Single-molecule diffusometry reveals the nucleotide-dependent oligomerization pathways of Nicotiana tabacum Rubisco activase

The oligomerization state of many proteins is critical to their function.¹ It forms the basis for substrate recognition for numerous enzymes² and underpins the rich biophysical phenomena of allostery³ and cooperativity.^4,5 On the other hand, erroneous aggregation is involved in the pathogenesis of various amyloid diseases (e.g., Alzheimer’s, Parkinson’s, Huntington’s, etc.).^6,7 Understanding the mechanism and pathways by which proteins assemble into higher-order structures is of fundamental and biomedical significance. Several biochemical and biophysical methods are available to examine the oligomerization behavior of proteins but all have certain limitations. Methods based on liquid chromatography⁸ and sedimentation^9,10 result in at least partial fractionation of the sample, an effect that is not suitable for complexes that undergo rapid exchange during measurements. Scattering-based techniques like small angle X-ray scattering (SAXS)¹¹ and dynamic light scattering (DLS)¹² rely on model-dependent fitting, which can be error-prone for highly polydispersed and/or interacting samples. Recently, native mass spectrometry has shown great promise in characterizing the quaternary structure of proteins, but the degree of perturbation introduced by electrospraying macromolecules into the gas phase is still debated.¹³ 

On the other hand, useful methods with single-molecule sensitivity have been developed. Electron microscopy offers structural information at the single-particle level but requires samples to be moved out of equilibrium onto a surface and dried in a vacuum. Recently, fluorescence correlation spectroscopy (FCS) has been applied to resolve oligomerization states of proteins^14,15 in solution, but this method has limited resolution for a mixture of assembly states.¹⁶ Other techniques, such as two-color coincidence detection,¹⁷ stepwise photobleaching,¹⁸ and single-molecule Förster resonance energy transfer (FRET),¹⁹ have been applied to characterize oligomerization under the assumption that the perceived fluorescence brightness signal reflects differences in oligomer size. Here, we describe a more direct method to measure the oligomer distribution of proteins in solution, with single-molecule resolution under equilibrium conditions. The method is based on measuring the hydrodynamic radius of individual molecular oligomers contained in an aqueous-phase Anti-Brownian ELectrokinetic (ABEL) trap in real time (we call it “ABEL-oligo”). We use this method to dissect the oligomerization pathway of tobacco Rubisco activase, a catalytic chaperone essential for carbon fixation.

Rubisco activase (Rca) is a chaperone-like, AAA+ ATPase that rescues the main carbon fixation enzyme Rubisco [ribulose-1,5-bisphosphate (RuBP) carboxylase/oxygenase] from premature inhibition.^20–22 Despite its importance, the detailed mechanism of Rubisco activation by Rca is poorly understood, in part due to the complicated oligomerization behavior of Rca observed in vitro. Previous biophysical characterization of several higher plant Rca proteins in vitro using a suite of established methods⁸ revealed a highly heterogeneous and transient oligomer distribution, but these approaches could not be used to establish a quantitative oligomerization pathway. Surprisingly, sedimentation velocity experiments on tobacco (Nicotiniana tabacum, Nt) Rca^9,10 suggested that the oligomerization of Nt-Rca does not depend on nucleotide. On the contrary, the assembly of spinach Rca was shown to be strongly dependent on the type of nucleotide,²⁰ consistent with earlier work reporting that both ATP and ATPγS promote the formation of higher-order oligomers (MW > 550 kDa), whereas ADP does not.²³ Additionally, the oligomerization behavior of cotton β-Rca, as monitored by FCS, appeared to be a function of nucleotide.²⁴ Meanwhile, several structural studies suggested hexamers as the functional species^22,25,26 [Fig. 1(b)]. Here we set out to apply our ABEL-oligo method to define and characterize the assembly pathway of Nt-Rca in vitro and examine the effects of ATP binding (with Mg·ATPγS) and hydrolysis (with Mg·ADP).

The gene coding for Nicotiana tabacum (tobacco, Nt) Rca (383 amino acid residues, MW 42.764 kDa, theoretical PI 5.5) was transferred into the pHUE plasmid^27,28 following the published procedures.²⁹ The QuikChange mutagenesis kit (Agilent Genomics) was used to introduce the S379C substitution near the C-terminus of Nt-Rca, as confirmed by DNA sequencing. Nt-Rca-S379C was expressed in E. coli BL21*(DE3) (Invitrogen) as a 6His-tagged ubiquitin fusion protein and purified via affinity chromatography on a Ni²⁺-nitrilotriacetic acid (Ni-NTA) column (Qiagen). Subsequently, the N-terminal 6His-ubiquitin tag was cleaved using a deubiquitylating enzyme, the sample was dialyzed and reapplied to a Ni-NTA column, as described previously.²⁹ Protein fractions were spin-concentrated and buffer-exchanged into 25 mM HEPES pH 7.5, 250 mM KCl, 5 mM MgCl₂, 10% glycerol, and 2 mM ADP. Protein preparations were quantitated using the Bradford method with bovine serum albumin as a standard. Aliquots were flash frozen in liquid N₂ and stored at −80 °C.

The conjugation of the Alexa Fluor 647 C₂-maleimide fluorophore (Molecular Probes, Eugene, OR) to the engineered cysteine residue of Nt-Rca-S379C was carried out as described previously.^24,29 Briefly, labeling was carried out in the presence of 5 mM ADP at a dye:protein ratio of 3.5 in HEPES buffer pH 7.2. In control reactions containing wild-type Nt-Rca instead of Nt-Rca-S379C, the covalent labeling of protein was not observed (Fig. S1 of the supplementary material), indicating that the cysteine residue in position 379 is specifically targeted by the maleimide functional group. Alexa-labeled Nt-Rca-S379C was purified by gel filtration, and then analyzed by reverse-phase high-performance liquid chromatography (HPLC) (Agilent Technologies 1100) on a C18 analytical column (Vydac 218 TP) using a linear water/acetonitrile gradient with 0.1% trifluoroacetic acid (TFA). Elution of protein and dye was monitored by optical density (OD) at 220, 280, and 650 nm. Protein eluted at 46.9% acetonitrile as a single peak and carried the Alexa-dye label, whereas free Alexa-dye was not detected (expected elution at 16.4% acetonitrile). Alexa/protein ratios were determined by absorbance measurements on collected HPLC fractions as described previously and provided a 1:1 molar ratio of protein to dye.¹⁴ Aliquots of labeled (88.2 μM) and label-free (268.0 μM) Nt-Rca-S379C in buffer containing 25 mM HEPES–NaOH, pH 7.5, 250 mM KCl, 2 mM ADP, 5 mM MgCl₂, 10% glycerol were flash frozen in liquid N₂ and stored at −80 °C.

On the day of the experiment, samples were thawed on ice. Labeled Rca was first diluted to 3 nM concentration in a buffer containing no Mg²⁺ or nucleotide (50 mM HEPES, pH 7.5, 150 mM KCl) and kept on ice for ∼20 min. Subsequently, a reaction mixture containing ∼10 pM labeled Rca and various amounts (specified in the Results section) of unlabeled Rca was created in the reaction buffer containing Mg²⁺ and nucleotide (50 mM HEPES, 150 mM KCl, 3 mM Trolox, 0 or 5 mM MgCl₂, 0 or 2 mM nucleotide, 10% glycerol). The mixture was incubated on ice for ∼20 min, deoxygenated (see below), and injected into the ABEL trap for single-molecule measurements. For cotton Rca, the 20 min incubation time has been shown to be sufficient to reach equilibrium.¹⁴ The concentration of labeled Rca was kept at ∼10 pM in all measurements. In experiments with 2 mM ATPγS, the final solution also contained between 0 and 80 μM ADP, which was carried over from the storage buffer of the unlabeled Rca preparation. Likewise, the low-ADP condition contained between 0 and 80 μM ADP.

The ABEL trap was implemented as previously described.³⁰ Briefly, the apparatus consists of a position-sensing module, a signal-processing unit, and a microfluidic sample holder. To sense the position of a single labeled biomolecule or oligomer in real time, we used fast laser scanning in a deterministic pattern and photon-by-photon molecule localization defined by the position of the laser spot at the moment of detection of a fluorescence photon. The signal-processing unit computes the most-probable position of the diffusing molecule given all available physical information. It is implemented on a field programmable gate array (FPGA) board for minimum computational delay. Given estimated position values, feedback voltages were generated, amplified, and applied to the solution in a microfluidic sample holder to compensate for the Brownian motion of a single molecule. Our ABEL trap suppresses the Brownian motion of a single molecule in 2D (xy) only; diffusion in z is confined by the axial thickness of the microfluidic cell to ∼700 nm.

The diffusion coefficient and electrokinetic mobility of individual trapped molecules were extracted using a maximum-likelihood estimator as previously described.³¹ This works even though the molecule is prevented from diffusing away because at any given time, the molecule is still driven by thermal diffusion as well as by the fields used for trapping. Briefly, the algorithm uses the photon-stamped position sequence of a single trapped molecule and the feedback voltages as inputs and utilizes an iterative procedure to reconstruct the motion trajectory. It then separates the in-trap motions of a single molecule into a voltage-dependent component and a stochastic component to estimate electrokinetic mobility and diffusion coefficient, respectively. The algorithm can be configured to run in real time, or with higher precision offline, with a tunable time resolution between 50 ms and ∼1 s. The estimated diffusion coefficient has a Gaussian distributed uncertainty that scales with $1/Nph$ (N_ph being the number of photons detected during the estimation window; supplementary material). In general, both single-molecule hydrodynamic radius (from diffusion coefficient) and charge (from electrokinetic mobility) information can be extracted, as previously demonstrated for DNA hybridization.³¹ In this work, the electrokinetic mobility is dominated by the electro-osmotic response of the bulk solution [resulting from the polyelectrolyte multi-layer coating (see below) to prevent protein sticking] and contains little information about protein charge. We thus utilized only the diffusion information to identify the different oligomers.

We used the 633 nm line from a HeNe laser as the excitation source with an average excitation intensity of 1.5 kW/cm². Trapping experiments were performed in the reaction buffer (50 mM HEPES, pH 7.5, 150 mM KCl, 3 mM Trolox, 0 or 5 mM MgCl₂, 0 or 2 mM nucleotide, 10% glycerol) with an added oxygen scavenging system³² (2 mM protocatechuic acid and 50 nM protocatechuate 3,4-dioxygenase) to prevent premature photobleaching of the dye Alexa 647. For experiments designed to probe the dynamical assembly/disassembly processes in real time, the excitation intensity was lowered by a factor of two (0.75 kW/cm²) and the oxygen scavenging system was prepared fresh before the measurements. To prevent protein sticking to the microfluidic surfaces, the sample chamber was passivated by polyelectrolyte multilayers using layer-by-layer deposition.^33,34 Sticking is easily detected by observation of the appearance of non-random behavior in the trapping field direction (i.e., the mean value of the feedback voltages being non-zero).

In the experiment, we only probe protein complexes containing at least one fluorescent subunit. We need to establish a link between the measurable quantities and the underlying oligomer distribution. The probability of an n-mer incorporating exactly one labeled subunit (L = 1) assuming independence is given by the binomial distribution,

where the denominator is the total concentration of an n-mer, and p is the ratio between labeled and total Rca subunits in solution. Labeled Rca was kept at ∼10 pM and the total unlabeled Rca subunit concentration was varied between 200 nM and 11 μM. Under these conditions, p lies between 10⁻⁶ and 10⁻⁴ (e.g., 10 pM/200 nM = 0.5 × 10⁻⁴), so (1 − p)ⁿ⁻¹ ≈ 1 and oligomers with more than one labeled subunit can be ignored.

We can now relate the fractional concentration (P⁽ⁿ⁾) of the different oligomers to the apparent percentage concentrations measured using the fluorescent subset,

From the above equation, it becomes clear that by measuring the fluorescent subset of particles and the oligomeric stoichiometries n, we can extract the fractional concentrations of the different oligomers.

The changes in diffusion coefficient were detected by a change-point algorithm similar to Ref. 35. Given that the estimation uncertainty follows a Gaussian distribution,³¹ we modified the algorithm to use a Gaussian-based log-likelihood ratio test instead of a Poisson model as in the original publication. Two-dimensional densities were estimated from a scatter plot using a Gaussian-smoothing algorithm (2D kernel density estimation³⁶).

For a monomer-dimer-tetramer-hexamer assembly model, we have for the equilibrium constants (Fig. 3)

The K_d values were determined by a least-squares procedure minimizing the following error function:

where $P̃i(n)$ is the experimentally extracted fractional concentration of a n-mer at the i-th concentration point (Fig. S3 and Tables S1–S3 of the supplementary material) and $Pi(k)$ is the fractional concentration predicted by the equilibrium [Eq. (3)] at the corresponding protomer concentration; $P̃i(>2)$ represents the unresolved tetramer/hexamer mixture seen in the experiment. This procedure can reliably extract $Kd1–2$ and $Kd2–4$ but not $Kd4–6$ (the error function is insensitive to the value of $Kd4–6$⁠, Fig. S2 of the supplementary material). As a result, only $Kd1–2$ and $Kd2–4$ are interpreted in this work.

To resolve the different oligomerization states of Nt-Rca in solution, we first incubated labeled Nt-Rca (∼10 pM) with unlabeled proteins (0-11 μM of protomer concentration). After equilibrium was reached, labeled Nt-Rca was randomly incorporated into the different assembly states. We then injected the equilibrated sample into a microfluidic-based single-molecule trapping system known as the Anti-Brownian ELectrokinetic (ABEL) trap described in the section titled Materials and methods [Fig. 1(a)]. Inside the trap, individual fluorescently tagged proteins were detected in solution with their 2D diffusive motion suppressed to within 2 μm from the trap center, resulting in long-time monitoring of single-molecule behavior in aqueous buffer solution. In the trap, individual proteins enter the center (“trapping region”) of the microfluidic chamber by diffusion and, if they carry a fluorescent subunit, are detected and captured by the feedback electrokinetic forces. These trapping events manifested themselves as brightness plateaus lasting ∼1 s [Fig. 1(a)]. Termination of a trapping event is most likely due to photobleaching of the fluorescent label (Alexa 647). Almost all observed molecules (>99%) entered the trap with a brightness that corresponds to one Alexa 647 molecule and left the trap with a single photo-bleaching step, supporting the assumption that there is at most one labeled subunit per complex (Materials and methods). When a single protein was captured, its diffusion coefficient and electrokinetic mobility were estimated with ∼100 ms time resolution using a parameter estimation framework developed in our previous work.³¹ Briefly, the raw particle trajectory in the trap is fitted with a Langevin-type dynamic model with proper treatment of measurement noise. The diffusion coefficient (D) is related to the hydrodynamic radius (r) of the molecule through the Stokes-Einstein relation (D = k_BT/6πηr) and directly probes oligomerization state. Here, the electrokinetic mobility (μ) is only weakly dependent on molecular charge because trapping was dominated by electro-osmotic forces resulting from charged, anti-fouling surface coatings of the microfluidic chamber (Materials and methods). As a result, mobility values were expected to be similar for all measured molecules, regardless of their physical properties, as observed in Figs. 1(a) and 2.

The ability to measure single-molecule size allowed us to probe the full distribution of oligomerization states in solution. We thus time averaged all D values and all μ values collected for each individual protein complex. For all measured molecules, the average values were displayed as single points on a D-μ parameter space and also shown as a density distribution plot using a 2D adaptive kernel density estimation method.³⁶ In the example data shown in Fig. 1(a), obtained with 400 nM Nt-Rca in the presence of ATPγS and Mg²⁺, we found three components in equilibrium with each other, as the projected D histogram requires a minimum of three Gaussian distributions to fit. Sub-stoichiometric labeling and single-molecule diffusometry thus provide a powerful means to dissect the distribution of oligomerization states in solution under equilibrium conditions.

We conducted single-molecule diffusometry measurements with varying protein (subunit) concentrations (Rca) from 0 to 11 μM, under three different nucleotide conditions. The results are summarized in Fig. 2. Below 200 nM, we saw exclusively a single population with a diffusion coefficient of ∼50 μm²/s, which we identified to be the monomeric species. The value of the assigned monomer species (50 μm²/s) is consistent with theoretical modeling of hydrodynamic properties³⁷ based on the crystal structure (supplementary material). This behavior of Nt Rca is similar to cotton β-Rca, which was observed to be exclusively monomeric below 300 nM.¹⁴ At intermediate protein concentrations between 0.4 μM and 2 μM with no Mg²⁺ and low ADP [Fig. S3 (left) of the supplementary material], we observed multi-modal distributions, indicating assembly into well-defined oligomeric states. When fitting the projected D histogram [Fig. 1(a) and Fig. S3 of the supplementary material] using three Gaussians, we consistently resolved a distinct population with a D that is 80% of the monomer. We identified this population to be Rca dimers: to a first-order approximation, the diffusion coefficient of a protein scales with (MW)^−1/3 (Ref. 38), where MW is the molecular weight (i.e., for dimers, the diffusion coefficient should be 1/2^1/3 = 0.79 times that of the monomer, consistent with the 0.8 value observed here). We thus identified the dimer as an assembly intermediate under the low-ADP condition. We note that although this simple scaling law between diffusivity and molecular weight is derived by approximating proteins as spheres, a consideration of more realistic protein shapes or structural models has been demonstrated to result in only minor corrections (D₂/D₁ from 0.79 to 0.80).¹⁴ Similarly, following the same analysis, dimers were also identified as assembly intermediates under Mg·ATPγS [Fig. S3 (middle) of the supplementary material] and Mg·ADP [Fig. S3 (right) of the supplementary material] conditions.

Upon visual inspection, the oligomerization pathways are distinct for the different nucleotide conditions (Fig. 2). Here, the diffusion coefficient is normalized by that of the monomer extracted from the fit [Fig. S3 of the supplementary material; in cases where few monomers were observed, the D of the monomer species was estimated from the dimer population by D₁ = D₂(1/2)^−1/3]. To understand these differences more quantitatively, we highlight the expected D positions of the monomer, dimer, tetramer, and hexamer species on the D-μ map using the D_n/D₁ = (1/n)^1/3 scaling law. Again, although these estimates were made by the simple “ball-filling” model, considering realistic protein shapes results in very small corrections (<5%; supplementary material). In the first condition (Fig. 2, left), Mg²⁺ was omitted and residual ADP concentrations ranged from 0 to 80 μM. Since the K_i value for ADP inhibition of Nt Rca has recently been determined to be 37 μM,³⁹ we estimate that on average, about half of the active sites were occupied with nucleotide. Under these conditions, we observed the buildup of dimers and the appearance of tetramers in the range of 0.4–0.8 μM Rca. At 2 μM, Nt-Rca seemed to exist in a complex equilibrium containing monomers, dimers, tetramers, and hexamers. Above 4 μM, we observed near complete depletion of the monomer species while most of the population was found to exist as a mixture of tetramers and hexamers. At 11 μM, oligomers larger than hexamers were observed. On the other hand, in the presence of 5 mM Mg²⁺ and 2 mM ATPγS, we observed a dramatically different oligomerization progression. Between 0.4 and 0.8 μM, we observed weak dimer buildup and rapid formation of a D ∼ 0.6D₁ species. We identified this state to be a mixture of tetramers and hexamers, given the limited resolution in measuring D values³¹ compared to the relatively small difference in D between tetramers and hexamers (∼14% difference, similar to the limit of our resolution; Fig. S6 of the supplementary material). Between 2 and 11 μM, we observed a gradual sharpening of the tetramer/hexamer population. At 11 μM, Nt-Rca seemed to be “locked” into the tetramer/hexamer form, instead of continuing to form higher oligomers, as observed for the low-ADP condition (Fig. 2, left). We then conducted measurements in the presence of 5 mM Mg²⁺ and 2 mM ADP, which may mimic the post-hydrolysis state of the enzyme complex. In this case, we observed an oligomerization progression very similar to that of the low-ADP protein. Dimers populate at sub-μM concentrations, tetramers and hexamers exist between 1 and 4 μM, and aggregates larger than hexamers form at >10 μM.

To further quantify the different oligomerization behavior as a function of nucleotide binding, we first extracted the fractional concentrations of the different oligomers, defined to be $P(n)=nCntot/∑kkCktot$ (⁠$Cntot$ is the total concentration of an n-mer), directly from fitting the projected D histogram with three Gaussian components [Fig. 1(a) and Fig. S3 and Tables 1–3 of the supplementary material]. The acquired fractional concentrations (⁠$P(n)$⁠) were then fit to an assembly model defined by the monomer-dimer-tetramer-hexamer pathway using an error minimization procedure (Material and methods). Due to our limited ability to separate tetramers and hexamers, we combined them into one species in the quantification (Fig. 3, blue circles). We extracted a monomer-dimer dissociation constant of $Kd1–2=1.2μM$ with Mg·ATPγS and $Kd1–2=0.5μM$ with Mg·ADP (residuals of the fit are plotted in Fig. S4 of the supplementary material). On the other hand, the dimer-tetramer/hexamer dissociation constants were an order of magnitude smaller with Mg·ATPγS compared to Mg·ADP (⁠$Kd2–4=0.2μM$ with Mg·ATPγS and $Kd2–4=5.7μM$ with Mg·ADP), indicating that ATPγS induces rapid formation of tetramers/hexamers from dimers.

Single-molecule diffusometry with the ABEL-oligo method offers the unique opportunity to observe dynamic assembly/disassembly processes that underpin an equilibrium, by measuring the time-dependent changes in the hydrodynamic radius of a single protein complex. To observe these processes in real time, we optimized experimental conditions (Materials and method) to greatly extend the trapping duration of single Alexa 647-labeled Nt-Rca proteins (from ∼1 s to ∼5-10 s). We indeed observed many molecules showing dynamic transitions in diffusivity with a time resolution of ∼150 ms. Figure 4(a) shows two examples, both taken with Mg·ADP. In the top panel, the molecule entered the trap as a tetramer (D ∼ 30 μm²/s or 0.6D₁ from 0.2 to 0.8 s), disassembled into a monomer (D ∼ 50 μm²/s from 0.8 to 1.8 s), and then underwent sequential assembly (monomer-dimer-tetramer/hexamer, from 1.8 s to the end of the trace). In the bottom panel, the molecule most likely entered as a dimer (D ∼ 42 μm²/s or 0.8D₁) and underwent multiple rounds of assembly and disassembly processes. By pooling all transitions detected under the Mg·ADP condition, we created a transition map by plotting the ending D state against the starting D state [Fig. 4(b)]. As illustrated by the transition map, the most encountered transitions are between dimers and tetramers/hexamers. The second most populated transition is between monomers and dimers. These results highlight the dynamic nature of the dimer-tetramer/hexamer pathway and further support the 1-2-4-6 assembly model we used to fit the equilibrium data. Interestingly, dynamic assembly and disassembly processes were also observed with Mg·ATPγS but with much decreased frequency, as quantified in Fig. 4(c). From the frequencies of the observed transitions, we estimated the rate of subunit exchange to be ∼0.1 s⁻¹ with ATPγS and >0.2 s⁻¹ with ADP. These data suggest that binding of ATPγS locks the protein into a more stable oligomer while ADP binding destabilizes the oligomers. It also suggests that some of the smearing observed in Fig. 2, especially in the Mg·ADP case (for example, the trimer-like density maxima at 0.8 μM), could be due to assembly/disassembly dynamics.

We have shown that single-molecule diffusometry is a novel and powerful method to resolve protein oligomerization pathways in solution, under equilibrium conditions. It is a single-molecule (here single-complex) method, thus applicable to sub-μM concentrations and dynamically interacting samples, where traditional bulk techniques fail. The method is based on measuring a physical property (the hydrodynamic radius) of individual objects, thus providing a more direct readout compared to other single-molecule fluorescence techniques (e.g., stepwise photobleaching, FRET, etc.). The method only requires fluorescent labeling of a small fraction of the species in equilibrium and can monitor oligomerization reactions at physiological protein concentrations (>1 μM), unlimited by the concentration barrier⁴⁰ for conventional single-molecule optical detection. Here, we demonstrated sufficient resolution to resolve monomers, dimers, tetramer/hexamers, and higher oligomers. Indeed, the method will be most powerful in the small-oligomer regime, where the hydrodynamic sizes of the different species differ by >20%. It should be noted that with a sufficient amount of detected photons, much smaller differences could in principle be resolved.³¹ Taken together, we envision the ABEL-oligo method described here to be widely applicable to other oligomerization systems and more generally to problems that involve biomolecular interactions.

The biological system studied here is the assembly pathway of tobacco Rubisco activase. Our data are consistent with previous measurements on the bulk level (Fig. S5 of the supplementary material) and reveal rich new details at the single-molecule level. We find that Nt-Rca assembles through a monomer-dimer-tetramer-hexamer pathway, which is consistent with the recent biophysical characterization of cotton Rca using FCS^14,24 and related AAA+ family proteins.⁴¹ We find that ATP binding (mimicked by non-hydrolyzable ATPγS in our experiments) strongly influences the assembly pathway. In particular, ATP binding promotes assembly of higher-order oligomers in the form of tetramers and hexamers. Notably, the extracted equilibrium constant for hexamer formation from dimers (K_d^2–4 × K_d^4–6), 0.2 μM², is identical to that reported for cotton β-Rca in the presence of ATPγS.²⁴ Quantitative analysis of the equilibrium composition and dynamic assembly/disassembly transitions of Nt Rca reveal the dimer as a critical intermediate that responds differentially to ATPγS and ADP binding. ATPγS binding drastically reduces the dimer-tetramer/hexamer dissociation constant and suppresses the rate of subunit exchange, leading to efficient and stable formation of tetramers/hexamers that dominate the equilibrium at 1-10 μM protein concentration.

We hypothesize that ATP binding at the subunit-subunit interface induces a conformational change at the dimer level that “primes” the protein for stable formation of higher-order, functional complexes. Nucleotide binding could serve as a regulatory mechanism that controls the concentration of functional oligomers. We anticipate future experiments on mutants and under different ATP/ADP ratios, together with correlated activity assays, to reveal deeper structural-functional insight into this system.

See supplementary material for additional sample characterization, analysis details, and measurement statistics.

This work is supported in part by the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences of the U.S. Department of Energy through Grant Nos. DE-FG02-07ER15892 (to W.E.M.) and DE-FG02-09-ER16123 (to R.M.W.). Q.W. acknowledges support from the Lewis-Sigler Fellowship from Princeton University.