Chapter fouR
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin Klaartje Houben, Mae Saldajeno, Grant Ganshaw, Richard Bott and Rolf Boelens
Abstract Osmolytes are small organic solvent additives that can stabilize protein structures. Propanediol is an osmolyte that is known to stabilize Bacillus lentus subtilisin, as it is used for storage of the enzyme before use as protein degrading component in washing powders. We have studied the effect of 1,2-propanediol on the structure and dynamics of BLS from NMR chemical shift perturbations, saturation transfer experiments and 15N relaxation rates. Chemical shift changes and enhanced saturation transfer are mainly observed for residues located in the substrate-binding interface. In the presence of 20% 1,2-propanediol the overall rotational correlation time of the protein is increased, due to the higher viscosity of the solution. A slight rigidification is observed in one of the loops that align the S4 substratebinding cleft. This rigidification may cause a reduced proteolytic activity and thus reduced autolytic cleavage as well. Therefore, part of the stabilizing effect of 1,2-propanediol on BLS presumably may be ascribed to the reduction of autolysis. 1,2-propanediol H 2C
CH
OH OH
CH3
TMAO CH3 H 3C
δ+
N
CH3
urea O
δ-
O
H 2N
C
NH2
Chapter 4
Introduction The 269-residue protein Bacillus lentus subtilisin (BLS) is a member of the superfamily of subtilisin-like serine proteases. These enzymes have attracted significant research interest due to their amenability for both structural and functional studies as well as their considerable industrial importance as protein-degrading components in washing powders. Several threedimensional structures of subtilisins are available, determined with X-ray crystallography (1STB: Alden et al., 1971; 1SVN: Betzel et al., 1992; 1GCI: Kuhn et al., 1998) as well as with NMR spectroscopy (1AH2: Martin et al., 1997). The secondary structure elements can be divided into an internal core, consisting of a seven-stranded parallel β sheet (e1-e7) and two buried α helices (hC and hF). This core is encapsulated by seven amphiphatic α helices and two antiparallel β strands (Figure 1). The solution and crystal structures are very similar, only in solution a small helix (hH) is absent, as was confirmed by observed flexibility from NMR 15N relaxation rates and rapid hydrogen exchange for residues in that segment (Martin et al., 1997). Subtilisins have been the focus of considerable efforts in protein engineering (see for example Wells and Estell, 1988; Siezen et al., 1991). This has led to several variants with dramatically altered enzymatic activity, specificity and pH optima. Generally, however, little differences are observed in the three-dimensional structures of the various variants. Structural studies of subtilisins over a wide range of pH or in the presence of organic solvents have also shown little variation in the protein structures under these conditions (Fitzpatrick et al., 1993; Lange et al., 1994). Interestingly, a more detailed analysis that focused on protein dynamics has shown differences in internal flexibility among mutant subtilisins. For two engineered variants of BLS with increased enzymatic activity, altered flexibility of the protein structure was observed, using X-ray crystallography (Bott et al., 1992; Graycar et al., 1999) as well as NMR spectroscopy (Mulder et al., 1999). For commercial application autolytic proteolysis of subtilisin is a serious problem since it leads to lower effective product activity. Therefore BLS is stored in the presence of 1,2-propanediol that stabilizes the protein before its actual use as a protein degrading compound (Becker et al., 2003; Bott et al., 2003; Castro and Knubovets, 2003). Thus far the mechanism of stabilization has not been unraveled, and will be the subject of our studies. The stabilization of subtilisin can originate from mainly two different effects: (i) intrinsic stabilization by an osmolytic effect; (ii) reduced enzymatic activity. Organic osmolytes can be grouped according to their ‘compatible’ or ‘counteracting’ stabilizing effects (Yancey et al., 1982). Compatible osmolytes, such as amino acids and polyols, stabilize proteins without substantially affecting protein functional activity, whereas counteracting osmolytes, such as methylamines, increase protein stability and can as well counteract detrimental effects of urea on both protein structure and function (Yancey et al., 1982). In general, the driving force of protein stabilization by both types of osmolytes is thought to be the unfavorable interaction of the osmolyte with the protein backbone, resulting in preferential hydration of the structured protein (for reviews see Timasheff, 1993; Bolen and Baskakov, 2001). The reduced activity and concomitant reduced autolysis can be another factor in the stabilization effect of 1,2-propanediol on BLS. 66
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
��
��
�� ��
��
��
�� ��
�����
����� ��
��
������
�� ��
�� ��
��
��
��
��
Figure 1. Ribbon representation of the 0.78 Å crystal structure (1GCI) of Bacillus lentus subtilisin. The structure is composed of two β-sheets (A: e1-e7 and B: e8-e9) and nine helices (hA-hI). In solution two helices are absent (hA and hH). The three residues that make up the active site catalytic triad are indicated as well as the residues that coordinate the two calcium ions.
To arrive at a better understanding of the specific effect of 1,2-propanediol on BLS we have identified the regions that are affected by the presence of 1,2-propanediol in the protein solution from proton and nitrogen chemical shift perturbations, as well as saturation transfer experiments. In order to see whether 1,2-propanediol influences the protein backbone flexibility, we have measured NMR 15N relaxation rates in the presence of 20% 1,2propanediol. For comparison we have studied the effects of either trimethylamine N-oxide (TMAO), a counteracting osmolyte, and urea, which destabilizes proteins, on the NMR chemical shifts and protein backbone dynamics as well.
Results Thermal stability In order to see whether the thermal stability of BLS is affected by the presence of 1,2-propandiol in the protein solution, we have performed differential scanning calorimetry (DSC) runs, both in the absence an in the presence of 1,2-propandiol. In Table 1 the melting temperatures of BLS are given. The melting temperature of BLS in the absence of a solvent additive is 66.2 ˚C, which is practically unaltered in the presence of 1,2-propanediol. Autolytic cleavage The effect of 1,2-propanediol on the autolytic activity of BLS was probed by incubation of the protein for 24 hours at 42 ˚C in the absence or the presence of 1,2-propanediol. The 67
Chapter 4
amount of intact protein was assayed through an activity test (Hsia et al., 1996). In the presence of 1,2-propanediol the remaining activity after 24 hours is significantly higher than in the absence of propanediol (Table 1). This indicates that the BLS is less prone to autolytic cleavage when 1,2-propanediol is present in the protein solution. Table 1. Melting temperatures and autolytic cleavage at 42 ˚C of BLS in the absence or the presence of 1,2-propanediol. Solvent Additive1
Melting Point [˚C] 2
Initial Activity3
Activity3 after 24 hrs4
Activity remaining [%]
-
66.2
6.7 6.6
3.1 3.0
46 45
20 % 1,2-propanediol
66.3
7.1 7.4
5.6 5.4
79 73
Both samples contained a 25 mM acetate buffer at pH 5. Using 0.36 mg of protein. 3 Activity in (U/ml). 4 Incubation for 24 hours at 42 ˚C. 1 2
Chemical shift perturbation NMR chemical shifts are highly sensitive probes of the local electronic environment of nuclei. Changes in the chemical shifts therefore provide very useful information about the chemical structure of biomolecules, such as local geometry and interaction with other molecules. In Figure 2A & D the changes in backbone amide proton and nitrogen chemical shifts due to the presence of 20% 1,2-propanediol (2.7 M) in the protein solution are shown. A gradual up-field chemical shift change is observed for the amide protons and nitrogens of most of the residues, indicating a global effect of 1,2-propanediol on the protein. Distinct chemical shift changes are however observed for a sub-set of residues and are indicated on the protein structure in Figure 3A, where large and medium changes are represented by large and small black spheres, respectively. Most residues are located in the face of the enzyme that represents the substrate-binding cleft (Ala96, Ser97, Gly100, Val102, Leu124, Ser126, Gly205, Ala209, Gly213; using a linear numbering from 1-269), as indicated in Figure 3B, where the protein backbone is colored either black or dark grey for residues participating in enzyme-substrate interactions. Residues that line-up the S1 and S4 substrate-binding pockets, which are the predominant determinants of substrate binding, are represented by the black ribbon. Additionally, large chemical shift changes are observed for residues in other flexible loops (Gln176, Leu251) as well as for the NHε signal in the side chain of Arg269. In the presence of 1 M TMAO the solubility of BLS is lower, since approximately 2530% of the protein precipitated. Chemical shift changes in the 1H-15N HSQC spectrum upon addition of TMAO to the protein solution are indicated in Figure 2B & E. Large and medium chemical shift changes are indicated on the structural model in Figure 3A by large and small dark-grey spheres, respectively. The chemical shift changes in the presence of TMAO are less pronounced than in the presence of 1,2-propanediol; clearly the global change in chemical 68
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
shifts is absent in the presence of TMAO. The largest chemical shift changes are found for residues in the flexible loops around Ser55 and Ser253 as well as for residues around the low affinity calcium-binding site, a turn at Ala83 and in helix hG (His243, Asn246 and Thr247). Smaller chemical shift changes are observed in the small N- (hA, hB) and C-terminal (hI) helices. Clearly the effect of TMAO on the protein structure is very different from what is observed for 1,2-propanediol, since none of the residues in the substrate-binding cleft display large chemical shift changes in the presence of TMAO. Apart from chemical shift changes, an increase in intensity of a few weak signals in BLS are observed in the presence of TMAO, which is the case for Ser76 and Ala131 and the NHε signals of Arg164 and Arg241. This is presumably due to a reduced conformational exchange broadening of the signals. In the crystal structure (1GCI) the NH2η side chains of these two arginines are hydrogen bonded to oxygen atoms that coordinate the calcium in the weak calcium site. In the ensemble of BLS solution structures, however, the side chains of Arg164 and Arg241 have several different orientations and are in most structures exposed to the solvent. The increased NHε signal intensity is therefore likely due to a change in the ensemble of side chain conformations for these two arginines, resulting in a reduction of exchange broadening.
�������������������������������
�� �� ���� ����
��
��
��
��
��
��
��
��
��
�� �� ��
��
��
�� ��
���������������������
A
����������
B
������������
C
� �����
���� � �����
���� � ����� �����
�
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��������������
��������������������������������
�� �� ��� ��� ��� ���� ����
��
��
��
��
��
��
��
��
��
�� �� ��
��
��
�� ��
���������������������
D
��� ��� ���� ����
����������
E
��� ��� ���� ���� ����
������������
F
�
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��������������
Figure 2. Proton (A-C) and nitrogen (D-F) chemical shift changes in the presence of 1,2-propanediol, TMAO and urea. The difference between BLS chemical shifts in the presence of solute minus the chemical shifts in absence of solute are given. Secondary structure elements are indicated on top, where rectangulars represent α helices and arrows represent β strands. β sheet A is composed of the black strands and B of the grey strands. The two buried α helices are indicated in grey.
69
Chapter 4
The effect of urea, a protein-destabilizer, on BLS was also probed by proton and nitrogen chemical shift perturbations. These perturbations are displayed in Figure 2C & F and are indicated by the white spheres in Figure 3A. In proximity of the substrate-binding interface large chemical shift changes are observed for Ala96, Ser182 and Asn212, where smaller changes are observed for Gly100, Gly125 and the active site serine 215. Additional changes are found for Ser36, for Ile43 and Ser48 at the edges of β-strand e1 and for Gly52 in the following flexible loop. In addition to residues that show chemical shift changes direct after addition of urea, several residues around the weak calcium site display chemical shift changes in time after addition of urea (indicated by dashed lines in Figure 2 & 3), presumably due to changes in the calcium occupancy of the calcium-binding site. Moreover, during time signal intensities of a few amides reduce significantly and are indicated by an asterisk in Figure 3, and several signals appear at random coil positions, indicative of protein breakdown. The residues that lose signal intensity are located in surface exposed loops that are flexible, line up the S4 binding pocket, or are part of the calcium binding-sites, and presumably these regions represent sites that are prone to autolytic cleavage.
A ��
��
B
���
��
� �
��
��
��
��
��
�
��
��
�
���
��
��
���
���
��� ���
���
���
� ���
�� ��
���
���
��
�
���
��
� ���
��
� �
��
�
���
��
���
�
���
���
���
��� ��� ���
��� ��� ���
���
���
��
�
�
�
���
��� ���
���
���
Figure 3. Chemical shift changes, flexible parts and substrate-binding pockets of Bacillus lentus subtilisin indicated on the ribbon representation of the three-dimensional structure (1GCI) of Bacillus lentus subtilisin. (A) Large and small spheres represent large and medium chemical shift changes. Chemical shift changes observed in the presence of 1,2-propanediol are indicated in black, for TMAO in dark grey and for urea in white. Dashed black lines indicate residues that display chemical shift changes in time after addition of urea to the protein solution and stars represent places in the structure where loss of signal intensity is observed in time after addition of urea. (B) Flexible residues in either of the three solvents are indicated by grey spheres to facilitate the discussion in the text. Dark-grey spheres are residues that have a contribution of conformational exchange in either of the solutions. Regions involved in substrate binding are colored darkgrey or black. Black regions line-up the S1 and S4 substrate-binding pockets.
70
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
The only residue that displays distinct chemical shift changes in all three different solvents is Ala166 in the weak calcium-binding site. Interestingly, the observed chemical shift changes in the presence of urea are opposite to what is observed for both TMAO and 1,2-propanediol. Opposite chemical shift changes between TMAO and urea are also observed for residues His243, Asn246 and Thr247 in helix hG. In the case of 1,2-propanediol and urea, opposite chemical shift changes are observed for Gly100, Asn212 and Gly213, which are all residues located in the substrate-binding interface of the enzyme, whereas changes in the same direction are observed for Ala96. Between 1,2-propanediol and TMAO a distinct opposite behavior is observed for Ser55. The temperature dependence of the proton chemical shifts of BLS, measured over a range from 300 to 320 K (data not shown), showed no difference in the absence or in the presence of either 1,2-propanediol or TMAO, indicating no dramatic differences in hydrogen bond strengths. Saturation transfer To identify whether 1,2-propanediol transiently interacts with BLS, we have performed saturation transfer experiments. Upon irradiation of either water, and thus also the hydroxyl groups of 1,2-propanediol, or the methyl group of 1,2-propanediol, the intensities of many signals in the 1H-15N SE-HSQC decrease (Figure 4A & B). There are, however, some differences in the intensity changes in the absence and in the presence of 1,2-propanediol (Figure 4C). In the case of water saturation (Figure 4A & C) a decrease in saturation transfer is observed due to the presence of 1,2-propanediol for a few residues in the β strands (e4, e5, e6) and α helix (hF) in the core of the protein (Ala120, Val148, Val171, Ala226), as well as for Ser126 in the substrate binding cleft and some amides at the protein surface (Ala131, Ala166 and Gln239). This could indicate that these positions are less accessible for either water or 1,2-propanediol, when 1,2-propanediol is present in the solution. An increase of saturation transfer is observed for residues in β strands e1 and e2 (Gly46, Tyr89), in close proximity to the active site (Thr64 and Met216), the backbone segment that is known to weakly bind the leaving component of the substrate (Thr202, Tyr203, Ser206) and for Gln185. This increase can be due to an increase in saturation transfer from water to the protein, as well as due to transfer of saturation from the hydroxyl groups of 1,2-propanediol to the protein. When saturating the methyl resonance of 1,2-propanediol, methyl resonances of the protein itself are saturated as well, which causes several residues in the absence as well as in the presence of 1,2-propanediol to decrease in intensity (Figure 4B). However, when comparing the relative signal intensities, some distinct differences are observed in the presence of 1,2-propanediol (Figure 4C). Increased saturation transfer is observed for Gly46, Gly52, Ser154, Gly155, Phe183 and Ser206 at the protein surface, as well as for Thr64 and Thr214 adjacent to the active site His62 and Ser215, respectively. Many of these residues are located in regions where also an increase of saturation transfer is observed when the water resonance is irradiated. We interpret this as that in the latter case saturation transfer in these regions predominantly originates from the 1,2-propandiol hydroxyls, rather than from water. Most of the residues for which an increase of saturation transfer is observed due to the 71
Chapter 4
presence of 1,2-propanediol in the solution are located in the substrate binding interface, in which area also the predominant chemical shift changes are observed, as mentioned above. This strongly suggests that 1,2-propanediol indeed preferentially resides at this site of the enzyme.
A
������������������������������������
���
���������
��� ��� ���
��� ����������������������������
��� ���
�
��
��
�� ��
B
���
��
��
��
��
��
��
��
��
��
���
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��
��
��
����������������������������
��
��
��
�� �� ��
��
��
�� ��
�����������������������������������������������������
���������
��� ��� ��� ��� ���
������������
C
�
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
����������������������������������������������
���
����������������
�����������������
��� ��� ���� ����
�
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��������������
Figure 4. Saturation transfer in the absence and the presence of 1,2-propanediol. (A) Relative amide signal intensities in the HSQC spectrum in the absence (white circles) or in the presence (black circles) of 1,2propanediol, while irradiating the water resonance for 1 second prior to the experiment. (B) Relative amide signal intensities in the HSQC spectrum in the absence (white squares) or in the presence (black squares) of 1,2-propanediol, while irradiating the 1,2-propanediol methyl resonance during 2 second prior to the experiment. (C) Difference in relative intensities due to the presence of 1,2-propanediol in the case of water irradiation (dark-grey circles) or methyl irradiation (light-grey squares).
Dynamics from 15N relaxation rates It appears that 1,2-propanediol transiently binds in the substrate-binding interface of BLS. In this way it could affect the internal dynamics at that site, where previously changes in dynamics were related to altered enzyme activity (Bott et al., 1992; Graycar et al., 1999; Mulder et al., 1999). Therefore we measured the backbone dynamics of BLS from NMR relaxation rates. The 15N relaxation rates of BLS without solvent additives, as well as in the presence of either 1,2-propanediol or TMAO are shown in Figure 5. In the presence of 1,2-propanediol the average R1 and R2 relaxation rates for residues located in secondary structure elements have changed to 0.91 ± 0.02 s-1 (versus 1.56 ± 0.05 s-1) and 18.1 ± 0.8 (versus 10.2 ± 0.4 s-1), giving an apparent rotational correlation time of 14.0 ± 0.7 ns (versus 72
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
7.5 ± 0.4 ns). The increase by a factor of 1.87 of the rotational correlation time correlates well with the approximate 1.85 fold increase of the solvent viscosity in the presence of 20% 1,2propanediol at 315 K (Yaws, 1995-1997). The apparent rotational correlation time has also increased in the presence of TMAO (8.6 ± 0.5 ns). Molecular dynamics simulations have shown that TMAO decreases water diffusion and increases the number as well as lifetime of water hydrogen bonds (Zou et al., 2002). Here we find that this increased water structure probably also affects the rotational diffusion of the protein.
�� ��
��
��
��
��
��
��
��
��
��
�� �� ��
��
��
�� ��
��������
��� ��� ��� ���
��� ����������������� ����������������������������
���
�
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��
��
��
��
��
��
��
��
��
��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� ���
��������
��� ��� ��� ��� �
��������
�� �� �� �
�
��������������
Figure 5. 15N R1, R2 and {1H}-NOE relaxation parameters for BLS in the presence and absence of solutes at 600 MHz and 315 K. Ser253 falls off the scale in the upper panel for BLS in the presence of both 1,2-propanediol and TMAO and is missing for BLS in absence of solutes due to severe overlap.
The relaxation rates have been analyzed using the reduced spectral density approach (Peng and Wagner, 1992; Farrow et al., 1995; Ishima et al., 1995) and spectral density plots are shown in Figure 6. To facilitate the discussion below several residues that are indicated in Figure 6 are represented by spheres on the protein structure in Figure 3B. While the rotational diffusion time-scales clearly differ in the presence of solute, the time-scales for internal motions are very similar and around 300 ps (see Material and Methods). The spectral density values for BLS are displayed in Figure 6A & B. Residues in the most flexible regions of the protein structure, as were identified previously (Remerowski et al., 1996; Martin et al., 1997; Mulder et al., 1999), are indicated by the black diamonds and are located in the two stretches that line-up the S4 substrate-binding pocket, as well as in some loops (see Figure 6). The relative high J(ωN) value for Ser101 indicates the presence of internal dynamics on a slightly slower timescale. The two white squares are residues (G52, Ala131) for which the higher J(0) values indicate the presence of conformational exchange. For Ala131 this is 73
Chapter 4
also reflected in the observed line-broadening in the 1H-15N HSQC spectrum due to a high 1 H R2 relaxation rate, which in turn causes the low precision of the relaxation rates for this residue. In addition to the very flexible residues indicated by the black diamonds, grey circles that have both lower J(0) and J(ωN) values than the core of the protein, indicate additional flexible residues (Gly45, Arg180, Ala188, Ser250, Leu251, Arg269).
���
�����������
��� ���
��� ��������������� �������
������ ���
�������������������������
C
���
E
������
�� ��� ��� ���
��
��� ���
���
��������������
A
������
��� ��� ��
�������
��
��� ��� ���
��� ���
���
���
���
��� �� �� �������
��� ���
B
��������������
���
D
���
F
���
��� ���
���
��� ���
���
���
���
���
���
��� ���
��
���
���
��� ��� �� ��� ��
�
�
�� ���������������
�
�
�
�
���������
�
�
�
�
�
�
���������
�
�
�
�
�
���������
�
�
Figure 6. Reduced spectral density plots of J(0) versus J(ωN) (upper panels) and J(0) versus J(ωH+ωN) (lower panels) for BLS in without solvent additives (A,B), in the presence of 1M TMAO (C,D) and in the presence of 2.7 M 1,2-propanediol (E,F).
The spectral density values in the presence of TMAO are presented in Figure 6C & D. Changes are observed for both Gly52 and Ala131, where there is no evidence for conformational exchange in the presence of TMAO. Ala131 has increased signal intensity in the 1H-15N HSQC spectrum in the presence of TMAO, which indicates also a reduced contribution of conformational exchange to the 1H R2 rate. A contribution of conformational exchange is instead found for Leu80, Ser154 and Met216. The latter two residues are located adjacent to residues involved in substrate interactions, whereas Leu80 is located in close proximity of the high affinity calcium-binding site. Ser101, which seems to be mobile on a slightly slower internal time frame than other residues for BLS in the absence of a solvent additive, has in the presence of TMAO a similar motional time-scale as other flexible residues. It should be noted here that residue Ser253, which is the most flexible amide in the protein, had to be left out of the analysis of BLS without solvent additive, due to severe overlap. The same is true for Ser55 in the presence of TMAO. 74
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
Figure 6E & F display the spectral density plots for BLS in the presence of 1,2-propanediol. None of the residues seem to have a considerable contribution of conformational exchange, with the exception of Ala131, which signal intensity is, as in absence of 1,2-propanediol, low in the 1H-15N HSQC. Like for BLS in the absence of solvent additives, the time-scale of internal motion for Ser101 seems to be slightly slower than for other residues, and a similar slower time-scale correlates with the spectral density values for Thr132. Both Ser126 and Gly252 have moved along the dashed line towards slightly higher J(0) and J(ωN) values, indicating a slight rigidification of these loops in the presence of 1,2-propanediol.
Discussion Several studies have focused on the origin of the protein stabilizing effect of TMAO. This osmolyte is thought to have an indirect stabilizing effect, where the unfavorable interaction between TMAO and the amide unit in the peptide backbone makes the dominant contribution to stabilization (Wang and Bolen, 1997). Molecular dynamics simulations revealed an enhanced water structure in the presence of TMAO (Zou et al., 2002). In contrast, the favorable interaction of urea with the protein backbone provides the dominant driving force for protein unfolding. Urea was shown to weaken water-water interactions and likely TMAO can counteract the destabilizing effect of urea through enhancement of water structure (Zou et al., 2002). The presence of the osmolyte TMAO in solution showed changes in proton and nitrogen chemical shifts for a limited number of residues of BLS, located in flexible loops, helices and a turn as well as around the weak calcium site. The chemical shift changes due to TMAO around the calcium site as well as in the amphiphatic helix hG are opposite to what is observed in the presence of urea. This gives an indication that these are regions of the protein that are presumably stabilized by TMAO, while destabilized by urea. The amides affected in helix hG are involved in hydrogen bonds and reside at the hydrophilic site of the helix. The effect of TMAO on the backbone dynamics of BLS is mainly reflected in a slower rotational diffusion as well as in an increase of ms-μs conformational exchange for Leu80, Ser154 and Met216 and a reduction of conformational exchange for Gly52 and Ala131. While Leu80 is located in the loop that coordinates the calcium in the high affinity-binding site, the other two residues are located in the face of the enzyme that harbors the active site, with Ser154 adjacent to the oxyanion residue Asn153 and Met216 adjacent to the active site serine 215. Both Ser154 and Met216 are involved in a hydrogen bond at the surface of the protein, while the amides of Gly52 and Ala131, for which conformational exchange is reduced, are solvent exposed. The presence of TMAO in the solution might stabilize the interaction of Gly52 and Ala131 with water, while in the case of Ser154 and Met216 interaction with water is competing with intramolecular hydrogen bonds, causing the exchange broadening for these amides. Tryptophan phosphorescence studies showed no effect of TMAO on the conformational dynamics for the native fold of several proteins (Gonnelli and Strambini, 2001). Here we found, however, that in the case of BLS the rotational diffusion of the protein is altered, and 75
Chapter 4
small local changes in dynamics in the presence of TMAO could be identified. The effect of the solute 1,2-propanediol on BLS is distinct from both TMAO and urea. Chemical shift perturbations are mainly observed for residues located in loops in the face of the enzyme that contains the active site. Several amides in the substrate-binding loops are affected by the presence of 1,2-propanediol in the solvent, which might be due to a specific interaction of 1,2-propanediol with the protein in this region. Saturation transfer experiments where either the water, and thereby also the hydroxyl groups of 1,2-propanediol, or the methyl resonance of 1,2-propanediol were irradiated, showed an increased of saturation transfer for several residues also located in the face of the enzyme that contains the active site and substrate-binding clefts. Both chemical shift changes and saturation transfer experiments thus suggest that 1,2-propandiol preferentially resides at this face of the enzyme. Analysis of the 15N relaxation rates shows that BLS tumbles slower in the presence of 20% 1,2-propanediol, in agreement with the increased viscosity of the solvent. For most residues, however, the internal dynamics are unaltered with exception of Ser126 and Gly252, for which a slight rigidification is observed. The rigidification of S126 is especially interesting, since it participates in both S1 and S4 substrate-binding pockets. In the presence of 1,2-propandiol a reduction in saturation transfer when irradiating the water signal was also observed for this residue. Unfortunately, few other probes are available in this stretch, due to the presence of Pro127 and Pro129, and severe overlap for Ser128. In an engineered variant of BLS (RSYSA), having increased enzymatic activity, residues 123-125 were shown to have increased mobility (Graycar et al., 1999). This increased mobility presumably accounted for the enhanced proteolytic activity by enabling the binding cleft to adapt more readily to different reactive sites on a protein substrate surface. Rigidification of this stretch could therefore have the opposite effect, resulting in a lower proteolytic activity of the protein, and thereby reducing autolysis. In view of the fact that both chemical shift changes as well as increased saturation transfer effects in the presence of 1,2-propanediol are also mainly observed for residues in the substrate-binding interface, it is likely that 1,2-propanediol indeed transiently binds at this face of the enzyme and thereby affects the dynamics and the activity of the enzyme. Therefore the stabilizing effect of 1,2-propanediol in the storage medium of BLS presumably mainly resides in the reduction of autolysis, rather than in an increase of the intrinsic stability of the protein structure. This correlates well with the unaltered melting temperature of BLS in the presence of 1,2-propanediol, as well as the higher amount of protein left after incubation in the presence of 1,2-propanediol. Since severe dilution of the enzyme solution occurs before its use as protein degrading agent in washing powders, the effect of 1,2-propanediol on the activity will be minimal once applied.
Material & methods NMR samples 15 N labeled samples of Bacillus lentus subtilisin were produced as described previously (Mulder et al., 1999) and were inhibited by DFP (diisopropylfluorophosphate). The samples 76
Towards understanding the stabilizing effect of propanediol on Bacillus lentus subtilisin
contained typically 1 mM protein in a 25 mM deuterated acetate buffer at pH 5.0 with 5% D2O. After addition of 1,2-propanediol (CAS 57-55-6; propylene glycol) to the protein solution, the volume was reduced back to 500 μl, resulting in a 20% v/v (2.7 M) 1,2-propanediol solution. A 5.5 M TMAO solution was prepared by dissolving d9-trimethyl N-oxide (CAS 1184-78-7) in acetate buffer, which was set to pH 5.0 by adding HCl. This was added to the protein samples in steps to reach a final concentration of approximately 1 M TMAO. An 8 M urea (CAS 57-13-6) solution set to pH 5 was added to the protein solution to reach a final concentration of 2.3 M urea. NMR experiments All NMR experiments were recorded on a Bruker AVANCE 600 MHz spectrometer (1H frequency of either 600.13 or 600.28 MHz) equipped with a TXI probe with z-gradients, at a temperature of 315K. The different solutes were added in steps to the protein sample and after each addition a 1H-15N SE-HSQC spectrum was recorded, using typically 4 scans, spectral widths of 12 x 2.5 kHz and 896 x 220 complex points, for 1H and 15N respectively. 15 N relaxation experiments (R1, R2 and 1H-15N heteronuclear NOE) were recorded as described by Farrow et al. (1994) using pulsed field gradients. Relaxation delays for the R1 experiment were 0, 100, 200, 300, 400 (2x), 600, 800, 1000, 1200, 1600 ms and for the R2 CPMG experiment 0, 16, 32, 48, 64 (2x), 80, 96, 112, 144, 176 (2x) ms using 5 kHz 15N 180˚ pulses with a pulse spacing of 0.9 ms. In the heteronuclear NOE experiment 120˚ pulses were used to saturate proton. No relaxation experiments were recorded in the presence of urea due to rapid degradation of the protein sample by autolysis. Saturation transfer experiments were recorded using a 1H-15N SE-HSQC pulse sequence, where prior to the experiment either the water or methyl resonance of 1,2-propanediol (1.13 ppm) was irradiated using a field strength of 53 Hz during 1 or 2 seconds, respectively. A reference experiment was recorded where the irradiating field was omitted. Data analysis NMR spectra were processed with NMRPipe (Delaglio et al., 1995) and analyzed with NMRView (Johnson and Blevins, 1994). Fitting of relaxation curves was performed with Curvefit (Palmer et al., 1991b) (http:// cpmcnet.columbia.edu/dept/gsas/biochem/labs/palmer/software/curvefit.html) using a twoparameter fit, with the exception of Leu80 in the presence of TMAO, where a three-parameter fitting was used. Due to resonance overlap the relaxation rates of a number of residues could not be accurately determined and are therefore left out of the analysis (e.g. Ser253 in BLS). The relaxation rates have been analyzed using the reduced spectral density approach (Farrow et al., 1995; Ishima et al., 1995). The black lines in the spectral density plots (Figure 6) represent theoretical curves for isotropic rigid body tumbling: ��� � �
� �� � �� ��� � � �
Linear fitting of grey and black J(0) versus J(ωN) data-points, indicated by the dashed lines 77
Chapter 4
in Figure 6, results in a rotational correlation time and a correlation time for internal motion of 7.8 ns and 280 ps for BLS in the absence of solvent additives, 9.0 ns and 330 ps in the presence of TMAO and 14.7 ns and 240 ps in the presence of 1,2-propanediol, respectively. Differential scanning calometry Calorimetric experiments to determine the melting temperatures of BLS in the absence and in the presence of 1,2-propanediol were performed in Palo Alto on a MicroCalTM VP-DSC MicroCalorimeter instrument, using capillary cells of 0.60 ml volume under a constant pressure of 30 psi and a heating rate of 90ºC/hr. The samples contained approximately 0.36 mg of protein in a 25 mM acetate buffer of pH 5 and the same concentration of solvent additives was used as described for the NMR samples. Due to autolytic cleavage of the protein upon heat treatment DSC chromatograms were not reversible, which prohibited further thermodynamic analysis of the data. Autolytic activity assay Protein samples containing 3000 ppm subtilisin in 1.5 ml were prepared using the same conditions as described above. Prior to and after 24 hours of incubation at 42 ˚C using a Eppendorf Thermomixer at 300 rpm, the activity was assayed using the protocol described previously (Hsia et al., 1996).
Acknowledgements Dick Schipper is gratefully acknowledged for inhibition of the subtilisin samples with DFP. This project was funded by the Research Council for the Chemical Sciences of the Netherlands Organization for Scientific Research (NWO-CW).
78
