LncRNA EPR controls epithelial proliferation by coordinating Cdkn1a transcription and mRNA decay response to TGF-β

Long noncoding RNAs (lncRNAs) are emerging as regulators of fundamental biological processes. Here we report on the characterization of an intergenic lncRNA expressed in epithelial tissues which we termed EPR (Epithelial cell Program Regulator). EPR is rapidly downregulated by TGF-β and its sustained expression largely reshapes the transcriptome, favors the acquisition of epithelial traits, and reduces cell proliferation in cultured mammary gland cells as well as in an animal model of orthotopic transplantation. EPR generates a small peptide that localizes at epithelial cell junctions but the RNA molecule per se accounts for the vast majority of EPR-induced gene expression changes. Mechanistically, EPR interacts with chromatin and regulates Cdkn1a gene expression by affecting both its transcription and mRNA decay through its association with SMAD3 and the mRNA decay-promoting factor KHSRP, respectively. We propose that EPR enables epithelial cells to control proliferation by modulating waves of gene expression in response to TGF-β.

direct effect of Smads? This can be addressed in part by asking if the downregulation occurs in the absence of new protein synthesis (cycloheximide treatment), or by demonstrating Smad3 binding to the Linc-EPR promoter. -As apparent from the title, the authors propose that Linc-EPR antagonizes TGF-b effects. This needs to be evaluated with some caution. Linc-EPR may antagonize the EMT program initiated by TGF-b, but not directly the response to TGF-b, and this may occur through its effect on cell proliferation (in part through the control of Cdkn1a expression). Are there other direct TGFb/Smad target genes that are controlled directly by Linc-EPR? Does Linc-EPR control induction of Cdkn1a in response to other stimuli? And in cells that do not undergo EMT? - Fig. 1b, S1g and corresponding text: The conclusion that Linc-EPR expression is restricted to epithelial cells is premature, based on the whole organ/tissue PCR data presented. All the organs/tissues analyzed by RT-PCR contain a mixture of cells. In situ hybridization data of tissue slices showing epithelial and other cells is needed to evaluate whether Linc-EPR expression is restricted to epithelial cells. -Fig. 1f, S1h and other figure panels: The NMuMG cells that are not transfected with Linc-EPR do not look epithelial, yet should look epithelial. Additionally, Fig. 3c does not show detectable Ecadherin expression, whereas they should be epithelial and express E-cadherin. As the cells were apparently not treated with TGF-b, I am concerned that they are not behaving the way they should and have drifted toward a mesenchymal phenotype.
-The transcriptome analyses led the authors to conclude that a large number of genes are controlled by Lnc-EPR. However, the cell cycle analyses in Fig. 3i show major changes in the fractions of cells in different stages of the cell cycle, which may be due in part to changes in Cdkn1a expression. Therefore the major changes in the transcriptome might result from the cell cycle changes and the relative changes in fractions of cells in different cell cycle stages, and not from Lnc-EPR expression per se. This is an important issue, and might be addressed by evaluating cells that are in the same phase of the cell cycle. Point-by-point list of the revisions. REVIEWER 1 We are pleased to read that Reviewer 1 appreciated our effort to provide a comprehensive portrait of EPR biology and we have done our best to provide solid experimental evidence supporting the mechanistic functional model that we propose.
First of all we apologize for the confusion mainly created by the Methods section. We fully agree with the Reviewer that it is not correct to compare transfected cells with cells not subject to the transfection procedure. Indeed, we did not perform such comparison in our work but we recognize that we did not describe with sufficient clarity our experimental procedure. In this revised manuscript (Methods, Cell Transfections sub-section [page 31] and first sub-section of the Results [page 6]) we now clearly indicate that the transfectants that we generated in distinct mammary gland cell lines (using plasmids with different backbones) were compared with "real" mock transfectants (i.e. all the EPR transfectants were always analyzed together with the respective mock cells transfected with the corresponding empty vector -thus including the same promoter-and subject to the same selection procedure).
Further, we have diligently worked to add experimental evidence to prove that the mutated EPR version unable to generate the peptide displays the majority of functions exerted by the non-mutated EPR. The new data are presented in Figure 4a Finally, in order to support the data presented in our manuscript, we would also like to share with the Reviewer an additional set of data obtained using NMuMG cells stably transfected with either a shorter versions of EPR (nt. 1-560 of the EPR sequence which include the ORF herein referred as to 5'+EPRp) or its derivative carrying a STOP mutation in the second codon (5'+EPRpSTOPE). Also in this case (please see the Figure below) both transfected plasmids yield superimposable results in terms of CDKN1A expression and cell proliferation. These experiments are part of an effort to define the EPR regions responsible for the effects described in this manuscript and will be part of a different study.
a. Immunoblot analysis of total cell extracts from either mock, 5'+EPRp-, and 5'+EPRpSTOPE-overexpressing NMuMG cells. (b) Proliferation analysis of either either mock, 5'+EPRp-, and 5'+EPRpSTOPE-overexpressing NMuMG cells. showing that TGF-β treatment for 1 hour does not change the interaction of EPR with Cdkn1a promoter (as for the technique used to demonstrate direct interaction between EPR and the Cdkn1a promoter, please see our response to minor point # 9). The new results, together with the experimental data already included in the manuscript, allow us to propose the following model. TGFβ induces an early wave of Cdkn1a expression due, in part, to an increased SMAD complexdependent gene transcription. A prolonged TGF-β treatment causes the return of Cdkn1a levels to the baseline. EPR, which is similarly bound to Cdkn1a promoter either in cells untreated or treated with TGF-β for 1 hour, recruits SMAD3 molecules when they accumulate into the nucleus upon treatment with the cytokine and this leads to rapid Cdkn1a gene transcription. Thus, we believe that the limiting step in the rapid transcriptional induction of Cdkn1a is represented by the enhanced availability of SMAD3 upon TGF-β treatment and its recruitment by EPR. In parallel, EPR interacts with KHSRP limiting its association with Cdkn1a mRNA and this results in the stabilization of the transcript. We propose that EPR down-regulation upon 6 hours of TGF-β treatment causes SMAD3 dismissal from Cdkn1a promoter that results in a return of Cdkn1a transcription to basal levels and, in parallel, enables KHSRP to interact with Cdkn1a mRNA and to destabilize it. We propose that EPR-regulated molecular events shape the rapid wave of Cdkn1a expression in response to TGF-β in NMuMG cells. The evidence that CDKN1A is abundant in cells overexpressing EPR even in the absence of TGF-β treatment allows us to hypothesize that overexpressed EPR is able to recruit SMAD3 molecules already present in cell nuclei to Cdkn1a promoter region and, possibly, to distal enhancers as well as to block KHSRP-induced Cdkn1 mRNA degradation. log2 fold changes >|2.0|" has been replaced with "|log2 fold change| >2.0.
We thank the Reviewer for her/his suggestion and we adopted the new nomenclature renaming the linc-EPR as EPR and the peptide as EPRp.
We added a new Supplementary Figure 3c showing that the mutated EPRpSTOPM (STOP mutation in the first Methionine) is not expressed. Further, we show in this revised manuscript that the overexpression of EPRSTOPM affects gene expression similarly to EPRSTOPE.
We are sorry for not being clear enough. We have previously reported that KHSRP is predominantly We apologize for the poor labeling of Figure 1e (and of other panels). We amended our mistake in the revised Figures 1e (and also 4a, 4f, 5f, 6b, 6e, and 6f).
In the legend to  promoter as well as with the promoter of an additional EPR target gene that is currently under investigation in our laboratory (Cdx2, please see also our response to Reviewer 2).
As for the technique used in order to show the interaction of EPR with the Cdkn1a promoter, we would like to respectfully underline that we adopted ChIRP because, since its original discovery and application by Dr. Chang's laboratory, it has been recognized and validated as one of the methods of choice to map lncRNA interactions with DNA (as well as with protein factors). We hope that the Reviewer will appreciate our effort in setting up and optimizing this complex technique in NMuMG cells.
Screenshots of Integrative Genomics Viewer windows showing representative ChIRP-Seq data analysis. We thank the Reviewer for asking us to better clarify this point. The new Supplementary Figure 4g shows that cytoplasmic Cdkn1a mRNA is relatively stable in mock cells and its decay rate is not significantly affected by EPR overexpression.
Following Reviewer 1 and 2 suggestion, we removed the sentence regarding p53.
We greatly appreciate the overall positive comment of this Reviewer and her/his constructive criticisms that allowed us to improve our manuscript.
(1) We observed that down-regulation of EPR expression occurs in colon carcinoma cells LS180 upon TGF-β treatment also in the absence of any modulation of factors relevant during EMT. We show these data in the Figure below for the Reviewer's convenience. (2) Following Reviewer's suggestion, we include in this revised manuscript (Supplementary Figure 1e) new data showing that SMAD3 interacts with the EPR promoter region and that this interaction is modulated by TGF-β treatment. This observation suggests that EPR expression is under the direct control of the TGF-β/SMAD signaling pathway. Experiments aimed at investigating the detailed molecular mechanisms underlying EPR regulation by TGF-β are in progress in our laboratory and will be part of further studies.
We thank again the Reviewer for giving us the opportunity to better charachterize some mechanistic aspects of EPR function.
-In the course of the analysis of ChIRP-Seq datasets (please see also our response to Reviewer 1) we have identified the gene encoding the transcription factor CDX2 as a direct target of TGF- letter, we believe that EPR expression/function is part of the TGF-β-SMAD signaling. However, we highly evaluate the Reviewer's concern and we have decided to change the title of the manuscript by removing the notion of "antagonism" between EPR and the TGF-β signaling that can lead to misunderstandings.
The Reviewer is perfectly right. We re-analyzed the RNA-Seq data derived from different subpopulations of normal breast cells isolated by FACS analysis from reduction mammoplasty specimens (presented in Figure 1d; Of course we cannot exclude that EPR is expressed also in non-epithelial cells of other tissues and, as a consequence, we have moderated our statements throughout the entire manuscript by eliminating the notion of "epithelial-restricted" EPR expression. In this revised version we provide new phase contrast microscopy images (new Figure 1g) showing more confluent cell cultures since we realized that the epithelial morphology of NMuMG cells can be less evident at lower confluence. Also, we substituted the anti-CDH1 immunoblot shown in Figure   3c with a new one (longer exposure) in which the CDH1 expression is evident also in mock cells. We would also like to share with the Reviewer additional evidence that NMuMG cells used for this study (purchased at ATCC, no. CRL-1636, maintained in culture no longer than two weeks after thawing, and repeatedly frozen) display the expected epithelial characteristics. In the Figure below is shown an example of the preliminary characterization that has been performed before using the cells for the experiments. Epithelial markers such as Occludin and Afadin are properly expressed and localized.
Immunofluorescence analysis (using the indicated antibodies) of mock-transfected NMuMG cells cultured to confluence.
. We thanks the Reviewer for her/his criticism that prompted us to include new data into our revised manuscripts. Data presented in the new Figure 3g indicate that the gene expression changes induced by either EPR or EPRSTOPE in G1-enriched cells are superimposable to those observed in the total cell population (please compare Supplementary Fig. 3g with Fig. 3b).
. As requested also by Reviewer 1 we have deleted the statement about p53. -.

Occludin Afadin
We are sorry for the lack of clarity. As described in the original protocol from Dr. Chang's laboratory, we denominated Even and Odd two distinct sets of tiling biotinylated oligonucleotides that are used in order to isolate complexes between the lncRNAs and target genomic regions. The use of two distinct sets of tiling oligonucleotides contributes to obtain specificity in the procedure. We have added a schematic on the top of Figure 5a in order to make more clear the ChIRP procedure and the position of the oligonucleotides.
The authors have addressed the concerns from the previous round of review in a satisfactory way, and the newly added data provide substantial support to their model. The only remaining minor comment (that the authors can be trusted to address by themselves) is to add as a last panel in the paper an illustration of the proposed model.
Reviewer #2 (Remarks to the Author): The authors identify a lnc-RNA, named Linc-EPR, that is downregulated during TGF-b-induced EMT, is responsible for substantial changes in gene expression and promotes the epithelial phenotype. Focusing on one target gene, they show that Linc-EPR controls the transcription activation of the Cdkn1a gene (encoding p21Cip1/Waf1) and Cdkn1a mRNA degradation, which correlates with changes in proliferation. They propose that Linc-EPR antagonizes the TGF-b response, is under the control of TGF-b, and controls epithelial-mesenchymal transition and cell proliferation.
As I mentioned in my review of the previous version, this is overall a good manuscript with attractive observations. I previously requested additional data to strengthen the conclusions, and feel that this was done only to some extent, and that, in some cases the authors preferred to weaken their conclusions rather than to strengthen their data. Furthermore, some figure panels are not sufficiently informative about controls, what was done and how the reader should evaluate the data. In providing my comments previously, I hoped that this manuscript would have been better revised, rather than just adding or changing the yellow highlighted sentences (and corresponding data panels). One question that sill stands out, and that I steered towards in my previous comments, relates to the TGF-b-inducibility of this system. Based on what I see in the data and my experience, the EPR activity will be TGF-b dependent; however, the authors do not show convincing data in that regard. Indeed, the data without adding TGF-b seem to reflect a high level of autocrine TGF-b signaling. In other words, they often compare low with higher TGF-b signaling, rather than no TGF-b versus +TGF-b. Blocking (autocrine) TGF-b signaling is easily achieved using the TbRI kinase inhibitor SB431542, but this was not done. Comparing +SB431542 and +TGF-b would give much better results. Maybe I was not sufficiently clear on how to address the TGF-b dependence, but I feel reluctant to prescribe exactly what experiments need to be done and how to do things. Finally, in their rebuttal, the authors show me some data that would benefit the manuscript. Incorporating these into the manuscript would strengthen it. After all, I am reviewing it from the standpoint of a critical reader, and the reader does not consult the reviews.  Fig. S1e is not convincing. Any controls? Also, the untreated sample shows very high Smad3 binding. Is this basal autocrine TGF-b signaling? If so, why does it decrease so much at 6 h? What is the level if you block TGF-b signaling using SB431542. These data do not look good, and do not allow for conclusions.
-As requested before, is the gene encoding EPR a direct target of TGF-b/Smad signaling? Is it induced (and then downregulated) in response to TGF-b + cycloheximide to block new protein synthesis? This should be a simple experiment.
-line 117: Why do the authors refer to Fig. S1g? What is shown? I cannot figure it out from the legend, nor do I understand it.
-line 117: Fig. 1c has no controls. What are the controls for the different compartments? If Neat1 and tRNA-Lys are seen as controls, then it looks to me that the fractionation was not good at all. The standard controls for subcellular fractionation are needed.
-lines 123-124: Any possibility to show that EPR is made in other epithelia? For example, colon, which is high in Fig. 1b. The luminal epithelial expression is interesting, but that result has only a very narrow scope. It would be nice to see more evidence that it is expressed in epithelial or differentiated epithelial cells.
-line 132: Fig. S1i is not informative since the cells are just too small to be seen.  Fig. 3b; however, these are not comparable panels.
-lines 262 and 265: "either EPR or EPRSTOPE overexpression" should be changed to "overexpression of either EPR or EPRSTOPE". It is unclear the way it currently reads -lines 303-314, related to Fig. 5b and Fig. S4c, d: The high level of Smad3 binding to the Cdkn1a promoter most likely reflects autocrine TGF-b signaling, which is then also probably the reason why the Smad3 association at the promoter seems constitutive. Blocking autocrine TGF-b signaling using SB431542 will most likely show the absence of Smad3 in the nucleus and at the promoter, and should reveal that the interaction of Smad3, and possibly of EPR is TGF-b-induced (in contrast to what we see now).
-lines 417-419: Please do not talk about the "TGF-b paradox" and the "switch". These terms are coined by those who do not have sufficient insight into the cancer biology of TGF-b, yet want to coin a quotable term. There is no paradox or switch. General comments. We followed the Reviewer's suggestion and performed a series of experiments using the inhibitor SB431542 to unambiguously demonstrate the ability of TGF-to modulated EPR expression (please see our point-by-point answers).
Finally, in their rebuttal, the authors show me some data that would benefit the manuscript.
Incorporating these into the manuscript would strengthen it. After all, I am reviewing it from the standpoint of a critical reader, and the reader does not consult the reviews.
According to the Reviewer's indication we have incorporated in this revised version some data that have been originally presented only to Reviewers (please see our point-by-point answers).
Point-by-point answers.

its legend are insufficiently informative. How can I see that TGF-b modulates the interaction of KHSRP with 67 lncRNAs?
The reviewer is right. In order to make clear to readers the results of our transcriptome-wide studies we added two new Tables (Supplementary Table 1a and Supplementary Table 1b) that include the complete list of lncRNAs whose expression levels are influenced by TGFtreatment (new Supplementary Table 1a) and a complete list of lncRNAs that interact with Rossi et al. Table 1b). Consequently, we have deleted previous Supplementary Figures 1a and 1b. -line 108: In Fig. S1c, what does P.I. stand for?

KHSRP in a TGF--regulated manner (new Supplementary
We are sorry for the inappropriate Figure Supplementary Fig. 1c) causes a limited increase of EPR levels that is not statistically significant. Thus, thanks to the Reviewer's question, we could rule out that basal autocrine TGF-signaling plays a significant role in the modulation of EPR expression in NMuMG cells. Further, as requested by the Reviewer, we present a positive control (Serpine1, lower section of panel d in the new Supplementary Fig. 1) and a negative control (Mettl9 panel e in the same Figure) for the ChIP-qPCR experiments. Importantly -prompted by the Reviewer's criticisms-we adopted an alternative ChIP protocol in order to reduce the background signal (see the revised Methods Section and the new reference 55). The new ChIP protocol allowed us to improve the specific signal-to-noise ratio in NMuMG cells. As shown in the new Supplementary Fig.   1d, SB431542 treatment completely abrogated the TGF--dependent induction of SMAD3 interaction with both EPR and Serpine1 promoters. The strong interaction of SMAD3 with the EPR promoter after 6 hours of TGF-treatment (upper section of Supplementary Fig.   1d) allows us to hypothesize that a transcriptional repressor complex including SMAD3 might associate with the promoter region to rapidly down-regulate its EPR transcription.
Future studies will be needed to obtain further molecular details on the transcriptional regulation of EPR gene expression by TGF-signaling.
-As requested before, is the gene encoding EPR a direct target of TGF-b/Smad signaling?

Is it induced (and then downregulated) in response to TGF-b + cycloheximide to block new protein synthesis? This should be a simple experiment.
To elucidate whether de novo protein synthesis is required for TGF--induced downregulation of EPR expression, we performed experiments using cycloheximide. Data presented in the new Supplementary Fig. 1f indicate that the suppression of EPR expression by TGF-does not involve de novo protein synthesis suggesting that, differently from Zeb2/SIP1 (control in the new Supplementary Fig. 1f), EPR represents a direct target of TGF-/SMAD signaling.  although in this tissue its expression is not exclusive of epithelial cells.
Single cell mRNA sequencing was performed in mouse small intestine. Avg. Difference is the log2 fold expression change between the specified group and the other groups.
To this respect, we would like to thank again the Reviewer for pushing us to moderate our previous statement (epithelial-restricted) and to refer to EPR as an epithelial-enriched lncRNA.
-line 132: Fig. S1i is not informative since the cells are just too small to be seen.
We resized the panel (new Supplementary Figure 2c) in order to improve its visibility.
-line 134: For the cobblestone morphology, the authors refer to Fig. 1f, which shows a gel.
The Reviewer is right, the sentence was too compact and, as a consequence, misleading.
We have re-written the sentence in order to avoid confusion. We apologize for the confusion. The Reviewer's remarks pushed us to completely re-write the sentence in order to clarify our points. Cells presented in Figure 1g were not treated with TGF-.
-lines 139-143: The conclusion that correlates EPR expression with EMT-related factors in normal breast (Fig. S1m)  We replaced the Immunoblot presented in the previous Supplementary Figure 2c Figure   2g and this fact prevents us from investigating Reviewer's question.
Rossi et al.
We have amended the mistake.
-lines 252-255: the authors invite the reader to compare Fig. S3g with Fig. 3b; however, these are not comparable panels.
The Reviewer is right. The sentence was not correct and we removed it. The purpose of the experiments presented in Supplementary Fig. 4g is to show that the expression changes induced in G1-enriched cells by overexpression of either EPR or EPRSTOPE are superimposable to those observed in the total cell population and, in order to make our point clearer to the readers, we ameliorated the new Supplementary Fig. 4g by showing the expression analysis of Cdh1, Tjp1, Ocln, and Tnc transcripts that are all included in Fig. 3b.
-lines 262 and 265: "either EPR or EPRSTOPE overexpression" should be changed to "overexpression of either EPR or EPRSTOPE". It is unclear the way it currently reads We have amended our mistake.
-lines 303-314, related to Fig. 5b and Fig. S4c We thank again the Reviewer for suggesting us to explore the possibility of an autocrine TGF-signaling. First, also in this case, the improvement and refinement of our ChIP protocol allowed us to increase the specific signal-to-noise ratio of the SMAD3/Cdkn1a promoter interaction. Second, the results of ChIP experiments performed in cells treated with SB431542 demonstrate that the TGF--dependent enhancement of SMAD3 interaction with Cdkn1a promoter is blunted by the compound (please compare the new Fig. 5b with the new Supplementary Fig. 5d). Notably, SB431542 does not affect the increased levels of SMAD3/Cdkn1a promoter interaction observed in cells overexpressing either EPR or EPRSTOPE when compared to mock cells (new Supplementary Fig. 5d). Our interpretation of the existence of some interaction between SMAD3 and Cdkn1a promoter in untreated EPR-overexpressing cells -which is paralleled by the increased expression of Cdkn1a-is Rossi et al. 8 that the interaction between overexpressed EPR and the Cdkn1a promoter might favor the recruitment of SMAD3 from the limited number of molecules already present in the nucleus of cells not treated with TGF-. The presence of a limited but sizeable amount of SMAD molecules in the absence of any treatment is supported by some literature (Ref. 34). To be clearer, we completely re-wrote this part (last paragraph of page 11).
-lines 417-419: Please do not talk about the "TGF-b paradox" and the "switch". These terms are coined by those who do not have sufficient insight into the cancer biology of TGF-b, yet want to coin a quotable term. There is no paradox or switch.
We followed Reviewer's suggestions and removed the words "TGF-paradox" and "switch".
Accordingly, we deleted previous reference 39 and substituted it with two new references (new Refs. 42 and 43).
We have amended our mistakes.