoplast system (FIG. 5B). A similar fluorescence pattern indicates that the interaction does not change protein localization as seen for the individually expressed proteins.

DISCUSSION
Interacting proteins exhibit synergistic effects on carotene di-hydroxylation.
Using a bacterial assay system, we showed that di-hydroxylation of α-carotene to lutein requires co-expression of CYP97A and CYP97C, enzymes that interact in planta and interact in vitro in pull-down assays. In contrast, lutein does not form in the case of enzymes that do not exhibit interaction, such as HYD and CYP97C. We hypothesize that a synergistic interaction between CYP97A and CYP97C is required to drive lutein biosynthesis for the purpose of channeling pathway substrates, stabilizing the enzyme-substrate complex and/or promoting interaction with other enzymes or components. We also found that the most efficient dihydroxylation of β-carotene to zeaxanthin was achieved by HYD, an enzyme that could form homodimers. CYP97A, which could not form homodimers, was less efficient in dihydroxylation of β-carotene to zeaxanthin. Although further research is needed to understand the connection between interaction and efficiency of di-hydroxylation in planta, we hypothesize that the ability to form a protein complex improves efficiency of hydroxylation of dual-ringed carotene substrates.

Do CYP97A and CYP97C work simultaneously or sequentially?
The hypothesis based on plant mutants is that in carotenoid biosynthesis, CYP97A functions first to produce the monohydroxylated carotene, zeinoxanthin, which is transferred by some unknown mechanism as the substrate for CYP97C (Kim and DellaPenna, 2006;Kim et al., 2009). In those studies, CYP97C mutants accumulated substantially higher levels of the monohydroxylated carotene, compared to CYP97A mutants, although the levels were not what would be expected on the basis of wild type levels of lutein formed in leaf tissue. In CYP97A mutants, the monohydroxylated α-cryptoxanthin barely accumulated, indicating that CYP97C was unable to hydroxylate the ε-ring of α-carotene. Using the bacterial assay system, we found that CYP97A produced ~20-fold more monohydroxylated carotene (zeinoxanthin) as compared to levels of monohydroxylated carotene (α-cryptoxanthin) catalyzed by CYP97C, when either enzyme was expressed alone in bacterial cells producing α− and β−ring-containing carotenes.
These results show that in bacteria, CYP97A can accept the α-carotene substrate to produce the monohydroxylated product, while CYP97C is limited. Of course, there are many reasons why CYP97C might show poor activity in E. coli (e.g. non-optimized codon usage, missing cofactors). However, the fact that CYP97C can function in bacteria when co-expressed with CYP97A, suggests that CYP97A is the missing factor that must be present in order for CYP97C to function, and together with CYP97A, to produce lutein. The observed interaction between these two enzymes may reflect a stabilizing complex required in the case of CYP97A activity, and a multi-enzyme structure needed to channel the zeinoxanthin substrate, in the case of CYP97C. We propose that protein-protein interaction between CYP97A and CYP97C facilitates recruitment of CYP97C2 to access the zeinoxanthin substrate. Such interaction might also serve in a regulatory role to control pathway flux. It would be intriguing to learn whether these enzymes exist only as a heterodimer or if other proteins, including other carotenoid enzymes, also participate in formation of a metabolon in vivo.
Substrate specificity? From genetic and functional complementation studies (Tian et al., 2004;Kim and DellaPenna, 2006;Quinlan et al., 2007), it is generally accepted that CYP97C is an ε-ring hydroxylase and CYP97A is a β-ring hydroxylase. Mutant phenotypes suggested that CYP97A hydroxylates the β-rings of α-carotene to form zeinoxanthin which is the preferred substrate for ε-ring hydroxylation by CYP97C (Kim and DellaPenna, 2006).
However, biochemical phenotypes of mutants carrying only a subset of carotene ring hydroxylases suggest broader substrate specificity for these enzymes. Arabidopsis mutant plants with only CYP97C (and not CYP97A or the two nonheme carotene hydroxylases) produced significant levels of lutein, ~75% of wild type (wt), and plants with only CYP97A contained lutein about 5% of wt (Kim and DellaPenna, 2006;Kim et al., 2009). The explanation given for the significant level of lutein, which requires both ring-specific hydroxylases though only one is present, is that the remaining CYP has additional activity towards the other ring type. CYP97C, in particular, was thought to have significant β-ring hydroxylase activity, given that the triple mutants still produced 75% wt levels of lutein (Kim et al., 2009). Broad substrate specificity for CYPs was also suggested by the results of CYP97A overexpression (Kim et al., 2010). However, in our earlier E. coli functional complementation studies (Quinlan et al., 2007), we found no evidence that the ε-ring hydroxylase CYP97C could hydroxylate β-rings and in the present study β-ring hydroxylation by CYP97C was minimal.
One explanation (Kim et al., 2009) for the disparity between apparent function of CYP97C in planta and in E. coli was that the engineered E. coli contained only the individual β-ring or εring substrates but not the mixed ring substrates found in plants. Therefore, if CYP97C did indeed hydroxylate both βand εrings as postulated, we should have obtained lutein accumulation in bacteria that produce both βand εring substrates, as evidence of such postulated broad substrate specificity. However, even when simultaneously presented with both the βand ε-rings, neither CYP97C nor CYP97A, when expressed alone, produced detectable levels of lutein, and in general barely produced a di-hydroxylated product. Therefore, we found no evidence for broad substrate specificity to explain lutein formation by a single CYP enzyme as suggested by the plant studies.
The question remains, why is lutein still formed in mutants containing only CYP97C?
Recent studies (Kim et al., 2010) suggest, that another poorly studied paralog CYP97B, which is evolutionarily-related to CYP97A and CYP97C, might exhibit a carotene β-ring hydroxylase activity. If so, then an explanation for the production of lutein in the Arabidopsis triple mutants is the functional CYP97B that is present in the mutant genetic background. If CYP97B is indeed a β-ring hydroxylase, it might function together with CYP97C to form lutein. Similarly, if CYP97B has some minor ε-ring hydroxylase activity, CYP97B could form lutein in partnership with CYP97A. In fact, in triple Arabidopsis mutants, levels of lutein were lower when the "only" ring hydroxylase was CYP97A as compared to when the "only" ring hydroxylase was CYP97C. Therefore, one could speculate that CYP97B is more efficient in hydroxylating βrings as compared to ε-rings. Furthermore, if CYP97B is indeed another βcarotene ring hydroxylase, we might expect CYP97B to form functional enzyme partnerships, as we observed for CYP97A and CYP97C. That is, expression of CYP97B alone would be predicted to be insufficient to mediate carotene di-hydroxylation (e.g. to produce xanthophylls such as lutein). In support of this hypothesis, quadruple mutants (CYP97A, CYP97C and the two nonheme carotene hydroxylases) of Arabidopsis were albino and showed 10% wild-type levels of carotenes but xanthophylls were completely blocked (Kim et al., 2009). Absence of a requisite partner enzyme could explain why xanthophylls are not produced in the quadruple mutant even in the presence of CYP97B. Further enzyme analysis of CYP97B and studies to test for synergistic interaction with the other known carotene hydroxylases are warranted.

Enzyme interchangeability and pathway dead-ends. Since both HYD and CYP97A
have the ability to produce zeinoxanthin by hydroxylation of the β-ring of α-carotene, CYP97A and HYD were expected to be functionally interchangeable. However, results of enzyme coexpression did not support this. Combined expression of HYD and CYP97C was not productive in lutein formation in bacteria, showing that HYD could not substitute for CYP97A, which is consistent with a similar conclusion based on the biochemical phenotypes of plant mutants (Kim and DellaPenna, 2006). If interaction between CYP97C with CYP97A is needed to stabilize carotene ring hydroxylation by CYP97C (e.g. by providing zeinoxanthin substrate), then inability to form a complex could account for the inability of HYD with CYP97C to function together. It is possible that a non-interaction has non-biological explanations.
However, the absence of interaction was consistent with the lack of synergy seen in bacterial co-expression and may be further explained by the fact that the enzymes exist in different membrane settings. Therefore, if and when HYD catalyzes formation of zeinoxanthin from αcarotene in planta, this zeinoxanthin will be a pathway dead-end as far as further conversion to lutein.

CONCLUSION
Based on our results, we conclude that in the branched pathway in plants, the primary route to formation of lutein from α-carotene is mediated by interacting CYP97 enzymes, whereas the primary route for zeaxanthin formation from β-carotene is mediated by HYD enzymes, with some contribution from CYP97A. Our studies also support the widely held notion that carotenoid biosynthesis involves protein complexes to maximize pathway flux. Such an interaction between proteins is a useful regulatory mechanism that allows plants to direct the pathway towards various metabolites as required in certain tissues or conditions. Our studies showing interaction of these proteins for the first time lay the foundation for further investigations into the roles and topologies of the putative carotenoid metabolons. Further understanding of protein-protein interactions in the pathway will provide insight for more efficient engineering of carotenoid composition to avoid dead-end products and improve plant stress responses or nutritional content.

pUC35S-GUS-Nos constructs used for in planta localization:
A full-length cDNA of CYP97A4 without a stop codon was amplified from the pRT-A4   For protein expression and purification, CYP97C2 was PCR-amplified from pTnT-C2 with primers 3209, 3210 and cloned into pET23b+ (Promega) to give a translational fusion with His-tag. The resulting plasmid was named pET23-CYP97C2.
For carotenoid analyses, overnight cultures in LB medium were diluted 50-fold into 50 ml fresh medium, then grown in the dark at 250 rpm at 37°C until OD 0.6 and induced with 10 mM IPTG and further cultured for a total of three days. Negative controls never generated any hydroxylated products.

Extraction of carotenoids from E. coli cells, HPLC and LC-MS analysis
50 ml cultures containing antibiotics for selection of plasmids (Ampicillin, 50 μg/ml; chloramphenicol, 34 μg/ml; kanamycin, 30 μg/ml; streptomycin, 30 μg/ml) were centrifuged at 3000 g, 10 min. Bacterial cell pellets were extracted in 5 ml of methanol using a sonicator (Vibra Cell), and pelleted down at 3000 g, 10 min. Supernatants were transferred to 100 ml Pyrex flasks and evaporated under nitrogen gas, then dissolved in 300 µl of methanol and frozen at -80°C for 30 minutes and pelleted down at 4°C. Supernatants were transferred to HPLC vials (Waters).

HPLC separation was carried out using a Waters system equipped with a 2695
Alliance separation module, a 996 photodiode array detector, a column heater, a fraction collector II, Empower software (Millipore), and a Develosil expressing genes encoding carotenoid biosynthetic enzymes (Cunningham Jr. et al., 1996;Cunningham et al., 2007). Integrated peak areas for extracted metabolites were calculated and carotenoids were quantified as a % of total carotenoids. All data were collected at 450nm. LC-MS was performed on a Waters 2695 HPLC equipped with a 2998 PDA detector coupled to a Waters LCT Premiere XE TOF-MS system using electrospray ionization in positive ion mode. Separation was performed as described above.

Chloroplast isolation and in vitro import
Chloroplasts used in import assays were isolated from 10-14 day old pea plants as described (Bruce et al., 1994). Approx. 25g of leaves were homogenized at 4°C with a blender in 75ml of cold grinding buffer (50mM HEPES pH 8, 0.33M sorbitol, 1mM MgCl 2, 1mM MnCl 2, 2mM Na 2 EDTA pH 8, 0.1% BSA, 0.1% Na-ascorbate) by 3-5 bursts of 1s each. All further operations were performed on ice using cold buffers. The homogenate was filtered through 2 layers of cheesecloth and 1 layer of Nylon mesh (60 µm) and the filtrate was Reaction mixtures were prepared containing purified chloroplasts (0.5 mg/ml), 1X import buffer, 4mM methionine, 4 mM ATP, 4 mM MgCl 2 , 10 mM KAc, 10 mM NaHCO 3 and 10 µl of reticulocyte lysate translation product in a total volume of 150 µl. Reaction mixtures were incubated for 25 min at 25°C in light. Import reactions were stopped by adding 500 µl of import buffer and centrifuged at 800 g, 2 min, 4°C to obtain pellet of intact chloroplasts. Pellets were resuspended in 200 µl import buffer, supplemented by 1mM CaCl 2 and each reaction mixture was divided into two equal aliquots. Thermolysin was added to one aliquot to a concentration of 125 ng/ul and incubated for 30 min at 4°C. The reaction was terminated by addition of EDTA to a concentration of 10 mM. The other aliquot was used as control of the import reaction. For fractionation, chloroplasts were washed twice with import buffer, then diluted with HL buffer (10 mM HEPES-KOH, 10 mM MgCl 2 , pH 8); the total mixture was frozen in liquid nitrogen/thawed 3 times, and then centrifuged at 16 000 g, 20 min. Alkaline treatment of membrane fractions was performed with 200 mM Na 2 CO 3 , pH>10, 10 min on ice, and pellets containing treated membranes were separated from the supernatant by centrifugation at 16 000 g, 20 min. All fractions including soluble, membrane, and purified membrane pellets were analyzed by SDS-PAGE. Radiolabeled protein bands were visualized using a Storm Phosphorimager (Amersham Biosciences).

Isolation and transformation of maize protoplasts
Isolation and transformation of maize protoplasts was performed using protocols from (Sheen, 1991;van Bokhoven et al., 1993)  Transformational efficiency for protoplasts was 80-90%.

Confocal microscopy
Transient expression of GFP or YFP fusion proteins was visualized using a DM16000B inverted confocal microscope with TCS SP5 system (Leica Microsystems CMS, Germany). Oil or water immersion objective (63X) was used in all cases. 488 nm argon laser was used to excite the fluorescence of GFP and chlorophyll. Chlorophyll autofluorescence was detected between 664 and 696 nm. GFP fluorescence was detected between 500 and 539 nm and always confirmed by recording the emission spectrum by wavelength scanning (lambda scan) between 500 and 600 nm with a 3-nm detection window. A 514 nm line of argon laser was used to excite the fluorescence of YFP, and the emission spectrum was detected and confirmed by lambda scan between 524 and 575 nm. LAS AF software (Leica Microsystems CMS, Germany) was used for image acquisition. Images were obtained by combining several confocal Z-planes. Dodecyl β-D-maltoside), 0.5 mM TCEP and 0.12% AEBSF. Samples were rotated at 4°C for one hour. Insoluble material was removed by centrifugation at 5000g, 20 minutes, 4°C. Lysate, containing CYP97C2, was mixed with 600 µl of equilibrated Ni-NTA agarose slurry (Qiagen) and incubated overnight, 4°C. 200 µl of the Ni-NTA agarose slurry containing immobilized CYP97C2 was used for the pull down assay.
CYP97A4 solution was mixed with the CYP97C2 agarose slurry and incubated for 2h at 4°C.
The same amount of CYP97A4 was mixed with 200 µl of pure Ni-NTA agarose slurry as a negative control. After incubation, slurries were loaded onto 1 ml polypropylene columns (Qiagen) and washed three times with Wash Buffer (25mM HEPES pH 7.8, 25 mM NaCl, 40mM imidazole). Interacting proteins were eluted with the same buffer but containing 200mM of imidazole and 5% glycerol. Samples were analyzed by SDS PAGE, the gel was stained with Coomassie Blue for total protein and dried before phosphorimaging.    precursors. Chloroplasts harboring imported proteins were then re-isolated and subjected to thermolysin treatment to distinguish between proteins that were peripherally bound to the outer chloroplast envelope, and those that had been imported thus processed to remove the transit peptide. The mature proteins were recovered as protease-resistant forms (arrow), confirming import of these proteins into chloroplasts.
Chloroplasts containing imported proteins were hyptonically lysed and fractionated into soluble and membrane fractions. The pellet fractions were then treated with an alkaline buffer to wash away peripherally associated membrane proteins. Purity of fractions was controlled by import and fractionation analysis of a chloroplast lumen protein, tpsOE16:: GFP (Marques et al., 2003) and integral thylakoid membrane-bound protein, LHCP (Tan et al., 2001). SDS-PAGE analysis chloroplasts and their fractions indicated that the CYP97A4 and CYP97C2 were synthesized as precursors of about 69 kDa and 62 kDa, and then processed to 64 and 59 kDa respectively. HYD4 was synthesized as a precursor of roughly 34 kDa, and processed to 27 kDa. P, translation products; I, imported protein; (+), thermolysin treatment; S, soluble proteins; M, membrane proteins; MA, alkaline-treated membrane fraction.            . Chloroplast import assays of CYP97 and diiron HYD proteins. Isolated pea chloroplasts were used for in vitro import of 35 S-methionine radio-labeled protein precursors. Chloroplasts harboring imported proteins were then re-isolated and subjected to thermolysin treatment to distinguish between proteins that were peripherally bound to the outer chloroplast envelope, and those that had been imported thus processed to remove the transit peptide. The mature proteins were recovered as protease-resistant forms (arrow), confirming import of these proteins into chloroplasts. Chloroplasts containing imported proteins were hyptonically lysed and fractionated into soluble and membrane fractions. The pellet fractions were then treated with an alkaline buffer to wash away peripherally associated membrane proteins. Purity of fractions was controlled by import and fractionation analysis of a chloroplast lumen protein, tpsOE16:: GFP (Marques et al., 2003) and integral thylakoid membrane protein, LHCP (Tan et al., 2001). SDS-PAGE analysis chloroplasts and their fractions indicated that the CYP97A4 and CYP97C2 were synthesized as precursors of about 69 kDa and 62 kDa, and then processed to 64 and 59 kDa respectively. HYD4 was synthesized as a precursor of roughly 34 kDa, and processed to 27 kDa. P, translation products; I, imported protein; (+), thermolysin treatment; S, soluble proteins; M, membrane proteins; MA, alkaline-treated membrane fraction.  are interacting with each other as seen by restored YFP fluorescence. Fusions of nYFP and cYFP with ChrD protein from cucumber, which is known to form homodimer complexes in plastids (Libal-Weksler et al., 1997), were used as a positive control. B, Transient expression of GFP-fused proteins in maize protoplasts. CYP97 proteins are localized throughout etioplasts, and concentrated at the spot of red chlorophyll autofluorescence of prolamellar bodies, as would be expected for proteins with stromal/weak peripheral membrane association. HYD4 is strictly co-localized with prolamellar bodies consistent with integral thylakoid membrane binding. Chlorophyll, chlorophyll autofluorescence. Scale bar = 10µm.