IL-23–Th17 Circuits in Barrier-Tissue Autoimmunity

Separating Initiation from Maintenance

In barrier-tissue autoimmunity, detecting IL-23, Th17 cells, and downstream cytokines together does not by itself show which process started inflammation or which process sustains it. A useful mechanistic model keeps those roles distinct. An initiating signal can generate or recruit a Th17-like population; IL-23 can then support the persistence, expansion, or pathogenic activity of IL-23-responsive cells. Those cells produce downstream cytokines, which may contribute to tissue inflammation.

This distinction is relevant in psoriasis and inflammatory bowel disease (IBD), but the same circuit should not be interpreted identically in every tissue. In psoriasis, IL-17-family cytokines from Th17 cells can stimulate keratinocytes to produce inflammatory mediators. In IBD, IL-23-responsive T-cell activity can contribute to disease, while IL-17 also has barrier-protective functions in some contexts. The circuit below is a deliberately compact representation of one mechanistic component, not a complete model of either disease.

Definition: In this tutorial, initiation means the process that creates an initial Th17-like cell pool or supplies new cells to it. Maintenance means continued survival, expansion, or activity of that pool after the initiating input has diminished. A signal may contribute to both processes biologically, but the model represents them as separate terms so their roles can be tested rather than assumed.

A Minimal IL-23–Th17–Cytokine Model

Let \(A(t)\) denote local IL-23 concentration, \(T(t)\) the abundance of an IL-23-responsive Th17-like cell pool, and \(C(t)\) the concentration of one downstream cytokine, such as IL-17A. These are tissue-level states; \(T\) is not a count of every T cell in the organism. One compact model is:

\frac{dA}{dt}=p_A U(t)-d_A A

\frac{dT}{dt}=S(t)+\left[r_{23}\frac{A}{K_A+A}-d_T\right]T-qT^2

\frac{dC}{dt}=p_C T-d_C C

Here \(U(t)\) is an upstream tissue-activation input that promotes IL-23 production by local antigen-presenting or other myeloid cells. \(S(t)\) is an initiating input that supplies newly differentiated or recruited Th17-like cells; it is not assumed to be generated by IL-23 in this minimal model. The saturating term \(A/(K_A+A)\) represents the fraction of maximal IL-23-dependent support, with \(K_A\) the concentration giving half-maximal support. The parameter \(r_{23}\) is the maximum IL-23-dependent growth or support rate, \(d_T\) is the loss rate of the cell pool, and \(q\) limits its growth at high abundance. For the cytokine, \(p_C\) is production per unit of \(T\), and \(d_C\) is its first-order removal rate. All rate constants have units consistent with time, and all concentrations and cell-pool units must be specified for a particular application.

This representation assumes that the measured tissue can be approximated by a well-mixed compartment; that the cell pool can be treated as one effective population; that IL-23 support is saturable; and that cytokine production is proportional to \(T\) over the modeled range. It also assumes that \(S(t)\) and \(U(t)\) can be specified independently for the question being studied. These are modeling choices, not universal biological facts. In particular, different downstream cytokines may have distinct sources, regulation, and effects, so they should be represented by separate \(C_j\) states when those distinctions matter.

Key distinction: \(S(t)\) supplies the initial cell pool, whereas \(A(t)\) changes its net maintenance rate. A model that puts every effect into one undifferentiated “inflammation” term cannot, by itself, distinguish a transient initiating event from a support signal that keeps a response active.

When Can IL-23 Support Maintenance?

After the initiating input has ended, set \(S(t)=0\). If \(A\) is approximately constant over the interval of interest, define the net low-abundance growth rate \(a\) as:

a=r_{23}\frac{A}{K_A+A}-d_T

The cell-pool equation becomes \(dT/dt=aT-qT^2\). At low abundance, the sign of \(a\) determines whether the pool initially grows or declines. If \(a\) is positive, the model has a stable positive equilibrium; if it is negative, the pool declines toward zero. This is a statement about the assumed model, not a guarantee that IL-23 alone is sufficient to maintain a real tissue response.

The threshold can be derived directly. A positive net growth rate requires:

r_{23}\frac{A}{K_A+A}>d_T

If \(r_{23}>d_T\), multiplying by \(K_A+A\), which is positive, and rearranging gives:

A>A_{\mathrm{crit}}=\frac{K_A d_T}{r_{23}-d_T}

If \(r_{23}\le d_T\), no finite \(A\) can make the bracketed net rate positive in this model: even maximal IL-23 support cannot exceed cell loss. This threshold depends on the chosen model structure and parameter values; it is not a clinical cutoff.

Derivation of the maintenance equilibrium: With \(S=0\), an equilibrium satisfies \(0=T(a-qT)\). The equilibria are \(T=0\) and, when \(a>0\), \(T^*=a/q\). The derivative of the right-hand side \(aT-qT^2\) with respect to \(T\) is \(a-2qT\). At \(T=0\) it equals \(a\), so zero is unstable when \(a>0\); at \(T^*=a/q\) it equals \(-a<0\), so the positive equilibrium is stable. Thus a brief initiating input can seed a pool that approaches \(a/q\) if IL-23 support remains sufficiently high.

Worked Example: Finding the IL-23 Maintenance Threshold

Consider a hypothetical tissue model with \(r_{23}=0.80\ \text{day}^{-1}\), \(d_T=0.30\ \text{day}^{-1}\), and \(K_A=2.0\) concentration units. The model assumes \(r_{23}>d_T\), so a finite threshold exists.

Calculate the threshold.

A_{\mathrm{crit}}=\frac{K_A d_T}{r_{23}-d_T} =\frac{(2.0)(0.30)}{0.80-0.30} =\frac{0.60}{0.50} =1.20\ \text{concentration units}

At \(A=2.0\) concentration units, the IL-23-dependent rate is \(0.80(2.0)/(2.0+2.0)=0.40\ \text{day}^{-1}\). Thus \(a=0.40-0.30=0.10\ \text{day}^{-1}\), which is positive: the model permits maintenance. At \(A=0.50\), the support rate is \(0.80(0.50)/(2.0+0.50)=0.16\ \text{day}^{-1}\), giving \(a=0.16-0.30=-0.14\ \text{day}^{-1}\). The cell pool then declines after initiation has stopped. These conclusions concern this hypothetical parameterization, not measured IL-23 thresholds in patients.

Worked Examples: Persistence and Tissue Cytokine Output

Worked Example: A Transient Initiation Followed by Different Maintenance Conditions

Suppose a brief initiating input leaves \(T(0)=1.0\) normalized cell-pool unit, after which \(S(t)=0\). Use the two net rates from the preceding example and \(q=0.050\) per normalized cell-pool unit per day. For constant \(a\), the equation \(dT/dt=aT-qT^2\) has solution \(T(t)=T_0e^{at}/[1+(qT_0/a)(e^{at}-1)]\) when \(a\ne0\).

With high IL-23 support, \(a=0.10\ \text{day}^{-1}\). After five days:

T(5)=\frac{(1.0)e^{0.10(5)}}{1+\left(\frac{0.050(1.0)}{0.10}\right)(e^{0.10(5)}-1)} =\frac{1.6487}{1+0.5(0.6487)} =1.245

The pool has increased from 1.0 to about 1.245 units and tends toward \(a/q=0.10/0.050=2.0\) units. With low IL-23 support, \(a=-0.14\ \text{day}^{-1}\):

T(5)=\frac{(1.0)e^{-0.14(5)}}{1+\left(\frac{0.050(1.0)}{-0.14}\right)(e^{-0.14(5)}-1)} =\frac{0.4966}{1+(-0.3571)(-0.5034)} =0.421

The same initial pool therefore grows under the first assumed condition and declines under the second. This comparison illustrates maintenance, not initiation: both scenarios begin with the same \(T(0)\), and the initiating input is off in both.

Worked Example: Calculating a Downstream Cytokine Trajectory

In a separate hypothetical calculation, hold the Th17-like pool at \(T=1.5\) units. Let \(p_C=2.0\) cytokine units per cell-pool unit per day, \(d_C=0.50\ \text{day}^{-1}\), and \(C(0)=0\). If \(T\) is constant, the cytokine equilibrium is \(C^*=p_C T/d_C\), and the trajectory is \(C(t)=C^*+[C(0)-C^*]e^{-d_Ct}\).

First, \(C^*=(2.0)(1.5)/0.50=6.0\) cytokine units. After two days:

C(2)=6.0+(0-6.0)e^{-0.50(2)} =6.0(1-e^{-1}) =3.793\ \text{cytokine units}

The modeled cytokine has risen toward its steady value but has not reached it. This lag follows from finite cytokine removal; it does not establish that the cytokine caused the initiating cell response. In an application, the cytokine, cell-pool units, and tissue compartment would need to be defined to make the value interpretable.

Interpreting the Circuit in Psoriasis and IBD

For psoriasis, the model can represent an IL-23-supported Th17-like pool whose IL-17-family output contributes to keratinocyte activation and local inflammation. It is a useful abstraction when the question concerns whether support can sustain an already seeded population. It does not represent every cellular source of IL-17, every epidermal process, or the full set of mediators that produce plaques.

For IBD, the same mathematical structure can represent local IL-23 support of an IL-23-responsive T-cell component and production of a selected downstream cytokine. However, it should not be read as saying that all intestinal inflammation is Th17-mediated, or that every downstream cytokine is harmful. IL-17 can contribute to inflammation in some settings and support epithelial barrier defense in others. A model that treats all cytokine output as a single damaging variable would conceal this ambiguity rather than resolve it.

To compare tissues, one could use the same model structure but estimate tissue-specific production, loss, and support parameters, or add states when evidence requires them. A shared structure does not imply shared parameter values or identical biological effects. In either disease, an observed association between IL-23 and downstream cytokines alone cannot establish the direction of causation; time-resolved measurements and appropriate mechanistic evidence are needed to discriminate candidate explanations.

Assumptions, Diagnostics, and Alternatives

Before applying the model, check whether the data and question support its simplifications. The single-pool representation is questionable if distinct IL-23-responsive cell populations have different kinetics or cytokine outputs. The proportional \(p_C T\) production term is questionable if cytokine production per cell changes substantially with activation state. A constant or externally specified \(U(t)\) may be inadequate if upstream activation changes in response to tissue inflammation. These limitations are reasons to compare model structures, not reasons to interpret a fitted parameter as a direct biological measurement.

Useful diagnostics include checking whether the model reproduces the timing and direction of measured IL-23, cell abundance, and cytokine changes; examining whether residuals show persistent time patterns; and testing whether estimates of \(r_{23}\), \(d_T\), and \(q\) are separately identifiable from the available observations. If only cytokine concentrations are measured, a rise in \(C\) could reflect more \(T\), greater production per cell, slower clearance, or a different source. The simple model cannot distinguish these possibilities without additional data or assumptions.

Two alternatives clarify what the minimal model does and does not claim. A model in which IL-23 directly drives the initiation term \(S(t)\) would assign IL-23 a role in creating new Th17-like cells, not only supporting an existing pool. That structure should be chosen only when the question and evidence support it; it is not interchangeable with the maintenance formulation. At the other extreme, a static association model can describe covariance between IL-23 and cytokines but cannot represent persistence after an initiating input ends. Comparing such alternatives requires data informative about the timing of the initiating signal, cell abundance, and cytokine output.

Common pitfall: “IL-23 is elevated when Th17-associated cytokines are elevated, so IL-23 must have initiated the response.” This confuses co-occurrence with causal order and ignores the possibility that IL-23 supports a population established by another input. In the model, initiation is represented by \(S(t)\), while IL-23 changes the net maintenance rate. The distinction is a hypothesis to evaluate, not a conclusion established by a cross-sectional correlation.

The model also does not, by itself, predict the clinical response to an intervention or prove that a cytokine is a disease driver. Such conclusions require evidence beyond the circuit’s structure. Its immediate value is narrower: it makes explicit how an initiating input, IL-23-dependent maintenance, and downstream cytokine production can be separated and checked against data.

Check Your Understanding

Answer each question, then compare with the solutions below.

Using \(r_{23}=0.60\ \text{day}^{-1}\), \(d_T=0.20\ \text{day}^{-1}\), and \(K_A=1.5\) concentration units, calculate \(A_{\mathrm{crit}}\).
In the model, what does it mean if \(S(t)=0\) but \(a>0\)? Does it mean IL-23 initiated the response?
For constant \(T=2.0\) units, \(p_C=1.5\) cytokine units per cell-pool unit per day, and \(d_C=0.50\ \text{day}^{-1}\), calculate the steady-state cytokine concentration.
Why should a model of IBD not automatically interpret every increase in IL-17 as increased tissue damage?
Name one observation that could help distinguish increased cell abundance from reduced cytokine clearance as explanations for a higher measured cytokine concentration.

Solutions. (1) \(A_{\mathrm{crit}}=(1.5)(0.20)/(0.60-0.20)=0.75\) concentration units. (2) The initiating input is off, but the net growth rate is positive, so the seeded pool can expand toward its positive equilibrium. This does not mean IL-23 created the initial pool; that is represented separately by \(S(t)\). (3) \(C^*=p_C T/d_C=(1.5)(2.0)/0.50=6.0\) cytokine units. (4) IL-17 may have inflammatory effects and may also contribute to barrier defense, depending on tissue context; concentration alone does not specify its net effect. (5) Measure the cell pool alongside cytokine concentrations over time, or obtain an independent estimate of cytokine clearance. Either can help separate production-related explanations from slower removal.