Enzymatic elaboration of oxime-linked glycoconjugates in solution and on liposomes

Oxime formation is a convenient one-step method for ligating reducing sugars to surfaces, producing a mixture of closed ring α- and β-anomers along with open-chain (E)- and (Z)-isomers. Here we show that despite existing as a mixture of isomers, N-acetylglucosamine (GlcNAc) oximes can still be substrates for β(1,4)-galactosyltransferase (β4GalT1). β4GalT1 catalysed the galactosylation of GlcNAc oximes by a galactose donor (UDP-Gal) both in solution and in situ on the surface of liposomes, with conversions up to 60% in solution and ca. 15–20% at the liposome surface. It is proposed that the β-anomer is consumed preferentially but long reaction times allow this isomer to be replenished by equilibration from the remaining isomers. Adding further enzymes gave more complex oligosaccharides, with a combination of α-1,3-fucosyltransferase, β4GalT1 and the corresponding sugar donors providing Lewis X coated liposomes. However, sialylation using T. cruzi trans-sialidase and sialyllactose provided only very small amounts of sialyl Lewis X (sLex) capped lipid. These observations show that combining oxime formation with enzymatic elaboration will be a useful method for the high-throughput surface modification of drug delivery vehicles, such as liposomes, with cell-targeting oligosaccharides.


Materials
Reagents were purchased from Sigma-Aldrich Co. Ltd., Dorset, UK with some exceptions. DMPC and DOPC were purchased from Avanti Polar Lipids Inc., Alabama, USA. Lissamine™ rhodamine DHPE and BioDesignDialysis Tubing ® were purchased from ThermoFisher Scientific Inc., USA. All lectins were purchased from Vector Labs. PD-10 gel permeation chromatography (GPC) columns were purchased from GE Healthcare. Vivaspin ® 500 was obtained from Sartorius Stedin Biotech.

Instrumentation
NMR measurements were made on either a Bruker Ultrashield 400 or a Bruker 500 MHz spectrometer at RT (295 K) in a suitable deuterated solvent (typically CDCl3, D2O or CD3OD). Chemical shifts () were measured in parts per million (ppm) and coupling constants (J) in Hertz (Hz). 1 H-and 13 C-NMR spectroscopy are referenced relative to the residual solvent peaks. Multiplicities are indicated as singlet (s) doublet (d), doublet of doublets (dd), triplet (t), quartet (q), and multiplet (m). Spectra assignments were made by 2D COSY, HMQC, HSQC, DEPT-90 and DEPT-135 measurements to aid peak identification. NMR data were processed and analysed using MestreNova.
Electrospray mass spectrometry was performed on an Agilent 6560 IMS Qtof MS with an Agilent 1290 Infinity 2 series LC. High resolution mass spectrometry was performed on a Water Q-TOF micro with an ES+/-ion source. For GC-MS experiments an Agilent 1200 series 6510 Q-TOF LC-MS was used. Reversed-phase HPLC purification was performed either on an Agilent 1100 series system, LC system with either a G1315B diode-array detector or a G1365B multi-wavelength detector. LC−MS measurements were made using an Agilent 1100 LC/MSD Ion Trap System with an Electrospray Source and positive ionisation. The following columns were used: Agilent Zorbax Eclipse XDB-C8 (150 × 4.4 mm, 5 μm; used for the quantification of galactosylation of 2 by LC-MS) fitted with a guard column, Macherey-Nagel Nucleosil C18 (150 × 4.6 mm, 5 μm, used for the transformation of GlcNAc-PNP) fitted with a guard column, Agilent Eclipse XDB-C8 (250 × 9.4 mm, 5 μm; used to purify 2 and 13) and Agilent Eclipse XDB-C18 (250 × 9.4 mm, 5 μm; used to purify other compounds).
FTIR spectra were recorded on a Bruker Alpha-P instrument and analysed using OPUS 6.5 software between 4000 and 500 cm -1 . An ATR platinum diamond detector was used. The acquisition was over 64 scans at a resolution of 2 cm -1 . Absorbance and fluorescence spectra were measured on an Infinite® 200 PRO NanoQuant plate reader from TECAN and a CLARIOstar plate reader. UV-visible spectroscopy S7 was measured using a Jasco V-660 spectrophotometer at 400 nm/min in a 200-800 nm range. Elemental analysis was performed on a Thermo Scientific FLASH 2000 series CHNS/O analyser. DLS were performed on a Malvern Zetasizer Nano ZSP 633-nm laser at 25 °C. The  potential was measured on a Malvern Zetasizer Nano ZSP at 25 °C using the diffusion barrier method. 1 Fluorescence images were acquired using a Zeiss Axio Imager A1 fluorescence microscope with a Canon Powershot G6 digital camera attached.

Synthesis of N-alkoxyamines 1 and 3
Scheme S1: Synthesis of lipid 1 and tether 3 To anchor these synthetic oxime-glycolipids to the liposomal membrane, a cholesteryl anchor was selected. 2a A reactive N-alkoxyamine terminus was linked to the cholesteryl unit through a triethyleneglycol spacer, which was hoped to facilitate access of enzymes and lectins to the ligated sugars when the lipid is embedded in a liposomal membrane. The trifluoroacetate salt of Nalkoxyamine lipid 1, which combines these features, was synthesised with 20% overall yield in three steps from commercially available reagents (Scheme S1).
Based on a methodology described by Thakar et al., 2b the tether 3 was synthesised over two reaction steps using similar synthetic methodology.
The synthesis of the isomers of 6 was accomplished by mixing 1 eq. of O-ethylhydroxylamine hydrochloride with 1.5 eq. of GlcNAc in methanol (10 mL). The solution stirred under reflux conditions overnight (65 C, N2 atmosphere). The solvent was removed under reduced pressure and the crude mixture was purified by HPLC on a 60 min gradient ranging from 5% to 95% MeCN in water with a flow rate of 1.0 mL/min. The sample was collected and freeze-dried to afford the product as oil.   Data matched that reported previously. 5

Isomer interconversion from 6 enriched in E-isomer
An alternative purification of the reaction mixture of glycoconjugates 6 by HPLC afforded a fraction with a large proportion of open chain (E)-oxime (E/Z/α/β, 93:4:0: 3). Assuming that open-chain (E)oxime is not a substrate for the β4GalT1 enzyme whereas the β-anomer is, access to this enriched fraction allowed assessment of whether equilibration between isomers allows open-chain oximes to form substrate for this enzyme.
To determine the amount of time required to produce the ring-closed form in significant amounts, the selected fraction (open-chain (E)-oxime) was dissolved in buffered D2O (50 mM MES buffer, pH 7) and monitored by 1 H-NMR spectroscopy over time at 37 C ( Figure S1a). After six days, the (E)-oxime was still the predominant isomer (61%) in the mixture but the (Z)-oxime and cyclic β-anomer were detected in significant amounts. Based on the integration of the respective doublets, the (Z)-oxime S13 and the cyclic β-anomer proportions were 21% and 18%, respectively. The -pyranoside isomer was not observed during the course of the experiment (i.e. E/Z/α/β was 61:21:0:18). The 2D COSY spectrum of the mixture before and after 6 days shows the increased intensity of the crosspeak corresponding to the β-anomeric doublet ( Figure S1b,c). These observations suggest that the (E)oxime isomer can interconvert into the cyclic β-anomer to re-establish equilibrium. Figure S1. Equilibration of glycoconjugate 6 (fraction predominantly (E)-oxime) over time monitored by NMR spectroscopy (buffered D2O, 295 K, 400 MHz). (A) Interconversion of acyclic/cyclic adducts is achieved over 6 days. (B) 2D COSY NMR spectrum (buffered D2O, 295 K, 400 MHz) before and (C) after 6 days shows a cross-peak corresponding to the anomeric proton of the -oxime isomer. S14
The GlcNAc/N-ethoxyamine adduct 6 (0.02 mmol, 4.6 mg) was dissolved in MES buffer (1 mL, 50 mM, pH 7.0) and mixed with UDP-Gal (0.03 mmol, 18.3 mg), MnCl2 (3 μL of a 1.0 M solution in water) and β4GalT1 (17.19 μL of a 0.54 mg/mL solution). The mixture was vortex mixed and incubated overnight at 37 °C. The final product was purified by HPLC on a 40 min gradient ranging from 5% to 100% MeCN in water with a flow rate of 1.0 mL/min. The sample was collected and freeze-dried to afford a 1:1 mixture of 6 and the product 11 (3 mg in total, 35% mass recovery).

Enzymatic galactosylation of 6 enriched in E-isomer
A portion of 6 enriched in the open-chain (E)-oxime (93%, see Section 3.2 above) was then subjected to standard β4GalT1 enzymatic transformation conditions for an extended period of 6 days at 37 C. The reaction mixture after transformation was then purified by HPLC. The 1 H-NMR spectrum of the purified mixture showed a doublet at 4.4 ppm that corresponds to the proton attached to the anomeric carbon of Gal with a -1,4 linkage ( Figure S2a). The relative integration of this proton environment compared to the CH3 of the N-ethoxyamine portion of the molecule (set to 3.00) was S15 0.60 (60% extent of conversion into the LacNAc adduct). Successful conversion to the LacNAc product was confirmed by the observation of the product peak in the positive ion liquid chromatography-mass spectrometry (LC-MS); [11 + Na] + ( Figure S2b). From the isomer interconversion experiment on 6 (Section 3.2), the proportion of β-anomer at equilibrium in buffered D2O was 0.18 ( Figure S1). This means that up to 18% of GlcNAc should be available to conjugate with Gal after 144 h; however, 60% of LacNAc adduct was, in fact, produced. This difference may be related to the subsequent conjugation of Gal to GlcNAc adducts by the 4GalT1 enzyme during the equilibration period. This means that the presence of the enzyme may pull the mixture out of equilibrium, leading to fresh generation of more β-adduct.
Although a reactant mixture of 6 consisting of pure -pyranoside could not be obtained to confirm it was the best (or only) substrate for 4GalT1, these observations suggest that the -anomer of the adduct is the best substrate for the enzyme.  The synthesis of compound 4 was accomplished by mixing 1 eq. of 3 with 2 eq. of GlcNAc in methanol (10 mL). The solution was stirred under reflux conditions overnight (65 C, N2 atmosphere). The solvent was removed under reduced pressure and the crude mixture was purified by HPLC on a 60 min gradient ranging from 5% to 95% MeCN in water with a flow rate of 1.0 mL/min. The sample was collected and freeze-dried to afford the product as an oil.

Monomeric adducts with tether 3
General experimental procedure for adducts 7- 10 The synthesis of compounds 7-10 was accomplished by mixing 1 eq. of 3 with 0.5 eq. of the respective reducing sugar (lactose, glucose, galactose or mannose) in methanol (10 mL). The solution was stirred under reflux conditions overnight (65 C, N2 atmosphere). The solvent was removed under reduced pressure and the crude mixture was purified by HPLC on a 60 min gradient ranging from 5% to 95% MeCN in water with a flow rate of 1.0 mL/min. The sample was collected and freeze-dried to afford the product as oil.

Mixtures of monomeric and dimeric adducts with tether 3
General experimental procedure for adducts S5-S22 The synthesis of compounds S5-S22 was accomplished by mixing 1 eq. of 3 with 0.5 eq. of the respective reducing sugar (N-acetylglucosamine, L-fucose, glucosamine, glucose-6-phosphate, 2deoxyglucose, N-acetyllactosamine or 3'sialyllactose) in methanol (10 mL). The solution stirred under reflux conditions overnight (65 C, N2 atmosphere). The solvent was removed under reduced pressure and the crude mixture was purified by HPLC on a 60 min gradient ranging from 5% to 95% MeCN in water with a flow rate of 1.0 mL/min. The sample was collected and freeze-dried to afford the product as oil.       Lipid 1 (1 eq.) and GlcNAc (2 eq.) were dissolved in MeOH (10 mL) under reflux conditions (N2 atmosphere, 65 C, overnight). The next day, the solvent was removed under reduced pressure and the crude mixture was purified by silica column chromatography (DCM/MeCN/MeOH 5:5:1 changed to MeOH to recover the product). TLC Rf 0 (DCM/MeCN/MeOH 5:5:1).  Lipid 1 (1 eq.) and LacNAc (2 eq.) were dissolved in MeOH (10 mL) under reflux conditions (N2 atmosphere, 65 C, overnight). The next day, the solvent was removed under reduced pressure and the crude mixture was purified by silica column chromatography (DCM/MeCN/MeOH 5:5:1 changed to MeOH to recover the product). TLC Rf 0 (DCM/MeCN/MeOH 5:5:1).

Overview
Given the successful use of β4GalT1 to transform DMPC/2 liposomes and the failure to directly condense 3'SL with lipid 1, it was hoped that the use of multiple glycosyltransferases would provide more complex glycolipids. Applying multienzyme synthetic sequences to synthetic sugars has been reported to give difficult-to-access bioactive oligosaccharides. 7 Such complex oligosaccharides on the surface of drug-loaded liposomes may produce highly specific targeting of particular cell types.
Combinations of three enzymes with glycosyltransferase activity were used, namely β4GalT1, TcTS and α-1,3-fucosyltransferase (α1,3-FucT). Different combinations of these enzymes could provide four oligosaccharides (Scheme S3). The methodology was first validated using soluble GlcNAc-PNP S18, as the p-nitrophenyl (PNP) chromophore permits reaction monitoring by HPLC and thereby provides quantitative data on the reaction timecourse.
The conversion of S18 to S20 with β4GalT1/UDP-Gal and TcTS/3'-SL (Scheme S3, right) using a "onepot" procedure has already been reported, giving up to 70% of S20 within an hour with concomitant 25% conversion to S19. 8 Product S20 has a Neu5Ac(2-3)Gal(β1-4)GlcNAc sequence that is similar to the Neu5Ac(2-3)Gal(β1-4)Glc sequence found on GM3 glycolipid. 9 A combination of β4GalT1 and α-1,3-FucT may elaborate a GlcNAc terminus into Lewis X (Le x ), which mediates several important cellular functions in neural development and in immunity. 10 This would also be a new combination of enzymes for "one pot" liposome modification methodology (Scheme S3, left). The glycosyl donor substrate for α1,3-FucT is GDP-fucose. Although commercially available, GDPfucose is expensive, so FKP enzyme (L-fucokinase/GDP-fucose pyrophosphorylase) was used to S34 convert L-fucose into GDP-Fuc, with ATP and GTP added as donors. 11 The enzymatic synthesis of Le x -PNP was investigated by both sequential and "one-pot" approach using β4GalT1 and α1,3-FucT enzymes (Scheme S4). The sequential approach comprised an initial incubation of GlcNAc-PNP with β4GalT1 enzyme. HPLC analysis revealed a (92 ± 2)% conversion after 5 h, which reached a maximum (96 ± 1)% conversion after 16 h. After this time, the solution was incubated with α1,3-FucT enzyme. After only 3 h, the Le x -PNP product S21 was obtained with approximately complete conversion. However, a 16 h "one-pot" reaction, where GlcNAc-PNP S18 is incubated with both β4GalT1 and α1,3-FucT enzymes, showed a maximum conversion onto Le x -PNP of about (75 ± 10)% with GlcNAc-PNP S18 was still present ((29 ± 6)%). This observation suggest the activity of β4GalT1 is compromised when mixed with α1,3-FucT and respective reagents, although the absence of intermediate S19 suggests that α1,3-FucT activity is not significantly affected.
Sialylation of S21 by TcTS might provide sialyl Lewis X (sLe x ), a key cell surface saccharide that plays key roles in inflammatory and immune processes. 12 In the natural biosynthetic pathway, 13 sialylation generally occurs before fucosylation, 14 so the synthesis of S22 was investigated using either a sequential or a "one-pot" mix of β4GalT1 and TcTs enzymes, which was followed by reaction with α1,3-FucT (Scheme S3, bottom). The "one-pot"/α1,3-FucT approach involved a first step incubating with both 4GalT1 and TcTS 15 over 16 h, then the final step catalysed by α1,3-FucT for 16 h ( Figure  S3). Analysis after the first "one-pot" step showed the sialylation step was less effective with this batch of TcTS/3'-SL; after 16 h the relative proportions were 26% product S20, 16 59% intermediate S19 and 5% starting GlcNAc-PNP S18. Then the final step, catalysed by α1,3-FucT, was carried out for 16 h to give final proportions of sLe x -PNP S22 and Le x -PNP S21 of (30 ± 1)% and (67 ± 1)%, respectively. These proportions show that fucosylation of S19 and S20 was very effective. The sequential approach comprised an initial incubation of GlcNAc-PNP with β4GalT1 enzyme, then after 16 h the solution was incubated with TcTS for a further 16 h. Subsequently, this solution was incubated with α1,3-FucT (16 h). After 2 h, the proportions of S22 (34 ± 13)% and S21 (65 ± 13%) were at their highest values. These two approaches gave no significant difference in the proportions of oligosaccharides obtained, so the shorter "one-pot"/α1,3-FucT methodology was subsequently applied to DMPC/2 liposomes. Figure S3: (a) Enzymatic transformation of a mixture of S19 and S20 (~60% and ~30% respectively, formed by the "one-pot" enzymatic approach) using α1,3-FucT/FKP/Fuc, affording S21 and S22. (b) The reaction was monitored over 16 h by HPLC and the mass of the compounds in each peak was determined by LC-MS. (c) Time course for each compound. Each point represents the mean ± SD (n = 3).

Sequential and "one-pot" formation of LeX-PNP
Two approaches were investigated to afford Le X -PNP S21, a sequential and a 'one-pot' enzymatic reaction (Scheme S4).

S36
Scheme S4: Synthesis of PNP-Le x S21 by an enzymatic methodology. Two approaches were investigated: the sequential reaction using (a) 4GalT1 and α1,3-FucT enzymes; and the (c) 'one-pot' enzymatic transformation using 4GalT1/ α1,3-FucT enzymes in the same reaction. Sugar symbols according to SNFG.
The sequential enzymatic transformation of GlcNAc-PNP S18 was carried out in two steps at 37 C to afford Le x -PNP S21. The first step of the reaction involved mixing GlcNAc-PNP S18 (200 µM) with 4GalT1 (0.54 mg/mL)/UDP-Gal (10 mM) to catalyse the transfer of Gal residues and afford S19 ( Figure  S3).
The second step of the sequential reaction involved using α1,3-FucT (3.7 mg/mL)/FKP (8 mg/mL)/Fuc (10 mM) to transform LacNAc-PNP S19 into Le x -PNP S21. The two expected peaks were observed in the HPLC traces ( Figure S4a), namely Le x -PNP S21 (11.6 min) and LacNAc-PNP S19 (12.6 min). Quantification of all compounds was carried out ( Figure S4b), which shows an instant decrease of LacNAc-PNP S19 that coincides with an increase in Le x -PNP S21. After 2 h of reaction, all LacNAc-PNP S19 is consumed and Lex-PNP S18 was produced with 100% conversion.
The fucosylation of LacNAc-PNP S19 using FucT was surprisingly fast and it only took a total of 2 h for the full consumption of the substrate. Therefore, the fucosylation step was repeated and the reaction analysed using shorter time points. In addition, the order that the enzymes (FKP and FucT) were added into the solution was also assessed. Specifically, FKP addition at the same time as α1,3-FucT or 20 min before α1,3-FucT enzyme (after all the other reagents) was investigated. The results obtained ( Figure  S5) suggested that the order that each enzyme is added into the solution is crucial to the outcome of the reaction. When FKP enzyme is added into the solution 20 min before the α1,3-FucT enzyme, the synthesis of Le x -PNP S21 occurs straight away ( Figure S5a). However, when both enzymes are added at the same time in the solution, the enzymatic transformation of LacNAc-PNP S19 into Le x -PNP S21 takes longer and at 40 min it reaches only a 50% conversion ( Figure S5b). Since FKP is the accessory enzyme for the production of donor nucleotides and it is used in excess in the reaction (compared with the α1,3-FucT), it is suggested that adding the FKP enzyme into the solution 20 min before the addition of α1,3-FucT, results in a decrease of the second enzymatic transformation time from 2 h to 20 min. The 'one-pot' enzymatic transformation was then investigated, which consists of a one-step protocol at 37 C. GlcNAc-PNP S18 (200 µM) was mixed with 4GalT1 (0.54 mg/mL)/UDP-Gal (10 mM) and α1,3-FucT (3.7 mg/mL)/FKP (8 mg/mL)/Fuc (10 mM) in the same reaction step, which resulted in the synthesis of Le x -PNP S21. The reaction was monitored by HPLC ( Figure S6a) and two peaks were identified as being Lex-PNP S21 (11.6 min) and GlcNAc-PNP S18 (13.0 min). The peak corresponding to LacNAc-PNP S19 was not found during the course of the reaction, which suggests LacNAc-PNP S19 is consumed as soon as it forms. The proportion of each molecule over time ( Figure S6b) revealed a decrease of GlcNAc-PNP S18 and an increase of Le x -PNP S21 in the first three hours, after which both molecules reach a plateau corresponding to proportions of 29% and 71%, respectively. The outcome of the reaction suggests that the activity of 4GalT1 enzyme is compromised by the presence of α1,3-FucT enzyme (or a reagent from the fucosylation reaction). Previously the galactosylation reaction using 4GalT1 resulted in a 96% conversion. However, in this experiment 29% of starting material GlcNAc-PNP S18 was detected. Le x -PNP S21 was afforded by two methods and the results revealed that the sequential approach afforded higher proportion overall conversion to Le x -PNP S21 (~96% over two steps) than the 'onepot' approach (71%, Figure S6). In addition, the 'one-pot' method also gave unreacted S18 at the end of the experiment, which was detrimental to the Le x -PNP S21 yield. Also, by adding the FKP enzyme 20 min before the addition of α1,3-FucT, the reaction time of the sequential approach could be considerably decreased.

Procedures for enzymatic transformation of PNP derivatives
The enzymatic transformation of GlcNAc-PNP was monitored by reverse-phase HPLC using a 20 min gradient method ranging from 5 to 30% MeCN/IPA (4:1) in water with a flow rate of 0.5 mL/min. The eluent was monitored by setting the HPLC DAD module to record the phenyl group absorbance at 300 nm. A Macherey-Nagel Nucleosil C18 (150 × 4.6 mm, 5 μm) fitted with a guard column was used. The GlcNAc-PNP S18 eluted at 13.0 min, LacNAc-PNP S19 at 12.6 min, Neu5Ac-LacNAc-PNP S20 at 11.4 min, sLe x -PNP S22 at 10.3 min and Le x -PNP S21 at 11.6 min. The m/z of each peak was analysed by LC-MS using the same method.
Preparation of fluorescently labelled liposomes: Labelled liposomes were prepared by mixing 1.8 μmol of DMPC, 0.2 μmol of synthetic glycolipid and 0.002 μmol (0.1% mol/mol) rhodamine DHPE (6.7 µL of a stock solution in chloroform) in CH3OH/CHCl3 (1:1 v/v). After solvent evaporation, the resulting thin film was hydrated and the lipid suspension was extruded following the standard liposome preparation method described above.
Preparation of DOPC and DOPC/2 liposomes: DOPC liposomes were prepared by mixing 2 μmol of DOPC in CH3OH/CHCl3 (1:1 v/v), whereas DOPC/2 liposomes were prepared by mixing 1.8 μmol DOPC and 0.2 μmol of glycolipid 2 in CH3OH/CHCl3 (1:1 v/v). After solvent evaporation, the resulting thin films were hydrated with citrate buffer (1 mL, 300 mM, pH 4.0) and the lipid suspension was extruded following the standard liposome preparation method described above. Agglutination assay using DLS The liposome suspension was transferred to disposable cuvettes (ZEN0040). Each sample was analysed by DLS before and after the addition of the respective lectin (WGA or ECL) at a concentration of 0.1 mg/mL.
Agglutination assay using Fluorescence Microscopy Liposomes were labelled with rhodamine DHPE following the protocol described in the standard procedure above. To facilitate visualisation by fluorescence microscopy, liposomes were extruded using a membrane of 800 nm pore size to afford LUVs. Liposomes were imaged before and after incubation with the respective lectin (WGA or ECL at 0.1 mg/mL) using an oil immersion 100x objective.

Measurement of galactosylation in enzymatically transformed liposomes
Quantification by galactose oxidase assay Both DMPC/2 and DMPC/13 liposome formulations were analysed using the galactose oxidase assay. This assay is a simple approach for the direct measure of Gal in solution. Galactose oxidase (GOase) is a fungal enzyme able to oxidise the C-6 hydroxymethyl group of Gal, forming an aldehyde (Scheme S5). During the catalysis, oxygen is reduced to hydrogen peroxide (H2O2). Adding horseradish peroxidase (HRP) and 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) into the solution results in the oxidation of the latter by the H2O2. Finally, a coloured ABTS positive cation (ABTS) + is formed, in a quantity proportional to the number of Gal present in solution. The GOase assay is not only suitable for the detection of free reducing Gal residues, but also to galactosyl derivatives, including Gal-or LacNAc-capped glycolipids. 20,21,22 In addition, the stereospecificity of GOase is susceptible to the orientation of the C-4 hydroxyl group, hence it does not oxidise Glc residues. 22 Samples of DMPC/2 and DMPC/13 liposomes were treated according to the supplier's instructions in the galactose assay kit from Sigma. The DMPC/2 liposomes were analysed as the control experiment. The fluorescence of the samples was measured (λex = 535 nm and λem = 587 nm) and extrapolated from a standard curve that showed the relationship between known concentrations of synthesised glycolipid 13 and the respective fluorescence values. The results obtained (see manuscript Figure 5a) S45 suggest that the degree of Gal conversion in DMPC/13 liposomes is about 16%, a comparable result to that obtained by LC-MS (see manuscript Figure 5b).
Scheme S5: Oxidation of Gal by galactose oxidase (GOase) and concomitant production of the chromophore ABTS by horse radish peroxidase (HRP).

Estimation by LC-MS
Liquid chromatography-mass spectrometry (LC-MS) was considered for the quantification of the Gal conversion since glycolipids 2 and 13 lack chromophores, which limits analysis by UV detection (e.g. HPLC). Mass spectrometry is highly sensitive and tolerant to mixtures; however, is not usually used as a quantitative technique because ionisation efficiencies vary among different molecules. 23 Therefore, some considerations were deliberated to carry out quantitative analysis using LC-MS, namely method optimisation (including the choice of solvents and column), reference solutions for MS calibration (purine 121 ion solution and ESI tuning mix), number of sample replicates (set up to three), analysis of blanks (MES buffer) and the interference by the enzyme (since samples were not purified). It was hoped that a standard curve showing the relationship between known concentrations of glycolipids 2 and 13 and their MS peak area could be obtained. For the purpose of the standard curve, glycolipid 13 was chemically synthesised following the same methodology used to afford glycolipid 2.
In the context of method optimisation, samples containing DMPC/2 and DMPC/13 liposomes were used. Due to the glycolipid's hydrophobic nature, a C8 column was chosen. The best compromise between the appearance of shoulders in the peak corresponding to 13 and the ionisation level of 2 was achieved using a 10 min gradient ranging from 80% to 100% of a mixture of organic solvents (50% MeCN/50% IPA supplemented with 0.1% formic acid) in water (with 0.1% formic acid) with a flow rate of 1.0 mL/min. The extracted chromatograms of [2+Na] + and [13+Na] + showed that molecule 2 eluted at 6.0 min and the molecule 13 eluted at 5.6 min. Analysis of samples containing the enzyme 4GalT1 did not show any interference with the glycolipid MS peaks.
For the quantification of Gal conversion, liposome samples containing DMPC/2 (before enzymatic transformation) or DMPC/13 (after enzymatic transformation) were diluted (0.02 mM) and transferred to mass spectroscopy tubes. The tubes were analysed by LC-MS using the optimised conditions. Six replicates were investigated from two independent experiments and the standard curve repeated for each one. Before enzymatic transformation, DMPC/2 liposomes showed one peak at 6.0 min that correspond to the m/z of the glycolipid 2 (see manuscript Figure 5c). On the other hand, liposomes that were incubated with 4GalT1 gave rise to two peaks (see manuscript Figure 5d), namely at 6.0 min that corresponds to the m/z of 2 ([2+Na] + ), and at 5.6 min that corresponds to an m/z of 13 ([13+Na] + ). The peak area of 2 is smaller in enzymatically transformed samples DMPC/(2+13) than in DMPC/2 suggesting a partial consumption of 2 after the enzymatic reaction. The respective peak areas were then extrapolated from the standard curve ( Figure S11) and a 19% conversion to 13 was estimated.
Methodology: Aliquots (5 μL) of DMPC/2 (before enzymatic transformation) and DMPC/(2+13) (after enzymatic transformation) liposome suspensions (0.02 mM concentration in MES buffer) were analysed by LC-MS using a 10 min gradient ranging from 80% to 100% of a mixture of organic solvents (50% MeCN/50% IPA supplemented with 0.1% formic acid) in water (with 0.1% formic acid) with a flow rate of 1.0 mL/min. Due to the glycolipid amphiphilic nature, an Agilent Zorbax Eclipse XDB C8 column (150 × 4.4 mm, 5 μm) fitted with a guard column was used. A vial containing only buffer was used to wash the system. Analysis of samples containing the 4GalT1 enzyme did not show any interference with the glycolipid MS peaks. Molecule 2 eluted at 6.0 min and molecule 13 eluted at 5.6 min based on the extracted chromatograms of [2+Na] + and [13+Na] + , respectively. The area of each S47 peak (in counts) was normalised as a ratio of enzymatic transformed product (EC1093) in terms of nontransformed product (EC931) using the Equation S1 . The values obtained were then extrapolated from a standard curve ( Figure S11) to obtain the percentage of Gal conversion in each sample. The standard curve consisted of known concentrations of glycolipid 2 mixed with the chemically synthesised glycolipid 13 (at the ratio of 0:100, 25:75, 50:50, 72:25 and 100:0), plotted against the normalised peak area from the respective extracted chromatogram. A total of two experiment were performed (each with three replicates) and for each one a new standard curve was prepared as well as standard MS calibration (using solutions of purine 121 ion solution and ESI tuning mix).

Quantification of available N-alkoxyamino groups on liposome surfaces
The extent to which this class of lipid embedded into bilayers was assessed using the 2,4,6trinitrobenzenesulfonic acid (TNBS) assay on lipid 1. The TNBS assay is a colorimetric test that consists of reacting TNBS with free primary amines, which result in an orange coloured trinitrophenyl derivative with absorbance at 320 nm (Scheme S6). 1 TNBS does not react with secondary (R2NH) or tertiary (R3N), making this assay highly specific for detecting primary (RNH2) amines.

S48
Scheme S6: Colour generating reaction in the TNBS assay. 1 For the purpose of the experiment, liposomes (200 nm diameter) were prepared bearing amino lipid 1 because the free primary amine can be detected by the TNBS assay. The liposomal suspension was transferred to a Vivaspin ® 500 centrifugal concentrator (cut-off of 10 kDa) and after centrifugation, both top and bottom solutions were collected and analysed (Figure S12 A). It was expected that nonembedded lipid 1 would move to the bottom compartment of the Vivaspin while lipid 1 embedded within the bilayers would stay on top, as they should be unable to cross the filter. The collected solutions were analysed according to the TNBS methodology described by Davidenko et al. (see below). 24 The absorbance values obtained for each sample was extrapolated from a standard curve that consisted of the relationship between known concentrations of glycine (one primary amine per molecule) and the respective absorbance at 320 nm (see below).
The results obtained ( Figure S12 C) show that the Vivaspin top compartment contained about 0.02 µmol of amine groups (100 µL at 0.2 mM), whereas on the bottom compartment no amine content was detected. Glycolipid 1 was not detected in the bottom compartment, which means that full incorporation of glycolipid 1 (0.2 mM) within the membrane bilayer of the liposomes was achieved. This result indicates that this class of synthetic N-(alkyloxy)amine-terminated lipid does, in fact, mix well with the ld phase of DMPC phospholipid bilayers.
Methodology: Liposomes were prepared bearing amino groups (lipid 1) because the free primary amine can be detected by the TNBS assay. The liposomes suspension was transferred to a Vivaspin ® 500 (cut-off of 10 kDa, 2.9 nm pore size) and centrifuged (4000 rpm, 20 min). It was expected that nonembedded lipid 1 would move to the bottom compartment of the Vivaspin while the molecules embedded within the bilayers would stay on top as they should be unable to cross the filter (Figure S12 A). Both top and bottom solutions were collected and analysed following the same methodology. Then 0.5 mL of freshly made 0.05% (v/v) TNBS was added to each tube, followed by a 2 h incubation at 40 C. In order to stabilise the TNBS-complex, 1.5 mL of HCl (6 M) was added to each solution and samples were further incubated for 90 min at 60 C. Finally, samples were diluted with 2.5 mL of distilled water. An aliquot of 100 µL of each sample was transferred to a clear 96 well-plate (Corning) and the absorbance was measured at 320 nm using a plate reader. The absorbance values obtained were extrapolated from a standard curve that consisted of glycine solution at different known concentrations (0.001 -0.167 mmol/well) plotted against the respective absorbance (Figure S12 B). Blank samples were used to subtract auto absorbance of the samples.

Solubility of glycolipid 2 in buffer
The alternative approach of the synthesis of 13 was considered, where the enzymatic reaction would be performed in buffer in the absence of liposomes, which we hoped would decrease the amount of steric hindrance encountered by the enzyme and increase conversion. However, mixing the glycolipid 2 in buffer (MES buffer, pH 7.0), which are the enzymatic reaction conditions, results in the formation of insoluble particles and large aggregates (40-2000 nm diameter as shown by DLS, Figure S13), due to the amphiphilic nature of the glycolipid. Given this observation, this approach was abandoned.

Monitoring of multienzyme transformation of 2 embedded in liposomes by LC-MS.
Glycolipid 2 was mixed with phospholipids prior to vesicle formation, which results in half of the lipid becoming located in the inner leaflet of the bilayer. Therefore, at least half of glycolipid 2, which is located in the inner face of the liposomes, will not be easily accessible to externally added regents The proportion available will depend on the rate of flip-flop of 2; should flip-flop be slow then only half will be available.

Studies of liposome stability over time
Stability is an essential parameter during liposome production, storage and administration. It includes phospholipid and ligand chemical stability, the preservation of liposome size and structure, and cargo retention. 25 In this sense, the storage stability of DMPC/13 liposomes (in MES buffer, 0.2 mM, 200 nm diameter) was investigated at two different storage temperatures, namely 4 C and 37 C (to mimic cell culture temperature) over 24 days using DLS (hydrodynamic size and PdI). Figure S15 compares the performance of DMPC, DMPC/2 and DMPC/13 liposomes based on their size and PdI when stored at 4 C and 37 C. Among these formulations, DMPC/2 liposomes show the best performance throughout the 24 days, with consistent diameter size and PdI regardless of the temperature. In the case of DMPC/13 liposomes, they are stable at 37 C; however, at 4 C their diameter changes considerably from 164 nm to 230 nm (PdI of 0.3). Similar behaviour was observed with DMPC, in which at 37 C the liposomes were more stable than at 4 C, although the PdI values remain unchanged in both temperatures. In summary, these results show that liposomes were more stable at 37 C than 4 C, which is beneficial for cell culture use. Importantly, the absence of liposomal agglomeration during the 24 days in all formulations reveals that liposomes were sufficiently charged to repel each other, despite their low  potentials (displayed in Table S2 and Table S3) Methodology: Liposomes (DMPC, DMPC/2 and DMPC/13 liposomes) were prepared in MES buffer (100 nm diameter, 0.02 mM concentration) and transferred to a disposable cuvette (ZEN0040). Samples were stored at two different temperatures namely 4 C (to mimic the storage temperature) and 37 C (to mimic cell culture temperature). Samples were analysed by DLS and the respective hydrodynamic size and PdI recorded over 24 days. Before each measurement, each sample was pipetted up and down to resuspend any sedimented liposomes in the cuvette. Table S2: Zeta potential of DMPC and DMPC/2 liposomes in MES (50 mM, pH 7.0) and HEPES (25 mM, 150 mM NaCl, pH 7.5) buffer using the diffusion barrier method. 19 A total of five measurements were performed. Table S3: Zeta potential of DMPC/13 liposomes prepared in MES and HEPES buffer using the diffusion barrier method. 19 Five measurements were performed for each sample.

Procedure for the active encapsulation of doxorubicin in liposomes
Active encapsulation consists of adding DOX to liposomes under an ion gradient that is established by an acidic pH inside and a basic pH outside the liposome ( Figure S17). As a result, the drug crosses the bilayer and, due to the acidic environment, precipitates inside the liposome causing its entrapment. This approach has been used in the preparation of liposomes loaded with DOX for clinical use. 27 The preparation of liposomes in acidic buffer (citrate buffer) to establish a pH gradient for the encapsulation of DOX was shown to be incompatible with the enzymatic transformation of DMPC/2 liposomes using the 4GalT1 enzyme. Therefore DOX encapsulation would have to occur prior (on DMPC/2 liposomes) to the enzymatic transformation due to the acidic conditions.
Following the literature pH gradient procedure for the active loading of DOX, 27 liposomes (DMPC and DMPC/2, ca. 100 nm diameter, 2.0 mM lipid, prepared according to details in Section 7) were prepared in citrate buffer (300 mM, pH 4.0), followed by the change of the external buffer to HEPES (25 mM with 150 mM NaCl, pH 7.5) using a gel permeation chromatography (GPC) column. DOX solution (final molar ratio of 4:1 phospholipid to the drug) was added to the liposomes and incubated for 1 h at T>Tm. Finally, non-encapsulated DOX was removed by a GPC column.
Liposomes were loaded at three different temperatures, namely 29 C, 28 40 C and 60 C, and the loading efficacy investigated by comparing the absorbance (DOX absorbance peak at 480 nm) of liposomes before and after the purification step (GPC column). UV-visible spectroscopy data ( Figure  S16) reveals complete loss of DOX absorbance after the purification step, regardless of the loading temperature. These data suggest that the encapsulation of DOX in DMPC and DMPC/2 liposomes was unsuccessful. Methodology: Doxorubicin (DOX) was encapsulated into liposomes using an active loading approach (based on a pH gradient) described in the literature ( Figure S17). 27 Figure S17: Active loading of DOX into liposomes by a pH gradient using the citrate loading procedure. 27 (A) Liposomes are prepared in 300 mM citrate buffer, pH 4.0, followed by a (B) buffer exchanged to 25 mM HEPES buffer (with 150 mM NaCl), pH 7.5. (C) Finally, DOX crosses the bilayer and, due to the internal acidic pH, precipitates inside the liposome, causing its entrapment

Efficiency of doxorubicin encapsulation in liposomes
The concentration of DOX in each liposomal formulation before and after purification was determined by interpolating each sample absorbance values from a standard curve (relationship between known concentrations of free DOX and the respective absorbance values, see methodology below).
The concentration of DOX detected in each liposomal formulation before and after GPC purification is shown in Figure S18.
Once DOX is encapsulated inside liposomes, their spherical shape tends to change to an oval shape due to the precipitation of DOX in their core. 30 Therefore, the hydrodynamic size of the formulated liposomes was verified by DLS before and after loading of DOX. DMPC/chol, DMPC/chol/2, DOPC and S55 DOPC/2 liposomes show that the hydrodynamic size was the same within the experimental error (before and after drug encapsulation), which means that no significant change in diameter size occurred due to DOX precipitation.
Methodology: Aliquots (100 µL) of each sample (before and after GPC column purification) were transferred to a clear 96 well-plate (Corning) and the absorbance measured at 480 nm. The concentration of DOX in each sample was determined by extrapolating the sample's absorbance values from a standard curve ( Figure S19). The standard curve consisted of known concentrations of free DOX (0.01-0.60 mM) plotted against the respective absorbance values. Bare liposomes were considered as blank samples to account for liposomal scattering. In parallel, all liposome suspensions were analysed by DLS before and after loading of DOX and their hydrodynamic size compared. The values obtained were the same within the experimental error and no differences were observed between the diameter size of liposomes before and after encapsulation with DOX ( Figure S20).

Release of doxorubicin from loaded liposomes
The rate of release of doxorubicin from DMPC/chol+DOX, DMPC/chol/2+DOX, DOPC+DOX and DOPC/2+DOX liposomes was investigated using dialysis. Samples were transferred into a dialysis bag (BioDesignDialysis Tubing ® , 14000 MWCO) and at specific time intervals, aliquots of the liposomal solution were collected. The absorbance of each sample was measured ( = 480 nm) and the percentage of DOX inside the liposomes calculated. The initial percentage of loaded DOX for each liposomal formulation was set as the loading efficiency calculated earlier ( Figure S18), i.e. 85% for DMPC/chol, 68% for DMPC/chol/2, 89% for DOPC, and 83% for DOPC/2. Free unencapsulated DOX was used as the experiment control.
The release profile of DOX is presented in Figure S21. All the liposomal formulations were able to retain their cargo for a longer period than the control (the free drug) and were retained in the dialysis bag. Although DOPC based liposomes showed the highest encapsulation efficiencies, namely 89% for DOPC and 83% for DOPC/2, the ability to retain their cargo was poor and at 48 h there was no DOX detected in both formulations. Since DOPC has a Tm of -17C, DOPC and DOPC/2 liposomes are more prone to leakage at room temperature, which might explain the premature release of their cargo.  S58

Cell toxicity assays
The cytocompatibility of enzymatically transformed liposomes was explored. DMPC, DMPC/2 and DMPC/(2+13) liposomes were incubated with HepG2 cells, and their cell toxicity determined using the Alamar Blue ® assay.
HepG2 cells (2.5 × 10 4 cells/cm 2 , 24 h cell adhesion) were incubated with sterile DMPC, DMPC/2 and DMPC/(2+13) liposomes (100 nm diameter) at different concentrations. After 24 h incubation, the media was replaced by media containing resazurin and incubated for further 4 h. The fluorescence of each sample was recorded (λex = 530 nm and λem = 590 nm) and interpolated into a standard curve that showed the relationship between known concentrations of living cells (in the absence of liposomes) and the respective resorufin fluorescence values. Cells incubated in the absence of liposomes were included in the experiment as a control.
Liposomes at 5.3, 13.3 and 26.7 µg/mL do not alter cell viability (independent of the liposome formulation) since cell numbers were similar to the control ( Figure S22). At higher liposomal concentrations (53.4 µg/mL), both DMPC and DMPC/2 liposomes showed similar cell numbers with the control, but DMPC/(2+13) liposomes cause a reduction of the cell numbers to 0.75 × 10 4 cells/cm 2 ; however, it was not considered significantly different from the control by the non-paired Student's t test (p value <0.05 considered significant). 33 Based on the mean decrease of the sample (DMPC/(2+13) at 53.4 µg/mL), it is hypothesised that if higher concentrations were tested (e.g. 106.8 µg/mL), an even lower cell density would be observed and likely to be significant different from the control. These results indicate that all concentrations of liposomes should be suitable for cell culture experiments using HepG2 cells.

S59
Procedure for culture of human hepatocellular carcinoma cell line (HepG2) Cells were cultured in Eagle's minimum essential medium (EMEM) supplemented with 10% (v/v) foetal bovine serum (FBS), 2 mM L-glutamine, 1% non-essential amino acids and 1% penicillin/streptomycin solution. Cells were cultured in T25 culture flasks with 5 mL media and incubated at 37 °C with a 5% CO2 atmosphere. Media was renewed every 48/72 h. When cells became 80% confluent, the media was removed and the flask washed with PBS in order to remove the non-adherent cells. Then, the cells were incubated with 1 mL of accutase for 5 min at 37 °C. The cell suspension was aspirated back and forth in order to separate cell aggregates. Finally, the cell suspension was distributed among three T25 flasks (1:3 subcultivation ratio).

Procedure for measuring cell viability using Alamar blue
Liposomes were prepared (100 nm) and their hydrodynamic size and monodispersity verified by DLS ( Figure S23). Small liposomes were used to facilitate liposomes sterilisation by filtration (0.2 µm pore size filters), which is the common sterilisation technique used to prepare pharmaceutical products. 25 Liposomes were enzymatically transformed using 4GalT1 enzyme as described before. After reaction, 4GalT1 enzyme and UDP-Gal were removed from the liposomes solution (200 µL) by centrifugation (13000 rpm for 5 min). The supernatant was removed (150 µL) and fresh MES buffer (150 µL) was added to resuspend the pellet (by vortex). Each supernatant collected was analysed by UV-visible spectroscopy to determine the amount of enzyme and UDP-Gal present (peak at 280 nm) still present in the suspension. A total of 4 washes were necessary to remove all 4GalT1 and UDP-Gal present. HepG2 cells were seeded on a 96-well plate at 2.7 × 10 4 cells/mL and incubated for 24 h for cell adhesion. Then, sterilised liposomes at concentrations of 5.3, 13.3, 26.7 and 53.4 µg/mL were added to the cells. After 24 h incubation at 37 C, the media was replaced by a solution of 1:10 of resazurin (0.55 mM in media without serum). The composition of fetal bovine serum (FBS) may vary between batches and, considering reproducibility of the experiment, this supplement was not included. Also, in the absence of serum, any liposomal interactions with albumin and other serum components are prevented. Cell proliferation is also likely to be prevented. 34 The plate was incubated for 4 h at 37 C, followed by the transfer of 200 µL of cell media to a 96-well black plate (Greiner). The fluorescence of each well was recorder (λex = 530 nm and λem = 590 nm). In parallel, a standard curve was prepared S60 using known concentrations of cells (0-2.7 × 10 4 cells/mL) incubated in the absence of liposomes plotted against fluorescence intensity ( Figure S24).