Why Contact Duration Changes Bridge Formation
A cell engager can bind both cells and form a bridge, but the cells may separate before that bridge forms or persists. A model that calculates binding at equilibrium can therefore overstate assembly during brief encounters. The relevant quantity depends on the question: it may be the probability of being bridged at the end of a contact, the probability that a contact is still present and bridged, or the probability that a bridge formed at least once before separation. These quantities are not interchangeable.
We begin with a minimal kinetic model for a contact that remains available for a specified duration. Let \(U\) denote an unbridged, bridge-competent cell pair and \(B\) a bridged pair. Bridge formation occurs from \(U\) at rate \(a\), and bridge dissociation returns the pair from \(B\) to \(U\) at rate \(b\). Both rates have units of inverse time. The rates are treated as constant during the contact; in particular, engager availability and effective target density are held fixed.
This is a conditional model: it does not itself include cell separation. The contact duration \(T\) is imposed from outside, and \(P_B(T)\) describes bridge occupancy at the end of that available window. The rate \(a\) is an effective formation rate, not necessarily a microscopic association constant; it can summarize binding steps and depend on engager concentration, antigen abundance, and geometry.
Bridge Occupancy During a Fixed Contact Window
Probability flows from \(U\) to \(B\) at rate \(aP_U(t)\), and from \(B\) back to \(U\) at rate \(bP_B(t)\). Since \(P_U(t)=1-P_B(t)\), the bridge probability satisfies
For \(a+b>0\), solving this first-order equation with \(P_B(0)=0\) gives
If \(a=b=0\), no transition occurs and \(P_B(T)=0\). If \(a=0\) and \(b>0\), the bridge probability is also zero. When \(a>0\), the long-duration limit is \(a/(a+b)\), the equilibrium fraction in state \(B\). A finite contact does not generally reach this limit: the exponential term quantifies the remaining transient.
Worked Example: A Short Contact Does Not Reach Equilibrium
Suppose hypothetical measurements support an effective bridge-formation rate \(a=0.12\ \text{min}^{-1}\) and a bridge-dissociation rate \(b=0.08\ \text{min}^{-1}\). A cell-pair contact lasts \(T=5\) minutes. Assume the pair begins unbridged and the two-state rates remain constant during contact.
The equilibrium bridge fraction would be \(a/(a+b)=0.12/0.20=0.600\). The finite-time probability is
Thus, about 0.379 of such pairs are predicted to be bridged at the 5-minute endpoint under this model, below the equilibrium fraction of 0.600. For a 20-minute contact, the calculation is
The longer contact approaches, but does not exceed, the equilibrium occupancy. The rate sum \(a+b=0.20\ \text{min}^{-1}\) sets the relaxation timescale; the equilibrium fraction alone would not reveal how much assembly occurs during a short encounter.
Adding a Stochastic Contact Duration
Some contacts do not have one fixed duration. A simple alternative assumes that contact termination occurs independently of bridge state with a constant hazard \(m\), measured in inverse time. Contact duration is then exponentially distributed, with survival probability \(\exp(-mT)\) through time \(T\). This assumption means that, at any instant, the chance of separation over a short interval depends on the interval length but not on how long the contact has already lasted.
Because this contact-loss hazard acts equally in \(U\) and \(B\), the probability of still having contact at time \(T\) is \(\exp(-mT)\). Conditional on still having contact, the bridge-state kinetics are unchanged from the fixed-window model. Therefore the joint probability of being both in contact and bridged at time \(T\), starting in \(U\), is
When \(a=b=0\), this joint probability is zero. If \(m=0\), contact never terminates in this model and the expression reduces to fixed-duration bridge occupancy. The probability conditional on contact still being present is \(P_B(T)\); the joint probability is smaller because it also requires survival of the contact.
The probability of having formed a bridge at least once is a different endpoint. Under the same independent contact-loss assumption, the first bridge forms from \(U\) at rate \(a\), while separation occurs at rate \(m\). Starting unbridged, the probability that bridge formation happens before separation is \(a/(a+m)\) when \(a+m>0\). This “ever formed” event remains true even if the bridge later dissociates; it should not be reported as endpoint bridge occupancy.
Worked Example: Contact Survival and Endpoint Bridge Probability
Use the hypothetical rates \(a=0.12\ \text{min}^{-1}\), \(b=0.08\ \text{min}^{-1}\), and an independent contact-loss rate \(m=0.05\ \text{min}^{-1}\). Evaluate the probability that a pair remains in contact and is bridged at \(T=10\) minutes.
First, the fixed-window bridge probability is
The probability contact persists to 10 minutes is \(e^{-0.05(10)}=e^{-0.5}\approx0.6065\). Independence of contact termination from bridge state gives
The result is about 0.315, whereas about 0.519 is the bridge probability conditional on contact still being present. The difference is consequential: the first quantity counts all initially considered pairs, while the second conditions on contact survival. The probability that a bridge forms at least once before separation is \(0.12/(0.12+0.05)\approx0.7059\), a third quantity that includes bridges that may subsequently dissociate.
When Contact Loss Depends on the Bridged State
A different mechanistic assumption is that contact loss occurs specifically from the bridged state. This can represent a model in which an unbridged encounter remains available until bridge formation, while a bridged pair can either dissociate back to \(U\) or leave the modeled contact at rate \(m\). It is not equivalent to an independent contact-duration hazard: it assigns no contact-loss transition to \(U\). Use it only when that state dependence is intended.
Let \(U(t)\) and \(B(t)\) now denote probabilities of occupying these transient states, starting from \(U(0)=1\), \(B(0)=0\). The absorbing state represents termination of the modeled contact. The equations are
For exponential solutions with decay rate \(r\), the transient-state matrix has characteristic equation
Its roots are
The discriminant is nonnegative because \((a+b+m)^2-4am=(a-m)^2+b^2+2b(a+m)\geq0\). For nonnegative rates, both roots are nonnegative; they are both positive when \(a>0\) and \(m>0\). If \(a=0\) or \(m=0\), one root is zero, so the zero-rate boundary case must not be described as two positive decay rates.
When \(r_1<r_2\), solving the equations with the stated initial conditions yields
If \(r_1=r_2\), the limiting expression is \(B(t)=at e^{-r_1t}\). If \(a=0\), then \(B(t)=0\). If \(m=0\) and \(a+b>0\), the model has no absorbing loss and the expression reduces to the fixed two-state bridge probability \(a[1-e^{-(a+b)t}]/(a+b)\). If \(a=b=0\), \(B(t)=0\). These boundaries matter because formulas with a rate difference in the denominator should not be used without checking that the denominator is nonzero.
Worked Example: A Bridged-State Contact-Loss Model
Consider hypothetical rates \(a=0.20\ \text{min}^{-1}\), \(b=0.10\ \text{min}^{-1}\), and \(m=0.05\ \text{min}^{-1}\), with the bridged-state-only contact-loss assumption. The rate sum is \(a+b+m=0.35\ \text{min}^{-1}\), and the product is \(am=0.0100\ \text{min}^{-2}\). Therefore
giving \(r_1\approx0.03138\ \text{min}^{-1}\) and \(r_2\approx0.3186\ \text{min}^{-1}\). At 10 minutes,
The unbridged probability is obtained from \(U(0)=1\) and \(U'(0)=-a=-0.20\ \text{min}^{-1}\), giving \(U(t)\approx0.4129e^{-r_1t}+0.5871e^{-r_2t}\). Hence \(U(10)\approx0.326\), and the probability of remaining in either transient state is \(S(10)=U(10)+B(10)\approx0.806\). This survival result is specific to loss from \(B\) only; it should not be substituted for survival under a contact-loss hazard that also acts in \(U\).
Choosing and Checking the Contact Model
The models differ in their assumptions and in the quantities they produce. They are alternatives, not successive levels of one universally correct formula.
| Model | Contact-loss assumption | Useful output | Key limitation |
|---|---|---|---|
| Fixed window, \(U\leftrightarrow B\) | Contact is available for a specified duration \(T\) | Bridge occupancy \(P_B(T)\) | Does not represent variation in contact duration |
| Independent exponential loss | Same constant loss hazard in \(U\) and \(B\) | Joint probability of contact and bridge at \(T\) | Assumes memoryless, state-independent separation |
| Bridged-state-only loss | Loss at rate \(m\) from \(B\), none from \(U\) | Transient bridge probability and contact survival | May be inappropriate if unbridged encounters also separate |
If contact termination occurs from both states but at different rates, use explicit loss rates \(m_U\) and \(m_B\). The transient equations become \(U'=-(a+m_U)U+bB\) and \(B'=aU-(b+m_B)B\). The predicted state probabilities then follow from this system with the chosen initial conditions; neither the independent-hazard factorization nor the bridged-state-only formula generally applies. More complex or time-varying rates can be handled by numerically integrating the corresponding state equations.
Diagnostics should focus on both assumptions and numerical behavior. Check that rates and durations use compatible time units, that all rates are nonnegative, and that probabilities remain between zero and one. In the fixed-window model, \(P_B(0)=0\), and for \(a>0\) the initial slope is \(P'_B(0)=a\); these are useful implementation checks. In the bridged-state-only model, \(S'(t)=-mB(t)\leq0\), so survival cannot increase. If a fitted model implies increasing contact survival or negative probabilities, inspect the state transitions, rate signs, and numerical solver settings.
Observed bridge occupancy at a single duration may not identify \(a\) and \(b\) separately: many rate combinations can yield similar \(P_B(T)\). Measurements across multiple contact durations, or independent information on binding and dissociation, help distinguish rapid formation followed by loss from slow formation that has not yet approached equilibrium. When contact durations vary, recording or modeling their distribution is preferable to applying a single mean duration without checking the consequences of that simplification.
Common Mistakes and Practical Meaning
- Using equilibrium occupancy for a brief contact. The equilibrium value \(a/(a+b)\) assumes enough time for relaxation. Use the finite-time expression when the contact window is short relative to \(1/(a+b)\).
- Confusing occupancy with ever-formed probability. \(P_B(T)\) concerns the state at the endpoint; an “ever formed” probability also counts bridges that later dissociate.
- Combining incompatible contact-loss assumptions. Multiplying by \(e^{-mT}\) assumes independent loss from either state. It is not the solution when loss is allowed only from \(B\).
- Ignoring boundary cases. Check zero rates and repeated roots before using a closed form with a denominator such as \(a+b\) or \(r_2-r_1\).
Contact-time dependence matters because a cell pair may have enough opportunity to assemble a bridge only during sustained contact. A small endpoint bridge probability can reflect a low formation rate, a short available window, appreciable dissociation, or contact loss. These mechanisms have different implications for model interpretation, so state the assumed contact process and the probability being predicted before drawing a conclusion about synapse assembly.
Check Your Understanding
Answer each question using the stated model assumptions. Brief solutions follow the questions.
- Under the fixed-duration model, let \(a=0.10\ \text{min}^{-1}\), \(b=0.10\ \text{min}^{-1}\), and \(T=5\) minutes. What is the bridge probability at the endpoint?
- For an independent contact-loss rate \(m\), explain the difference between the joint probability of being in contact and bridged at \(T\), and the bridge probability conditional on still being in contact.
- In the bridged-state-only model, what happens to the bridge probability when \(a=0\)? Why is it incorrect to call both decay rates positive in this boundary case?
- For the fixed-window model, what does the probability \(a/(a+b)\) represent, and when should it not be used as the endpoint bridge probability?
- If contact can terminate from both \(U\) and \(B\) at different rates, which model change is required?
Solutions. (1) \(P_B(5)=0.5[1-e^{-1}]\approx0.3161\). (2) The joint probability includes both survival of contact and bridge occupancy; the conditional probability is \(P_B(T)\) among pairs whose contact remains. Under an independent hazard the joint value is \(e^{-mT}P_B(T)\). (3) No bridge forms, so \(B(t)=0\). The characteristic equation has product \(am=0\), hence one root is zero rather than positive. (4) It is the long-time equilibrium bridge fraction for \(a+b>0\); it is not generally the endpoint probability for a finite, short contact. (5) Include separate loss rates \(m_U\) and \(m_B\) in the state equations and solve the resulting transient system.