Nicotiana benthamiana plants were grown at 21°C and 60–70% humidity in the Unifarm greenhouses of Wageningen University (Westerhof et al., 2014).

All chemicals were obtained from Sigma (Germany) if not indicated otherwise. Fluorescent NBD-lipids and pure lipids were purchased from Avanti (Alabama, United States). Antibodies purchased from Abcam (Cambridge, MA, United States).

The fluorescent ABP directed against α-galactosidases was produced based on previously described synthesis (Willems et al., 2014). Alpha-Galactopyranose configured cyclophellitol-aziridine was grafted with Cy5 as fluorophore, as described in detail in the Supplementary Material (Scheme 1). The newly synthesized probe was examined on reactivity with human recombinant α-GAL (Fabrazyme, Sanofi-Genzyme, Cambridge, MA, United States) (Supplementary Figures S1A,B).

Both GLA (ID: 2717) and NAGA (ID: 4668) were amplified from MegaMan Human Transcriptome cDNA library (Stratagene). The coding region of genes was amplified by Phusion^® HighFidelity PCR MasterMix (BioLabs) using the following primers: GLA sense: 5′-CTCATGAGTGCCAAGACCAACCTCTTCCTCTTCCTCATCTTCTCCCTCCTGCTCTCCCTC TCCTCCGCCCTGGACAATGGATTGGCAAG-3′, GLAantisense: 5′-CCCGTACGTTAAAG TAAGTCT-3′, NAGA sense: 5′-CTCATGAGTGCCAAGACCAACCTCTTCCTCTTCCTCATC TTCTCCCTCCTGCTCTCCCTCTCCTCCGCC-3′, NAGA antisense: 5′-CCGTACGTCACTGCT GGGACA-3′. The genes were amplified between BspHI and BsiWI restriction sites, lacking the sequence encoding for their native signal peptide. Instead, the native signal peptide of Arabidopsis thaliana chitinase was added at primer sequences. Insertion of mutations S188E and A191L in NAGA gene was achieved via overlap extension PCR. PCR reactions were performed using the following primers: NAGA mutant antisense: 5′-CCTTCATAGAGTGGCCACTCGCAGGAG AAGGC-3′, NAGA mutant sense: 5′-GCCTTCTCCTGCGAGTGGCCACTCTATGAAGG-3′. The elements were cloned into pGEM -T Easy Vector System (Promega), following electroporation of DH5α Escherichia coli cells. After clone selection and confirmation of sequences by sequencing (Macrogen), the complete open reading frames (ORFs) of GLA, NAGA, and NAGA^EL were inserted in plant expression vectors. A modified version of pMDC32 expression vector, named pHYG (Westerhof et al., 2012), was used as the main plant expression vector during all experiments. The vector was digested with Acc65I and NcoI, leaving compatible overhangs for ligation of the elements flanked between BspHI and BsiWI restriction sites. All constructs were under the control of cauliflower mosaic virus 35S constitutive promoter, with duplicate enhancer (d35S) and the nopaline synthase terminator (Tnos) derived from Agrobacterium tumefaciens. The pHYG vectors harboring the genes were used for transformation of Agrobacterium tumefaciens strain MOG101, following N. benthamiana plant leaf infiltrations.

Agrobacterium tumefaciens cultures were grown as previously described (Westerhof et al., 2012, 2014). The constructs were co-expressed or not with the tomato bushy stunt virus silencing inhibitor p19 (Voinnet et al., 2003), by mixing the bacterial cultures 1:1 following 1–2 h incubation at room temperature.

The inoculated bacterial cultures were used for infiltration of the two youngest leaves of 5–6 weeks old N. benthamiana plants, to ensure optimum protein expression, as previously described (Westerhof et al., 2014). Whole leaves and leaf disks (50 mm) were harvested at different days post-infiltration (dpi), snap-frozen and homogenized in liquid nitrogen. Homogenization was performed with a tissue lyser (AKA Qiagen TissueLyser II) at 30 rounds/sec for 1 min. Extraction of total soluble proteins of samples was achieved using 2–3 ml per leaf gram of extraction buffer (30 mM citrate-phosphate buffer, pH 6.3, containing 2% (w/v) polyvinylpolypyrrolidone, 0.1% (v/v) Tween 20, 0.5 M NaCl, and protease inhibitor by Roche, EDTA free) following homogenization at same tissue lyser program. When whole leaves were used, homogenization was performed by grinding with a mortar and pestle. Samples were next centrifuged for 15 min at 15000 g (F13S 14 × 50 cy, Sorvall rotor), at 4°C. The supernatant was collected and used for further analysis or it was stored at -80°C after desalting at G25-Sephadex column.

As first step in the purification of all enzymes, a 5 ml column of Concanavalin-A-Sepharose (GE healthcare Bio-Sciences) was used. The column was equilibrated with 40 ml washing buffer (0.1 M sodium acetate, 0.1 M NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 1 mM MnCl₂, pH 6.0). Then, protein sample (lysate) was applied to the column, 1:1 diluted in washing buffer (0.5 ml/min loading conditions). For α-GAL 49 ml of protein lysate was applied on the column, for α-NAGAL 38 ml and α-NAGAL^EL 25 ml. Proteins were eluted with 30 ml elution buffer (washing buffer supplemented with 0.9 M methyl-α-mannoside and 0.9 M methyl-α-glucoside). Ten fractions of 1 ml showing highest levels of enzymatic activity were pooled (post-ConA). Next, for α-GAL purification, 10 ml of post-ConA was loaded on an ultrafiltration device (Centricon Plus-20, 15 ml with 10 kDa molecular cutoff, Millipore, Bedford, MA, United States) to remove methyl-monosaccharides and buffer exchanged to 20 mM sodium acetate, pH 5.0, until obtaining a volume of 3 ml. Dialysis of the sample was continued in 20 mM sodium acetate buffer, pH 5.0, 4°C, overnight. A HiTrap SP HP column (1 ml; GE healthcare Bio-Sciences) was equilibrated with 15 ml binding buffer: 20 mM sodium acetate buffer, pH 5.0. The sample, 3 ml, was applied on the column which was extensively washed with 10 ml binding buffer afterward. Then, protein was eluted using a gradient of 0–300 mM NaCl, 15 ml. Enzyme eluted in 0.5 ml fractions 2–4, which were pooled. For wild type α-NAGAL and modified α-NAGAL^EL, post-ConA, 2–3 ml, was buffer exchanged via ultrafiltration to 10 mM potassium phosphate buffer, KPi (pH 6.3) with 10 mM NaCl. Next, sample was dialyzed in the same buffer to further remove any methyl-monosaccharides. After this, sample was applied on a HiTrap Q HP column (1 ml; GE healthcare Bio-Sciences). Anion exchange chromatography was performed using a gradient of 10–600 mM NaCl in 10 mM KPi (pH 6.3). Enzymes eluted in 0.5 ml fractions 2–4, which were pooled. As final purification step for all enzymes gel filtration was used. A Superdex^TM 200 Increase 10/300 GL (GE healthcare Bio-Sciences) column was prior equilibrated with 50 ml of 20 mM sodium acetate buffer with 150 mM NaCl, pH 5.0. Samples were applied and chromatography performed at a flow 0.75 ml/min. 0.2 ml fractions were collected and analyzed on enzymatic activity. The final material was snap frozen in liquid nitrogen and stored at -80°C until further use.

Vectors for human cell line HEK293 were produced as follows: the complete OPFs of human α-GAL, α-NAGAL, and α-NAGAL^EL were amplified from previous pGEMt easy vectors harboring the genes with their native signal peptide, via PFX50 high fidelity DNA polymerase (Invitrogen), using the primers below. GLA sense: 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTACCACCATGCA GCTGAGGAA CCCAGA-3′, GLA antisense: 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTC TTAAAGTAAGTCTTTTAATGACATC-3′, NAGA sense:5′-GGGGACAAGT TTGTACA AAAA AGCAGGCTACCACCATGCTGCTGAAGACAGTGCTC-3′, NAGA antisense:5′-GGGGACCA CTTTGTACAAGAAAGCTGGGTCTCAGCTGGGACATCTCCAG-3′. The elements were cloned via Gateway cloning into the donor vector pDONR221 (BP reaction), following transformation of DH5α E. coli strain. Clones were collected upon ampicillin selection and their sequences were confirmed. Next, they were cloned into the destination vector, pcdna-DEST40-zeo (LP reaction), forming the final construct to transfect HEK293 cells. The HEK293 cells were cultured in six well plates in 2.5 ml Dulbecco’s Modified Eagles Medium (DMEM, Sigma) supplemented with 10% fetal calf serum, 1% penicillin/streptomycin and glutamax at 37°C, at 7% CO₂. Three μg of pcdna-DEST40-zeo DNA was transfected into cultured HEK293 cells using polyethylenimine, primarily mixed with serum free DMEM medium. After 2 days of cell culture, 200 μg/μl of zeocin was added to the culture media for selection of clones harboring the gene of interest. The cells grew for 3 weeks at 37°C, at 7% CO₂ in the presence of antibiotics and lysed in 20 mM KPi buffer (pH 6.5), 0.1% Triton with additional protease inhibitor by Roche. The cells were stored at -150°C in DMEM, 20% fetal calf serum, 10% DMSO until further use.

Both α-galactosidase A and α-N-acetyl-galactosaminidase activities of samples were examined with corresponding 4-methylumbelliferyl (4MU) substrates. For α-galactosidase activity measurement samples were incubated for 1 h at 37°C with a final concentration of 1.2 mg/ml 4MU-α-D-galactosylpyranoside (4MU-α-GAL) in 150 mM citrate-phosphate buffer pH 4.6, supplemented with 0.1% (w/v) BSA and released 4MU was quantified as described earlier (Blom et al., 2003).

Activity of plant produced recombinant enzymes and Fabrazyme toward 5 μM of NBD-C12-Gb3 substrate was measured for 3 h at 37°C in 150 mM citrate-phosphate buffer (pH 4.6) containing 0.05% (v/v) Triton X-100 and 0.2% (w/v) sodium taurocholate, pH 4.6. Lipids were extracted with the Bligh and Dyer procedure, applied to HPTLC. The plate was scanned for fluorescent lipids with a Typhoon FLA 9500. Protein concentrations in assay were for Fabrazyme 3.8 μg/ml, α-GAL 3.2 μg/ml, α-NAGAL 40 μg/ml, and α-NAGAL^EL 16 μg/ml. Activity of recombinant enzymes toward natural C18: Gb3 was measured using 500 pmol of lipid in (100 μl total volume) 150 mM citrate-phosphate buffer with 0.05% (v/v) Triton X-100 and 0.2% (w/v) sodium taurocholate for 3 h at 37°C. The pH was 4.2 for α-GALs and 5.2 for α-NAGALs. After incubation, neutral lipids were extracted by the Folch method. Deacetylation of lipids was performed as described earlier (Ferraz et al., 2016b). After deacetylation, samples were dried and further cleaned by adding 1:1 butanol: water. The butanol phase was collected, dried and lipids were dissolved in 200 μl methanol, ready for LC–MS/MS injection. The conversion of Gb3 to lactosylceramide (Lac Cer) was measured in LC–MS/MS using C17-Gb3 and C17-Lac Cer as controls and C17-Sphinganine as internal standard. The same assay conditions were used when applying natural Lyso-Gb3 as substrate. Lipids were extracted according to Bligh and Dyer and subjected to LC–MS/MS injection as above. Isotope ¹³C-Lyso-Gb3 was used as an internal standard for monitoring the conversion of Lyso-Gb3 into Lyso-Lac Cer. Protein concentrations in assays were for Fabrazyme 1.9 μg/ml, α-GAL 1.6 μg/ml, α-NAGAL 20 μg/ml, and α-NAGAL^EL 8 μg/ml.

K_m, V_max, and k_cat values were determined using 4MU substrates. Reactions were performed for 1 h or 30 min at 37°C at 10 different 4MU-α-GAL and 4MU-α-NAGAL concentrations in 150 mM citrate-phosphate buffer pH 4.6 supplemented with 0.1% (w/v) BSA. The 4MU-α-GAL concentrations in the assays ranged from 0.074 to 4.72 mM; for 4MU-α-NAGAL from 0.022 to 0.91 mM. Protein concentrations in the assays were: for α-GAL 7 ng/ml, Fabrazyme 2 ng/ml, α-NAGAL 16 ng/ml, when using 4MU-α-NAGAL substrate and 100 ng/ml, when using 4MU-α-GAL substrate, and for α-NAGAL^EL 16 ng/ml. Parameters were calculated using GraphPad Prism6.

Proteins were deglycosylated with EndoH and PNGase F (Wilbers et al., 2016). MALDI-TOF analysis performed as previously described (Wilbers et al., 2016).

To examine purity and molecular mass of recombinant α-galactosidases, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS- PAGE) (10% polyacrylamide) was used (Witte et al., 2010). Samples were run under reduced conditions. After SDS-PAGE analysis, gels were stained with Coomassie Brilliant Blue staining or silver stain using the PhastSystem system (GE Healthcare).

The α-GAL Cy5 ABP was diluted in 150 mM citrate-phosphate buffer, pH 5 and applied to lysates or pure protein preparations at 0.5 or 0.25 μM final concentration, respectively. The samples were incubated for 30 min, at 37°C and the reaction was stopped via the addition of 4x gel loading buffer (containing β-mercaptoethanol). The samples were then boiled for 5 min at 95°C and loaded on 10% SDS-PAGE gels. Subsequently, western blotting and fluorescence scanning was performed exactly as described earlier (Witte et al., 2010; Willems et al., 2014).

Protein concentrations were determined with a Micro BCA Protein Assay Reagent Kit (PIERCE), and bovine serum albumin was used as a standard, according to the supplier’s protocol. In addition, pure protein concentrations were measured in Nano Drop 2000c (Thermo Scientific) via adjusting the molecular weight and extinction coefficient parameters.

Samples were incubated for 1 h 4°C while rolling, with 2 μl Ab⁺ FD serum or normal serum of an FD patient and α-galactosidase activity was measured with 4MU-α-GAL and 4MU-α-NAGAL substrate (Linthorst et al., 2004).

To serum from two different FD patients pre-ERT, recombinant enzymes (8 μmol artificial substrate hydrolysis/h/ml) were added and incubated overnight at 37°C. Then, lipids were extracted as described before and applied to LC–MS/MS.

We first established the optimal conditions for production of α-galactosidases α-GAL, α-NAGAL, and α-NAGAL^EL in a transient expression system in N. benthamiana. In the constructs the sequences of mature α-galactosidases were preceded by that of the Arabidopsis thaliana chitinase signal peptide. Agrobacterium tumefaciens infiltrated leaves of 5–6 weeks old N. benthamiana plants were collected at 1, 3, and 5 days post-infiltration (dpi), lysates were prepared and enzymatic activities of α-galactosidases were measured using the respective 4MU substrate (Figure 1A). For each enzyme, an optimal yield was observed at 5 dpi, while co-expressing the p19 silencing suppressor of tomato bushy stunt virus. This time point was used throughout all later experiments for harvesting. The amounts of active recombinant α-GAL, α-NAGAL and α-NAGAL^EL at 5 dpi was 71, 5 and 7 nmol/h/μg total protein, respectively, as measured with 4MU-α-GAL substrate. No 4MU-α-NAGAL activity was detected for the α-NAGAL^EL enzyme, whereas the wild type reached 33 nmol/h/μg at 5 dpi (Figure 1A). The amount of enzyme activity was proportional to the amount of active enzyme molecules detected with the fluorescent ABP (Figure 1B) and that detected with western blotting using anti-α-GAL and α-NAGAL antibodies (Figure 1C). The molecular weight of α-GAL was about 44 kDa. Those of the two α-NAGALs were about 49 kDa, the higher mass being due to the known presence of one extra N-glycan (Clark and Garman, 2009).

Recombinant α-galactosidases were purified from lysates of N. benthamiana leaves by sequential chromatography during which the presence of enzymes was monitored by measurement of enzymatic activities with 4MU-substrates. As generic first step we used Concanavalin A-Sepharose chromatography exploiting the presence of mannose-containing N-glycans in the recombinant α-galactosidases ensuring high affinity binding to the immobilized lectin (Yasuda et al., 2004). The recovery in this step was high (>40%) for each enzyme and led to >50-fold purification (Table 1). The next step in purification was ion exchange chromatography performed at pH 5.0 for α-GAL and at pH 6.3 for α-NAGAL and α-NAGAL^EL as described in experimental procedures. This led for each enzyme to considerable further purification with acceptable recovery (Table 1). The final purification step for all enzymes was gel filtration, resulting in an apparently pure protein as judged by silver staining of protein resolved by SDS-PAGE (Figure 2A). About 1 μg α-GAL, α-NAGAL, and α-NAGAL^EL was purified from 1 mg of soluble lysate protein.

The apparent molecular weights of all three α-galactosidases as determined by gel filtration coincided with dimers (113 kDa for α-GAL, 134 kDa for α-NAGAL, and 132 kDa for α-NAGAL^EL) (Supplementary Figures S2A,B). Of note, the specific activity of plant-produced α-GAL (2.35 mmol/h/mg) was quite comparable to that of Fabrazyme (2.18 mmol/h/mg) (Table 2) (Tajima et al., 2009). After purification, further enzyme characterization was performed. Equal μg amounts of pure enzymes were incubated with fluorescent ABP followed by SDS-PAGE and fluorescence scanning analysis (Figure 2B). The intensity of the fluorescently labeled bands was the same for the three enzymes, revealing similar amounts of active enzyme molecules in each purified protein preparation. Western blotting using anti-α-GAL and anti-α-NAGAL antibodies showed that α-NAGAL^EL was not recognized by anti-α-GAL antibody (Figure 2B).

Next, the N-glycan composition of the purified α-galactosidases was examined using deglycosylation with Endo H and PNGase F endoglycosidases following SDS-PAGE analysis. Fabrazyme, produced in CHO cells, was used as a control. Consistent with literature reports, PNGase F led to conversion of Fabrazyme to a molecular mass of 39 kDa (Figure 3A). In the case of plant produced enzymes the reductions in molecular mass with EndoH or PNGase F were less pronounced (Figure 3A). This difference can be attributed to the presence of α1-3 core fucose residues, attached to most of the N-acetyl-glucosamine backbones (GlcNAc), inhibiting digestion by the endoglycosidases. The higher molecular weight of Fabrazyme compared to the plant produced enzymes could be due to the different N-glycan composition of this enzyme, mainly consisting of complex N-glycan structures, with additional mannose 6 phosphate residues (Sakuraba et al., 2005). To get more insight in the N-glycan composition of the plant produced enzymes, all N-linked glycans were released with PNGase A (not inhibited by the presence of α1-3 fucose) and subjected to MALDI-TOF MS (Figures 3B–D). The most prominent N-glycan type of α-GAL recombinant protein was paucimannose structure (Man₃GlcNAc₂) with α1-3 fucose attached on the core GlcNAc and β1-2 xylose on the first mannose. Conceivably, the enzyme was secreted to the apoplastic fluid and hexosaminidases cleave the GlcNAc molecules yielding the pausimanosidic structure. On the other hand, both α-NAGALs have oligo-mannose type N-glycans (Man₇GlcNAc₂) as the most abundant N-glycan type, suggesting their retention in Golgi and /or ER compartments.

We next examined the cross reactivity of recombinant α-galactosidases toward neutralizing antibodies in the serum of an Ab-positive male FD patient receiving ERT for 6 years (T = 1). As a control, serum was used of the same patient before ERT (T = 0) and still lacking antibodies against therapeutic α-GAL. Inhibition of enzymatic activity by presence of Ab-positive serum was determined (Figure 4A). Activity of plant produced α-GAL and Fabrazyme was clearly inhibited, but activities of α-NAGALs, as determined with 4MU-α-GAL and/or 4MU-α-NAGAL substrates were not significantly influenced. The stability of the plant-produced galactosidases and Fabrazyme in human plasma was next studied by incubating enzymes at 37°C for different time periods and detection of residual enzymatic 4MU-α-GAL activity (Figure 4B). Alpha-NAGAL^EL, like α-NAGAL, was more stable than both α-GALs (Figure 4B). The finding was made several times with independent incubations.

We next studied the substrate specificity of the various recombinant α-galactosidases. First, activity was measured toward artificial 4MU-α-GAL and 4MU-α-NAGAL substrates. The amino acid substitutions Ser188Glu and Ala191Leu in the catalytic pocket of α-NAGAL (α-NAGAL^EL) abolished activity toward 4MU-α-NAGAL as earlier reported (Tajima et al., 2009; Tomasic et al., 2010) (Table 2). The noted increase in hydrolytic specific activity toward 4MU-α-GAL (about a doubling) was less prominent as earlier reported by Tajima et al. (2009). The kinetic parameters kcat and Km determined with 4MU-α-GAL as substrate were improved by the amino acid substitutions in α-NAGAL^EL (Table 3). The plant produced α-GAL showed comparable kinetic parameters to Fabrazyme (Table 3) (Hopkin et al., 2003; Lee et al., 2003). To determine whether production in N. benthamiana leaves influences kinetic features, the various recombinant enzymes were also generated in human HEK293 cells as described in experimental procedures (Figure 5A). ABP-labeling showed that similar proteins were produced in N. benthamiana and HEK293 cells (Figure 5B). The specific activity toward 4MU-α-GAL was also comparable, being relatively low for α-NAGAL and increased (threefold) by the amino substitutions in α-NAGAL^EL (Figure 5C).

The activity of plant produced galactosidases toward lipid substrates was next determined. NBD-Gb3 was found to be degraded by Fabrazyme and plant-produced α-GAL on a par (Figure 6A). Of note, significant degradation of NBD-Gb3 to NBD-Lac Cer by α-NAGAL and α-NAGAL^EL was detected (Figure 6A). To further substantiate the findings, we incubated wild type and mutant α-NAGALs with NBD-Gb3 in the absence or presence of 2.5 mM N-acetylgalactosamine, a potent inhibitor of α-NAGAL activity (Figure 6B). The wild type α-NAGAL was inhibited, resulting in loss of NBD-Lac Cer formation, whereas α-NAGAL^EL was not. Next, activity toward 5 μM of natural Gb3 and its deacetylated form Lyso-Gb3, was determined using detection of metabolites with LC–MS/MS. The analysis revealed a 2–5-fold increased activity of α-NAGAL^EL compared to α-NAGAL with both substrates (Figure 6C). The pH optimum was pH 4.2 for α-GALs and pH 5.2 for both α-NAGALs (Supplementary Figure S3).

Finally, serum of male FD patients, containing high amount of Lyso-Gb3, was incubated overnight with recombinant galactosidases (at a concentration of 8 μmol/h/ml activity mimicking levels reached during ERT). Of note, the rate of degradation of Lyso-Gb3 in the serum sample was highest with α-NAGAL^EL (Figure 6D). Likely this reflects the greater stability and ongoing activity of α-NAGAL^EL at neutral pH as compared to that of α-GALs.