Application Note
SeQure DX™ Enabled Pre-Clinical Risk Assessment of PCSK9 Guide RNAs
Background
Ischemic heart disease, caused by atherosclerosis, remains the leading cause of death worldwide. High levels of LDL cholesterol in the bloodstream are a contributory factor in the development of atherosclerosis. Long-term reduction of cholesterol levels may reduce the development of atherosclerotic plaques and the risk of cardiovascular events such as heart attacks and strokes.
The LDL receptor is responsible for removing LDL cholesterol from the bloodstream. Proprotein convertase subtilisin kexin type 9 (PCSK9) promotes the degradation of LDL receptors on liver cells, making PCSK9 a target for cholesterol-lowering treatment. Two cholesterol-lowering, anti-PCSK9 monoclonal antibodies are available, requiring repeated dosing every two to four weeks.1
Gene editing is a promising alternative to existing therapies, offering the potential of one-off delivery for durable gene knockdown. Musunuru et al. recently reported a potential in vivo gene therapy strategy using base editing to reduce PCSK9 expression by disrupting splice donor and acceptor sites in the PCSK9 gene.2 The authors identified three guide RNAs (gRNAs) with relatively high in vitro editing efficiency in primary human hepatocytes: PCSK9-1, PCSK9-3 and PCSK9-4. This possible therapeutic approach was successful in cynomolgus monkeys, with stable (at least eight months) reduction in circulating PCSK9 (90%) and LDL (60%) levels after a single treatment with lipid nanoparticles formulated with adenine base editor mRNA (ABE8.8) and PCSK9-1 gRNA.
Therapeutic developers must submit an IND filing (USA), CTA (EU) or other regional regulatory application to progress to human clinical trials. The US FDA’s guidance to developers of gene- and genome-editing therapies includes recommendations on the non-clinical safety assessment that should be included in an IND application.3 These safety studies can include identifying, enumerating and characterizing on- and off-target editing events. Additional recommendations include using multiple (orthogonal) high-sensitivity methods to detect low-frequency events, considering genomic heterogeneity, and assessing the functional consequences of on- and off-target editing.
Aim
We used SeQure DX Guide Profiler™ to provide an initial ranking of eight of the gRNAs reported by Musunuru et al.2 Next, we used Guide Select™ for detailed profiling of the on- and off-target risk of the three gRNAs that guided the highest reported editing activity in primary human hepatocytes.2 Finally, ONE-seq™ provided the highest sensitivity in silico and in vitro characterization of the best (lowest risk) guide RNA and nominated off-target sites for later in vivo confirmation.
Method overview
Figure 1 provides a Guide Profiler workflow overview. Guide Profiler is an in silico assay that screens candidate gRNAs (up to six differences) to identify potential off-targets, assess their biological impact, and analyze on-target variation to provide a gRNA risk profile for early informed gRNA selection. Off-target loci are identified by sequence similarity-based in silico searching of the hg38 genome and 3,502 haplotype-phased human genomes from the 1000 Genomes Project and Human Genome Diversity Project.
Workflow overview of Guide Profiler
Figure 1: (1) Computational (in silico) search for sequence matches, (2) on-target variant analysis, (3) off-target library size benchmarking, (4) analysis of the biological impact of cutting at putative off-target loci and (5) summary and report of guide RNA risk profile
Figure 2 provides an overview of the workflow for Guide Select and ONE-seq. Guide Select combines in silico analysis of a limited search space with a biochemical assay for in vitro, off-target cutting with multiplexed gRNAs, enabling rapid variant-aware risk assessment for program prioritization and gRNA selection.
ONE-seq nomination assays evaluate individual candidate gRNAs with a genome-wide, in silico, variant-aware search for putative off-target enumeration and a comprehensive biochemical assessment of in vitro cutting for high-sensitivity detection of low-frequency off-target events.4
Workflow overview of Guide Select and ONE-seq
Figure 2: (1) In silico search for sequence matches with ≤ 4 (Guide Select) or ≤ 6 (ONE-seq) differences, (2) target oligos synthesized on DNA chip (50,000 – 240,000), (3) uniform ONE-seq library, (4) in vitro editing with multiplexed (Guide Select) or single (ONE-seq) gRNAs, (5) deep Next-Generation Sequencing (NGS) and (6) analysis of biological impacts and off-target nomination
Variant-aware risk profiling is achieved by screening against the human reference genome, hg38, and 3,502 haplotype-phased genomes from 1000 genome and HGDP initiatives with up to six sequence differences. Figure 3 represents the geographical origin of the 3,502 haplotype-phased genomes. Knowledge of the geographical origin of genomic sequences is critical to evaluating variations that might impact on- and off-target editing in "intent-to-treat" populations.
Variant-aware gRNA profiling screens against genomic sequences from all seven superpopulations
Figure 3: The geographical origin of the haplotype-phased genomes used in SeQure DX in silico screening. The circle size is proportional to the number of individual genomes in the superpopulation indicated.
Results
Guide Profiler enabled rapid risk evaluation and gRNA ranking by Risk Profiler
Eight guide RNAs targeting the PCSK9 gene (Table 1) were analyzed in silico using Guide Profiler to identify on-target population variants and potential off-target editing sites.
Genetic variation at on-target sites can negatively impact intended gene editing, and on-target variation is a critical factor in determining gRNA risk scores. Early identification and evaluation of on-target variants is crucial for successful therapeutic development. Guide Profiler analysis revealed genomic variants that might impact on-target editing guided by five of the gRNAs (Table 2).
Between 25% and 35% of putative off-target edits were identified only by screening variant genomes (Figure 4), demonstrating the importance of considering genomic diversity in gene editing therapy risk assessment. The most off-target sites were identified for PCSK9-7, and the fewest were identified for PCSK9-3.
While the number of putative off-target edits contributes to the editing risk profile, the location of off-target sites must also be considered. Guide Profiler provided a rapid analysis of the potential biological impact of off-target editing for the eight guides at two risk thresholds (≤ 3 differences and ≤ 6 differences). The guides were scored and ranked by relative risk using the criteria indicated (Table 2). PCSK9-1 is ranked as the best gRNA despite having many candidate off-targets (≤ 6 differences). Musunuru et al. reported the in vitro editing efficiency using the gRNAs tested.2 The three leading gRNAs, PCSK9-1, PCSK9-3 and PCSK9-4, have low to medium risk ranked by Risk Profiler and were advanced for further assessment.
| Guide ID | Guide Sequence |
|---|---|
| PCSK9-1 | CCCGCACCTTGGCGCAGCGGtgg |
| PCSK9-2 | GGTGGCTCACCAGCTCCAGCagg |
| PCSK9-3 | GCTTACCTGTCTGTGGAAGCggg |
| PCSK9-4 | TGCTTACCTGTCTGTGGAAGcgg |
| PCSK9-5 | TTGGAAAGACGGAGGCAGCCtgg |
| PCSK9-6 | GAAAGACGGAGGCAGCCTGGtgg |
| PCSK9-7 | TCCCAGGCCTGGAGTTTATTcgg |
| PCSK9-8 | AGCACCTACCTCGGGAGCTGagg |
Table 1: PCSK9-targeting gRNA sequences
Figure 4: Guide Profiler enumerates putative off-target loci. The count of potential off-target sites found for each gRNA in the reference genome hg38 (blue) and any of the 3,502 haplotype-phased variant genomes (green) is shown
Table 2: Guide Profiler risk preview for eight gRNAs targeting PCSK9. Guide RNAs are ranked according to their risk profile. The total of putative off-target sites with up to three and up to six differences is shown. Guide Profiler provides further information regarding the risk of potential off-target sites (≤ 3 differences). Risk Profiler assigns a risk score for each guide, indicating their relative off-target risk. Cells are colored according to the risk as indicated.
Guide Select enabled rapid variant-aware in silico and biochemical evaluation for guide ranking
In silico screening of the three leading gRNAs by Guide Select identified a combined total of 6,642 putative off-target sequences with up to four differences. These oligos were synthesized on a DNA chip, and an in vitro cutting assay with pooled gRNAs and Cas9 was performed in triplicate. Cleaved sites were classified using the Edits of Relative Concern (ERC) framework (Table 3). The ERC framework simplifies the interpretation of the potential biological impact of cleavage by classifying cuts according to multiple criteria, described in Table 3.
Table 3: The biological risk of putative off-targets is ranked with reference to the ERC framework. Potential off-target sites are bucketed into tiers according to the criteria noted.
Cross-referencing with the Cancer Gene Census (CGC) and ClinGen databases (Table 4) provides further evaluation of nominated off-target edits. PCSK9 has the lowest risk within the ERC framework and the fewest off-target sites identified with reference to the CGC and ClinGen databases.
| Guide | CGC Tier 1 | CGC Tier 2 | ClinGen database | Cancer gene list |
|---|---|---|---|---|
| PCSK9-1 | 1 | 0 | 4 | COL2A1 |
| PCSK9-3 | 2 | 0 | 10 | CAMTA1 |
| PCSK9-4 | 2 | 0 | 13 | APC, MLLT3 |
Table 4: Putative off-target sites are classified according to their potential clinical significance. Candidate off-target loci in Tiers 1 and 2 (according to the ERC framework) linked to genes annotated in the CGC and ClinGen database are noted.
The proximity of off-target loci to the on-target site can profoundly impact editing safety. High off-target editing frequencies at sites close to the on-target locus increase the probability of intrachromosomal rearrangements. Figure 5 shows the location of off-target loci (≤ 4 differences) from Tier 1 and Tier 2a (high cleavage frequency) on the same chromosome as the on-target locus. PCSK9-1 has the lowest risk from off-target proximity. Guide Select identified PCSK9-1 as the gRNA with the most favorable on-target profile and off-target risk. PCSK9-1 progressed to ONEseq screening for an in-depth, high-sensitivity analysis profile.
Figure 5: Putative off-target on the same chromosome as the on-target. Candidate off-target loci from Tier 1 and Tier 2a (high cleavage frequency) are shown. Due to scaling, arrows indicating off-target sites very close in location may not be individually discernible.
ONE-seq nominated off-target sites for the lead gRNA, PCSK9-1
A high-stringency (≤ 6 differences) in silico search identified 486,844 putative off-target sequences for PCSK9-1. These sequences were synthesized and screened at three different ratios in triplicate for in vitro cleavage by Cas9 and PCSK9-1. As described earlier, cleaved off-target sites were classified with reference to the ERC framework. Figure 6 is a graphical representation of these sites.
Most of the in vitro cleaved Tier 1 and Tier 2 off-target sequences are present only in variant genomes (Figure 7) and would be missed in a search limited to the human reference genome, highlighting the importance of variant-aware screening.
ONE-seq nominated 13 high-risk (Tier 1 and 2) off-target sites and identified 30 Tier 3 sites for further consideration.
Figure 6: Graphic representation of ONE-seq off-target cleavage by tier for PCSK9-1. Higher ONE-seq scores indicate higher cleavage frequencies in the ONE-seq assay. Annotation concern scores summarize genomic, functional and known disease associations of loci.
Figure 7: The contribution of variant genomes to the list of nominated off-target loci. Off-target sites cleaved in the in vitro cutting assay using PCSK9-1 gRNA were classified in ERC tiers. The number of Tier 1 and 2 cuts in the reference genome (blue) and variant genomes (green) are shown.
Summary
Here, we presented a complete workflow from Guide Profiler to ONE-seq for guide prioritization, selection and off-target nomination for a series of therapeutically relevant guide RNAs targeting the PCSK9 gene.
Guide Profiler’s in silico ranking provides rapid insight into potential off-target risk for candidate guide RNAs, enabling developers to avoid investment in guides with problematic on-target (e.g., variation in intent-to-treat populations) and off-target (e.g., high-risk sites in cancer genes) profiles and to prioritize lower-risk guides for further analysis with Guide Select or ONE-seq.
Guide Select and ONE-seq leverage computational tools and biochemical assays to nominate candidate off-target sites across thousands of genomes. Guide Select provides rapid, cost-effective multiplexed analysis that can streamline the selection process for multiple candidate guides. ONE-seq is the highest sensitivity assay used for off-target nomination for the gRNA(s) selected for therapeutic editing.
Despite having a relatively high number of putative off-target loci, PCSK9-1 gRNA has the best Risk Profiler rank, which was confirmed by Guide Select screening. This highlights the importance of factors beyond the enumeration of off-target sites for in silico risk determination. Less than one-third of the nominated off-target loci for PCSK9-1 were identified in the human reference genome, confirming the need for therapeutic developers to consider the genetic ancestry of donors, intent-to-treat populations and clinical trial participants.
References
- Mohamed F, Mansfield B, Raal FJ. Targeting PCSK9 and Beyond for the Management of Low-Density Lipoprotein Cholesterol. Journal of Clinical Medicine. 2023; 12(15):5082. https://doi.org/10.3390/jcm12155082
- Musunuru, K, Chadwick, AC, Mizoguchi, T. et al. In vivo CRISPR base editing of PCSK9 durably lowers cholesterol in primates. Nature 593, 429–434 (2021). https://doi.org/10.1038/s41586-021-03534-yUS
- Human Gene Therapy Products Incorporating Human Genome Editing; Guidance for Industry, January 2024, https://www.fda.gov/media/156894/download
- Petri K, Kim DY, Sasaki KE. et al. Global-scale CRISPR gene editor specificity profiling by ONE-seq identifies population specific, variant off-target effects. bioRxiv 2021.04.05.438458. doi: https://doi.org/10.1101/2021.04.05.43845
