To examine the metabolic impact of the three RIBA isoforms identified in A. thaliana, the relative transcript amounts of the three AtRIBA homologs were assessed using qRT-PCR analyses in different tissues and developmental stages of wild-type plants. Transcripts of the three RIBA homologs accumulate in all analyzed tissues, but transcript levels were different (Figure 2A). The accumulation of AtRIBA1 mRNA exceeded the transcript levels of AtRIBA2 and AtRIBA3 in all organs and developmental stages analyzed. However, the widely parallel accumulation of AtRIBA transcripts does not indicate a strong tissue specificity of gene expression of single members of the AtRIBA gene family.

A strong down-regulation of the AtRIBA1 gene expression in the Arabidopsis rfd1 mutant correlated with a bleached phenotype [18]. Seedlings were unable to grow photoautotrophically on soil and their growth was abandoned in sugar-supplemented media after several weeks. Interestingly, the bleaching phenotype of the seedlings is not counterbalanced by the simultaneous AtRIBA2 and AtRIBA3 expression, which remains at wild-type levels in rfd1[17]. For a gradual reduction of riboflavin biosynthesis, several transgenic lines with AtRIBA1 antisense RNA expression under control of the CaMV 35S promoter were generated (Figure 3A). RIBA mRNA levels were determined in rosette leaves of two representative AtRIBA antisense lines (A1#2 and A1#5) displaying different degrees of pigment deficiency. While the intensity of the phenotype in the two selected lines correlates with reduced AtRIBA1 transcript levels, AtRIBA2 and AtRIBA3 are expressed at levels comparable to wild-type tissue (Figure 2B). Hence, the expression of the latter isoforms does not prevent the deficiency in riboflavin biosynthesis and the decline of AtRIBA1 expression in Arabidopsis was not compensated for at the transcriptional level by modified activity of the homologous genes.

The phenotypic alterations of individuals with a severe phenotype, including line A1#2, were compared between three leaf sections. While the leaf base (Sample I in Figure 3B,C) was similar to a wild type, a progressive loss of pigmentation was observed towards the leaf tip (samples II and III). In these leaf sections, a gradual reduction of riboflavin (Figure 3C) was observed. AtRIBA1 protein levels were strongly decreased in A1#2 leaves (Figure 3D), agreeing with the observed reduction in mRNA amounts.

Since all three AtRIBA genes are constantly expressed, the observed AtRIBA1 depletion phenotype of rfd1[18] and the antisense lines depicted in Figure 3A could be explained by spatial separation of AtRIBA proteins in different subcellular compartments.

To address the localization of the three AtRIBA proteins, iPSORT [19], TargetP [20] and Predotar [21] algorithm were used for targeting predictions of their subcellular localization (Figure 4A). The prediction hinted at the existence of an N-terminal RIBA transit peptide and either a plastidic or mitochondrial localizations. We examined in situ targeting of the three RIBA homologs by CLSM-mediated visualization of transiently expressed green fluorescent protein (GFP) fusion proteins after transformation of Nicotiana benthamiana leaves. Based on predictions by TargetP, three different gene constructs encoding the putative transit peptides of AtRIBA1-3 were fused to 5′-end of the GFP-encoding sequence (Figure 4A). The co-localization of chlorophyll and GFP fluorescence observed for all three RIBA-GFP fusions (Figure 4B) clearly demonstrates that all AtRIBA N-termini contain plastid targeting signals.

For detailed comparative characterization of AtRIBA enzymes in vitro, all three RIBA-encoding sequences were expressed in E. coli. The design of the artificial RIBA sequences included adaptation to E. coli codon usage and addition of an N-terminal His–tag (Data S1). All RIBA proteins specified by newly generated plasmid constructs lacked the plant-specific N-terminal sequence including the transit peptide. Thus, in comparison to the wild-type protein precursor sequences the recombinant proteins lacked the first 127 (AtRIBA1), 105 (AtRIBA2) or 100 amino acids (AtRIBA3), respectively (cf. alignment in Figure S1B in [17]).

All three over-produced recombinant RIBA proteins were purified by FPLC using metal affinity chromatography. SDS polyacrylamide gel electrophoresis of the purified recombinant AtRIBA1-3 proteins confirmed the predicted molecular weights of 47.6, 42.3 and 46.5 kDa, respectively (Figure 5A). The purified AtRIBA proteins were assayed for GCHII and DHBPS activity in vitro (Figure 5B,C). A GCHII enzymatic function was clearly demonstrated for RIBA1 as well as for RIBA3, whereas RIBA2 did not display a detectable activity (Figure 5B). A DHBPS activity was determined for RIBA1 and RIBA2. Here, RIBA3 did not display a measurable DHBPS enzymatic function (Figure 5C). Taken together, the data obtained from in vitro assays clearly indicate that both recombinant AtRIBA2 and AtRIBA3 isoforms are able to carry out only one of two enzymatic activities of bifunctional RIBA proteins.

In E. coli, ribA and ribB genes encode monofunctional GCHII and DHBPS enzymes, respectively. Corresponding E. coli mutant strains were employed to study complementation of respective enzymatic functions by RIBA1-3 in situ and corroborate the data obtained in vitro.

AtRIBA sequences designed for expression in E. coli (Data S1) were cloned into pACYC184 and transformed into E. coli ribA and ribB mutants, respectively. The vector pACYC184 was also used as blank control. We assayed the growth rate (OD₆₀₀) of resulting E. coli transformants in M9 minimal medium liquid cultures (Figure 5D,E). Since ribA and ribB are riboflavin auxotrophs, precultures were grown in LB medium containing 0.4 g/L riboflavin. At OD₆₀₀ of 0.6, the cells were pelleted, washed and resuspended in M9 minimal medium (OD₆₀₀ = 0.1) and grown at 37 °C for 24 h.

Cultures of both E. coli mutants containing plasmid pACYC184 reached in M9 minimal medium a stationary phase at OD₆₀₀ of approx. 0.2 (Figure 5D,E). Cultures of the ribA mutant (deficiency in GCHII) revealed an enhanced growth rate when transformed with RIBA1 or RIBA3 constructs, while the growth rate of E. coli ribA cells transformed with pACYC-AtRIBA2 did not differ from the blank control strain (Figure 5D). The culture of the DHBPS-deficient mutant strain ribB displayed an enhanced growth when transformed with either the RIBA1 or RIBA2 encoding plasmid. In contrast, pACYC-AtRIBA3 was not able to improve the growth rate in comparison to the empty control plasmid (Figure 5E).

Based on the description of rfd1[17,18], the specific function of the three homologous AtRIBA genes in Arabidopsis riboflavin biosynthesis was further investigated. Transgenic lines were generated that constitutively express AtRIBA1 antisense RNA. These new lines displayed various degrees of a bleaching leaf phenotype (Figure 3A) occurring at different stages of plant development. An early bleaching of individual lines, like A1#2, phenotypically resembled rfd1. These lines were unable to survive under photoautotrophic conditions. The leaves of other transgenic lines including A1#5 bleached later during plant development (Figure 3A). Here, the first leaves developed with wild-type like pigmentation, while new leaves in the Arabidopsis rosette stage turned white. The loss of pigmentation corresponds to inactivation of AtRIBA1 expression as demonstrated by quantitative PCR analysis (Figures 2B and 3D).

Numerous Arabidopsis tissues were analyzed by qPCR to compare the abundance of AtRIBA mRNAs. The transcript profile of all three homologous RIBA genes (Figure 2A) reveals a largely constitutive expression with AtRIBA1 being the most abundant transcript in all tested tissues. This implies that AtRIBA2 as well as AtRIBA3 are ubiquitously expressed and their expression has at least the potential to partially complement AtRIBA1 deficiency. The transcript analysis of antisense lines (Figure 2B), however, demonstrates that there is no concerted transcriptional regulation of the RIBA gene family members in Arabidopsis, since the isogenes did not show altered transcript accumulation in AtRIBA1 antisense lines.

It could be speculated that the characteristic bleaching phenotype is explained by an insufficient dosage of expressed RIBA protein as the overall amount of RIBA transcripts and proteins in antisense plants is severely reduced (Figures 2B and 3C,D). Alternatively, the lack of complementation of AtRIBA1 deficiency might be caused by differences in subcellular localization or enzymatic functions among the homologs.

The localization of the RIBA isoforms was tested employing GFP fusions in transiently transformed Nicotiana leaf cells. All three AtRIBA proteins possess N-terminal sequences that exclusively direct GFP fusions to plastids (Figure 4). These findings agree with mass spectrometry data available for all three proteins from plastid proteome projects as summarized in the SubCellular Proteomic Database (SUBA, http://suba.plantenergy.uwa.edu.au/) [22] as well with data available for further enzymes of riboflavin synthesis (SUBA entries for At4g20960, At2g44050, At2g20690) and, thus, corroborates the assumption that plant riboflavin biosynthesis is exclusively localized in plastids [9]. In conclusion, the plastid localization of all three RIBA isoforms does not explain the inability of RIBA2 and RIBA3 to complement RIBA1 deficiency in Arabidopsis.

Differences in enzymatic activities of the bifunctional A. thaliana RIBA homologs have previously been suggested when the protein primary structures were compared with sequences of bacterial RibA proteins (Figure S1 in [17]). Amino acids known to be indispensable for either binding of zinc ions (i.e. for GCHII activity) or substrate binding and catalysis (i.e., for DHBPS activity) are lacking in AtRIBA2 and AtRIBA3, respectively [23,24]. The recent characterization of a N. benthamiana RIBA homolog underlined the functional importance of specific residues in plant RIBA sequences [25].

The present work aimed at comparatively examining the enzymatic activities of all three AtRIBA homologs by two alternative approaches. First, purified recombinant RIBA proteins were employed in in vitro assays (Figure 5A–C). Second, the complementation of E. coli ribA and ribB mutants substantiated the results in situ (Figure 5D,E). A ribB (DHBPS)-deficient strain showed improved growth characteristics when complemented with the AtRIBA1 and AtRIBA2 sequences. Although the complementation was only partial, the elevated growth rate was characteristic in comparison to that of the same mutant containing the AtRIBA3-expressing plasmid. The complementation of ribA deficient E. coli cells was more effective by AtRIBA1 than by AtRIBA3. However, the latter exceeded the effect of AtRIBA2 and the pACYC184 vector control.

In conclusion, both the in vitro and the in situ approaches confirmed the hypothesis that AtRIBA2 as well as AtRIBA3 represent only mono- instead of bifunctional enzymes for riboflavin biosynthesis: while AtRIBA2 is lacking GCHII activity, AtRIBA3 does not display DHBPS function.

The architecture of bifunctional enzymes has been suggested to favor the coordination of expression and catalysis [26]. GCHII and DHBPS are the first committed enzymes of a converged pathway, in which two molecules of the DHBPS product and one of GCHII (9 and 2 in Figure 1) are required to finally synthesize riboflavin. Interestingly, in vitro enzyme assays revealed a two times higher activity of recombinant DHBPS compared to GCHII [10]. Hence, the expression of a fused DHBPS-GCHII protein could ensure catalysis of adequate amounts of reaction products required in the riboflavin biosynthetic pathway [26].

We make the point that the Arabidopsis genome contains two genes encoding bipartite proteins with only one enzymatic function in the riboflavin biosynthetic pathway. We assert, with our results, that expression of both monofunctional enzymes is insufficient for a replacement of bifunctional AtRIBA1. Different reasons may account for the lack of complementation by means of AtRIBA2 and AtRIBA3: their lower expression strength in comparison to AtRIBA1, steric hindrance of the two monofunctional proteins or an involvement of the latter in other metabolic pathways.

In addition, we can not entirely exclude potential mechanistic and structural functions of RIBA2 and RIBA3 in the formation of multienzymatic complexes in riboflavin biosynthesis. Thus, sequestration and protection of labile metabolic intermediates can improve metabolic channeling as demonstrated recently for bifunctional BIO3-BIO1 in Arabidopsis biotin synthesis [27].

Interestingly, the occurrence of at least three RIBA homologs is conserved among all angiosperms of which complete genome data are available. The sequence similarity of the different isoforms was investigated by phylogenetic analysis (Figure 6). As indicated, Selaginella, an early tracheophyte, represents the first plant harboring more than one RIBA gene. Already one of the two isoforms found in the genome of this lycopodiophyte (designated SmRIBA2 in Figure 6) is situated in a RIBA3-specific clade of the phylogenetic tree. A comparison of protein primary sequences reveals an exchange of essential amino acids in all members of this RIBA3 clade and thus implies a loss of DHBPS function in this group early in higher plant evolution (Figure S1). This hints at a specific requirement for an independent GCHII activity that became irreplaceable due to an acquisition of special functions in plant metabolism. Interestingly, evolutionary changes of GCHII activities have been reported recently for three GCHII isogenes in Streptomyces[28].

It is noteworthy that Arabidopsis possesses an unusual set of RIBA isoforms with AtRIBA2 being monofunctional in riboflavin biosynthesis. All other inspected angiosperm species comprise RIBA2 genes that encode all necessary amino acids to fulfill DHBPS as well as GCHII function as illustrated by the amino acid alignment depicted in Figure S2. Due to the close relationship of RIBA1 and RIBA2 isoforms they are forming a common clade in the phylogenetic tree. However, the unusual evolutionary situation of AtRIBA2 is reflected by an extended branch length (Figure 6).

At present, we hypothesize that the loss of bifunctionality for two out of three AtRIBA isoforms hints either at novel metabolic functions or at a structural role in the spatial organization of plant riboflavin biosynthesis. To investigate the specific impact of a loss of AtRIBA2 and AtRIBA3 function on plant metabolism is hence a challenging task to be addressed in the future.

The AtRIBA1 coding region was amplified using primers 17 and 18 (Table S1). The product was cut using SmaI and inserted into binary vector pGL1, which was derived from pGPTV-bar [30] by removing GUS and introducing a 35S CaMV promoter and a multiple cloning site. A. thaliana was transformed using standard procedures.

Enzymatically essential RIBA regions were identified based on alignments with prokaryotic RibA and RibB ([17], therein Figure S1). Arabidopsis sequences were adapted to E. coli codon usage and N-terminal His-tags integrated (Data S1). Sequences were provided as pUC57 subclones by GenScript (Piscataway, NJ, USA). NdeI/HindIII fragments were cloned into pET22b(+) (Merck, Darmstadt, Germany). Expression in ArcticExpress™ (DE3) RIL Competent Cells (Agilent, Santa Clara, CA, USA) was induced with 1 mM IPTG at 13 °C for 24 h. Soluble recombinant proteins were purified via FPLC using HisTrap HP columns (GE Healthcare, Uppsala, Sweden) and dialyzed with SnakeSkin^TM Pleated Dialysis Tubing (10,000 MWCO, Thermo Scientific, Waltham, MA, USA) in 20 mM Tris-HCl, 200 mM NaCl, 5% glycerol, pH 8.4. The concentration of recombinant RIBA protein fractions was determined by comparison to Bovine Serum Albumin standards on Coomassie stained gels.

The enzyme assays were performed according to [31] with minor modifications. For GCHII activity: Assay mixtures (100 μL) containing 100 mM Tris-HCl pH 7.8, 5 mM MgCl₂, 5 mM DTT, 100 μM GTP and either RibA from B. subtilis (0.5 mg/mL) or RIBA1-3 from A. thaliana (0.975, 0.15, 0.175 μg/μL) were incubated 1 h at 37 °C. After addition of EDTA (10 mM) and diacetyl (5 mM), the samples were incubated 1 h at 37 °C. Then, 100 μL of TCA (300 mM) were added and samples centrifuged for 2 min at 13,000 rpm and 10 °C. 20 μL from the supernatant were loaded on RP18 column (Lichrospher100 RP18, 5 μL, 250 × 4 mm, flow rate 1 mL min⁻¹) and eluted isocratically with methanol/water (v/v 4:6). Effluent fluorescence of 6,7-dimethylpterin was monitored (λ_ex 340 nm; λ_em 400 nm). For DHBPS activity: assay mixtures (100 μL) containing 100 mM Tris-HCl pH 7.8, 5 mM MgCl₂, 5 mM DTT, 100 μM d-ribose 5-phosphate, phosphoriboisomerase (0.5 U/mL), 100 μM, 5-diamino-6-ribitylamino-2,4(1H,3H) pyrimidinedione, lumazine synthase of B. subtilis (0.5 U/mL) and either B. subtilis RibA (0.5 mg/mL) or A. thaliana RIBA1-3 (0.975, 0.15, 0.175 μg/μL) were incubated 2 h at 37 °C. 20 μL from the supernatant were loaded on RP18 column and eluted isocratically with methanol/water/formic acid (26:234:1). Effluent fluorescence of 6,7-dimethyl-8- ribityllumazine was monitored (λ_ex 408 nm; λ_em 490 nm). Synthetic dimethylpterin and 6,7-dimethyl- 8-ribityllumazine have been used as calibration standards.

RIBA1 and RIBA3 sequences were cut from pUC57 (see above) with Ecl136II/HincII (RIBA1) or Ecl136II/SmaI (RIBA3). For RIBA2, primers 1 and 2 (Table S1) were applied using RIBA2 in pUC57 as template; the resulting product was cut with PsiI. All RIBA fragments were inserted into the EcoRV site of pACYC184 (NEB, Ipswich, MA, USA). Constructs were transformed into E. coli knock-out strains BSV18 (ribA18::Tn5; CGSC# 6992) and BSV11 (ribB11::Tn5; CGSC# 6991) defective in RibA (GCHII) and RibB (DHBPS), respectively, [32] obtained from the E. coli Genetic Stock Center (http://cgsc.biology.yale.edu/).

For growth assays, LB cultures with 0.4 g/L riboflavin were inoculated with an overnight culture and grown at 37 °C, 250 rpm to an OD₆₀₀ of 0.6. The mutants require high concentrations of externally added flavins due to the lack of specific riboflavin import mechanisms in E. coli. Cells were pelleted and washed three times in M9 minimal medium containing 20% glucose [33]. M9 cultures containing antibiotics were inoculated with washed cells to give an initial OD₆₀₀ of 0.1.

Total RNA was isolated using innuPREP Plant RNA Kit (Analytic Jena, Jena, Germany). 0.4 μg of DNAseI-pretreated total RNA was reverse transcribed with oligo dT18 using RevertAid RT (Thermo Scientific, Waltham, MA, USA). cDNA was amplified with SensiMix SYBR No-ROX kit (Bioline, London, UK) on a CFX96 Real-Time System (Bio-Rad, Hercules, CA, USA) using the oligonucleotides listed in Table S1. Expression rates were calculated relative to SAND (At2g28390 [34]) according to the 2^−ΔCT method [35,36].

Plant material (0.1 g) harvested from rosette leaves of 8-week-old plants grown under short day conditions (10 h light/14 h dark) at 120 μmole photons m⁻² s⁻¹ was ground in liquid N₂, resuspended in 0.5 mL of methanol/methylen chloride (9:10), incubated for 2 h under gentle agitation at 4 °C and centrifuged. Samples were analyzed on HPLC system 1100 (Agilent, Santa Clara, CA, USA) using a NovaPak C18 column (150 mm; 3.9 mm diameter; 4 μm particle size) and eluted with a linear gradient of 50 mM ammonium acetate (pH 6) and methanol from 100% to 47% over 20 min at 0.8 mL/min. Riboflavin, FMN and FAD were detected by fluorescence (λ_ex 265 nm, λ_em 530 nm) and confirmed using authentic standards (Sigma-Aldrich, St. Louis, MO, USA).

N-termini of RIBA proteins were amplified from cDNA using primers 3/4, 5/6 and 7/8 (Table S1). Products were cut by SmaI (RIBA1) or KpnI/SmaI (RIBA2, RIBA3) restrictions and ligated into modified pCF203. [37] Resulting plasmids were transformed into A. tumefaciens pGV2260 and used to infiltrate Nicotiana benthamiana leaves [38]. Transient GFP expression was visualized by confocal laser scanning microscopy (CLSM) (λ_ex 488 nm, GFP λ_em 500–550 nm, chlorophyll λ_em 600–700nm).

Proteins were separated on 12% polyacrylamide gels as described [39]. Immune-detection of AtRIBA used a polyclonal anti-RIBA1 antiserum (Biogenes, Berlin, Germany) which was affinity-purified on nitrocellulose-bound antigen. Immune-detection used HRP-conjugated anti-rabbit antibody (Sigma-Aldrich, St. Louis, MO, USA), ECL Western Blotting Detection System (GE Healthcare, Uppsala, Sweden) and a Stella 3200 (raytest Isotopenmessgeräte GmbH, Straubenhardt, Germany).