Codon Modification and PAMPs in Clinical AAV Vectors: The Tortoise or the Hare?

Main Text

Recombinant AAV (rAAV) vectors are a key component of an emergent therapeutic paradigm with a demonstrated definitive benefit for genetic diseases. Immune responses are among the most challenging barriers to human gene therapy. Viral vectors are highly advantageous because of their evolved ability to negotiate complex intracellular pathways to efficiently deliver a DNA payload; however, the human immune system has evolved multiple pathways, including innate pathogen-associated molecular pattern (PAMP) sensors that trigger adaptive effector functions to eliminate infected cells.¹ While rAAV product developers cannot avoid using AAV capsid, which is the source of viral peptides that render transduced cells targets for capsid-specific cytotoxic T lymphocytes (CTLs), the elimination of immune co-stimulatory features is important. Unmethylated CpG dinucleotide-based motifs (CpGs) are known PAMPs that bind and dimerize monomeric TLR9 expressed in human dendritic cells^2,3 and have been shown to cause activation of the Toll-like receptor (TLR)9-MyD88 signaling pathway, thereby promoting CTL responses to AAV vectors in non-clinical models.4, 5, 6 Polynucleotides containing unmethylated CpGs are adjuvants used in vaccine development to stimulate strong cellular immune responses.⁷ Details now available for eight hemophilia B gene therapy trials that used differing codon-modification strategies resulting in a broad range of CpG content (0- to 5-fold of wild type) in the factor IX (FIX) open reading frame (ORF) reveal that low CpG correlates strongly to long-term expression. Herein is provided a perspective that unmethylated CpG content in AAV vectors is the “key” attribute that triggers transgene expression-limiting immune responses in humans and that novel clinical vector production strategies to increase CpG methylation should be developed.

Clinical Evidence

A discussion of the role of CpGs in AAV vectors and their contribution to immunotoxicity and loss of transgene expression in hemophilia gene therapy was catalyzed by the report by Chapin and colleagues (J. Chapin et al., 2018, 14^th Workshop on Novel Technologies and Gene Therapies for Hemophilia, conference), which reported unexpected loss of FIX expression in 7 of 8 patients in their clinical trial using AAV8-FIXsc (Padua) investigational product BAX335. They hypothesized that CpG enrichment resulting from “codon-optimization” of the FIX ORF was the root cause of the CTL formation that eliminated transduced cells. This hypothesis is supported by results from seven other AAV-based gene therapy trials for hemophilia B reporting long-term follow-up, as summarized in Table 1. Among the variables, including serotype, expression cassette configuration, production method, vector genome (vg) and estimated total capsid dose, and the use of immune-suppression, low CpG content is the only parameter that fully correlates with long-term FIX expression. Codon modification was used to remove the 19 CpGs present in wild-type FIX cDNA in all four trials that reported durable FIX expression in all (33 combined) subjects, as well as the absence of or modest CTL responses that were easily controlled by transient immune suppression. In contrast, a different codon modification approach that aimed to increase the translational kinetics of the expression cassette and, in the process, increasing CpGs by approximately 5-fold over wild-type cDNA, was used in the three trials that reported stronger CTL responses that were not well-controlled by immune suppression. In the two studies that published outcomes, loss of FIX expression in all but one of 14 subjects was reported. These data are consistent with unmethylated CpGs in AAV vectors as the primary trigger for efficacy-limiting CTL responses in humans. Higher doses that render more transduced cells targets for capsid-specific CTLs and AAV serotypes that are more efficiently taken up by TLR9-expressing dendritic cells are likely important contributing factors.

Table 1.

AAV Gene Therapy Clinical Trials for Hemophilia B

Sponsor	Serotype/Configurationa	No. of CpG in ORF	Production	Dose (×1012)		Immunology		Outcomes
				(vg/kg)	(∼cp/kg)	ISb	CTLc	Peak FIX	Duration
CHOP, Stanford Avigen	AAV2-FIX/ss	19d (WT)	HEK	2	2	–	++	12% (n = 1)	<3 months
UCL, St Jude	AAV8-FIX/sc	0d	HEK	0.2–2	1–10	+	+	2%–11% (n = 10)	>1 year
Shire (BAX335)	AAV8-FIX Padua/sc	99d	HEK	0.2–3	ND	++	++	4%–45% (n = 8)	<3 months
CHOP	AAV8-FIX19/ss	94e	ND	1–2	ND	++	++e	ND	ND
Pfizer (SPK-9001)	AAVSPK-FIX Padua/ss	0d	HEK	0.5	1.5–2.5	+	+	34% (n = 10)	>1 year
Uniqure (AMT060)	AAV5-FIX/sc	0d	Bac	20	40	+	+	7% (n = 5)	>1 year
Dimension (DTX101)	AAVrh10-FIX/ss	96d	HEK	1.6–5	ND	++	++	3%–8% (n = 6)	<3 months
Uniqure (AMT061)	AAV5-FIX Padua/sc	0d	Bac	20	40	–	+	47% (n = 3)	>1 year

^aGenome configuration: ss, single-stranded genome; sc, self-complementary genome.

^bImmune suppression: –, not used; +, minority of subjects; ++, majority of subjects.

^cCapsid-specific CTLs by IFN-γ ELISPOT: +, minority of subjects; ++, majority of subjects.

^dNathwani, 2019, American Society for Hematology Annual Meeting, Ham Wasserman Lecture

^eHigh and Anguela, 2016, USTPO 20160375110

Codon Optimization

The term “codon optimization” has been used to describe various codon modifications in rAAV vectors, often without clear definition. Given that (1) the number of DNA sequences that can encode a single protein is large, (2) the poor clinical outcomes associated with codon modifications that increase CpG content as shown in Table 1, and (3) the likely exclusion of benefit from future AAV products due to seroconversion⁸ for human subjects that receive vectors that fail to achieve durable therapeutic effect, it is clear that guidelines for codon modification of rAAV vectors for in vivo gene therapy should be developed and shared. The algorithms previously used to increase the efficiency of recombinant human protein expression in heterologous production cells that increase CpG content^9,10 should not be used. Clinical experience for hemophilia B supports the concept that codon modification to remove CpGs is beneficial and should be used for other rAAV investigational products where the route of administration is immunologically responsive. The sharing of CpG content in rAAV expression cassettes by sponsors developing gene therapies for other diseases (e.g., hemophilia A and DMD) would further help define best practices for the field and optimal benefit for human subjects. Since safety and long-term expression are the key attributes for most AAV gene therapies, a strategy of “hasten slowly” for codon modification is prudent—prioritize first and foremost the elimination and avoidance of PAMPs.

Hypomethylation of CpG Dinucleotides in AAV Vectors

Two additional lines of evidence support the idea that hypomethylation is an important product attribute flaw in the current generation of AAV vectors. The first is that efficacy-limiting CTLs and loss of expression were observed in the first liver-directed AAV-FIX clinical trial (CHOP, Avigen)¹¹ in which the FIX cDNA (i.e., with wild-type CpG content) was used, albeit in the absence of immune suppression regimens that were subsequently developed. The second line of evidence stems from direct biochemical analysis of marked CpG hypomethylation in AAV generated using transient transfection¹². Maneuvers used in the clinic to mitigate this AAV vector product immunostimulatory feature—namely reduction of CpGs by codon modification—and immune suppression of subjects until capsid peptides have cleared from transduced tissues have not always prevented CTL generation and loss of transgene expression.

Vector Production Approaches

A novel approach that takes more direct aim at the root cause of hypomethylation in AAV vectors is to increase CpG methylation by development of improved production technologies. While such strategies would vary depending on production cell line type (e.g., mammalian or insect) and mode of introduction (e.g., transfection or infection) of the genes required for vector generation, the main objective would be to provide enough targeted methyl transferase during production of input vector DNA (e.g., plasmid) and during vector genome replication in production cells to achieve ∼75% CpG methylation, comparable to the level in human DNA. To meet this goal, accurate understanding of the provenance of AAV expression cassette DNA is required. For example, in the generation of AAV2 vectors by transient transfection with plasmid DNA in HEK293 cells, a fraction of packaged genomes was reported to be of plasmid DNA origin, i.e., rescued directly from vector plasmid.¹³ A model for AAV expression cassette rescue and packaging from plasmid DNA showing pathways for genomes excised from plasmid (left) and canonical¹⁴ replication-derived AAV genomes (right) is shown in Figure 1. The prokaryotic DNA genomes are expected to contain only unmethylated CpGs, while the genomes replicated in the mammalian cell would contain some methylated CpGs, though few because of insufficient methyl transferase available during the rapid kinetics of replication and packaging.¹² Calculation shows that, for typical transient transfection methods, vector plasmid copy number input is comparable to the number of AAV packaged genomes produced, supporting that both pathways shown in Figure 1 should be considered in the development of strategies to increase CpG methylation. In reference to Figure 1, CpG methylation of the vector plasmid (a) and during AAV genome replication (b) represent strategies for methyl transferase supplementation. Similar considerations can be used for other AAV vector production systems, for example, to address CpG methylation limitations in Holometabola.¹⁵ The frequency of unmethylated CpG motifs in the genomes of AAV vectors prepared for human gene therapy is likely a critical quality attribute for vectors intended for in vivo administration, and specifications to ensure adequate innate immune histocompatibility should be established.

Figure 1.

Model for AAV Vector Genome Rescue, Replication, and Packaging

The AAV genome within the vector plasmid DNA is rescued at Holliday junctions (diagonal arrows). Resolution at the terminal resolution sites (trs), and ITR repair leads to paired complementary strands. These can be packaged directly (left) to give AAV particles containing predominantly prokaryotic DNA vector genome with no CpG methylation or provide a replication template leading to a eukaryotic DNA genome that can then be packaged (right) to give an AAV particle containing limited CpG methylation. ‘a’ and ‘b’ are proposed sites for introduction of methyl transferase.