Preclinical lentiviral hematopoietic stem cell gene therapy corrects Pompe disease-related muscle and neurological manifestations

Abstract

Pompe disease, a rare genetic neuromuscular disorder, is caused by a deficiency of acid alpha-glucosidase (GAA), leading to an accumulation of glycogen in lysosomes, and resulting in the progressive development of muscle weakness. The current standard treatment, enzyme replacement therapy (ERT), is not curative and has limitations such as poor penetration into skeletal muscle and both the central and peripheral nervous systems, a risk of immune responses against the recombinant enzyme, and the requirement for high doses and frequent infusions. To overcome these limitations, lentiviral vector-mediated hematopoietic stem and progenitor cell (HSPC) gene therapy has been proposed as a next-generation approach for treating Pompe disease. This study demonstrates the potential of lentiviral HSPC gene therapy to reverse the pathological effects of Pompe disease in a preclinical mouse model. It includes a comprehensive safety assessment via integration site analysis, along with single-cell RNA sequencing analysis of central nervous tissue samples to gain insights into the underlying mechanisms of phenotype correction.

Graphical abstract

Lentiviral vector-mediated hematopoietic stem and progenitor cell gene therapy, using an insulin-like growth factor 2 tag with R37A mutein, demonstrated long-term reversal of pathological effects in central nervous system, muscles, and improvement of locomotor function, including safety by integrations site assessment and single-cell RNA-sequencing analysis of brain.

Introduction

Pompe disease is a hereditary metabolic myopathy caused by a deficiency of the acid alpha-glucosidase (GAA) enzyme.^1,2 It is characterized by the accumulation of glycogen in lysosomes, primarily in the heart, skeletal muscles, and central nervous system (CNS).^1,2,3,4 Infantile-onset Pompe disease (IOPD) patients have very low levels of GAA activity and experience severe muscle weakness. On the other hand, late-onset Pompe disease (LOPD) patients have higher residual GAA activity, leading to a slower disease progression, but often become dependent on wheelchairs and ventilators, with a reduced life expectancy.

The current standard of care (SOC) for Pompe disease is enzyme replacement therapy (ERT), which involves regular administration of recombinant human GAA (rhGAA).^5,6,7 While ERT can extend the lives of Pompe patients, it does not guarantee long-term symptom-free survival. The effectiveness of ERT is limited by several factors, such as poor penetration of rhGAA into affected muscle cells, low levels of mannose-6-phosphate (M6P) required for efficient cellular uptake, and abnormal M6P trafficking.

Alternative developments to treat Pompe disease include adeno-associated virus (AAV) gene therapy, as well as allogeneic hematopoietic stem and progenitor cell (HSPC) therapy, which utilizes the hematopoietic system to produce rhGAA enzyme.² Unlike ERT, allogeneic HSPC transplantation has shown potential in promoting immune tolerance induction, which has shown to enhance the efficacy of recombinant GAA enzyme infusions in preclinical studies.^8,9,10 Additionally, liver-directed AAV gene therapy has also shown promising results in immune tolerance induction to rhGAA.¹¹ However, AAV serotypes targeting specific tissues, such as the liver, may not effectively deliver the transgene product to other affected areas, such as the CNS.

Lentiviral-mediated HSPC gene therapy holds promise in treating (neuro)metabolic diseases. To enhance the efficacy of HSPC gene therapy, we employed a lentiviral vector containing a glycosylation-independent lysosomal targeting (GILT) tag sequence derived from the insulin-like growth factor 2 (IGF2) peptide, which specifically binds with high affinity to the IGF2 receptor (IGF2R) fused to a truncated catalytic domain of GAA. As we and others have previously demonstrated, this fusion peptide improves enzyme secretion and enhances the reduction of glycogen, myofiber, and CNS vacuolation in critical tissues.^12,13,14 Additionally, the GILT-tagged vector sequence included a R37A mutein, which has been proven to significantly diminish insulin receptor signaling mediated by the IGF2-tag.¹⁴

In this study, we have demonstrated consistent and elevated therapeutic enzyme activity, leading to complete or nearly complete reduction in the accumulation of disease-related substances in critical target tissues, including the CNS. Our analysis of integration sites revealed typical patterns associated with lentiviral vectors, without any indications of clonal expansions or biased selection of oncogenic sites. In addition, by performing single-cell RNA sequencing of microglial cells that were isolated using fluorescence-activated cell sorting (FACS), we observed that our therapeutic approach was capable of restoring these cells to a state similar to wild-type (WT) cells when compared with untreated glial cells or a non-GILT-GAA vector. In summary, these findings underscore the potential of lentiviral HSPC gene therapy as a promising strategy for the treatment of Pompe disease, addressing the limitations of current therapeutic options.

Results

MND promoter provides robust expression compared with housekeeping gene promoters

The myeloproliferative sarcoma virus enhancer, negative control region deleted, and dl587rev primer-binding site substituted (MND) promoter was selected for its robust expression, because for Pompe disease, high levels of recombinant protein expression are required as compared with other lysosomal storage diseases, such as Fabry or Gaucher disease, to achieve therapeutic responses.¹³ In a first study (study 1; Figure S1), we used green fluorescent protein (GFP) vectors containing the MND promoter, and assessed the robustness of GFP fluorescent signal compared with housekeeping gene promoters commonly used in clinical trials, the phosphoglycerate kinase (PGK) and elongation factor 1 alpha short (EFS).^15,16 To evaluate the strength of these different promoter sequences in vivo, peripheral blood, bone marrow, thymus, and spleen, but also microglia-like cells in the brain were collected and measured by flow cytometry. The MND promoter clearly showed the highest level of GFP mean fluorescence intensity in all hematopoietic tissues, compared with PGK and EFS promoters (Figures S2A–S2D). In these groups, the vector copy number (VCN) values in bone marrow were similar between groups (Figure S2E). In the CNS tissue, GFP percentage was more clearly detectable in the the MND promoter group, than the PGK or EFS groups (Figure S2F). The robust expression of the MND promoter in microglia-like cells, was ∼3.9-fold higher in GFP mean fluorescence intensity compared with EFS and ∼2.4-fold to PGK promoter (Figure S2G). VCN measured in microglia-like cells was also similar among the three groups (Figure S2H).

Hematopoietic reconstitution of genetically modified stem cells

In a previous study, a codon-optimized GILT-R37A-tagged GAA transgene (subsequently called GILT) was selected as a lead vector.¹⁴ This vector contained the MND promoter to drive GILT expression. A control vector containing GFP was included as a reference for hematopoietic reconstitution compared with the therapeutic vector.

We performed a combined (“hybrid”) preclinical long-term efficacy and safety study using the GILT-vector. The study design and lentiviral vectors used are shown in Figure S1 (study 2). In this study we assessed long-term expression and efficacy in Gaa^−/− mice by a lentiviral vector dose titration on Gaa^−/− Lin cells (multiplicity of infection [MOI]: 0.75, 1.5, and 3), which were transplanted in both 9- to 12-week-old 7.5-Gy irradiated male and female Gaa^−/− recipients.

In the pretransplant Lin cells, the VCN consistently increased in both bulk male and female cell cultures (Table S1). To track donor cells after transplantation, male and female cells were transplanted in the opposite sex, as done before.^9,13,14,17 Notably, the MOI 3 condition in the female Lin-bone marrow cells resulted in fewer colony-forming units (CFU), and after lineage enrichment, the percentage of Lin-Sca-1+c-Kit+ cells was higher in the male bone marrow cells (14.68% ± 0.58, n = 6) than the female bone marrow cells (9.53% ± 0.71, n = 6). Finally, the bodyweight of the male mice was higher compared with the female mice (Table S1).

In the gene therapy-treated Gaa^−/− mice, supraphysiological plasma GAA activity levels were sustained up to the end of the study (week 31; Figures 1A and 1B). In Gaa^+/+ control plasma, there was no specific GAA activity signal detected as reported before (data not shown).¹⁴ Peripheral blood white blood cell (WBC) GAA activity was generally increased in the higher vector dose groups (MOI 1.5 and 3) compared with the lower MOI group (MOI 0.75) at weeks 4, 16, and 31 (Figure S3A), as well as the GAA activity at termination in bone marrow at termination week 32 (Figure 1C).

Figure 1.

Hematopoietic reconstitution

(A) Specific enzyme activity in plasma at weeks 4, 16, and 31. Background signal detected in Gaa^−/− mice was subtracted from enzyme activity measured to show specific activity. (B) Enzyme activity in WBCs at week 31. (C) Enzyme activity in bone marrow at 8 months after transplantation. (D) VCN/diploid genome in bone marrow determined by qPCR. (E) Chimerism based on nuclear detection of lentiviral vectors using DNAscope. (F) VCN quantified by nuclear vector-positive stained dots per cell using DNAscope. Bone marrow was collected 8 months after transplantation (n = 9–13). Statistical analysis; (B)–(F): individual values, group medians, and interquartile ranges shown. Exact Wilcoxon Rank Sum p values for group comparisons; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Peripheral blood WBC = white blood cell. WT, wildtype; KO, knockout; MOI, multiplicity of infection; and UT, untransduced.

Using qPCR to determine VCN, WBC (male groups MOI 0.75–3, medians per group 1.9–3.8 VCN and female groups 2.1–5 at week 4 and 1.9–3.3 and 1.8–3.4 at week 28, respectively; Figure S3B) showed similar patterns as the VCN in bone marrow samples (male treatment groups medians VCN 2.1–3.6 and females 1.9–3.6) (Figure 1D). The VCN in the WBC in the male group (MOI 3) was consistently lower than the group (MOI 1.5) in WBC over time (Figure S3B). This was also observed in the bone marrow (Figure 1D). The bone marrow VCN in the male and female GFP groups was 2.9 and 1.8, respectively (Figure 1D).

Using in situ hybridization assay (DNAscope) to detect single nuclear lentiviral vector integrations, quantification showed a partial chimerism of vector-positive bone marrow cells with a range of medians of 46%–67% in males and 40%–66% in females (Figure 1E). The chimerism in the male GILT group with MOI 3 was lower than the male GILT group with MOI 1.5. Quantification of the number of positive vector integrations per cell by DNAscope, showed that average nuclear single dots per bone marrow cell was slightly lower than quantified by qPCR (male treatment groups: medians 1.3–2.1 and female treatment groups 1.1–2.2; Figure 1F). The average VCN in the transduced bone marrow cells <5, and the distribution of vector copies per cell similarly distributed between treatment groups and the GFP control group (Figure S3C). The majority of vector-positive cells contained a VCN of 1 (mean 33%–49%), and cells with >5 VCN ranged from 9.2%–13.9% in males and 10.6%–15.6% in females.

The GILT treatment group (MOI 3) was used to determine VCN in peripheral tissues, and compared with the GFP group (MOI 1.5). In spleen and thymus (VCN medians 2.8–3.3), VCN was similar to WBC and bone marrow in the GILT treatment group (Figure S3D). In non-hematopoietic tissues lung, liver, and kidney VCN were ∼3.5–17-fold lower. In reproductive organs testis and ovaries low VCNs were detected of 0.036 and 0.44, respectively (Figure S3D). VCN in the brain of treated mice was low in the GILT group (MOI 3; medians VCN in male mice: 0.034 and in female mice: 0.04), similar to the GFP group (MOI 1.5; medians VCN in male: 0.026 and in female mice: 0.02; Figure S3D), which showed the low engraftment of gene-modified cells in the brain (Figure S3D).

The fact that we transplanted sex mismatched Lin cells into opposite sex recipients may have contributed to a lower average VCN and chimerism levels in the hematopoietic compartment in the male GILT group MOI 3 compared with the MOI 1.5 group. However, all the transplanted mice in this study reconstituted genetically modified Lin cells, and VCNs were in a clinically acceptable range.

Complete correction of heart parameters

In study 2 (Figure S1), heart parameters were measured at month 7 in gene therapy-treated male and female mice. Heart GAA activity was restored to normal levels in WT mice (WT median values were 3.8 in male and 4.5 in female Gaa^+/+ mice) and ranged from low MOI to high MOI 3.22–5.45 nmol/h/mg (male) and 3.1–7.7 nmol/h/mg (female) in the treatment groups (Figure 2A).

Figure 2.

Cardiac function restoration

(A) GAA enzyme activity, and (B) glycogen in heart. Heart tissue was collected 8 months after transplantation (n = 9–13). (C) Heart weights, (D) left ventricle (LV) mass, (E) LV mass index, (F) ejection fraction, (G) fractional shortening, (H) and isovolumic contraction time (IVCT). Analysis was performed 7 months after transplantation (n = 9–13). Statistical analysis; (A)–(H): individual values, group medians, and interquartile ranges shown. Exact Wilcoxon Rank Sum p values for group comparisons; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. WT, wildtype; KO, knockout; and MOI, multiplicity of infection.

Consequently, biochemical glycogen assessment showed complete elimination to undetectable levels (Figure 2B), which was confirmed by periodic acid-Schiff (PAS) quantification (Figures S4A and S4B). As a consequence, vacuolation was also markedly reduced compared with male Gaa^−/− mice, which all had severity scores of 2, while all treated mice were restored to nomal levels in the MOI 3 group, with >80% of the mice in the lower MOI groups restored to WT levels (Figure S5). In the female groups, ∼70% had a severity score of 3, and >80% of mice had no detectable tissue pathology after treatment (Figure S5).

Cardiac remodeling and contractile function were affected in Pompe mice as previously described.⁹ In our study, GILT-vector treatment resulted in normalization of heart weight (Figure 2C). Left ventricle (LV) mass and index were also determined by cardiac echography, and also showed reduction to normal values (Figures 2D and 2E). As for functional analysis, ejection fraction and fractional shortening were also significantly reduced after GILT-vector treatment in all groups (Figures 2F and 2G), an indication of correction of hypertrophic cardiomyopathy. In addition, isovolumic contraction time (IVCT) was reduced in male recipient Gaa^−/− mice (Figure 2H). Other cardiac parameters, such as LVVs, LVVd, LVPWs, LVPWd, IVSs, IVDd, LVIDs and LVIDd (abbreviations in Table S2), which showed changes in size, mass geometry of the heart in Pompe mice, were improved and restored to normal ranges in GILT-treated mice (Figures S6A–S6H).

A low VCN was sufficient to drastically restore most cardiac parameters confirming the susceptibility to remodeling of the heart after GILT-vector treatment.

Correction of skeletal muscle pathology and improvement of locomotor function

In study 2 (Figure S1), the skeletal muscles diaphragm, as well as the hindleg muscles quadriceps, and gastrocnemius in all treatment groups consisted of higher GAA enzyme activities than the WT mice, i.e., in males this was 1.3–7.2-fold higher and females 1.3–7.9-fold higher (Figures 3A–3D). In the tibialis only the highest MOI 3 resulted in a 2.3-fold increase in males and 1.9-fold increase in females. Consequently, glycogen in diaphragm was reduced >98% in Gaa^−/− treated male mice and >96% in Gaa^−/− treated female mice compared with Gaa^−/− mice (Figure 3E). In the hindleg muscles in Gaa^−/− treated male mice, quadriceps, gastrocnemius (both >92%) and tibialis (>95%), glycogen was also significantly reduced (Figures 3F–3H). In Gaa^−/− treated female mice, quadriceps, gastrocnemius (both >85%), and tibialis (>93%) all showed significant reduction in glycogen levels.

Figure 3.

Skeletal muscle rescue

GAA activity in skeletal muscle diaphragm (A), quadriceps (B), gastrocnemius (C), and tibialis muscle (D). Glycogen in muscle diaphragm (E), quadriceps (F), gastrocnemius (G), and tibialis muscle (H) at month 8 after transplantation (n = 9–13). Functional correction showing gait analysis (I) and wire hang at month 7 (J). Statistical analysis; (A)–(H): individual values, group medians, and interquartile ranges shown. Exact Wilcoxon Rank Sum p values for group comparisons; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. WT, wildtype; KO, knockout; and MOI, multiplicity of infection.

Quantification of PAS staining confirmed the glycogen content measured by the biochemical analysis of skeletal muscle lysates (Figure S4). Vacuolation scores in the diaphragm of both male and female gene therapy treatment groups were all lower than the Gaa^−/− group, mostly with severity scores of 0 (Figure S5). In the hindleg muscles, Gaa^−/− mice always showed the most severe scores, while the gene therapy groups consistently demonstrated correction of skeletal muscle pathology, with higher MOI groups trending toward lower severity scores.

Locomoter function was improved in the gene therapy groups, as assessed by gait analysis and wire hang (Figures 3I and 3J). The effects on gait were more pronounced in the female cohort (Figure 3I). Latency to fall was significantly increased in all treated mouse groups, however not fully restored to latency levels seen in WT mice (Figure 3J).

Correction of CNS pathology

In study 2 (Figure S1), the median values in GAA enzyme activity in WT mice was 5.0 in cerebrum, 5.4 in cerebellum, and 8.2 in spinal cord in males (Figures 4A–4C). In the females this was and 3.3 in cerebrum, 4.6 in cerebellum, and 6.5 in spinal cord (Figures 4A–4C). The robust expression of using the MND promoter, in conjunction with the GILT-tag, showed significant increases of GAA activity in these CNS tissues, but median values were lower than WT GAA enzyme activity levels, e.g., in cerebrum (males: 10%–13%, females: 19%–28%), in cerebellum (males: 6%–10%, females: 7%–15%), spinal cord (males: 12%–28%, females: 22%–28%; Figures 4A–4C).

Figure 4.

Restoration of function in the CNS

GAA enzyme activity in CNS: (A) cerebrum, (B) cerebellum, and (C) thoracic spinal cord. Reduction of glycogen in (D) cerebrum, (E) cerebellum, and (F) thoracic spinal cord. Tissues were analyzed at 8 months after transplantation (n = 9–13). Exact Wilcoxon Rank Sum p values for group comparisons; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. WT, wildtype; KO, knockout; and MOI, multiplicity of infection.

Nonetheless, biochemical analysis of glycogen showed more than 99% reduction in cerebrum, cerebellum, and spinal cord compared with Gaa^−/− mice (Figures 4D–4F). Quantification of PAS staining signal confirmed the reduction in glycogen content in the CNS, as a similar reduction was observed in cerebrum, cerebellum, hippocampus, and spinal cord (Figures S4A and S4B), as well as sciatic nerve (Figure S4B). CNS tissue pathology was clearly detectable in the Gaa^−/− group with abundant cytoplasmic vacuoles in visible neurons with a maximum severity grading in cerebrum, cerebellum, and spinal cord gray matter (Figure S5). Remarkably, almost all gene therapy-treated groups, including the low MOI 0.75 group, showed complete resolution of cytoplasmic vacuoles in neurons (Figure S5).

Brain VCN engraftment levels were 72- to 84-fold lower than in hematopoietic tissue, such as spleen, but sufficient for almost complete elimination of glycogen in CNS tissue.

Genomic detection of integrated payloads and integration site analysis

To assess genotoxicity we performed integration site analysis (ISA)¹⁸ using an established protocol applying our computational pipeline¹⁹ to genomic bone marrow samples. Samples were selected from two studies (studies 2 and 3; Figure S1).

First, samples were collected from male and female Gaa^−/− mice from the long-term 32-week study to have a late time point after hematopoietic reconstitution in which the mice were conditioned with irradiation before infusion of transduced Lin-bone marrow cells (study 2; Figure S1). For this 32-week study, VCN per diploid genome of in vivo bone marrow samples selected for ISA from the GILT group ranged from 3.05–5.61 (data not shown). For GFP-vector group bone marrow samples selected for ISA, the range of VCN was 1.90–5.33 copies VCN per diploid genome (study 2, Figure S1).

In study 3 using the GILT-vector (Figure S1), Busulfex-conditioned mice were used, implementing the reagent that will be used in clinical setting to create space in the bone marrow for efficient engraftment. This study was terminated at 16 weeks after infusion of the drug product, and included GILT and GFP-vector treated groups, and both male and female Gaa^−/− mice.

Characterization of the transduction efficiency of the lentiviral vector lots used for these study is shown in Figures S7A–S7C, as well as the GAA activity measured in WBC and plasma at 4 and 12 weeks, in spleen and thymus at 8 (intermediate termination) and 16 weeks (Figure S7D). Hematopoietic reconstitution resulted in restoration of GAA activity in diaphragm heart and quadriceps (Figure S7E). Also, VCNs in irradiated mice were similar in hematopoietic tissues spleen and thymus were, as expected, higher than in non-hematopoietic tissues (Figures S8A and S8B). Digital droplet PCR (ddPCR) was used to determine the VCN of in vitro samples (range 4.18–4.68 copies/genome) and of in vivo test samples from group 2 (range 3.36–8.71 copies/genome) (data not shown).

At takedown, ISA analysis in bone marrow gDNA of GILT mice conditioned with either irradiation or Busulfex showed a polyclonal insertion site pattern as measured by the number of insertion sites retrieved (Figure S9) and Shannon diversity (Figures 5A–5C, S10A, and S10B). Results were similar between the GILT and GFP control group with no sign of aberrant clonal expansions in any of the mice analyzed (i.e., no integration in the GILT group contributed with more than 7.2% or 6.32% to the total sequence pool for the irradiated mice or Busulfex-conditioned mice, respectively). For both the irradiation and Busulfex experiments, we did not observe any overlap among the top 10 most abundant integration loci observed in each animal, suggesting that there was no obvious vector-driven selective growth associated to a specific gene locus (Figure S9). One animal had an integration inside Kansl1l, contributing 20.84% to the overall sequence reads pool. Additionally, in the Busulfex experiment, the only reoccurring gene in the vicinity of a top 10 integration was Prkd1 for two animals. Of note, these are not the same chromosomal positions, but only the same gene. When looking at the insertion profile, we observed in both groups higher representation of integrations in a distance of 100–1,000 kb relative to CpG islands, marking actively transcribed regions but not in the direct vicinity of CpG islands (1–10 kb) (Figure S10C). The vector generally disfavored GC-rich regions (100 bp – 10 kb, marking promoter regions of genes), longer gene widths or intergenic regions. These observations are in line with the classical insertional preference for gene bodies and underrepresentation of integrations in promoter regions described for lentiviral vectors.²⁰ In addition, no selection for proto-oncogenic integration sites was observed in vivo in either of the groups (Figure 5D).

Figure 5.

Integration site analysis

(A) Shannon diversity index and number of insertions collected at 8 months after transplantation in each mouse in the GILT (upper panel) or (B) GFP (lower panel) group from the study using irradiation conditioning. (C) Comparison of Shannon diversity indexes calculated on insertion sites collected from the GILT (blue bars) vs. GFP (gray bars) group (n = 4–10). (D) Percent of vector insertions detected in proto-oncogenes in the GILT group (blue bar), GFP group (gray bar) and a randomly generated insertion site dataset (white bar) (n = 4–10). Mann-Whitney U p values for group comparisons; ∗p < 0.05, ∗∗p < 0.01. ns, not significant.

Overall, these data suggest that regardless of the transgene and conditioning system used, our lentiviral vector displayed a polyclonal and typical insertional profile in vitro and in vivo.

Furthermore, because we used the IGF2-tag, which could potentially bind the insulin receptor, we monitored the glucose levels in the study 2 groups (Figure S1) up to 6 months after transplantation of genetically modified cells. No significant changes were observed over time in both the male and female treatment groups compared with control Gaa^−/− mice (Figure S10D).

Brain single-cell transcriptional profiling

In a final study 4 (Figure S1), we further investigated the ability of using GILT-tag to achieve phenotypic correction of microglia in the brain. The MND promoter was replaced by another strong viral promoter, the spleen focus forming virus (SFFV) regulatory elements, which are comparable in promoter strength and have shown similar therapeutic outcome in muscle.¹³ In this study, lentiviral vectors with the SFFV promoter to drive GILT-tag or GAA (SFFV.GILT or SFFV.GAAco vector; lentiviral vectors depicted in Figure S1; study 4) were used to transduce Lin cells, and subsequently infused in Busulfex-treated mice.

VCNs of Lin cells, as well as in peripheral blood, bone marrow, and brain are shown in Figures S11A and S11B. Immunohistochemistry for GAA protein revealed presence in the brain, but with distinct morphological patterns (Figure S11C). The GAA protein in the GAAco group was much more pronounced, compared with the GILT group, which may be caused by the reduced ability of GAA protein to be secreted and subsequent cellular retainment as proposed by Dogan et al.¹⁴ In case of microglia cells, cellular retainment of the transgene product, may correct the transduced microglia-like cell, but not all of these cells are transduced, and cross-correction of non-transduced microglia-like cells may still be required.

At 32 weeks after infusion, FACS-sorted CD45+CD11b+CXCR3+ microglia cells from the brain of the GILT vs. GAAco-treated groups (two mice per group, two samples per mouse) were analyzed by single-cell RNA sequencing (scRNA-seq) (10× genomics). Immunophenotyping of single-cell suspensions of glial cells showed an overall decrease in the percentage of microglia and increase in the proportion of astrocytes in untreated Gaa^−/− and in GAAco-treated animals (Figures 6A and S11D). In contrast, the percentage of microglial cells and astrocytes was restored to WT levels in the GILT treatment group (Figure 6A). Endothelial cells and neurons were not affected (Figure S11E). Astrogliosis has been previously described in the Pompe mouse model used in these studies.²¹ In line with previous reports, we observed a higher percentage of astrocytes in nontreated Gaa^−/− and GAAco-treated animals, compared with WT animals. GILT-treated mice had comparable percentages of astrocytes relative to WT animals. Microglial cells were sorted based on CD45, CX3CR1, and CD11b expression, and prepared for scRNA-seq analysis (Figure S11D). We generated the combined single-cell transcriptional map (Figure 6B) where we identified five main clusters based on transcriptional similarities (see materials and methods), which we annotated based on the differentially expressed genes reported in Table S3. Notably, the microglia from GILT and GAAco mapped separately. The microglia from the GILT-treated mice displayed an enrichment of clusters 1, 2, and 3 (generally corresponding to a more homeostatic defined signature) while the microglia from the GAAco-treated group was highly concentrated in clusters 4 and 5 (reminiscent of disease-associated microglia signatures previously reported)^22,23,24,25 (Figure 6C). We observed higher expression of ribosomal genes in the microglia from GAAco while the GILT mice showed a statistically higher expression of genes normally associated with MHC complex expression, such as H2Eb1, H2-Ab1, Cd74, and H2-A (Figure 6D). This phenotype was previously reported to be linked to border-associated macrophages which are generally yolk-sac-derived macrophages localized near the blood-brain barrier (BBB).^25,26 When looking for vector LTR presence in the mRNA of the single cells isolated from each experimental group, we could clearly observe a higher density of LTR+ cells in the microglia isolated from the GILT mice (Figure 6E), where ∼33% of the microglia were detected as vector positive (VCN 5.8, n = 2), vs. 2.1% found in the GAAco group (VCN 2.4, n = 2). This near 16-fold higher level of LTR-positive cells may indicate a preferential survival advantage for GILT-GAAco-corrected microglia, which, as far as we know, has not been observed in other preclinical lysosomal storage disorders.

Figure 6.

Single-cell RNA sequencing profile of microglial cells

(A) Percentage of microglial cells and astrocytes gated on live cells, measured via flow cytometry from enzymatically digested brain samples from mice 8 months after transplantation (n = 4–9). (B) UMAP plot generated from a total of 53,421 single cells (n = 4 mice, two mice per group) clustered based on transcriptional similarities. Clusters are annotated on the top right based on their specific gene signatures listed in Table S3.

(C) UMAP plots showing single-cell distribution in the GILT-treated mice (left plot) or GAAco-treated mice (right plot). Pie charts on the top right of each plot show the relative frequency of cells belonging to each cluster in the two groups. Percentages are reported for the most represented clusters in each group. (D) Bar plots displaying the average expression of ribosomal genes and homeostatic microglia-associated genes in microglial cells from GILT and GAAco-treated mice. (E) Density plots showing the distribution of LTR+ (in red) and LTR− (in black) cells. The bar plot on the right shows the percent of LTR-positive (LTRpos) and LTR-negative (LTRneg) in the GILT (dark gray bars) and GAAco (light gray bars) groups. (F) Density plots showing the distribution of all single cells in the map (in gray) vs. the single cells expressing the WT microglia signature reported in Figure S8D. WT microglia signature expression scores calculated in the GILT vs. GAAco group are reported in the violin plot on the right. (Mann-Whitney U p values for group comparisons; ∗∗∗∗p < 0.0001). WT, wildtype and KO, knockout.

Accordingly, analyzing the transcriptional profile of LTR+ vs. LTR cells (Figure S12A) we could again identify gene signatures of activated microglia in the LTR cells (more enriched in the GAAco group), while the LTR+ cells (more enriched in the GILT mice) showed a transcriptome more typical of homeostatic microglia. Based on these observations, we wanted to formally confirm that the cells isolated from the GILT mice carried a transcriptional profile more similar to normal microglia when compared with the GAAco group, as a result of a more efficient genetic correction driven by the GILT-vector.

To this aim, in a second small-scale experiment, we analyzed by scRNA-seq the microglia isolated from untreated WT B6129SF1/J control mice or Pompe mice (two mice per group), generated the corresponding single-cell map (Figures S12B and S12C) and identified a list of genes specifically expressed by WT microglia (Figure S12D). When projecting this signature onto the GILT vs. GAAco map we could clearly observe that the WT signature was highly enriched in the area of the map where the GILT cells were localized (Figure 6F). Accordingly, the WT signature score was statistically higher in the GILT vs. the GAAco-treated mice suggesting that the GILT cluster contained a higher frequency of cells with a physiological microglia phenotype as compared with the GAAco group (Figure 6F).

Discussion

In this study, long-term efficacy and safety were observed in a mouse model of Pompe disease using a GILT containing lentiviral construct at low VCN. Heart parameters, which are relatively easy to normalize, compared with skeletal muscle, were shown to respond well to the therapeutic approach of using HSPC with GILT technology in both male and female Pompe mice. Skeletal muscle involvement in Pompe disease is critical to improve disease pathology, but is more difficult to achieve with SOC. In our lentiviral approach, skeletal muscle parameters responded well after long-term gene therapy exposure, similar to what has also been observed in liver-directed AAV gene therapy.²⁷ In addition, novel capsids generated by library selection, such as AAVB1, may address both muscular and neuronal deficits in Pompe disease more efficiently compared with classical serotypes such as AAV9.²⁸

CNS involvement and pathology and its contribution to disease phenotype has become more apparent in recent years, hence Pompe disease is generally considered a neuromuscular disorder. There are numerous reports that the CNS and peripheral nervous system (PNS) are associated with white matter abnormalities, cognitive impairment, and respiratory deficits. Subsequently, the neurological component plays a critical role in disease progression, which is also not addressed by SOC.^{3,4,29,30,31,32,33} Both the CNS and phrenic nerve dysfunction contributes to the respiratory insufficiency and failure, the most common cause of death in untreated IOPD and LOPD patients.³⁴ Moreover, neuropathology has also been investigated in Pompe mice, and double Gaa^−/− transgenic mice expressing hGAA in skeletal muscle showed that skeletal muscle is not the sole contributor to ventilatory dysfunction.^35,36 Furthermore, other investigators specifically investigated correction of neuropathology and respiratory dysfunction in Pompe mice.^27,37,38 An advantage of the followed approach using HSPC transplantation is that biodistribution of genetically modified cells is broad throughout the body and includes effective delivery to the CNS.³⁹ Furthermore, the MND promoter showed robust expression in microglia-like cells in the brain compared with two commonly used housekeeping gene promoters for hematopoietic stem cell (HSC) gene therapy, the EFS and PGK promoter. However, we only compared expression differences between the promoter variants with GFP as a model transgene. We did not investigate whether similar therapeutic effects could be obtained in the Gaa^−/− mice using EFS- or PGK-driven constructs in this study. Nevertheless, it is known from clinical application of ERT, that much higher doses of recombinant protein are required in Pompe disease patients than in Fabry or Gaucher patients, i.e., ∼40× higher to reach therapeutic efficacy, especially in targeting the skeletal muscle.²

Furthermore, in preclinical studies, in which the EFS promoter was used in combination with β-globin enhancer elements to enhance expression in erythroid cells, as an alternative to using a strong viral promoter, results showed partial correction of the heart, but skeletal muscle was resistant to therapy.⁴⁰ Another study assessed the use of the PGK promoter with codon-optimized GAA and showed no correction of locomotor tests, and no glycogen reduction in skeletal muscles and heart at high vector copy numbers. The results were inferior to using stronger viral promoters, such as MND and SFFV promoter, supporting the challenge of delivering sufficient transgene product to the required target tissues with physiological promoters.⁴¹ Nevertheless, the approach was not tested with the GILT-tag, and this would be of interest to pursue in follow-up studies.

Besides, in another study in which a lentiviral vector with codon-optimized GAA was driven by the MND promoter, CNS pathology was not resolved.¹⁴ Hence, usage of GILT-tag may improve local delivery of the therapeutic protein, but tag technology may also enhance delivery through the BBB.^13,42,43 We did not observe a major difference in VCN in hematopoietic tissues, CNS, and other organs comparing GFP with GILT-vector. Interestingly, isolation of microglia-like cells showed that GILT-treated mice had a major increase in the contribution of genetically modified cells. Of note, both astrocytes and microglia-like cell distribution appeared affected in Pompe mice, and the GAAco construct was unable to rescue CNS pathology, as well as microglia activation, which was normalized in GILT-vector treated mice.

Lentiviral HSC gene therapy has shown polyclonal integration profiles in metabolic diseases.^20,44,45,46 In preclinical models for Pompe disease, lentiviral vectors with the strong SFFV promoter showed no significantly increased risk of in integration frequency near oncogenes compared with eukaryotic promoters.⁹ However, in one mouse in our study, a vector integration into Kansl1l contributed to 20.84% of the overall sequence reads. Importantly, this integration was found in gDNA of bone marrow of a mouse in the irradiation experiment and GFP-vector group. As known from previous literature,^47,48,49,50 a certain degree of clonal selection is expected in these experimental models as a result of a stochastic selection due to the conditioning regimen combined with transplantation of Lin-bone marrow cells. This animal did not show any specific sign of growth defects, or aberrant immune profile/blood cell counts as compared with the animals hosting a more diverse insertional profile. Therefore a clonal dominance should be interpreted differently in this group vs. one occurring in the mouse group treated with Busulfex, which represents an experimental setting closer to clinical translation, and no such occurrences were detected. Kansl1l is also not a known oncogene and has no relation to leukemogenesis.

In clinical trials, a gamma retroviral vector with adenosine deaminase (ADA) cDNA under control of MND enhancer/promoter sequences has been used for ADA severe combined immunodeficiency (ADA-SCID). This demonstrated long-term enduring efficacy, without developing malignancies, but also highlighted risks of genotoxicity with gammaretroviral vectors, indicated by numerous common integration sites near proto-oncogenes and by increased abundance of clones with integrations near MECOM and LMO2.^51,52 These clones showed stable behavior over multiple years and never expanded to the point of dominance or dysplasia.⁵² In addition, the Food and Drug Administration approved the product elivaldogene autotemcel (eli-cel or Skysona) for cerebral adrenoleukodystrophy (C-ALD), which uses a replication-incompetent lentiviral vector with ABCD1 cDNA under the control of an internal MND promoter to transduce CD34+ HSPCs.^53,54,55 However, three patients treated with eli-cel in the ALD-102 (ClinicalTrial.gov ID: NCT01896102) and ALD-104 (ClinicalTrial.gov ID: NCT03852498) studies, were diagnosed with myelodysplastic syndrome (MDS) at 14 months, 2 years and 7.5 years after after Skysona administration (3 out of 67 patients treated),^56,57,58 and the hematologic malignancies appear to be a result of the Skysona lentiviral vector, Lenti-D, integration in proto-oncogenes, and particularly LTR-transactivation of MECOM (https://www.fda.gov/vaccines-blood-biologics/skysona). In our study, ISA showed typical lentiviral insertion patterns in transplanted mice conditioned with sublethal irradiation as well as Busulfex conditioning. Furthermore, genotoxic events were absent during the course of the study in HPSC transplanted mice containing GFP or GILT vectors, but this may also point out limitations to the sensitivity of testing vector genotoxicity in mice, and complementary assays such as IVIM/SAGA may aid to inform potential risk profile during lentiviral vector selection for clinical programs.⁵⁹

Altogether, the ex vivo HSC gene therapy approach using the GILT-tag with R37A substitution demonstrated long-term efficacy in both CNS and skeletal muscle with a favorable safety profile in preclinical studies, at clinically relevant dose, which warrants further investigation.

Materials and methods

Plasmid construction and lentiviral vector production

Lentiviral vectors containing the myeloproliferative sarcoma virus enhancer, negative control region deleted, dl587rev primer-binding site substituted (MND) promoter to express human codon-optimized GAA (GAAco), and GILT-R37A-tagged GAA coding sequence were previously described.¹⁴ In other experiments the SFFV promoter was used for therapeutic transgene expression. Furthermore, lentiviral vectors expressing GFP were constructed with the following promoters: MND, PGK, and EFS.

All third-generation self-inactivating LV vectors described above were produced by transient transfection of 293T cells using packaging and transfer plasmids at Vectorbuilder. Titers were determined on human osteosarcoma (HOS) cells (ATCC; Cat# CRL-1543) by qPCR and were subsequently titrated on lineage negative bone marrow cells to determine the MOI for subsequent studies.

Mice

The Gaa^tm1Rabn/J mice (Gaa^−/− mice, Pompe mice) were used for the study.⁶⁰ Mice were obtained from an on-site breeding colony (kindly provided by Dr. Nina Raben, NIH, MD). Control B6129SF1/J mice (Gaa^+/+ mice, WT mice) were obtained from Jackson Laboratories (Stock No. 101043). All mice were maintained in clean rooms and fed with irradiated certified commercial chow and sterile acidified water ad libitum. Assessment of animal health status, body weight, and examinations during the in-life term of the study were conducted by veterinarian personnel and documented. All protocols were approved by the Institutional Animal Use and Care Committee at the Canadian Council on Animal Care or the Charles River Accelerator and Development Lab (CRADL), Cambridge, MA.

Assessment of dosing formulations and infusion in conditioned Gaa^−/− mice

After in vitro assessment of vector preparations, GILT-tagged GAA containing lentiviral vectors were used for in vivo testing in male and female Gaa^−/− mice mice. Bone marrow cells were harvested from femurs and tibias of 6- to 12-week-old male or female Gaa^−/− donor mice, and Lin-enriched for HPSCs using RoboSep (StemCell Technologies). After enrichment, cells were overnight transduced with each respective lentiviral vector (Figure S1), at a density of 10⁶ cells/mL in serum free StemMACS medium containing 100 ng/mL recombinant murine stem cell factor (SCF); 50 ng/mL recombinant human FMS-like tyrosine kinase 3 ligand (Flt-3) and 10 ng/mL recombinant human thrombopoietin (TPO; StemCell Technologies). After sublethal conditioning of 9- to 12-week-old Gaa^−/− mice with 7.5 Gy gamma-irradiation (1.6 Gy/min) or by 4 × 25 mg/kg Busulfex (Otsuka Pharmaceutical) injections days −4 to −1 before dosing of 5 × 10⁵ cells, the mice were injected intravenously with enriched donor Lin- HSPCs transduced with lentiviral vectors. Female and male Gaa^−/− (knockout, KO) and Gaa^+/+ (wildtype, WT) mice nontreated and GFP-vector control groups were included.

Female or male Gaa^−/− recipients that were irradiated were injected intravenously with 0.5 × 10⁶ cells per mouse from the opposite sex and monitored for 32 weeks with interim blood collections. Following dosing, Methocult preparation (CFU analysis) was performed, and remaining transduced Lin-bone marrow cells were put back in culture. In addition, cell pellets of 0.5 × 10⁶ untransduced and transduced cells were prepared for VCN analysis. Aliquots were also sent out for ISA assay. After immunomagnetic bone marrow cell isolation, cells were stained with anti-mouse Hematopoietic Lineage antibody cocktail (Invitrogen, cat #: 22-7770-72), rat-anti-mouse CD117-APC (c-Kit; cat #: 553356), rat anti-mouse Ly-6A/E-PE (Sca1; cat #: 553108) and a viability dye (FVD506) to evaluate the efficiency of the cell enrichment by flow cytometry analysis.

For Busulfex-conditioned mice, same-sex Lin cells were transplanted, and mice were followed up to 16 weeks after transplantation or 32 weeks for the brain scRNA-seq experiment. At interim time points after cell infusion, WBC and plasma GAA activity, blood glucose, and complete blood counts using a Hemavet (Drew Scientific) analyzer were measured in the irradiation experiment.

Echocardiography assessment

Complete echocardiographic examination was performed in all animals at month 7 after treatment (∼week 24–28) at the Lady Davis Institute for medical research (Jewish General Hospital Sir Mortimer B. Davis, 3755 chemin de la Cote Sainte-Catherine, Montréal, Québec, H3T 1E2), using a FUJIFILM VisualSonics echocardiography system (Vevo 3100, FUJIFILM VisualSonics, Inc. Canada). The echocardiography results were evaluated by the technical specialist at the Lady Davis Institute.

Animals were transferred to the treatment facility in a transport vehicle with controlled temperature. Sterile transport bins were modified to allow up to eight animals/bin with food pellets and gel packs. Clinical signs were monitored immediately before and after transportation. At the test site, the animals were placed in dorsal recumbency on a heated surface and allowed to reach steady-state resting heart rate (500–550 bpm) prior to echocardiography examination.

Echocardiography was performed from a left parasternal approach using standards from American Society of Echocardiography for measures. Analyses were done using M-mode and pulse wave Doppler signal with a short-axis view and an apical two-chamber view. Echocardiography was performed as described previously with some modifications.^61,62 Briefly, the mice were anesthetized with 3.5% isoflurane mixed with O₂ at 1 L/min. The depth of anesthesia was confirmed by the rear foot squeezing. The mice were secured lightly in supine position on a warming pad adjusted to 38°C, and their anterior chest was shaved. The echocardiography was performed under isoflurane anesthesia using a VEVO 3100 ultrasound machine and a MX550S transducer with a center frequency of 40 MHz (FUJIFILM VisualSonics Inc., Toronto, ON, Canada). The body temperature and heart rate (HR) were maintained respectively to 36°C–37°C and 500–550 beats/minute during the study. Body temperature was measured using a rectal probe and maintained by adjusting the intensity of an infrared lamp that is positioned over the mouse. HR could be influenced by changes in body temperature, with lower HR with temperature below 36°C and higher HR with temperature above 37°C. Furthermore, even if body temperature was within the required range, mice could present bradycardia. This was corrected by decreasing the % of isoflurane up to 2%. However, if HR increased too much, the % of isoflurane was re-increased gradually up to 3.5%.

Left ventricle (LV) systolic function was assessed as follows. Two-dimensional guided M-mode images were obtained from a short-axis view at the papillary muscle level to determine the LV structure and systolic function. The LV internal diameter in diastole (LVIDd) and the interventricular septum and LV posterior wall thickness in diastole (IVSd and LVPWd) were measured. LV mass, LV mass index, and fractional shortening (FS) were calculated as follows. LV mass (mg) = [(LVIDd + IVSd + LVPWd)³ − LVIDd³] × 1.055 × 0.8, where 1.055 is the density of the rat myocardium (in mg/mm³) (17) and 0.8 a correcting factor to compensate for the overestimation of LV mass. LV mass index (mg/g) = LV mass/body weight. FS (%) = [(LVIDd + LVIDs)/LVIDd] × 100. LV mass index (LVMi) was determined as follows: LVMi (mg/g) = LV mass × 0.8/body weight.

Locomotor function tests

The gait analysis was performed to evaluate the locomotor function in mice. For that, the entire underside of all toes and the center of the feet of the animal were fully covered in paint. After that, the mouse was placed on a piece of white paper at the start of the white plastic corridor and allowed to walk all the way into the black box (safety place). At least two steps consistently spaced with clear, non-smudged footprints were measured from each foot for fore-to-hind paw distance. The male animals were tested before the female animals for any testing day. The test was performed as per standard operating procedure (SOP) including habituation runs on two occasions for the groups on month 6 and 7.

The wire hang test was performed to assess the muscular strength of mice. For that, the animal was placed on the cage top, which was then inverted and suspended above the home cage. The latency until the animals fall was recorded. The average performance for each session was presented as the average of the three trials. The test was performed as per SOP on three occasions for all animals in month 7.

Blood glucose monitoring

Blood glucose was measured as previously described¹⁴ at different time points during the in-life study and prior to scheduled termination with glucometer Accu-chek AVIVA. Animals were fasted overnight prior to collections.

Scheduled termination procedure

Scheduled termination was either weeks 16 or 32 post-transplant. Animals were fasted 8–12 h prior to euthanasia and anesthetized with isoflurane. After cardiac puncture blood collection, the animals were perfused with phosphate-buffered saline (PBS) pH 7.4 until internal organs were pale in appearance. The heart mass was weighed, and the dissected tissues were snap frozen and stored at −80°C for GAA activity and glycogen measurement or processed for histopathology. Therapeutic endpoints included biochemical GAA activity, tissue glycogen content, VCN analysis, histological evaluation, blood glucose, and immunohistochemistry in brain sections.

Measurement of GAA enzyme activity and glycogen by biochemical analysis

For GAA enzyme activity and glycogen measurement, tissue samples of heart, diaphragm, gastrocnemius, quadriceps femoris, tibialis anterior, cerebellum, cerebrum and thoracic spinal cord were collected at necropsy and processed chilled to homogenization in sterile dH₂O, centrifugation, removal of clear supernatants, and then stored at −80°C until the selected assays were performed. Similarly, terminal plasma and cell pellets from peripheral blood, spleen, and bone marrow samples were collected and kept at −80°C for GAA enzyme activity.

The GAA enzyme activity was measured similar as described.¹⁴ Study samples were assayed in a 96-well plate using fluorescent synthetic substrate 4-methylumbelliferyl-alpha-D-glucosidase (4-MU) at 6 mM concentration, in presence of 9 μM acarbose and 90 min incubation. A 0.5 M carbonate buffer pH 10.7 was used to stop the reaction and assay plates were read at 365-nm excitation and 450-nm emission in SpectraMax i3X (Molecular Devices, CA). This method was qualified over the range of 0.1–81 nmol/mL. Results were normalized for protein concentration in the sample using Bicinchoninic Acid (BCA) kit (Pierce, ThermoScientific).

The glycogen in tissues was estimated treating samples with and without Aspergillus niger amyloglucosidase that generates β-D-glucose, which is oxidized to release gluconic acid and hydrogen peroxide.¹⁴ In presence of horseradish peroxidase, hydrogen peroxide reacts with o-dianisidine hydrochloride to generate a colored product detectable at 540 nm in a spectrophotometer and is proportional to the glucose concentration in the sample. The qualified assay had a dynamic range of 5.6–160 μg/mL. Results were normalized for protein concentration in the sample using BCA kit (Pierce, ThermoScientific).

VCN analysis

VCN quantification was completed either by quantitative PCR (qPCR) or digital droplet PCR (ddPCR). All qPCR assays consisted of oligonucleotide primers and probe mixes containing either a TaqMan 6-carboxyfluorescenin (FAM) or VIC fluorescent probe designed to amplify the HIV Psi vector sequence or housekeeping genes glycosyltransferase like domain containing 1 (Gtdc1) or transferrin receptor protein 1 (Tfrc) as previously described.^14,39 For qPCR VCN assessment, a single plasmid containing both sequences was used as a reference standard in a range of 50 to 5 × 10⁷ copies. Data were reported as VCN/diploid genome.

For ddPCR, specific primers targeting the HIV Psi element were used to detect the integrated lentiviral vector, along with specific primers targeting genomic reference sequences Gtdc1.

Immunohistochemistry

Brain hemispheres of GAAco and GILT-vector treated mice were perfused fixed in 4% formaldehyde overnight at 4°C, then washed 3× and stored in PBS. The specimens were embedded and arranged for coronal sectioning in a gelatin matrix using MultiBrain Technology (NeuroScience Associates, Knoxville, TN). After curing with a formaldehyde solution, the blocks were rapidly frozen by immersion in 2-methylbutane chilled with crushed dry ice and mounted on a freezing stage of an AO 860 microtome. The MultiBrain blocks were sectioned coronally with a setting on the microtome of 35 μm. All sections were cut through the entire cortex of the brain and collected sequentially into a series of cups. All cups contained Antigen Preserve solution (50-parts PBS pH7.0, 50-parts ethylene glycol, 1-part polyvinyl pyrrolidone).

For immunohistochemistry, selected sections were stained free-floating. All incubation solutions from the primary antibody onward used Tris-buffered saline (TBS) with Triton X-100 as the vehicle; all rinses were with TBS. After a hydrogen peroxide treatment and rinses, each section was immunostained with primary antibody overnight at room temperature. Following rinses, the primary rabbit antibody against GAA (ab240102 Abcam) and secondary goat anti-rabbit biotinylated (Vector Laboratories, BA-1000) was applied. After further rinses Vector Lab’s ABC solution Catalog # PK-6100 (avidin-biotin-HRP complex for VECTASTAIN Elite ABC, Vector, Burlingame, CA) was applied. The sections were again rinsed, then treated with a chromagen:diaminobenzidine tetrahydrochloride (DAB), nickel (II) sulfate and hydrogen peroxide to create a visible reaction product. Following further rinses, the sections were mounted on gelatin-coated glass slides, air dried, then dehydrated in alcohol, cleared in xylene, and coverslipped. Brightfield stainings were scanned using a Huron Digital Pathology LE120 TissueScope. Slides were scanned using a 20× objective for a final image resolution of 0.4 μm/pixel.

PAS and H&E staining of mouse tissues

Following scheduled necropsy, tissues from Gaa^−/− and Gaa^+/+ mice were divided for different endpoint analyses, and a portion of each collected, preserved, trimmed, and placed in separate cassettes. PAS and hematoxylin and eosin (H&E) stainings were performed as described.¹⁴ Tissues were fixed in 10% NBF for up to 32 h and post-fixed in 1% periodic acid (PA)/10% neutral buffered formalin (NBF) for 48 h at 4°C. One hemisphere from each brain sample along with heart, diaphragm, tibialis anterior, gastrocnemius, and quadriceps femoris tissues were fixed in 10% NBF for up to 32 h and further post-fixed in 1% PA/10% NBF for 48 h at 4°C and processed into FFPE blocks. Blocks of cerebral cortex, cerebellum, hippocampus, and/or brainstem, cervical spinal cord, heart, quadriceps femoris, diaphragm, gastrocnemius, and tibialis anterior were sectioned at 4 μm, mounted onto glass slides, and stained with PAS and H&E, and evaluated for glycogen accumulation and vacuolation by light microscopy. Tissues stained via PAS/H protocols were scanned at ×20 magnification using a Hamamatsu Nanozoomer whole-slide scanner. Scan files were imported into Visiopharm for quantitative image analysis. A region of interest (ROI) surrounding individual tissue sections was automatically applied to all scan files using a Visiopharm Analysis Protocol Package (APP). Automated ROIs were manually refined to optimize anatomic homology across animals. Automated image analysis APPs were used to detect the dark purple PAS+ signal contrasted against a pink PAS-background. The following equations were performed in Visiopharm to generate quantitative endpoints: $Totalarea=PAS+area+PAS-area,PAS+fraction=\frac{PAS+area}{Totalarea},\text{and}PAS+percentage=PAS+fraction\times 100$. Quantitative immunofluorescent data were exported from Visiopharm as excel spreadsheets.

Vacuolation in CNS tissue was limited to specific cell types and sub-anatomic sites. Therefore, severity grading was assigned based on the estimated number of visible neurons with abundant cytoplasmic vacuoles within the most affected anatomic regions (e.g., brain stem, spinal cord gray matter, sciatic nerve). Scores were assigned as: N (vacuolation absent); 1 (<10% of neurons heavily vacuolated); 2 (≥10% and up to 25% of neurons heavily vacuolated); 3 (≥25% and up to 50% of neurons heavily vacuolated; or less than 25% total with heavily vacuolation of specific brain stem nuclei, e.g., gracile nucleus), 4 (≥50% and up to 75% of neurons heavily vacuolated); 5 (≥75% of neurons heavily vacuolated). Vacuolation within myofibers in muscle (cardiac, skeletal; smooth muscle) was scored as: N (vacuolation absent); 1 (minimal vacuolation visible throughout section, requiring at least 20× microscope objective to confirm vacuoles were present); 2 (vacuolation clearly visible with a 10× objective, vacuoles are limited within individual myofibers and do not form large aggregates); 3 (vacuoles clearly visible at 10× and lower objectives with aggregates forming around nuclei and center portions of myofibers); 4 (large vacuolation aggregates are prominent in many myofibers); 5 (most myofibers contained large vacuolation aggregates).

DNAscope for nuclear detection of lentiviral copy number

RNAscope Assay (Advanced Cell Diagnostics, Newark, CA) is an RNA in situ hybridization (ISH) approach that allows simultaneous signal amplification and background suppression to achieve single-molecule visualization while preserving tissue morphology.^63,64 The approach was modified to visualize lentiviral vector DNA (DNAscope). DNAscope target retrieval conditions were optimized for detection of single lentiviral vector DNA using oligoprobes in bone marrow cells of treated mice. DNAscope pretreatment conditions were as follows: Epitope Retrieval (LS ER2): 15 min at 88°C, Protease III: 5 min at 40°C, Customized denaturation DNAscope Protocol DNA in situ hybridization for detection of DNA was performed on the Leica automation platform using the RNAscope 2.5 LS Red Reagent Kit (Advanced Cell Diagnostics, Newark, CA) along with custom accessory reagents according to the manufacturer’s instructions. Briefly, samples on cytospin slides were pretreated with heat and protease prior to hybridization with the target oligo probe. Preamplifier, amplifier, and alkaline phosphatase-labeled oligos were then hybridized sequentially, followed by chromogenic precipitate development. Specific DNA staining signal was identified as red punctate dots. Samples were counterstained with Hematoxylin. Slides were scanned with the 3DHistech Panoramic SCAN II digital scanner to perform HALO image analysis to detect and quantify lentiviral vector-positive cells throughout the entire bone marrow cytospin preparation and copy number measurement as binned dots per cell.

Integration site analysis

Samples from both in vivo studies were used to perform integration site analysis. Bone marrow isolated from 16 transplanted male and female Gaa^−/− mice (n = 14 with the GILT-vector and n = 2 with non-transduced cells) at 8 months after transplantation from 7.5 Gy irradiated mice. Additionally, eight in vitro samples (cultured lineage negative cells), 7 days after transduction with GILT-vector (two samples each of MOI 0.75, 1.5, and 3) or MND.GFP (two samples of MOI 1.5) lentiviral vectors. Mice were selected for ISA based on the highest VCN/diploid genome (dg) measured at week 12 and week 16 in the peripheral blood DNA (determined by qPCR).

The second study included the bone marrow samples obtained from male and female Gaa^−/− mice from the Busulfex study (n = 5/sex) with transduction was done at an MOI of 5. The mice were culled at 6 weeks after Lin-bone marrow cell transplantation. Three Gaa^−/− mice administered non-transduced Lin-bone marrow cells (male n = 1; female n = 2) were also included in this analysis as controls.

INSPIIRED sample processing was performed as described in Sherman et al.,¹⁸ and adapted and described in Ha et al.⁶⁵ DNA samples in nuclease-free water were subject to shearing (Covaris 220) for 60 s at a peak power of 50 Watts, 5% duty factor, 200 cycles/burst, 4°C water temperature. Samples were AMPure purified (0.7-fold bead to sample ratio) and used for end preparation of fragmented DNA using the NEBNext Ultra End Repair/dA-Tailing Module. Previously generated linkers (linker blunt + sample-specific linker) were introduced with the NEBNext Ultra Ligation Module. Samples were again AMPure purified (0.7-fold bead to sample ratio) and used in exponential PCR1 with linker-specific primers (300 nM), and LTR specific primer 1 (300 nM), blocking oligo (1 μM), dNTPs (200 nM), Clontech Advantage PCR buffer (1×) and mix (1×) (TaKaRa), linker-ligated DNA (0.6 v/v ratio), water (0.172 v/v ratio) and a thermocycler program of 1 min at 95°, and then 20 cycles of 30 s at 95°C, 30 s at 80°C and 90 s at 70°C for exponential amplification and finally 4 min at 72°C. For the second nested PCR, the primers, blocking oligos, dNTPs, Clontech Advantage PCR mix, and buffer had the same concentration as in PCR1, and the exponential amplification step in PCR2 was performed with 15 cycles. For all reactions, specific index primers and sample-specific linker primers were used. Afterward, PCR2 reactions were mixed in equal volumes to generate the final library 210303_INSPIIRED_RUN40. The library was first column purified prior to two AMPure purifications with a 0.7-fold, and 0.6-fold ratio of beads to sample volume. Libraries were transferred to the research core unit genomics (RCUG) of Hannover Medical School for quality control via Bioanalyzer and analysis by Illumina sequencing on flow cells with 15 million clusters.

Bioinformatic analysis was generally performed as described by Berry and colleagues.⁶⁶ The analysis files necessary to run the INSPIIRED pipeline were downloaded from GitHub (https://github.com/BushmanLab/INSPIIRED). Individual sequence files were aligned and annotated to the mouse genome (mm9). The plasmid vector sequences served as a reference for LTR regions and vector trimming. The processing and alignment statistics were exported before uploading the results to a local database using inSiteUploader.R. The data frame with all integration site data was exported and used for customized post-processing steps in Microsoft Excel 2016, GraphPad Prism (Version 5), and R (3.6).

Isolation of microglia/microglia-like cells and scRNA-Seq

After transcardial perfusion, the brains of four female mice, two from SFFV.GILT and two from the SFFV.GAAco experimental groups were processed by enzymatic digestion (Neural Tissue Dissociation Kit [P], Miltenyi Biotec). The resulting single-cell suspensions were stained with viability dye and antibodies for microglia cells, neurons, astrocytes and endothelial cells (LIVE/DEAD Fixable Aqua Dead Cell Stain Kit, Thermo Fisher; CD45.1 clone A20 in BV421 (BD); CD45.2 clone 104 in BV421 (BD); CD11b clone M1/70 in APC-780 (Thermo Fisher); CX3CR1 clone SA011F11 in BV605 (BioLegend); PECy7 Thy1.1 (also known as CD90.1) clone OX-7 in PE-Cy7 (BD); Thy1.2 (also known as CD90.2) clone 53–2.1 in PE-Cy7 (BD); ACSA-2 clone IH3-18A3 in PE (Miltenyi Biotec); CD31 clone 390 in PerCP/Cy5.5 (BioLegend). Microglia cells, defined as CD45+ CD11b+ Cx3cr1+, were FACSsorted using an MA900 Multi-Application Cell Sorter (Sony Biotechnology). For the generation of single-cell transcriptomes, a target cell number of 2.5 × 10³ or 5 × 10³ cells from each sorted population were run using the Chromium Controller (10× Genomics) using Chromium Next GEM Single-Cell 3′ Reagent Kits (10× Genomics). The libraries generated were then run on the NextSeq 550 (WT vs. Pompe microglia datasets) or the NovaSeq 6000 (GILT vs. GAAco datasets) Sequencing System (Illumina) using either the using NextSeq 500/550 High Output v2.5 (150 cycles) Kit or the NovaSeq 6000 S2 Reagent Kit v1.5 (100 cycles) Kit (Illumina). The Illumina raw BCL sequencing files were processed through the CellRanger software (10× Genomics) for generating FASTQ files and count matrixes (https://support.10xgenomics.com/single-cell-gene-expression/software/overview/welcome), which were then used as input for the SEURAT V4.0 (https://satijalab.org/seurat/) R tool for single-cell genomics analyses. Briefly, single-cell barcodes were filtered for the ones containing mitochondrial gene content lower than 15%. Expression data then were normalized, scaled, and searched for variable features using the SCTransform function of SEURAT V4.0 followed by UMAP dimensionality reduction and clustering using the FindClusters function with resolution set at 0.2. The maps shown in Figures 6B, 6C, S12B, and S12C were generated using the ggplot RPackage based on the UMAP coordinates from the SEURAT package. The differentially expressed genes shown in Figures 6D, S12A, and S12D were identified through the FindMarkers function of SEURAT and plotted either using Prism 9 (GraphPad Software, LLC.) or through the pheatmap R package (https://cran.r-project.org/web/packages/pheatmap/pheatmap.pdf). The signature scores of Figure 6F have been calculated through the UCell R package (https://github.com/carmonalab/UCell) using as signature the WT-specific genes shown in Figure S12D.

Statistical analysis

Statistical analysis was performed using either Prism (GraphPad) or R. For scRNA-seq data we used GSEAPreranked analysis to identify the gene sets that are enriched in LTR-negative and LTR-positive samples. To perform this analysis, we first derived a score using log2FoldChange and p value from differential gene expression analysis (score = −s∗log10(p value)) with s = −1 if the log2FoldChange of gene and p value from differential gene expression analysis (score = −s∗log10(p value)) with s = −1 if the log2FoldChange of gene. For all other data, a Student’s t test (when only two groups were present), one-way ANOVA with Tukey’s post hoc analysis (when multiple groups were present), or linear regression was used. Significance was defined as p < 0.05.

Experimental groups were sized to allow for statistical analysis; not all the animals were included in the analysis, and select outliers were excluded. Mice were assigned randomly to experimental groups based on weights. All other statistical analysis was analyzed by VERISTAT, Inc or internally by applying median and interquartile range using Kruskal-Wallis and exact Wilcoxon Rank Sum tests for groups comparison and SAS/STAT software v9.4. A result of <0.05 was indicative of significant difference in the groups. For correlation analysis, the Pearson R coefficient and p value were used.

Quantitative immunofluorescent data were collected in Visiopharm and statistical analysis was performed using SAS software (v9.4). Continuous variables were analyzed via Levene’s test for equality of variance. Where Levene’s test was not significant, one-way ANOVA was used to detect differences among three more group means. Tukey’s test was used to explore pairwise group comparisons when one-way ANOVA was significant. Where Levene’s test was significant, the Kruskal-Wallis test was used to detect differences among three or more group means. The Dwass-Steel-Critchlow-Fligner method was used to explore pairwise group comparisons when Kruskal-Wallis test was significant. Significance was set to p < 0.05 for all statistical tests. In the figures ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001.

Data and code availability

Data supporting the studies presented in this manuscript can be found in the main text or the supplemental material. Transcriptomic data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (GEO) database and are available through GEO: GSE247070. Additional information, as appropriate, may be made available by request directed to the corresponding author and, as appropriate, following execution of a suitable confidentiality agreement with AVROBIO, Inc. AVROBIO, Inc. acquired Tectonic Therapeutic in a reverse-merger in June 2024 and subsequently changed its name to Tectonic Therapeutic. All references to “AVROBIO” should be read to refer to Tectonic Therapeutic, despite the fact that the research and work described herein occurred before the closing of the reverse-merger and the subsequent name change.

Acknowledgments

We thank all the members of AVROBIO, Inc. for their continued support of our work. We would like to thank Charles River Laboratories for study execution. Additionally, KCAS Bioanalytical & Biomarker Services for their contributions to sample analysis, NeuroScience Associates (NSA) for the immunohistochemistry of brain, and VERISTAT, Inc, for biostatistical analysis. Finally, we would like to thank Nina Raben (National Institutes of Health, MD) for supplying the Gaa knockout mice.

Author contributions

J.K.Y. contributed to the experimental design and execution, biochemical and molecular assays development and qualification, data analysis and interpretation, and wrote the manuscript; J.W.S., M.L., M.P.D., M.E.J., C.T., R.N.P., A.Y., Y.D., C.B., V.P.C., C.F., F.H., L.B., Z.U., D.I., R.H.K., S.G., V.M., M.R., and A.S. conducted experiments, and performed data analysis, M.R. and A.S. contributed to experimental design and analysis; T.M. performed statistical analysis; C.O., R.P. contributed to assay development and study logistics; C.M. contributed to the design and the reporting of the work presented; L.B. and N.P.v.T. contributed to the design, data analysis, interpretation and writing of the manuscript.

Declaration of interests

All authors were former employees of AVROBIO, Inc., Cambridge, MA, USA during the conception and writing of the manuscript, except V.M., M.R., and A.S. N.P.v.T. and C.M. are inventors on patents in the field of HSC gene therapy. AVROBIO, Inc., has a preclinical gene therapy program for Pompe disease (AVR-RD-03) based on a genetically modified HSPC platform using lentiviral vectors. Collection of data and analysis was performed as part of the program. This research received no external funding and was sponsored by AVROBIO, Inc.