Laforin is a human dual-specificity phosphatase (DSP) involved in glycogen metabolism regulation containing a carbohydrate-binding module (CBM). Mutations in the gene coding for laforin are responsible for the development of Lafora disease, a progressive fatal myoclonus epilepsy with early onset, characterized by the intracellular deposition of abnormally branched, hyperphosphorylated insoluble glycogen-like polymers, called Lafora bodies. Despite the known importance of the CBM domain of laforin in the regulation of glycogen metabolism, the molecular mechanism of laforin–glycogen interaction is still poorly understood. Recently, the structure of laforin with bound maltohexaose was determined and despite the importance of such breakthrough, some molecular interaction details remained missing. We herein report a thorough biophysical characterization of laforin–carbohydrate interaction using soluble glycans. We demonstrated an increased preference of laforin for the interaction with glycans with higher order of polymerization and confirmed the importance of tryptophan residues for glycan interaction. Moreover, and in line with what has been described for other CBMs and lectins, our results confirmed that laforin–glycan interactions occur with a favourable enthalpic contribution counter-balanced by an unfavourable entropic contribution. The analysis of laforin–glycan interaction through the glycan side by saturation transfer difference (STD)–NMR has shown that the CBM-binding site can accommodate between 5 and 6 sugar units, which is in line with the recently obtained crystal structure of laforin. Overall, the work in the present study complements the structural characterization of laforin and sheds light on the molecular mechanism of laforin–glycan interaction, which is a pivotal requisite to understand the physiological and pathological roles of laforin.
Laforin is a unique human dual-specificity phosphatase (DSP), as it contains an N-terminal carbohydrate-binding module (CBM) and a C-terminal HCXXGXXR(S/T) phosphatase catalytic active-site motif [1–3]. EPM2A, the gene coding for laforin was first identified as associated with an early onset, fatal progressive myoclonous epilepsy with autosomal recessive inheritance called Lafora disease [4–6]. The pathological hallmark of the disease is the appearance of deposits of insoluble hyperphosphorylated glycogen exhibiting sparse branching in the cytoplasm of almost every cell type (called Lafora bodies), suggesting the involvement of laforin in glycogen dephosphorylation in vivo [4–7].
Laforin has been shown to bind glycogen and its hyperphosphorylated form in Lafora bodies, as well as plant starch and amylopectin through its N-terminal CBM [2,8–10], which belongs to the CBM20 family  (CAZy database: www.cazy.org/CBM20.html). Typically these CBMs, also known as starch-binding domains , have seven β-strands forming an open-sided distorted β-barrel [3,12], where the glycan binding and recognition is invariantly mediated by aromatic amino acid side chains, such as those from tryptophan, tyrosine and, less frequently, phenylalanine, forming the hydrophobic scaffold of CBM-binding sites [11,13]. Apart from their classification in different families, a structural and functional classification has been proposed for CBMs grouping them into: ‘surface-binding’ CBMs (type A), ‘glycan chain-binding’ CBMs (type B) and ‘small-sugar-binding’ CBMs (type C), with laforin and the rest of CBM20 family members being classified as type B–glycan chain binders [11,13].
In addition to its ability to bind glycogen, laforin was shown to down-regulate various proteins involved in glycogen synthesis, in particular by forming a complex with malin, an E3 ubiquitin ligase . This laforin–malin complex was shown to be responsible for down-regulating the cellular glucose uptake through modulation of the subcellular localization of glucose transporters , as well as for negatively regulating the levels of several glycogenic proteins. Among known targets are R5/PTG, the glycogen targeting subunit of type 1 protein phosphatase (PP1) [16–19]; neuronatin ; R6, another PP1 glycogenic targeting subunit ; glycogen debranching enzyme (GDE/AGL) ; and the glycogen synthase itself . However, most of these targets, which were first identified in cell-culture experiments, were not confirmed in either laforin or malin knockout mice [23–26].
Nevertheless, the involvement of laforin and malin in regulating glycogen synthesis is evidenced by the rescue of symptoms and other disease hallmarks when the glycogen synthase gene was knocked out in a Lafora disease animal model . The importance of regulating neuronal glycogen synthesis surpasses Lafora disease, as anomalous neuronal glycogen deposits are a common hallmark of other conditions and diseases, such as aging, anoxia, ischaemia, schizophrenia, Alzheimer's disease and others [28,29].
Interestingly, laforin CBM seems to be of utmost importance for this glycogenesis regulatory activity, as nearly one-third of known EPM2A mutations are concentrated in exon 1 which encodes most of the CBM of laforin . Moreover, the retained ability of a phosphatase-deficient laforin mutant to recruit malin, target it to glycogen-related enzymes [17,19] and rescue EPM2A−/− mice from Lafora disease , suggests that phosphatase activity is dispensable for the down-regulation of glycogenesis by laforin, further reinforcing the relevance of its CBM for this regulatory function.
The interaction of laforin with glycans is still poorly characterized as most binding studies have been limited to laforin interaction with large and complex glycans or inferred from glycan interaction studies with other CMB20 containing enzymes . More recently, the crystallographic structure of laforin in the presence of maltohexaose was determined . The structure clearly evidences the importance of the CBM residues Trp32, Lys87, Trp99, Asp107 and Glu103 for the CBM–glycan interaction and the authors were able to demonstrate that laforin has a stable dimeric quaternary structure, mediated by the DSP domain . Nevertheless, a better understanding of the mode and dynamics of interaction between laforin and glycans is essential to elucidate how laforin interacts with glycogen, the molecular mechanisms governing glycogen metabolism and if/how these can be eventually modulated to ameliorate Lafora disease phenotypes.
With this in mind, we have biophysically characterized in detail the interaction of laforin with a series of soluble glycans by the standard approaches used to study such interactions. We have demonstrated that, in accordance with other CBM20 members, laforin has an increased preference for glycans with increasing degree of polarization and that binding is mediated by favourable enthalpic and unfavourable entropic contributions. We have also established the binding constants for a series of malto-oligosacharides, along with a detailed characterization of the interaction from the perspective of the glycans.
Glycans (sucrose, trehalose, cellobiose, maltotriose, maltotetraose, maltopentaose, maltohexaose, maltoheptaose, α-, β- and γ-cyclodextrins) were purchased from Carbosynth Limited.
Protein expression and purification
Expression and purification of full-length laforin was performed as previously described . Briefly, full-length laforin cDNA cloned into pET21a(+) and devoid of any fusion tag was expressed in Escherichia coli BL21 star (DE3; Thermo Fisher Scientific) as inclusion bodies. After extensive washing with TN (50 mM Tris, 150 mM NaCl, pH 7.4) and TNT (50 mM Tris, 150 mM NaCl, pH 7.4, 0.1% Triton X-100) buffers, the inclusion bodies were dissolved into 8 M urea buffer (8 M urea, 0.1 M Tris, 1 mM glycine, 1 mM EDTA, pH 10.5) and 2-mercaptoethanol added to a final concentration of 100 mM. The denatured protein was rapidly diluted (20-fold) into refolding buffer (20 mM Tris, 0.5 mM oxidized glutathione, 1.25 mM reduced glutathione, 0.5 mM DTT) and the pH slowly adjusted to 8.0. The protein was then kept at 4°C until purification.
Protein purification was achieved by first concentrating the protein by tangential flow ultrafiltration (Sartocon Slice, Sartorius), followed by N2 pressurized stirred cell concentrator (Amicon 8200, Millipore). Concentrated protein was clarified by ultracentrifugation (100000 g) and applied to a HiLoad 26/60 Superdex 200 pg size exclusion column (GE Healthcare Life Sciences), equilibrated with 20 mM Tris, 0.4 M urea, pH 8.0, to remove aggregated protein. The collected fractions with phosphatase activity were pooled and applied to a Mono Q HR 5/50 GL anion exchange column (GE Healthcare Life Sciences), equilibrated with 20 mM Tris, 0.4 M urea, pH 8.0, the protein being eluted by a linear 0–0.5 M NaCl gradient.
Determination of protein concentration
The concentration of purified protein was determined by measurement of A280 with a Nanodrop 1000 (Thermo Scientific) spectrophotometer using a molar absorption coefficient of 73340 M−1·cm−1 calculated by the software Vector NTI Advance 11 (Thermo Fisher Scientific) using the protein amino acid sequence as input.
Microscale thermophoresis (MST) experiments were performed using a Monolith NT.LabelFree instrument (Nanotemper Technologies GmbH) as reviewed by Jerabek-Willemsen et al. . Laforin was mixed with appropriate buffer to a final concentration of 0.5 μM and glycans were titrated in a 1:1, 16-fold, serial dilution and loaded into standard treated glass capillaries. After 15 min of sample equilibration, a thermal gradient of about 2°C–3°C was induced over a period of 30 s using an IR laser (20% intensity setting) at room temperature . Intrinsic tryptophan and tyrosine fluorescence was excited at 280 nm (light-emitting diode set to 20% intensity) and emission at 360 nm was recorded. The intrinsic fluorescence signal was used to measure the thermophoretic movement of laforin. The thermophoretic movement of laforin alone differs significantly from the thermophoresis of a laforin–sugar complex due to sugar binding-induced changes in size, charge and solvation energy. This glycan binding-induced change of thermophoresis was measured during the glycan titration series, plotted against the respective glycan concentrations and then used to derive the dissociation constant Kd by fitting the obtained data with the law of mass action . For buffer screening, laforin affinity towards maltoheptaose was measured in 50 mM sodium phosphate buffer, pH 6.5–7.5, and in 50 mM Tris/HCl buffer, pH 7.5–8.5. Laforin glycan size preference was measured using linear glycans ranging from 2 to 7 glucose units and cyclodextrins with 6–8 sugar units. Results were analysed using the NanoTemper Analysis version 1.2.231.
Absorbance scans were collected with a Varian Cary 100 UV/visible spectrophotometer between selected wavelengths with a spectral bandwidth of 0.5 nm, an average integration time of 0.2 s, a data interval of 0.2 nm and temperature set to 20°C. UV-difference scans of laforin (4.3 μM) in the absence and in the presence of 5 mM γ-cyclodextrin or maltoheptaose were obtained between 270 and 300 nm in 50 mM phosphate buffer, pH 7.5, and corrected for buffer-only baseline or buffer with γ-cyclodextrin prior to UV difference calculation.
UV-difference titrations were performed using 7 μM laforin in a 3.5 ml quartz-suprasil cuvette (Hellma) containing 2 ml volume, in 50 mM phosphate buffer, pH 7.5, under stirring with stepwise addition of titrating glycan. Difference spectra were obtained by first collecting an initial baseline spectrum of 7 μM laforin that was baseline corrected for buffer and subsequent scans collected after sequential addition of ligand. The samples were allowed to equilibrate for 2 min after addition of the ligand. Each scan was first corrected for dilution and subtracted for the scan at zero-ligand concentration. The trough to peak heights at ΔA293–ΔA289 were plotted compared with the total ligand concentration and dissociation constants were calculated by fitting the data to a one-site binding non-linear regression equation using Graph Pad Prism 4 software (Graph Pad software).
The secondary structure of laforin in solution in the absence and presence of glycans was checked by CD on an Olis DSM-20 CD spectrophotopolarimeter controlled by the GlobalWorks software. Far-UV CD spectra were recorded between 190 and 260 nm with a step-resolution of 1 nm, a slit-width of 0.6 nm and an integration time of 7 s. Acquisition was performed at 25°C using a 0.2-mm path length cuvette with a protein concentration of approximately 0.3 mg/ml. Spectra were run in the absence of glycans and also in the presence 5 mM maltoheptaose or 5 mM γ-cyclodextrin. The spectra were averaged over two scans and corrected by subtraction of the buffer signal.
The near-UV titration of laforin with maltoheptaose and γ-cyclodextrin was performed on an Applied Photophysics (Leatherhead) Chirascan-plus controlled by Pro-Data Chirascan v4 software. Near-UV spectra were recorded between 250 and 320 nm with a step-resolution of 0.1 nm and a slit-width of 1 nm and an integration time of 2 s. Acquisition was performed in triplicate at 25°C using a 1 cm path length cuvette with a protein concentration of approximately 0.4 mg/ml. The spectra were averaged over three scans and corrected by subtraction of the buffer signal. The titration was done by consecutive addition of 5 mM oligosaccharide in 0.4 mg/ml laforin to keep laforin concentration constant, followed by 1 min stirring and 3 min equilibration prior to spectra collection. Data analysis and correction was performed using Applied Photophysics Pro-Data Viewer and the variation of CD signal at 283 nm upon glycan addition was plotted, with fitting and affinity constant calculations performed by GraphPadPrism 4.0 (GraphPad Software) by fitting the data to a one-site binding non-linear regression equation.
The CD thermal melt of laforin in the absence and presence of glycans was performed on an automated Applied Photophysics Chirascan-plus ACD system controlled by Pro-Data v5 software with autosampler module. Far-UV spectra were recorded between 200 and 250 nm with a step-resolution of 1 nm and a slit-width of 1 nm and an integration time of 0.5 s. Acquisition was performed in duplicate from 20°C to 93°C in 1°C interval, using a 0.4 cm path length cuvette with a protein concentration of approximately 0.4 mg/ml. Data analysis was performed using Applied Photophysics Global 3 Analysis software.
Isothermal titration calorimetry
Isothermal titration calorimetry (ITC) experiments were performed with an iTC200 (Microcal/Malvern). Glycans were diluted in 20 mM phosphate buffer, pH 7.5, and titrated into 433 μM laforin, equilibrated in the same buffer. Titrations were performed by injecting 2 μl of aliquots of either 4 mM γ-cyclodextrin or 4 mM maltoheptaose solutions into the sample cell (volume 200 μl). The injections were performed at 25°C, stirring at 1000 rpm and with a 150 s recovery interval. An initial injection of ligand (0.4 μl) was made to avoid equilibration artifacts and discarded during data analysis. All ITC data were corrected for dilution and the binding stoichiometry (n), enthalpy (ΔH) and dissociation constant (Kd) were determined by fitting the corrected data to a single binding site model using the Microcal ITC200 Origin software provided by the manufacturer.
Differential scanning calorimetry
The thermal unfolding profiles of laforin were measured using an automated capillary differential scanning calorimeter (VP-capDSC; Microcal/Malvern). Prior to analysis, laforin was desalted using 5 ml HiTrap desalting column (GE Healthcare Life Sciences) and the protein concentration adjusted to 0.5 mg/ml. For DSC analysis of laforin in the presence of glycans, each glycan was weighted and diluted in either desalting buffer (20 mM sodium phosphate buffer, pH 7.4, 50 mM NaCl) in order to act as baseline buffer; or directly in desalted protein solution. Scans were performed from 10°C to 95°C at 1.8°C/min using no feedback mode, with three buffer compared with buffer scans obtained prior to each protein compared with buffer scan. Laforin differential scanning calorimetry (DSC) data were corrected for buffer baselines, using the third buffer compared with buffer run to correct each protein run. The standard Microcal Origin software (version 7) was used for data acquisition and analysis.
Initially the ligand-based 1H-NMR experiments were carried out at 5°C using a Bruker Avance III 500 NMR spectrometer with a TCI cryoprobe and a Agilent VNMRS-600 NMR Spectrometer equipped with a RT 5-mm pulse field gradient (PFG) triple resonance inverse probe, with 1H-frequencies of 500.132 MHz and 599.72 MHz respectively. For purposes of increased sensitivity and resolution, the TOCSY, saturation transfer difference (STD) and STD–TOCSY experiments were repeated using a Bruker DRX 700 NMR spectrometer with a 5 mm TCI cryoprobe, 1H-13C/15N/D with shielded Z-gradient, with 1H-frequency of 700.13 MHz (CERM, University of Florence). The software used for acquisition was TopSpin 1.3 PL 10.
All samples for ligand-based NMR experiments were made up to 200 μl in 3 mm (for the 500 MHz), to 600 μl in 5 mm (for the 600 MHz) or to 290 μl in 3 mm (for the 700 MHz) NMR tubes containing 0.1 mM (trimethylsilyl)-propionic acid-d4 (TSP) or 0.1 mM sodium 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS), for calibration purposes.
For each ligand sample a 1D 1H spectrum was obtained and full spectral assignments were obtained for maltose and maltotriose, using 2D gCOSY, TOCSY and 1H-13C HMQC spectra, whereas for maltotetraose, maltopentaose and maltoheptaose, the majority, but not all the assignments, were obtained due to many signal overlaps which could not be resolved (see Supplementary Information). Spectra were acquired with water suppression using a W5 Watergate (WATER suppression by GrAdient-Tailored Excitation) gradient spin-echo pulse sequence for the 500 MHz spectrometer  and a double PFG spin echo (DPFGSE) sequence  for the 600 and 700 MHz spectrometers. 1D STD NMR experiments [37,38] were performed at 500, 600 and 700 MHz. On the 500 MHz spectrometer, STD experiments employed a 40 ms selective Gaussian 180° shaped pulse at a frequency alternating between ‘on resonance’ (0.5 ppm) and ‘off resonance’ (~80 ppm) after every scan. On the 600 MHz spectrometer, STD spectra were acquired directly from phase cycling, employing a 50 ms selective Gaussian 180° shaped pulse at a frequency alternating between ‘on resonance’ (–2.515 Hz, 0.8 ppm) and ‘off resonance’ (22.000 Hz, ~40 ppm) after every scan. Typical acquisition parameters included equal spectrometer gain value, a spectral window of 8.389 Hz (14 ppm), 2048 scans (preceded of 64 dummy scans), with 2 s acquisition and a repetition delay of 1.5 s. A spin-lock filter (T1ρ, previously calibrated) of 30 ms at 30 dB was applied to remove protein resonances. On the 700 MHz spectrometer, the 1D STD experiments used a spin-lock filter at ~7 kHz with spin lock time of 30 ms and a selective Gaussian 180° pulse with saturation time of 2 s and ‘on’ and ‘off resonance’ frequencies of 0.6 ppm and 30 ppm respectively and 2048 scans were acquired. For reasons of sensitivity and resolution, STD–TOCSY spectra  were acquired on the 700 MHz spectrometer, with a saturation time of 2.5 s, ‘on’ and ‘off resonance’ frequencies of 0.3 and 30 ppm respectively, an acquisition matrix of 2048×256 points and 96 scans. Normalization for different number of acquisitions was performed as previously described . Typical final concentrations for STD–NMR experiments were 1.2 mM of glycan, 24 μM of laforin in 20 mM phosphate buffer, pH 7.5, containing 10% 2H2O for field/frequency lock. Relaxation-edited Carr–Purcell–Meiboom–Gill (CPMG) NMR experiments [41,42] incorporated a CPMG spin-lock time of 200 ms before the acquisition period.
RESULTS AND DISCUSSION
Laforin has a preference to interact with glycans with higher polymerization degree
Laforin glycan binding was first evaluated by MST, as this technique allows immobilization-free quantitative analysis of protein interactions in solution with very low sample consumption [33,43]. For protein–glycan interactions there is the additional advantage of label-free interaction analysis, as protein intrinsic tryptophan fluorescence can be used to monitor the thermophoresis process [33,43]. To test the optimum interaction buffer conditions, the analysis of dissociation constants for laforin interaction with maltoheptaose in buffers of variable pH was performed, using a Monolith NT.Label-Free MST (Nanotemper, GmbH) system, which revealed a preference for the interaction at pH 7.5 (Figure 1A; Supplementary Table S1), consistent with the physiological pH of the cytosol and nucleus  where laforin localizes .
Laforin–glycan interaction affinity measurements by MST
After determining the optimum pH of interaction, we then assessed laforin binding with increasing chain length linear and cyclic glycans. As shown in Figure 1(B) and Supplementary Table S2, an increased affinity towards glycans with higher degree of polymerization (DP) was observed. No binding was detected for sucrose, trehalose or cellobiose (result not shown), which is consistent with the binding affinities reported for other CBM20 family members [11,13], where increased affinity up to hexasaccharides has been frequently demonstrated, with negligible interaction with glycans with a DP of three or less [11,13].
In the recent report describing laforin structure concordant data was obtained when protein melting temperature was measured in the presence of increasing DP glycans . Despite no affinity constants being measured, the authors noticed negligible protein stabilization determined for laforin incubated with glycans with DP 2–4 and increasing protein stability when incubated with increasing DP .
Laforin CBM has been grouped along with other non-amylolytic low-affinity CBM20 containing enzymes . For this subgroup of enzymes, glycan-binding affinity data is only available for Arabidopsis glucan water dikinase 3 (GWD3), which has been shown to bind to cyclodextrins with affinities in the same order of magnitude as those herein obtained for laforin . However, GWD3 CBM displayed a clear preference towards α-cyclodextrin and no detectable affinity for linear malto-oligosaccharides with a DP between 3 and 7 , which is clearly different from what we and others  have obtained for laforin. Supported by these findings and after addressing the best conditions and the best CBM ligands, we proceeded with a detailed analysis of laforin–glycan interaction for the two ligands with highest affinities, namely maltoheptaose (Kd=173±9 μM) and γ-cyclodextrin (Kd=114±10 μM).
Laforin has a low-affinity glycan-binding CBM
Glycan recognition very often involves aromatic amino acid side chains in the binding site of proteins, such as tryptophan and tyrosine. The recent report describing laforin structure interacting with maltohexaose confirmed the involvement of various tryptophan side chains (Trp32 and Trp99), along with Lys87, Asp107 and the backbone of Glu103 on the glycan-interaction site .
Near-UV CD and UV-difference are two spectroscopic techniques that are sensitive to micro-environmental changes around the side chains of aromatic amino acids and thus suitable to obtain some degree of structural insight of the interaction [46,47]. Based on this, both techniques were used to further characterize laforin–glycan interaction.
As shown in Figure 2, the incubation of laforin with saturating amounts (5 mM) of γ-cyclodextrin induced significant changes in the spectra of laforin, with equivalent results obtained for the interaction with maltoheptaose (result not shown). The near-UV CD spectra, although arising from aromatic amino acid side chains, is not readily amenable to detailed interpretation, as possible contributions from disulfide bonds or non-protein cofactors might absorb in the same spectral region ; however, the UV-difference spectra, which measures the change in the absorbance of UV light by laforin induced by the glycan binding, can indeed provide information regarding the nature of the amino acids involved in the interaction, as the patterns of change detected in these spectra are influenced by the type of aromatic side chain involved . Our results show that the UV difference signals of laforin with saturating amounts of either γ-cyclodextrin (Figure 2B) or maltoheptaose (result not shown) are composed entirely of features arising from tryptophan residues, with three peaks at 275, 285 and 293 nm and two distinctive troughs at 278 and 289 nm [46,48]. These results confirmed the participation of laforin tryptophan residues Trp32, Trp85 and Trp99 at the CBM interaction site, which has been recently reported for laforin  and is already known for other CBM20 members [11,13]. It also shows that the binding events measured are mainly due to the CBM and not to DSP binding, as the DSP-binding pocket is devoid of tryptophan residues and has two tyrosine residues (Tyr146, Tyr304), which would influence the UV-difference spectra. This was somehow expected, as W32G laforin mutation, which abolishes laforin glycan binding, renders the DSP domain insensitive to inhibition by either glycogen or malto-oligosaccharides . The presence of a maltohexaose in the DSP pocket of laforin crystal structure, is thus, most probably due to the high concentration of glycan used during the crystallization experiments .
Laforin–glycan interaction spectral changes
The laforin spectral changes induced by glycan interaction were further explored to determine CBM-binding affinities by titration with both glycans. The variations of CD and UV-difference spectra upon addition of γ-cyclodextrin or maltoheptaose were plotted (Figure 3) and Kd values were determined. The Kd values obtained for γ-cyclodextrin (106±11 μM by near-UV CD and 86±12 μM by UV-difference) and maltoheptaose (170±24 μM by near-UV CD and 258±26 μM by UV-difference) were in agreement with the results obtained by MST, pointing towards a low-affinity glycan binding and a marginal preference towards γ-cyclodextrin (DP=8) over maltoheptaose (DP=7).
Laforin–glycan interaction affinity determination
To check if laforin–glycan interaction induced any major structural rearrangement, we have checked laforin secondary structure by far-UV CD in the absence and in the presence of saturating amounts (5 mM) of both glycans, as this spectral region corresponds to peptide bond absorption, thereby giving the protein content of regular secondary structural features such as α-helix and β-sheet . The results revealed that no significant difference in laforin spectra was induced upon glycan interaction (result not shown).
Nevertheless, when far-UV CD was used to monitor laforin thermal melt in the absence and presence of maltoheptaose and γ-cyclodextrin, a significant increase in laforin melting temperature was measured in the presence of both glycans (Supplementary Figure S1A). Laforin's melting temperature increased from 52.5±0.1°C (van't Hoff enthalpy: 472.6±9.4 kJ/mol), when tested in the absence of glycans, to 55.6±0.1°C (van't Hoff enthalpy: 588.2±15.4 kJ/mol) when incubated with 5 mM maltoheptaose and to 56.6±0.1°C (van't Hoff enthalpy: 667.9±14.0 kJ/mol) when incubated with 5 mM γ-cyclodextrin, thereby indicating a tighter binding of the protein to γ-cyclodextrin over maltoheptaose.
Stability of laforin was also described in the two recent papers describing laforin  or its DSP crystallographic structures . In the first report , laforin Tm was evaluated by differential scanning fluorimetry (also known as thermofluor), with laforin Tm measured at 49.7°C, close to our results and a ΔTm for maltoheptaose incubation of about 1°C, in contrast with our ΔTm of 3.1°C. Although we have obtained a similar Tm for laforin, the ΔTm differences in the presence of glycan might be a result of the higher amounts of glycans (5 mM compared with 1 mM) we used in our assays, which definitely should induce higher Tm shifts.
In the second paper , protein stability was measured by CD, with authors reporting a Tm for laforin of 60°C, which is considerably higher than our and others  results (52.5°C and 49.7°C respectively). The authors used a spectropolarimeter equipped with a refrigerated recirculator for the melting experiments and don't describe the use of any temperature probe to accurately measure the temperature inside the cuvette. If the authors used as assay temperature, the temperature set at the recirculator, they might have experienced a temperature drop between the recirculator set temperature and the actual temperature at the cuvette, which might account for the increased Tm measured. This paper also reports a ΔTm of 2°C for 5 mM α-cyclodextrin incubation and an incredible ΔTm of 8°C for 10 mM maltose, in clear opposition to the results of the first paper, where 1 mM maltose incubation (DP of 2) induced no significant change in laforin Tm .
In order to clarify these discrepancies, we have evaluated the melting temperature of laforin in the absence and presence of glycans by DSC. Our results (Figure 4) are in accordance with the results of the first paper; we have not detected a significant increase on the Tm of laforin (53.44±0.05°C) when incubated with 10 mM maltose (53.22±0.05°C). We have measured a ΔTm of 4°C when laforin was incubated with 5 mM α-cyclodextrin and confirmed our previous CD results, with Tms of 56.94±0.07°C and 57.87±0.06°C for laforin incubated with maltoheptaose and γ-cyclodextrin respectively.
Thermal unfolding of laforin in the presence of different glycans
Laforin glycan binding is driven by favourable enthalpic and unfavourable entropic contributions
To provide a quantitative and qualitative characterization of the driving energies governing laforin–glycan interaction, we have performed ITC assays, collecting calorimetric data from the titration of laforin (433 μM) with maltoheptaose (4 mM) or γ-cyclodextrin (4 mM; Figure 5). Data interpretation yielded the binding parameters (Table 1) which evidenced a spontaneous reaction. This is indicated by the negative Gibbs free energy which is dominated by favourable changes in enthalpy (ΔH), mainly resulting from burying the type-B CBM apolar surface area , counterbalanced by an unfavourable entropic contribution (TΔS), resulting from the additive effect of the favourable entropic effect upon the release of ordered water molecules and the unfavourable entropic effect resulting from the loss of freedom of the binding partners .
Calorimetric data for ITC titration of laforin
|Glucan||Kd (μM)||ΔG (kJ/mol)||ΔH (kJ/mol)||TΔS (kJ/mol)||n|
|Glucan||Kd (μM)||ΔG (kJ/mol)||ΔH (kJ/mol)||TΔS (kJ/mol)||n|
Although low c-values were used during the ITC titrations (c=2 for maltoheptaose and c=3.6 for γ-cyclodextrin), which might introduce errors in the determination of stoichiometry and enthalpy , similar favourable enthalpic and unfavourable entropic contributions have been described by numerous groups characterizing glycan interactions by lectins and CBMs in similar circumstances [11,13,46,52].
The affinities measured by ITC for the interaction of laforin with maltoheptaose and γ-cyclodextrin should not be affected by the low c value  and were once again in good agreement with the affinities measured either by MST, CD or UV difference, clearly categorizing laforin CBM as a low-affinity CBM. This is in line with what has been previously described for other members of the CBM20 family  and interpreted as a means to facilitate regulatory roles of CBM20-containing enzymes [11,45], further reinforcing the role of laforin as a glycogen metabolism regulator.
Epitope mapping of glycans interacting with laforin revealed that all maltopentaose glucose units are in contact with laforin
To further reinforce the characterization of the molecular mechanism involved in laforin–glycan interaction, we then corroborated ligand binding of the strongest binders to laforin, γ-cyclodextrin and maltoheptaose, using two distinct ligand-based NMR experiments, STD (Supplementary Figures S2A and S2C [37,38] and CPMG [41,42] NMR techniques (Supplementary Figures S2B and S2D). The binding of both compounds to laforin was detected unambiguously in the two ligand-based NMR experiments. Due to the severe resonance overlap observed in the proton NMR spectra of this kind of molecule (Figure 6; Supplementary Information), quantitative group epitope mapping (GEM) and STD amplification factors (ASTD) cannot be accurately measured from the STD–NMR experiments to disclose the precise protein-binding profile for these glycans. Instead, the definition of a qualitative epitope was attempted based on the structural data obtained from the crystal structure of the maltohexaose–laforin complex . As the results from 1D STD–NMR experiments are rather limiting and erroneous when such an extensive resonance overlap exists, we used the 2D variant of this experiment, the STD–TOCSY.
We sought to answer the question: are the laforin–glycan interactions occurring in solution via the terminal parts of the glycans or are these driven from their core? Although resonance overlap limits the γ-cyclodextrin interrogation, representative signals to distinguish the terminal from the core rings of the linear glycans were possible to obtain with this approach (Figure 6; Supplementary Information). Based on these experiments, we can follow several spectral regions for the glycans and determine the binding epitope with a greater level of certainty. The two upfield triplets inform of both extremities of the glycans, region I on the non-reducing end (ring C) and region II on the reducing end (ring A; Figure 6, I and II) and the resonances close to 3.9 ppm inform the binding of the central rings B (Figure 6, III). We also have the information about the glycoside bond region of the central rings B and the terminal ring C from the β-anomer H1 protons appearing overlapped at region VI, whose intensity is dominated by the B rings (Figure 6, VI). Region IV corresponds to the H5 protons of the central rings B and the H6s of all rings in the peripheral region of the glycan chain (Figure 6, IV). Another overlapped region V includes many core (rings B) and peripheral (rings A and C) protons (Figure 6, V). With these spectral regions identified we could then interrogate how the interaction is occurring. Furthermore, the Kd values obtained from the MST data indicate that the order of magnitude changing-step in the binding affinity is observed from a DP of 5–6 (Figure 1B), so we decided to study the binding epitope for the maltotetraose, maltopentaose and maltoheptaose interactions with laforin.
1H—NMR (600 MHz) spectrum of 20 mM maltotetraose, with spectral assignment and regions to allow the GEM
For maltoheptaose, most of the resonances appear on the STD–TOCSY when compared with its TOCSY spectra (Figure 7A; Supplementary Figure S3A), highlighting that most of the protons of the linear glycan are indeed interacting with the protein. In fact, regions III, IV and V, corresponding to most protons of the B rings, as well as regions I (H3 of ring C) and VI (H1 of rings B and C) give strong signals. However the H2 proton from ring A in the reducing end (region II) gives no STD signal. When compared with maltopentaose one can see a small decrease in the intensities of the resonances at regions III and IV (Figure 7B), in agreement with a small affinity reduction, but this glycan interacts with the protein through protons of the B rings, as well as protons of the terminal rings C (region I) and A (region II). Thus, whereas maltopentaose shows a protein interaction spanning from the centre to both extremities of the molecule including all five glucosyl residues (Figures 7A and 7B), maltoheptaose shows no detectable STD interaction with the protein through the reducing terminal residue. The high affinity of these glycans to laforin results from a strong central glycan contribution (Figures 7A and 7B, dotted box with regions III, IV and V), whereas the interactions with the terminal rings C and A (regions I and II respectively) are weaker in maltopentaose and undetected for ring A in maltoheptaose. Contrastingly, the maltotetraose STD–TOCSY spectrum shows the weakened signals of the protons from the B rings at the core of the glycan and very weak interaction with the terminal C ring and undetectable interaction with ring A (Figure 6 C), probably due to its low interaction affinity.
TOSCY and STD–TOCSY experiments to obtain the epitope for the glycan's interactions with Laforin
Although the ligand-based NMR data clearly detects this type of glycan–protein interaction, the observed major increase in the relative intensity of the effect between the four- and five-membered glycans parallels their observed affinity increase, as discussed before.
These STD–NMR data obtained for the maltotetraose, maltopentaose and maltoheptaose in solution and the corresponding indications for their laforin-binding epitope, can be compared with the structural data obtained from the crystal structure of the maltohexaose–laforin complex . Because in that study the crystals were obtained at pH 5, which is far from the pH at which the relative affinity of the glycan to the protein is optimal (pH ~7.5), as determined by us, the formation of the glycan–laforin complex at that pH was forced by using a very high glycan concentration (250 mM). As a consequence, several glycan molecules were bound to the laforin dimer in the crystal, in particular at both the DSP and the CBM domains .
We studied in detail the interaction site of maltohexaose with the laforin CBM domain in the published crystal structure , as in the experimental conditions used is the major binding site, as discussed above. Supplementary Table S3 lists the glycan–laforin proton–proton distances shorter than 3.5 Å (1 Å=0.1 nm) for each of the six glucosyl units of maltohexaose, which could produce STD–NMR signals. Thus, many shorter distances involving exchangeable hydroxy protons were excluded. These interactions are also illustrated in Supplementary Figure S4. It can be seen that, whereas two central glucosyl units (2 and 3 from the reducing end) show a large number (14 and 7, respectively) of short distances to the protein, the number of such interactions sharply decreases towards the non-reducing (glucosyl unit 1) and reducing (glucosyl unit 5) ends of the maltohexaose and the reducing end (glucosyl unit 6) shows no such short distances. This is in agreement with the NMR observation that the maltopentaose glycan interacts with the laforin CBM site through all its glucosyl units and the stronger STD effects observed for the central glucosyl glycan units relative to the terminal ones. The absence of a detected interaction of the glucosyl unit in the reducing end of maltoheptaose with laforin can be rationalized by the predicted absence of amino acid side chains in the laforin CBM in the crystal structure within 3.5 Å of that glucosyl unit.
The recent paper describing the long sought deter-mination of laforin 3D structure shed some light on the mechanism of carbohydrate interaction, as the structure was obtained with a bound glycan . Nevertheless, the structural model obtained only represents a still picture of such interaction, in the conditions amenable to crystallize the protein, and a complementary biophysical characterization of laforin–glycan interaction was still missing.
We have used some of the standard biophysical techniques used to characterize glycan binding to thoroughly characterize laforin–glycan interaction. We have shown that laforin has an optimum pH for the interaction with glycans at physiological pH and that it is a low-affinity glycan binder, as described for other CBM20 family members. We have also put the numbers on laforin glycan affinities, complementing what was already known on the increased affinity of laforin towards glycans with increasing DP. Our data confirmed the involvement of tryptophan residues in the glycan interaction surface of laforin CBM and confirmed that, according to what was expected, laforin interaction with glycans is driven by favourable enthalpic contribution with an unfavourable entropic contribution.
STD–NMR proved to be the right tool to evaluate the interaction from the side of the glycan ligands with 4–7 glucose units. Although the interaction with the protein as probed by STD–NMR was much stronger for glycans with DP higher than four, all the sugar units interacted with the laforin CBM binding site up to DP of six, while the central region of the glycan had increasing importance in the stability of the interaction as the chain length increased. For longer glycan chains, it is predicted that the reducing end will no longer be in close contact with the protein CBM surface. In conclusion, the results described in the present work complement the recent structural characterization of laforin with a rigorous and thorough biophysical characterization of laforin carbohydrate interaction in aqueous solution, which shed light on the molecular mechanisms of physiological laforin glycogen interaction, hopefully providing novel tools for the understanding of the pathological mechanisms involved in Lafora disease, which is the first step towards the development of future therapies for such devastating disease.
David Dias, Joana Furtado, Emeric Wasielewski, Rui Cruz, Bernard Costello, Lindsay Cole, Tiago Faria and Pedro Castanheira performed the experiments. Philipp Baaske, Rui Brito, Alessio Ciulli, Isaura Simões, Sandra Macedo-Ribeiro, Carlos Faro, Carlos Geraldes and Pedro Castanheira designed the experiments. Isaura Simões, Carlos Geraldes and Pedro Castanheira wrote the paper.
We thank Massimo Lucci for kind assistance on running the Bruker 700 NMR spectrometer during the visits to CERM.
This work was funded by Fundo Europeu de Desenvolvimento Regional (FEDER) funds through the Operational Competitiveness Program (COMPETE) by Portuguese National funds through the Fundação para a Ciência e Tecnologia (FCT) [PTDC/BIA-PRO/111141/2009] [FCOMP-01-0124-FEDER-014323], under which JF and RC were the recipients of FCT BI fellowships. DMD was the recipient of a FCT PhD grant [SFRH/BD/81735/2011]. Some NMR experiments were performed using the Varian VNMRS 600 MHz spectrometer from the Portuguese National NMR Network which was purchased and maintained within the framework of the National Program for Scientific Re-equipment [REDE/1517/RMN/2005], with funds from POCI 2010 (FEDER) and FCT. Financial support in the form of access to the Bio-NMR Research infrastructure at the CERM NMR laboratory at the University of Florence, co-funded under the 7th Framework Programme of the EC (FP7/2007-2013) grant agreement 261863 for conducting the research was also obtained. Authors would also like to acknowledge grant [PEst-C/SAU/LA0001/2013-2014].
These authors contributed equally to this work.