Non-sequence-specific interactions can account for the compaction of proteins unfolded under “native” conditions

Abstract

Proteins unfolded by high concentrations of chemical denaturants adopt expanded, largely structure-free ensembles of conformations well approximated as random coils. In contrast, globular proteins unfolded under less denaturing conditions (via mutations, or transiently unfolded after a rapid jump to native conditions) and molten globules (arising due to mutations or co-solvents) are often compact. Here we explore the origins of this compaction using a truncated, equilibrium-unfolded variant of the 57-residue FynSH3 domain. As monitored by far-UV circular dichroism, NMR spectroscopy and hydrogen exchange kinetics, a four-residue carboxy-terminal deletion of FynSH3 (CΔ4) appears to be largely unfolded even in the absence of denaturant. Nevertheless, CΔ4 is quite compact under these conditions, with a hydrodynamic radius only slightly larger than that of the native protein. In order to understand the origins of this molten globule-like compaction we have characterized a random sequence polypeptide of identical amino acid composition to CΔ4. Notably, we find that the hydrodynamic radius of this random sequence polypeptide also approaches that of the native protein. Thus, while native-like interactions may contribute to the formation of compact “unfolded” states, it appears that non-sequence-specific monomer-monomer interactions can also account for the dramatic compaction observed for molten globules and the “physiological” unfolded state.

Introduction

Due to the relative ease with which they are studied, chemically denatured proteins have remained the “gold standard” unfolded state for more than 60 years (1–3). Spectroscopic and small angle scattering studies of such proteins have generally upheld the longstanding view that the chemically-denatured state occupies an ensemble of highly expanded, effectively random coil conformations (4, 5). Consequently, the random-coil model remains the starting point for many theoretical models of the folding process (e.g., 6, 7).

In vivo, however, proteins fold in the absence of denaturant and thus the behavior of proteins unfolded under non-denaturing conditions is pertinent to our understanding of folding in the cell. Previous studies of globular proteins unfolded by mutation indicate that the unfolded states populated under more physiological conditions are generally compact (8, 9). For example, unfolded forms of λ-repressor, staphylococcal nuclease and drkSH3 populated in the absence of denaturant are expanded only 8%, 30% and 40% relative to their respective folded forms, markedly less than the ~2-fold expansions observed for proteins of similar size under more highly denaturing conditions (8, 10, 11). Consistent with this, some (12–14), though not all (15, 16), proteins undergo transient compaction prior to folding. This compaction, and other compelling spectroscopic evidence for native-like order in proteins unfolded in the absence of denaturant (5), suggests that the physiological unfolded state may contain significant residual structure (17–20). If present such structure could restrict a newly synthesized protein to a native-like topology from which it is preconfigured to fold rapidly (17). The extent to which the collapse of unfolded proteins under physiological conditions is driven by native-like interactions, however, has seen only limited investigation (21–23) and remains controversial (24–26)

Here we explore the relationship between the chemically denatured state and the more compact unfolded state populated in the absence of denaturants. We do so by comparing the spectroscopic features and molecular dimensions of guanidine hydrochloride-(GuHCl)-unfolded FynSH3 to a variant unfolded by mutation. We also compare these states with the properties of a random sequence polypeptide of identical amino acid composition as our unfolded FynSH3 variant. These comparisons allow us to investigate the extent to which generic heteropolymer behavior, that is composition-driven interactions, is sufficient to account for the compaction of proteins unfolded under near-physiological conditions.

Results

In order to investigate the origins of the observed compaction of proteins under physiological conditions it was first necessary to produce a variant protein that is unfolded at room temperature in the absence of denaturants. Scattered but abundant literature reports suggest that terminal deletion mutations of single domain proteins produce unfolded states after removal of several residues from either terminus (27–30). Fyn-SH3, a physically and biologically well characterized, single domain protein of known structure was selected as our test system (31) and, based on the aforementioned studies, subjected to sequential carboxy-terminal deletions. Folding free-energies were determined for each variant (data not shown) in order to identify the least truncated variant, CΔ4 (Fig. 1), that lacks a clear denaturant unfolding transition.

Figure 1.

The sequence of full-length FynSH3 (WT), a four-reside carboxy-terminal truncation (CΔ4) and a random-sequence variant (RS) of identical sequence composition to CΔ4. CΔ4 and RS share less than 4% sequence identity, and BLAST searches indicate that the sequence of RS is not statistically significantly related to that of any other protein. Not shown are unstructured, 21-residue his-tag/linkers at the amino-terminus of all three constructs.

The spectral features of CΔ4 suggest this variant is significantly unfolded at room temperature in the absence of denaturant. For example, while distinct from the circular dichroism (CD) spectrum of the chemically denatured protein, the CD spectrum of CΔ4 differs significantly from that of the native SH3 fold (Fig. 2). While the unusual CD signature of the SH3 fold renders it is difficult to interpret these differences in terms of secondary structure content, the diminished ellipticity in the far-UV clearly indicates a significant loss in structure; in particular, the large, positive peak at 220 nm that presumably arises due to tertiary interactions is almost entirely abolished. Similarly, the 1D NMR spectrum of CΔ4 exhibits a significant loss of peak dispersion, including the effectively complete loss of a highly shifted methyl resonance (Fig. 3) at −0.55 ppm that is characteristic of native SH3 domains (32). The ¹H, ¹⁵N HSQC spectrum of CΔ4 exhibit similarly limited spectral dispersion, further confirming the largely unfolded character of this variant (Fig. 4). Finally, the amide-protons of CΔ4 fully exchange within the 7 minute dead time of a manual solvent exchange experiment, which is at least an order of magnitude more rapid than the exchange rate of the native protein (Fig. 5) and suggests that, even in the absence of denaturant, the fixed secondary structure content of CΔ4 is significantly diminished. This said, readily apparent line broadening in its 1D NMR spectra, non-trivial dispersion in its ¹H, ¹⁵N HSQC spectra and experimentally-significant ellipticity differences from the chemically denatured wild-type protein suggest that, rather than being fully unfolded, CΔ4 resides somewhere towards the unfolded end of the molten globule spectrum.

Figure 2.

While the far-UV CD spectrum of CΔ4 exhibits slightly greater ellipticity than the chemically denatured protein, it exhibits significantly less ellipticity than the native protein under the same conditions, suggesting a significant reduction in structure. In contrast the spectrum of the RS protein collected under non-denaturing conditions is nearly indistinguishable from that of the chemically denatured, wild type protein, which is thought to adopt a fully unfolded, random-coil configuration (4). The spectrum of GuHCl-denatured protein is truncated at 220 nm due to the strong absorbance of the denaturant below this wavelength.

Figure 3.

CΔ4 exhibits little chemical shift dispersion, suggesting that it lacks fixed tertiary structure. In particular, a strongly shifted methyl resonance at −0.6 that is a signature of the SH3 fold is effectively entirely lost. However, CΔ4 exhibits moderate line broadening (relative to both the native and GuHCl-unfolded protein), suggesting that it may adopt a molten globule-like conformation. RS, in contrast, lacks both significant spectral dispersion and line broadening, suggesting that it is nearly completely unfolded even in the absence of denaturant. A large peak arising from TSP (used as an internal standard) was deleted from the RS spectrum, accounting for the data gap around 0 ppm.

Figure 4.

The ¹⁵N HSQC of native FynSH3 exhibits significantly greater chemical shift dispersion than CΔ4, indicating that the latter is relatively unfolded under these non-denaturing conditions. The chemical shift dispersion of FynSH3 in 6M GuHCl is, however, still further reduced. This suggests that CΔ4 resides somewhere along the molten-globule spectrum.

Figure 5.

Hydrogen exchange protection-studies suggest that CΔ4 is largely unfolded. Whereas the native protein has not fully exchanged after 1 hour at pH 7, CΔ4 appears to exchange completely within the 7minute dead time of this manual mixing experiment. Shown are the ratios of total amide-to-aromatic 1H intensities (normalized to unity for the unexchanged proteins) at intervals after transfer into fully deuterated buffer at pD 7.

Despite the above evidence that CΔ4 is largely unfolded in the absence of denaturant, the variant protein is quite compact under these conditions. Specifically, the hydrodynamic radius (R_h) of CΔ4 is 19.4 ± 0.3 Å (95% confidence intervals), which is within error of the 18.4 ± 1.0 Å R_h measured for the full-length protein (Fig. 6; Table 1). We note, however, that both the full-length FynSH3 and CΔ4 constructs contain highly hydrophilic, unstructured 21-residue amino-terminal tails (retained to facilitate purification and handling) and, because of this, our measurements presumably underestimate the true relative expansion of CΔ4. (Given that this tail is extremely hydrophilic, it appears a reasonable assumption that it does not interact with the collapsed protein and remains a random coil.) The observed compaction of CΔ4 thus appears consistent with the 10 – 30% expansion previously reported for molten globules (9). In the presence of 6M GuHCl the R_h of CΔ4 increases to 23.0 ± 1.0 Å (Fig. 6), which is slightly less than the 25.2 ± 0.3 Å observed for the (four-residue longer) full-length protein under the same conditions (Table 1). These values correspond closely to the R_h predicted for chemically denatured proteins of the same lengths (33).

Figure 6.

CΔ4 is relatively compact in the absence of denaturant. Shown is the Pulsed Field Gradient-NMR signal attenuation of CΔ4 and dioxane, an internal standard, measured as functions of field gradient strength in the absence and presence of GuHCl. The small discrepancy between the dioxane measurements arises due to the increased viscosity of GuHCl solutions. The larger discrepancy observed between CΔ4 measurements reflects an additional increase in diffusion constant due to expansion of the chain. The fitted curves (equation 1, 2) predict radii of hydration of 19.4 ± 0.3 Åand 23.0 ± 1.0 Å for CΔ4 in the absence and presence of GuHCl respectively.

Table 1.

Hydrodynamic radii of FynSH3 variants in the presence and absence of denaturant.

FynSH3 variant	Non-denaturing Conditionsa	Denaturing Conditionsa,
WT	18.4±1.0	25.2±0.3
CΔ4	19.4±0.3	23.0±1.0
RS	20.5±1.0	24.4±1.8

^aMean and 95% confidence intervals.

The observed compaction of CΔ4 and other “physiologically unfolded” and molten globual proteins could arise as a consequence of native-like interactions. Alternatively, the observed collapse could reflect a less specific, Flory-like coil-to-globule transition driven by non-sequence-specific interactions (34, 35). In order to discriminate between these possibilities we have characterized a random-sequence polypeptide (RS) of identical amino acid composition to CΔ4 under the assumption that, whereas native-like interactions are sequence specific a Flory-type collapse will depend only on sequence composition. The circular dichroism spectrum of the randomized RS protein in aqueous buffer is closely similar to that of FynSH3 in 6M GuHCl (Fig. 2) and its 1D NMR spectrum shows little chemical shift dispersion (Fig. 3). However, despite the strong spectroscopic evidence that RS is unfolded under these conditions its R_h also approaches that of the native protein (Table 1). In 6M GuHCl, in contrast, its R_h expands to 24.4 ± 1.8 Å, which is effectively indistinguishable from that of the naturally-occurring sequence under the same conditions (Table 1).

Discussion

CΔ4 appears effectively unfolded as monitored by CD, NMR and hydrogen exchange kinetics and yet its dimensions are close to those of the native protein. Similar, albeit somewhat less dramatic compaction has been observed in the unfolded states of λ-repressor, staphylococcal nuclease and drkSH3 that are populated in the absence of denaturant (8, 10, 11), suggesting that the effect is relatively common. Such compaction could be the result of native-like interactions leading to a contraction of CΔ4 relative to the random-coil configuration. A random sequence polypeptide of identical amino acid composition, however, is similarly compact despite exhibiting even less spectroscopic evidence of structure.

CΔ4 might best be thought of as a molten globule. For example, while CΔ4’s NMR spectra and hydrogen exchange kinetics strongly argue that it lacks fixed tertiary structure, the protein’s NMR spectra exhibit line-broadening characteristic of molten globules (9). Consistent with this, the observed compaction of CΔ4 is similar to that of other molten globules, which typically exhibit molecular dimensions quite close to those of the native state (36). Examples include the molten globule states of α-lactalbumin, for which the radius of gyration, R_g, is expanded only 10% relative to the native protein (36), and apomyoblobin, with an R_g expanded 30% over that of the native protein (37). Nevertheless, it is perhaps surprising that CΔ4 should populate a molten globule given that the “structured” regions of this protein span only 53 residues. We are not aware of any other protein this small for which a stable molten-globule state has been reported.

In contrast to CΔ4, the random sequence variant RS does not exhibit the characteristic spectral fingerprints of a molten globule and appears to be fully unfolded even in the absence of denaturant. Under these same conditions, however, the R_h of RS closely approaches that of both CΔ4 and the native protein. This observation is consistent with the work of Urabe and coworkers, who report that a largely random, 141-residue sequence (selected for solubility rather than any attribute more specifically associated with “foldedness” and not corresponding to any naturally occurring protein) also exhibits molten globule-like compaction (38). It is also consistent with the work of Schuler and co-workers, who have measured pairwise distances across the protein cspB unfolded at very low denaturant and find the monotonic relationship between through-space distance and sequence separation expected for a randomly collapsed state(23). Finally, all of these observations are consistent with classic heteropolymer theory (35), which predicts that a polypeptide chain will undergo a coil-to-globule transition whenever monomer-monomer interactions are more favorable than those between monomers and solvent (as is presumably the case under physiological conditions) (24, 35, 39). Such non-sequence-specific, composition-dependant interactions have been invoked to explain the rapid compaction observed during the early stages of folding for RNase A and cytochrome c (24, 25, 40). The observed compaction of RS is consistent with these claims, as are simulations that predict similar sequence-independent compaction for randomized variants of ubiquitin and λ–repressor (41). Because meaningful native-like interactions are presumably sequence-specific (rather than simply composition-dependant), the compaction observed for random sequences such as these demonstrates that molten globule dimensions can be achieved without the formation of sequence-specific, native-like interactions.

The results presented here do not contradict claims that native-like interactions drive the compaction of non-native states. That is, while we have shown that sequence-independent interactions are sufficient to account for the observed compaction, native-like interactions may also be playing a role in the compaction of CΔ4 and other compact non-native states. Moreover, even if compaction is dominated by sequence-independent interactions it is important to note that such interactions may nevertheless accelerate folding by reducing the conformational search to more compact states (42, 43). Thus the formation of the even non-specific compact states may play an important role in protein folding.

Methods

Truncation variants

The coding sequence for full-length Fyn-SH3 was directionally cloned into a pET-15b (Novagen, Inc.) expression vector introducing an amino-terminal 6xHis tag to facilitate purification. QuikChange PCR (Stratagene, Inc.) was used to generate single amino acid truncations from the carboxy-terminus through successive introductions of a stop-codon. The identity of each truncation clone was then confirmed by DNA sequencing (data not shown). Mutant plasmids were transformed into BL21(DE3) pLysS (Stratagene, Inc.) and expressed by induction with isopropyl-β-D-thiogalactopyranoside. The protein was purified by Ni-NTA agarose (Qiagen, Inc.) affinity chromatography in 50mM potassium phosphate buffer, pH 7 (termed “aqueous buffer” below) plus 2M GuHCl, which was added in order to minimize loss from aggregation. The final yield of CΔ4 was approximately 1mg/l.

Design and preparation of the random sequence protein

The random sequence was designed by sequentially numbering each codon of the CΔ4 gene sequence and then using a random number generator to create five unique, randomized sequences. Each of the five sequences was analyzed using the ProtParam Tool (Swiss Institute of Bioimformatics) to assess expressability in E. coli (44). The sequence determined to be the most “expressible,” termed RS (Fig. 1), was then optimized for codon usage (45) and the corresponding synthetic gene was synthesized (Genescript, Inc.) and subcloned into pET-15b (Novagen, Inc.). Expression, purification and approximate yield were as for CΔ4.

Circular dichroism

CD spectra from 200–270nm were collected on an AVIV model 202 CD-spectrometer (AVIV, Inc.) for full-length FynSH3 in aqueous buffer plus 6M GuHCl, and for CΔ4 and RS in aqueous buffer (Fig. 2). Sample concentrations were 10μM as determined by UV-absorbance (assuming ε = 16500 M⁻¹ cm⁻¹).

1D NMR

¹H NMR spectra were collected for wild-type Fyn-SH3, CΔ4 and RS in aqueous buffer and for wild-type Fyn-SH3 in aqueous buffer plus 6M GuHCl (Fig. 3). Spectra were collected in 10% D₂O at 25°C on a Varian 600 MHz spectrometer (Varian, Inc.) over 128 scans of 1024 points.

HSQC

¹⁵N HSQC spectra were collected for CΔ4 in aqueous buffer, and wild-type FynSH3 in aqueous buffer and in aqueous buffer plus 6M GuHCl (Fig. 4). ¹⁵N labeled protein was produced by growing FynSH3 and CΔ4 expressing BL21(DE3) pLysS (Stratagene, Inc.) E. coli in M9 minimal media enriched with 1g/l ¹⁵N ammonium chloride (Cambridge Isotopes Laboratory, Inc.). The labeled proteins were then purified as described above. Spectra were collected in 10% D₂O at 25°C on a Varian 600 MHz spectrometer (Varian, Inc.) over 8 scans, 128 increments.

Hydrogen exchange

¹H NMR spectra were collected for CΔ4 and wild-type Fyn-SH3 in 10% D₂O, 50mM potassium phosphate (pH 7). Protein samples were then rapidly exchanged into 100% D₂O, 50mM potassium phosphate (pD 7) by sephadex-G25 (Sigma, Inc.) spin-column gel-filtration (7 minute dead-time between exchange and collection of first spectrum). Spectra were collected at 25°C on a Varian 600 MHz spectrometer (Varian, Inc.) over 64 scans of 9600 points.

Pulsed-field Gradient NMR

Hydrodynamic radii (R_h) for wild-type Fyn-SH3, CΔ4 and RS were measured in aqueous buffer and in aqueous buffer plus GuHCl (Fig. 6, Table 1) using pulsed-field gradient NMR using the watersLED pulse sequence (46). Experiments were conducted in 100% D₂O at 25°C on a Varian 600 MHz spectrometer (Varian, Inc.) over 32 scans of 9600 points. (Of note, these experiments were conducted in 100 D₂O in order to minimize interference from signals arising due to water. Control experiments conducted in protonated buffer in the absence of added denaturant, however, produce an R_h for CΔ4 that is effectively indistinguishable from that observed in 100% D₂O, indicating that deuteration of the solvent is not responsible for the observed compaction –data not shown.) Signal attenuation of the protein and the reference molecule (dioxane at 1mM) were measured as a function of field gradient strength. Hydrodynamic radii were calculated by fitting the relative decay rates of the sample molecule and the dioxane reference to the relationship (33):

$$I=exp[-{(\gamma \text{g}\delta )}^2(\mathrm{\Delta }-\delta /3)\text{D}]$$ (1)

Where I is the peak height, γ is the proton gyromagnetic ratio, g is gradient strength, δ and Δ are delays and D is the diffusion constant of the molecule, which is the only fitted parameter. Using the latter parameter we obtain the hydrodynamic radius, R_h., of the molecule under investigation using the relationship:

$${\text{R}}_{\text{h}}=D_{\mathit{ref}}/D_{\mathit{prot}}({\text{R}}_{\text{h}\text{ref}})$$ (2)

where D_ref and D_prot are the diffusion constants of the reference molecule and the protein respectively, and R_{h ref} is the known hydrodynamic radius of dioxane.

Abbreviations

CΔ4

4-residue carboxy-terminal deletion variants of FynSH3

R_h

hydrodynamic radius

GuHCl

guanidine hydrochloride

RS

random sequence polypeptide of identical amino acid composition to CΔ4