AAV Biodistribution and Tissue-Specific Expression

Why Tissue-Resolved Modeling Matters

An AAV vector can reach several tissues, yet the tissue with the most vector genomes need not produce the most transgene product. Tissue vector-genome assays measure vector DNA present in a sample; they do not, by themselves, establish how much of that DNA is in productively transduced cells or how strongly those cells express the transgene. A quantitative model should therefore separate distribution, productive transduction, genome loss, and product formation.

The earlier material on deriving vector genome clearance and transgene expression kinetics provides the basic clearance and expression framework. Here, the key extension is to assign tissue-specific parameters and compare tissue measurements on a clearly defined scale. All values in the examples below are hypothetical.

Definition: Let \(A_i(t)\) be the amount of assay-detectable vector genomes in tissue \(i\), in vector genomes (vg), and let \(M_i\) be that tissue's mass in grams. The tissue genome abundance is \(Q_i(t)=A_i(t)/M_i\), in vg/g. Let \(\phi_i\) be the fraction of tissue-associated genomes treated as productively transduced, \(r_i\) the product formation rate per productive genome, and \(B_i(t)\) the amount of transgene product in the tissue, in pg. These quantities and parameters may differ across tissues.

A Tissue-Specific Model

For a compact model, assume vector distribution to tissues is rapid relative to the observation times. Let \(D\) be the administered vector-genome amount and \(f_i\) the fraction assigned to tissue \(i\) after distribution. The initial amount in that tissue is \(A_i(0)=f_iD\). The fractions need not sum to one if some administered vector remains outside the modeled tissues or is otherwise unaccounted for.

After the initial distribution, assume tissue vector genomes decline by first-order loss with rate constant \(k_{g,i}\), measured in day\(^{-1}\). A fixed fraction \(\phi_i\) is taken to be productive over the modeled interval. Transgene product is formed at rate \(r_i\) per productive genome and lost with first-order rate constant \(k_{y,i}\), also in day\(^{-1}\). Under these assumptions:

\frac{dA_i}{dt}=-k_{g,i}A_i,\qquad Q_i(t)=\frac{A_i(t)}{M_i},\qquad P_i(t)=\phi_iA_i(t)

\frac{dB_i}{dt}=r_iP_i(t)-k_{y,i}B_i(t)

Here \(P_i(t)\) is the modeled amount of productive genomes, in vg. The units of \(r_i\) are pg/(vg·day), so \(r_iP_i\) is pg/day. This model distinguishes a genome measurement from a transduction assumption: \(A_i\) includes assay-detectable vector genomes, while \(\phi_i\) determines what fraction contributes to modeled product formation. The model does not imply that an assay directly measures \(\phi_i\).

The rapid-distribution assumption is not suitable when tissue delivery continues over an appreciable period. In that case, use a time-dependent tissue input \(U_i(t)\), in vg/day, and replace the first equation with \(dA_i/dt=U_i(t)-k_{g,i}A_i(t)\). The input can differ by tissue, representing distinct uptake or delivery. The remaining expression equation can be retained if the productive fraction and expression parameters are adequately represented as constant.

Deriving the Product Trajectory

With rapid distribution, \(A_i(t)=A_i(0)e^{-k_{g,i}t}\), so the product equation is a linear differential equation driven by a declining source. If \(B_i(0)=0\), multiplying the equation by the integrating factor \(e^{k_{y,i}t}\) gives:

\frac{d}{dt}\left(e^{k_{y,i}t}B_i(t)\right) =r_i\phi_iA_i(0)e^{(k_{y,i}-k_{g,i})t}

Integrating from 0 to \(t\), then dividing by \(e^{k_{y,i}t}\), yields the product amount. For unequal loss rates, the result is:

B_i(t)=r_i\phi_iA_i(0) \frac{e^{-k_{g,i}t}-e^{-k_{y,i}t}}{k_{y,i}-k_{g,i}}, \qquad k_{y,i}\ne k_{g,i}

If the two rates are equal, the limit of this expression is \(B_i(t)=r_i\phi_iA_i(0)t e^{-k_{g,i}t}\). This expression also satisfies the initial condition \(B_i(0)=0\). Product can initially accumulate even while vector genomes are declining, because formation continues from the genomes that remain. The product trajectory depends on both genome loss and product loss, not just on genome abundance at the time of measurement.

Worked Example: Comparing Initial Genome Abundance Across Tissues

Suppose a hypothetical subject receives \(D=1.00\times10^{13}\) vg. After rapid distribution, 12% is assigned to liver and 3% to muscle. Let the modeled liver mass be 1,500 g and the muscle mass be 8,000 g. Calculate each tissue's initial vector-genome amount and abundance.

For liver, \(A_{\text{liver}}(0)=0.12(1.00\times10^{13})=1.20\times10^{12}\) vg. Therefore:

Q_{\text{liver}}(0)=\frac{1.20\times10^{12}\ \mathrm{vg}}{1{,}500\ \mathrm{g}} =8.00\times10^8\ \mathrm{vg/g}

For muscle, \(A_{\text{muscle}}(0)=0.03(1.00\times10^{13})=3.00\times10^{11}\) vg, giving:

Q_{\text{muscle}}(0)=\frac{3.00\times10^{11}\ \mathrm{vg}}{8{,}000\ \mathrm{g}} =3.75\times10^7\ \mathrm{vg/g}

The initial liver abundance is \(8.00\times10^8/(3.75\times10^7)=21.3\) times the muscle abundance. This comparison is per gram; it is not a comparison of total genomes in the two whole tissues. The assigned fractions total 15%, leaving the remaining 85% outside this two-tissue allocation.

Worked Example: Genome Abundance and Instantaneous Product Formation

Continue with the initial values above. Suppose \(k_{g,\text{liver}}=0.20\) day\(^{-1}\) and \(k_{g,\text{muscle}}=0.05\) day\(^{-1}\). At day 7, let liver have \(\phi=0.01\) and \(r=0.0005\) pg/(vg·day), while muscle has \(\phi=0.20\) and \(r=0.002\) pg/(vg·day). Calculate vector-genome abundance and product formation rate per gram.

First, the genome abundances are \(Q_i(7)=Q_i(0)e^{-k_{g,i}7}\). Using \(e^{-1.4}=0.2465979\) and \(e^{-0.35}=0.7046881\):

Q_{\text{liver}}(7)=(8.00\times10^8)(0.2465979) =1.973\times10^8\ \mathrm{vg/g}

Q_{\text{muscle}}(7)=(3.75\times10^7)(0.7046881) =2.6426\times10^7\ \mathrm{vg/g}

The product formation rate per gram is \(r_i\phi_iQ_i(7)\). For liver, this is \((0.0005)(0.01)(1.973\times10^8)=986.4\) pg/(g·day), or \(0.9864\) ng/(g·day). For muscle, it is \((0.002)(0.20)(2.6426\times10^7)=10{,}570\) pg/(g·day), or \(10.57\) ng/(g·day).

At day 7 the liver still has about 7.47 times more vector genomes per gram, but the modeled muscle product formation rate per gram is about 10.7 times higher. The reversal follows from the assumed differences in productive fraction and expression rate. It is a consequence of the hypothetical parameters, not a general prediction that muscle expresses AAV transgenes more strongly than liver.

Worked Example: Accumulated Product Can Reflect Earlier Expression

Use the same initial tissue amounts and masses. Suppose liver has \(\phi=0.01\), \(r=0.0005\) pg/(vg·day), \(k_g=0.20\) day\(^{-1}\), and \(k_y=0.10\) day\(^{-1}\). Suppose muscle has \(\phi=0.20\), \(r=0.002\) pg/(vg·day), \(k_g=0.05\) day\(^{-1}\), and \(k_y=0.10\) day\(^{-1}\). With no product at time zero, calculate product per gram at day 7.

For liver, the initial product formation rate per gram is \(r\phi Q(0)=(0.0005)(0.01)(8.00\times10^8)=4{,}000\) pg/(g·day). Thus:

\frac{B_{\text{liver}}(7)}{M_{\text{liver}}} =4{,}000\frac{e^{-0.20(7)}-e^{-0.10(7)}}{0.10-0.20} =4{,}000\frac{0.246598-0.496585}{-0.10} =9{,}999.5\ \mathrm{pg/g}

This is \(10.00\) ng/g. For muscle, the initial formation rate per gram is \((0.002)(0.20)(3.75\times10^7)=15{,}000\) pg/(g·day). Therefore:

\frac{B_{\text{muscle}}(7)}{M_{\text{muscle}}} =15{,}000\frac{e^{-0.05(7)}-e^{-0.10(7)}}{0.10-0.05} =15{,}000\frac{0.704688-0.496585}{0.05} =62{,}430.9\ \mathrm{pg/g}

This is \(62.43\) ng/g. In this scenario the liver has more vector genomes per gram at day 7, whereas modeled accumulated product per gram is higher in muscle. The product amount reflects expression over the preceding interval as well as product loss. To compare whole-tissue product amounts rather than concentrations per gram, multiply each result in pg/g by its tissue mass in grams.

Assumptions, Diagnostics, and Alternatives

The model's interpretation depends on its assumptions. Rapid distribution treats tissue delivery as complete at the chosen time zero. First-order genome loss assumes a constant proportional loss rate over the interval. The fixed \(\phi_i\) assumes that the productive fraction does not change with time, and constant \(r_i\) assumes a stable average product formation rate per productive genome. First-order product loss and the initial condition \(B_i(0)=0\) are also assumptions, not automatic properties of every assay or transgene.

Useful checks include comparing observed genome time courses with the predicted exponential decline, examining whether product rises or falls as predicted, and checking whether estimates change substantially when sampling times or tissue definitions change. Tissue samples may differ in cell composition, and a value per gram can obscure cell-type-specific localization. Vector DNA assays and product assays may also have different sampling, sensitivity, and normalization properties. A discrepancy between them is a reason to examine model assumptions and measurement definitions; it is not, by itself, proof of a particular biological mechanism.

A static partition model is parsimonious when early tissue distribution is fast and only post-distribution observations matter. A time-dependent uptake model is preferable when delivery into tissues is prolonged or when early measurements show continuing accumulation. A more detailed model could allow productive fraction or expression rate to vary by cell type or time, but such added parameters require data that can distinguish them. With only a single tissue genome measurement and a single product measurement, uptake, productive fraction, expression rate, and loss rates may not be separately identifiable.

Common pitfall: “The liver has more vector genomes per gram, so it must produce more transgene product per gram.” This conclusion does not follow from genome abundance alone. In the model, formation also depends on \(\phi_i\) and \(r_i\), while measured product additionally reflects its loss rate and prior formation. Compare measurements on the same tissue and time scale, and state whether the endpoint is per gram or whole-tissue amount.

These models describe distribution and expression, not therapeutic effect. A product concentration or formation rate is not itself a clinical response. Interpretation should also distinguish the modeled quantities from the assay readouts used to estimate them: vector-genome abundance is not a direct measurement of productive genomes, and product abundance is not a direct measurement of product formation rate.

Check Your Understanding

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

If a tissue contains \(2.4\times10^{11}\) vg in a mass of 600 g, what is its vector-genome abundance in vg/g?
Using the liver parameters in the first worked example, calculate liver abundance at day 4, given \(k_g=0.20\) day\(^{-1}\).
Which model assumption is most directly relaxed by adding an input \(U_i(t)\) to the tissue genome equation?
Two tissues have the same \(Q_i(t)\), but different \(\phi_i\) and \(r_i\). Must their instantaneous product formation rates per gram be equal? Explain.

Solutions. 1. \(Q=A/M=(2.4\times10^{11}\ \mathrm{vg})/(600\ \mathrm{g})=4.00\times10^8\) vg/g. 2. \(Q(4)=(8.00\times10^8)e^{-0.20(4)}=(8.00\times10^8)(0.449329)=3.595\times10^8\) vg/g. 3. It relaxes the assumption that tissue distribution is complete at time zero; \(U_i(t)\) allows continuing tissue input. 4. No. The formation rate per gram is \(r_i\phi_iQ_i(t)\), so equal genome abundance does not imply equal formation when \(r_i\) or \(\phi_i\) differs.

AAV Biodistribution and Tissue-Specific Expression

What You'll Learn