Focal lesions following intracerebral gene therapy for mucopolysaccharidosis IIIA

Abstract

Objective

Mucopolysaccharidosis type IIIA (MPSIIIA) caused by recessive SGSH variants results in sulfamidase deficiency, leading to neurocognitive decline and death. No disease‐modifying therapy is available. The AAVance gene therapy trial investigates AAVrh.10 overexpressing human sulfamidase (LYS‐SAF302) delivered by intracerebral injection in children with MPSIIIA. Post‐treatment MRI monitoring revealed lesions around injection sites. Investigations were initiated in one patient to determine the cause.

Methods

Clinical and MRI details were reviewed. Stereotactic needle biopsies of a lesion were performed; blood and CSF were sampled. All samples were used for viral studies. Immunohistochemistry, electron microscopy, and transcriptome analysis were performed on brain tissue of the patient and various controls.

Results

MRI revealed focal lesions around injection sites with onset from 3 months after therapy, progression until 7 months post therapy with subsequent stabilization and some regression. The patient had transient slight neurological signs and is following near‐normal development. No evidence of viral or immunological/inflammatory cause was found. Immunohistochemistry showed immature oligodendrocytes and astrocytes, oligodendrocyte apoptosis, strong intracellular and extracellular sulfamidase expression and hardly detectable intracellular or extracellular heparan sulfate. No activation of the unfolded protein response was found.

Interpretation

Results suggest that intracerebral gene therapy with local sulfamidase overexpression leads to dysfunction of transduced cells close to injection sites, with extracellular spilling of lysosomal enzymes. This alters extracellular matrix composition, depletes heparan sulfate, impairs astrocyte and oligodendrocyte function, and causes cystic white matter degeneration at the site of highest gene expression. The AAVance trial results will reveal the potential benefit–risk ratio of this therapy.

Introduction

Mucopolysaccharidosis type IIIA (MPSIIIA or Sanfilippo syndrome type A) is an autosomal recessive lysosomal storage disorder primarily characterized by central nervous system (CNS) degeneration leading to neurocognitive decline and early death. ¹ MPSIIIA is caused by deficient activity of the lysosomal enzyme sulfamidase (heparan‐N‐sulfatase; E.C.3.10.1.1) due to pathogenic variants in the SGSH gene. The enzyme is involved in lysosomal degradation of the glycosaminoglycan heparan sulfate; its deficiency results in intra‐ and extra‐lysosomal accumulation of heparan sulfate causing defects in signaling pathways, calcium homeostasis, lipid metabolism, intracellular trafficking, and autophagy, leading to progressive cellular and organ disease. ¹ , ² Prevalence of MPSIIIA is estimated at 1:100,000 live births. ³ Generally, MPSIIIA manifests at age 2 to 3 years with developmental delay, followed by behavioral problems. ¹ Coarse facial features, recurrent upper respiratory tract infections, and hepatomegaly can be early signs. After a plateau phase, relentless cognitive and motor decline occurs. In patients with the most common, rapidly progressing phenotype, cognitive age equivalent rarely surpasses 36 months. ³ , ⁴ The mechanisms by which accumulating heparan sulfate causes neurodegeneration have not been fully elucidated but include interference with multiple cellular pathways leading to neuroinflammation, disturbed autophagy, impaired cellular signaling, and oxidative stress. ⁵ , ⁶ Pathologic studies in MPSIIIA show that throughout the brain neurons have enlarged cell bodies with foamy cytoplasm and distended processes, and mild gliosis in the adjacent parenchyma. The neuronal changes are particularly severe in the cerebral and cerebellar cortex and substantia nigra ⁶ and lead to progressive cortical atrophy. ⁷

Currently, no disease‐modifying treatment is available for this devastating disease. Multiple therapeutic attempts, including hematopoietic stem cell transplantation (HSCT), ⁸ , ⁹ oral genistein to reduce heparan sulfate accumulation, ¹⁰ , ¹¹ intrathecal recombinant sulfamidase ¹² and intravenous chemically modified recombinant sulfamidase ¹³ have proved largely ineffective. Direct intracerebral administration of an AAVrh.10 vector carrying the human SGSH cDNA with a phosphoglycerate kinase promoter (PGK) proved safe in four MPSIIIA patients and suggested a benefit in neurocognitive development in the youngest patient. ¹⁴

These results prompted Lysogene (Paris, France) to initiate a multicenter open‐label clinical trial, named AAVance (NCT03612869), using intracerebral administration of AAVrh.10 vector with a stronger promoter to drive SGSH cDNA expression and higher vector dose. During standard post‐treatment monitoring by MRI of a study subject, development of lesions around the injection sites was observed. Investigations were initiated to determine the cause. First, we considered a viral infection. The vector used is replication‐incompetent but could hypothetically be mobilized by superinfection of a helper virus or wild‐type AAV. Second, we considered an immune response to an injected ingredient. Third, we considered that overexpression of sulfamidase might cause cellular toxicity, such as activation of the unfolded protein response (UPR) and integrated stress response (ISR) in transduced cells or spilling of lysosomal enzymes with detrimental effects on surrounding tissue.

Methods and Materials

AAVance trial protocol

This phase 2/3 clinical trial (ClinicalTrials.gov Identifier: NCT03612869) was approved in the Netherlands by the Central Committee for Research involving Human Subjects (CCMO). The investigational product, LYS‐SAF302 (INN olenasufligene relduparvovec), is a defective adeno‐associated virus (AAV) 2 vector serotype rh.10, carrying the cytomegalovirus enhancer fused to a chicken β‐actin promoter/rabbit β‐globin intron (CAG promoter) to drive high levels of expression of human SGSH cDNA. ¹⁵ A single dose of 7.2E+12 viral genomes (vg) per patient divided over six 500‐μL injections at three sites per cerebral hemisphere was delivered using the ClearPoint Neuro SmartFlow® canula. This dose was 10 times higher than used in the previous phase 1–2 study with the first‐generation vector LYS‐SAF301. A neuronavigation system (Surgiscope; Isis) was used to select injection sites in the white matter anterior, lateral, and posterior to the basal ganglia.

Immunosuppression to prevent and suppress immune response to the AAVrh.10 vector preparations and the transgene product was started 7 days before surgery and consisted of mycophenolate mofetil (600 mg/m²/b.i.d.), continued for a period of 2 months, and tacrolimus (0.15 mg/kg/day) aiming at levels of 8–10 ng/mL during the first 3 months and 5–8 ng/mL thereafter for 1 year. In addition, prednisolone (1 mg/kg/day) was started 1 day before surgery and continued for 10 days.

Inclusion of patients in the trial was based on a DNA‐confirmed diagnosis of MPSIIIA and a cognitive developmental quotient on the Bayley Scales of Infant and Toddler Development‐Third Edition (BSID‐III) of 50% or higher. Patients homozygous for the p.Ser298Pro variant, conveying a slowly progressive phenotype, and patients with a non‐severe form of MPSIIIA based on the investigators judgment were excluded.

MRI monitoring of trial patients occurred at baseline and at 3, 6, 9, 12, 18, and 24 months post‐treatment. The MRI protocol included 3D T1‐ and T2‐weighted sequences with multiplanar reconstruction, and diffusion tensor imaging. To monitor the lesions, a 3D‐FLAIR sequence was added.

Study subject

A Dutch female, who is an AAVance trial subject, from here on referred to as study patient, was diagnosed with MPSIIIA at 8 months of age by detection of high heparan sulfate in the urine. Metabolic studies were performed because of unexplained persisting mildly increased ASAT and ALAT levels detected in the follow‐up of a parechovirus meningitis at age 1 month. Physical examination was normal, without hepatomegaly or facial features of MPSIII. Sulfamidase enzyme activity was undetectable in leukocytes. Mutation analysis revealed compound heterozygosity for two variants in the SGSH gene (NM_000199.4), previously reported in MPSIIIA: c.734G > A, p.(Arg245His) and c.892 T > C, p.(Ser298Pro).

The patient was included in the trial and received intracerebral injections of LYS‐SAF302 at 11 months of age. The procedure was uncomplicated, and no neurological problems were observed after the procedure. The MRI after 3 months revealed emerging lesions around the injection sites, which increased at 7 months after therapy (Fig. 1).

Figure 1.

Sequential MRIs in the patient of this study. At age 10 months, just prior to gene therapy, T2‐weighted images (left) and diffusion‐weighted images (right) showed no abnormalities. At 3 months after gene therapy, T2‐weighted images (left) showed the first lesion, most prominent in the right parietal region. Diffusion‐weighted images (right) revealed no abnormalities. At 7 months after gene therapy, FLAIR images (first 3 images from the left; right parasagittal, axial and left parasagittal) showed enlarging lesions that were rarefied in the center. Diffusion‐weighted images (right) showed diffusion restriction in the rim of some of the lesions. At 13 months after gene therapy, FLAIR images (first 3 images from left; right parasagittal, axial and left parasagittal) showed that the lesions had decreased in size but were more cystic The lesion in the right parietal region was cystic after biopsy. Diffusion‐weighted images (fourth) revealed no diffusion restriction.

Sample collection

Eight months after gene therapy, the study patient underwent stereotactic needle biopsy of the right parieto‐temporal deep white matter under CT‐guidance. The procedure was uncomplicated, and a post‐biopsy MRI showed no damage outside the lesion. Eight 2 × 6 mm samples were obtained: two for histology, two for viral studies, one for transmission electron microscopy and three were frozen.

For comparison, postmortem brain white matter from two untreated MPSIIIA patients, two childhood‐onset Vanishing White Matter (VWM) patients, and two controls without local confounding neuropathology was examined. All controls were matched for age as much as possible. Additionally, brain white matter tissue obtained at epilepsy surgery in one infant and two young adults was used for control. The demographics of the tissues used are reported in Table 1.

Table 1.

Human brain tissue samples analyzed.

Sample identifier	Tissue type	Age	Sex	Diagnosis
MPSIIIA‐VT, study patient	Biopsy, parieto‐temporal white matter	1.5y	f	MPSIIIA
MPSIIIA 1 a	Postmortem, frontal white matter	11y	f	MPSIIIA
MPSIIIA 2 a	Postmortem, frontal white matter	12y	f	MPSIIIA
VWM 1	Postmortem b , frontal white matter	9y	f	VWM, EIF2B5 homozygous c.271A > G, p.Thr91Ala
VWM 2	Postmortem b , frontal white matter	10y	f	VWM, EIF2B5 compound heterozygous c.338G > A, p.Arg113His and c.1208C > T, p.Ala403Val
Control 1	Postmortem b , frontal white matter	1d	f	congenital brain stem tumor
Control 2	Postmortem b , frontal white matter	1 m	f	SUDI
Control 3	Biopsy, temporal white matter	21y	f	MMCD
Control 4	Biopsy, temporal white matter	33y	m	MMCD
Control 5	Biopsy, temporal white matter	1y	m	MMCD

d, days; m, months; MMCD, mild malformation of cortical development; SUDI, sudden unexplained death in infancy; VWM, vanishing white matter; y, years.

Please note the tissue used contained no cortical derangements.

^a Neuropathologically confirmed disease obtained through the NIH NeuroBioBank (https://neurobiobank.nih.gov).

^b Postmortem delay below 6 h.

During anesthesia, blood and CSF were also collected.

Viral studies

CSF, blood, and brain biopsy samples were studied by extensive routine viral testing, including PCR for herpes simplex virus (HSV), Epstein–Barr virus (EBV), cytomegalovirus (CMV), and John Cunningham (JC) virus.

In total, 22 CSF, serum and brain biopsy samples were subjected to viral metagenomics, including a virus discovery pipeline, to detect known and unknown viruses using VIDISCA‐NGS. ¹⁶ Samples of multiple time points from the study patient and two others included in the current trial, and from three patients included in the safety and tolerability study with LYS‐SAF301 were investigated (Table 2). ¹⁴ As a positive control, a brain biopsy from a dog that had received an intracerebral injection of LYS‐SAF302 as part of a preclinical bio‐distribution study ¹⁵ containing 10⁷–10⁸ AAVrh.10 vector copies/μg genomic DNA was included.

Table 2.

Human body fluid and dog brain tissue samples analyzed.

Material	Study	ID/sample info	Time point	Date	Pre‐treatment	Number of reads	Anello‐viridae	Canine circo virus	AAV
CSF	New	AAVance trial subject 1	Screening	03/09/2019	DNase	7697
CSF	New	AAVance trial subject 1	Screening	03/09/2019	No	7573
Serum	New	AAVance trial subject 1	Screening	03/09/2019	DNase	29368	14
Serum	New	AAVance trial subject 1	M9	21/03/2020	DNase	21187	7
CSF	New	AAVance trial subject 1	M12	05/10/2020	DNase	9620
CSF	New	AAVance trial subject 1	M12	05/10/2020	No	12588
Serum	New	AAVance trial subject 1	M12	05/10/2020	DNase	34273	6
Serum	New	AAVance trial subject 2	Screening	24/09/2019	DNase	21469
Serum	New	AAVance trial subject 2	M12	14/09/2020	DNase	14898
CSF	New	Current study patient	Screening	18/03/2020	DNase	5642
CSF	New	Current study patient	Screening	18/03/2020	No	13566
Serum	New	Current study patient	Screening	18/03/2020	DNase	18466
CSF	New	Current study patient	M6	05/10/2020	DNase	3958
CSF	New	Current study patient	M6	05/10/2020	No	25036
Serum	New	Current study patient	M6	05/10/2020	DNase	2
CSF	New	Current study patient	M8	06/11/2020	DNase	8080
CSF	New	Current study patient	M8	06/11/2020	No	14775
Plasma	New	Current study patient	M8	06/11/2020	DNase	11468	1221
Plasma	New	Current study patient	M8	06/11/2020	DNase	42531	2021
Dog brain Wash	New	403			DNase	77667		2
Dog brain	New	403			Dnase	43629		1	1
CSF	Old	02‐M‐M V2	Inclusion	09/11/2011	Dnase	2683
CSF	Old	02‐M‐M V2	Inclusion	09/11/2011	No	16153
Serum	Old	02‐M‐M V2	Inclusion	09/11/2011	DNase	10463
CSF	Old	635/01 E11901‐16 04‐F‐G	Inclusion		DNase	1580
CSF	Old	635/01 E11901‐16 04‐F‐G	Inclusion		No	4268
Serum	Old	635/01 E11901‐16 04‐F‐G	Inclusion		DNase	12103	5
CSF	Old	635/01 E11901‐27 J2		12/12/2012	DNase	1721	1
CSF	Old	635/01 E11901‐27 J2		12/12/2012	No	12114
Serum	Old	635/01 E11901‐27 J2		12/12/2012	DNase	2981	24
CSF	Old	635/01 P1‐SAF. 301 M12.V19		29/05/2013	DNase	972
CSF	Old	635/01 P1‐SAF. 301 M12.V19		29/05/2013	No	1636
Serum	Old	635/01 P1‐SAF. 301 M12.V19		29/05/2013	DNase	13458	164

Tissue samples were soaked in 200 μL Universal Transport Medium (UTM) and incubated for 1 h, after which 110 μL of liquid was withdrawn. The remaining biopsy tissue was homogenized using a pestle (DWK Life Sciences Inc. (Kimble Chase, KIM‐749520‐0090)). Both withdrawn liquid and homogenized tissue were used for viral metagenomics. CSF was diluted (1:1) in UTM. VIDISCA‐NGS was performed as previously described. ¹⁶

Presence of SGSH transcript from LYS‐SAF302 vector in brain biopsies

qPCR was conducted, as described, ¹⁷ on cDNA from two biopsies from the study patient to determine LYS‐SAF302 vector‐specific transgene expression. The vector‐specific oligonucleotide sequence used is listed in Table 3. The amplification cycle was repeated 40 times. cDNA from two LYS‐SAF302 vector‐treated dogs was included as positive controls. An H₂O control and cDNA from an untreated human cell line were included as negative controls. Quantification of LYS‐SAF302 vector expression was determined using a standard curve with the LYS‐SAF302 vector plasmid ranging from 1.42E+11vg/mL to 1.42E+6vg/mL.

Table 3.

Oligonucleotide primers for qPCR.

Gene	Forward primer (5′ → 3′)	Reverse primer (5′ → 3′)
ATF4	TCAGTCCCTCCAACAACAGC	TCTGGCATGGTTTCCAGGTC
CHOP/DDIT3	GTACCTATGTTTCACCTCCTG	CTGGAATCTGGAGAGTGAG
TRIB3	CTTTGTCTTCGCTGACCGTGA	CCCACAGGGAATCATCTGGC
PDIA3	CCGCAAAACCTTTAGCCATGA	AACCTCTCCAGAGCCTTCCC
DNAJC3	GCCACACACCTTTCCTCCTC	ACGGCAGCATGAAACTGAGA
XBP1s	TCCGCAGCAGGTGCAG	CCAAGTTGTCCAGAATGCCC
XBP1u	GCAGCACTCAGACTACGTG	CCAAGTTGTCCAGAATGCCC
LYS‐SAF302	CCAGCCCCTCCACAATGA	CACTGGAGTGGCAACTTCCA
GAPDH	ACAGTCAGCCGCATCTTCTT	GACTCCGACCTTCACCTTCC
AKT1	TTGTGAAGGAGGGTTGGCTG	TTGAGGAGGAAGTAGCGTGG

Histology, immunohistochemistry, and electron microscopy

Formalin‐fixed paraffin‐embedded tissue sections were cut at 5 μm and routinely stained with hematoxylin and eosin (H&E) and Kluver‐Periodic Acid Schiff (PAS). Immunohistochemical stains were performed as described, ¹⁸ using antibodies against the mature myelin protein proteolipid protein (PLP, Bio‐Rad, MCA839G, 1:3000), oligodendrocyte marker OLIG2 (Sigma‐Aldrich, AB9610, 1:100), astrocyte proteins glial fibrillary acidic protein (GFAP; Dako, Z0334, 1:1000; and Sigma‐Aldrich, AB5541, 1:500) and S100 (Dako, Z311, 1:400), activated microglia marker HLA‐DR (Ebioscience, 14‐9956‐82, 1:750), T lymphocyte marker CD3 (Dako A0452, 1:200), B lymphocyte marker CD20 (Dako, M0755, 1:50), UPR/ISR marker 4EBP1 (Cell signaling, 53H11, 1:400), apoptosis marker caspase 3 (Cell signaling, 9661L, 1:500), and the enzyme sulfamidase (SGSH, Invitrogen, PA5‐82831, 1:200).

Immunohistochemistry was performed to detect presence and concentrations of heparan sulfate, using phage display derived (scFv)‐primary antibodies directed against heavily sulfated heparan sulfate structures, including heparan (AO4B0820, 1:5), ¹⁹ and preferentially low‐sulfated heparan sulfate motifs (HS4E420, 1:5). ¹⁹ scFv‐antibodies containing a vesicular stomatitis virus (VSV)‐tag were used to allow detection by a secondary anti‐VSV tag antibody P5D4 (Sigma, 1:200) in 0.1% BSA/PBS overnight at room temperature. Co‐staining was performed with GFAP. Sections were subsequently incubated for 1 h at room temperature with tertiary fluorophore antibodies (Alexa 488‐ and ‐594 tagged, 1:400) and counterstained with DAPI (Fluoromount G mounting medium, Invitrogen). Photographs were taken using a Leica DM5000B microscope (Leica Microsystems).

For ultrastructural examination, tissue was fixed in 2% glutaraldehyde, 4% paraformaldehyde in 0.1 mol/L sodium cacodylate buffer (pH 7.4), post‐fixed in 1% osmium tetroxide, 1% potassium ferrocyanide, dehydrated, and embedded in epoxy resin. Ultrathin sections were contrasted with uranyl acetate and lead citrate and examined.

UPR transcriptome analysis

Total RNA was isolated in parallel from similar volumes of white matter (Table 1). One tissue sample was processed for each individual, except for the study patient, for whom two biopsies were used as duplicates. RNA isolation, cDNA synthesis, and SYBR‐Green qPCR were performed as described. ¹⁷ Table 3 lists oligonucleotide sequences used for qPCR. GAPDH and AKT were used as reference.

Results

Patient and MRI

The study patient was clinically examined multiple times before and after gene therapy; at latest examination, she was 36 months old. Before therapy, neurological examination showed no evidence of developmental delay or focal brain abnormalities. At 18 months, when the cerebral lesions were at their maximum size, she displayed preferential use of her right hand and left‐sided hyperreflexia, but no Babinski sign. At 24 months, neurological examination had normalized and has remained normal since.

Analysis of CSF obtained during the surgery revealed absence of leukocytes, normal glucose and a slightly elevated protein concentration of 0.56 g/L. IgG index was within the normal range.

Sequential MRIs are illustrated in Fig. 1. The first MRI, obtained at 10 months, shortly before therapy, showed no abnormalities. At 14 months, 3 months after injection, well demarcated signal abnormalities (total volume 5030 mm³) within the cerebral white matter were seen around 4 of the 6 injection trajectories without diffusion restriction. At 18 months, 7 months after injection, the lesion volume had increased to 15391 mm³, with lesion extension into the posterior limb of the internal capsule on the left. FLAIR images revealed rarefaction and cystic decay of the lesions. Diffusion restriction was present in part of the lesion area. After gadolinium administration, no abnormal enhancement was present. At 20 months, 9 months after the procedure, the lesion volume had slightly decreased to 10911 mm³ and diffusion restriction had decreased. At 2 years of age, 13 months after the procedure, the lesion volume had decreased to 8547 mm³. Six and 12 months later, the total lesion volume remained similar (8759 and 7137 mm³, respectively).

Virology

Brain biopsy material, blood, and CSF of the study patient were negative for adenovirus, CMV, EBV, HSV, and JC. Similarly, no viruses were detected by VIDISCA‐NGS in the CSF of any patient sample analyzed (Table 2) or in the brain biopsy of the study patient. The only viruses that were detected in the patients were detected in serum. These were anelloviruses, identified in at least one time point in all trial patients (results not shown). A canine circovirus and AAV sequences were detected in the canine brain biopsy. The detected AAV sequence in the canine brain did not align with the vector sequence but aligned with AAV capsid protein (VP1) sequence. No other AAV sequences were identified. Because anelloviruses and circoviruses are generally considered commensal viruses and non‐pathogenic, no further studies were performed.

SGSH transcript from LYS‐SAF302 vector in the brain biopsies

qPCR for LYS‐SAF302 vector transcript showed the presence of SGSH‐specific mRNA expression at the time of biopsy (Fig. 2). The study patient samples contained approximately 1.40E+10 copies/mL, which was comparable to two samples from vector‐treated dogs. No vector‐derived transcripts were detected in untreated controls.

Figure 2.

AAV vector‐derived SGSH transgene expression is observed in biopsies of a patient with MPSIIIA 8 months after treatment. SGSH‐specific signal of two samples of AAV vector‐treated (VT) dogs (pos ctrl canine 1 and 2) and two biopsies obtained from a VT MPSIIIA patient (IIIA‐VT biopsy 1 and 2) was compared to a standard curve using the LYS‐SAF302 vector plasmid. Lower Cycle Threshold (CT) values indicate higher SGSH signal. Amplification cycles did not increase the CT value of a non‐treated human sample (neg ctrl human) sufficiently to indicate SGSH signal, which is therefore displayed on the x‐axis.

Histology, immunohistochemistry, and electron microscopy

Two needle biopsies of the study patient were used for histology and immunohistochemistry. They only contained white matter. The tissue was not rarefied nor edematous. It showed an increased cellularity with some degree of cell clustering around the blood vessels. Axons had normal appearance. Myelin was not reduced in amounts, but expressed no PLP, indicating immaturity (Fig. 3). The density of OLIG2‐expressing immature oligodendrocyte precursor cells (OPCs) was increased, more than expected for the age of the child. About 40%–50% of these oligodendrocytes expressed cleaved Caspase 3, indicating apoptosis. The astrocytes displayed a gemistocytic morphology with abundant cytoplasmic S100 expression sparing the nuclei, indicating immaturity. No overexpression of GFAP and no elongation or overlap of the astrocytic cell processes were noted, arguing against florid reactive gliosis.

Figure 3.

Neuropathology of MPSIIIA, untreated and vector‐treated. Hematoxylin & Eosin stain of untreated patients with MPSIIIA (A) shows normal white matter with thin‐walled blood vessels and normal cellularity. In the lesioned white matter of vector‐treated (VT) MPSIIIA patient (B), subject of the current study, some cellular clustering is seen around blood vessels. These cells have the morphology of lymphocytes. Kluver‐periodic acid Schiff (PAS) stain for myelin (blue) shows more intense staining in the untreated (C) than the vector‐treated patient (D). Immunostaining against the major myelin protein proteolipid protein (PLP) of untreated patients' white matter (E) shows normal mature myelin content; by contrast, the affected white matter of the vector‐treated patient (F) virtually contains no PLP. Stain against the pan‐oligodendrocyte marker Olig2 (G) shows an increase in Olig2‐positive cells in the lesioned white matter of the vector‐treated patient. Approximately 50% of these cells express caspase3 (H), a marker of apoptosis. Stain against the astrocyte marker glial fibrillary acidic protein (GFAP) of the white matter of the untreated patients with MPSIIIA (I) shows normal density of GFAP‐positive cells, arguing against reactive gliosis. The white matter of the vector‐treated patient (J) also shows no reactive gliosis. Only a subset of GFAP‐positive cells express the immaturity marker S100 (K) in the untreated patients. However, the density of S100‐expressing immature astrocytes (L) is clearly increased in the vector‐treated patient. Staining against the activated microglia marker HLA‐DR shows no appreciable activation in untreated patients (M), and some degree of microglia activation in the vector‐treated patient (N). Immunostaining against lymphocytes of the vector‐treated patient shows presence of some T‐cells (O) and B cells (P) around the blood vessels and in the parenchyma. Original magnifications (A–P) 200× bar: 10 μm.

Activated microglia with a sporadical amoeboid morphology were sparse in the tissue., Some T lymphocytes were identified, mostly around blood vessels. There were also a few perivascular B lymphocytes. Blood vessel walls were normal.

The UPR/ISR marker 4EBP1 was expressed only in lymphocytes (Fig. 4). There was no 4EBP1 expression in other cell types.

Figure 4.

White matter lesions of vector‐treated patient with MPSIIIA show no unfolded protein response activation, but profound depletion of heparan sulfate. Immunohistochemical staining against the unfolder protein/integrated stress response marker 4EBP1 shows no positive cells in the control (A) and untreated MPSIIIA white matter (B). In the vector‐treated MPSIIIA patient lymphocytes express 4EBP1 (C). Strongly 4EBP1‐positive astrocytes in Vanishing White Matter (VWM) serve as control (D). Electron microscopy image of a control (E) shows normal ensheathed axons, while an image of the vector‐treated patient (F) shows several axons ensheathed by slightly thinned, though normally compacted myelin. On the left, the nucleus of an astrocyte is seen. No lysosomes are visible. Staining against sulfamidase (SGSH) shows normal enzyme expression in the cellular cytoplasm of control white matter (G), near depletion of enzyme immunoreactivity in the untreated MPSIIIA white matter (H), and enzyme overexpression in the vector‐treated MPSIIIA patient (I). Fluorescence immunostaining against the highly sulfated heparan sulfate motifs (HS‐High, green) combined with the astrocyte marker glial fibrillary acidic protein (GFAP, red) shows that, compared to control white matter (J) HS‐High is markedly accumulated in the white matter of untreated MPSIIIA patients (K), but virtually devoid in the vector‐treated MPSIIIA white matter (L). Original magnifications (A–D; G–I) 200× bar: 10 μm; (E,F) 18 500× bar: 1 μm (J–L) 400× bar: 5 μm.

Histology and immunohistochemistry of the two untreated MPSIIIA patient samples and of the surgical controls showed normally myelinated white matter, without inflammation and with little reactive gliosis. Neuropathology of the two VWM patient samples was consistent with previous observations in this disease, and included tissue rarefaction, lack of myelin, increased density of oligodendrocytes, and dysmorphic astrocytes intensely expressing 4EBP1. ²⁰ , ²¹

In controls, staining for sulfamidase showed expression in the cytoplasm of all white matter cells, while complete lack of enzyme expression was observed in the untreated MPSIIIA patient samples. Abundant immunoreactivity to sulfamidase detected in virtually all white matter cells and in the extracellular matrix (ECM) of the vector‐treated study patient, was consistent with strong sulfamidase expression. Staining against highly sulfated heparan sulfate motifs revealed that heparan sulfate was highly abundant in the ECM and inside astrocytes and macrophages of untreated MPSIIIA patient samples, but barely detectable in the ECM of the study patient, with levels far below those of the controls. Staining against the low‐sulfated heparan sulfate motifs showed the same trend, although with higher signal in blood vessel walls of the vector‐treated study patient as compared to untreated patients and the controls.

Ultrastructural analysis of the study patient showed relative thinning of myelin sheaths without changes in compaction (Fig. 4). Lysosomes were very scarce In the cytoplasm of most cells (Fig. 3E). In some glial cells, the rough endoplasmic reticulum had lost its perinuclear localization and was arranged in concentric circles.

UPR transcriptome analysis

qPCR analyses of vector‐treatment effects were in line with neuropathological findings. Biopsies from the study patient did not show increased expression of UPR mRNA markers compared to untreated MPSIIIA patients and epilepsy‐surgery controls (Fig. 5). Increased expression of some markers was clearly visible in postmortem white matter from VWM patients as previously observed, ²¹ validating the method.

Figure 5.

UPR activation is not observed in study patient with MPSIIIA after vector‐treatment. White matter tissue was analyzed for expression of UPR/ISR markers by qPCR. Graphs show individual data points of prototypical mRNA markers for UPR activation. AKT and GAPDH were taken as reference mRNAs. White matter is from postmortem samples or surgical biopsies, as indicated. White matter tissue from VWM patients shows increased expression of ATF4‐regulated UPR/ISR markers but not ATF6 or IRE1α regulated UPR markers. ²¹

Discussion

MPSIIIA is a lysosomal disease, primarily affecting gray matter of the CNS, with only mild involvement of other tissues. ¹ , ³ , ⁴ Several earlier therapy trials ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ were unsuccessful in halting the disease. Currently, no disease‐modifying treatment is available. In vivo gene therapy is a promising alternative. Safety and tolerability of intracerebral administration of AAVrh.10 vector carrying the human SGSH gene with a PGK promoter have been demonstrated in four MPSIIIA patients. ¹⁴ Data from that study suggested that the youngest participating patient experienced a benefit in neurocognitive development. ¹⁴ As the next step and to enhance efficacy, a modified transgene was developed with the more potent CAG promoter to drive higher levels of gene expression and a 10‐fold higher vector dose was used.

The patient reported in this study received intracerebral gene therapy at the relatively early age of 11 months. Between 3 and 7 months after therapy, increasing lesions occurred along four of the six injection sites, with local rarefaction and cystic decay. Her lesions were mostly clinically silent; one lesion, which extended into the posterior limb of the internal capsule, was associated with subtle signs of pyramidal dysfunction, which were transient and did not recur. Other male and female subjects in the AAVance trial, whose MRIs we reviewed, developed similar local white matter abnormalities around injection trajectories. The lesions were progressive during several months, but from 12 to 18 months after injection stable or decreased lesion volumes were observed (data not shown). Of the reviewed AAVance trial subjects, the current study patient had the largest lesions at the time of review. A stereotactic biopsy was taken from a lesion during the phase of lesion progression and blood and CSF were collected to study pathomechanisms.

First, we considered a local viral infection. LYS‐SAF302 is a recombinant AAV2 genome with serotype AAV.rh10 that is replication incompetent, yet theoretically may be mobilized when co‐infection with wild‐type AAV or helper virus occurs. ²² We found no evidence of a DNA‐virus with the capacity to help adeno‐associated viruses by PCR in blood, CSF, and brain tissue of the study patient. VIDISCA‐NGS found no link with a virus. VIDISCA‐NGS is a viromics technique that can detect any RNA or DNA virus in CSF and blood when concentrations reach >10E4 copies/mL. ¹⁶ , ²³ Electron microscopy revealed no viral particles. The signs of inflammation in the biopsy tissue were limited and CSF obtained during the biopsy revealed no evidence of an ongoing inflammatory response. The MRI characteristics, location, extent, and course over time of the white matter lesions in the study patient were very similar to those in other AAVance trial subjects with a viral encephalitis, more variability would be expected. Thus, a viral infection was regarded unlikely.

Second, we considered a local immune response to AAV vector injected or the transgene product sulfamidase. The wild‐type human sulfamidase could contain novel epitopes for patients and elicit an immune response, and the same holds true for the AAV vector components ²⁴ , ²⁵ , ²⁶ , ²⁷ and other injected ingredients. A toxicology study investigating administration of an AAVrh.10 vector expressing human arylsulfatase A, and a null AAVrh.10 vector into the white matter of nonhuman primates ²⁸ revealed inflammatory lesions with moderate to high numbers of T and B cells, microglia and activated macrophages at injection sites. Serum samples of the study patient were negative for anti‐sulfamidase antibodies at baseline, 1 and 6 months, while they were positive at 9, 12, and 18 months. However, she had no detectable anti‐SGSH T‐cell response (data not shown), and immune activation was only very mild in the brain biopsy, without prominent T‐ or B‐cell infiltration or activated microglial response, as would have been expected for an immune‐mediated white matter lesion. ²⁹ The consistent MRI characteristics for the study patient and other reviewed AAVance trial subjects also argues against immune‐mediated encephalitis, for which more variability would be expected.

The third option considered effects of local sulfamidase overexpression. qPCR showed robust vector‐derived SGSH transgene expression in the biopsies of the study patient. Sulfamidase overexpression could activate proteotoxic stress within the endoplasmic reticulum in transduced cells, thereby activating the UPR and ISR, contributing to cell death. We used VWM brain tissue as positive control for ISR activation. VWM is a leukodystrophy, characterized by a dysregulated ISR as central disease mechanisms. ²¹ A constitutive abnormal ISR activation is observed in VWM white matter cells, especially astrocytes. ²¹ Unlike in VWM tissue, both immunohistochemistry and qPCR revealed no evidence of activation of the UPR and ISR in white matter cells in the biopsy tissue of the study patient, ruling out this explanation. Electron microscopy of brain tissue of the study patient showed sporadic glial cells, in which the rough endoplasmic reticulum had lost its perinuclear localization and was arranged in concentric circles, indicating endoplasmic reticulum stress, but the number of such cells was very low.

We next investigated the option that sulfamidase overproduction would alter lysosomal enzyme trafficking resulting in excessive secretion into the ECM, or cause cell damage and death leading to release of lysosomal enzymes into the ECM. Staining for sulfamidase showed complete lack of enzyme expression in the untreated MPSIIIA patients, whereas the staining was markedly positive in the vector‐treated study patient, consistent with high levels of sulfamidase expression, both intracellularly and diffusely in the ECM around the injection site. While ECM heparan sulfate was increased in untreated MPSIIIA patients, it was barely present in biopsy tissue of the study patient, with levels far below those of normal control subjects. Although sulfamidase has optimal function at low pH as present in lysosomes, ³⁰ it is also active at higher pH. ³¹ Complete degradation of heparan sulfate cannot occur by the action of sulfamidase alone but instead requires concerted action of a series of lysosomal enzymes. ³² Possibly, sulfamidase overexpression led to cellular damage and spilling of lysosomal enzymes into the ECM, with subsequent digestion of ECM components, thus altering the normal ECM composition, which in the study patient was devoid of heparan sulfate.

The brain's ECM is a dynamic macromolecular network that supports cell growth and viability and plays crucial roles in regulating structural development, integrity, homeostasis, and repair. ³³ The ECM is composed of water, proteins, and mucopolysaccharides, a heterogeneous family of linear polysaccharides composed of repeating disaccharide units, comprising chondroitin sulfate, heparan sulfate, dermatan sulfate, keratan sulfate, and hyaluronic acid. In the ECM, mucopolysaccharides exist as glycol‐conjugates covalently linked to a core protein to form proteoglycans (PGs). Mucopolysaccharides are highly active components of the ECM, each with different properties and functions. ³⁴ Heparan sulfate‐PGs (HSPGs) are major ECM constituents and include glypicans, syndecans, perlecan, agrin, collagen XVIII, CD44, neuropilin‐1, betaglycan, and serglycin. HSPGs serve crucial roles in development and are upregulated upon injury. In the white matter, HSPGs may influence oligodendrogenesis. In the normal brain, oligodendrocytes differentially express HSPGs at various stages of differentiation, where they support OPC proliferation and differentiation and impact on myelin thickness. ³⁵

Derangements in the composition of the ECM are associated with disease. In white matter injury of premature infants and multiple sclerosis (MS) lesions, high molecular weight hyaluronan is strikingly elevated in the affected white matter, inhibiting maturation and differentiation of OPCs recruited to the damaged white matter, contributing to failure of myelination or remyelination. ³⁶ , ³⁷ , ³⁸

The ECM mucopolysaccharide composition is also altered in VWM. ³⁹ The MRI characteristics of the cerebral white matter abnormalities of the vector‐treated MPSIIIA patients and patients with VWM are strikingly similar, both displaying tissue rarefaction and cystic decay, although these findings were diffuse in VWM and focal in vector‐treated MPSIIIA. ⁴⁰ , ⁴¹ Pathology findings are also similar. In the brain biopsy of the study patient, the white matter displayed an abundance of OPCs concurring with lack of mature myelin, and 50% of the OPCs showed signs of apoptosis. Astrogliosis is lacking. The pathology of VWM is characterized by increased numbers of OPCs, while mature myelin is lacking and astrogliosis is meager, causing cystic white matter decay. ³⁹ , ⁴¹ , ⁴² In VWM, these features are caused by immaturity of oligodendrocytes and astrocytes, which are unable to perform their mature functions of (re)myelination and reactive gliosis, respectively. ³⁹ , ⁴² The immature astrocytes heavily impact on the ECM composition in VWM. Similar to their fetal counterparts, they secrete abundant amounts of heparan sulfate‐, chondroitin sulfate‐, and dermatan sulfate‐mucopolysaccharides and high molecular weight hyaluronan, which, like in the fetal brain, result in a highly water‐enriched ECM that is prone to cavitation, and hampers maturation of OPCs into myelin‐forming cells. ³⁹

In conclusion, it is likely that the intracerebral gene therapy in this study, with high local levels of sulfamidase expression, leads to dysfunction of transduced cells with spilling of lysosomal enzymes into the ECM, altering the composition of the ECM around the transduced cells, locally depriving the ECM from heparan sulfate, impairing astrocyte and oligodendrocyte maturation and function, and causing local cystic white matter decay. The similar characteristics and time course of the white matter lesions on MRI for the study patient and other subjects of the AAVance study would be consistent with this sequence of events. The fact that the first four patients of the Lysogene phase 1 study who underwent gene therapy with a first‐generation vector administered at a significantly lower dose and carrying a weaker promoter did not experience similar white matter lesions on early or on prolonged follow‐up ¹⁴ is in line with an overexpression‐related complication.

Importantly, the lesions around the injection sites in patients of the AAVance study were self‐limiting, remained local, stopped progressing after 12–18 months, and were not associated with clinically significant sequels in the study patient or other AAVance study subjects. Studies of intracerebral administration of the same gene therapy vector in dogs and nonhuman primates showed that vector genome copy levels are highest near the sites of injection and decrease proportionally to the distance from the sites of injection. ¹⁵ These results support the hypothesis of a local effect related to overproduction of sulfamidase that specifically affects cells in the vicinity of the injection trajectory.

This study implicates that intracerebral gene therapy carries a risk of excessively high local levels of transgene expression near the sites of vector injection, causing local tissue damage. The issue of intracerebral gene therapy‐related brain tissue damage, also observed in other clinical trials for different diseases, was discussed at a recent FDA Advisory Committee Meeting. ⁴³ FDA Committee members highlighted the need for frequent monitoring by MRI and stressed that evaluation should be made on a case‐by‐case basis, weighing benefits versus risk, in the context of the natural history of the disease. ⁴⁴ MPSIIIA is a devastating disease, leading to inexorable neurological decline and death, without curative therapy options. The forthcoming analysis of the therapeutic efficacy of LYS‐SAF302 in the AAVance trial will inform on the potential benefit of treatment versus the risk of localized and apparently clinically silent white matter lesions.

Conflict of Interest

Marianna Bugiani, Truus E.M. Abbink, Arthur W.D. Edridge, Lia van der Hoek, Anne E. J. Hillen, Niek P. van Til, Gino V. Hu‐A‐Ng, Marjolein Breur, and Marjo S. van der Knaap have no conflicts of interest. Karen Aiach, Philippe Drevot, Michaël Hocquemiller, Ralph Laufer, are employees and shareholders of Lysogene. Frits A. Wijburg is principal investigator in the AAVance trial, sponsored by Lysogene.