Therapeutic levels of FVIII following a single peripheral vein administration of rAAV vector encoding a novel human factor VIII variant

Abstract

Recombinant adeno-associated virus (rAAV) vectors encoding human factor VIII (hFVIII) were systematically evaluated for hemophilia A (HA) gene therapy. A 5.7-kb rAAV-expression cassette (rAAV-HLP-codop-hFVIII-N6) containing a codon-optimized hFVIII cDNA in which a 226 amino acid (aa) B-domain spacer replaced the entire B domain and a hybrid liver-specific promoter (HLP) mediated 10-fold higher hFVIII levels in mice compared with non-codon-optimized variants. A further twofold improvement in potency was achieved by replacing the 226-aa N6 spacer with a novel 17-aa peptide (V3) in which 6 glycosylation triplets from the B domain were juxtaposed. The resulting 5.2-kb rAAV-HLP-codop-hFVIII-V3 cassette was more efficiently packaged within AAV virions and mediated supraphysiologic hFVIII expression (732 ± 162% of normal) in HA knock-out mice following administration of 2 × 10(12) vector genomes/kg, a vector dose shown to be safe in subjects with hemophilia B. Stable hFVIII expression at 15 ± 4% of normal was observed at this dose in a nonhuman primate. hFVIII expression above 100% was observed in 3 macaques that received a higher dose of either this vector or the N6 variant. These animals developed neutralizing anti-FVIII antibodies that were abrogated with transient immunosuppression. Therefore, rAAV-HLP-codop-hFVIII-V3 substantially improves the prospects of effective HA gene therapy.

Figure 1

Comparison of AAV hFVIII variants. (A) Schematics of 4 rAAV vector genomes encoding hFVIII variants under the control of the LP1 or the smaller HLP regulatory element and upstream of either rabbit globin or synthetic polyadenylation sites. Shown are the A1, A2, A3, C1, and C2 domains of hFVIII either with the SQ B-domain sequences (top schematic) or with the proximal 226-aa region of the B domain (N6). The hFVIII cDNA sequence in the lower 2 constructs are codop. (B) Alkaline gel analysis of viral genomes extracted from 5 × 10¹⁰ rAAV5-LP1-codop-hFVIII-N6 (lane 1) and rAAV5-HLP-codop-hFVIII-N6 (lane 2) viral particles. (C) Mean hFVIII:Ag levels ± SD in murine plasma at 6 weeks after a single tail vein administration of rAAV-hFVIII constructs pseudotyped with serotype 5 capsid in wild-type C57Bl/6 mice (N = 3, dose = 2 × 10¹³ vg/kg). (D) Mean (± SD) proviral copy number in murine liver transduced with rAAV5-hFVIII variants.

Figure 2

Evaluation of transduction with rAAV-HLP-codop-hFVIII-N6 in immunocompetent mice. (A) Kinetics of hFVIII:Ag expression following a single tail vein administration of 2 × 10¹³ vg/mouse of rAAV-HLP-codop-hFVIII-N6 pseudotyped with serotype 5 or 8 capsid in wild-type C57Bl/6 mice (N = 3/group). Shown are mean FVIII:Ag levels ± SD. (B, upper left) Southern blot of liver genomic DNA derived from mice (M1 and M2) transduced with rAAV8-HLP-codop-hFVIII-N6 was digested with Kpn I, a double cutter. Standards consist of rAAV-HLP-codop-hFVIII-N6 plasmid DNA spiked into naive mouse liver genomic DNA at the stipulated genome copies per cell concentration and then digested with Kpn I. (Upper right) Uncut DNA or DNA digested with a single cutter (Not I). HH, head-to-head and HT head-to-tail expected band size from proviral DNA in concatemeric configurations, respectively. (Bottom) A schematic showing the relative positions of Kpn I and Not I endonucleases in rAAV-HLP-codop-hFVIII-N6 when in a head to tail concatemeric configuration. (C) Western blot showing a single ∼210-kd band in the plasma of mice transduced with rAAV8-HLP-codop-hFVIII-N6, which is not present in naive mouse plasma. Positive control (+ve control) consisting of full-length recombinant hFVIII (Helixate) diluted in mouse plasma.

Figure 3

Comparison of smaller rAAV-HLP-codop-hFVIII constructs. (A) Schematic of smaller codon-optimized hFVIII expression cassettes. Top schematic consists of B domain–deleted variant, rAAV-HLP-codop-BDD-hFVIII, which is ∼5.1 kb in size. It comprises A1-A2-A3-C1-C2 FVIII domains and contains the SQ sequences between the A2 and A3 domains. rAAV-HLP-codop-hFVIII-V1 and rAAV-HLP-codop-hFVIII-V3 are identical to rAAV-HLP-codop-BDD-hFVIII except that peptides V1 or V3 have been placed within the SQ sequence. Both peptides contain the same 6 N-linked glycosylation triplets. (B) Analysis of viral DNA extracted from 5 × 10¹⁰ particles of each of the codon-optimized hFVIII vector preparation following separation on an alkaline agarose gel and run in duplicate showing bands of ∼5 kb, the expected size for the rAAV-HLP-codop-BDD-hFVIII (BDD) and rAAV-HLP-codop-hFVIII-V3 (V). In comparison, a rather diffuse signal was observed for the genomes extracted from AAV8-HLP-codop-hFVIII-N6 (N6), suggesting the packaging of a more heterogeneous proviral species. (Right) Size standards.

Figure 4

Evaluation of AAV FVIII variants in vivo. (A) Mean hFVIII:Ag levels ± SD in murine plasma derived from male wild-type C57Bl/6 mice (N = 6) at 6 weeks after tail vein injection of 2 × 10¹² vg/kg of rAAV8-HLP-codop-hFVIII variants. (B) Sodium dodecyl sulfonate polyacrylamide gel electrophoresis/western blotting of murine plasma following gene transfer with rAAV-HLP-codop-BDD-hFVIII (BDD) and rAAV8-HLP-codop-hFVIII-V3 (V3) using a polyclonal rhesus anti-hFVIII antibody showing the heavy (∼90 kDa) and light (∼80 kDa) chains as well as a 170-kDa nonprocessed, primary translation product. Negative control (−ve) is naive mouse plasma and positive control consists of recombinant BDD hFVIII (ReFacto, first lane) diluted in murine plasma containing protease inhibitors. (C) Relationship between rAAV8-HLP-codop-hFVIII-V3 dose and hFVIII:Ag levels (mean ± SD) in murine plasma and (D) transgene copy number (mean ± SD) at 12 weeks following gene transfer.

Figure 5

Expression of hFVIII expression in F8^−/− mice. (A-D) Kinetics of hFVIII expression shown as activity (hFVIII:C; mean ± SD) and antigen (hFVIII:Ag; mean ± SD) levels in male F8^−/− mice following a single tail vein administration of low (2 × 10¹² vg/kg) and high dose (2 × 10¹³ vg/kg) of (A) rAAV-HLP-codop-BDD-hFVIII (BDD), (B) rAAV-HLP-codop-hFVIII-N6 (N6), or (C) rAAV-HLP-codop-hFVIII-V3 (V3). Standards consisted of normal human pool plasma (NHP) diluted in murine plasma. (D) Transgene copy number in the liver at 8 weeks following gene transfer of rAAV-HLP-codop-BDD-hFVIII, rAAV-HLP-codop-hFVIII-N6, and rAAV-HLP-codop-hFVIII-V3. (E) Blood loss in F8^−/− mice following gene transfer with rAAV-HLP-codop-BDD-hFVIII, rAAV-HLP-codop-hFVIII-N6, and rAAV-HLP-codop-hFVIII-V3 (V3) compared with that in F8^−/− mice treated with vehicle alone or recombinant FVIII.

Figure 6

Evaluation of AAV FVIII variants in non-human primates. (A) hFVIII:Ag levels (solid black line; mean ± SD, left y-axis) and anti-hFVIII antibody titers (dashed gray line; relative units, right y-axis) as assessed by immunocapture assays over time in 4 2- to 4-year-old male monkeys following a single peripheral vein tail vein administration of either rAAV8-HLP-codop-hFVIII-N6 (N6) or rAAV8-HLP-codop-hFVIII-V3 (V3). Animals that developed neutralizing anti-hFVIII antibodies were treated with 4 monthly cycles of rituximab and cyclophosphamide shown by arrows. (B) Southern blot analysis showing molecular configuration of rAAV8-HLP-codop-hFVIII-N6 genome in macaque liver. Approximately 15 µg of DNA from each time point (1, 4, and 40 weeks after gene transfer) was electrophoresed uncut (uncut) or following digestion with Not I (single cutter) or Kpn I (double cutter). Shown are bands representing high-molecular-weight concatemers (HMWC) in the head-tail (HT) and head-to-head (HH) formation. (Bottom) A schematic of rAAV-HLP-codop-hFVIII-N6 in the HT configuration together with the cleavage sites for Kpn I and Not I.