Ligand Binding Modulates the Mechanical Stability of Dihydrofolate Reductase

INTRODUCTION

Dihydrofolate reductase (DHFR) is an essential enzyme that converts 7,8-dihydrofolate (DHF) to 5,6,7,8, tetrahydrofolate (THF) in the presence of nicotinamide adenine dihydrogen phosphate (NADPH) (1,2). THF serves as the vital one-carbon donor in the syntheses of thymidylate and purine nucleosides. Inhibiting DHFR blocks DNA synthesis and kills the cell. Anti-folate drugs such as methotrexate (MTX), which bind to DHFR more strongly than the natural substrate DHF, are common cancer therapeutics (3). The binding sites of NADPH and DHF/MTX are located in two different regions of DHFR, but, at the enzyme's active site, the two ligands come into close proximity to enable hydride transfer from NADPH to DHF (4).

Ligand-binding-induced conformational changes in proteins are ubiquitous (5-8). Structural changes accompanied by ligand binding are important in the enzymatic function of GTP-binding proteins (5), ion channels (6), and calcium binding proteins (7,8) to name a few. Ligand binding also affects the thermodynamic stability of proteins (7,8). Ligand binding to DHFR also causes large changes in thermodynamic stability (9). The midpoint urea-induced unfolding transition of human DHFR is shifted from 1.4 M urea to 2.8 M urea in the presence of NADP^sup +^/folate (9).

The large changes in thermodynamic stability induced by the binding of MTX have made DHFR the molecule of choice in studying protein translocation through protein channels like those of the mitochondrial membrane (10,11) as well as those of the adenosine triphosphate (ATP)-dependent proteases found in prokaryotes and eukaryotes (12,13). These channel pores are narrow, ranging between 10-22 [Angstrom] (14-16). Hence, folded globular proteins, ranging in size upwards from ~5 nm, must be unfolded before they can enter the protein channels of the mitochondrial membrane import motor (11,17) and ATP-dependent proteases ( 17-20). Various studies have proposed that the mitochondrial import motors (Hsp70) are capable of doing mechanical work either as a Brownian ratchet (16), or as molecular motors (21) or ratchet-motor mixtures (22). These publications suggest that mechanical unfolding of the targeted protein is an essential step in translocation through protein channels. DHFR has played a significant role in these studies. In the absence of MTX, DHFR readily traverses the translocation protein channels in mitochondria (10,11) and the degradation protein channel in the proteasome (12,23), whereas addition of MTX pronouncedly slows the rate of both the mitochondrial import (10,11) and the proteasomal degradation of DHFR (12,23). These observations suggest that MTX increases the mechanical stability of DHFR. In this article, we used protein engineering combined with single-molecule atomic force microscopy (AFM) techniques (24-29) to test this hypothesis.

A detailed sequence of events taking place in a typical polyprotein stretching experiment by single-molecule atomic force microscopy is depicted in Fig. 1 A and the resulting sawtooth pattern force-extension curve in Fig. 1 B. The mechanical stability of the protein being measured can be readily determined from the average peak force required to unfold each module (28).

Our experimental results demonstrate that DHFR unravels easily at forces averaging 27 pN, and binding ligands or inhibitor increases its unfolding force to 83 pN. The increased mechanical stability directly explains the large reduction in the degradation rate of DHFR when bound to one of its ligands (12,23), supporting the view that mechanical unfolding is a required step before protein translocation or degradation (23,30). Our study is the first demonstration that ligand binding can strengthen a protein against mechanical stress. Together with the recent discovery that the mechanical stability of a protein depends on the direction of the applied force (31,32), ligand modulation of mechanical stability as reported here opens a previously unrecognized perspective in cell biology.

METHODS

Protein engineering

(DHFR)^sub 8^ and (127-DHFR)^sub 4^ were constructed following the methods described previously (29). Briefly, the Chinese hamster ovary DHFR (CHODHFR) gene, which is [approximate]90% homologous to the human gene, was amplified by PCR with 5' BawHI restriction site and 3' Bg/II and KpnI restriction site and cloned into the vector pTVBlue (Novagen, Madison, WI). In the case of the (127-DHFR)^sub 4^, we used the vector pT7Blue, which already contained the clone of human cardiac immunoglobulin I27 (24). Iterative cloning was used to make the synthetic genes of (127-DHFR)^sub 4^ and (DHFR)^sub 8^. These genes were cloned into the expression vector pQE16 (Qiagen, Hilden, Germany). Protein expression was done in BLR(DE3) cells (Novagen) in the presence of 1 mM IPTG. The proteins were purified with Ni^sup 2+^ affinity chromatography and eluted with a buffer (pH 7.0) containing 250 mM imidazole, 50 mM sodium phosphate, and 300 mM sodium chloride.

Chemicals

MTX, NADPH, and DHF were purchased from Sigma-Aldrich (St. Louis, MO) and used without further purification. Phosphate-buffered saline (PBS) was used for making stock solutions. MTX was first dissolved in a minimum amount of 0.1 N NaOH and then diluted to the desired concentration with PBS buffer. The pH of the resulting solution was adjusted to 7.4.

Single-molecule AFM experiment

The details of the atomic force microscope and its mode of operation have been described elsewhere (33). The spring constant of the cantilevers that we used was [approximate]40 pN/nm, measured using the equipartition theorem (34). Unless stated otherwise, we used a pulling rate of 400 nm/s and the peak-to-peak noise in the recordings was 15 ± 4 pN.

RESULTS

The DHFR-MTX complex is mechanically stable

To examine the mechanical stability of DHFR we used protein-engineering techniques to construct a polyprotein (35) made of eight identical repeats of DHFR (see Methods, above). We use polyproteins to identify the molecule being pulled by the AFM. Polyproteins give a characteristic sawtooth pattern in the force-extension curves, serving as a fingerprint to identify the molecule that is being pulled. This is essential given that a majority of the force-extension curves arise from pulling unidentified molecules and only 1-10% of the pulls result in sawtooth patterns (36). Despite these considerations, force-extension curves obtained by stretching the DHFR^sub 8^ polyprotein were featureless and never showed the characteristics of a sawtooth pattern (regularly spaced peaks of similar amplitude). These data could be interpreted as an indication that the DHFR protein lacks mechanical stability. However, the data show a negative result and therefore we cannot be certain whether DHFR is mechanically unstable, or we simply failed to correctly pick up the native form of the protein (Fig. 2 A). A very different result was observed when we added 1.2 mM MTX to the bathing solution. Under these conditions we repeatedly observed force-extension curves displaying sawtooth patterns with peak unfolding forces averaging 78 ± 14 pN (n = 72) (Fig. 2 B; see also Table 1). The increase in contour length of the unfolding molecule was determined by fitting the wormlike chain (WLC) model of polymer elasticity to the segment of the force-extension curve leading up to each peak (Fig. 2 B; thin lines). We measured a spacing between peaks of 67.3 ± 0.5 nm (n = 72). This contour length increment is in close agreement with an expected value of ~65 nm, calculated from the number of amino acids in the DHFR sequence (186 aa × 0.36 nm/aa = 67 nm) minus the end-to-end distance of the folded DHFR protein (1.8 nm for Human DHFR).

It was clear from the experiments that the DHFR^sub 8^ polyprotein could not be used if the protein became mechanically weak in the absence of the MTX ligand. Under those conditions, the lack of a clear fingerprint prevented us from positively identifying the molecules being pulled. Hence, we engineered protein chimeras where DHFR was combined with an immunoglobulin-like module (127 from human cardiac titin) that would provide a clear sawtooth pattern fingerprint even when MTX was absent (29,37). Accordingly, we constructed a (DHFR-I27)^sub 4^ polyprotein. The 127 protein module serves as fingerprint with a regular spacing of 28.1 nm and an unfolding force of ~200 pN (29,35).

We started by stretching the (DHFR-I27)^sub 4^ polyprotein chimera, in a PBS solution containing 190 µM MTX. The resulting force-extension curves showed sawtooth patterns that contained two distinct regions (Fig. 3 A). The earliest region contained unfolding peaks of 82 ± 18 pN (n = 277; Fig. 3 B), which were equally spaced by contour length increments of 67.4 ± 1.0 nm (n = 277; Fig. 3 C) closely corresponding to the sawtooth patterns observed for the unfolding of the DHFR^sub 8^ polyprotein (Fig. 2 B). The later part of the sawtooth pattern was clearly marked by unfolding peaks of 220 ± 36 pN (n = 322), which were equally spaced with contour length increments of 28.0 ± 0.7 nm (n = 322), identifying them as 127 unfolding events (29,35). In the force-extension recordings of Fig. 3 A, we can clearly distinguish three or four 127 unfolding events, establishing unambiguously that the three earlier peaks arise from the unfolding of DHFR (see (29) for a discussion on the usage of the chimera approach). It is clear from the experiments that DHFR in the presence of MTX requires 78-82 pN to unfold. This unfolding force was found to be relatively independent of the MTX concentration in the range between 19 µM and 1.2 mM MTX (Table 2). We have done experiments at a lower pulling speed (80 nm/s) on (DHFR-I27)^sub 4^ in the presence of 190 µM MTX. At 80 nm/s, the DHFR unfolding force was 62 ± 7 pN. Under the same conditions the unfolding force for the I27 was 181 ± 24 pN, in close agreement with earlier results (28).

DHFR is mechanically weak

Force-extension curves obtained by stretching the (I27-DHFR)^sub 4^ polyprotein in the absence of any ligand are shown in Fig. 4 A. In contrast to the recordings obtained with MTX, the force-extension curves now show a featureless spacer followed by a set of I27 unfolding events. The I27 unfolding events are recognized by their characteristic unfolding force and contour length increment between peaks (209 ± 34 pN and 28.1 ± 1.3 nm, respectively; n = 226). The top two traces shown in Fig. 4 A show four I27 unfolding events implying that at least three DHFR molecules must have been stretched. In the bottom two traces showing three I27 unfolding events, we must have extended at least two DHFR proteins (see (29) for a discussion of this point). Most of the recordings obtained without MTX show similar results with only a long featureless spacer before the I27 fingerprint. However, in some cases, we observed either a sawtooth pattern at very low force (e.g., Fig. 4 A; second trace from the top) or isolated peaks in the region of the extension that should contain DHFR unfolding events. To estimate the force required to unfold DHFR in the absence of MTX, we constructed a series of WLC curves with contour length increments of ?L^sub c^ = 67 nm that end with the first I27 unfolding event (dotted lines in Fig. 4 A). The number of WLC curves was set to the minimum number of DHFR unfolding events expected for a given number of I27 unfolding events (see above). The highest force at which the WLC curves intersected the experimental trace was taken as the unfolding force of DHFR. A histogram (n = 163) of unfolding forces of DHFR is shown in Fig. 4 B, with a mean unfolding force of 27 pN, which is <50% of the force for the DHFR-MTX complex and it is close to the resolution of our apparatus (see Methods, above). Even at a higher pulling rate (4000 nm/s) we did not observe a mechanical fingerprint for DHFR. The force-extension curves obtained at 400 nm/s and 4000 nm/s were similar. The putative unfolding force of DHFR was higher at 4000 nm/s than at 400 nm/s (53 vs. 27 pN). Although the putative unfolding force doubled at 4000 nm/s, the peak-to-peak noise tripled on increasing the pulling speed (from 15 ± 4 pN to 48 ± 12 pN). At 4000 nm/s, 60% of the putative unfolding events occurred within the noise level (<50 pN) and the unfolding force histogram showed no clear feature above the noise level. Hence, the observed difference may not be significant. By contrast, we observed a clear difference in the unfolding force of the I27 module, which was found to be 262 ± 55 pN, in close agreement with those previously reported (28). Hence, when there is no ligand, DHFR does not show a significant mechanical stability. This is not surprising, given that there are proteins, with a well-defined folded structure, which were nonetheless shown not to have a measurable mechanical stability. For example, both calmodulin and barnase polyproteins were shown to extend readily at a low force without any unfolding force peaks (25,38). In both of these cases, the engineered polyproteins were shown to have a thermodynamic stability that was similar to that of the monomers (21,25,38). We have not measured the thermodynamic stability of the DHFR polyproteins due to the difficulty in expressing them in sufficiently large amounts for bulk experiments. However, the DHFR protein has been used extensively as a fusion protein with a variety of other proteins and peptides, without altering the stability of the protein (12,21,22,30,39,40).

The cumulative unfolding probability, obtained by integrating and normalizing the unfolding force histograms of Fig. 3 B and Fig. 4 B is shown in Fig. 5. Remarkably, at 40 pN, 80% of ligand-less DHFR have already unfolded (thick line in Fig. 5), whereas only <10% of DHFR have unfolded in the presence of 190 µM MTX (dashed line in Fig. 5). These results show that upon binding MTX, the DHFR protein becomes mechanically stable.

DHF and NADPH ligands mechanically stabilize DHFR

The natural substrate DHF binds to DHFR at the same site as MTX, but the dissociation constant of human DHFR-DHF (K^sub d^ [asymptotically =] 580 nM) (41) is almost two orders-of-magnitude larger than the dissociation constant of DHFR-MTX (K^sub d^ <10 nM) (42). The mechanical properties of DHFR in the presence of 180 µM DHF are given in Table 2. The force required to unfold DHFR-DHF is 83 ± 16 pN (n = 36), which is similar to that of DHFR-MTX. In these experiments, the ligand concentration (180-190 µM) was always higher than the concentration of DHFR (~1 µM). Under these conditions, most of the DHFR (>99%) is in ligandbound form and >99% of the binding sites of DHFR molecules are occupied.

The coenzyme NADPH has a different binding site from MTX or DHF (1) with a dissociation constant of K^sub d^ [asymptotically =] 45 nM (42). The unfolding force of DHFR in the presence of 210 µM NADPH is 98 ± 15 pN (n = 86), which is slightly larger than the rupture force obtained with MTX or DHF. However, this difference is within the margin of error of the measurement and may not be significant. Therefore, although MTX and NADPH bind at different sites in the protein, they induce similar mechanical stability in DHFR.

Interestingly, we found that MTX and NADPH were not additive in their stabilizing effects. We measured the force required to unfold DHFR in the presence of 190 µM MTX and 210 µM NADPH (Fig. 6). The figure shows that DHFR in the presence of both NADPH and MTX still requires 83 ± 13 pN (n = 136) to unfold, which is similar to the force required to unfold the DHFR-MTX complex or the DHFR-NADPH complex (Table 2). Despite the fact that the binding of MTX and NADPH to DHFR is cooperative (43), no additional mechanical stability was found when DHFR was occupied by both ligands simultaneously.

DISCUSSION

Mechanism of DHFR stabilization by MTX, DHF, and NADPH

A ligand (MTX, DHF, or NADPH) may stabilize DHFR against mechanical unfolding by several plausible general mechanisms. The increased mechanical stability could be the result of ligand-protein specific interactions, conformational changes triggered by ligand binding.

Ligand binding could introduce specific DHFR-ligand interactions, such as hydrogen bonds or van der Waals interactions, and these interactions contribute to directly resisting the applied mechanical force. In such case, we expect a one-to-one relationship between binding strength and mechanical resistance. However, we can clearly rule this out, given that the force required to unfold the DHFR-MTX complex is the same as that required to unfold the DHFR-DHF complex, and that the dissociation constant of DHFR-MTX is >50-fold smaller than that of DHFR-DHF (< 10 vs. 580 nM, see above). Similarly, the addition of a second ligand fails to increase the force required to unfold the DHFR complex.

Ligand binding could cause a conformational change in DHFR, thereby bringing the protein into a mechanically stable state. It is already known that ligand binding induces conformational changes in the DHFR molecule (44). Studies of the crystal structure of Escherichia coli DHFR before and after binding a ligand show large conformational changes in the Met20, G-H, and F-G loops (44). These extensive conformational rearrangements upon binding different ligands suggest that the flexibility of DHFR is modulated by ligand binding. It has been speculated that the structural fluctuations observed in E. coli DHFR are necessary to accommodate the intermediates that form during the catalytic cycle (44). Using hydrogen/deuterium exchange to probe the amplitude of these fluctuations Yamamoto and colleagues (45) found that the binding of either folate or NADPH reduced hydrogen/deuterium exchange, indicating that the structural fluctuations of DHFR were reduced by ligand binding. Furthermore, they found that the magnitude of the reduction in the amplitude of the fluctuations was not additive, when both folate and NADPH were simultaneously bound to DHFR (45). It is tempting to conclude that the fluctuations in the E. coli DHFR structure, as well as their reduction in the presence of ligands, are correlated with the changes in mechanical stability that we observe in our single-molecule AFM studies. However, since the CHO-DHFR used in our studies is <30% homologous to its bacterial counterpart, it may be misleading to compare the DHFRs from the two species. Instead, it should be more appropriate to compare the CHO-DHFR with the human version, because the protein sequences are 90% identical. In urea denaturation, human DHFR is only marginally stable in the absence of ligands, with ?G^sub U-N^ = 2.4 kcal/mol (9). Binding of folate and NADP^sup +^ stabilize human DHFR by [approximate]3.5 kcal/mol (9). High concentrations of human apo-DHFR have a strong tendency to aggregate (9), and human DHFR has been crystallized only when it is bound to ligands (4).

DHFR translocation across the mitochondrial membrane

Nearly two decades ago, Eiler and Schatz (10) demonstrated that MTX blocks the translocation of DHFR through the mitochondrial membrane by preventing the unfolding of the protein. Since then, MTX-bound DHFR has served as the benchmark molecule in protein translocation across mitochondrial membrane, as well as in protein degradation by ATP-dependent proteases. It is widely believed that, to fit into the narrow translocation or degradation channel, at least a portion of the folded protein must be converted to a threadlike conformation (14-16,46,47). However, there is little consensus on how this conversion occurs. Our single-molecule results show conclusively that ligand binding stabilizes DHFR mechanically. Therefore, if the rate-determining step in protein translocation were the force-induced unraveling of a protein, then our results would explain why the MTX-DHFR complex is resistant to translocation.

DHFR degradation by the proteasome

The significance of the finding that the mechanical stability of DHFR is dependent on ligand binding is best illustrated when considering proteasomal degradation. The proteasome degrades ubiquitin-tagged DHFR at a rate ~0.05 min^sup -1^ (12,23). In these experiments it was clearly established that the rate-limiting step was the unfolding of DHFR (23). Furthermore, as it had been shown earlier for translocation of DHFR across the mitochondrial membrane, addition of MTX reduced the rate of degradation by >10-fold (12,23). It has been recently proposed that a mechanical force generated by the proteasomal ATPase motor triggers the unfolding of the targeted protein (12,48). This model is equivalent to a medieval rack, where the ATPase motor pulls on DHFR against a ubiquitin chain bound to the proteasome, until DHFR unfolds. This model predicts that there should be a close correlation between the effect of MTX on the proteasomal degradation of DHFR, and its mechanical stability.

The proteasomal motor is an AAA ATPase. The AAA ATPases are ringlike hexameric motor proteins that are thought to convert conformational changes of the ring into a pulling force along a linear processive motion (50). Although there are no direct measurements of the mechanical capacity of the proteasomal motor, we use, as a model, the bacteriophage portal motor-a similar ringlike ATPase that also converts conformational changes of its ring into a linear translocation, and is capable of generating average forces of 57 pN (51). Fig. 5 shows the change in the cumulative unfolding probability of DHFR, triggered by the binding of MTX. At 57 pN, DHFR will be mostly unfolded (P^sub u^ = 0.87), whereas the DHFR-MTX complex will remain mostly folded (P^sub u^ = 0.13). It is interesting to note that the widest gap in the plot of Fig. 5 occurs precisely around the range of forces known to be generated by an AAA ATPase motor.

We thank Dr. Hongbin Li and Dr. Atom Sarkar for stimulating discussions.

The Chinese hamster DHFR gene is a generous gift from Professor L. A. Chasin. This work has been supported by National Institutes of Health grants to J.M.F. L.L. is a Damon Runyon Fellow (DRG-No. 1792-03).

© 2005 Biophysical Society Provided by ProQuest LLC. All Rights Reserved.