Distinct cell-specific expression of homospermidine synthase involved in pyrrolizidine alkaloid biosynthesis in three species of the Boraginales. 1

Homospermidine synthase (HSS) is the first specific enzyme in pyrrolizidine alkaloid (PA) biosynthesis, a pathway involved in the plant’s chemical defense. HSS has been shown to be recruited repeatedly by duplication of a gene involved in primary metabolism. Within the lineage of the Boraginales, only one gene duplication event gave rise to HSS. Here, we demonstrate that the tissue-specific expression of HSS in three boraginaceous species, Heliotropium indicum , Symphytum officinale , and Cynoglossum officinale , is unique with respect to plant organ, tissue, and cell-type. Within H. indicum , HSS is expressed exclusively in non-specialized cells of the lower epidermis of young leaves and shoots. In S. officinale , HSS expression has been detected in the cells of the root endodermis and in leaves directly underneath developing inflorescences. In young roots of C. officinale , HSS is detected only in cells of the endodermis, but in a later developmental stage, additionally in the pericycle. The individual expression patterns are compared to those within the Senecioneae lineage (Asteraceae), where HSS expression is reproducibly found in specific cells of the endodermis and the adjacent cortex parenchyma of the roots. The individual expression patterns within the Boraginales species are discussed as being a requirement for the successful recruitment of HSS after gene duplication. The diversity of HSS expression within this lineage adds a further facet to the already diverse patterns of expression that have been observed for HSS in other PA-producing plant lineages, making this PA-specific enzyme one of the most diverse expressed proteins described in the literature.


INTRODUCTION
Pathways involved in the biosynthesis of alkaloids are characterized by an efficient coordination of a large number of enzymes. Molecular approaches have shown that the cellular localization of the alkaloid pathways is remarkably diverse and complex, often including the translocation of intermediates between multiple cell types (Facchini and St-Pierre, 2005;Ziegler and Facchini, 2008). The alkaloids are accumulated in cell types and tissues that are in most cases distinct from those that are involved in alkaloid biosynthesis. In the opium poppy (Papaver somniferum), alkaloids are synthesized in sieve elements of the phloem and accumulated in laticifers (Bird et al., 2003). Moreover vindoline, a monoterpenoid indole alkaloid of Catharanthus roseus, is accumulated in laticifers and also in idioblasts, but biosynthesis involves at least three different cell types (St-Pierre et al., 1999;Burlat et al., 2004). Nicotine and tropane alkaloids, such as hyoscyamine, are synthesized in roots of certain solanaceous species in at least two cell types and are translocated into the shoots (Nakajima and Hashimoto, 1999;Suzuki et al., 1999a). The same is true for the pyrrolizidine alkaloids (PAs), which Senecio species produce exclusively in the roots before their transport to the shoots via the phloem (Hartmann et al., 1989;Witte et al., 1990).
For PAs that are constitutively produced by the plant as part of its chemical defense against herbivores, the only unequivocally identified enzyme involved in their biosynthesis is homospermidine synthase (HSS). HSS catalyzes the formation of the first pathway-specific intermediate, homospermidine, which is incorporated exclusively into the necine base, the characteristic bicyclic structure of PAs (Böttcher et al., 1993;Böttcher et al., 1994). HSS has been shown to be recruited by gene duplication from deoxyhypusine synthase (DHS), an enzyme involved in the posttranslational activation of the eukaryotic initiation factor 5A (eIF5A). Phylogenetic analyses suggest that this recruitment occurred at least five times independently in the various angiosperm lineages that are described to produce PAs, i.e., once within protein-modifying activity while retaining all other properties, including the former "side activity" of the ancestor to catalyze the formation of homospermidine (Ober et al., 2003b). Thus, in the case of HSS evolution, the loss of DHS activity, i.e., the ability to bind the protein substrate, resulted in an enzyme with a new function.
Several examples in the literature show that partitioning of the primary gene function between two daughter genes is common with respect to gene expression (Force et al., 1999;Papp et al., 2003;Zhang, 2003;Bridgham et al., 2008). Comparative analyses of HSS and DHS expression in various PA-producing species have shown that the expression pattern of each enzyme differs, suggesting that, with respect to gene regulation, the gene duplication event was followed by subfunctionalization (Moll et al., 2002;Anke et al., 2004;Anke et al., 2008;Ober and Kaltenegger, 2009).
Whereas DHS is expressed uniformly in all analyzed tissues, irrespective of the species, HSS shows highly specific expression patterns. In Senecio vernalis (Asteraceae, Senecioneae), HSS is expressed in groups of specific cells of the endodermis and the adjacent cortex parenchyma that are opposite to the phloem (Moll et al., 2002). The vicinity to the phloem suggests that these cells express not only HSS, but most probably also the complete PA pathway, as the phloem is the tissue by which the PAs are translocated from the roots to the shoots. In contrast, HSS of Eupatorium cannabinum (Asteraceae, Eupatorieae) is expressed in all cells of the root cortex parenchyma, but not in the cells of the endodermis (Anke et al., 2004). This individual expression pattern of the key enzyme of an identical pathway in these two lineages has been interpreted to support the polyphyletic origin of PA biosynthesis within the Asteraceae (Anke et al., 2004). Furthermore, in E. cannabinum, HSS expression has been shown to be dependent on the developmental stage of the plant. HSS expression is highest as long as the plant is growing. As soon as the flower buds open, HSS expression diminishes and is not detectable when the fruits are produced (Anke et al., 2004). This mechanism ensures relatively constant PA levels per biomass, as alkaloids are only produced when the plant is producing biomass by its growth. A similar developmental expression pattern has been observed in the orchid Phalaenopsis, a PA-producing member of the monocots. PA biosynthesis and HSS expression have been detected in mitotically active cells of the apical meristem of aerial roots, most probably providing PAs for the vegetative tissues of the plant, e.g., young leaves and the shoot meristem. As soon as the orchid develops an inflorescence, HSS expression and PA biosynthesis is additionally detected in young flower buds, providing accessory alkaloids for the developing reproductive tissues. Although HSS expression diminishes with increasing age of the flower buds and is not detectable when the flower opens, PA levels are still high in the open flower, at about 1 mg/g fresh weight (Anke et al., 2008).
With respect to the Boraginales, we have studied three species in detail, i.e., Heliotropium indicum (indian heliotrope, perennial), Symphytum officinale (common comfrey, perennial), and Cynoglossum officinale (houndstongue, biennial). Based on the molecular analyses of this taxon, H. indicum is grouped in the family Heliotropiaceae, and the other two species in the Boraginaceae (Gottschling et al., 2001). Phylogenetic analyses of HSS-and DHS-coding cDNA sequences suggest that the gene duplication resulting in HSS occurred early in this lineage, well before the separation of the Heliotropiaceae from the Boraginaceae (Reimann et al., 2004).
Tracer feeding experiments have shown that, first, the core pathway for PAs from putrescine and spermidine via homospermidine is the same as the pathway in Senecio species and, second, despite the common origin of HSS within the Boraginales lineage, different species differ with respect to their site of PA biosynthesis: PAs are produced exclusively in the shoots of H. indicum (Heliotropiaceae), exclusively in the roots of S. officinale (Boraginaceae), and in shoots and roots of C. officinale (Boraginaceae) (Van Dam et al., 1995;Frölich et al., 2007). To analyze these variations at the cellular level, we have studied the expression patterns of HSS and DHS, which share a common ancestor. Our results show that despite the identical function of HSS in PA biosynthesis, the HSS expression pattern is distinct in all three analyzed species. This observation is discussed with respect to the individual selection pressures that acted after gene duplication during recruitment of HSS for PA biosynthesis in the lineages of the three analyzed species. A comparative analysis of a further member of the Senecioneae lineage (Asteraceae) supports the view that this variability of expression patterns is a unique characteristic of the Boraginales lineage and not a general concept valid for all PA-producing lineages.

RESULTS
To analyze hss and dhs gene expression, we used semiquantitative reverse transcription (RT)-polymerase chain reaction (PCR) to localize the transcript and, in the case of HSS, immunoblots to localize the respective protein. RT-PCR was performed with primers highly specific for the cDNAs of HSS and DHS to test for the presence or absence of the respective transcripts in the various plant tissues and to allow, at best, a rough estimate of transcript levels. Subsequently, the various tissues were tested by immunoblots for the expression of the HSS protein, before those tissues with the highest HSS expression were used for immunolocalization experiments.

Tissue-Specific Expression of HSS in H. indicum
In H. indicum, transcripts of the hss gene were detectable in all above-ground tissues tested, including young and older stems, leaves, and the flower from the bud to fruit stage (Fig. 1A). This result confirmed previous tracer feeding studies that identified the shoot of H. indicum as the site of PA biosynthesis. Roots were shown to be unable to incorporate tracer into the alkaloid structures, but to synthesize low levels of homospermidine (Frölich et al., 2007). For the tissues taken from the inflorescence, we were unable to exclude a positive signal because of adhering tissue of the inflorescence axis. Therefore, in an additional experiment cDNA was prepared separately from the inflorescence axis, the open flowers, and the fruits.
High levels of hss transcript were detectable in the axis and the flowers, but almost no transcript was found in the fruits (Fig. 1B). In contrast, in both experiments, transcript of dhs was detectable in all tested tissues. The presence of the dhs transcript in all tested tissues raised the question as to whether the ability of the H.
indicum roots to synthesize homospermidine that was described by (Frölich et al., 2007) was attributable to low levels of HSS in the roots as suggested by the faint signal detected by the PCR containing the cDNA of roots or to the known side activity of DHS (Ober and Hartmann, 1999a;Ober et al., 2003a;Reimann et al., 2004).
Immunoblots showed a more specific pattern of HSS expression ( Fig. 1C left panel).
Only protein extracts of young leaves revealed an intense signal, a weak signal being detectable in the extracts of young stems and of flower buds. No signal was detectable in roots, older stems, the open flower, the fruits, or older leaves. These data suggest that HSS expression is dependent on the developmental stage of the leaves. This was tested in an additional immunoblot with extracts of leaves covering all developmental stages (categorized according to leaf size). The strongest signal was detectable in young leaves, steadily decreasing with the increasing size of the leaves ( Fig. 1C right panel). This expression pattern of HSS correlates with the tracer feeding experiments by Frölich (2007) showing that PA biosynthesis is most efficient in young leaves. Of note, the epidermis peeled from young stems also showed a distinct signal for HSS expression.

Immunolocalization of HSS in Shoots of H. indicum
The immunoblots suggested that HSS was expressed mainly in young leaves and to a lesser extent in the epidermis of young shoots and flower buds. Therefore, we decided to use young leaves and young stems of H. indicum for immunolocalization studies. Figure   showed no signal (Fig. 3B). We speculated that the stage of plant development might influence HSS expression. Therefore, we analyzed young terminal leaves that differed in their position on the plant. Type I leaves lay directly beyond a terminal inflorescence with fully opened flowers, type II leaves occurred at the same position, but with flower buds still closed, and type III leaves were on stems without any inflorescences. The extract of the young root served as a positive control. The immuno-blot confirmed the expression of HSS in leaves of type II, whereas leaves of type I and III were devoid of any label. This result supports the idea that HSS expression in the shoot of S. officinale depends on the developmental stage of the leaves.

Immunolocalization of HSS in Roots of S. officinale
Young white roots that developed during the same year were cut and embedded in resin, before being labeled with HSS-specific antibody. Figure

Tissue-Specific Expression of HSS in C. officinale
C. officinale is the only plant species for which PA biosynthesis was detected by tracer feeding experiments in both shoot and root (Van Dam et al., 1995). Instead, the RT-PCR experiment and the immuno-blot analysis showed that the HSS encoding transcript and HSS protein were found only within the roots, occurring only as traces, if at all, within the other tissues (Fig. 4). This observation was independent of the age of the plant, e.g., whether the samples had been taken off the rosette plant in the first year of development or off the flowering plant in the second year. Again, as in H. indicum and S. officinale, the dhs transcript was detectable in all tested tissues.

Immunolocalization of HSS in Roots of C. officinale
For immunolocalization experiments, young white roots of the rosette plant and the flowering plant in their first and second year of development, respectively, were embedded in resin. For detection with an HSS-specific antibody for C. officinale, the various zones of differentiation were analyzed, i.e. the region that lies a few millimeters behind the root tip and in which the vascular cylinder is in the procambial stage ( Fig. 5A), the region after complete differentiation (Fig. 5B), and the region that is characterized by emerging lateral roots (Fig. 5C+D). The intense green color of the fluorescein isothiocyanate (FITC)-labeled HSS was not detectable before the fate of the cells was determined in the completely differentiated root. Here, expression was restricted to the layer of the endodermis that was unequivocally identified by the presence of the Casparian strip showing yellow autofluorescence. In the region characterized by emerging lateral roots, HSS was detectable not only in the endodermis but also in the pericycle. As shown in Figure 5C, cambial cells were detectable (labeled by asterisks) between the xylem and the phloem tissue indicating that secondary growth started in this region of the root. Figure 5D shows that HSS expression is switched off in the area of an emerging side root, a phenomenon previously described for roots of S. officinale. The expression pattern of HSS in the analyzed zones of root differentiation was the same in roots taken from the rosette stage or the flowering stage of the plant in the first and second year of development, respectively (data not shown).

Comparative Localization of HSS in Roots of J. vulgaris
Despite of the monophyletic origin of HSS within the lineage of the Boraginaceae, our data showed individual expression patterns of HSS in the three analyzed species.
This observation raised the question as to whether such diversity might also be found within other lineages, for which an independent origin of HSS was reported (Reimann et al., 2004). Therefore, we decided to compare the tissue-specific expression pattern of HSS in two species of the Senecioneae lineage, one of the two lineages within the Asteraceae for which an independent origin of HSS was proposed (Reimann et al., 2004). We selected the species Jacobaea vulgaris (syn. Senecio jacobaea), which was formerly classified into the genus Senecio sect. Jacobaea. Recent molecular systematic studies have shown that species that belong to this section form a wellsupported cluster, which is only distantly related to other species usually attributed to Senecio, such as S. vernalis (Pelser et al., 2006;Pelser et al., 2007). For S. vernalis, we have been able, in a previous study, to show that HSS expression is root-specific and restricted to distinct groups of endodermis and neighboring cortex cells located opposite to the phloem (Moll et al., 2002). Figure 6A shows an immunoblot confirming the root-specific expression of HSS for J. vulgaris. Figure 5E shows a cross section of a young root of J. vulgaris, HSS being expressed in the same cells as described previously for S. vernalis, i.e., only in those cells of the endodermis and the adjacent cortex parenchyma opposite to the phloem tissue. The phloem cells are located between the strands of the tetrarch xylem tissue. A longitudinal section of a root of the same age shows that the label runs parallel to the phloem tissue in those cortex cells that are closest to the central cylinder (Fig. 5F).

DISCUSSION
Our research group uses the model system of PA biosynthesis to study the evolution of a pathway in plant secondary metabolism. It is known that the first specific enzyme in PA biosynthesis, namely HSS, originated by duplication of a gene encoding an ancestral DHS, an enzyme that is involved in primary metabolism and that is highly conserved within eukaryotes (Park et al., 1997;Ober and Hartmann, 1999a, b). This duplication event occurred in an almost identical manner several times independently in various plant lineages during angiosperm evolution and, in particular, once early in the Boraginales lineage (Reimann et al., 2004). In consideration of the polyphyletic origin of HSS, the resulting PA-producing pathways are surprisingly identical with respect to the biosynthetic sequences and the structures of the resulting products (Frölich et al., 2007). Although HSS is the only identified and characterized enzyme of PA biosynthesis, convergent evolution is likely to have formed these pathways because of similar selection pressures, i.e., the defense against herbivores.
As the ability to synthesize PAs evolved independently in various angiosperm lineages, PA biosynthesis offers an instructive system for studying comparatively the repeated evolution of this pathway. Here, we have compared the cell-specific expression patterns of HSS in three species belonging to the Boraginales and show that, in all three species, HSS shows a unique and specific expression pattern (Fig.   7). In contrast, DHS transcription is detectable in all analyzed tissues, an observation that has been described previously for other tested species (Moll et al., 2002;Anke et al., 2004;Nurhayati and Ober, 2005;Anke et al., 2008;Ober and Kaltenegger, 2009). Plant secondary metabolism is well-known for its diversity concerning the huge array of chemical structures that are produced. Our studies show that secondary metabolism is also diverse with respect to its regulation. Although HSS catalyzes an identical reaction in PA biosynthesis, it is one of the most diverse enzymes described so far for plant metabolism with respect to its expression pattern. Furthermore, we have demonstrated that, despite the monophyletic origin of an enzyme, its expression pattern can be highly specific in different species and thus does not provide a robust argument for conclusions about the evolutionary origin of this specific gene (Anke et al., 2004). This has previously proposed for the enzymes that are involved in benzylisoquinoline alkaloid biosynthesis; these enzymes have been localized to different tissues in the opium poppy and T. flavum, and a monophyletic origin early in the evolution of the Angiosperms has been proposed for them (Liscombe et al., 2005;Samanani et al., 2005). For the evolution of HSS within the Convolvulaceae, we have been able to show that the duplication of the dhs gene was followed by a time of strong purifying selection on both copies, most likely because of protein-protein interactions involved in the mechanism of DHS and HSS (Kaltenegger and Ober, unpublished). These involve the binding of the eIF5A precursor protein as the substrate of the DHS and the interaction of subunits, as both enzymes are homotetramers. Mutations within one of the gene copies would have detrimental effects on the essential function of DHS, resulting in a strong purifying selection on both copies. Subfunctionalization with respect to tissue-specific expression has been postulated to precede the diversification of the structural gene. Within the Convolvulaceae, positive Darwinian selection has been detected to have shaped one of the copies to become a HSS later in evolution, most probably after subfunctionalization of the gene copy regulation (Kaltenegger and Ober, unpublished). A similar scenario is likely within the Boraginales, suggesting that speciation events might have preceded the lineage-specific modification of the regulatory elements of one of the gene copies, i.e., the gene that developed to encode the HSS. The high cell specificity of HSS expression might be a tool for a straight forward approach for the identification of new enzymes and regulatory elements involved in PA biosynthesis by differential or cell-specific approaches. The knowledge of these additional elements is essential if we are to shed light on the independent evolution of the pathway in various plant lineages.

Plant Material
H. indicum plants were grown in the greenhouse at approx. 25°C and were propagated by shoot cuttings. S. officinale and C. officinale plants were grown in the institute garden, some of them being potted to allow easy sampling of the roots.

RNA Isolation and Semiquantitative RT-PCR
Samples of various plant tissues were pulverized in liquid nitrogen in a mortar before total RNA was extracted with the RNeasy Plant Mini Kit (Qiagen). For each sample, 1 µg of total RNA was used as a template for reverse transcription with an oligo(dT) 17 primer (Supplemental Table 1

Recombinant Protein Expression
To generate and to test the specificity of polyclonal antibodies against HSS of H.  Table 1). The expression vector pET23a was used only for expression of HSS and DHS of J. vulgaris.
Expression constructs were introduced into Escherichia coli BL21(DE3) and expressed as described previously (Ober and Hartmann, 1999a). E. coli cells were harvested by centrifugation, suspended in lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 20 mM imidazole), and broken up by sonication. His-tagged proteins were purified with nickel-nitrilotriacetic acid-agarose (Qiagen, Hilden, Germany) according to the manufacturer's instructions. A cobalt-containing resin (Talon Metal Affinity Resin, Clontech) was used only in the case of the HSS of S. officinale.

Polyclonal Antibody Preparation and Affinity Purification
Recombinant HSS of H. indicum, S. officinale, C. officinale, and J. vulgaris were used to raise polyclonal serum in rabbits (Bioscience, Göttingen, Germany). For affinity purification, 1-5 mg of purified recombinant HSS was coupled to CNBr-activated Sepharose 4B (GE Healthcare). The resulting matrix was used to purify the polyclonal antibodies as described previously (Anke et al., 2004). The antibodies used in this study against HSS of H. indicum, S. officinale, C. officinale, and J. vulgaris were tested in immunoblots for their cross-reactivity to DHS of the same species. Two gels for SDS-polyacrylamide gel electrophoresis (SDS-PAGE) were prepared, each with identical amounts of purified recombinant HSS and DHS of H.
indicum, S. officinale, C. officinale, and J. vulgaris respectively. After the blotting step, one membrane was used for protein staining with Indian ink to ensure equal loading of HSS and DHS protein samples, and the other was developed as an immunoblot with the antibody against HSS of the respective species. In all analyses, the antibody against HSS was also able to bind to the DHS of the same species, but with a 5-to 10-fold (H. indicum), 5-fold (S. officinale), 10-fold (C. officinale), and 5-fold (J. vulgaris) lower affinity than to HSS.

Protein Gel-Blot Analysis
Various tissues were sampled and immediately frozen in liquid nitrogen and stored in -80°C until use. For protein extraction, the samples were pulverized in a mortar in the presence of liquid nitrogen and extracted twice in phosphate-buffered saline supplemented with 5% (m/v) polyvinylpyrrolidone and 2.5% (w/v) sodium ascorbate.
Addition of polyvinylpyrrolidone and sodium ascorbate reduced protein precipitation during sample preparation by polyphenols present in large amounts in tissues of Boraginaceae plants. Protein separation by SDS-PAGE, semidry blotting, and immunodetection was carried out as described in Anke et al. (2008). To estimate the mass of protein fragments, the protein molecular weight marker 14.4 to 116 kDa (Fermentas) was used.

Immunocytochemical Localization
Small segments of various plant organs (0.5-1.0 cm) were immersed for 2 h under reduced pressure in ice-cold buffered fixative according to Anke et al. (2004). To avoid protein precipitation by polyphenols, sodium ascorbate (2.5%, m/v) was added as an antioxidant to the fixative just before the samples were added. Following dehydration in a graded ethanol series, the tissues were embedded in Technovit 7100 resin (Hereaus-Kulzer, Hanau, Germany) for light microscopy analyses or in Unicryl resin (Plano, Wetzlar, Germany) for transmission electron microscopy, according to the manufacturer's instructions. Sections of 3 to 4 µm in thickness were cut by using a microtome (HM3555S, Microm) and mounted on glass slides coated with Teflon (Roth, Karlsruhe, Germany) or on adhesion microscope slides (SuperFrost, Menzel, Braunschweig, Germany). Blocking of the sections with protein, incubation with antibodies, and detection was performed as described previously (Anke et al., 2008) with the modification that, for the UV detection step, a goat-anti rabbit AlexaFluor488 antibody (1:100, Molecular Probes, Invitrogen) was used as secondary antibody. The specificity of the label was confirmed by the incubation of successive sections with HSS-specific antibody of the respective plant to which increasing amounts of soluble HSS were added. These preincubations resulted in a decreasing intensity of the label. Preincubations with soluble DHS and bovine serum albumin (BSA) had no effect on labeling intensity, excluding any cross-detection of DHS and any non-specific protein labeling (Supplemental Fig. 1-5).

Detection of Unspecific Proteinases in Protein Preparations for Immunoblots
In some of our immunoblots, mainly of root tissue, incubation with HSS-specific antibody resulted not only in the detection of the HSS signal of approx. 44 kDa, but also in the detection of two additional bands of 18 and 26 kDa (Fig. 3B, 4B, 6A). We speculated that these two bands might be the result of cleavage of the HSS protein, e.g., by proteolytic activity from contamination of residual soil adhering to the fine young roots harvested for the analyses or by an endogenous regulatory process within the root. The latter idea resulted from the observation that, in initial experiments, we observed the cleavage mainly in samples containing older roots. To test these hypotheses, we used a root tissue culture of J. vulgaris grown under axenic conditions, divided the sample into three aliquots and extracted one aliquot according to the standard protocol. To the extraction buffer of the second aliquot, soil material was added that had been taken from the field in which the J. vulgaris plants of this study had been grown. The third aliquot was extracted under the same conditions as the second aliquot but with addition of the proteinase inhibitor phenylmethylsulfonyl fluoride (PMSF, final concentration 0.2 mM). Figure 6B shows a strong signal of HSS at 44 kDa. Cleavage of HSS was found in the second aliquot by detection of an additional 26 kDa signal. Cleavage was efficiently inhibited by the addition of PMSF in the third aliquot, supporting our hypothesis that degradation of HSS was attributable to contamination from soil material.