Drug Effects on Expression Versus Protein Stability

Why the Mechanism Matters

A lower protein concentration after treatment does not, by itself, reveal how the drug caused the change. The drug might reduce the production of mRNA, reduce translation of existing mRNA, or increase protein removal. These mechanisms can produce similar long-term protein levels while generating different time courses and different mRNA responses. Distinguishing them matters when interpreting a pharmacodynamic measurement, choosing which biomarker to measure, or deciding whether a simple turnover model is adequate.

The comparison here uses the coupled first-order mRNA–protein model and step-response ideas from the earlier material on mRNA and protein turnover. It focuses on three changes introduced at treatment time: a reduction in transcription, a reduction in translation, or an increase in protein degradation. The model assumes no feedback from protein to gene expression and no drug-induced change in other rates.

Baseline Model and Assumptions

Let \(m(t)\) be mRNA abundance and \(p(t)\) be protein abundance. In the absence of drug, mRNA is synthesized at rate \(k_m\) and degraded at first-order rate \(k_{dm}m\). Protein is translated at rate \(k_t m\) and degraded at first-order rate \(k_{dp}p\). Thus,

\frac{dm}{dt}=k_m-k_{dm}m,\qquad \frac{dp}{dt}=k_t m-k_{dp}p.

Here \(k_m\) has units of mRNA amount per time, \(k_t\) has units of protein amount per mRNA amount per time, and \(k_{dm}\) and \(k_{dp}\) have units of inverse time. Assume all rates are nonnegative and, for the steady-state formulas below, specifically require \(k_{dm}>0\) and \(k_{dp}>0\). The system is at its pre-treatment steady state at \(t=0\), so

m_0=\frac{k_m}{k_{dm}},\qquad p_0=\frac{k_t m_0}{k_{dp}}.

The normalized trajectories \(m(t)/m_0\) and \(p(t)/p_0\) require positive baseline abundances. In the worked normalized comparisons, assume \(k_m>0\), \(k_t>0\), and therefore \(m_0>0\) and \(p_0>0\). If a baseline abundance is zero, its normalized trajectory is undefined; use the unnormalized differential equations instead. For example, if \(k_m=0\) with \(k_{dm}>0\), the steady-state mRNA abundance is zero, not a positive reference for normalization.

The drug effect is treated as an immediate, constant step at \(t=0\). Let \(a\) be the residual fraction of transcription, \(b\) the residual fraction of translation, and \(g\) the multiplier on protein degradation. We use \(0\leq a\leq1\), \(0\leq b\leq1\), and \(g\geq1\). The model assumes these effects do not change mRNA degradation, protein translation machinery beyond the stated multiplier, or other processes. These are mechanistic assumptions, not facts guaranteed by observing a drug response.

Three Mechanisms and Their Predicted Trajectories

For transcriptional inhibition, the mRNA production rate changes from \(k_m\) to \(a k_m\), while the other rates remain unchanged. The resulting mRNA trajectory is

\frac{m(t)}{m_0}=a+(1-a)e^{-k_{dm}t}.

The new steady-state mRNA level is \(a m_0\). Protein production depends on the changing mRNA abundance, so protein generally responds more slowly. Writing \(\alpha=k_{dm}\) and \(\beta=k_{dp}\), the normalized protein trajectory, when \(\alpha\ne\beta\), is

\frac{p(t)}{p_0} =a+(1-a)\frac{\beta e^{-\alpha t}-\alpha e^{-\beta t}}{\beta-\alpha}.

When the rates are equal, \(\alpha=\beta\), the expression has the finite limit

\frac{p(t)}{p_0}=a+(1-a)(1+\beta t)e^{-\beta t}.

For translational inhibition, mRNA production and degradation are unchanged, so \(m(t)=m_0\). Translation becomes \(b k_t m\), giving

\frac{p(t)}{p_0}=b+(1-b)e^{-k_{dp}t}.

For accelerated protein degradation, mRNA and translation remain unchanged, but the protein degradation rate becomes \(g k_{dp}\). The predictions are

m(t)=m_0,\qquad \frac{p(t)}{p_0}=\frac{1}{g}+\left(1-\frac{1}{g}\right)e^{-g k_{dp}t}.

Both translational inhibition and accelerated degradation leave mRNA unchanged in this model. Their protein trajectories differ in their final levels and time constants: translation inhibition approaches the residual fraction \(b\) with time constant \(1/k_{dp}\), while degradation acceleration approaches \(1/g\) with time constant \(1/(g k_{dp})\).

Why Transcriptional Inhibition Can Have a Delayed Protein Response

The key feature of the transcriptional case is that protein initially remains at its baseline level even though mRNA starts to fall. This follows directly from the protein balance at \(t=0\): before treatment, \(k_t m_0=k_{dp}p_0\), so the initial protein slope after transcription is inhibited is still zero. Protein production falls only as mRNA falls, creating a lag in the protein response.

A short derivation gives the full trajectory. Set \(y=p/p_0\) and \(f=m/m_0=a+(1-a)e^{-\alpha t}\). Since \(p_0=k_t m_0/\beta\), the protein equation becomes

\frac{dy}{dt}=\beta(f-y),\qquad y(0)=1.

Subtract the new steady-state fraction \(a\), and let \(z=y-a\). Then

\frac{dz}{dt}+\beta z=\beta(1-a)e^{-\alpha t},\qquad z(0)=1-a.

Multiplying by the integrating factor \(e^{\beta t}\) and integrating from \(0\) to \(t\) gives

e^{\beta t}z(t)-(1-a) =\beta(1-a)\int_0^t e^{(\beta-\alpha)s}\,ds.

For \(\alpha\ne\beta\), evaluating the integral and rearranging yields \(z(t)=(1-a)(\beta e^{-\alpha t}-\alpha e^{-\beta t})/(\beta-\alpha)\), which gives the stated protein trajectory. When \(\alpha=\beta\), the integral is \(t\), producing the equal-rate expression. This derivation also shows why both turnover rates contribute to the protein time course.

Worked Example: Transcriptional Inhibition With a Slow Protein Response

Suppose baseline mRNA is \(100\) copies per cell, with \(k_m=50\) copies per cell per day and \(k_{dm}=0.50\) per day. Let \(k_t=0.20\) protein copies per mRNA copy per day and \(k_{dp}=0.10\) per day. A treatment reduces transcription to \(a=0.40\) of baseline. Find the relative mRNA and protein abundances after 2 days.

The baseline values are consistent with steady state: \(m_0=k_m/k_{dm}=50/0.50=100\) copies per cell, and \(p_0=k_t m_0/k_{dp}=(0.20)(100)/0.10=200\) protein copies per cell. The required degradation rates are positive, and \(p_0>0\), so normalization is valid.

With \(t=2\) days, \(k_{dm}t=1\). Therefore,

\frac{m(2)}{m_0} =0.40+(1-0.40)e^{-1} =0.40+0.60(0.3679) =0.6207.

The mRNA abundance is \(0.6207(100)=62.07\) copies per cell. For protein, \(\alpha=0.50\) per day and \(\beta=0.10\) per day, so

\frac{p(2)}{p_0} =0.40+0.60 \frac{0.10e^{-1}-0.50e^{-0.2}}{0.10-0.50} \approx0.40+0.60(0.9314436) \approx0.9589.

Thus protein is \(0.9589(200)=191.8\) copies per cell after 2 days. The modest early protein decline is consistent with the relatively slow protein turnover and the lagged response to falling mRNA; it does not imply that transcription was only weakly inhibited.

Worked Example: Translational Inhibition

Suppose \(m_0=80\) mRNA copies per cell and \(p_0=500\) protein copies per cell at baseline. Translation is reduced to \(b=0.30\) of baseline, while \(k_{dp}=0.20\) per day. Calculate mRNA and protein after 5 days.

This mechanism leaves mRNA synthesis and degradation unchanged, so \(m(5)=80\) copies per cell. The model requires no change in \(k_{dp}\), and the positive baseline abundances permit normalization. For protein,

\frac{p(5)}{p_0} =0.30+(1-0.30)e^{-(0.20)(5)} =0.30+0.70e^{-1} =0.5575.

The predicted protein abundance is \(0.5575(500)=278.8\) copies per cell. The eventual steady-state protein fraction is \(0.30\), but the value at day 5 is higher because pre-existing protein has not yet been fully removed.

Worked Example: Accelerated Protein Degradation

A treatment triples protein degradation, so \(g=3\), from a baseline rate \(k_{dp}=0.10\) per day. Let baseline mRNA and protein be \(m_0=60\) copies per cell and \(p_0=300\) copies per cell. Find the relative and absolute protein abundance after 4 days.

The mRNA trajectory remains \(m(t)=60\) copies per cell under this mechanism. Because the baseline abundances are positive and \(k_{dp}>0\), the normalized protein formula applies. The new degradation rate is \(gk_{dp}=0.30\) per day, giving

\frac{p(4)}{p_0} =\frac{1}{3}+\frac{2}{3}e^{-(0.30)(4)} \approx0.3333+0.6667(0.3012) \approx0.5341.

The predicted protein abundance is \(0.5341(300)=160.2\) copies per cell. The long-run fraction is \(1/3\), and the accelerated degradation makes the approach to that lower level faster than the baseline protein turnover time course.

Comparing the Predictions and Checking Model Fit

Mechanism	mRNA after treatment	Long-run protein fraction	Protein response feature
Transcriptional inhibition	Declines toward \(a m_0\)	\(a\)	Typically lags behind the mRNA decline
Translational inhibition	Unchanged at \(m_0\)	\(b\)	Approaches \(b\) with rate \(k_{dp}\)
Accelerated protein degradation	Unchanged at \(m_0\)	\(1/g\)	Approaches \(1/g\) with rate \(gk_{dp}\)

At the instant of treatment, the predicted slopes also differ. Under transcriptional inhibition, \(dm/dt=-(1-a)k_m\), while the protein slope is initially zero because protein production and degradation still balance at baseline. Under translational inhibition, the initial protein slope is \(-(1-b)k_t m_0\). Under accelerated degradation it is \(-(g-1)k_{dp}p_0\). These differences may be informative if sampling begins early enough and measurements are sufficiently precise.

Paired mRNA and protein measurements are therefore more informative than protein alone for separating transcriptional inhibition from the other two mechanisms under this model. However, unchanged mRNA does not distinguish reduced translation from faster protein degradation by itself. Their steady-state protein fractions can even match: if \(b=1/g\), both predict the same long-run protein level. Time-course data may separate them because their response rates differ, but a single protein measurement cannot generally identify the mechanism and its parameters uniquely.

Diagnostics should ask whether the observed data agree with the model’s assumptions. A changing mRNA level when only translation or degradation was hypothesized suggests either a different mechanism, indirect regulation, measurement variability, or failure of the no-feedback assumption. A protein trajectory that is not a smooth approach to the predicted steady state may indicate time-varying drug exposure, delayed onset, additional compartments, or non-first-order turnover. Sampling only after a long time can miss the lag under transcriptional inhibition; sampling only very early may miss the eventual response. Measurements of mRNA and protein should therefore be interpreted with their collection times and assay properties in view.

The main alternatives answer different questions. A single-exponential fit to protein can summarize an observed decline but does not, by itself, identify whether translation or degradation changed. An endpoint-only comparison can estimate a long-run effect if steady state has actually been reached, but discards the transient information that may help distinguish mechanisms. The coupled model links mRNA and protein trajectories through explicit processes, at the cost of assuming constant rates, an instantaneous step effect, and no unmodeled feedback. If those assumptions are implausible, fitted mechanistic parameters should not be treated as uniquely established by the data.

Common Mistakes and Limitations

Equating a protein decline with reduced transcription. Reduced translation and faster degradation can also lower protein. Check mRNA and the time course, not just the protein endpoint.
Expecting protein to fall immediately after transcription is inhibited. At the treatment instant, protein production still balances degradation because mRNA has not yet changed. Protein responds as mRNA declines.
Confusing residual fractions and inhibition fractions. Here \(a\) and \(b\) are the fractions remaining after treatment. For example, \(a=0.40\) means 40% of baseline transcription remains, or a 60% reduction.
Using normalized trajectories with a zero baseline. Ratios such as \(p(t)/p_0\) require \(p_0>0\). If \(p_0=0\), solve and interpret the unnormalized equations instead.
Interpreting matching steady states as proof of the same mechanism. Distinct mechanisms can have the same long-run protein fraction while predicting different mRNA or transient behavior.

Key takeaway: In the uncoupled first-order model, transcriptional inhibition changes mRNA first and protein later; translational inhibition and accelerated degradation leave mRNA unchanged but produce different protein response rates and steady-state fractions. These are model-based predictions whose interpretation depends on the stated assumptions.

Check Your Understanding

Try each question, then compare your reasoning with the solutions below.

For transcriptional inhibition with \(a=0.50\), \(k_{dm}=k_{dp}=0.25\) per day, what are the normalized mRNA and protein levels after 4 days?
For translational inhibition with \(b=0.60\), \(k_{dp}=0.40\) per day, what is the protein fraction after 2 days, and what happens to mRNA?
If degradation doubles and \(k_{dp}=0.30\) per day, what is the protein fraction after 3 days and its long-run fraction?
Can a single late protein measurement always distinguish translational inhibition from accelerated degradation? Explain.
Why is the ratio \(p(t)/p_0\) invalid when the baseline protein abundance is zero?

Solutions. (1) Here the rates are equal, so \(m(4)/m_0=0.50+0.50e^{-1}=0.6839\), and \(p(4)/p_0=0.50+0.50(1+1)e^{-1}=0.8679\). (2) The protein fraction is \(0.60+0.40e^{-(0.40)(2)}=0.7797\); mRNA remains at baseline. (3) With \(g=2\), \(p(3)/p_0=0.50+0.50e^{-(0.60)(3)}=0.5826\), approaching \(1/g=0.50\). (4) No. A single protein fraction may be produced by either mechanism; for example, their long-run fractions match when \(b=1/g\). Time-course and mRNA data can add distinguishing information under the model assumptions. (5) Division by zero is undefined, so the normalized ratio has no meaning; the unnormalized abundance and differential equations must be used.