Development of β-globin gene correction in human hematopoietic stem cells as a potential durable treatment for sickle cell disease

Abstract

Sickle cell disease (SCD) is the most common monogenic serious disease with 300,000 births annually worldwide. SCD is autosomal recessive from a single point mutation in codon six of the β-globin gene (HBB) resulting in sickle hemoglobin. Ex vivo β-globin gene correction in autologous patient-derived hematopoietic stem and progenitor cells (HSPCs) might be an ideal treatment of SCD. We previously developed an HBB gene targeting strategy that utilizes high-fidelity Cas9 precomplexed with chemically modified guide RNAs to induce rAAV6-mediated gene correction of the SCD-causing mutation in HSPCs. Here we present foundational translational data that demonstrate the pre-clinical feasibility, efficacy, and toxicology of HBB gene correction in plerixafor-mobilized CD34+ cells from healthy and SCD patient donors (Drug Product-gcHBB-SCD). Notably, we achieved up to 60% HBB allelic correction in clinical-scale gcHBB-SCD manufacturing and long-term engraftment in immunodeficient NSG mice, with multi-lineage allele gene correction frequencies of 20% in multiple hematopoietic organs. The long-term safety tumorigenicity/toxicology study demonstrated no evidence of abnormal hematopoiesis, genotoxicity or tumorigenicity from the engrafted gcHBB-SCD Drug Product. Altogether, this preclinical data supports the safety, efficacy, and reproducibility of a gene correction strategy for initiation of a Phase I/II clinical trial for SCD patients.

One sentence summary

The pre-clinical efficacy and safety data support the initiation of a phase I/II clinical trial for β-globin gene correction in patient-derived hematopoietic stem cells for the treatment of sickle cell disease.

Introduction

Sickle cell disease (SCD) is an autosomal recessive monogenic blood disorder, caused by a single point mutation (A>T) in the sixth codon of the β-globin gene. This missense mutation changes the amino acid from glutamic acid (E) to Valine (V) and the formation of sickle β-chains (HbS). HbS forms abnormal hydrophobic interactions through axial and lateral contacts within sickling hemoglobin (HgbS) in the deoxygenated state leading to dysfunctional hemoglobin polymerization, the typical deformed sickle-like shape or red blood cells, and a hemolytic anemia (1). SCD affects millions of people worldwide causing both a high medical burden to patients, families, and communities and high financial burden to these same groups as well as the broader health care system (https://www.nhlbi.nih.gov/health-topics/sickle-cell-disease). The main disease manifestations of SCD are vaso-capillary ischemia, such as vaso-occlusive pain crisis (VOC) and acute chest crisis (ACS). The current standard of care focuses on symptomatic relief involving pain control, blood transfusions, hydroxyurea administration, and recently approved drugs that inhibit HgbS polymerization (2) or reduce pain crisis (3). The only approved curative treatment to date is allogeneic human leukocyte antigen (HLA)-matched hematopoietic stem cell transplantation (Allo-HSCT) (4, 5). However, a matched sibling donor is only available in 10–15% of cases (5, 6) and the use of haploidentical or unrelated donors is still considered experimental (7).

The gene therapy field has recently seen unprecedented efforts to provide new and potentially curative treatments for SCD by taking advantage of ex vivo lentiviral transduction of autologous hematopoietic stem and progenitor cells (HSPCs). These therapies in development are focused on restoring hemoglobin levels through expression of an additional globin gene, either a non-sickling β-chain variant (8, 9) or the γ-chain for fetal Hgb (HgbF) formation (10). The reinfusion of autologous genetically-modified HSPCs (auto-HSCT) post-patient myeloablation obviates the need to find donors for transplantation and decreases the risk of immune related complications due to HLA–mismatch between donor and recipient (e.g. rejection, graft-versus host disease, poor immune reconstitution). Another crucial improvement in HSCT protocols for sickle cell patients has been the introduction of plerixafor (Mozobil, Genzyme) as an alternative HSPC-mobilizing agent to G-CSF which causes toxicity and deaths in SCD patients (11, 12). Plerixafor is a bicyclam molecule that antagonizes the binding of chemokine stromal cell-derived factor-1 (SDF-1) to its cognate receptor CXCR4, and results in rapid and reversible mobilization of HSPCs into peripheral circulation (13, 14). Currently, there are four ongoing clinical trials (at clinicaltrials.gov: NCT02186418, NCT02247843, NCT02140554, NCT03964792), some of which already report therapeutic safety and efficacy (15); however, lentiviral vector related safety concerns and transduction efficiency still continues to be a limitation (16, 17), warranting further development of gene therapies for SCD.

The CRISPR/Cas9 system (18) has proven to be versatile, simple to engineer, and efficacious in human HSPCs (19–22). It can be used to induce precise genetic alterations, by exploiting either the non-homologous end joining (NHEJ) DNA-repair pathway, or to perform “gene correction” and “gene addition” through the homologous recombination (HR) repair pathway, enabled by the supply of a homologous DNA donor template (23). Several groups have demonstrated an increase in HR through recombinant adeno-associated virus serotype 6-mediated (rAAV6) delivery of the donor template (20, 24–34). Presently, several genome editing approaches in autologous HSPCs utilizing the CRISPR/Cas9 system are being developed for SCD. Some strategies are aimed at reactivating the expression of the cognate fetal hemoglobin by interfering with the function of the β-globin switch master regulator BCL11A (35). Such strategies have involved drastically reducing BCL11A expression (36, 37), or by recreating DNA edits in the β-globin locus mimicking hereditary persistence of fetal hemoglobin (38, 39) and altering BCL11A binding sites at the γ-globin promoter (40). The expected outcome would be a switch in hemoglobin expression from the adult mutated β-chain (HbS) to the fetal γ-chain (HbG). Currently, there is one open clinical trial (clinicaltrials.gov NCT03745287) exploring this approach for SCD and two more for β-Thalassemia major (clinicaltrials.gov NCT03432364, NCT03655678). However, an alternative approach and potentially more ideally curative is to directly correct the pathogenic sickle mutation in HBB gene, thus preserving physiologic regulation of gene expression. By directly correcting the endogenous HBB gene, this will also remove expression of pathologic HbS from the cell, thus avoiding the problem of competing with the highly expressed pathologic protein. We developed an HBB gene correction strategy that combined ribonucleoprotein delivery of a high-fidelity Cas9 variant with rAAV6 DNA donor delivery that preserved HSC biological functions following the manufacturing process (20, 26).

Here we describe the preclinical studies with HBB gene corrected CD34+ HSPCs to treat SCD (gcHBB-SCD), focusing on efficacy, safety, toxicology and tumorigenicity, and clinical-scale manufacturing feasibility. In particular, we demonstrate: 1) reproducible high frequency HBB gene correction in plerixafor-mobilized HSPCs from healthy donors and SCD patients, 2) long-term engraftment of HBB gene corrected HSPCs, 3) absence of genotoxicity and tumorigenic potential, and 4) high feasibility of manufacturing a full human clinical dose. Altogether these studies are the foundation for the preclinical study supporting an Investigational New Drug (IND)-submission to the Food and Drug Administration (FDA) for a Phase I/II trial for the treatment of SCD.

Results

HBB gene correction in long-term engrafting plerixafor-mobilized hematopoietic stem and progenitor cells from healthy donors

We previously demonstrated the proof-of-concept of successful HBB targeting with a reporter donor cassette in long-term engrafting HSCs isolated from mobilized peripheral blood (20). We also showed feasibility of HBB gene correction for the E6V point mutation by using an AAV6 DNA donor (20, 41). Here, we performed the gene correction process in clinically relevant plerixafor mobilized-HSPCs (PlxHSPCs) from healthy donors (21) (Fig.1A). HBB gene correction protocol proved to be highly reproducible in five different HSPCs donors, with a mean of 80% viability and 92% CD34+ cells by FACS, and 22% HBB allelic gene correction (gcHBB) assessed by droplet-digital PCR (ddPCR) (Supplemental Fig.1A). We next tested targeted HSPC engraftment potential and hematopoietic reconstitution in sublethally-irradiated NSG mice. We compared the systemic intravenous (tail vein, IV) and the intra-bone (intrafemorally, IF) route of delivery by assessing for biodistribution and kinetics of HSPCs marked by a luciferase expression cassette targeted to the HBB locus. Gene targeted HSPCs showed faster engraftment with IF injection but IV injection resulted in a broader biodistribution, with HSPCs homing to multiple bone marrow sites (Supplemental Fig 2). Subsequently, we manufactured PlxHSPCs for HBB gene correction (gcHBB-SCD) and administered the cells in a dose escalation into NSG mice using both IV and IF injection methods (Fig. S1B). Human cells collected from bone marrow of NSG mice at week 16 post-injection showed bi-lineage reconstitution of lymphoid and myeloid cells (Fig. S1C) and ~5% of long-term HSPCs retained gcHBB alleles, independent of route of administration (Fig. S1D). These results provided the basis to proceed with performing IV injections in subsequent human HSPC engraftment studies using NSG mice. Furthermore, bone marrow residing HSPCs were also differentiated towards the erythroid lineage ex vivo for 14–16 days (Fig. 1A, Fig. S3A–B) and showed 2.5% and 16% gcHBB alleles in two independent experiments (Fig. S3C). These results suggest that it is possible to target HBB for E6V gene correction in long-term engrafting plxHSPCs, albeit at frequencies that might be disease modifying rather than curative in an autologous transplant setting.

Fig. 1. Therapeutic gene correction protocol optimization in healthy donor derived HSPCs.

(A) Schematics of cell manufacturing protocol, in vitro readouts and experimental design of in vivo NSG mouse studies. (B) Percentage of in vitro gene corrected alleles (gcHBB) (left), viability and CD34 content (right) in up to 13 cell donors; black lines indicate mean values. (C) Human chimerism in bone marrow of NSG mice at week-16 post-injection. Control cells were i. electroporated only (Mock), ii. coupled with either RNP (Cas9 only) or iii. AAV6 (AAV6 only). One-way ANOVA Kruskal-Wallis test plus Dunn’s multiple comparisons test; ns, not significant. (D) Percent distribution of human hematopoietic lineages within the human cell population. (E) Percent distribution of gcHBB alleles in the human cell population (bulk) and in the respective hematopoietic lineages. (F) Percent gcHBB alleles distribution in bulk population and in a mouse-by-mouse analysis of human lineages in bone marrow samples, which show either good representation of all lineages (left) or mostly myeloid cells composition (right). (G) Quantification of HBB alleles distribution in single CFU colonies. WT: wild type, INDEL: insertion/deletions, HR: homologous recombination. Dot colors indicate HBB cellular outcome: white “neutral”, red “deficient”, green “corrective”. Brackets group genotypes per cellular outcomes; black bars indicate mean values. (H) Percent of gcHBB alleles in vivo (bone marrow) linked to genotype of single sorted HSPC derived CFU. Pearson r test, p<0.001. Black bars and lines indicate median values if not differently specified.

Manufacturing protocol optimization to achieve higher frequencies of in vivo HBB gene correction with very low detectable off-target activity

Based on the above results, we optimized the genome editing protocol to achieve higher in vivo retention of gene correction alleles with limited off-target activity. We investigated the following parameters: 1) the addition of UM171 molecule to the culture media, which has been shown to promote HSC cycling and self-renewal (42), to increase AAV6 transduction efficiency and HSCs recovery (43), and to reduce ROS generation during cell cycle (44); 2) implementation of the low-density culture conditions to promote HSCs cycling (25, 45); and 3) use of a high-fidelity Cas9 (R691A) variant that works well in the RNP format (26).

While the addition of UM171 to the culture medium did not impact the in vitro gene targeting frequencies (Fig. S4A) nor the in vivo engraftment potential of human CD34+ cells (Fig. S4B), it increased in vivo retention of gcHBB alleles with respect to the in vitro population (median ratio of 0.75 (in vivo divided by in vitro)) (Fig. S4C) in the NSG bone marrow of transplanted mice. When calculating the absolute number of human cells with corrected alleles on a mouse-by-mouse basis, we measured a ~3-fold increase of engrafted gcHBB-SCD in the UM171 treated group. Because of the large and known heterogeneity in the NSG model, the result was not statistically significant (Fig. S4D). All gcHBB-SCD cell populations were also able to engraft in secondary transplant experiments (Fig. S4E). As previously demonstrated, low-density cell culture condition primes multi-lineage repopulating HSPCs for HR (Fig. S5A–B) (25, 45). Interestingly, these findings corresponded to a decrease in pro-inflammatory cytokines in the cell culture media (45, 46) (Fig. S5C). Cytokines such as TGFA, IL-8, IL-1RA, IP-10, HGF and MCP-1 were at significantly lower concentration in the media at day 2, indicating a potential decreased effect of inhibitory feedback signals on cell growth (45). Finally, we wanted to improve the editing safety profile by using a high-fidelity (HiFi) Cas9 protein. This modified Cas9 variant has been shown to retain the high on-target activity of WT Cas9 while reducing off-target editing (26). When used in the gene targeting protocol, HiFi Cas9 was equivalent to WT protein in on-target activity at the HBB gene, generating an average of 71% HR when coupled with AAV6 transduction (Fig. S5D). Importantly, HiFi Cas9 dramatically decreased INDEL frequency at OT1 below the limit of detection in both the in vitro culture and the in vivo engrafted HSPCs (Fig. S5E).

gcHBB-SCD contains long-term engrafting HBB corrected HSPCs above the putative therapeutic threshold for SCD

We investigated whether the optimized manufacturing protocol using UM171, low density culture condition and HiFi Cas9 (Fig. 1A) would increase the ~5% gcHBB allele frequency in long-term engrafting plxHSPCs. The gcHBB-SCD were manufactured using plxHSPC from healthy donors and resulted in an average of 65% gcHBB alleles, while consistently showing >80% viability and >90% CD34+ cellular content (Fig. 1B). When studying the kinetics of AAV6 transduction, we found an 8-hour minimum period of AAV6 incubation necessary for efficient allele correction frequency, while the accumulation of gcHBB alleles was a function of time likely related to the completion of the HR-DNA repair process (Fig. S6A–B). NSG mice were transplanted with cells harvested 48 hours after RNP delivery and AAV transduction. gcHBB-SCD cells showed similar engraftment frequencies (28% median engraftment, n=15) to control cells, including electroporated only (mock, n=9), RNP-only (n=4) and AAV6-only (n=4) (Fig. 1C). Engrafted human cells derived from transplanted gcHBB-SCD were sorted from bone marrow at week 16 using human lineage-specific markers (Fig. S7). The human graft was multi-lineage and consisted mostly of CD19+ B-cells and CD33+ myeloid cells (Fig. 1D). Genomic DNA was extracted from the different cell types and gcHBB alleles frequencies were analyzed in specific lineages. Across all the mice transplanted (n=15), there was a median gcHBB allele frequency of 21%, 20%, 31%, 27%, and 24% in bulk human cell population, B-cells, myeloid cells, T/NK cells, and HSPCs, respectively (Fig. 1E). The rest of the HBB alleles analyzed in the bulk human cell population were ~60% INDELs and ~20% WT (Fig. S8A). A mouse-by-mouse analysis highlighted that the majority of NSG mice had several human hematopoietic lineages derived from transplanted gcHBB-SCD, with similar frequencies of gcHBB alleles among the different lineages (Fig. 1F, left). Interestingly, a subset of mice had reduced B cell gene correction frequencies but sustained higher correction frequencies in the myeloid lineage (Fig. 1F, right). To estimate the percent of HSPCs that had at least one gcHBB allele in vivo, we single-cell sorted CD34+ HSPCs harvested from bone marrow of NSG mice (n=11) into methylcellulose and genotyped single CFU colonies. Positive colonies for the gcHBB allele(s) were an average mean of 30% frequency (up to 50% in some donors), with an average mean of 12% carrying a bi-allelic gcHBB (Fig. 1G and Fig. S8B). Notably, this frequency of gcHBB positive colonies was approximately 1.3 times more than the alleles frequency in the bulk cell population from NSG bone marrow (Fig. 1H). These results indicate that gcHBB-SCD contains long-term repopulating HSCs with ~30% of cells carrying an HBB corrected allele. This cell frequency of having at least one allele corrected matches the threshold of 30% donor cell chimerism as important in reducing SCD symptoms following allo-HSCT (47).

Plerixafor mobilized long-term engrafting HSPCs from sickle cell disease patients can be gene corrected at high frequencies

To understand if the optimized plxHSPC manufacturing from healthy donors translated to sickle cell HSPCs, we obtained cryopreserved CD34+ plxHSPCs from two different SCD patients (NCT03226691) and assessed the gcHBB-SCD manufacturing reproducibility, and ability to differentiate into erythrocytes and express HgbA tetramers. In six independent experiments, we achieved >60% gcHBB alleles (Fig. 2A) with minimal effect on cell viability (>80% viable cells) and expression level of CD34 (>90% CD34+ cells) (Fig. 2B), showing feasibility of the gene correction protocol in SCD derived HSPCs. We plated gcHBB-SCD and mock control (electroporated HSPCs) into an erythroid differentiation medium and obtained >85% CD45−/CD34−CD71+/GPA+ red blood cells (Fig. 2C, Fig. S9A). gcHBB-SCD derived erythroblast in vitro generated a restored output of ‘corrective’ hemoglobin (<10% HgbS, >25% HgbF, >65% HgbA) in contrast with the mock control derived erythroblasts (>85% HgbS, >10 HgbF, 0% HgbA) (Fig. 2 D and Fig. S9B). As expected, the level of HgbA restoration increased with the higher gcHBB allele frequencies (Fig. 2A). We also efficiently manufactured non-mobilized peripheral blood CD34+ cells from SCD patients (an average of 57% gcHBB alleles) (20) and differentiated them towards the erythroid lineage, hence making similar findings that gcHBB alleles produced >90% HgbA (Fig. S9C). Collectively, these data indicate that high frequencies of gene correction (for E6V HBB gene mutation) can be achieved in SCD patient HSPCs, which results in a restoration of non-pathologic HgbA expression in vitro.

Fig. 2. Effective gene correction, recovery of normal hemoglobin output, and engraftment capability in SCD patient derived HSPCs.

(A) Percent allele modification in six gcHBB-SCD biological replicates (Exp 1–6). (B) Viability and CD34 content in vitro. (C) Percent of in vitro differentiated erythroid cells displaying CD71+ CD235a+ markers (see Fig. S9A). (D) Quantification of hemoglobin tetramers (HPLC) on erythroid differentiated population (n=4). (E) Human chimerism in the bone marrow of NSG mice analyzed at week-16 post-injection. Mann-Whitney test; **, p<0.01. Black bars indicate median values. (F) Percent distribution of human hematopoietic lineages in the NSG bone marrow. (G) Percent of gcHBB alleles in the human cell population (bulk) in vitro and in vivo mouse-by-mouse, along with the respective human hematopoietic lineages collected from NSG bone marrow. Mann-Whitney test; *, p<0.05. Black bars and lines indicate mean (± SD) if not differently specified.

We tested gcHBB-SCD manufactured from SCD patients to engraft long-term and reconstitute hematopoiesis. gcHBB-SCD along with mock controls were injected IV into sub-lethally irradiated NSG mice. Bone marrow human chimerism as well as gcHBB allele frequencies within the human cell compartment were evaluated 16–20 weeks post-transplantation. Overall, SCD patient derived plxHSPCs showed robust engraftment capacity with a median human chimerism of 80% and 42% for mock electroporated HSPCs and gcHBB-SCD, respectively (Fig. 2E). The human graft was multi-lineage and consisted of mostly CD19+ B cells and CD33+ myeloid cells (Fig. 2F). The bulk human cell population in the bone marrow of mice injected with gcHBB-SCD showed a robust percentage of gcHBB alleles (median 30%). Furthermore, we isolated B cells, myeloid cells, red blood cells and HSPCs from the bulk human population and assessed similar gene correction frequencies in all the human hematopoietic lineages (Fig. 2G). These findings demonstrate long-term reconstitution of HBB gene corrected HSCs with multi-lineage potential. Interestingly, in the red blood cells compartment (GPA+) gene correction frequencies were higher with a median of 60% gcHBB alleles (Fig. 2G). This frequency suggests there is in vivo enrichment of gene corrected erythroid progenitors, which might have a better fitness compared to the noncorrected sickle cells or cells with bi-allelic INDELs leading to a beta-thalassemia phenotype. NSG mice engrafted with gcHBB-SCD generated from non-mobilized patient-derived peripheral blood HSPCs showed no decrease in gcHBB alleles in the bone marrow graft with respect to the in vitro cell population (Fig. S9D–E). Engraftment with these cells was almost entirely myeloid, suggesting that the patient derived non-mobilized CD34+ HSPCs had a significant myeloid bias (48). Nonetheless, even the non-mobilized CD34+ HSPCs retained long-term engraftment potential following transplantation into NSG mice. Overall, these results demonstrate the therapeutic potential of the gene correction platform to treat sickle cell disease (and directly revert the E6V mutation).

No evidence of in vivo oncogenic potential of gcHBB-SCD manufactured at clinical-scale

To support the development of a first-in-human clinical trial for a sickle cell disease HBB gene correction therapy, we performed IND-enabling studies aimed at defining gcHBB-SCD clinical-scale manufacturing protocol that met specifications for cell product potency (gcHBB alleles), identity and purity (CD34+), and cell viability. To ensure that manufacturing would be substantially similar to the clinical process, we converted from cell culture reagents to current good manufacturing practice (cGMP) compliant products and engaged with key contract development and manufacturing organizations (CDMOs) for large-scale provisions of key manufacturing reagents (see Methods). We adapted a protocol for AAV6 production from previously published 2-plasmids system technology (49). In particular, we developed the scale-up production process up to 3L scale in a suspension culture of HEK-293T. Several mid-scale AAV6 lots were produced at Stanford Laboratory of Cell and Gene Medicine (LCGM). Two of these (lot#3 and #4) were used in large-scale manufacturing of human HSPCs for gcHBB-SCD IND-enabling studies (Fig. S10). We assessed the following parameters for clinical-scale manufactured gcHBB-SCD: 1) feasibility of CD34+ harvest and manufacturing from plerixafor mobilized healthy donors, 2) efficacy of gcHBB-SCD cell product manufactured at large-scale, and 3) long-term toxicological and tumorigenic potential in NSG mice.

Apheresis products of plerixafor mobilized-peripheral blood (on average 50-billion of total nucleated cells) were purchased and freshly manipulated. Once plated, cells were cultured for two days at a concentration of 2.5E+5 cells/mL in cytokine supplemented media. At day 2, cells were collected and underwent the gene correction protocol adapted for large-scale manipulation. The final product gcHBB-SCD was then harvested at day 4 and underwent quality control and cryopreservation. We demonstrated the feasibility of manufacturing at clinically relevant cell numbers (2.3E+7 – 1.3E+08 CD34+ HSPCs) with a mean of 77% viability, 94% CD34+ cells, and 45% gcHBB alleles (Table 1, Fig. 3A). The colony forming-unit ability (CFU-assay) of edited cells was 2-fold lower compared to untreated controls without lineage skewing (Fig. 3B, Fig. S11A–B). Genotype analysis of colonies showed a higher percent of gene-corrected HSPCs than gcHBB alleles (Fig. 3C), with an average 56% of cells having at least one corrected gcHBB allele (Fig. 3D and Fig. S11C), of which 32% and 24% were mono- and bi-allelic events, respectively (Fig. 3D).

Table 1.

Successful medium-scale cell manufacturing process.

Assay	Specifications	Run 1 LCGM-18019	Run 2 LCGM-18020	Run 3 LCGM-18021	Run 4 LCGM-18022	Run 5 LCGM-18023	Run 6 LCGM-18025	Average	SD
Viability (%)	≥70%	90	79	65	75	75	80	77	8.1
Viable Cell Count	Report	1.13E+08	2.37E+07	2.91E+07	7.30E+07	1.11E+08	1.26E+08	7.80E+07	4.36E+07
Percent Cell Recovery (%)	Report	80.8	51.5	40.3	37.4	74	108.6	65.4	27.5
Phenotypic Analysis (%)	>80% CD34	85	99.8	99.7	93	92	92	93.6	5.6
Percent Allele Correction (%)	>20%	56	41	39	37	45	52	45	7.6
Percent Cells Corrected (%)	Report	65	47	53	54	NA	61	56	7.1

Scale-up process development was performed at the Stanford – LCGM. Final gcHBB-SCD drug products from six manufacturing runs (classified as LCGM-18019–23 and −25) were tested and met specifications for the listed quality release criteria (rows). Last two columns report average and SD demonstrating a reproducible manufacturing process.

Fig. 3. Efficient scale-up cell manufacturing in clinically relevant CD34+ cells with lack of tumorigenicity in a long-term toxicology study.

(A) Percent of allele modification in six medium-scale cell manufacturing runs (Run 1–6). (B) CFU frequency for gcHBB-SCD cell product and untreated cell counterpart. Paired t test. *, p<0.05. (C) Scatter plot linking gene correction of HBB alleles in bulk cell population to single genotyped colonies. Pearson r test *, p<0.05. (D) Quantification of HBB alleles distribution in single CFU colonies (genotype). WT: wild type, INDEL: insertion/deletions, HR: homologous recombination. Dot colors indicate HBB cellular outcome: white “neutral’, red ‘deficient’, green ‘corrective’. Brackets group genotypes per cellular outcomes. (E) Human chimerism in hematopoietic tissues (peripheral blood, bone marrow and spleen) of NSG mice in long-term toxicology study. Untreated HSPCs (in blue) and gcHBB-SCD (in green) groups were segregated by sex. One-way ANOVA Kruskal-Wallis test plus Dunn’s multiple comparisons test; ns, not significant; *, p<0.05; **, p<0.001; ***, p<0.0001. (F) Percent gcHBB alleles in hematopoietic tissues collected at study end from gcHBB-SCD NSG cohort (n=39). (G) Percent human engraftment and (H) percent gcHBB alleles in bone marrow of NSG mice injected with cell manufacturing run 5 at Stanford lab (n=10). Black bars and lines indicate median values.

The long-term toxicology and tumorigenicity study was conducted using the gcHBB-SCD product generated from the 5 pre-clinical process qualification runs. Manufactured gcHBB-SCD cell products were thawed and injected in sub-lethally irradiated adult NSG mice. Vehicle control (PBS injected) and untreated HSPCs were also injected to account for radiation toxicity and influence of treatment (Fig. S11D). Mice were observed for the development of any adverse clinical signs and in life parameters (Supplementary Methods), with no differences seen between controls, mice injected with untreated HSPCs and mice injected with gcHBB-SCD. Peripheral blood was collected (8- and 16-weeks post-injection) and at termination (20-weeks post injection or at early euthanasia for clinical criteria) and samples were submitted to the Comparative Pathology Laboratory (CPL) for complete blood count and leukocyte differential. Comparing vehicle control and untreated HSPCs groups to the gcHBB-SCD group, there were no added effects of the gcHBB-SCD treatment on hematological endpoints of the mice transplanted (Fig. S12A). Importantly, extensive histopathology observations performed in >40 different body tissues showed no gross lesions or adverse effects attributable to cellular infusion in all animals analyzed. All NSG mice transplanted were assessed for engraftment of human cells in the peripheral blood (Week 8, 16 and 20), bone marrow (Week 20) and spleen (Week 20). In the mice transplanted with gcHBB-SCD, there was a mean engraftment of 3.3%, 0.5% and 0.8% human engraftment, while mice injected with untreated HSPCs displayed a mean of 19%, 3.3% and 7.3% human engraftment at week 8, 16 and 20, respectively. Human chimerism in bone marrow was a mean of 3.3% and 20.0% human cells in mice transplanted with gcHBB-SCD (n=38 mice analyzed) and untreated HSPCs (n=23 mice analyzed), respectively. Interestingly, there was increased human chimerism in the spleen for both mice groups, mean of 25.1% for gcHBB-SCD group and 66.0% for untreated HSPCs group. We further analyzed engraftment data separating female and male mice. Differences seen between mice transplanted with untreated HSPC and gcHBB-SCD were statistically significant in most of the grouped analysis (Fig. 3E). Albeit, engraftment data in spleen were more similar to those of bone marrow chimerism presented before (in Fig. S1, Fig. S4, Fig. S9, Fig. 1 and Fig. 2), and indicate preferential homing of HSPCs to the spleen in these cohorts of mice (one-way ANOVA and Friedman test). Several human hematopoietic lineages were represented within the human CD45 population in BM and spleen, with a prevalence of myeloid cells (Fig. S12B–C). Human cells engrafted in NSG mice in the toxicology study (n=39 mice analyzed), had a percent gcHBB alleles within ranges previously shown, with a median of 7.0%, 23% and 20% in peripheral blood, bone marrow and spleen, respectively (Fig. 3F). We further confirmed robust engraftment potential of gcHBB-SCD when injecting in parallel cells from toxicology manufacturing run 5 (see Table 1) in NSG mice at a Stanford laboratory (Fig. 3G–H, Fig. S12D). In conclusion, gcHBB-SCD manufactured at clinical-scale with cGMP compatible reagents are capable of long-term engraftment into multiple sites of NSG mouse hematopoiesis with retention of gcHBB alleles at frequencies similar to our small-scale manufacturing protocols. There was no evidence of tumorigenicity or abnormal hematopoiesis from the manufactured product. There was continued evidence that the genetically modified cells did not engraft as well as control cells, a finding consistent with all genetic engineering approaches, including lentivirus (50, 51).

Low molecular genotoxicity in gcHBB-SCD investigational cell product

While the in vivo mouse study described above did not show any functional tumorigenicity or abnormal hematopoiesis, a concern for genome editing using engineered nucleases is the possibility of generating double strand breaks at off-target sites (OTs) in the genome that can create INDELs and/or translocations at genes regulating critical cell functions. The gcHBB-SCD lots manufactured for the long-term toxicology study were analyzed at molecular level for genomic alterations. First, cells were karyotyped, and chromosomal analysis was performed. From 150 cell metaphases there was no detectable clonal gross genomic aberrations (Fig. S13A–B). One gcHBB-SCD lot (LCGM-10819) had three metaphase anomalies reported, however the cytogenetic report concluded that in the absence of additional like cells, none were considered clonal in nature and were likely technical artifacts. Nonetheless, to determine if these findings were of clonal nature, we screened 15 additional metaphase spreads from the same gcHBB-SCD lot, and all showed normal karyotypes (Fig S13B). Of note, this gcHBB-SCD lot (LCGM-10819) showed long-term engraftment with no signs of tumorigenicity or abnormal hematopoiesis when injected in NSG mice in the toxicology study.

We performed an extensive off-target analysis in the same gcHBB-SCD lots through NGS measurement of INDELs of a master list of putative OTs. The list was developed by cross-referencing three state-of-the-art methods for identifying potential off-target sites of sgRNAs (GUIDE-seq, CIRCLE-seq, and COSMID) (Table S1) (52, 26, 53, 54). We created a master list of 67 OTs to analyze by amplicon-based sequencing in gcHBB-SCD cells generated from the toxicology lots (Table S1).

The analysis showed that only a single off-target site (OT1) had measurable INDELs in all samples (0.5–2.9%) (Fig. 4A). This site is located in Chr9:101,833,584–101,833,606 that is in an intergenic region (site is ~270,000nt 3’ to closest gene--GRIN3A, Chr9:101,569,353), was identified by all three methodologies for off-targets identification and has been previously identified (20, 55). The HiFi Cas9 utilized in these studies significantly reduced the frequency of INDELs at OT1 (~2% and ~30% for HiFi and WT nuclease, respectively), while maintaining high on-target activity as previously described (26).

Fig. 4. Next generation sequencing (NGS) based-techniques assess very minimal genotoxicity in gcHBB-SCD cell product.

(A) NGS results of the 67 off-target rank list assessed in the six medium-scale manufacturing runs for toxicology studies (Run 1–6). On-target activity=percent INDELs+HR. Dotted line indicates the assay detection threshold. (B) Circos plots of genome-wide prey junctions binned into 5-Mb regions (black bars) are plotted on a log scale with indicated ticks; frequency ranges are colored from light orange (10 – 100) to increasingly darker orange colors by factors of 10. Red arrow connects the HBB bait site on chromosome 11 to the off-target hotspot on chromosome 9. (C) IGV plot of HBB off-target on chromosome 9 (coordinates bottom left); junctions are shown on a logarithmic scale. Red and blue numbers indicate the number of junctions from the specified region that translocated in the plus and minus orientation, respectively. The bar (bottom right) indicates 50 base pairs length. (D) Junction frequency at the main HBB-OT (OT1, Chr9) for WT and HiFi Cas9 (mean ± SD). Paired t-test; ***, p<0.001.

We also used high-throughput, genome-wide, translocation sequencing (HTGTS) analysis, which is based on linear-amplification-mediated PCR (LAM-PCR) to look for gross chromosomal rearrangements in cell products manufactured using the WT and HiFi Cas9 (56, 57) (Fig. 4B and Fig. S14A–B). HTGTS identified OT1 on chromosome 9 as the main off-target site (Fig. 4B). HTGTS demonstrated that HiFi Cas9 generated a 30-fold reduction in translocation frequency between HBB locus and OT1 (Fig. 4C–D).

A portion of junctional reads displayed a frequency that was independent of Cas9 specificity. These events included deletions of various size spanning 4kb downstream of the break site (Fig. S14D–E). A small number of events included junctions from HBB to HBD gene and to a breakpoint cluster region (BPC). Overall, the frequency of gross chromosomal changes is significantly reduced by using HiFi Cas9.

Discussion

We describe the development of a therapeutic cell product using ex vivo Cas9-RNP and rAAV6 for HBB correction in human hematopoietic stem/progenitor cells to promote the correction of the SCD causing point mutation in HBB.

We optimized the gene correction protocol in clinically relevant plerixafor mobilized HSPCs which preserved the viability and HSCs repopulation capacity in a xenotransplantation model. The overall gene correction frequencies and preservation of gene corrected cells following transplantation into NSG mice exceed what has been previously published when using single-stranded oligonucleotide delivery (40, 48, 58).

Common findings amongst genome editing strategies across different groups that exploit HR have been low numbers of gene edited cells and poor retention of gene edited cells after transplantation into NSG mice (20, 40, 59, 60). We introduced modifications which gave high levels of percent gcHBB alleles in vitro and increased the long-term engraftment of human cells with gcHBB alleles in vivo. In particular, the human engrafted population in NSG mice increased ~4-fold (from ~8% to ~30%) as well as the frequency of gene corrected alleles (from ~5% to ~20% gcHBB alleles) using healthy donor HSPCs. Notably, the percent human engraftment and gcHBB alleles were even higher in HSPCs derived from SCD patients, with a median of 40% and 30%, respectively. In the majority of NSG mice analyzed at study termination, we further observed an enrichment of gene correction in GPA+ erythroid cells (median of 50% gcHBB alleles), indicating a selective advantage of gene corrected red blood cells in vivo. The in vivo findings combined with finding >90% HgbA production in vitro derived from gene corrected sickle cell HSPCs, provide strong evidence that erythroid correction with greater than 70% HgbA can be achieved in vivo in humans. In patients undergoing allogenic transplant, subsequent long-term persistent mixed donor chimerism, ranging from 10–20%, has generated beneficial clinical results by keeping HgbS levels below 50% (6, 47, 61, 62). Likewise, even 1–5% normal donor HSPCs can result in the production of >70% non-sickling RBCs in the periphery due to the survival advantage (>10–30fold) of non-sickling RBCs over sickling RBCs (63, 64). These findings support the hypothesis that long-term engraftment with 10–20% ‘corrected’ CD34+ HSPC may be beneficial in reducing SCD symptomology. Estimates, however, of long-term engraftment from allo-HSCT may not completely reflect the biological process of engraftment of autologous gene-edited cells. Of note, in the xenotransplantation studies, all hematopoietic lineages were represented in the engrafted cell population, with a robust myeloid cell representation even in animals with poorer engraftment (see Fig. 1F). Myeloid lineage displayed sustained levels of HBB gene correction in bone marrow sorted cells and at single CFU level with 30% of colonies carrying at least one corrected HBB allele (see Fig. 1E–G). Herein lies a differentiating aspect of this gene correction therapy where we believe each cell with at least one corrected HBB allele will produce non-pathologic HgbA and less pathologic HgbS (47).

The scale up of manufacturing of gcHBB-SCD was performed with plerixafor mobilized healthy donor cells rather than SCD patient cells because we did not feel it was ethical at this point to subject SCD patients to the process given the potential risks. The challenge we encountered, especially in product manufacturing runs 2–3 (LCGM-18020 and LCGM-18021) (Table 1), is that healthy donors do not mobilize CD34+ HSPCs well after a single dose of plerixafor and the pheresis products had high red blood contamination that could negatively impact the subsequent manufacturing, highlighting the importance of beginning with high quality starting material.

The in vivo tumorgenicity and toxicology study demonstrated no evidence of tumors or abnormal hematopoiesis but did show reduced engraftment compared to control cell populations. While the drug product did engraft, the reduced engraftment highlights the importance of assuring a good cell dose when it is infused into patients in a phase I trial to reduce the probability of non-engraftment. The molecular analysis of genotoxicity demonstrated measurable off-target changes but included off-target INDELs at a single intergenic region and rare translocations between the on and single off-target site. The risk of tumorigenicity from these detectable changes is predicted to be low as the OT1 lies in an inter-genic region of no known function and there is no predictable oncogenic potential of the translocation. In human leukemias, for example, translocations into the HBB gene have never been described.

In summary, given the still tremendous unmet medical need, this pre-clinical data demonstrates a safety and efficacy profile justifying continued testing in phase I/II trials with severe sickle cell disease patients.

METHODS

AAV6 vector design, production, and purification

Adeno-associated virus, serotype 6 (AAV6) vector plasmids were cloned into the pAAV-MCS plasmid (Agilent Technologies, Santa Clara, CA, USA), comprising of inverted terminal repeats (ITRs) derived from AAV2. Gibson Assembly Mastermix (New England Biolabs, Ipswich, MA, USA) was used for the creation of the gcHBB-SCD vector as per manufacturer’s instructions. The gcHBB-SCD vector includes 2.4kb of homologous sequence to HBB with aforementioned SNP variations (20). Additionally, an AAV6 vector, designed for tracking of targeted HSPCs post-transplantation, included a bi-cistronic Nluc cDNA-T2A-TurboGFP cDNA cassette driven by SFFV with 400bp homology arms to HBB. AAV6 vectors were produced as described previously (25, 65). Final AAV6 vector was aliquoted with 0.001% F-68 pluronic acid and stored at −80 °C until further use. Additionally, gcHBB-SCD purified AAV6 vector preps were also purchased from Vigene Biosciences.

The mid-scale AAV6 lots were produced at Stanford’s Laboratory of Cell and Gene Medicine (LCGM). A production process in a cell bank HEK293T cells in suspension was developed for gcSCD-AAV6. Cells were thawed at 37°C and expanded in Freestyle F17 media supplemented with 5mM GlutaMAX and 0.2% (w/v) Pluronic F68. The cells were expanded up to a final volume of 3L in spinner flasks and co-transfected with same plasmids described above, using a polyethylenimine (PEI)-based transfection method. Two days after transfection, HEK293T cells producing gcSCD-AAV6 were harvest by centrifugation. Supernatant was removed and the cells were re-suspended in Dulbecco’s phosphate buffered saline (DPBS) prior to storing at −80±20°C. The total quantity of gcSCD-AAV6 in the harvested cells was determined by qPCR. Four in parallel replicates of 3L production were defined as one batch. All replicates were pooled prior to the cell lysis. The cell suspension underwent chemical lysis using Triton™ X-100 and clarification by depth filtration. gcSCD-AAV6 was then purified with an affinity column chromatography (AVB sepharose resin) followed by ultrafiltration on a tangential flow filtration system. During the final ultrafiltration step, the product was exchanged into the final formulation buffer, DPBS with 0.001% Pluronic F68. The product was filtered with a sterilizing grade filter (Fig. S10). In-process testing was performed on the eluate fractions from the purification, including silver stain SDS-PAGE analysis, bioburden analysis, and viral genomic titer determination by qPCR. The final gcSCD-AAV6 products were an average titer of 8×10¹¹ vg/mL.

All gcSCD-AAV6 lots were stored at −80±20°C until use. All AAV6 vectors were further tittered using ddPCR to measure number of vector genomes as previously described (66).

In vitro culture of CD34⁺ HSPCs

Human CD34⁺ HSPCs were cultured in small-scale conditions as previously described (67, 20, 25, 68, 69). Frozen purified CD34⁺ HSPCs were sourced from plerixafor- and/or G-CSF-mobilized peripheral blood (AllCells, Alameda, CA, USA and STEMCELL Technologies, Vancouver, Canada). Frozen purified CD34⁺ HSPCs from plerixafor-mobilized peripheral blood of patients with SCD were kindly provided by Dr. John F. Tisdale, National Institutes of Health, Bethesda, MD. CD34⁺ HSPCs were cultured at 1.0×10⁵ - 2.5×10⁵ – 5×10⁵ - 1.0×10⁵ cells/mL (as indicated in results) in StemSpan SFEM II (STEMCELL Technologies, Vancouver, Canada) or GMP Stem Cell Growth Medium (SCGM, CellGenix®, Freiburg, Germany) supplemented with 4 human cytokines (Peprotech, Rocky Hill, NJ United States) cocktail, stem cell factor (SCF) (100ng/mL), thrombopoietin (TPO) (100ng/mL), FLT3–ligand (100ng/mL), IL-6 (100ng/mL) and 20mg/mL streptomycin, and 20U/mL penicillin. StemRegenin1 (SR1, STEMCell Technologies, Vancouver, Canada) 0.75 μM and/or UM171 35nM (STEMCell Technologies, Vancouver, Canada) were added to the culture medium depending on experimental requirements (as indicated in results).

The medium-scale cell manufacture for use in the non-clinical tumorigenicity/toxicology study took place in the unclassified process development labs according to standard operating procedures (SOPs) at the Stanford-LCGM. In brief, the manufacturing process for the gcHBB-SCD drug substance started with the immunomagnetic selection of CD34+ HSPCs from G-CSF or Plerixafor mobilized apheresis products from healthy donor (AllCells and Hemacare). Prior to CD34+ selection, the apheresis product was washed with CliniMACS® (Miltenyi, Bergisch Gladbach, Germany) PBS/EDTA buffer containing 0.5% (v/v) HSA to remove platelets. This wash step was performed on the LOVO Cell Processing System. CD34+ HSPCs were then isolated by positive selection on the CliniMACS® Plus instrument (Myltenyi, Bergisch Gladbach, Germany) and were either freshly plated in culture (for manufacturing runs 1–4) or frozen before manufacturing (for manufacturing runs 5 and 6). The cell incubator conditions were 37°C, 5% CO2, and 5% O2.

Gene correction procedure

The HBB synthetic chemically modified sgRNA, purified by high-performance liquid chromatography (HPLC) were purchased from Synthego (Menlo Park, CA, USA). A large-scale production of sgRNA was provided by Agilent according to manufacturer proprietary process development (Agilent, Santa Clara, USA). The synthetic modifications comprise of the 2′-O-methyl-3′-phosphorothioate at the three terminal nucleotides of the 5′ and 3′ ends described previously (19). The target sequence for the HBB gRNA (R-02) is as follows: 5′-CTTGCCCCACAGGGCAGTAACGG-3′ (54, 70). The wild-type and HiFi Cas9 protein variants (Alt-R S.p. Cas9 Nuclease V1, V2, V3) used was purchased from Integrated DNA Technologies (Coralville, Iowa, USA) and Aldevron (Fargo, North Dakota, USA). Preceding electroporation, RNPs were complexed at a Cas9: sgRNA molar ratio of 1:2.5 at 25°C for 10min. Next, CD34⁺ cells were resuspended in P3 buffer (Lonza, Basel, Switzerland) with complexed RNPs and subsequently, electroporated using the Lonza 4D Nucleofector in (micro)Nucleocuvette™ and 4D-Nucleofector LV Unit in 1mL Nucleocuvette™ Cartridge (medium-scale) (program DZ-100). Electroporated cells were then plated at 2.5×10⁵ cells/mL in the cytokine-supplemented media. Immediately following, rAAV6 was dispensed onto cells at 2.5×10³ – 1.0×10⁴ vector genomes/cell based on titers determined by ddPCR and incubated for 24 hours. After, a medium addition or medium exchange was performed to remove any residual gcSCD-AAV6, and the CD34⁺ HSPCs were cultured for an additional 12–24 hours.

LAM-HTGTS

CD34⁺ cells expressing either WT or HiFi Cas9 and HBB gRNA were collected and isolated for genomic DNA as previously described (57). HBB locus-specific oligos for HTGTS library preparation are listed in Table S2. For each replicate, 18ug of genomic DNA (~5E+6 CD34+ cells) was used and libraries were generated, aligned on the hg19 genome build, and analyzed as previously described (71) but with the exclusion of the blocking digest step. Filtered HTGTS junctions were normalized down to the smallest replicate library and subsequently pooled for analysis (Fig. S14A). Hotspots for pooled libraries were called when the majority of replicate libraries displayed the same statistically enriched regions measured by MACS2 (72) with FDR-adjusted p-values >10⁻¹⁰.

Off-target activity analysis by rhAmpSeq

A list of target sites (on- and off-target editing) for total editing using the HBB sgRNA was compiled based on previously published information (GuideSeq/CircleSeq) and bioinformatic prediction (COSMID) tools (26, 53, 54). In total, 68 target sites (1 on-target, 67 off-target) were submitted for rhAmpSeq amplicon design on the Integrated DNA Technologies website (www.idtdna.com). Amplicons were requested based on minimum and maximum insert sizes of 120 and 200 bp, respectively, and a single pool was delivered containing rhAmpSeq sequencing primers for all 68 amplicons. Sequencing libraries were generated with the rhAmpSeq Library Kit and Illumina i5/i7 indexing primers. PCR amplicons were sequenced on an Illumina MiSeq (v2 chemistry; 2 × 150) and data demultiplexed using Picard tools v2.9 (https://github.com/broadinstitute/picard). NGS data were analyzed using a custom-built pipeline. Forward and reverse reads were merged into extended amplicons (flash v1.2.11)(73) before being aligned against the GRCh38 genomic reference (minimap2 v2.12)(74). Reads were assigned to targets in the multiplex primer pool (bedtools tags v2.25)(75) and re-aligned to the target, favoring alignment choices with INDELs near the Cas9 predicted cut site. At each target, editing was calculated as the percentage of total reads containing an INDEL within a 4 bp window of the cut site.

Hemoglobin tetramer analysis

Approximately 1×10⁶ cells differentiated erythrocytes were lysed using water equivalent to three volumes of pelleted cells. The mixture was incubated at room temperature for 15min, followed by 30s sonication. To separate lysate from erythrocyte ghosts, cells underwent centrifugation at 13,000 RPM for 5min. HPLC analysis of hemoglobin in their native form were analyzed on a weak cation-exchange PolyCAT A column (100 × 4.6-mm, 3μm, 1,000Å) (PolyLC Inc., Columbia, MD, USA) using a Shimadzu UFLC system at room temperature. Mobile phase A (MPA) consists of 20mM Bis-tris + 2mM KCN, pH 6.96. Mobile phase B (MPB) consists of 20mM Bis-tris + 2mM KCN + 200mM NaCl, pH 6.55. Clear hemolysate was diluted four times in MPA, and then 20μL was injected onto the column. A flow rate of 1.5mL/min and the following gradients were used in time (min)/%B organic solvent: (0/10%; 8/40%; 17/90%; 20/10%; 30/stop).

Methylcellulose CFU assessment

HSPCs were stained with CD34 APC (561; BioLegend, San Diego, CA, USA), Ghost Dye Red 780 (Tonbo Biosciences, San Diego, CA, USA) and live CD34⁺ were single cell sorted into 96-well plates containing MethoCult Optimum (STEMCELL Technologies, Vancouver, Canada). After 2 weeks, colonies were appropriately scored based on external appearance in a blinded manner. Subsequently, single cell colonies were individually harvested and genotyped as mentioned previously for targeted allelic analysis (20, 21).

CD34⁺ HSPC transplantation into immunodeficient NSG mice

Six- to eight-week-old NSG mice (Jackson Laboratory, Bar Harbor, ME, USA) were irradiated using 200rads of radiation 24hr prior to transplantation with electroporated HSPCs (48hr post-electroporation) via intra-femoral or tail-vein injections. Approximately 2.5×10⁵ - 4.3×10⁶ electroporated HSPCs (exact number noted in figures) were injected using an insulin syringe with a 27G, 0.5-inch (12.7mm) needle. This experimental protocol was approved by Stanford University’s Administrative Panel on Laboratory Animal Care.

Statistical analysis

All statistical tests on experimental groups were done using Prism7 GraphPad Software. The exact statistical tests used for each comparison are noted in the figure legends. For making multiple comparisons we used one-way ANOVA Kruskall-Wallis multiple-comparisons test with one variable and two-way ANOVA with Bonferroni post-test in case of two variables. For comparing the average mean of two sample groups, we used the unpaired Student’s t-test to reject the null hypothesis (P < 0.05).