CUPRA-ZYME: An assay for measuring carbohydrate-active enzyme activities, pathways and substrate specificities

Abstract

Carbohydrate-Active enZymes (CAZymes) are involved in the synthesis, degradation and modification of carbohydrates. They play critical roles in diverse physiological and pathophysiological processes, have important industrial and biotechnological applications, are important drug targets and represent promising biomarkers for the diagnosis of a variety of diseases. Measurements of their activities, catalytic pathways and substrate specificities are essential to a comprehensive understanding of the biological functions of CAZymes and exploiting these enzymes for industrial and biomedical applications. For glycosyl hydrolases a variety of sensitive and quantitative spectrophotometric techniques are available. However, measuring the activity of glycosyltransferases is considerably more challenging. Here, we introduce CUPRA-ZYME, a versatile and quantitative electrospray ionization mass spectrometry (ESI-MS) assay for measuring the kinetic parameters of CAZymes, monitoring reaction pathways and profiling substrate specificities. The method employs the recently developed competitive universal proxy receptor assay (CUPRA), implemented in a time-resolved manner. Measurements of the hydrolysis kinetics of CUPRA substrates containing ganglioside oligosaccharides by the glycosyl hydrolase human neuraminidase 3 served to validate the reliability of kinetic parameters measured by CUPRA-ZYME and highlight its use in establishing catalytic pathways. Applications to libraries of substrates demonstrate the potential of the assay for quantitative profiling of the substrate specificities glycosidases and glycosyltransferases. Finally, we show how the comparison of the reactivity of CUPRA substrates and glycan substrates present on glycoproteins, measured simultaneously, affords a unique opportunity to quantitatively study how the structure and protein environment of natural glycoconjugate substrates influences CAZyme activity.

Graphical Abstract

Introduction

Carbohydrate-Active enZymes (CAZymes) are a large group of enzymes involved in the synthesis, degradation and modification of carbohydrates. The CAZyme database, (CAZy; www.cazy.org) contains sequence information for >340,000 enzymes, which are organized into families defined by sequence similarity, structure and function.¹ The four principal families are glycosyl hydrolases (glycosidases), glycosyltransferases, polysaccharide lyases and carbohydrate esterases.² Carbohydrate binding modules and auxiliary activity families (e.g., ligninolytic and lytic polysaccharide mono-oxygenases) are also part of this enzyme classification system. However, substrate specificity can’t be predicted entirely from sequence, so for the majority of CAZymes their substrate is not known. The most direct route to ascribing a function to a putative CAZyme involves demonstrating enzymatic activity on carbohydrate-based substrates.³

The CAZymes, which exist as both intra- and extracellular enzymes,^4,5 play critical roles in diverse physiological processes in humans. For example, endogenous CAZymes control the glycosylation of proteins and lipids, while exogenous enzymes associated with bacteria are responsible for the digestion of complex carbohydrates in the gut.⁶ CAZymes also have important industrial and biotechnological applications, are important drug targets and represent promising biomarkers for the diagnosis of a variety of diseases, including certain cancers.^7–9 Measurements of the activities, pathways and substrate specificities are fundamental to a comprehensive understanding of the biological functions of CAZymes and exploiting these enzymes for industrial and biomedical applications.^10–13 Despite their importance, deficiencies in the available analytical tools hinder kinetic studies of CAZymes and, as a result, knowledge of the substrate specificities of most of these enzymes is incomplete. In fact, it is estimated that fewer than 1% of the putative CAZymes predicted from DNA sequence data within the CAZy database are associated with any experimental data.¹⁴

Most naturally occurring glycans do not exhibit optical properties and, therefore, enzymatic processing of glycan substrates typically can’t be directly monitored using spectrophotometry-based kinetic assays. Application of optical methods generally requires the introduction of a suitable chromophore (e.g. fluorophore) into the substrate. However, such modifications, depending on their nature and location, may influence CAZyme activity. Moreover, optical methods are susceptible to fluorescence quenching at high concentrations and not readily amenable to analyzing multiple substrates simultaneously. Alternatively, glycan processing kinetics may be monitored using a secondary chemical or enzymatic reaction, which generates an optically detectable signal from the CAZyme product. For example, the release of sialic acid from glycans by a sialidase can be monitored by reacting the monosaccharide with malononitrile to produce a fluorescent product.¹⁵ Multi-enzyme schemes, wherein a second enzyme acts rapidly on the CAZyme product to produce a coloured/fluorescent secondary product, are frequently used for glycosyl hydrolases. However, developing coupled assays for glycosyltransferases is more challenging.¹⁶ Some of the reported methods rely on the detection of the donor leaving group (e.g., the nucleotide diphosphate, UDP) by displacing a corresponding fluorophore-labeled donor product from an immobilized antibody or through some other fluorescence-based readout achieved using coupled chemical/enzymatic reactions.^17,18 DNA-linked enzyme-coupled assays have also been demonstrated for some glycosyltransferases.^19,20 These methods, however, require multiple and typically expensive reagents (i.e., antibodies or enzymes) to implement, may be susceptible to background interferences and may not be readily applied to mixtures of substrates. Radiometric methods, which employ substrates containing a radioactive isotope, have been widely used to study CAZyme-catalyzed reactions in the past.²¹ In such assays, the change in radiation intensity is the evaluated signal. Radio-labeling is attractive in that it does not change the kinetic property of the natural substrate (provided the label does not lead to a significant kinetic isotope effect).²² However, an extraction (or chromatography) step is required in order to separate radiolabeled substrate from the product. Additionally, the availability of radioactive-labelled glycan substrates compounds is often limited and use of the assay is restricted to safety certified labs. Therefore, the demand for more versatile and, in particular, label-free assays for measuring CAZyme activities is high, particularly for enzymes involved in building up glycans, namely the glycosyltransferases.²³

Electrospray ionization mass spectrometry (ESI-MS) has emerged as a versatile, label-free method for monitoring enzyme reactions.^23–33 A unique advantage of ESI-MS is that it allows for the simultaneous detection of substrate and product and, if present at observable concentrations, intermediates. Also, it can be readily implemented on multiple substrates provided they have distinct molecular weights (MWs), thus allowing libraries of substrates to be screened simultaneously. Such an approach is ideal for profiling substrate specificity. However, because of differences in ESI-MS response factors, correlating the measured substrate and product ion abundances to solution concentrations represents a challenge and quantitative monitoring of the enzyme-catalyzed reaction progress generally requires the use of internal standards, which can be expensive or difficult to synthesize, or calibration curves.^28–33

Here, we introduce a versatile and quantitative assay for measuring the kinetic parameters of CAZyme reactions, establishing their catalytic pathways and profiling their substrate specificities. The method, referred to as CUPRA-ZYME (Figure 1a), is based on the recently developed Competitive Universal Proxy Receptor Assay (CUPRA) and time-resolved ESI-MS measurements.³⁴ CUPRA, which employs oligosaccharides labeled at the reducing end with an affinity tag (examples of three different CUPRA linker (CL)-affinity tag combinations, which were used in the present work, are given in Figure 1b) that is recognized by a protein receptor (referred to as the universal proxy receptor, ^UniP_proxy), and direct ESI-MS analysis of the non-covalent complexes of the heterobifunctional glycan ligands (called CUPRA ligands) and ^UniP_proxy. Changes in relative abundances (as measured by ESI-MS) of specific ^UniP_proxy-CUPRA ligand complexes upon introduction of a glycan-binding protein (GBP) serves to identify ligand binding and allow affinities to be quantified. When performed in a time-resolved fashion and in the presence of a CAZyme, CUPRA allows for time-dependent changes in the concentrations of CUPRA ligands that are substrates (i.e., CUPRA substrates, CS (Figure 1c)) to be measured. Because of the ESI-MS response factor-independent manner in which substrate concentration is measured, CUPRA-ZYME eliminates the need for calibration curves or internal standards, which are generally required with ESI-MS-based enzyme kinetics.^28–33 Moreover, the concentrations of products and any intermediates that retain the affinity tag can also be quantified, independent of the nature of the chemical modification catalyzed by the CAZyme.

Figure 1.

CUPRA-ZYME kinetic assay. (a) Workflow of CUPRA-ZYME method, wherein the time-dependent CUPRA substrate (CS) and product (CP) concentrations are determined from the relative abundances of CS and CP bound to a universal proxy protein (^UniP_proxy) measured by ESI-MS. (b) Structures of the three CUPRA linker-affinity tags (CL) used in this work. (c) Structures of three CS used in this work; the glycan moiety is shown in red, the affinity tag in blue and the linker in black.

Kinetic measurements of the hydrolysis of CS prepared from ganglioside oligosaccharides by the glycosyl hydrolase human neuraminidase 3 (NEU3) were used to establish the reliability of CUPRA-ZYME for measuring kinetic parameters and highlight its use for studying multistep enzyme reactions. Measurements performed on human sialytransferases (ST6Gal1 and ST3Gal4) with libraries of CS acceptors demonstrate the potential of the assay to profile the substrate specificities of glycosyltransferases rapidly and quantitatively. Finally, we illustrate how kinetic measurements performed simultaneously on CS and glycans present on glycoproteins offers unprecedented insight into influence of the substrate environment in natural glycoconjugates on CAZyme activity.

Experimental Methods

Materials and Methods

Details on the proteins, CAZymes, carbohydrates and other materials used in this study, as well as the method used to prepare the CUPRA substrates (Table S1 and Figures S1–S3) are given as Supporting Information.

Mass spectrometry

ESI-MS measurements were carried out in positive ion mode using three instruments, a Synapt G2S ESI quadrupole-ion mobility separation-time-of-flight (Q-IMS-TOF) mass spectrometer (Waters, Manchester, UK), a Q-Exactive Ultra High Mass Range (UHMR) Hybrid Quadrupole-Orbitrap and a Q Exactive Classic Hybrid Quadrupole-Orbitrap mass spectrometers (Thermo Fisher Scientific, U.K.), each equipped with a nanoflow ESI (nanoESI) source. A description of the instrumental parameters used for the direct ESI-MS and CUPRA-ZYME assays, as well as procedures used for analyzing the time-resolved mass spectra are given as Supporting Information.

CUPRA-ZYME

To implement the CUPRA-ZYME assay (Figure 1a) aliquots of stock solutions of CAZyme, ^UniP_proxy, CS and Lyz (Lysozyme, P_ref), all in 200 mM aqueous ammonium acetate, were manually mixed in an eppendorf tube and then transferred to the nanoESI tip. Unless otherwise noted, measurements were carried out at pH 7 and 25 °C. ESI mass spectra were collected continuously starting 3 min after mixing. Where required, the mass spectrum was corrected for the occurrence of nonspecific CS-^UniP_proxy binding formed during the ESI process using the reference protein method, which has been described in detail elsewhere.^35–37 Lyz served as P_ref in all kinetic experiments.

Progress curves (plots of concentrations of CS and CUPRA product(s) (CP) versus time) were constructed from the time-resolved ESI mass spectra. The concentrations of CS and CP were calculated according to the total abundances (Ab) of the gas-phase ions (considering all detected charge states) of the corresponding ^UniP_proxy complexes ((^UniP_proxy+CS) and (^UniP_proxy+CP)) and the known initial CS concentration ([CS]₀), eqs 1a and 1b:

$$[\text{CP}]=\frac{{[\text{CS}]}_{0}Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CP})}{Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CP})+Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CS})}$$ (1a)

$$[\text{CS}]=\frac{{[\text{CS}]}_{0}Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CS})}{Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CP})+Ab({}^{\text{Uni}}{\text{P}}_{\text{proxy}}+\text{CS})}$$ (1b)

Kinetic parameters (K_M and v_max) were established by fitting the Michaelis-Menten equation to the initial rates measured over a series of initial CS concentrations, eq 2:

$$v_0=\frac{v_{\max }{[\text{CS}]}_0}{K_M+{[\text{CS}]}_0}$$ (2)

where v₀ is initial velocity, v_max is maximum velocity and K_M is substrate concentration at half v_max. The initial rates were determined from linear best fits of the [CP] versus reaction time plots; unless otherwise noted, data acquired from 3 min to 10 min were used.

Results and discussion

Hydrolysis of 3′-sialyllactose by NEU3

To demonstrate the reliability of CUPRA-ZYME for quantitatively monitoring CAZyme reactions, we studied the human glycosyl hydrolase NEU3. Within cells, NEU3 is a plasma membrane-associated neuraminidase, which preferentially cleaves terminal α2–3-linked Neu5Ac residues on both glycoproteins and glycolipids.^38–40 The ganglioside GM3, which contains α2–3-linked Neu5Ac, is believed to be the natural substrate of NEU3.⁴ Accordingly, we monitored the catalytic activity of human NEU3 towards ${\text{CS}}_{3\text{SL}}^{\text{S}1}$, which contains the trisaccharide Neu5Acα2–3Glcβ1–4GlcNAc (3’-sialyllactose; 3SL).

With the goal of determining K_M, CUPRA-ZYME measurements were performed using ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and human NEU3 at seven different initial concentrations, ranging from 10 μM to 500 μM. Shown in Figure 2a are representative ESI mass spectra measured for aqueous ammonium acetate solutions (200 mM, pH 7 and 25 °C) of NEU3, ${\text{CS}}_{3\text{SL}}^{\text{S}1}$, ^UniP_proxy (hCA1) and P_ref measured at reaction times (t) of 20 min and 40 min. Also shown is the corresponding mass spectrum measured for the same solution but with no NEU3 present (i.e., t = 0 min). Protonated ions corresponding to free ^UniP_proxy and the (^UniP_proxy + ${\text{CS}}_{3\text{SL}}^{\text{S}1}$) complex, at charge states +9 and +10, are evident in the mass spectra. Also, ions of the ^UniP_proxy bound to the lactose-containing CUPRA reaction product, (^UniP_proxy + ${\text{CP}}_{\text{Lac}}^{\text{S}1}$) were also detected at 20 min and 40 min reaction times, as were ions (protonated and sodiated) corresponding to the other reaction product, Neu5Ac. As expected, the relative abundances of the (^UniP_proxy + ${\text{CS}}_{3\text{SL}}^{\text{S}1}$) ions decrease with reaction time, while those of the (^UniP_proxy + ${\text{CP}}_{\text{Lac}}^{\text{S}1}$) and Neu5Ac ions increase. Qualitatively similar results were obtained from mass spectra acquired at higher initial substrate concentrations. However, at concentrations above 100 μM, signal corresponding to nonspecific interactions between ^UniP_proxy and ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ and ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ was evident. For example, as illustrated in Figure S4a, at an initial ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ concentration of 200 μM, signal corresponding to ^UniP_proxy bound to two ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ was observed at early reaction times, while later in the reaction ^UniP_proxy bound to two ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ were detected. At intermediate reaction times, ^UniP_proxy ions bound simultaneously to ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ were detected. To minimize error in the time-dependent ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ concentration measurements resulting from nonspecific binding, the measured distributions of P_ref bound to ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ (Figure S4b) were used to correct the mass spectra.^35–37

Figure 2.

Human neuraminidase enzyme kinetics measured by CUPRA-ZYME. (a) Representative ESI mass spectra acquired for aqueous ammonium acetate solutions (200 mM, pH 7 and 25 °C) of NEU3 (6.25 μg mL⁻¹), ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ (10 μM) and ^UniP_proxy (20 μM). (b) Progress curves ([${\text{CP}}_{\text{Lac}}^{\text{S}1}$] versus reaction time) measured using seven different initial concentration of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ (10 μM, 50 μM, 100 μM, 150 μM, 200 μM, 350 μM and 500 μM). (c) Initial rates determined by linear fitting of progress curves shown in (b) from 3 min to 10 min. (d) Plot of initial rates versus initial ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ concentrations. Red curve corresponds to best fit of eq 2 to the experimental data.

Shown in Figure 2b are the progress curves (time-dependent concentration of ${\text{CP}}_{\text{Lac}}^{\text{S}1}$, calculated from the abundances of the (^UniP_proxy + ${\text{CS}}_{3\text{SL}}^{\text{S}1}$) and (^UniP_proxy + ${\text{CP}}_{\text{Lac}}^{\text{S}1}$) ions using eq 1a and 1b) measured at reaction times ranging from 3 min to 85 min for the seven different initial concentrations (from 10 μM to 500 μM) of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$. Because of ion suppression effects and extensive nonspecific binding, measurements at higher concentrations were not performed. The portions of the progress curves (3 min to 10 min) used to determine v₀ are shown in Figure 2c. Non-linear fitting of the Michaelis-Menten equation (eq 2) to plots of initial rates versus initial substrate concentrations was used to establish the kinetic parameters (Figure 2d). This analysis gave a Michaelis constant (K_M) of 210 ± 62 μM and v_max of 9.8 ± 1.3 μM min⁻¹. In principle, the catalytic constant (k_cat) can be calculated from the nominal molar concentration of NEU3 used for these measurements. However, because the concentration of active enzyme is most certainly lower than the nominal concentration, this approach is likely to underestimate k_cat significantly.

The K_M measured by CUPRA-ZYME agrees, within combined error, with a value (140 ± 30 μM) reported for 3SL functionalized, at the reducing end, with boron-dipyrromethene (BODIPY),⁴¹ and also a value (166 μM) measured for the GM3 ganglioside using a radiometric assay.⁴⁰ It is important to note, however, that reported values were determined at acidic pH, 4.5 and 3.8, respectively, and it is known that the activity of NEU3 is sensitive to pH.^40,42 Therefore, to more conclusively establish that CUPRA-ZYME correctly reports on the time-dependent CS and CP concentrations, the aforementioned measurements were repeated in the presence of known concentration of N-acetyl-D-neuraminic acid-1,2,3-¹³C₃ (Neu5Ac-¹³C₃), which served as an internal standard for the Neu5Ac product. As can be seen from the mass spectra data acquired using an initial ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ concentration of 50 μM (Figure S5a), the total abundance of Neu5Ac ions increases, relative to that of the Neu5Ac-¹³C₃ ions, with reaction time. Based on the reasonable assumption that Neu5Ac and Neu5Ac-¹³C₃ have identical ESI-MS response factors, the time-dependent concentration of Neu5Ac was calculated from the abundance ratio of Neu5Ac and Neu5Ac-¹³C₃ ions and the initial concentration of Neu5Ac-¹³C₃ (eq S1). Shown in Figure S5b is a plot of Neu5Ac concentration versus reaction time. Also shown is the time-dependent concentration of ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ measured independently using CUPRA-ZYME. Notably, the two progress curves are essentially indistinguishable. Similar agreement (between product concentrations measured by CUPRA-ZYME and the internal standard) was obtained using higher initial concentrations of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ (data not shown). Together, these results establish that CUPRA-ZYME provides an accurate measure of the time-dependent CS and CP concentrations in solution.

Because the affinity of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ for the ^UniP_proxy hCA1 is quite high (13 ± 4 μM in 200 mM aqueous ammonium acetate at pH 7 and 25 ºC),³⁴ a substantial fraction of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ will be bound to the ^UniP_proxy, particularly at low initial concentrations. For example, at 10 μΜ ${\text{CS}}_{3\text{SL}}^{\text{S}1}$, 98% of the substrate will be bound to ^UniP_proxy at the start of the reaction. To establish that ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ binding to ^UniP_proxy (hCA1) does not influence the NEU3 kinetics, measurements were carried out in the absence of ^UniP_proxy but in the presence of Neu5Ac-¹³C₃. The reaction progress curves measured for initial ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ concentrations ranging from 10 μM to 500 μM and the corresponding regions from which the initial rates were determined are shown in Figures S6a and S6b, respectively. Fitting of the Michaelis-Menten equation to these data gives a K_M of 274 ± 76 μM and v_max of 9.7 ± 1.3 μM min⁻¹. Importantly, the K_M value agrees, within combined error, with the value measured by CUPRA-ZYME under the same solution conditions, indicating that binding of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ to ^UniP_proxy has little or no effect on the NEU3 activity. This finding may reflect the fast (relative to hydrolysis) on-off kinetics of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ binding to ^UniP_proxy, wherein only free ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ is processed, or suggest that NEU3 is indifferent to ^UniP_proxy binding the substrate. As discussed in more detail below, the latter view is supported by the kinetic data acquired for ${\text{CS}}_{3\text{SL}}^{\text{B}}$.

pH profile of NEU3.

Because the affinity of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ (and the resulting ${\text{CP}}_{\text{Lac}}^{\text{S}1}$) for hCA1 decreases substantially with decreasing pH, CUPRA-ZYME measurements are not easily performed near the optimal pH (approximately 4.8) of NEU3 or other neuraminidases using CS^S1 and hCA1 as ^UniP_proxy.^43,44 This pH restriction was overcome using a CS^B substrate (${\text{CS}}_{3\text{SL}}^{\text{B}}$), which contains a biotin affinity tag (Figure 1) and mSA as the ^UniP_proxy. Although the affinity of ${\text{CS}}_{3\text{SL}}^{\text{B}}$ for mSA wasn’t measured in the present work, the reported affinity of biotin is 7.7 nM and is expected to remain high over a wide range of pH.⁴⁵ Prior to applying CUPRA-ZYME with ${\text{CS}}_{3\text{SL}}^{\text{B}}$ and mSA to assess the activity pH profile of NEU3, we first sought to compare the NEU3 hydrolysis kinetics of ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CS}}_{3\text{SL}}^{\text{B}}$ at neutral pH. To do this, the assay was performed on a solution containing both ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CS}}_{3\text{SL}}^{\text{B}}$, along with their corresponding ^UniP_proxy (mSA and hCA1). The results, which are summarized in Figure S7, show that the progress curves measured with both substrate-^UniP_proxy pairs are very similar. This finding suggests that, at least in the case of NEU3, the structural differences between ${\text{CS}}_{3\text{SL}}^{\text{S}1}$ and ${\text{CS}}_{3\text{SL}}^{\text{B}}$ and the interaction with ^UniP_proxy do not significantly influence enzyme activity.

To establish a NEU3 activity pH profile, CUPRA-ZYME was performed on aqueous ammonium acetate solutions (25 °C) of NEU3, ^UniP_proxy (mSA) and ${\text{CS}}_{3\text{SL}}^{\text{B}}$ at pH of 4.0, 4.5, 5.0, 6.0 and 7.0. Shown in Figure S8a are representative mass spectra measured under the most acidic condition, at 3 min, 20 min and 40 min reaction times. Because of the high affinity of mSA for the substrate and resulting CP (${\text{CP}}_{\text{Lac}}^{\text{B}}$), only bound ^UniP_proxy was observed in the mass spectra. The progress curve measured at each solution pH is shown in Figure S8b. Using the initial rates determined from these curves, the influence of pH on the relative activity of NEU3 was assessed (Figure S8c). It can be seen that NEU3 activity is highest at pH 5.0, which is consistent with previous reports of the optimal pH of this enzyme being in the 4.5 to 5.0 range.^38,40,42 While there is only a modest decrease in activity from pH 5.0 to pH 4.5, the enzyme loses almost all activity at pH 4.0. It can also be seen that NEU3 remains quite active at higher pH, with relative activities (compared to pH 5.0) of approximately 70% and 60% at pH 6.0 and 7.0, respectively.

Hydrolysis of GD3 tetrasaccharide by NEU3: evidence of endoneuraminidase activity

A unique feature of CUPRA-ZYME is the ability to simultaneously quantify not only the substrate(s) and product(s) of a given CAZyme but also reaction intermediates (i.e., enzymatic products that are also substrates of the CAZyme) that are produced. Moreover, the assay can be used to study CAZyme pathways that involve multiple enzymes and enzymatic steps. To illustrate the ability of CUPRA-ZYME to monitor a multistep CAZyme reaction, we applied the assay to the hydrolysis of the CS^S1 substrate containing the GD3 ganglioside tetrasaccharide Neu5Acα2–8Neu5Acα2–3Galβ1–4Glcβ-OH (${\text{CS}}_{\text{GD}3}^{\text{S}1}$). This tetrasaccharide contains both an external α2–8- and an internal α2–3-linked Neu5Ac. NEU3 is generally considered to be an exoneuraminidase and is expected to first cleave the terminal Neu5Ac of ${\text{CS}}_{\text{GD}3}^{\text{S}1}$ to give ${\text{CP}}_{3\text{SL}}^{\text{S}1}$, which itself is a substrate (i.e., ${\text{CP}}_{3\text{SL}}^{\text{S}1}$ ≡ ${\text{CS}}_{3\text{SL}}^{\text{S}1}$) and will be converted to ${\text{CP}}_{\text{Lac}}^{\text{S}1}$.²

Shown in Figure 3a are representative mass spectra measured for aqueous ammonium acetate solutions (200 mM, pH 7 and 25 °C) of NEU3 (6.25 μg mL⁻¹), ${\text{CS}}_{\text{GD}3}^{\text{S}1}$ (90 μM) and ^UniP_proxy (hCA1, 20 μM) measured at 3 min, 40 min and 80 min. Analysis of the mass spectra show that consumption of ${\text{CS}}_{\text{GD}3}^{\text{S}1}$ is accompanied by the formation of ${\text{CP}}_{\text{Lac}}^{\text{S}1}$, as well as ${\text{CP}}_{3\text{SL}}^{\text{S}1}$. Inspection of the corresponding plots of concentration versus reaction time (Figure 3c) reveals that the concentration of ${\text{CP}}_{3\text{SL}}^{\text{S}1}$ initially increases and then decays at longer times. This is accompanied by a slight delay (induction period) in the appearance of ${\text{CP}}_{\text{Lac}}^{\text{S}1}$. These findings are consistent with the stepwise loss of Neu5Ac from ${\text{CS}}_{\text{GD}3}^{\text{S}1}$. However, inspection of the low m/z region of the mass spectra reveals the presence of ions (protonated and sodiated and protonated with loss of water) corresponding to dineuraminic acid (Neu5Acα2–8Neu5Ac) (Figure 3b). Dineuraminic acid can only be formed by cleavage of the internal α2–3-linkage. The concentration of dineuraminic acid could not be reliably determined from the mass spectra because a suitable (and stable) internal standard was not available. Nevertheless, it can be seen that the relative abundance of dineuraminic acid ions initially increase and then decrease with time. This behavior is consistent with dineuraminic acid being a substrate of NEU3 and converted to Neu5Ac. To the best of our knowledge, this represents the first report of a human neuraminidase exhibiting endoneuraminidase activity.

Figure 3.

Endoneuraminidase activity of NEU3 revealed by CUPRA-ZYME. (a) Representative mass spectra acquired for aqueous ammonium acetate solutions (200 mM, pH 7 and 25 °C) of NEU3 (6.25 μg mL⁻¹), ^UniP_proxy (hCA1, 20 μM) and ${\text{CS}}_{\text{GD}3}^{\text{S}1}$ (90 μM) measured at 3 min, 20 min and 40 min reaction time. (b) The region shown (m/z 580 – m/z 630) contains the dineuraminic acid ions. (c) Time-dependent concentrations of ${\text{CS}}_{\text{GD}3}^{\text{S}1}$ (shown in black), ${\text{CP}}_{3\text{SL}}^{\text{S}1}$ (blue) and ${\text{CP}}_{\text{Lac}}^{\text{S}1}$ (red).

Substrate specificities of human sialyltransferases

Because CUPRA-ZYME allows multiple substrates to be monitored simultaneously, provided they have different MWs, the assay is ideally suited to quantitatively profile the substrate specificities of CAZymes. This approach, which we recently described for NEU2 and NEU3,³⁴ is illustrated here for two human sialyltransferase, ST6Gal1, which transfers Neu5Ac in an α2–6-linkage from a CMP-Neu5Ac donor to a Galβ1–4GlcNAc acceptor,⁴⁶ and human sialyltransferase ST3Gal4, which transfers Neu5Ac in an α2–3-linkage from CMP-Neu5Ac to Galβ1–4GlcNAc, Galβ1–3GlcNAc or Galβ1–3GalNAc acceptors.^47,48 A library of 20 CS (5 μM for each, structures shown in Table S1) was split into 5 smaller libraries according to their MWs and each was incubated with ^UniP_proxy (hCA1, 20 μM), ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ (5 μM, which served as the reference substrate) and ST6Gal1 or ST3Gal4 at pH 7 and 25 °C. The relative (to ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$) reactivities of all tested CUPRA substrates, determined from the first 5 min of the reaction, are listed in Table S2.

The time-resolved CUPRA-ZYME data (Figure 4) show that, of the CS tested, only ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ and ${\text{CS}}_{\text{LNnT}}^{\text{S}1}$ (reactivity of 60% compared to ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$) serve as acceptors for ST6Gal1. These results are consistent with the reported finding that ST6Gal1 is highly specific for substrates possessing a terminal Galβ1–4GlcNAc disaccharide.⁴⁶ Moreover, the lack of reactivity observed for ${\text{CS}}_{\text{LxTria}}^{\text{S}1},{\text{CS}}_{\text{LxTetra}}^{\text{S}1}$ and ${\text{CS}}_{\text{LxHexa}}^{\text{S}1}$ indicate that fucosylation of the GlcNAc residue introduces steric effects not tolerated by ST6Gal1. ST3Gal4 requires the participation of divalent cations for optimal activity (data not shown),^49–51 thus MnCl₂ (125 μM) was added to the solutions to test this enzyme. ST3Gal4 was found to exhibit somewhat more relaxed substrate specificities. But, like ST6Gal1, ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ is preferred and ${\text{CS}}_{\text{LNnT}}^{\text{S}1}$ is a reasonably good substrate (relative reactivity 60%). However, unlike ST6Gal1, ST3Gal4 is able to transfer Neu5Ac with reasonably high efficiency to substrates containing terminal Galβ1–4Glc (${\text{CS}}_{\text{Lac}}^{\text{S}1}$, 69%; ${\text{CS}}_{\text{LNT}}^{\text{S}1}$, 9%) and Galβ1–3GlcNAc (${\text{CS}}_{\text{LNAc}2}^{\text{S}1}$, 12%; ${\text{CS}}_{\text{Gb}5}^{\text{S}1}$, 4%) motifs.

Figure 4.

Substrate specificities of human sialyltransferases ST6Gal1 and ST3Gal4 measured by CUPRA-ZYME. Time-dependent fractional concentrations of 21 CS (Table S1, Supporting Information) in aqueous ammonium acetate solutions (200 mM, pH 7 and 25 °C) of CS (each 5 μM), ^UniP_proxy (hCA1, 20 μM) and (a) ST6Gal1 (42 μg mL⁻¹) or (b) ST3Gal4 (200 μg mL⁻¹) and MnCl₂ (125 μM).

Comparison of CUPRA and glycoprotein substrates

The functional characterization of CAZymes is usually performed using model substrates, which typically comprise the minimal glycan structure recognized by the enzyme. However, there have been few attempts to correlate the kinetics measured for these model substrates and those of the natural glycans, particularly in the case of N- and O-linked glycans associated with proteins. Because of the amenability of CUPRA-ZYME to be performed simultaneously with non-CUPRA substrates (e.g. glycoproteins), it is ideally suited to quantitatively assess the relative reactivities of model substrates and natural glycan substrates.

To illustrate how CUPRA-ZYME, combined with high resolution ESI-MS, can be leveraged to gain insight into the influence of substrate environment in natural glycoconjugates on CAZyme activity, we first compared the hydrolysis kinetics of human prostate specific antigen (PSA) by NEU3 with those of CS possessing α2–3- and α2–6-linked Neu5Ac. PSA, a 28.4 kDa glycoprotein (237 amino acids) produced by the prostate epithelial cells, is a widely used biomarker for prostate cancer screening.^52,53 The protein possesses a single glycosylation site at Asn 45 occupied by bi-antennary complex N-linked glycan.^54,55 The commercial sample of PSA consists of two major species with MWs of 28400.50 Da and 28283.93 Da, which correspond to bi-antennary fucosylated (Hex₅HexNAc₄Fuc₁α-Neu5Ac₂) and non-fucosylated (Hex₅HexNAc₄α-Neu5Ac₂) PSA (Figure S9). The Neu5Ac residues in PSA are known to be both α2–3- and α2–6-linked.

To assess the reactivity of the Neu5Ac-containing N-glycan substrates in PSA at neutral pH, NEU3 kinetic measurements were performed on a solution containing PSA, ${\text{CS}}_{6\text{SL}}^{\text{S}1}$ and ${\text{CS}}_{3\text{SLNAc}}^{\text{S}2}$, together with hCA1 (Figure S10). As expected, the hydrolysis of ${\text{CS}}_{3\text{SLNAc}}^{\text{S}2}$ is faster, approximately 3.5 times, than of ${\text{CS}}_{6\text{SL}}^{\text{S}1}$ (Figure 5a).³⁴ The progress curve measured for PSA exhibits behavior that is consistent with the presence of both the more reactive α−2–3-linked Neu5Ac and the less reactive α2–6-linked Neu5Ac. Analysis of the kinetic data indicates that α2–3-linked Neu5Ac in the bi-antennary N-glycans of PSA, which represents 51% of the Neu5Ac in the glycoprotein, is 5.2 times more reactive than ${\text{CS}}_{3\text{SLNAc}}^{\text{S}2}$, while the α2,6-linked Neu5Ac is 2.7 times less reactive than ${\text{CS}}_{6\text{SL}}^{\text{S}1}$ (Table S3). These results show that the relative hydrolysis kinetics of α2–3- and α2–6-linked Neu5Ac within glycoproteins may differ substantially. In the case of PSA, there is a 50-fold difference between the reactivity of α2–3- and α2–6-linked Neu5Ac. In contrast, for the CS investigated, there is only a 3.6-fold difference.

Figure 5.

Comparison of reactivity of CS and natural glycoconjugates evaluated by CUPRA-ZYME and ESI-MS. (a) Time-dependent fractional concentration of Neu5Ac in the N-glycans of PSA measured in an aqueous ammonium acetate solution (200 mM, pH 7 and 25 °C) of PSA (10 μM), ^UniP_proxy (hCA1, 10 μM), ${\text{CS}}_{6\text{SL}}^{\text{S}1}$ (20 μM), ${\text{CS}}_{3\text{SLNAc}}^{\text{S}2}$ (20 μM) and NEU3 (6.25 μg mL⁻¹). (b) Time-dependent fractional concentration of N-glycan acceptor substrate in asialo-PSA measured in an aqueous ammonium acetate solution (200 mM, pH 7 and 25 °C) of asialo-PSA (5 μM), ^UniP_proxy (hCA1, 5 μM), MnCl₂ (125 μM), ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ (5 μM), CMP-Neu5Ac (200 μM) and ST3Gal4 (200 μg mL⁻¹). (c) Time-dependent fractional concentration of N-glycan acceptor substrate in asialo-PSA measured in an aqueous ammonium acetate solution (200 mM, pH 7 and 25 °C) of asialo-PSA (5 μM), ^UniP_proxy (hCA1, 5 μM), MnCl₂ (125 μM), ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ (5 μM), CMP-Neu5Ac (200 μM) and ST6Gal1 (42 μg mL⁻¹).

To demonstrate how this same approach is applicable to glycosyltransferases, we compared the rates of Neu5Ac transfer from CMP-Neu5Ac to asialo-PSA and ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ by ST6Gal1 and ST3Gal4 at neutral pH. The asialo-PSA was prepared by incubating PSA with the bacterial NEU overnight, followed by mild heating to deactivate the enzyme. Analysis of the kinetic data in Figure 5b reveals that asialo-PSA and ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ exhibit similar reactivity (i.e., k = 1.4k₁, where k is the rate constant for transfer of α2,3-linked Neu5Ac to ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ and k₁ is the apparent rate constant for transfer of α2,3-linked Neu5Ac to asialo-PSA with ST3Gal4 (Table S3). However, it can also be seen that conversion of monoasialo-PSA to di-sialylated PSA proceeds with very poor efficiency (k₂ = 0.017k₁, where k₂ is the apparent rate constant for transfer of α2,3-linked Neu5Ac to monoasialo-PSA). The corresponding mass spectra show that mono-sialylated PSA is the only product detected at long reaction times (Figure S11). Moreover, the progress curve for the formation of ${\text{CP}}_{3\text{SLNAc}}^{\text{S}2}$ (Figure 5b) confirms that the absence of PSA product is not due to a loss in ST3Gal4 activity. Rather, the results indicate that the remaining branch of mono-sialylated PSA is a very poor substrate for ST3Gal4. Because this behavior was not observed (data not shown) for free bi-antennary N-glycan substrates (Galβ1–4GlcNAcβ1–2Manα1–3(Galβ1–4GlcNAcβ1–2Manα1–6)Manβ1–4GlcNAcβ1–4GlcNAc and Neu5Acα2–6Galβ1–4GlcNAcβ1–2Manα1–3(Galβ1–4GlcNAcβ1–2Manα1–6)Manβ1–4GlcNAcβ1–4GlcNAc), the present results suggest that PSA is responsible for blocking the transfer of the second Neu5Ac, presumably through steric effects.

Inspection of the mass spectra shown in Figure S12, which reveals both mono-sialylated and di-sialylated PSA ions at long reaction times, suggests that ST6Gal1 is significantly more efficient than ST3Gal4 at transferring two molecules of Neu5Ac to the N-glycan substrates of asialo-PSA. However, analysis of the time-resolved data (Figure 5c) shows that the apparent rate constant for the addition of the second Neu5Ac (i.e., k₂) is only 3-fold larger than that for ST3Gal4 (Table S3). The greater efficiency by which ST6Gal1 produces di-sialylated PSA originates primarily from the much faster transfer of the first Neu5Ac (k₁ = 75k₂). This strong preference of ST6Gal1 for the asialo-PSA substrate is also reflected in the much slower conversion kinetics measured for ${\text{CS}}_{\text{LNAc}}^{\text{S}2}$ (k = 0.16k₁, Table S3).

Taken together, the results obtained for the CS and monoasialo-PSA and PSA demonstrate that the reactivity of simple model substrates may differ significantly from that of natural glycoconjugates. The current work also highlights how CUPRA-ZYME, combined with high resolution ESI-MS, provides a straightforward approach to quantifying the influence of N-glycan structure and local (protein) environment on CAZyme activity.

Conclusions

This work introduces a powerful MS-based assay for measuring CAZyme activities, monitoring reaction pathways and identifying the formation of intermediates and quantitatively profiling their substrate specificities. Because of the novel manner in which the glycan substrate and corresponding product concentrations are measured, the assay is insensitive to differences in their ESI-MS response factors, regardless of the type of the chemical modification catalyzed by the enzyme. As a result, there is no requirement for calibration curves or the use of internal standards. Currently, due to ion suppression effects and nonspecific association during the ESI process, implementation of the assay is restricted to initial CS concentrations ≤500 μM. However, efforts to extend the accessible concentration range through the use of sub-μm nanoESI emitters are ongoing.^56–58

Measurements of the hydrolysis kinetics of a CS containing the GM3 oligosaccharide by NEU3 established the reliability of CUPRA-ZYME for measuring kinetic parameters. Application of the assay to a CS composed of the GD3 oligosaccharide, which provided the first direct evidence of endoneuraminidase activity by NEU3, highlighted the advantages of CUPRA-ZYME for studying multistep reactions. The implementation of the assay using human sialyltransferases ST6Gal1 and ST3Gal4 and libraries of CS served to demonstrate the ease with which the substrate specificities of CAZymes can be quantitatively profiled. Finally, the potential of CUPRA-ZYME, when combined with high resolution ESI-MS analysis, to quantify the relative reactivity of CS and natural glycoconjugates was illustrated for the desialylation of PSA (by NEU3) and sialylation of asialo-PSA (by ST6Gal1 and ST3Gal4). Notably, these results showed that, in some cases, the reactivity of the bi-antennary N-glycan substrates exhibit significant, kinetic differences compared to the structurally simpler CS tested.