The single berberine bridge enzyme homolog of Physcomitrella patens is a cellobiose oxidase

The berberine bridge enzyme from the California poppy Eschscholzia californica (Ec BBE) catalyzes the oxidative cyclization of (S)‐reticuline to (S)‐scoulerine, that is, the formation of the berberine bridge in the biosynthesis of benzylisoquinoline alkaloids. Interestingly, a large number of BBE‐like genes have been identified in plants that lack alkaloid biosynthesis. This finding raised the question of the primordial role of BBE in the plant kingdom, which prompted us to investigate the closest relative of Ec BBE in Physcomitrella patens (Pp BBE1), the most basal plant harboring a BBE‐like gene. Here, we report the biochemical, structural, and in vivo characterization of Pp BBE1. Our studies revealed that Pp BBE1 is structurally and biochemically very similar to Ec BBE. In contrast to Ec BBE, we found that Pp BBE1 catalyzes the oxidation of the disaccharide cellobiose to the corresponding lactone, that is, Pp BBE1 is a cellobiose oxidase. The enzymatic reaction mechanism was characterized by a structure‐guided mutagenesis approach that enabled us to assign a catalytic role to amino acid residues in the active site of Pp BBE1. In vivo experiments revealed the highest level of PpBBE1 expression in chloronema, the earliest stage of the plant's life cycle, where carbon metabolism is strongly upregulated. It was also shown that the enzyme is secreted to the extracellular space, where it may be involved in later steps of cellulose degradation, thereby allowing the moss to make use of cellulose for energy production. Overall, our results suggest that the primordial role of BBE‐like enzymes in plants revolved around primary metabolic reactions in carbohydrate utilization. Database Structural data are available in the PDB under the accession numbers 6EO4 and 6EO5.


Introduction
The berberine bridge enzyme (BBE)-like protein family is a large enzyme family found in bacteria, fungi, and plants, named after its best characterized member, the BBE from Eschscholzia californica (EcBBE, EC 1.21.3. 3). In Californian poppy (E. californica), BBE participates in the biosynthesis of benzylisoquinoline alkaloids, where it catalyzes the oxidative formation of the berberine bridge resulting in the conversion of (S)-reticuline to (S)-scoulerine [1]. Whole-genome sequencing efforts have revealed the occurrence of many BBE-like genes in virtually all plants even those that lack alkaloid biosynthesis [2]. In the model plant Arabidopsis thaliana, for example, 27 genes were identified to encode BBE-like enzymes, with an even higher number found in Citrus clementina (41 genes), Glycine max (43 genes), and Populus trichocarpa (65 genes). Despite the ubiquitous and abundant presence of BBE-like proteins in the plant kingdom, the biochemical and physiological function of the enzymes remain largely unknown. Based on multiple sequence alignments, the 27 BBE homologs of A. thaliana were classified into seven subfamilies [3]. Recent studies have successfully assigned a function to two of these subfamilies, demonstrating oxidation of indole cyanohydrins and monolignols to the corresponding keto and aldehyde products, respectively [3,4]. Besides, earlier reports on enzymes involved in plant defense have implicated BBE-like proteins in the oxidation of mono-and polysaccharides derived from glucose as well as from galactose [5]. A similar activity was also assigned to the pollen allergen Phl p 4 [6] and to nectarin 5, a BBE-like protein from ornamental tobacco [7].
Unlike most other plants, the genome of the moss Physcomitrella patens harbors only two genes that can be assigned to the BBE-like family [8]. Gene #1 (Pp3c12_2640V3.1) shares the highest sequence identity with fungal carbohydrate oxidases (ca. 30% identity on the amino acid sequence level) and is closely related to EcBBE and other plant enzymes from the BBE-like family (25-28% identity on the amino acid sequence level). In contrast, gene #2 (Pp3c2_12420V3.1) has a much lower similarity to EcBBE (< 20% sequence identity) and is more closely related to the ecdysteroid-22-oxidase from Metarhizium rileyi (29.3%; Uniprot: I0J0L0).
Because of the role of P. patens as a paradigm for the conquest of land by plants [9], we reasoned that unraveling the function of the moss BBE-like protein would provide insight into the evolutionary origin of this ubiquitous multigene family in plants. Toward this goal, we have produced the BBE homolog, termed PpBBE1 and encoded by gene #1 (Pp3c12_2640V3.1), in the methylotrophic yeast Komagataella phaffii and subsequently characterized the purified enzyme biochemically and structurally. In addition, expression and secretion of the wild-type protein as well as the phenotypical characteristics of a knockout strain were studied in order to reveal the in vivo function of the enzyme. Here, we report that PpBBE1 is an efficient cellobiose oxidase that is very similar to fungal BBElike oxidases, such as glucooligosaccharide oxidase (GOOX), in terms of the active site composition as well as the enzymatic reaction mechanism. Our in vitro findings are further supported by the data obtained from in vivo experiments since the highest level of expression could be detected in chloronema, an early stage of the plant's life cycle, where carbohydrate metabolism is strongly upregulated [10,11]. Thus, our results suggest that the diversity of enzymatic reactions found in the BBE-like protein family in the plant kingdom originated from this 'primordial' carbohydrate oxidase.

Results
Gene organization, expression, and secretion of Physcomitrella patens BBEs According to the latest release of the genome V3.3 available on cosmoss.org, PpBBE1 is located on chromosome 12 and represented by the gene model Pp3c12_2640V3.1 encoding a protein of 501 amino acids. The full-length gene comprises 3243 bp, of which 2808 bp belong to the coding sequence. The latter consists of a 5 0 -UTR of 498 bp, a transcript of 1506 bp and a 3 0 -UTR of 804 bp, with the transcript being formed by three exons (99, 256, and 1151 bp for exon 1, 2, and 3, respectively) and two introns. PpBBE3 (note that PpBBE2 was identified as a pseudogene) is located on chromosome 2 and represented by Pp3c2_12420V3.1, consisting of two exons and an intron, encoding a protein of 586 amino acids (Fig. 1A).
The expression of PpBBE1 throughout the whole life cycle and upon different growth conditions was analyzed in silico using publicly available transcriptomic data based on microarray experiments [12,13]. PpBBE1 is highly expressed in protoplasts and chloronema cells, which both harbor large and numerous chloroplasts. Analysis of the caulonema, the fastgrowing cell type of protonema from which buds arise and which is characterized by fewer chloroplasts and slimmer cells with oblique cell walls, in contrast, did not reveal any expression. In the gametophore, expression of PpBBE1 was restricted to the green parts, whereas in rhizoids, no expression was detected. During sporophyte development, PpBBE1 expression was only detectable in unfertilized archegonia, the female sexual organs, and the mature spore capsule while PpBBE3 expression was restricted to the late stages of sporophyte development ( [12]; Fig. 1B). On dedifferentiation of leaves to chloronema as well as after incubation in darkness, PpBBE1 expression was upregulated [13].
PpBBE1 is about 54 kDa and harbors a predicted N-terminal signal peptide  according to SignalP [14] as well as an internal FAD-binding domain (78-216, PF01565) and a C-terminal BBE domain (457-499, PF08031) according to the PFAM protein domain annotation [15]. Furthermore, the PpBBE1 was identified in the supernatant of P. patens liquid cultures by mass spectrometry whereas PpBBE3 was not detected.

Production and purification of PpBBE1 and basic characterization
The gene of the P. patens BBE-like protein encoded by locus Pp3c12_2640V3.1 (gene #1) was expressed in the methylotrophic yeast Komagataella paffii as described in Experimental procedures. The protein (molecular mass 54 kDa) was secreted into the culture medium and purified by Ni-NTA affinity chromatography using the C-terminal octahistidine tag. Typical yields of pure protein varied between 5 and 7 mgÁL À1 of culture. The UV-Vis absorption spectrum of purified PpBBE1 exhibited two peaks at 375 and 443 nm indicating the presence of a flavin chromophore (Fig. 2, black line). Denaturation of the protein resulted in a single peak at 440 nm (Fig. 2, red line). Similar spectral changes were reported for other members of the BBE family and reflect the bicovalent linkage of the isoalloxazine ring via the C6-and 8a-position to a cysteine and histidine residue, respectively [2,[16][17][18]. As both residues are conserved in the sequence of  PpBBE1 (left) and PpBBE3 (right) expression throughout the whole lifecycle of Physcomitrella patens visualized using the Physcomitrella eFP Browser available on http://bar.utoronto.ca/efp_physcomitrella/cgi-bin/efpWeb.cgi [12]. PpBBE1 (H111 and C172), the observed spectral properties are in accordance with the formation of both covalent linkages. To further substantiate the presence of two covalent linkages, the protein was photoreduced under anoxic conditions. As shown in Fig. 3, reduction initially yielded the anionic flavin semiquinone, which was then further reduced to the flavin hydroquinone (Fig. 3A,B). Reoxidation of the fully reduced sample yielded absorption characteristics typical for a 6-thioflavin with a sharp peak at 425 nm and a broad absorption centered at 750 nm (Fig. 3B,C). The underlying cleavage of the carbon-sulfur bond in the covalently linked cysteine residue was previously observed for EcBBE [16] and appears to occur in all bicovalently linked flavins found in the BBE family. Yet, another hallmark of bicovalently linked flavins is their unusually high redox potential, which is typically more positive than +100 mV, for example, +132 AE 4 mV for EcBBE [17]. Thus, we also determined the redox potential for PpBBE1 using the dye equilibration method [19]. As shown in Fig. 4, the reporting dye and the flavin were reduced synchronously allowing to plot the log of the ratio of oxidized vs reduced dye against the log of the ratio of oxidized vs reduced enzyme-linked FAD (Fig. 4A,B). From this Nernst plot, a redox potential of +121 AE 1 mV was obtained, which is comparable with other BBE-like enzymes [2,17,20].

Identification of substrates
Based on the fact that the majority of BBE-like enzymes in bacteria and fungi oxidize alcohol groups of a variety of complex saccharide structures, an initial screening for enzymatic activity focused on mono-, di-, and polysaccharides (for a complete list of compounds, see Experimental procedures). This promptly revealed that PpBBE1 oxidizes the two 1,4-b-linked disaccharides cellobiose and lactose. A more detailed kinetic analysis by steady-state kinetics yielded a higher apparent k cat and lower apparent K M for cellobiose, that is, 48 s À1 vs 28 s À1 and 0.6 mM vs 4.2 mM, respectively, indicating that the catalytic efficiency with cellobiose is 10 times higher than with lactose (k cat / K M = 8 9 10 4 M À1 Ás À1 and 7 9 10 3 M À1 Ás À1 for cellobiose and lactose, respectively; Fig. 5). In order to identify the oxidation product, cellobiose was used in a scale-up experiment to obtain sufficient amounts of product, which was subsequently analyzed by NMR spectroscopy (see Experimental procedures for details). The NMR spectra clearly demonstrate the formation of cellobionolactone, that is, the free anomeric center of cellobiose is oxidized to the corresponding lactone, which spontaneously hydrolyses to the ring-open acid form (see Fig. S1). Finally, we determined the pHdependence of the reaction under steady-state conditions. As shown in Fig. 6, the activity of PpBBE1 reached a maximum at pH 5.8 (black squares).

Presteady-state experiments with cellobiose
To further evaluate the reductive and oxidative halfreaction of the enzyme, we conducted presteady-state experiments in the stopped-flow apparatus (see Experimental procedures). First, the rate of reduction in PpBBE1 was measured as a function of cellobiose concentration. The rate of reduction increased linearly with the cellobiose concentration and saturation was not observed up to 1 mM. At higher concentrations, the observed rate constants could not be extracted reliably due to the dead time restrictions of the instrument. Due to this limitation, we were unable to determine or fit a limiting rate of reduction and a dissociation constant for cellobiose. Therefore, we have used the determined bimolecular rate constant of  . Steady-state kinetics performed with PpBBE1 using cellobiose (A) and lactose (B) as substrates. Reaction rates were determined at pH 6, near the pH-optimum of PpBBE1 wild-type, keeping the enzyme concentration constant (10 nM) and varying the substrate concentration (cellobiose: 50-1500 lM; lactose: 250-8000 lM). From these data, it can be concluded that PpBBE1 shows an about 10-fold higher catalytic efficiency with cellobiose than with lactose (apparent k cat /K M = 8 9 10 4 AE 7 9 10 3 and 7 9 10 3 AE 1 9 10 3 M À1 Ás À1 from three determinations, respectively; the standard deviations are displayed as error bars).

1927
The 1.7 9 10 5 M À1 Ás À1 for further comparison and discussion. The rate of reoxidation of the reduced flavin by dioxygen was determined at a final oxygen concentration of 140 lM to 0.5 9 10 5 M À1 Ás À1 and is thus ca. 3 times slower than the rate of reduction.

X-ray crystallographic structure of PpBBE1
In order to gain further insights into the structure-function relationships in the active site of PpBBE1 and to develop a more detailed enzymatic reaction mechanism, we solved the three-dimensional structure by X-ray crystallography. Hanging drop vapor-diffusion crystallization setups was used to obtain light yellow, tetragonal bipyramidal crystals of wild-type protein as well as a protein variant (D396N) that was generated in the course of our site-directed mutagenesis study (see sitedirected mutagenesis of putative active site residues). Crystals of the D396N variant generally diffracted to higher resolution and data generated from these crystals was, therefore, used for phasing of the structure. Including data up to a resolution of 2.6 A, the structure was solved by molecular replacement using a homology model based on AtBBE-like 15 (PDB: 4UD8; [3]) as template, confirming a protein fold, that had previously been shown for BBE-like enzymes (Fig. 7) [2]. An identical structure was then confirmed for the wild-type data refined to 2. 9 A after phasing with the initial model of the variant. PpBBE1 crystallized as a loose dimer in the asymmetric unit with a dimerization interface of no biological significance according to PISA analysis [21]. In line with gel filtration experiments, PpBBE1 is, therefore, expected to be functional as a monomer. Individual protomers also showed no substantial structural differences with a root-mean-square deviation of 0.15 A for 404 aligned backbone Ca residues. A closer inspection of the active site revealed a relatively large active site with clear interpretable electron density of all active site residues. In addition, continuous electron density between the 8a-and C6-positions of the flavin cofactor and residues H111 and C172, respectively, confirmed the bicovalent linkage of the flavin cofactor. Both wildtype and the D396N structure featured a pronounced density at the expected site of substrate binding that could not be interpreted at the current resolution limits. Interestingly, the density projects toward the C4a atom of the flavin cofactor and might thereby explain the unexpectedly weak yellow coloration of the crystals. Further studies are needed to identify the source of this copurified ligand species.
The closer inspection of the active site revealed six residues with a potential role in substrate binding and activation (Fig. 8). Among these residues, two are likely to be required for substrate activation. D396 forms a hydrogen bond with Y461 and thereby may assist in the deprotonation of the tyrosine residue, which, in turn, could serve as catalytic base. Q424, which is in close vicinity to Y461 and D396, might also contribute to the activation of Y461, but could also play a role in substrate coordination or activation. The role of W186 and S422 is less obvious. Being able to interact with peptide backbone of L397 and the side chain of Y398, respectively, these two residues might be responsible for the correct positioning of D396 relative to Y461, but from the crystal structure, their importance is difficult to predict.

Site-directed mutagenesis of putative active site residues
In order to obtain more detailed information on the role of various active site residues, we initiated a site-directed mutagenesis program. Thus, all six amino acid residues found in close vicinity of the flavin cofactor were replaced, that is, tyrosine 113 and 461 to phenylalanine (Y113F and Y461F), aspartic acid 396 to alanine and asparagine (D396A and D396N), glutamine 424 to alanine (Q424A), tryptophan 186 to phenylalanine (W186F), and serine 422 to alanine (S422A). In addition, the residue held responsible for controlling the Fig. 6. Comparison of the pH-profile of wild-type PpBBE1 (black) and the D396A variant (red). Reaction rates were determined in the pH-range from 4 to 8, using cellobiose (500 lM) as substrate and DCPIP (100 lM) as final electron acceptor. All reactions were started upon addition of the enzyme (wild-type: 10 nM; D396A: 200 nM) and the decrease in absorption at 600 nm was monitored for 120 s. The wild-type enzyme showed highest activity at pH 5.8, whereas the pH-optimum of the variant was shifted by more than one pH-unit to pH~7.2 (turnover rates at the different pH-values were determined in triplicates; the standard deviations are displayed as error bars). oxygen reactivity on the si-side of the isoalloxazine ring, that is, valine 175, was replaced by leucine (V175L; Fig. 8). As shown in the summary of kinetic parameters in Table 1, the replacement of Y461, D396, W186, S422, and Q424 significantly affected the catalytic rates whereas replacement of Y113 caused only marginal changes compared to the wild-type enzyme. These results further support the hypothesis that D396 acts as an active site base and therefore determines the pH-profile of the enzyme. This concept is also in agreement with the observation that replacement of D396 by an alanine resulted in a substantial shift of the pH-optimum from 5.8 to~7.2 ( Fig. 6, red spheres). Previously, we established that the oxygen reactivity of the reduced flavin in the BBE-like enzyme family is controlled by a single 'gate keeper' residue on the siside of the isoalloxazine ring. In the case of PpBBE1, V175 is found in the pertinent 'gate keeper' position, and thus, the reduced enzyme is expected to rapidly react with dioxygen, as confirmed by the high rate of reoxidation, that is, 0.5 9 10 5 M -1 s -1 . In accordance with our 'gate keeper' model, replacement of this valine by leucine diminished the rate of reoxidation to 25 M À1 Ás À1 , thus basically suppressing the reaction of the reduced flavin with dioxygen as previously found for other members of this enzyme family [3,20,22].

Molecular docking and molecular dynamics simulations
As the substantial decrease in enzymatic activity of some variants could not be fully explained based on the crystal structure alone, molecular docking and subsequent molecular dynamics simulations with PpBBE1 and cellobiose were performed. The docking experiment resulted in a reasonable binding of cellobiose to the active site of PpBBE1, though some distances between the ligand and interacting residues were not in a favorable range. Therefore, two independent molecular dynamics simulations based on the previously obtained docking results were conducted (Fig. 8A).
The resulting average structures demonstrated a high flexibility of Y461, as the distance between D396 and the tyrosine as well as between the tyrosine and the anomeric hydroxyl group of cellobiose strongly differed in the two simulations ( Fig. 8B,C). In one simulation, Y461 is in hydrogen-bonding distance with D396, thus indicating the activation of tyrosine for catalysis, whereas the second simulation better represents the deprotonation of the anomeric hydroxyl group of the substrate molecule.
Furthermore, the significant decrease in enzymatic activity resulting from the exchange of Q424 by alanine can be better explained based on our in silico data. The amide nitrogen of Q424 directly points toward the C1* hydroxyl group of cellobiose and thus may have a role in stabilization of the catalytic transition state. In addition, the simulations indicated the involvement of several other amino acid residues in substrate binding: Q298 and K256 are both in hydrogen-bonding distance to the C4 hydroxyl group of bound cellobiose (Fig. 8B) suggesting that they are involved in substrate recognition. Their suggested role in substrate binding also explains the higher K M of lactose compared to cellobiose, as the C4 hydroxyl of lactose is an axial position and thus points away from these two amino acids.

Generation and analysis of BBE-deficient moss mutants
To analyze the function of PpBBE1 in vivo, we generated moss DPpBBE1 knockout mutants deleting the  coding sequence almost completely. A 3.6 kb long fragment was amplified from genomic DNA and the central part of this fragment comprising nearly the whole-coding sequence leaving only 38 bp of exons 1 and 37 bp of exon 3 was replaced by the nptII selection cassette (Fig. 9A). After poly(ethylene glycol)-mediated protoplast transformation (done in six independent approaches), antibiotic selection, and regeneration [23], 915 surviving colonies were picked. In an initial screening of 28 colonies by leaflet PCR [24], 12 putative DPpBBE1 lines were identified and validated by RT-PCR (Fig. 9B). Flow cytometry analysis [25] confirmed that these lines were haploid. Subsequently, the copy number of the targeting construct was analyzed with a qPCR-based method on genomic DNA ( Fig. 9C; [26]). For all further experiments, three DPpBBE1 single integration lines (#80, #83 and #100) were used.
Growth and morphology of DPpBBE1 plants were unchanged compared to WT throughout the whole life cycle from protoplast regeneration to sporophyte development. Also, upon salt stress, neither gametophores on solid plates nor protonema in liquid culture showed a phenotype differing from WT. In order to study the effect of DPpBBE1 on the degradation of the carbohydrate cellobiose, the plants were grown on media supplemented with either 0.5% cellobiose or 0.5% glucose, which is known to delay differentiation by prolongation of juvenile stages [27] and also promotes growth of Physcomitrella cultures in low light conditions [28]. Growth and development of DPpBBE1 lines were studied with and without supplementation of carbohydrates in standard growth conditions (16 h light, 8 h darkness) or darkness. Therefore, protonema cultures were grown in liquid medium (Fig. 10A) and protonema spot inocula (Fig. 11A) or single gametophores (Fig. 11B) were used as starting material on solid medium. In all conditions, no difference between DPpBBE1 and WT was detected regarding growth rate and development. The addition of cellobiose to  Table 1. Presteady-state and steady-state parameters determined for wild-type PpBBE1 and the protein variants. All measurements were performed in 50 mM HEPES buffer, pH 6, that is, near the pH optimum found for wild-type PpBBE1. In the presteady-state experiments, no saturation could be observed up to 1 mM cellobiose, where the reaction was complete within the dead time of the instrument. Thus, a linear fit was applied in the range from 100 to 800/1000 lM in order to allow comparison with the variants. The slope obtained from this fit is presented as a bimolecular rate constant k red (M À1 Ás À1 ). The turnover rate (k obs ) refers to the reaction rates determined under steady-state conditions, using 1 mM cellobiose as substrate and 100 lM DCPIP as final electron acceptor. cultures grown in the dark led to an intermediate phenotype between the control without a carbohydrate source and media supplemented with glucose, but again no significant difference between the WT and the knockout strain could be observed. In addition to the phenotypical analysis, qualitative and quantitative analysis of the glucose and cellobiose content in the supernatant of the liquid cultures was performed. After 3 weeks of incubation, the presence of reducing sugars was assayed with the Benedict's test, before conducting a quantitative analysis using the 3,5-dinitrosalicylic acid (DNS) assay (Fig. 10B) and glucose test strips (for details the Experimental procedures). As before, no differences between DPpBBE1 and WT cultures were observed. Similarly, cultures that were grown on plates for 4-6 weeks were analyzed in terms of the carbohydrate content in the solid media using congo red staining. Again, this analysis did not reveal any difference between DPpBBE1 and WT.

Discussion
The successful expression of gene #1 (Pp3c12_2640V3.1) in K. paffii led to the production of suitable amounts of PpBBE1 and enabled us to carry out the biochemical and structural characterization of the enzyme. In the course of our studies, we identified the disaccharide cellobiose as the best substrate, which is oxidized at the free anomeric center to the corresponding cellobionolactone and spontaneously hydrolyses to cellobionate. The rate of cellobiose oxidation is very fast and exceeds the rate observed for many other flavin-dependent oxidases [29][30][31][32]. In fact, the observed bimolecular rate constant already at rather low substrate concentrations is greater than the rate of reoxidation of the reduced flavin by dioxygen, which was also observed for other enzymes belonging to the BBE-like protein family. Thus, PpBBE1 clearly harbors a very efficient active site for the oxidation of cellobiose. In order to understand the catalytic contributions of the amino acids in the active site, we have generated a set of seven variants and determined the relevant kinetic parameters (Table 1). In conjunction with the crystal structure of PpBBE1, we are now in a position to propose an enzymatic reaction mechanism for the oxidation of cellobiose. As shown in Scheme 1, we propose that the side chain carboxylate group of D396 deprotonates the neighboring side chain of Y461, which in turn abstracts a proton from the anomeric hydroxyl group of the substrate leading to the transfer of a hydride to the N5-position of the flavin. The relevance of this catalytic diade is in accordance with the observed effects on the rate of substrate oxidation (Table 1) and the changes in the pH-profile found for the D396A variant (Fig. 6). The positioning of the anomeric center in the vicinity of the side chain of Y461 and the N5-position of the flavin was confirmed by substrate docking to our crystallographic structure of the protein. In addition, the characterization of the W186F and the S422A variant also revealed that these residues play an important role in organizing the active site, probably by ensuring the correct positioning of the catalytic base D396. This view is supported by the structural network established by these residues, which apparently stabilize the structural topology by interacting with the backbone carbonyl of L397 (W186) and the hydroxyl group of Y398 (S422 via a water molecule), respectively. Q424, on the other hand, may form a hydrogen bond with Y113, but in view of the total loss of enzymatic activity in the Q424A variant, direct stabilization of the transition state seems to be the primary function of this amino acid side chain. This finding is in line with the observation that, according to the crystal structure, Q424 forms only one hydrogen bond, whereas a possible second interaction with the hydroxyl group at the anomeric center of cellobiose is revealed in our docking simulations.
As mentioned in the introduction, we were particularly curious to identify a substrate (or a group of substrates) for PpBBE1 because this would provide information on the 'primordial' function of the BBElike enzyme family in plants. In this context, the identification of cellobiose and lactose as substrates of PpBBE1 is very intriguing since most members of the BBE-like enzyme family in bacteria and fungi oxidize hydroxyl groups of various carbohydrate substrates, for example, aclacinomycin oxidoreductase from Streptomyces species [33] and GOOX from the rot fungus Acremonium strictum [18]. Thus, it is apparent that this activity was already present in early land plants represented by mosses and constituted the starting point for the evolution of BBE-like enzymes in vascular plants, where members of the family are engaged in the oxidation of cyanohydrines and monolignols, as described for A. thaliana [3,4]. The unusual oxidative formation of the berberine bridge in benzylisoquinoline biosynthesis and the cyclization of the monoterpene moiety of cannabigerolic acid to D 1 -tetrahydrocannabinolic acid (THCA), catalyzed by EcBBE and THCA synthase from Cannabis sativa, respectively, have clearly taken this enzyme family to an unprecedented new level of catalytic sophistication and reflect the diverse array of secondary metabolites found in several plant lineages. On the other hand, the observation that some plants possess BBE-like enzymes with glucose oxidase activity, such as nectarin V from tobacco [7] indicates that the ability to oxidize carbohydrates was also conserved in plants. These observations are in tune with the current concept that biosynthetic pathways have evolved in part by gene duplication and subsequent neofunctionalization of one of the genes to support novel enzymatic modifications of metabolites leading to the enormous variety of compounds found in plants [9,34]. The increasing number of genes encoding members of the BBE-like enzyme family during plant evolution as well as distinct changes in the active site composition suggests that this process was relevant for the evolutionary diversification of this enzyme family in the plant kingdom [2,3].
Our in vitro findings are also in good accordance with the in vivo data obtained from the characterization of a WT and a PpBBE1-knockout strain. The expression of PpBBE1 was high in defined developmental stages, namely chloronema and protoplasts, which originate from protonema tissue. The chloronema harbors high photosynthetic activity characterized by a high number of chloroplasts of large size. On the other hand, caulonema cells possess fewer and smaller chloroplasts that are induced later by auxin and allow spreading of the colony by fast, radial growth relying on the availability of energy and carbon supply [28]. The expression pattern of PpBBE1 reflects the differences in the two protonema tissue types, being highly expressed in chloronema, but not detectable in caulonema. This finding nicely correlates with the recently performed transcriptome analysis of protoplasts [35] as well as of chloronema cells [10,11] that have shown a significant upregulation of carbohydrate and energy metabolism in these early stages of the plant's life cycle. Our hypothesis that PpBBE is involved in carbohydrate degradation is further supported by the fact that also fungi use cellobiose for energy production, which is initially metabolized extracellularly to cellobionic acid and then further degraded by cellobionic acid phosphorylases after transport into the cytoplasm. The resulting products, gluconate and glucose-1-phosphate, are then converted to 6-phosphogluconate and glucose-6phosphate before entering the pentose phosphate pathway and glycolysis, respectively [36]. Even though this catabolic pathway has never been reported for plants, it is possible that P. patens has a similar set of enzymes allowing it to most efficiently use cellulose for energy production [37].
Similarly, PpBBE1 expression was detected in the green parts of the gametophore but not the rhizoids, root-like structures with a similar morphology as caulonema filaments. On transition of the gametophyte to the sporophyte, a transcriptomic switch occurs regarding carbohydrate metabolism and photosynthesis, as the gametophyte has a primary role in photosynthesis, resource acquisition, and long-term maintenance of the plant, while the sporophyte predominantly relies on the gametophyte for both carbon and energy [10,38,39]. This switch is also reflected by the expression of PpBBE1, which is very low or not detectable Scheme 1. Proposed reaction mechanism for the oxidation of cellobiose to the corresponding cellobionolactone catalyzed by PpBBE1. D396 is expected to deprotonate the side chain -OH of Tyr461, which thereby gets activated to act as catalytic base. By abstracting the proton from the anomeric hydroxyl group (reducing end), lactone formation is triggered and a hydride in transferred from the anomeric carbon to the N5-position of the FAD cofactor. The latter thereby gets reduced and is regenerated upon reoxidation by molecular oxygen. throughout the sporophyte development to the mature spore. Interestingly, the knockout of PpBBE1 showed no apparent impact on plant morphology and development throughout the whole life cycle in all conditions tested in this study. Even upon growth in darkness, when assimilation via photosynthesis does not take place and supplementation of the media with cellobiose or glucose as carbohydrate source is required for heterotrophic growth, no differences were detected comparing DPpBBE1 and WT plants. The effect of glucose and sucrose on Physcomitrella grown in the dark leading to increased caulonema formation has been well studied [28,40]. In our study, the phenotype of cultures grown in cellobiose containing growth media is an intermediate between controls without a carbohydrate source and glucose supplemented cultures although, also here, no difference between the WT and the knockout strains was detected. These findings indicate that cellobiose is metabolized like other carbohydrates such as glucose, used as positive control, or sucrose. As upon heterotrophic growth, not only the phenotype and development of DPpBBE1 compared to WT plants but also the concentration of cellobiose or glucose in culture supernatants was unchanged, we conclude that PpBBE1, for which we identified cellobiose as best substrate, is not the only enzyme active in cellobiose metabolism in Physcomitrella. Therefore, we propose that the deletion of PpBBE1 is compensated by the action of other enzymes involved in cellobiose turnover.

Materials
All chemicals and media ingredients were purchased from Sigma-Aldrich (St. Louis, MO, USA), Roth (Karlsruhe, Germany), Merck (Darmstadt, Germany), Fluka (Buchs, Switzerland) or Becton, Dickinson and Company (Franklin, Lakes, NJ, USA) and were of the highest grade commercially available.
All restriction enzymes used were ordered form Thermo Scientific/Fermentas (St. Leon-Rot, Germany) or New England Biolabs (Ipswich, MA, USA), while the Ni-sepharose column material was obtained from GE Healthcare (Chalfont St. Giles, UK).

Cloning and expression
For heterologous expression of the PpBBE-gene in K. phaffii, the respective gene was purchased from Thermo Scientific. On ordering, the native sequence was codon optimized for K. phaffii and nucleotides coding for an octahistidine tag were fused to the 3 0 -end of the gene.
After subcloning into the Escherichia coli pJET vector, PCR amplification was performed in order to remove the native signal sequence identified using SignalP. This PCR product was then cloned into the K. phaffii vector pPICZa, to enable secretion of PpBBE1 in the non-native host. Having linearized the DNA using SacI, electroporation following the instructions provided by the EasySelect TM ExpressionKit (Invitrogen, Waltham, MA, USA) was applied to transform KM71H cells with the gene of interest. These cells were also transformed with pPICK-PDI plasmid DNA to be able to coexpress of the protein disulfide isomerase from Saccharomyces cerevisiae.
Having identified stable and well-expressing clones in small-scale screenings, following the instructions given by [41], large-scale expression in a BBI CT5-2 fermenter (Sartorius, G€ ottingen, Germany) was performed as described by [42]. After 96 h of MeOH feed (induction time), the pH was set to 8, and cells and medium were separated by centrifugation (3500 g for 20 min). The cells were discarded and the supernatant was incubated with 30 mL of Ni-Sepharose TM (GE Healthcare) Fast Flow column material, which had been equilibrated with binding buffer (50 mM NaH 2 PO 4 , 150 mM NaCl pH 8) prior to use. To ensure binding of all tagged protein, 1 h of stirring at 4°C was required, before the column material was filtered and packed into an empty column (XK 26). Having washed with~150 mL of wash buffer (50 mM NaH 2 PO 4 , 150 mM NaCl, 20 mM imidazole pH 8) to remove all the unspecifically bound proteins, PpBBE1 was eluted with elution buffer (50 mM NaH 2 PO 4 , 150 mM NaCl, 300 mM imidazole pH 8). SDS/PAGE analysis was then used to determine the quality of the purification, before pooling the fractions with the highest protein content and lowest amount of impurities (typical yields: 5-7 mgÁL À1 of culture;~98% purity). Subsequently, the protein solution was concentrated and the buffer was exchanged to 50 mM HEPES buffer pH 6 using CentriprepsÒ 30 kDa MWCO (Millipore Merck, Burlington, MA, USA).

Protein denaturation
To be able to determine the protein concentration based on the characteristic absorption of FAD at about 450 nm, a spectrum of the native as well as of denatured PpBBE1 was recorded. Assuming that the extinction coefficient of the flavin species found in denatured PpBBE1 is equal to the one of 6-S-cysteinyl FMN, an extinction coefficient of 9800 M À1 Ácm À1 could be calculated for native PpBBE1 at 443 nm.

Photoreduction
Photoreduction was performed as described by Massey et al. [43]. Under anaerobic conditions, quartz cuvettes were filled with 20 lM enzyme solution, which also contained 5 mM EDTA as photosubstrate as well as 5 lM 5-deaza-FMN. Having recorded a UV/VIS spectrum of the native, oxidized enzyme, the sample was irradiated and further spectra were recorded at different time points. In order to avoid heat denaturation of the protein, the cuvettes were constantly cooled to 15°C.

Determination of the redox potential
The redox potential of the flavin cofactor was determined using the dye-equilibrium method based on the xanthine/ xanthine oxidase system first reported by Massey [19]. Two separate solutions, one containing 500 lM xanthin, 5 lM methyl viologen, and 30 lM PpBBE1 and a second one with xanthin oxidase (about 5 ng) and toluyene blue (A 647 = 0.3; E°: +115 mV) were prepared in 50 mM HEPES buffer pH 7. By mixing these solutions in a stopped flow device (SF-61DX2; Hi-Tech, TgK Scientific Limited, Bradford-on-Avon, UK), placed in an anaerobic glove box, the reaction was started. To monitor the reduction in the enzyme and the dye, 300 absorption spectra (350-700 nm) were recorded with a KinetaScan T diode array detector (MG-6560) from Hi-tech, within 50 min. The log ([ox]/ [red]) of the enzyme was plotted against the log ([ox]/[red]) of the dye allowing calculation of the redox potential as described by Minnaert [44].

pH-optimum
Having identified two well-accepted substrates for PpBBE1, the pH-optimum was determined in order to be able to study the enzyme characteristic kinetic properties at the optimal pH. Therefore, different buffers in the pH-range of 4-9 were prepared and substrate conversion was analyzed for the wild-type as well as for the variant enzymes. Since both substrate and product do not have any characteristic spectral properties, a coupled assay with DCPIP (100 lM final concentration) was established. By recording the change in absorption at 600 nm, substrate conversion could be monitored and the pH-optimum could be determined. All measurements were performed with constant substrate concentration (1 mM cellobiose) and an enzyme concentration of 10 nM for the wildtype and 200 nM for the variant enzymes, respectively.

Steady-state kinetics
Steady-state kinetics of PpBBE1 were measured with cellobiose and lactose for the wild-type enzyme and all active site variants were subsequently only measured with the more efficiently converted cellobiose substrate. A coupled assay with DCPIP (100 lM final concentration in 50 mM HEPES pH 6) was used to determine the initial rates observed upon altering the substrate concentration (cellobiose: 50-1500 lM; lactose: 250-8000 lM), while keeping the enzyme concentration constant. As for the analysis of the pH-optimum, final enzyme concentrations of 10 nM for the wild-type and the Y113F variant and of 200 nM for the remaining variants were used. All reactions were monitored at 600 nm for 120 s and were performed in triplicates.
By plotting the extracted initial velocities vs the respective substrate concentration, v max as well as apparent K M values could be determined.
Due to the low reactivity of protein variants, no full Michaelis-Menten curve was recorded. Instead, conversion rates at 1 mM cellobiose were used for an internal comparison and measured in triplicates.

Presteady-state analysis
For analysis of the reductive and the oxidative half-reaction, a stopped-flow device from Hi-Tech (SF-61DX2) installed in a glove box (Belle Technology, Cambridge, UK) was used. In both cases, time-dependent spectral changes of the flavin absorption were monitored either with a KinetaScanT diode array detector (MG-6560) or a photomultiplier (PM-61s). Data analysis was then performed with the Kinetic Studio Software (TgK Scientific Limited) by fitting the data points recorded at 443 nm.
To study the reductive half-reaction, observed rate constants for flavin reduction at six different substrate concentrations (100-1000 lM) were determined in 50 mM HEPES pH 6 (at higher substrate concentrations, the reaction was too fast to reliably fit the data to extract rate constants). By plotting k obs vs the corresponding substrate concentration, a linear increase in reaction rate constants could be observed, making it impossible to fit a limiting reductive rate. In order to allow comparison with the active site variants, a linear fit was applied and the slopes were used for data interpretation.
The oxidative half-reaction was studied by following the reoxidation of the flavin upon mixing enzyme, reduced with 1.2 eq of substrate, with air-saturated buffer (20°C). To obtain bimolecular rate constants, the rates determined in this experiment were divided by the amount of oxygen dissolved in the buffer (final concentration: 140 lM).

Site-directed mutagenesis
In order to study the reaction mechanism in more detail, active site variants were produced. All constructs were generated from pPICZa-PpBBE1 wild-type expression strain applying polymerase chain reaction-based mutagenesis. The amino acid exchanges were introduced with the forward and reverse primers that both carried the desired mutations in the nucleotide sequence (for primer sequences, see Table 2). After successful cloning and transformation of the mutant strains, expression and purification of the PpBBE1 variants were performed as described for the wild-type protein.

Crystallization and data collection
Protein crystals were grown in hanging drop vapor diffusion setups using EasyXtalÒ 15-Well Tools (Qiagen, Hilden, Germany) and the corresponding crystallization supports (Qiagen). Protein solution (200 lM in 50 mM HEPES buffer pH 6) and crystallization reagent (0.1 M Mg (OOCH) 2 , 13/14% poly(ethylene glycol) 3350 for wild-type and variant, respectively) were mixed in 2 : 1 or 1.5 : 1 ratio and incubated at 20°C. After overnight equilibration, streak seeding with crystals from the initial hit was performed. Ten to 14 days later, bipyramidal crystals appeared, which reached their final size after 3 weeks. For cryoprotecting crystals when flash freezing them in liquid nitrogen, 3 lL of 0.1 M Mg(OOCH) 2 , 13/14% poly(ethylene glycol) 3350, and 40% glycerol were added to the drop containing the PpBBE1 crystals a few minutes prior to freezing. Crystals were stored in liquid nitrogen and data were collected at 100 K at beamline P11 of the German Electron Synchrotron (DESY, Hamburg, Germany).

Data processing and structure elucidation
Data from the tetragonal bipyramidal crystals were processed with the XDS package [45]. A unique molecular replacement solution was obtained in space group P4 1 2 1 2 in Phenix Phaser [46], using a homology model generated on the SWISS-MODEL sever [47][48][49] based on the crystal structure of AtBBE-like 15 (pdb code: 4UD8; [3]) as template. The initial solution was refined in cycles of maximum-likelihood least-squares refinement of models modified with Coot [50] using r A -weighted 2mF o À DF c and F o À F c electron density maps and including an initial simulated annealing (torsion) step. Details of data collection, processing, and refinement are summarized in Table 3.

Molecular docking
The crystal structure of PpBBE1 wild-type (WT; PDB: 6EO4) in complex with the bicovalently linked FAD cofactor was used as a basis for substrate-docking experiments. All molecular docking experiments and subsequent MD simulations were performed with the YASARA Structure suite (YASARA structure version 17.3.30 (Yasara Biosciences [51]). Crystal waters were removed from the structure prior to the docking experiment. Molecular docking was conducted utilizing the Autodock Vina Plugin [52] of the Yasara Structure suite. Docking was performed with the rigid receptor while the ligand was flexible with a docking cell lining the active site cavity of PpBBE1. The ligand chosen for the docking experiments was the previously identified substrate of PpBBE1, cellobiose. Two hundred and fifty docking runs were performed with an RMSD cutoff of 1 A. The resulting docking poses were inspected visually. Fwd

Molecular dynamics simulation
The enzyme complex consisting of PpBBE1, the bicovalently linked FAD cofactor and the previously docked substrate cellobiose were used as the starting structure for molecular dynamics (MD) simulations. MD simulations and analyses were performed using the YASARA Structure suite, version 17.3.30 (YASARA Biosciences [51]). A periodic simulation cell that enveloped the whole enzyme with an additional 5 A margin in each dimension was used with explicit solvent. The AMBER14 force field [53] was applied and long-range electrostatic potentials were calculated using the Particle Mesh Ewald (PME) method and a cutoff of 7.864 A [54,55]. AutoSMILES utility was used to attribute force field parameters to PpBBE1 and FAD [56]. The hydrogen-bonding network optimization was carried out using the method of Hooft and coworkers [57], and pK a values at pH 6 were assigned [58]. The simulation cell was filled with water containing 0.9% of NaCl (density of 0.993 gÁmL À1 ). Solvent was relaxed and the system was subsequently energy minimized using steepest descent minimization to remove conformational stress, followed by a simulated annealing minimization to convergence (< 0.05 kJÁmol À1 per 200 steps). Integration time steps were 1.33 and 4 fs for intra-and intermolecular forces, respectively. Subsequently, MD simulations at 310 K were performed, whereby integration time steps for intramolecular and intermolecular forces of 1.25 fs and 2.5 fs were applied, respectively. MD simulations were carried out over a time span of 12.5 ns. A restrain was added to the force field to maintain a catalytically competent complex at the active site. A restraining spring with a force of 100 SFC was applied combined with a distance restrain of 3 A between the catalytically relevant hydrogen (H1*) of the substrate and the N5 of FAD. Additionally, an angle of 105°was set between N10-N5 and the previously mentioned hydrogen atom (H1*) of the substrate. The restraints were set following the findings of defining general distances and angles for catalysis-competent complexes of flavoenzymes and their respective substrates [59]. Analysis of MD simulation results included measuring general distance traces and hydrogen-bonding events of cellobiose with predefined amino acid side chains of PpBBE1 that determine the structure of the active site.

Secretome analysis
For secretome analysis, the liquid culture supernatant of a P. patens line secreting human recombinant Erythropoietin (IMSC40216, www.moss-stock-center.org) [60,61] was precipitated with TCA as described in [61]. Sample preparation and subsequent SDS/PAGE as well as mass spectrometry were performed as described in [62]. Raw data processing and database search were performed against the V1.6 protein models (www.cosmoss.org) using Mascot software (Matrix Science, Boston, MA, USA).

Physcomitrella patens moss material and growth conditions
For all experiments, P. patens (Hedw.) B.S. ecotype 'Gransden 2004 0 was used. This is the same strain that was used for genome sequencing [63] and was deposited at the International Moss Stock Centre (IMSC, http://www.mossstock-center.org) under accession number 40001. The DPpBBE1 lines described in this study were cryopreserved according to [64] and are available from the International Moss Stock Center (IMSC) with the following accessions: 40829 for DPpBBE1#80, 40830 for DPpBBE1#83, and 40831 for DPpBBE1#100.
The material was axenically cultured in Knop medium (250 mgÁL À1 KH 2 PO 4 , 250 mgÁL À1 KCl, 250 mgÁL À1 MgSO 4 9 7 H2O, 1 gÁL À1 , Ca(NO 3 ) 2 9 4 H 2 O, and 12.5 mgÁL À1 FeSO 4 9 7H 2 O, pH 5.8; [65]) supplemented with microelements [66] cultivated in a growth chamber under controlled conditions (25°C) with a 16 h : 8 h, light:dark regime at a light intensity of 70 lmolÁm À2 Ás À1 as described in [67]. For supplementation of media with carbohydrates, cellobiose and glucose were added to a final concentration of 0.5% (w/v). For salt stress experiments, NaCl was added to the solid culture media in concentrations of 500 mM, 750 mM, and 1 M. For the adjustment of dry weight in P. patens, 10 mL of liquid culture was removed prior to subculturing and filtered through gauze (Miracloth, Calbiochem, Schwalbach, Germany). The dry weight was determined after drying the sample at 105°C for 2 h. For the induction of sporophytes, the Petri dishes were transferred to inducing conditions according to [68]. For dark treatment, protonema in liquid medium were grown for 3 weeks, whereas gametophores on solid medium were incubated in a light-tight box in a growth chamber for 4-6 weeks. Additionally, spot inocula of protonema were grown in darkness for 4 weeks either directly after inoculation or after preculture under standard light conditions for 2 weeks.

Generation and validation of DPpBBE lines
For construction of the PpBBE1-knockout construct, genomic DNA was isolated according to [69] and used as template in an overlap extension PCR. Fragments of exon 1 and exon 3, together with the respective~850 bp flanking regions, were amplified in separate PCR reactions, with the reverse primer for exon 1 (first 38 bp) and the forward primer for exon 3 (last 37 bp) carrying an overlapping sequence with the flanking regions of the ntpII selection cassette, which should replace the missing gene fragment (Fig. 9A, Table 4). In a second PCR reaction, all three fragments were fused, before subcloning the newly generated DNA to pJET and transforming the recombinant plasmid with E. coli top 10 cells for further amplification. Since sequencing of the subcloned DNA revealed a mutation, the error was removed in an additional mutagenesis PCR reaction (Primers see Table 4). Prior to the transformation procedure, the PpBBE1knockout construct was cut from the pJET vector using SphI and PacI. Protoplast isolation, transformation, and regeneration to obtain stable knockout plants were performed as previously described [23] using 40-50 lg of DNA. Selection was done in two successive cycles by transferring the cultures to solid media complemented with 25 mgÁL À1 G418 (Promega, Mannheim, Germany)-release period of 2 weeks between the two selections as described in [23]. Plants were screened for correct 5 0 -and 3 0 -integration of the knockout construct via direct PCR [24] using the primer combinations BBE_5 0 Int_f and RT1 or RT4 and BBE_3 0 Int_r, respectively. As a positive control for successful extraction of DNA, the primers EF1a_fw and EF1a_rev were used. The loss of transcript in the putative DPpBBE1 lines was verified by direct PCR, using the primer couple BBE_E2_f and BBE_E3I_r, which generates a transcript of 546 bp length in WT. As control for successful cDNA synthesis, C45_fwd und C45_rev were used, amplifying a part of the constitutively expressed mRNA of the ribosomal protein L21.
To estimate the number of integrations of PpBBE1targeting constructs, a qPCR-based method [26] was used. The 5 0 -HR region and the 3 0 -HR region were amplified using the primer couples qPCR_BBE_5HR_fw and qPCR_BBE_5HR_rev and qPCR_BBE_3HR_fw and qPCR_BBE_3HR_rev, respectively, whereas transcripts of the ntpII selection cassette and the single copy gene pCLF (internal control) were generated using the primer couples q_npt_fw and q_npt_rev and pCLF_5915_qf and pCLF_5981_qr or pCLF_7739_qf and pCLF_7804_qr, respectively. From these results, the copy number of PpBBE1-knockout genes could be determined.
As control lines, a WT strain and a line with a confirmed single integration of the nptII selection cassette were used. The complete list of primers used for the cloning of the PpBBE1-targeting construct and molecular validation of the transgenic plants is shown in Table 5.
As the DNA content of Physcomitrella is known, haploidy of the DPpBBE1 lines could be confirmed in a fluorescence-based assay. Nuclei prepared from 10 to 30 mg freshly chopped protonema material were stained with a 4 0 ,6-diamidino-2-phenylindole (DAPI)-containing buffer [25] and analyzed using a CyflowÒ Space flow cytometry system (Partec, M€ unster, Germany) as described in [70].

Qualitative and quantitative determination of cellobiose and glucose
In order to qualitatively assay the cellobiose and glucose content of solid media, plates were stained with a solution of 0.1% (w/v) congo red. After washing with a 0.5 M Table 4. Primers used for the generation of the DPpBBE1 strain.
Primer Sequence NaCl solution for five times, for better visualization of lighter stained zones, the pH was adjusted to 2. Benedict's test, using 1 mL of the sample and 200 lL of Benedict's reagent, performed according to [71], was used to qualitatively analyze the liquid culture supernatants. Here, a color change from blue to red, caused by the precipitation of copper sulfate, indicates the presence of reducing sugars in the solution. For quantification of cellobiose and glucose in the liquid media, a DNS assay was performed as described [72]-250 lL of sample was mixed with 250 lL of 0.005 M sodium acetate buffer pH 4.8 and 750 lL of DNS reagent. Photometric quantification was done at 575 nm for all samples and appropriate standard dilution series for each sugar. Additionally, sugars were quantified using the glucose test strips of the Reflectoquant system (Merck).

Image acquisition
Images were acquired with an ICc1 or Mrc5 CCD camera (Zeiss, Oberkochen, Germany) at an Olympus SZX7 binocular microscope (Olympus, Tokyo, Japan) or an Axiovert S100 (Zeiss). AXIOVISION Software Version 4.8 (Zeiss) was used for imaging. Table 5. Primers used for the cloning of the PpBBE1-targeting construct and the subsequent molecular validation of the transgenic plants.