Interpreting Discordant T-Cell Expansion and Tumor Response

When T-Cell Expansion and Tumor Response Disagree

A rise in measured T-cell counts does not, by itself, imply that a tumor should shrink. Counts may be measured in blood while tumor killing depends on cells entering tumor tissue; even cells present in the tumor may have reduced function, encounter antigen-negative tumor cells, or kill susceptible cells too slowly to offset tumor growth. These are distinct explanations for the same apparent mismatch between immune expansion and tumor response.

The useful modeling question is not simply whether T cells expanded, but which link between expansion and tumor-cell removal is limiting. The earlier material on diagnosing priming-limited versus effector-limited resistance motivates separating T-cell supply from activity at the tumor site. Here, the focus is on interpreting discordance after an increase in T-cell abundance.

A Minimal Model of Access, Function, and Tumor Growth

Let \(P\) denote the peripheral abundance of a tumor-reactive T-cell population and \(E\) the intratumoral abundance, scaled so that one unit of \(E\) is a chosen reference density. Let \(B_+\) and \(B_-\) denote the densities of antigen-positive and antigen-negative tumor cells, respectively. Antigen-positive cells express the target; antigen-negative cells are not recognized through that target. Let \(f\), between zero and one, be the functional fraction of the intratumoral T-cell population. A value of one represents full modeled function, and a low value represents diminished function, for example through exhaustion.

One simple representation of tumor entry is:

Here \(m\) is the entry rate per unit of peripheral abundance, \(g_E\) is the local T-cell proliferation rate, and \(\delta_E\) is the loss rate from tumor tissue. Thus, a large or rising \(P\) need not produce a large \(E\): if entry is limited, or cells are lost from the tissue, intratumoral abundance can remain low. The equation is a model assumption, not a claim that all T-cell populations follow identical kinetics.

For antigen-positive tumor cells, let \(a\) denote effective target availability, scaled from zero to one, and \(k\) the killing strength per functional T-cell unit. The per-cell killing hazard is \(h=kfEa\), with units of inverse time. With antigen-negative cells not killed through this target, a basic tumor model is:

Here \(r_+\) and \(r_-\) are the net growth rates in the absence of this T-cell killing, and \(\mu\) is a rate of transition from the antigen-positive to antigen-negative population. These equations assume that the modeled killing term acts only on antigen-positive cells and that the transition transfers cells between the two modeled populations. Other tumor-cell states or immune mechanisms could be added, but should not be inferred from these equations unless they are represented.

If the tumor is homogeneous and antigen-positive, and antigen availability is constant, then \(\mu=0\) and \(B=B_+\). The model becomes \(\frac{dB}{dt}=(r-kfEa)B\). Dividing by \(B\) gives the instantaneous relative growth rate:

This yields a local control condition: tumor burden decreases when \(kfEa>r\), is momentarily stable when \(kfEa=r\), and increases when \(kfEa<r\). It is a condition on the modeled tumor-site killing hazard, not on peripheral T-cell counts alone. It also describes the instantaneous direction of change; if the quantities vary over time, it does not by itself guarantee sustained control.

Four Mechanisms That Can Produce the Same Discordance

The model separates four potential bottlenecks. Insufficient tumor entry means \(E\) remains too small despite an increase in \(P\). Exhaustion or another functional impairment means \(f\) is low even when \(E\) is substantial. Antigen loss means an increasing share of the tumor is in \(B_-\), which is not killed by this target-specific mechanism. Weak per-cell killing means \(k\) is too small to offset growth even when entry and the functional fraction are adequate.

These mechanisms can coexist. For example, a tumor may contain both antigen-negative cells and exhausted T cells. The distinctions are useful because they point to different model changes: altering entry changes \(E\); altering functional state changes \(f\); altering antigen composition changes \(B_+\) and \(B_-\); and altering killing potency changes \(k\). Treating all four as a single “T-cell failure” parameter may fit tumor data but obscures what the model is claiming.

Interpreting the Model and Its Evidence

The four scenarios differ in their model variables, but a tumor trajectory alone may not tell them apart. In a homogeneous-tumor model, the observed net rate identifies the difference \(r-kfEa\), not each factor separately. For example, the same value of \(kfEa\) can arise from low \(E\) and high \(k\), or high \(E\) and low \(k\). A fit to tumor burden may therefore describe the net effect well while leaving the mechanism uncertain.

A mechanistic interpretation should check whether the observations align with the proposed bottleneck. A peripheral count increase paired with persistently low modeled \(E\) is consistent with an entry limitation, but a blood count alone does not measure entry. High \(E\) with low modeled \(f\) is consistent with impaired function, but abundance is not a direct measure of function. A growing antigen-negative compartment supports antigen-loss escape in this model, whereas a single total-tumor measurement cannot establish the antigen composition. High \(E\), adequate \(f\), and retained antigen with continuing growth can motivate testing a low \(k\), but other omitted growth or suppression processes could produce a similar pattern.

A useful model diagnostic is to compare predicted and observed trajectories for each state that the model represents, rather than checking only total tumor burden. If antigen-positive and antigen-negative tumor cells are modeled separately, their predicted fractions should be compared with available observations. If the model predicts abundant intratumoral cells but the observed tissue data are persistently low, its entry or retention assumptions deserve review. If it predicts effective killing while the tumor continues growing, inspect the assumed antigen availability, functional fraction, killing strength, and tumor growth rate. A mismatch is a reason to question assumptions, not automatic proof of one preferred alternative.

The main alternative is a more aggregated model that combines T-cell abundance, function, and recognition into one effective killing term, such as \(qE\), where \(q\) absorbs several processes. This can be useful when only total tumor burden and an aggregate T-cell count are available: it uses fewer parameters and may be easier to estimate. Its limitation is interpretability. A change in \(q\) cannot distinguish reduced function from weak per-cell killing or reduced antigen recognition. A state-separated model can represent those mechanisms explicitly, but needs enough information to constrain the added states and parameters. Greater mechanistic detail does not guarantee that the mechanism is identifiable.

Assumptions, Limitations, and a Common Mistake

The threshold \(kfEa>r\) assumes the modeled quantities are on compatible time and abundance scales, that the killing term is proportional to \(E\), \(f\), and \(a\), and that the net tumor growth rate is represented by \(r\). Real systems may have time-varying rates, spatially uneven T-cell distribution, saturation of killing, additional suppressive processes, or tumor-cell states not captured here. The antigen-negative compartment is not necessarily permanent: antigen expression may change over time, and the simple transition rate \(\mu\) does not describe every route to escape.

Common mistake: “T cells doubled, so the tumor should have shrunk; continued growth proves the T cells are exhausted.” This reasoning treats peripheral abundance as equivalent to tumor entry, assumes antigen is retained, and assigns a specific functional defect without evidence. The correction is to use the local balance \(kfEa\) versus \(r\), check which components are represented and supported, and recognize that the same tumor trajectory can be generated by different combinations of access, function, antigen status, and killing strength.