The cytochrome P450 enzyme CYP96A15 is the midchain alkane hydroxylase responsible for formation of secondary alcohols and ketones in stem cuticular wax of Arabidopsis.

Most aerial surfaces of plants are covered by cuticular wax that is synthesized in epidermal cells. The wax mixture on the inflorescence stems of Arabidopsis (Arabidopsis thaliana) is dominated by alkanes, secondary alcohols, and ketones, all thought to be formed sequentially in the decarbonylation pathway of wax biosynthesis. Here, we used a reverse-genetic approach to identify a cytochrome P450 enzyme (CYP96A15) involved in wax biosynthesis and characterized it as a midchain alkane hydroxylase (MAH1). Stem wax of T-DNA insertional mutant alleles was found to be devoid of secondary alcohols and ketones (mah1-1) or to contain much lower levels of these components (mah1-2 and mah1-3) than wild type. All mutant lines also had increased alkane amounts, partially or fully compensating for the loss of other compound classes. In spite of the chemical variation between mutant and wild-type waxes, there were no discernible differences in the epicuticular wax crystals on the stem surfaces. Mutant stem wax phenotypes could be partially rescued by expression of wild-type MAH1 under the control of the native promoter as well as the cauliflower mosaic virus 35S promoter. Cauliflower mosaic virus 35S-driven overexpression of MAH1 led to ectopic accumulation of secondary alcohols and ketones in Arabidopsis leaf wax, where only traces of these compounds are found in the wild type. The newly formed leaf alcohols and ketones had midchain functional groups on or next to the central carbon, thus matching those compounds in wild-type stem wax. Taken together, mutant analyses and ectopic expression of MAH1 in leaves suggest that this enzyme can catalyze the hydroxylation reaction leading from alkanes to secondary alcohols and possibly also a second hydroxylation leading to the corresponding ketones. MAH1 expression was largely restricted to the expanding regions of the inflorescence stems, specifically to the epidermal pavement cells, but not in trichomes and guard cells. MAH1-green fluorescent protein fusion proteins localized to the endoplasmic reticulum, providing evidence that both intermediate and final products of the decarbonylation pathway are generated in this subcellular compartment and must subsequently be delivered to the plasma membrane for export toward the cuticle.

Plant cuticles are composed of the fatty acid polyester cutin as well as complex mixtures of very-long-chain (VLC) aliphatic lipids, which form the component known as cuticular wax. Wax that is localized within the cutin matrix is designated intracuticular wax, whereas wax that is deposited as a film or as microcrystals on the outer surface of the cutin polymer is known as epicuticular wax (Jeffree, 2006). Cuticular waxes contain species-characteristic compound classes and chain length patterns . The cuticular wax mixture of Arabidopsis (Arabidopsis thaliana) leaves consists of VLC alkanes, aldehydes, fatty acids, primary alcohols, corresponding alkyl esters, and small amounts of triterpenoids; the wax on inflorescence stems, although similar, contains high amounts of secondary alcohol and ketone constituents (Rashotte et al., 1997). In fact, alkanes, secondary alcohols, and ketones comprise 80% to 90% of total Arabidopsis stem wax (Rashotte et al., 1997) and it has been postulated that all three compounds are involved in forming the epicuticular wax crystals found on stem surfaces (Rashotte and Feldmann, 1998;Jetter et al., 2006). Cuticular wax components are formed by elongation of saturated fatty acyl chains and their modification either by the acyl-reduction pathway, yielding primary alcohols and esters, or by the decarbonylation pathway, leading to aldehydes, alkanes, secondary alcohols, and ketones ( Fig. 1; Kunst et al., 2006). Forward-genetic approaches have successfully identified some of the genes involved in plant wax biosynthesis from several plant species. For example, visual screenings of mutagenized Arabidopsis populations have yielded a set of 21 eceriferum (cer) mutants with characteristic changes in the composition of inflorescence stem wax (Koornneef et al., 1989;McNevin et al., 1993). Some of the genes mutated in these cer lines have been cloned and characterized, including those coding for the ketoacyl-CoA synthase CER6 (Millar et al., 1999;Fiebig et al., 2000;Hooker et al., 2002) and the enoyl-CoA reductase CER10 (Zheng et al., 2005) involved in fatty acid elongation, as well as the fatty acyl-CoA reductase CER4 that is responsible for primary alcohol synthesis (Rowland et al., 2006). However, not a single step in the decarbonylation pathway has been confirmed through the identification, cloning, and characterization of the predicted enzyme. A few Arabidopsis genes causing mutant wax phenotypes have been putatively linked to early steps in the decarbonylation pathway; among them, CER1 and CER3 have been cloned (Aarts et al., 1995;Rowland et al., 2007), but their biochemical functions are still not established.
The last two steps in the decarbonylation pathway are thought to proceed by consecutive oxidation reactions first leading from alkanes to secondary alcohols, and then from secondary alcohols to ketones . The biosynthetic link between these three compound classes was first substantiated by biochemical data showing that both alkanes and secondary alcohols can be transformed into ketones in Brassica oleracea leaf disc assays (Kolattukudy and Liu, 1970;Kolattukudy et al., 1971Kolattukudy et al., , 1973. Based on these findings, it was postulated that alkane hydroxylation and subsequent alcohol oxidation steps are catalyzed by one or two mixed-function oxidases. Further indirect evidence for this reaction sequence came from Arabidopsis Figure 1. Biosynthetic pathways leading to Arabidopsis cuticular wax components. The acyl reduction pathway yields primary alcohols with even carbon numbers as well as the corresponding wax esters. The decarbonylation pathway leads to the formation of alkanes, secondary alcohols, and ketones with odd carbon numbers. In the decarbonylation pathway, only the reactions modifying 30:0 acyl CoA are shown to illustrate the formation of products with prevalent chain lengths. The proposed action of MAH1 in the decarbonylation pathway is indicated. studies in which deficiencies in alkanes were correlated with a lack of secondary alcohols and ketones (Millar et al., 1999;Chen et al., 2005); also, cer20 mutants were reported to have reduced levels of secondary alcohols and ketones, but not alkanes (Rashotte et al., 2001), suggesting that the defective gene product is involved in the alkane hydroxylation step, or possibly also in secondary alcohol oxidation. Unfortunately, no genes or enzymes involved in the process have been identified to date, and therefore the formation of these major compound classes of Arabidopsis wax remain to be investigated on the molecular level.
In the current project, we complemented the previous forward-genetic experiments with a reversegenetic approach to identify genes responsible for ketone formation in Arabidopsis stems. Prime candidates for the mixed-function oxidases (or monooxygenases) were expected to be in the family of cytochrome P450dependent enzymes because prior studies had delineated roles for cytochrome P450 oxidation of various lipophilic plant products. Examples include hydroxylation steps in the biosynthesis of terpenoids (Kim et al., 2005;Morikawa et al., 2006;Mau and Croteau, 2006), alkaloids (Collu et al., 2001(Collu et al., , 2002, and phenolics (Whitbred and Schuler, 2000;Ehlting et al., 2006). Most notably, cytochrome P450 enzymes have recently also been shown to be involved in the synthesis of cutin monomers by hydroxylating methyl groups at the v-chain terminus of fatty acids (for review, see Kandel et al., 2006). It stands to reason, then, that a cytochrome P450 enzyme could also be involved in the hydroxylation of alkanes during wax biosynthesis.
The Arabidopsis genome contains 272 cytochrome P450 genes (including 26 pseudogenes; Rhee et al., 2003;Wortman et al., 2003) of which 33 have been grouped together in the CYP86 clan of non-A-type P450s, which include the subfamilies CYP86, CYP94, CYP96, and CYP704 (Durst and Nelson, 1995;Nelson et al., 2004). Members from this clan have previously been shown to be involved in the hydroxylation reactions of fatty acyl-derived substrates (Kahn et al., 2001;Wellesen et al., 2001;Werck-Reichhart et al., 2002) and therefore genes within this clan may be regarded as candidates for cuticle biosynthetic enzymes. To further narrow the choice of gene candidates, expression data have to be taken into account. In particular, a recent microarray experiment revealed the genes up-regulated in epidermal cells in stem regions of active wax synthesis and secretion (Suh et al., 2005). A cytochrome P450 gene (CYP96A15, At1g57750) with currently unknown function showed 2-to 3-fold up-regulation in the stem epidermis when compared with total stem tissue and we speculated that it may encode the mixed-function oxidase responsible for formation of secondary alcohols and/or ketones found in Arabidopsis stem wax. To test this hypothesis, we performed experiments addressing the following questions: Is this gene involved in the decarbonylation pathway of wax biosynthesis? Can the corresponding gene product catalyze the oxidation from alkanes to secondary alcohols and/or the oxidation from secondary alcohols to ketones? Is this the sole enzyme involved in production of secondary alcohols and ketones or are there other related enzymes? Where in Arabidopsis stem epidermal cells is the enzyme localized?

RESULTS
To investigate whether CYP96A15 (At1g57750) codes for an enzyme in the decarbonylation pathway catalyzing the cuticular wax secondary alcohols and ketones, three sets of experiments were carried out. First, the wax composition and load of selected T-DNA insertional mutant lines was assessed. Second, ectopic expression of the gene was employed to determine its biochemical function. Finally, gene expression patterns and subcellular localization of the gene product were examined.

Molecular Characterization of the CYP96A15 Gene and of T-DNA Insertional Mutant Lines
To test whether the CYP96A15 gene is involved in wax biosynthesis, three T-DNA insertional mutants were obtained and characterized. Because we hypothesized that the corresponding gene product catalyzes a reaction introducing a hydroxyl group on a methylene unit in the middle of alkane molecules, we called the three mutant lines mah1-1, mah1-2, and mah1-3 (for midchain alkane hydroxylase). Accordingly, the CYP96A15 gene was tentatively designated as MAH1.
Initially, the wild-type MAH1 gene was sequenced, corroborating previously published genomic information. Genomic DNA and cDNA templates gave identical full-length PCR products, thus confirming predictions that the MAH1 gene does not contain any introns. To determine the nature and extent of gene disruption in the three T-DNA insertional lines, MAH1 genomic regions were PCR amplified and sequenced. Sequencing confirmed the respective published T-DNA insert locations ( Fig. 2A). Briefly, mah1-1 plants had a T-DNA insertion within the coding region of the MAH1 gene, 692 bp downstream from the translational start site; mah1-2 plants were found to have a T-DNA insertion located in the 3#-untranslated region (UTR) of the gene, 1,601 bp downstream from the translational start site; and mah1-3 plants contained a T-DNA insertion in the promoter region of the gene, 216 bp upstream of the translational start site. Semiquantitative reverse transcription (RT)-PCR assays using stem total RNA from these same mutants revealed normal MAH1 steadystate transcript levels in mah1-2, reduced transcript levels in mah1-3, and truncated transcripts in mah1-1 (Fig. 2B). All three mutant lines were observed to bolt and flower earlier than wild-type controls under the growth conditions used in this study.

Stem Cuticular Wax Phenotypes of mah1 Mutants
To verify the involvement of MAH1 in wax biosynthesis, cuticular wax on inflorescence stems of homozygous mah1 mutant plants was analyzed and compared with the corresponding wild type. Only mah1-3 plants were found to have total wax loads of 27 mg/cm 2 , identical to the wild type (Table I). In contrast, mah1-1 and mah1-2 stems had wax coverages of 16 mg/cm 2 and 21 mg/cm 2 , respectively, corresponding to approximately 40% and 15% reductions in wax amounts. mah1-1 exhibited the most severe changes not only in overall wax amounts, but also in wax composition (Table I). Secondary alcohols and ketones were reduced from 2 mg/cm 2 and 6 mg/cm 2 in the wild type to 0.1 mg/cm 2 and , 0.05 mg/cm 2 in mah1-1, respectively, whereas the alkane fraction was found to be slightly increased from 10 mg/cm 2 in the wild type to 11 mg/cm 2 in the mah1-1 mutant. Coverage of aldehydes, fatty acids, and primary alcohols was also reduced to some extent, whereas ester and triterpenoid amounts were similar to wild-type stem wax.
The other two mutants, mah1-2 and mah1-3, also differed significantly from wild type in the levels of the major products of the decarbonylation pathway, whereas showing relatively small and mostly nonsignificant changes in the minor compound classes (Table  I). mah1-2 secondary alcohol and ketone levels were reduced to 71% and 25% of the wild-type loads, respectively, but these reductions were not as severe as those seen in mah1-1. At the same time, the mah1-2 mutant also had alkane coverage increased above that of the wild type, with amounts similar to mah1-1. , the 3#-UTR sequence is indicated by diagonal hatching down and to the right, and the 5#-UTR sequence is indicated by diagonal hatching down and to the left. Arrows and numbers above them indicate the positions of the T-DNA insertions (triangles). B, Analysis of MAH1 transcript levels in mutant stems. MAH1 expression was analyzed by RT-PCR using total RNA and ACTIN2 as a control. Two different sets of primers were employed to screen mah1-1 to check for transcriptional knockout: Gene-specific results depicted in mah1-1a correspond to a unique set of primers ranging from the start site of the gene to approximately 60 bp upstream of the T-DNA insertion; mah1-1b results correspond to a set of primers spanning from 34 to 468 bp downstream of the insertion. Table I. Cuticular wax composition of wild-type and mutant inflorescence stems Total wax amounts and coverage of individual compound classes (mg/cm 2 ) are given as mean values 6 SE for individual compound classes (n 5 9 in three batches). Different letters within the compound class show that wax loads differed significantly between respective lines (mixed-effect ANOVA; Tukey-Kramer posthoc test, P , 0.05). Trace and 0.0 signify values that were below 0.05 mg/cm 2 . Mixed-effect ANOVA (a 5 0.05) was conducted using treatment as a fixed effect and batch nested within treatment as a random effect. Pertinent results for each compound class were as follows: Total load, F3, 8.132 5 11.083, P 5 0.003, b 5 0.026; alkanes, F3, 8.108 Although mah1-3 had wild-type levels of secondary alcohols and the least severe ketone reduction (down to 53% of wild-type loads), this mutant had the highest accumulation of alkanes of all three mutant lines. The increase in alkane amounts within this mutant very nearly compensated for the reduction in ketones, resulting in an overall wax coverage similar to that of the wild type. The chain length patterns within wax compound classes were very similar for all three mutant lines ( Fig.  3), showing dramatic increases of the C 29 alkane over the wild-type level, whereas alkanes of other chain lengths were not affected. Taken together, the observed changes in the three major classes of stem wax compounds synthesized by the decarbonylation pathway in all three mutant lines clearly point to the involvement of MAH1 in this branch of wax biosynthesis.
Remarkably, in spite of all the aforementioned quantitative (wax amounts) and qualitative (compositional) chemical variation between mutant and wild-type stem waxes, there were no discernible differences regarding the numbers, arrangement, size, or shape of epicuticular wax crystals on the stems (Supplemental Fig. S1).

Mutant Complementation and Ectopic Overexpression
To confirm the involvement of the MAH1 gene in wax biosynthesis, MAH1 was expressed as a GFP fusion (MAH1:GFP) in the mah1-1 mutant background under the control of either the native MAH1 or the cauliflower mosaic virus (CaMV) 35S promoter. Multiple transgenic plants were analyzed and all were found to have significant restoration of the cuticular wax phenotype. Thin-layer chromatography (TLC) analyses of stem wax extracts showed a characteristic pattern of compound classes for the wild type from which mah1-1 differed by the lack of secondary alcohols and ketones (Fig. 4). In contrast, stem wax from transgenics expressing MAH1 (in the mutant background) showed a TLC pattern similar to wild type, containing secondary alcohols and ketones together with all other stem wax compound classes.
Further gas chromatography (GC) analyses revealed that a representative line expressing a C-terminal MAH1:GFP fusion under the control of the 35S promoter had stem wax containing ketones and secondary alcohols at 44% and 62% of wild-type levels, respectively; alkane coverage was found to be only 13% higher than in the wild type, thus well below the mah1-1 levels (Table I). Similarly, the stem wax mixture of a line in which MAH1 was expressed under the control of its native promoter contained ketones and secondary alcohols at approximately 30% and 77% of wild-type stem levels, respectively, and alkane coverage approximately 15% above wild type. Altogether, these results confirm that the mutant wax phenotypes affecting products of the decarbonylation pathway are due to a defective MAH1 gene.
Leaves of transgenic plants expressing MAH1-GFP under the control of the CaMV 35S promoter were used to further study biochemical activity of MAH1. Whereas alkanes dominated the leaf waxes of both the wild type and the mah1 mutant, secondary alcohols and ketones could not be detected by TLC in either line (Fig. 4). Trace levels of the latter two compound classes could be identified by GC-mass spectrometry (MS) in wild-type leaf wax, but not in the mutant (data not shown). Thus, leaves represent an ideal tool for biochemical characterization of the MAH1 enzyme. MAH1 was strongly expressed in leaves under the control of the CaMV 35S promoter (see Fig. 7F). TLC analysis showed that leaf wax of transgenics contained not only the characteristic wild-type wax constituents, but also substantial amounts of compound classes that coeluted with secondary alcohol and ketone standards (Fig. 4). The new leaf wax components were further studied by GC-MS and determined to be midchain secondary alcohols and ketones with functional groups either on or next to the central carbon of the chain, thus matching those found in stem wax (data not shown). Therefore, the results of ectopic expression confirm our hypothesis that MAH1 (CYP96A15) codes for the enzyme catalyzing the conversion of wax alkane substrate into midchain alcohols and ketones and validate the gene's designation as MAH1. Further scanning electron microscopy (SEM) comparisons of leaves from transgenics expressing MAH1-GFP under the control of the CaMV 35S promoter and from mah1-1 and wild-type controls revealed no morphological differences (cell shape or wax crystals) caused by the accumulation of these additional compounds (data not shown).

MAH1 Gene Expression Patterns
To investigate the expression patterns of the MAH1 gene on the organ level, experiments using GUS fusion constructs and RT-PCR were performed. When expressed under the control of the native MAH1 promoter in the wild-type background, GUS protein activity was found to be mainly present in nascent stems (Fig. 5A), petioles (Fig. 5, B and C), and developing siliques (Fig.  5, B, E, and F). Detailed analyses of stem expression revealed a pronounced gradient, where GUS activity was strongest in the upper 4 to 5 cm of the stem, still present in the next 10 to 12 cm, and conspicuously absent in stem regions more than 17 cm away from the meristem. Depending on the stage of development, floral organs also stained differentially for GUS, with nascent tissues exhibiting no staining (Fig. 5C) and older pistils revealing various levels of GUS activity (Fig. 5,D and E). Petals and sepals showed no GUS expression, with the exception of faint expression in the sepals of some older flowers (Fig. 5D). Under the same staining conditions, expression of GUS was absent from rosette leaves (data not shown), cauline leaves (Fig. 5A), seeds (Fig. 5F), and roots (Fig. 5G).
Stem-specific expression of MAH1 was further verified by RT-PCR. Transcript levels of MAH1 were found to be high in stems and flower buds of 4-week-old plants, whereas expression levels were relatively low (but consistently detectable) in roots and leaves (Fig. 6).
To further study MAH1 expression in tissues within organs, the MAH1:GFP fusion construct under the control of the MAH1 native promoter in mah1-1 was employed. Stem cross sections revealed that MAH1 expression was confined primarily to the layer of epidermal cells (Fig. 7A). MAH1-associated GFP fluorescence was localized in the pavement cells of the stem and petiole epidermis, but was absent from the guard cells and trichomes of these organs (Fig. 7, B and C). The same MAH1:GFP construct, when expressed under the control of the CaMV 35S promoter instead of the native promoter, was found in pavement cells as well as guard cells (Fig. 7D) and trichomes (data not shown) of the stem. This finding underscores that the native MAH1 promoter alone causes the pavement cell-specific expression of this gene. Under the control of the native promoter, the fusion protein was only weakly expressed in some sepal tissues (Fig. 7B) and was absent from leaves ( Fig. 7E). In contrast, intense GFP fluorescence could be detected in pavement, guard, and trichome cells of transgenic leaves expressing the fusion driven by the 35S promoter (Fig. 7F).

Subcellular Localization of the MAH1 Gene Product
GFP fusion constructs were also used to study the subcellular localization of the MAH1 gene product and with it the site of the final steps in the decarbonylation pathway of wax biosynthesis. Under the control of either the native or the CaMV 35S promoters, GFP fluorescence in the stem epidermis of MAH1:GFP Arabidopsis transgenic lines was most intense within reticulate networks typical of the endoplasmic reticulum (ER; Fig. 8). Subsequent treatment with rhodamine B hexyl ester, a dye capable of staining the ER, confirmed ER localization of the MAH1 protein (Supplemental Fig. S2). When expressed under the control of the 35S promoter, MAH1 fusion protein was also localized in the ER of Arabidopsis leaf epidermal cells. Transient expression of CaMV 35S-expressed MAH1:GFP in wild-type tobacco (Nicotiana tabacum) leaves also displayed a subcellular pattern akin to that observed in mah1-1 plants (data not shown).

DISCUSSION
The principal goal of this study was to identify and characterize a gene encoding a wax biosynthesis enzyme involved in the decarbonylation pathway. We chose the cytochrome P450 gene CYP96A15 as a primary candidate and hypothesized that the corresponding enzyme catalyzes the oxidation reactions yielding the secondary alcohols and ketones that accumulate in the stem wax of Arabidopsis.

MAH1 (CYP96A15) Is a Midchain Alkane Hydroxylase Involved in the Formation of Secondary Alcohols and Ketones in Arabidopsis Stems
Several allelic T-DNA insertion mutants of the MAH1 (CYP96A15) gene all showed cuticular wax phenotypes, with secondary alcohols and ketones either missing entirely or present in significantly smaller quantities than on the wild-type stem. Wax of the mutant lines also contained alkanes in higher relative (% of total wax) and absolute (mg/cm 2 ) amounts than  the wild type, which in turn partially or fully compensated for the loss of secondary alcohols and ketones. Other changes in mutant wax composition were relatively minor. Secondary alcohol and ketone deficiencies were partially rescued by expression of MAH1 under the control of either the native promoter or the CaMV 35S promoter. Taken together, these results clearly show that the cytochrome P450 enzyme encoded by MAH1 (CYP96A15) is involved in the formation of wax secondary alcohols and ketones.
In combination with previous reports in the literature, our results can be used to specifically define substrates utilized by the CYP96A15 enzyme. Prior studies had demonstrated that exogenous nonacosane (C 29 alkane) is incorporated into secondary alcohols and ketones by B. oleracea leaves (Kolattukudy and Liu, 1970;Kolattukudy et al., 1973). As a result, alkane is assumed to be the in vivo substrate for the hydroxylation reactions in the wax biosynthetic pathway yielding secondary alcohols and ketones ( Fig. 1; Kolattukudy, 1996;Rashotte et al., 2001;Jenks et al., 2002). However, because these biosynthetic steps had not been confirmed to date, the alternative scenario could not be ruled out that acyl or aldehyde precursors may also be hydroxylated and the resulting hydroxyacids or hydroxyaldehydes transformed into secondary alcohols . According to this hypothesis, hydroxylation would occur before decarbonylation and secondary alcohols and alkanes would be formed as parallel, rather than direct, sequential products in the pathway. However, none of the hydroxylated intermediate products that would be produced by this alternative scenario (in particular, hydroxyaldehydes) have been reported in Arabidopsis cuticular lipids to date. Our finding that reduced levels of secondary alcohols and ketones occur in tandem with increased levels of alkanes also points to a direct precursor-product relationship between these compound classes, arguing further against aldehyde or acid hydroxylation. Thus, it appears most likely that the CYP96A15 enzyme functions as an alkane hydroxylase.
Ectopic expression of MAH1 in Arabidopsis leaves led to the formation of secondary alcohols and ketones with chain lengths and isomer compositions matching those on wild-type stems. This result, together with the absence of those compounds in mah1 mutants, shows that the enzyme is specifically hydroxylating the central carbons of its alkane substrates, thus confirming its designation as a midchain alkane hydroxylase. To our knowledge, this is the first alkane hydroxylase reported from plants. Only few such enzymes had previously been described in bacteria and yeast (Saccharomyces cerevisiae; van Beilen et al., 2003;van Beilen and Funhoff, 2007) and only a few of these are able to metabolize medium-to long-chain alkane (#C 16 ) substrates. Furthermore, among the few long-chain alkane hydroxylases that have been successfully cloned and characterized, none are reported to be cytochrome P450 enzymes (Tani et al., 2001;van Beilen and Funhoff, 2007). MAH1 (CYP96A15) is currently classified as a member of the CYP86 clan of non-A-type P450 enzymes, most of which are thought to use fatty acids as substrates (Nelson et al., 2004). The biochemical action of MAH1 is quite unique because it is able to hydroxylate long-chain alkanes in the center of the molecule (midchain). This distinguishes it from most other known P450 enzymes, alkane hydroxylases, and fatty acid hydroxylases, which can only catalyze reactions on terminal or subterminal carbons of the substrates (e.g. v-1 or v-2; Whyte et al., 1998;van Beilen et al., 2003;Kandel et al., 2005Kandel et al., , 2006. To date, only one other P450 enzyme with in-chain hydroxylase activity has been described and it prefers C12 fatty acid substrates very different from the VLC alkane substrates of MAH1 (Morant et al., 2007). The ability of MAH1 to catalyze midchain alkane hydroxylation suggests a unique active-site morphology for this enzyme, a unique folding of the alkane substrate, or perhaps a combination of these two conditions.
Current gene ontology classifies MAH1 as a eukaryotictype P450 enzyme (Mulder et al., 2007). Based on its sequence, the protein is predicted to have a hydrophobic transmembrane (Transmembrane Hidden Markov Model, version 2; Sonnhammer et al., 1998) N-terminal signal anchor (0.903 probability using SignalP 3.0 with defaults; Bendtsen et al., 2004;Emanuelsson et al., 2007) with secretory/ER targeting (TargetP 1.1 5 0.970; WoLF PSORT 5 5.0 with defaults; Nakai and Horton, 1999;Bendtsen et al., 2004;Emanuelsson et al., 2007). Our subcellular localization results confirm these bioinformatic predictions. E-type class II P450 enzymes associated with the ER most frequently have a cluster of Pros (Pro-Pro-X-Pro) preceded by a cluster of basic residues (the halt-transfer signal) between the hydrophobic amino-terminal membrane-anchoring segment and the globular part of the protein (Werck-Reichhart and Feyereisen, 2000); curiously, MAH1 possesses such a Pro-rich ER-targeting region, but it is located in the middle of the protein primary sequence at position 365 to 369 of the 497amino acid protein. It should be noted that MAH1 also contains a group I domain similar to some other P450 enzymes and proteins of diverse other families.

MAH1 Is Involved in Secondary Alcohol and Ketone Biosynthesis
The reaction sequence leading from alkanes to secondary alcohols and ketones involves two distinct steps that might be catalyzed either by the same cytochrome P450 or by two separate enzymes, with at least the first of them being a cytochrome P450. This raises the question of whether one or both of these steps are catalyzed by MAH1 in Arabidopsis. Regarding the first reaction, the stem wax of mah1-1 mutants was found to lack not only the ketone end products of the pathway, but also the secondary alcohol intermediates. Furthermore, the ectopic expression of MAH1 in Arabidopsis leaves led to the formation of secondary alcohols above the levels found in wild-type wax. These results both unambiguously show that the enzyme must be involved in the initial hydroxylation transforming alkanes into secondary alcohols.
Our ectopic expression data also provide information on whether MAH1 further catalyzes the second oxidation from secondary alcohols to ketones. Expression of MAH1 in the leaf epidermis led to the accumulation of ketones (together with secondary alcohols) well above the trace amounts found in the leaf wax of the wild type. The appearance of ketones may be due to (1) MAH1 carrying out both steps of the reaction sequence; (2) a specific leaf enzyme being dedicated to this step; or (3) an unspecific enzyme catalyzing the further oxidation of secondary alcohols in the leaf.
The existence of a leaf enzyme capable of oxidizing wax secondary alcohols to ketones might explain the low levels of ketones that have been reported for the waxes of the Landsberg erecta and Wassilewskija wildtype leaves (Rashotte et al., 1997(Rashotte et al., , 2001 and found here for the Columbia (Col) wild type. However, it is important to note that both new products accumulated in the transgenic leaf wax in a ratio similar to that on the wild-type stem, making it unlikely that a nonspecific leaf enzyme catalyzed the second reaction while MAH1 carried out the first step (scenario 3).
The second oxidation (secondary alcohols to ketones) can be likened to the initial step (alkanes to secondary alcohols) if both reactions are described as hydroxylations occurring on the same carbon atom of the two substrates. The second hydroxylation then transforms an alcohol (-CHOH-) into a geminal diol (-C(OH) 2 -) that will be spontaneously dehydrated into the final ketone (-CO-) product. Both reactions are thus very similar, each replacing one hydrogen atom of the methylene unit by an OH group. Therefore, it seems plausible that a single enzyme, capable of carrying out hydroxylations on the central carbon, should catalyze both consecutive reactions leading from alkanes to secondary alcohols and on to ketones. This hypothesis argues against the presence of a specific enzyme carrying out the second oxidation reaction in the leaves, with MAH1 catalyzing only the first step (scenario 2).
Further evidence against scenario 2 is provided by a comparison between the secondary alcohol and ketone amounts in the stem waxes of wild type and the mah1-2 and mah1-3 lines. If the two reactions were carried out by two independent enzymes, then the second enzyme should have slightly lower activity than (approximately 75% of) the first one (MAH1) to account for the 1:3 ratio of secondary alcohols to ketones in the wild type. Consequently, mah1 mutations affecting only the first step of the sequence should reduce secondary alcohol levels much more than ketone levels. In stark contrast to this scenario, all the mah1 mutants have more severely reduced ketone than secondary alcohol levels. The mah1-3 mutant can serve as an extreme example where secondary alcohol levels are identical to the wild type, whereas ketone levels are reduced by approximately 50%. Of all the above arguments, we favor the hypothesis that MAH1 acts both as an alkane hydroxylase and as a secondary alcohol hydroxylase (scenario 1).
The double activity of MAH1 is reminiscent of a few other cytochrome P450 enzymes that also oxidize hydrocarbons to ketones (Turnbull et al., 2004;Jungmann et al., 2005;Ortiz de Montellano and De Voss, 2005). However, possible double activity of MAH1 would differ from these ketone-forming enzymes because it produces a characteristic mixture of ketone and secondary alcohol products. This mixture occurred at various MAH1 enzyme levels, both in wild-type leaves with weak expression of the MAH1 gene and in leaves ectopically expressing it at high levels. These results are in good accordance with the hypothesis that MAH1 catalyzes the second reaction step (scenario 1); however, they do not provide sufficient proof for it at this point.
The evidence that MAH1 is the enzyme required for formation of wax secondary alcohols and ketones in the stem wax of wild-type Arabidopsis sheds new light on other gene products that had previously been implied in late steps of the decarbonylation pathway. Most notably, cer20 stems had been found to show lower levels of secondary alcohols and ketones than the wild type (Rashotte et al., 2001;Jenks et al., 2002), thus resembling mah1 stem wax. Based on this phenotype, it had been suggested that CER20 might be an enzyme responsible for the midchain oxidation of alkanes. However, our work on MAH1 indicates that the involvement of other enzymes in the process is unlikely. CER20 and MAH1 are different genes located on Arabidopsis chromosomes 5 and 1, respectively, and therefore CER20 must have a nonenzymatic function affecting secondary alcohol and ketone accumulation. It may be speculated that CER20 codes for a protein associated with MAH1 (e.g. through redox coupling), a protein involved in wax secretion, or a regulator of the decarbonylation pathway. In this context, it might be interesting to note that the cer20 stem wax contained alkane amounts similar to the wild type, thus differing from the mah1 phenotype with its increased alkane concentration.
The varying total amounts of decarbonylation pathway products (alkanes, secondary alcohols, and ketones together) in mah1 mutants suggest that some form of feedback inhibition is occurring. Severe losses of ketone and secondary alcohol products were associated with significant increases in alkane amounts in all mutant lines, but only within a range of 11 to 14 mg/ cm 2 . Further reductions in the ketone and secondary alcohol levels resulted in decreases in total wax loads rather than more alkane accumulation. In all of these cases, other metabolites upstream from alkane formation in the wax pathways did not change in similar proportion to changes of ketone, secondary alcohol, and total wax loads. Overall, these results indicate that accumulation of alkanes to a characteristic threshold level prompts the down-regulation of upstream reactions in wax biosynthetic pathways.

Loss of MAH1 Activity and Its Secondary Alcohol and Ketone Products Does Not Affect Epicuticular Wax Crystals on Arabidopsis Stems
An interesting finding of this study is that all the mah1 mutant lines had glaucous inflorescence stems indistinguishable from the wild type. Even the mah1-1 allele, with the most severely reduced MAH1 transcript levels leading to a complete absence of secondary alcohols and ketones in the stem wax, showed visually normal surfaces. This result explains why previous visual screens for wax mutants with glossy stems had not yielded any candidate lines with defects in the MAH1 gene.
The glaucous appearance of mah1 stems prompted us to further investigate the epicuticular wax crystals on these mutant cuticles with SEM. The stem surfaces of all three mutant lines were covered with similar numbers of wax crystals as the wild type and the crystal shapes were also indistinguishable from those on the wild type. These findings are noteworthy because alkanes, secondary alcohols, and ketones had all been implicated in the formation of wax crystals on Arabidopsis stem surfaces (Rashotte and Feldmann, 1998;Jetter et al., 2006) and changes in the amounts of these compounds in the (total) wax mixture might therefore be expected to affect the appearance of the epicuticular wax crystals. To understand why the surface structures instead appeared unaltered in the mutants, the exact composition of wild-type and mutant crystals will have to be assessed directly.
Final Steps of the Decarbonylation Pathway Are Localized in the ER of Epidermal Pavement Cells of the Arabidopsis Inflorescence Stem We found that MAH1 is expressed most strongly in the inflorescence stems, petioles, and siliques, but not weakly in the rosette and cauline leaves, seeds, and roots of Arabidopsis. These results are in agreement with published predictions for gene expression (Zimmermann et al., 2004). MAH1 expression was found to be restricted to the epidermal layer of stems, again confirming previous predictions based on microarray data (Suh et al., 2005). Finally, MAH1 gene activity was highest in the actively growing parts of the inflorescence stems in the same regions where rapid cuticle formation occurs. Chemical analyses had shown that this is also the part of the plant surface where secondary alcohols and ketones accumulate most rapidly (Suh et al., 2005), whereas they are almost absent from the leaf wax (Jenks et al., 2002). Based on the perfect correlation between expression patterns and metabolite occurrence, it can be concluded that MAH1 functions mainly, or even exclusively, as a hydroxylase in the decarbonylation pathway of stem wax biosynthesis.
MAH1 expression was limited to the pavement cells of the stem epidermis, whereas it was absent from guard cells (and trichomes). Based on this finding, it can be expected that the wax composition of these types of epidermal cells will differ at least in the amounts of secondary alcohols and ketones. It is currently not known whether the different types of Arabidopsis stem epidermal cells are also autonomous in the expression of other genes involved in cuticle formation, possibly causing further differences in their surface compositions. In this context, it is interesting to note that guard cells lacked the epicuticular wax crystals typical for the pavement cells of wild-type and mah1 mutant stems (data not shown). Hence, both epidermal cell types differ in surface composition and structure, creating a heterogeneous surface patchwork that will, in turn, cause locally varying properties and affect the biological functions of the Arabidopsis stem surface.
On a subcellular level, MAH1 was confined to the ER of stem epidermal cells. This localization of the enzyme implies that the final steps of the decarbonylation pathway occur in this compartment and that alkanes, secondary alcohols, and ketones are likely present at substantial concentrations in the ER membranes during biosynthesis. The current results thus define the subcellular compartment where all the major cuticular wax components are being generated, including intermediates and end products. It follows that the wax metabolites must be picked up in the ER for transport to the plasma membrane from which ATP-binding cassette transporters are thought to export them toward the cuticle (Pighin et al., 2004;Bird et al., 2007). Even though intracellular trafficking of wax molecules is crucial for successful cuticle formation, and hence for plant survival, nothing is known about the mechanisms involved. The current data at least allow us to pinpoint the starting location of wax trafficking within the cell and might consequently help future studies into its mechanism. CONCLUSION We have identified and characterized the Arabidopsis cytochrome P450 enzyme CYP96A15 and found it to function as a unique midchain alkane hydroxylase (MAH1). The enzyme was demonstrated to be involved in two consecutive steps of the decarbonylation pathway of cuticular wax biosynthesis, catalyzing the hydroxylation of alkanes to secondary alcohols and possibly also to ketones. Interestingly, neither reduced concentrations nor a total loss of secondary alcohols and ketones in mah1 mutant lines affected the appearance of epicuticular wax crystals on the stems. MAH1 was expressed predominantly in epidermal pavement cells (but not guard cells) of inflorescence stems, and the MAH1 enzyme was localized to the ER of the epidermis cells. Because all wax biosynthetic enzymes identified and studied to date have been localized to the ER, it seems likely that the entire process of wax biosynthesis is confined to a single subcellular compartment. This hypothesis leads to the prediction that subcellular trafficking must start with all the wax products in the ER and transport mechanisms must be operating that shunt the metabolites to the plasma membrane and on to the cuticle.

Plant Material
The T-DNA insertional mutant line mah1-1 (flanking sequence tag no. 427D09; Samson et al., 2002) was obtained from the Institut National de la Recherche Agronomique. T-DNA insertional mutant lines mah1-2 (SALK_049943) and mah1-3 (SALK_133155; Alonso et al., 2003) were obtained from the Arabidopsis Biological Resource Center. Seeds from all lines were initially planted and screened for allele homozygosity using genomic DNA extraction from leaves (Berendzen et al., 2005) and touchdown PCR (Don et al., 1991). Primers for screening and sequencing were designed with the aid of sequence-indexed Arabidopsis (Arabidopsis thaliana) T-DNA insertion data supplied by the Salk Institute for Genomic Analysis Web sites (http:// signal.salk.edu/cgi-bin/tdnaexpress and http://signal.salk.edu/isects.html). Locations of reported insertion sites were confirmed by direct sequencing of flanking PCR products (utilizing a left-border T-DNA-specific forward primer in conjunction with a gene-specific reverse primer). Seeds from the positively identified homozygous lines were used to grow plants for all subsequent experiments.
Seeds were spread upon Arabidopsis agar plates (Somerville and Ogren, 1982)

Wax Extraction and Chemical Characterization
Leaves or stems were harvested from plants 4 to 7 weeks after plating. Total cuticular wax mixtures were extracted by immersing whole organs twice for 30 s into chloroform (CHCl 3 ). The two solutions were combined, n-tetracosane (C 24 alkane) was added as an internal standard, and the solvent was completely evaporated under vacuum. In TLC analyses, approximately 2 mg of wax were separated on silica gel with chloroform mobile phase using the sandwich technique (Tantisewie et al., 1969) and visualized by staining with primuline and UV light.
For GC analyses, samples were resuspended in approximately 300 mL of CHCl 3 , transferred to a GC autosampler vial, dried under nitrogen, and derivatized with 10 mL of N,O-bis(trimethylsilyl) trifluoroacetamide (Sigma) and 10 mL pyridine (Fluka) for 60 min at 70°C. Wax composition was analyzed using a capillary GC (5890 N; Agilent; column 30-m HP-1, 0.32-mm i.d., d f 5 0.1 mm; Agilent) with He carrier gas inlet pressure programmed for constant flow of 1.4 mL min 21 with a MS detector (5973 N; Agilent). GC was carried out with temperature-programmed on-column injection and oven temperature set at 50°C for 2 min, raised by 40°C min 21 to 200°C, held for 2 min at 200°C, raised by 3°C min 21 to 320°C, and held for 30 min at 320°C. Individual wax components were identified by comparing their mass spectra with those of authentic standards and literature data. Quantitative analysis of wax mixtures was carried out using capillary GC with flame ionization detector under the same conditions as above, but with H 2 carrier gas inlet pressure regulated for constant flow of 2 mL min 21 .
Wax loads were determined by comparing GC-flame ionization detector peak areas against internal standard and dividing by the surface area extracted for the corresponding sample. Total leaf surface areas were calculated with ImageJ software (Abramoff et al., 2004) by measuring the apparent leaf areas in digital photographs and multiplying by 2. Stem surface areas were calculated by measuring the projected two-dimensional stem areas in photographs and multiplying by p.

Semiquantitative RT-PCR
Total RNA was extracted from stems, roots, buds, and leaves of 4-week-old plants grown in soil. Tissues were ground up thoroughly in 200 mL of RNA later (Ambion/Applied Biosystems), using a tube and motorized pestle. The lysate was processed immediately by a Qiagen RNeasy plant mini kit and then used as template for RT by Moloney murine leukemia virus reverse transcriptase (New England Biolabs). cDNAs used for semiquantitative RT-PCR were normalized based on the intensity of PCR-amplified ACTIN2 fragments generated by the primers 5#-CCAGAAGGATGCATATGTTGGTGA-3# and 5#-GAGGAGCCTCGGTAAGAAGA-3# (yielding an approximately 250-bp fragment). MAH1 gene-specific primers 5#-AACTTTGTGCCCGCTTGGAA-3# and 5#-ACAGCTTTGGCCACTGTCAA-3# (generating a 434-bp fragment) were used in reactions conducted simultaneously under identical conditions as ACTIN2 controls. Because these primers amplify a downstream region of the gene stretching from 1726 to 11,160 and because the mah1-1 T-DNA insert had been proposed to be localized approximately at position 1692, we used an additional set of primers, 5#-ATGGCGATGCTAGGTTTTTACGTA-3# and 5#-TTCGCCAATATCCGCAGCTT-3# ranging from 11 to 1638 to determine mah1-1 steady-state transcript levels.

Cryo-SEM
Segments from the apical 4 to 6 cm of stems were mounted onto cryo-SEM stubs using graphite paste and plunged into liquid nitrogen. Frozen stems were transferred into an Emitech K1250 cryosystem and water sublimed for 10 min at 2110°C. Samples were viewed with a Hitachi S4700 field emission SEM (Nissei Sangyo America) using an accelerating voltage of 1.5 kV and a working distance of 12 mm.

Cloning of MAH1 and Construction of Vectors for Expression of MAH1:GFP and MAH1 Promoter:GUS Fusions
With aid from The Arabidopsis Information Resource SeqViewer (http:// www.arabidopsis.org), 3,126 bp of Arabidopsis chromosome 1 surrounding and including the At1g57750 coding sequence (GenBank accession no. AY090941) was PCR amplified (Phusion DNA pol; Finnzymes/NEB) using isolated genomic Col-0 DNA as a template with primers 5#-GCCGTTGGAT-GATGAATATGCACGACT-3# and 5#-TTACAAAGATTCGAGGACCGGGCA-3#. The resulting product included the proposed 5#-UTR, 3#-UTR, the entire open reading frame (which lacks introns) of MAH1 (CYP96A15), and the additional nucleotide sequence stretching 1,310 bp upstream of the 5#-UTR. This genomic fragment was cloned into pGEM-EZ (Promega), then sequenced and found to perfectly match The Institute for Genomic Research sequence published on The Arabidopsis Information Resource. All subsequent constructs were made by using this clone as template.
Resultant PCR products were placed into vector pDONR221 (Invitrogen) by performing a BP reaction (attB 3 attP / attL 1 attR) using Integrase and Integration Host Factor in the form of BP Clonase II (Invitrogen). Integrated constructs were then inserted in frame into binary vectors for fusion with either GFP (pGWB4 or pGWB5) or GUS (pGWB3) by performing an attL 3 attR / attB 1 attP reaction (LR reaction with Integrase, Integration Host Factor, and excisionase in the form of LR Clonase II from Invitrogen) between the entry clone (pDONR221) and the appropriate pGWB acceptor vector. Finished binary vector constructs were sequenced to confirm that the sequence was maintained.

GUS and GFP Visualization
Expression of GUS in transgenic pGWB3-MAH1P:GUS wild-type (Col-0) plants was assayed by submerging whole-plant tissues in acetone under vacuum for 30 min and then washing in buffer composed of 0.1% Triton X-100, 0.25 mM K 4 Fe(CN) 6 , 3H 2 O, 0.25 mM K 3 Fe(CN) 6 , 3H 2 O, and 50 mM phosphate buffer, pH 7.0, three times for 5 min each. Washed tissues were subsequently stained by incubation in this same buffer with the addition of 1 mM 5-bromo-4-chloro-3-indolyl-b-D-glucuronide under vacuum for 30 min, and then at ambient pressure (with gentle agitation) at 37°C for 16 h. Afterward, stained tissues were placed in 70% ethanol and imaged using a Stemi 2000-C dissecting microscope (Zeiss) mounted with a QCAM digital camera (QImaging) and Openlab 4.01 software (Improvision).
Arabidopsis plants were immersed for 10 to 30 min either in FM4-64 (8.2 mM) solution (Vida and Emr, 1995) for plasma membrane staining or in rhodamine B hexyl ester solution (1.6 mM) for ER staining. Autofluorescence, as well as GFP, rhodamine B hexyl ester, and FM4-64 fluorescence was examined with a Zeiss LSM 5 Pascal confocal laser-scanning microscope. The excitation wavelength for GFP was 488 nm with the emission filter set at 505 to 530 nm; autofluorescence was detected using an emission filter set at 600 to 650 nm. The excitation wavelength for rhodamine B hexyl ester was 568 nm with the emission filter set at 600 to 650 nm. The excitation wavelength for FM4-64 was 514 nm with the emission filter set at 600 to 650 nm.

Statistics
Comparisons of wild-type and mutant wax data utilizing mixed-effect univariate ANOVA (a 5 0.05; treatment as a fixed effect, batch nested within treatment as a random effect), Dunnett's t, and Tukey-Kramer posthoc tests (a 5 0.05; using harmonic means in cases of unequal n) were conducted with SPSS 11.0 software. When necessary to meet assumptions of normality or equal variance, datasets were transformed into normal scores using Tukey's formula (r 2 1/3)/(w 1 1/3), where r is the rank and w is the sum of the case weights (Tukey, 1962). For each analysis, type II error statistics (b) were also calculated. Values for b range from 0 to 1 and correspond directly to the chance of committing type II error in an ANOVA F test. The power of an ANOVA F test can be calculated as 1 2 b, yielding the probability that the F test will detect the differences between groups equal to those implied by sample differences (Sokal and Rohlf, 1995).

Supplemental Data
The following materials are available in the online version of this article.