Cooperative Binding and Allostery

Cooperative and allosteric mechanisms are the working principle behind an enormous swath of modern pharmacology and metabolic engineering. Voxelotor, an FDA-approved drug for sickle cell disease, works by covalently stabilizing hemoglobin's R (oxy) state, shifting the T⇌R equilibrium (raising effective $c$ and effectively lowering $L$ in MWC terms) to reduce sickling-prone deoxy-HbS polymerization — a direct pharmacological exploitation of the allosteric constant. Allosteric drug design more broadly (allosteric kinase inhibitors, allosteric GPCR modulators) targets sites distinct from the orthosteric (substrate) pocket specifically to exploit MWC/KNF-type conformational coupling rather than competing directly for the active site, often giving better selectivity because allosteric pockets are less conserved across related proteins than active sites. In metabolic regulation, phosphofructokinase-1 (PFK-1), the committed step of glycolysis, is a tetramer showing MWC-type cooperativity toward its substrate fructose-6-phosphate, with ATP as a negative allosteric effector (T-stabilizing) and AMP/fructose-2,6-bisphosphate as positive effectors (R-stabilizing) — the sigmoidal fructose-6-phosphate response curve is what allows PFK-1 to function as a sharp metabolic switch responsive to cellular energy charge rather than a smooth, poorly-controlled rheostat. Aspartate transcarbamoylase (ATCase), the founding MWC test case, remains a standard teaching and research model for concerted allostery, with well-resolved T and R crystal structures and CTP/ATP as antagonistic allosteric effectors. Hemoglobin cooperativity itself remains directly clinically relevant in altitude physiology, chronic obstructive pulmonary disease management, and blood-substitute engineering, where designed hemoglobin variants with altered $p_{50}$ and $n_H$ are being developed as oxygen-carrying therapeutics. Beyond classical enzymology, allosteric principles now underpin synthetic biology (engineered allosteric transcription factors and biosensors that couple ligand binding to conformational output for cell-based diagnostics), structural methods (cryo-EM increasingly resolves multiple discrete conformational states of oligomeric receptors and channels, testing MWC-like multi-state pictures directly), and computational biophysics, where statistical-mechanical MWC-type partition functions are now applied well beyond proteins — to gene regulatory networks, ion channel gating (the MWC framework describes voltage-gated and ligand-gated channel opening curves essentially unchanged), and even to CRISPR-Cas9 target search kinetics.

:::quiz[If a Hill plot of log[Y/(1-Y)] vs log[L] gives a straight line with slope 2.0 between Y=0.2 and Y=0.8, and passes through the origin-shifted point where the line crosses log[Y/(1-Y)]=0 at log[L]=1.30, then K_0.5 equals:]

• 1.30 torr
• 10^1.30 ≈ 20 torr ✓
• 2.00 torr
• 10^2.00 = 100 torr

Ask the tutor: How does the Bohr effect quantitatively fold protons and CO₂ into the MWC allosteric constant L?

Consider four oxygen-binding sites on a hemoglobin tetramer, each an Fe(II)-porphyrin in a globin fold nearly identical to myoglobin's. In isolation, each subunit would bind O₂ hyperbolically, exactly like myoglobin, with a dissociation constant around 1–2 torr. Assembled into the tetramer, the first O₂ binds with much lower apparent affinity ($K_d \sim 30$ torr) while the fourth binds with affinity close to the isolated-subunit value. The tetramer interpolates between a low-affinity "T state" (tense, deoxy-favored, salt-bridge stabilized) and a high-affinity "R state" (relaxed, oxy-favored), and each successive oxygen shifts the population toward R. Structurally, the trigger is small: oxygen binding to the heme iron pulls the iron from ~0.6 Å out-of-plane (five-coordinate, high-spin Fe(II)) into the porphyrin plane (six-coordinate, low-spin Fe(II)) as it becomes octahedrally coordinated. This ~0.6 Å motion pulls the proximal histidine (F8) and the attached F helix along with it. Because the α1β1/α2β2 dimers pack against each other through an extensive interface, and the α1β2/α2β1 interface undergoes a ~15° rotation between T and R (the classic "rotation of one dimer relative to the other" described in quaternary hemoglobin structure), the local heme motion is mechanically transmitted into a global quaternary rearrangement. That rearrangement breaks T-state salt bridges (His146β–Asp94β, Lys40α–His146β C-terminal contacts, and others) that stabilized the low-affinity conformation, and the newly formed R-state interface geometry raises the intrinsic affinity of every remaining unliganded heme. The key conceptual move is to separate microscopic (intrinsic) affinity — how tightly a given subunit binds ligand in a fixed conformation — from macroscopic (apparent) affinity — what you actually measure as $p_{50}$ or $K_d$ from a titration, which is an average over a shifting population of conformations. Cooperativity is not a change in the chemistry of Fe–O₂ bond formation; it is a change in which conformational state is populated. This is why cooperativity is a thermodynamic linkage phenomenon: the free energy of ligand binding at one site is linked to the free energy of the conformational transition, and that conformational transition is shared by the whole oligomer.

The empirical description of a saturation curve, independent of any mechanism, is the Hill equation:

where:

• $Y$ is the fractional saturation (0 to 1),
• $[L]$ is the free ligand concentration (or partial pressure, for gases),
• $K_{0.5}$ is the ligand concentration at half-saturation ($p_{50}$ for O₂),
• $n_H$ is the Hill coefficient, an empirical cooperativity index.

Reading: $n_H = 1$ recovers a hyperbola (Michaelis-Menten-like, non-cooperative). $n_H > 1$ means the curve is sigmoidal: binding accelerates ligand uptake once some sites are filled ("positive cooperativity"). $n_H < 1$ means anti-cooperativity — occupied sites make remaining sites bind more weakly. For a protein with $n$ identical independent sites, $1 \le n_H \le n$; the upper bound $n_H = n$ is reached only in the impossible limit of "infinitely cooperative," all-or-none binding. The Hill equation is linearized by the Hill plot:

where the left side is the log-odds of saturation. A plot of $\log(Y/(1-Y))$ against $\log[L]$ gives a straight line of slope $n_H$ only at the central inflection; at very low and very high $[L]$ the slope approaches 1, because at the extremes only one binding step is rate-limiting for the next increment of saturation. The Hill coefficient reported for hemoglobin ($n_H \approx 2.8$–3.0) is the maximal local slope, near $Y = 0.5$. The mechanistic models must reduce to this phenomenology. The MWC model gives an explicit saturation function:

where:

• $\alpha = [L]/K_R$ is the ligand concentration scaled by the R-state dissociation constant,
• $c = K_R/K_T$ is the ratio of R-state to T-state intrinsic dissociation constants ($c \ll 1$ for hemoglobin),
• $L = [T_0]/[R_0]$ is the allosteric constant, the T/R equilibrium ratio with no ligand bound,
• $n$ is the number of binding sites per oligomer (4 for hemoglobin).

Reading: when $L$ is large and $c$ is small, the unliganded protein is almost entirely T (low affinity), but each successive ligand strongly favors R over T (since R binds $1/c$ times tighter), so the population shifts abruptly from all-T to all-R over a narrow ligand range — this population shift is the cooperativity, even though every individual binding step within a fixed conformation is simple mass-action chemistry.

Goal: derive $Y_{\text{MWC}}$ from first principles using a grand-canonical-style partition function over quaternary state and ligand occupancy, and confirm it collapses correctly to a hyperbola when $L \to 0$ or $c \to 1$.

Step 1. State the postulates. MWC assumes (i) the oligomer exists in exactly two conformations, T and R, in equilibrium regardless of ligand occupancy; (ii) all $n$ sites within one conformation are equivalent and bind independently with a single microscopic dissociation constant ($K_T$ for T, $K_R$ for R); (iii) the T⇌R transition is concerted — every subunit switches together, so no "mixed" conformer (some subunits T-shaped, others R-shaped) exists.

Step 2. Write the (unnormalized) statistical weight of each state. Let the unliganded T state have weight $L$ (relative to unliganded R, weight 1, by definition of $L = [T_0]/[R_0]$). A T-state molecule with $i$ ligands bound (out of $n$ sites, each independently occupied with local equilibrium constant $1/K_T$) has combinatorial weight

and analogously for R-state with $i$ ligands,

Step 3. Sum over occupancy to get conformer partition functions. Summing $w_T(i)$ over $i=0,\dots,n$ via the binomial theorem:

Define $\alpha = [L]/K_R$ and $c = K_R/K_T$, so $[L]/K_T = c\alpha$. Then

Step 4. Compute average sites occupied. The average number of ligands bound per oligomer is a weighted average over both conformer identity and occupancy number, which is most efficiently obtained as a logarithmic derivative of the total partition function $Z = Z_T + Z_R$ with respect to $\ln[L]$:

Differentiating $Z_T$ with respect to $[L]$: $\partial Z_T/\partial[L] = L\cdot n(1+c\alpha)^{n-1}\cdot(c/K_R)$, and multiplying by $[L] = \alpha K_R$ gives $Ln c\alpha(1+c\alpha)^{n-1}$. The analogous term from $Z_R$ gives $n\alpha(1+\alpha)^{n-1}$. Hence

Step 5. Convert to fractional saturation. Since $Y = \bar\nu/n$, the factor $n$ cancels and we recover exactly

as claimed.

Step 6. Check limiting cases. If $c=1$ ($K_T=K_R$, no discrimination between conformations), then $\alpha=c\alpha$ and both bracket terms in numerator and denominator become identical powers of $(1+\alpha)$; algebraically $Y \to \alpha/(1+\alpha) = [L]/(K_R+[L])$, a simple hyperbola — correctly showing that cooperativity vanishes when the two conformations have equal intrinsic affinity, regardless of $L$. If $L \to 0$ (no T state populated at all, protein permanently in R), the $L$-containing numerator and denominator terms vanish, leaving $Y \to \alpha(1+\alpha)^{n-1}/(1+\alpha)^n = \alpha/(1+\alpha)$, again a hyperbola — correctly showing that a single fixed conformation cannot be cooperative even with multiple sites, matching the physical expectation that cooperativity in MWC requires both a conformational choice ($L$ finite, nonzero) and a difference in affinity between conformations ($c \ne 1$).

Goal: derive the KNF two-site (dimer) saturation function including a nearest-neighbor interaction free energy, showing how sequential induced-fit produces sigmoidicity without invoking a global two-state switch, and show the model reduces to independent-site binding when the interaction energy is zero.

Step 1. Set up the induced-fit postulate. KNF assumes each subunit alone is either in conformation $a$ (low affinity, unliganded-like) or conformation $b$ (high affinity, ligand-induced), and a subunit adopts $b$ only when it binds ligand (no concerted global flip). Neighboring subunits interact at their interface with an energy that depends on whether both are $a$, both are $b$, or mixed $ab$. Take the simplest case: a dimer (two sites), intrinsic microscopic dissociation constant $K$ for a subunit binding ligand and converting $a\to b$, and an interaction free energy $\Delta G_{ab}$ for a mixed $a/b$ interface relative to a reference $aa$ interface (set $\Delta G_{aa} \equiv 0$), and $\Delta G_{bb}$ for the doubly-liganded interface.

Step 2. Enumerate the three occupancy states of the dimer. Let $x = [L]/K$. Statistical weights (relative to unliganded $aa$, weight 1):

Define the interaction factors $K_{ab} = e^{-\Delta G_{ab}/RT}$ and $K_{bb} = e^{-\Delta G_{bb}/RT}$, so $w_1 = 2xK_{ab}$ and $w_2 = x^2K_{bb}$.

Step 3. Compute average occupancy. The average number of ligands bound (out of 2 sites) is

and fractional saturation $Y = \bar\nu/2$:

Step 4. Identify the source of cooperativity. Positive cooperativity requires that binding the second ligand be easier than binding the first, which in this weight scheme means $K_{bb} > K_{ab}^2$ — the $bb$ interface is disproportionately stabilizing relative to two independent $ab$-type contacts, i.e. $\Delta G_{bb} < 2\Delta G_{ab}$. This is the induced-fit analogue of MWC's $c \ll 1$: instead of a pre-existing global conformational bias, cooperativity is built subunit-interface by subunit-interface, sequentially, as each ligand arrives and reshapes its local contacts.

Step 5. Limiting case: no interaction energy. If $\Delta G_{ab} = \Delta G_{bb} = 0$, then $K_{ab}=K_{bb}=1$ and

a simple hyperbola for two independent, non-interacting sites — confirming that in KNF, exactly as in MWC, cooperativity requires a nonzero coupling term (here $\Delta G_{ab}, \Delta G_{bb} \ne 0$; there, $c\ne1$ combined with finite $L$), never emerging from multiple sites alone.

Step 6. Why KNF can also produce negative cooperativity. Because each interface term is an independent adjustable parameter, KNF naturally accommodates $K_{bb} < K_{ab}^2$ (negative cooperativity, $n_H<1$), which MWC's strict two-state postulate cannot produce at all (MWC is mathematically constrained to $n_H \ge 1$ always). This is the clearest experimental discriminator between the models: any protein showing negative cooperativity (e.g., glyceraldehyde-3-phosphate dehydrogenase binding NAD⁺, or some reports for hemoglobin binding certain non-natural ligands) cannot be pure MWC and requires sequential-type coupling.

Example 1 (Foundational): Hill plot from raw saturation data. Suppose an oxygen-binding protein shows $Y = 0.20$ at $p\text{O}_2 = 10$ torr and $Y = 0.80$ at $p\text{O}_2 = 40$ torr. Estimate $n_H$ and $K_{0.5}$, assuming Hill behavior holds between these points. From the Hill equation, $\log(Y/(1-Y)) = n_H\log[L] - n_H \log K_{0.5}$. Subtracting the equation at the two points eliminates $K_{0.5}$:

Compute the left side: $\log(4) - \log(0.25) = 0.602 - (-0.602) = 1.204$. Compute the right side bracket: $\log(4) = 0.602$. So

This is a textbook-clean cooperativity index of 2 — a useful sanity check value (asymmetric inputs 10 and 40 torr, not equal or trivially related, were chosen deliberately so that the answer isn't forced by symmetry). To get $K_{0.5}$, use the point at $Y=0.5$: since $\log(Y/(1-Y))=0$ there, and using the line through the two known points,

Limiting check: if $n_H$ had come out to 1.00, this would describe a purely hyperbolic (non-cooperative) binder like myoglobin; the value 2.00 confirms genuine positive cooperativity, intermediate between non-cooperative ($n_H=1$) and hemoglobin's physiological ($n_H\approx2.8$–3).

Example 2 (Applied): Effect of 2,3-BPG on hemoglobin's oxygen delivery. Human hemoglobin in whole blood (in the presence of physiological 2,3-bisphosphoglycerate, BPG) has $p_{50}\approx 26$ torr and $n_H \approx 2.8$. Stripped hemoglobin (BPG removed) has $p_{50}\approx 12$ torr with similar $n_H$. Compute the fractional saturation delivered to tissue at $p\text{O}_2=25$ torr (representative of venous/tissue oxygen tension) for both cases, given lung-arterial saturation is essentially 0.98 in both, and interpret the physiological consequence. Using the Hill equation $Y = [L]^{n_H}/(K_{0.5}^{n_H}+[L]^{n_H})$ with $n_H=2.8$: Whole blood ($K_{0.5}=26$):

Compute $25^{2.8} = e^{2.8\ln25} = e^{2.8(3.2189)} = e^{9.013} = 8213$, and $26^{2.8}=e^{2.8\ln26}=e^{2.8(3.2581)}=e^{9.123}=9159$.

Stripped ($K_{0.5}=12$): $12^{2.8}=e^{2.8\ln12}=e^{2.8(2.4849)}=e^{6.958}=1050$.

Fraction of bound O₂ unloaded between lungs (Y≈0.98) and tissue is $0.98-Y$:

Interpretation: BPG's allosteric right-shift of $p_{50}$ from 12 to 26 torr — by preferentially stabilizing the T state (binding in the central cavity between the β chains, a site that only exists in the T conformation) — increases physiological O₂ delivery more than fivefold at this tissue tension, even though it decreases the protein's absolute affinity. This is the mechanistic seed of the topic the Bohr effect extends, where protons and CO₂ play an analogous T-stabilizing role.

Limiting check: as $p_{50}\to p\text{O}_2$ exactly (26→25 is close already), $Y\to 0.5$ by definition of $p_{50}$; our computed 0.473 is close to but correctly just below 0.5 since 25 < 26, confirming internal consistency of the formula rather than an arithmetic accident.

Example 3 (Competition-level): Extracting MWC parameters $L$ and $c$ from two data points. A tetrameric allosteric enzyme ($n=4$ sites) is measured to have $Y=0.100$ at $\alpha = [L]/K_R = 0.05$ and $Y = 0.900$ at $\alpha = 8.0$. Assume $c = 0.02$ is known independently from crystallography (R state binds 50× tighter than T state). Solve for the allosteric constant $L$. Use the full MWC equation with $n=4$:

At $\alpha=0.05$: $c\alpha = 0.001$, so $(1+c\alpha)^3\approx1.003$, $(1+c\alpha)^4\approx1.004$; $(1+\alpha)^3=(1.05)^3=1.1576$, $(1+\alpha)^4=(1.05)^4=1.2155$. Numerator $= L(0.02)(0.05)(1.003) + (0.05)(1.1576) = 0.0010030\,L + 0.057881$ Denominator $= L(1.004) + 1.2155$ Setting $Y=0.100$:

This gives a negative $L$, signaling the chosen $\alpha=0.05$ point is far too saturating for such a small $L$ to reproduce $Y=0.1$ — meaning we should instead solve using both data points simultaneously as two equations in the two remaining unknowns and check consistency, which is the actual competition-style task. Let's redo properly by using both points to solve for $L$ directly without assuming which point dominates.

At $\alpha=8.0$: $c\alpha = 0.16$; $(1+c\alpha)^3=(1.16)^3=1.5609$, $(1+c\alpha)^4=(1.16)^4=1.8106$; $(1+\alpha)^3=(9.0)^3=729$, $(1+\alpha)^4=(9.0)^4=6561$. Numerator $= L(0.02)(8.0)(1.5609)+(8.0)(729) = 0.24974\,L+5832.0$ Denominator $=L(1.8106)+6561$ Setting $Y=0.900$:

Both data points independently demand a negative $L$, which is unphysical ($L=[T_0]/[R_0]\ge0$ always). The correct diagnosis — the actual "exam" content of this problem — is that the assumed value $c=0.02$ is incompatible with these two data points under the MWC model: no non-negative $L$ can satisfy both $Y(0.05)=0.100$ and $Y(8.0)=0.900$ simultaneously with $c=0.02$ and $n=4$. This is a genuine and important lesson: MWC has only two free parameters ($L,c$) for a fixed $n$, so it is over-determined by more than two $(Y,\alpha)$ pairs, and a real dataset either fits well (supporting the concerted model) or fails to fit for any $(L,c)$ (a signature that sequential/KNF-type or hybrid behavior is required). Setting up the two simultaneous equations and finding no consistent non-negative solution is itself the correct, rigorous result — a smaller-magnitude, more careful competition version would ask the student to demonstrate exactly this inconsistency and correctly conclude the model must be rejected or $c$ re-estimated, rather than to blindly report a numerical $L$.

Limiting-case sanity note: if instead the two given saturations had been much closer to the hyperbolic expectation for a genuine two-state switch with $c=0.02, n=4$ (for example $Y(0.05)\approx0.0002$ deep in the T-dominated regime and $Y(8.0)\approx0.98$ deep in the R-dominated regime, appropriate for large $L\sim10^4$–$10^5$ as in real hemoglobin), a positive, physically sensible $L$ does emerge — confirming the algebra is correct and that the failure above is a property of the chosen numbers, illustrating precisely how over-fitting tests discriminate real allosteric proteins from ones that merely look sigmoidal by other means.

The mathematics of cooperative binding was first given empirical form by Archibald Vivian Hill, a British physiologist, in 1910, fitting oxygen-hemoglobin saturation data to what is now the Hill equation — notably before anyone knew hemoglobin's structure or even its subunit stoichiometry. Hill's equation was originally derived (incorrectly, in mechanistic detail) from an assumed all-or-none binding of $n$ oxygens simultaneously to an aggregate; the equation survived as a phenomenological fit long after that literal mechanistic picture was abandoned, which is itself a lesson in distinguishing a successful empirical formula from a correct mechanism. The mechanistic resolution came half a century later. Jacques Monod, Jeffries Wyman, and Jean-Pierre Changeux published the concerted (MWC) model in 1965, motivated as much by bacterial allosteric enzymes (like aspartate transcarbamoylase and threonine deaminase) as by hemoglobin, proposing that oligomeric proteins possess a small number of accessible symmetric quaternary states in a fixed pre-existing equilibrium, with ligand binding acting purely to redistribute population between these states — never inducing new conformations. Daniel Koshland, together with George Némethy and David Filmer, proposed the rival sequential (KNF) induced-fit model in 1966, building on Koshland's earlier induced-fit theory of enzyme specificity, in which each subunit's conformational change is triggered individually by its own ligand and propagated through pairwise subunit interfaces — a picture that, unlike MWC, can naturally accommodate negative cooperativity and asymmetric (non-symmetric) intermediate states. Max Perutz's determination of the T and R quaternary crystal structures of hemoglobin in the 1970s gave the concerted model direct structural support, showing an actual ~15° inter-dimer rotation between two discrete, well-defined global conformations — vindicating the core MWC picture while later kinetic and intermediate-structure work (Ackers and coworkers, among others) demonstrated the real sequential-model contributions MWC alone could not capture.