Quantification of CpG Motifs in rAAV Genomes: Avoiding the Toll
2020; Elsevier BV; Volume: 28; Issue: 8 Linguagem: Inglês
10.1016/j.ymthe.2020.07.006
ISSN1525-0024
Autores Tópico(s)Viral gastroenteritis research and epidemiology
ResumoGene therapy using recombinant adeno-associated virus (rAAV) vectors has demonstrated definitive benefits for genetic diseases and has enormous future potential, representing an important part of the next paradigm of human therapeutics. However, many clinical trials with rAAV have reported varying degrees of immunotoxicity, potentially including that associated with the recently reported deaths of two subjects in a clinical trial for X-linked myotubular myopathy. Analogous to the identification and removal of immunogenic features of early-era monoclonal antibodies, thereby "humanizing" those products, while recognizing the immutable viral nature of the vector capsid, a similar strategy of humanizing addressable features of AAV vectors during their design is an opportunity to accelerate successful clinical product development. The synergistic nature of the multiple pathways that comprise human innate and adaptive immune responses combined with the consequences of failure to adequately control them after AAV-mediated gene delivery, including immunotoxicity, potential loss of transgene expression, and AAV antibody seroconversion preventing re-administration, support the need to identify and remove "microbial legacy" immunostimulatory features such as pathogen-associated molecular patterns (PAMPs).1Barton G.M. Kagan J.C. A cell biological view of Toll-like receptor function: regulation through compartmentalization.Nat. Rev. Immunol. 2009; 9: 535-542Crossref PubMed Scopus (522) Google Scholar This commentary focuses on one such PAMP, the unmethylated CpG motifs (PAMP CpG) commonly found in AAV vectors due to hypomethylation of vector genomes during their production2Tóth R. Mészáros I. Hüser D. Forró B. Marton S. Olasz F. Bányai K. Heilbronn R. Zádori Z. Methylation status of the adeno-associated virus type 2 (AA2).Viruses. 2019; 11: 38Crossref Scopus (10) Google Scholar and presence of expression cassette elements of microbial origin that are rich in CpGs. Herein we describe approaches to quantify the Toll-like receptor 9 (TLR9) innate immune pathway activation risk for DNA sequences of interest, e.g., rAAV expression cassettes under consideration as investigational products based on CpG/motif content and methylation, providing a tool to assess and guide reduction of TLR9-associated immunogenicity. PAMP CpG binds and dimerizes TLR9 molecules3Ohto U. Shibata T. Tanji H. Ishida H. Krayukhina E. Uchiyama S. Miyake K. Shimizu T. Structural basis of CpG and inhibitory DNA recognition by Toll-like receptor 9.Nature. 2015; 520: 702-705Crossref PubMed Scopus (237) Google Scholar expressed in plasmacytoid dendritic cells (pDCs), leading via MyD88 to cellular immune responses.4Hartmann G. Weiner G.J. Krieg A.M. CpG DNA: a potent signal for growth, activation, and maturation of human dendritic cells.Proc. Natl. Acad. Sci. USA. 1999; 96: 9305-9310Crossref PubMed Scopus (559) Google Scholar An established and growing body of non-clinical5Zhu J. Huang X. Yang Y. The TLR9-MyD88 pathway is critical for adaptive immune responses to adeno-associated virus gene therapy vectors in mice.J. Clin. Invest. 2009; 119: 2388-2398Crossref PubMed Scopus (207) Google Scholar, 6Faust S.M. Bell P. Cutler B.J. Ashley S.N. Zhu Y. Rabinowitz J.E. Wilson J.M. CpG-depleted adeno-associated virus vectors evade immune detection.J. Clin. Invest. 2013; 123: 2994-3001Crossref PubMed Scopus (145) Google Scholar, 7Shirley J.L. Keeler G.D. Sherman A. Zolotukhin I. Markusic D.M. Hoffman B.E. Morel L.M. Wallet M.A. Terhorst C. Herzog R.W. Type 1 IFN sensing by cDCs and CD4+ T cell help are both requisite for cross-priming of AAV capsid-specific CD8+ cells.Mol. Ther. 2020; 28: 758-770Abstract Full Text Full Text PDF PubMed Scopus (34) Google Scholar, 8Xiang Z. Kurupati R.K. Li Y. Kuranda K. Zhou X. Mingozzi F. High K.A. Ertl H.C.J. The effect of CpG sequences on capsid-specific CD8+ cell responses to AAV vector gene therapy.Mol. Ther. 2020; 28: 771-783Abstract Full Text Full Text PDF PubMed Scopus (25) Google Scholar and clinical9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar evidence unsurprisingly supports the model that a viral capsid containing a genome with PAMP CpG stimulates innate and adaptive immune pathways, leading to the formation of capsid-specific cytotoxic T lymphoctes (CTLs). Figure 1 illustrates how high PAMP CpG levels in the expression cassette of an AAV vector lead to CTLs that eliminate transduced hepatocytes (Figure 1A), while a vector genome with a sub-threshold PAMP CpG level does not activate the TLR9-MyD88 pathway and spares transduced cells, leading to durable transgene expression (Figure 1B). Factors not shown in Figure 1 certainly contribute to these pathways, e.g., higher vector doses would be expected to increase the severity of the hepato-immunotoxicity shown in Figure 1A. Transient immune suppression is frequently used and partially effective in managing CTL responses in rAAV clinical studies but adds complexity and risk.10Samelson-Jones B.J. Finn J.D. Favaro P. Wright J.F. Arruda V.R. Timing of intensive immunosuppression impacts risk of transgene antibodies after AAV gene therapy in nonhuman primates.Mol. Ther. Methods Clin. Dev. 2020; 17: 1129-1138Abstract Full Text Full Text PDF PubMed Scopus (24) Google Scholar Avoidance of TLR9 activation by reducing PAMP CpG in AAV vector genomes during investigational product design is a promising approach to directly address the root cause. With recognition of the innate immunogenic risk of PAMP CpG in rAAV, vector design strategies, including codon modification of open reading frames and sequence changes in non-coding elements to reduce CpG dinucleotides, are becoming best practices. Complete CpG removal from an expression cassette is possible but has the potential to cause transprotein misfolding due to non-wild-type translational kinetics11Mauro V.P. Chappell S.A. A critical analysis of codon optimization in human therapeutics.Trends Mol. Med. 2014; 20: 604-613Abstract Full Text Full Text PDF PubMed Scopus (150) Google Scholar and adversely affect the performance of expression cassette elements, such as inverted terminal repeats (ITRs). Understanding the PAMP CpG threshold for human TLR9 activation, coupled with a method to quantify the TLR9 activation potential ("KTLR9") in candidate expression cassettes, would be helpful to guide clinical vector design. Three risk factor (RF) equations were developed and used to estimate KTLR9 in 15 relevant DNA test sequences, with the results shown in Table 1. The equations progressively incorporate three attributes of DNA sequences known to activate the TLR9-MyD88 pathway. RF1 considers just the fraction (f) of total CpG dinucleotides (CpGT) divided by the nucleotide length (nt) for each DNA sequence, which ranged from 0.965% in the human genome (suppressed compared to 6.25% for random nucleotide utilization) to 9.42% in the bacterium K. peumoniae genome in the 15 sequences analyzed. RF2 multiplies RF1 by the estimated fraction of CpG dinucleotides that are unmethylated (CpGMeneg / CpGT) in each type of DNA test sequence:12Tost J. DNA methylation: an introduction to the biology and the disease-associated changes of a promising biomarker.Mol. Biotechnol. 2010; 44: 71-81Crossref PubMed Scopus (170) Google Scholar ∼0.25 for human DNA (RF2 = 0.25RF1), 1.0 for bacterial DNA (RF2 = RF1), and ∼0.95 for the viruses and rAAV vectors2Tóth R. Mészáros I. Hüser D. Forró B. Marton S. Olasz F. Bányai K. Heilbronn R. Zádori Z. Methylation status of the adeno-associated virus type 2 (AA2).Viruses. 2019; 11: 38Crossref Scopus (10) Google Scholar listed in Table 1 (RF2 = 0.95 RF1). RF3 modifies RF2 to incorporate known immune-stimulatory (S4) and -inhibitory (I4) tetranucleotide CpG motifs reported by vaccine research aiming to enhance cellular immune responses using oligonucleotide adjuvants.3Ohto U. Shibata T. Tanji H. Ishida H. Krayukhina E. Uchiyama S. Miyake K. Shimizu T. Structural basis of CpG and inhibitory DNA recognition by Toll-like receptor 9.Nature. 2015; 520: 702-705Crossref PubMed Scopus (237) Google Scholar,13Bode C. Zhao G. Steinhagen F. Kinjo T. Klinman D.M. CpG DNA as a vaccine adjuvant.Expert Rev. Vaccines. 2011; 10: 499-511Crossref PubMed Scopus (557) Google Scholar,14Pohar J. Yamamoto C. Fukui R. Cajnko M.M. Miyake K. Jerala R. Benčina M. Selectivity of human TLR9 for double CpG motifs and implications for the recognition of genomic DNA.J. Immunol. 2017; 198: 2093-2104Crossref PubMed Scopus (34) Google Scholar The S4 and I4 CpG motifs were enumerated and summed for each test DNA sequence. The motif sequences selected and their TLR9 activation "weights" used for the RF3 equation in Table 2 are preliminary and directional. A broader CpG motif selection and more accurate, data-based, motif weighting factors would improve the predictive potential. A similar formula that incorporated immune-stimulatory (S6) and -inhibitory (I6) hexanucleotide CpG motifs gave a comparable range of values and the same relative ranking of the test sequences as obtained using RF3 (not shown). A normalized value for RF3 (NRF3) was calculated by dividing the RF3 value for each DNA test article by that for the human genome (0.191), i.e., the sequence assumed to represent the lowest risk of TLR9 pathway activation. The NRF3 for the complete human genome is, by definition, unity, with values ranging from 0.92 to 2.68 for selected human genes and a CpG-rich portion of chromosome 1, providing an indication of intragenomic variation. In contrast, an average NRF3 value of 20.7 was measured for three bacterial genomes known to be strongly TLR9 activating.15Dalpke A. Frank J. Peter M. Heeg K. Activation of toll-like receptor 9 by DNA from different bacterial species.Infect. Immun. 2006; 74: 940-946Crossref PubMed Scopus (131) Google Scholar Together, the human and bacterial genome data define a NRF3 range from 1 to ∼20 corresponding from negligible (−) to high (+++) values for TLR9 activation potential. The NRF3 values for the genomes of helper viruses used in rAAV production ranged from 13.2 to 28.1, demonstrating the PAMP CpG risk represented by residual helper virus DNA impurities in purified AAV preparations. While it is challenging to obtain complete expression cassette DNA sequences for clinical vectors, the availability of sequences, clinical immunotoxicity, and therapeutic outcomes for the four AAV-FIX vectors9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar listed in Table 1 provide an opportunity to further qualify the NRF3 equation. The better clinical performance of AAVSPK-FIX Padua/ss and AAV8-FIX/sc, including long-term transgene expression and lower incidences of CTLs and immunotoxicity,9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar correspond to lower NRF3 values of 3.09 and 6.80, respectively. The higher NRF3 values of 7.80 and 12.7 calculated for AAV2-FIX/ss and AAV8-FIX19/ss, respectively, correspond to vectors that gave higher immunotoxicity without durable transgene expression.9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar These data support that AAV vectors with lower NRF3 scores approaching a "humanized" value have lower immunotoxicity and better long-term clinical benefit, while those with scores above a threshold value of ∼7 are associated with deleterious immune responses not well-controlled by immune suppression, leading to loss of transgene expression. Use of such quantitative tools to evaluate TLR9 activation potential after their further refinement and validation with genome sequences from other clinical constructs represents an approach to improve AAV vectors by reducing their potential to cause immunotoxicity.Table 1TLR9 Activation Risk Factors for Selected DNA SequencesDNA Test ArticleReferenceRF1RF2RF3NRF3KTLR9HumanComplete genomeNCBI Homo sapiens GRCh380.9650.2410.1911.00−F8 geneNCBI: NG_0114030.9210.2300.1790.94−F9 geneNCBI: NC_0000230.7490.1870.1760.92−Dystrophin geneNCBI: NG_0122320.7970.1990.2001.05−Chr1 CpG-rich fragmentaChromosome 1, nucleotides 1,000,000 to 2,000,000.NCBI: NC_0000013.7040.9260.5112.68−Clinical rAAVAAVSPK-FIX Padua/ssWright9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar1.0270.9760.5903.09−AAV8-FIX/scWright9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar1.7571.6691.2986.80+AAV2-FIX/ssWright9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar2.0371.9361.4907.80++AAV8-FIX19/ssWright9Wright J.F. Codon modification and PAMPs in clinical AAV vectors: the tortoise or the hare?.Mol. Ther. 2020; 28: 701-703Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar3.5303.3542.41812.7+++BacterialEscherichia coliNCBI: CP_0096857.4717.4714.68324.5+++Klebsiella pneumoniaeNCBI: FO_8349069.4219.4214.42123.1+++Staphylococcus aureusNCBI: NC_0077952.5482.5482.75014.4+++Helper VirusesAAV2NCBI: NC-0406715.8475.5554.96826.0+++Adenovirus5NCBI: AC_0000086.7176.3812.52213.2+++Autographa californicaNCBI: NC_0016236.1835.8735.36328.1+++a Chromosome 1, nucleotides 1,000,000 to 2,000,000. Open table in a new tab Table 2CpG Motifs and Equations Used to Calculate Risk FactorsMotif NameCpG Tetranucleotides IncludedKTLR9 WeightingCpGS4Σ ACGT, TCGT, CCGT+2∗CpGExCpGT - CpGS4 - CpGI4+1CpGI4Σ GCGG, CCGC, GCGC−1RF1=f[CpGT/ nt]×100%RF2=f[CpGT/ nt]×f[CpGMeneg/ CpGT]×100%RF3=f[*CpGEx+2CpGS4–CpGI4/nt]×f [CpGMeneg/ CpGT]×100% =f[CpGT+CpGS4–2CpGI4/nt]×f[CpGMeneg/CpGT]×100%NRF3=RF3(testarticle)/ RF3(humangenome) Open table in a new tab RF1=f[CpGT/ nt]×100%RF2=f[CpGT/ nt]×f[CpGMeneg/ CpGT]×100%RF3=f[*CpGEx+2CpGS4–CpGI4/nt]×f [CpGMeneg/ CpGT]×100% =f[CpGT+CpGS4–2CpGI4/nt]×f[CpGMeneg/CpGT]×100%NRF3=RF3(testarticle)/ RF3(humangenome) The author thanks Thomas Chalberg for enumeration of CpG, S4, S6, I4, and I6 motifs in the human genome (hg38 assembly) using a Python-based scripting program, Cecile Martin for science graphic artist support for Figure 1, and Maria-Grazia Roncarolo, Thomas Chalberg, and Bradley Hamilton for review of the manuscript and helpful comments.
Referência(s)