Discovery and mechanism of a pH-dependent dual-binding-site switch in the interaction of a pair of protein modules.

Many important proteins undergo pH-dependent conformational changes resulting in "on-off" switches for protein function, which are essential for regulation of life processes and have wide application potential. Here, we report a pair of cellulosomal assembly modules, comprising a cohesin and a dockerin from Clostridium acetobutylicum, which interact together following a unique pH-dependent switch between two functional sites rather than on-off states. The two cohesin-binding sites on the dockerin are switched from one to the other at pH 4.8 and 7.5 with a 180° rotation of the bound dockerin. Combined analysis by nuclear magnetic resonance spectroscopy, crystal structure determination, mutagenesis, and isothermal titration calorimetry elucidates the chemical and structural mechanism of the pH-dependent switching of the binding sites. The pH-dependent dual-binding-site switch not only represents an elegant example of biological regulation but also provides a new approach for developing pH-dependent protein devices and biomaterials beyond an on-off switch for biotechnological applications.

INTRODUCTION

The pH of a solution is important for the chemical reactions involved in life processes. pH-dependent conformational switches in proteins are involved in many biological processes, and pH-dependent biomolecules can also be developed as sensors and switches in biotechnology (–). Most pH-dependent conformational changes in proteins result in an “on-off” switch, which is useful in protein design and the development of functional biomaterials (, ). Discovery of new types of pH-dependent protein-protein interactions is of great value in both protein chemistry and biotechnological applications.

During the study of cellulosome assembly in Clostridium acetobutylicum, we discovered that the interaction of two assembling modules, a cohesin and a dockerin, named CaCohA2 and CaDoc0917, shows an unusual pH-dependent mode. The two cohesin-binding sites on the dockerin are switched from one to the other at pH 4.8 and 7.5, which is accompanied by a difference in affinity of two orders of magnitude; such a phenomenon has not been reported previously for any protein-protein interaction.

Cellulosomes are multienzyme complexes secreted by a number of anaerobic bacteria for lignocellulose degradation (). Cellulosomes are assembled using cohesin modules from scaffolding proteins (scaffoldins) and dockerin modules from enzymes and scaffoldins, and the two kinds of assembly modules are classified into types I, II, and III according to their sequences and functions in cellulosome assembly (). The cohesins and dockerins of C. acetobutylicum are annotated as type I modules, but they differ phylogenetically from known types I, II, and III cellulosomal assembly modules (), and their structures and interactions have not been extensively studied. In general, type I assembly modules have a dual-binding interaction mode, i.e., the dockerin has two symmetric sites to bind the cohesin in two alternative orientations by 180° rotation. These dual-binding interaction modes are considered to play roles in the plasticity and flexibility of cellulosomes (–). Nevertheless, a few exceptions were also discovered in which some cohesin-dockerin interactions show a single-binding mode, which was proposed to play regulatory roles for cellulosomes (). However, the pH-dependent feature has not been reported for cohesin-dockerin interactions.

We characterized the pH-dependent features of the CaCohA2-CaDoc0917 interaction, using a combination of nuclear magnetic resonance (NMR) and isothermal titration calorimetry (ITC). The structural basis and mechanism of the unusual pH-dependent features were elucidated by mutagenesis, crystal structure determination, NMR titration, ITC, and molecular dynamics (MD) simulation. This is the first report of a pH-dependent protein-protein interaction switch between two binding sites and has potential application in biotechnological development in the future.

RESULTS

The affinity between CaCohA2 and CaDoc0917 is pH dependent

To study the interaction between CaCohA2 and CaDoc0917, their thermodynamic parameters and stoichiometry were measured by ITC. The results (Table 1) indicated that CaDoc0917 interacts with CaCohA2 in a stoichiometric ratio of 1:1. The affinity between CaCohA2 and CaDoc0917 greatly varies with pH. The equilibrium binding constant (KA) at pH 4.8 is about 6 and 100 times greater than at pH 6.5 and 7.5, respectively. The KA of the CaCohA2-CaDoc0917 interaction at pH 4.8 is similar to those of the other reported type I cohesin-dockerin interactions (~107 to 108 M−1), such as the interaction between the second cohesin of ScaA (CipA) and the dockerin of the xylanase from Clostridium thermocellum (pH 7.5, 65°C) (), as well as that between the first cohesin of ScaA (CipC) and the dockerin of cellulase Cel5A from Clostridium cellulolyticum (pH 7.5, 35°C) (). We also measured the affinity of a pair of cohesin and dockerin molecules (CtCohA2, the second cohesin module of ScaA, and CtDoc48S, the dockerin module of Cel48S) from C. thermocellum (Table 1). The results indicate that they have similar affinities at both low and high pH. Therefore, the pH-dependent affinity of the C. acetobutylicum cohesin-dockerin modules is unusual compared with known cohesin-dockerin interactions.

Table 1

Thermodynamic parameters measured by ITC at 298 K.

	pH	KA (M−1)	ΔH (kcal/mol)	TΔS (kcal/mol)	N
CaCohA2/CaDoc0917	4.8	14.8 ± 2.4 × 107	−7.37 ± 0.04	3.78	0.950 ± 0.002
	6.5	2.4 ± 0.4 × 107	−4.67 ± 0.03	5.40	0.862 ± 0.003
	7.5	1.8 ± 0.1 × 106	−9.54 ± 0.13	−1.02	1.090 ± 0.011
CtCohA2/CtDoc48S	4.8	2.49 ± 1.51 × 109	−14.04 ± 0.06	−1.25	0.940 ± 0.002
	7.5	1.17 ± 0.95 × 109	−11.39 ± 0.06	0.98	0.924 ± 0.002

Solution structures of CaCohA2 and CaDoc0917 at pH 4.8

On the basis of the NMR assignments of CaCohA2 and CaDoc0917 (named CohA2 and DocA, respectively, in a previous study) at pH 4.8 (), we determined the solution structures of CaCohA2 and CaDoc0917 separately (Fig. 1 and table S1). The structure of CaCohA2 consists of nine β strands forming an elongated β barrel similar to the known type I cohesin structures (Fig. 1A and fig. S1). There is a helical insertion in β8 of CaCohA2, compared with which β8 of most other type I cohesins comprises a single elongated strand, except the cohesin (AcCohC3) from Acetivibrio cellulolyticus [Protein Data Bank (PDB) codes 4UYP and 4UYQ], which also contains a helical insertion in β8. The structure of CaDoc0917 consists of three helices, similar to the other known dockerin structures (Fig. 1B and fig. S1). Helix 1 and helix 3 are arranged in an antiparallel orientation, forming two F-hand motifs with two calcium-binding loops, respectively. Helix 2 is a 310 helix rather than an α helix; both α helices and 310 helices have been observed in helix 2 of other type I dockerins (–, ).

Fig. 1

Solution structures and pH-dependent binding-site switch of CaCohA2 and CaDoc0917.

(A) Cartoon representation of CaCohA2 structure. The β strands and loops are in green and gray, respectively. (B) Cartoon representation of the CaDoc0917 structure. The helices and loops are in red and gray, respectively. The calcium ions are shown as yellow spheres. (C) Superposition of helix 1 (yellow) and helix 3 (green) of the CaDoc0917 structure. (D) Sequence alignment of the N- and C-terminal parts of CaDoc0917. The sequences of helixes are shown in red. The calcium-binding sites are indicated by orange arrows. The key sites for species specificity and binding orientation are indicated by black arrows and a black rectangle. The important residues in the calcium-binding loop for the following NMR study are shaded by a green rectangle. (E and F) The 1H-15N heteronuclear single-quantum coherence spectroscopy (HSQC) spectral regions for the residues N14 and N47 of CaDoc0917 titrated with CaCohA2 at pH 4.8 (E) and pH 7.5 (F). (G) 1H-15N HSQC spectra of CaDoc0917 in complex with CaCohA2 at a molar ratio of 1:1 at different values of pH. ppm, parts per million.

Previous studies revealed that most dockerins have duplicated sequences to form a symmetric structure with two antiparallel helices and two calcium-binding loops, resulting in the dual cohesin-binding sites (–, ). The calcium cations are coordinated by residues n, n + 2, n + 4, n + 6 (main chain O atom), and n + 11 of the loops. Residues n + 9 and n + 10 in the calcium-binding loops are the key residues that characterize the binding site and species specificity of cohesin-dockerin interactions (). Residues n + 9 and n + 10 of CaDoc0917 calcium-binding loops are identical, Gly15Arg16 and Gly48Arg49 (Fig. 1, C and D), thus suggesting that the interaction of CaDoc0917 and CaCohA2 probably exhibits a dual binding mode. To simplify the subsequent description here, we define the binding sites using residues Gly15Arg16 and Gly48Arg49 as site 1 and site 2, and the corresponding orientations of CaDoc0917 in the complex as orientation 1 and orientation 2. Analysis of the CaDoc0917 structure indicates that other residues of helix 1 and helix 3 lack internal symmetry (Fig. 1, C and D), such as Arg22 and Ile55, which should be symmetric positions of the two binding sites. Therefore, the binding mode of CaDoc0917 and CaCohA2 required further experimental investigation.

The orientation of CaDoc0917 binding to CaCohA2 is pH dependent

NMR spectra of dockerins generally display good peak dispersion in 1H-15N heteronuclear single-quantum coherence spectroscopy (HSQC) spectra, and several peaks of the residues from the two calcium-binding loops of the dockerin are located in the downfield regions remote from the other peaks (, , ). Therefore, it is possible to use NMR titration to identify the cohesin-binding site on dockerin. Since previous studies have shown that most cohesin-dockerin pairs from C. thermocellum have a conserved dual-binding mode (), we first used CtCohA2 and CtDoc48S from C. thermocellum to validate the NMR titration method. NMR titration experiments indicated that two sets of peaks coexist in the solution when CtCohA2 and 15N-CtDoc48S form a complex in 1:1 molar ratio and under different pH conditions from 4.9 to 7.9 (fig. S2). Therefore, the NMR experiments provide direct evidence for the dual-binding mode of the CtCohA2-CtDoc48S interaction in solution, and the dual-binding mode exists at all measured pH values. These results also demonstrated that the NMR method is an excellent approach to investigate the binding mode of the cohesin-dockerin interaction in solution.

We then used NMR titration to study the binding of CaDoc0917 and CaCohA2. In the 1H-15N HSQC spectrum of CaDoc0917 at pH 4.8, peaks of N14 and N47 from the two calcium-binding loops are located in the downfield region. When CaCohA2 was gradually added into the solution of CaDoc0917 with a molar ratio of CaDoc0917:CaCohA2 from 1:0 to 1:1, the original peak of N47 became weaker and finally disappeared, and a new peak of N47 appeared with a large position shift (Fig. 1E). However, the peak of N14 only shifted slightly. The total peak number of the spectrum at the molar ratio 1:1 is almost the same as that of free CaDoc0917. Therefore, the result indicates that at pH 4.8, the interaction of CaDoc0917 and CaCohA2 displays a single-binding mode, using site 2 of CaDoc0917, i.e., in the complex, CaDoc0917 is in orientation 2.

When we used NMR titration to investigate the interaction of CaDoc0917 and CaCohA2 at pH 7.5, the N14 residue rather than N47 displayed a large chemical shift change during the titration (Fig. 1F). Therefore, at pH 7.5, the interaction between CaDoc0917 and CaCohA2 switched to a single-binding mode using site 1. To further confirm this pH-dependent binding-site change, we performed a pH titration using the CaCohA2 and 15N-CaDoc0917 complex at a ratio of 1:1 (Fig. 1G). The change of the peaks of N14 and N47 demonstrated that the cohesin-binding site on CaDoc0917 was gradually switched from site 2 to site 1 at pH values from 4.8 to 7.8, through a dual-binding mode at pH around 6.6. These results indicate that the two orientations of CaDoc0917 in the complex can exist in solution, but their populations vary with pH. Considering that the affinity of the CaDoc0917-CaCohA2 interaction is pH dependent according to the ITC experiments (Table 1) and the two binding sites are competitive in solution, we proposed a dynamic model for the observed phenomenon (Fig. 2A). According to this model, the two binding orientations have different affinity changes according to the pH, which resulted in the switching of the dominant orientation under different pH conditions.

Fig. 2

The two binding sites of CaDoc0917 have different affinity changes as a response to pH change.

(A) Proposed dynamic model of the pH-dependent features of the CaDoc0917-CaCohA2 interaction. An arbitrary sigmoid curve was used for the pH-dependent affinity change for each orientation. (B and C) NMR titrations of CaDoc0917 (left panels), CaDoc0917(R16D) (middle panels), and CaDoc0917(R49D) (right panels) with CaCohA2 at pH 4.8 (B) and pH 7.5 (C). The spectra of CaDoc0917:CaCohA2 at molar ratios of 1:0, 1:0.2, 1:0.4, 1:0.6, 1:0.8, and 1:1.0 are shown in black, blue, green, cyan, pink, and red, respectively. (D) Results of ITC experiments of CaDoc0917(R16D) and CaDoc0917(R49D) binding with CaCohA2 under different pH conditions.

Validation of the dynamic binding model using mutation of the Arg residue at the n + 10 positions of the calcium-binding loops

Although a single-binding mode has been observed in other cohesin-dockerin interactions, all of them have asymmetric key binding residues (i.e., the n + 9 and n + 10 positions of the calcium-binding loop) (). CaDoc0917 exhibits internal symmetry at these positions (Gly15Arg16 and Gly48Arg49) but showed a single-binding mode at both low and high pH, which raises the question of whether these residues still play key roles in the binding. Because Gly is a small residue without a side chain, the Arg residue is likely more important than the Gly residue in the interaction. We therefore constructed two mutants of CaDoc0917 in which one of the Arg residues was mutated to Asp to investigate the role of the Arg residues.

NMR titration experiments of the CaDoc0917 mutants with CaCohA2 indicated that CaDoc0917(R49D) bound to CaCohA2 using only site 1, owing to the large shift of the Asn14 peak, whereas mutant CaDoc0917(R16D) bound to CaCohA2 using only site 2, owing to the large shift of the Asn47 peak at both pH 4.8 and pH 7.5 (Fig. 2, B and C). Therefore, CaDoc0917 can bind to CaCohA2 in both orientations at both low and high pH with Arg16 and Arg49 as the essential residues.

Because both CaDoc0917 mutants bind CaCohA2 in a single orientation at both low and high pH, their binding affinities can be used to represent those of one orientation of the wild-type dockerin with CaCohA2 under different pH conditions. ITC experiments indicated that the KA between CaDoc0917(R16D) and CaCohA2 at pH 4.8 is about 600 times greater than that at pH 7.5, while the KA between mutant CaDoc0917(R49D) and CaCohA2 at pH 4.8 is about 31 times greater than that at pH 7.5 (Fig. 2D and Table 2). Therefore, the two orientations have pH-dependent affinities but respond differently to pH change. Orientation 2 exhibits higher affinity than orientation 1 at low pH but lower affinity at high pH, which validates the proposed dynamic model shown in Fig. 2A.

Table 2

Binding affinities of CaCohA2 and CaDoc0917 with mutations measured by ITC.

CaCohA2	CaDoc0917	KA (M−1) at pH 4.8	KA (M−1) at pH 7.5	Affinity ratio
Wild type	R49D	4.3 ± 0.3 × 107	13.9 ± 0.4 × 105	31
E128Q	R49D	2.1 ± 0.2 × 107	15.7 ± 0.7 × 105	13
D130N	R49D	2.1 ± 0.3 × 107	42.4 ± 1.8 × 105	5
E128QD130N	R49D	1.1 ± 0.1 × 107	23.2 ± 1.3 × 105	5
Wild type	R49DV19M	51.5 ± 21.9 × 107	19.8 ± 2.4 × 105	260
Wild type	R49DR22I	2.1 ± 0.2 × 107	1.8 ± 0.3 × 105	117
Wild type	R16D	10.3 ± 1.6 × 107	1.7 ± 0.3 × 105	606
E128Q	R16D	2.2 ± 0.3 × 107	1.2 ± 1.0 × 105	183
D130N	R16D	0.8 ± 0.1 × 107	3.2 ± 0.7 × 105	25
E128QD130N	R16D	0.36 ± 0.03 × 107	4.6 ± 0.3 × 105	8
Wild type	R16DM52V	2.9 ± 0.4 × 107	2.9 ± 0.5 × 105	100
Wild type	R16DI55R	13.2 ± 1.5 × 107	10.4 ± 0.5 × 105	127
Wild type	R16DE53Q	5.0 ± 0.9 × 107	19.3 ± 1.3 × 105	26

Overall structure of the complex of CaCohA2 and CaDoc0917(R16D/R49D) at low and high pH

Our attempt to obtain the crystal structure of the wild-type CaCohA2-CaDoc0917 complex was unsuccessful after extensive screenings. However, using the R49D and R16D mutants of CaDoc0917, we successfully obtained crystals and structures of the CaCohA2-CaDoc0917(R16D) and CaCohA2-CaDoc0917(R49D) complexes at pH 5.5 and 8.0, respectively. We further soaked the crystals of CaCohA2-CaDoc0917(R16D) obtained at pH 5.5 into high-pH buffers and, conversely, the crystals of CaCohA2-CaDoc0917(R49D) obtained at pH 8.0 into low-pH buffers, and lastly obtained also the crystal structures of CaCohA2-CaDoc0917(R16D) at pH 8.2 and CaCohA2-CaDoc0917(R49D) at pH 5.4. The four structures of the two orientations at both low and high pH were subsequently solved to resolutions of 1.6 to 2.3 Å (table S2).

The crystal structures of CaCohA2-CaDoc0917(R49D) at pH 5.4 and pH 8.0 consist of one heterodimer complex in one asymmetric unit (ASU). The crystal structures of CaCohA2-CaDoc0917(R16D) at pH 5.5 and pH 8.2 consist of two heterodimer complexes in one ASU. As expected, the binding of CaDoc0917(R49D) and CaDoc0917(R16D) to CaCohA2 adopted orientation 1 and orientation 2, respectively (fig. S3). The binding of CaDoc0917(R16D) to CaCohA2 in the two heterodimer structures in one ASU shows a slight orientation difference, suggesting that the complex in orientation 2 may be less rigid than in orientation 1 (figs. S3B and S4). In the four crystal structures, the binding interface of CaCohA2 involves the β3, β5, β6, and β8 strands, which is similar to other cohesin-dockerin complexes from C. thermocellum, C. cellulolyticum, and A. cellulolyticus (–, , , ). However, if the structures of the complexes from different species are superimposed by the cohesins, the orientations of the dockerins showed considerable differences (fig. S3, F and G). This is reasonable because the residues at the binding surfaces of both cohesin and dockerin are largely not conserved in different species (fig. S5). The overall structures of the CaDoc0917 mutants and CaCohA2 in the structures of the complexes are essentially identical to the corresponding individual solution structures of the wild-type CaDoc0917 and CaCohA2. Therefore, the affinity difference of these complexes appears to be caused by the detailed interactions at the interface.

Structural differences of the complexes at low and high pH: The origin of pH-dependent affinity

By comparing detailed interactions for orientation 1 and orientation 2 under different pH conditions, we found that the major differences are the polar interactions of the Asp127-Asp132 loop of CaCohA2, while the other interactions are almost identical at low and high pH. The side chain conformations of the Asp127-Asp132 loop vary greatly at high and low pH (Fig. 3, A and B). The electron densities of the side chains of some residues in this loop are observed at low pH but not under high-pH conditions for either orientation. Therefore, the loop should have dynamic changes dependent on low to high pH values. The Asp127-Asp132 loop contains four negatively charged residues, the neighboring charged residues of which may have pKa shifts according to previous studies (, ). Because the pKa value of a specific residue in the clustered charged residues is difficult to measure experimentally, we performed state-of-the-art continuous constant pH MD (CpHMD) simulations () to estimate the pKa values of interfacial charged residues. The results revealed that Asp84, Glu128, Asp130, Asp132, and Asp135 have substantial pKa shifts in both isolated CaCohA2 and CaCohA2 in complex with CaDoc0917 (table S3), which leads to their protonation at low pH. On the other hand, Asp135 neither participates in the interaction with CaDoc0917 nor exhibits any conformational change in the structures at different pH values. Thus, it is unlikely that this residue plays an important role in the pH-dependent interaction. The protonated Glu128 forms a hydrogen bond with Asp84 and could possibly stabilize the conformation of the loop. When the pH was increased, Asp84 and Glu128 became negatively charged, which resulted in the loss of the hydrogen bond and high dynamic conformation of the loop. These changes result in weakened interactions between the loop and the dockerin at high pH compared with those at low pH and explains the observed change in affinity under the different pH conditions. Asp130 participates in the interaction with CaDoc0917 by forming a hydrogen bond, which could be weakened by the dynamic increase along with its deprotonation.

Fig. 3

Dynamic changes of the Asp127-Asp132 loop of CaCohA2 and differences in interaction between the two orientations.

(A) Region of the Asp127-Asp132 loop in the structures of the CaDoc0917(R49D)-CaCohA2 complex at pH 5.4 and pH 8.0. (B) The region of the Asp127-Asp132 loop in the structures of the CaDoc0917(R16D)-CaCohA2 complex at pH 5.5 and pH 8.2. Electron densities of each structure are contoured at 1.0σ. (C and D) Superposition of three structures: The CaDoc0917(R49D)-CaCohA2 complex [orientation (ori.) 1] and two CaDoc0917(R16D)-CaCohA2 complexes in one ASU (orientations 2-1 and 2-2) at low pH (5.4 or 5.5) and high pH (8.0 or 8.2). Key residues are shown as sticks, and water molecules (wt) are shown as balls. The sites of Glu53/Gln23 and Met52/Val19 are shown in (C), and the sites of Arg22/Ile55 are shown in (D).

Because Asp84 and Asp135 do not directly interact with CaDoc0917 and are partially buried, we propose that Glu128 and Asp130 comprise the major residues that play key roles in the pH-dependent affinity. This was further validated by mutagenetic analysis. ITC experiments indicated that mutation of Glu128 or Asp130 to a neutral residue can dramatically reduce the difference in affinity at low and high pH for both orientations (Table 2). The ratio of affinity for the orientation 1 complex [i.e., the CaCohA2 versus the CaDoc0917(R49D) complex] at low and high pH is 35, compared with 606 for the orientation 2 complex [i.e., the CaCohA2 versus the CaDoc0917(R16D) complex]. Mutation of Glu128 to Gln reduced the affinity ratios to 13 and 183, and the D130N mutation reduced the affinity ratios to 5 and 25, for the orientation 1 and orientation 2 complexes, respectively. With the double mutation E128Q/D130N, the affinity ratios decreased to 5 and 8 for the two orientations. Therefore, the pKa shifts of Glu128 and Asp130 are the major sources of the observed pH-dependent affinity.

Interfacial differences between two orientations at the same pH: The structural basis of the pH-dependent orientation switch

As we have demonstrated that the pH-dependent affinity for the two orientations is mainly determined by the pKa shifts of Glu128 and Asp130 of CaCohA2, the structural basis of the pH-dependent orientation switch was further analyzed. The pH-dependent switch should reflect the difference in interaction between CaCohA2 and the two orientations of the dockerin, since CaDoc0917 has two asymmetric cohesin-binding sites. By comparing the structures of the complexes for each orientation at low or high pH, we observed similar hydrophobic interactions between CaCohA2 and Leu51/Ile18 of CaDoc0917 and three major differences at the binding interface, which are the interactions between CaCohA2 and the asymmetric Glu53/Gln23, Met52/Val19, and Ile55/Arg22 of CaDoc0917 (Fig. 3, C and D).

Glu53 and Gln23 of CaDoc0917 are not at symmetric positions in the sequence (Fig. 1D), but they form similar hydrogen bonds with Asp130 of CaCohA2 in the low-pH structures of the two orientations. At high pH, the hydrogen bond between Asp130 and Gln23 in orientation 1 is weakened or lost, but Asp130 and Glu53 in orientation 2 not only completely lose the hydrogen bond between them but become repulsive owing to deprotonation of Asp130. This difference causes the observed increase in affinity ratio of orientation 2 at low and high pH, compared with that of orientation 1. Mutation of Glu53 to Gln in CaDoc0917(R16D) resulted in a slight change of affinity with CaCohA2 at low pH but a markedly increased affinity (about 11.4-fold) at high pH (Table 2), which confirms the repulsive interaction between Glu53 and Asp130 at high pH. The affinity ratio at low and high pH for the interaction between CaCohA2 and CaDoc0917(R16DE53Q) becomes 26, similar to the ratio (31) for CaCohA2 and CaDoc0917(R49D). Therefore, the Glu53 and Gln23 residues at asymmetric positions of CaDoc0917 are responsible for the difference in affinity ratio for its two binding orientations.

Met52 and Val19 of CaDoc0917 are at the equivalent positions in orientation 2 and orientation 1 complexes, respectively, and form hydrophobic interactions with CaCohA2. In the orientation 2 complex, Met52 of CaDoc0917(R16D) is completely buried by Leu36, Leu87, Thr83, and the backbone atoms of residues Gly35 and Glu128-Asp130 of CaCohA2, forming a strong hydrophobic interaction. However, in the orientation 1 complex, Val19 of CaDoc0917(R49D) is surrounded by several water molecules in addition to the hydrophobic residues Leu36 and Leu87 of CaCohA2, suggesting that its side chain may be too short to form strong hydrophobic interactions with CaCohA2. It is interesting that the Met52 and Val19 sites are also surrounded by Glu128-Asp130 of CaCohA2, thus providing additional pH-dependent effects on the interaction. At low pH, the rigid Glu128-Asp130 region protects the strong hydrophobic interaction of Met52 and CaCohA2, resulting in substantial affinity differences between Met52 and V19 in binding to CaCohA2. However, at high pH, the Glu128-Asp130 region becomes more dynamic and provides less protective effects on the hydrophobic interactions of Met52/Val19, making the difference between Met52 and Val19 obscure for binding to CaCohA2. ITC experiments indicated that at low pH, the V19M mutation of CaDoc0917(R49D) increases the affinity by about 12-fold, while the M52V mutation of CaDoc0917(R16D) decreases the affinity by a factor of 3.6. However, at high pH, both mutants showed only slight changes in binding affinity (less than twofold) (Table 2). Therefore, the difference in hydrophobic interaction between Val19 and Met52 plays a role in the difference in affinity between the two dockerin orientations at low pH.

Arg22 and Ile55 are also at equivalent symmetrical positions of CaDoc0917 and are involved in the interaction with CaCohA2 in both orientation 2 and orientation 1 complexes. Both residues form similar hydrophobic interactions with CaCohA2 in both orientations, but Arg22 has different polar interactions in the two orientations. Two additional hydrogen bonds are formed by Arg22 of CaDoc0917(R49D) and Tyr66/Thr83 of CaCohA2 in the orientation 1 complex at both low and high pH. However, in the orientation 2 complex, Arg22 forms either a water-mediated hydrogen bond with Tyr66/Thr83 of CaCohA2 or a cation-pi interaction with Tyr66 of CaCohA2 in only one of the two structures in one ASU, which suggests that the interactions are weak. ITC experiments using CaDoc0917(R49D/R22I) and CaDoc0917(R16D/I55R) showed that the asymmetric Arg22/Ile55 residues at the symmetric site benefit the orientation 1 interaction only slightly at low pH but has greater effects at high pH (Table 2).

On the basis of the above results of structure determination and interaction analysis, we propose a model for the pH-dependent affinity and binding-site switch of the CaDoc0917 interaction with CaCohA2 (Fig. 4). The pKa shifts of Glu128 and Asp130 in CaCohA2 are the source of the observed pH-dependent affinity. At low pH, Glu128 and Asp130 are protonated, and the Asp127-Asp132 loop is rigid to protect the hydrophobic interactions between the two proteins. At high pH, however, Glu128 and Asp130 become negatively charged, which makes the loop more dynamic and less protective for the hydrophobic interaction, resulting in lower affinity at higher pH. The two asymmetric binding sites on CaDoc0917 are the source of the pH-dependent switch between the two orientations because they have a different response to the pH-induced change of the CaCohA2 Asp127-Asp132 loop. The different response of the two binding sites reflects the combined effect of different hydrophobic, hydrogen-bonding, and electrostatic interactions.

Fig. 4

Cartoon model of the pH-dependent affinity and binding-site switch of the interaction between CaDoc0917 and CaCohA2.

(A) CaCohA2 exhibits pH-dependent flexibility and changes in charge reflecting pKa shifts of Glu128 and Asp130. CaDoc0917 has two asymmetric sites, which contribute to cohesin binding (see text for details). (B) Differences between the two orientations at low and high pH, resulting in the pH-dependent binding-site switch.

DISCUSSION

In this study, we found that a pair of cohesin and dockerin modules (CaCohA2 and CaDoc0917) from C. acetobutylicum displays a pH-dependent binding-site switch by an elaborate multiplicity of interactions. The pH dependence originates from a negatively charged loop of CaCohA2 at the interface with the dockerin, in which Glu128 and Asp130 exhibit remarkable pKa shifts to values near 7.0. The two binding sites on CaDoc0917 are asymmetric and have a differential response to the pH-induced charge and conformational changes of the Asp127-Asp132 loop. The different response of the two binding sites on the dockerin is caused by the combined effects of different hydrophobic, hydrogen-bonding, and electrostatic interactions, which are more complicated and elaborate than previous reports that detail pH-dependent conformational changes and on-off switches. Sequence analysis of all cohesins and dockerins in the genome of C. acetobutylicum () indicated that the pH-dependent affinity of CaCohA2 should be conserved in all cohesins of ScaA of C. acetobutylicum, but the other eight dockerins lack the asymmetric features of CaDoc0917 (fig. S6). Therefore, the pH-dependent binding-site switch of the CaDoc0917-CaCohA2 interaction is unique among the cellulosomal assembly modules of C. acetobutylicum.

Our structure and MD analysis have revealed the mechanism of the pH-dependent binding-site switch in the CaCohA2-CaDoc0917 interaction. In principle, changes in the protonation state of the interfacial residues between the two protein modules can affect their interaction, thereby resulting in a pH-dependent phenomenon. Such a phenomenon is observed frequently when a histidine residue is located in the interface between two proteins, because of its pKa at about 6.0 (). However, if no histidine exists in the interface and a pH-dependent interaction occurs between pH 5 and 8, one would expect the occurrence of pKa shifts, usually in aspartic and glutamic acids, which is the case observed here in the CaCohA2-CaDoc0917 interaction. The pKa shifts in protein-protein interactions could occur in either the unbound or bound form, and the latter (i.e., pKa changes along with the binding interaction) has been studied more thoroughly (). However, the pKa shifts in the CaCohA2-CaDoc0917 interaction belong to the former type, i.e., the pKa shifts exist in the unbound form protein, and no notable pKa change occurs during binding. Therefore, the CaCohA2-CaDoc0917 interaction provides an interesting example for studying the different mechanisms of pH-dependent protein-protein interactions.

The unique pH-dependent feature of cohesin-dockerin modules from C. acetobutylicum reported in this study provides a previously unknown diversification in cellulosome assembly. A previous study demonstrated that the cohesin and dockerin modules from C. acetobutylicum exhibit species specificity, i.e., they have no interaction with modules from other cellulosome-producing species (), which makes them potentially useful in the construction of designer cellulosomes. The species specificity can be explained by the structures reported in this study because the key interaction of cohesin-dockerin modules from C. acetobutylicum is based on electrostatic interactions, which are different from the key hydrophobic or hydrogen-bonding interactions in C. cellulolyticum and C. thermocellum. The pH-dependent features of the cohesin and dockerin modules from C. acetobutylicum will provide novel and interesting modular components for assembly of designer cellulosomes and for heterologous applications. More broadly, the pH-dependent feature of CaCohA2 and CaDoc0917 can potentially be used in synthetic biology and biotechnological development. Thus, the pH-dependent binding-site switch provides a new approach to design a more complex signal switch than a simple on-off switch based on a single pH-dependent conformational change. Likewise, these features can be applied to the field of functional materials, where the pH-dependent properties of CaCohA2 and CaDoc0917 could be further improved by protein engineering to generate higher affinity or a more complete switch.

C. acetobutylicum is a well-known solventogenesis bacterium for ABE (acetone, butanol, and ethanol) fermentation (). The growth of this bacterium has been characterized to have two stages under different pH conditions, i.e., an acidogenesis stage at relatively high pH values of about 6 and a solventogenesis stage at low pH of about 4.5 (). Intriguingly, C. acetobutylicum cannot use cellulose, as the secreted cellulosomes have very low activity, and the reason for this enigma is still unknown (). It is not clear whether the pH-dependent feature of the cellulosomal modules is related to the two pH-related growth stages. In this context, it is interesting to speculate whether the higher affinity of the cohesin-dockerin interactions benefits cellulosome stability at the low-pH solventogenesis stage with high ABE concentration. Further work is required to address this premise after future clarification of the possible physiological roles of the cellulosome in C. acetobutylicum.

MATERIALS AND METHODS

NMR structure determination of CaCohA2 and CaDoc0917

The procedure of protein expression, purification, and NMR chemical shift assignments of CaCohA2 and CaDoc0917 has been published previously (). NMR samples consisted of 0.6 mM protein in 20 mM sodium acetate buffer (pH 4.8) with 10% (v/v) D2O, 100 mM KCl, 1 mM CaCl2, 0.01% (w/v) NaN3, and 0.05% (w/v) sodium 2,2-dimethylsilapentane-5-sulfonate (DSS). The initial structures were calculated using the CANDID module of the program CYANA () with the peak lists from the 15N-edited and 13C-edited nuclear Overhauser effect (NOE) spectroscopy (NOESY)–HSQC datasets and the dihedral angle restraints predicted by the program TALOS-N () using the chemical shifts. The structures were refined using the program Crystallography & NMR System (CNS) () and RECOORDScripts () with the distance restraints derived by the semiautomatic NOE-assignment program Structure Assisted NOE Evaluation (SANE) () and the dihedral angle restraints. The distance restraints for calcium ions in CaDoc0917 (2.4 ± 0.2 Å between calcium and coordinative oxygen atoms) were introduced according to the homologous dockerin structures and the sequence alignments. Hydrogen bond restraints were also introduced according to the secondary structure elements in the late stage of the refinement. A family of 100 structures was generated by CNS, and 50 structures with the lowest energies were subjected to the refinement in explicit water using the program CNS and RECOORDScript. A set of 20 structures with the lowest energies was selected to analyze the violations and the structure quality. The iterations of the refinement procedures were performed until there were no NOE violation greater than 0.2 Å and no dihedral angle violation greater than 5°.

Plasmid construction and protein purification

The CaCohA2 and CaDoc0917 proteins used for NMR structure determination have the two additional cloning artificial residues (Thr-Ser) at the C terminus (), which could potentially interfere with the interaction between CaCohA2 and CaDoc0917. Therefore, new plasmids were constructed to express the proteins without these C-terminal artificial residues. The DNA fragments of CaCohA2 (UniProt Q977Y4, residues 611 to 759) and CaDoc0917 (UniProt Q97KK2, residues 477 to 537) were amplified by polymerase chain reaction (PCR) from the previous plasmids pET28aNS-CaCohA2 and pET28aNS-CaDoc0917, respectively. The DNA fragments were digested by the Bam HI and Xho I restriction enzymes and then ligated into the vector pET28a-SMT3 (), forming the plasmids pET28a-SMT3-CaCohA2 and pET28a-SMT3-CaDoc0917. CaCohA2 and CaDoc0917 mutants were constructed using the QuikChange Site-Directed Mutagenesis protocol (). The encoded target proteins in these plasmids contain an N-terminal His6-SMT3 tag, which can be removed by ULP1 protease treatment in the purification procedures.

The CaDoc0917 and CaCohA2 proteins were expressed in Escherichia coli Rosetta(DE3) and E. coli BL21(DE3) cells, respectively. The cells were cultured overnight in 20 ml of LB medium at 37°C and then amplified into 1 liter of LB medium with kanamycin sulfate (50 mg/ml). When the optical density at 600 nm reached 1.0 to 1.2, the pH of the culture was adjusted to ~7.5 using NaOH solution, and then 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) was added into the culture to induce the protein expression. The cells were grown for an additional 4 to 6 hours and harvested by centrifugation at 6000g at 4°C for 2 min. Cell pellets were resuspended in 20 ml of buffer A (50 mM tris and 500 mM NaCl, pH 8.0) and lysed by ultrasonication. The proteins were applied on a His-trap HP column (GE healthcare) and were then eluted with buffer A containing 500 mM imidazole. The eluate was exchanged with buffer A and then treated with ULP1 protease for 30 min at 25°C. The solution was then passed through the His-trap HP column again to remove the cleaved His6-SMT3 tag and the ULP1 protease. The flow-through solution was concentrated into 2 ml and was then applied on a Superdex 75 column (GE Healthcare) preequilibrated in a buffer containing 20 mM sodium acetate/acetic acid (pH 4.8), 100 mM KCl, and 5 mM CaCl2 or a buffer with 20 mM tris-HCl (pH 7.5), 100 mM KCl, and 5 mM CaCl2. Fractions containing target proteins were collected and concentrated to 0.5 to 1.0 ml. The mutants were expressed and purified following the same procedures. Protein concentration was determined by the ultraviolet absorption at 280 nm using a theoretical molar extinction coefficient.

The 15N-labeled CtDoc48S protein was purified by affinity chromatography and gel filtration, as described in a previous paper (). The DNA fragments of CtCohA2 (residues 183 to 322 of CipA, GenBank no. WP_020458017.1) were amplified by PCR from the genome of C. thermocellum ATCC 27405. The DNA fragments were digested by the Eco RI and Xho I restriction enzymes and then ligated into the vector pET28a, forming the plasmid pET28a-CtCohA2. The CtCohA2 proteins were expressed and purified following the same procedures as CaCohA2, except that the N-terminal His6-tag of CtCohA2 was not removed.

Isothermal titration calorimetry

The affinity between CaCohA2 and CaDoc0917 or their mutants was measured by using an isothermal titration calorimeter (VP-ITC, Microcal Northampton, MA) at 25°C. Protein solutions for ITC were dialyzed overnight against the same buffer. The low-pH buffer contained 20 mM sodium acetate/acetic acid, 100 mM KCl, and 5 mM CaCl2, at pH 4.8. The high-pH buffer contained 20 mM tris-HCl, 100 mM KCl, and 5 mM CaCl2, at pH 7.5. The ligand protein solution (28 drops, 10 μl per drop) was added from a rotating stirrer syringe to the reaction cell with the protein concentrations of 10 μM in the cell and ligand protein concentrations of 100 μM in the syringe. Calorimetric data analysis was carried out with Origin 7.0 (OriginLab Corporation, USA). Binding parameters were determined by fitting the experimental binding isotherms using a single-site model.

NMR titration experiments

The 15N-labeled proteins for NMR titration experiments were expressed in M9 medium containing 15N-labeled NH4Cl as the sole nitrogen source and purified by the same procedures for nonlabeled proteins. The NMR titration experiments were performed on a Bruker 600-MHz NMR spectrometer at 298 K with standard 1H-15N HSQC pulse program. The low- and high-pH buffers same as those in the ITC experiments were used in the NMR titration experiments, except that 10% D2O was added to the buffer to provide the NMR field-locking signal. For pH titration experiments, the pH of the sample was adjusted from pH 4.8 by gradually adding a small amount of 1 M tris solution at pH 9.0, and an InLab Micro pH meter (Mettler-Toledo) was used to monitor the pH changes.

Crystallization, data collection, structure determination, and refinement

The protein complexes CaDoc0917(R16D)-CaCohA2 and CaDoc0917(R49D)-CaCohA2 were prepared by gel filtration using a mix of the dockerin and cohesin with a molecular ratio of 1.5:1.0. The fractions of the protein complex were concentrated to ~20 mg/ml in 10 mM Hepes buffer with 10 mM NaCl at pH 7.0. Crystallization conditions of the complex were screened by using commercial screening kits and NT8 and were then optimized in 24-well crystallization plates. Crystals of the CaDoc0917(R16D)-CaCohA2 complex were obtained with 0.2 M MgCl2, 0.1 M sodium acetate, and 25% (w/v) PEG 3350 at pH 5.5. Crystals of the CaDoc0917(R49D)-CaCohA2 complex were obtained with 0.2 M sodium acetate, 0.1 M tris-HCl, and 30% (w/v) PEG4000 at pH 8.0. The crystals of CaDoc0917(R16D)-CaCohA2 at high-pH condition were obtained by soaking the crystals grown at pH 5.5 into a buffer containing 0.2 M sodium acetate, 0.1 M tris-HCl, and 30% (w/v) PEG 4000 at pH 8.2 for 10 min. The crystals of CaDoc0917(R49D)-CaCohA2 at low-pH condition were obtained by soaking the crystal grown at pH 8.0 into a buffer containing 0.2 M MgCl2, 0.1 M sodium acetate, and 25% (w/v) PEG 3350 at pH 5.4 for 10 min. The data were collected at the Shanghai Synchrotron Radiation Facility (SSRF), beamlines BL17U1 and BL19U1 (–), in a 100-K nitrogen stream. Data indexing, integration, and scaling were conducted using X-ray Detector Software (XDS) (). The initial phase was determined by molecular replacement using PHENIX (), and the search models for CaCohA2 and CaDoc0917(R16/49D) were the isolated cohesin and dockerin structures, respectively, from the structure of a cohesin-dockerin complex from C. thermocellum (PDB 1OHZ) (). Structure refinements were iteratively performed using the programs COOT () and PHENIX.

MD simulation

The x-ray crystal structures CaDoc0917(R16D)-CaCohA2 at pH 5.5 and CaDoc0917(R49D)-CaCohA2 at pH 8.0 were used as the starting models in simulation. The replica-exchange continuous CpHMD simulation () implemented in Amber 18 () was used to calculate the pKa of interfacial residues. The procedure of parameter setting and calculations followed the manual of Amber18 and the literature (, , ). Each system was equilibrated by 1000 steps of energy minimization and followed by 4-ns pH replica-exchange CpHMD simulation. The pH ranged from 2.5 to 8, with an interval of 0.5, and replicas attempted a pH exchange every 250 MD steps. The pKa values were obtained by fitting the data to the Henderson-Hasselbalch equation.