Inducible, virus-free direct lineage reprogramming enhances scalable generation of human inner ear hair cell-like cells

Reviewer #1 (Public review):

Summary:

In this manuscript, Rainey et al investigated the effects of transcription factors, ATOH1, POU4F3, GFI1, and SIX1 on the induction of hair cells from human pluripotent stem cells. The authors used a doxycycline-inducible system to control transgene expression and demonstrated significant improvement in the efficiency of MYO7A+ hair cell differentiation compared to the retrovirus-mediated approach. Next, they characterized differentiated cells using single-cell RNA-seq and identified a population of hair cell-like cells with gene expression profiles similar to the fetal human vestibular hair cells. Finally, they revealed the electrophysiological properties of induced cells consistent with those of mechanosensitive hair cells.

Strengths:

A key finding in this study is the rapid induction of cells expressing multiple hair cell markers that takes place within 21 days after overexpression of the four transcription factors. Additionally, the authors demonstrate that doxycycline-mediated gene overexpression outperforms retroviral-mediated gene transfer in terms of both the efficiency and reproducibility of hair cell induction. Furthermore, the authors demonstrate that these induced hair cells can be used to study hair cell protection from cisplatin ototoxicity.

Weaknesses:

The authors conclude that the induced cells lack distinct hair cell subtypes. However, the characterization of generated hair cells in single-cell RNA-seq data is insufficient. Additional vestibular or cochlear hair cell-enriched marker gene and protein expression should be analyzed. Moreover, the morphological features and mechanotransduction channel activity of the induced hair cells have not been analyzed.

Reviewer #2 (Public review):

Summary:

The study employs a specific set of transcription factors to promote lineage conversion of pluripotent stem cells into fetal hair cells. In pluripotent stem cells, an inducible expression system containing SIX1, ATOH1, POU4F3, and GFI1 (SAPG) was inserted into a safe harbor site. The stable cell line allows for doxycycline-inducible expression of transcription factors to generate induced hair cells (iHCs). These changes were observed in gene expression and electrophysiological properties. Comparing the transcriptome with iHCs derived from fibroblasts or primary human inner ear tissue suggested that it is similar to human hair cells. Although the iHCs did not have hair bundles - a key morphological feature of hair cells - the cellular system has immense potential for the field. The defined transcription factors allow for the dissection of gene regulatory networks and provide a molecular handle for the lineage conversion process. The results also suggest that the pluripotent stem cells were not directly converted into iHCs. Instead, there are several transitional cell states. These observations indicate that lineage conversion may still be hampered by yet undefined molecular obstacles and may help identify and overcome these in future work. The stable cell line allows for repeatable and large-scale screening studies, which is not feasible using primary human cells.

Strengths:

The cellular system is well-designed, with clearly described expression of the defined factors. Transient expression of the exogenous transcription factors SIX1, ATOH1, POU4F3, and GFI1 (SAPG) upon doxycycline induction is well-documented. Increased expression of endogenous SAPG factors suggests activation of self-regulatory feedback pathways during conversion. The stable iPS cell line provides a tool for the field to study lineage conversion or generate large numbers of iHCs.

Single-nuclear RNA-seq distinguishes distinct cell clusters and cellular transition states, validating the system's utility. A comparison of previously published data from iHCs and human fetal hair cells also suggested that iHCs are similar to developing human hair cells at the transcriptome level. Whole-cell patch clamp recordings show the generation of excitable cells with heterogeneous ion channel properties, which suggests a change in the cell type.

Weaknesses:

The interpretation of the snRNA-seq results could be strengthened by explaining the three distinct clusters for uninduced cells and how they transition into the iHC trajectory.

Although the analysis focuses on the cell cluster that represents iHCs (R5), a short discussion on what clusters R1-R4 (Figure 3B) represent would be useful. These cells do not express high levels of the SAPG factors even after 21 days of continuous doxycycline induction and may provide insight into hurdles that hamper lineage conversion.

RNA velocity analysis on single-nuclear RNA-seq is impressive but requires clarification on inferring the pseudotime trajectory. Some rationale and explanation on how the ratio of unspliced to spliced mRNA in the nucleus can be used to infer the differentiation trajectory would strengthen the discussion.

Reviewer #3 (Public review):

Summary:

In this manuscript, Robert N. Rainey et. al. reported a new approach to induce hair cell-like cells from a human induced pluripotent stem cell line. Based on the previously identified key transcription factors SIX1, ATOH1, POU4F3, and GFI1 (SAPG), which are essential for the conversion into induced hair cell-like cells in mice. The manuscript represents an advance over the authors' previous published work, which used the same transcription factors but viral gene delivery.

Strengths:

The manuscript is clear and well-written. The background is easy to follow for people outside of the field. The data are well-organized and well-described. The evidence is strong.

Weaknesses:

General comments:

(1) The manuscript generated multiple valuable datasets for the field. However, the data are not deposited in the hearing field central resource for gene expression (umgear.org), and links are not provided in the figure legends to datasets or dataset collections in the gEAR. This is a major comment as it significantly decreases the utility of the datasets generated in the manuscript and decreases the ease of reuse of the data. This is a flaw that could be easily addressed by uploading the data and generating links to datasets in the body of the manuscript.

(2) If a pulse of Dox induces the SAPG and starts the conversion process, it is not clear why the analyzed cells were treated for 21 days - a duration that can negatively affect the fate of converting hair cells.

(3) Foxj1 is listed as a supporting cell-specific gene; however, it is expressed in the cochlear hair cells until the end of the first postnatal week.

(4) It is not clear why cells were sorted for analysis of the retrovirally induced cells but not in the stable cell line, which also expressed tdTomato.

(5) Figure 1D and Supplementary Figure 2: the authors state that the endogenous ATOH1 and POU4F3 expressions decrease after 7d. Should the authors have stats on the graphs?

(6) Supplementary Figure 4: OCT4 should be replaced by POU5F1 (or vice versa) for consistency.

(7) The authors show the induction or decrease of the exogenous transcription factor expressions by RT-qPCR. It would be nice, if possible, to also see either WB or immuno with antibodies directed against the tags.

Bioinformatic comments:

(1) In the previous study (Menendez et al. 2020), ATAC-seq and regulatory elements are employed in the analysis, while a similar analysis is missing in this study. It will be informative to show the motif enrichment analysis at promoter regions of differentially expressed genes (DEGs) in the most hair cell-like cluster 3 (RV-R3).

(2) In the previous study (Menendez et al. 2020), it was stated that SAPG can convert supporting cells to hair cells, while in this study, the authors stated that "reprogramming with SAPG does not activate supporting cell networks in the stable cell line". Can the authors provide more analysis/comments on this difference?

(3) The approach in this study tends to generate a very similar level of expression for the SAPG factors, while the real levels of expression might be different for actual transcriptional regulation, eg, Figure 1C. How will this very close expression level of SAPG affect the features of the induced hair cell?

(4) Figure 5B, missing color bar to show the DEG strength in the heatmap. Why are Six1 and Gfi1 not shown in this heatmap?