The longitudinal kinetics of AAV5 vector integration profiles and evaluation of clonal expansion in mice

Abstract

Adeno-associated virus (AAV)-based vectors are used clinically for gene transfer and persist as extrachromosomal episomes. A small fraction of vector genomes integrate into the host genome, but the theoretical risk of tumorigenesis depends on vector regulatory features. A mouse model was used to investigate integration profiles of an AAV serotype 5 (AAV5) vector produced using Sf and HEK293 cells that mimic key features of valoctocogene roxaparvovec (AAV5-hFVIII-SQ), a gene therapy for severe hemophilia A. The majority (95%) of vector genome reads were derived from episomes, and mean (± standard deviation) integration frequency was 2.70 ± 1.26 and 1.79 ± 0.86 integrations per 1,000 cells for Sf- and HEK293-produced vector. Longitudinal integration analysis suggested integrations occur primarily within 1 week, at low frequency, and their abundance was stable over time. Integration profiles were polyclonal and randomly distributed. No major differences in integration profiles were observed for either vector production platform, and no integrations were associated with clonal expansion. Integrations were enriched near transcription start sites of genes highly expressed in the liver (p = 1 × 10^-4) and less enriched for genes of lower expression. We found no evidence of tumorigenesis or fibrosis caused by the vector integrations.

Keywords: AAV5; AAV5-hFVIII-SQ; Spodoptera frugiperda; clonal expansion; common integration site; episomes; human embryonic kidney 293; target enrichment sequencing; valoctocogene roxaparvovec; vector integration.

Graphical abstract

Figure 1

Dynamics of vector integration in long-term vector-treated mice (A) Study design. (B) Integration frequency in vector-transduced mouse livers determined by TES. (C) Correlation between VCN and integration frequency. (D) Frequency of episomal vector genomes determined by TES. (B and D) Comparisons between Sf- and HEK293-produced vector were not significant unless noted. The integration frequency is the measure of vector-host genome reads, and concatamer frequency is the measure of vector-vector genome reads obtained from the TES results. Data are presented as the mean ± SD. ∗p < 0.05 using a one-way ANOVA with Tukey’s multiple comparison test. n = 6 per time point, with each dot representing a single liver sample from each mouse (average of technical duplicate). AAV5-hA1AT, adeno-associated virus serotype 5-human alpha-1 anti-trypsin; ANOVA, analysis of variance; HEK293, human embryonic kidney 293; HLP, human liver-specific promoter; IS, integration site; ITR, inverted terminal repeat; nt, nucleotide; SD, standard deviation; Sf, Spodoptera frugiperda; TES, target enrichment sequencing; VCN, vector copy number.

Figure 2

PMD clonal framework-based evaluation of clonality (A) Sf- and (B) HEK293 vector-treated samples. Based on the properties of Rényi entropies, a clonality plane was constructed based on two extreme components of diversity: richness (the total number of integration sites) and evenness (the relative number of reads of each integration site). Richness represents the upper bound for evenness, meaning that evenness can never be higher than richness. For this reason, when evenness equals richness, the sample is considered perfectly polyclonal, and it will sit on the line that bisects the first quadrant. On the other hand, the lower the evenness, the closer the sample will be to the x axis. A monoclonal sample is defined by having an evenness of 0. The PMD index reports the ratio of the distance from the theoretical threshold for maximal polyclonality and monoclonality (see figure insert on top-left side of the figure for further details).^. Dist, distance; Dm, distance from monoclonality; Dp, distance from maximum polyclonality; HEK293, human embryonic kidney 293; max, maximum; PMD, polyclonal-monoclonal distance; Sf, Spodoptera frugiperda.

Figure 3

Observed and expected fraction of integrations for Sf- and HEK293 vector-treated mice near regions of open chromatin (A) all regions of open chromatin and (B) open chromatin in close proximity to a TSS. The red line represents the proportion of observed integrations that fall within 20 kb of an ATAC-seq peak, and the histogram represents the distribution that would be expected by chance based on the median simulated value. The observed integrations are a single value because they represent the total number of ISs observed in all samples combined found near open chromatin windows. The simulated values are represented by a histogram to reflect the distribution of each individual simulation from the 10,000 total simulations used in the analysis. ATAC-seq, assay for transposase-accessible chromatin with sequencing; FC, fold change; HEK293, human embryonic kidney 293; IS, integration site; Sf, Spodoptera frugiperda; TSS, transcription start site.

Figure 4

Observed and expected fraction of integrations near genes of different expression levels (A) all samples combined and for samples from (B) Sf- and (C) HEK293 vector-treated mice. The red line represents the proportion of observed integrations that fall within 10 kb of a TSS for a protein-coding gene, and the boxplots represent the distribution of integrations that would be expected by chance based on the median simulated value. Genes with the highest expression profile in the liver (90th percentile) according to the GTEx liver data are classified as highly expressed. p values represent 1 minus the percentile of the observed value within the expected distribution, and the z scores were used as a proxy for enrichment. Boxes represent the interquartile range, whiskers represent the range up to 1.5× the interquartile range, black horizontal lines represent the median. FE, fold enrichment; GTEx, genotype-tissue expression; HEK293, human embryonic kidney 293; Sf, Spodoptera frugiperda; TSS, transcription start site.

Figure 5

Comparison of integrated vector genome fragments between Sf- and HEK293 vector-treated mice (A) ITR, (B) promoter and hA1AT, (C) stuffer sequence, and (D) poly(A) regions. Data are presented as the mean ± SD with n = 6 per time point. Each dot represents a single liver sample from each mouse, and the data were analyzed using a one-way ANOVA with Tukey’s multiple comparison test. ANOVA, analysis of variance; hA1AT, human alpha-1 anti-trypsin; HEK293, human embryonic kidney 293; ITR, inverted terminal repeat; poly(A), polyadenylation; SD, standard deviation; Sf, Spodoptera frugiperda.