Gene expression changes in Tay-Sachs disease begin early in fetal brain development

Abstract

Treatment of monogenic disorders has historically relied on symptomatic management with limited ability to target primary molecular deficits. However, recent advances in gene therapy and related technologies aim to correct these underlying deficiencies, raising the possibility of disease management or even prevention for diseases that can be treated pre-symptomatically. Tay-Sachs disease (TSD) would be one such candidate, however very little is known about the presymptomatic stage of TSD. To better understand the effects of TSD on brain development, we evaluated the transcriptomes of human fetal brain samples with biallelic pathogenic variants in HEXA. We identified dramatic changes in the transcriptome, suggesting a perturbation of normal development. We also observed a shift in the expression of the sphingolipid metabolic pathway away from production of the HEXA substrate, GM2 ganglioside, presumptively to compensate for dysfunction of the enzyme. However, we do not observe transcriptomic signatures of end stage disease, suggesting that developmental perturbations precede neurodegeneration. To our knowledge, this is the first report of the relationship between fetal disease pathology in juvenile onset TSD and the analysis of gene expression in fetal TSD tissues. This study highlights the need to better understand the “pre-symptomatic” stage of disease in order to set realistic expectations for patients receiving early therapeutic intervention.

INTRODUCTION

Comprehensive understandings of disease mechanisms have been critical to developing therapies for genetic diseases, including understanding the therapeutic window to maximize benefit. However, due to incomplete understanding of disease pathogenesis and inability to target their molecular mechanism, most genetic disorders are managed only symptomatically. Advances in gene therapy¹ have raised the possibility of treating monogenic disorders without requiring such detailed understanding of pathogenic mechanisms. However, understanding the therapeutic window, the state of disease at that time of therapy, and the extent to which disease progression is reversible will still be critical for developing effective therapies.

Tay-Sachs disease (TSD, MIM#272800) is a lysosomal storage disorder (LSD) caused by loss of function of HEXA, which encodes the alpha subunit of the beta-hexosaminidase enzyme (HEXA). Deficient HEXA activity results in accumulation of the HEXA substrate, GM2 ganglioside, within lysosomes and results in neurodegeneration. Infantile TSD presents shortly after birth with severe neurological deterioration and death occurring by four years². Juvenile³ and adult-onset⁴ forms demonstrate a spectrum of disease severity that roughly correlate with the level of residual enzyme activity. In even the infantile form, newborns appear clinically well at birth², suggesting that restoring HEXA activity early in life could provide significant clinical benefit. Gene transfer⁵ and small molecule⁶ therapies are currently under investigation even though storage of GM2 ganglioside has been reported in human fetal tissues as early as the second trimester⁷. Lysosomal storage and disease pathology have also been noted during fetal development in Gaucher disease⁸, raising the possibility of developmental perturbations in other LSDs that present with storage at birth, which can be observed in the form of cherry red maculae (Sup Table 1)⁹.

In this study, we utilized RNAseq to evaluate the transcriptomes of human fetal brains with genotypes associated with juvenile onset TSD. We find that the transcriptomes are perturbed by 17 week’s gestation, suggesting abnormal neurodevelopment, which has implications for therapeutic intervention for TSD. If brains have not developed properly in utero, then therapy administered even immediately after birth may have limited potential. This work raises important questions about the state of brain development at birth and the therapeutic window required for optimizing outcomes.

MATERIALS AND METHODS

Study approval and sample collection

The index case was evaluated at the NIH Clinical Center under protocol (NCT00029965, 02-HG-0107). Fetal brain samples were received frozen from the clinics where the terminations were performed. Affected brain samples included brainstem, cerebellum, cortex, and white matter. Samples obtained from NeuroBioBank included brainstem, cerebellum, cortex, thalamus, and white matter. Complete sample list is included as supplemental table 2.

Ganglioside analysis of human brain by high-performance thin-layer chromatography (HPTLC)

10 mg of tissue sample from affected and control cortex were weighed, and the total lipids were sequentially extracted by 1:1, v/v and 1:2, v/v of chloroform-methanol (C-M), and 30:60:8, v/v/v of chloroform-methanol-water (C-M-W). Total lipids were dried under N₂ gas and separated to polar (upper) and non-polar (lower) phases by Folch’s method¹⁰. To enrich gangliosides, mild alkaline treatment of upper phase was performed: the organic solvent in upper phase was evaporated and incubated in 0.1 N sodium hydroxide (NaOH) at 40°C for 2 h. The solution was neutralized in acetic acid, diluted twice with 0.1 M potassium chloride (KCl) and desalinated by Sep-Pak C18 column (Waters, Milford, MA, USA, Part # WAT020805). Prior to sample application, Sep-Pak C18 column was washed by 5 column volume each of methanol and water and 2 column volumes of 0.1M KCl. After sample application, the salt was washed out by 5 column volumes of water and gangliosides were eluted by 2 ml of methanol and 6 ml of 1:1, v/v of C-M. The eluent was dried, dissolved proportionally to brain weight in 30:60:8, v/v/v of C-M-W (0.5 mg/ml) and the aliquot (10 μl, gangliosides in 5 mg brain tissue) was applied to HPTLC followed by development in 55:45:10, v/v/v of C-M-0.2% CaCl₂. HPTLC plate was dried and sprayed with resorcinol reagent. HPTLC was tightly sealed by a glass plate using clips and the plate was located facing down in 90 C oven for 15 min. The gangliosides were visualized as blue-violet bands. The standards (Matraya LLC, State College, PA, USA, Cat # 1065&1508) were applied in the same HPTLC plate and Rf values were compared to bands in samples. Resorcinol reagent was prepared by adding 80 ml of concentrated hydrochloric (HCl) acid and 0.25 ml of 0.1 M copper sulfate (CuSO₄) to aqueous resorcinol solution (200 mg resorcinol (Sigma-Aldrich, St. Louis, MO, USA, Cat # R1000) in 10 ml of water) and adjusting final volume to 100 ml with water.

RNA extraction

Total RNA was extracted from tissue samples using the Maxwell^® RSC simplyRNA Tissue Kit (Promega, Madison, WI, USA, Cat #AS1340) according to manufacturer’s protocol. RNA quality was assessed using a Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and samples with RNA Integrity Numbers below 7 were discarded.

Sequencing

Library preps for next generation sequencing were performed using the Illumina TruSeq Stranded Total RNA with Ribo-Zero Globin kit (Illumina, San Diego, CA, USA, Cat # 20020613) and sequenced on a Novoseq 6000 with 150 bp paired end reads at an average depth of 1.3E8 reads per sample.

Alignment and quantification

Paired-end 150bp reads were aligned to the reference human genome (GRCh38.104) using STAR¹¹ v2.7.8 with 149bp junction gaps. Sorted bam files were quantified using RSEM¹² v1.3.2, where the transcriptome reference was generated using the GRCh38.104.gtf file. Disease state was confirmed by visualization in IGV confirming the presence of both HEXA variants in each sample’s bam file. After TMM normalization, differential expression analysis was performed using edgeR¹³ v3.30.0. Statistically significant differentially expressed genes were those with Benjamini-Hochberg-corrected p-value (FDR) < 0.05.

Pathway analyses

Ingenuity Pathway Analysis (IPA)¹⁴ was performed using the edgeR output table and running a core analysis for pathway enrichment.

Gene Set Enrichment Analysis (GSEA v4.1.0)^15,16 was performed using raw count tables generated by RSEM and GSEA’s pre-processing recommendations for each sample generated by RSEM against KEGG gene sets.

BiNGO¹⁷ analysis was performed using the edgeR output table. DE genes were mapped to their Swiss-Prot or longest TrEMBL protein using biomaRt (). BiNGO was performed against Gene Ontology “Biological Process” gene sets using the ontology file (.obo) and human annotation file (.gaf) from the GO website (2022-07-01 release).

Bulk deconvolution

Bisque¹⁸ was used to perform marker-based bulk deconvolution on TPMs output by RSEM. Cell type markers were derived from scRNA-seq studies of human fetal brain tissue, specifically Polioudakiset al 2019¹⁹and Nowakowski et al 2017²⁰. Average log2-fold changes were used as weights for the analysis.

RESULTS

In this study we evaluated a patient with juvenile-onset TSD (under study NCT00029965) and the transcriptomes of two second trimester TSD fetal brains from the same family, which were compared to two age matched controls obtained from the NIH NeuroBioBank. The index case was born to parents of non-Ashkenazi heritage and diagnosed with juvenile-onset TSD at 3.5 years old, 1.5 years after the recognition of symptoms. Genetic testing revealed compound heterozygosity for likely pathogenic variants in HEXA: NM_000520 c.346+1G>T and c.1559G>A (Fig 1A, Sup Fig 1). Prenatal testing in two subsequent pregnancies revealed that both fetuses carried both pathogenic HEXA alleles. The pregnancies were terminated at 17 weeks and brain tissues were sampled at multiple sites (Sup Table 2). We confirmed GM2 ganglioside accumulation in affected brain tissue by HPTLC (Fig 1B). Here, we demonstrate that disease processes have begun during fetal development even for this less severe form of disease.

Figure 1. Fetal brains with mild TSD genotypes demonstrate signs of disease.

A) Pedigree demonstrating segregation of pathogenic alleles and basic clinical findings which are consistent with juvenile onset TSD. B) HPTLC of gangliosides extracted from control and TSD brains demonstrates excess GM2 ganglioside in TSD affected brain. C) Expression of the components of SPT is decreased in affected and unaffected samples from multiple regions of each brain. Each point represents TPM for a single RNA sample and p values are adjusted for multiple testing from edgeR analysis. D) Expression of genes across sphingolipid metabolism are generally decreased and include ganglioside synthesis up to the point of GM3 ganglioside.

We extracted total RNA from each tissue sample and performed RNAseq with 150bp paired end reads at an average depth of 1.3E8 reads/sample. We utilized STAR¹¹ and RSEM¹² to map and quantify the reads against the GRCh38 human reference genome, and edgeR¹³ to perform differential gene expression analysis (additional details in supplemental methods). We identified more than 7,000 differentially expressed genes (DEGs) with FDR < 0.05 (Sup Fig 2). Due to the limited size of this study, changes in individual genes are likely more indicative of interpersonal variation than underlying physiology. Thus, we found analyzing the differential expression patterns for multiple genes in the same pathway to be much more robust. Gene ontology (GO) analysis using BiNGO¹⁷ revealed decreased expression of key genes involved in the ganglioside biosynthetic pathway²¹ (Sup Fig 3), including the formation of the precursor lipid, ceramide. We saw decreased expression of the three genes encoding the subunits of serine palmitoyltransferase (SPTLC1, SPTLC2, SPTSSA), the rate limiting commitment step of de novo sphingolipid synthesis (Fig 1C). Importantly, we observed coordinated downregulation of the first three glycosyltransferase genes (UGCG, B4GALT5, and ST3GAL5) in the ganglioside biosynthetic pathway, which strongly suggests reduced synthesis of GM2 ganglioside (Fig 1D). We also observed a slight increase in SGPL1, which irreversibly catalyzes the degradation of S1P and regulates the exit of sphingolipid substrate from the metabolic pathway²². These results demonstrate that HEXA deficiency has modulated the transcriptome even in the less severe juvenile-onset form of disease and even as early as the second trimester.

Global evaluation of these transcriptomes suggested that the TSD fetal brain is less developed compared to age-matched controls. Principal component analysis (PCA) distinguished TSD samples from most control samples based on a single PC, which accounted for 37.5% of the variance in the data (Fig 2A). Specifically, we found that TSD samples cluster with control forebrain regardless of the brain region from which they were extracted, suggesting that TSD samples have genetic signatures that are inappropriate for their sampling location. Evaluation of cell type specific markers (Sup Table 3) using Bisque¹⁸ suggested decreased abundance of oligodendrocyte precursors, astrocytes, endothelia, and pericytes in TSD samples (Sup Fig 4). The decreased markers of endothelia and pericytes may be evocative of decreased S1P signaling²³, which may be a byproduct of decreased sphingolipid metabolism. BiNGO analysis also revealed increased expression of genes involved in cell cycle progression, suggesting increased proliferation (Fig 2B) consistent with previous results reported in organoids²⁴, mice²⁵, and iPSCs²⁶. Additionally, BiNGO predicted increased activity of stem cell maintenance pathways, another carefully regulated process in the developing brain (Fig 2C). Taken together, these data suggest that developmental processes may be perturbed due to an imbalance of differentiation and proliferation.

Figure 2. HEXA deficiency impairs brain development.

A) PCA separates samples based on PC1, with control forebrain samples clustering near TSD samples. B) Expression of genes associated with GO terms of cell division and proliferation are increased. C) Genes from the Somatic Stem Cell Population Maintenance GO term demonstrate modest elevation. D) Genes involved in CLEAR signaling are largely decreased in TSD samples as well as the control cortex and white matter. Values are normalized within each row and columns are grouped by brain region and affected state.

While our data suggest disease pathology begins in utero, we also observe a lack of pathology associated with late-stage disease. We did not observe upregulation of immune pathways, implying that neuroinflammatory processes have not yet begun. We also did not find upregulation of TFEB, the master regulator of lysosomal activity²⁷, which has been implicated in multiple LSDs²⁸. Ingenuity Pathway Analysis (IPA)¹⁴ identified the CLEAR signaling network upstream of TFEB as the most significant pathway (Sup Table 4), and strongly differentially expressed CLEAR genes (FDR < 0.05, abs log₂FC > 1) were almost exclusively decreased (Fig 2D). We also observed a concurrent decrease in expression of many TFEB target genes (Sup Fig 5) as well as decreased expression of lysosomal genes (Sup Fig 6) when evaluated by GSEA^15,16. Interestingly, CLEAR network and TFEB target genes have lower expression in control forebrain samples, implying that these patterns may be characteristic of the developmental state of those brain regions. These analyses suggest that developmental perturbations have begun before the onset of neurodegeneration.

In summary, we observe shifts in the expression of genes directly related to HEXA deficiency as well as global effects that suggest abnormal neurodevelopment. However, we do not observe hallmarks of neurodegeneration such as significant inflammation or apoptosis. Taken together, these results suggest that neurodevelopmental perturbations precede neurodegeneration in TSD, and that these processes begin early in fetal brain development.

DISCUSSION

This study is the first since the 1970s to evaluate the effects of TSD on fetal development. We confirmed the basic findings from those early studies that lysosomal storage begins early in gestation and applied a genome wide analysis to evaluate how the disease has impacted brain development.

We have demonstrated that accumulation of GM2 ganglioside begins by the second trimester even in juvenile onset disease and that fetal brains affected with TSD exhibit abnormal transcriptomes. Transcriptomic profiles of TSD brains suggested metabolic compensation for GM2 ganglioside accumulation through decreased expression of most genes involved in the ganglioside biosynthetic pathway. This has implications beyond preventing generation of new gangliosides, as sphingolipids are integral to many other cellular processes, including the signaling pathways of S1P. While we have insufficient sample to directly measure S1P, other transcriptomic signatures suggest that S1P signaling may be impacted. Perturbation of other pathways downstream of ceramide and sphingolipid metabolism likely have more global effects which we observe as blunted development relative to controls. It is also possible that additional compensatory transcriptomic alterations are induced to prevent more extensive damage and maintain physiologic homeostasis. While we cannot probe the extent of compensation, these mechanisms are not effective in counteracting the loss of HEXA activity as demonstrated clinically by slowed then arrested development and eventual neurologic decline in juvenile TSD patients like the proband described in this report. The more rapid clinical presentation seen in infantile TSD suggests that brain development may be even more significantly affected when HEXA activity is completely absent.

Signs of overt neurodegeneration, which importantly include neuroinflammation, are not yet present in the transciptome, possibly providing a therapeutic window for treatment. Reducing or preventing neuroinflammation is a primary therapeutic focus for a significant portion of neurodegenerative disease research. Future therapies, including gene therapy, administered prenatally may represent a tractable therapeutic window for treating TSD. Prenatal diagnosis of TSD has become less common due to the success of carrier screening for common pathogenic variants in HEXA in high-risk populations. While in utero therapy may have significant potential for those individuals who have known family history of TSD, this will not be available to couples who do not know they are carriers. However, advances in prenatal genetics²⁹ may improve the ability to identify affected fetuses in the near future. Understanding the state of the brain at birth will be critical to determining the optimal therapeutic window for treating TSD.

Further studies are obviously needed to discern how the transcriptomic changes we observe manifest in disease and to understand to what extent these differences might be reversible. Notably, these considerations are not unique to TSD or to LSDs, and altered fetal development should be considered for any monogenic disease whose pathogenic expression in utero cannot be compensated for by maternal function or fetal-specific paralogs. In utero therapy may be needed to effectively treat disorders and such studies are currently underway. A clinical trial to evaluate in utero enzyme replacement therapy for some LSDs has begun already been initiated (NCT04532047)³⁰, and prenatal modulator therapy appears to have been successful at preserving pancreatic function in a fetus affected with cystic fibrosis³¹. This study highlights the complex pathogenesis of congenital disorders, the need to carefully establish outcome measures for clinical studies, and most importantly to set reasonable expectations of treatment for families of affected patients.

Synopsis:

Enzyme deficiency during gestation perturbs the transcriptomes of developing human fetal brains. These findings may influence clinical trial design and help set expectations for clinical outcomes even with early postnatal intervention.

Abbreviations

TSD

Tay Sachs disease

LSD

Lysosomal storage disorder

HEXA

Gene encoding the alpha subunit of the beta hexosimidase enzyme

HEXA

Beta hexosimidase enzyme

DEG

Differentially expressed gene

SPT

Serine palmoityl transferase

S1P

Sphingosine-1-phosphate

GO

Gene ontology

IPA

Ingenuity Pathway Analysis

GSEA

Gene Set Enrichment Analysis

CLEAR

Coordinated Lysosomal Expression and Regulation gene network

DATA AVAILABILITY

Raw fastq files are available at SRA as BioProject PRJNA931973. Raw gene counts (after STAR alignment and RSEM quantification) are available at GEO as accession GSE224860.