Binding of HMGN proteins to cell specific enhancers stabilizes cell identity

The dynamic nature of the chromatin epigenetic landscape plays a key role in the establishment and maintenance of cell identity, yet the factors that affect the dynamics of the epigenome are not fully known. Here we find that the ubiquitous nucleosome binding proteins HMGN1 and HMGN2 preferentially colocalize with epigenetic marks of active chromatin, and with cell-type specific enhancers. Loss of HMGNs enhances the rate of OSKM induced reprogramming of mouse embryonic fibroblasts (MEFs) into induced pluripotent stem cells (iPSCs), and the ASCL1 induced conversion of fibroblast into neurons. During transcription factor induced reprogramming to pluripotency, loss of HMGNs accelerates the erasure of the MEF-specific epigenetic landscape and the establishment of an iPSCs-specific chromatin landscape, without affecting the pluripotency potential and the differentiation potential of the reprogrammed cells. Thus, HMGN proteins modulate the plasticity of the chromatin epigenetic landscape thereby stabilizing, rather than determining cell identity.


Introduction
Page 3, Line 58. Define epigenetic plasticity Page 4, line 67. Instead of epigenetic marks maybe histone modifications could be used instead.
Page 4, line 69. The authors write "play a role in cell fate decisions". Presumably you mean transcription factor binding?

Figure 1
The meta-analysis of HMGN binding is interesting but I would like to see HMGN binding over some specific genomic regions. Ideally in regions of the genome where there might be differences between cell types.
Page 6, line 127. Instead of epigenetic marks maybe histone modifications could be used instead.
Panel 1g on super-enhancers should be moved to a supplementary figure Page 7, line 174. Maybe a better term that "epigenetic reorganization" could be found. Page 8,line 196. The discussion about initial engagement of TFs sound very speculative. What is the evidence? This might be better saved for the discussion.  Page 11, Line 271. What do the authors mean by "HMGNs safeguard cell identity"? I am sure this could be better worded.
Page 11, line 282. I think it is best to avoid a phrase like "HMGNs stabilise cell identity". HMGNs influence chromatin structure, that in turn regulates transcription factor access, and this then alters cell identity. The results in the manuscript are sufficiently clear that it is not necessary to "hype" the role of HMGNs.
We thank the reviewers for taking the time to review the manuscript and for their constructive remarks.
Below is the answer to each of the comments by the 3 reviewers.

REVIEWER #1 (Remarks to the Author):
This is a very interesting manuscript identifying the binding pattern and functional influence of constitutive expressed chromatin binding proteins, HMGN1 and 2, on somatic and mouse pluripotent stem cell states.
The authors show high quality CHIP-seq analysis and that they bind enhances of somatic cell identity and when knockdown is conducted, this shuts down more efficiently somatic cell identity and enhances iPSC formation. The manuscript is well written, and data is of high quality and support the conclusions made (except in one case) which i detail below. References are accurate and methods are well detailed.
I support publishing this work following addressing this point: 1) the results of changing iN cell efficiency in the last figure are not convincing and no p values are indicated at all. is the effect significant and does HMGN knockdown enhance iN transdifferentiation? any result is fine, but their need to be statistical analysis and conclusions adjusted accordingly. Answer: We have assessed reprograming efficiency by both immunofluorescence and quantitative PCR of 3 neuronal markers. As requested, in Fig. 6 we have added p values in panels e, g and the legend. The data show statistically significant changes starting day 3 after induction of neuronal transdifferentiation.

REVIEWER #2 (Remarks to the Author):
In this manuscript, using induced pluripotent cells as a model, Bing and colleagues found that, the loss of ubiquitous nucleosome binding protein HMGNs can enhanced the efficiency of iPSC generation. They further found that the loss of HMGNs could accelerate the erasure of the MEF-specific epigenetic landscape and the establishment of ESC-specific chromatin landscape. Based on this observation, they proposed that, HMGN can stabilize the cell fate but not determine the cell fate, and they also used the ASCL1 induced conversion of fibroblast into neurons to support their idea. In general, these findings are of potential interest and could be useful to the field of somatic cell reprograming, and also offer insights for the function of ubiquitous proteins in cell reprogramming. However, the quality of the data should be improved significantly, and more evidence is needed to support the authors' conclusion. Major concern: 1 Pluripotency assay: AP straining or SSEA1/EpCAM FACS is not a suitable way to assess pluripotency or efficiency. The author should use Oct4-GFP colonies number or other more stringent standards to estimate the reprogramming efficiency (Figure 2a-2d). In addition, how about the quality of the generated iPSC colonies? Could they give rise to chimera? The ability of germ line transmission? ANSWER: We assessed pluripotency not only using ALP staining and SSEA1/EpCAM FACS sorting, but also used Flow cytometry to measure Oct4-GFP expression in 8 day cultures generated by fibroblast obtained from genetically altered mice carrying Col1a1::tetOP-OSKM, R26-M2rtTA and Oct4-GFP, as shown in Fig 2f. As previously demonstrated by others (see refence 31 and new reference 38), in these cells, Oct4_GFP expression following Dox induced OSKM expression is a reliable measure of their pluripotency. Significantly, we downregulated HMGN levels by treating cells with specific siRNAs and found that within 8 days of reprogramming induction, downregulation of HMGNs leads to a significant increase in the percentage of Oct 4 positive cells (Fig 2 f). In addition, the data in Fig 2h and  supplementary figure 3 indicates that the DKO cells express several pluripotency markers just as wild type cells do, albeit the upregulation is faster in DKO cells. Furthermore, injection of the DKO cells into nude mice gave raise to all 3 germ layers and both the growth rate the transcription profiles of the DKO teratomas was the same as that of the WT teratomas. Thus, by several accepted criteria, the pluripotency of the DKO cells is the same as that of the WT cells, which are known to produce chimera. In the revised manuscript we have modified panels d,e, and f in Fig 2. In d,e we added the description of the transfected cells, in panel f we now indicate that the data is from day 8 cultures. In addition, we have added a new paragraph at the bottom of page 13 which emphasizes all the points mentioned here.
The major emphasis of this manuscript is the role of HMGNs in maintaining cell fate. Although our manuscript is not concerned with the technical and practical aspects of iPSCs, we certainly agree with the reviewer that the quality of the generated iPSCs are an extremely important point, especially since in principle our results may have practical applications. Our data shows that the transcription profile and morphology of iPSCs generated from WT cells are very similar to those generated from DKO MEFs (Fig 5 a-c). In addition, our studies with nude mice clearly show that by several criteria, including the generation of 3 germ layers and transcription profile, the teratomas generated from both WT and DKO iPSCs are indistinguishable (Fig 5d-g). Furthermore, we already reported that mice lacking HMGN1 and HMGN2 are viable (Reference 22). It is well documented that iPSCs generated from WT MEFs can yield chimera capable of germ line transmission. Taken together the available data strongly suggest that the DKO iPSCs will also generate chimera and there will be germ line transmission. For practical uses, tests need to be done with human cells. We thank the reviewer for raising these important questions. In the revised version we discuss these points in the new paragraph that starts at the at the bottom of page 13.

The role of HMGNs in cell identity:
The authors show that HMGNs are mainly co-located with the active epigenetic modification and binding to cell specific enhances and also observed the shift of the epigenetic landscape from somatic cell fate to pluripotent ones during iPS generation. These would be consistent with the notion that HMGNs can serve to stabilize cell identity. Two points should be further explained. First, the binding of HMGNs to cell specific enhances may due to the open state of the loci in chromatin, the authors should explain the relationship between the binding of cell specific enhance and the open-close state of the chromatin state in depth. Second, if HMGNs can stabilize cell identity, they should also be required to establish iPSCs. This is contradictory to the results obtained. Furthermore, in the DKO iPSCs, how can they maintain identity? ANSWER: Regarding the first point. We added a paragraph (2 nd paragraph on PAGE 12) to point out that the binding of HMGNs to enhancers may be due to specific properties of active chromatin. Regarding the second point, as discussed in the manuscript, HMGNs stabilize rather than determine cell identity. Thus, the DKO cells can maintain their identity but when the identity is challenged by overexpressing OSKM or other transcription factors, they lose their identity faster than WT cells. Indeed, it has been previously shown by others that chromatin remodelers can affect the stability of cell identity. For example, in a landmark manuscript (reference 38) it has been shown that the histone chaperon CAF-1 has similar properties. Just like we found with HMGNs, MEFs lacking CAF-1 are more easily converted to iPSCs and mice lacking CAF-1 survive. It has been demonstrated (reference 22,45, see also http://tools.mouseclinic.de/phenomap/jsp/annotation/public/phenomap.jsf ) that mice lacking HMGNs are more susceptible to various biological stresses. Thus, as was shown with other chromatin modifiers, cells lacking HMGNs survive and maintain their identity, but their ability to maintain cell identity is weakened making them more susceptible to cell fate alterations, including cancer. The difference between stabilizing and establishing cell identity is mentioned in several places in the manuscript including the abstract. In the revised manuscript we added a sentence to the second paragraph of page 13 to further emphasize the distinction and to compare it to the effects of CAF-1.

Both over-expression and DKO
HMGNs impact the reprogramming process, but how do these factors co-operate with other factors to be located into the cell specific enhances and then regulate the gene expression? Could there be an alternative explanation?

ANSWER:
We have not shown that overexpression of HMGN affects the reprograming factor. What we show is that overexpression of HMGN in DKO cells rescues the wild type phenotype (Fig 2e). This is a very important control to show that phenotype seen is indeed due to loss of HMGNs rather than to some minor genotypic differences between WT and DKO cells. This is mentioned in the first paragraph of page 7.

The effect for ASCL1 induced conversion of fibroblast into neurons seems weak.
The author should calculate the significant difference with strong statistics. In addition, it's better to prove the induced "Neurons" are real neurons in physiology function. Minor concern: 1、Figure 1a,1e, no scale bar. 2、Figure 4d. The Data range of the ChIP-seq data for represented genes in IGV-Screen should keep same at different time point. ANSWER: As requested, we added p-values to the relevant panels in figure 6 to show that the data is statistically significant. Indeed, for use of HMGN treatment in practical medical procedures it is imperative to show that the neurons are physiologically competent. As elaborated above, this needs to be done in human cells. Similar to the experiments done in the manuscript describing the effects of CAF1 on cell identity (reference #38), our experiments with neuronal system serve to strengthen the notion that HMGNs stabilize cell identity. Fig 2a and 2e rather than Fig 1a,1e. These images are from 12-well culture plates containing iPSCs colonies. Scale bars are not generally added to images of culture plates. Nevertheless, as requested we now added scale bars.

2.
It is not possible to keep the ranges at the same level because the IGV represent raw data of histone modifications at different days of differentiation. It is well documented that the level of the histone modifications changes during differentiation. Thus, the RPKM showing the levels of modification at a certain locus changes during differentiation. Our study compares WT and DKO cells therefore the data range of WT and DKO cells are indeed kept the same thereby allowing comparison of histone modification in WT to those in DKO cells during differentiation. In the IGVs shown in figure 4d, the data range for all the modifications at all time points of WT cells are identical to those of DKO cells.

Introduction
Page 3, Line 58. Define epigenetic plasticity ANSWER: according to the dictionary "plasticity" is defined as: "the quality of being easily shaped or molded". This definition describes well our finding that loss of HMGN affect the ease at which chromatin can be changed as measured by several epigenetic marks. This term has been used in the literature before (see: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4039141/). As requested, in the revised manuscript we added a short descriptor immediately after the word "plasticity" (see page 3 last paragraph).
Page 4, line 67. Instead of epigenetic marks maybe histone modifications could be used instead. ANSWER: We also see differences in ATAC sensitivity and we previously reported changes in DNase I (reference #22). Previous studies from several laboratories showed that HMGN compete with H1 therefore it is possible that loss of HMGN leads to additional "epigenetic "changes.
Page 4, line 69. The authors write "play a role in cell fate decisions". Presumably you mean transcription factor binding? ANSWER: As elaborated above there are other factors than just transcription factor binding that play a role in cell fate decision. We mentioned CAF-1 (reference #38), H1 (reference #13), HMGA (reference # 16), however there are several other chromatin modifiers that play a role in this process. In the introduction we did not want to limit ourselves to one single possible scenario.

Figure 1
The meta-analysis of HMGN binding is interesting but I would like to see HMGN binding over some specific genomic regions. Ideally in regions of the genome where there might be differences between cell types. ANSWER: Supplementary Fig 2 e,f show IGVs of cell specific HMGN binding at specific genomic loci.

Michael Bustin, PhD
Therefore, in this letter I am asking you to reconsider your request for performing the chimera assay for the following major reasons: 1. Granted, the chimera assay is a very good test for iPSCs pluripotency. However, the manuscript already contains several assays showing cell pluripotency in both Wild type and Knock-out cells, including alkaline phosphatase staining (Fig 2), expression of pluripotency marker genes, teratoma assay showing differentiation into the three germ layers (Fig. 5), and detailed RNAs-seq analyses (6 biological replicates!) showing identical transcription profiles of WT and HMGN knock-out teratomas (Fig 5). By all accepted criteria, the iPSCs generated from WT cells are indistinguishable from those generated from the HMGN knock-out cells. Furthermore, it is well documented that mice lacking HMGNs are born. Given these findings and the known ability of WT iPSCs to generate chimera mice, it would not be meaningful to test whether the knock-out iPSCs can also generate chimera mice. Thus, showing that our iPSCs, can also generate chimera would not be meaningful and would not add any new, unexpected information.
2. Chimera is indeed a good assay for iPSCs but we do not study the generation of iPSCs. The manuscript contains two major new findings. First, that the ubiquitous HMGN protein localize to chromatin regulatory regions, including to the cell-type specific super-enhancers (Most of data in Fig 1). Second, that HMGN proteins stabilize cell identity. Loss of HMGN accelerates transcription factor mediated changes in cell identity in two separate experimental system. In other words, loss of HMGNs destabilizes the ability of MEFs to maintain their identity and facilitates the establishment of new cell identities. Neither of these two major points require experimental proof that we can generate viable chimera. We study the kinetics of epigenetic changes associated with the changes in cell fate and demonstrate that HMGNs stabilize the existing MEF-specific cell fate. We do not develop new methods of generating iPSCs, do not characterize new properties of iPSCs, or improve the therapeutic potential of such cells. Thus, showing that the iPSCs can generate chimera would not strengthen the major findings of the manuscript.
3. The chimeras will be prepared by our animal facility, they are not generated in our laboratory. We were informed by our animal facility that obtaining approval from the NIH animal committee and then generating the chimera would take over 4 months, perhaps even longer. Thus, generating chimera will significantly delay the publication of this manuscript, but not add significant new information or provide additional experimental support for the major findings of our manuscript.
Please note that neither reviewer 1 nor reviewer 3 felt that this is necessary. In fact, even reviewer #2 writes: "The authors may be right that given what is known, those cells should give rise to chimera. But, a chimera assay is easier than the teratoma". Thus, this reviewer does not suspect that our findings are wrong he just would like to get more information which, as mentioned above, will not strengthen the manuscript.