Harrison C. Brown

Enhanced biosynthesis of coagulation factor VIII through diminished engagement of the unfolded protein response.

Human and porcine coagulation factor VIII (fVIII) display a biosynthetic efficiency differential that is being exploited for the development of new protein and gene transfer-based therapies for hemophilia A. The cellular and/or molecular mechanism(s) responsible for this phenomenon have yet to be uncovered, although it has been temporally localized to post-translational biosynthetic steps. The unfolded protein response (UPR) is a cellular adaptation to structurally distinct (e.g. misfolded) or excess protein in the endoplasmic reticulum and is known to be induced by heterologous expression of recombinant human fVIII. Therefore, it is plausible that the biosynthetic differential between human and porcine fVIII results from differential UPR activation. In the current study, UPR induction was examined in the context of ongoing fVIII expression. UPR activation was greater during human fVIII expression when compared with porcine fVIII expression as determined by ER response element (ERSE)-luciferase reporter activity, X-box-binding protein 1 (XBP1) splicing, and immunoglobulin-binding protein (BiP) up-regulation. Immunofluorescence microscopy of fVIII expressing cells revealed that human fVIII was notably absent in the Golgi apparatus, confirming that endoplasmic reticulum to Golgi transport is rate-limiting. In contrast, a significant proportion of porcine fVIII was localized to the Golgi indicating efficient transit through the secretory pathway. Overexpression of BiP, an integral UPR protein, reduced the secretion of human fVIII by 50%, but had no effect on porcine fVIII biosynthesis. In contrast, expression of BiP shRNA increased human fVIII expression levels. The current data support the model of differential engagement of UPR by human and porcine fVIII as a non-traditional mechanism for regulation of gene product biosynthesis.

Enhancing the pharmaceutical properties of protein drugs by ancestral sequence reconstruction

Optimization of a proteins pharmaceutical properties is usually carried out by rational design and/or directed evolution. Here we test an alternative approach based on ancestral sequence reconstruction. Using available genomic sequence data on coagulation factor VIII and predictive models of molecular evolution, we engineer protein variants with improved activity, stability, and biosynthesis potential and reduced inhibition by anti-drug antibodies. In principle, this approach can be applied to any protein drug based on a conserved gene sequence.

Factor VIII A3 domain substitution N1922S results in hemophilia A due to domain-specific misfolding and hyposecretion of functional protein

A point mutation leading to amino acid substitution N1922S in the A3 domain of factor VIII (fVIII) results in moderate to severe hemophilia A. A heterologous expression system comparing N1922S-fVIII and wild-type fVIII (wt-fVIII) demonstrated similar specific coagulant activities but poor secretion of N1922S-fVIII. Immunocytochemical analysis revealed that intracellular levels of N1922S-fVIII were similar to those of wt-fVIII. The specific activity of intracellular N1922S-fVIII was 10% of that of wt-fVIII, indicating the presence of large amounts of a nonfunctional N1922S-fVIII-folding intermediate. wt-fVIII colocalized with both endoplasmic reticulum (ER)- and Golgi-resident proteins. In contrast, N1922S-fVIII colocalized only with ER-resident proteins, indicating a block in transit from the ER to the Golgi. A panel of conformation-dependent monoclonal antibodies was used to determine native or nonnative folding of N1922S-fVIII. Intracellular N1922S-fVIII but not secreted N1922S-fVIII displayed abnormal folding in the A3 and C1 domains, indicating that the A1, A2, and C2 domains fold independently into antigenically intact tertiary structures, but that folding is stalled in the mutant A3 and its contiguous C1 domain. In summary, the N1922S substitution results in poor secretion of a functional protein, and the domain-specific defect in folding and intracellular trafficking of N1922S-fVIII is a novel mechanism for secretion defects leading to hemophilia A.

Bioengineered coagulation factor VIII enables long-term correction of murine hemophilia A following liver-directed adeno-associated viral vector delivery

Clinical data support the feasibility and safety of adeno-associated viral (AAV) vectors in gene therapy applications. Despite several clinical trials of AAV-based gene transfer for hemophilia B, a unique set of obstacles impede the development of a similar approach for hemophilia A. These include (i) the size of the factor VIII (fVIII) transgene, (ii) humoral immune responses to fVIII, (iii) inefficient biosynthesis of human fVIII, and (iv) AAV vector immunity. Through bioengineering approaches, a novel fVIII molecule, designated ET3, was developed and shown to improve biosynthetic efficiency 10- to 100-fold. In this study, the utility of ET3 was assessed in the context of liver-directed, AAV-mediated gene transfer into hemophilia A mice. Due to the large size of the expression cassette, AAV-ET3 genomes packaged into viral particles as partial genome fragments. Despite this potential limitation, a single peripheral vein administration of AAV-ET3 into immune-competent hemophilia A mice resulted in correction of the fVIII deficiency at lower vector doses than previously reported for similarly oversized AAV-fVIII vectors. Therefore, ET3 appears to improve vector potency and mitigate at least one of the critical barriers to AAV-based clinical gene therapy for hemophilia A.

Effects of FVIII immunity on hepatocyte and hematopoietic stem cell–directed gene therapy of murine hemophilia A

Immune responses to coagulation factors VIII (FVIII) and IX (FIX) represent primary obstacles to hemophilia treatment. Previously, we showed that hematopoietic stem cell (HSC) retroviral gene therapy induces immune nonresponsiveness to FVIII in both naive and preimmunized murine hemophilia A settings. Liver-directed adeno-associated viral (AAV)-FIX vector gene transfer achieved similar results in preclinical hemophilia B models. However, as clinical immune responses to FVIII and FIX differ, we investigated the ability of liver-directed AAV-FVIII gene therapy to affect FVIII immunity in hemophilia A mice. Both FVIII naive and preimmunized mice were administered recombinant AAV8 encoding a liver-directed bioengineered FVIII expression cassette. Naive animals receiving high or mid-doses subsequently achieved near normal FVIII activity levels. However, challenge with adjuvant-free recombinant FVIII induced loss of FVIII activity and anti-FVIII antibodies in mid-dose, but not high-dose AAV or HSC lentiviral (LV) vector gene therapy cohorts. Furthermore, unlike what was shown previously for FIX gene transfer, AAV-FVIII administration to hemophilia A inhibitor mice conferred no effect on anti-FVIII antibody or inhibitory titers. These data suggest that functional differences exist in the immune modulation achieved to FVIII or FIX in hemophilia mice by gene therapy approaches incorporating liver-directed AAV vectors or HSC-directed LV.

728. Tissue-Directed Transgene Engineering for AAV and Lentivector Gene Therapy Approaches

Clinical gene therapy frequently is encumbered by low transgene product biosynthesis at predictably safe vector doses. It has been hypothesized that the presence of rare codons may regulate transgene product expression through depletion of the available cognate tRNA pool. Codon optimization is the prominent strategy utilized to overcome this hypothesized limitation and involves replacing rare, presumably translation rate-limiting, codons with the most frequent ones. Typical algorithms attempt to match the codon usage frequency of the transgene coding sequence to that of target organisms total mRNA pool, which has been shown to approximate the overall available tRNA concentrations. Upon closer examination, it appears that both codon frequency and tRNA content vary between tissue types. Therefore, we hypothesize that codon-optimization can be improved by tailoring to the codon-frequency of the most highly expressed genes present in target cell types. Our tissue-directed codon optimization algorithm utilizes novel codon usage indices generated from target cell gene expression data. As proof of concept, we developed tissue-codon optimized variants of coagulation factor VIII (FVIII) to be utilized in lenti- and adeno-associated viral (LV and AAV, respectively) vectors. These two vectors are the leading platforms for clinical gene therapy of hemophilia A. LV is utilized to target autologous hematopoietic cell types ex vivo while AAV is delivered intravenously to genetically-modify hepatocytes. However to date, development-stage LV and AAV gene therapy products for hemophilia A have been characterized by low-level FVIII transgene product biosynthesis. Initially, we designed human hepatocyte-, monocyte- and standard overall human-optimized FVIII constructs (LCO, MCO and HCO, respectively) to be compared to wild-type FVIII. Upon initial examination, expression of LCO-FVIII was shown to be 3-fold greater than either MCO- or wild-type FVIII from the human hepatoma cell line, HepG2, following transient transfection. In contrast, LCO-FVIII and MCO-FVIII expression was diminished 12 and 4-fold, respectively, compared to wild-type in the ‘neutral’ human embryonic kidney 293T cell line. Furthermore, following hydrodynamic injection of naked plasmid DNA into hemophilia A mice, LCO-FVIII exhibited a sustained 10-fold increase in FVIII expression relative to the HCO-FVIII comparator. In attempt to generate a lead candidate for clinical translation, we utilized several of our most promising vector components to construct a liver-optimized AAV8 vector consisting of LCO-ET3, our previously described high-expression bioengineered FVIII variant, transcriptionally driven by a novel 146bp liver-directed promoter and adjacent MVM intron. Following intravenous delivery into hemophilia A mice, vector doses of 1e12 and 1e11 vector genomes per kg achieved sustained, predictable curative plasma FVIII levels of 200% and 20% of normal human levels, respectively. These initial results support the utility of our novel approach of clinical tissue-directed transgene optimization.

Target-Cell-Directed Bioengineering Approaches for Gene Therapy of Hemophilia A

Potency is a key optimization parameter for hemophilia A gene therapy product candidates. Optimization strategies include promoter engineering to increase transcription, codon optimization of mRNA to improve translation, and amino-acid substitution to promote secretion. Herein, we describe both rational and empirical design approaches to the development of a minimally sized, highly potent AAV-fVIII vector that incorporates three unique elements: a liver-directed 146-nt transcription regulatory module, a target-cell-specific codon optimization algorithm, and a high-expression bioengineered fVIII variant. The minimal synthetic promoter allows for the smallest AAV-fVIII vector genome known at 4,832 nt, while the tissue-directed codon optimization strategy facilitates increased fVIII transgene product expression in target cell types, e.g., hepatocytes, over traditional genome-level codon optimization strategies. As a tertiary approach, we incorporated ancient and orthologous fVIII sequence elements previously shown to facilitate improved biosynthesis through post-translational mechanisms. Together, these technologies contribute to an AAV-fVIII vector that confers sustained, curative levels of fVIII at a minimal dose in hemophilia A mice. Moreover, the first two technologies should be generalizable to all liver-directed gene therapy vector designs.

456. Transgene Bioengineering Through Ancestral Protein Reconstruction

Bioengineering of the transgene often is a critical component of preclinical gene therapy R&D. Transgenes and their products represent the active agent in nucleic acid pharmaceuticals and similar to small molecule pharmaceuticals, they can be modified to possess improved pharmacological properties. However as they are significantly more complex than small molecules, the available strategies for bioengineering, such as in silico rational design, directed evolution and homolog/ortholog-scanning mutagenesis, are less robust. Herein, we propose combined ancestral sequence and protein reconstruction (ASR and APR, respectively) as newly accessible approaches to transgene/transgene product bioengineering. ASR is the prediction of ancient sequences from extant ones and well developed ASR methods and tools now exist. Furthermore, the availability of de novo custom DNA synthesis and recombinant protein expression systems now facilitates APR to complement and extend ASR findings. Previously through the study of extant FVIII orthologs, we discovered that differential molecular, biochemical and immunological properties with exist and could have a positive pharmacological impact upon engineering into human FVIII. For example, porcine FVIII was shown to display 10-100-fold more efficient biosynthesis than human FVIII in vitro and in vivo, while murine FVIII displays 5 - 10-fold greater stability following thrombin activation. Ovine FVIII displays intermediate biosynthesis and stability, but strikingly reduced cross-reactivity to anti-human FVIII inhibitory antibodies. APR provides a high-resolution mapping solution to these ortholog sequence-activity relationships and also takes advantage of the observation that ancient proteins often have unpredicted and/or expanded functionalities that can be efficiently mapped to specific amino acid residues through comparisons of ancestral proteins and genes within an evolutionary lineage. Therefore, we sought to validate APR as a FVIII discovery/bioengineering platform with the expectation that this approach can be successfully applied to essentially all hemostatic, as well as non-hemostatic, gene therapies. Initially, we employed ASR/APR to resurrect 14 ancestral (An) FVIII molecules. Each An-FVIII was shown to be active in standard coagulation assays using human plasma demonstrating evolutionary compatibility. To study biosynthetic efficiency, secreted An-FVIII activity and mRNA transcript levels were analyzed from stably transfected cells demonstrating that, An-53, an ancestral primate sequence with 95% identity to extant human FVIII, displayed the greatest biosynthetic efficiency equivalent to porcine FVIII and our lead bioengineered high expression FVIII, ET3. As a proxy for AAV gene therapy, hemophilia A mice were administered several doses of a liver-directed An-53 AAV plasmid DNA cassette via hydrodynamic injection resulting in peak plasma FVIII activity levels ≥12-fold higher than observed with the ET3 transgene. In addition to superior biosynthetic efficiency, we have identified An-FVIII variants with 2 - 3 fold improved specific activity and stability greater or equal to murine FVIII. Furthermore, we have identified An-FVIII molecules that display reduced immune reactivity and have used these constructs to define functional epitopes to the single amino acid level. Currently, we are refining this approach to identify the key functional residues responsible for each property with the goal of improving the pharmacology of the human FVIII transgenes.

244. Factor VIII Immunogenicity Is a Barrier to AAV Gene Therapy for Hemophilia A

Liver-directed adeno-associated virus (AAV) gene therapy utilizing a bioengineered, high-expression coagulation factor VIII (FVIII) transgene (ET3) has been demonstrated by our laboratory to produce therapeutic FVIII levels in a naive, immune-competent murine model of hemophilia A (HA). Recently, other groups have shown that similar liver directed gene therapy approaches also have the potential to reverse pre-existing immunity, including inhibitors, to coagulation factor IX (FIX) in murine models of hemophilia B (HB). In the current study, we sought to determine if, similarly, ET3-AAV8 could reverse pre-existing FVIII immunity in HA mice. Cohorts of mice were immunized with either recombinant ET3 or human FVIII (hFVIII). Robust anti-FVIII immune responses were elicited as expected and characterized by ELISA (total IgG) and Bethesda (inhibitor) titers ranging from 6000-15,400 arbitrary units (AU) and 22-286 BU, respectively, for ET3 immunized mice, while hFVIII immunized mice showed respective ELISA and Bethesda titers ranging from 460-1480 AU and 23-63 BU. Subsequently, the mice were administered ET3-AAV8 at a dose (4e12 vp/kg) previously shown to produce complete correction of the FVIII deficiency (>50% normal human FVIII activity) and monitored for 16 weeks. No significant change was observed in either ELISA or Bethesda titers for either cohort indicating that the pre-existing immunity was not modulated substantively by FVIII-AAV8 (p>0.5). As Arruda and colleagues previously showed that liver-directed FVIII-AAV8 can eradicate pre-existing anti-FVIII antibodies in a canine model of HA harboring low Bethesda titers (£12 BU), we hypothesized that the capacity for ET3-AAV8 to eliminate pre-existing antibodies may exist under a certain immune threshold quantitatively linked to the inhibitor titer. To test this hypothesis, cohorts of HA mice were established with pre-existing Bethesda titers ranging from 0-25 BU (n=10). Following administration of ET3-AAV8, no significant changes in titers have been observed to date. A corollary question then arose regarding the immune modulation potential of FVIII-AAV8 in naive HA mice, which has been studied extensively for FIX-AAV in HB mice. Specifically, we tested if a mid (4e12vp/kg) or high (2e13vp/kg) dose of ET3-AAV8 could tolerize naive HA mice to recombinant FVIII infusions. FVIII production was dose dependent with the mid cohort averaging 70% normal human levels (0.7 IU/ml) and the high dose cohort averaging 200% normal levels (2 IU/ml) over 5 months. Mice were then challenged with weekly intravenous injections of ET3. All mice treated with the mid AAV dose developed anti-FVIII ELISA and Bethesda titers similar to challenged controls with a concomitant loss of detectable plasma FVIII activity. However, mice in the high dose cohort maintained FVIII activity indicating a possible antigen production requirement for immune modulation consistent with current models of active tolerance. Collectively, these preclinical data indicate that current clinical FIX-AAV gene therapy data may not predict clinical outcomes for similar FVIII-AAV products, especially as trials progress into previously untreated and inhibitor populations.

682. Applied In Silico Engineering of Factor VIII-AAV Transgene Cassettes

Efforts to develop a clinical adeno-associated virus (AAV) vector product for hemophilia A have been hampered by many factors including low-level factor VIII (FVIII) expression, inefficient vector manufacture, and dose limiting vector toxicity. We recently described pre-clinical testing of a 5.86kb AAV2/8 vector encoding a bioengineered FVIII transgene (ET3) that confers enhanced FVIII secretion efficiency. FVIII-deficient (hemophilia A) mice treated at vector doses ranging from 5e11-2e13 vp/kg were corrected of their FVIII deficiency and bleeding phenotype. However due its oversized genome, the vector suffered from low titer manufacture and substantial inter-particle heterogeneity. Although the packaging capacity of AAV vectors is debated, vector genome sizes 4.7-5.0kb are manufactured to higher yield and consistency than those exceeding 5.0kb. The B-domain deleted FVIII coding sequence is 4.4kb, and with the addition of necessary viral and regulatory control elements, FVIII-AAV genomes exceed the packaging capacity of the virus by 10-25%. We therefore sought to engineer a FVIII-AAV genome <5.0kb in length allowing for both enhanced FVIII expression and efficient packaging. Although ET3 already exhibits enhanced secretion over human FVIII due to the incorporation of specific non-native amino acids, others have shown synonymous codon optimization to benefit hepatic FVIII-AAV expression as well. Therefore, we examined the codon usage of 43 highly expressed, liver-specific genes and used the resulting information to generate a novel codon usage table. Using this table, we synthesized codon-optimized versions of both BDD human FVIII and ET3, the former of which demonstrated 3-fold superior expression in vitro. Likewise, this novel codon optimization algorithm provided the ET3-AAV transgene cassette with a 4-fold in vitro and 14-fold in vivo enhancement of expression. However, the large size of the codon optimized ET3-AAV genome remained incompatible with efficient viral vector packaging. To address this design limitation, we used a combinatorial transcription factor binding site assembly approach to create a panel of liver-specific promoters ranging in size from 62-172 bases. These promoters represent a 30-90% size reduction over currently utilized liver specific promoters such as HLP and HCR-hAAT, which range in size from 250 to over 700 bases. Critically, a small subset of these promoters drive comparable transgene expression levels and specificity to that observed with HLP and HCR-hAAT. As a last design change, we eliminated nonessential viral genomic DNA and cloning remnants that are present in many earlier AAV designs. This removed an additional 80 bases. Collectively, these approaches enabled the development of a high expression FVIII-AAV genome of only 4.8kb in length. The exceptionally small size of this FVIII transgene may allow for the addition of desirable regulatory control features such as efficient introns or scaffold matrix attachment regions that could provide additional benefit. Furthermore, these technologies should be generalizable and likely enabling to other liver-directed AAV gene therapies.