Preclinical safety and biodistribution of CRISPR targeting SIV in non-human primates

Development of safe and effective viral specific gRNAs

CRISPR gRNA designer tools were used to identify gRNAs that target LTR and Gag with minimal chance of unintended effects for the previously described HIV-directed guides [6, 10]. Previous research demonstrated efficacy in targeting HIV in culture and using an infected humanized mouse model with guide RNAs targeting LTR-1 and GagD [6]. For this work in rhesus macaques, gRNAs targeting SIV were chosen to represent the same regions as the HIV guides for these biodistribution and safety studies.

The bioinformatics tool COTANA (CRISPR Off-Target Nomination & Analysis) exhaustively searched the rhesus macaque reference genome (NCBI genome assembly Mmul_10) for the most similar chromosomal sequences to the target site. COTANA output each nominated site with the specified number of differences between the gRNA and DNA sequence (Table 1). The output for the exhaustive search, including mismatches and possible bulges, contains few nominated sites in the rhesus macaque genome and very few with less than 4 differences (Supplementary materials and methods). The list of output sites also has few with low penalty match scores, which generally indicate a more possible chance of locating sequence-verified off-target sites. Penalty match scores were calculated based on the increasing number, type, and locations of the mismatches and bulges between the gRNA and chromosomal sequence (Fig. 1). When using gene editing to correct disease-causing mutations, one is generally limited to the closest few target sites. When targeting viral sequences for inactivation, there is much greater flexibility as multiple targeting strategies and series of target sites can be bioinformatically scanned to choose conserved viral target sites that have a much lower number and similarity of chromosomal locations [11,12,13]. Those gRNAs output many fewer sites than gRNAs targeting coding sequences, as seen in previous publications and when studying EBT-101 [11, 12].

All sites in the bioinformatic output from the rhesus macaque genome searches using SIV guide LTR allowing three differences and Gag allowing four differences are listed. The guide (colored) and PAM sequence (gray) are listed on top, the chromosomal sequence on the bottom with the locations and identity of mismatches shown in boxes. The two types of bulges are also displayed. Gaps in the chromosomal DNA relative the guide RNA are shown in red boxes with dashes. Additional nucleotides in the chromosomal DNA relative to the guide are shown in red triangles. The number of mismatches and bulges and match score for each chromosomal locations are listed in the right column. No loci were identified without mismatches in the proximal 8 nucleotides where differences are less tolerated, outlined in red. Moreover, none of the nominated off-target loci have a penalty match score less than 2 (right column). Bp base pair, chr chromosome, LTR long terminal repeats, MM mismatch, PAM protospacer-adjacent motif, SIV simian immunodeficiency virus.

Assessment of EBT-001 in NHP model of HIV

In two matching studies, SIV-infected and antiretroviral-treated NHP were used as a large animal model to test AAV9-delivered, EBT-001’s ability to remove integrated SIV DNA. In study one, a total of 10 male rhesus macaques (Macaca mulatta) were divided into 3 groups: Group 1 had no EBT-001 treatment, Group 2 with 1.4 × 10¹² genome copies (GC)/kg of EBT-001, and Group 3 with 1.4 × 10¹³ GC/kg of EBT-001 (Fig. 2A). In study two, a higher dose (1.4 × 10¹⁴ GC/kg) of EBT-001 was given to 2 animals in Group 4 (Fig. 2B). Both studies examined NHPs infected with SIVmac239 (200 TCID₅₀) via the i.v. route and treated with a triple ART regimen (tenofovir (TDF), emtricitabine (FTC), and DTG) via the s.q. route (Fig. 2C). This regimen was selected for its similarity to how HIV-infected humans are treated to control HIV infection. ART began 28 days post-infection and all 12 animals remained on ART for the entire course of the study. Once viral levels were reduced to minimum levels in the blood, animals were maintained on only ART (Group 1) or treated with one of 3 doses of EBT-001 delivered by a single IV infusion (Fig. 3C, Supplementary material and methods, Table S2). All monkeys received full doses of EBT-001. One monkey (CK49) in Group 3 experienced shortness of breath and hypoxia after administration of anesthesia that worsened after infusion of EBT-001. The infusion of EBT-001 was paused while the monkey was treated for acute pulmonary edema and recovered. The remaining dose of EBT-001 was delivered several days later without any complications. This was an adverse reaction due to the anesthetic agent, Dexdomitor because the re-challenge of EBT-001 was well-tolerated and the symptoms were consistent with adverse reactions to Dexdomitor. One animal in Group 2 developed a skin rash after EBT-001 infusion. Plasma viral RNA was measured throughout the course of infection and after EBT-001 (Fig. 2D, E).

Initial study design n = 10 (A), follow-up study design n = 2 (B), study chart with groups 1-4 identified (C), Individual plasma viral loads (log10 copies/mL of plasma) (D) and mean and SEM of plasma viral loads of 4 Groups (E). Black is untreated Group 1; red is Group 2, 10¹² GC/kg; light blue is Group 3, 10¹³ GC/kg; dark blue is Group 4, 10¹⁴ GC/kg. Mean and SEM are shown for the group at days post infection. ART anti-retroviral therapy, CHOP Children’s Hospital of Philadelphia, GC genome copies, IV intravenous, SIV simian immunodeficiency virus, sq subcutaneous, TCID₅₀ median tissue culture infectious dose, UNC University of North Carolina.

EBT-001 vector DNA biodistribution as log10 copies of EBT-001 DNA per µg of monkey gDNA in major tissue reservoirs of HIV (A) and other tissues, including various brain regions (B). Symbols represent individual animals and mean and SEM are shown. gDNA, genomic DNA; SEM, standard error of the mean. Untreated 3 mos- untreated animals did not receive EBT-001 but where necropsied at the same time as the EBT-001 animals treated for 3 months. ND not determined.

Plasma samples were assayed for neutralizing antibodies to AAV9 prior to the study and at necropsy (Table S4). EBT-001 treatment was associated with increases in AAV9 neutralizing antibodies 3 or 6 months after treatment relative to pre-dose values. The pre-dose titers (serum reciprocal dilution values) were low (<5 and 80), except CF63, which had a pre-dose titer of 320. The values at necropsy ranged from 80-5120 in EBT-001 treated animals and were highest in the monkeys sacrificed at 6 months post-EBT-001. Of note, CF63’s titer did not change after EBT-001 treatment even though it was the highest at baseline.

All monkeys survived until the scheduled euthanasia (3 or 6 months post EBT-001) when a full necropsy was performed on all animals (Supplementary material and methods). Overall, the gross necropsy observations were within normal range, except for a single animal (1T6, Group 1) from the ART-only treated group, who was dehydrated and had a thin body condition (2/5 body composition score). All findings were consistent with SIV infection (Table S5). No EBT-001-related gross or microscopic pathology findings were noted in a comprehensive histopathological evaluation of major organs performed by Charles River Laboratories, LLC. General microscopic findings attributed to chronic SIV infection were observed within all treatment groups and included minimal to mild lymphoid hyperplasia of the germinal centers and mantle zone of the mesenteric lymph node, lymphoid hyperplasia of the secondary lymphoid follicles in the white pulp of the spleen, and mononuclear cell infiltration of the heart, kidney, and lung. Additionally, there was lymphoid hyperplasia of the periarteriolar lymphoid sheaths of the spleen in animal BM79 (Group 4). There were no apparent differences in the frequency or severity of these findings in monkeys treated with ART only or in monkeys treated with ART in combination with EBT-001. In a separate neuropathological evaluation performed by Experimental Pathology Laboratories, Inc., findings in all SIV-infected rhesus macaques were few and consisted primarily of focal minimal findings of microgliosis and less commonly astrocytosis and/or sparse perivascular mononuclear cell infiltrates. Overall, the microscopic findings were consistent with chronic SIV infection and were observed with similar frequency in control SIV-infected animals examined at the same timepoints from another EBT-001 study. The findings were therefore not attributed to treatment with EBT-001.

Detection of vector DNA in tissues

To assess the biodistribution of the vector in tissues at necropsy, a TaqMan-based qPCR assay was used to quantitate EBT-001 (vector) DNA (Charles River) (Fig. 3A, B). All DNA biodistribution and vector shedding samples collected from the untreated animals (Group 1) were negative, demonstrating the contamination was well controlled from animal dosing, necropsy, DNA extraction, and qPCR analysis in this study. Tissues that are considered major HIV/SIV reservoirs, including spleen, lymph nodes, colon, bone marrow compartment, and blood were examined for vector biodistribution (Fig. 3A). Levels of vector DNA in spleen ranged from 4.82–6.44 (log10 copies of EBT-001 DNA/ug of monkey DNA). For the mesenteric lymph nodes, the value ranged from 4.01–6.42 and for colon 2.52–4.24 (log10 copies of EBT-001 DNA/ug of monkey DNA). In bone marrow, vector DNA was undetectable in Group 2, 1.75–3.55 (log10 copies of EBT-001 DNA/ug of monkey DNA) in Group 3, and 2 logs higher in Group 4 with average 5.5 (log10 copies of EBT-001 DNA/ug of monkey DNA). In blood, where cells turnover rapidly, vector DNA was only found in the 2 animals from Group 3 that were necropsied at 3 months post-EBT-001 and was absent in animals in the same group that were necropsied at 6 months and in the higher dosed animal also sacrificed at 6 months (Group 4) (Fig. 3A). However, when biodistribution was examined serially in blood starting at 2 weeks post-EBT-001, vector DNA was detected in Groups 2 and 3 and decreased over time (Figure S5). The highest concentration of EBT-001 DNA was detected in the liver samples from Groups 3 and 4 (Fig. 3B). For the brain and other tissues, Group 2 was low or undetectable and distribution was increased in Group 3 but interestingly, levels in Group 4 were in the range of 4 (log10 copies of EBT-001 DNA/ug of monkey DNA). No evidence of vector shedding in stool or urine was observed (not shown). The qPCR results confirm that EBT-001 was distributed widely to blood and all major organs and tissues tested in animals dosed with the vector. Biodistribution of EBT-001 (DNA levels) was dose dependent, and the vector levels in each tissue and blood decreased over time, although persisting 6 months after injection in many tissues.

Editing of proviral DNA fragments from blood cells and various solid organs

SIV excision activity was assayed across treatment groups and tissues. The 2 gRNAs in EBT-001 target 3 locations in SIV and therefore excise 3 large intervening integrated proviral SIV segments. The LTR gRNA cuts both the 5′ and 3′LTR. As this guide is co-delivered along with the Gag gRNA, there is the possible removal of the intervening integrated proviral SIV DNA sequence between the 5′LTR to Gag, Gag to the 3′LTR and between the 5′LTR and 3′LTR. Two assays were used to demonstrate 5′LTR to Gag (5G) and Gag to 3′LTR (G3) excision in tissues and blood (see Supplementary material and methods and Supplementary Table 3). Evidence of excision was detected by the observation of nested PCR products of distinct DNA fragments of 268 or 171 bp resulting from the removal of intervening DNA sequences between 5′LTR to Gag (5G) or Gag to 3′LTR (G3), respectively (Fig. 4A, C). The 5 G assay also amplified the full-length SIV at an amplicon size of 1282 bp. Due to extensive homology between the 5′ and 3′LTR sequences and the lack of standardized flanking sequences due to the random chromosomal integration of SIV proviral DNA, it was not possible to measure the third excision product which removes the entire 5′LTR to 3′LTR portion. Both PCR-based excision assays (5G and G3) were completed (in duplicate by 2 operators) for blood and small sections of tissues and organs, including: bone marrow, brain (brain stem, cerebellum, frontal, occipital, parietal, prefrontal, temporal cortices), colon, duodenum, heart (left ventricle, left atrium, right ventricle, right atrium), kidneys, liver, lungs, mesenteric lymph nodes, spleen, and testes (Fig. 4, Tables S6, and S7). The 5G and G3 data are summarized in Table S6 for tissues and Table S7 for blood. In Fig. 4B for tissues and Fig. 4D for serial blood draws, positive excision, defined by the detection of the nested PCR product of distinct DNA fragments of 268 bp (5G) or 171 bp (G3) or both, is represented by a black box. No excision, defined by no detection of either of the nested PCR products of 268 bp (5G) or 171 bp (G3), is represented by a gray box. If no full-length (top band of 1282 bp in the 5G only) or excised product could be amplified (in both 5G and G3), then this is represented by a white box. This analysis demonstrated SIV DNA cleavage in a broad range of tissues, 3 and 6 months after vector treatment. SIV viral excision, 5′LTR to Gag or Gag to 3′LTR, was observed in some tissue for all EBT-001 treated animals. There was individual animal variability observed, as well as differences in excision in tissues. Excision was detected at 3 months and in some tissues at 6 months. No virus was detected in some tissue samples at the highest dose of EBT-001 (Group 4) at 6 months post-EBT-001, which may indicate the elimination of viral DNA from those sections of tissues. The results of the excision activity analysis showed SIV DNA cleavage in a broad range of tissues 3 and 6 months after vector treatment, and in blood 2–10 weeks after vector treatment, including the presence of SIV excision in the blood of one animal (BM79) at necropsy (Fig. 4D). Since the 5G EXA assay amplifies both the full-length virus (1282 bp) and the amplicon (268 bp) of the excision product after the removal of intervening DNA sequences between 5′LTR to Gag, we calculated the percent efficiency of the 5G excision (Tables S8 and S9, Fig. S6, Supplementary material and methods). We understand the limitations of this PCR-based assay, as it is semi-quantitative, and it is possible that the smaller excision products may be amplified more efficiently than the intact virus. In addition, this assay reflects only one of 3 possible excision products that can occur through EBT-001 cutting. Despite the limitations of assay detection, the results show varying excision abilities with several animals reported to have detectable excision in tissue sections examined, even at the lower doses.

Example of 5G and G3 PCR of necropsy tissues from monkey MA285 (A). Presence of viral excision (displayed as black boxes), absence of viral excision (gray boxes), or absence of full-length and excised product (white boxes) in multiple tissues at necropsy (B). Example of 5G and G3 PCR from blood of animals BM79 and CP26 pre-EBT-001 and 4, 5, and 6 months post-EBT-001 (C). Presence of viral excision (black box) or absence of viral excision (gray boxes) from the blood of animals after EBT-001 (D). Dotted line represents the necropsy timepoint of 3 months or 6 months. No blood was obtained from the 10¹⁴ GC/kg animals at 0.5-3.0 months post-EBT-001 due to COVID shutdown. GC genome counts, mes ln mesenteric lymph node, PCR polymerase chain reaction, SIV simian immunodeficiency virus.

Weights, lymphocyte and monocyte counts, blood chemistries and cytokine analyses

Weight (kg) was measured in each monkey over the course of the study. We observed that Group 1 (untreated, maintained on ART) animals continued to lose weight during the study, whereas the EBT-001 treated groups kept on their trajectory of weight gain as is expected for their normal group (treated vs untreated, P = 0.029) with the highest dose animals having the largest weight gains (Fig. 5A). Among the cohort receiving the highest dose of EBT-001, neither animal experienced weight loss and both had increases in absolute lymphocyte count (Fig. 5B) without an increase in monocytes (Fig. 5C). No changes were seen in cholesterol, glucose, phosphorous, calcium, globin, and total protein in EBT-001 treated animals when comparing pre-infection, pre-EBT-001 to 3 months post-EBT-001 (Fig. 5D–I). However, the ART-only treated animal, IT6, had severe leukopenia (Fig. 5B), moderately severe hypophosphatemia (Fig. 5F), moderate hypocalcemia (Fig. 5G), severe anemia, and hypoproteinemia (Fig. 5H, I). In addition, untreated animal, 7S0, had marked hypoglycemia (Fig. 5E).

The mean (SEM) of animal’s weights are shown as percent change of pre-EBT-001 weight. Weights from EBT-001 treated (red, light blue, and dark blue lines) are significantly increased compared to untreated (black) (P = 0.029). A linear mixed-effects model was used to compare percent change in weight from pre-EBT-001 over time between treated vs untreated groups. Statistical significance was based on a 2-sided alpha level of less than 0.05 (A) The fold change in lymphocyte count from the CBC data over the average pre-EBT-001 values are shown. Each symbol is one animal and shown as the mean and SEM (B). The fold change in monocyte count from the CBC data over the average pre-EBT-001 values are shown. Each symbol is one animal and shown as the mean and SEM (C). The scatter dot plots show the mean and SEM, where each symbol represents an animal from pre-infection, pre-EBT-001 and 3 months post-EBT-001 for cholesterol (D), glucose (E), phosphorus (F), calcium (G), globin (H), total protein (I), ALP (J), ALT (K), and AST (L). Cytokines measured by MSD over time post-EBT-001 are shown for individual animals for plasma IFN-gamma (M), plasma IL-7 (N), plasma IL-15 (O). ALP alkaline phosphatase, ALT alanine transaminase, ART anti-retroviral therapy, AST aspartate transaminase, CBC complete blood count, IL interleukin, Mo month, MSD mesoscale discovery, SEM standard error of the mean, SIV simian immunodeficiency virus.

No notable changes in clinical chemistry or hematology parameters were observed in EBT-001 treated animals at pre-infection, pre-EBT-001, or 3 months post EBT-001 (Fig. 5J-L). However, the Day 6 post-EBT-001 dose laboratory evaluation of the 2 animals in Group 4 (highest dose) showed transient increases in serum concentrations of the liver enzymes, alkaline phosphatase (ALP), alanine transaminase (ALT) and aspartate aminotransferase (AST), and in total serum bilirubin (collectively referred to as liver function tests) (Fig. S7). Importantly, the elevated liver function tests returned to baseline range values in both NHP within 1 to 8 weeks (Table S10), and no evidence of liver necrosis or other signs of liver injury was observed by microscopic examination at the scheduled termination of the study at 6 months following dosing. Interferon (IFN)-gamma, interleukin (IL)-7, and IL-15 (Fig. 5M–O) cytokines were examined in plasma in animals after EBT-001 to check for adverse cytokine responses to AAV9 and immune responses. IFN-gamma was only elevated in one untreated animal and none of the EBT-001 treated group (Fig. 5M). IL-7, important in T cell development and HIV killing, was increased in one animal from Group 4 and IL-15, which activates natural killer (NK) cell-mediated viral responses, trended to increase in the treated animals (Fig. 5N–O).

Off-target assessment using whole genome sequencing

The potential for unintended off-target editing in the EBT-001 treated NHPs was assayed in lymph nodes using whole genome sequencing analysis (WGS) to provide a means to identify possible unintended editing events across the genome. WGS was conducted on genomic DNA samples from lymph node biopsies collected from 2 NHPs, CK49 and CH97, before receiving EBT-001 and from the same animals after 3 months or 6 months of treatment with EBT-001, respectively. WGS was performed using 2 × 150 bp paired end sequencing and averaged ~30× coverage depth. To effectively align and process the sequencing reads, a pipeline was designed in the High-performance Integrated Virtual Environment (HIVE), a cloud-based environment optimized for storage and analysis of extra-large data, such as NGS (9; Supplementary text). When assessing for unintended off-target editing, we focused on the nominated sites with relatively higher homology to the guide sequences. The indel rates were calculated for genomic loci within 10 bp of hypothetical cut sites of each nominated locus. When we compared the treated with untreated samples from lymph node, the initial screen identified 5 sites to have significantly different numbers of SNPs (Table 2) using two-tailed chi-squared test with the Benjamini-Hochberg procedure to control false discovery rate (FDR). We observed that all these SNPs occur at a distant location from the hypothetical cut site (7–10 bases away) with SNPs showing in both treated and untreated samples. Further examination of these sites has shown a lack of evidence to attribute these changes to gene editing due to alignments being too noisy, having long stretches of Ts around the potential site that may result in ambiguous alignment, or coverage being low (Supplementary text). Additionally, it was noted that the relative change of the SNPs frequency was not consistent between the two select animals across all 5 sites. Hence, we conclude that there is lack of evidence for off-target editing activity. We also screened for structural variants among these nominated sites, we detected no read supporting partial alignments within 40 bp windows surrounding the nominated cut sites. Additionally, we looked at potential AAV integration events by searching for chimeric reads with partial alignment. Compared to samples before treatment, the analysis found only 2 potential reads in CH97 post-treatment and 1 read in CK49 post-treatment. The 2 reads from CH97 post treatment were identical to each other, possibly PCR duplicates, and showed partial alignment to Chr7:128862755–128862834 with partial alignment to Cas9 coding region (EBT001: 2636–2565). The 1 read in CK49 post treatment also showed partial alignment to Cas9 coding region (EBT001: 2944–3037), and partially aligned to Chr12:28738543–28738586 (Fig. S8). However, it was noted that among these three reads, none of their respective paired end read supported the partial alignment results. Taken together with the extremely little amount of supporting reads, we concluded that none of the 3 split-read jumps can be attributed as evidence for AAV integration in the lymph nodes after EBT-001 treatment.