Premyogenic progenitors derived from human pluripotent stem cells expand in floating culture and differentiate into transplantable myogenic progenitors

Human induced pluripotent stem cells (hiPSCs) are a potential source for cell therapy of Duchenne muscular dystrophy. To reliably obtain skeletal muscle progenitors from hiPSCs, we treated hiPS cells with a Wnt activator, CHIR-99021 and a BMP receptor inhibitor, LDN-193189, and then induced skeletal muscle cells using a previously reported sphere-based culture. This protocol greatly improved sphere formation efficiency and stably induced the differentiation of myogenic cells from hiPS cells generated from both healthy donors and a patient with congenital myasthenic syndrome. hiPSC-derived myogenic progenitors were enriched in the CD57(−) CD108(−) CD271(+) ERBB3(+) cell fraction, and their differentiation was greatly promoted by TGF-β inhibitors. TGF-β inhibitors down-regulated the NFIX transcription factor, and NFIX short hairpin RNA (shRNA) improved the differentiation of iPS cell-derived myogenic progenitors. These results suggest that NFIX inhibited differentiation of myogenic progenitors. hiPSC-derived myogenic cells differentiated into myofibers in muscles of NSG-mdx4Cv mice after direct transplantation. Our results indicate that our new muscle induction protocol is useful for cell therapy of muscular dystrophies.


Results
Continuously stirred floating culture scaled up derivation of myogenic cells from human iPS cells. To obtain sufficient numbers of myogenic cells for cell therapy, we first combined the EZ sphere method 21 with a continuously stirred floating culture system using a bioreactor (Supplementary Figure 1A). As expected, the cell yield was increased (average 5.8-fold, maximum 16.4-fold) by continuous low-speed stirring (Supplementary Figure 1B), but there was no increase in the percentage of myogenic spheres by the stirred suspension culture compared to the original method (Supplementary Figure 1D). In addition, the four iPS cells (253G4, 201B7, 454E2, and 409B2) formed multinucleated myotubes with quite different efficiencies (Supplementary Figure 1C).

Reproducible induction of premyogenic progenitors from human iPS cells using CHIR-99021 and LDN-193189.
For efficient induction of myogenic cells, we thought that induction of the paraxial mesoderm was the most critical step. Therefore, we investigated whether the dual modulation of Wnt and BMP pathways using CHIR-99021 and LDN-193189, recently reported by Chal et al., increases the efficiency of sphere-based muscle induction. First, we examined whether the reported method correctly induces expression of the genes that are expressed in the paraxial mesoderm. For this analysis, we used two normal hiPSC lines (201B7 and 454E2) and two hiPSC lines derived from a patient with congenital myasthenic syndrome (CMS) due to a GFPT1 mutation (GFPT1 #3 and GFPT1 #8). T (BRACHYURY) was transiently expressed in Di-CL medium. TBX6 was induced in Di-CL medium and downregulated in DK-HIFL medium. PAX3 expression was induced in DK-HIFL medium. Finally, PAX7 was induced in all iPSC clones cultured in DK-I medium with good reproducibility (Fig. 1).

Premyogenic progenitors efficiently differentiated into myogenic cells in floating culture.
After CHIR-99021 and LDN-193189 treatment, we collected the differentiating cells using a cell scraper, transferred them to a floating culture at four different time points (protocols 1-4 in Fig. 2A), and cultured them as floating spheres as described by Hosoyama et al. 21 . Then, to evaluate the myogenic activity, we plated one sphere per well in 10% FBS/DMEM medium on collagen-coated 24-well plates. After 4 weeks adhesion culture, the cells were stained for muscle myosin heavy chain (MF20) and MYOGENIN, which is expressed in differentiating myoblasts and myotubes (Fig. 2). Importantly, CHIR-99021 and LDN-193189 treatment increased the numbers of spheres formed compared with the original EZ sphere method. In contrast, the original method gave rise to much fewer and bigger spheres with irregular shapes (Fig. 2B). The highest percentage of myogenic spheres was obtained using protocol 3 or 4, although 409B2 cells efficiently differentiated into myogenic cells using protocol 2 or 3 (Fig. 2B). The average muscle induction efficiency of the new protocol was higher than those of original EZ sphere methods, although there was no statistically significant difference due to the large variation in the inductions using the original method (

Sorting myogenic cells using cell surface markers.
To analyze the properties of myogenic progenitors, we examined cell surface markers on sphere cells derived from the four hiPSC clones (201B7, 253G4, 409B2, and 454E2) at various time points using more than 20 antibodies (data not shown). After suspension culture, all cells were negative for TRA-1-60, TRA-1-81, and SSEA4, suggesting that no undifferentiated iPS cells remained in the culture (data not shown). When examined after six-week sphere culture and four-week adhesion culture, CD271, which was expressed on postnatal myoblasts but not fibroblasts in our preliminary FACS screening for candidates of cell surface markers (data not shown), was expressed on more than 60% of iPSC-derived sphere cells (Fig. 3). After cell sorting, myotubes were formed exclusively by the CD271-positive fraction, but many CD271-positive fraction cells were non-myogenic, indicating that CD271 is not sufficient for purification of myogenic progenitors.
SCIeNTIfIC REPORTS | (2018) 8:6555 | DOI:10.1038/s41598-018-24959-y ERBB3, which has been recently demonstrated to be an excellent cell surface marker on PAX7+ muscle progenitor cells derived from hPSCs using two muscle induction protocols 20 , was expressed on a small fraction of sphere cells. Myotubes were formed exclusively by the ERBB3-positive fraction. Previously, we reported that the combination of CD56 (NCAM) and CD82 effectively enriched hiPSC-derived myogenic cells 23 , but ERBB3 enriched myogenic cells much more powerfully than the combination of CD82 and CD56. Sphere cells contained both CD57-positive and -negative cells. MF20-positive myotubes were formed exclusively by CD57-negative cells (Fig. 3B). Screening of surface markers of adult myoblasts using BD Lyoplate ™ Screening Panels revealed that CD108 was expressed on both fibroblasts and myoblasts from adult skeletal muscle (Nishiyama et al., unpublished observation). Nonetheless, hiPSC-derived myogenic cells were highly enriched in the CD108-negative fraction (Fig. 3B). At present we have no explanation for this discrepancy. CD108 might be up-regulated in myogenic cells postnatally. A commercial M-cadherin monoclonal antibody recognized myogenic cells after non-enzymatic dissociation by Gibco ® Cell Dissociation buffer but did not recognize the molecule after trypsin treatment, making it difficult to isolate myogenic progenitors from tightly packed spheres (data not shown). Fractionation with CD146/MCAM or CD318 failed to enrich myogenic cells (data not shown). We next sorted CD57(−) CD108(−) CD271(+) ERBB3(+) cells (Fig. 3C), and examined their PAX7, MYOD, MYOGENIN, and NFIX expressions by immunocytostaining (Fig. 3D,E). MYOD was expressed in most of the cells. PAX7 was expressed in 30-40% of the sorted cells when fixed within 48 hours after FACS sorting (Fig. 3D). After plating onto collagen-coated dishes in 10% FBS/DMEM, hiPSC-derived CD57(−) CD108(−) CD271(+) ERBB3(+) cells fused to form multinucleated myotubes. MYOGENIN was expressed mainly in multinucleated myotubes (Fig. 3E). In contrast, PAX7 was expressed in mononuclear cells between myotubes (Fig. 3E).  Four different protocols (protocol 1-4) of induction were tested for its efficiency using four iPS cell lines (201B7, 409B2, GFPT1 #3, and GFPT1 #8). After treatment with CHIR-99021 and LDN-193189, cells were cultured as spheres as described 21 . Each sphere was then plated onto collagen-coated 24-well plates at one sphere per well and cultured in 10% FBS/DMEM for 4 weeks. (B) 201B7-Myf5-tdTomato cells, a subline of 201B7, were plated onto iMatrix-coated 6-well plates at 2.2 × 10 5 cells/well. Half of them were cultured in mTeSR1 for 6 days. Another half were cultured in CL for 3 days, CLF for 3 days, and HIFL for 2 days. Then both groups were transferred to floating culture, and 3 days later, the numbers of spheres b and the diameter of all spheres (C) were measured using ImageJ. n   20 , suggesting that our cells have the phenotype of fetal myoblasts, and probably equivalent to their hiPSC-skeletal muscle progenitor cells (SMPCs). RT-qPCR analysis revealed that MRF4, MSTN, MYH2(fast IIa), and MYH8 (fetal/perinatal MHC) were up-regulated by SB431542. Interestingly, SB431542 reduced the expression of NFIX, but unexpectedly did not increase mRNA levels of MYOD or MYOGENIN. Although Notch signal is reported to inhibit terminal differentiation of muscle stem cells in mice 25 , a Notch inhibitor, DAPT (γ-secretase inhibitor), showed no significant effect on muscle differentiation of hiPSC-derived muscle progenitors. Interestingly, the expression of PAX7 was not down-regulated by the treatment of SB431542, but was significantly reduced by dual treatment of SB431542 and DAPT (Fig. 4,  Supplementary Figure 4). Addition of an ERBB3 ligand, a recombinant neuregulin-beta1, showed no significant effects on differentiation of iPSC-derived myogenic progenitors (data not shown).
NFIX maintained undifferentiated state of hiPSC-derived myogenic progenitors. As soon as 24 hours after SB431542 administration, NFIX was downregulated (Fig. 5). To know the role of NFIX, we knock-downed NFIX using shRNA plasmids. Two shRNAs effectively reduced the NFIX mRNA level (about 50-55% compared with the control). The reduction of NFIX expression was almost the same as that of SB431542 treatment in 454E2 iPSC-derived myogenic cells (Fig. 5D) and 201B7 iPSC-derived myogenic cells (data not shown). The knockdown of NFIX promoted myogenic differentiation of hiPSC-derived myogenic cells to some extent (Fig. 5E), further suggesting that NFIX plays a role in maintenance of the undifferentiated state of myogenic progenitors.
hiPSC-derived myogenic cells efficiently fused with adult myoblasts. We next examined whether human iPSC-myogenic cells fuse with adult human myoblasts. To this end, we added hiPSC-derived myogenic cells (unfractionated and fractionated cells) labelled with copGFP to a differentiating culture of human adult myoblasts labelled with fluorescent TagRFP protein. Soon after the start of co-culture, the two kinds of myogenic cells started to fuse (Fig. 6A). The fusion efficiency of fractionated hiPSC-derived myogenic cells was higher than that of unfractionated cells (Fig. 6B). Importantly, SB431542 treatment again improved the fusion between human iPSC-myogenic cells fuse with adult human myoblasts (Fig. 6B).   into the peritoneal cavity of the recipient mice for four consecutive days to enhance differentiation of the transplanted cells 20 (Fig. 6C). After transplantation of 1 × 10 5 FACS-sorted cells, 12-13 human lamin A/C-positive (nuclear membrane) and human spectrin-positive (sarcolemma) myofibers were constantly detected on transverse sections; transplantation of 1 × 10 6 unfractionated cells showed less efficiency (Fig. 6D,F). Human lamin A/C-positive, human spectrin-positive myofibers were also positive for dystrophin (Fig. 6E). Pretreatment of the cells with recombinant neuregulin1, a ligand of ERBB3, reduced the efficiency of transplantation of purified cells (Fig. 6F). This result suggests that ERBB3 signaling might suppress the differentiation of hiPSC-derived myogenic progenitors, although further analysis is needed to conclude this.

Stable induction of myogenic progenitors from hiPS cells by new sphere method.
To prepare a large quantity of myogenic cells from hiPS cells, we first tried to improve the EZ sphere method 21 by continuous low-speed stirring of the culture using bioreactors. This method greatly increased cell yield, but the system was unstable and the average percentage of myogenic spheres of all experiments was not increased (Supplementary Figure 1). Therefore, to obtain myogenic cells more efficiently, we combined the EZ sphere method with a recently reported step-wise paraxial mesoderm induction protocol 18,19 . The first two or three steps of the directed differentiation induction using two small molecules (CHIR-99021 and LDN-193189) and three promyogenic growth factors (FGF-2, HGF, IGF-I) greatly increased the efficiency of sphere formation at the start of floating culture (Fig. 2B) and constantly induced PAX3+ PAX7+ premyogenic progenitors from the four hiPSC lines. Transfer of progenitors at the dermomyotomal or premyogenic stage in adhesion culture (i.e., protocol 3 or protocol 4 in Fig. 2) to floating culture showed high efficiency of muscle induction, although the protocol needed minor adjustments for each iPS clone (Fig. 2D). Sphere culture longer than six weeks reduced the percentage of myogenic spheres (Supplementary Figure 3). This is probably due to the faster growth of non-myogenic cells than differentiating muscle precursors at this stage.

TGF-β signaling regulated differentiation of hiPSC-derived myogenic progenitors.
Mouse embryonic myoblasts differentiate into myotubes in the presence of TGF-β. In contrast, differentiation of fetal myoblasts is strongly inhibited by TGF-β 24 . Recently, Hicks et al. reported that hPSC-derived myogenic progenitors express TGF-β signaling genes at higher levels than fetal myoblasts and that SB431542 induces the maturation of hPSC-derived skeletal muscle 20 . We observed that TGF-β inhibitors also promoted the differentiation of hiPSC-derived mononuclear myogenic progenitors, which otherwise form myotubes poorly in 10% FBS/DMEM medium. Interestingly, the blockage of TGF-β signaling promoted the differentiation of FACS-sorted, highly purified myogenic progenitors. Therefore, it is likely that muscle progenitors secrete TGF-β (or related factors) to regulate the timing of differentiation themselves. Gene expression analysis and knockdown experiments with NFIX shRNA suggested that inhibition of TGF-β signals promotes muscle differentiation of hiPSC-derived myogenic progenitors partly by reducing NFIX expression. Since a similar regulation is likely operating in vivo, the mechanisms by which NFIX inhibit differentiation of hiPSC-derived myogenic progenitors are of great interest and need to be elucidated.

Human iPS cell-derived myogenic cells were transplantable. Transplantation experiments showed
that myogenic cells derived from hiPSCs are transplantable and form dystrophin-positive myofibers. We also confirmed that hiPSC-derived myogenic cells efficiently fused with adult myoblasts in vitro. These data suggest that myogenic cells derived from hiPSCs using the sphere-based method are a good candidate cell for cell therapy of DMD. However, transplanted cells differentiated into host myofibers with low efficiency (10-15 myofibers/section with 1 × 10 5 myogenic progenitors). In a recent report, Hicks et al. injected 1 × 10 6 cells/TA muscle (ten times our cell numbers) and administered SB431542 directly into skeletal muscle every three days for two weeks 20 . Transplantation of larger numbers of the cells and direct SB431542 administration into muscle for a longer period might improve the efficiency of engraftment of our hiPSC-derived myogenic cells. The factors inhibiting the engraftment of transplanted cells and their fusion with host dystrophic myofibers, other than the TGF-β signal, also should be investigated in future studies.

Experimental Procedures
Ethical statement. Informed consent was obtained for all human iPS cells used in this study. Samples were anonymized upon leaving the clinic. The study was approved by the Ethics Committee of the National Center of Neurology and Psychiatry, and all methods were carried out in accordance with the guidelines.

Cells.
Human iPS cells (253G4, 201B7, 409B2, and 454E2) cultures established from healthy donors were provided from S. Yamanaka at the Center for iPS Cell Research and Application (CiRA), Kyoto University. Lines 253G4 and 201B7 were generated using retroviral vectors 6,26 , and 409B2 and 454E2 were generated using episomal vectors and proven to be integration-free 27  Disease-specific iPS cells (GFPT1 #3 and #8) or DOX7 were established in our lab from fibroblasts of congenital myasthenic syndrome (CMS) patients. Ullrich-iPSCs were induced in our lab from fibroblasts carrying heterozygous mutations in the COL6A2 gene (GM23778) obtained from the Coriell Institute.
Paraxial mesoderm induction. Paraxial mesoderm was induced from human iPS cells as described 18,19 .
Sphere culture. EZ sphere culture was performed as described previously 21  Muscle differentiation. After floating culture, spheres were plated on collagen-coated 24-well plates (Iwaki) at one sphere per well and induced to differentiate into myotubes in DMEM containing 10% fetal bovine serum (FBS) (Gibco) up to 4 wk. DMEM supplemented with 10% FBS medium supported cell growth and after reaching confluence, cells started to fuse to form multinucleated myotubes. For FACS, the spheres were plated onto 100-mm dishes for a week.
RNA isolation, cDNA synthesis, and qPCR array. Total RNA was isolated from cells with Trizol (Invitrogen) or an RNeasy Mini Kit (Qiagen), reverse-transcribed into cDNA using PrimeScript RT reagent kit (Perfect Real Time, Takara), and amplified by primer sets (Supplementary Table 2) and SYBR Premix EX Taq II (Til RNaseH Plus, Takara). SYBER green signals were monitored by a CFX Connect system (Bio-Rad), and ΔCt (1/2^(Cq of the gene-median Cq)) was calculated.