Selenoglycosides as Lectin Ligands: ⁷⁷Se‐Edited CPMG‐HSQMBC NMR Spectroscopy To Monitor Biomedically Relevant Interactions

Abstract

The fundamental importance of protein–glycan recognition calls for specific and sensitive high‐resolution techniques for their detailed analysis. After the introduction of ¹⁹F NMR spectroscopy to study the recognition of fluorinated glycans, a new ⁷⁷Se NMR spectroscopy method is presented for complementary studies of selenoglycans with optimised resolution and sensitivity, in which direct NMR spectroscopy detection on ⁷⁷Se is replaced by its indirect observation in a 2D ¹H,⁷⁷Se HSQMBC spectrum. In contrast to OH/F substitution, O/Se exchange allows the glycosidic bond to be targeted. As an example, selenodigalactoside recognition by three human galectins and a plant toxin is readily indicated by signal attenuation and line broadening in the 2D ¹H,⁷⁷Se HSQMBC spectrum, in which CPMG‐INEPT long‐range transfer ensures maximal detection sensitivity, clean signal phases, and reliable ligand ranking. By monitoring competitive displacement of a selenated spy ligand, the selective ⁷⁷Se NMR spectroscopy approach may also be used to screen non‐selenated compounds. Finally, ¹H,⁷⁷Se CPMG‐INEPT transfer allows further NMR sensors of molecular interaction to be combined with the specificity and resolution of ⁷⁷Se NMR spectroscopy.

Glycans are increasingly recognised for their capacity to encode biological information at high density, thus attracting interest in understanding how it is specifically read and translated into a plethora of physiological effects.1 For instance, the recognition of cellular glycoconjugates by various endogenous sugar receptors2 underlies specific bridging in the cell–cell/matrix adhesion and initiation of signalling in diverse regulatory processes such as the induction of anoikis/apoptosis, or mediator release in degenerative and inflammatory diseases like (osteo)arthritis.3 The analysis of such clinically relevant molecular recognition processes and screening for potent antagonists with maximal specificity, sensitivity, and resolution are still great experimental challenges on the way to cracking the sugar code and its biomedical implications.2c, 4

Solution‐state NMR spectroscopy can probe molecular interactions with atomic resolution over a wide range of affinities,5 and NMR screening is now widely used in pharmaceutical research. It may be implemented with detection of suitable proteins, which requires their isotopic labelling, evinces binding sites, and includes irreversible binding. Alternatively, the unlabelled small ligands are observed, which allows faster throughput by pooling and spectrally separating compounds, and thus, is particularly suited to detect weak binding, but often overlooks strong binding because only the dissociated state is detected. Inherent problems of ¹H NMR spectroscopy, where signals often show poor dispersion, complex fine structure (from extensive homonuclear J(H,H) coupling), and overlap (aggravated by a vast spectral background and intense solvent signals), further complicate ligand identification, especially in mixtures of chemically related compounds. These shortcomings are avoided upon selecting another NMR‐active nucleus with superior signal dispersion and no natural background, which may be introduced by chemical derivation. For glycans with their particularly complex and similar ¹H NMR spectra,6 the substitution of hydroxyl groups for fluorine or oxygen for selenium atoms are established modifications, for example, to inhibit enzymatic degradation. This also introduces the extremely well dispersed spin‐$1/2$ NMR isotopes ¹⁹F and ⁷⁷Se, with chemical shift ranges of δ≈300 and 3000 ppm, respectively. For ¹⁹F, high natural abundance (100 %) and sensitivity (83 % of ¹H) facilitate direct NMR detection and have boosted the development and broad application of diverse ¹⁹F NMR spectroscopy techniques for efficient screening and detailed analysis of molecular interactions.7 For the rare (7.63 % abundance) and insensitive (0.7 % of ¹H) ⁷⁷Se isotope,8 in contrast, direct ⁷⁷Se NMR detection9 is unfavourable. To enhance the sensitivity substantially and enable reliable ligand ranking by affinities, we propose indirect ⁷⁷Se detection and initial ¹H→⁷⁷Se polarisation transfer, as implemented in the 2D ¹H,⁷⁷Se HSQMBC experiment with CPMG‐INEPT out‐and‐back transfer.

Despite great success, broad applications, and the advanced technical state of ¹⁹F NMR based interaction studies, analogous ⁷⁷Se NMR approaches may become an important complementary technique because the required OH/F and O/Se substitutions label different molecular sites for NMR observation and with distinct biochemical impacts. In particular, whereas OH/F substitution removes a potential hydrogen‐bond donor that may impair or even abolish molecular recognition, O/Se exchange in the glycosidic bond has no known adverse effects, but increases its stability against hydrolysis. Several previous findings indicate that selenoglycosides are excellent biomimetics of natural O‐glycosides and exhibit similar dynamic and conformational properties to those of thioglycosides,10 whereas molecular docking analyses predict full biocompatibility with O‐glycosides.11 Methyl selenoglycosides co‐crystallise readily with their cognate bacterial, fungal, and human lectins to enable multi‐wavelength anomalous diffraction (MAD) phasing of their complex crystal structures.12 The diselenodiglycoside of β‐GlcNAc was shown to be a ligand for wheat germ agglutinin by means of ¹H saturation transfer difference (STD) NMR spectroscopy, whereas its presence in serum was traced by means of ⁷⁷Se NMR spectroscopy.13 Similar to their thio analogues,10, 14 selenodigalactoside (SeDGal; Scheme 1) and diselenodigalactoside (DSeDGal) were recognised by human galectins and a ricin‐like plant toxin,15 in contrast to diselenodiglucoside (DSeDGlc). At the same time, selenoglycoside synthesis16 has advanced greatly, and ⁷⁷Se NMR signal shifts were shown to be sensitive indicators for phenylselenyl acetate binding to α‐chymotrypsin17 and selenomethyl glycoside binding to plant lectins.9 All studies, however, employed very insensitive direct ⁷⁷Se NMR detection.

Scheme 1.

Structures of the selenodiglycosides used in this study.

To further explore the NMR spectroscopy potential of selenated compounds far more efficiently, we herein propose indirect ⁷⁷Se detection through ¹H and initial ¹H→⁷⁷Se polarisation transfer to increase the sensitivity by 12=(γ _H/γ ${}_{{}^{77}\mathrm{Se}}$ )^3/2 and 5.23=(γ _H/γ ${}_{{}^{77}\mathrm{Se}}$ ), respectively, that is, by a factor of 63 overall. Yet, the approach relies on moderate scalar two‐ and three‐bond ^2, 3 J(H,⁷⁷Se) couplings (10–30 Hz) for out‐and‐back transfer of ¹H coherence to a ⁷⁷Se spin in the glycosidic bond that require rather long transfer periods 2Δ≈(^2,3 J(H,⁷⁷Se))⁻¹=30–100 ms. Transfer efficiency is then degraded by transverse ¹H R ₂ relaxation and by competing evolution of homonuclear J(H,H) coupling that causes periodic signal phase and amplitude modulation. The use of CPMG‐INEPT18 for ¹H→⁷⁷Se coherence transfer can alleviate these deleterious effects of COSY‐type J(H,H) evolution by enforcing TOCSY‐type evolution instead, which is purely in‐phase and eventually leads to equipartition of magnetisation in the coupled ¹H spin system. This removes signal phase modulation and greatly reduces intensity losses from co‐evolving J(H,H) coupling. Moreover, CPMG‐INEPT can suppress line broadening from chemical exchange (R _exch) that often increases the net ¹H R ₂* relaxation rate in cases of transient interactions provided that the CPMG echo delay (τ in Figure 1) is shorter than the lifetime of interchanging states (τ _exch).19 If this condition is maintained for all ligands binding to the same site of a protein, relaxation losses are minimised and R ₂* is reduced to a population‐weighted sum of slow R _2,f and fast R _2,b relaxation rates for the free (f) and bound (b) ligand state, respectively [Eq. (1)]:

$${\displaystyle R{}_2{}^{*}=p{}_{\mathrm{f}}R{}_{2,\mathrm{f}}+p{}_{\mathrm{b}}R{}_{2,\mathrm{b}}=R{}_{2,\mathrm{f}}+p{}_{\mathrm{b}}(R{}_{2,\mathrm{protein}}-R{}_{2,\mathrm{f}})}$$ (1)

Figure 1.

Pulse sequence of the 2D ¹H,⁷⁷Se CPMG‐HSQMBC experiment. Narrow and broad bars indicate 90 and 180° pulses, respectively, applied with phase x, unless indicated otherwise. ϕ ₁ is incremented according to the XY16 scheme20 used in the CPMG‐INEPT module. Further phases: ϕ ₂=y; ϕ ₃=x, −x; ϕ ₄=x, x, −x, −x; ϕ ₅=x, x, x, x, −x, −x, −x, −x; ϕ _rec=x, −x, x, −x, −x, x, −x, x. The last 90° pulse on ⁷⁷Se (grey) is an optional clean in‐phase (CLIP) purge pulse21 to destroy residual 2 H_x,ySe_z anti‐phase components. The CPMG echo delay, τ, was set to 150 μs, while the number, n, of CPMG echos was adjusted to achieve the desired ^2,3 J(H,⁷⁷Se) coupling evolution time, Δ≤0.5/^2, 3 J(H,⁷⁷Se). Clean, phase‐sensitive ⁷⁷Se coherence selection is achieved by setting gradient pulse strengths G ₂/G ₄=80: ±15.257 (with alternating sign between serial free induction decays (FIDs)). The z‐spoil gradients G ₁ and G ₃ have arbitrary but distinct strengths. Sine bell‐shaped gradient pulses of 1 ms duration were followed by a recovery delay of 200 μs. For ⁷⁷Se composite pulse decoupling (CPD) during FID acquisition, the WALTZ16 (or 64) scheme with a 90° pulse length of 0.5/BW(⁷⁷Se) was used, in which BW is the bandwidth [Hz] of the ⁷⁷Se spectrum.

For small ligands (R _2,f«R _2,protein) in the weak binding regime (p _b≈K _A), R ₂* then becomes proportional to the affinity constant, K _A [Eq. (2)]:

$${\displaystyle R{}_2{}^{*}\approx R{}_{2,\mathrm{f}}+K{}_{\mathrm{A}}R{}_{2,\mathrm{protein}}}$$ (2)

Thus, the signal intensity (I) of a ligand in the presence of a weakly binding protein is attenuated according to Equation (3):

$${\displaystyle I\propto \exp (-2\Delta [R{}_{2,\mathrm{f}}+K{}_{\mathrm{A}}R{}_{2,\mathrm{protein}}])}$$ (3)

whereas the intensity in the absence of the protein (I ₀) is attenuated only by slow R _2,f relaxation [Eq. (4)]:

$${\displaystyle I{}_0\propto \exp (-2\Delta R{}_{2,\mathrm{f}})}$$ (4)

The R _2,f dependence then cancels out in the ratio of the signal intensity of a ligand in the presence and absence of a cognate protein [Eq. (5)]:

$${\displaystyle I/I{}_0\approx \exp (-2\Delta K{}_{\mathrm{A}}R{}_{2,\mathrm{protein}})}$$ (5)

Thus, relative signal attenuations from binding allow a ranking of two ligands, a and b, that bind to the same site of a protein, and an estimation of the ratio of their affinity constants [Eq. (6)]:

$${\displaystyle K{}_{\mathrm{A},\mathrm{a}}/K{}_{\mathrm{A},\mathrm{b}}\approx \ln (I/I{}_0){}_{\mathrm{a}}/\ln (I/I{}_0){}_{\mathrm{b}}}$$ (6)

This approximation holds for small ligands in the weak binding regime [Eq. (2)] if line broadening from chemical exchange (R _exch) is suppressed. By ensuring this as well as undistorted signal shapes, CPMG‐INEPT enhances sensitivity and is critical for reliable ligand ranking by relative affinities. Nevertheless, R _exch is not suppressed during chemical shift evolution, thus adding to R ₂* relaxation [Eq. (2)] of ⁷⁷Se (t ₁) and ¹H (t ₂) as extra line broadening. Although this enhances indication sensitivity for molecular interactions, it biases signal intensities, but not their integrals. Thus, Equations (4) to (6) strictly apply to signal integrals, but are also approximately valid for signal intensities.

The 2D ¹H,⁷⁷Se CPMG‐HSQMBC (Figure 1) is derived from our previously presented 2D CPMG‐HSQMBC experiment with composite 90°_y−180°_x−90°_y inversion pulses for increased CPMG bandwidth at reduced power and clean coherence selection by echo/antiecho gradients (G ₂, G ₄).22 Contrary to the original experiment, which was developed to accurately measure long‐range coupling constants, a refocusing second CPMG‐INEPT module is inserted prior to FID acquisition to enable ⁷⁷Se decoupling in order to obtain sharp and purely in‐phase signals for easy quantification of signal intensities or integrals.

A 2D ¹H,⁷⁷Se HSQMBC spectrum of SeDGal and DSeDGal (Figure 2) illustrates the significant phase distortions (in the ¹H dimension) and sensitivity losses due to amplitude modulation from COSY‐type co‐evolution of homonuclear ⁿ J(H,H) couplings during conventional INEPT transfer. In contrast, CPMG‐INEPT suppresses these deleterious effects and allows a quantitative analysis of the substantially enhanced signal intensities.

Figure 2.

The 2D ¹H,⁷⁷Se HSQMBC spectra of DSeDGal (red, 10 mm) and SeDGal (blue, 10 mm), implemented with INEPT (left) or CPMG‐INEPT (right) modules. The ¹H→⁷⁷Se polarisation transfer delay, Δ, was set to 54.5 ms. The 1D ¹H (F2) traces extracted at both ⁷⁷Se resonances are shown below the contour plots to illustrate the suppression of signal phase and amplitude modulation from co‐evolution of J(H,H) couplings during 2 Δ afforded by CPMG‐INEPT. Both spectra were recorded and plotted with identical parameters.

We then recorded 2D ¹H,⁷⁷Se CPMG‐HSQMBC spectra to analyse binding of SeDGal, DSeDGal, and SeDGlc (Scheme 1) to three human galectins and to Viscum album agglutinin (VAA). Both selenogalactosides show distinct signal attenuation after addition to Gal‐3, an anti‐apoptotic and pro‐inflammatory effector and Gal‐1 antagonist,16 at a molar excess of 20:1 (Figure 3, right, top) or 60:1 (Figure 3, right, bottom). The extracted ¹H[⁷⁷Se] traces readily show strong signal attenuation for SeDGal (blue: I/I ₀≈8 % at 20:1 molar ratio) and weak attenuation for DSeDGal (red: I/I ₀≈84 % at 20:1 molar ratio), whereas the SeDGlc signals (black) remain unaffected. The extent of binding‐induced signal attenuations agree well with previously reported IC₅₀ values,15a and Equation (6) suggests about 15‐fold higher affinity for SeDGal than for DSeDGal. In the presence of Gal‐1 (Figure S1 in the Supporting Information), the signal attenuations for SeDGal (at 20:1 molar ratio: I/I ₀≈10 %) and DSeDGal (I/I ₀≈90 %) are very similar to those caused by Gal‐3, thus confirming a similar differential, albeit somewhat weaker, recognition by this related human galectin. In contrast, Gal‐7 (Figure S2) provokes less attenuation of SeDGal signals (at 20:1 molar ratio: I/I ₀≈36 %), whereas DSeDGal signals are again attenuated only weakly (I/I ₀≈92 %), thus suggesting that Gal‐7 discriminates less between SeDGal and DSeDGal than Gal‐1 or Gal‐3. Finally, VAA causes comparable signal attenuation for SeDGal and DSeDGal (Figure S3: I/I ₀≈22 and 27 %, respectively), which confirms a very similar affinity of VAA for these two ligands, in contrast to the galectins.15 As a negative control, none of the four lectins affected SeDGlc signal intensities, in line with their specific galactose, but not glucose, recognition. In summary, induced signal attenuations observed in 2D ¹H,⁷⁷Se CPMG‐HSQMBC spectra appear to be a valid indicator of ligand binding and allow their reliable ranking by relative affinities.

Figure 3.

Selenoglycoside binding to Gal‐3 monitored by 2D ¹H,⁷⁷Se CPMG‐HSQMBC. The 2D spectra of the ligand mixture containing DSeDGal (red), SeDGal (blue), and SeDGlc (black) at 2.5 mm concentrations each were recorded before (left) and after (right) adding Gal‐3 to a protein/ligand ratio of 1:20 (top) and 1:60 (bottom). ¹H (F2) traces extracted at the three ⁷⁷Se signals are shown next to the contour plots. The indicated attenuated signal intensities from binding are relative to the free ligand intensity. Each 2D ¹H,⁷⁷Se CPMG‐HSQMBC spectrum was measured and plotted with identical parameters. Further experimental details can be found in the Supporting Information.

Our experiments with mixtures of three selenoglycosides implicitly register their competitive displacement if they share a common binding site. To verify this assumption, we also recorded signal attenuations in the 2D ¹H,⁷⁷Se CPMG‐HSQMBC spectrum after addition of only DSeDGal (2.5 mm) to a lectin (125 μm; 30 μm for Gal‐3 to avoid its precipitation), and after subsequent coaddition of SeDGal (2.5 mm).The latter caused a significant rebound of the DSeDGal signal attenuation (I/I ₀) with all four test lectins, that is, from 55 to 91 % for Gal‐1 (Figure S4), 67 to 86 % for Gal‐3 (Figure 4), 89 to 97 % for Gal‐7 (Figure S5), and 12 to 37 % for VAA (Figure S4). Thus, SeDGal displaces DSeDGal from all four lectins, which confirms their competitive binding at the same recognition site. This makes DSeDGal a suitable spy ligand to indirectly monitor galectin binding by means of 2D ¹H,⁷⁷Se CPMG‐HSQMBC through its competitive displacement, which allows application of this specific and sensitive detection method to be extended to non‐selenated compounds as well. In such assays with a single selenated spy ligand, the 1D version of the experiment, that is, a ¹H[⁷⁷Se‐filtered] CPMG‐HSQMBC, should be more efficient than the 2D version. However, the latter is generally preferable for mixtures of selenated compounds to resolve ¹H signal overlap as the most important source of information bias or loss.

Figure 4.

Competitive displacement of diselenodigalactoside (DSeDGal) binding to Gal‐3, monitored by 2D ¹H,⁷⁷Se CPMG‐HSQMBC. A) ¹H (F2) traces of DSeDGal (red, top) and SeDGal (blue, bottom) in the absence of Gal‐3, yielding the reference signal intensities, I ₀=100 %. B) The ¹H (F2) trace of DSeDGal (2.5 mm) after adding Gal‐3 (30 μm, that is, molar ratio=0.75:60) reveals a binding‐induced signal attenuation to I/I ₀=67 %. C) Equimolar addition of SeDGal (2.5 mm) causes a rebound of the attenuated DSeDGal signal to 86 % (red, top), thus indicating its competitive displacement. The SeDGal spectrum is conversely attenuated to 57 % (blue, bottom), which confirms its preferred binding by Gal‐3.

The proposed HSQMBC scheme for indirect ⁷⁷Se detection with initial ¹H→⁷⁷Se polarisation transfer by CPMG‐INEPT may easily be combined with other sensors of molecular interaction than the signal attenuation (from increased R ₂ relaxation) exploited here. For instance, intermolecular ¹H_protein→¹H_ligand saturation transfer can be combined with the extreme spectral resolution, clarity, and specificity of ⁷⁷Se editing in a 2D STD‐¹H,⁷⁷Se‐HSQMBC experiment in order to study transient molecular interactions without spectral background and artefacts. For analyses of stable protein–ligand complexes, the same CPMG‐HSQMBC building block may be used, for example, in a diagonal free 2D ¹H[¹³C],¹H[⁷⁷Se] NOESY experiment to specifically observe intermolecular NOE contacts between a ¹³C‐labelled protein and a selenated ligand.

In conclusion, we have presented an optimised general scheme for the specific study of selenated compounds by means of ⁷⁷Se NMR spectroscopy, with substantially increased sensitivity and resolution compared with that of the previously proposed direct 1D ⁷⁷Se detection method.9 The proposed 2D ¹H,⁷⁷Se CPMG‐HSQMBC experiment instead detects ⁷⁷Se indirectly in a second spectral dimension, by employing ¹H→⁷⁷Se CPMG‐INEPT out‐and‐back transfer through ^2,3 J(H,⁷⁷Se) long‐range coupling. As shown for the example of selenoglycoside recognition by three clinically relevant human galectins and the plant toxin VAA, this experiment enables efficient screening of selenated ligands, sensitive detection of molecular binding, and reliable ligand ranking by relative affinities. The range of applications is vast and similar to that of ¹⁹F NMR‐based techniques, whereas selenium introduction through O/Se exchange has far less potential impact on molecular interactions and can target other sites than OH/F exchange. This may be particularly important and beneficial in glycosciences, where O/Se exchange in the glycosidic bond is now well established and allows glycan/protein recognition to be studied directly at this often critical linkage, and without risk of impairing molecular interactions, in contrast to OH/F exchange.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

M. Raics, I. Timári, T. Diercks, L. Szilágyi, H.-J. Gabius, K. E. Kövér, ChemBioChem 2019, 20, 1688.