Antibody Neutralization and Viral Rebound

Why Antibody Neutralization Changes Viral Growth

Antibodies can bind free viral particles and prevent them from initiating new infections. In a within-host model, this effect can be represented as an added loss of infectious free virus. If antibody concentration falls, or if a virus phenotype is less sensitive to the antibody, that loss may no longer be sufficient to keep viral replication below its growth threshold. This provides a mechanism for viral rebound: renewed growth of viral burden after a period of suppression.

A rebound is not, by itself, proof that antibody waning or reduced sensitivity caused it. Target-cell availability, infected-cell dynamics, other immune mechanisms, and changes in viral production can also affect the trajectory. The model below isolates antibody-mediated neutralization so that its predictions and assumptions are explicit.

Definition: In a target-cell-limited viral model, let \(T\) be susceptible target cells per millilitre, \(I\) infected cells per millilitre, \(V\) infectious free-virus particles per millilitre, and \(A\) antibody concentration. The parameters \(\beta\), \(p\), \(\delta\), and \(c\) are, respectively, the infection rate constant, virus production per infected cell, infected-cell loss rate, and baseline infectious-virus clearance rate. Let \(k_N(A)\) be the antibody-dependent neutralization rate, with units of inverse time. A basic model is

\(\frac{dT}{dt}=-\beta TV,\quad \frac{dI}{dt}=\beta TV-\delta I,\quad \frac{dV}{dt}=pI-[c+k_N(A)]V.\)

The equations assume that infection is proportional to both \(T\) and \(V\), infected cells are lost at a first-order rate, and infectious virus is produced at a constant rate per infected cell. Antibody acts on free infectious virus through a first-order loss term; it does not directly kill infected cells or change their virus production in this model. The parameters are treated as constant except for \(A\), and the compartments are assumed sufficiently well mixed for these rates to apply.

A common concentration-response form is \(k_N(A)=k_{\max}A/(IC_{50}+A)\), where \(k_{\max}\) is the maximum neutralization rate and \(IC_{50}\) is the antibody concentration that gives half of that maximum in this model. This saturating expression is a modelling choice, not a universal law. It presumes a monotone response that approaches zero at zero antibody and approaches \(k_{\max}\) at high concentration.

Antibody-Dependent Growth Threshold

To assess whether infection can grow, hold \(T\) and \(A\) approximately constant over a short interval. This is a local approximation: it is most defensible when target-cell depletion and antibody change are slow relative to the viral and infected-cell dynamics being examined. The effective reproductive number is the expected number of newly infected cells generated by one infected cell in this fixed environment. An infected cell produces virus at rate \(p\) and remains infected for an average \(1/\delta\) time, yielding \(p/\delta\) particles on average. Each infectious particle remains in the free-virus compartment for an average \(1/[c+k_N(A)]\) time and infects target cells at rate \(\beta T\). Therefore,

R_{\mathrm{eff}}(A,T)=\frac{\beta T p}{\delta[c+k_N(A)]}.

When \(R_{\mathrm{eff}}>1\), infection can grow under the fixed-environment approximation; when \(R_{\mathrm{eff}}<1\), it declines asymptotically in that approximation. The equality \(R_{\mathrm{eff}}=1\) is the threshold, not a prediction that viral load changes immediately at that instant. In particular, the current quantities \(I\) and \(V\) retain their prior history.

Threshold result: For fixed \(T\), viral growth is possible when \(k_N(A)<\beta Tp/\delta-c\). If the right-hand side is non-positive, even the model with no antibody neutralization has \(R_{\mathrm{eff}}\leq 1\) at that target-cell level. If \(k_N(A)\) is bounded above, its maximum must also be large enough to cross the threshold for antibody alone to suppress growth.

The threshold also follows directly from the linearized equations. With \(T\) and \(A\) fixed, perturbations in \(I\) and \(V\) evolve according to a two-compartment linear system. A mode proportional to \(e^{\lambda t}\) satisfies

(\lambda+\delta)(\lambda+c+k_N(A))-\beta Tp=0.

At \(\lambda=0\), the left side is \(\delta[c+k_N(A)]-\beta Tp\). If \(R_{\mathrm{eff}}>1\), this value is negative, while the expression tends to positive infinity as \(\lambda\) increases. The quadratic has one positive root, so a growing mode exists. If \(R_{\mathrm{eff}}<1\), both coefficients of the quadratic are positive and its roots have negative real parts; perturbations decay. This justifies the threshold for this two-compartment model under fixed \(T\) and \(A\).

Worked Example: Neutralization Suppresses Growth

Suppose a hypothetical infection has \(\beta=2.0\times10^{-7}\) mL per virion per day, \(T=1.0\times10^6\) cells/mL, \(p=100\) virions per cell per day, \(\delta=0.50\) per day, and \(c=10\) per day. Let \(k_{\max}=50\) per day and \(A=9IC_{50}\). Determine whether the model predicts growth.

First, \(\beta T=(2.0\times10^{-7})(1.0\times10^6)=0.20\) per day. With no antibody, \(R_0=\beta Tp/(\delta c)=(0.20)(100)/[(0.50)(10)]=4.00\).

At \(A=9IC_{50}\), the neutralization rate is \(k_N=50(9)/(1+9)=45\) per day. Thus,

R_{\mathrm{eff}}=\frac{(0.20)(100)}{(0.50)(10+45)} =\frac{20}{27.5}=0.7273.

Because this value is below 1, the model predicts decline in the fixed-\(T\), fixed-\(A\) approximation. The calculation does not say that viral load must fall at a particular rate, nor that the same antibody concentration will suppress every virus phenotype.

Waning Antibody and the Timing of Rebound

After antibody production or dosing decreases, concentration may be approximated over a limited interval by first-order waning, \(dA/dt=-k_A A\), where \(k_A\) is the antibody loss rate. The solution is \(A(t)=A_0e^{-k_A t}\), with \(A_0\) the concentration at the start of the interval. If neutralization increases monotonically with \(A\), then \(k_N[A(t)]\) falls as antibody wanes, and \(R_{\mathrm{eff}}\) rises, provided \(T\) is approximately fixed.

For the saturating neutralization function, the threshold can be translated into a concentration. If \(K=\beta Tp/\delta-c\) and \(0<K<k_{\max}\), the threshold \(k_N(A)=K\) occurs at \(A=IC_{50}K/(k_{\max}-K)\). This follows by rearranging \(K=k_{\max}A/(IC_{50}+A)\). Concentrations below this value give \(R_{\mathrm{eff}}>1\) under the model assumptions. If \(K\leq0\), no positive antibody concentration is needed to keep the fixed-environment reproductive number at or below 1. If \(K\geq k_{\max}\), the specified antibody response cannot drive the rate below the threshold, even at arbitrarily high concentration.

Worked Example: Calculating a Waning Threshold

Use the parameters in the preceding example, but suppose antibody starts at \(A_0=8IC_{50}\) and wanes with \(k_A=0.15\) per day. Find when the model crosses from suppression to possible growth, assuming \(T\) remains at \(1.0\times10^6\) cells/mL.

The threshold neutralization rate follows from \(R_{\mathrm{eff}}=1\): \(k_N=\beta Tp/\delta-c=(20/0.50)-10=30\) per day. Set \(50A/(IC_{50}+A)=30\). Dividing by 50 gives \(A/(IC_{50}+A)=0.60\), so \(A=1.5IC_{50}\).

The crossing time satisfies \(8IC_{50}e^{-0.15t}=1.5IC_{50}\). Cancelling \(IC_{50}\) and taking logarithms gives

t=\frac{\ln(8/1.5)}{0.15}=11.16\ \text{days}.

At about 11.16 days the local reproductive number reaches 1; after that, it exceeds 1 if the other assumptions remain valid. This is a threshold-crossing time, not necessarily the time when a viral-load assay first detects rebound. Viral burden may remain low for a while, or the local approximation may cease to hold as \(T\) or \(A\) changes.

Reduced Sensitivity to Antibody

A distinct route to renewed growth is a viral phenotype that is less affected by the same antibody concentration. One simple representation assigns phenotype \(j\) its own neutralization parameters, for example \(k_{N,j}(A)=s_jk_{\max}A/(IC_{50,j}+A)\), where \(s_j\) is a sensitivity multiplier between 0 and 1. A smaller \(s_j\), or a larger \(IC_{50,j}\), means less neutralization at a given concentration. This formulation compares phenotypes; it does not explain how a less-sensitive phenotype arose or became prevalent.

For an illustrative comparison, assume both phenotypes share the production and clearance parameters in the first example. At \(A=9IC_{50}\), let the sensitive phenotype have \(k_N=45\) per day and a less-sensitive phenotype have one-quarter that rate, \(11.25\) per day. The second phenotype has

R_{\mathrm{eff}}=\frac{20}{(0.50)(10+11.25)} =\frac{20}{10.625}=1.882.

Thus the same antibody concentration predicts suppression of the sensitive phenotype but permits growth of the less-sensitive one. This is a conditional model prediction, not evidence that an observed rebound was caused by viral escape. Establishing that explanation requires evidence about the virus and its sensitivity, as well as consideration of other causes of changing viral burden.

Worked Example: Comparing Two Neutralization Profiles

Suppose two hypothetical phenotypes are tested at \(A=4\ \mu\text{g}/\text{mL}\). Both share \(k_{\max}=40\) per day. Phenotype S has \(IC_{50,S}=1\ \mu\text{g}/\text{mL}\), while phenotype L has \(IC_{50,L}=8\ \mu\text{g}/\text{mL}\). In a host environment where \(\beta Tp/\delta=24\) per day and \(c=8\) per day, calculate each reproductive number.

For S, \(k_{N,S}=40(4)/(1+4)=32\) per day. Therefore, \(R_{\mathrm{eff},S}=24/(8+32)=0.600\), so the model predicts decline in the fixed environment.

For L, \(k_{N,L}=40(4)/(8+4)=13.33\) per day. Therefore, \(R_{\mathrm{eff},L}=24/(8+13.33)=1.125\), so the model permits growth. The comparison demonstrates that a concentration alone does not determine the outcome: the neutralization profile and the viral growth parameters both matter.

Model Checks, Alternatives, and Limitations

A fitted model should be checked against the structure of the data, not only against a calculated threshold. If the model predicts a single smooth decline while measurements show sustained growth, changing slopes, or different responses across phenotypes, the assumed neutralization and viral-dynamics structure may be inadequate. Measurements of antibody concentration, neutralization activity, and viral burden at multiple times can help distinguish a waning response from a change in sensitivity, but the mechanisms may be difficult to identify if only viral load is observed.

Two representations of neutralization should not be confused. The model here removes infectious particles through an additional rate \(k_N(A)V\). An alternative is to reduce the infectivity of particles while leaving their physical clearance unchanged, replacing \(\beta\) by an antibody-dependent \(\beta_{\mathrm{eff}}(A)\). Either can represent reduced infection potential, but they imply different interpretations and may have different consequences for measurements of total versus infectious virus. Applying both effects without evidence can count the same neutralization mechanism twice.

The threshold calculation is also sensitive to assumptions about \(T\). If target cells decline, the factor \(\beta T\) decreases and viral growth may be limited even as antibody wanes. If the immune response has a continuing source, a simple exponential decay may be inappropriate. More detailed time-varying models can integrate the differential equations rather than treating \(T\) and \(A\) as fixed. Those models still require data to distinguish parameters: for example, a change in viral load alone may not identify whether \(p\), \(c\), or \(k_N(A)\) changed.

Check units. The terms \(c\) and \(k_N\) must both have units of inverse time; \(R_{\mathrm{eff}}\) must be dimensionless.
Check the concentration response. The chosen neutralization function should be consistent with the intended assay or biological interpretation, including its plausible range.
Check the local approximation. A fixed-\(T\), fixed-\(A\) threshold is less informative when those quantities change substantially over the interval of interest.
Check the viral measurement. A measurement of total particles need not equal infectious virus, the state represented by \(V\) in this model.

Common pitfall: “Antibody concentration fell below \(IC_{50}\), so viral rebound must occur.” This is not generally correct. \(IC_{50}\) describes a feature of the selected neutralization function; it is not automatically the host-specific growth threshold. Rebound potential depends on the balance \(\beta Tp/\delta\) versus \(c+k_N(A)\), and an observed rise also depends on the evolving states and other mechanisms.

Check Your Understanding

Answer each question using the assumptions stated in the model. Brief solutions follow the questions.

If \(k_{\max}=40\) per day and \(A=3IC_{50}\), what is \(k_N(A)\)? If \(\beta Tp/\delta=24\) per day and \(c=8\) per day, calculate \(R_{\mathrm{eff}}\) and state the predicted direction of change.
Using the parameters in Question 1, suppose \(A(t)=6IC_{50}e^{-0.10t}\). When does \(R_{\mathrm{eff}}\) cross 1?
Why can a viral-load increase occur later than the time at which the instantaneous \(R_{\mathrm{eff}}\) first exceeds 1?
Name one assumption that would make the fixed-\(T\), fixed-\(A\) threshold unreliable during a long observation period.

Solutions. (1) \(k_N=40(3)/(1+3)=30\) per day. Then \(R_{\mathrm{eff}}=24/(8+30)=24/38=0.6316\), so the model predicts decline in the fixed environment. (2) The threshold requires \(k_N=24-8=16\) per day. Solving \(40A/(IC_{50}+A)=16\) gives \(A/(IC_{50}+A)=0.4\), hence \(A=(2/3)IC_{50}\). Therefore, \(6e^{-0.10t}=2/3\), and \(t=\ln(9)/0.10=21.97\) days. (3) The infected-cell and virus amounts retain their prior values and change continuously; growth potential crossing 1 does not instantly produce a measurable rise. A detectable rebound may require time for the growing mode to increase viral burden. (4) For example, substantial target-cell depletion changes \(T\), and continuing antibody production or non-exponential waning changes \(A\). Either violates the assumption that those quantities are fixed for the threshold calculation.