On 2017 Jul 07, JEANNIE LEE commented:

Authors: John E. Froberg,* Chen-Yu Wang,* Roy Blum, Yesu Jeon, and Jeannie T. Lee**

equal contribution * corresponding author **

We appreciate the response of Chen et al. to our Technical Comment. However, we do not believe the response satisfies our concerns regarding the genotypes of the cells used in Chen et al. We outline our critique below and invite commentary from the community:

1. Chen et al. continued to maintain that ∆LBS is a true deletion and stated that they confirmed ∆LBS by Sanger sequencing. They should provide FASTA files containing the Sanger sequence data from the same clone used in the study.

2. Chen et al. erroneously stated that our idea of an LBS inversion was based on the presence of discordant reads within the Xist locus. The “inversion reads” are a specific subset of discordant reads with one end mapped within the LBS region and the other just outside and where both reads align to the same strand. Whereas discordant reads can indeed be found outside of the LBS region in both WT and ∆LBS cells, such inversion reads can only be detected in ∆LBS and ∆LBS rescue, but not in WT, LBR knockdown, and SHARP knockdown samples. Thus, “inversion reads” are specific to ∆LBS and ∆LBS rescue and not simply an artifact of the discordant RAP reads.

3. In Figure 2, Chen et al. proposed that the discordant reads that suggest an inversion are simply an artifact of sequencing the RAP capture probes. This cannot be the case. The inversion reads were also found in input samples for both ∆LBS and ∆LBS rescue. The input was set aside before addition of capture probes, therefore the presence of inversion reads in input cannot be explained by sequencing the probes. Thus, the inversion reads likely arose from a genuine inverted sequence at the endogenous locus, not from capture probe contamination.

4. To explain why we were able to align many reads to the LBS region, Chen et al. proposed that LBS inserted into a repetitive region of chromosome 12 (and that the region was not flipped in situ). The authors provided no data to support this. We were also unable to see an enrichment of chr12 (or any other autosome) among the discordant reads. The authors must provide the inverse PCR Sanger sequence data and confirmation of a chr12 insertion by FISH, southern blot, or other means.

5. Chen at al. stated that our analysis of the LBS deletion is flawed because we used the wrong coordinates for ∆LBS. We used chrX:100676777-100677593 (mm9). These coordinates completely overlap and are nearly identical to the ∆LBS mm9 coordinates provided by Chen et al. (chrX: 100676791-100677575). Thus, the coordinates are essentially the same and cannot explain our differences.

6. We raised the major concern that the authors’ CLIP data lacked crossover reads that must be present if a deletion is present. In response, Chen et al. suggested that there were 4 crossover reads. We dispute that these 4 reads cross over the deleted region. Read 1 (1928049) could potentially cross over “∆LBS”, but showed deletions and mismatches that gave the read a poor alignment score (right side pair cigar: 21M3D14M). Read 2 (4677231) could also potentially cross over, but its quality score was also so low that it was filtered out by our pipeline. Read 3 (1928051) does not cross over at all, and was also filtered out due to a low quality score. Read 4 (4677228) also does not cross over and instead aligned upstream of the LBS region. Thus, the 4 reads do not qualify as crossovers, leaving unanswered why the CLIP data failed to reveal a deletion, inversion, or intact sequence.

7. The RNA FISH experiment in Figure 1F the authors use to argue against inversion may not have worked. Chen et al. argued that the absence of RNA FISH signal with probes antisense to LBS to argue that LBS is deleted and not inverted. The antisense probe should have produced one pinpoint spot in wild-type cells due to Tsix transcription. However, there is no pinpoint spot in wild-type cells.

8. The ∆A cell line they used for RAP does not have a Repeat A deletion, at least at the endogenous locus. Chen et al. admitted that they used the wrong cell line and published an incorrect dataset. They should provide full characterization of the cells actually used in the published experiment and explain how the data coincidentally supported their conclusions.

9. Chen et al. claimed that our RAP analysis is different from theirs because we did not properly account for probe sequences. This suggestion fails to explain the differences between our analyses. First, Chen et al. did not describe how they “account for probe sequences”. Second, our RAP patterns were derived after excluding discordant reads. This means that we already excluded the reads Chen et al. claim cause the discrepancy between our RAP patterns and theirs. Finally, excluding probe sequences will not change the global X-chromosomal RAP pattern because probe sequences are only found within Xist.

10. We question how Chen et al. scaled RAP coverage tracks when comparing different RAP samples. As our analysis showed that, excepting the WT RAP, all RAP experiments showed very limited coverage on the X outside of the Xist locus — thereby strongly indicating a failure of RAP experiments. These observation are not consistent with the continued assertion of Chen et al. that reduced Xist binding is found in ∆LBS and ∆A HPRT (even though they admittedly used the wrong cell line), but not in ∆LBS rescue. It remains unclear how they scaled and normalized different RAPs and how they arrived at these conclusions.

11. Furthermore, the possibility that the observed phenotypes arise from experimental variability cannot be excluded. Chen et al. argue that the fact that multiple Xist mutants (LBR KD, ∆LBR, an un-described ∆A line) all show the same phenotype eliminates the necessity of replicates. This is simply not true; it is possible that all mutants show a Xist localization defect because all of the mutant RAPs failed. The authors also argue that replicates aren’t necessary since they compare average profiles across all active vs. inactive regions of the chromosome. However, this comparison does not take into account RAP efficiency, and comparing RAP signal between active vs. inactive regions in an inefficient RAP is not biologically meaningful. Biological replicates of the same mutants in parallel with wild-type are absolutely essential for evaluating RAP phenotypes.

12. Chen et al. utilized FISH to show that perturbing LBR-Xist interaction abolishes the ability of Xist to internalize active genes into the Xist cloud (Chen et al. 2016), a phenotype similar to the ΔA Xist mutants (Chaumeil et al. 2006). Strikingly, the distance from X-linked genes to the ΔA or ΔLBS Xist clouds measured by Chen et al. spanned ~ ½ nuclear radius, far greater than what was described in Chaumeil et al. This very large nuclear distance calls into question the specificity of the FISH experiments.

13. In the final paragraph, the authors questioned the nucleation site model. However, no evidence was provided. The references they cited in connection to this statement are not relevant to the role of YY1 or to the Repeat F region for Xist nucleation. In fact, the original LBR study of Chen et al. argued that the nucleation site (a.k.a. “LBS region”) is needed for proper Xist spreading. This argument (notwithstanding the questionable genotypes) would be entirely consistent with the original findings of Jeon & Lee in 2011, which reported that deleting this region of Xist inhibits Xist spreading as a consequence of failed nucleation.

On 2017 Jul 07, JEANNIE LEE commented:

Authors: John E. Froberg,* Chen-Yu Wang,* Roy Blum, Yesu Jeon, and Jeannie T. Lee**

equal contribution * corresponding author **

We appreciate the response of Chen et al. to our Technical Comment. However, we do not believe the response satisfies our concerns regarding the genotypes of the cells used in Chen et al. We outline our critique below and invite commentary from the community:

1. Chen et al. continued to maintain that ∆LBS is a true deletion and stated that they confirmed ∆LBS by Sanger sequencing. They should provide FASTA files containing the Sanger sequence data from the same clone used in the study.

2. Chen et al. erroneously stated that our idea of an LBS inversion was based on the presence of discordant reads within the Xist locus. The “inversion reads” are a specific subset of discordant reads with one end mapped within the LBS region and the other just outside and where both reads align to the same strand. Whereas discordant reads can indeed be found outside of the LBS region in both WT and ∆LBS cells, such inversion reads can only be detected in ∆LBS and ∆LBS rescue, but not in WT, LBR knockdown, and SHARP knockdown samples. Thus, “inversion reads” are specific to ∆LBS and ∆LBS rescue and not simply an artifact of the discordant RAP reads.

3. In Figure 2, Chen et al. proposed that the discordant reads that suggest an inversion are simply an artifact of sequencing the RAP capture probes. This cannot be the case. The inversion reads were also found in input samples for both ∆LBS and ∆LBS rescue. The input was set aside before addition of capture probes, therefore the presence of inversion reads in input cannot be explained by sequencing the probes. Thus, the inversion reads likely arose from a genuine inverted sequence at the endogenous locus, not from capture probe contamination.

4. To explain why we were able to align many reads to the LBS region, Chen et al. proposed that LBS inserted into a repetitive region of chromosome 12 (and that the region was not flipped in situ). The authors provided no data to support this. We were also unable to see an enrichment of chr12 (or any other autosome) among the discordant reads. The authors must provide the inverse PCR Sanger sequence data and confirmation of a chr12 insertion by FISH, southern blot, or other means.

5. Chen at al. stated that our analysis of the LBS deletion is flawed because we used the wrong coordinates for ∆LBS. We used chrX:100676777-100677593 (mm9). These coordinates completely overlap and are nearly identical to the ∆LBS mm9 coordinates provided by Chen et al. (chrX: 100676791-100677575). Thus, the coordinates are essentially the same and cannot explain our differences.

6. We raised the major concern that the authors’ CLIP data lacked crossover reads that must be present if a deletion is present. In response, Chen et al. suggested that there were 4 crossover reads. We dispute that these 4 reads cross over the deleted region. Read 1 (1928049) could potentially cross over “∆LBS”, but showed deletions and mismatches that gave the read a poor alignment score (right side pair cigar: 21M3D14M). Read 2 (4677231) could also potentially cross over, but its quality score was also so low that it was filtered out by our pipeline. Read 3 (1928051) does not cross over at all, and was also filtered out due to a low quality score. Read 4 (4677228) also does not cross over and instead aligned upstream of the LBS region. Thus, the 4 reads do not qualify as crossovers, leaving unanswered why the CLIP data failed to reveal a deletion, inversion, or intact sequence.

7. The RNA FISH experiment in Figure 1F the authors use to argue against inversion may not have worked. Chen et al. argued that the absence of RNA FISH signal with probes antisense to LBS to argue that LBS is deleted and not inverted. The antisense probe should have produced one pinpoint spot in wild-type cells due to Tsix transcription. However, there is no pinpoint spot in wild-type cells.

8. The ∆A cell line they used for RAP does not have a Repeat A deletion, at least at the endogenous locus. Chen et al. admitted that they used the wrong cell line and published an incorrect dataset. They should provide full characterization of the cells actually used in the published experiment and explain how the data coincidentally supported their conclusions.

9. Chen et al. claimed that our RAP analysis is different from theirs because we did not properly account for probe sequences. This suggestion fails to explain the differences between our analyses. First, Chen et al. did not describe how they “account for probe sequences”. Second, our RAP patterns were derived after excluding discordant reads. This means that we already excluded the reads Chen et al. claim cause the discrepancy between our RAP patterns and theirs. Finally, excluding probe sequences will not change the global X-chromosomal RAP pattern because probe sequences are only found within Xist.

10. We question how Chen et al. scaled RAP coverage tracks when comparing different RAP samples. As our analysis showed that, excepting the WT RAP, all RAP experiments showed very limited coverage on the X outside of the Xist locus — thereby strongly indicating a failure of RAP experiments. These observation are not consistent with the continued assertion of Chen et al. that reduced Xist binding is found in ∆LBS and ∆A HPRT (even though they admittedly used the wrong cell line), but not in ∆LBS rescue. It remains unclear how they scaled and normalized different RAPs and how they arrived at these conclusions.

11. Furthermore, the possibility that the observed phenotypes arise from experimental variability cannot be excluded. Chen et al. argue that the fact that multiple Xist mutants (LBR KD, ∆LBR, an un-described ∆A line) all show the same phenotype eliminates the necessity of replicates. This is simply not true; it is possible that all mutants show a Xist localization defect because all of the mutant RAPs failed. The authors also argue that replicates aren’t necessary since they compare average profiles across all active vs. inactive regions of the chromosome. However, this comparison does not take into account RAP efficiency, and comparing RAP signal between active vs. inactive regions in an inefficient RAP is not biologically meaningful. Biological replicates of the same mutants in parallel with wild-type are absolutely essential for evaluating RAP phenotypes.

12. Chen et al. utilized FISH to show that perturbing LBR-Xist interaction abolishes the ability of Xist to internalize active genes into the Xist cloud (Chen et al. 2016), a phenotype similar to the ΔA Xist mutants (Chaumeil et al. 2006). Strikingly, the distance from X-linked genes to the ΔA or ΔLBS Xist clouds measured by Chen et al. spanned ~ ½ nuclear radius, far greater than what was described in Chaumeil et al. This very large nuclear distance calls into question the specificity of the FISH experiments.

13. In the final paragraph, the authors questioned the nucleation site model. However, no evidence was provided. The references they cited in connection to this statement are not relevant to the role of YY1 or to the Repeat F region for Xist nucleation. In fact, the original LBR study of Chen et al. argued that the nucleation site (a.k.a. “LBS region”) is needed for proper Xist spreading. This argument (notwithstanding the questionable genotypes) would be entirely consistent with the original findings of Jeon & Lee in 2011, which reported that deleting this region of Xist inhibits Xist spreading as a consequence of failed nucleation.