The authors have declared that no competing interests exist.
Conceived and designed the experiments: JTS KRJ. Performed the experiments: JTS DAS. Analyzed the data: JTS DAS. Contributed reagents/materials/analysis tools: JTS DAS. Wrote the paper: JTS DAS DS KRJ.
Most chronic viral infections are managed with small molecule therapies that inhibit replication but are not curative because non-replicating viral forms can persist despite decades of suppressive treatment. There are therefore numerous strategies in development to eradicate all non-replicating viruses from the body. We are currently engineering DNA cleavage enzymes that specifically target hepatitis B virus covalently closed circular DNA (HBV cccDNA), the episomal form of the virus that persists despite potent antiviral therapies. DNA cleavage enzymes, including homing endonucleases or meganucleases, zinc-finger nucleases (ZFNs), TAL effector nucleases (TALENs), and CRISPR-associated system 9 (Cas9) proteins, can disrupt specific regions of viral DNA. Because DNA repair is error prone, the virus can be neutralized after repeated cleavage events when a target sequence becomes mutated. DNA cleavage enzymes will be delivered as genes within viral vectors that enter hepatocytes. Here we develop mathematical models that describe the delivery and intracellular activity of DNA cleavage enzymes. Model simulations predict that high vector to target cell ratio, limited removal of delivery vectors by humoral immunity, and avid binding between enzyme and its DNA target will promote the highest level of cccDNA disruption. Development of de novo resistance to cleavage enzymes may occur if DNA cleavage and error prone repair does not render the viral episome replication incompetent: our model predicts that concurrent delivery of multiple enzymes which target different vital cccDNA regions, or sequential delivery of different enzymes, are both potentially useful strategies for avoiding multi-enzyme resistance. The underlying dynamics of cccDNA persistence are unlikely to impact the probability of cure provided that antiviral therapy is given concurrently during eradication trials. We conclude by describing experiments that can be used to validate the model, which will in turn provide vital information for dose selection for potential curative trials in animals and ultimately humans.
Innovative new approaches are being developed to eradicate viral infections that until recently were considered incurable. We are interested in engineering DNA cleavage enzymes that can cut and incapacitate persistent viruses. One hurdle is that these enzymes must be delivered to infected cells as genes within viral vectors that are not harmful to humans. In this paper, we developed a series of equations that describe the delivery of these enzymes to their intended targets, as well the activity of DNA cutting within the cell. While our mathematical model is catered towards hepatitis B virus infection, it is widely applicable to other infections such as HIV, as well as oncologic and metabolic diseases characterized by aberrant gene expression. Certain enzymes may bind DNA more avidly than others, while different enzymes may also bind cooperatively if targeted to different regions of viral DNA. We predict that such enzymes, if delivered efficiently to a high proportion of infected cells, will be critical to increase the probability of cure. We also demonstrate that our equations will serve as a useful tool for identifying the most important features of a curative regimen, and ultimately for guiding clinical trial dosing schedules to ensure hepatitis B eradication with the smallest number of possible doses.
To date, cure of most chronic viral infections has remained an impossible goal. Replicating forms of hepatitis B virus (HBV), Herpes Simplex Virus (HSV) and Human Immunodeficiency Virus (HIV) can be targeted with potent small molecule therapies, thereby decreasing the burden of disease associated with these pathogens
Curative approaches to these infections will need to target persistent, non-replicating viral genomes. DNA cleavage enzymes, including homing endonucleases (HE) or meganucleases, zinc-finger nucleases (ZFN), transcription activator-like (TALEN) effector nucleases and CRISPR-associated system 9 (cas9) proteins represent a promising new therapeutic approach for targeting these viral forms
Zinc-finger nucleases are currently being used successfully
Mathematical models are crucial tools for identifying dynamics of active infection, and for designing antiviral regimens that maximize potency and avoid drug resistance
To this end, we developed theoretical models that capture different critical components of viral cure approaches with DNA cleavage enzymes. Our initial models and analyses focus on HBV infection. HBV infects hepatocytes, which are highly accessible to gene therapy delivery vectors and which can be assessed serially for clearance of non-replicating virus. For this reason, HBV may be the most promising initial target for cure. However, the model is easily expanded to account for parameters that govern HIV-1 and HSV infections.
We describe the mathematics of viral vector delivery to hepatocytes, and enzyme - substrate kinetics in the setting of heterogeneous density of episomal infection per hepatocyte. The theoretical problem of
Our simulations suggest that therapeutic outcome is likely to hinge on four key factors: percent vector delivery to target cells per dose (which in turn depends on what proportion of vectors are removed by humoral immune mechanisms), enzyme-DNA target binding affinity and cleavage efficiency, degree of binding cooperativity between cleavage enzymes and target DNA, and number of transgenes delivered per vector. We predict that re-accumulation of the latent pool of HBV is unlikely to occur rapidly enough to overcome weekly dosing of delivery vectors, provided that viral replication is concurrently suppressed with available antiviral therapy. If cleavage enzymes that target single regions within the viral genome are used,
The HBV genome can exist in various states within a cell according to stage of replication. The persistent viral form is covalently closed circular (cccDNA), which is maintained with a half-life of months to years in cells
While we believe that parameters of gene therapy vector delivery and intracellular pharmacodynamics described in our models can ultimately be precisely identified, the parameter values that govern dynamics of HBV cccDNA persistence are likely to remain undetermined when our therapies are tested in animal and human trials. With these uncertainties in mind, the goal of our models is not to prove or disprove competing dynamical hypotheses of HBV cccDNA persistence
Gene therapy vectors can be delivered intravenously, allowing random dispersion to target hepatocytes. Entry into target cells can be achieved by utilizing vectors that are engineered to preferentially bind chosen cell surface receptors, such as sodium taurocholate cotransporting polypeptide, which are specific to HBV target cells
In addition, the model is structured such that gene product, or DNA cleavage enzyme concentration in the cell nucleus is assumed to be directly proportional to the number of delivery vectors with successful entry and gene expression. Recombinant adeno-associated virus (AAV) can be produced in high titers in plasmid transfected cells
The heterogeneous distribution of gene therapy vectors to cells can be captured with an adjusted multiplicity of infection formula Pv = [(σ*m)v * e−(σ*m)]/v!, where m is the ratio of delivery vectors to target cells in the human body (including uninfected hepatocytes which presumably also take a high number of viral vectors), v is the number of vectors delivered per target cell, and Pv is probability of v transduced vectors per cell (
(a) There is a probability, Pv = [(σ*m)v * e−(σ*m)]/v!, of different amounts of vector (red) being delivered to and transduced within each cell containing the target virus (green). (b & c) The percentage of cells with different amounts of transduction will vary according to functional multiplicity of infection (fMOI) which is equal to the ratio of transduced delivery vectors to target hepatocytes multiplied by the proportion of vectors which are transduced, or fMOI = m * σ.
Parameter σ is included to account for the fact that most vectors will not transduce their intended target cells. As a function of the development process, >90% of vector capsids lack viral DNA
σ will take on a value of one if transduction of all dosed vectors occurs, and zero if no gene expression is achieved. This parameter value may be lower for infections such as HIV where latently infected cells potentially exist in anatomic sanctuaries such as the nervous system, as compared to HBV where vectors encounter the liver during first pass metabolism. Certain delivery vectors such as adenovirus (ADV) are immunogenic and delivery of identical serotypes will decrease with successive doses
The delivery equation reveals a wide distribution of vector delivery and transduction when σ*m >1. If HBV infection is modeled with 2*1011 hepatocytes, even if 1012 vectors are delivered successfully (m = 1200, σ = 0.004, σ*m = 5), there is no transduction within a small percentage (
The latter condition is unlikely to promote complete eradication of HBV cccDNA: if we make the simplifying and overly optimistic assumptions that delivery of one or more vectors automatically leads to lethal mutation of all viral genomes within a target cell, that no immunity to the viral vector or enzyme develops with successive doses of delivery vectors, and that there is no replenishment of infected hepatocytes or HBV cccDNA between doses, then the number of doses prior to eradication can be estimated with the formula Nn = N0 * (1−P(v>0))n where N0 is initial number of infected cells, Nn is the remaining number of infected cells following n doses, and cure occurs when Nn<1. The number of necessary doses increases dramatically if 50% delivery is not achieved while delivery greater than 99% dramatically decreases number of doses needed for cure (
Delivery to target cells | Effective MOI (σ * m) | HBV | HIV | HSV |
13.8 | 2 | 2 | 1 | |
11.5 | 3 | 2 | 2 | |
9.2 | 3 | 2 | 2 | |
6.9 | 4 | 3 | 2 | |
4.6 | 6 | 4 | 4 | |
3.0 | 9 | 6 | 5 | |
2.3 | 12 | 8 | 7 | |
1.4 | 19 | 14 | 12 | |
0.69 | 38 | 26 | 22 | |
0.29 | 91 | 59 | 50 | |
0.10 | 247 | 154 | 132 |
Here we assume that successful delivery and gene transduction automatically leads to inactivation of all viral genomes within an infected cell.
This analysis highlights the importance of high vector to target cell ratio, even under favorable assumptions regarding intracellular pharmacodynamics. Because the value of parameter m will be known as a function of dose, the key unknown parameter of delivery is σ, the proportion of vectors that enter target cells and are transduced.
Two factors will drive outcome of an infected cell following delivery of transgene-carrying vectors: the number of viral vectors transduced in the cell and the strength of the enzyme-substrate interaction. The critical biophysical interactions are the binding affinity between enzyme and substrate, the efficiency of enzyme cleaving following binding and the efficiency of precise DNA repair. These processes are captured indirectly with constant d in the formula λo = 1/(1+(v/d)) where v is number of vectors transduced in the cell and λo is probability that the genome will remain uncleaved. In this formula, d is scaled according to vector gene expression value per cell under the assumption that intracellular enzyme concentration is directly proportional to v
A possible hurdle to disruption of latent genomes is resistance to the cleavage enzyme in question. cccDNA molecules may contain pre-existing mutations. The HBV mutation rate is relatively low
Most persistent HBV exists as multiple non-replicating episomes within infected cells. For this reason, outcomes for a cell with intra-nuclear cleavage enzyme expression include only partial inactivation of genomes, as well as development of
(a) An HBV infected cell with three cccDNA molecules (green circles), and delivery of DNA cleavage enzyme containing vectors (red viruses) can transition to several states where none, some, or all of the episomes are eliminated and/or become resistant to the cleavage enzyme. Arrow thickness denotes the relative probability of each event. (b) Cleavage enzymes (red wavy lines) may bind HBV cccDNA molecules cooperatively, whereby binding of one enzyme to its target sequence enhances binding of other enzymes to the same target on separate episomes. (c) Cleavage enzymes (multi-colored wavy lines) that target separate regions within episomes (thick colored lines of corresponding color) may bind HBV cccDNA molecules cooperatively, whereby binding of one enzyme to its target sequence enhances binding of other enzymes to separate sequences on the same episome.
Enhanced cooperative binding between HIV directed antiviral agents and their multivalent viral enzyme targets has been demonstrated as a key determinant in antiviral agent potency. For example at equivalent drug concentrations, HIV protease inhibitors can be 100,000 times more potent than HIV nucleoside reverse transcriptase inhibitors
The mechanism to determine whether cooperative binding is present is generation of log-converted dose response curves with a particular emphasis on the slope of the curve, which translates to Hill coefficient (h*z) in the formula λo = 1/(1+(v/d)h*z). Parameter h represents enhanced binding of one enzyme product to multiple intranuclear episomes (
Parameter z represents the possibility of enhanced or impaired binding of multiple enzymes products to one viral genome at separate binding sites (
Any cleavage event that renders the molecule replication incompetent is a terminal event for the episome; induced resistance at one site still leaves other potential target sites susceptible to cleavage. Arrow thickness denotes the relative probability of a certain event.
Assuming that potency of a single DNA cleavage enzyme (z = 1) on an individual cccDNA episome level is captured with the equation λo = 1/(1+(v/d)h) where λo is probability of the episome remaining uncleaved, total cleavage enzyme activity within a single cell is represented by Pc(i) = (Si) * (1−λo)i * (λo)(S−i) where a cell has S enzyme susceptible cccDNA genomes and Pc(i) represents the probability of cleaving
To synthesize these concepts for HBV infection, we created a three-dimensional matrix. This model tracks total number of cells occupying different states over time. Between cleavage enzyme doses, the numbers of cells with every possible combination of replication competent enzyme susceptible (S) and enzyme resistant (R) genomes are measured. A third dimension is incorporated following each infusion of therapy, and accounts for different doses of vector transduction: each item within the matrix represents the total number of infected cells with a certain value for S, R and v. Transition probabilities are calculated for each cell according to Pc(i) and Pr(k). The matrix is updated accordingly following each dose (
Initial data suggest that enzyme activity and DNA mutations accrue over a week following vector delivery
Strategies to bypass enzyme resistance will be analogous to those employed for antiviral therapy, namely design of cleavage enzymes that target separate regions within episomal HBV cccDNA (
If two enzymes targeting separate sites are split among vectors, then “antagonistic potency” may occur: while the likelihood of
One enzyme/vector, single enzyme (q = 1) | One enzyme/vector, multiple enzymes (q≥1), sequential delivery1 | Multiple enzymes (q≥1)/vector2 | |
Pv = [(σ*m)v * e−(σ*m)]/v! | Pv = [(σ*m)v * e−(σ*m)]/v! | Pv = [(σ*m)v * e−(σ*m)]/v! | |
λ0 = 1/(1+(v/d)h) | λ0 = 1/(1+(v/d)h) | λ0 = 1/(1+(v/d)h*z) | |
Pc(i) = (Si) (1−λ0)i(λ0)(S−i) | |||
Pr(k) = (ik) (Ψ)k(1−Ψ)(i−k) | Pr(k) = (ik) (Ψ)k(1−Ψ)(i−k) | Pr(k) = (ik) (Ψ)k(1−Ψ)(i−k) |
1 = Delivery occurs between each of q successive enzyme exposure.
2 = Delivery accounts for cellular levels of each of q multiple co-packaged enzymes.
q = number of cleavage enzymes, number of viral sites targeted by cleavage enzymes.
S = number of fully susceptible episomes per cell.
R1….q = number of episomes resistant to 1…q cleavage enzymes per cell.
v = # vectors delivered per cell.
m = vector dose/# of target cells.
σ = proportion of vectors that reach uninfected or infected target cells following delivery.
λ0 = probability of a single episome remaining uncleaved.
d = binding coefficient.
h = “between episome” Hill coefficient for a single enzyme.
z = “within episome” Hill coefficient for multiple enzymes.
Pc(i) = probability that
Stot = S+R1+R2+Rq−1.
Ψ = resistance rate per cleavage event.
Pr(k) = probability that
While a very high effective fMOI (>10 in
Each cell within the liver may harbor different numbers of episomes with zero, single and multiple resistant sites (
Example of one theoretical cell containing 6, 3, 2 and 1 episomes with 0, 1, 2, and 3 mutations respectively, as well as 3 delivery vectors.
The model output is constructed in one of two ways: either the number of episomes with any resistance (Stot−S) are plotted against number of fully susceptible episomes (S); or the number of episomes with total resistance to all episomes (SRtot) are plotted against number of remaining episomes without total resistance (Stot−SRtot). Only after q doses are given is it possible to have SRtot>0 due to totally resistant episomes to each of the q available enzymes. Following q doses, our model assumes repeated dosing of the finally dosed enzyme, as the least number of resistant episomes will exist against this enzyme.
If vectors such as ADV with higher gene payload capacities are utilized, then two or more separate enzymes can be delivered and transduced within the same vector. This approach has the theoretical advantage of increasing per cell dose of cleavage enzyme, and has the potential to increase the proportion of targets that receive multiple cleavage enzymes, a process we term “synergistic potency” (
Our model allows for analysis of potential benefits gained from delivery of multiple transgenes within a single ADV vector. The delivery equation is unchanged from prior simulations, as the number of vectors and therefore proportion of cells with no vector transduction (Pv = 0) remain the same. If a vector carries
To demonstrate characteristics of the model, we conducted simulations under different assumptions of vector delivery (fMOI), enzyme-substrate binding avidity/cleavage efficiency (binding dissociation constant or d), and cooperative binding of enzymes to multiple episomes (Hill coefficient or h). Initial simulations assumed a single transgene per vector and ignored
First, we performed a multi-parameter sensitivity analysis with parameter values drawn randomly from a pre-determined wide range (fMOI 0.5–5, binding dissociation constant 0.008–5, and Hill coefficient 0.2–5) using Monte Carlo selection methods. We generated 200 parameter sets and simulated the model to identify parameter effects on therapeutic outcome. Increasing fMOI (R2 = 0.50), and decreasing binding dissociation constant (R2 = 0.24) predicted lower remaining numbers of infected cells to a greater extent than increasing the Hill coefficient (R2 = 0.03).
To obtain a more mechanistic understanding of how model parameters interact to impact the extent of episome disruption, we created 80 parameter sets derived from 4 possible values for fMOI, 4 possible values for the Hill coefficient, and 5 possible values for binding dissociation constant. Model simulations were stochastic but produced equivalent results for repeat experiments with each parameter set. At low fMOI (m*σ = 0.5), decreasing the dissociation constant and/or increasing the Hill coefficient only allowed for a slight relative decrease in number of infected cells following 10 doses; at higher levels of vector delivery, each 5-fold decrease in the dissociation constant (change in color in
All simulations of HBV eradication show results of ten weekly doses of therapy. A single enzyme is used and
If we assumed that humoral immunity removed an increasing proportion of vectors prior to delivery with each dose (successive decreases in parameter σ), then a greater number of cells retained replication competent episomes following 10 doses even with a potent regimen (
However, pre-treatment burden of infection as measured by median number of cccDNA episomes per cell prior to initiation of gene therapy, had only a small impact on remaining number of infected cells (
If
(a) Potent regimens with high fMOI (m*σ = 5), high enzyme – DNA binding avidity (d = 0.04), and positive binding cooperativity (h = 2) will allow for high levels of simulated resistance and predominance of resistant episomes following only 2 to 3 doses; a higher resistance rate (5% versus 1%) will promote a higher number of infected cells containing enzyme resistant episomes. (b) Infected cells containing enzyme resistant episomes will ultimately achieve equivalent levels assuming equal resistant rates whether a potent (m*σ = 5, d = 0.004 & h = 2) or less potent (m*σ = 1, d = 1 & h = 2) regimen is used. (c) If successive enzymes are dosed that target different regions within HBV cccDNA episomes, then the number of remaining episomes following multiple doses decreases accordingly; susceptible and resistant replication competent genomes are summed; by 60 days, all remaining episomes are resistant to each of the dosed enzymes (not shown in diagram).
To avoid cleavage enzyme resistance, we next considered sequential delivery of 1, 2, 3, 4, or 5 enzymes in separate, weekly doses. A new enzyme was given each week until no new enzymes remained (at the sixth dose for the 5 enzyme condition, for example). At this point, the final enzyme was repeatedly redosed. Simulations assumed favorable potency parameters and a resistance rate of 1%. The addition of extra enzymes increased the time until enzyme resistant forms predominated, and lowered the steady state of cells retaining replication competent HBV cccDNA by ∼0.5 log with addition of each enzyme (
We next simulated trials with a single dose of a multi-payload vector such as ADV carrying 1, 2, or 3 transgenes concurrently under different assumptions of fMOI and cooperative binding of enzymes to multiple episomes. A favorable enzyme-substrate binding avidity/cleavage efficiency was assumed for each simulation. Results from simulations with 36 pre-selected parameter sets (all following a single dose with assumed resistance rate = 1%) show that total remaining cccDNA genomes decreased with increasing fMOI, and that maximizing the transgene payload (blue line,
Simulation of HBV eradication employing gene therapy following a single dose. Each data point represents number of remaining infected cells (y-axis) after a simulation with one of 36 unique parameter sets. x-axis is functional multiplicity of infection (fMOI). Enzyme-DNA binding avidity is fixed (d = 0.04). Color represents number of enzymes delivered per vector (orange, green and blue = 1,2 & 3 respectively). Hill coefficient is 1, 2 and 5 (square, diamond and circle). The simulation assumes no pre-existing resistance. (a) Addition of multiple DNA cleavage enzymes within single vectors decreases the number of total remaining infected cells, particularly when vector delivery is high and intracellular binding cooperativity is present. (b) Addition of multiple DNA cleavage enzymes within single vectors decreases the number of total remaining infected cells harboring HBV cccDNA with any
If cccDNA levels reconstitute at a meaningful rate and several day intervals are required between doses to allow effects of DNA cleavage enzymes to accrue, then this may imply the need for more prolonged therapeutic courses. We therefore examined the effects of underlying dynamics of HBV cccDNA survival, as well as the possible delayed effects of DNA cleavage enzymes following target cell entry. Several factors may drive changes in levels of HBV cccDNA during suppressive antiviral therapy. Hepatocytes with HBV cccDNA molecules periodically die at a rate equivalent to that of an uninfected hepatocyte (
(a) Cell death inducing decay of incorporated episome. (b) Episomal degradation. (c) cccDNA expansion despite suppressive antiviral therapy. (d) Hepatocyte replication with equal dispersion of cccDNA molecules between cells. (e) Hepatocyte replication with replication of cccDNA molecules between cells. Only mechanism (c) could increase number of doses needed prior to inactivation while other mechanisms (a, b and d) may allow more rapid cure.
In all simulations in
Simulation of HBV inactivation following 10 doses of therapy given every two weeks. Unlike prior simulations, these simulations assume that DNA cleavage activity accrues evenly over a week rather than instantaneously. Parameters reflect high potency (m*σ = 5, d = 0.004, h = 2). Traces represent no cccDNA activity between does (black lines), residual cccDNA replication between doses at rate = 0.01/day (red lines), hepatocyte replication with dispersion of cccDNA between cells (orange lines) and hepatocyte death with concurrent death of episomes (blue lines). Solid lines are cells with susceptible episomes and would represent therapeutic outcomes in the absence of de novo resistance (Ψ = 0). Dotted lines are cells with resistant episomes and would represent therapeutic outcomes with de novo resistance (Ψ = 0.01). Effects on therapeutic outcomes are minimal unless fairly high levels of cccDNA turnover are assumed (red lines) as may occur in the absence of fully suppressive antiviral therapy.
We describe mathematical models that aim to capture critical features of DNA cleavage enzyme therapy for eradication of HBV. Our results identify potentially critical parameters that will determine whether cure will be feasible with available vector cleavage enzyme constructs. In particular, successful vector delivery to the majority of target cells with each infusion, and favorable intracellular binding kinetics between enzymes and DNA target sites appear to be pre-requisites for successful regimens. Cooperative binding of enzymes between multiple episomal targets could also potentially limit the number of doses needed prior to cure, particularly if enzyme concentration in cells only marginally exceed binding coefficient values.
While multiple doses of gene therapy will likely be required for cure, the first dose appears to be particularly critical. In order to enhance potency and limit resistance, this dose should have a high vector to target cell ratio, and if possible, multiple enzymes should be packaged within each delivery vector. Sequential use of different enzymes appears to be another useful strategy to avoid
While our integrated therapeutic model is relatively complex, its individual components (vector delivery, intracellular pharmacodynamics, resistance) are quite manageable. In total, the model contains only five unknown parameter values including 1) proportion of vectors removed prior to entry into target cells, 2) enzyme-DNA binding coefficient, 3) vector-DNA cleavage dose response slope (Hill coefficient), 4) resistance rate per DNA cleavage event and 5) dose response slope within a single episome if multiple enzymes are present in the cell nucleus. Each of these parameter values can be identified via specific experimental approaches for all vectors and cleavage enzymes of interest, which will allow for testing and refining of the model.
Vector delivery to target cells is best estimated initially in animal model studies. Humanized mouse models of HBV hold promise for this indication
A critical caveat of the functional MOI (fMOI) is that the vector to target cell ratio assumed in parameter m is inclusive of all cells that may serve as targets for vector entry, rather than only HBV infected cells. If a particular delivery vector also efficiently enters other intrahepatic cells such as Kupffer cells, endothelial cells or cells in other organs, then the fMOI will decrease accordingly. In effect, these cells will serve as vector sponges and will decrease the probability of high vector delivery to infected cells containing HBV cccDNA. Therefore, vector receptor specificity is critical not only to avoid untoward toxicity, but also to ensure that precious vector is not wasted.
A key experimental goal should be to determine which enzymes achieve avid binding and DNA cleavage activity (low values for d) and positive cooperative binding (h>1 or z>1) to their DNA targets. Dose response curve slope and enzyme-substrate binding coefficients can be obtained from cell culture models of HBV cccDNA infection in which infected cell lines are exposed to delivery vectors dosed at different multiplicities of infections. Using high throughput sequencing of the DNA target site, it will be possible to measure the proportion of target genomes with terminally disrupted DNA for each vector dose. Experimental dose response curves can be tested against our models describing enzyme DNA binding kinetics. If multiple enzymes are delivered concurrently in a single vector, then similar curves can be used to assess cooperative binding between several sites within a single episome.
To obtain a conservative upper limit for resistance rate per cleavage event, it will first be necessary to identify cells with confirmed vector delivery and HBV cccDNA cleavage. One possibility is to sort for vector transduced cells that are HBV e antigen positive and then look for mutation events within the cleaved open reading frame. For the purposes of informing clinical trial dose design, this estimate will be useful to ensure that doses exceed predicted thresholds for viral persistence.
When all unknown parameter values are estimated and a model structure is selected that best represents available data regarding vector delivery, enzyme/DNA substrate kinetics and resistance rate, then it will be possible to design regimens that maximize probability of cure while limiting excess dosing and possible toxicity. While it will be necessary to characterize all available delivery vectors and cleavage enzymes prior to predicting likelihood of therapeutic success, certain strategies are promising based on
A key challenge will be measuring therapeutic outcome. For HBV, it is difficult to take serial quantitative measures of episomal reservoirs of infection. While active viral replication can be tracked with quantitative PCR, burden of quiescent viral episomes can only be assessed with liver biopsy and tissue quantitation of uncleaved HBV cccDNA using sequencing. Even a tiny number of latently infected cells may theoretically be enough to reactivate and repopulate the reservoir. Because serial biopsies are likely to be feasible only in animal models of infection, therapeutic efficacy will ultimately need to be evaluated with close clinical follow up after cessation of antiviral therapy. For this reason, we make conservative assumptions in our model, so that the dosing schedule exceeds the presumed threshold for cure.
While we have focused on eradication of HBV, our model is easily adjusted to account for potential cure of other chronic viral infections such as HIV or HSV-2. The burden and properties of non-replicating viral stores differ dramatically between HBV, HIV and HSV
HSV latency exists within a relatively low number of neuronal cell bodies in either the trigeminal or dorsal root ganglia
In summary, we present a model to capture the effects of gene therapy with DNA cleavage enzymes for chronic HBV infection. The model helps identify key therapeutic parameters that will be necessary for cure, and outlines appropriate experimental steps to identify dosing regimens that are most likely to disrupt all latent viral DNA following a minimal number of gene therapy doses.
Simulations were performed on C++ and using Microsoft Excel.
(WMV)
The authors would like to thank Jun Sellers for assistance in manuscript preparation.