CTLA-4 Control of T-Cell Priming

Why Model CTLA-4 at the Priming Stage?

Priming is not simply a switch that turns a T cell on. Antigen recognition supplies a signal, while co-stimulatory and inhibitory inputs shape whether that signal produces activation and expansion. CD28 and CTLA-4 can both bind the B7 ligands CD80 and CD86 on antigen-presenting cells. A model that treats their access to B7 as competitive makes one part of this regulation explicit: when CTLA-4 occupies more B7, less is available for CD28-mediated co-stimulation.

The modeling distinction matters. CTLA-4 can alter how many T cells become activated and how many subsequently expand. That does not, by itself, imply that each activated T cell kills tumor cells at a different rate. A QSP model should place a CTLA-4 effect in the priming and expansion terms unless there is separate evidence for a change in the downstream killing process.

Competitive Binding and Its Assumptions

Let \(B_{\mathrm{tot}}\) be the total available B7 concentration on an antigen-presenting cell. Let \(C\) and \(H\) denote the free concentrations of CD28 and CTLA-4 available to bind B7, and let \(K_C\) and \(K_H\) be their respective dissociation constants. Lower \(K\) means stronger binding affinity. The model assumes rapid reversible binding relative to the timescale of priming, a shared accessible ligand pool, mutually exclusive binding, and approximately constant free receptor concentrations during the binding calculation.

The last assumption is a useful approximation when receptor is in sufficient excess, or when the free receptor concentrations are otherwise known. If ligand binding substantially depletes free receptor, receptor mass-balance equations are needed instead. The model also treats the B7 pool as fixed over the binding calculation; it does not include ligand synthesis, internalization, or removal.

With free receptor concentrations treated as fixed, the binding equations give \([BC]=[B]C/K_C\) and \([BH]=[B]H/K_H\). Define the dimensionless binding weights \(w_C=C/K_C\) and \(w_H=H/K_H\). Conservation of the B7 pool then gives the free and bound fractions:

The fractions sum to one, and the concentrations in the three states are \(B_{\mathrm{tot}}\theta_{\mathrm{free}}\), \(B_{\mathrm{tot}}\theta_C\), and \(B_{\mathrm{tot}}\theta_H\). For example, increasing \(w_H\) lowers \(\theta_C\) when \(w_C\) is nonzero: CTLA-4 does not merely add an inhibitory signal; it also competes with CD28 for the shared ligand.

From Occupancy to Activation and Expansion

Occupancy is not itself an activation outcome. A simple mechanistic bridge is to define a dimensionless priming drive \(q\), with antigen contribution \(A\), co-stimulation gain \(\alpha\), inhibition gain \(\beta\), and reference B7 concentration \(B_{\mathrm{ref}}\):

Here \(A\), \(\alpha\), and \(\beta\) are nonnegative dimensionless quantities. This is a phenomenological signal combination: its parameters must be estimated or calibrated for the model context, and the equation does not claim that biochemical signaling is literally additive. One possible bounded activation response is a logistic function:

The slope \(s>0\) and midpoint \(q_{50}\) set the response shape. In a simple priming model, if \(P\) is the number of available precursor T cells and \(k_{\mathrm{prim}}\) is an effective priming rate, the input of activated cells can be written \(k_{\mathrm{prim}}p_{\mathrm{act}}P\). An activated-cell compartment \(E\) might then follow \(dE/dt=k_{\mathrm{prim}}p_{\mathrm{act}}P+r_{\mathrm{exp}}E-d_EE\), where \(r_{\mathrm{exp}}\) is an expansion rate and \(d_E\) is a loss rate. This formulation locates the CTLA-4 effect upstream, through priming and potentially the resulting cell pool.

Why Priming Is Not the Same as Tumor-Cell Killing

A model should distinguish the number and state of available effector T cells from their tumor-cell killing rate. For example, a downstream tumor equation could contain a separate per-cell killing parameter \(k_{\mathrm{kill}}\), with a term whose magnitude depends on \(k_{\mathrm{kill}}\) and the effector population \(E\). A CTLA-4 priming effect can change \(E\) through \(p_{\mathrm{act}}\) and expansion while leaving \(k_{\mathrm{kill}}\) unchanged. Consequently, total tumor-cell loss may change because there are more or fewer effectors, even when the per-cell killing parameter is identical.

This separation is also useful for interpreting data. If a treatment changes activated-cell counts but does not change a per-cell killing assay under matched effector conditions, that pattern is consistent with a priming or expansion effect rather than a direct change in killing potency. Conversely, a measured change in killing at matched effector number would require a separate downstream mechanism or parameter. The comparison is conceptual here; estimating tumor-killing dynamics requires its own model and data.

Derivation, Diagnostics, and Limits

The competition result follows directly from ligand conservation. Under the stated equilibrium assumptions, the three B7 concentrations are \([B]\), \([BC]=[B]w_C\), and \([BH]=[B]w_H\). Adding them gives \(B_{\mathrm{tot}}=[B](1+w_C+w_H)\). Dividing each state concentration by \(B_{\mathrm{tot}}\) yields the occupancy fractions. This also shows why increasing \(w_H\) lowers \(\theta_C\): the denominator increases while the CD28 numerator stays fixed.

The model has useful checks. All weights should be nonnegative and dimensionless; the occupancy fractions must lie between zero and one and sum to one. With no CTLA-4 binding (\(w_H=0\)), \(\theta_C=w_C/(1+w_C)\). With no CD28 (\(w_C=0\)), \(\theta_C=0\). At very high CTLA-4 weight, \(\theta_H\) approaches one and \(\theta_C\) approaches zero. If calculated fractions violate these bounds, check units, the denominator, and whether receptor concentration was confused with receptor occupancy.

When model predictions disagree with measurements, first inspect whether the assumed accessible B7 pool, receptor concentrations, and binding affinities are plausible in the relevant priming context. If increasing CTLA-4 abundance barely changes predicted CD28 occupancy, the model may be in a regime where other weights dominate, or its inputs may be insufficiently identified. Measurements of B7 availability and receptor abundance can help assess the binding component; activation and expansion measurements are needed to assess the signal-to-response component. Occupancy alone does not establish functional rescue.

A simpler alternative is a phenomenological inhibition term that reduces activation without explicitly representing B7 binding. It can be useful when only response data are available and binding parameters cannot be identified. The explicit competition model is more interpretable when receptor and ligand measurements are available, but it introduces parameters and assumptions that may not be supported by the data. Neither approach proves that CTLA-4 is the only cause of altered priming.

The competition-only formulation has important limitations. It omits possible changes in B7 abundance, ligand removal by CTLA-4-mediated processes, receptor trafficking, cell-to-cell heterogeneity, and other inhibitory or co-stimulatory inputs. It also assumes that the binding quantities used are appropriate for the relevant cell surface and time. These effects should be added only when the data and scientific question support them; otherwise, extra mechanisms can make an already weakly identified model harder to interpret.

Common Pitfall

A common mistake is to reason, “CTLA-4 blockade increases CD28 access to B7, therefore it directly increases tumor-cell killing.” The first part can follow from competitive binding under the model assumptions; the second does not. Increased CD28 occupancy can raise priming and expansion in the chosen response model, changing the number of effectors available later. A direct change in per-cell tumor killing requires separate evidence and a separate downstream model term.