Targeted gene correction of FKRP by CRISPR/Cas9 restores functional glycosylation of α-dystroglycan in cortical neurons derived from human induced pluripotent stem cells

Mutations in genes required for functional glycosylation of α-dystroglycan cause a group of congenital muscular dystrophies associated with brain malformations, referred to as dystroglycanopathies. The lack of isogenic, physiology-relevant human cellular models has limited our understanding of the cortical abnormalities in dystroglycanopathies. Here we generate induced pluripotent stem cells (iPSCs) from a severe dystroglycanopathy patient with homozygous mutations in the ribitol-5-phosphate transferase gene, FKRP. We carry out targeted gene correction in FKRP-iPSCs using CRISPR/Cas9-mediated genome editing. We characterise the directed differentiation of FKRP- and corrected-iPSCs to neural stem cells, cortical progenitors and cortical neurons. Importantly, we show that targeted gene correction of FKRP restores functional glycosylation of α-dystroglycan in iPSC-derived cortical neurons. We independently validate this result by showing targeted gene mutation of FKRP disrupts functional glycosylation of α-dystroglycan. This work demonstrates the feasibility of using CRISPR/Cas9-engineered human iPSCs for modelling dystroglycanopathies and provides a foundation for therapeutic development. Highlights Generation of FKRP-iPSCs for modelling cortical abnormalities in dystroglycanopathies Precise gene correction by CRISPR/Cas9-mediated genome editing Directed differentiation of isogenic control and FKRP-iPSC to cortical neurons Functional glycosylation of α-dystroglycan is restored in cortical neurons derived from CRISPR/Cas9-corrected iPSCs Targeted gene mutation of FKRP disrupts functional glycosylation of α-dystroglycan in cortical neurons


Introduction
Post-translational processing of dystroglycan is critical for its function as a cell surface receptor in a variety of fetal and adult tissues. The dystroglycan precursor is cleaved into non-covalently associated α-and β-subunits, forming an integral component of a multiprotein complex. The mature α-dystroglycan is a heavily glycosylated peripheral membrane protein whose molecular weight ranges from 100 to 156 kDa, depending on the tissue-specific glycosylation. In contrast, the βdystroglycan is a 43 kDa transmembrane protein that links to the actin cytoskeleton via interaction with dystrophin (Barresi and Campbell, 2006). Defective O-linked glycosylation of α-dystroglycan is a common pathological hallmark associated with a number of genetic syndromes, encompassing symptoms from muscular dystrophies to ocular defects, cognitive deficits, and cortical malformations (cobblestone lissencephaly) in the central nervous system (CNS). This group of autosomal recessive disorders are commonly referred to as secondary dystroglycanopathies. (Godfrey et al., 2011). Currently there is no cure or effective treatment for dystroglycanopathies.
The effort of identifying causative gene mutations in dystroglycanopathies has shed light on a novel mammalian glycosylation pathway (reviewed in Freeze, 2013;Yoshida-Moriguchi and Campbell, 2015).
Allelic FKRP mutations lead to the widest spectrum of clinical severities, ranging from the mild lateonset limb-girdle muscular dystrophy without neurological deficits (e.g. LGMD2I) to congenital muscular dystrophy with severe CNS abnormalities (e.g. Walker-Warburg syndrome) (Godfrey et al., 2007;Godfrey et al., 2011). Specifically, malformations of cortical development are pathological features at the severe end of the dystroglycanopathy spectrum (Barkovich et al., 2012;Beltran-Valero de Bernabe et al., 2004;Devisme et al., 2012). The absence of physiology-relevant human cellular models has hindered our understanding of mechanisms underlying CNS involvement in dystroglycanopathies and our ability to test potential drug targets in a neural-specific context. In addition, the patient's neural tissue is not easily accessible and there is no appropriate isogenic control for the patient's tissue. Induced pluripotent stem cells (iPSCs), which are remarkably similar to embryonic stem cells (ESCs), can be generated from human somatic tissues (Takahashi et al., 2007). Recent studies have demonstrated success in using patient-specific iPSCs to model a variety of neurodevelopment and neurodegenerative disease Ryan et al., 2013;Shi et al., 2012a). These studies suggest that patient-specific iPSC-derived cortical neurons will allow us to study mechanisms underlying CNS involvement in the severe forms of dystroglycanopathies.
In this study we test the hypothesis that patient-specific iPSC-derived cortical neurons can recapitulate pathological hallmarks in dystroglycanopathies. We generate the first human FKRP-iPSCs from a congenital muscular dystrophy patient with severe CNS abnormalities. We carry out CRISPR/Cas9-mediated genome editing (Cong et al., 2013;Mali et al., 2013) and differentiate isogenic pairs of iPSCs to cortical neurons. We show that, for the first time, targeted gene correction of FKRP restores α-dystroglycan functional glycosylation in iPSC-derived cortical neurons, whereas targeted gene mutation of FKRP disrupts α-dystroglycan glycosylation. Taken together, our isogenic pairs of iPSC-derived cellular models will further elucidate mechanisms underlying the CNS involvement in FKRP-deficient dystroglycanopathies.

Generation and characterisation of FKRP-iPSCs derived from a dystroglycanopathy patient with CNS abnormalities
We obtained dermal fibroblasts from a female individual, previously diagnosed congenital muscular dystrophy with CNS abnormalities, including marked cognitive delay, microcephaly, cerebellar cysts and cerebellar dysplasia. Genetic analysis revealed homozygous FKRP c.1364C>A (p.A455D) mutations in this affected individual. Both parents were heterozygous carriers of this mutation. To generate patient-specific iPSCs, we reprogrammed the fibroblasts with the FKRP c.1364C>A mutations using a six-factor reprogramming technology based on a doxycycline-inducible system (Wang et al., 2011). The initial characterisation of several independent FKRP-iPSC lines by qPCR revealed the gene expression of pluripotency markers, such as NANOG, OCT4 and REX1 ( Figure   S1). Among these patient-specific FKRP-iPSC lines, we focused on the FKRP-iPSC line 1-6, which exhibited the normal karyotype, suggesting these iPSCs are genetically stable ( Figure 1B).
Immunocytochemistry confirmed the expression of pluripotency markers, such as NANOG, OCT4, Tra-1-60 and SSEA4 in the established FKRP-iPSC line ( Figure 1C). In addition, in vitro differentiation of the FKRP-iPSCs formed embryoid bodies with cell types representing all three embryonic germ layers, confirmed by immunocytochemistry in specific cell lineages, such as β-III tubulin (Tuj1) for ectoderm, smooth muscle actin (SMA) for mesoderm and α-fetoprotein (AFP) for endoderm ( Figure 1D).

Targeted gene correction of FKRP-iPSCs using CRISPR/Cas9 mediated genome editing
A common issue of patient-specific iPSCs in disease modelling is the lack of appropriate isogenic control cells, causing concerns about the effect of genetic backgrounds on phenotypic variability. To overcome this issue, we applied a precise genome editing strategy to correct the FKRP c.1364C>A (p.A455D) mutation using site-specific endonuclease CRISPR/Cas9 stimulated homologous recombination (Cong et al., 2013;Mali et al., 2013), which consists a single guide RNA (sgRNA), the Cas9 nuclease and a donor targeting vector. To avoid potential off-target effects, we used an optimised computational algorithm (http://crispr.mit.edu) to identify appropriate sgRNAs for utilizing the CRISPR/Cas9 to generate DNA double strand breaks (DSB) near the FKRP mutation ( Figure   2A). This sgRNA has high on-target and low off-target scores (Table S1). To facilitate the screening process for homologous recombination events, we used an approach based on the transposon piggyBac (PGK-puro∆tk) selection cassette in the targeting donor vector, which enables both positive and negative selections (Yusa, 2013).
We PCR-amplified two 1-kb fragments flanking the CRISPR/Cas9 target site close to the FKRP c.1364C>A allele, which was simultaneously corrected (FKRP c.1364C). The two fragments (homology left and right arms) flanking a piggyBac (PGK-puro∆tk) selection cassette were assembled together into a targeting donor vector ( Figure 2B). DNA sequences at the junctions between piggyBac (PGK-puro∆tk) selection cassette and homology arms were modified to accommodate TTAA sequences, yet coding the same amino acids after excision of the selection cassette ( Figure 2B). We electroporated the site-specific CRISPR/Cas9 plasmids with the targeting donor vector (FKRP c.1364C) into the iPSCs. By positive selection with puromycin, iPSCs with an integrated donor vector formed puromycin-resistant clones, which were picked for rapid PCR genotyping ( Figure S2). We identified 3 homozygously and 27 heterozygously targeted independent clones (Table 1), which were confirmed by sequencing (data not shown). To excise the piggyBac (PGK-puro∆tk) selection cassette, 2 homozygously targeted FKRP (c.1364C)-iPSC clones (5.19 and 3.16) were electroporated with piggyBac transposase expression plasmid, followed by negative selection in culture media containing 1-(2-Deoxy-2-fluoro-β-D-arabinofuranosyl)-5-iodo-2,4(1H,3H)-Pyrimidinedione (FIAU), a thymidine analogue that would be processed to toxic metabolites in the presence of piggyBac (PGK-puro∆tk). PCR genotyping identified 11 biallelicly corrected-iPSC clones that had the selection cassette completely excised without re-integration ( Figure S3 and Table 1). The biallelicly corrected-iPSC clones (5D17, 5D23 and 3B17) were sequenced to confirm a clean correction of the FKRP mutation and the engineered selection cassette excision site ( Figure   2C). The biallelicly corrected iPSC lines retained normal karyotypes ( Figure 2D) and pluripotency (data not shown). In addition, we sequenced the top 5 potential off-target sites ( Table S1) and confirmed that no mutations were introduced during genome editing (data not shown).

corrected-iPSCs
To investigate the potential of iPSCs for modelling neural pathogenesis in dystroglycanopathies, we used a serum-free neural induction medium to derive primitive neural stem cells (NSCs) from iPSCs (Yan et al., 2013). Immunocytochemistry confirmed that NSCs derived from FKRP-and CRISPR/Cas9 corrected-iPSC lines express classic NSC markers, including SOX1, SOX2 and nestin ( Figure 3A,B). In terms of the efficiency of neural induction, we did not observe discernable difference between NSCs derived from FKRP-and three corrected-iPSC lines, 5D17, 5D23 and 3B17 ( Figure 3C,D). Quantification of SOX1 and SOX2 positive NSCs showed >99% efficiency from both FKRP-and three corrected iPSC lines ( Figure 3C,D).
Subsequently, we directed the differentiation of FKRP-and corrected-NSC lines towards cortical projection neurons using a well-established protocol, which recapitulates important stages in human cortical development (Shi et al., 2012b). After switching to the Neural Maintenance Medium for one week, we confirmed the identity of FKRP-and corrected-iPSC derived cortical stem and progenitor cells, which express classic cortical stem cell markers, PAX6, OTX2 and vimentin ( Figure

Functional glycosylation of α-dystroglycan is restored in cortical neurons derived from CRISPR/Cas9 corrected-iPSCs
Next, we investigated whether we could detect functional glycosylation of α-dystroglycan in cortical neurons derived from our corrected-iPSC lines. To do this, we performed immunoblot using the IIH6 antibody, which recognises the laminin-binding glyco-epitope on α-dystroglycan (Michele et al., 2002). Wildtype mouse muscle and brain lysates were used as positive controls to show differential glycosylation of α-dystroglycan in a tissue-specific manner. In addition, we included lysate of cortical neurons from a non-isogenic, wildtype iPSC20 line (hereinafter called WT-iPSCs) (Wang et al., 2011). We showed that the molecular weight of glycosylated α-dystroglycan in WT-iPSC derived cortical neurons is similar to that in mouse brain lysate (~120 kDa) and less than that in muscle lysate (~156 kDa) ( Figure 4A), consistent with previously reported tissue-specific glycosylation of αdystroglycan ( Barresi and Campbell, 2006). Furthermore, IIH6 reactivity on immunoblot was almost not detected in cortical neurons derived from FKRP-iPSCs ( Figure 4A and B). In contrast, cortical neurons derived from CRISPR/Cas9 corrected-iPSC lines clearly showed IIH6 reactivity, indicating restored glycosylation of α-dystroglycan ( Figure 4B).
Following the detection of glycosylated α-dystroglycan in corrected cortical neurons, we investigated whether the laminin-binding activity is associated with the IIH6 glyco-epitope on α-dystroglycan. We confirmed that laminin-binding is disrupted in cortical neurons derived from FKRP-iPSCs ( Figure   4D), whereas cortical neurons derived from corrected-iPSC lines showed strong laminin-binding activities ( Figure 4D). Together, these results demonstrated a restoration of functional glycosylation of α-dystroglycan in CRISPR/Cas9-corrected corrected cortical neurons.

Targeted gene mutation of FKRP disrupts functional glycosylation of α-dystroglycan in cortical neurons derived from iPSCs
To independently validate the results from our FKRP-and corrected-iPSC derived cortical neurons, we used the same genome editing strategy ( Figure S4A) to knock-in the FKRP c.1364C>A (p.A455D) mutation into the WT-iPSCs (Wang et al., 2011). We constructed a donor targeting vector carrying the FKRP mutation for CRISPR/Cas9-mediated homologous recombination. We electroporated the site-specific CRISPR/Cas9 plasmids with the targeting donor vector (FKRP c.1364C>A) into the WT-iPSCs ( Figure S4B). Using puromycin positive selection and PCR genotyping ( Figure S2), we identified 2 homozygously targeted clones (a15 and c43), which were expanded for selection cassette excision (Table 2). Following FIAU negative selection and PCR genotyping ( Figure S3), we identified 2 biallelicly FKRP mutated-iPSC clones (C29 and C31) that have the selection cassette completely excised without re-integration ( Table 2). The FKRP mutated-iPSC clones (C29 and C31) were sequenced to confirm the biallelic knock-in of FKRP c.1364C>A (p.A455D) mutations and the engineered selection cassette excision site ( Figure S4C). The wildtype-and FKRP mutated-iPSC lines were differentiated to NSCs ( Figure S5) and subsequently cortical neurons for functional analysis. Compared with cortical neurons derived from WT-iPSCs, targeted gene mutation of FKRP disrupted the IIH6 reactivity ( Figure 4A and C) in cortical neurons derived from FKRP mutated-iPSC lines, indicating loss of functional glycosylation of α-dystroglycan.
Together, our results demonstrated that target gene mutation of FKRP by CRISPR/Cas9-mediated genome editing could recapitulate pathological hallmarks of dystroglycanopathy in cortical neurons.

Discussion
The combination of patient-specific iPSCs and CRISPR/Cas9-mediated genome editing has enormous potential for disease modelling, as well as for drug discovery and development. We report here that CRISPR/Cas9-induced homologous recombination together with the piggyBac positive/negative selection cassette is a powerful and versatile strategy, which allows precise modification of the mammalian genome at a single base-pair level without leaving footprints. For the first time, we demonstrate a precise genome editing at the FKRP locus in human iPSCs. Using our genome editing strategy, we were able to achieve 7-10% bialleic targeting efficiency at the FKRP locus (Table 1) and no mutations were observed at the top 5 predicted off-target sites (data not shown). According to Leiden Open Variation Database (http://www.lovd.nl/), at least 95% reported FKRP DNA variants are missense mutations (969 out of 1018 cases). Homozygous null FKRP mutations are exceptionally rare and a single case of Walker-Warburg syndrome with a mutation affecting the translational start site was reported (Van Reeuwijk et al., 2010). Most patients carry heterozygous or homozygous FKRP mutations or, less frequently, a combination of a nonsense or frameshift mutation with a missense mutation (5 out of 38 patients in our research cohort). Among these mutations, FKRP c.826C>A (L276I) mutation is one of the most common variants and causes a mild form of dystroglycanopathy (LGMD2I). Our genome editing strategy can be applied to precisely correct these FKRP missense mutations and by implication small insertion/deletion mutations (INDELs), as well as similar type of mutations in other dystroglycanopathy genes. Recent advances in the delivery of Cas9-sgRNA ribonucleoproteins (Cas9 RNPs) (Liu et al., 2015), may further improve the efficiency of biallelic gene targeting. The advantage of using Cas9 RNPs for genome editing is the rapid Cas9 cleavage action and protein turnover in the cells within 24 h of delivery, which increases the gene targeting efficiency and reduces the off-target mutation rate that is critical for future therapeutic application.
The process of cellular reprogramming from fibroblasts to iPSCs followed by directed differentiation from iPSCs to NSCs and subsequently to cortical neurons involves dramatic changes of gene expression profile of a cell. The functional glycosylation of α-dystroglycan is a consequence of a complex biosynthetic pathway orchestrated by at least 17 known enzymes to form the ECM ligandbinding moiety (Yoshida-Moriguchi and Campbell, 2015). Our results indicate that genes involved in functional glycosylation of α-dystroglycan are expressed during in vitro corticogenesis. Interestingly, the molecular weight of α-dystroglycan varies due to tissue-specific O-glycosylation (Barresi and Campbell, 2006). Consistent with previous studies, we show that the molecular weight of glycosylated α-dystroglycan in cortical neurons derived from human iPSCs is similar to that in the mouse brain and less than that in mouse muscle (Figure 4). Importantly, we demonstrate that Mutations in FKRP and other known causative genes together account for approximately 50-60% of severe forms of dystroglycanopathies (Roscioli et al., 2012;Vuillaumier-Barrot et al., 2012).
Previously a zebrafish based study showed cooperative interaction between ISPD, FKTN and FKRP (Roscioli et al., 2012). Recent studies demonstrated the enzymatic function of ISPD, FKTN and FKRP, which synergistically add tandem ribito-5-phosphate onto α-dystroglycan (Gerin et al., 2016;Kanagawa et al., 2016). Despite these advances, the effects of FKRP mutations and defective glycosylation of α-dystroglycan in a disease context still remain poorly understood. Our isogenic pairs of control-and FKRP-iPSC derived cortical neurons will further elucidate molecular mechanisms underlying cortical development in health and disease using a systems-based approach, similar to transcriptome analysis of in vitro corticogenesis from human ESCs (van de Leemput et al., 2014). Importantly, this work will not only elucidate the role of glycosylated dystroglycan in the brain, but will also have significant impacts on studying its role in other tissues.
Apart from muscle degeneration, dilated cardiomyopathy is a frequent associated feature of FKRPdeficient dystroglycanopathy (Mercuri and Muntoni, 2013). The isogenic pairs of human iPSCs generated in this study have enormous potential in many other research avenues. In addition to neural differentiation, the isogenic pairs of control-and FKRP-iPSC lines can be used for myogenic and cardiac differentiation with established protocols (Bellin et al., 2013;Chal et al., 2016) for studying pathological mechanisms underlying muscle degeneration and dilated cardiomyopathy in muscular dystrophy patients. In addition, one can envisage the use of progenitor cells derived from CRISPR/Cas9-corrected iPSCs for autologous cell replacement therapies in the future. The advantage of using autologous, CRISPR/Cas9-corrected progenitor cells is the reduced risk of immunological rejection and need for immunosuppression.
In conclusion, we have established two isogenic pairs of human iPSC models for a severe form of FKRP-deficient dystroglycanopathy. A significant impact of this study is that other human iPSCbased dystroglycanopathy models will be generated. Collectively, these iPSC models will provide powerful in vitro platforms that can be exploited for setting up high-content drug screens using quantifiable phenotypes as readouts. Importantly, while more than 90% of drugs fail during development (DiMasi et al., 1991), our isogenic pairs of human iPSC lines will provide an invaluable resource for systematic chemical and genetic screens capable of reverting pathophysiological defects in physiology-relevant cellular models. This will facilitate the drug discovery and development for treating dystroglycanopathies.     neurons is similar to that in the mouse brain (~120 kD) and lower than that in the mouse muscle (~156kD). IIH6 reactivity was almost not detected in cortical neurons derived from FKRP-iPSCs.   (Wang et al., 2011). In vitro differentiation and analysis were carried out as described (Wang et al., 2011).

Vector construction and CRISPR/Cas9 mediated genome editing
For targeted gene correction, we used Gibson Assembly (New England BioLabs) to construct the targeting donor vector. Briefly, the 1-kb left and right homology arms (LHA and RHA) were PCRamplifed from parental FKRP fibroblast. The FKRP c.1364C>A (p.A455D) mutation on the LHA was simultaneously corrected with a modified primer (Table S2). The piggyBac (PGK-puro∆tk) selection cassette and vector backbone were PCR-amplified from the pMCS-AAT_PB-PGKpuroTK plasmid (Yusa et al., 2011). The four PCR fragments have 40-bp overlappings from end to end (primers are listed in Table S2) and joined together using Gibson Assembly Master Mix. The hCas9 plasmid was a gift from George Church (Addgene plasmid # 41815) (Mali et al., 2013). Short DNA fragments containing the FKRP sgRNA target sequence (Table S1) were generated by annealing two primers (Table S2), which create two overhangs ready for cloning into BsaI-digested p1261_U6_BsaI_gRNA_plasmid (kind gift of Sebastian Gerety, Wellcome Trust Sanger Institute).
The expression of sgRNA is under the control of U6 promoter. The cell suspension was transferred to a cuvette and electroporated using Amaxa Nucleofection 2b device with program B16. For targeted gene mutation, the same strategy was used to construct the donor targeting vector carrying the FKRP c.1364C>A (p.A455D) mutation on LHA using primers in Table S2.
When reaching 90% confluence (approximately within 3-4days), cells were re-seeded (60,000/cm 2 ) in Neural Maintenance Medium (with 5 µM Y-27632 dihydrochloride for 24 h), followed by media change every other day. In vitro cortigenesis occurs in the following weeks. Cultured cells were then harvested or fixed at specific time points for further analysis.

Immunocytochemistry
Cells were fixed with 4% PFA for 15 mins at room temperature (RT). Prior to immunocytochemistry, cells were permeabilized with 0.1% Triton/phosphate-buffered saline (PBS) for 15min at RT and blocked with either 10% goat serum/PBS or 10% horse serum/PBS. Primary antibodies used were: OCT4 ( SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher) was applied to PVDF membranes and signals were visualized using a Bio-Rad ChemiDoc MP imaging system.

Laminin overlay analysis
Protein samples were extracted, resolved and transferred to PVDF membranes as described above.
The PVDF membrane was then blocked with 5% skimmed milk in laminin binding buffer (LBB), containing 10 mM triethanolamine, 140 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 (pH adjusted to 7.6), followed by incubation with laminins from Engelbreth-Holm-Swarm murine sarcoma basement membrane (Sigma, L2020) in LBB with a final concentration of 5 µg/ml at 4 o C overnight. The PVDF membrane was then incubated with the pan-laminin antibody (1:1000; Sigma, L9393), washed with 1x LBB with 0.1% Tween and incubated with HRP conjugated anti-rabbit IgG (Jackson laboratories). Signal detection using chemiluminescence substrate was performed as above.