This article provides the input received from anonymous peer reviewers and the authors' responses to them. Our Inside the Paper series aims to give readers a glimpse at what goes on behind the scenes to get a paper from submission to publication.

Nature editor's overview

Finding the cell population that forms the developing heart could identify clues to fundamental biology, congenital heart disease and effective cell therapies. Here, Bu et al. identify a set of heart cells in the human fetal heart that express the marker ISL1, a developmental transcription factor previously shown to be expressed in mouse cardiac progenitors. This marker is expressed at the earliest stages of human heart formation as well as at the stage when progenitors commit to differentiated cell types. Once cells become fully differentiated, however, the expression of the marker shuts down.

To see if these ISL1+ cells were capable of forming the various cell types in the human heart, Bu et al. made transgenic human embryonic stem (hES) cell lines in which the ISL1 progenitors were genetically tagged, allowing the researchers to track whether differentiated cells arose from this population. They found that, in vitro, the primordial ISL1 cells could self-renew, expand and generate the three major cardiac lineages: cardiomyocytes, endothelial cells and smooth muscle cells. They also showed that Wnt ligands can both promote the expansion and maintain the multipotency of the hES cell–derived ISL1 progenitors. Together, the work supports the theory that the developing heart is built from the retained population of multipotent stem cells. This suggests a potential pivotal role for a primordial ISL1 progenitor in human congenital heart disease as well as providing a promising cell type for therapies.

Bu, L. et al. Human ISL1 heart progenitors generate diverse multipotent cardiovascular cell lineages. Nature 460, 113-117 (2009) | Article |

Below are comments from the three anonymous experts who reviewed the submitted manuscript, with responses from the paper's authors in italics.

[Editor's note: Page and figure numbers refer to those in the draft manuscript provided to the reviewers and may not correspond to those in the final, published manuscript.]

Cardiac development expert's comments

ES cell and cardiac progenitor expert's comments

Wnt signaling in development expert's comments

Reviewer 1: cardiac development expert

The manuscript of Bu et al. identifies and characterizes an ISL1 cardiovascular precursor population derived from hES cells. The Chien group has defined an ISL1 cardiovascular precursor population derived from murine ES cells. They have shown that this progenitor population is multipotent and can be expanded on a Wnt-expressing feeder layer. This manuscript builds on these prior studies by showing that a similar multipotent ISL1 progenitor population can be generated from hES cells. Furthermore, the ISL1 progenitors can be identified in human fetal hearts. These findings represent a significant advance, as they open the door for further work on the hES-derived ISL1 cells to understand their contribution to congenital heart malformations as well as to develop their application for regeneration of damaged heart tissue. Overall, the study is well-designed and the experiments are largely clear and conclusive. However, some refinements would improve the quality of the manuscript.

Comment 1: The manuscript refers to a population of cells that "express ISL1alone". This term should be more precisely defined. In some cases, "ISL1 alone" appears to mean ISL1/Nkx2-5–/KDR– (for example, in the first paragraph of the manuscript). In other cases (Figure 1c), "ISL1 alone" appears to indicate a lack of differentiated lineage markers. The term should be used consistently, and better data should be provided to support the claimed pattern of marker expression. The authors state that the "ISL1 alone" population accounted for a substantial amount of the ISL1 precursor pool." However, this conclusion cannot be made based on measuring pairs of markers (Figure 1a). For example, an ISL1/cTNT– cell might be KDR . Therefore, quantitation of this population requires simultaneous staining for all relevant markers. In Figure 1c, the "ISL1 alone" cells appear in locations outside of the myocardium. Do they share an anatomical location (such as the adventitial tissue surrounding the base of the great vessels)?

Authors' response: The co-staining of ISL1 and either cTNT, SMMHC or PECAM1 has been performed in adjacent sections to better define the distribution of different ISL1 progenitor pools in a specific heart region, which has been included in Figure S1. Based on these stainings, there is a considerable percentage of ISL1 cells that do not coexpress either the cardiac lineage differentiation markers (cTNT, SMMHC or PECAM1) or the cardiac progenitor marker NKX2-5.

Credit: original painting by Jon Marquette

We did observe ISL1 expression in locations outside the myocardium in human fetal heart tissue. A significant amount of these progenitors was found in the adventitial layer separating the great vessels as the referee pointed out, where an accumulation of ISL1+ cells and some intermediates of ISL1/SMMHC and ISL1/KDR can also be detected. We have clarified the use of the term "ISL1 alone" in the body of the manuscript.

Comment 2: The table in Figure 1a is poorly laid out, making it hard to interpret. Why are specific intermediates only quantitated in specific compartments (for example, there are no ISL1 /cTNT intermediates measured in OFT)? For completeness, it would be helpful to quantify areas with fewer ISL1 cells (for example, the right and left ventricles).

Authors' response: We agree that the table in Figure 1a is difficult to interpret due to the complexity of the human fetal tissue analyzed in terms of the distribution of different subsets of ISL1 populations across areas of the heart. We did stain similar regions in the right (Figure 1) and left (Figure S1) atrium for the three intermediate populations (cTNT, SM-MHC and PECAM). However, there is a large variation in cell composition and distribution in different heart regions; for example, we could not detect cardiomyocytes in the upper levels of the great vessels, and no VSMCs or ECs [vascular smooth muscle cells or endothelial cells]were detected in basal locations of the OF [outflow tract]T. In this regard, the three intermediate populations vary in different heart compartments (for example, there were very few ISL1 /cTNT in the outflow tract), which makes it difficult to quantify all three intermediates in the same compartment. Therefore, we have elected to remove the table in Figure 1a. Instead, we examined ISL1 cells at different developmental stages, 11 and 18 weeks specifically, which have been presented as text in the body of the manuscript and in supplementary materials to reviewers, clearly describing the results. Moreover, we were not able to detect any ISL1 progenitor in any ventricle sections as shown in additional materials at the earliest stage (11 weeks of gestation).

Comment 3: Throughout the manuscript, KDR is used as both a marker of an early mesodermal population and as a marker of differentiated endothelial cells. A more definitive differentiated endothelial cell marker should also be used to demonstrate an endothelial fate, such as CDH5. KDR staining in Figure 1c appears to lack specificity.

Authors' response: We have repeated all the experiments with the definitive endothelial markers CDH5 and/or PECAM1 to better distinguish between the early mesodermal progenitors and the differentiated endothelium. We have also included PECAM1 staining in Figure S1 to present the endothelial lineage in the human fetal heart. The KDR data has been removed from the manuscript as it does not add substantively to the PECAM1 data.

Comment 4: How many ISL1-bgeo transgenic ES cell lines were studied to exclude potential integration site effects? The usefulness of this ES cell line relies on LacZ expression faithfully recapitulating ISL1 expression. Therefore, more convincing data are needed than those in Figure 2b, which is subject to selection bias.

The authors use the ISL1-bgeo transgenic ES cells to show that human ISL1 progenitors can be propagated on CMCs [cardiomyocytes], which this group previously showed in the murine system. To support this point, the authors demonstrate persistent LacZ expression in EB [embryoid body]-derived cells cultured for five days on CMCs. However, the authors also show that human cells can simultaneously express ISL1 and differentiated markers. Therefore, to better support this point, the authors should show that these ISL1+ cells retain their undifferentiated state and are not differentiated cardiac intermediates.

Authors' response: Thirty-two ISL1-bgeo transgenic ES cell lines have been studied to exclude potential site effects caused by random integrations. Two of the thirty-two lines were used in this study. In both lines, LacZ is coexpressed with ISL1 and represents at least 30–40% of the endogenous ISL1 expression as documented by immunostaining assays.

After culturing five days on CMC feeders, many ISL1+ cells retain their undifferentiated state and coexpress NKX2-5. However, CMC feeders can't fully keep human ISL1+ progenitors from differentiation. The progenitors propagate on CMC feeders and spontaneously differentiate into cardiomyocytes and smooth muscle cells in 10–14 days when cultured on CMC feeders. These results are consistent with our findings in the murine system.

Comment 5: The authors generated ISL1-Cre-IRES-puro knock-in ES cells. As these cells are critical reagents in this and future studies, the authors should show that Cre expression recapitulates endogenous ISL1 expression. The majority of beating EBs were DsRed-. Does this indicate that most cardiomyocytes are not derived from ISL1+ precursors?

The authors indicate that "a second increase in ISL1 expression around EB day 10 suggested that ISL1 might be involved in the onset of other lineages during development". This raises concerns that beyond EB day 10, ISL1 and Cre marking are no longer specific indicators of cardiac progenitors. For example, what then is the significance of the finding that 2% of cells are DsRed+ after 30 days of in vitro differentiation?

Authors' response: We now provide new evidence by ISL1 and Cre antibody co-staining to show that ISL1-driven Cre expression represents endogenous ISL1 protein levels (Figure S2b). DsRed expression relies on the expression of the Cre recombinase from the ISL1 endogenous promoter and the excision of the transcriptional stop codon. Since excision of the transcriptional stopper may not be completely efficient, not all ISL1+ progenitors and their progeny will be labelled by DsRed. Nonetheless, we have confirmed that all DsRed+ cells dissociated from day 8 EBs also express ISL1. Based on the time-course study, we always sort DsRed+ cells no later than EB day 10 for the clonal assays, differentiation studies and Wnt3a expansion studies to eliminate possible contaminations from other noncardiac ISL1 lineages.

We agree with the reviewer that there is no relevant significance to show 2% of cells are DsRed+ after 30 days of in vitro differentiation. This description has been removed from the manuscript.

Comment 6: Demonstration of multipotency rests on observed colonies being derived from single cells. This was not rigorously demonstrated.

Authors' response: We have now used a two-colour system to address this concern. We generated a new transgenic hES cell line carrying a pCAG-flox-eGFP reporter in the ISL1-Cre knock-in background. DsRed+ and GFP+ cells were purified on EB day 8. Equal numbers of red and green cells were mixed and plated at up to fourfold of the cell density compared to that for original clonal assays. None of the 101 colonies/clusters formed after 10 days of plating was found to contain mixed colours (Figure S4a–b). A separate cell-dose experiment shows that the number of DsRed+ and eGFP+ colonies has a linear relationship to the number of input cells (Figure S4c). These findings strongly imply that the colonies developed on MEF feeders and used for the further assays are derived from single ISL1+ cardiovascular progenitors.

Comment 7: The relationship of the authors' ISL1+/KDR– population to the previously reported KDRlow/ckit– population (Yang et al.) is unclear. On page six, the authors suggest the ISL1+/KDR– population is downstream of this pool. On page seven, they suggest that it is upstream of this pool. The argument on page seven rests on the re-expression of KDR in the DsRed+ cells. However, KDR expression in this late population might simply reflect endothelial differentiation (as suggested by Figure 4d). This again gets to the confusing dual use of KDR as an endothelial and a progenitor marker.

Authors' response: The hES cell lines and differentiation protocols used in Yang's and our studies are essentially different, which makes it difficult to directly relate the progenitor populations isolated in the two studies. In order to avoid confusion between KDR+ progenitor cells and KDR+ endothelial cells, we have used PECAM1 and/or CDH5 as endothelial markers in all revision experiments (Figures S1, 3e and 4c).

Comment 8: The percentage of clones that differentiated into cTNT+, smMHC+ and VE-Cad+ lineages (and double and triple combinations of these markers) should be quantified.

Authors' response: The relative number of clones that differentiated into either single or multiple lineages is important to demonstrating the multipotency of the ISL1 progenitors. An RT-PCR–based clonal assay has been performed in an additional 30 single DsRed+ cell–derived colonies, and this brought the total tested clone number close to 100. The percentage of clones that differentiated into single and multiple lineages has been summarized and presented in the new table in Figure 3f. Moreover, new data based on intracellular sorting using cTNT, SMNT and PECAM1 antibodies are provided in Figure 3g as additional evidence to address the differentiation efficiency of the ISL1 progenitors.

Comment 9: Multipotency after Wnt3a culture should be demonstrated more conclusively.

Authors' response: We analyzed 32 additional clones in the revised manuscript. We used RT-PCR to demonstrate that clones cultured on a Wnt3a feeder layer keep their cardiac potential. Results are summarized in Figure 4c of the revised manuscript.

Back to top

Reviewer 2: ES cell and cardiac progenitor expert

The manuscript concerns an area of pluripotent stem cell research in which there is a paucity of well-characterized, genetically marked cells for lineage tracing and cell-type selection from mixed populations. The data here represent a potential tool for cardiovascular research as ISL1+ cells appear, in the mouse heart, to mark cardiac progenitors that can give rise to cardiomyocytes and endothelial cells, although they are also expressed in progenitors of multiple other lineages. However, although interesting, the data raise a number of issues that weaken the conclusions.

Comment 1: In the data for Figure 1, the authors examined the coexpression of ISL1 with markers for cardiomyocytes, VSMCs and ECs in the fetal human heart. For staining ECs, an antibody for KDR was used. Because KDR may also mark early cardiac progenitors, it is essential to include another EC marker (for example, VE-cadherin). Why did the authors not stain for all markers in the same region of the heart? The percentage of ISL1+ cells is very high (close to 30%); what percentage of cells express ISL1 only (that is, what percentage is negative for lineage markers and Nkx2.5) and thus are potential cardiac progenitor cells? Based on Figures 1 and S1, the decreased intensity of ISL1 is not evident. How was this determined? There are so many ISL1+ cells that it is important to include more negative controls. For example, the left ventricle is not shown, and, according to the lineage data in mice, this area should lack ISL1+ cells. Is this the case? Also, what does the staining with this antibody look like in the mouse heart, and does it overlap with that of the ISL1 reporter?

Authors' response: We agree with the referee in regards to the potential confusion of using KDR as a marker for endothelial cells and have now applied PECAM1 (CD31) as a more definitive marker for endothelium staining.

We did stain similar regions in the right (Figure 1) and left (Figure S1) atrium for cTNT, SMMHC and PECAM1. However, there is a large variation in cell composition and distribution in different heart regions. As a result, we could not detect cardiomyocytes in the upper levels of the great vessels, and no VSMCs or ECs were detected in basal locations of the OFT. Therefore, the ISL1 /cTNT and ISL1 /KDR intermediates were counted from the SVC/RA region, whereas the ISL1 /SMMHC intermediates were counted from the OFT. Both regions contain the expected intermediates. The co-staining of ISL1 and either cTNT, SMMHC or PECAM1 has been performed in adjacent sections to better define the distribution of different ISL1 progenitor pools in a specific heart region, which has been included in Figure S1. Based on these stainings, there is a considerable percentage of ISL1 cells that do not coexpress either the cardiac lineage differentiation markers (cTNT, SMMHC or PECAM1) or the cardiac progenitor marker NKX2-5. Given this, Figure 1a is unclear due to the complexity of the human fetal tissue analyzed in terms of the distribution of different subsets of ISL1 populations across areas of the heart. Therefore, we have elected to remove it from the revised manuscript. Instead, we examined ISL1 cells at different development points, 11 and 18 weeks specifically, showing ISL1 cells significantly decrease during human heart development. These data have been presented in the text of the revised manuscript and in Figure S1.

We agree with the referee that, at this stage, there is not an evident correlation between the intensity of ISL1 expression and the potential progenitor state. However, the fact that there is a dramatic decrease in the ISL1 cells between week 11 and week 18 of human gestation is highly consistent with the hES cell data, and this observation is noted in the text of the manuscript.

In order to address the reviewer's concern that the high percentage of ISL1 cells in the human fetal heart based on immunohistochemistry assays is not an artefact, we have included all the relevant controls as additional materials for reviewers. The ISL1 antibody employed in this study has been published by multiple groups previously and has a high degree of specificity in mouse embryos, murine and hES cells, cell-derived ISL1 progenitors and human heart tissue (Cai et al. 2003; Laugwitz et al. 2005; Moretti et al. 2006; Sunfu et al. 2007; Qyang et al. 2007). Moreover, we were not able to detect any ISL1 cells in any ventricle sections, as shown in additional materials for reviewers at the earliest stage (11 weeks of gestation).

Comment 2: In the data for Figure 2, it is unclear whether the ISL1-bgeo BAC transgenic hES cell line truly recapitulates endogenous ISL1 expression. What is the percentage of overlapping (and non-overlapping) expression between LacZ and ISL1 antibody staining (Figure 2b shows LacZ + cells that are ISL1–)? What percentage of cells in day 6 EBs are LacZ before clonal analysis experiments? Is there enrichment for LacZ clones? What is the molecular identity of the other clones? This is 90% of the clones and may represent other cardiac progenitors.

Authors' response: Thirty-two ISL1-bgeo transgenic ES cell lines have been studied to exclude potential random integration site effects. Two of the thirty-two lines were used in this study. In both lines, LacZ expression overlaps with ISL1 and represents at least 30–40% of the endogenous ISL1 expression that is demonstrated by immunostaining assays. As expected, LacZ staining was cytoplasmic and ISL1 staining was nuclear. Irradiated feeder cells (unstained) are labelled (Figure 2a). There are very few LacZ+ cells (<0.1%) in day 6 EBs. After expansion on CMC, there is an enrichment of LacZ clones. A few ISL1 clones stained positive for NKX2-5. However, both ISL1 and NKX2-5 antibodies could label only around 50% of clones. The identity of the rest colonies is under further investigation and is outside the scope of the current manuscript.

Comment 3: In Figure 2c, an inducible Cre transgenic hES cell line for the ISL1 locus was generated. This creates heterozygous expression of ISL1. How will this affect cardiac progenitor cells and cardiac differentiation (particularly in the context where intermediate phenotypes have lower expression of ISL1)?

Authors' response: Studies from multiple laboratories reveal no discernable phenotype of either embryos or cardiac progenitors in mice that are heterozygous for ISL1. Furthermore, the studies described here demonstrate that the hES cell lines that have only one copy ISL1 have similar capabilities for self-renewal and multipotency — key features of heart progenitors. Although genetic variation in the ISL1 locus might be related to the development of congenital heart disease, a detailed analysis of this possibility is outside the scope of this manuscript.

Comment 4: In Figure 2e, ISL1-Cre hES cells were transfected with two different plasmids. How were colonies picked? Are these stable lines under antibiotic selection pressure? This is at least not indicated in the figure.

Authors' response: ISL1-Cre hES cells were electroporated with linearized pCAG-flox-DsRed and circular pCAG-Flpase plasmids. Electroporated cells were plated and cultured for two days before being selected with 0.5 μg/ml puromycin for the stable transfection of pCAG-flox-DsRed plasmid. Puromycin selection was applied for 2 days, and stable ISL1-Cre DsRed cell lines were established after 2–3 weeks culture in. Forty-eight lines were characterized to confirm the coexpression of ISL1 and DsRed. Two of these lines were used for all the experiments performed in this study. pCAG-Flpase was transiently transfected into these cell lines to remove the FRT-Antibiotics-FRT cassette in the initial knock-in construct. PCR analysis confirmed that the pCAG-Flpase plasmid was not present after the initial transfection. The removal of the FRT-Antibiotics-FRT cassette from the ISL1-Cre DsRed cell lines used in this study has also been confirmed by PCR.

Comment 5: On page 5, 20–30% of the beating EBs are described as expressing DsRed. This means that 70–80% of the beating EBs are negative for DsRed and, because the transgenic line results in irreversible expression of DsRed, that the majority of the cardiomyocytes and progenitor cells did not express ISL1 during differentiation. How can the authors explain this? Again, what is the overlapping percentage of ISL1-immunoreactivity and DsRed expression? Do these cells express KDR or Nkx2.5? Furthermore, what is the percentage of DsRed cells that coexpress α-actinin or other lineage markers? This will give an answer to how many ISL1 cells eventually give rise to cardiomyocytes or other cardiovascular cells.

Authors' response: In our knock-in cell line, DsRed expression relies on the expression of the Cre recombinase from the endogenous ISL1 locus and on the excision of the translational stop after the CAG promoter. Since this process might not be completely efficient, some ISL1 progenitors as well as their differentiated progeny will not express DsRed.

Another consideration is that there is a significant delay between the time of DsRed translation and its maturation into a fluorescently active marker (Yarbrough et al. PNAS, 98(2): 462–467), and this time lag also contributes to the presence of ISL1 +/DsRed– cells. Thus, although all DsRed + cells from day 8 EBs were ISL1+ , approximately 20% of ISL1+ cells were DsRed–. Finally, it should be noted that ISL1 is expressed in SHF[second heart field] progenitors but not in FHF [first heart field] progenitors, another source of cardiomyocytes.

As shown in Figure 3c, DsRed+ cells isolated from day 8 EBs have an eightfold enrichment of NKX2-5 expression compared to that of the DsRed– population. It indicates that a portion of the cells is NKX2-5+. But they didn't express KDR at the time of isolation, shown by the RT-PCR result presented in Figure S3a–b.

Intracellular FACS experiments included in the revised manuscript demonstrate that 4% of DsRed cells coexpress cTNT, 44% co-express SMMHC and 3% express PECAM1. These results have been included in the revised manuscript as Figure S3g.

Comment 6: In Figure 3b, the DsRed+ cells are barely visible. The appropriate negative controls for FACS analysis are missing.

Authors' response: DsRed + cells corresponded to approximately 0.1% of the total cells on differentiation day 6–8. DsRed expression relies on the expression of the Cre recombinase from the ISL1 endogenous promoter and the excision of the transcriptional stop. Since excision of the transcriptional stopper may not be completely efficient, not all the ISL1+ progenitors and their progeny will be labelled by DsRed. Meanwhile, the slow maturation of DsRed tetramer protein may cause a notable straggler between the activation of the ISL1 promoter and the expression of DsRed fluorescence reporter. According to our observations, DsRed fluorescence becomes much stronger and the DsRed + population increases significantly after EB day 9. In order to study early ISL1 progenitors, EB day 8 DsRed + progenitors were isolated and used for all experiments.

Comment 7: The data presented in this paper provide no evidence as to whether a subset of cardiac progenitors may be downstream of the previously described progenitors or whether they represent a separate group.

Authors' response: We agree with the reviewer that the relationship between the ISL1 cardiac progenitors isolated from this study and those described in the Yang et al. paper remains poorly understood. The manuscript has been revised to reflect this.

Comment 8: For Figure 3d, the cloning efficiency of DsRed-sorted cells is approximately 0.2–0.5%. Is this higher than the cloning efficiency of the previous experiment with unsorted cells (and analyzed for LacZ+ cells)? If ISL1+ cells represent a true cardiac progenitor cell population, it would be expected that this percentage be higher. Furthermore, can these cells be expanded (and if so, for how many passages) and stored as aliquots? The expression of late-differentiation markers suggests (Figure 3e–f) that progenitor cells are no longer present (and cannot be expanded) and that differentiation has taken place. A more quantitative approach should be used for Figure 3e–f.

Authors' response: DsRed + cells have a slightly higher cloning efficiency than unsorted cells. Usually, 3,050 clones are formed from 100,000 unsorted cells, which is about 0.03005%.

We did not perform the passage study on these cells given the concern that our culture condition might create biases toward certain lineages and that the niche is essentially missing in the in vitro culture. However, we did culture these FACS-sorted cells on MEF [mouse embryonic fibroblast] feeders for a week, then sorted the DsRed + cells again and replated them at very low density to isolate single cell–derived clones. This resulted in most clones being positive for both ISL1 and NKX2-5.

In the clonal assay experiment, DsRed+ cells were sorted at EB day 8, plated on MEF feeders and cultured with human B27 medium. These culture conditions favour cardiac differentiation. Approximately 50% of the clones were ISL1 +/NKX2-5+, indicating that most clones keep at least partial progenitor features (Figure 3e–f). This is consistent with our results from human fetal heart immunohistochemistry. We have also documented many intermediates that coexpress the cardiac progenitor markers ISL1 and NKX2-5 as well as the cardiovascular lineage–specific markers cTNT, SMMHC and PECAM1 (Figures 1a–b and S1a–b). This suggests that cardiac intermediates may persist for a significant period of time during development.

Comment 9: Is the differentiation of the DsRed+ cells to cardiovascular cells more efficient than that of the DsRed– cells?

Authors' response: According to the intracellular FACS analysis, the DsRed + cells are a little more efficient to differentiate into cardiovascular cells than wild-type hES cells (<0.1%–2.0%) under our culture conditions.

Comment 10: For Figure 4, in order to test the hypothesis that Wnt signaling plays an important role in expansion of cardiac progenitor cells, ISL1-bgeo BAC transgenic hES cells were treated with BIO. The authors claim a 30% increase in the number of ISL1+ clones. However, this is more likely to be an effect of survival than expansion of cells. For Figure 4b, why were sorted cells not used in order to establish the role of Wnt on expansion of ISL1+cells? Why are the percentages of DsRed + cells so low in Figure 4b?

Authors' response: BIO- or Wnt3a-conditioned medium was applied three days after the dissociated cells were plated. At that time, most cells had already formed colonies, and all unattached nonviable cells had been rinsed away. Moreover, we did not observed an increase of cloning efficiency for sorted DsRed + cells that were plated on MEF feeders supplied with Wnt3a-conditioned medium (Figure 4c). Therefore, the increase of ISL1+ clones does not appear to be due to a survival benefit of the BIO- or Wnt3a-conditioned medium. The sorted cells were used to perform a clonal assay to explore the role of Wnt3a on the expansion of ISL1+ cells (Figure 4c). DsRed + cells were not expanded as much on the control feeder cells in Figure 4. These observations were consistent across five independent experiments.

Back to top

Reviewer 3: Wnt signaling in development expert

In their manuscript, the authors undertake a multifaceted molecular approach to defining the progressive lineage commitment of human cardiovascular cells during embryonic and fetal development. Specifically, they describe the distribution and molecular phenotype of ISL1+ cells in fetal heart, as well as in differentiating hES cell derivatives, using transgenic and knock-in hES cell lines in which ISL1 expression can be monitored directly or using irreversibly-labelled ISL1+ cells that can be followed throughout subsequent lineage commitment. Using these tools, the authors have generated data to support their conclusions that, in the developing human heart, ISL1 is expressed in 'primordial' cardiac progenitors that are negative for previously described early cardiovascular markers, self-renew in the presence of Wnt signaling and give rise to diverse cardiac lineages, including epicardium. The elucidation of human embryonic cardiovascular progenitors and the characterization of their progressive commitment to the numerous lineages that comprise a functional heart is essential to understanding both congenital malformations as well as potential treatments of defects and disease using pluripotent cell–based therapies. Although the work presented in this manuscript offers compelling evidence for the authors' conclusions to this end, some outstanding questions and problems remain to be resolved. Because the standard of evidence required for proof should be commensurate with the pre-eminent forum that Nature represents, in the opinion of this reviewer, this manuscript should not be accepted before fully addressing the questions listed below.

Comment 1: Fetal heart immunohistochemistry: the methods state, "Human fetal hearts were obtained...", and the results section describes data obtained from one heart in "the first trimester". Greater detail as to the timing and number of fetal hearts that were analyzed is essential. This reviewer can understand that perhaps the authors' only intention with this figure was to establish that ISL1 is expressed in the fetal heart at some stage in the first trimester, and that this expression colocalized in some instances with cardiovascular markers. Yet considering that this serves as a foundation for subsequent results, as well as the only piece of in vivo data, it is important that the authors at least put their results into developmental context. Indeed, the implication made from these data was that "ISL1 is expressed in the developing human heart and suggest[s] the possibility that a unique subset of primordial ISL1 progenitors might give rise to downstream multipotent heart progenitors...". Lacking any additional temporal stages during fetal heart development, and provided that this fetal heart represents an unspecified time point during the first trimester, the authors' claims of progressive lineage commitment are unwarranted.

This reviewer was also confused by the quantitative data in Figure 1. Firstly, the authors state the percentage of ISL1+ cells in the superior vena cava, posterior vena cava and outflow tract to be 29.2%, 28.6% and 25.9%, respectively. How do the authors obtain this number? Do they define the total number of cells by pan-nuclear counterstain? These data are not provided. Furthermore, the authors enumerate the percentage of ISL1+ cells that express either cTNT, KDR or SMMHC. Yet the highest number of ISL1+ cells shown to express one of these differentiated markers is 9.1% of the 25.9% ISL1+ cells (~2.5% of total cells) that express SMMHC in the outflow tract. Does this mean that more than 80% of the ISL1+ population (at least) does not express any of the three markers? If so, does that mean that the majority of the population consists of 'primordial' ISL1+ cells? The authors should address this, especially in light of the fact that there is no developmental context given for the fetal heart in question.

Authors' response: We agree with the reviewer regarding the need of including a greater number of embryonic samples in order to draw any conclusions about the ISL1 progenitor pool in the developing human heart. The purpose of the data set shown in this manuscript is to demonstrate the existence of ISL1+ cells in the human fetal heart. Nonetheless, we have analyzed multiple fetal hearts of the first trimester with immunohistochemical or biochemical studies. These results demonstrate a similar pattern of ISL1+ cell distribution and also confirm that different subpopulations of ISL1+ cells also express cTNT, SM-MHC (Martin-Puig et al., ongoing experiments). In order to document the temporal distribution of ISL1 progenitor populations in the developing human heart, we analyzed fetal hearts at 11 and 18 weeks of gestation (Figure 1 and Figure S1c–d). These results demonstrate that, compared to the 11-week-old heart, the 18-week-old heart has relatively fewer ISL1+ cells, especially the cells expressing only ISL1. Taken together, these results support the role of ISL1 as a human cardiac progenitor marker and are consistent with previously published murine data (Cai et al. 2003; Laugwitz et al. 2005; Moretti et al. 2006; Sunfu et al. 2007; Qyang et al. 2007).

The original percentages of ISL1+ cells indicated as 29.2%, 28.6% and 25.9% were obtained by direct counting of ISL1+ cells over the total number of cells (determined by nuclear counterstain quantification). We thank the reviewer for pointing out the confusion about the percentages of cells coexpressing ISL1 and differentiation markers. The percentages are of the total cells, but not of ISL1+ cells counted. Since the data in the table (Figure 1a) is difficult to interpret, we now present the data in the body of the manuscript. In order to determine more clearly the percentage of the different intermediates in a certain heart region, we stained a serial of adjacent sections with ISL1 and either cTNT, SMMHC or PECAM1 (Figure S1c–d). A high percentage of ISL1+ cells in the heart at 11 weeks of gestation do not express any of the 3 lineage-specific markers (Figure S1c). But in lack of direct isolation and differentiation on these cells, the statement that the majority of the population consists of "primordial" ISL1+ cells is hard to conclude from the human fetal heart studies.

Comment 2: Bgeo BAC Transgenic hESC line: hES cell lines that have been ISL1-genetically altered using targeted BAC vectors present a number of potential risks. The human BAC clone CTD-2314G24 contains approximately 125kb of sequence, which the authors have chosen carefully due to the relative absence of annotated genes, ESTs and/or microRNA sequences. However, there remains a possibility that uncharacterized regulatory sequences exist in this BAC and its exogenous addition may contribute to the phenotype. Furthermore, random insertion of the BAC sequence may disrupt genes and thus pervert the phenotype of the hES cell line that was used. However unlikely these events may be, they cannot be excluded and, therefore, the authors should at least corroborate the data obtained using one or more additional ISL1 lines. Even provided that collateral damage from BAC transgene insertion does not meaningfully affect the phenotype of the hESC line and/or its derivatives, the evidence provided for the specificity of the report is scant. The rudimentary quantification of the activity ("After five to six days of in vitro differentiation, cells were detected by X-gal staining...co-staining of X-gal and ISL1 was performed," and later, "single cell derived clones were observed after five days of in vitro expansion. Of these clones, 10+/- 5% had bgeo activity.") was not at all convincing. The authors could perform flow cytometric analysis and sorting to isolate and analyze the bgeo+ population; or they could provide quantitative data for the colocalization of X-gal and ISL1. Otherwise, this reviewer is not convinced that ISL1+ cells are faithfully and consistently indicated by bgeo activity.

Authors' response: We greatly thank the reviewer for many critical advices on the BAC transgene. Thirty-two ISL1-bgeo transgenic ES cell lines were examined to exclude potential site effects caused by random integrations. Two of these were used in this study. In both lines, LacZ is co-expressed with ISL1 and represents at least 3040% of the endogenous ISL1 expression as documented by immunostaining assays. The bgeo activity in these BAC transgenic lines was not sufficient for FDG staining for flow cytometry analysis and the Cre/LoxP system was used to amplify ISL1 signal as described.

Comment 3: Knock-in ISL1-Cre/pCAG-flox-DsRed hES cell line: the creation of a knock-in ISL1-Cre hES cell line is a remarkable achievement, yet there are potential risks inherent to the experiments the authors have conducted. First, presumably the integration of the Cre recombinase results in nullification of one allele of ISL1. Is it possible that this affects the phenotype of the hES cell lines? More importantly, though, the authors introduce the DsRed reporter using a transgene, and thus run the same risk discussed above. In these experiments, the authors wisely targeted more than one parental line (and presumably reproduced their results in both), yet these potential risks should still be addressed.

With regard to the data corroborating the fidelity/specificity of the DsRed report, again, the authors provide scant and not entirely convincing data. First, the authors state, "20–30% of beating EBs of the ISL1-Cre DsRed knock-in cells expressed DsRed." This is vague. Do the authors mean that 20–30% contained at least one DsRed cell? And, provided that all positive contractile EBs appear like the one shown in Figure 2h, how do the authors account for the 70% of contractile EBs that did not have DsRed cells? If the parental hES cell lines were derived from clones, do the authors think that the other contractile EBs did not contain ISL1 cells? Also, were there DsRed cells that were not associated with contractile EBs? Given the incompleteness of the data, this reviewer is left to conclude that either the authors' knock-in line is inefficient in its report of ISL1 or that cardiovascular progenitors arise by molecular mechanisms that do not involve ISL1. The authors need to address this.

The authors base the fidelity of their knock-in ISL1-cre/DsRed line on data shown in Figure 2f. Considering that this reporter is the basis for the majority of the authors' conclusions, the immunohistochemical data provided here are not sufficient. In fact, there are DsRed cells that seem to have low levels of ISL1, which could be explained by the irreversible nature of DsRed expression, but why are there cells that have high ISL1 but no DsRed? Along the same lines, how do the authors account for cells that are ACTN2 and DsRed? Did they arise from a molecular mechanism not involving ISL1? Given the seeming inconsistency/infidelity of this reporter and its place as the cornerstone of the authors' conclusions, it would seem that more rigorous immunocytological and quantitative data should be provided. Specifically, the authors should define the presence/absence of ISL1 and DsRed cells at discreet time points during hES cell differentiation. For example, why didn't the authors perform a temporal analysis like that found in Figure 3a for isolated DsRed cells? Why didn't the authors exploit their other BAC transgenic-based reporter line that allows direct assessment of ISL1 expression in live cells?

Authors' response: We thank the reviewer for acknowledging the technical difficulties in creating knock-in hES cell lines; to our knowledge, this represents the first example of a human knock-in ES cell line. [Note: There have been several knock in hESC lines reported eg oct4, even for specific lineages eg Mixl1] We established 48 transgenic lines with pCAG-flox-DsRed reporter in human ISL1-Cre knock-in background. These were validated by in vitro differentiation, flow cytometry and RT-PCR analyses. All experiments were performed in replicate in at least two or three independent transgenic lines.

In ISL1-Cre DsRed knock-in cell lines, approximately 20–30% of EBs with beating clusters were DsRed . As shown in Figure S2d and Movie 1, DsRed cells always occurred in clusters and never as single cells.

The labelling efficiency of ISL1 cardiac progenitors in the ISL1-Cre/DsRed system relies on activities of the ISL1 endogenous promoter, Cre recombinase and pCAG promoter. Additional evidence by ISL1 and Cre antibody co-staining has been provided to address that ISL1-driven Cre expression represents the endogenous ISL1 protein (Figure S2b). Besides immunohistochemical data, qPCR and RT-PCR analyses confirmed that DsRed cells are in fact ISL1 progenitors and their derivatives. There is a time delay, however, between the initial onset of ISL1 expression and the onset of DsRed fluorescence. This is due to the time delay associated with excision of the floxed transcriptional stop, the transcription of DsRed and the translation and maturation of the DsRed tetramer (Yarbrough et al. PNAS, 98(2): 462–467). Immunostaining experiments confirmed that all DsRed cells from day 8 EBs are ISL1 but about 20% of ISL1 cells are DsRed–. In order to isolate the most immature cardiac progenitors, we isolated DsRed cells from EB day 8, prior to the peak fluorescence of DsRed.

BAC-based reporter approaches did not allow for the isolation of cardiac progenitors as the ISL1 promoter was not sufficient for FACS purification. The ISL1/DsRed lineage-tracing system takes the advantages of the irreversible mark of DsRed expression. All the clones that we analyzed were confirmed under fluorescent microscope.

Comment 4: Clonal assays for measuring differentiation potential of ISL progenitors: the authors state, "To assess the differentiation potential of human ISL1 progenitors, we performed clonal assays on single DsRed cells." Clonal assays are the gold standard for measurement of multilineage potential; however, this reviewer believes that the methods used by the authors present a significant risk that what are called "clones" are not in fact derived from single cells. First of all, the authors do not include exclusion criteria in the flow cytometric analysis other than DAPI to exclude dead cells. This is especially important considering that the retrieval and inclusion of doublets would drastically undermine the results of "clonal" assays. The authors should use more sophisticated flow cytometric analysis, including doublet exclusion, by plotting forward and side scatter height versus width. This would ensure that the starting population consists of single cells. The potential for artefactual data collection is further magnified by the plating methods used by the authors: "We isolated DsRed cells from day 8 EBs by FACS and plated them at low density on a MEF feeder layer. We then picked individual colonies...within seven days at a cloning efficiency of 0.2–0.5%." Plating numerous cells in the same well, even at low density, presents the opportunity for cells to adhere to each other in suspension either before or upon settlement on the feeder layer. Also, a cloning efficiency of 0.5% is quite low and may suggest that apparent clones were derived from more than one cell. In light of this possibility, it is not surprising that the micrograph the authors provide of a "clone" at seven days seems to contain much more than 100 cells. This would suggest that, if they were single-cell derived, they would have undergone at least seven population doublings, or one per day. Do the authors contend that these cells proliferate at this high a rate, especially considering that many of these clones express differentiated markers and not only the ISL1 that identifies the primordial cardiac progenitor?

Authors' response: We used the standard protocol for all FACS sorting and analysis. The gating strategy in flow cytometry analysis includes not only the employment of the forward scatter (FSC) versus the side scatter (SSC) to exclude cell debris and clamps, but also the application of the forward scatter height (FSC) versus the width (FSC-W) to exclude doublets. In addition, nonviable cells were excluded by DAPI staining. True red cells were distinguished from autoflourescent cells by gating on PE versus PE-Cy5.

Upon further inspection of the original data, we have found that the original figure legend was mistakenly labelled. The clone presents a typical clone right before the expression assay, which is after 14 days of culture. The figure has been revised.

Comment 5: Wnt3a-Feeder layer expansion of DsRed -putative ISL1 progenitors: the data presented along these lines are dependent on the fidelity of the reporter that the authors constructed, which, according to this reviewer, have multiple problems that have already been mentioned. Additionally, this reviewer has problems with the transcriptional analysis presented in Figure S3c. The data lack convincing quantification — there are no error bars, and the transcript level is expressed in fold change relative to DsRed– cells. But this rests on the presumption that DsRed cells represent a faithful reflection of ancestrally ISL1 cells. This, as this reviewer has mentioned, cannot be taken for granted. Furthermore, defining the transcript level in terms of fold change relative to DsRed– cells is confusing. How were DsRed– cells separated from the feeder cells? What were the threshold cycles for amplification during the qPCR reaction? That is, was there a significant level of any of those transcripts present in DsRed– cells, or does a threefold increase in the DsRed population amount to an increase from 1 mRNA to 3 mRNA molecules (that is, a biologically irrelevant increase)?

Authors' response: In order to avoid the selectional bias, we performed the similar experiments on wild-type hES cells. We dissociated day 6 EBs into single cells and plated them at a low density on mouse CMC feeders in the presence of BIO or Wnt3a feeder layer. BIO treatment resulted in an over tenfold increase in ISL1 expression compared to the CMC feeder control. Coculturing with Wnt3a-secreting feeder for five days resulted in a sixfold increase in ISL1 cluster number compared to those cultured on MEF feeders. These results confirmed that hES cell–derived ISL1 cardiac progenitors could be expanded in response to Wnt signaling. This result is consistent with that we obtained from human transgenic and knock-in ES cell lines. The Wnt3a feeders were irradiated and therefore were incapable of undergoing cell division. In contrast, the hES cells have a short doubling time under these conditions. Nevertheless, given the reviewer's concern regarding the transcriptional analysis, we have kept it in the supplementary materials as supporting evidence.

Back to top