Referee 2's review:

In this manuscript the authors explore the phenotype of differentiating human ESC that have been engineered to over express the transcription factor PAX4. They have specifically examined whether the constitutive expression of PAX4 enhanced the propensity for differentiating HESC to 'spontaneously' give rise to pancreatic β cells in an embryoid body suspension differentiation. The experiments are based on the prior observations that PAX4 deficient mice failed to develop pancreatic β cells and that over expression of PAX4 in mouse ESC enhanced the expression of β cell genes.

The data shown suggests a similar outcome in the human ESC PAX4 over expressing clones. The clones constitutively express PAX4 message and protein and differentiation of theses clones is associated with increased expression of pancreatic endocrine markers, induction of voltage-gated calcium channels and an increase in Newport Green staining cells that were enriched in insulin C-peptide positive putative endocrine b cell precursors.

I have some concerns with the methodology of the paper that should be addressed before the work is published. The choice of differentiation medium is puzzling - it seems that no attempt has been made to direct differentiation along endodermal lineages through the inclusion of tgf b growth factors (activin, bmp4) and retinoic acid, as published previously by a number of investigators. For ES cells differentiating in KOSR (? it is not stated whether the FGF was removed) it is unclear what lineages will be represented in the differentiating EBs. 'Spontaneous' differentiation yielding progeny of all three germ layers is usually seen with serum-based differentiation media and there is a possibility that much of the differentiation in KOSR might be along the neurectodermal lineage.

Generally it would be anticipated that cells developing into pancreatic endoderm would express markers of primitive streak and definitive endoderm prior to expression of pancreatic genes. Therefore, at the very least, the cDNA timecourse samples (Figure 2) should be analysed for expression of primitive streak genes such as brachyury, mixl1 goosecoid, and definitive endoderm markers such as foxa2 and sox17 to demonstrate this and provide some circumstantial evidence that the expression of insulin was actually in a definitive endoderm derivative and not a neural cell type.

Additional specific comments:
1. Abstract of manuscript is confusing and written in poor English.

2. Isotype controls were not described for immunofluorescence stains and unstained cells are not acceptable as a negative control for flow cytometry.

3. It is not clear how many experiments were performed to give these results - how many of the five clones were used and how often were the experiments repeated?

4. Fig 1 needs to show low and high power images of H7.Px4 cells stained with anti PAX4 alone as well as the overlay and include a similar image of H7 parental cells as a control. Ideally these images should be complemented with intracellular FACs for PAX4 if the antiPAX4 antibody is monoclonal. This would provide confirmation of the uniformity of the cell population that was overexpressing PAX4.

2. In Fig 3 how many experiments used to generate panel C - not stated

3. Fig 4 should show profiles of Newport green staining differences between H7 control and H7.Px4 clones

4. Fig 4C what do the errors on the dCt values represent?

5. In Figure 5 what proportion of the sorted as Newport green high cells expressed pro insulin? How many experiments were done that yielded this result?

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.