- Email: [email protected]
299. Directed Evolution and Rational Design Combined for Enhanced AAV Vectors
AAV VECTOR DEVELOPMENT 298. Characterization of an AAV Capsid Library Using PacBio CCS Single Molecule Sequencing Damien Marsic,1 Sergei Zolotukhin.1 1 Pediatrics, University of Florida, Gainesville, FL.
Sequencing a library that potentially contains millions of variants represents a challenge, especially when the DNA fragment to be sequenced is much larger than the read length generated by most next-gen sequencing platforms. We analyzed an AAV capsid library, in the form of a 1399 bp fragment of the capsid gene containing 182 variable nucleotide positions distributed in 8 regions, using PacBio, the only large-scale sequencing platform to date capable of generating the long reads required for this kind of application. Pacbio’s most serious limitations are a relatively low throughput compared with other next-gen platforms, as well as a very low accuracy. We decided to use the Circular Consensus Sequencing (CCS) mode, in which circularized single molecules are continuously read and a consensus sequence is computed if at least 3 passes have been completed. A total of 52474 post-filter (but pre-CCS computing) reads were obtained, with a mean read length of 3401 nt. Therefore, the average depth of coverage was only 2.4x, which means that a majority of reads were discarded for not reaching the threshold of 3 passes. The number of final CCS reads was 19455, of which 19053 reads had a length that was within 10% of the reference length. However, not a single read could be found that matched the reference sequence, indicating a high prevalence of sequencing errors. Because no software support existed to analyze our particular dataset, a dedicated code (caplib, publicly available at https://sourceforge.net/projects/caplib/) was developed. The program attempts to align conserved regions of the CCS reads to a reference sequence, and retrieves the variable regions only if they match exactly the reference sequence (including ambiguous nucleotides). The corrected reads generated by caplib are therefore derived from the CCS reads that likely have no errors in any of the variable regions. Obviously, this approach excludes all sequences which could have genuine insertions or deletions in the variable regions, and ignores any possible variation in the conserved regions. Only 840 sequences were recovered, which nevertheless provided invaluable information on the mutability of particular amino acid positions in the capsid, thus revealing surface areas susceptible to directed evolution as well as regions critical for structural integrity. Although the number of recovered sequences might seem very small compared to the sequencing output size, it only required a fraction of the cost, time and efforts that would have been necessary to reach the same results with traditional Sanger sequencing.
299. Directed Evolution and Rational Design Combined for Enhanced AAV Vectors
Hanen Khabou,1,2,3 Peggy Barbe,1,2,3 Gwenaëlle Auregan,4 Mélissa Desrosiers,1,2,3 Alexis-Pierre Bemelmans,4 Deniz Dalkara.1,2,3 1 INSERM UMR_S 968, Institut de la Vision, Paris, France; 2 Sorbonne Universités, UPMC Univ Paris 06, UMR_S 968, Institut de la Vision, Paris, France; 3CNRS, UMR_7210, Paris, France; 4 Molecular Imaging Research Center (MIRCen) and CNRS URA2210, Commissariat à l’Energie Atomique et aux Energies Alternatives (CEA), Département des Sciences du Vivant (DSV), Institut d’Imagerie Biomédicale (I2BM),. Adeno-associated virus (AAV) is currently the most efficient vehicle for retinal gene delivery, as shown by the successful clinical trials for the treatment of Leber congenital amaurosis. However, first generation vector technology used in these clinical trials needs improvements, especially for widespread gene delivery to the neural retina. Recently, a novel AAV variant that transduces the entirety of the neural retina and RPE was obtained using in-vivo directed evolution. In this directed evolution screen, a transgenic mouse line expressing GFP in its photoreceptors was employed to select for viral Molecular Therapy Volume 22, Supplement 1, May 2014 Copyright © The American Society of Gene & Cell Therapy
variants that can penetrate from the vitreous into this deeper layer of the neural retina where photoreceptors are located. The successful variant isolated from this screen, called AAV2.7m8, is characterized by an insertion of a heptamer (LGETTRP) near the three-fold axis of symmetry on the AAV2 capsid. The peptide insertion is responsible for the modification of virus tropism leading to significantly improved retinal penetration properties, and high level transduction of photoreceptors. In this study, we investigated whether the insertion of this particular peptide is responsible for better transduction of deep retinal layers and better spread. To do this, the 7m8 peptide was used to create novel AAV variants based on AAV5, AAV8, AAV9 and AAVrh10. These serotypes were chosen based on desired properties they have such as faster kinetics and/or higher initial photoreceptor tropism compared to AAV2. These novel 7m8 insertion serotypes were then evaluated for their ability for retinal gene delivery in comparison with the unmodified AAV5, 8, 9 and rh10 by intravitreal administration in the mouse retina. We also investigated the efficiency of gene delivery and the diffusion capacities of our AAV variants in the brain after intracerebral administration. Intravitreally delivered AAV9.7m8 had higher efficiency of transduction than that of its parental serotype as previously shown with AAV2.7m8, while 7m8 insertion did not markedly improve the tropism of AAV5, 8 or rh10. Altogether our results indicate that improvement of transduction following 7m8 insertion is capsid-dependent.
300. Novel AAV Vectors Mediate Efficient Intramuscular Gene Delivery
Taihra Ul-Hasan,1 Laura J. Smith,1 Manasa Chandra,1 K. K. Wong,2 Saswati Chatterjee.1 1 Virology, City of Hope, Duarte, CA; 2Hematology/Stem Cell Transplantation, City of Hope, Duarte, CA. Novel AAV capsids isolated from CD34+ cells of healthy adult peripheral blood, termed AAVHSCs, have been found to efficiently transduce CD34+ human hematopoietic stem cells from cord and peripheral blood and support high levels of stable luciferase expression in murine xenograft models. Some AAVHSCs also efficiently transduce specific organs such as liver, heart, lung, and muscle. We evaluated the ability of one of these isolates, AAVHSC15, to support long-term transgene expression following intramuscular injection as compared with AAV8 in both immunocompetent C57BL/6 and immunodeficient NOD/SCID mice. Mice were injected in the muscle with 1e10, 5e10, or 1e11 particles of firefly luciferaseencoding vectors. Luciferase expression was evaluated by serial bioluminescent imaging up to >188 days after injection. AAV8 and AAVHSC15 both supported sustained long-term expression in vivo. Transient luciferase expression was observed for about 4 weeks in the liver, likely due to leakage of vector into the circulation. Luciferase expression was readily detectable in the muscle from day 3. While both AAV8 and AAVHSC15 supported sustained long-term expression in muscle, luciferase expression was 3 to 5-fold higher in AAVHSC15 injected mice. Interestingly, several differences in long term transgene expression were noted between immunocompetent and immunodeficient mice. These included: 1. Greater fluctuation in early transgene expression levels in immunocompetent mice, for up to 8 weeks; 2. Stable expression levels were approximately >10-fold higher in immunodeficient mice than in C57BL/6; 3. A greater differential in stable intramuscular expression levels was observed between AAVHSC15 and AAV8 in NOD/SCID mice, with AAVHSC15 supporting >10-fold higher levels than AAV8. To explore the underlying causes of these differences, we evaluated anti-capsid neutralizing antibodies in the sera of AAV injected mice. While no neutralizing antibodies were detected in naïve pre-injection sera, anti-AAV8 antibodies were detectable by 4 and 11 weeks postinjection. Importantly, no anti-AAVHSC15 antibodies were detected S115
Sequencing a library that potentially contains millions of variants represents a challenge, especially when the DNA fragment to be sequenced is much larger than the read length generated by most next-gen sequencing platforms. We analyzed an AAV capsid library, in the form of a 1399 bp fragment of the capsid gene containing 182 variable nucleotide positions distributed in 8 regions, using PacBio, the only large-scale sequencing platform to date capable of generating the long reads required for this kind of application. Pacbio’s most serious limitations are a relatively low throughput compared with other next-gen platforms, as well as a very low accuracy. We decided to use the Circular Consensus Sequencing (CCS) mode, in which circularized single molecules are continuously read and a consensus sequence is computed if at least 3 passes have been completed. A total of 52474 post-filter (but pre-CCS computing) reads were obtained, with a mean read length of 3401 nt. Therefore, the average depth of coverage was only 2.4x, which means that a majority of reads were discarded for not reaching the threshold of 3 passes. The number of final CCS reads was 19455, of which 19053 reads had a length that was within 10% of the reference length. However, not a single read could be found that matched the reference sequence, indicating a high prevalence of sequencing errors. Because no software support existed to analyze our particular dataset, a dedicated code (caplib, publicly available at https://sourceforge.net/projects/caplib/) was developed. The program attempts to align conserved regions of the CCS reads to a reference sequence, and retrieves the variable regions only if they match exactly the reference sequence (including ambiguous nucleotides). The corrected reads generated by caplib are therefore derived from the CCS reads that likely have no errors in any of the variable regions. Obviously, this approach excludes all sequences which could have genuine insertions or deletions in the variable regions, and ignores any possible variation in the conserved regions. Only 840 sequences were recovered, which nevertheless provided invaluable information on the mutability of particular amino acid positions in the capsid, thus revealing surface areas susceptible to directed evolution as well as regions critical for structural integrity. Although the number of recovered sequences might seem very small compared to the sequencing output size, it only required a fraction of the cost, time and efforts that would have been necessary to reach the same results with traditional Sanger sequencing.
299. Directed Evolution and Rational Design Combined for Enhanced AAV Vectors
Hanen Khabou,1,2,3 Peggy Barbe,1,2,3 Gwenaëlle Auregan,4 Mélissa Desrosiers,1,2,3 Alexis-Pierre Bemelmans,4 Deniz Dalkara.1,2,3 1 INSERM UMR_S 968, Institut de la Vision, Paris, France; 2 Sorbonne Universités, UPMC Univ Paris 06, UMR_S 968, Institut de la Vision, Paris, France; 3CNRS, UMR_7210, Paris, France; 4 Molecular Imaging Research Center (MIRCen) and CNRS URA2210, Commissariat à l’Energie Atomique et aux Energies Alternatives (CEA), Département des Sciences du Vivant (DSV), Institut d’Imagerie Biomédicale (I2BM),. Adeno-associated virus (AAV) is currently the most efficient vehicle for retinal gene delivery, as shown by the successful clinical trials for the treatment of Leber congenital amaurosis. However, first generation vector technology used in these clinical trials needs improvements, especially for widespread gene delivery to the neural retina. Recently, a novel AAV variant that transduces the entirety of the neural retina and RPE was obtained using in-vivo directed evolution. In this directed evolution screen, a transgenic mouse line expressing GFP in its photoreceptors was employed to select for viral Molecular Therapy Volume 22, Supplement 1, May 2014 Copyright © The American Society of Gene & Cell Therapy
variants that can penetrate from the vitreous into this deeper layer of the neural retina where photoreceptors are located. The successful variant isolated from this screen, called AAV2.7m8, is characterized by an insertion of a heptamer (LGETTRP) near the three-fold axis of symmetry on the AAV2 capsid. The peptide insertion is responsible for the modification of virus tropism leading to significantly improved retinal penetration properties, and high level transduction of photoreceptors. In this study, we investigated whether the insertion of this particular peptide is responsible for better transduction of deep retinal layers and better spread. To do this, the 7m8 peptide was used to create novel AAV variants based on AAV5, AAV8, AAV9 and AAVrh10. These serotypes were chosen based on desired properties they have such as faster kinetics and/or higher initial photoreceptor tropism compared to AAV2. These novel 7m8 insertion serotypes were then evaluated for their ability for retinal gene delivery in comparison with the unmodified AAV5, 8, 9 and rh10 by intravitreal administration in the mouse retina. We also investigated the efficiency of gene delivery and the diffusion capacities of our AAV variants in the brain after intracerebral administration. Intravitreally delivered AAV9.7m8 had higher efficiency of transduction than that of its parental serotype as previously shown with AAV2.7m8, while 7m8 insertion did not markedly improve the tropism of AAV5, 8 or rh10. Altogether our results indicate that improvement of transduction following 7m8 insertion is capsid-dependent.
300. Novel AAV Vectors Mediate Efficient Intramuscular Gene Delivery
Taihra Ul-Hasan,1 Laura J. Smith,1 Manasa Chandra,1 K. K. Wong,2 Saswati Chatterjee.1 1 Virology, City of Hope, Duarte, CA; 2Hematology/Stem Cell Transplantation, City of Hope, Duarte, CA. Novel AAV capsids isolated from CD34+ cells of healthy adult peripheral blood, termed AAVHSCs, have been found to efficiently transduce CD34+ human hematopoietic stem cells from cord and peripheral blood and support high levels of stable luciferase expression in murine xenograft models. Some AAVHSCs also efficiently transduce specific organs such as liver, heart, lung, and muscle. We evaluated the ability of one of these isolates, AAVHSC15, to support long-term transgene expression following intramuscular injection as compared with AAV8 in both immunocompetent C57BL/6 and immunodeficient NOD/SCID mice. Mice were injected in the muscle with 1e10, 5e10, or 1e11 particles of firefly luciferaseencoding vectors. Luciferase expression was evaluated by serial bioluminescent imaging up to >188 days after injection. AAV8 and AAVHSC15 both supported sustained long-term expression in vivo. Transient luciferase expression was observed for about 4 weeks in the liver, likely due to leakage of vector into the circulation. Luciferase expression was readily detectable in the muscle from day 3. While both AAV8 and AAVHSC15 supported sustained long-term expression in muscle, luciferase expression was 3 to 5-fold higher in AAVHSC15 injected mice. Interestingly, several differences in long term transgene expression were noted between immunocompetent and immunodeficient mice. These included: 1. Greater fluctuation in early transgene expression levels in immunocompetent mice, for up to 8 weeks; 2. Stable expression levels were approximately >10-fold higher in immunodeficient mice than in C57BL/6; 3. A greater differential in stable intramuscular expression levels was observed between AAVHSC15 and AAV8 in NOD/SCID mice, with AAVHSC15 supporting >10-fold higher levels than AAV8. To explore the underlying causes of these differences, we evaluated anti-capsid neutralizing antibodies in the sera of AAV injected mice. While no neutralizing antibodies were detected in naïve pre-injection sera, anti-AAV8 antibodies were detectable by 4 and 11 weeks postinjection. Importantly, no anti-AAVHSC15 antibodies were detected S115