Chemical combinations potentiate human pluripotent stem cell-derived 3D pancreatic progenitor clusters toward functional β cells

Human pluripotent stem cell (hPSC)-derived pancreatic β cells are an attractive cell source for treating diabetes. However, current derivation methods remain inefficient, heterogeneous, and cell line dependent. To address these issues, we first devised a strategy to efficiently cluster hPSC-derived pancreatic progenitors into 3D structures. Through a systematic study, we discovered 10 chemicals that not only retain the pancreatic progenitors in 3D clusters but also enhance their potentiality towards NKX6.1+/INS+ β cells. We further systematically screened signaling pathway modulators in the three steps from pancreatic progenitors toward β cells. The implementation of all these strategies and chemical combinations resulted in generating β cells from different sources of hPSCs with high efficiency. The derived β cells are functional and can reverse hyperglycemia in mice within two weeks. Our protocol provides a robust platform for studying human β cells and developing hPSC-derived β cells for cell replacement therapy.

screening conditions and rationale for the optimization of stage-7. For example, how does supplementation of HGF, IGFI and PD173074 into stage 7 medium promote stage 6 endocrine progenitors toward commitment to immature β-cells? Since it is emphasized by the authors that they used a "late-stage readout strategy" to optimize the differentiation system, the authors should do retrospective analysis on stage 7 cells by examination of key transcription marker genes. 7. The authors should include the experimental methods for the GSIS study ( Figure 2E) in the "Material and Method" to describe how the authors did this static GSIS assay in vitro. For example, how many stage 8 cells were used for one assay and how long were cells incubated the prior to collection of the supernatant for hormone measurement? Aside from the data presented as "fold over basal" shown in Figure 2E, the authors should also show the raw values of C-peptide concentration. Equally importantly, did the authors examine the total insulin or total C-peptide content of stage 8 cells? As positive controls, the authors should examine human C-peptide secretion from primary human islets assessed using the same methods. Dynamic GSIS assay of stage 8 cells with different secretagogue stimulation (e.g. exendin-4 and KCl) would be more convincing evidence to demonstrate their in vitro functionality. 8. Regarding Figure.2F and 2G, how many stage 8 cells and fibroblasts were transplanted under the kidney capsule of STZ-induced diabetic NSG mice? Is the "no treatment" group a sham surgery? What glucose dose were animals given? The authors should include this critical information in the manuscript.
Minor points: 1. Line 33-34 and Line 64 -"The derived β-cells were physiologically functional" is not accurate, since the GSIS study ( Figure 2E) adopted a supra-physiological concentration of glucose challenge (16.7 mM) as the stimulation. 2. Line 126-127 -"Further characterization of PP-10C-treated PPs proved that most NKX6.1+ cells expressed PDX1". In addition to the flow cytometry data shown in Figure.1M, it would be further strengthened if the authors could show immunostaining data of NKX6.1 with PDX1 to draw that conclusion. 3. Line 271-272 -the authors used "3ug/ml chir99021" for stage 1 day 1 and "0.3ug/ml chir99021" for stage 1 day 2. By converting into uM, these are 6uM CHIR99021 and 0.6uM CHIR99021. Compared with many other published protocols, 3uM CHIR99021 is the most commonly used concentration at stage 1. On the other hand, it seems the authors only tested 1uM and 3uM CHIR99021 as listed in the Supplementary Table 1. Can the authors explain the rationale of deciding to use a higher concentration of CHIR99021 (6uM)? 4. Line 281-284 -the authors missed providing the concentration of KGF used in the stage 3 recipe. 5. Line 297 and Line 322 -do the authors really use "ZnS04 (10mM)" in the differentiation protocol, or should it be 10uM?
Reviewer #2 (Remarks to the Author): The reporter cell line is on the H1 hESC background. H1 has a high propensity towards generating pancreatic lineage among hPSC lines. The authors show flow cytometry plots of two hIPSC lines differentiated towards INS/NKX6/1+ cells using their approach. How many hPSC lines were tested together? What was the heterogeneity between lines? Importantly, how similar were independent differentiation batches for the same hPSC lien and among different tested lines.
The authors hypothesize that premature NKX6.1 loss causes the lower efficiency of beta cell induction with other protocols. Another likely scenario is that NKX6.1 expression is induced only in a portion of PDX1+ PPs cells, especially since PDX1 expression is very efficiently induced (even in ~ 90% of total live cells) at this stage of differentiation. Therefore, it is essential to show which experimental data, like lineage tracing or by other methods, support this claim. Further, at what stage is the expression of NKX6.1 decreasing? How reproducible is the described protocol? The data showing INS/NKX6.1 induction from several rounds of independent differentiation preferable for more than H1 cell line, would be critical as variability is a common problem.
Have the authors assessed the long-term efficacy and safety of stage 8 grafts? What was the longest time in vivo tested?  Review: In this manuscript by Liu et al the authors report that a new protocol used to develop organoids with 10 chemicals leads to a greater efficiency of formation of beta-like cells compared to the exiting approaches. They report that the beta-like cells are functionally competent and able to reverse hyperglycemia in a model of STZ-induced diabetes. The data are interesting but lack a number of experimental and technical details and don't provide a mechanistic explanation for the intriguing findings. Critique: 1. Introduction -last line should be rewritten. 2. What was the rational for using the custom library of compounds? 3. Under Results section ( Figure 2) the authors should spell out in mM the exact glucose concentration when they mention "low" and "high". Makes it easier for the reader rather than having to go back forth between text and figure to see the values. Figure 2E actually shows C-peptide. Please show insulin. Please include absolute levels of C-peptide and insulin so the readers can appreciate the levels. How does this value compare to C-peptide secretion from native islets at this glucose concentration. It would be useful to plot C-peptide and insulin secretion data in the same bar chart so one can compare differential secretory response between the organoid-derived b-cells and the native islet beta cells to glucose. The glucose level of 16.7 mM is pretty high. While this concentration is routinely used in the community it will be useful to know whether these cells respond to 7-8 mM glucose (i.e. post-prandial levels)? Now, that would be a very important advance in the field! 4. The authors have focused only on glucose as a secretagogue. While glucose is one of the most important what about responses to GLP-1 stimulation? 5. Considering there are double hormone+ cells, did the authors examine glucagon levels in the secretory functional responses? These cells have been reported to occur in most protocols. Did these cells functionally respond in this organoid approach? 6. The Nkx6.1/GFP merged cells in Fig 1C does not appear uniform in contrast to Fig 1B. Is there an explanation? 7. In Figure 1I the % cells don't add up to 100%! 8. In Figure 1J there a large empty spaces in the middle in the clusters in #s 1, 4, 15 and 19. Are these artefacts? 9. In Figure 2 F, a control model with native islets being transplanted would provide a useful comparison to assess how the organoid-derived beta-like cells act to counter the hyperglycemia. This is especially important since the mouse recovery from the stress after the surgery typically takes 10 days. The fact that the blood glucose comes down to less than 250 mg/dl within a week after transplanting the urganoid-derived beta-like cells is a remarkable observation. 10. Not sure if this Reviewer missed it -how many beta-like cells were transplanted? How does this compared with the typical transplantation of human IEQs in a similar STZ model setting? 11. While the data are intriguing no mechanistic explanations are provided as to how the organoid approach is able to promote the efficiency and/or the secretory function of the beta-like cells.

Reviewer #1 (Remarks to the Author):
In this manuscript, the authors describe a new differentiation protocol and report a chemical recipe for increasing the production efficiency of hPSC-derived NKX6.1+/INS+ β-cells. The authors state that their method incorporated several optimizations, including: 1) a V-bottom plate-based aggregation strategy to make 3D pancreatic progenitor clusters; 2) a 10-chemical/factor cocktail (they term "PP-10C") to "poise" pancreatic progenitor clusters; and 3) an additional 3-step chemical combination of signaling modulators to guide pancreatic progenitor clusters toward functional β-cells. Finally, the derived β-cells responded to high glucose stimulation in vitro under static GSIS assay and were able to reduce hyperglycemia in STZ-induced diabetic mice within two weeks post transplantation. Overall, while the high efficiency and functionality of the differentiations is an impressive achievement, the manuscript is lacking important details or data to further substantiate many of the key claims. Therefore, the manuscript could be further strengthened if the following points are addressed.
Major points: 1. Line 88-89 -Results provided do not fully support the conclusion that "The resulting chemically defined protocol produced a monolayer of PPs with an efficiency of ~80% ( Figure.1C, S1A, S1B and S1C)". This conclusion appears to be based upon detecting NKX6.1 (or GFP) expression without examining PDX1 expression. The authors should immunostain PDX1 with NKX6.1 and quantify the percentage of PDX1+/NKX6.1+ double positive cells among the whole cell populations at the end of stage 4.

Response:
We followed the reviewer's suggestion to perform the immunostaining of PDX1 with NKX6.1 in sections from the aggregate at the end of stage 4, which is shown in the new Figure S1D. In addition, we performed FACS analysis to quantify the percentage of PDX1+/NKX6.1+ double positive cells at the end of stage 4 (see newly added data in Figure S1F). The results of three biological replications showed an average of ~80% PDX1+/NKX6.1+ double positive cells among the whole cell populations at the end of stage 4. These data collectively lend additional support to our conclusion that "The resulting chemically defined protocol produced a monolayer of PPs with an efficiency of ~80%".
Similarly, it is interesting to see the dose of Activin-A at stage 1 impacts NKX6.1 expression at stage 4 (Figure.S1D). It is equally important to examine whether this gradient treatment of Activin-A at stage 1 impacts PDX1 expression, and more importantly, the co-expression of PDX1 and NKX6.1.

Response:
We performed the additional analysis and included these data in the new Supplementary Table  1. When the decreasing gradient (day1-day2-day3: 115-110-110 ng/ml) dose of Activin-A was administered at stage 1, of all NKX6.1-expressing cells, more than 98.5% of them also express PDX1 at the end of stage 4. This suggests that NKX6.1-expressing cells usually are a subset of PDX1-expressing cells. When constant low-dose or high-dose of Activin-A was used at stage-1, both the percentage of PDX1-expressing cells and NKX6.1-expressing cells at stage-4 were affected. Below is a summary of how the doses of Activin-A affect PDX1 and/or NKX6.1 expressions.  Figure.1E-F, the clusters seem to be ~1 mm in diameter. Were these relatively large clusters examined for dead cells and a necrotic core? Experiments with Live/Dead imaging would help address this.

Response:
We agree with the reviewer that "optimal size" is not a clear term in this context. We determined the number of cells and the size of the 3D PP clusters based on the following factors: (1) the cells should form a compact cluster without significant loss of cells after aggregation; and (2) the clusters are in the size that can be easily handled manually by an opened 100ul pipette tip or tweezers in the medium for transferring onto new air-liquid interfaces when beginning the next differentiation stages. For this, we tested a different number of seeding cells and found about 0.1-0.4 million cells, which generated clusters with a diameter ~0.5-~3mm and fulfilled the above-mentioned criteria. If more than 0.5 million cells were used, significant cells did not completely integrate into the clusters during the aggregation process. If less than 0.1 million cells were used for making clusters, the clusters were too small for later stage manual handling. Typically, we used 0.2 million cells for making one cluster. It's noteworthy that these pp-3D-clusters will become a little flat (like a very thick disk) after further culturing in the air-liquid interface.
We also followed the reviewer's suggestion of testing cell survival in the clusters. We examined these clusters after culturing for 4 days in different conditions. To identify cell death by live/dead imaging, we applied the TUNEL assay and co-stained with NKX6.1 and DAPI. The results revealed very few apoptotic cells in PP-10C conditions (see newly added data in Figure S1H, condition #13). In contrast, we did observe more cell deaths in other conditions. Moreover, we found necrotic cores (with a lot of pervading DAPI, dying/dead cells and cell debris) in some other conditions, such as conditions #4, #5, and #19 in Figure S1H. Very occasionally, we saw a small cyst inside PP-10C treated clusters (condition #13) (No cyst in Figure 1J or 1K; a small cyst in Figure S1H). As the small cyst in condition #13 has sharp boundaries and contains very few dead or dying cells, or cell debris/fragments, we believe these cysts might be early-stage pancreatic duct-like structures instead of necrotic cores.
Also regarding Figure.1F, it is hard to appreciate the NKX6. Response: According to the reviewer's suggestion, we added undifferentiated NKX6.1-NLS-GFP hPSCs aggregations as a negative control (newly added data as Figure 1F). In addition, we also generated a highresolution image showing PDX1-expression in PP-3D-clusters ( Figure S1I).
3. Figure.1H -can the authors explain why the purple area shown in the top left corner of the "GCG" image does not appear in the "merged with DAPI" image?

Response:
We have corrected this mistake in the revision.

Line 112: "Consistent with previous reports, most INS+/NKX6.1-cells expressed GCG". The authors ran flow analysis for NKX6.1 and INS, and this should be presented with GCG too.
Response: According to the reviewer's suggestion, we have newly added the result of INS/GCG flow cytometry analysis from the end-stage of the Method-1 differentiation cells. Together with the data provided previously ( Figure 1H and 1I), this clearly supports that "most INS+/NKX6.1-cells expressed GCG".
5. Figure.1J -By screening 21 conditions, the authors found condition #13 best maintained NKX6.1 expression (Line 120-121). However, the PP 3D clusters are relatively large and based on the images shown in Figure 1J, many clusters have empty appearing regions (especially core regions) without NKX6.1 staining. As the authors didn't show DAPI nuclear staining, it is not clear whether cells in the core were still there but didn't maintain NKX6.1 expression or cells were dead/dying in the core and therefore losing NKX6.1. Thus, there is a possibility that condition #13 stands out as the best condition to retain NKX6.1 expression via promoting cell survival within these giant 3D PP clusters.

Response:
To address this question, we performed NKX6.1/TUNEL/DAPI staining for ten conditions (including #13 and the other nine ones). The TUNEL staining showed very few dying/dead cells in 10-C conditions (#13 condition) (see newly added data in Figure S1H), but we did observe more dying/dead cells in most of other conditions. Moreover, in some other conditions ( Figure S1H, condition #4, #5, and #19) we found necrotic cores (with lots of pervading DAPI stain, dying/dead cells, and cell debris). As mentioned in the previous answer, in some occasional cases, we saw a small cyst inside PP-10C treated clusters in the condition #13 (No cyst in Figure 1J or 1K; a small cyst in Figure S1H). As the small cyst in the condition #13 has sharp boundaries and contains very few dead or dying cells, or cell debris/fragments, we believe these cysts might be pancreatic duct-like structures instead of necrotic cores.
In light of these observations, we agree with the reviewer that condition #13 could retain NKX6.1 expression via promoting cell survival. But we also observed that in conditions #1 and #15 ( Figure S1H), a lot of live cells lost their expression of NKX6.1 despite the fact that cells survived well and the aggregates had no obvious necrotic core. Thus, we believe that besides promoting cell survival, condition #13 also employs some other mechanisms to retain NKX6.1-expression.

Response:
We are glad to hear that the reviewer appreciated our efforts in optimizing the different stages of the differentiation processes. For stage-7, we applied different treatments on stage-7 cells, and then subject cells for stage-8 differentiation. We analyzed NKX6.1+/INS+ cells percentage at the end of stage 7 and 8 and performed GSIS at the end of stage-8.
The key chemical combinations we tested at stage 7 were iβ -9C minus LDN, T3, Repsox, GSIXX, RA, HGF, IGF1, or PD, respectively. We found that LDN, T3, Repsox, GSIXX, RA, HGF, IGF1, and PD is all required for maintaining a high percentage of INS+ and NKX6.1+ cells at the end of stage 8, and the removal of some of them also affects the GSIS function of the final product-β cells. The addition of i β-9C is used for the first time at stage-7 but not in other published protocols. However, compared to our analysis of other stages, the screening experiments for stage-7 were not systematic but instead were done in multiple small experiments. Therefore, we found it difficult to summarize them into a single table. Although we agree that analyzing how these iβ -9C factors affect cells' transcriptional factors at stage 7 is interesting. Such a detailed analysis will be better suited to a follow-up study. Figure 2E) in the "Material and Method" to describe how the authors did this static GSIS assay in vitro. For example, how many stage 8 cells were used for one assay and how long were cells incubated the prior to collection of the supernatant for hormone measurement?

Response:
We have added the following method details in the updated manuscript. For static glucosestimulated insulin secretion assays, Stage-8 cells (usually 1-2 clusters, equivalent to 0.2~0.4 million cells in total) ( Figure 3A), or 15 primary human islets ( Figure 3A)  Aside from the data presented as "fold over basal" shown in Figure 2E, the authors should also show the raw values of C-peptide concentration.
Response: According to the comments from this reviewer and other reviewers, we presented the absolute values of C-peptide concentrations in the revised figures ( Figure 3A). These values are normalized to account for cell number differences between tests.
Equally importantly, did the authors examine the total insulin or total C-peptide content of stage 8 cells?
Response: According to the reviewer's suggestion, we examined the total insulin content of stage 8 cells ( Figure S3B), which is ~ 62ng per 10000 cells. This value is comparable to that of human primary islets.
As positive controls, the authors should examine human C-peptide secretion from primary human islets assessed using the same methods. Dynamic GSIS assay of stage 8 cells with different secretagogue stimulation (e.g. exendin-4 and KCl) would be more convincing evidence to demonstrate their in vitro functionality.
Response: According to the reviewers' suggestion and to better demonstrate the functionalities of β cells from our new protocol, we added primary human islet as a positive control and performed additional secretagogue stimulations including Exendin-4 and KCl in our stage-8 cells. This new data showed that β cells from our new protocol performed very similarly to primary islet cells in static GSIS assay ( Figure 3A and S3A). For Dynamic GSIS assay, we had tried but were not able to find suitable equipment in the surrounding areas. Although dynamic GSIS assay could provide a little more information for assessing the function of our cells, our static GSIS assay data should be sufficient to support the conclusion that the beta cells from our new protocol are functional.
8. Regarding Figure.2F and 2G, how many stage 8 cells and fibroblasts were transplanted under the kidney capsule of STZ-induced diabetic NSG mice? Is the "no treatment" group a sham surgery? What glucose dose were animals given? The authors should include this critical information in the manuscript.

Response:
We routinely transplanted about 1.6 million cells (about 8 clusters) under the kidney capsule for each diabetic mouse. "No treatment" is a sham surgery. For the in vivo glucose-stimulated c-peptide secretion assay, 2g glucose/kg body weight (2g/kg; 30% solution) was used after 16hr fasting. This information has been added to the new manuscript.
Minor points: 1. Line 33-34 and Line 64 -"The derived β-cells were physiologically functional" is not accurate, since the GSIS study ( Figure 2E) adopted a supra-physiological concentration of glucose challenge (16.7 mM) as the stimulation.

Response:
We have removed the word "physiologically" from the manuscript. Figure.1M, it would be further strengthened if the authors could show immunostaining data of NKX6.1 with PDX1 to draw that conclusion.

Response:
We have added this new data in Figure S1I.
3. Line 271-272 -the authors used "3ug/ml chir99021" for stage 1 day 1 and "0.3ug/ml chir99021" for stage 1  Response: We are sorry for these typos. The CHIR99021 concentration should be "3 µM" for day 1 and "0.3 µM" for day 2. We have corrected these typos in the updated manuscript. We thank the reviewer for pointing these out.

Response:
We have added this information (KGF, 50ng/ml) in the updated manuscript.

Line 297 and Line 322 -do the authors really use "ZnS04 (10mM)" in the differentiation protocol, or should it be 10uM?
Response: We are sorry for this typo. It should be "10 µM". We have corrected this in the updated manuscript. We thank the reviewer for pointing this out.

The reporter cell line is on the H1 hESC background. H1 has a high propensity towards generating pancreatic lineage among hPSC lines. The authors show flow cytometry plots of two hIPSC lines differentiated towards INS/NKX6/1+ cells using their approach. How many hPSC lines were tested together? What was the heterogeneity between lines? Importantly, how similar were independent differentiation batches for the same hPSC lien and among different tested lines;
Response: We have tested in total five cell lines (including H1, H1-NKX6.1-GFP; and three integrationfree hiPSC lines (hiPSC1, hiPSC2, and hiPSC3) for our protocol. We were able to routinely generate >60% beta cells from them. Below is a summary table of the tests done in these five cell lines. In addition, our protocol is reproducible for the same cell line with low batch variations. These results suggest our protocol would be suitable for many ES and high-quality iPS cell lines. These data have been added to the new manuscript as Supplementary Table 3. Response: There might be some misunderstanding here. For all testing after stage 4, we started with a high percentage of a Pdx1+/NKX6.1+ (~80%) ( Figure 1C, 1G, S1C, and S1D. We then subjected these high quality PPs (~80% NKX6.1+ cells) to the last three steps of the R-protocol (we termed this combined protocol as "Method 1"). However, this trial yielded only 14% NKX6.1+/INS+ cells ( figure 1l). Moreover, the total "NKX6.1+ cells" drop to 19% ( Figure 1I). Based on this observation, we speculated that the last three steps of the R-protocol cannot maintain NKX6.1 expression and primarily led to the INS+/GCG+/NKX6.1-cell fate. Importantly, when PP 3D clusters were firstly incubated with PP-10C condition for four days before being subjected to the final three steps of the R-protocol, NKX6.1 expression in the whole cell population was preserved. This suggested that NKX6.1 expression is not stable after induction at the PP stage and can be lost prematurely during the early phase of later differentiation. We have clarified this in the new manuscript. Using our optimized protocol, we did not see a significant decrease of NKX6.1 expression after it was turned on. However, its expression could decrease in a few days if the condition is not ideal, such as many examined conditions in Figure 1J and S1H.

How reproducible is the described protocol? The data showing INS/NKX6.1 induction from several rounds of independent differentiation preferable for more than H1 cell line, would be critical as variability is a common problem.
Response: As we showed in the above table, we have tested in five cell lines (including H1, H1-NKX6.1-GFP, hiPSC1, hiPSC2, and hiPSC3) and multiple rounds for our protocol. We can routinely generate >60% beta cells in those cases.

Have the authors assessed the long-term efficacy and safety of stage 8 grafts? What was the longest time in vivo tested?
Response: The longest time we tested in vivo was 8 months after grafting of stage 8 cells. The ES-derived beta cells could reverse diabetes and maintain normal glucose levels for at least 8 months, and did not form tumors. Response: It is normal to see few (if not absolutely none) "NGN3" or "NEUROD1" in Figure 1k and 1L, as they should not be efficiently induced under condition #13 at that stage. Figure S2D has a lot of "NEUROD1" is because it was intentionally induced. We revised the descriptions to make it more clear.  Figure 3A). These values are normalized to account for cell number differences between tests. In addition, following the reviewers' suggestions and to better demonstrate the functionalities of beta cells from our new protocol, we added primary human islets as positive controls and performed additional secretagogue stimulations, including Exendin-4 and KCl in our stage-8 cells ( Figure S3A). Collectively, this data showed that beta cells from our new protocol performed very similarly in static GSIS assays as primary islet cells ( Figure 3A and S3A). Moreover, we also examined the total insulin content of stage 8 cells with controls from primary human islets ( Figure S3B) Response: The recent protocols from Melton, Hebrok, or Millman all have made some breakthroughs and addressed issues in their interests. In this study, we focused more on the special issues to achieve high efficiency, homogeneity, and cell line independency. That was why we strived to develop our own protocol. It will be not very feasible to compare cells differentiating from all different protocols for in vivo transplantation experiments, considering that all those protocols involve multiple steps and several dozen chemicals/factors and require high amounts of differentiated cells for in vivo assays. Therefore, it would be extremely time-consuming and cost prohibitive to repeat those protocols merely for the purposes of making them as control. We think such comparisons will be better pursued through collaborations in the future. Obviously, this is beyond the scope of our study at the moment. To acknowledge their contributions to the field, we have cited these recent publications in our revised manuscript as collective efforts to advance the field.
We agree with the reviewer that including a human islet positive control is important. Due to the current COVID19 pandemic, we are facing a lot of difficulty in securing a sufficient number of samples as well as lab hours. Specifically, we rely on external providers for human islet samples. As far as we know, the main supplier of US human islet distribution was completely shut down for several months since the start of pandemic outbreak, and has recently resumed operations but with a much reduced capacity. In addition, our Institute is still operating on a reduced capacity, especially in regards to the working hours in animal facilities. Despite these difficulties, we did our best to obtain human islets and conduct the in vitro GSIS assays and total insulin content for human islet controls, as we show in the new figure 3A and 3B.
Nonetheless, COVID19 should not be an excuse to lower the bar. We have some thoughts on the necessity of performing the in vivo experiments with human cadaver islets and would like to share them with you here. After going over the literature and communicating with several experts in the field, we have the impression that the results of such transplantation assays of human islets have been accepted as a scientific fact in the field. Consistent with this, we noted that several important publications on hESCdifferentiated beta cells recently published in high profile journals, including Vegas Communication, 2019, did not provide primary human islets as such in vivo controls, and mainly presented the critical comparison between transplantation of hPSC-derived beta cells and no treatments (in addition to that, we also included a more rigorous control of transplantation of fibroblasts). We think that this practice will become more popular in the coming future as it saves time and cost.

Review: In this manuscript by Liu et al the authors report that a new protocol used to develop organoids with 10 chemicals leads to a greater efficiency of formation of beta-like cells compared to the exiting approaches. They report that the beta-like cells are functionally competent and able to reverse hyperglycemia in a model of STZ-induced diabetes. The data are interesting but lack a number of experimental and technical details and don't provide a mechanistic explanation for the intriguing findings.
Critique:

Introduction -last line should be rewritten.
Response: This sentence had grammar issues, and we have rewritten the sentence to "The resulting β cells were functional and capable of reversing hyperglycemia of diabetic model mice within two weeks."

What was the rational for using the custom library of compounds?
Response: Inducing β cells from pluripotent stem cells are very complicated and expensive experiments, which involves multiple stages and chemicals/ growth factors for each stage, thus making a very largescale screen not feasible. While small commercial chemical libraries that contain hundreds of chemicals are also available, they usually fail to contain chemicals/growth factors regulating all classic developmental pathways. To reduce cost and save time, we generated a custom screen library comprised of more than one hundred chemicals/growth factors that could modulate (activate or inhibit) most of the known development and differentiation-related signaling pathways (former Supplementary Table 1, now  Supplementary Table 2). To this end, we managed to screen > 2000 conditions using different combinations of chemicals/factors from the library to finally build up our own protocol. Figure 2) the authors should spell out in mM the exact glucose concentration when they mention "low" and "high". Makes it easier for the reader rather than having to go back forth between text and figure to see the values. Figure 2E actually shows C-peptide. Please show insulin. Please include absolute levels of C-peptide and insulin so the readers can appreciate the levels. How does this value compare to C-peptide secretion from native islets at this glucose concentration. It would be useful to plot C-peptide and insulin secretion data in the same bar chart so one can compare differential secretory response between the organoid-derived b-cells and the native islet beta cells to glucose. The glucose level of 16.7 mM is pretty high. While this concentration is routinely used in the community it will be useful to know whether these cells respond to 7-8 mM glucose (i.e. post-prandial levels)? Now, that would be a very important advance in the field! Response: In the revised manuscript, we have substituted the overly-general terms "low" or "high" glucose concentration with specific numeric descriptions. For the former Figure 2E, although we call it glucose stimulated insulin secretion (GSIS) assay; we actually measure the secreted form of insulin, which is c-peptide. To alleviate this misunderstanding, we changed the assay name "GSIS" to "glucose stimulated C-peptide secretion" assay. As other reviewers also have the same concern, we now use absolute levels of C-peptide for previous Figure 2E (Now Figure 3A). Additionally, we also added primary human pancreatic islets as the control in these assays. These values of our stage 8 cells are comparable to that of native human islets (see the new Figure 3A). As 16.7 mM glucose is most commonly used in the research community, we stuck with this concentration. We did stimulate our stage 8 cells using 8mM glucose for one of the prior experiments and observed C-peptide induction but a little lower level than that with 16.7mM glucose.

The authors have focused only on glucose as a secretagogue. While glucose is one of the most important what about responses to GLP-1 stimulation?
Response: We did not try GLP-1 stimulation, but instead we have tried Extendin-4, a very potent GLP-1 receptor agonist. We showed that our stage 8 cells also responded to Extendin-4 (new Fig S3A), so I would expect that our stage 8 cells would also respond to GLP-1.  Figure S2E), only ~5% double hormone+ cells existed at the end of Stage 8. Therefore, we did not further examine the function of these double hormone+ cells rigorously. Based on our previously experiences, as long as the cells still keep a hybrid phenotype (expressing both GCG and INS), we do not think that our approach or other approaches will make these double hormone+ cells functional. Fig 1C does not appear uniform in contrast to Fig 1B. Is there an explanation?

The Nkx6.1/GFP merged cells in
Response: Due to the low brightness of DAPI staining in the previous Figure 1B, the total cell number of the sample was not clear and therefore it could have given the wrong impression that the NKX6.1 expression is uniform across the sample. We have increased the intensity of the DAPI signal in the revised Figure 1B. It is now clear that quite a number of cells were not expressing NKX6.1. Compared to the NKX6.1 expression in Figure 1C, Figure 1B has a lower percentage of NKX6.1+ cells and appears less uniform. Figure 1I the % cells don't add up to 100%!

Response:
We have corrected this mistake in the figure.
8. In Figure 1J there a large empty spaces in the middle in the clusters in #s 1, 4, 15 and 19. Are these artefacts?
Response: We are sorry for the confusion. For these empty spaces in the middle of the clusters in Figure  1J, most of them were filled with cells, while only a few of them have empty spaces or dead cells. We have added several images (including samples from #1, 4, 15 and 19) with DAPI and TUNEL staining to show their conditions in the revised Figure S1H.
9. In Figure 2 F, a control model with native islets being transplanted would provide a useful comparison to assess how the organoid-derived beta-like cells act to counter the hyperglycemia. This is especially important since the mouse recovery from the stress after the surgery typically takes 10 days. The fact that the blood glucose comes down to less than 250 mg/dl within a week after transplanting the organoidderived beta-like cells is a remarkable observation.

Response:
We agree that including a human islet positive control is informative. However, due to the current COVID19 pandemic, we are facing a lot of difficulty in securing a sufficient number of samples as well as lab hours. Specifically, we rely on external providers for human islet samples. As far as we know, the main supplier of US human islet distribution was completely shut down for several months since the outbreak of the pandemic and has recently resumed operations but with much reduced capacity. In addition, our Institute is still operating on a reduced capacity. Despite these difficulties, we did our best to obtain the human islets and conduct the in vitro GSIS assay for the human islet control and present this data in the new figure 3A and 3B.
To get around this hurdle, we had contacted several groups in the diabetes field to seek their collaboration for such an in vivo experiment. However, it turned out to be extremely difficult to convince anyone to take on such an expensive and labor-intensive experiment with current COVID19 restrictions. Therefore, we think it is not very likely that we can perform the in vivo experiments in the foreseeable future. Nonetheless, COVID19 should not be an excuse to lower the bar. We have some thoughts on the necessity of performing the in vivo experiments and would like to share them with you here. After going over the literature and communicating with some experts in the field, we have a common thought that the results of such transplantation assays of human islets have been accepted as a scientific fact in the field. Consistent with this, we noted that several important publications on hESC- included a more rigorous control of transplantation of fibroblasts). We think this practice will become more reasonable and popular in the coming future as it saves overall cost and time.

Not sure if this Reviewer missed it -how many beta-like cells were transplanted? How does this compared with the typical transplantation of human IEQs in a similar STZ model setting;
Response: About 1.6 million cells (about 8 clusters) were transplanted under the kidney capsule of each diabetic mouse. For the transplantation of human islets in a similar STZ model setting, typically 500-4000 IEQs were used. On average, each islet contains about 1560 cells; 500-4000 IEQs have 0.8-6 million cells. From the literature and our past experiences, we feel more human islet cells (rather than ESderived beta cells) are needed to achieve similar in vivo function partially because primary islet cells do not survive very well after transplantation.
11. While the data are intriguing no mechanistic explanations are provided as to how the organoid approach is able to promote the efficiency and/or the secretory function of the beta-like cells.

Response:
We are also very interested in the mechanistic explanations but feel such a study is beyond the scope of the current work and better suited for a follow-up.

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): Responses from the authors addressed most of the previous concerns and the quality of the revised manuscript is improved. Nevertheless, the authors still leave some major questions unresolved in the current version of manuscript.
Major points: 1. In the response letter, the authors agreed that "compared to our analysis of other stages, the screening experiments for stage-7 were not systematic but instead were done in multiple small experiments" and they also commented "LDN, T3, Repsox, GSIXX, RA, HGF, IGF1, and PD is all required for maintaining a high percentage of INS+ and NKX6.1+ cells at the end of stage 8". Given the focus of the paper, as indicated in the title, is "Novel chemical combinations potentiate human pluripotent stem cell-derived 3D pancreatic progenitor clusters toward functional β-cells ", it is important to provide sufficient details and evidence to allow readers to see the data supporting the conclusions and be able to reproduce the differentiation method as reported by the authors. The revised manuscript is still lacking in data supporting stage 7 optimization and is only mentioned in one sentence (on Page 6 Line 163-165). Is it possible for the authors to provide data like Figure 2B or provide a table summarizing their optimizing trials to support their point that all the iβ-9C components are indeed necessary? 2. The authors wish to highlight that the novel combinations of several chemicals/factors are being reported for the first time in this manuscript. However, they didn't provide any mechanistic explanations and should delete the statement "discovered new mechanisms" in Line 50. The authors argued in the response letter that "they feel such a study is beyond the scope of the current work and better suited for a follow-up". However, at least in the Discussion section the authors should provide discussions on the potential mechanisms of the key chemicals they screened for the protocol and cite relevant previous studies to support their points. For example, the authors commented on FSK (a cAMP pathway activator) and CI-1033 (a pan-ErbB inhibitor) in promoting the stage 6 differentiation but they didn't give further explanations on if and how these pathways are related to beta cell function or differentiation. The same concerns are also applied to other "novel" compounds as highlighted in this manuscript, including GABA, HGF, IGF1, PD173074, G-1, Deza and ZM447439. If the space is limited in the manuscript, the authors can provide these information in supplementary materials. 3. Another remaining major concern is the lack of quantification and statistical analysis of data in main Figures 1-2 regarding protocol optimization and cell characterization. It is appreciated that the authors provided nice representative data (flow cytometry and immunostaining), however, they didn't show any statistics on these analyses. As a result, it is hard to appreciate the level of consistency or heterogeneity within the same differentiation and across different differentiations, especially for the Method-3 as developed by the authors in this report. Although the authors added new Supplementary Table 3 in this regard, why not just add quantification results and graphs in the main figures? For example, it is recommended to include quantification of results relevant to Figure 1G, 1I, 1M and 1O, and Figure 2B (at least the complete Fβ-7C). In this way, readers can readily judge the advantages of the Method-3 protocol over previous Method-1/2 protocols. 4. Pertaining to the optimized differentiation protocol established in this study (as shown in Figure S2), can the authors comment on the rationale of the following compounds in terms of their temporal dosing manner? λ RA: used on stage 3, stage 5-7 but not on stage 4 -Why do the authors think RA is not required at stage 4? Is there an explanation for dosing the RA with a gap at stage 4? λ SANT1: used on stage 3, stage 5-6 -The same question described above is applied here. λ ZnSO4: used on stage 5 and stage 7, but not on stage 6 or stage 8 -Is there an explanation for this intermittent dosing of zinc?