DOI: 10.1002/chem.201500708
Full Paper
& Oligosaccharides
Chemoenzymatic Syntheses of Sialylated Oligosaccharides Containing C5-Modified Neuraminic Acids for Dual Inhibition of Hemagglutinins and Neuraminidases L¦monia Birikaki,[a, b] St¦phanie Pradeau,[a, b] Sylvie Armand,[a, b] Bernard Priem,[a, b] Luis Mrquez-Domnguez,[c] Julio Reyes-Leyva,[c] Gerardo Santos-Lûpez,[c] Eric Samain,[a, b] Hugues Driguez,[a, b] and S¦bastien Fort*[a, b] Abstract: A fast chemoenzymatic synthesis of sialylated oligosaccharides containing C5-modified neuraminic acids is reported. Analogues of GM3 and GM2 ganglioside saccharidic portions where the acetyl group of NeuNAc has been replaced by a phenylacetyl (PhAc) or a propanoyl (Prop) moiety have been efficiently prepared with metabolically engineered E. coli bacteria. GM3 analogues were either obtained by chemoselective modification of biosynthetic Nacetyl-sialyllactoside (GM3NAc) or by direct bacterial synthesis using C5-modified neuraminic acid precursors. The latter
Introduction N-Acetylneuraminic acid (NeuNAc) is the common terminal carbohydrate unit of glycoproteins and glycolipids expressed at the surface of mammalian cells. Generally appended to a galactose or a N-acetyl-galactosamine unit by an a-2,3- or a-2,6-glycosidic linkage, NeuNAc is ideally positioned to interact with proteins. It is thus involved in a variety of normal and pathological processes such as cell–cell adhesion, cell recognition, bacterial infection, viral invasion, inflammation and cancer metastasis.[1, 2] The biological significance of sialylated molecules makes them excellent targets for the design of carbohydrate-based drugs. In particular, glycoconjugates derived from a-2,3-sialyllactose, the saccharidic portion of gangliosides can be used to target certain types of cancer and viruses. [a] Dr. L. Birikaki, S. Pradeau, Dr. S. Armand, Dr. B. Priem, Dr. E. Samain, Dr. H. Driguez, Dr. S. Fort Univ. Grenoble Alpes, CERMAV, 38000 Grenoble (France) Fax: (+ 33) 4-76-54-72-03 E-mail: [email protected] [b] Dr. L. Birikaki, S. Pradeau, Dr. S. Armand, Dr. B. Priem, Dr. E. Samain, Dr. H. Driguez, Dr. S. Fort CNRS, CERMAV, 38000 Grenoble (France) [c] L. Mrquez-Domnguez, Dr. J. Reyes-Leyva, Dr. G. Santos-Lûpez Laboratorio de Biologa Molecular y Virologa Centro de Investigaciûn Biom¦dica de Oriente Instituto Mexicano del Seguro Social Metepec, Puebla (M¦xico) Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/chem.201500708. Chem. Eur. J. 2015, 21, 10903 – 10912
strategy proved to be very versatile as it led to an efficient synthesis of GM2 analogues. These glycomimetics were assessed against hemagglutinins and sialidases. In particular, the GM3NPhAc displayed a binding affinity for Maackia amurensis agglutinin (MAA) similar to that of GM3NAc, while being resistant to hydrolysis by Vibrio cholerae (VC) neuraminidase. A preliminary study with influenza viruses also confirmed a selective inhibition of N1 neuraminidase by GM3NPhAc, suggesting potential developments for the detection of flu viruses and for fighting them.
The modification of cellular glycosylation corresponds indeed to a common phenotypic modification of cancer cells, which mainly affects the outer part of glycans, leading to the expression of tumor-associated carbohydrate antigens (TACAs).[3] These TACAs have become important molecular targets for the development of anti-cancer vaccines.[4–6] GM3 [(Neua2-3)Galb1-4Glc] and GM2 [GalNAcb1-4(NeuAca2-3)Galb4Glc] gangliosides are overexpressed in various types of cancer, namely those of breast, melanoma, ovary, prostate and lung and thus, they have emerged as promising target for cancer vaccines.[7] Synthetic glycoconjugates combining GM2 with an immunogenic carrier (KLH,[8] poliovirus polypeptide sequence[9]) have been synthesized and were shown to induce cell-specific antibodies for human tumor. A GM2-KLH vaccine, administrated with the adjuvant QS-21, induced not only high IgM titers, but also durable IgG antibodies that were shown to induce complement-mediated lysis of GM2-positive tumor cells in most patients.[8] Recently, new potential carbohydrate-based anti-cancer vaccines incorporating unnatural sialic acids have been proposed to address the immunotolerance issues. Indeed, N-modified GM3,[10, 11] GD3[12] and polysialic acid[13] antigens have showed increased immunogenicity, when compared to the N-acetyl oligosaccharides counterparts and have thus been proposed as promising vaccine candidates against melanoma or lung cancers. In this emerging therapy for cancer cells, a synthetic vaccine targeting a TACA analogue would be administrated together with a sugar precursor meant to induce the expression of the artificial TACA derivative in place of the native TACA.[14] Among various N-modified sialic acid
10903
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper containing antigens, the N-phenylacetyl (NPhAc) derivative generally induced the strongest immune response and their anti-sera were the least reactive with the natural GM3 glycans. These results suggest that the GM3NPhAc is an excellent candidate for an antitumor vaccine presenting relatively low autoimmune reactions.[15, 16] Sialylated oligosaccharides are not only attractive targets for active immunotherapy, but they also play a major role in viral infections. Influenza viruses cause seasonal epidemics and occasional threats, as seen with “swine” influenza H1N1, avian influenza H5N1[17] and the very recent H7N9[18] viruses that induced a significant morbidity and mortality in humans. H1N1 flu virus, which caused a world-wide pandemic in 2009, is now a human seasonal flu virus that also circulates in pigs. 2014 was the first year since 2009 where H1N1 has been so predominant in the United States.[19] H7N9 virus was reported for the first time in early 2013, and was confirmed to be a low pathogenic avian influenza virus in poultry. The reported cases of attack by this novel virus have been increasing since the summer of 2013. The virus poses a potential threat to public health, particularly in China, and for this reason, it is attracting a worldwide broad attention. Influenza virus infections are initiated when the virus adsorbs on respiratory epithelial cells by the interaction of its hemagglutinin with the a-2,6 or a-2,3-sialoglycoconjugates. The release and spreading of the virus is subsequently promoted by a neuraminidase.[20] So far, the prevention and treatment against influenza consist of seasonal vaccination and the administration of sialidase inhibitors namely, oseltamivir[21] (Tamiflu) and zanamivir (Relenza),[22] respectively. Despite these treatments, new virus strains are constantly emerging, especially resistant to oseltamivir.[23] Unlike oseltamivir, which can be orally administrated, zanamivir can only be inhaled due to its low bioavailability, limiting the use of this drug. Interfering with both hemagglutinins and sialidases has emerged as a possible alternative strategy to prevent influenza infection. Indeed, liposomes made of sialoglycolipids in which the C3 position of the sialic acid was modified with
an axial fluorine atom inhibited influenza infection at micromolar concentration.[24, 25] This experimental result confirmed that such glycomimetics not only inhibit the attachment of viruses to the cell surface receptor but also disturbs the release of the progeny viruses from infected Mardin–Darby canine kidney (MDCK) cells by inhibiting both hemagglutinin and sialidase. Sialylated galactosides containing C5-diversified sialic acid have also proven efficient tools to conduct specificity studies of sialidases.[26] The ability of various influenza viruses including human and avian influenza A strains to hydrolyze these glycomimetics has been tested, revealing that it showed marked differences depending on the C5 substitution. These results should certainly contribute to the design of new antiviral drugs designed to target specific influenza strains. Interest in modified sialic acids is not restricted to the study of influenza viruses. The adenovirus serotype Ad37, which causes epidemic keratoconjunctivitis, binds to and infects human corneal epithelial (HCE) cells by attaching itself to cellular glycoproteins carrying terminal sialic acids. N-Acyl modified GM3 were chemically synthesized and assayed in Ad37 cell-binding. In addition, compounds bearing small substituents, in particular GM3NProp, were shown to be as effective inhibitors as sialic acid.[27] Taken together, these observations highlight the potential of using modified sialylated oligosaccharides, but the access to these molecules by an efficient process remains an important issue. Common synthetic strategies developed for the production of sialoglycoconjugates and their mimetics generally rely on time-consuming multistep chemical approaches.[11, 27] Alternatively, in vitro enzymatic processes involving NeuAc-aldolase have been successfully developed.[28–30] The latter technique, while being more rapid since it does not require the handling of protecting groups, requires the use of expensive mannosamine precursors, thus hampering syntheses at a multigram scale. In the present work we report a convenient access to C5-modified sialosides either by chemical modification of GM3NAc produced by metabolically engineered E. coli or by direct bacterial synthesis from modified neuraminic acid pre-
Figure 1. Chemo-enzymatic routes towards GM3 and GM2 analogues, incorporating C5-modified neuraminic acids. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10904
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper cursors (Figure 1). Beyond the challenge raised by the synthesis of such molecules, our final goal was to investigate the influence of these C5-modified sialosides on both their binding to hemagglutinins and their stability against neuraminidases.
Results and Discussion Synthesis of GM3 and GM2 oligosaccharide analogues The synthesis of GM3 analogues was undertaken following two different routes. The first approach consisted in the enzymatic synthesis of GM3NAc by metabolically engineered E. coli using b-allyl lactoside and N-acetyl-neuraminic acid (NeuNAc) exogenous acceptors, followed by the chemoselective modification of the neuraminic moiety. Metabolically engineered bacteria have emerged as efficient tools to synthesize complex oligosaccharides at a multigram scale.[31] Engineered E. coli cells behave as living factories, producing oligosaccharides in a onepot procedure, starting from cheap substrates, without requiring the isolation of intermediates or enzymes. The efficiency of such a bacterial approach is far above that of classical in vitro enzymatic processes. The range of oligosaccharides produced by this technique has considerably expanded over the last ten years: among others, they include the production of gangliosides,[32, 33] of blood group antigens,[34, 35] as well as of chitinoligosaccharides.[36, 37] Conjugatable forms of GM3NAc containing an alkyne or alkene-terminated spacer[38] have also been advantageously prepared by this technique, thus leading to a rapid and efficient access to neo-glycoconjugates, such as glycopolymers[39, 40] and glycoproteins,[41] for diagnosis or therapeutic applications. As seen in Figure 2, GM3NAc 1 was prepared by bacterial synthesis from N-acetyl-neuraminic acid and allyl b-lactoside as reported previously.[38] The, de-N-acetylation was then conducted by alkaline hydrolysis affording 2 in 93 % yield. Evidence for this transformation was given by NMR spectroscopy with the disappearance of the signals of both the methyl and carbonyl groups together with the upfield shift of the H-5 proton of the neuraminic residue centered at 2.91 ppm. It is noteworthy that although the acetamido hydrolysis requires strongly alkaline conditions, the allyl group re-
mained intact and was still available for subsequent linking to another biomacromolecule. The selective N-acylation of 2 with propionic anhydride or phenylacetylchloride in methanol afforded 3 and 4 in 89 and 88 % yield respectively. Both compounds displayed their expected characteristics, when analyzed by mass and NMR spectrometry and no by-product resulting from the O-acylation was observed. An alternative approach to the synthesis of GM3NProp 3 and GM3NPhAc 4 was envisaged by internalization and glycosylation of C5-modified sialic acids in E. coli. Enzymatic in vitro glycosylation of modified sialic acids onto glycoprotein glycans by sialyltransferases has already been reported. Synthetic analogues of sialic acids where the C-5 acetyl group had been replaced by formyl-, trifluoroacetyl-, benzyloxycarbonyl-, or aminoacetyl groups have been successfully activated into CMP-derivatives by CMP-sialic acid synthase and transferred by a-2,6 and a-2,3 sialyltransferases onto antifreeze glycoprotein or asialofetuin. The glycosylation efficiency of these modified sialic acids was comparable to that of the unmodified NeuNAc, with only some exceptions.[42] CMP-NeuNProp was also efficiently prepared in vitro using CMP sialate synthase, albeit on a 10 mg scale only.[43] The promiscuity of sialic acid-modifying enzymes has triggered intensive research on in vivo N-acyl engineering of cell-surface sialoglycoconjugates[44, 45] but the large-scale synthesis of N-modified sialooligosaccharides has been poorly investigated so far. NeuNProp 9 and NeuNPhAc 10 were thus prepared by chemical modification of NeuNAc (Figure 3) and tested for the in vivo production of sialomimetics in E. coli. The de-N-acetylated neuraminic acid was previously reported to form an internal Schiff base in water, by condensation of the amino group at C5 with the carbonyl group at C2, yielding 4-hydroxy-5-(1,2,3,4-tetrahydroxybutyl)-D1-pyrroline-2-carboxylic acid thus preventing an efficient acylation reaction.[46] Therefore, NeuNAc was first converted into its a-methyl glycoside. Methyl esterification and O-methyl glycosylation of NeuNAc were conducted in a one-pot procedure with Dowex 50 (H + ) resin in MeOH.[47] The resulting O-methyl sialoside 5 was subsequently de-N-acetylated under alkaline conditions. Saponifica-
Figure 2. Synthesis of GM3NProp and GM3NPhAc oligosaccharides by chemical modification of recombinant GM3NAc. i) KOH, H2O, reflux, 93 %, ii) (CH3CH2CO)2O, MeOH, 89 %, iii) PhCH2COCl, Et3N, MeOH, 88 %. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10905
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper tion into a more polar compound, which did not diffuse outside of the bacterium, and therefore became accumulated within the intracellular fraction. Mass spectrometry analysis confirmed the formation of the GM2 analogue with a peak related to [M + Na] + at m/z 935 in the intracellular fraction exclusively. A high-density culture in Figure 3. Chemical synthesis of NeuNProp 9 and NeuNPhAc 10. i) MeOH, Dowex 50 (H + ), reflux, 37 %; ii) KOH, two-liter reactor was next carEtOH, H2O, reflux; iii) (CH3CH2CO)2O, MeOH, RT, 72 % from 5 or PhCH2COCl, Et3N, MeOH, 0 8C for 10 min and then ried out and the GM2 analogue RT, 50 % from 5; iv) AcOH, H2O, 80 8C, 55 % for 9 and 75 % for 10. 11 (Figure 4) was isolated in 23 % yield (based on the engagtion of the methyl ester first occurred within the first hours ed NeuNprop 9) and characterized by NMR after purification while the de-N-acetylated methyl sialoside 6 was obtained on silica gel chromatography. The 1H NMR spectrum (see Supafter heating under reflux overnight. N-Acylation reactions porting Information S1) of 11 compared with the spectrum of were carried out, using either propionic anhydride or phenylaallyl GM2[38] was characterized by the presence of only one sincetyl chloride in MeOH. NeuNProp 9 and NeuNPhAc 10 were figlet at 2.03 ppm relative to the acetyl group of the GalNAc nally isolated after acid hydrolysis of the methyl glycoside. In unit as well as two additional signals of the propionyl group: spite of moderate yields for both protection and de-protection one triplet centered at 1.13 ppm and a quadruplet centered at of the anomeric position, this synthesis involved only three pu2.32 ppm. Similar modifications are visible on the 13C NMR rification steps and was easily achievable on a gram scale. spectrum with the presence of three characteristic peaks cenNMR data of the products were in good agreement with those tered at 23.9, 10.1 and 29.9 ppm attributed to the acetyl group reported earlier, but resulting from a more complicated synof the GalNAc and propionyl groups, respectively. thetic route.[48] The preparation of N-phenylacetyl GM3 and GM2 analogues NeuNProp 9 was then used as a substrate in bacterial synwas then investigated. Cultures of low cell density with LB1 theses with the E. coli strains LB1 (previously referred to as and TA02 strains did not yield the desired products under conJM107-nanA¢)[31] and TA02[32] producing GM3 and GM2 oligosacventional conditions. The absence of NeuNPhAc 10 in the charides respectively. Low cell density cultures in triple baffled mass spectrum of the intracellular fraction, led to the hypotheErlenmeyer flask were first started to assess the ability of the sis that either this modified neuraminic acid could not be modified acceptor to be internalized and to be glycosylated. transported by the sialic permease (NanT), or that the expresNeuNProp 9 and allyl b-lactoside were added exogenously to sion of the permease was not at all induced by the modified the E. coli culture, at the same time as the protein expression sialic acid, since NanT permease is normally induced by inducer (IPTG) and the culture left to grow for 19 h. After cenNeuNAc. Therefore, two new strains (DC25 and ZWU11), includtrifugation and cell lysis, both intracellular fraction and culture ing a new plasmid where the nanT gene was under the control medium were partially purified on activated charcoal and analyzed by TLC and mass spectrometry. The presence of GM3NProp 3 was evidenced in both the intracellular and extracellular fractions with a characteristic peak of [M + Na] + ion at m/z 732. This finding supports an active internalization of the acceptor into the bacteria, its efficient activation into a CMP donor and its glycosylation by N. meningitidis a-2,3sialyltransferase. The presence of the product in the extracellular medium was explained by passive diffusion across the bacterial membrane, a commonly observed phenomenon during the production of trisaccharides.[49] Culture at low cell density of the GM2 producing strain in presence of NeuNProp 9 and allyl b-lactoside produced the expected tetrasaccharide as well. The GalNAc residue was provided by the endogenous bacterial pool of GlcNAc, which is epimerized by the UDP-GlcNAc C4 epimerase.[32] Monitoring of the reaction by TLC showed an accumulation of the Figure 4. GM3 (3, 4) and GM2 (11, 12) analogues obtained by in vivo bacterial synthesis acceptors into the cytoplasm and their transforma- in E. coli. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10906
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper of the Lac promoter, which is induced by IPTG, were develropoietin (EPO) glycosylated with N-propanoyl or N-pentanoyl oped. Low density culture of these new strains in presence of sialic acids showed a higher resistance to neuraminidase treatNeuNPhAc 10 and allyl b-lactoside efficiently produced the exment than EPO glycosylated with N-acetyl sialic acid.[51] pected tri- and tetrasaccharides as evidenced by TLC and MS These results prompted us to assess the influence of N-acyl spectrometry. Once again the GM3 4 analogue was isolated modification in compounds 3 and 4 on both the Maackia from both intracellular and extracellular fraction while the GM2 amurensis agglutinin (MAA) binding and the stability against derivative 12 accumulated in the cytoplasm only. High density Vibrio cholerae neuraminidase. MAA, which is specific for a-2,3linked sialic acids[52] often serves as a model protein for sialylatculture produced 4 and 12 in 28 and 18 % yield, respectively from NeuNPhAc 10 after purification by ion exchange chromaed oligosaccharides binding studies. The affinity of GM3 anatography and silica gel chromatography. While the production logues 3 and 4 for MAA was probed by solid-phase Enzymeof GM3NPhAc 4 was comparable to that of GM3NAc 1, the synLinked Lectin Assay (ELLA). Bovine fetuin, a highly sialylated thesis of GM2NPhAc 12 was substantially less efficient than glycoprotein was first immobilized into a microtiter plate. Then that of the GM2NAc counterpart.41 Since no remaining starting horseradish peroxidase-labelled MAA was incubated in presence of increasing concentrations of the competitive ligands 1, material was detectable by TLC at the end of the reaction, it is hypothesized that the exogenous acceptors are partially de3 and 4 (Figure 5). Quantification of MAA binding to fetuin graded or transformed into side-products during the fermentawas then revealed by colorimetric reaction in presence of ophenylenediamine dihydrochloride. These analyses were pertion. A minor loss of material during the purification steps is not excluded either. Nevertheless, this bacterial process reformed in triplicate. mains highly competitive in comparison to others enzymatic syntheses because it is particularly cheap, fast and easy to implement on a gram scale. The spectroscopic data of 4 obtained by the fermentation process were identical to those of the trisaccharide obtained by chemical modification of 1. 1H NMR spectrum of GM2NPhAc 12 showed characteristic peaks of additional GalNAc residue. One could notice an additional doublet centered at 4.74 ppm and a singlet at 2.03 ppm corresponding to the proton at the anomeric position and of the N-acetyl group, respectively. Moreover, as expected, the aromatic protons were visible in the downfield region, from 7.31 to 7.44 ppm. The 13C NMR spectrum of 12 compared to the respective spectrum of GM3NPhAc 4 showed an additional peak at 175.5 ppm, corresponding to the Figure 5. Inhibition of bovine fetuin–MAA binding measured by ELLA. carbonyl carbon of the GalNAc residue. Likewise, at the anomeric region four signals were visible, with the additional peak corresponding to the carbon of the GalNAc residue at 103.4 ppm. Other characteristic peaks inComplete inhibition of lectin binding is seldom observed in ELLA assay.[53–55] In the present study, levels of inhibition clude those of the acetyl group of GalNAc and of the aromatic group at 23.3, 128.0 and 135.6 ppm respectively. reached 70 % at 1 mm of inhibitor. Interestingly, both N-Prop and N-PhAc analogues 3 and 4 showed an affinity for MAA similar to that of GM3NAc 1. Even, the bulky aromatic moiety Biological testing of 4 did not hamper the binding to the lectin. These findings Developing compounds able to interact with sialic acid recepsuggest that 3 and 4 might efficiently mimic natural sialoglycotors, but less susceptible to sialidase hydrolysis, are attractive conjugates and could be used as synthetic decoys for biologifor multiple purposes. Both properties are particularly interestcal systems. More importantly, 3 and 4 were stable and remained noning when considering the detection or inhibition of influenza viruses since both sialic acid hemagglutinins and sialidases are hydrolyzed when incubated in presence of Vibrio cholerae (VC) involved in infection by influenza viruses. Such molecules may neuraminidase. GM3NPhAc 4 competitively inhibited the hyalso avoid a rapid clearance from the blood by the asialoglycodrolysis of the fluorogenic substrate 4-methylumbelliferyl-aprotein receptor and would present a long-lasting activity in NeuNAc by the sialidase, with a Ki of 1.04 mm (see Supporting vivo. Previous studies in the literature suggest that modificaInformation S2.1). Since GM3NPhAc 4 is a substrate analogue tion of the N-acyl part of sialic acid can improve the binding to rather than a transition state inhibitor, it is not surprising that biological receptors and the resistance to sialidases. For examits inhibition constant is in the millimolar range only. Indeed, ple, the binding of N-fluoroacetylneuraminic acid to Myelin-asthis value is in good agreement with the Michaelis constant resociated glycoprotein (MAG) is about 17-fold better than that ported for the GM3NAc (Km 1.2 mm).[56] However, the inhibition [50] of N-acetylneuraminic acid. In a recent study, human erythprofile exhibited in Figure 6 clearly indicates a time dependent Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10907
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper inhibition, often referred to as a slow binding inhibition. Similar mode of action has been already observed for neuraminidase targeting molecules.[57, 58] This phenomenon generally appears to be a consequence of the very slow rate of dissociation of the inhibitor from the enzyme. In spite of a poor inhibition constant, 4 durably inactivates VC neuraminidase. Indeed, the hydrolysis of GM3NAc 1 (1.15 mm) was totally abolished after pre-incubation with GM3NPhAc 4 at 144 mm (see Supporting Information S2.2). Simultaneous incubation of substrate and inhibitor with the sialidase led to total inactivation of the enzyme after a transient period of time directly related to the concentration of inhibitor. These results suggest that this new glycomimetic may have long lasting activity in vivo and could certainly be used against influenza viruses. However, further analyses, including structural data, would be required to elucidate the mode of action of GM3NPhAc. One can notice that similar mode of action might also be attributed to GM3NProp 3 although to a lesser extent. Indeed, hydrolysis of GM3NAc by VC neuraminidase is 7-fold more important in presence of the inhibitor GM3NProp 3 than with GM3NPhAc 4 at similar concentration (see Supporting Information S2.2).
Table 1. Sialidase activity of different influenza virus strains on GM3NAc 1 and GM3NPhAc 4.
Substrate fetuin (1 mm) GM3NAc 1 (0.25 mm) GM3NPhAc 4 (0.25 mm)
H1N1 100 700 13
Relative activity [%] H5N2 100 220
Full Paper
& Oligosaccharides
Chemoenzymatic Syntheses of Sialylated Oligosaccharides Containing C5-Modified Neuraminic Acids for Dual Inhibition of Hemagglutinins and Neuraminidases L¦monia Birikaki,[a, b] St¦phanie Pradeau,[a, b] Sylvie Armand,[a, b] Bernard Priem,[a, b] Luis Mrquez-Domnguez,[c] Julio Reyes-Leyva,[c] Gerardo Santos-Lûpez,[c] Eric Samain,[a, b] Hugues Driguez,[a, b] and S¦bastien Fort*[a, b] Abstract: A fast chemoenzymatic synthesis of sialylated oligosaccharides containing C5-modified neuraminic acids is reported. Analogues of GM3 and GM2 ganglioside saccharidic portions where the acetyl group of NeuNAc has been replaced by a phenylacetyl (PhAc) or a propanoyl (Prop) moiety have been efficiently prepared with metabolically engineered E. coli bacteria. GM3 analogues were either obtained by chemoselective modification of biosynthetic Nacetyl-sialyllactoside (GM3NAc) or by direct bacterial synthesis using C5-modified neuraminic acid precursors. The latter
Introduction N-Acetylneuraminic acid (NeuNAc) is the common terminal carbohydrate unit of glycoproteins and glycolipids expressed at the surface of mammalian cells. Generally appended to a galactose or a N-acetyl-galactosamine unit by an a-2,3- or a-2,6-glycosidic linkage, NeuNAc is ideally positioned to interact with proteins. It is thus involved in a variety of normal and pathological processes such as cell–cell adhesion, cell recognition, bacterial infection, viral invasion, inflammation and cancer metastasis.[1, 2] The biological significance of sialylated molecules makes them excellent targets for the design of carbohydrate-based drugs. In particular, glycoconjugates derived from a-2,3-sialyllactose, the saccharidic portion of gangliosides can be used to target certain types of cancer and viruses. [a] Dr. L. Birikaki, S. Pradeau, Dr. S. Armand, Dr. B. Priem, Dr. E. Samain, Dr. H. Driguez, Dr. S. Fort Univ. Grenoble Alpes, CERMAV, 38000 Grenoble (France) Fax: (+ 33) 4-76-54-72-03 E-mail: [email protected] [b] Dr. L. Birikaki, S. Pradeau, Dr. S. Armand, Dr. B. Priem, Dr. E. Samain, Dr. H. Driguez, Dr. S. Fort CNRS, CERMAV, 38000 Grenoble (France) [c] L. Mrquez-Domnguez, Dr. J. Reyes-Leyva, Dr. G. Santos-Lûpez Laboratorio de Biologa Molecular y Virologa Centro de Investigaciûn Biom¦dica de Oriente Instituto Mexicano del Seguro Social Metepec, Puebla (M¦xico) Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/chem.201500708. Chem. Eur. J. 2015, 21, 10903 – 10912
strategy proved to be very versatile as it led to an efficient synthesis of GM2 analogues. These glycomimetics were assessed against hemagglutinins and sialidases. In particular, the GM3NPhAc displayed a binding affinity for Maackia amurensis agglutinin (MAA) similar to that of GM3NAc, while being resistant to hydrolysis by Vibrio cholerae (VC) neuraminidase. A preliminary study with influenza viruses also confirmed a selective inhibition of N1 neuraminidase by GM3NPhAc, suggesting potential developments for the detection of flu viruses and for fighting them.
The modification of cellular glycosylation corresponds indeed to a common phenotypic modification of cancer cells, which mainly affects the outer part of glycans, leading to the expression of tumor-associated carbohydrate antigens (TACAs).[3] These TACAs have become important molecular targets for the development of anti-cancer vaccines.[4–6] GM3 [(Neua2-3)Galb1-4Glc] and GM2 [GalNAcb1-4(NeuAca2-3)Galb4Glc] gangliosides are overexpressed in various types of cancer, namely those of breast, melanoma, ovary, prostate and lung and thus, they have emerged as promising target for cancer vaccines.[7] Synthetic glycoconjugates combining GM2 with an immunogenic carrier (KLH,[8] poliovirus polypeptide sequence[9]) have been synthesized and were shown to induce cell-specific antibodies for human tumor. A GM2-KLH vaccine, administrated with the adjuvant QS-21, induced not only high IgM titers, but also durable IgG antibodies that were shown to induce complement-mediated lysis of GM2-positive tumor cells in most patients.[8] Recently, new potential carbohydrate-based anti-cancer vaccines incorporating unnatural sialic acids have been proposed to address the immunotolerance issues. Indeed, N-modified GM3,[10, 11] GD3[12] and polysialic acid[13] antigens have showed increased immunogenicity, when compared to the N-acetyl oligosaccharides counterparts and have thus been proposed as promising vaccine candidates against melanoma or lung cancers. In this emerging therapy for cancer cells, a synthetic vaccine targeting a TACA analogue would be administrated together with a sugar precursor meant to induce the expression of the artificial TACA derivative in place of the native TACA.[14] Among various N-modified sialic acid
10903
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper containing antigens, the N-phenylacetyl (NPhAc) derivative generally induced the strongest immune response and their anti-sera were the least reactive with the natural GM3 glycans. These results suggest that the GM3NPhAc is an excellent candidate for an antitumor vaccine presenting relatively low autoimmune reactions.[15, 16] Sialylated oligosaccharides are not only attractive targets for active immunotherapy, but they also play a major role in viral infections. Influenza viruses cause seasonal epidemics and occasional threats, as seen with “swine” influenza H1N1, avian influenza H5N1[17] and the very recent H7N9[18] viruses that induced a significant morbidity and mortality in humans. H1N1 flu virus, which caused a world-wide pandemic in 2009, is now a human seasonal flu virus that also circulates in pigs. 2014 was the first year since 2009 where H1N1 has been so predominant in the United States.[19] H7N9 virus was reported for the first time in early 2013, and was confirmed to be a low pathogenic avian influenza virus in poultry. The reported cases of attack by this novel virus have been increasing since the summer of 2013. The virus poses a potential threat to public health, particularly in China, and for this reason, it is attracting a worldwide broad attention. Influenza virus infections are initiated when the virus adsorbs on respiratory epithelial cells by the interaction of its hemagglutinin with the a-2,6 or a-2,3-sialoglycoconjugates. The release and spreading of the virus is subsequently promoted by a neuraminidase.[20] So far, the prevention and treatment against influenza consist of seasonal vaccination and the administration of sialidase inhibitors namely, oseltamivir[21] (Tamiflu) and zanamivir (Relenza),[22] respectively. Despite these treatments, new virus strains are constantly emerging, especially resistant to oseltamivir.[23] Unlike oseltamivir, which can be orally administrated, zanamivir can only be inhaled due to its low bioavailability, limiting the use of this drug. Interfering with both hemagglutinins and sialidases has emerged as a possible alternative strategy to prevent influenza infection. Indeed, liposomes made of sialoglycolipids in which the C3 position of the sialic acid was modified with
an axial fluorine atom inhibited influenza infection at micromolar concentration.[24, 25] This experimental result confirmed that such glycomimetics not only inhibit the attachment of viruses to the cell surface receptor but also disturbs the release of the progeny viruses from infected Mardin–Darby canine kidney (MDCK) cells by inhibiting both hemagglutinin and sialidase. Sialylated galactosides containing C5-diversified sialic acid have also proven efficient tools to conduct specificity studies of sialidases.[26] The ability of various influenza viruses including human and avian influenza A strains to hydrolyze these glycomimetics has been tested, revealing that it showed marked differences depending on the C5 substitution. These results should certainly contribute to the design of new antiviral drugs designed to target specific influenza strains. Interest in modified sialic acids is not restricted to the study of influenza viruses. The adenovirus serotype Ad37, which causes epidemic keratoconjunctivitis, binds to and infects human corneal epithelial (HCE) cells by attaching itself to cellular glycoproteins carrying terminal sialic acids. N-Acyl modified GM3 were chemically synthesized and assayed in Ad37 cell-binding. In addition, compounds bearing small substituents, in particular GM3NProp, were shown to be as effective inhibitors as sialic acid.[27] Taken together, these observations highlight the potential of using modified sialylated oligosaccharides, but the access to these molecules by an efficient process remains an important issue. Common synthetic strategies developed for the production of sialoglycoconjugates and their mimetics generally rely on time-consuming multistep chemical approaches.[11, 27] Alternatively, in vitro enzymatic processes involving NeuAc-aldolase have been successfully developed.[28–30] The latter technique, while being more rapid since it does not require the handling of protecting groups, requires the use of expensive mannosamine precursors, thus hampering syntheses at a multigram scale. In the present work we report a convenient access to C5-modified sialosides either by chemical modification of GM3NAc produced by metabolically engineered E. coli or by direct bacterial synthesis from modified neuraminic acid pre-
Figure 1. Chemo-enzymatic routes towards GM3 and GM2 analogues, incorporating C5-modified neuraminic acids. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10904
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper cursors (Figure 1). Beyond the challenge raised by the synthesis of such molecules, our final goal was to investigate the influence of these C5-modified sialosides on both their binding to hemagglutinins and their stability against neuraminidases.
Results and Discussion Synthesis of GM3 and GM2 oligosaccharide analogues The synthesis of GM3 analogues was undertaken following two different routes. The first approach consisted in the enzymatic synthesis of GM3NAc by metabolically engineered E. coli using b-allyl lactoside and N-acetyl-neuraminic acid (NeuNAc) exogenous acceptors, followed by the chemoselective modification of the neuraminic moiety. Metabolically engineered bacteria have emerged as efficient tools to synthesize complex oligosaccharides at a multigram scale.[31] Engineered E. coli cells behave as living factories, producing oligosaccharides in a onepot procedure, starting from cheap substrates, without requiring the isolation of intermediates or enzymes. The efficiency of such a bacterial approach is far above that of classical in vitro enzymatic processes. The range of oligosaccharides produced by this technique has considerably expanded over the last ten years: among others, they include the production of gangliosides,[32, 33] of blood group antigens,[34, 35] as well as of chitinoligosaccharides.[36, 37] Conjugatable forms of GM3NAc containing an alkyne or alkene-terminated spacer[38] have also been advantageously prepared by this technique, thus leading to a rapid and efficient access to neo-glycoconjugates, such as glycopolymers[39, 40] and glycoproteins,[41] for diagnosis or therapeutic applications. As seen in Figure 2, GM3NAc 1 was prepared by bacterial synthesis from N-acetyl-neuraminic acid and allyl b-lactoside as reported previously.[38] The, de-N-acetylation was then conducted by alkaline hydrolysis affording 2 in 93 % yield. Evidence for this transformation was given by NMR spectroscopy with the disappearance of the signals of both the methyl and carbonyl groups together with the upfield shift of the H-5 proton of the neuraminic residue centered at 2.91 ppm. It is noteworthy that although the acetamido hydrolysis requires strongly alkaline conditions, the allyl group re-
mained intact and was still available for subsequent linking to another biomacromolecule. The selective N-acylation of 2 with propionic anhydride or phenylacetylchloride in methanol afforded 3 and 4 in 89 and 88 % yield respectively. Both compounds displayed their expected characteristics, when analyzed by mass and NMR spectrometry and no by-product resulting from the O-acylation was observed. An alternative approach to the synthesis of GM3NProp 3 and GM3NPhAc 4 was envisaged by internalization and glycosylation of C5-modified sialic acids in E. coli. Enzymatic in vitro glycosylation of modified sialic acids onto glycoprotein glycans by sialyltransferases has already been reported. Synthetic analogues of sialic acids where the C-5 acetyl group had been replaced by formyl-, trifluoroacetyl-, benzyloxycarbonyl-, or aminoacetyl groups have been successfully activated into CMP-derivatives by CMP-sialic acid synthase and transferred by a-2,6 and a-2,3 sialyltransferases onto antifreeze glycoprotein or asialofetuin. The glycosylation efficiency of these modified sialic acids was comparable to that of the unmodified NeuNAc, with only some exceptions.[42] CMP-NeuNProp was also efficiently prepared in vitro using CMP sialate synthase, albeit on a 10 mg scale only.[43] The promiscuity of sialic acid-modifying enzymes has triggered intensive research on in vivo N-acyl engineering of cell-surface sialoglycoconjugates[44, 45] but the large-scale synthesis of N-modified sialooligosaccharides has been poorly investigated so far. NeuNProp 9 and NeuNPhAc 10 were thus prepared by chemical modification of NeuNAc (Figure 3) and tested for the in vivo production of sialomimetics in E. coli. The de-N-acetylated neuraminic acid was previously reported to form an internal Schiff base in water, by condensation of the amino group at C5 with the carbonyl group at C2, yielding 4-hydroxy-5-(1,2,3,4-tetrahydroxybutyl)-D1-pyrroline-2-carboxylic acid thus preventing an efficient acylation reaction.[46] Therefore, NeuNAc was first converted into its a-methyl glycoside. Methyl esterification and O-methyl glycosylation of NeuNAc were conducted in a one-pot procedure with Dowex 50 (H + ) resin in MeOH.[47] The resulting O-methyl sialoside 5 was subsequently de-N-acetylated under alkaline conditions. Saponifica-
Figure 2. Synthesis of GM3NProp and GM3NPhAc oligosaccharides by chemical modification of recombinant GM3NAc. i) KOH, H2O, reflux, 93 %, ii) (CH3CH2CO)2O, MeOH, 89 %, iii) PhCH2COCl, Et3N, MeOH, 88 %. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10905
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper tion into a more polar compound, which did not diffuse outside of the bacterium, and therefore became accumulated within the intracellular fraction. Mass spectrometry analysis confirmed the formation of the GM2 analogue with a peak related to [M + Na] + at m/z 935 in the intracellular fraction exclusively. A high-density culture in Figure 3. Chemical synthesis of NeuNProp 9 and NeuNPhAc 10. i) MeOH, Dowex 50 (H + ), reflux, 37 %; ii) KOH, two-liter reactor was next carEtOH, H2O, reflux; iii) (CH3CH2CO)2O, MeOH, RT, 72 % from 5 or PhCH2COCl, Et3N, MeOH, 0 8C for 10 min and then ried out and the GM2 analogue RT, 50 % from 5; iv) AcOH, H2O, 80 8C, 55 % for 9 and 75 % for 10. 11 (Figure 4) was isolated in 23 % yield (based on the engagtion of the methyl ester first occurred within the first hours ed NeuNprop 9) and characterized by NMR after purification while the de-N-acetylated methyl sialoside 6 was obtained on silica gel chromatography. The 1H NMR spectrum (see Supafter heating under reflux overnight. N-Acylation reactions porting Information S1) of 11 compared with the spectrum of were carried out, using either propionic anhydride or phenylaallyl GM2[38] was characterized by the presence of only one sincetyl chloride in MeOH. NeuNProp 9 and NeuNPhAc 10 were figlet at 2.03 ppm relative to the acetyl group of the GalNAc nally isolated after acid hydrolysis of the methyl glycoside. In unit as well as two additional signals of the propionyl group: spite of moderate yields for both protection and de-protection one triplet centered at 1.13 ppm and a quadruplet centered at of the anomeric position, this synthesis involved only three pu2.32 ppm. Similar modifications are visible on the 13C NMR rification steps and was easily achievable on a gram scale. spectrum with the presence of three characteristic peaks cenNMR data of the products were in good agreement with those tered at 23.9, 10.1 and 29.9 ppm attributed to the acetyl group reported earlier, but resulting from a more complicated synof the GalNAc and propionyl groups, respectively. thetic route.[48] The preparation of N-phenylacetyl GM3 and GM2 analogues NeuNProp 9 was then used as a substrate in bacterial synwas then investigated. Cultures of low cell density with LB1 theses with the E. coli strains LB1 (previously referred to as and TA02 strains did not yield the desired products under conJM107-nanA¢)[31] and TA02[32] producing GM3 and GM2 oligosacventional conditions. The absence of NeuNPhAc 10 in the charides respectively. Low cell density cultures in triple baffled mass spectrum of the intracellular fraction, led to the hypotheErlenmeyer flask were first started to assess the ability of the sis that either this modified neuraminic acid could not be modified acceptor to be internalized and to be glycosylated. transported by the sialic permease (NanT), or that the expresNeuNProp 9 and allyl b-lactoside were added exogenously to sion of the permease was not at all induced by the modified the E. coli culture, at the same time as the protein expression sialic acid, since NanT permease is normally induced by inducer (IPTG) and the culture left to grow for 19 h. After cenNeuNAc. Therefore, two new strains (DC25 and ZWU11), includtrifugation and cell lysis, both intracellular fraction and culture ing a new plasmid where the nanT gene was under the control medium were partially purified on activated charcoal and analyzed by TLC and mass spectrometry. The presence of GM3NProp 3 was evidenced in both the intracellular and extracellular fractions with a characteristic peak of [M + Na] + ion at m/z 732. This finding supports an active internalization of the acceptor into the bacteria, its efficient activation into a CMP donor and its glycosylation by N. meningitidis a-2,3sialyltransferase. The presence of the product in the extracellular medium was explained by passive diffusion across the bacterial membrane, a commonly observed phenomenon during the production of trisaccharides.[49] Culture at low cell density of the GM2 producing strain in presence of NeuNProp 9 and allyl b-lactoside produced the expected tetrasaccharide as well. The GalNAc residue was provided by the endogenous bacterial pool of GlcNAc, which is epimerized by the UDP-GlcNAc C4 epimerase.[32] Monitoring of the reaction by TLC showed an accumulation of the Figure 4. GM3 (3, 4) and GM2 (11, 12) analogues obtained by in vivo bacterial synthesis acceptors into the cytoplasm and their transforma- in E. coli. Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10906
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper of the Lac promoter, which is induced by IPTG, were develropoietin (EPO) glycosylated with N-propanoyl or N-pentanoyl oped. Low density culture of these new strains in presence of sialic acids showed a higher resistance to neuraminidase treatNeuNPhAc 10 and allyl b-lactoside efficiently produced the exment than EPO glycosylated with N-acetyl sialic acid.[51] pected tri- and tetrasaccharides as evidenced by TLC and MS These results prompted us to assess the influence of N-acyl spectrometry. Once again the GM3 4 analogue was isolated modification in compounds 3 and 4 on both the Maackia from both intracellular and extracellular fraction while the GM2 amurensis agglutinin (MAA) binding and the stability against derivative 12 accumulated in the cytoplasm only. High density Vibrio cholerae neuraminidase. MAA, which is specific for a-2,3linked sialic acids[52] often serves as a model protein for sialylatculture produced 4 and 12 in 28 and 18 % yield, respectively from NeuNPhAc 10 after purification by ion exchange chromaed oligosaccharides binding studies. The affinity of GM3 anatography and silica gel chromatography. While the production logues 3 and 4 for MAA was probed by solid-phase Enzymeof GM3NPhAc 4 was comparable to that of GM3NAc 1, the synLinked Lectin Assay (ELLA). Bovine fetuin, a highly sialylated thesis of GM2NPhAc 12 was substantially less efficient than glycoprotein was first immobilized into a microtiter plate. Then that of the GM2NAc counterpart.41 Since no remaining starting horseradish peroxidase-labelled MAA was incubated in presence of increasing concentrations of the competitive ligands 1, material was detectable by TLC at the end of the reaction, it is hypothesized that the exogenous acceptors are partially de3 and 4 (Figure 5). Quantification of MAA binding to fetuin graded or transformed into side-products during the fermentawas then revealed by colorimetric reaction in presence of ophenylenediamine dihydrochloride. These analyses were pertion. A minor loss of material during the purification steps is not excluded either. Nevertheless, this bacterial process reformed in triplicate. mains highly competitive in comparison to others enzymatic syntheses because it is particularly cheap, fast and easy to implement on a gram scale. The spectroscopic data of 4 obtained by the fermentation process were identical to those of the trisaccharide obtained by chemical modification of 1. 1H NMR spectrum of GM2NPhAc 12 showed characteristic peaks of additional GalNAc residue. One could notice an additional doublet centered at 4.74 ppm and a singlet at 2.03 ppm corresponding to the proton at the anomeric position and of the N-acetyl group, respectively. Moreover, as expected, the aromatic protons were visible in the downfield region, from 7.31 to 7.44 ppm. The 13C NMR spectrum of 12 compared to the respective spectrum of GM3NPhAc 4 showed an additional peak at 175.5 ppm, corresponding to the Figure 5. Inhibition of bovine fetuin–MAA binding measured by ELLA. carbonyl carbon of the GalNAc residue. Likewise, at the anomeric region four signals were visible, with the additional peak corresponding to the carbon of the GalNAc residue at 103.4 ppm. Other characteristic peaks inComplete inhibition of lectin binding is seldom observed in ELLA assay.[53–55] In the present study, levels of inhibition clude those of the acetyl group of GalNAc and of the aromatic group at 23.3, 128.0 and 135.6 ppm respectively. reached 70 % at 1 mm of inhibitor. Interestingly, both N-Prop and N-PhAc analogues 3 and 4 showed an affinity for MAA similar to that of GM3NAc 1. Even, the bulky aromatic moiety Biological testing of 4 did not hamper the binding to the lectin. These findings Developing compounds able to interact with sialic acid recepsuggest that 3 and 4 might efficiently mimic natural sialoglycotors, but less susceptible to sialidase hydrolysis, are attractive conjugates and could be used as synthetic decoys for biologifor multiple purposes. Both properties are particularly interestcal systems. More importantly, 3 and 4 were stable and remained noning when considering the detection or inhibition of influenza viruses since both sialic acid hemagglutinins and sialidases are hydrolyzed when incubated in presence of Vibrio cholerae (VC) involved in infection by influenza viruses. Such molecules may neuraminidase. GM3NPhAc 4 competitively inhibited the hyalso avoid a rapid clearance from the blood by the asialoglycodrolysis of the fluorogenic substrate 4-methylumbelliferyl-aprotein receptor and would present a long-lasting activity in NeuNAc by the sialidase, with a Ki of 1.04 mm (see Supporting vivo. Previous studies in the literature suggest that modificaInformation S2.1). Since GM3NPhAc 4 is a substrate analogue tion of the N-acyl part of sialic acid can improve the binding to rather than a transition state inhibitor, it is not surprising that biological receptors and the resistance to sialidases. For examits inhibition constant is in the millimolar range only. Indeed, ple, the binding of N-fluoroacetylneuraminic acid to Myelin-asthis value is in good agreement with the Michaelis constant resociated glycoprotein (MAG) is about 17-fold better than that ported for the GM3NAc (Km 1.2 mm).[56] However, the inhibition [50] of N-acetylneuraminic acid. In a recent study, human erythprofile exhibited in Figure 6 clearly indicates a time dependent Chem. Eur. J. 2015, 21, 10903 – 10912
www.chemeurj.org
10907
Ó 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Full Paper inhibition, often referred to as a slow binding inhibition. Similar mode of action has been already observed for neuraminidase targeting molecules.[57, 58] This phenomenon generally appears to be a consequence of the very slow rate of dissociation of the inhibitor from the enzyme. In spite of a poor inhibition constant, 4 durably inactivates VC neuraminidase. Indeed, the hydrolysis of GM3NAc 1 (1.15 mm) was totally abolished after pre-incubation with GM3NPhAc 4 at 144 mm (see Supporting Information S2.2). Simultaneous incubation of substrate and inhibitor with the sialidase led to total inactivation of the enzyme after a transient period of time directly related to the concentration of inhibitor. These results suggest that this new glycomimetic may have long lasting activity in vivo and could certainly be used against influenza viruses. However, further analyses, including structural data, would be required to elucidate the mode of action of GM3NPhAc. One can notice that similar mode of action might also be attributed to GM3NProp 3 although to a lesser extent. Indeed, hydrolysis of GM3NAc by VC neuraminidase is 7-fold more important in presence of the inhibitor GM3NProp 3 than with GM3NPhAc 4 at similar concentration (see Supporting Information S2.2).
Table 1. Sialidase activity of different influenza virus strains on GM3NAc 1 and GM3NPhAc 4.
Substrate fetuin (1 mm) GM3NAc 1 (0.25 mm) GM3NPhAc 4 (0.25 mm)
H1N1 100 700 13
Relative activity [%] H5N2 100 220
