Menu

Close

SWI/SNF remains localized to chromatin in the presence of SCHLAP1

Jesse R Raab, Keriayn N Smith, Camarie C Spear, Carl J Manner, J. Mauro Calabrese, Terry Magnuson

Preprint posted on May 14, 2018 https://doi.org/10.1101/322065

Article now published in Nature Genetics at http://dx.doi.org/10.1038/s41588-018-0272-z

A powerful call for reproduction studies: overexpression of a prostate cancer specific non-coding RNA does not function by depletion of SWI/SNF from chromatin to increase invasiveness.

Selected by Carmen Adriaens

The preprint and its context

In recent years, the scientific community together with funding agencies and publishing companies more strongly recognizes the necessity for reproduction studies. Fortunately, the perceived value of these studies lies not only in calling out irreproducible work or revealing scientific misconduct, but also in strengthening observations from colleagues in the field and in helping to put forward different interpretations for existing work when more information is obtained and new ways of proving a hypothesis are developed.

In 2013, the group of Prof. Chinnaiyan at the Michigan Center for Translational Pathology published a study showing that high expression of a novel long non-coding RNA (lncRNA), SChLAP1, can be used as a prognostic factor for poor outcomes in prostate cancer patients (Presner et al., Nat. Gen., 2013). Using knockdown and overexpression assays, the authors elegantly demonstrated that SChLAP1 increased the invasiveness of prostate cancer cells both in vitro and in vivo. Mechanistically, they proposed that this lncRNA is bound by the SWI/SNF chromatin remodeling complex known to both positively and negatively influence gene expression. Through SChLAP1 binding, the complex would be depleted from the chromatin to increase the expression of pro-invasion and pro-metastasis genes, hence causing the phenotype. SChLAP1 functioning was interpreted based on two observations: (1) the lncRNA binds directly to the SNF5 (SMARCB1) subunit of the SWI/SNF complex; (2) SChLAP1 overexpression causes genome-wide loss of SNF5 binding to chromatin as assessed by ChIP-sequencing.

In a recent preprint, the group of Prof. Magnuson at the University of North Carolina partially challenged and partially corrobated the above findings. On the one hand, they provided additional evidence that SChLAP1 has an important role in the proliferation, invasion and metastasis of prostate cancer cells and they validated that the SWI/SNF subunit SNF5 physically interacts with this lncRNA. On the other hand, the authors performed a series of experiments to determine whether SWI/SNF is globally depleted from the chromatin. They found that SChLAP1 overexpression did not affect broad SWI/SNF chromatin localization and binding both with the same experimental approach in the same conditions and with the same and additional reagents as in the original study. Probing changes in chromatin accessibility by ATAC-sequencing, the authors then found that the chromatin state of a set of pro-invasive genes (GO terms NFKB signaling, epithelial to mesenchymal transitions, and nucleotide metabolism) is more open, although they could not pinpoint a direct role for the SWI/SNF complex at these loci nor find evidence in the microarray data from Presner et al. for changes in their expression. Finally, the authors demonstrated that the observed binding of SChLAP1 to SWI/SNF is due to a rather promiscuous binding ability of SWI/SNF to RNA rather than to a specific effect between this lncRNA and the complex.

My opinion

Why this preprint caught my attention is because it is different from most scientific literature: it is a reproduction study. Although the scientific community agrees that these kinds of studies are necessary, still very few are conducted, and even less are published in high-visibility journals. The reasons for this are manifold. For instance, it may often make the challenger unpopular with the author(s) of the initial work, or, there is no money available to conduct a study much less flashy and novel than when original research is performed. However, the current preprint proves they are a must.

Indeed, this preprint is an excellent example of a rigorous reproduction study: although not all the results were reproduced, part of them are, and the validity of specifically these findings are strengthened independently. In the case of consistent SChLAP1 upregulation in aggressive prostate cancer, and its role in the metastatic propensity of the tumor, this could for instance prompt a company to confidently develop SChLAP1 detection as a biomarker to determine prostate cancer malignancy. Conversely, the authors took several diverse experimental approaches to show that the initially proposed mechanism of action was probably not real. Therefore, as the authors stated in the beginning of the work, an alternative hypothesis needs to be explored and I look forward to discovering what the actual mechanism of action of SChLAP1 is.

Finally, I really like the way the authors approached their biological question. I like to imagine that the concept under study is a physical object (e.g. a cube), and each face provides a different angle to look at it. Here, SChLAP1 tethering away the SWI/SNF complex from the DNA was the cube. Maybe the authors of the original article only looked at one or two faces of it, and through looking at it from multiple angles in this preprint, a different, more accurate and more complete view was obtained, and the cube was not at all what it originally seemed to be. (Figure 1)

Figure 1: Two (left) vs the many (right) faces of a “conceptual cube” in science give a very different view on what the object looks like.

Questions and concluding remarks

What prompted the authors to conduct this study? Which event or observation or results made them doubt the proposed mechanism of action?

On a more philosophical note, would access to the peer review file have helped to understand the publishing of the results? Would the absolute requirement for mechanism in high-impact factor journals lead to the risk of either overinterpretation of the work, or a potentially less careful evaluation of a striking result? (Admittedly, this is a different debate.)

From a more experimental point of view, to assess whether the dynamic chromatin sites also change in expression, the authors used the microarray data from Presner et al. I wonder if it would not be necessary to simultaneously probe gene expression changes and chromatin dynamics to be able to reach the conclusion that the one doesn’t influence the other, especially because the GO terms in the ATAC experiment point to the “opening” of genes that could explain the phenotype.

Finally, the biology of SChLAP1 seems to be different in immortalized and transformed cells. Maybe both papers use the wrong cell type to assess the mechanism, and a system that also displays the invasion phenotype would be more appropriate to find out how this lncRNA works?

 

Posted on: 15th June 2018

Read preprint (No Ratings Yet)




  • Author's response

    Dr. Jesse Raab shared

    Response to questions
    – What prompted the authors to conduct this study? Which event or observation or results made them doubt the proposed mechanism of action?
    SWI/SNF composition can vary based on inclusion of some mutually exclusive subunits (BRG1 and BRM for example). I was very interested in the hypothesis that not all forms of SWI/SNF were disrupted. In prostate cancer, it seems SWI/SNF Is surprisingly rarely mutated, so I though this might suggest a mechanism aside from the whole complex being disrupted. I thought loss of SNF5 ( one SWI/SNF subunit ), might allow a change in the composition or function of SWI/SNF. I was thinking along the lines of a nice paper a few years ago, and some continuation of that work from Cigall Kadoch and Jerry Crabtree’s labs where they showed a fusion protein found in synovial sarcoma (SS18-SSX) displaced SNF5 and affected targeting of the complex. So I really wasn’t trying to replicate the Prensner paper, we were trying to take some further steps to understand some other roles SWI/SNF complexes might play. ChIP-seq of SWI/SNF can be a little tricky, so it wasn’t really until we did the biochemical experiments that I was convinced the reported mechanism wasn’t quite right.
    – On a more philosophical note, would access to the peer review file have helped to understand the publishing of the results? Would the absolute requirement for mechanism in high-impact factor journals lead to the risk of either overinterpretation of the work, or a potentially less careful evaluation of a striking result? (Admittedly, this is a different debate.)
    It might have helped, my personal preference is journals publish those files because it would help to know if anyone had thought about the sorts of issues we were thinking about.
    – From a more experimental point of view, to assess whether the dynamic chromatin sites also change in expression, the authors used the microarray data from Presner et al. I wonder if it would not be necessary to simultaneously probe gene expression changes and chromatin dynamics to be able to reach the conclusion that the one doesn’t influence the other, especially because the GO terms in the ATAC experiment point to the “opening” of genes that could explain the phenotype.
    We interpreted the array data to at least partially support the ATAC data, and we wouldn’t really expect all ATAC and microarray expression changes to align perfectly. So I wasn’t real convinced doing more expression analysis would change our conclusion, but I agree with the suggestion and doing this might strengthen our results. To further strengthen our main result, it might make sense to knockout SNF5 from RWPE1 and see if that differs from the SCHLAP1 overexpression. If it differs, it would support our conclusions, if they look the same it would support the previous paper.
    -Finally, the biology of SChLAP1 seems to be different in immortalized and transformed cells. Maybe both papers use the wrong cell type to assess the mechanism, and a system that also displays the invasion phenotype would be more appropriate to find out how this lncRNA works?
    I think both cell types have the invasive phenotype, and it correlates with the expression of SCHLAP1. I am curious whether the benign epithelial cells (RWPE1) have the correct signaling pathways and things that might be present in tumors to really see a strong phenotype. We purposely chose to work mostly in that cell type so that we would have a relatively normal cell, but if the lncRNA needs co-factors not present in that cell type we would miss them.

    Have your say

    Your email address will not be published. Required fields are marked *

    This site uses Akismet to reduce spam. Learn how your comment data is processed.

    Sign up to customise the site to your preferences and to receive alerts

    Register here
    Close