Antigen Presentation and Priming of Tumor-Reactive T Cells

From Tumor Antigen to T-Cell Priming

A tumor-reactive T cell cannot be primed by tumor-antigen abundance alone. Antigen must become available for uptake, dendritic cells (DCs) must carry and present it in a lymph node, and a matching naïve T-cell precursor must encounter that presentation. A QSP model makes these links explicit, helping distinguish a failure to generate antigen from a failure to transport or present it.

The earlier material on T-cell state compartments in tumor-immune models provides the broader compartment framework. Here the focus is the upstream chain from tumor antigen to priming, followed by a compact representation of activated-cell expansion. Co-stimulation and inhibitory signals are not resolved here; their influence is treated as part of effective priming and expansion parameters, rather than explained mechanistically.

A Linked Compartment Model

Let \(A(t)\) be available tumor antigen, measured in ng; \(B(t)\) be antigen-bearing DC cargo in the draining lymph node, measured in ng antigen-equivalent; and \(P(t)\) be a presentation level, measured in presentation units. Let \(N(t)\) and \(E(t)\) denote numbers of tumor-reactive naïve and activated T cells. A minimal model is:

Here \(q_A\) is antigen release into the available pool (ng/day), \(k_A\) is antigen loss (day\(^{-1}\)), and \(U_{\max}\) and \(K_A\) set the maximum DC uptake flux (ng/day) and its half-saturation level (ng). The fraction \(f\), between zero and one, is the fraction of taken-up antigen that reaches the modeled lymph-node cargo pool. The rates \(k_B\) and \(k_P\) describe cargo loss and presentation loss; \(k_{\mathrm{proc}}\) converts cargo into presentation units per day. The priming rate constant \(k_{\mathrm{prim}}\), death rate \(d_E\), and expansion rate \(r_E\) have units day\(^{-1}\). \(K_P\) is in presentation units, and \(E_{\max}\) is in cells.

The equations assume a well-mixed antigen pool, constant parameters over the time window of interest, first-order loss of cargo and presentation, and a saturable uptake flux. They also assume that one aggregate \(P\) adequately represents relevant peptide-MHC presentation and that the tumor-reactive precursor pool can be represented by a single \(N\). In a short-term application, parameters may be treated as fixed even though DC number, antigen processing, and T-cell composition can vary biologically.

The model does not imply that presentation alone determines activation in every setting. The priming equation is an effective description conditional on the biological context represented by its parameters. It does not distinguish individual DC subsets, peptide identities, spatial encounters, or the separate signals involved in activation.

Steady States and a Presentation Delay

Suppose antigen availability \(A\) is held constant. The antigen uptake flux is \(U(A)=U_{\max}A/(K_A+A)\), and the input to the lymph-node cargo pool is \(J_B=fU(A)\). At steady state, inflow balances loss in each downstream compartment:

These expressions follow by setting \(dB/dt=0\) and \(dP/dt=0\) in their respective equations and solving. If \(A\) itself is dynamic, its steady state instead satisfies \(q_A=k_AA+U(A)\). With \(q_A>0\) and \(k_A>0\), the right-hand loss grows without bound as \(A\) grows, so there is a finite positive steady state. The downstream steady states can then be calculated using that value of \(A\).

The chain also produces a delay: antigen uptake can begin before lymph-node cargo and presentation approach their steady levels. For a constant input \(J_B\), initially empty \(B\) and \(P\), and positive rates \(k_B\ne k_P\), solving the two linear equations gives:

The expression for \(P(t)\) is the accumulated effect of processing cargo over time, minus presentation loss. It is zero at \(t=0\) and tends to \(k_{\mathrm{proc}}J_B/(k_Bk_P)\), the steady state above. When \(k_B=k_P\), the expression with \(k_P-k_B\) in the denominator is replaced by its continuous limit; the underlying differential equations remain well-defined.

What the Model Can and Cannot Distinguish

Explicit compartments are useful when measurements or interventions can separate stages of the pathway. For example, antigen measurements can inform \(A\), DC cargo or migration measurements can inform \(B\) and \(f\), presentation assays can inform \(P\), and T-cell measurements can inform priming and expansion. If only activated T-cell counts are observed, a low response could reflect low antigen release, poor uptake, inefficient transport, rapid presentation loss, few precursors, or a small priming rate. Those mechanisms may produce similar outputs and cannot necessarily be estimated separately from that observation alone.

A simpler alternative is to replace the antigen and DC compartments with a measured or fitted presentation input. This reduces the number of parameters and may be appropriate when intermediate data are unavailable, but it cannot predict how an intervention on uptake or DC migration changes presentation. Conversely, adding separate DC subsets, processing steps, and peptide-specific states is justified only when data can support those distinctions. A more elaborate model is not automatically more informative.

The steady-state formula shows why uptake, transport, processing, and presentation loss can be difficult to disentangle: \(P^*\) depends on their combined effect. Time-course measurements can help because the approach to steady state reflects \(k_B\) and \(k_P\), but adequate sampling and informative perturbations are still needed. Fitting a model to T-cell expansion alone does not establish that the upstream mechanism is correct.

Diagnostics, Failure Modes, and Limitations

• Check units and signs. Every influx and loss term should have units of amount or cells per day. Loss terms should reduce the relevant state when that state is positive.
• Check positivity and plausible scale. With nonnegative inputs and positive loss rates, the antigen, cargo, and presentation states should remain nonnegative. Negative simulated presentation suggests a coding, parameter, or solver problem.
• Look for implausible accumulation. Under constant antigen input, saturation of the uptake flux alone does not guarantee bounded antigen. If \(k_A=0\) and \(q_A>U_{\max}\), then at high \(A\), \(dA/dt\) approaches \(q_A-U_{\max}>0\), so antigen continues to accumulate. A predicted presentation level that rises without bound under constant input points to missing loss or another mechanism that makes the net flux nonpositive at high presentation levels; saturation of the input alone is not sufficient.
• Inspect timing as well as endpoint values. A model may match steady presentation but predict an implausibly fast response if it omits a meaningful transport or processing delay.
• Do not equate presentation with cell activation in every context. A high \(P\) predicts a high value of \(g(P)\) in this model, but it does not prove that the biological conditions for priming are otherwise met.

The equations compress spatial movement and cell-to-cell encounters into rates, and the \(P\) state combines many possible peptide-MHC species into one level. The logistic expansion term is phenomenological; it does not identify the biological source of a finite expansion limit. The model is therefore most useful for testing pathway-level hypotheses and organizing measurements, not for claiming that every mechanistic detail has been resolved.

Common Mistake

A common error is to argue that saturating antigen uptake must make antigen abundance bounded. Saturation only caps the uptake flux; it does not cap the antigen pool itself. For example, if the antigen source is \(q_A=5\) ng/day, the maximum uptake is \(U_{\max}=3\) ng/day, and there is no other antigen loss, then at high antigen levels the net change approaches \(5-3=2\) ng/day. The pool keeps growing. A bounded prediction requires a loss process or another mechanism that makes net input nonpositive at high abundance. This check applies to presentation as well: a saturated input into \(P\) still accumulates without bound if presentation has no loss.