o-glycosylation and stability : unfolding of glucoamylase induced by heat and guanidine...

Eur. J. Biochem. 207, 661 -670 (1992) $3 FEBS 1992

0-Glycosylation and stability Unfolding of glucoamylase induced by heat and guanidine hydrochloride

Gary WILLIAMSON', Nigel J. BELSHAW ', Timothy R. NOEL', Stephen G. RING' and Michael P. WILLIAMSON'

* Department of Molecular Biology and Biotechnology, University of Sheffield, England

(Received January 28/April8, 1992) - EJB 92 01 11

AFRC Institute of Food Research, Norwich Laboratory, Norwich Research Park, England

We have examined the stabilities of the catalytic and binding domains of glucoamylase 1 from Aspergillus niger and how these stabilities are affected by the 0-glycosylated linker glycopeptide which separates the domains. On heating, the catalytic domain unfolds irreversibly, whereas the binding domain unfolds reversibly as shown by differential scanning calorimetry and by 'H NMR. The stability of three functional peptides, derived from glucoamylase 1, containing the binding domain alone and with 10 or 38 residues of the linker glycopeptide [Williamson, C., Belshaw, N. J. and Williamson, M. (1992) Biochem. J . 282,423 -4281 was examined. Refolding in each case was reversible after thermal or chemical denaturation. P-Cyclodextrin stabilised the binding domain by the same amount when it was part of glucoamylase 1 or an isolated domain. The thermal stability of the catalytic domain was not affected by the binding domain; however, the catalytic domain increased the melting temperature of the binding domain. Furthermore, the linker glycopeptide stabilised the binding domain against reversible thermal and chemical denaturation by about 10 kJ/mol, but only a portion of the 0-glycosylated residues were required for stabilisation. On a simple molecular mass basis, the linker glycopeptide does not contribute as much as expected to the denaturational enthalpy of glucoamylase 1 and, in addition, shows only a small conformational change on chemical or thermal denaturation; this supports an extended structure for the linker. The results demonstrate that the unfolding pathway of glucoamylase 1 depends on the concentration of P-cyclodextrin and that the presence of the catalytic domain and/or the linker glycopeptide stabilises the binding domain.

Glucoamylase from Aspergillus niger consists of two functional domains, a catalytic domain of M , z 52000 (residues 1 - 470) and a binding domain of M , z 1 1 000 (residues 509 - 616), as deduced from proteolytic studies [I], production of genetically truncated forms [2] and amino acid sequence [3]. The two domains are separated by a region of about 40 amino acids, of which 28 are serine or threonine residues O-glycosylated with mono-, di- or trisaccharides [4]. This linker region acts as a semi-flexible rod and the tumbling time of the binding domain is substantially faster than the catalytic domain as demonstrated by 'H NMR 151. In addition, the linker forms an extended structure owing to steric effects of the carbohydrate attached to the peptide [6]. This is analogous to that seen in

a diverse range of proteins containing heavily 0-glycosylated linker regions, such as mammalian mucins [7], mammalian low-density-lipoprotein receptor [8] and fungal cellulases [9].

Partially deglycosylated glucoamylase loses catalytic activity more rapidly than the native form both on storage at 30 "C [lo] and at higher temperatures [ll]. In order to elucidate the molecular mechanisms involved in determining the stability of glucoamylase 1, and in particular the relevance of the 0-glycosylated linker in facilitating communication and modulating the properties of the two domains, we have examined the effect of the 0-glycosylated linker peptide on the conformational stability of glucoamylase and its isolated functional domains (Scheme 1).

Correspondence to G. Williamson, AFRC Institute of Food Re- search, Norwich Laboratory, Norwich Research Park, Colney, Norwich NR4 7UA, England

Fux: +44603 507723. Abbreviations and symbols. Glucoamylase-1-(471- 616), residues

471 - 61 6 from glucoamylase 1 ; glucoamylase-I -(499 - 616), residues 499- 61 6 from glucoamylase 1 ; glucoamylase-1-(509- 616), residues 509 - 61 6 from glucoamylase 1 ; DSC, differential scanning calorimetry; Tm, melting temperature; GdnCI, guanidine hydrochloride; AH,, molar enthalpy change on unfolding; Qd, enthalpy change; AC,, change in heat capacity on unfolding; C,,,, maximum deflection in heat capacity from the baseline; AH,, van't Hoff enthalpy change; Tk, melting temperature in presence of ligand L.

Enzymes. Glucoamylase, 1,4-a-~-glucan glucohydrolase (EC 3.2.1.3); a-amylase, 1,4-a-~-glucan-4-glucanohydrolase (EC 3.2.1 .I).

MATERIALS AND METHODS

Protein purification and assays

Glucoamylase 1 and 2 were purified by published methods [12]. Clucoamylase-l-(471- 616), glucoamylase-l-(499-616) and glucoamylase-1-(509 - 616) were prepared and purified from glucoamylase 1 as previously reported [5]. Concentration of protein was measured using the following absorption coef- ficients : glucoamylase 1, 137 mM - cm- ' [ 131; glucoamylase 2, 109 mM-' cm-' [14]; glucoamylase-2-(471-616), glucoamylase-1-(499-616) and glucoamylase-I-(509 -616), 30.7 mM-' cm-' [5].

662

I I Glucoamylase 1

I I Glucoamylase 2

G1C I 4

GI eo9

Scheme 1. Structures of glucoamylase fragments. GI C, glucoamylase- 1 -(471 - 616); G 1 C499, glucoamylase-l-(499-616); GlCSo9, glucoamylase-l-(509 - 616); CAT, catalytic domain; BIND, binding domain.

Differential scanning calorimetry

DSC experiments were carried out using a Setaram microDSC (Lyon, France), a heat-flow Calvet calorimeter operating under controlled temperature (resolution = 0.01 K). Calibration constants for temperature and heat flow were as supplied by Setaram. Protein (1 - 10 mg/ml) was dia- lysed for 16 h against 10 mM potassium phosphate buffer at the desired pH before DSC. Samples (about 500 mg) of glucoamylase solution were hermetically sealed in re-usable DSC sample cells. An equal weight of buffer was sealed in a matching cell and used as a reference. Samples were analysed using a programmed heating rate of 1 "C min-' over a temperature range 20- 100°C. After heating to 1OO"C, samples were cooled at 1°C min-' and re-scanned to observe any reversible transitions. A scan of the empty DSC cells was subtracted from each data set before integration of the endotherm, using a linear baseline, to obtain the enthalpy change of& [15J From this value, the molar enthalpy change on unfolding (AH, ) was calculated. The van't Hoff enthalpy changes for denaturation were calculated using the equation:

A H , = 4 R l-2 ( cmax/Qd>

where T, is the temperature of maximum deflection in heat capacity from the fitted baseline, C,,, is the maximum deflection in heat capacity, (2d is the specific enthalpy change obtained from the area of the DSC curve [I51 and R the gas constant. The change in heat capacity on unfolding, AC,, was estimated by linear extrapolation of the pre- and post- denaturation baselines.

Denaturation by guanidine hydrochloride

The concentration of guanidine hydrochloride (GdnC1) was measured refractometrically [ 161. All protein denaturation experiments were carried out in 40 mM potassium phosphate pH 7.8. For equilibrium studies, the protein and denaturant (total volume 0.1 ml) were incubated for 48 h at 30"C, and cooled to 20 OC for 2 h before spectrophotometry. Initial experiments demonstrated that 2 h was sufficiently long to attain equilibrium at 20°C. Absorption spectra were recorded using a Beckman DU70 spectrophotometer equipped with cells able to measure 50 pl of solution.

0 10 20 Elution volume (ml)

Fig. 1 . Size exclusion chromatography on a Tosoh G3000SW column (0.75 x 30 cm) in 6 M guanidine hydrochloride. Glucoamylase 1 after DSC at (a) pH 7.8 and (b) 7.0. The arrow indicates the elution position of unheated glucoamylase 1.

'H-NMR spectroscopy

Data were collected on a Bruker AMX-500 spectrometer into 8 K data points, over a spectral width of 6000 Hz, using presaturation for 1.2 s to suppress te residual solvent signal. The data were processed using a mild Lorentz-to-Gaussian window function to increase resolution. Solutions of glucoamylase 1 (0.7 mM) and glucoamylase-I-(499 - 616) (1.3 mM) in D 2 0 at pH 8.0 were run at the indicated temperature.

RESULTS

Thermal denaturation of glucoamylase 1 : preliminary experiments

The effect of pH on the thermal unfolding of glucoamylase 1 was examined in order to determine the best pH for denaturation experiments. In 10 mM potassium phosphate pH 7.0, an exothermic peak, probably caused by aggregation, was observed between 335 ~ 340 K, at the low-temperature side of the unfolding endotherm. This effect was more pro- nounced in 10 mM sodium acetate pH 4.5, closer to the isoelectric point of the enzyme (PI = 4.0 [17]). At pH 7.8, 10 mM potassium phosphate, no exothermic peak was observed, but at all three pH values, no catalytic activity remained after heating to 100OC. The extent of aggregation of glucoamylase 1 after heating to 100°C was examined by gel filtration in 6 M guanidine hydrochloride/O. 1 mM dithiothreitol (Fig. 1). At pH 7.0, almost all of the protein was aggregated. At pH 7.8, about 50% remained in a monomeric form. Since guanidine hydrochloride disrupts hydrophobic interactions and dithiothreitol breaks disulphide bonds, any aggregation observed arises from covalent aggregation. Munch and Tritsch [I81 found that glucoamylase 1, after heat treatment at 7 0 T , showed considerable disulphide bond exchange leading to aggregation at pH 5.5 and also peptide bond hydrolysis at lower pH. Under our conditions we found no evidence for peptide bond hydrolysis. Since disulphide bonds were reduced in 0.1 mM dithiothreitol, aggregation must be caused by reactions giving non-disulphide-bonded covalent linkages. The reactions that give rise to this aggregation are unknown, but could be a Maillard-type reaction between the covalently bound sugars and protein amides,

663

Table 1. Dependence of thermal denaturation of glucoamylase 1 on its concentration at pH 7.8.

Concentration T,,,

mgiml K kJ/mol J g-' K - '

A Hd AH" 4

~~

12.0 341.0 1730 305 0.33 10.0 339.6 1850 308 0.39 8.33 338.7 2150 299 0.44 4.0 337.1 1680 321 0.45

Mean (k SD) 1850 210 308 9 0.40 0.05

although this is speculation. Refolding of free or immobilised glucoamylase 2 after denaturation in GdnCl shows hysteresis, and this effect is strongly dependent on the pH [19]. The hysteresis also depends on the method of measurement (activity, fluorescence emission or elipticity), demonstrating that the renaturation of glucoamylase 2 is a complex reaction probably involving metastable intermediates. At pH values less than 6, the overall mechanism is determined by the kinetic competition between folding and aggregation but, at higher pH, the yield increased with pH to about 80% [19]. This is in agreement with our thermal denaturation results and, for these reasons, the thermal behaviour of glucoamylase 1 was studied at pH 7.8 where no aggregation exotherm was observed.

Calorimetric studies

Table 1 shows the data obtained from the calorimetric experiments as a function of glucoamylase 1 concentration at pH 7.8. T,,, exhibited a small decrease ( ~ 4 K) with decreasing protein concentration (from 12 to 4mg/ml). AH, and AH, showed no significant trend with protein concentration. AHd for glucoamylase 1 of 1850 kJ mol-' is unusually large for a globular protein, but is comparable to that determined for a-amylase from Aspergillus oryzae of 2246 kJ mol-' under equivalent conditions [20]. The ratio AHd/AHv for glucoamylase 1 was 6.0 and is the same as that reported for a- amylase from A . oryzae, although only the latter contains bound Ca2+ [20]. For most proteins which obey a two-state model for unfolding, AHd/AH, is about 1 [15]. When this ratio is significantly greater than unity, one explanation is that the protein consists of more than one independent cooperative domain. However since d f i d / d H , is essentially an analysis of the shape of the DSC curve, then secondary effects, such as aggregation, which decrease or prevent reversibility of unfolding, will also affect this ratio. Since both a-amylase and glucoamylase aggregate on thermal denaturation, the process is irreversible and the DSC experiment cannot give a thermodynamically meaningful AH,. As with a-amylase from A . oryzae, the high values of AHd/AH, may be partly due to aggregation. However, subsequent experiments demonstrated that there is also a contribution to this ratio due to unfolding of separate domains in glucoamylase 1.

The reversibility of the thermally induced unfolding of glucoamylase 1 at pH 7.8 is shown in Fig. 2a and b. The first scan exhibited a shoulder at the low-temperature side of the endothermic peak. After one cycle of heating and cooling, a further DSC scan showed a greatly reduced peak with T,,, = 333.1 K. One possible explanation for this is that one domain

of glucoamylase 1 unfolds reversibly. Thermal denaturation was also examined in the presence of P-cyclodextrin (Fig. 2c

n

I I I

330 340 350 Temperature ( K

Fig. 2. Calorimetric scans of 0.122 mM glucoamylase 1 in 0.01 M potassium phosphate pH 7.8. (a) Glucoamylase 1 alone, first scan; (b) rescan of (a) after cooling at 1 Kimin; (c) glucoamylase 1 and 5 mM p- cyclodextrin, first scan; (d) rescan of (c), after cooling a t 1 K/min.

and d). In the first DSC curve, the shoulder observed in the absence of P-cyclodextrin shifted from the low-temperature side of the peak to the high-temperature side. The rescan showed a peak with T , = 347.0 K. Intact glucoamylase 1 has been observed to bind /I-cyclodextrin with no loss of catalytic activity [21]. After site-specific proteolysis, the isolated binding domain is able to bind P-cyclodextrin, whereas glucoamylase 2 (the catalytic domain) binds this ligand weakly or not all 1221. The shift in T, of the peak observed on re- scanning suggests that this peak is a result of the domain that binds to P-cyclodextrin re-folding on cooling before re- scanning, and that the shoulder in the first scan arises from unfolding of the same domain.

'H-NMR spectroscopy of heat-treated proteins

Confirmation that only the binding domain refolded on cooling was achieved by obtaining 'H-NMR spectra of glucoamylase 1 before and after heating, and comparing these to the isolated binding domain (glucoamylase-1-(499 - 61 6) (Fig. 3).

Glucoamylase 1 exhibited several resonances between 0.6 and -0.3 ppm with relatively small linewidths (typically 8- 10 Hz), which are sharper than expected for a protein of M , ~ 8 0 0 0 0 . The resonances between -0.3 and -0.8 ppm are broader and more characteristic of a protein the size of glucoamylase 1.

Comparison with the 'H-NMR spectrum of the binding domain showed clearly that most of the sharper resonances in glucoamylase 1 are due to the binding domain and so the broader resonances must be due to the larger catalytic domain. The tumbling times of the two domains are therefore different, in agreement with data from two-dimensional NMR experiments [5] . On heating glucoamylase 1 to T > T,, the resonances upfield from 0.6 ppm disappear, showing that the protein has completely unfolded. On cooling, only the resonances due to the binding domain re-appeared, proving that only the binding domain refolds after temperature-induced unfolding. The chemical shifts of the refolded binding domain in

664

Table 2. Thermal denaturation of glucoamylase 2 (GZ), glucoamylase 1-1-(471- 616) (GIC), glucoamylase-1-(499 -616) (GIC) and glucoamylase- I-(509-616) (GlC509). Protein concentrations were 0.122 niM at pH 7.8; aa = amino acid.

Protein T,,, A H d AHd Carbo- A Hd A Hv ~ AH, ACp ACP 4G<293)

hydrate A H , content [5]

K J/g kJ/mol YO (by mass) kJ/Mol aa kJ/mol J g- ’ K-’ kJ mol-’ K - ’ kJ/mol

G2 338.9 23.9 1650 18 3.21 340 4.8 0.44 30 (125 30) G1C 330.6 14.5 360 36 2.48 460 0.79 0.45 11 1 6 k 4 GlC499 330.5 18.8 310 22 2.65 500 0.62 0.59 10 13 k 10 GlC5’” 327.9 24.8 310 5 2.90 430 0.72 0.57 7 1 9 i 7

I I I I I I I

0.4 0.2 0.0 -02 -04 -0.6 -0.8 Chemical shift (PPM)

Fig. 3. Upfield region of ‘H-NMR spectra at pH 8.0. (a-c) 0.7 mM glucoamylase 1, at (a) 320 K, (b) after heating to 340 K and (c) after cooling to 320 K ; (d) 1.3 mM glucoamylase-l-(499-616). Spectrum (d) is virtually unchanged after heating to 340 K and cooling (data not shown).

glucoamylase 1 were identical to those of the proteolytically cleaved binding domain, showing that this refolding is indeed reversible. Addition of P-cyclodextrin (1 mol/mol) had little effect on the upfield signals, demonstrating that the conformation of the protein is not greatly affected by the presence of this ligand. Repetition of the thermal denaturation experiments in the presence of fi-cyclodextrin confirmed an increase in T, for reversible denaturation of the P-cyclodextrin-binding domain.

DSC of isolated functional domains

Table 2 shows the data obtained from DSC experiments on the catalytic domain with the full length of the 0-glycosylated linker at the C-terminus (glucoamylase 2) and for glucoamylase-I -(471- 616), glucoamylase-1-(499 ~ 616) and glucoamylase-l-(509 ~ 616) at equivalent molar protein concentrations.

The T, for glucoamylase 2 was similar to glucoamylase 1 but dH,/dH, was 4.8 (cf. 6.0 for glucoamylase 1). Since, as

expected from the results above, the shoulder due to the binding domain was not observed, we suggest that the difference of 1.2 in A H d / d H v is due solely to the absence of the binding domain. Additionally, because the T, of glucoamylase 1 is not significantly different to glucoamylase 2, it is clear that the presence of the binding domain does not affect T, of the catalytic domain.

Contribution of the 0-glycosylated linker to thermal unfolding

The values of T,,, for glucoamylase-l-(471-616) and glucoamylase-1-(499 ~ 61 6) were not significantly different, but T, for glucoamylase-1-(509 - 616), however, was about 2.5 K lower. This shows that the 0-glycosylated decapeptide at the N-terminus of the binding domain increases T,, but the presence of a further 27-residue peptide, also heavily 0- glycosylated, makes no further difference to T,,, of the binding domain.

AHd increased from g~ucoamylase-l-(471- 61 6) through to glucoamylase-1-(509 - 616). This increase (measured in J/g) was proportional to the decrease in carbohydrate content. This indicates that the contribution of the covalently bound carbohydrate to the total enthalpy is low compared to that of the protein component. If A H , is recalculated with respect to protein content, the enthalpy of denaturation becomes 22.7, 24.1 and 26.1 J/g protein for glucoamylase-I-(471- 616), glucoamylase-1-(499 ~ 616) and glucoamylase-1-(509 - 616) respectively. Since, from above, the T,,, values of glucoamylase-I-(471 - 616) and glucoamylase-I-(499 - 616) are the same, implying equal stabilities of the binding domain portion, then the contribution of the whole 0-glycosylated region to AHd can be approximated by extrapolation. Comparison between AHd for g~ucoamylase-l-(471- 616) (360 kJ/mol) and for glucoaniylase-I-(499-616) (310 kJ/mol) reveals a difference of only 50 kJ/mol even though the M , are 24900 and 16400 respectively. Thus by extrapolation to the binding domain alone ( M , = l l 600), 54% of the molecular mass (i.e. the 0-glycosylated region) contributes only 16% to AH,, of glucoamylase-l-(471- 616). Consistent with this extrapolation is the fact that there does not appear to be a significant difference between the denaturational enthalpies for glucoamylase-I -(499 - 61 6) and glucoamylase-I-(509 - 61 6) even though the stabilities are different. This implies that the 0- glycosylated residues 500 - 509 contribute very little to the AHd ofthe protein, although they clearly influence its stability.

If the 0-glycosylated linker does not contribute towards AH,, then it might be expected that the sum of the denaturational enthalpies for the binding domain plus the enthalpy for the catalytic domain would not be very different to AH, for glucoamylase 1. From Table 2, the sum of the

665

I I t I

31 320 330 340 350

Temp ( K Fig. 4. Differences between calorimetric scans at pH 7.8. (a) The scans of 0.122 mM glucoamylase 2 and 0.122 mM glucoamylase-l-(509- 616) were added together (-) and compared to 0.122 mM glucoarnylase 1 (----). The dotted line shows the baseline extrapolation for estimation of A C, for glucoamylase 1. (c) The difference between the scans in (b).

denaturational enthalpies for glucoamylase-I-(471- 616) and for glucoamylase 2 is 2010 kJ/mol, which is only 8% higher than the denaturational enthalpy for glucoamylase 1 (1850 kJ/ mol, Table 1). Since the summation accounts for a binding domain, a catalytic domain and effectively two linker regions, the contribution of molecular mass of the two linkers to this total is 31%. As a rough estimate from this calculation, the linker contributes to AHd only about a quarter of that predicted from a simple molecular mass. This compares to a contribution of 32% calculated from the experiments on isolated domains above. In addition, the sum of the DSC curves for glucoamylase-1-(471- 616) and for glucoamylase 2 are qualitatively similar (results not shown). However, summation of the DSC responses for glucoamylase 2 plus glucoamylase-l- (509-616) is qualitatively different to that of glucoamylase 1 (Fig. 4).

At first sight, it might be expected that this summation should represent the intact glucoamylase 1 more accurately than the glucoamylase 2 + glucoamylase-l-(471-616) summation above. Examination of the data in Table 2 and the GdnCl data presented below, however, shows that the difference arises indirectly because the 0-glycosylated peptide stabilises the binding domain and hence the absence of the linker shifts the DSC peak of glucoamylase-1-(509-616) to a lower temperature.

pH dependence of thermal denaturation of glucoam y lase- 1 -(47 1 - 6 16)

The change in heat capacity on protein unfolding, AC,, determines the temperature dependence of the enthalpy of denaturation as described by Kirchoff s relation :

AAHd -- - AC,

AT

and a plot of AHd against T, at a range of pH values gives a straight line with slope = AC, [15]. This equation assumes that AC, and AHd alone do not vary too greatly with pH over the range of pH employed.

Tm (K)

Fig. 5. Determination of AC, for glucoamylase-l-(471-616). AHd and T, were determined (A) at pH 7.80, pH 7.48, pH 6.95 and pH 6.12. The value of AC, as estimated from each calorimetric scan at these pH values by extrapolation of the baselines for folded and unfolded protein was 0.45 0.05 (SD) kJ mol-' K - I . They intercept (see text) was varied to obtain the best fit of the slope (= AC,) (-) to the data points, and the error i n this slope is plotted (----).

Fig. 5 shows a plot of AHd against T, between pH ~ 6 - 8 for glucoamylase-I-(471- 616). The same DSC curves were also used to determine AC, by extrapolation of the baselines in the pre- and post-transition regions. The mean of these determinations (plotted as the slope with fitted y intercept) agrees well with the individual data points. This confirms that AC, as calculated from extrapolation of baselines is correct (11.0 kJ mol-' K-').

Effect of P-cyclodextrin on denaturation

The effect of P-cyclodextrin on the thermal behaviour of glucoamylase-I-(471- 616) was measured and compared to that of p-cyclodextrin on glucoamylase 1. T,, AHd and AC, of glucoamylase-I-(471- 616) all increased with increasing p- cyclodextrin concentration (Table 3). On addition of p- cyclodextrin, the ratio AHd/AH, increased; at 2 mM, the ratio was close to 1 which approximates to a two-step denaturation.

666

Table 3. Effect of 8-cyclodextrin on thermal denaturation. Protein concentration was 0.122 mM; aa = amino acid.

Protein fl-Cyclodextrin T, A H d A Hd A Hv ~ A Hd ACP ACP A Hv

kJ/mol aa kJ/mol J g - ' K- ' kJ K - ' mol-' mM K kJ/mol

0.50 336.5 470 3.24 410 0.99 0.52 13 2.00 340.9 560 3.86 580 0.96 0.56 14 5.18 343.5 650 4.48 570 1.14 0.62 16

Gluco- 0 339.6 1850 3.00 300 6.5 0.40 32 amylase 1 2.00 340.8 1880 3.05 343 5.5 0.47 38

5.00 343.0 1840 3.00 310 5.9 0.51 41 0" 333.1 88 0.82 2.00" 343.1 150 1.40 -

5.00" 346.5 300 2.80 -

G l C 0 330.6 360 2.48 460 0.79 0.45 11

~ - - -

- - -

- - -

a Data were obtained from the rescan of the sample after one cycle of heating and cooling at 1 K/min

Table 3 also shows the thermal unfolding parameters for glucoamylase 1 in the presence of P-cyclodextrin. Since the catalytic domain has about 6 times the molecular mass of the binding domain, the changes in the latter are not large enough to be detected when overlapped by the DSC curve of the intact molecule. However, the rescans are more informative. For the rescan of glucoamylase 1, the increase in T, with P- cyclodextrin concentrations of 2.0 mM and 5.18 mM were 10.0 K and 13.0 K, respectively. For the isolated glucoamylase-1-(471- 616) the corresponding increases in T, were 10.3 K and 12.9 K respectively.

The change in T, on addition of ligand L with concentration [L], is related to the dissociation constant & by:

T, T: R AT, = ~ In f [Ll/Kdl

where T, is the melting temperature in the absence of ligand, Tk is the melting temperature in the presence of L and AH, is the denaturational enthalpy in the absence of ligand [23]. Since the dependence of AT, on [L] is the same, within experimental error, for glucoamylase-l-(471-616) alone and for the rescan of glucoamylase 1, then it is clear that the presence or absence of the unfolded catalytic domain does not affect Kd for the binding domain.

The effect of the catalytic domain on the unfolding of the binding domain can be estimated by comparing the DSC parameters of the rescan of the DSC curve for glucoamylase 1 to the DSC curve for glucoamylase-I-(471- 616) both in the presence and absence of P-cyclodextrin (Table 3). T, for the binding domain attached to the catalytic domain, less T, for the binding domain alone, was 2.5 K, 2.2 K and 3.0 K for 0, 2 mM and z.5 mM P-cyclodextrin, respectively. Thus, although the binding domain has little effect on the unfolding of the catalytic domain, it is in fact stabilised by the catalytic domain by 2 - 3 K. In the absence of P-cyclodextrin the catalytic domain is predominantly in the folded conformation during most of the unfolding transition of the binding domain. Conversely, with 5 mM P-cyclodextrin, the catalytic domain is mainly in the unfolded state at the T, of the binding domain. In both of these cases, whether the catalytic domain is predominantly folded or not, the effect on binding domain T, is similar. We therefore conclude that the binding domain is stabilised by 2-3 K by the presence of the catalytic domain, but that this stabilisation is not dependent on the foldedness of the catalytic domain.

250 260 270 280 290 300 310

Wavelength (nm)

Fig. 6. Absorption spectra of glucoamylase-l-(509 - 616) in 4.92 M GdnCI/4O mM potassium phosphate pH 7.8, after (a) 65 s, (b) 97 s, (c) 132 s, (d) 270 s, (e) 515 s, (f) 810 s.

Kinetics of denaturation in GdnCl

On incubation of glucoamylase-I-(509-616) in 4.92 M GdnC1, the ultraviolet absorption spectrum showed an increase in absorbance with a peak at 262 nm, and a decrease with a trough at 291 nm (Fig. 6). There is an isosbestic point at 275 nm, which shows that the absorbance change arises predominantly from the existence of two states, presumably folded and unfolded as supported by the thermal denaturation studies. The rate of decrease of absorbance follows first-order kinetics, and the rate constant ( k ) for glucoamylase-1-(471- 616), glucoamylase-1-(499 - 616) and glucoamylase-I-(509 - 616) is given in Table 4. It can be seen that k for glucoamylase- I-(509 - 616) is higher than that for glucoamylase-1-(499 - 616) or glucoamylase-l-(471-616).

Estimation of using GdnCl

The absorbance change as a function of GdnCl concentration was used to calculate AG,,,,, by extrapolation of the data to a concentration of zero guanidine hydrochloride (Fig. 7). The data were fitted to an equation that takes the

667

00 1.0 2.0 30 4.0 5.0 6.0 [Gdn.HCq (M)

Fig. 7. Effect of GdnCl on the A289/AZ75 ratio (0) and AG (m) of 0.14 mg/ml glucoamylase-1-(509-616).

Table 4. Unfolding of glucoamylase-1-(471- 616) (GlC), glucoamylase-1-(499 - 616) (G1C499) and glucoamylase-l-(509- 616) (GlCSo9) in GdnCI. k is the first-order rate constant for unfolding in 4.92 M GdnC1/40 mM potassium phosphate pH 7.8 at 293 K. Other values were determined by linear extrapolation to zero GdnCI.

mg-' kJ/mol kJ mol-' M - ' MI"

G1C 1.9 32 f 2 - 9.6 3.36 G1C499 1.9 3 5 f 3 - 10.0 3.46 G1CSo9 4.3 23 & 2 - 8.2 2.77

-501 I I I I I I I 280 290 300 310 320 330 340 350

Temp ( K )

Fig. 8. Stability of glucoamylase-1-(471- 616) as derived from DSC (----) and from denaturation in GdnCl at 293 K (A). The curve was generated using the Gibbs-Helmoltz equation (see text) using the values in Table 2.

slope and intercept of the pre- and post-transition baselines (Y,, M,, Y,,, Mu respectively) into account:

The absorbance changes in this region arise because of the transfer of tryptophan and tyrosine residues into a different environment. Since all of the Trp and Tyr residues are located between 527-616 (using the numbering of intact glucoamylase 1) in all glucoamylase-1-(471- 616) species, the GdnC1-induced reaction only measures the unfolding of the binding domain (i.e. residues 509-616). Changes in the conformation of the U-glycosylated region would not be detected except indirectly as a result of the influence of the U- glycosylated region on the stability of the binding domain. The conclusion from this experiment, therefore, is that the

where [D] is the concentration of denaturant, Yobs is measured absorbance, and m is the slope of the plot of AG against [D] [24]. The solid line in Fig. 7 is simulated from the above equation using Yf = 0.7717, Mf = -0.00020, Y,, = 0.6887, Mu = -0.00369, m = -8.23 kJ mol-' M- ' dG(H20) = 22.8 kJ/mol. The midpoint of denaturant concentration from the titration ([GdnC1]l,2) is 2.77 M. Table 4 shows that there is no significant difference between A G ( H Z o ) for glucoamylase- 1-(471- 616) and glucoamylase-1-(499 - 616). However, glucoamylase-l-(509 ~ 61 6) is less stable thermodynamically, as shown by a lower dG(H20) and a lower [GdnC1]l,2.

N-terminal glycodecapeptide (499 - 508 by glucoamylase 1 numbering), present in glucoamylase-I-(499 ~ 61 6) but not glucoamylase-1 -(SO9 - 616), stabilises the binding domain by about 10 kJ/mol. The presence of a further 26-amino-acid glycopeptide has no additional effect on the stability of the binding domain.

Comparison of AC from thermal and GdnCl denaturation

The stability (AG) of the binding domain to denaturation as a function of temperature (Fig. 8) was predicted using data

668

- [ 8-cyclodextrin]

Scheme 2. Unfolding pathway of glucoamylase 1. P-Cyclodextrin increases the stability of the binding domain (BIND) relative to that of the catalytic domain (CAT).

from the DSC experiments and a modified form of the Gibbs- Helmholtz equation:

AGO) AH, (1 - T/T,) - AC, (T, - T + Tln TIT,)

where is the enthalpy of denaturation, T, is the melting temperature and AC, is the heat capacity change between folded and unfolded forms of the protein [25]. The relationship makes the assumption that AH, and AC, are independent of temperature over the temperature range considered. Although both parameters must, in reality, vary with temperature, the accuracy of the experimental data is such that the relationship holds to within normal measurement errors. The validity of this assumption has been discussed [15, 251.

The thermal unfolding data give values of AG(393) for glucoamylase-1-(471- 616), glucoamylase-1-(499 - 616) and glucoamylase-l-(509 - 61 6) that are not significantly different (Table 2). This is because A C, has a comparatively large effect on AC at T 6 T,, but is the most difficult parameter to measure accurately when the amount of sample is limited. However, the pH dependence of glucoamylase-1-(471- 616) thermal denaturation was used to provide an internal confirmation of AC, for glucoamylase-1-(471-616), although in- sufficient sample was available to confirm AC, in the same way for glucoamylase-I -(499 - 616) and glucoamylase-l- (509 - 616). A C ( 2 9 3 , H 2 O ) was estimated more accurately by using GdnCl denaturation as above.

In the presence of ligand, the free energy change on unfolding depends on the dissociation constant, Kd, and on the ljgand concentration [L] [26]. If AG&] is the free energy change in the presence of L, then:

AAC(T, = AC(T) - AGtk] = RTln(1 + [L]/Kd).

The binding constant of glucoamylase-1-(471-616) for p- cyclodextrin was measured by ultraviolet difference spectroscopy. At pH 7.8, Kd = 11.9 k 0.1 pM, determined as described previously [22]. The expected change in free energy, when [P-cyclodextrin] = 5.18 mM, is 14.8 kJ/mol. The ther- modynamic values refer to protein unfolding, assuming Kd of

amylase-1-(471 - 616), estimated from the Gibbs-Helmholtz equation relating AG(293) to T,, AH, and AC,, was 16 k 8 kJ/ mol (at 5.18 mM P-cyclodextrin), which despite the relatively high error, is in agreement with the calculated value.

From the thermal unfolding data for glucoamylase-l- (471 -616), AC(293) was 16 5 4 kJ/mol at 293 K, which compares with a value of 32 i 2 kJ/mol at 293 K from denaturd- tion in GdnCl (Fig. 8). The discrepancy does not arise from experimental error. Denaturation by GdnCl is measured by changes in absorbance of the aromatic amino acid residues, which are due solely to the binding domain. On the other hand, the DSC calculations are based on an M , of 24900 for glucoamylase-1-(471- 616) which includes the binding domain and the 0-glycosylated linker region. We have shown that the latter plays little part in the unfolding of glucoamylase 1 or glucoamylase-1-(471- 616), and thus the discrepancy between the two methods may arise from the presence of the 0-glycosylated peptide. Alternatively, it is possible that the structure of the ‘unfolded’ protein is different after thermal or chemical denaturation, which would also give rise to a discrepancy between values of AG. However, it is probable that this effect would not produce such a large discrepancy in AG as seen in Fig. 8 [27].

DISCUSSION

The unfolding of glucoamylase 1 can be represented as in Scheme 2. Either the catalytic domain or the binding domain unfolds first on increasing the temperature, depending on the concentration of P-cyclodextrin. The unfolding of the catalytic domain is irreversible, but the unfolding of the binding domain is reversible. Although the presence of the catalytic domain stabilises the binding domain, the state of foldedness of the catalytic domain has no significant effect on the T, of the binding domain and so the reversible steps in Scheme 2 have equilibrium constants which are not significantly different. Also, the presence of the binding domain has no effect on T, of the catalytic domain. These results imply that the interac- tion free energy, AGAB [26] between the two domains approximates to zero, since a pre-requisite for this model is that only a folded domain stabilises a partner domain, and that on denaturation of one domain, AC becomes zero. However, we have not performed a rigorous analysis of the data of the intact enzyme, since the irreversibility of the unfolding of the catalytic domain precludes calculation of a thermodynamically meaningful AGAH. The mechanism of stabilisation of the binding domain by the catalytic domain may be entropic: with a large peptide at the other end of the linker, the conformation of the binding domain would become more restricted in the unfolded state. As a consequence, AS for unfolding would be less positive, and thus AC would be less negative and manifest by an increase in T,.

An 0-glycosylated linker peptide between two domains is found in a wide range of proteins and is thought to adopt a semi-rigid and extended structure [5, 6, 28, 291. The relatively small enthalpy change associated with ‘unfolding’ of this portion is consistent with this structure exhibiting only a small conformational change on denaturation. The results of Shogren et al. [7] are of interest in this context. Using light scattering of ovine submaxillary mucin, they found that the values of the root-mean-square radius of gyration and of the hydrodynamic radius were large both in GdnCl(6 M) and in NaCl (0.1 M), implying very little conformational difference in these two solvents. Unlike the sugars in glucoamylase, the - -

ligand binding is temperature invariant. A AG(293) for gluco- 0-glycosylation in mucins is predominantly the disaccharide

a-N-acetylneuraminic acid-(2,6)-a-N-acetylgalactosamine-O- Ser/Thr, and is therefore charged. However, electrostatic interactions were shown to be much less important than steric interactions, and so the results can be reasonably extrapolated to the glucoamylase linker peptide.

The stabilisation of the binding domain by a portion of the 0-glycosylated region warrants further discussion, since the mechanism of conformational stabilisation of proteins by glycosylation has not received much attention in the literature. Proteins unfold both reversibly and irreversibly, as represented by: N + U + I

where N is the native form of the protein, U is the reversibly unfolded form and I is the irreversibly unfolded form. I arises from a modification of U such as aggregation, covalent modification, peptide bond hydrolysis, etc. There are many examples in the literature where deglycosylation has clearly altered the U --f I reaction. For example, glucose oxidase from Aspergillus niger and yeast invertase are less soluble and more prone to aggregate after deglycosylation [30, 311. Deglyco- sylated hen ovomucoid and low-density-lipoprotein receptor [32, 331 are more susceptible to proteolysis. There are very few examples where glycosylation has been shown to alter the true conformational stability, i.e. reversible unfolding of the protein, although it is clear that glycosylation, especially when densely packed, can have a dramatic effect on the conformation of the peptide backbone [6,34,35]. It has been reported for galactose oxidase from Dactylium dendroides that deglycosylation affects the rate of reversible denaturation induced by guanidine hydrochloride although no values of Gibbs free energy were measured [36].

The reversible transition between native and unfolded states is described by differences in enthalpy, entropy and Gibbs energy. Experimentally, AG is described by the Gibbs- Helmholtz equation in which stability (i.e. AG) is a function of T,, AH, and AC,. Thus experimentally, glycosylation may affect any one or more of these parameters. T, is decreased by deglycosylation, as shown in Table 2. On the other hand, AHd and AC, for glucoamylase-1-(499-616) and glucoamylase-1-(509- 616) are identical to within experimental error on a molar basis. AC, correlates to the number of con- tacts between non-polar groups in native proteins [25] and the heat capacity change for the folding of globular proteins is proportional to the reduction in water-accessible non-polar surface area [37]. Since the only difference between glucoamylase-1-(499 - 616) and glucoamylase-1-(509 ~ 616) is a hydrophilic glycopeptide, no significant difference in AC, would be predicted. On this basis, it appears that most of the stabilising effect of glycosylation on AG as measured by denaturation by GdnCl is not due to changes in AC,. Further- more, the unfolding of the binding domain approximates to a two-state transition, and the AHd/AH, ratio is not affected by glycosylation. One possible stabilisation mechanism is by effects on the refolding reaction. It is well documented that glycosylation is essential for the in vivo folding of some proteins, such as yeast acid phosphatases 1381. However, only N- glycosylation has been shown to be important in this role. Another possibility may be that glycosylation limits the conformational space available to the unfolded peptide, i.e. glycosylation raises its conformational free energy by decreasing its entropy. In this way, the presence of organised structure in 0-glycosylated ‘denatured’ proteins may help to stabilise the folded protein. At present, however, the exact role of 0-glycosylated in conformational stabilisation remains speculative.

669

The stability of the catalytic domain has also been examined by producing genetically truncated forms of glucoamylase 2 [2]. Although the stability was measured only by catalytic activity after incubation over a range of temperatures and no thermal data were given, it is clear that a small region of the 0-glycosylated linker region is essential for stability (up to residue 482), but that the rest of this region does not affect thermal denaturation. It is not apparent whether this change affects the conformational stability (N + U) or the irreversible change (U 4 I) in the catalytic domain. However, from these results, in combination with those presented here, it can be concluded that the presence of 0-glycosylated linker (z 10 amino acids) is essential for maximum stability of both domains of glucoamylase, but that this effect is brought about primarily by the portion of the linker region closest to the relevant domain.

We thank the Agriculture and Food Research Council (UK) and the Science and Engineering Research Council (UK) for financial support.

REFERENCES 1.

2.

3.

4.

5.

6. I.

8.

9.

10.

11.

12. 13.

14.

15. 16. 17.

18.

19.

20

21.

22

23 24 25

Belshaw, N. J. & Williamson, G. (1990) FEBS Lett. 269, 350-

Evans, R., Ford, C., Sierks, M., Nikolov, Z. & Svensson, B. (1990)

Svensson, B., Larsen, K., Svendsen, I. & Boel, E. (1983) Carlsherg

Gunnarsson, A., Svensson, B., Nilsson, B. & Svensson, S. (1984)

Williamson, G., Belshaw, N. J. &Williamson, M. P. (1991) Bio-

Jentoft, N. (1990) Trends Biochem. Sci. 15, 291 -294. Shogren, R., Gerken, T. A. & Jentoft, N. (1989) Biochemistry 28,

Yamamoto, T., Davis, C. G., Brown, M. S., Schneider, W. J., Casey, M. L., Coldstein, J. L. & Russell, D. W. (1984) Ce1139,

Knowles, J., Teeri, T. T., Lehtovadra, P., Penttila, M. & Saloheimo, M. (1 988) Biochemistry and genetics of cellulose degradation, Academic Press, New York.

Lineback, D. R., Russell, I. J. & Rasmussen, C. (1969) Arch. Biochem. Biophys. 134, 539-550.

Shenoy, B. C., Rao, A. G. A. & Rao, M. R. R. (1984) J . Biosci.

Subbaramaiah, K. & Sharma, R. (1988) Starch 40,182- 185. Clarke, A. & Svensson, B. (1984) Carlsherg Res. Commun. 49,

Svensson, B., Clarke, A. J., Svendsen, I. & Miller, H. (1990) Eur.

Privalov, P. L. (1979) Adv. Protein Chem. 33, 167 - 241. Nozaki, Y. (1970) Methods Enzymol. 26, 43-50. Svensson, B., Pedersen, T. G., Svendsen, I., Sakai, T. & Ottesen,

M. (1982) Carlsherg Res. Commun. 47, 55-69. Munch, 0. & Tritsch, D. (1990) Biochim. Biophys. Acta 2041,

Gottschalk, N. & Jaenicke, R. (1991) Biotechnol. Appl. Biochem.

Fukada, H., Takahashi, K. & Sturtevant, J. M. (1987) Biochemis-

Savel’ev, A. N., Sergeev, V. R. & Firon, L. M. (1989) Biochemistry

Belshaw, N. J. &Williamson, G. (1991) Biochim. Biophys. Acta

Schellman, J. A. (1975) Bzopolymers 14, 999-1018. Pace, C. N. (1990) Trends Biotechnol. 8, 93-97. Privalov, P. L. & Gill, S. J. (1988) Adv. Protein Chem. 39, 191 -

353.

Gene91, 131-134.

Res. Commun. 48,529 - 544.

Eur. J . Biochem. 145,463 -468.

chcm. J . 282,423 -428.

5525-5536.

27-38.

6, 601-611.

113-122.

J . Biochem. 188, 29-38.

111 -1 16.

14,324-335.

try 26, 4063 - 4068.

(Engl. Transl. Biokhimiya) 54, 1411 -1416.

1078, 117-120.

234.

670

26. Brandts, J . F.. Hu, C.-Q., Lin, L.-N. & Mas, M. T. (1989) Bio-

27. Alonso, 0. V. & Dill, K. A. (1991) Biochemistry 30, 5974-5985. 28. Shen, H., Schmuck, M., Pilz, I., Gilkes, N. R., Kilburn, D. G.,

Miller, R. C. & Warren, R. A. J . (1991) J . Bid. Chem. 266,

29. Gerken, T. A., Butenhof, K. J. & Shogren, R. (1989) Biochemistry

30. Takegawa, K., Fujiwara, K., Iwahara, S., Yamamoto, K. & Tochikura, T. (1989) Biochem. Cell. Biol. 67, 460-464.

31. Schulke, N. & Schmid, F. X. (1988) J . Biol. Chem. 263, 8827- 8831.

32. Gu, J. X., Matsuda, T., Nakamura, R., Ishiguro, H., Sasaki, M. & Takahashi, N. (1989) J . Biochem. (Tokyo) 106,66-70.

chemistry 28, 8588 -8595.

1 1 335 - 1 1 340.

28, 5536-5543.

33. Kozarsky, K., Kingsley, D. & Krieger, M. (1988) Proc. Nut1 Acad.

34. Hollosi, M., Perczel, A. & Fasman, G. D. (1990) Biopolymers 29,

35. Walsh, M. T., Watzlawick, H., Putnam, F. W., Schmid, K. &

36. Mendonca, M. H. & Zancan, G. T. (1988) Arch. Biochem. Bio-

37. Livingstone, J. R., Spolar, R. S. & Record, M. T. Jr (1991)

38. Riederer, M. A. & Hinnen, A. (1991) J . Bacterial. 173, 3539-

Sci. USA 85,4335-4339.

1549 - 1564.

Brossmer, R. (1990) Biochemistry 29, 6250-6257.

phys. 266, 427-434.

Biochemistry 30,4231 -4244.

3546.

o-glycosylation and stability : unfolding of glucoamylase induced by heat and guanidine...

Documents