Why the Direction of Cell Flow Matters
Two cell populations can change in similar ways over a short observation period while following different mechanisms. A cell may move between phenotypes in both directions, or it may leave a precursor phenotype by differentiating into a state from which it does not return. These mechanisms have different implications for long-term composition: reversible transitions can settle into a stable mixture, whereas one-way differentiation can progressively deplete the precursor population.
The earlier material on phenotype switching introduced transitions between cell states. Here the distinction is whether flow can occur in both directions or only from a precursor into a terminal state. The equations below isolate that distinction by assuming no cell division, death, or external input. This makes the consequences of transition direction explicit; additional processes can be added when the data or biology require them.
Models and Assumptions
Let \(A(t)\), \(B(t)\), and \(D(t)\) denote cell counts at time \(t\). In the reversible model, let \(k_{AB}\) be the rate per cell for transition from \(A\) to \(B\), and \(k_{BA}\) the rate per cell for transition from \(B\) to \(A\). Both rates have units of inverse time. The transition flows per unit time are \(k_{AB}A\) and \(k_{BA}B\).
This model assumes that the transition rates are constant over the modeled interval, each flow is proportional to the number of cells in its source state, and cells are neither created nor lost. It also treats each state as sufficiently homogeneous for one rate to represent its cells. The equations are deterministic descriptions of population counts; stochastic variation may matter when counts are small.
For irreversible differentiation, let \(k_{AD}\) be the per-cell rate at which precursor cells in \(A\) differentiate into \(D\). Under the same no-division, no-death, no-input assumptions, the equations are:
The key structural difference is not simply that the labels differ. In the reversible model, each state is a source for one flow and a destination for another. In the differentiation model, \(D\) has incoming flow but no outgoing transition flow. Thus an apparent decline in \(A\) alone does not establish irreversible differentiation; the accompanying cell counts and the possibility of reverse flow matter.
Deriving the Distinct Cell-Flow Consequences
For reversible transitions, adding the two equations gives \(d(A+B)/dt=0\). Therefore the total \(N=A+B\) is conserved. Substituting \(B=N-A\) into the equation for \(A\) gives a single differential equation. Its equilibrium value is obtained by setting the derivative to zero.
If both rates are zero, then \(dA/dt=dB/dt=0\), so both counts remain at their initial values and every initial composition is an equilibrium. Subtracting the equilibrium value from both sides of the dynamic equation yields \(d(A-A_{\mathrm{eq}})/dt=-(k_{AB}+k_{BA})(A-A_{\mathrm{eq}})\). Solving this first-order equation, and using \(B=N-A\), gives:
When both rates are positive, the two-state system approaches a stable mixture. At equilibrium the directional flows are equal, \(k_{AB}A_{\mathrm{eq}}=k_{BA}B_{\mathrm{eq}}\). This does not mean that cells stop transitioning: both gross flows can remain positive while their difference, and hence the net change in either count, is zero.
For one-way differentiation, solving \(dA/dt=-k_{AD}A\) gives \(A(t)=A(0)e^{-k_{AD}t}\). Since \(A+D\) is conserved under the stated assumptions, the differentiated-cell count is \(D(t)=D(0)+A(0)-A(t)\). When \(k_{AD}>0\), \(A\) approaches zero and \(D\) approaches the initial combined count; when \(k_{AD}=0\), both counts remain constant. The transition flow \(k_{AD}A(t)\) declines as the precursor pool is depleted.
Worked Example: Reversible Switching Approaches a Stable Mixture
Suppose a hypothetical cell system has \(A(0)=800\) cells and \(B(0)=200\) cells. Cells transition from \(A\) to \(B\) at \(k_{AB}=0.30\) per day and from \(B\) to \(A\) at \(k_{BA}=0.10\) per day. Assume no division, death, or external input. Find the counts after 3 days.
The conserved total is \(N=800+200=1000\) cells. The equilibrium count in \(A\) is:
The decay rate toward equilibrium is \(0.30+0.10=0.40\) per day. Therefore:
The remaining count is \(B(3)=1000-415.7=584.3\) cells, using the unrounded values in the calculation. The total remains 1000 cells, and the counts are moving toward the equilibrium values of 250 cells in \(A\) and 750 cells in \(B\); they have not reached equilibrium by day 3.
Worked Example: One-Way Differentiation Depletes the Precursor
In a separate hypothetical system, \(A(0)=600\) precursor cells and \(D(0)=100\) terminal-state cells. Differentiation occurs at \(k_{AD}=0.20\) per day. With no cell division, death, or external input, calculate both counts after 4 days.
The precursor count is:
The total is conserved at \(600+100=700\) cells, so:
Approximately \(600-269.6=330.4\) precursor cells have entered \(D\) by day 4. That is cumulative differentiation in this no-death model, not the number of cells differentiating per day at day 4. The instantaneous differentiation flow at day 4 is \(k_{AD}A(4)=0.20(269.6)\approx53.9\) cells per day.
Gross Flows, Net Changes, and Trajectory Diagnostics
Counts show net change, not every transition. In the reversible model, the net rate of change in \(A\) is the flow into \(A\) minus the flow out of \(A\): \(dA/dt=k_{BA}B-k_{AB}A\). Consequently, a small observed change in \(A\) could reflect small flows in both directions, or large nearly balanced flows. The counts alone do not reveal the two gross rates unless the data contain enough information to identify them.
Worked Example: A Rising Count Can Coexist With Flow in Both Directions
Suppose \(A(0)=300\) cells and \(B(0)=700\) cells, with \(k_{AB}=0.20\) per day and \(k_{BA}=0.10\) per day. The total is 1000 cells, and the equilibrium count in \(A\) is \(0.10(1000)/0.30=333.3\) cells. Find \(A\) after 5 days and interpret the flows then.
Using the reversible solution:
Thus \(B(5)=1000-325.9=674.1\) cells. The directional flows at day 5 are approximately \(0.20(325.9)=65.2\) cells per day from \(A\) to \(B\), and \(0.10(674.1)=67.4\) cells per day from \(B\) to \(A\). Their difference is \(67.4-65.2=2.2\) cells per day into \(A\), consistent with \(A\) still increasing toward its equilibrium count. Observing that increase does not mean the \(A\)-to-\(B\) transition has stopped.
Under the constant-rate, two-state reversible model, \(A(t)\) approaches its equilibrium monotonically from its initial value: the sign of \(A(t)-A_{\mathrm{eq}}\) does not change in the solution. Under the simple one-way model, \(A\) cannot rise and \(D\) cannot fall. These are useful diagnostics, not universal biological laws. A trajectory that violates them can indicate proliferation, death, changing rates, external input, measurement error, or omitted states rather than proving one particular alternative mechanism.
A practical comparison is to fit or simulate both structures against repeated counts of all relevant states, checking whether their predicted directions and time courses match the observations. A constant total across \(A\) and \(B\) is consistent with reversible switching under the no-growth assumptions; a constant total across \(A\) and \(D\) is consistent with the corresponding no-loss differentiation model. But a single total or a few snapshots rarely establish transition direction. Lineage information or direct measurement of transitions may be needed to distinguish mechanisms that produce similar counts over a limited interval.
Choosing a Model and Recognizing Its Limits
| Feature | Reversible transition \(A\rightleftarrows B\) | Irreversible differentiation \(A\to D\) |
|---|---|---|
| Transition directions | Both states can be sources and destinations | \(A\) supplies \(D\); no return transition |
| Under the stated assumptions | \(A+B\) is conserved; counts approach a mixture | \(A+D\) is conserved; when \(k_{AD}>0\), \(A\) declines and \(D\) accumulates |
| Equilibrium behavior | Positive rates give a stable mixture with balanced flows | When \(k_{AD}>0\), all remaining cells eventually enter \(D\) |
| Useful evidence | Return transitions or persistent balanced directional flux | Lineage evidence of one-way entry and no return over the relevant timescale |
These are minimal models, not exhaustive descriptions of cell biology. If cells divide or die, those processes add terms to the equations and total cell count need not be conserved. For example, if cells in \(D\) die at a first-order rate \(\delta_D\), then \(dD/dt=k_{AD}A-\delta_DD\); in that model, \(D\) need not increase monotonically. If transition rates vary with time or depend on treatment or population density, constant-rate solutions no longer apply directly. More than two phenotypes, hidden intermediates, and cell-to-cell heterogeneity may also require additional states or a different model structure.
The main modeling choice is therefore mechanistic: does the biological question allow a cell in the destination state to return to the precursor or alternative phenotype? A reversible model is appropriate when return is plausible and relevant on the study timescale. A one-way model is appropriate when differentiation is effectively terminal on that timescale. Treating a reversible transition as irreversible can overstate precursor depletion; allowing a return flow when none exists can understate the accumulation of terminal cells. Both choices should be checked against biology and the observed trajectories.
Common Mistakes
- Equating net change with transition direction. A stable count in \(A\) does not imply no transitions. In the reversible model, \(k_{AB}A\) and \(k_{BA}B\) can both be positive and equal.
- Calling any declining precursor count irreversible differentiation. Death, reduced proliferation, or movement into an unmeasured state can also lower \(A\). The mechanism requires evidence beyond the decline itself.
- Assuming “terminal” means immortal. Terminal describes the absence of a return transition in the model. Death can still occur and must be modeled separately if relevant.
- Using conservation when other flows are present. \(A+B\) or \(A+D\) is constant only under the stated assumptions. Division, death, and external input change the balance.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A reversible system has \(A(0)=500\), \(B(0)=500\), \(k_{AB}=0.40\) per day, and \(k_{BA}=0.10\) per day. What is \(A(2)\), and is \(A\) initially increasing or decreasing?
- For one-way differentiation, \(A(0)=400\), \(D(0)=50\), and \(k_{AD}=0.25\) per day. Find \(A(2)\) and \(D(2)\), assuming no other processes.
- In a reversible model, the count in \(A\) is constant at one observation time. Does that establish that transitions have stopped? Explain.
- In a proposed one-way model with no other processes, repeated measurements show \(D\) falling while \(A\) rises. Is the model consistent with those observations? Give at least one possible explanation for the mismatch.
Solutions. 1. The total is 1000 cells and \(A_{\mathrm{eq}}=0.10(1000)/(0.40+0.10)=200\) cells. Hence \(A(2)=200+(500-200)e^{-1.0}=200+300(0.367879)=310.4\) cells. Initially, \(dA/dt=-0.40(500)+0.10(500)=-150\) cells per day, so \(A\) is decreasing. 2. \(A(2)=400e^{-0.25(2)}=400e^{-0.5}\approx242.6\) cells. The conserved total is 450 cells, so \(D(2)=450-242.6=207.4\) cells. 3. No. If \(A\) is known to be locally constant at that time, then \(dA/dt=0\), so the model implies \(k_{AB}A=k_{BA}B\); a single count observation cannot establish this. 4. No. With only \(A\to D\) and no other processes, \(A\) cannot rise and \(D\) cannot fall. The mismatch could reflect omitted proliferation or death, a return transition, external input, measurement error, or a model that combines states inappropriately.