Concomitant rupture of the HEAT repeat array at multiple sites under denaturing conditions

Mutations were made of the conserved valine residue at position 24 of the HEAT motifs and of non-conserved valine residues at other positions (Table S2). The remaining mutations, S142V, Q169P, S445P and H496V are to consensus residues, these being natural substitutions that ought to be accommodated by the structure of the repeat. Mutations in HEAT repeats 3–7 and HEAT repeats 14–15 (Figs. 2A,B; Table S2) shift the midpoint of the first unfolding transition to lower urea concentrations, indicating destabilization of the native state relative to the intermediate. Mutations in HEAT repeats 8–10 have little or no effect on the first unfolding transition when monitored by fluorescence; however, destabilisation of the native state relative to the intermediate is observed when unfolding of these mutants is monitored by CD (Fig. 2C; Table S3). The lack of fluorescence change associated with unfolding of HEAT repeats 8–10 is likely due to the absence of tryptophan residues in these repeats. Mutations in HEAT repeats 1–2 and 11–13 have little or no effect on the stability of the native state relative to the intermediate when measured by either fluorescence or CD (Fig. 2D; Tables S2 and S3). The simplest interpretation of the results is that the first transition is associated with the unfolding of HEAT 3–10 and HEAT 14–15, and the second transition is associated with the unfolding of HEAT 1–2 and HEAT 11–13, i.e. HEAT 1–2 and HEAT 11–13 are structured in the hyper-fluorescent intermediate and the other repeats are unstructured.

Open in a separate window

Figure 2

Equilibrium denaturation curves of mutant PR65/A

Denaturation curves of the mutants are in red and wild type is in blue. Effects of mutations in HEAT repeats 3–7 and 14–15 are shown measured by fluorescence (A) and by CD (B); for simplicity, only the CD data for the first unfolding transition are shown. Effects of mutations in HEAT repeats 8–10 are shown in (C). Effects of mutations in HEAT repeats 1–2 and 11–13 are shown in (D).

Truncations map the stability distribution across the repeat array and define subdomains

The one-dimensional, modular nature of repeat protein architectures allows the use of truncations to interrogate the structure and provide a complement to site-directed mutagenesis (4, 30, 31). We truncated the protein at the end of each repeat or in the middle of repeats in those cases where the truncation at the end of a repeat was not expressed in a soluble form. Several truncated variants (referred to as HEAT 1–13, HEAT 1–10, HEAT 1–9, HEAT 1–8, HEAT 1–5, HEAT 4–15, HEAT 5–15) were not soluble, and two variants (HEAT 1–6 and HEAT 2–15) were unstable and oligomerised under native conditions. A total of eight truncated variants (HEAT 1–14, HEAT 1–12, HEAT 1–11, HEAT 1–9.5, HEAT 1–8.5, HEAT 1–7, HEAT 3–15, HEAT 3–9.5) were monomeric and were analysed further (Table S4). Although the ellipticity signal at 222 nm could potentially provide information about how much native helical structure is maintained upon truncation, in the case of PR65/A the five tryptophan residues appear to contribute significantly to this ellipticity (as shown by our observation that mutations of the tryptophan residues in full-length PR65/A cause large changes in the 222 nm ellipticity (Fig. S5), and these changes are not the result of partial unfolding in the tryptophan mutants as gel-filtration shows that they are fully folded). It is therefore impossible to correlate the magnitude of the 222 nm ellipticity of the fragments, which contain varying numbers of tryptophan residues, with their helical content. Instead, urea-induced unfolding of the variants was used to compare stability and structural integrity across the whole fragment set. A pattern emerges from the CD-monitored denaturation profiles (Fig. 3A). Deletion of the most C-terminal repeat (HEAT 1–14) results in a large shift in the first unfolding transition to lower urea concentrations and a large decrease in the apparent m-value; upon further truncation (HEAT 1–12 and HEAT 1–11) this transition is shifted back to higher urea concentrations; a similar pattern is evident in the fluorescence-monitored denaturation profiles (Fig. 3B; Table S5). Deletion of HEAT 13 and HEAT 12 also results in a large shift to lower urea concentrations in the second unfolding transition monitored by fluorescence and less oligomerisation of the intermediate as judged by native PAGE (Fig. 3C); moreover, no oligomerisation is observed for constructs comprising fewer than ten repeats. These results are consistent with the site-directed mutagenesis data, and they show that HEAT repeats 11–13 are required for the stability of the oligomeric intermediate. A second transition in the CD-monitored denaturation profile is evident for HEAT 1–9.5, HEAT 1–8.5 and HEAT 1–7 and absent for HEAT 3–9.5, and therefore it most likely corresponds to the unfolding of HEAT 1–2. The CD-monitored denaturation profile does not change greatly between HEAT 1–9.5, HEAT 1–8.5 and HEAT 1–7. The CD- and fluorescence-monitored denaturation data of these three variants together suggest that HEAT repeats 8–9.5 are weakly structured in the absence of any C-terminal repeats.

Open in a separate window

Figure 3

Truncated variants of PR65/A

Denaturation curves of truncated variants measured by CD (A) and fluorescence (B). The truncations are in red and the full-length protein is in blue. The variant HEAT 3–9.5, which is both N- and C-terminally truncated is shown in yellow with HEAT 1–9.5 in red for comparison. The protein concentration used was between 0.5 µM and 3 µM, depending on the sample, and the data were normalized to 1 µM protein concentration. (C) Native gels of wild type and truncated variants. Samples were incubated in 0–8 M urea and then diluted with a loading dye, resulting in final urea concentrations of 0–4 M urea.

The truncations are consistent with the model for the unfolding of full-length PR65/A inferred from the mutants, in which the hyper-fluorescent intermediate has structured HEAT 1–2 and HEAT 11–13. The truncations help us to further delineate the subdomains with the HEAT-repeat array and provide information about the relative stabilities of these subdomains. Deletion of the N-terminal two repeats (variant HEAT 3–15) shifts the first transition to a much lower urea concentration, indicating that HEAT 1–2 acts as a cap for the HEAT 3–10 moiety. The variant HEAT 1–13 was insoluble, suggesting that HEAT 14–15 acts as a cap for the HEAT 11–13 moiety; however, the mutational data both at equilibrium and under kinetic conditions (see below) indicate that HEAT 11–13 remains folded when HEAT 14–15 is unfolded, suggesting that HEAT 11–13 is an independently stable moiety and that the HEAT 14–15 cap acts to protect the HEAT 11–13 moiety from oligomerisation rather than stabilising it against unfolding. Deletion of HEAT 15 destabilises the C-terminal moiety resulting in it unfolding at a much lower urea concentration in the variant HEAT 1–14 than in the full-length protein. Further deletions, of HEAT 13 and HEAT 12, destabilise the HEAT 11–13 moiety causing this moiety to unfold at very low urea concentrations in variants HEAT 1–12 and HEAT 1–11.

Additional intermediates revealed in the unfolding kinetics

The unfolding kinetic traces are characterized by three phases at urea concentrations above 5 M: a fast phase associated with a decrease in fluorescence, a slower phase associated with an increase in fluorescence and a very slow phase associated with a decrease in fluorescence (Fig. S6A). The fast and slower phases are of similar amplitudes and the slowest phase is much smaller in amplitude. At urea concentrations below 5 M the slowest phase is not present. However, the traces still fit to the sum of three exponential phases because a different slow phase appears, one that is associated with an increase in fluorescence (Fig. S6B); this phase is small in amplitude. The unfolding traces obtained by stopped-flow far-UV CD fit to a single exponential phase at higher urea concentrations and to the sum of two exponential phases at lower urea concentrations; the rate constants for these phases are similar to those of the fast and slower phases measured by fluorescence (at higher urea concentrations these two rate constants are very similar and they can only be distinguished in the fluorescence experiments because they have opposite signs). The fast phase is larger in amplitude than the slower phase, which is consistent with the assignment of these phases to the unfolding of specific subdomains of different sizes that we make below.

The fluorescence-monitored refolding kinetic traces can be fitted to the sum of three exponential phases (Fig. S6C). There is a fast phase associated with an increase in fluorescence and two slower phases associated with decreasing fluorescence. Interestingly, the rate constants increase with increasing urea concentrations between 1.9 M and 2.3 M urea, in contrast to the regime below 1.9 M urea and above 2.3 M urea where they show the usual decrease upon increase in urea concentration (Fig. 4A). Non-linearity can be caused by transient aggregation; however, in the case of PR65/A the refolding rate constants are independent of protein concentration between 0.2 µM and 5 µM (Fig. S7); the unfolding kinetics is also independent of protein concentration (Fig. S7). Another explanation is that a misfolded species accumulates that needs to unfold in order to progress to the native state. Due to the complexity of the refolding kinetics and the potential for additional difficulties in interpreting the data arising from the large number of proline residues (26), our subsequent analysis focuses instead on the unfolding kinetics. The plot of the logarithm of the rate constants for unfolding versus urea concentration is shown in Figure 4A.

Open in a separate window

Figure 4

(A) Unfolding and refolding kinetics of full-length PR65/A (A), truncated variants (B) and single-site mutations (C), monitored by stopped-flow fluorescence

The urea dependence of the rate constants is shown for the wild type in blue and the mutants and truncated variant HEAT 3–15 in red. The fast phase is in squares, the slow phase is in triangles, the slowest phase between 3.5 M and 5.5 M urea is in circles and between 5.5 M and 8.5 M urea in inverted triangles

Unfolding of HEAT 3–10 occurs rapidly

Upward curvature is evident in the urea dependence of the rate constants of the fast unfolding phase (Fig. 4A), which suggests that there are parallel pathways accessible, as has been observed observed previously for the large ankyrin repeat protein D34 (36). The additional, downward curvature evident in this limb and in the limb of the slower phase can be interpreted as a denaturant-dependent shift between two (or more) sequential transition states separated by a high-energy intermediate (sequential barriers model) (32) or a gradual movement of the top of a broad energy barrier along the reaction coordinate in accordance with Hammond behavior (33) (34). Mutations in HEAT repeat 3 (V103A) and in HEAT repeats 8–10 (V283A, V333A, V340A, V390A) speed up the fast phase, suggesting that it corresponds to the unfolding of these repeats (Fig. 4C). The mutation V103A does not show upward curvature, suggesting that only one pathway is now accessible. The mutation V340A also shows no upward curvature, suggesting rerouting of unfolding via the other pathway. Since V283A, V333A and V390A are not very destabilizing, neither pathway is abrogated and therefore there is upward curvature for these mutants as for the wild type. Mutations in HEAT 4–7 (S142V, Q169P, V200A, V219A, V258A) have only small effects on the fast unfolding phase and no significant effect on the other unfolding phases (Fig. S8).

A second intermediate is formed subsequently by the unfolding of HEAT 14–15

Very destabilizing mutations in HEAT repeats 14–15 (V535A, V575A) have the most dramatic effect on the slower unfolding phase, which is associated with an increase in fluorescence, suggesting that this phase arises from unfolding of these elements of structure. Interestingly, however, these mutations also increase the rate of the fast phase described in the previous section (Fig. 4C). This result can be understood if the mutations accelerate the rate of unfolding of repeats 14–15 such that this process becomes competitive with unfolding of 3–10, and the unfolding kinetics of these two regions can no longer be considered as sequential steps. A simple scheme that can explain the data allows for 3–10 to unfold first, followed by 14–15, or vice versa. The rate for the fast phase is then the sum of the unfolding rates of 14–15 and 3–10, and can therefore be increased by strongly destabilizing mutations of 14–15. The dominant slow phase is still unfolding of 14–15. For V535A, the slow phase is completely absent (Fig. 4C); this is most likely because the mutation is so destabilizing that the initial state observed has repeats 14–15 already unfolded – because either they are already unfolded at low denaturant as suggested by the equilibrium fluorescence, or they unfold in the dead time of the instrument. A similar scenario is most likely responsible for the absence of the slow phase for V575A at higher urea concentrations. For V532A the slow phase is speeded up to a lesser extent and therefore can still be observed. Later we describe in detail how such a scheme quantitatively fits the kinetic traces for all the above mutants.

A truncated variant comprising HEAT 13–15 was made in order to simplify the unfolding kinetics; W477 provides a spectroscopic probe for this variant. The unfolding traces of HEAT 13–15 are monophasic and associated with an increase in fluorescence (Fig. S9A); this result is consistent with our assignment of the slow phase for full-length PR65/A to the unfolding of HEAT 14–15. These kinetic data are also consistent with the equilibrium data showing that HEAT 14–15 is unfolded in the hyper-fluorescent intermediate. Interestingly, V535A has an effect on the slow unfolding kinetics also: whereas for the wild type there is a slow unfolding phase below 5 M urea associated with an increase in fluorescence (corresponding to unfolding of HEAT 1–2, see below), this phase is absent in the mutant V535A and instead there is a slow phase associated with a decrease in fluorescence.

HEAT 1–2 unfolds to form a third intermediate

The biggest effects of mutations V25A and V44A (HEAT 1 and HEAT 2) were on the slowest unfolding phase, observed at urea concentrations below 5 M, which is associated with an increase in fluorescence (Fig. 4C). The rate constants for this phase are speeded up by these mutations. Moreover, the phase is absent in the N-terminally truncated variant, HEAT 3–15 (Fig. 4C). The results suggest that this phase corresponds to the unfolding of HEAT 1–2, which occurs separately from HEAT 3–10 below 5 M urea whereas HEAT 1–2 unfolds concomitantly with HEAT 3–10 above 5 M urea. Interestingly, V44A also speeds up the phase assigned to the unfolding of HEAT 14–15.

Unfolding of HEAT 11–13 is the slowest step

V435A (HEAT 11) and V454A (HEAT 12) speed up the slowest unfolding phase observed for the wild type above 5.5 M urea (Fig. 4C). The absence of this phase at urea concentrations below 5.5 M urea is consistent with it corresponding to unfolding of the HEAT 11–13 moiety: the equilibrium experiments show that this moiety is folded at lower urea concentrations. V435A also has an effect on the other phases; the slow phase, corresponding to unfolding of HEAT 1–2 and associated with increasing fluorescence, is absent in its kinetics.

Unfolding kinetics of truncated variants

The unfolding traces of C-terminally truncated variants HEAT 1–9.5, HEAT 1–8.5 and HEAT 1–7 (all of which have W140 and W257) show a decrease in fluorescence (Fig. S9B). The absence of the phase observed for full-length PR65/A associated with an increase in fluorescence suggests that the hyperfluorescence of the intermediate arises from the unfolding of repeats in the C-terminal moiety and it is consistent with the absence of a hyper-fluorescent intermediate for these truncations at equilibrium. The unfolding of HEAT 1–9.5 can be fitted to the sum of three exponential phases (Fig. S9B and Fig. 4B). There is a very fast phase that is faster than any of the phases observed for the full-length protein, a fast phase that has the same rate constants and urea dependence as the fast phase of the full-length protein, and also a slow phase. The slow phase is very low in amplitude (~5%). The very fast phase is the major phase at high urea concentrations whereas the fast phase is the major phase at lower urea concentrations. The unfolding of HEAT 1–8.5 can be fitted to the sum of two exponentials and shows the same very fast and fast phases as HEAT 1–9.5; the fast phase has very low amplitude and the slow phase observed for HEAT 1–9.5 is absent (Fig. 4B). The unfolding kinetics of HEAT 1–7 shows a single, very fast phase, which is similar to what is observed for HEAT 1–8.5 and HEAT 1–9.5 (Fig. 4B). The presence of a fast phase for variants HEAT 1–8.5 and HEAT 1–9.5, similar to that observed for full-length PR65/A, and its absence in the kinetics of HEAT 1–7, is consistent with the assignment of this phase to the unfolding of HEAT 8–10. We therefore propose the following to explain the appearance of the very fast unfolding phase in these three truncated variants and the presence at low amplitudes in HEAT 1–8.5 and HEAT 1–9.5 of the fast unfolding phase characteristic of full-length PR65/A: in the absence of HEAT 8–10, the HEAT 1–7 variant is able to unfold in a single very rapid step. For HEAT 1–8.5 and HEAT 1–9.5 there is a mixture of ‘native’ states, arising because the HEAT 8–10 moiety is truncated; the truncated moiety is unstructured in the major population and so unfolding proceeds as it does for HEAT 1–7, whereas the moiety is structured in the minor population and so unfolding proceeds as it does for the full-length protein.

Unfolding of HEAT 4–7 in a very fast unfolding step revealed by double-jump experiments

Two observations suggest that HEAT 4–7 unfolds after the unfolding of HEAT 8–10 in a very fast step, which, since it is faster than the previous step, is not detected in the kinetic measurements. First, mutations in HEAT repeats 4–7 do not have a clear effect on the any of the five kinetic unfolding phases observed for full-length PR65/A, whereas mutations elsewhere in the protein have clear effects that allow each phase to be assigned to the unfolding of a specific subdomain. Second, when repeats adjacent to HEAT 4–7 are deleted (i.e. the truncated variants HEAT 1–7, HEAT 1–8.5, HEAT 1–9), a very fast phase is observed that is absent in the unfolding of the full-length protein (Fig. 4). To test whether there is an additional phase that is missed in the single-jump method we performed a double-jump experiment in which unfolded protein was allowed to refold for a delay time of between 30 ms and 50 s before unfolding was initiated. The unfolding kinetics of this refolded protein was multiphasic and all but one of the phases had similar rate constants to those observed by single-jump unfolding of native protein (Fig. S10). The new phase is very fast and its amplitude first increases and then decreases with increasing refolding delay time, consistent with the accumulation and decay of an intermediate. We therefore propose that this very fast phase is the ‘missing’ phase that corresponds to the unfolding of HEAT 4–7.

Kinetic model for unfolding of PR65/A

The above mutation and truncation studies establish qualitatively that the unfolding of PR65/A can be understood in terms of a predominant order of unfolding events. We have used these data to construct a six-state phenomenological kinetic model that explains both the wild-type data and the data for mutants over a wide range of denaturant concentrations. The scheme is summarized in Fig. 5A and includes the native and unfolded proteins as well as four intermediates, labeled I–IV. The initial unfolding of the wild-type protein occurs mainly via unfolding of repeats 3–10 (rate k₁), followed by unfolding of 14–15 (rate k₂). However, an alternative possibility is that these two regions could unfold in the reverse order, and, as described below, these parallel pathways are able to explain several features of the kinetic data. The last two steps are unfolding of repeats 1–2 (rate k₃), followed by 11–13 (rates k₄, k₋₄). Unfolding of 11–13 has been treated as being reversible in the model, since it does not occur at equilibrium below ~5 M urea for the wild-type protein. To allow for downward curvature in the denaturant-dependence of the rate k₁, it was modeled using a sequential barriers scheme in which unfolding from native to intermediate I occurs via a high energy intermediate X, i.e. k₁ = k_nxk_xi/(k_xu+k_xn) (32). This assumption is reasonable, as k₁ describes the folding of the largest contiguous “block” of repeats (3–10), and a similar model has been used to describe downward curvature in the unfolding of other long repeat arrays (7, 35). Note that this extra complication is relatively unimportant, and is only necessary to describe the unfolding data below ~4 M urea concentration. The scheme has five kinetic eigenvalues corresponding to the rates k₁, k₂ and k₃ and the sums k₁+k₂ and k₄+k₋₄ (see SI for details).

Open in a separate window

Figure 5

Kinetic model for PR65 unfolding

(A) Kinetic scheme in which the native protein unfolds via a sequence of (parallel) intermediates to the unfolded state. Names of species are given in boxes and rate coefficients for their interconversion are shown next to the arrows. Color code refers to the region which is unfolding in each step. (B) Lower: Relaxation rates from kinetic model (lines) overlaid on the rates obtained by fitting of exponentials to raw data; color code corresponds to that in (A), as far as possible. Upper: Amplitudes corresponding to each relaxation. Positive amplitude corresponds to a decrease of fluorescence with time. (C) Left axes: Raw fluorescence traces (blue symbols) overlaid with the global fit from the kinetic model (red lines); Right axes: Time evolution of the populations of the species shown in (A). A range of concentrations are shown for wild-type in the left column, and in the right column are shown representative traces for five mutants chosen from the different structural regions.

The fluorescence signal was modeled as a sum over populations of each species, weighted by their relative fluorescence quantum yield. The model was globally fit to the raw fluorescence traces at all urea concentrations for the wild-type protein and several mutants: V25A representing repeats 1–2, V333A for repeats 3–10, V436A for repeats 11–13, and V532A and V575A for repeats 14–15. The fit parameters for the wild type were the rates at 4 M urea, their denaturant dependences (m-values) and the quantum yields for each species. The mutants shared all the parameters with wild type, with only the rate coefficient(s) pertaining to unfolding of the relevant region being allowed to vary. However, for V436A it was found to be necessary to optimize separately the relative quantum yields; in retrospect, this can easily be rationalized by a direct Van der Waals contact between V436 and W477 in the crystal structure, resulting in changes in environment and solvent exposure of that tryptophan on mutation. The final optimized parameters are summarized in Table S6.

The model formally has five relaxation time scales, which are plotted versus urea concentration in Fig. 5B (lines). In Fig. 5C are shown the raw fluorescence traces fitted to the model. For clarity, we will describe separately the situation at urea concentrations below and above 5 M. At low denaturant concentrations, the protein unfolds via a series of hypofluorescent intermediates (I,II,III) to the hyperfluorescent intermediate IV which is stable up to ~5M, resulting in a “dip” in the fluorescence with time. This can clearly be seen to be correlated with the population of the intermediates also plotted in Fig. 5C. There are three exponential phases evident in the raw data (the relaxation rates for these are plotted as symbols in Fig. 5B). In the model, the fastest (green) phase is associated with the formation of intermediates I and II with rate k₁+k₂, which has a large positive amplitude, i.e. causes a decrease in fluorescence (upper part of Fig. 5B). Decay of II occurs with a very similar rate k₁≈ k₁+k₂ with small negative amplitude, but because of the similarity of time scales in the wild type, and the small population of intermediate II, it cannot be distinguished in the fluorescence traces. The second (magenta) exponential phase arises from decay of the populated intermediate I with rate k₂, and has a significant negative amplitude. This is primarily responsible for the rise in fluorescence at the end of the “dip”. Finally, a third (cyan) exponential phase with very small negative amplitude comes from conversion of intermediate III to IV at long times.

Above 5 M urea, the fastest phase is still associated with the formation of intermediates I and II with rate k₁+k₂. However, upward curvature is seen in this phase, resulting from the fact that at (very) high urea concentration, k₂>k₁ and there is a switch of pathway toward that going via intermediate II. As a result of the larger k₂, the second fastest phase now contains contributions from the decay of both intermediates I and II, since intermediate II is now significantly populated and the rates k₁ and k₂ are now more comparable. At very high denaturant, the second fastest phase overlays well with the rate k₁ as this phase is now mostly associated with decay of intermediate II, which has become larger in population than I. The relaxation with rate k₃ is still present, but less evident in the fluorescence traces, because above 5 M urea, intermediate IV decays further to the unfolded state with rate ~k₄. The amplitude of this final relaxation is of course initially zero and only emerges with significant positive amplitude at urea concentrations above ~6 M (Fig. 5B, upper panel).

The fluorescence traces for the mutants can be well described by the same model, by changing only the rates associated with the unfolding of the relevant subdomain. The fits for representative urea concentrations are shown in Fig. 5C. For example, the V25A data can be reproduced by an increase of rate k₃. The result is a faster depletion of intermediate III such that the fluorescence has almost reached a “plateau” value at long times. The V333A data are explained by an increase in k₁ (via k_nx), such that intermediate I is populated more rapidly leading to a faster initial fluorescence decay and a larger overall “dip” in the total fluorescence. Mutation V436A results in a large increase in k₄ and decrease in k₋₄. Consequently, the population of intermediates is much reduced at high denaturant concentration (being converted quickly to unfolded), and the overall signal appears almost single-exponential. Mutations V532A and V575A affect rate k₂. Because of the parallel pathways, an increase in k₂ can have multiple effects. The V532A results in the smaller increase of k₂ and the primary effect of this is that intermediate I decays more rapidly, such that the rise in fluorescence occurs earlier and the overall signal amplitude is smaller. The effect of V575A on k₂ is much larger, such that k₂ becomes larger than k₁. This also increases the rate of the initial fluorescence decay (with rate k₁+k₂), and results in intermediate II being the predominantly populated one. Because II has a larger fluorescence quantum yield than I, and decays with a fast rate (k₁), the result is that the magnitude of the “dip” in fluorescence is very small, and signal to noise ratio is small.

Origins and consequences of cooperativity in repeat proteins

Based on the order of unfolding events inferred from mutant studies, we have constructed a six-state kinetic model. Using a common set of rate coefficients and relative quantum yields for the different states, we are able to fit the kinetic data quantitatively over a wide-range of denaturant concentrations, for both wild-type and mutant proteins. The only parameter that is varied between the wild-type and each mutant protein is the unfolding rate of the region in which the mutation is located. Although there is a predominant order of unfolding, the model includes parallel unfolding pathways, which explain both the upward curvature in the fastest relaxation rate at high denaturant, as well as the fact that some mutations are able to alter more than one relaxation rate. HEAT 3–10 unfolds most rapidly, followed by HEAT 14–15, and then HEAT 1–2 and HEAT 11–13 unfold in slower steps. Because the distribution of stability between different subdomains of the array is very uneven, this order of unfolding events is strongly preferred, such that only very destabilizing mutations can shift the order, for example making unfolding of HEAT 14–15 competitive with HEAT 3–10. Similarly, our group and others have shown that alternative kinetic (un)folding routes are accessible for a number of ankyrin-repeat proteins such as myotrophin (7), D34 (35) and Notch (38), as was originally predicted based on the symmetric topology of repeat proteins (39). When intermediates are destabilized, unfolding of the HEAT repeats can occur in a single step, as occurs for example in the truncated variant HEAT 1–7.

The origins and limits of cooperativity have been explored for ankyrin and TPR repeats as well as more recently for LRRs, using both experimental and computational approaches (4, 5, 8, 39–42). These studies indicate that repeats that are distant from one another and therefore not directly in contact can nevertheless unfold in a coupled manner if they and the intervening units have similar intrinsic, low stabilities and the interfaces between them are strong. Cooperativity is thereby maintained over the medium range (4). If the distribution of stabilities is uneven, an intermediate state will be populated; for example in D34, a protein comprising twelve ankyrin repeats, there is an intermediate in which the C-terminal six repeats are structured and the N-terminal six repeats are unstructured (8). Moreover, cooperativity in repeat proteins can be easily modulated in a rational way, by changing the stabilities of the individual units; thus, we were able to make variants of D34 in which the unfolding of ten of the twelve repeats was coupled (8). In PR65/A, the stability is unevenly distributed across this large array with the result that multiple partly folded intermediates are populated.

Protein expression and purification

Wild-type PR65/A, mutants and truncated variants were expressed and purified as described (22). Protein concentration was determined spectrophotometrically (50). The protein was frozen after purification and stored at −80 °C.

Fluorescence spectroscopy

Fluorescence was measured using a Perkin Elmer luminescence spectrometer LS55. Urea samples were prepared using a Hamilton Microlab 500 Series. Protein stock was added to a final concentration of 0.5 µM in 50 mM MES buffer pH 6.5, 1 mM DTT. The samples were equilibrated at 25 °C for at least two hours before measurement. The 1 cm pathlength cuvette was thermostatted using a water bath (Thermo Neslab). An excitation wavelength of 280 nm was used and the excitation and emission bandwidths were 5 nm. Wavelength scans between 320 nm and 370 nm were performed for each sample at a scan speed of 1 nm s⁻¹.

Circular dichroism

Sample preparation was as for fluorescence except that the final protein concentration was 1.5 µM. Samples were incubated at 25 °C for 2 h in 25 mM MES buffer pH 6.5, 1 mM DTE and different urea concentrations. Far-UV CD spectra between 215 nm and 225 nm were then collected on a Chirascan™ circular dichroism spectrometer (Applied Photophysics) using a 0.2 cm pathlength cuvette thermostatted with a waterbath.

Native PAGE

Protein samples at 5 µM concentration were incubated in 50 mM MES buffer pH 6.5 containing 0 M, 1 M, 2 M, 3 M, 4 M, 5 M, 6 M, 7 M or 8 M urea for 2 h and then diluted with a loading dye resulting in final urea concentrations of 0 M, 0.5 M, 1 M, 1.5 M, 2 M, 2.5 M, 3 M, 3.5 M and 4 M urea. Samples were subsequently loaded on a Novex® Tris-glycine gel (Invitrogen) and run for 3 h.

Limited proteolysis

PR65/A, either in buffer or after 2 h incubation in 2.5 M urea, was digested with Proteinase K (dilution of 5 units per ml) at room temperature. The reaction was then terminated by boiling the samples for 10 minutes. Digestion fragments were resolved by SDS-PAGE and blotted on a membrane for N-terminal sequencing (performed by the Department of Biochemistry, University of Cambridge).

We thank the Center of Theoretical Biological Physics for computational resources. P.G.W. and P.O.C. acknowledge NIH grants GM071862 and R01 GM44557. L.S.I., M.T. and E.S. are supported by the Medical Research Council of the UK (MRC) (grant G1002329) and the Medical Research Foundation. R.B.B. is supported by a Royal Society University Research Fellowship.