doi: 10.1016/S0022-2836(02)00137-7 available online at http://www.idealibrary.com on
w B
J. Mol. Biol. (2002) 318, 679–695
NMR Studies of the Backbone Flexibility and Structure of Human Growth Hormone: A Comparison of High and Low pH Conformations Marina R. Kasimova1, Søren M. Kristensen1, Peter W.A. Howe1 Thorkild Christensen2, Finn Matthiesen3, Jørgen Petersen4 Hans H. Sørensen3 and Jens J. Led1* 1
Department of Chemistry University of Copenhagen Universitetsparken 5, DK-2100 Copenhagen Ø, Denmark 2 Protein Process Development Novo Nordisk A/S, Novo Alle´ DK-2880 Bagsværd, Denmark 3
Protein Chemistry, Novo Nordisk A/S, Hagedornsvej 1 DK-2820 Gentofte, Denmark 4
BioProcess Pilot Plant, Novo Nordisk A/S, Hagedornsvej 1 DK-2820 Gentofte, Denmark
15
N NMR relaxation parameters and amide 1H/2H-exchange rates have been used to characterize the structural flexibility of human growth hormone (rhGH) at neutral and acidic pH. Our results show that the rigidity of the molecule is strongly affected by the solution conditions. At pH 7.0 the backbone dynamics parameters of rhGH are uniform along the polypeptide chain and their values are similar to those of other folded proteins. In contrast, at pH 2.7 the overall backbone flexibility increases substantially compared to neutral pH and the average order parameter approaches the lower limit expected for a folded protein. However, a significant variation of the backbone dynamics through the molecule indicates that under acidic conditions the mobility of the residues becomes more dependent on their location within the secondary structure units. In particular, the order parameters of certain loop regions decrease dramatically and become comparable to those found in unfolded proteins. Furthermore, the HN-exchange rates at low pH reveal that the residues most protected from exchange are clustered at one end of the helical bundle, forming a stable nucleus. We suggest that this nucleus maintains the overall fold of the protein under destabilizing conditions. We therefore conclude that the acid state of rhGH consists of a structurally conserved, but dynamically more flexible helical core surrounded by an aura of highly mobile, unstructured loops. However, in spite of its prominent flexibility the acid state of rhGH cannot be considered a “molten globule” state because of its high stability. It appears from our work that under certain conditions, a protein can tolerate a considerable increase in flexibility of its backbone, along with an increased penetration of water into its core, while still maintaining a stable folded conformation. q 2002 Elsevier Science Ltd. All rights reserved
*Corresponding author
Keywords: backbone dynamics; amide proton exchange; human growth hormone; NMR; secondary structure
Present addresses: P. W. A. Howe, Analytical Sciences, Syngenta, Jealott’s Hill Research Station, Bracknell, Berkshire RG42 6EY, UK; F. Matthiesen, NatImmune A/S, Symbion, Fruebjergvej 3, DK-2100 Copenhagen Ø, Denmark. Abbreviations used: rhGH, recombinant human growth hormone; hGHbp2, human growth hormone binding protein dimer; LIF, leukemia inhibitory factor; EPO, erythropoietin; IL, interleukin; NOE, nuclear Overhauser effect; HSQC, heteronuclear single-quantum coherence; CSI, chemical shift indices; DLS, dynamic light scattering; 3HHR, hGH/hGHbp2 complex PDB excess code; 1HGU, the PDB excess code for non-bound rhGH (with seven mutations); TSP, 3-(trimethylsilyl) propionate, sodium salt. E-mail address of the corresponding author:
[email protected]
Introduction Human growth hormone (rhGH) is a 191-residue protein folded in a four-a-helical bundle, typical for growth hormones and other helical cytokines. Previously determined structures include that of the receptor-bound conformation1 – 4 and a nonbound mutant,5 both solved by X-ray crystallography. No solution structure of human growth hormone has been determined so far. The structure of receptor bound native rhGH shows that the four major helices are connected by several shorter helices and loops (Figure 1). While crystal structures are extremely valuable for the elucidation of physico-chemical properties of rhGH, it remains
0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved
680
Figure 1. Connectivity diagram for rhGH, a four-ahelical bundle with up – up – down– down connectivity and long crossover loops. According to the crystal structure of the rhGH/receptor complex (3HHR) the bundle consists of four major helices: first helix (A) from residue 9 to 34, second helix (B) from residue 72 to 92 and from residue 94 to 100, third helix (C) from residue 106 to 128, and fourth helix (D) from residue 155 to 184. The four main helices are referred to as the core of the protein. The surrounding loops are classified as the periphery. The mini-helices A0 , 38 – 47, and A00 , 64 – 70, are also attributed to the conventional periphery, because of their location and because of the uncertainty about their presence in solution. The “north” and the “south” ends of the molecule are defined as indicated on the Figure.
highly desirable to obtain information on the solution conformation of the non-bound molecule. In addition, over the past years it has become increasingly clear that in many cases the function of a protein cannot be explained solely on the basis of its static 3D structure. “Macromolecular function is, in many cases, highly dependent on excursions to excited molecular states and hence intimately coupled to flexibility”.6 In this context it should be noted that protein function is not limited to biological activity. As pointed out by Wetlaufer,7 proteins in general have multiple functions, with folding being the first, and degradation the last. Here we present an example of close relation between protein folding and dynamics. The question of flexibility is most relevant for rhGH, because this protein has an ability to populate different non-native states, both partially folded and alternatively folded (see DeFelippis et al.8 and references therein). In terms of its biological activity, rhGH is also very “flexible”, since it has a high affinity for both its own receptor and the prolactin receptor.9 Therefore, many aspects of structure– functional behavior of rhGH are related to the “excursions to excited molecular states”.
Backbone Flexibility of Human Growth Hormone
Here we describe an investigation of the structural and the dynamic properties of rhGH by NMR. The secondary structure of rhGH in solution was characterized by means of secondary chemical shifts, which are sensitive probes for the regular secondary structure. The dynamic characteristics of the protein were assessed through measurements of the 15N NMR backbone relaxation parameters and amide hydrogen exchange rates. 15N relaxation measurements explore fast atomic motions ranging from tens of pico- to tens of nanoseconds, whereas the hydrogen exchange reactions reflect the breaking and reformation of hydrogen bonds in a secondary structure and therefore probe the motions on a slower, millisecond to second, timescale.10 The combination of these two approaches provides information about a wide range of molecular motions. The investigation of the structure and dynamics of rhGH was performed in neutral and acidic solutions. Under these conditions the protein is known to populate two distinct conformations.8 A comparison of the dynamics of the hormone in solutions of different acidity allowed an identification of the changes in flexibility that accompany the transition of the protein from one conformational state to the other. Based on this information we describe an alternative state of the hormone populated at low pH.
Results NMR assignments 1
H, 15N and 13C NMR assignment of rhGH at pH 2.7 and 6.8 –7.0 was obtained by a combination and of standard [1H,15N]-double-resonance 1 15 13 [ H, N, C]-triple-resonance backbone assignment procedures.11 – 13 Of the 182 backbone amide protons in rhGH, which in principle may be observed, 136 were assigned at pH 2.7 while 108 were assigned at pH 6.8– 7.0. Thus, about a total of 75% of the backbone nuclei of rhGH were assigned, which is about the level of assignment obtainable for this protein with standard double and tripleresonance techniques, combined with uniform 15 N,13C-double labeling of the protein. The relatively low fraction of assigned backbone resonances may seem surprising, since there are many examples of four-helical bundle proteins whose structures were solved by NMR, including human erythropoietin (EPO),14 human IL-6,15 IL-4,16 GCSF17 and mLIF.18 These proteins are of comparable size and topology, which makes us believe that neither the size of rhGH nor its helicity are major obstacles for solving its structure. If spectral overlap is not the reason for the incomplete assignment, it is most probably the increased mobility of certain regions of the polypeptide chain. The structural heterogeneity of these regions could lead to extreme broadening of the NMR signals and hence to the difficulties in their detection.
Backbone Flexibility of Human Growth Hormone
681
Figure 2. The CSI for rhGH at pH 2.7 and pH 7.0. (a) and (b) 1Ha and the 13Ca CSI, respectively. In both panels the data at the two pH values are combined to illustrate the similarity of the secondary structure pattern at the two pH values. (c) A comparison of the helical structure predicted from the CSI data with the information available from structures of the rhGH/receptor complex (3HHR) and the crystal structure of rhGH mutant (1HGU). The helical regions in 3HHR and 1HGU are shown as gray horizontal bars and plotted as a function of residue number. The CSI data at pH 7.0 and at pH 2.7 are plotted above and below the X-ray data, respectively. The black horizontal bars represent the residues participating in the helices, while the open squares are the assigned residues that are not helical, or for which the data were insufficient to deduce the structure.
Secondary structure of rhGH The secondary structure of rhGH in solution was evaluated from the secondary chemical shift indices19 of the Ca and Ha at pH 7.0 and pH 2.7. Figure 2 summarizes the secondary structure prediction for the protein at neutral and acidic pH. According to these data, rhGH is predominantly a helical protein in solution, consistent with the crystal structures. The presence of unassigned regions hinders the full use of the chemical shift data, because in some cases the length of an assigned backbone segment is too short (less than four residues) to satisfy all criteria for a helical structure. Nevertheless, the information provided by the chemical shift indices (CSI) allows a comparison of the solution structure of the protein with the crystal structures of the free and receptor-bound forms (see Figure 2(c)). For example, it appears that the first helix of rhGH in solution is one helical turn longer than in its recep-
tor-bound conformation known from the crystal structure of the hGH/hGHbp2 complex. In solution, the first helix begins at the sixth residue, whereas in the complex it starts at residue 9. The same difference is observed between X-ray crystallographic structures of the receptor-bound (3HHR) and the free (1HGU) hormones; that is, the solution structure of the first helix is the same as in the crystal structure of the free protein. This suggests that the binding between rhGH and its receptor induces a slight unwinding at the beginning of the first helix. Although the solution structure and the crystal structure of the unbound hormone are similar in the first helix, this is not the case for the whole molecule. According to the crystallographic data, the free hormone lacks several mini helices that are present in the rhGH/hGHbp2 complex. Using our CSI results we can verify the existence of one of these helices, located at residues 94 –100. Thus, even though the solution conformation of the
682 Backbone Flexibility of Human Growth Hormone
Figure 3. Experimental relaxation parameters for 15N-labeled rhGH at 32 8C: left panel, at pH 7.0; right panel, at pH 2.7. The data are: (a) and (e) R1; (b) and (f) R2; (c) and (g) NOE; and (d) and (h) R1/R2. The helices, observed in 3HHR PDB file are indicated by black horizontal bars on the x-axis.
683
Backbone Flexibility of Human Growth Hormone
protein is expected to comply with the non-bound crystal form (as in the 1HGU crystal structure), the presence in the solution structure of a helix in this region is consistent only with the bound conformation of the hormone. The other noticeable difference between the NMR data at pH 7.0 and the X-ray data is the formation of a helical structure at residues 103 –107. None of the crystal structures, receptor-bound or free, show this helix. In the bound conformation (3HHR), these residues belong to a turn between the second and the third major helices, whereas in the free hormone (1HGU) this part is unstructured, forming a short inter-helical loop. Whether the residues 103 –107 belong to a longer helix or they form an independent short helix requires further structural investigation. However, the existence of this helical region indicates that none of the crystal structures represent the solution conformation of the non-bound hormone. The secondary chemical shift data from pH 7.0 and 2.7 are plotted in Figure 2(a) and (b). According to these data, the secondary structure content of rhGH is virtually independent of the solution conditions, which agrees well with the previous observations8 that the helical content of the protein is invariant from pH 2.5 to 8. Backbone dynamics Experimental R1, R2 and nuclear Overhauser effect (NOE) values were obtained for 99 backbone amides at pH 7.0, for 136 backbone amides at pH 2.7, and for the indole amide of the single tryptophan (W86) at both pH. These data are summarized in Figure 3. At pH 7.0, both the R1 and R2 relaxation rates are more or less uniform throughout the entire molecule with average values of 1.38 s21 and 19.7 s21, respectively. The transverse relaxation rate, R2, appears to be insensitive to the secondary structure. The only residues with elevated R2 rates, 6, 50, 56, 104, and 137, all belong to the N terminus and the loops, suggesting conformational exchange and/or structural heterogeneity in these regions. The lowest R2 value of 10.2 s21 is found for the C-terminal Phe191. Together with the negative NOE value, the low R2 indicates a high mobility of this residue on the ps to ns timescale. The NOE is the only dynamic parameter affected by the secondary structure environment of the backbone, thus its value is increasing within the major helices while decreasing towards the termini and in the loop regions. The average NOE value at pH 7.0 is 0.66. At pH 2.7 the average of the measured R1 relaxation rates is 1.35 s21, and thus is very similar to the value found at higher pH. In contrast, the R2 rates are highly heterogeneous, and their average, 13.8 s21, is considerably lower than at pH 7.0. Both termini and the loop between helices C and D (see Figure 1) are characterized by low R2 relaxation rates and small NOE values. These regions are the most mobile parts of the protein. Transverse
relaxation rates more than 1.5 times greater than the average are observed for the indole nitrogen of Trp86 and for the backbone amide of Gly161. Because a positive contribution to the R2 results from slow-motion exchange, the magnitude of these rates implies that Trp86 and Gly161 are involved in conformational exchange processes. The pattern of NOE values along the protein sequence (Figure 3) is also different at the two pH values. First, the average NOE value of 0.66 at pH 7.0 is higher than the average NOE value of 0.51 at pH 2.7. Second, at higher pH the NOEs are less variable across the backbone and seem to correlate better with the secondary structure elements, as defined in the crystal structure 3HHR. Molecular rotational correlation time Human growth hormone has a prolate shape. Therefore, the relaxation data were analyzed within the framework of an axially symmetric diffusor. At both pH 7.0 and pH 2.7, the inclusion of the axially symmetric diffusion model is justified as judged from the high F values of 11.3 and 4.4, respectively, when compared with the simple isotopic diffusion model. At both pH values, the estimated orientation of the principal component of the rotational diffusion tensor is nearly parallel to the principal axis of the inertia tensor and the helix bundle axis, as expected. The overall correlation time, tm, of rhGH obtained from the analysis of the relaxation data at pH 7.0 and 32 8C was 13.9 ns and the rotational diffusion anisotropy was 1.20. To exclude the possible influence of the unstructured portions of the protein, the tm value was recalculated using only the residues from the helical regions. Under these conditions a value of 13.8 ns was obtained. Thus, it was concluded that loops and termini are relatively rigid and that their mobility does not interfere with the overall rotation of the molecule. The tm value of 13.9 ns is rather high for a protein of 22.5 kDa. The theoretical value calculated using standard formulae for a spherical molecule20 of this size at 32 8C is 9.0 ns. Estimation of tm using correlation times reported for other members of the four-helical bundle family (hIl-3,21 hIl-416 and mLIF18) gives an approximate value of 10.5 ns. A possible explanation for the elevated rotational correlation time of rhGH is its propensity towards aggregation. Evidence that self-association of the hormone is indeed taking place was provided by NMR and dynamic light scattering (DLS) measurements. First, it was noticed that the NMR signals sharpen when the protein concentration is decreased from 0.90 to 0.45 mM, while an overall rotational correlation time, tm, of 12.6 ns was obtained from the backbone dynamics relaxation experiments at lower protein concentration. Both of these results are expected for a sample with an increased content of monomers. Second, the dynamic light scattering studies, performed at pH 6.0 at salt-free conditions, confirm that the
684
Backbone Flexibility of Human Growth Hormone
Figure 4. Column chart of calculated order parameters, S2 , for the solution structure of rhGH: (a) pH 7.0; (b) pH 2.7; (c) the difference, DS2 , between order parameters at neutral and acidic pH. The helices in the structure of the receptorbound rhGH are represented by black horizontal bars on the x-axis. At pH 7.0 residues 7 – 9, 14, 23, 56, 82, 83, 91, 104, 163, 171, 178, 180, 182 exhibit physically unrealistic values that exceed 1.0 (see the text).
apparent hydrodynamic radius, Rh,app, of rhGH depends on the concentration. At [rhGH] ¼ 0.1 mM, the Rh,app is estimated to be 2.5 nm, which is in agreement with its molecular mass of 22.5 kDa. With an increase of the rhGH concentration to 0.4 mM the Rh,app increases to 2.6 nm, while at [rhGH] ¼ 1.1 mM the hydrodynamic radius rises to 3.0 nm. However, at the latter concentration the reliability of the measurements becomes questionable, because of a progressive increase of Rh,app during the experiment. The change of the molecular size with time implies a time-dependent aggregation of the protein (data not shown). To estimate the degree of aggregation, we conducted a series of independent DLS measurements on the 15 N-labeled sample used in the NMR experiments, and on several freshly prepared unlabeled samples. In spite of filtration through a 22 mm filter, the high-molecular-weight component was present in all samples. The estimated amount of aggregates was between 1.2 and 2.5% in the investigated samples. At pH 2.7 the overall correlation time of rhGH is 11.9 ns and the rotational diffusion anisotropy is 1.33. The overall correlation time is still somewhat higher than expected, although it is more consistent with the size of the protein. Protein selfassociation cannot account for the elevated tm value at these conditions, as suggested by the concentration independence of the NMR spectra.22 Also, the DLS measurements performed under these conditions are consistent with rhGH being
monomeric (data not shown). The R2 =R1 ratios at pH 2.7 (Figure 3) show that the flexibility of the different segments of the protein is different. Thus, the values of R2 =R1 fluctuate between 3 and 20, indicating that some parts of the loops have significant conformational freedom and that their movement might not be well correlated with the overall rotation of the molecule. However, calculations of tm for the rhGH molecule with and without inclusion of the loops give similar results (11.9 ns versus 12.2 ns). The order parameters The order parameters were calculated using the model-free formalism of Lipari & Szabo.23,24 The results of the analysis are presented in Figure 4. Examination of the S2 values shows that this parameter is sensitive to the position of the residue in a secondary structure unit. It is also clear that pH of the solution has a profound effect on the residue-specific pattern of S2 values. At pH 7.0 the variation between the helical regions and the rest of the protein is less significant. Although there is a small decrease of S2 toward the termini and within inter-helical segments, the overall order parameters appear almost uniform with an average value of 0.93(^ 0.09). In the following analysis, we compare the flexibility of the major four-helical bundle, or the core of the protein, to the periphery (for the definition of the rhGH core, see Figure 1). If the order parameters
685
Backbone Flexibility of Human Growth Hormone
are calculated separately for the core and the noncore parts, average values of 0.96(^ 0.07) and 0.87(^ 0.11), respectively, are obtained. These values are sufficiently high to classify the whole protein, both its helices and the periphery, as uniformly rigid and tightly packed. With the decrease of pH to 2.7, the order parameters also decrease (Figure 4(b)) and reach an average S2 value of 0.76(^ 0.20). The difference in S2 values obtained at pH 7.0 and pH 2.7, designated as DS2 ; is plotted as a function of sequence in Figure 4(c). At low pH, the difference in dynamics of the four-helical bundle core and peripheral regions becomes larger. Calculation of the average S2 values performed separately for the core part and the rest of the protein gives 0.85(^ 0.12) and 0.64(^ 0.21), respectively. The variation of the S2 values throughout the molecule also becomes higher. For example, there are residues in the first helix that have S2 values as low as < 0.5, whereas in the loop between the second and the third helices the values fluctuate from < 0.2 to < 0.8 (Figure 4). Proton – deuterium exchange At pH 2.7, quantitative exchange rates were obtained for 69 backbone amide protons. Numerous amide protons exchange even prior to the beginning of the measurements, and their resonances are therefore absent in the HSQC spectrum recorded immediately after the period required for dissolving the protein and adjusting the pH. Also the amide protons of some residues exchanged too fast for an accurate rate determination, even though they are still present in the HSQC spectrum. Therefore, these exchange rates were not considered further. At pH 7.0 the exchange rates were obtained for 22 backbone amide protons. Table 1 and Figure 5 summarize the data obtained at the two pH values. Notwithstanding the sparsity of the data at pH 7.0, two important observations can be made. First, the values of the protection factors are much higher at this pH than at pH 2.7, and second, they are more uniformly distributed throughout the length of the bundle. The slow exchanging residues are found as frequently at one end of the molecule as at the other, which contrasts with the distribution of the slowly exchanging residues at pH 2.7, where most of them are clustered at the south end of the rhGH molecule (see Figures 1 and 5). The most noteworthy result of this experiment is the pH dependence of the amide exchange rate for Trp86 side-chain. This residue is located in the middle of the protein and, according to crystallographic data, it is completely buried in the protein core. The indole nitrogen of Trp86 forms a hydrogen bond1 to the Od1 of Asp169. Consistent with this, at pH 7.0 the indole hydrogen is highly protected from the exchange with the solvent (see Table 1). However, at pH 2.7 the Trp86 side-chain NH resonance is missing from the NMR spectra,
which suggests that it is fully exchanged prior to the data acquisition.
Discussion Solution conformation of rhGH The NMR data indicate that the overall solution conformation of rhGH is similar to that in the crystal phase. Thus, the secondary CSI confirm the presence of the four main helices forming the bundle and the long unstructured loops that connect the helices. According to the crystallographic data, the receptor-bound (3HHR) and the free (1HGU) hormones differ in the amount and the location of several shorter helices (see Figure 2(c)). In particular, this is the case in the region 90– 112 that forms a loop in the non-bound hormone, but belongs to the helices B and C in the rhGH/hGHbp2 complex. The variation between the different crystal structures suggests a high conformational mobility in this region, raising questions concerning its structure in solution. It is therefore interesting that the Ca CSI data at both pH values are consistent with the formation of short helices within the region 90 –112 (see Figure 2). The backbone dynamics, and in particular the high order parameters of several residues in this region (see Figures 3 and 4), also imply the presence of a stable structure at both pH values. This is further supported by the slow amide hydrogen exchange for Leu93 and Ala98 at pH 7.0, and for a number of residues at pH 2.7 (Table 1). Backbone and indole dynamics pH 7.0 Recombinant human growth hormone represents a unique example of a protein with an extreme variation of the backbone dynamics as a function of pH. The average generalized order parameter at physiological pH, S2average ¼ 0:93; is higher than seen for other globular proteins,25 which normally display average generalized order parameters between 0.80 and 0.91. The helical regions of the backbone have an S2average of 0.96(^ 0.07), which is somewhat higher than the values of 0.88 –0.92 obtained for bundle helices in other helical cytokines.16,21,26 The residue-specific S2 values are homogeneous throughout the molecule with only a small difference between the average values for helical and non-helical regions (0.96(^ 0.07) and 0.87(^ 0.11), respectively). This shows that at pH 7.0 the mobility of the long connecting loops is as restricted as in the helical bundle. Similar uniform distribution of backbone dynamics was found for LIF,26 a rigid four-helical bundle cytokine. Still, the average order parameter obtained here is surprisingly high and close to the maximum value. Moreover, for 15 residues (see Figure 4),
686
Backbone Flexibility of Human Growth Hormone
Table 1. Amide hydrogen exchange rates protection factors of rhGH at 32 8C pH 7.0 Residue R8 L9 F10 D11 N12 A13 M14 L15 R16 A17 R19 L20 L23 A24 Y28 E32 A34 L45 N47 Q49 L52 T60 R77 I78 S79 L80 L81 L82 I83 Q84 S85 W86b F92 L93 R94 V96 A98 G104 A105 D107 Y111 L113 L114 D116 L117 E118 E119 G120 I121 Q122 T123 L124 M125 G126 R127 S132 Q137 I138 S144 D154 K158 L162 D171 K172 V173 E174 T175 F176 L177 R178
21
kexch (h )
pH 2.7 Protection factor
0.15 ^ 0.04
209236 ^ 55796
0.124 ^ 0.016 0.43 ^ 0.15 0.026 ^ 0.008
63578 ^ 8201 27118 ^ 9459 679091 ^ 190783
0.014 ^ 0.009
21
kexch (h )
Protection factor
0.77 ^ 0.33 0.16 ^ 0.02 0.074 ^ 0.008 0.42 ^ 0.08 0.038 ^ 0.006 0.13 ^ 0.02 0.09 ^ 0.01 0.072 ^ 0.006 0.09 ^ 0.01 0.151 ^ 0.025 0.23 ^ 0.03 0.036 ^ 0.003 0.12 ^ 0.02 0.27 ^ 0.08
23.7 ^ 10.2 31.6 ^ 3.9 80.8 ^ 8.7 94.5 ^ 18.0 2325 ^ 367 201.0 ^ 30.9 115.2 ^ 12.8 63.45 ^ 5.29 90.3 ^ 10.0 126.0 ^ 20.9 250.2 ^ 32.6 140.3 ^ 11.7 43.00 ^ 7.17 59.0 ^ 17.5
0.076 ^ 0.006 0.31 ^ 0.06
201.1 ^ 15.9 82.1 ^ 15.9
0.100 ^ 0.015 0.09 ^ 0.04
361.4 ^ 54.2 89.1 ^ 39.6
407943 ^ 263000
0.320 ^ 0.273 0.023 ^ 0.004
31015 ^ 26463 402717 ^ 70160
0.025 ^ 0.004
86854 ^ 13874
0.014 ^ 0.003
1048576 ^ 225338
0.046 ^ 0.012
124157 ^ 32361
0.065 ^ 0.006
334052 ^ 30855
0.025 ^ 0.002
131458 ^ 10500
0.23 ^ 0.05
17989 ^ 3911
0.35 ^ 0.30 0.057 ^ 0.005
22525 ^ 19309 199919 ^ 17537
0.28 ^ 0.05
234191 ^ 41814
0.064 ^ 0.008 0.043 ^ 0.006
190787 ^ 23833 265008 ^ 36944
0.004 ^ 0.003 0.11 ^ 0.02
1058444 ^ 793040 103594 ^ 18842
a
a
0.047 ^ 0.006 0.043 ^ 0.005 0.038 ^ 0.004 0.038 ^ 0.004 0.032 ^ 0.005 0.022 ^ 0.005 0.015 ^ 0.006 0.020 ^ 0.006 0.026 ^ 0.004 0.064 ^ 0.007
107.2 ^ 13.7 189.1 ^ 22.0 80.69 ^ 8.49 290.2 ^ 30.6 167.5 ^ 26.2 193.7 ^ 44.0 284 ^ 114 111.5 ^ 33.4 232.8 ^ 35.8 450.0 ^ 49.2
a
a
0.070 ^ 0.012 0.039 ^ 0.005 0.22 ^ 0.09 0.44 ^ 0.23
126.7 ^ 21.7 93.95 ^ 12.0 36.96 ^ 15.1 8.92 ^ 4.66
a
a
a
a
0.023 ^ 0.004 0.032 ^ 0.009 0.20 ^ 0.02 0.054 ^ 0.005 0.040 ^ 0.010 0.079 ^ 0.006 0.016 ^ 0.004 0.022 ^ 0.004 0.037 ^ 0.005 0.041 ^ 0.004 0.037 ^ 0.005 0.033 ^ 0.004 0.074 ^ 0.005
1580 ^ 275 2140 ^ 602 28.80 ^ 2.88 193.8 ^ 17.9 107.9 ^ 27.3 578.9 ^ 44.0 654 ^ 164 470.4 ^ 85.5 856 ^ 116 896.3 ^ 87.4 138.3 ^ 18.7 183.4 ^ 22.2 144.8 ^ 9.8
0.048 ^ 0.011 0.14 ^ 0.02 0.11 ^ 0.02
142.6 ^ 32.7 154.2 ^ 22.0 172.9 ^ 31.4
a
a
0.08 ^ 0.02 0.082 ^ 0.007 0.051 ^ 0.007 0.12 ^ 0.04 0.17 ^ 0.08 0.14 ^ 0.06 0.62 ^ 0.25 0.065 ^ 0.011 0.034 ^ 0.005 0.030 ^ 0.004 0.039 ^ 0.005 0.032 ^ 0.008 0.051 ^ 0.004 0.038 ^ 0.004
257.6 ^ 64.4 37.60 ^ 3.21 405.0 ^ 55.6 1138 ^ 379 41.80 ^ 19.7 69.42 ^ 29.8 72.16 ^ 29.1 437.4 ^ 74.0 96.7 ^ 14.2 356.6 ^ 47.6 406.2 ^ 52.1 247.7 ^ 61.9 71.85 ^ 5.63 214.0 ^ 22.5 (continued)
687
Backbone Flexibility of Human Growth Hormone
Table 1 Continued pH 7.0 Residue I179 V180 Q181 C182 R183 S184 V185 a b
21
pH 2.7 21
kexch (h )
Protection factor
kexch (h )
Protection factor
0.049 ^ 0.005
119270 ^ 12160
0.092 ^ 0.016
170978 ^ 29748
0.056 ^ 0.005 0.021 ^ 0.007 0.066 ^ 0.009 0.16 ^ 0.03 0.32 ^ 0.08 0.30 ^ 0.04 0.42 ^ 0.07
54.76 ^ 4.89 73.77 ^ 24.6 127.7 ^ 17.4 253.0 ^ 47.4 78.37 ^ 19.6 98.93 ^ 13.2 9.34 ^ 1.56
The NH signal was not observed in the first HSQC spectrum at pH 2.7. Exchange data are available only for the indole NH.
the calculated order parameter exceeds the maximal physically realistic value of 1.0. These unrealistically high values can be explained as an artifact caused by a dynamic protein aggregation, with a reversible interaction between the monomeric and a multimeric species. This transient association of the monomers with an aggregate has a profound effect on the R2 values and subsequently on the order parameters. Computer simulations show that if the monomer is “visiting” high-molecular weight aggregates for 1 –2% of the time, and the resulting relaxation parameters are analyzed with standard model-free protocols,27 the estimated generalized order parameters can artificially increase by as much as 10% (data not shown). Another consequence of aggregation is the apparent overall correlation time of 13.9 ns, which is higher than expected for a 22.5 kDa nearly globular protein. In agreement with these observations, DLS measurements confirm the presence of a small amount of oligomers (see above). Therefore, we are convinced that this aggregation phenomenon is the cause of the unusually high average order parameters obtained here.
pH 2.7 At pH 2.7, the rhGH molecule loosens up, as shown by the decrease of the average order parameter to 0.76(^ 0.20). This value is comparable to the lowest average order parameter reported for a folded protein ðS2average ¼ 0:76ð^0:14Þ found for glucose permease IIA domain;28 for an overview of the relaxation dynamics parameters of proteins see Goodman et al.25). Furthermore, at pH 2.7 the S2 values are less uniform along the polypeptide chain than at pH 7.0 (see Figure 4), and with considerably smaller order parameters in the interhelical regions (0.64(^ 0.21)) than in the helical regions (0.85(^ 0.12)) as shown in Figure 4(c)). This clearly shows that the low pH mainly destabilizes the unstructured periphery of the hormone. The homogeneous and relatively small decrease of S2 within the major helices suggests that the flexibility of the helical bundle increases slightly. The larger changes of S2 in the loops and at the termini show that these regions gain a considerable motional freedom. Finally, the movement of the loops becomes more independent of the overall
Figure 5. The 3D representation of rhGH (3HHR), colored according to the residue-specific NH protection factors given in Table 1. The results obtained at pH 7.0 and pH 2.7 are shown on the left-hand side and on the right-hand side molecules, respectively. The color-coding is explained in the Figure. Residues colored gray are either unassigned or have undefined amide proton exchange rates. Residues not observed in the first HSQC spectrum of an exchange series due to fast exchange with the solvent, are colored red. The Trp86 side-chain is shown to emphasize its location within the protein core.
688
tumbling, as reflected by the non-uniform pattern of R2 =R1 ratios (Figure 3). Such partial unfolding of the periphery alters the hydrodynamic characteristics of the protein and explains the fact that the tumbling time, tm ¼ 11:9 ns; of rhGH at pH 2.7 is higher than expected. The widest range of backbone dynamics is observed for the CD-loop (Figures 1 and 4), where the S2 parameters range from < 0.2 for Gly131 and Gly136 to < 0.8 for Tyr143. The region 131– 136 is one of the most flexible parts of the protein with residue-specific backbone order parameters all below 0.4, which is comparable to the termini. An unrestricted motion of this magnitude is close to what is typically found for unfolded proteins.29,30 The enhanced motional freedom observed in this region is also supported by the NH-exchange rates, as the amide protons of all the residues in this region are fast-exchanging. The high flexibility and the fast amide proton exchange observed for some of the residues in the CD-loop imply that this region is disordered in the acid state of rhGH. Interestingly, the unfolding does not involve the whole loop. Thus, the order parameters of the residues following the Gly136 gradually increase to a maximum of 0.81 and 0.75 for Tyr143 and Ser144, respectively. In addition, the amide proton of Ser144 is well protected against NH exchange. These observations suggest that Tyr143 and Ser144 are involved in a secondary and/or tertiary interactions. In support of this, the part of the molecule that is spatially close to residues 143 and 144 also has a reduced mobility. This holds in particular for the residues 56 – 60 of the AB-loop, which in the crystal structure of the receptor-bound hormone (3HHR) are located in close proximity to 143 –144. Thus, even though Glu56, Ser57, Ile58 and Thr60 belong to a loop, they are less mobile than the average backbone, with order parameters ranging from 0.83 to 0.91. Moreover, the Thr60 amide hydrogen has a high protection factor (see Table 1). Therefore, both regions, 56 –60 and 143 –144, are conformationally restricted. Possibly they are involved in a mutual interaction which leads to a formation of a specific structure, such as the “irregular antiparallel b-sheet” found in a similar location of IL-4 by Redfield and co-workers.16 The overall distribution of the backbone flexibility of rhGH at low pH is similar to that of IL-4.31 This protein also belongs to the four-helical bundle family, but unlike native rhGH and LIF (see above) the loops of the native interleukin-4 are not rigid. At low pH IL-4 populates a very flexible, but highly ordered molten globule state,31 in which the loops and the C-terminal end of helix C are further unfolded. The average order parameter of the acid conformation of IL-4 is similar to that of growth hormone, 0.75(^ 0.20) versus 0.76(^ 0.20), respectively. Thus, the average flexibility of the acid state of rhGH is as high as of the molten globule state of IL-4. Nevertheless, the value of the Gibbs free energy of rhGH at
Backbone Flexibility of Human Growth Hormone
low pH, < 6 –10 kcal/mol, measured by chemical denaturation8 suggests a native-like stability, which contradicts the concept of molten globule. Therefore, the acid state of rhGH represents a type of conformation, which can be placed somewhere between the native and the molten globule states. The comparison of human growth hormone and interleukin-4 thus reveals a delicate balance between stabilizing intramolecular interactions and backbone flexibility that underlies the physical origins of cooperativity. The Trp86-Asp169 hydrogen bond The presence of a hydrogen bond connecting the Od1 of Asp169 to the indole nitrogen of Trp86 ˚ ) was first suggested by Bewley and Li32 and (2.9 A later confirmed by crystallographic data.1 This hydrogen bond is formed at pH values higher than , 4.5. In acidic solution the aspartate sidechain is most probably protonated and hence can no longer participate in the formation of an interresidual hydrogen bond. A higher conformational freedom is therefore expected for the Trp sidechain at lower pH, than at neutral pH. In agreement with this, the R2 =R1 ratio of the indole nitrogen of Trp86 at pH 2.7 is higher than at pH 7.0. In fact, its magnitude at pH 2.7 is the highest of all the residues, clearly indicating the presence of exchange processes and a conformational heterogeneity of the indole ring. Thus, the relaxation data suggest that at low pH Trp86 does not have a unique conformation, corresponding to the breaking of the hydrogen bond to Asp169. Further support for the existence of the pH-dependent hydrogen bond between Trp86 and Asp169 is provided by the dramatic decrease of the protection of the indole NH proton of Trp86 with decreasing pH (see Table 1). The acid state of rhGH Far UV CD measurements show that the helical content of rhGH remains essentially unperturbed in the pH range from 8 to 2, whereas the near UV CD indicates8 that the amount of tertiary structure decreases with pH in a cooperative fashion around pH 4.3. The puzzling contradiction of these observations is that while near UV CD indicates a partial unfolding of the molecule, the far UV CD data are consistent with complete preservation of the structural integrity of rhGH. A pH-induced conformational transition is not unique for rhGH. Two other helical bundle proteins, IL-2 and IL-4, undergo similar structural changes at low pH.33,34 In the case of IL-2 the resulting structure has the characteristics of a classic molten globule,33 while the acid conformation of IL-4 is a “highly ordered molten globule”.31 As mentioned above the level of the backbone flexibility of the latter is remarkably similar to the acid state of rhGH. Nevertheless, the term “molten globule”, as defined by Ptitsyn,35 is not applicable
Backbone Flexibility of Human Growth Hormone
689
Figure 6. The 3D representation of rhGH (3HHR) colored according to the amide proton exchange rates at pH 2.7. The color-coding is explained in the Figure. Red color designates the residues for which the exchange rates were too fast to be measured (see the text) or the peak intensities were too low. The unassigned residues are shown in gray. The side-chains are shown only for the aromatic residues, tyrosine, phenylalanine and the Trp86.
to rhGH, because the low-pH conformation exhibits several features incompatible with the notion of molten globules,35,36 i.e., a high stability towards chemical denaturation,8 as well as thermal denaturation and a cooperative unfolding behavior.37,38 The transition observed by near UV CD in human growth hormone reflects a decrease in anisotropy of the environment around various aromatic residues and, hence, indicates an increased access of water to the aromatic residues with decreasing pH. Let us consider the solvent accessibility of the rhGH core as seen from “water’s perspective”. For this purpose the experimentally measured amide proton exchange rates are preferred over the protection factors because of their unambiguous correlation with the hydrophobicity of the protein core, that is, faster exchange corresponds to lower hydrophobicity. Figure 6 shows the rhGH molecule at pH 2.7 colored according to the amide proton exchange rates. As is apparent from this Figure, one end of the protein (the north end, see Figure 1) shows faster exchange than the other end (the south end). This asymmetric access of water to the two ends of the bundle reveals that at pH 2.7 the hydrophobicity of the north end is decreased compared to the south end. However, the north end of the molecule is also where the majority of the aromatic residues are located as shown in Figure 6; that is, the region with high density of aromatic residues coincides with the region with high water accessibility. This observation immediately explains the results obtained by near UV CD, revealing a correlation between the decrease of the CD intensity and the location of fast-exchanging amide protons. Thus, the weakening of the tertiary interaction network involves only half of the molecule. The other half,
the south end, stays well protected from the NH exchange, which implies that the tertiary interactions between the four helices remain intact, thereby facilitating the conservation of the overall four-helical bundle fold of the protein. In other words, due to the remaining interactions at the south end of rhGH, the enhanced exchange at the north end does not lead to the disassembly of the whole bundle. The present example shows that the structural changes observed by near UV CD might not necessarily reflect the global unfolding of a protein. Rather they can be an indication of an increased local flexibility around aromatic chromophores. A similar, asymmetric distribution of the exchange-protected residues within a protein was found for another four-helical bundle cytokine, LIF.26 In LIF the protection factors increased slightly towards the N-terminal ends of the A and B helices, and the C-terminal ends of the C and D helices, that is, at the south end of the molecule. For the acid state of rhGH the difference between the protection factors of the two ends of the bundle is more pronounced than for LIF, suggesting that the amplitudes of the slow-timescale (ms – seconds) motions of the two halves are very different (see Figures 5 and 6). At the same time, the pico- to nanosecond dynamics of the four major helices in rhGH is uniform throughout the entire bundle, although the order parameters and thereby the rigidity are slightly lower at pH 2.7 than at pH 7.0. Thus, the combination of the uniform rigidity of helices with non-uniform NH exchange is a distinctive feature of the acidic state of rhGH. It shows that the conformation at low pH combines the ability to preserve the structure of the individual helices with the loosening of the tertiary interactions between them. The observed
690
Figure 7. The three-dimensional structure of rhGH (3HHR) showing the distribution of assigned and unassigned residues. The residues for which the backbone assignment was obtained at either pH 7.0 or pH 2.7 are colored gray. The residues with no backbone assignment at either pH are colored red (these are the residues 1, 2, 21, 22, 30, 37, 38, 48, 54, 55, 59, 61, 64 – 76, 86 – 89, 110, 145, 146, 164– 169). The Trp86 backbone is also colored red because assignment was obtained only for the indole NH. The side-chain of Trp86 is shown in black. The aromatic side-chains of Phe54 and Phe166 are also shown and are located in front of and behind Trp86, respectively.
asymmetric disruption of the tertiary packing appears as fraying or divergence of the helices from each other at the north end of the protein (see Figure 1). Backbone dynamics and biological activity of rhGH It is interesting to compare the dynamic properties of the individual residues of rhGH and the role they play in the biological activity of the protein. At pH 7.0, we found no correlation between the receptor binding epitope and the residuespecific internal motions of the molecule. Possibly the uniform rigidity of the protein obscures the difference in dynamics between functionally important and unimportant regions. However, the correlation becomes apparent upon destabilization of the protein at lower pH: the more stable south end was also shown to be important for the receptor binding.1 Using alanine scanning technique, de Vos and colleagues found that the first binding event between the hormone and its receptor is determined by the patch consisting of the three segments all located at the south end: the loop between residues 54 – 74, the C-terminal half of helix D and, to a lesser extent, the N-terminal region of helix A. Thus, the region of the bundle that contains the residues that play a significant role in the first binding event between the hormone
Backbone Flexibility of Human Growth Hormone
and its receptor remains well protected from the solvent, as indicated by the slow NH exchange observed in this region (see Figure 5, pH 2.7). This work also has an important implication for further investigations of the structure of rhGH in solution. Figure 7 shows that the unassigned residues are primarily clustered in two regions. One of these (residues 64 –76, cluster 1 in Figure 7) contains the mini-helix A00 , located at the first loop, and the beginning of the helix B. This particular region differs in the crystal structures of the receptor-bound (3HHR) and free (1HGU) rhGH by the presence of the mini-helix A00 . This helix presumably emerges as a result of the binding to the receptor. We believe that the difficulty in assigning the region may arise from conformational heterogeneity of this part of the polypeptide chain. It is possible that the intrinsic helical propensity of the non-bound protein leads to transient helix formation, which would explain the broadening of NMR signals in this region. The other unassigned cluster (residues 21, 22, 54, 55, 87 –89, 110, 145, 146 and 164 –169, cluster 2 in Figure 7) involves residues from all four helices. This cluster is centered around Trp86 and the two phenylalanine residues Phe54 and Phe166, dividing the four-helical bundle in two approximately even halves. Whether the lack of signals for this region arises from the increased local flexibility of the backbone needs further investigation. However, if this is the case, the enhanced flexibility of the central section of the molecule could have a biologically important function, serving as a hinge that provides rhGH with a transverse conformational freedom. Since the binding epitope spans most of the length of the helical bundle, the molecule may need an enhanced plasticity to adapt to the curvature of the receptor binding site. The same plasticity may play an important role in the ability of rhGH to interact with high affinity with both its own receptor and the prolactin receptor.9
Conclusions We have characterized the dynamics of human growth hormone by 15N relaxation and amide proton exchange. At pH 7.0, the rhGH molecule is rigid on both the ps –ns and the ms– seconds timescale. This also holds for the long inter-helical loops that are highly ordered even though without a regular secondary structure. In contrast, at pH 2.7 these loops are more unstructured and/or highly mobile, which appears as a progressive dissolution of the outer layer of the molecule. However, the core of the protein formed by the four major helices, although more flexible than at pH 7.0, still preserves its structural integrity. These conformational changes are in effect a partial unfolding, even though there is no detectable change in the secondary structure content. Thus, the results here show that the acidic state of
691
Backbone Flexibility of Human Growth Hormone
rhGH consists of the structurally conserved, but dynamically more flexible helical core surrounded by more mobile and less structured loops. The partial unfolding of the loops exposes the hydrophobic surface of the second and the fourth helices, resulting in an increased hydrophobicity of the acid state as detected previously by Nile red binding.8 It is important to notice that the loops do not unfold completely and do not expose the buried Trp86, as indicated by the absence of a red shift in the tryptophan fluorescence emission spectrum and the identical quantum yields at the two pH values.8 The dynamics data also indicate that only some segments of the loops unfold, while others stay folded. Thus, low backbone flexibility observed around the residues 56 –60 and 143– 144 suggests that these regions participate in the formation of a local secondary/tertiary structure. The pH-induced increase of the backbone flexibility of rhGH is accompanied by conformational destabilization of the molecule from < 11 –15 kcal/ mol at neutral pH to < 6 –10 kcal/mol at low pH.8 However, the native-like stability of the acid state of rhGH, which is still native-like, and its cooperative unfolding behavior37,38 suggest that the molecule maintains its globular conformation. It therefore appears from our work that under certain conditions, a protein can tolerate a considerable increase in flexibility of the backbone along with the increased amide hydrogen exchange between its core and the solvent without a considerable loss in its structure and stability.
Materials and Methods Materials Unlabeled, 15N-labeled, and [15N,13C]-double-labeled recombinant rhGH was kindly provided by Novo Nordisk. The [15N,13C]-double-labeled rhGH used for sequential assignment was dissolved in 90% H2O þ 10% 2 H2O. The pH of the sample was adjusted to either 2.65– 2.70 or to 6.8 –7.0 with HCl or NaOH, respectively. For the NH-exchange experiments at low pH, the lyophilized rhGH was dissolved in 99.996% 2H2O (Isotec Inc.) and the pH was adjusted with 2HCl to achieve the final pH of 2.7. For the similar experiment at pH 7.0 the concentrated protein solution at pH 7.0 was diluted with 99.996% 2H2O to the final experimental volume (see 1H exchange measurements section). For the backbone dynamics relaxation experiments the protein was dissolved in 10 mM phosphate or acetate buffers with the pH adjusted to 7.0 or 2.7. The final solution of rhGH was between 0.7 and 1 mM and contained 10% 2H2O. Dynamic light scattering Dynamic light scattering (DLS) was applied for estimation of the apparent hydrodynamic radius of hGH, Rh;app ; using a DynaPro99 with temperature-controlled microsampler (ProteinSolutions). Measurements were made at pH 2.7 and pH 6.0 at 25 8C for hGH concentrations between 0.1 mM and 1.1 mM. Before the measurements, samples were filtered through Whatman Anodisc 0.10 mm (Whatman 6809-7013) to remove inter-
fering dust particles. Approximately 100 data points were obtained during 30 minutes of analysis. Data were evaluated using the Dynamics V5.25 software (ProteinSolutions). After measurement, noisy data points were filtered using three standard criteria (amplitude . 0.1, baseline .1.010, and sum of squares # 50), and Rh;app was estimated as the mean value of the remaining data. Resonance assignment Heteronuclear double and triple-resonance spectra for the assignment of rhGH at pH 2.7 were recorded on a modified Bruker AM500 NMR spectrometer equipped with a 5 mm [1H,15N,13C] triple-resonance probe and two extra radio-frequency channels39 (four channels in all). The protein concentration was 0.9 mM for all uniformly labeled samples and no buffer substances were added. Double and triple-resonance spectra recorded on uniformly [15N,13C]-double-labeled samples in 99.99% 2H2O include [1H,13C]-HCCH-COSY, [1H,13C]HCCH-TOCSY, HCACO and HCA(CO)N spectra.40 Double-resonance spectra recorded on uniformly (15N)labeled samples in 90% H2O/10% 2H2O include [1H,15N]-HSQC, [1H,15N]-NOESY-HMQC, [1H,15N]HMQC-NOESY-HMQC and [1H,15N]-TOCSY-HMQC.11,41 Triple-resonance spectra recorded on [15N,13C]-doublelabeled samples in 90% H2O/10% 2H2O include HNCO, HN(CO)CA, HNCA, HN(CA)CO, and CBCA(CO)NH spectra.12,42 – 44 To further aid the resonance assignment, [1H,15N]-HSQC spectra were recorded on a specifically [15N]leucine-labeled sample.45 Heteronuclear double and triple-resonance spectra for the assignment of rhGH at pH 6.8 were recorded on 500 MHz and 750 MHz Varian UNITY Inova spectrometers equipped with 5 mm [1H,15N,13C] triple-resonance pulsed field gradient probes. All spectra were recorded on a 0.7 mM uniformly [15N,13C]-double-labeled rhGH sample in 90% H2O/10% 2H2O with 20 mM histidine, 1 mM NaF and 0.5 mM EDTA. The double and tripleresonance experiments recorded at pH 6.8 include [1H,15N]-HSQC, [1H,15N]-NOESY-HSQC, [1H,15N]-HMQCNOESY-HSQC, HNCA, HN(CO)CA, CBCA(CO)NH, HA(CA)NH, H(CACO)NH as described by Shang et al.46 These experiments were modified to include gradientselected sensitivity-enhancement and to avoid saturation of the water signal.44,47 – 49 The assignment at pH 6.8 – 7.0 was started using the assignment at pH 2.7 combined with a titration series.50 However, because of large changes in chemical shifts and protein precipitation during the titration, only a few peaks could be assigned using this approach, and the majority of the assignments were subsequently made using HNCA, HN(CO)CA, HA(CA)NH, H(CACO)NH and 3D NOESY spectra. Chemical shift indices 1 H, 13C and 15N chemical shifts were all referenced relative to water19 at 4.69 ppm at 32 8C or to TSP at 0.00 ppm. The 1Ha and 13Ca secondary chemical shift indices were calculated by subtracting the appropriate reference random-coil chemical shift values.19
1
H exchange measurements
The amide 1H/2H exchange was monitored by series of gradient-enhanced two-dimensional heteronuclear single-quantum coherence correlation (HSQC) spectra47
692
recorded at 32 8C on a Varian Unity Inova 500 MHz spectrometer with a pH value of either 2.7 or 7.0. A total of 1024 and 160 complex points were recorded in the t2 and t1 dimensions and with spectral widths of 10,000 Hz and 2000 Hz, respectively. At pH 2.7 a series of 117 spectra were recorded to monitor the amide proton exchange after dissolution of dry rhGH in 2H2O. The first 72 spectra were recorded with four scans per t1 increment corresponding to a duration of 17.32 minutes per spectrum, and the subsequent 45 spectra were recorded with 28 scans per t1 increment and a corresponding duration of 2.02 hours per spectrum. Prior to Fourier transformation, the recorded FIDs were zerofilled in both dimensions and processed with a Lorentzian-to-Gaussian resolution-enhancement filter in t2 and a shifted squared sine-bell filter in t1 : The residual water resonance was attenuated by deconvolution.51 The exchange rates were determined by a non-linear leastsquares fitting of the peak heights for the individual residues in the HSQC spectra to an exponential decay function, hðtÞ ¼ h0 expð2kobs tÞ; where h0 and kobs are the signal height at t ¼ 0 and the exchange rate constant, respectively. At pH 7.0 the experimental set up had to be redesigned, because dissolution of lyophilized rhGH directly in 2H20 at this pH resulted in 15N-correlated NMR spectra with almost no cross-peaks. Therefore, rhGH was initially dissolved in 200 ml of H2O-based phosphate buffer and then diluted to the final experimental volume of 600 ml with 2H2O. In this experimental setup, the final concentration of rhGH was defined by the solubility of the protein in the first step. Consequently the NMR spectra had lower intensity. Since rhGH has many weak and broad signals, the decrease in the intensity results in fewer detectable resonances, low data output and poor data quality. The exchange experiment was conducted as a series of three consecutive experiments, each containing either eight or four spectra. The first eight spectra were recorded with 16 scans per t1 increment corresponding to a duration of 145.4 minutes per spectrum. The following eight spectra were recorded with 48 scans per t1 increment and a corresponding duration of 7.27 hours per spectrum, and the final experiment had 144 scans and a duration of 21.81 hours. The analysis of the data was performed as described for pH 2.7. Protection factors were calculated as described by Bai et al.52
Relaxation measurements Standard HSQC experiments were used to measure the 15N longitudinal relaxation rate, R1 , the 15N transverse relaxation rate, R2 , and the steady-state [1H– 15N] NOE.53 All experiments were recorded with the gradient-enhanced and sensitivity-enhanced pulse schemes shown in Figure 3 of Farrow et al.54 15N R1 ; R2 and NOE relaxation series were recorded on 0.9 mM rhGH samples at a temperature of 305 K with pH adjusted to either 2.65 or 7.0. All relaxation experiments were recorded at 11.7 T with a 15N spectral width of 2 kHz and a 1H spectral width of 10 kHz. The 15N and 1H transmitters were positioned at 119 ppm and on the water resonance, respectively. Each individual 2D spectrum in a relaxation series was recorded as 160 complex t1 values times 1024 complex t2 values. F1 Quadrature was obtained by pulsed-field gradient coherence selection.47,54 The 15N pulse field strength was 5.2 kHz in the R1 and [1H – 15N] NOE experiments, whereas it was lowered to
Backbone Flexibility of Human Growth Hormone
3.4 kHz in the R2 experiments to reduce power dissipation in the sample during the long CPMG spin-lock periods. 15N was decoupled during data acquisition with the improved WALTZ-16 decoupling scheme55 with a field strength of 1 kHz. In all experiments, the 1H pulse field strength was 25 kHz for hard pulses. Pre-saturation of 1H magnetization in the steady-state [1H – 15N] NOE experiments were carried out with a series 12 kHz 1208 1H pulses separated by 5 ms delays.56 R1 relaxation rates were determined from a series of 12 spectra recorded with relaxation delays of 0.01005( £ 2), 0.06031, 0.13068, 0.23120, 0.34178, 0.48251( £ 2), 0.74387, 1.06554( £ 2) and 1.50785 seconds, where ( £ 2) indicates duplicate spectra. Dipole – dipole/CSA cross-correlation artifacts were suppressed by the application of 1H 1808 pulses spaced 5 ms apart throughout the relaxation delay.48,57,58 For the determination of R2 relaxation rates, ten spectra were recorded with relaxation delays of 0.0, 0.016832( £ 2), 0.033664, 0.050496, 0.067328( £ 2), 0.08416, 0.117824, and 0.151488 seconds. The relaxation delay was implemented as a train of CPMG sequences59 with a delay of 0.9 ms between subsequent 15N 1808 pulses and with dipole – dipole/CSA cross-correlation suppression implemented as in Figure 10(b) of Farrow et al.54 For both R1 and R2 experiments, the pre-scan delay was 1.5 seconds. R1 and R2 experiments were recorded with a pre-scan delay of 1.5 seconds and 32 transients for each t1 value. The corresponding total recording times were 21 hours and 18 hours, respectively. Steady-state [1H – 15N] NOEs were determined from 15N-excited HSQC spectra recorded with and without 1H presaturation. Spectra with presaturation were recorded with a pre-scan delay of two seconds followed by three seconds of presaturation and spectra without presaturation were recorded with a pre-scan delay of five seconds. [1H– 15N] NOE experiments were recorded with 64 transients for each t1 value, and a corresponding recording time of 50 hours.
Model-free analysis The protocol used for analysis of the relaxation data is analogous to that reported by Kristensen et al.60 and is based on the procedure previously described by Palmer et al.27 The overall correlation time was calculated from the 15% trimmed average local correlation times and a corresponding maximal NOE was calculated.53 Subsequently, the overall correlation time, tm ¼ ð2Dxx þ 2Dyy þ 2Dzz Þ21 ; the rotational diffusion anisotropy, h ¼ 2Dzz =ðDxx þ Dyy Þ; and the direction of the principal component of the rotational diffusion tensor in the molecular frame were simultaneously estimated by fitting experimental R2 =R1 ratios to values calculated for an axially symmetric diffusor.61,62 Residues with NOE values less than 75% of the maximal NOE, and residues for which 21 21 21 ðkR21 2 l 2 R2 Þ=kR2 l 2 ðkR1 l 2 R1 Þ=kR1 l . 1:5 £ SD were excluded from the calculation to avoid interference from residues affected by substantial internal dynamics and conformational exchange contributions to R2 ; respectively.62 The directions of N – H bond vectors relative to the molecular frame were calculated from the crystal structure of rhGH – receptor complex (PDB entry 3HHR) after calculating the amide hydrogen coordinates. Model-free parameters were calculated on a per residue basis by a grid search followed by a non-linear least-squares minimization procedure while keeping the overall rotational diffusion parameters constant.
693
Backbone Flexibility of Human Growth Hormone
Model-free parameter uncertainties were estimated by Metropolis Monte Carlo simulations. By analogy with the model-selection method introduced by Mandel,27,63 one of five models was selected for each site. The procedure uses Monte Carlo simulations combined with x2 and F-statistics testing for selecting the most appropriate model. Critical values of a ¼ 0:05 and a ¼ 0:2 were used in the x2 and F-statistics testing, respectively. The regression parameters of the five models27,63 are: (model 1: S2 ), (model 2: S2 ; ti ), (model 3: S2 ; Rex ), (model 4: S2 ; ti ; Rex ), and (model 5: S2 ; S2f ; ti ), where S2 is the squared order parameter, ti is the time scale of the internal motion, S2f is the squared order parameter of an internal motion on a timescale much faster than ti ; and Rex is the exchange contribution to the transverse relaxation rate. At pH 2.7 a total of 35, 22, 10, 18 and 52 NH sites were fitted with relaxation model 1, 2, 3, 4 and 5, respectively. At pH 7.0 a total of 25, 29, 17, 28 and one NH sites were fitted with relaxation model 1, 2, 3, 4 and 5, respectively. Among these 15 NH sites could not be fitted with physically reasonable squared order parameters and were re-analyzed allowing the order parameter to take values above 1.0. For all these NH sites, the estimated order parameters were above 1.0, but within two standard deviations and were therefore included in the data set.
5.
6. 7. 8.
9.
10.
Accession numbers The chemical shift and relaxation data obtained under salt-free conditions at pH 2.7 and pH 6.8 – 7.0 have been deposited in the BioMagResBank (http:// www.bmrb.wisc.edu) BMRB accession number 4689.
Acknowledgments The 750 MHz spectra were obtained at The Danish Instrument Center for NMR Spectroscopy of Biological Macromolecules. We are grateful to Dr Lixin Ma, Ms Else Philipp and Dr Jens Duus for technical assistance. Consulting assistance from Dr Rolf Tschudin in rebuilding the Bruker AM500 NMR spectrometer to a fourchannel system is also gratefully acknowledged. This work was financially supported by The Ministry of Industry (J. No. 85886), the Danish Natural Science Research Council (J. Nos. 9400351 and 9801801), Direktør Ib Henriksens Fond, Carlsbergfondet and Novo Nordisk Fond. Finally we acknowledge a postdoctoral stipend to M.R.K. from Novo Nordisk A/S.
11.
12.
13.
14.
15.
References 1. de Vos, A. M., Ultsch, M. & Kossiakoff, A. A. (1992). Human growth hormone and extracellular domain of its receptor: crystal structure of the complex. Science, 255, 306– 312. 2. Somers, W., Ultsch, M., de Vos, A. M. & Kossiakoff, A. A. (1994). The X-ray structure of a growth hormone prolactine receptor complex. Nature, 372, 478–481. 3. Atwell, S., Ultsch, M., de Vos, A. M. & Wells, J. A. (1997). Structural plasticity in a remodeled protein– protein interface. Science, 278, 1125– 1128. 4. Clackson, T., Ultsch, M. H., Wells, J. A. & de Vos, A. M. (1998). Structural and functional analysis of
16.
17.
18.
the 1:1 growth hormone:receptor complex reveals the molecular basis for receptor affinity. J. Mol. Biol. 277, 1111 – 1128. Ultsch, M., Somers, W., Kossiakoff, A. A. & de Vos, A. M. (1994). The crystal structure of affinity˚ resolution. matured human growth hormone at 2A J. Mol. Biol. 236, 286– 299. Kay, L. E. (1998). Protein dynamics from NMR. Nature Struc. Biol. NMR Suppl., 513– 617. Wetlaufer, (1984). The Protein Folding Problem (Wetlaufer, D. B., ed.), pp. 29 – 46, Westview, Boulder, CO. DeFelippis, M. R., Kilcomons, M. A., Lents, M. P., Youngman, K. M. & Havel, H. A. (1995). Acid stabilization of human growth hormone equilibrium folding intermediates. Biochim. Biophys. Acta, 1247, 35 – 45. Boutin, J. M., Edery, M., Shirota, M., Jolicoeur, C., Lesueur, L., Ali, S., Guould, D., Djiane, J. & Kelly, P. A. (1989). Identification of a cDNA encoding a long form of prolactin receptor in human hepatoma and breast cancer cells. Mol. Endocrinol. 3, 1455– 1461. Bru¨schweiler, R. (1994). Connections between NMR measurements and theoretical models of structural dynamics of biopolymers in solution. In Nuclear Magnetic Resonance Probes of Molecular Dynamics (Tycko, R., ed.), pp. 301– 334, Kluwer Academic Publishers, Dordrecht, The Netherlands. Marion, D., Driscoll, P. C., Kay, L. E., Wingfield, P. T., Bax, A., Gronenborn, A. M. & Clore, G. M. (1989). Overcoming the overlap problem in the assignment of 1H NMR spectra of larger proteins by use of 3-dimensional heteronuclear 1H– 15N Hartmann – Hahn multiple quantum coherence and nuclear Overhauser multiple quantum coherence spectroscopy—application to interleukin-1b. Biochemistry, 28, 6150– 6156. Ikura, M., Kay, L. E. & Bax, A. (1990). A novel approach for sequential assignment of 1H, 13C, and 15 N spectra of larger proteins—heteronuclear tripleresonance 3-dimensional NMR-spectroscopy—application to calmodulin. Biochemistry, 29, 4659– 4667. Grzesiek, S. & Bax, A. (1992). Correlating backbone amide and side-chain resonances in larger proteins by multiple relayed triple resonance NMR. J. Am. Chem. Soc. 114, 6291– 6293. Cheetham, J. C., Smith, D. M., Aoki, K. H., Stevenson, J. L., Hoeffel, T. J., Syed, R. S. et al. (1998). NMR structure of human erythropoietin and a comparison with its receptor bound conformation. Nature Struct. Biol. 5, 861– 866. Xu, G.-Y., Yu, H.-A., Hong, J., Stahl, M., McDonagh, T., Kay, L. E. & Cumming, D. A. (1997). Solution structure of recombinant human interleukin-6. J. Mol. Biol. 268, 468–481. Redfield, C., Boyd, J., Smith, L. J., Smith, A. G. & Dobson, C. M. (1992). Loop mobility in a four-helixbundle protein: 15N NMR relaxation measurements on human interleukin-4. Biochemistry, 31, 10431 – 10437. Zink, T., Ross, A., Lu¨ers, K., Cieslar, C., Rudolph, R. & Holak, T. A. (1994). Structure and dynamics of the human granulocyte colony-stimulating factor determined by NMR spectroscopy. Loop mobility in a four-helix bundle protein. Biochemistry, 33, 8453 –8463. Purvis, D. H. & Mabbutt, B. C. (1997). Solution dynamics and secondary structure of murine leukemia
694
19. 20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
inhibitory factor: a four-helix cytokine with rigid CD loop. Biochemistry, 36, 10146– 10154. Wishard, D. S. & Sykes, B. D. (1994). Chemical shifts as a tool for structure determination. Methods Enzymol. 239, 363– 392. Cantor, C. R. & Schimmel, P. R. (1980). Biophysical Chemistry, Part II: Techniques for the Study of Biological Structure and Function, pp. 461, W.H. Freeman and Co., San Francisco, CA. Feng, Y., Klein, B. K. & McWherter, A. C. (1996). Three-dimensional solution structure and backbone dynamics of a variant of human interleukin-3. J. Mol. Biol. 259, 524– 541. Abildgaard, F., Munk Jørgensen, A. M., Led, J. J., Christensen, T., Jensen, E. B., Junker, F. & Dalbøge, H. (1992). Characterization of tertiary interactions in a folded protein by NMR methods: studies of pH-induced structural changes in human growth hormone. Biochemistry, 31, 8587– 8596. Lipari, G. & Szabo, A. (1982). A model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 1. Theory and range of validity. J. Am. Chem. Soc. 104, 4545– 4549. Lipari, G. & Szabo, A. (1982). A model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 2. Analysis of experimental results. J. Am. Chem. Soc. 104, 4549– 4570. Goodman, J. L., Pagel, M. D. & Stone, M. (2000). Relationships between protein structure and dynamics from a database of NMR-derived backbone order parameters. J. Mol. Biol. 295, 963– 978. Yao, S., Smith, D. K., Hinds, M. G., Zhang, J.-G., Nicola, N. A. & Norton, R. S. (2000). Backbone dynamics measurements on leukemia inhibitory factor, a rigid four-helical bundle cytokine. Protein Sci. 9, 671– 682. Mandel, A. M., Akke, M. & Palmer, A. G. (1995). Backbone dynamics of Escherichia coli ribonuclease H: correlations with structure and function in an active enzyme. J. Mol. Biol. 246, 144– 163. Stone, M. L., Fairbrother, W. L., Palmer, A. G., III, Reizer, J., Saier, M. H. & Wright, P. E. (1992). Backbone dynamics of the Bacillus subtilis glucose permease IIA domain determined from 15N NMR relaxation measurements. Biochemistry, 31, 4394– 4406. Farrow, N. A., Zhang, O., Forman-Kay, J. D. & Kay, L. E. (1997). Characterization of backbone dynamics of folded and denatured states of an SH3 domain. Biochemistry, 36, 2390– 2402. Buck, M., Schwalbe, H. & Dobson, C. M. (1996). Main-chain dynamics of a partially folded protein: 15 N NMR relaxation measurements of hen egg white lysozyme denatured in trifluoroethanol. J. Mol. Biol. 257, 669– 683. Redfield, C., Smith, R. A. G. & Dobson, C. M. (1994). Structural characterization of a highly-ordered molten globule at low pH. Nature Struct. Biol. 1, 23 –29. Bewley, T. A. & Li, C. H. (1984). Conformational comparison of human pituitary growth hormone and human chorionic somatomammotropin (human placental lactogen) by second-order absorption spectroscopy. Arch. Biochem. Biophys. 233, 219–227. Dryden, D. & Weir, M. P. (1991). Evidence for an acid-induced molten globule state in interleukin-2; a fluorescence and circular dichroism study. Biochim. Biophys. Acta, 1078, 94 – 100.
Backbone Flexibility of Human Growth Hormone
34. Windsor, W. T., Syto, R., Le, H. V. & Trotta, P. P. (1991). Analysis of the conformation and stability of Escherichia coli derived recombinant human interleukin-4 by circular dichroism. Biochemistry, 30, 1259– 1264. 35. Ptitsyn, O. B. (1987). Protein folding: hypotheses and experiments. J. Protein Chem. 6, 273– 293. 36. Haynie, D. T. & Freire, E. (1993). Structural energetics of the molten globule state. Proteins, 16, 115– 140. 37. Kasimova, M. R., Milstein, S. J. & Freire, E. (1998). The conformational equilibrium of human growth hormone. J. Mol. Biol. 277, 409– 418. 38. Kasimova, M. R. (1999). Conformational equilibrium of human growth hormone: a mechanism for the self-association of the intermediate state, pp. 34 – 46. PhD thesis, The Johns Hopkins University. 39. Kay, L. E., Ikura, M., Tschudin, R. & Bax, A. (1990). 3-Dimensional triple-resonance NMR-spectroscopy of isotopically enriched proteins. J. Magn. Reson. 89, 496– 514. 40. Bax, A., Clore, G. M., Driscoll, P. C., Gronenborn, A. M., Ikura, M. & Kay, L. E. (1990). Practical aspects of proton carbon carbon proton 3-dimensional correlation spectroscopy of 13C-labeled proteins. J. Magn. Reson. 87, 620– 627. 41. Ikura, M., Bax, A., Clore, G. M. & Gronenborn (1990). Detection of nuclear Overhauser effects between degenerate amide proton resonances by heteronuclear 3-dimensional NMR-spectroscopy. J. Am. Chem. Soc. 112, 9020– 9022. 42. Grzesiek, S. & Bax, A. (1992). Improved 3D tripleresonance NMR techniques applied to a 31-KDa protein. J. Magn. Res. 96, 432– 440. 43. Clubb, R. T., Thanabal, V. & Wagner, G. (1992). A constant-time 3-dimensional triple-resonance pulse scheme to correlate intraresidue 1H, 15N, and 13C chemical shifts in 15N– 13C-labeled proteins. J. Magn. Reson. 97, 213– 217. 44. Muhandiram, D. R. & Kay, L. E. (1994). Gradientenhanced triple-resonance 3-dimensional NMR experiments with improved sensitivity. J. Magn. Reson. ser. B, 103, 203– 216. 45. Christensen, T., Petersen, J., Theisen, C. F., Bjerregaard, K., Kristensen, S. M. & Led, J. J. (1993). Specific 15N labeling of leucine residues in human growth hormone. Acta Chem. Scand. 47, 990– 993. 46. Shang, Z., Swapna, V. T., Rios, C. B. & Montelione, G. T. (1997). Sensitivity-enhancement of triple-resonance protein NMR spectra by proton evolution of multiple-quantum coherences using a simultaneous 1 H and 15N constant-time evolution period. J. Am. Chem. Soc. 119, 9274– 9278. 47. Kay, L. E., Keifer, P. & Saarinen, T. (1992). Pure absorption gradient enhanced heteronuclear single quantum correlation spectroscopy with improved sensitivity. J. Am. Chem. Soc. 114, 10663– 10665. 48. Kay, L. E., Nicholson, L. K., Delaglio, F., Bax, A. & Torchia, D. A. (1992). Pulse sequences for removal of the effects of cross-correlation between dipolar and chemical-shift anisotropy relaxation mechanism on the measurement of heteronuclear t1 and t2 values in proteins. J. Magn. Res. 97, 359–375. 49. Stonehouse, J., Clowes, R. T., Shaw, G. L., Keeler, J. & Laue, E. D. (1995). Minimisation of sensitivity losses due to the use of gradient pulses in triple-resonance NMR of proteins. J. Magn. Reson. 5, 226– 232. 50. Jørgensen, Th. B. (1995). Application of 2D and 3D NMR spectroscopy to protein studies (in Danish). MS thesis, University of Copenhagen.
Backbone Flexibility of Human Growth Hormone
51. Marion, D., Ikura, M. & Bax, A. (1989). Improved solvent suppression in one-dimensional and twodimensional NMR-spectra by convolution of timedomain data. J. Magn. Reson. 84, 425– 430. 52. Bai, Y., Milne, J. S., Mayne, L. & Englander, S. W. (1993). Primary structure effects on peptide group hydrogen exchange. Proteins: Struct. Funct. Genet. 17, 75– 86. 53. Kay, L. E., Torchia, D. A. & Bax, A. (1989). Backbone dynamics of proteins as studied by 15N inverse detected heteronuclear NMR-spectroscop—application to staphylococcal nuclease. Biochemistry, 28, 8972–8979. 54. Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal, S. M., Kay, C. M., Gish, G. et al. (1994). Backbone dynamics of a free and a phosphopeptide-complexed SRC homology-2 domain studied by 15N NMR relaxation. Biochemistry, 33, 5984– 6003. 55. Shaka, A. J., Keeler, J., Frenkiel, T. & Freeman, R. (1983). An improved sequence for broadband decopling: WALZ-16. J. Magn. Reson. 52, 335– 338. 56. Markley, J. L., Hornsley, J. W. & Klein, M. P. (1971). Spin-lattice relaxation measurements in slowly relaxing complex spectra. J. Chem. Phys. 55, 3604– 3605. 57. Boyd, J., Hommel, U. & Campbell, I. D. (1990). Influence of cross-correlation between dipolar and anisotropic chemical-shift relaxation mechanisms upon longitudinal relaxation rates of 15N in macromolecules. Chem. Phys. Lett. 175, 477– 482.
695
58. Palmer, A. G., Skelton, N. J., Chazin, W. J., Write, P. E. & Rance, M. (1992). Suppression of the effects of cross-correlation between dipolar and anisotropic chemical-shift relaxation mechanisms in the measurement of spin– spin relaxation rates. Mol. Phys. 75, 699– 711. 59. Meiboom, S. & Gill, D. (1958). Modified spin-echo method for measuring nuclear spin relaxation times. Rev. Scient. Instrum. 29, 688– 691. 60. Kristensen, S. M., Siegal, G., Sankar, A. & Driscoll, P. C. (2000). Backbone dynamics of the C-terminal SH2 domain of the p85a subunit of phosphoinositide 3-kinase: effect of phosphotyrosine– peptide binding and characterization of slow conformational exchange processes. J. Mol. Biol. 299, 771–788. 61. Tjandra, N., Feller, S. E., Pastor, R. W. & Bax, A. (1995). Rotational diffusion anisotropy of human ubiquitin from 15N NMR relaxation. J. Am. Chem. Soc. 117, 12562– 12566. 62. Tjandra, N., Wingfield, P., Stahl, S. & Bax, A. (1996). Anisotropic rotational diffusion of perdeuterated HIV protease from 15N NMR relaxation measurements at 2 magnetic fields. J. Biomol. NMR, 8, 273 –284. 63. Mandel, A. M., Akke, M. & Palmer, A. G. (1996). Dynamics of ribonuclease H: temperature dependence of motions on multiple time scales. Biochemistry, 35, 16009– 16023.
Edited by P. E. Wright (Received 21 November 2001; received in revised form 14 February 2002; accepted 19 February 2002)