Silva Comments on MBD3/NuRD Debate with Jacob Hanna

What’s the deal with the tension between José Silva and Jacob Hanna over pluripotency?

Researchers probing how to reprogram cells into powerful stem cells (induced pluripotent stem cells or IPSCs) and what the molecular mechanisms are behind this process have become entangled in a conflict.

Of course over the years scientists including stem cell scientists have had their share of disagreements and debates in the past including heated ones, but this reprogramming clash has taken it up a notch beyond past ones and has some unique elements to it.

For example, this scientific clash is unfolding largely in the public domain and involves post-publication review such as on PubPeer.

Looking under the “hood” of cells so to speak, their reprogramming involves a molecule called MBD3 and a corepressor complex called “NuRD” ( for more on NuRD see my review with my postdoc mentor Bob Eisenman on NuRD here in Cell). It seems that most everyone agrees that MBD3 and NuRD play some role in reprogramming.

However, what the exact roles of MBD3 and NuRD might be in this area are very unclear at this time and there are quite distinct views.

Jacob Hanna
Jacob Hanna.

On one side we have Jacob Hanna, a well-known stem cell scientist who was previously a postdoc in the lab of Rudy Jaenisch. Hanna (pictured at right) has had a number of very high profile papers including on induced pluripotency.

One major assertion from the Hanna Lab is that a repressive chromatin complex including MBD3 and NuRD proteins is inhibitory for pluripotency (e.g. in their Rais et al. paper that I discussed on this blog when it came out in 2013). I also had a chance to hear Hanna talk at the 2013 Till & McCulloch meeting, where I asked him some questions from the audience (see here).

On the other side we have José Silva, another top stem cell investigator. Silva did his post-doctoral work in the lab of Austin Smith. Of course Jaenisch and Smith are two of the top scholars in the stem cell field. The Silva lab has published a Cell Stem Cell paper coming to pretty much the opposite conclusion of Hanna.

Jose SilvaThis stem cell conundrum has been the subject of vigorous discussion on PubPeer. Now this MBD3 disagreement has also spilled over to bioRxiv where further back and forth is ongoing. On Silva’s side of the MBD3/NuRD fence are also well-known researchers Brian Hendrich and Paul Bertone.

In an effort to get to the bottom of this and build a bridge for constructive discussion, I have invited both Silva and Hanna to comment here on this situation.

Hanna has indicated that he would prefer to wait a month or so until additional work is published and then will provide comment. Stay tuned for that as my invitation to him stays open.

Silva has provided comments now, which are included verbatim below:

“Please find below the key elements (in no particular order) being discussed on the ongoing debate on MBD3/NuRD and reprogramming and my view on these:

1- Levels of Mbd3 expression and its relation to reprogramming.

Rais et al. evaluated the potential of Mbd3 depletion primarily in the Mbd3fl/- heterozygous background. Dr Hanna claimed that these cells express hypomorphic levels of Mbd3, corresponding to 20% that observed from wild type cells. Dr Hanna also claimed that these cells could reprogram with near 100% efficiency. However, our analysis of Dr Hanna’s data revealed that Mbd3fl/- cells express instead nearly wild type levels of Mbd3. This is also in line with the quantification of Mbd3 levels in Mbd3fl/- cells assessed in Dr Hendrich’s lab and mine. This evidence led us to question the effective depletion of Mbd3/NuRD function as having been a significant factor in the reported increase in reprogramming efficiency. (Please note that Dr Hanna’s Mbd3fl/- cells used in the Rais et al. study were generated by Dr Hendrich’s lab.)

2- Use of elevated copy number of Oct4-GFP reporter transgenes to assess ongoing reprogramming efficiency

The high copy number of randomly integrated Oct4-GFP reporter transgenes may lead to spurious GFP expression in non-reprogrammed cells.

3- Comparison of reprogramming efficiency to a control cell line harboring a deficient Oct4-GFP reporter transgene.

While Mbd3fl/- cells harbored an intact Oct4-GFP reporter known to be promiscuous for expression in a range of cell types, the control cells were transfected with an alternative Oct4-GFP reporter with much greater specificity for pluripotent cells. This finding indicates that key methodology was absent from the paper, and the comparison of reprogramming efficiency between cell lines representing compatible measurement conditions was potentially invalid.

4- Reproducibility of Dr Hanna’s claims. Rais et al. was regarded as a  landmark paper in stem cell biology and numerous labs around the world have tried to reproduce these results. Their attempts using independent systems have been unsuccessful. We do not dispute that Dr Hanna’s cells reprogram relatively efficiently. However, based on our own experimental work and on the detailed analysis led by Dr Bertone of genomic datasets published by Rais et al., we found no evidence to support the claim that depletion of Mbd3 leads to deterministic reprogramming.

José Silva

10 thoughts on “Silva Comments on MBD3/NuRD Debate with Jacob Hanna”

  1. How about Hanna’s claim that he can make mouse/human chimera? Not reproducible by his own postdoc mentor???

  2. A) In the following link, and in several papers by Austin Smith group, Mbd3/NuRD is clearly delineated as a repressor for pluripotency network (similar to TCF3).

    B) The paper claims that Mbd3 depletion in very EARLY stages of iPSC reprogramming has a dramatic negative effect, but on the other hand suggest a mechanism where Mbd3 contributes to the gateway activity of Nanog which operates at the FINAL stages of reprogramming. This is a major contradiction and it makes no sense at all. I think the authors were desperately trying to promote their crumbling “Nanog gateway activity”, which has been thoroughly invalidated.

    C) Figure 3D in Dos Santos et al. – can the authors exclude that Vitamin C and/or ALKi perturb Mbd3/NuRD complex composition or activity ?

    D) What about this paper Silva authored before:

    Proc Natl Acad Sci U S A. 2012 Oct 2;109(40):16202-7. doi:10.1073/pnas.1208533109.
    Zfp281 mediates Nanog autorepression through recruitment of the NuRD complex and inhibits somatic cell reprogramming.
    Fidalgo M1, Faiola F, Pereira CF, Ding J, Saunders A, Gingold J, Schaniel C, Lemischka IR, Silva JC, Wang J.

  3. Jacob Hanna tells the story of how he started working on Mbd3/NuRD role in pluripotency and reprogramming. Initially he, like many others in the field, was fascinated by the Nanog “gateway”concept (Silva et al. Cell 2009)(( The latter paper claimed that Nanog is not required for maintenance of pluripotency (Nanog KO ESCs can be generated from WT cells by two rounds of gene targeting), but is absolutely essential for establishing pluripotency from Nanog KO ICMs (via ES derivation) or fibroblasts (via IPS reprogramming).

    Hanna searched the literature for other cases where KO ESCs can be established by double targeting, but ESCs cannot be derived from embryos. Examples for that included Klf5 ( and Mbd3 (Kaji et al. Nature Cell Biology 2006, Development 2007). They received different Mbd3 modified ES cells from Hendrich, thinking that Mbd3 depleted somatic cells will be deficient in their reprogramming ability, and that Mbd3 is likely a partner for Nanog and/or Klf5 in making this “gateway” activity.

    However once they started working on this, they realized how many wrong conclusions were made by Hendrich and Silva:

    1) Hanna group was able to derive Nanog KO iPSCs even in standard reprogramming and growth conditions, but decided not to publish that. Now we know that is true following the recently published Eggan and Hochedlinger papers

    2) The Hanna group was able to derive Mbd3 KO ESCs from null embryos, opposite to the claims by Kaji and Hendrich, both in serum/LIF and 2i/LIF conditions. In Rais et al. they decided to be diplomatic about it, and blamed this discrepancy on the use of 2i/LIF. The latter results make sense, as Oct4+ cells can be observed in Mbd3 KO ICMs (Kaji et al. Development 2007), and Mbd3 KO ESCs are resistant to differentiation in the absence of Lif (opposite to Nanog KO ESCs which still highly tend to differentiate). The quote form the Smith and Silva review indicated above, questions the Hendrich results and conclusions as well.

    3) Finally they noted that Mbd3 flox/- cells reprogrammed extremely efficiently, opposite to their initial expectation. However this made sense because Mbd3 KO ESCs are “super”pluripotent, can grow without LIF and Mbd3 represses key naive pluripotency genes (Reynolds et al. Cell Stem Cell 2012). Further, the Hanna team noted that Mbd3-/- (complete knockout) MEFs and EpiSCs are basically not viable and abruptly stop proliferation upon complete depletion of Mbd3 protein. As indicated in their paper, even when they used Mbd3-/- cells, they were derived from Flox/- cells and ablation was done not before 48 hours following initiation of reprogramming, but still having residual and low Mbd3 levels in the first 2 days of reprogramming was critical. Obviously Jacob could not start his paper by debunking multiple paper by Smith, Silva and Hendrich, and presented his story differently for obvious “political” reasons (although i think he made a mistake).

    It is really unfortunate that Cell Stem Cell allows such a misleading paper to be published without addressing key issues. Dos Santos et al. do not use Mbd3 +/+ cells (talking about choosing appropriate controls as preached by many recently on this website), and their -/- cells were passaged multiple times before initiation of reprogramming. As the growth curves presented throughout the Silva study, Mbd3 -/- cells proliferate very poorly, and no validation for Mbd3-/- NPCs or EpiSCs is provided?? .
    Their entire results can be explained by a trivial abrupt inhibition of somatic cell and EpiSC proliferation, which is clearly discussed by Rais et al. 2013.

    Finally, the Dos Santos et al. claim that Mbd3 flox/- cells have normal protein levels of Mbd3 as seen in WT cells (Fig S7), is also misleading, because this is shown only for ES cells expanded in 2i/LIF (as over-emphasized by the authors). They did these tests, and noted that ES cells in serum/LIF, EpiSCs and MEFs (before and after OKSM induction) have 50% reduction in protein levels as reported by Rais et al. Nature 2013. Notably, 2i boosts Mbd3 expression in ES cells and that is why Silva overemphasized the sue of 2i in Figure S7, and intentionally avoided providing western blots in other conditions. It is important to note that Hanna protocol does not apply 2i in the first 48 hours and uses Mbd3 flox/- cells as a starting material.

    One would expect that “when you come out throwing dirt at people” you actually would have better data. Unfortunately, most reviewers and editors are too ignorant to read papers and can be fooled by a desperate group of scientists that were basically caught with their “pants down”.

    “Context dependent facilitation of reprogramming by Mbd3 = making sure you use “ghost” senescent cells as your starting material”.

    Just awful!!

  4. Dear Jose,

    1) one would expect that whenever you start throwing dirt, you should have a convincing data set, which the Do Santos dataset obviously not. People who are not able to reprogram cells in the absence of 2i (AKA you – Silva et al. 2008) should not be allowed to comment on reprogramming anyway and torture the world with their pre-IPS cells (whatever this is ….).

    2) Retract your (1) Nanog Gateway cell paper, (2)Nanog-Tet1 Nature gateway paper (which has so many shady western blots), and (3) Kaji Mbd3 Gateway paper in Development 2007, and then you can start preaching to others about controls, ethics etc….

  5. “Hanna has indicated that he would prefer to wait a month or so until additional work is published and then will provide comment. Stay tuned for that as my invitation to him stays open.”

    He’s publishing another paper when the other papers haven’t even been confirmed to be reproducible in other labs?

    #0 PhD papers (multiple papers with issues)
    #1 Forskolin (Hanna 2010),
    #2 Human primed cells that are labeled as naive (Gafni 2013), and now
    #3 Mbd3 (Rais 2013),


  6. The funny thing is that the Rais et al. and Gafni et al. paper showed this working in human cells, and in secondary human fibroblasts with deterministic efficiencies.

    If this phenomenon only works in one Mbd3 fl/- mouse from one lab, one can only speculate about these human figures!

    As I am entitled to my opinion, I don’t believe the human data at all.

  7. just another tadpole

    Yet another stem cell sad story.
    And nothing will change.
    While the world becomes more public and social, science insists on remaining agoraphobic. Unlike the panorama of 50 years ago (with Watson, Crick, Franklin and Wilkins) we live in space where collaborations and discussion can be fomented at a click of a computer, with no borders or barriers besides scientists themselves.

    The publication process is flawed and is failing to adapt to needs of an era of the now and the mediate. The peer reviewed process is still creating obstacles for advancement to come out while letting stories with plot holes out, open access publishing systems are excluding poorly funded labs out (got to to love the the Science AMA by Michael Eisen and Plos requests of bank statements of individual scientists) since very few grants have money allocated for the purpose of publication expenditures.

    Openness and collaborative efforts would have made this situation a non issue. Also, I think that if a lab has the intention to disprove published findings there should be a more methodical approach to the reproduction of the data, which it seems like it was not the case.

    And instead of flicking the door of communication open we are witnessing a conversation between 2 PIs, with publishing emails of former exchanges, and requesting wait times so other work is published instead of just taking this matter timely and face to face.


  8. Oof, using two different Oct4GFP reporter promoters within the same experiment is a Heck of a clanger.

    1. just another tadpole

      It is a sign that something didn’t quite work out along the generation of the cell lines, but, the pressure to “make it work” did.

Comments are closed.