Alternative REST Splicing Underappreciated

Synthesis

The article by Chen and Miller argues for a more comprehensive analysis of REST splicing - specifically in Lu et al 2014. Chen and Miller provide a detailed summary of the known, validated REST variants as well as some putative variants that may exist. This is an important piece of work as there are some concerns in the field regarding some of the claims in Lu et al. Chen and Miller focus on REST splicing which is appropriate given that the bar graph in figure 1b of Lu et al is critical to the manuscript yet shows inexplicable differences in REST levels depending upon the primer pair used. This coupled with the poor schematic detailing where the primers target REST, is a reasonable basis to query this figure. Chen and Miller also show typical westerns with 2 common REST antibodies. The whole blot is shown and therefore all the bands are visible. The blots show the messy nature of REST blots that we all see but that rarely get published. Similarly, Immunofluorescence using the 2 antibodies shows very different cellular localization, again highlighting the care with which such images should be interpreted. This is especially important given the known shuttling of REST between the nucleus and cytoplasm. My concerns with current manuscript mainly focus on the writing style. The paper is hard to read due to poor grammar throughout. Examples are: ‘Supported by that immunoblotting/-staining with different anti-REST antibodies yield inconsistent results, alternative splicing allows production of various structurally and functionally different REST protein isoforms in response to shifting physiological requirements, providing a reasonable explanation for the diverse, highly context-dependent REST function.’ “it was claimed in the addendum that 7 cases from the 171 cases labeled with IHC-00141 were labeled in separate slides with ab28018 and that ‘similar patterns of REST immunoreactivity were observed with increased nuclear REST in neurons in aged versus young adult cases and reduced nuclear REST in AD’, but the sample size (7 cases comprising at least 3 experimental groups, i.e. n<3 for 1-2 groups) is extremely too small to perform statistics”

Author Response

Nature Author #1 (Remarks to the Author)

The commentary by Chen and Miller suggests that alternative splice variants in REST may influence our findings. However, these authors have not performed any of the experiments in our paper (1) and do not provide evidence that alternative splice variants of REST contribute to our findings. Even without any new data, we can say with confidence that alternative REST splicing was underappreciated in a number of previous publications including Lu et al. What makes the paper of Lu et al. doubtful is that the authors ignored previously documented REST splice variants and that the employment of multiple anti-REST antibodies in their study is questionable, giving rise to the concern that nuclear REST difference between the experimental groups (young, aged, and AD) might be artificially generated by selective or preferential use of antibodies for different samples. Our concern was actually strengthened but not eliminated by the authors response to our commentary. 1. Chen and Miller note that they have described 45 or more potential alternative splice variants of REST in a previous publication (2) and that we have not considered these variants. However, most of these potential variants have not been shown to be translated into functional proteins. In our paper, we showed that a major established REST splice variant in brain, REST4, comprised <0.5% of REST mRNA in adult human prefrontal cortex. Chen and Miller suggest that we did not exclude expression of a putative new REST exon, “exon 5”, described by these authors. However, this new exon has also not been validated, and its theoretical presence would not change our conclusion that REST protein containing the canonical N- and C-terminal domains is increased in the aging human brain and decreased in AD. Chen and Miller also note that RT-PCR data in Fig. 1b of our paper “support the existence of alternative splicing in the aging brain”. This is misleading as there is no consistent change in the relative levels of PCR products of the different primers among aged brain samples to support an age-related change in REST splicing. The only consistent change is significantly increased PCR product, consistent with our conclusion that REST mRNA is increased in the aged brain. All the new REST exons and splice variants reported in our previous paper were sequence verified and deposited in NCBI GenBank, and primers for PCR assay of specific splice variants were clearly indicated in related Figures. So, I do not understand why the authors say that exon 5 is putative and has also not been validated. As mentioned in our commentary, not all REST protein isoforms predicted by alternative splicing have been validated; however, it is true that multiple REST protein isoforms (REST4, REST1 and RESTC) have been described. Particularly, the existence of RESTC provided strong evidence that E2-lacking mRNA variants (e.g. E1/E3/E4), which are widely expressed in mammalians and resistant to conventional REST knockout strategy (excision of E2), can be translated into functional proteins by using an alternative translation start site. As for the exon 5, it is predictive of REST4 which may confound findings of Lu et al. as mentioned below. Thus, we propose that alternative REST splicing was previously underappreciated, and that REST isoforms might be differentially assayed or manipulated, resulting in conflicting findings on REST. This presumption is well supported by that different anti-REST antibodies yield inconsistent immunoblotting/- staining results, for which we presented data for two antibodies (sc-25398 and ab21635) in our commentary. The RT-PCR data (Fig.1b) in Lu et al. clearly showed that expression of the 4 targeted regions (P1-P4) of REST changed differentially in most aged groups (71, 80, 88, 95 and >95 yr), and some regions (P1 and P4 in 71-yr, P1, P3 and P4 in 95-yr, and P2 in >95-yr) displayed similar levels of expression with that of the 25-yr groups. That is, P1-P4 regions were not expressed at the same level in aged groups. If only the full-length mRNA is expressed without alternative splicing, how can this differential expression be explained? By the way, we never claimed that a consistent expression change should be observed across the aged groups, but instead, we mentioned individual variation in alternative REST splicing in our previous study. 2. Chen and Miller claim that the REST4 splice variant might contribute to our findings in SH-SY5Y cells. This variant accounts for ~10-15% of the REST transcripts in SY5Y cells. It is important to note that an integral part of the experiments in SY5Y cells was overexpression of REST by transduction of REST cDNA. REST cDNA cannot generate the REST4 splice variant. The mRNA and protein expression of REST4 in SH-SY5Y cells, which was reported by 3 different groups, was totally ignored in Lu et al. Now the authors claim that REST4 accounts for ~10-15% of REST transcripts in SH-SY5Y cells. First, it is questionable how percentage of a specific isoform was calculated without considering other previously documented splice variants (i.e. E1/E3/E4). Second, the authors mentioned above that most of the REST mRNA variants identified in our previous study have not been shown to be translated into functional proteins, which means protein expression is more important than mRNA evidence; however, it is strange that without exception, the blot area informative of REST4 protein (predicted as 37kD but observed as 53kD) was excluded from all the immunoblots presented in Lu et al., implying that the authors had no intention to clarify whether or not REST4 protein is expressed in the tissue and cell samples, although REST4 expression in SH-SY5Y had been established. The authors claim that REST4 expression in SH-SY5Y will not affect their findings because REST overexpression was examined as a positive control. First, REST overexpression was not examined in some experiments (e.g. autophagy and FOXO1 expression). Second, in terms of REST knockdown, both REST4 and the full-length REST are targeted by the two shRNAs, how can the shRNA-induced effects be attributed to the full-length REST only? Third, as mentioned below, REST4 may be stained as same as the full-length REST. Moreover, alternative splicing is epigenetically regulated while REST itself is an epigenetic regulator, what if REST4 expression changes in response to REST overexpression or knockdown? So, as mentioned in our commentary, REST4 might be a major confounding factor in Lu et al. and its expression should not be ignored. 3. Chen and Miller suggest that the multiple REST antibodies we employed were “indiscriminate”. These antibodies were chosen to span the N- and C-terminal domains of the REST protein, and were screened by multiple antibody specificity controls, as described in our paper. In contrast, the immunocytochemical labeling presented in the Chen and Miller commentary (Fig. 2) is lacking in any specificity controls, used antibodies that were not used in our paper, and was performed in cell types not used in our paper. The question of antibody specificity is especially pertinent to their labeling with REST antibody ab21635, which they claim shows REST “colocalization with microtubules” (Fig. 2). To our knowledge, this has never been reported for REST. Chen and Miller cite 2 publications as evidence for this, which is misleading as the cited papers do not show this labeling pattern. Instead of clarifying how the multiple REST antibodies were assigned to each staining, which usually should be clearly indicated in the figure or legend, the authors insist that the antibodies employed in their study can exclusively recognize the full-length REST because they tested specificity of the antibodies. However, specificity data shown in Lu et al. was insufficient to exclude the existence of REST isoforms. As we all know, immunoblotting/-staining is based on binding of the antibody to its specific epitope, which however can be shared by multiple protein isoforms. For example, REST4, REST1 and the full-length REST share the same epitope (100-150AA) for sc-15118, such that: 1) all of the 3 isoforms can be indistinguishably stained by sc- 15118; 2) following electrophoresis separation and blotting with sc-15118, they can be observed as immunoreactive bands of different sizes; and 3) peptide(s) blocking the binding of sc-15118 to its epitope will work for all of the 3 isoforms. That is, REST isoforms sharing the same epitope can only be discriminated by blotting, but not by staining. Lu et al. tested specificity of immunostaining for 2 antibodies (IHC-00141 and 07-579?), but did not perform immunoblotting to exclude the existence of multiple REST isoforms which contain epitopes of the IHC antibodies. Thus, the existence of multiple REST protein isoforms, which are indistinguishable by immunostaining, was actually not excluded in Lu et al. Due to the existence of multiple REST isoforms, different antibodies may yield different staining patterns, like the data shown in our commentary. In our opinion, to avoid data misinterpretation, at least one antibody should be consistently used for samples of different groups while additional antibodies may be employed for data validation, otherwise the use of multiple antibodies will give rise to the concern that differences between the experimental groups might be artificially generated by selective or preferential use of the antibodies. It is notable that Lu et al. employed 3 antibodies (sc-15118, IHC-00141 and ab28018) for IHC but did not compare staining patterns between the antibodies, implying that the employment of multiple antibodies was not used for data validation. As such, what is the rationale for using multiple antibodies in their study? Lu et al. did not clearly indicate how the 3 IHC antibodies were assigned to each staining, making it possible that the antibodies were preferentially used for samples of different groups. Interpretation of immunoblotting/-staining results should consider not only the full-length REST, but also potential REST isoforms (e.g. REST4, REST1 and RESTC). Notably, in accordance with loss of conservation for internal sequences of exon 4, we previously showed that partial or complete skipping of exon 4 is very common, suggesting that the full-length REST might not be predominantly expressed. Although it is unknown whether non-specific binding contributes to co-localization of REST with microtubule stained by ab21635, it has been documented that REST associates with other factors (e.g. Huntingtin and dynactin p150Glued, of which the latter binds to microtubule) to form a cytoskeletal complex, which prevents REST from being translocated into nucleus. As an antibody against C-terminus of REST, ab21635 will recognize not only the full-length REST, but also truncated C-terminal REST isoforms (e.g. RESTC). Lacking zinc finger-5 which is reportedly important for nuclear targeting, the Cterminal REST isoforms may be predominantly localized in cytoplasm but still be capable of entering the nucleus through the putative NLS. In addition, our unpublished data show that staining with ab21635 responds to manipulation of alternative REST splicing. Thus, we would rather not exclude the possibility that a proportion of REST (full-length or truncated) co-localize with microtubules. 4. Chen and Miller suggest that our results on REST-KO mice could be confounded by the expression of a Cterminal REST isoform generated by skipping exon 2 with residual biological activity. We used the same REST gene targeting strategy utilized in the original REST-KO report and subsequently used by other labs (3,4). Furthermore, our paper showed that the vulnerability phenotype of neurons derived from Nestin- Cre:REST-KO mice was rescued by expression of wild-type REST, confirming REST loss of function. If there were some residual REST function in the REST-KO, it would suggest that the phenotype we observed is an underestimation of the true neuroprotective activity of REST. As such, it would not change any of the conclusions. It is still largely unknown about the diverse, context-dependent function of REST with numerous contradictions, so its not surprising that new findings in this field may greatly update our knowledge of REST. The existence of multiple REST mRNA and protein isoforms, which are yielded by alternative splicing, strongly suggests that REST isoforms might have been differentially assayed or manipulated in previous studies; however, in most cases, only the full-length REST was considered for data interpretation. With the remaining expression of RESTC isoform in conventional REST-KO mice having been recently described, we have no doubt that previous findings obtained from such kind of mice should be re-examined. Again, what makes the paper of Lu et al. doubtful is that the authors ignored previously documented REST splice variants and that the employment of multiple anti-REST antibodies in their study is questionable, giving rise to the concern that nuclear REST difference between the experimental groups might be artificially generated by selective or preferential use of antibodies.