Skip to main content
NCBI home page
eLife logoLink to eLife
. 2019 Nov 18;8:e47214. doi: 10.7554/eLife.47214

In vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation

Daniel Andergassen 1, Zachary D Smith 1, Jordan P Lewandowski 1, Chiara Gerhardinger 1, Alexander Meissner 1,2,✉, John L Rinn 3,✉
Editors: Jeannie T Lee4, Michael B Eisen5
PMCID: PMC6860989  PMID: 31738164

Abstract

Recent evidence has determined that the conserved X chromosome mega-structures controlled by the Firre and Dxz4 loci are not required for X chromosome inactivation (XCI) in cell lines. Here, we examined the in vivo contribution of these loci by generating mice carrying a single or double deletion of Firre and Dxz4. We found that these mutants are viable, fertile and show no defect in random or imprinted XCI. However, the lack of these elements results in many dysregulated genes on autosomes in an organ-specific manner. By comparing the dysregulated genes between the single and double deletion, we identified superloop, megadomain, and Firre locus-dependent gene sets. The largest transcriptional effect was observed in all strains lacking the Firre locus, indicating that this locus is the main driver for these autosomal expression signatures. Collectively, these findings suggest that these X-linked loci are involved in autosomal gene regulation rather than XCI biology.

Research organism: Mouse

Introduction

In female mammals, one of the two X chromosomes is inactivated to compensate for gene dosage between males and females (Lyon, 1961), a process termed X-chromosome inactivation (XCI). In mice, there are two successive waves of XCI: imprinted and random. Imprinted XCI starts early in development by specifically inactivating the paternal X chromosome (Xp) (Kay et al., 1994; Okamoto et al., 2004). While the Xp remains silenced in extra-embryonic lineages (Takagi and Sasaki, 1975), it is reactivated in the embryo during implantation, followed by random XCI (Mak et al., 2004). After this decision has been made, the inactive X (Xi) chromosome is epigenetically maintained throughout cell division as a compact chromatin structure known as the ‘Barr body’ (Barr and Bertram, 1949).

Recent studies using chromosome conformation capture (3C) based methods have identified that the Xi folds into two megadomains and forms a network of long-range interactions termed superloops that are directed by the non-coding loci Firre and Dxz4 (Rao et al., 2014; Horakova et al., 2012; Deng et al., 2015). Deletion of Dxz4 in cell lines leads to the loss of both megadomain and superloop (Giorgetti et al., 2016; Bonora et al., 2018; Froberg et al., 2018; Darrow et al., 2016), while deletion of Firre alone disrupts the superloop but has no impact on megadomain formation (Froberg et al., 2018; Barutcu et al., 2018). Moreover, the Firre locus is transcribed into the long non-coding RNA (lncRNA) Firre that escapes XCI and plays a role in nuclear organization (Hacisuleyman et al., 2014; Andergassen et al., 2017; Berletch et al., 2015; Bergmann et al., 2015; Yang et al., 2015). Recent studies in human and mouse cell line models of random XCI, found that deletion of these elements have minimal impact on X chromosome biology beyond the loss of these structures, though the phenotypic consequences of their deletion throughout mammalian development have not been addressed (Giorgetti et al., 2016; Bonora et al., 2018; Froberg et al., 2018; Darrow et al., 2016; Barutcu et al., 2018).

Here, we answer this question by generating mice lacking Firre and Dxz4 and performing an extensive transcriptomic analysis in embryonic, extraembryonic and adult organs. We determined that these elements are dispensable for mouse development and XCI biology. However, the absence of these loci results in organ-specific expression changes on autosomes, suggesting that these regions are involved in autosomal gene regulation rather than XCI biology.

Results

In order to test the in vivo role of Firre and Dxz4 loci both individually and in combination, we generated three knockout (KO) mouse strains: two carrying a single locus deletion (SKO) of either Firre (Lewandowski et al., 2019) or Dxz4 and one carrying a double locus deletion (DKO) of Firre in conjunction with Dxz4 (Figure 1A, Figure 1—figure supplement 1A, Materials and methods). Notably, we targeted the regions that have been previously reported to disrupt the superloop (Barutcu et al., 2018) and megadomain structures (Bonora et al., 2018). All founder mice were screened by PCR using primers that span the deleted region and identified mutants were confirmed by Sanger sequencing (Figure 1—figure supplement 1B–C, Materials and methods).

Figure 1. Mice carrying a single or double deletion of Firre and Dxz4 are viable and fertile and show the expected litter sizes and sex ratios.

(A) Schematic representation of the active (Xa) and inactive X (Xi) chromosomes. The deleted loci of the SKO and DKO mouse strains are indicated (Firre (red), Dxz4 (gray) and DKO (blue)). The Firre locus escapes random XCI resulting in a full-length transcript from Xa chromosome and multiple short isoforms from Xi. (B) Sex type results from homozygote intercrosses. The p-value was calculated based on binomial distribution using R (binom.test(x = 19, n = 43, p=0.5, alternative = ‘less’, conf.level = 0.95)) and the sex ratio by using the following formula: number of males/number of females x 100.

Figure 1.

Figure 1—figure supplement 1. Generation of mice that carry a single or double deletion of Firre and Dxz4.

Figure 1—figure supplement 1.

(A) Schematic overview of the three genome-editing strategies, more details in the Materials and methods section. (B) Genotype strategy and required primers to identify the KO and WT allele for Firre and Dxz4 (C) UCSC genome browser showing the Firre and Dxz4 region. The PCR product of the KO bands was Sanger sequenced, and the resulting sequence was aligned to the UCSC genome browser to confirm the deletion of these loci.
Open in a new tab

We found that homozygous mice of all three strains are viable and fertile, and by crossing males and females carrying a homozygous double deletion we observed the expected litter sizes and sex ratios (Figure 1B). To test whether the absence of the Firre and Dxz4 loci has an impact on random or imprinted (Xp chromosome inactive) XCI in vivo, we collected embryonic day 12.5 female F1 brains (random/skewed XCI), placentas and visceral yolk sac (imprinted XCI) from reciprocal crosses between our KO strains and Mus musculus castaneus (CAST), followed by RNA sequencing (RNA-seq) and allele-specific analysis using the Allelome.PRO pipeline (Andergassen et al., 2015) (Figure 2A). We collected the same organs from males to test the role of these loci on the X chromosome outside of XCI biology. Unsupervised clustering of the sequenced samples confirmed the identity of the tissues (Figure 2—figure supplement 1). We then investigated the Firre expression level in Firre SKO and DKO female brains, and detected approximately half of the wildtype levels as expected for random XCI (Figure 2B, Figure 2—figure supplement 2). Next, we inspected the Firre expression level in the placenta and visceral yolk sac (extra-embryonic organs that undergo imprinted XCI) to distinguish the phenotypic consequences of deletions if they are on the Xa (maternal inheritance of the deletion) versus the Xi (paternal inheritance of the deletion) chromosome. We did not detect Firre expression if the deletion was inherited maternally, while we detected Firre wildtype expression levels in the paternal deletion, demonstrating that Firre is exclusively expressed from the Xa (Figure 2B, Figure 2—figure supplement 2). Thus, in contrast to previous reports that identified Firre as a gene that escapes XCI in cell lines that model random XCI (Hacisuleyman et al., 2014; Andergassen et al., 2017; Berletch et al., 2015), we found that Firre does not escape imprinted XCI. Remarkably, in the Dxz4 SKO we detected a Firre wildtype pattern, suggesting that disruption of the superloop between the two loci or the absence of megadomains has no impact on Firre gene expression in either random or imprinted X inactivation (Figure 2B). We did not observe changes in Xist expression levels in the absence of these loci (Figure 2C).

Figure 2. Firre is exclusively expressed from the maternal X chromosome in extra-embryonic tissues.

(A) Allele-specific RNA-seq approach to test the functional impact of the maternal and paternal deletion on random and imprinted XCI using the Allelome.PRO pipeline (Andergassen et al., 2015). While the brain undergoes random/skewed XCI (Bl6 X chromosome inactive in 70% of cells), the placenta and visceral yolk sac undergo imprinted XCI (paternal X chromosome 100% inactive). Allele-specific analysis of the placenta and visceral yolk sac allows to distinguish the effects of maternal inheritance of the deletion (Xa, forward cross) versus paternal inheritance of the deletion (Xi, reverse cross). (B) Firre expression (mean and SD) in female and male brains, placentas and visceral yolk sacs for WT and KO mouse strain. Notably, Firre is approximately 4 times higher expressed from the Bl6 allele compared to the CAST allele, as observed by comparing the expression levels between the forward cross (Xa Bl6) and reverse cross (Xa CAST) in the placenta and visceral yolk sac. (C) Xist expression abundance in placenta for the Firre-Dxz4 double KO strain.

Figure 2.

Figure 2—figure supplement 1. RNA-seq quality control from F1 brains, placentas, and visceral yolk sacs.

Figure 2—figure supplement 1.

Heatmap showing unsupervised clustering of a Pearson correlation matrix (120 brain, placenta, and visceral yolk sac samples) from expression data (TPM), confirming the expected developmental relationship.

Figure 2—figure supplement 2. Firre, Dxz4 and Xist expression across tissue, sex and genotype.

Figure 2—figure supplement 2.

Firre, Dxz4 (4933407K13Rik) and Xist expression abundance in brain, visceral yolk sac and placenta, collected from the forward (WT and maternal inheritance of the deletion) and reverse cross (WT and paternal inheritance of the deletion) for all three strains. Only one of the replicates is shown.
Open in a new tab

Since Firre is only expressed from the Xa in the placenta and visceral yolk sac (imprinted XCI), we can use these tissues to disentangle the functional roles of Firre RNA from megadomain and superloop structures that only exist on the Xi. We hypothesized that: (1) females and males carrying a maternal Firre single or Firre-Dxz4 double deletion (deletion on Xa) lack the Firre lncRNA, (2) females carrying a paternal Firre deletion (deletion on Xi) lack the superloop and (3) females carrying a paternal Dxz4 single or Firre-Dxz4 double deletion (deletion on Xi) lack both the superloop and megadomains (Figure 3A left panel). To identify dysregulated X-linked genes for each possible combination of the deletion, we performed differential expression analysis and found that the only dysregulated gene on the X chromosome was the lncRNA Firre (FDR ≤ 0.1 and |log2FC| ≥ 1), suggesting that the mega-structures and the Firre lncRNA have no impact on imprinted XCI (Figure 3A right panel, Supplementary file 1 sheet A-B). Notably, by using the same criteria we detected only a few autosomal genes dysregulated in the DKO that were not changed in the SKO, suggesting that the mega-structures and the lncRNA Firre have no impact on autosomal gene regulation in the placenta (Supplementary file 1 sheet A-B). DNA methylation levels on CpG islands were also not affected in the absence of the mega-structures or the lncRNA Firre, suggesting that the epigenetic processes involved in establishing the inactive X proceed normally without either (Figure 3—figure supplement 1, Supplementary file 1 sheet C).

Figure 3. Mice carrying a single or double deletion of Firre and Dxz4 undergo normal random and imprinted X chromosome inactivation (XCI).

(A) Schematic overview showing the effect in the placenta of the deletion in females on Xa (top) or Xi (middle), and in males (bottom) for every KO strain (left). Log2FC across the X chromosome between wildtype and KO strains (right). Firre is the only differentially expressed gene on the X chromosome (DEseq2: FDR ≤ 0.1, |log2FC| ≥ 1). (B) Boxplot showing the allelic ratios for X-linked genes in the placenta and visceral yolk sac in WT and in the KO strains, for the forward cross (deletions on the maternal X = Xa) and reverse cross (deletions on the paternal X = Xi). (C) Scatter plot showing the allelic ratios for X-linked genes in the placenta and visceral yolk sac between WT and DKO on Xa and Xi. Pearson correlation coefficient r. Maternal ratios (red) and strain-specific escaping from the CAST (brown) and Bl6 (black) X chromosome are indicated (dashed line, escaper cutoff: allelic ratio < 0.2).

Figure 3.

Figure 3—figure supplement 1. Placentas lacking Firre and Dxz4 on Xaor Xi show a similar methylation levels as in wildtype.

Figure 3—figure supplement 1.

(A) Scatter plot illustrating the relationship between mean CpG islands methylation as measured by reduced representation bisulfite sequencing (RRBS) for wildtype placenta (x axis, n = 3) versus DKO deletion on Xa or Xi (y axis, n = 2). (B) Boxplot showing CpG islands methylation levels on autosomes and on the X chromosome between wildtype and DKO deletion on Xa and Xi.

Figure 3—figure supplement 2. Allelic ratio analysis from F1 brains, placentas and visceral yolk sacs.

Figure 3—figure supplement 2.

(A) Venn diagram showing common and strain specific escaper genes in the placenta. (B) Boxplot showing the allelic ratio for WT and each KO strain of X-linked genes (left) and Xist (right) for the brain, an organ that undergoes random/skewed XCI (Bl6 X chromosome in 70% of cells inactive). The allelic ratio of Xist is significantly skewed towards inactivation of the Bl6 X chromosome in the DKO (t-test, FDR-adjusted p-value=0.0108). (C) Boxplot showing the allelic ratio on autosomes for WT and the three KO strains in the brain, placenta, and visceral yolk sac.
Open in a new tab

Next, we performed an allele-specific expression analysis to test for deviations of the expected maternal ratios or a gain of gene escape in the presence of the deletions. We found that the median allelic ratio of all the X-linked genes was unchanged regardless if the deletions were on Xa or Xi (Figure 3B). Moreover, deletion of these loci on the Xa or Xi did not result in increased escape in the placenta or visceral yolk sac (Figure 3C, Supplementary file 1 sheet D-E). Of note, allele-specific analysis revealed strain-specific escaping with a greater number of gene escape from CAST Xi compared to Bl6 Xi (Figure 3C, Figure 3—figure supplement 2A, Supplementary file 1 sheet D-E). Random XCI in the brain also did not appear to be affected, since we detect the expected XCI skewing ratios in the presence of the deletions, a well-documented effect in female cells from crosses between Bl6 and CAST that results in the predominant inactivation of the Bl6 X chromosome (Calaway et al., 2013) (Figure 3—figure supplement 2B left panel). However, our DKO animals show significant skewing of the Xist allelic ratios (t-test, FDR-adjusted p-value=0.0108) (Figure 3—figure supplement 2B right panel), suggesting that the Bl6 chromosome is further biased toward silencing in mice lacking both Firre and Dxz4. In contrast, for autosomal genes we detect the expected biallelic ratios across all samples (Figure 3—figure supplement 2C).

To address whether the absence of both the mega-structures and Firre RNA has an impact on gene expression in an organ-specific manner, we generated a transcriptomic bodymap from adult females carrying a homozygous double Firre-Dxz4 deletion (Figure 4A, Figure 4—figure supplement 1A–B). We then performed differential expression analysis of spleen, brain, kidney, heart, lung and liver and found that most of the dysregulated genes show organ-specific expression changes, primarily on autosomes (autosomes: 98.15% n = 372 chrX: 1.85% n = 7), with only a few overlapping genes across organs (Figure 4B, Supplementary file 1 sheet F). The highest number of differentially expressed genes was detected in the spleen (n = 239, compared to the rest of the organs that on average had 30 dysregulated genes).

Figure 4. Homozygous deletion of Firre and Dxz4 loci results in organ-specific expression changes on autosomes.

(A) Cartoon illustrating the structural differences of the X chromosome between wildtype and DKO female mice (left), and the organs collected from 4 WT and 4 DKO six-weeks old adult females to generate the transcriptomic bodymap (right). (B) Overview of the differentially expressed genes in the female bodymap and their overlap across the six organs (DEseq2: FDR ≤ 0.1, |log2FC| ≥ 1). The bar plot shows the proportion of differentially expressed genes between autosomes and the X chromosome. Genes that are directly affected by the deletion are underlined (Firre, Dxz4 transcript (4933407K13Rik) and crossFirre (Gm35612, antisense to Firre). (C) Cartoon illustrating the structural differences of the X chromosome between WT and each of the KO strains (left), and the SKO organs collected from six-weeks old adult females (Firre = 3, Dxz4 = 2) (right). (D) Heatmap showing the fold changes of megadomain, superloop and Firre locus dependent gene sets in the spleen. (E) MA plot for all KO strains of the genes extracted from the top five gene sets (GSEA analysis) identified in the DKO and Firre SKO spleen (more details in the Materials and methods section and Supplementary file 1 sheet I). The black dot indicates significant differentially expressed genes (DEseq2: FDR ≤ 0.1).

Figure 4.

Figure 4—figure supplement 1. Analyzing adult female organs carrying a homozygous deletion of Firre and Dxz4.

Figure 4—figure supplement 1.

(A) Heatmap showing unsupervised clustering of a Pearson correlation matrix (48 samples, DKO bodymap) from expression data (TPM), confirming the expected developmental relationship. The DKO bodymap includes spleen, brain, kidney, heart, lung and liver from 6 weeks old females (4 WT and 4 DKO replicates). (B) Firre, Dxz4 (4933407K13Rik) and Xist expression abundance in the bodymap organs. One of the replicates is shown.

Figure 4—figure supplement 2. Analyzing adult female organs carrying a homozygous deletion of Firre and/or Dxz4.

Figure 4—figure supplement 2.

(A) Heatmap showing the log2FC of megadomain, superloop and Firre locus dependent gene sets in the liver (left). The major urinary protein gene cluster, which was previously found to show high interindividual expression variation (Thoß et al., 2015) shows megadomain-dependent upregulation (right). (B) Bar plot showing the Xist expression abundance in the bodymap organs. FDR values obtained by DEseq2 analysis. (C) Top five enriched dysregulated gene sets in the spleen of the DKO and of the Firre and Dxz4 SKO (p-value corrected for multiple testing using the p.adjust function in R according to the Benjamini and Hochberg method). The GSEA analysis was performed on DEseq2 test statistics with all GO gene sets (c5.all.v6.2.symbols). Asterix indicates the overlapping gene sets.
Open in a new tab

To validate and categorize the dysregulated genes identified in the DKO into superloop, megadomain, and Firre locus-dependent gene sets, we sequenced the spleen and the liver from independently generated Firre and Dxz4 SKO animals (Figure 4C). We hypothesized that dysregulated gene sets that are: (1) shared across the three strains are superloop specific (2) shared between Dxz4 SKO and DKO are megadomain-dependent and (3) shared between Firre SKO and the DKO are Firre locus specific. To identify gene sets for these categories, we selected the genes dysregulated in at least one of the three KO strains where the direction of the expression change in either one of the single KO agrees with that of the DKO (|log2FC| ≥ 1). We first used this approach for the liver, an organ with a low number of differentially expressed genes in the DKO, and identified 15 megadomain, four superloop, and 4 Firre locus dependent genes (Figure 4—figure supplement 2A, Supplementary file 1 sheet G). Notably, without applying the fold change cutoff, we observe megadomain dependence of Xist, which is significantly upregulated in the DKO brain and kidney, as well as in the Dxz4 SKO spleen (mean upregulation 30%, Figure 4—figure supplement 2B).

We next applied the same approach to the spleen, the organ with the highest number of differentially expressed genes in the DKO, and identified seven megadomain (4%), 26 superloop (14.8%) and 142 Firre locus (81.1%) dependent genes (Figure 4D, Supplementary file 1 sheet H). The Firre locus dependent gene set—the largest group—contains only downregulated genes, including the gene Ypel4 that was previously described to form interchromosomal interactions with the Firre locus (Hacisuleyman et al., 2014; Maass et al., 2018).

By gene set enrichment analysis (GSEA), we find that the top Firre SKO and DKO enriched gene sets are almost identical, sharing downregulated gene sets involved in chromosome structure and segregation (Figure 4—figure supplement 2C, Supplementary file 1 sheet I). Genes extracted from the enriched gene sets identified in the Firre SKO and DKO share a similar pattern of downregulation, which was not observed in the Dxz4 SKO (Figure 4E). In addition, fold changes in Firre SKO and DKO are strongly correlated (spearman rho = 0.87, p-value<2.2*10−16), indicating that the Firre locus is the main driver for downregulation of these gene sets (Figure 4E). To test whether this molecular phenotype can be uncoupled form X inactivation, we performed the same analysis in DKO male spleens. We found a similar pattern of downregulation as observed in females (Figure 4E). Across all strains carrying the Firre locus deletion, the enriched gene sets share the majority of genes with the greatest degree of dysregulation (Supplementary file 1 sheet I). Taken together these findings point to the Firre locus, independently of X inactivation, as the main driver of these autosomal expression signatures.

Discussion

The Firre and Dxz4 loci provide the platform for the formation of the X chromosome mega-structures and have been extensively studied in cell lines modeling random XCI (Rao et al., 2014; Horakova et al., 2012; Deng et al., 2015; Giorgetti et al., 2016; Bonora et al., 2018; Froberg et al., 2018; Darrow et al., 2016; Barutcu et al., 2018). Here we addressed the in vivo role of these elements by generating mice carrying a single or double deletion of these loci. In agreement with previous in vitro studies, we find that the loss of these loci in vivo does not affect random XCI. The lack of dysregulated X-linked genes in adult organs suggests that these mega-structures may also not be important for long-term maintenance of the Xi chromosome. Moreover, by studying the placenta and the visceral yolk sac we were able to show for the first time that loss of these loci also does not affect imprinted XCI.

Remarkably, deletion of these loci results in reproducible organ-specific expression changes on autosomes suggesting that structural changes of the Barr body may lead to autosomal gene dysregulation. Indeed, crosstalk between autosomes and the X chromosome has been proposed as a mechanism for X inactivation counting (Rastan, 1983). Whether these changes are directly regulated by the Barr body structure remains to be investigated. The largest transcriptional effect on autosomes is Firre locus-dependent and X inactivation-independent, suggesting a role for the Firre RNA or DNA locus in autosomal gene regulation. The major function of these dysregulated genes appears to be in ontologies associated with chromosome structure and segregation, which is in line with the known role for Firre lncRNA in nuclear organization (Hacisuleyman et al., 2014). This relationship points to an RNA dependent role, which may be either direct or indirect, on autosomal gene regulation. Collectively, our results indicate that the X-linked loci Firre and Dxz4, are involved in autosomal gene regulation rather than XCI biology in vivo.

Accession codes

Sequence data and alignments have been submitted to the Gene Expression Omnibus (GEO) database under accession code GSE127554.

Materials and methods

Mouse strains

Mice were housed under controlled pathogen-free conditions (Harvard University’s Biological Research Infrastructure). C57BL/6J (Bl6), B6D2F1/J (F1 Bl6 and DBA) and CAST/Ei (CAST) mice were purchased from the Jackson Laboratory.

The Firre deletion mouse (deletion mm10: chrX:50,555,286–50,637,116) was generated by inserting LoxP sites flanking the Firre gene body into JM8A ESCs (Bl6 background) (Hacisuleyman et al., 2014). Upon injecting the ESCs into blastocysts, the resulting founder mouse was crossed with CMV-CRE (B6.C-Tg(CMV-cre)1Cgn/J, JAX Stock no: 006054. The strain was generated from BALB/c-I ESCs and backcrossed to the C57BL/6J background for 10 generations). The resulting Firre SKO mouse was subsequently backcrossed three times with Bl6, as described in detail in Lewandowski et al. (2019) (Figure 1—figure supplement 1a).

The Dxz4 single deletion strain (chrX:75,721,164–75,764,733 mm10) was generated by co-injecting Cas9 mRNA (200 ng/µl) together with two guide RNA’s that span the Dxz4 locus (50 ng/µl each) into pronuclear stage 3 (PN3) zygotes isolated after mating super ovulated B6D2F1 female mice (Jackson labs) with Bl6 males as previously described (Wang et al., 2013) (Figure 1—figure supplement 1a).

The Firre-Dxz4 double deletion strain (Dxz4 deletion chrX:75,720,836–75,764,839, Firre deletion same as for Firre SKO), was generated by piezo-assisted Intracytoplasmic sperm injection of Firre SKO sperm into B6D2F1 oocytes (protocol described in Yoshida and Perry, 2007). Cas9 mRNA was co-injected with the two Dxz4 locus spanning gRNAs as described above (Figure 1—figure supplement 1a).

Embryos were cultured to the blastocyst stage, transferred into pseudopregnant CD-1 strain females (Charles River), and brought to term. Protospacer sequences (shown in Figure 1—figure supplement 1a) were identified using ChopChop (Labun et al., 2016) and single guide RNAs were synthesized from T7 promoter containing oligonucleotides using the MEGAshortscript in vitro transcription system (Invitrogen). Founder mice of the Dxz4 SKO and DKO strains were backcrossed at least two times with C57BL/6J to remove strain background or CRISPR-Cas9 off-target effects.

In order to control for strain background, all wildtype control mice were obtained from backcrossing of founder mice with C57BL/6J to match the KO strain background.

Tissue isolation and library preparation

To determine whether the deletion of Firre, Dxz4 or Firre-Dxz4 impact random or imprinted X chromosome inactivation, we collected embryonic day 12.5 brains, placentas and visceral yolk sac from reciprocal F1 crosses between the deletion strains and CAST/EiJ. For the brain and the placenta, we collected samples for all three strains from the forward cross (3 males and three females for WT and maternal deletion) and reverse cross (three females for WT and paternal deletion). For the placenta DKO we added an additional replicate of the maternal and paternal deletion. In addition, we collected visceral yolk sac samples from the DKO forward cross (three females WT and maternal deletion) and reverse cross (three females WT and paternal deletion). To test whether DNA methylation levels are altered in the absence of Firre and Dxz4 on either Xa or Xi, we collected placentas from the DKO forward cross (one female WT and two maternal deletion) and reverse cross (two females WT and two paternal deletion) to perform reduced representation bisulfite sequencing (RRBS) sequencing. To reduce the amount of maternal contamination we removed the decidua of the placentas. For the Firre-Dxz4 adult bodymap, we collected the spleen, brain, kidney, heart, lung and liver from 6 weeks old female mice carrying a homozygous double deletion (4 WT and 4 DKO replicates). To validate and classify the dysregulated genes form the DKO, we collected liver and spleen from independent generated female SKO strains (Firre 3 SKO and Dxz4 2 SKO replicates). To test whether the molecular phenotype observed in the female spleen can be uncoupled form X inactivation, we collected 6 weeks old spleens from DKO males (2 WT and 2 DKO replicates).

The collected tissues were snap frozen and stored at −80°C until further process. RNA was extracted from TRIzol lysates using RNeasy mini columns (Qiagen). The Illumina TruSeq kit was used to create polyA+ libraries from total RNA. We generated strand-specific libraries for the F1 placentas and brains (TruSeq stranded Illumina) and non-strand-specific TruSeq libraries for the adult organs and the visceral yolk sac. Libraries were quantified using a Qubit 2.0 Fluorometer, run on an Agilent 2100 Bioanalyzer to assess purity and fragment size, and sequenced on a HiSeq 2500 at Harvard University’s Bauer Sequencing Core (75 bp paired end).

Genomic DNA for reduced representation bisulfite sequencing (RRBS) was quantified using a Qubit 2.0 Fluorometer, and quality-assessed on an Agilent 2200 TapeStation D1000 ScreenTape. RRBS was performed on 10 ng of each sample using the NuGen Ovation RRBS Methyl-Seq System following the manufacturer's recommendations except that barcoded adapter-ligated samples were pooled in groups of 8 immediately prior to Bisulfite Conversion with the Qiagen EpiTect Fast Bisulfite Conversion kit. Library pools were purified with a 1X Agencourt RNA XP bead clean-up and sequenced on a HiSeq 4000. Sequenced reads were aligned to the mm10 reference genome using BSmap (Xi and Li, 2009) and methylation states were extracted using the MOABS mcall module (Sun et al., 2014).

RNA-seq alignment and analysis

The RNA data were aligned with STAR by using specific parameters to exclude reads mapping to multiple locations (STAR version 2.5.0 c: –outFilterMultimapNmax 1) (Dobin et al., 2013). The read counts for every isoform within the RefSeq gene annotation (downloaded February 2018) were calculated by using the Python script htseq-count (HTSeq version 0.6.1) (Anders et al., 2015).

Differential expression analysis was performed with the assumption of negative binomial distribution of the read counts and empirical estimation of variance by using the R packages DESeq2 (version 1.22.1) (Love et al., 2014) and fdrtool (Strimmer, 2008). Genes were called significant if their FDR-adjusted p-values were smaller or equal than 0.1.

Allele-specific expression was detected from RNA-seq by using the Allelome.PRO, as described in detail in Andergassen et al. (2015). Allelome.PRO uses the information of characterized single-nucleotide polymorphisms (SNPs) to assign sequencing reads to the corresponding strain in F1 crosses. For the SNP annotation we first extracted 20,606,390 high confidence SNPs between the CAST/EiJ (CAST) and C57BL6NJ (Bl6) form the Sanger database as described previously (Andergassen et al., 2015; Keane et al., 2011). Although we backcrossed all the founder mice to Bl6, to exclude potential strain background confounding effects, we performed the allele-specific analysis using only CAST/Bl6 SNPs where the Bl6 allele was shared between DBA, BALB/C and 129 (Final SNP number: 15,438,314 SNPs). For the Allelome.PRO analysis we only included SNPs that are covered by at least two reads by setting the ‘minread’ parameter to 2.

Superloop, megadomain, and Firre locus-dependent category assignment

To validate and categorize the dysregulated genes identified in the DKO bodymap analysis into superloop, megadomain, and Firre locus dependent gene sets, we selected genes dysregulated in at least one of the three KO strains (FDR ≤ 0.1) where the direction of the expression change in either one of the single KO agrees with that of the DKO (|log2FC| ≥ 1). Dysregulated gene sets that are: (1) shared across the three strains were categorized as superloop specific (2) shared between Dxz4 SKO and DKO were categorized as megadomain-dependent and (3) shared between Firre SKO and the DKO were categorized as Firre locus specific (Supplementary file 1 sheet G-H).

GSEA analysis

The GSEA analysis was performed on DEseq2 test statistics with all GO gene sets (c5.all.v6.2.symbols) available from MSigDB (Subramanian et al., 2005) after mapping genes to gene sets by gene symbols. The calculation was performed in R using the CAMERA package (Wu and Smyth, 2012). A gene set was called significant if the FDR-adjusted p-value (Benjamini and Hochberg method) is less than or equal to 0.1. Top five enriched dysregulated gene sets identified in the DKO and Firre SKO spleen (Figure 4—figure supplement 2C), were extracted (Supplementary file 1 sheet I) for each of the KO strains.

Acknowledgements

We thank Philipp Maass, Marta Melé, Rasim Barutcu, Kaia Mattioli, Gabrijela Dumbovic and Quanah Hudson for stimulating discussions and critical reading of the manuscript. Nydia Chang for assistance in the mouse facility. Sequencing was performed at the Bauer Core Facility at Harvard University.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Alexander Meissner, Email: meissner@molgen.mpg.de.

John L Rinn, Email: john.rinn@colorado.edu.

Jeannie T Lee, Massachusetts General Hospital, United States.

Michael B Eisen, HHMI, University of California, Berkeley, United States.

Funding Information

This paper was supported by the following grants:

  • National Institutes of Health P01 GM099117 to John L Rinn, Alexander Meissner.

  • Howard Hughes Medical Institute Faculty Scholar to John L Rinn.

  • Max-Planck-Gesellschaft to Alexander Meissner.

Additional information

Competing interests

No competing interests declared.

Author contributions

Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing.

Conceptualization, Resources, Investigation, Methodology, Writing—review and editing.

Resources, Methodology, Writing—review and editing.

Conceptualization, Supervision, Investigation, Methodology, Writing—review and editing.

Conceptualization, Supervision, Funding acquisition, Writing—review and editing.

Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review and editing.

Ethics

Animal experimentation: Mice used in these studies were handled according to approved institutional animal care and use committee (IACUC) protocols (#28-21) of Harvard University. Procedures were performed in accordance with the National Institutes of Health guidelines for the Care and Use of Laboratory Animals.

Additional files

Supplementary file 1. RNA and reduced representation bisulfite sequencing analysis.

(A) Log2FC (lfcMLE) and adjusted p-values (padj) from DEseq2 differential expression analysis, computed by comparing female WT placentas of the forward cross (n = 8) with each of the deletion strains (deletion on Xa, n = 3–4) and female WT placentas of the reverse cross (n = 9) with each of the deletion strains (deletion on Xi, n = 3–4). (B) Log2FC (lfcMLE) and adjusted p-values (padj) from DEseq2 differential expression analysis, computed by comparing male WT placentas of the forward cross (n = 9) with each of the KO strains (n = 3). (C) Methylation levels as measured by reduced representation bisulfite sequencing (RRBS) for WT placenta and DKO deletion on Xa or Xi. (D) Imprinted ratios of X-linked placenta genes for WT and each of the deletion strains (deletions on Xa or Xi, 0.5 = 100% maternal, −0.5 = 100% paternal). The allelic ratios for each replicate per genotype was combined by using the median. (E) Imprinted ratios of X-linked visceral yolk sac genes for WT and DKO (deletions on Xa or Xi, 0.5 = 100% maternal, −0.5 = 100% paternal). The allelic ratios for each replicate per genotype was combined by using the median. (F) Differentially expressed genes (DEseq2 differential expression analysis) in all tissues (bodymap) of DKO female and liver and spleen of SKO female. (G-H) Log2FC of megadomain, superloop and Firre locus dependent gene sets in the liver and spleen. (I) Top enriched gene sets identified in the DKO and Firre SKO spleen. (J) Information of every analyzed sample in this study.

elife-47214-supp1.xlsx (2.8MB, xlsx)
DOI: 10.7554/eLife.47214.013
Transparent reporting form
elife-47214-transrepform.pdf (318.6KB, pdf)
DOI: 10.7554/eLife.47214.014

Data availability

Sequence data and alignments have been submitted to the Gene Expression Omnibus (GEO) database under accession code GSE127554.

The following dataset was generated:

Andergassen D, Meissner A, Rinn JL. 2019. In vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation. NCBI Gene Expression Omnibus. GSE127554

References

  1. Andergassen D, Dotter CP, Kulinski TM, Guenzl PM, Bammer PC, Barlow DP, Pauler FM, Hudson QJ. Allelome.pro, a pipeline to define allele-specific genomic features from high-throughput sequencing data. Nucleic Acids Research. 2015;6:gkv727. doi: 10.1093/nar/gkv727. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Andergassen D, Dotter CP, Wenzel D, Sigl V, Bammer PC, Muckenhuber M, Mayer D, Kulinski TM, Theussl HC, Penninger JM, Bock C, Barlow DP, Pauler FM, Hudson QJ. Mapping the mouse allelome reveals tissue-specific regulation of allelic expression. eLife. 2017;6:e25125. doi: 10.7554/eLife.25125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Anders S, Pyl PT, Huber W. HTSeq--a Python framework to work with high-throughput sequencing data. Bioinformatics. 2015;31:166–169. doi: 10.1093/bioinformatics/btu638. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Barr ML, Bertram EG. A morphological distinction between neurones of the male and female, and the behaviour of the nucleolar satellite during accelerated nucleoprotein synthesis. Nature. 1949;163:676–677. doi: 10.1038/163676a0. [DOI] [PubMed] [Google Scholar]
  5. Barutcu AR, Maass PG, Lewandowski JP, Weiner CL, Rinn JL. A TAD boundary is preserved upon deletion of the CTCF-rich firre locus. Nature Communications. 2018;9:1444. doi: 10.1038/s41467-018-03614-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bergmann JH, Li J, Eckersley-Maslin MA, Rigo F, Freier SM, Spector DL. Regulation of the ESC transcriptome by nuclear long noncoding RNAs. Genome Research. 2015;25:1336–1346. doi: 10.1101/gr.189027.114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Berletch JB, Ma W, Yang F, Shendure J, Noble WS, Disteche CM, Deng X. Escape from X inactivation varies in mouse tissues. PLOS Genetics. 2015;11:e1005079. doi: 10.1371/journal.pgen.1005079. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bonora G, Deng X, Fang H, Ramani V, Qiu R, Berletch JB, Filippova GN, Duan Z, Shendure J, Noble WS, Disteche CM. Orientation-dependent Dxz4 contacts shape the 3D structure of the inactive X chromosome. Nature Communications. 2018;9:1445. doi: 10.1038/s41467-018-03694-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Calaway JD, Lenarcic AB, Didion JP, Wang JR, Searle JB, McMillan L, Valdar W, Pardo-Manuel de Villena F. Genetic architecture of skewed X inactivation in the laboratory mouse. PLOS Genetics. 2013;9:e1003853. doi: 10.1371/journal.pgen.1003853. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Darrow EM, Huntley MH, Dudchenko O, Stamenova EK, Durand NC, Sun Z, Huang SC, Sanborn AL, Machol I, Shamim M, Seberg AP, Lander ES, Chadwick BP, Aiden EL. Deletion of DXZ4 on the human inactive X chromosome alters higher-order genome architecture. PNAS. 2016;113:E4504–E4512. doi: 10.1073/pnas.1609643113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Deng X, Ma W, Ramani V, Hill A, Yang F, Ay F, Berletch JB, Blau CA, Shendure J, Duan Z, Noble WS, Disteche CM. Bipartite structure of the inactive mouse X chromosome. Genome Biology. 2015;16:152. doi: 10.1186/s13059-015-0728-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. doi: 10.1093/bioinformatics/bts635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Froberg JE, Pinter SF, Kriz AJ, Jégu T, Lee JT. Megadomains and superloops form dynamically but are dispensable for X-chromosome inactivation and gene escape. Nature Communications. 2018;9:5004. doi: 10.1038/s41467-018-07446-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Giorgetti L, Lajoie BR, Carter AC, Attia M, Zhan Y, Xu J, Chen CJ, Kaplan N, Chang HY, Heard E, Dekker J. Structural organization of the inactive X chromosome in the mouse. Nature. 2016;535:575–579. doi: 10.1038/nature18589. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hacisuleyman E, Goff LA, Trapnell C, Williams A, Henao-Mejia J, Sun L, McClanahan P, Hendrickson DG, Sauvageau M, Kelley DR, Morse M, Engreitz J, Lander ES, Guttman M, Lodish HF, Flavell R, Raj A, Rinn JL. Topological organization of multichromosomal regions by the long intergenic noncoding RNA firre. Nature Structural & Molecular Biology. 2014;21:198–206. doi: 10.1038/nsmb.2764. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Horakova AH, Moseley SC, McLaughlin CR, Tremblay DC, Chadwick BP. The macrosatellite DXZ4 mediates CTCF-dependent long-range intrachromosomal interactions on the human inactive X chromosome. Human Molecular Genetics. 2012;21:4367–4377. doi: 10.1093/hmg/dds270. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Kay GF, Barton SC, Surani MA, Rastan S. Imprinting and X chromosome counting mechanisms determine xist expression in early mouse development. Cell. 1994;77:639–650. doi: 10.1016/0092-8674(94)90049-3. [DOI] [PubMed] [Google Scholar]
  18. Keane TM, Goodstadt L, Danecek P, White MA, Wong K, Yalcin B, Heger A, Agam A, Slater G, Goodson M, Furlotte NA, Eskin E, Nellåker C, Whitley H, Cleak J, Janowitz D, Hernandez-Pliego P, Edwards A, Belgard TG, Oliver PL, McIntyre RE, Bhomra A, Nicod J, Gan X, Yuan W, van der Weyden L, Steward CA, Bala S, Stalker J, Mott R, Durbin R, Jackson IJ, Czechanski A, Guerra-Assunção JA, Donahue LR, Reinholdt LG, Payseur BA, Ponting CP, Birney E, Flint J, Adams DJ. Mouse genomic variation and its effect on phenotypes and gene regulation. Nature. 2011;477:289–294. doi: 10.1038/nature10413. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Labun K, Montague TG, Gagnon JA, Thyme SB, Valen E. CHOPCHOP v2: a web tool for the next generation of CRISPR genome engineering. Nucleic Acids Research. 2016;44:W272–W276. doi: 10.1093/nar/gkw398. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lewandowski JP, Lee JC, Hwang T, Sunwoo H, Goldstein JM, Groff AF, Chang NP, Mallard W, Williams A, Henao-Meija J, Flavell RA, Lee JT, Gerhardinger C, Wagers AJ, Rinn JL. The firre locus produces a trans-acting RNA molecule that functions in hematopoiesis. Nature Communications. 2019;10:5137. doi: 10.1038/s41467-019-12970-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology. 2014;15:550. doi: 10.1186/s13059-014-0550-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lyon MF. Gene action in the X-chromosome of the mouse (Mus Musculus L.) Nature. 1961;190:372–373. doi: 10.1038/190372a0. [DOI] [PubMed] [Google Scholar]
  23. Maass PG, Barutcu AR, Shechner DM, Weiner CL, Melé M, Rinn JL. Spatiotemporal allele organization by allele-specific CRISPR live-cell imaging (SNP-CLING) Nature Structural & Molecular Biology. 2018;25:176–184. doi: 10.1038/s41594-017-0015-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Mak W, Nesterova TB, de Napoles M, Appanah R, Yamanaka S, Otte AP, Brockdorff N. Reactivation of the paternal X chromosome in early mouse embryos. Science. 2004;303:666–669. doi: 10.1126/science.1092674. [DOI] [PubMed] [Google Scholar]
  25. Okamoto I, Otte AP, Allis CD, Reinberg D, Heard E. Epigenetic dynamics of imprinted X inactivation during early mouse development. Science. 2004;303:644–649. doi: 10.1126/science.1092727. [DOI] [PubMed] [Google Scholar]
  26. Rao SS, Huntley MH, Durand NC, Stamenova EK, Bochkov ID, Robinson JT, Sanborn AL, Machol I, Omer AD, Lander ES, Aiden EL. A 3D map of the human genome at Kilobase resolution reveals principles of chromatin looping. Cell. 2014;159:1665–1680. doi: 10.1016/j.cell.2014.11.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Rastan S. Non-random X-chromosome inactivation in mouse X-autosome translocation embryos-location of the inactivation centre. Journal of Embryology and Experimental Morphology. 1983;78:1–22. [PubMed] [Google Scholar]
  28. Strimmer K. Fdrtool: a versatile R package for estimating local and tail area-based false discovery rates. Bioinformatics. 2008;24:1461–1462. doi: 10.1093/bioinformatics/btn209. [DOI] [PubMed] [Google Scholar]
  29. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. PNAS. 2005;102:15545–15550. doi: 10.1073/pnas.0506580102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Sun D, Xi Y, Rodriguez B, Park HJ, Tong P, Meong M, Goodell MA, Li W. MOABS: model based analysis of bisulfite sequencing data. Genome Biology. 2014;15:R38. doi: 10.1186/gb-2014-15-2-r38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Takagi N, Sasaki M. Preferential inactivation of the paternally derived X chromosome in the extraembryonic membranes of the mouse. Nature. 1975;256:640–642. doi: 10.1038/256640a0. [DOI] [PubMed] [Google Scholar]
  32. Thoß M, Luzynski KC, Ante M, Miller I, Penn DJ. Major urinary protein (MUP) profiles show dynamic changes rather than individual ‘barcode’ signatures. Front Ecology and Evolution. 2015;3 doi: 10.3389/fevo.2015.00071. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Wang H, Yang H, Shivalila CS, Dawlaty MM, Cheng AW, Zhang F, Jaenisch R. One-step generation of mice carrying mutations in multiple genes by CRISPR/Cas-mediated genome engineering. Cell. 2013;153:910–918. doi: 10.1016/j.cell.2013.04.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Wu D, Smyth GK. Camera: a competitive gene set test accounting for inter-gene correlation. Nucleic Acids Research. 2012;40:e133. doi: 10.1093/nar/gks461. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Xi Y, Li W. BSMAP: whole genome bisulfite sequence MAPping program. BMC Bioinformatics. 2009;10:232. doi: 10.1186/1471-2105-10-232. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Yang F, Deng X, Ma W, Berletch JB, Rabaia N, Wei G, Moore JM, Filippova GN, Xu J, Liu Y, Noble WS, Shendure J, Disteche CM. The lncRNA firre anchors the inactive X chromosome to the nucleolus by binding CTCF and maintains H3K27me3 methylation. Genome Biology. 2015;16:52. doi: 10.1186/s13059-015-0618-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Yoshida N, Perry AC. Piezo-actuated mouse intracytoplasmic sperm injection (ICSI) Nature Protocols. 2007;2:296–304. doi: 10.1038/nprot.2007.7. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Jeannie T Lee1
Reviewed by: Jeannie T Lee2

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "in vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Jeannie T Lee as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Michael Eisen as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The study generates single or double deletions of Firre and Dxz4 in mice, and show that despite the repeats being conserved in mammals, these mutants are viable, fertile and show no defect in random or imprinted XCI. Instead, the lack of Firre and Dxz4 results in dysregulated genes on autosomes in an organ-specific manner. Although the authors present mostly negative data, the manuscript would be of interest to the field and is significant because (1) it analyzes the phenotype in vivo in a mouse model, contrasting with previous papers that performed experiments only in cell culture. (2) The current manuscript also sheds light on a previous claim by Giorgetti et al., (2016) that Dxz4 is required for genes to escape from XCI. (3) It is revealed that the loci also do not play a role in imprinted XCI. All three reviewers believe that the scope of the work is in principle in line with eLife. However, for publication in eLife, we would require several major revisions.

1) Better controls to rule out strain background or off-target artifacts, when looking at changes in autosomal gene expression. (See reviewer 3 notes).

2) More backcrosses and RNA-seq of placenta, as it is the only tissue with non-random XCI, and therefore the only tissue where the authors can distinguish between effects stemming from Firre/Dxz4's roles on the inactive vs. the active X. (See reviewer 2).

3) Direct versus indirect effects of Dxz4/Firre on autosomal expression. If the authors would like to claim a direct effect, CHART analyses would be required. Otherwise, the statements should be toned down. (See reviewer 1 notes).

4) The 2 main figures should be broken up. (See reviewer 1 notes).

5) Revision of statistical analyses. (See reviewer 2 notes).

6) Finally, to the extent possible, please try to address the additional concerns of reviewers 2 and 3.

Reviewer #1:

The paper from Andergassen et al., is a succinct report made up of mostly negative findings and confirms the negative results of several past publications. Although the knockouts rule out a role in XCI, the report is significant in that (1) it analyzes the phenotype in vivo in a mouse model, contrasting with previous papers that performed experiments only in cell culture; In the majority of past reports, it was shown that the conserved X chromosome "mega-structures" controlled by the Firre and Dxz4 alleles are not required for XCI in cell lines. (2) The current manuscript also sheds light on a previous claim by Giorgetti et al., (2016) that Dxz4 is required for genes to escape from XCI. No effect on gene escape is seen by Andergassen et al., in vivo. (3) The loci also do not play a role in imprinted XCI.

The authors generate mice carrying a single or double deletion of Firre and Dxz4, and show that despite the repeats being conserved in mammals, these mutants are viable, fertile and show no defect in random or imprinted XCI. One positive finding is that the lack of Firre and Dxz4 results in dysregulated genes on autosomes in an organ-specific manner. By comparing the dysregulated genes between the single and double deletion, they categorize superloop, megadomain, and Firre locus-dependent gene sets and see that Firre deletion has the greatest effect on autosomal expression signatures. The manuscript is within the scope of papers published in eLife and would be of interest to the field. Overall, the analysese are done well and the conclusions are mostly supported by the data. However, I recommend several revisions:

1) While I accept that there are changes in autosomal gene expression, I am less convinced that the effects are direct. I am also not convinced that X-chromosomal superstructures directly affect autosomal gene expression. If the authors would like to claim a direct role, they must perform additional analyses, including – for example – CHART analysis to correlate changes at specific loci with binding of the RNA, or 3C or Hi-C to determine if the loop domains interact with autosomal gene targets. Do these autosomal targets include the loci that the authors previously showed to interact with Firre?

2) If the authors do not wish to include additional work, they should tone down the conclusions regarding autosomal changes and acknowledge the possibility of indirect effects, per point 1.

3) The figures are very dense, difficult to read (panels are small), and should be broken up into additional figures. eLife allows more than 2 display panels. At least one figure should be devoted to a more careful and deeper analysis of XCI gene escape, since the lack of effect on escape is a major point of the paper. And another figure should likewise be devoted to a deeper analysis of imprinted XCI, since this is another major point of the paper.

Reviewer #2:

The manuscript by Andergassen et al., describes the generation of mouse single and double knockouts of two X-linked macrosatellites, Firre and Dxz4. The authors analyze sex ratio, litter size and expression phenotypes of these three knockouts relative to wild-type, i.e. F1 hybrids of musculus x castaneous crosses. Whether and how these macrosatellites may participate in X chromosome inactivation or escape from XCI is an important question, due to previously described roles of Firre (not referenced PMID: 26048247 and PMID:25887447) and Dxz4 (not referenced PMID:26248554) in gene expression and inactive X chromosome conformation, respectively. However, as presented, in vivo defects were non-comprehensively analyzed, and gene expression experiments improperly controlled and somewhat over-interpreted:

1) The authors main conclusion, that there is no sex ratio distortion in the double ko, is partially undermined by a miscalculated p-value (should be 0.2712 under a p=0.5 cumulative binomial) and only six reported litters. This table should also include the outcomes of the wildtype control and single homozygous ko matings.

2) Differences in the genetic background of wild-type (not specified), single knockouts (after only two backcrosses: ~12.5% BALB/C genome for Firre, and ~6.25% DBA for Dxz4), and the double ko (variable ~1.6% BALB/C and ~6.25% DBA), are not taken into account in the gene expression analysis. This would be a non-issue with more backcrosses to C57BL/6J, at least five (see Silver, mouse genetics book).

3) Low and inconsistent numbers of replicates in the RNA-seq that reduce statistical power. This is especially problematic for attributing differential expression to loss of megadomains and superloop, because this analysis depended on comparing across multiple differential gene lists, which shrink with low statistical power (see only two Dxz4 ko replicates). This could be addressed by additional replicates from the single ko's, to get at least four across the board for liver, spleen and placenta.

4) Homozygous animals also lack Firre RNA (and possibly Dxz4-associated transcripts) expressed from the active X. The analysis here assumes that there are no confounding interactions between differential genes attributed to one group or another. A deeper analysis of the placental data in heterozygous F1 hybrid animals, where such interactions can be excluded, would be preferable and likely yield more insight towards identifying megadomain- or superloop-specific gene expression, if any.

In conclusion, while the authors present a worthwhile question and set of experiments, more careful analysis and interpretation is necessary to address functions of Firre and Dxz4 in XCI and development. At minimum, the placental RNA-seq should be repeated and fully analyzed after more than 5 backcrosses, with at least 4 replicates for each of the 4 genotypes presented. Additional novelty could come from three interesting observations reported here: (1) enrichment for differential genes implicated in faithful chromosome segregation in mice lacking Firre expression, (2) Xist upregulation by 30% in Dxz4 single and double ko's (Figure 4—figure supplement 2B), and (3) shift in allelic ratio in the spleen of Dxz4 ko's (Figure 3—figure supplement 2B). Are there specific genes driving this or is the deletion skewing XCI?

Reviewer #3:

In their study, Andergassen et al., describe mouse mutants for the noncoding RNA-gene Firre and the Dxz4 Megadomain boundary element, which are important for X-chromosome 3D-structure. It has been an important question in the X-inactivation field, what the functional significance of these elements is and what their impact is on X-chromosome inactivation. Recent studies using cell lines have already demonstrated that deletion of these elements does not affect X-chromosome inactivation (XCI) in vitro. However, the current study is the first one to assess their importance in vivo and during imprinted XCI in the placenta. The data regarding a lack of an XCI phenotype in Firre and Dxz4-deletion mice is therefore important for the X-inactivation field.

The second main message of the paper is that nevertheless changes in autosomal gene expression are observed. Only few changes appear to be dependent on Dxz4-deletion, however more are observed in the Firre KO, which is expected from its role as a non-coding RNA, which could also have function away from the X-chromosome. However, as mentioned in my specific comments below, off-target or strain background effects can at the moment not be completely ruled out. Therefore, the authors should address these points, to ensure that the autosomal changes in gene expression are indeed caused by the Firre and Dxz4 mutation and are not artifacts.

Essential revisions:

1) The authors should perform Southern blotting and/or other assays to confirm that no autosomal off-target mutations were introduced when targeting the Firre and Dxz4 loci. PCR-screening alone as done at the moment (Figure 1—figure supplement 1) is not sufficient to rule that out especially considering the low number of back-crossings and the variability between replicates when looking at changes in autosomal gene expression (Figure 2D). This would be important especially as a main message of the paper is that there are Firre/Dxz4-dependent changes in autosomal gene expression.

2) Figure 2: The different knockout alleles have been generated in mixed strain backgrounds (129, BALB/C, DBA) and only relatively few backcrosses seem to have been performed to pure B6 background (>2). Where did the wildtype mice come from, to which autosomal gene expression levels were compared in the adult expression bodymap? Were they of equivalent mixed strain background, or were wildtype controls of pure strain background used? It is important that controls of matched strain background were used throughout the study, in order to avoid differences in autosomal gene expression due to strain-background effects.

3) Figure 3—figure supplement 2B and Figure 4—figure supplement 4C: Xist RNA seems to be upregulated dependent on Megadomain deletion (Figure 4—figure supplement 4C) and the allelic-ratio seems to skew random X-inactivation in the brain more towards the M Musculus (Dxz4-deleted) allele in Dxz4 and Firre/Dxz4 KO mice. The authors should check, if Xist skewing and allelic ratios of X-linked genes are significantly changed during random XCI in the Dxz4 and Firre/Dxz4 KOs.

4) Figure 2D: The megadomain-specific upregulation of urinary cluster genes in liver seems to be very variable between replicates. It would be good to add a third replicate for delta Dxz4 as at the moment one replicate shows strong upregulation, while the second one doesn't. Also, in the Firre/Dxz4 double-mutants half of the replicates show an effect while others don't. How do the authors explain this discrepancy between replicates?

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "in vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation" for further consideration at eLife. Your revised article has been favorably evaluated by Michael Eisen (Senior Editor), a Reviewing Editor, and three reviewers.

Reviewer 1 is satisfied with the revisions (no comments attached). Reviewer 3 also feels that the paper is much improved but remains concerned about strain background. We know that it may be difficult to completely rule out strain background issues, and thus do not consider this to be an impediment to publication. However, reviewer 2 has a number of remaining concerns. In particular:

1) Please include a supplementary table listing all RNA-seq samples as reviewer 2 suggested.

2) Please present a Venn diagram listing the CAST/BL6 -common and specific escaping genes in placental RNA-seq.

3) Please add a p-value for that difference in median X-linked expression and Xist differences.

4) Please report Dxz4 SKO and DKO sex ratios in Figure 1B.

5) Reviewer 2 also requested that you please re-analyze genome-wide differential expression without log2 FC cutoff in the placental RNA-seq data to distinguish the Firre locus from superloop and megadomain differentially expressed genes (if any) for Figure 4.

6) Resolve the incongruence between the genes listed as commonly differentially expressed in Figure 4 and the gene lists in the supplementary file. Please see comments below for more details.

Reviewer #2:

The revised manuscript by Andergassen et al., adds new placental RNA-seq data that was requested by reviewers so that the authors could exploit imprinted XCI to untangle the effects of Firre and Dxz4 deletions residing on the active vs. the inactive X. This is impossible to do in tissues with random XCI as analyzed in the "bodymap". The authors should state clearly that the conclusions of their bodymap data are possibly impacted by random XCI and unaccounted genetic variation (raised in the prior round of review). There appear to be very few consistently diff. expressed genes for any of the deletions in the placenta or other tissues (except the spleen). Moreover, the authors conclude that XCI and escape in the placenta are not affected, but apply a minimal log2FC cutoff of 1, the maximal differential that could be expected if XCI or escape are impacted (going from 1 to 2 active copies or vice versa). Allelic analyses which address this question directly is only presented in aggregated figures, supplementary tables, or overly compressed log-scales, obscuring gene-specific changes. These issues need to be addressed, not by new experiments, but a more in-depth and transparent gene expression analysis:

1) Please include a supplementary table listing all RNA-seq samples (indicating replicate groups) with read depths and% reads aligned. Please include PCA plots of all samples in this analysis, next to or in place of supplemental heatmaps. Supplementary tables on placental RNA-seq showed some genes with high log2FC that failed the significance cutoff – is this a replicate/variance issue? Was batch-correction applied to minimize the impact of different strain backgrounds? There is insufficient information in the methods to replicate this analysis.

2) Seeing more escaping genes on the CAST inactive X is an interesting observation, but not presented in meaningful depth. Allele-specific analysis of genes subject to and escaping XCI in the placental RNA-seq should be plotted on log2FC scale from -3 to +3 with Firre as an outlier (Figure 3A, rather than from -12 to +12). And in Figure 3B and C, there should be a Venn diagram listing the CAST/BL6 -common and specific escaping genes in placental RNA-seq.

3) Both DKO and Dxz4 show increased Xist expression and DKO animals have a very significant change towards more BL6-biased random XCI (Figure 3—figure supplement 2, please add a p-value for that difference in median X-linked expression). Figure 4—figure supplement 2B should be moved to main Figure 3, as Xist upregulation in the absence of Dxz4 appears to be consistent across the animal. Non-coding genes regulating Xist in cis and skewing XCI were not listed in the supplement – were any of them changed? Please include these genes the supplementary tables (use Gencode reference rather than RefSeq if necessary). In view of Xist and XCI skewing, it is essential that Dxz4 SKO and DKO sex ratios are reported in Figure 1B.

4) Please re-analyze genome-wide differential expression without log2FC cutoff in the placental RNA-seq data to distinguish Firre locus from superloop and megadomain differentially expressed genes (if any). Include this info in Figure 4, and discuss in the context of the other random XCI bodymap & GSEA analysis. If Firre is only relevant in the spleen, please state so clearly.

5) There is incongruence between the genes listed as commonly differentially expressed in Figure 4 and the gene lists in the supplementary file (only 8 genes match the table & the Venn diagram in 4E). And only 45 genes are diff. expressed in the spleens of both Firre and DKO mice. Please indicate clearly which GSEA-enriched gene sets originated from which comparisons (update Venn in Figure 4E), and include info listing these enriched diff. genes in the updated supplementary tables. There is insufficient information here to replicate this analysis.

Reviewer #3:

In their revised version the authors have added some new data (visceral yolk sac data, DNA-methylation data, more replicates) and addressed many of my concerns in changing the wording and splitting the figures, making the manuscript more readable. They have not completely addressed the concerns related to the mixed strain background, which is a main weakness of the paper. Nevertheless, the authors have explained the situation in more detail, providing indirect (but no new experimental) arguments that the phenotypes observed are indeed due to the gene modifications and not due to strain background effects. Some additional doubts remain about the genotyping strategies used (PCR only – cannot rule out off-target mutations elsewhere in the genome).

Nevertheless, the paper has generally improved from the initial version and adds an interesting piece of information to the X-inactivation field regarding functional non-requirement of Firre and X-megadomains for X-inactivation in vivo, but having rather potential functions in direct or indirect autosomal gene regulation.

eLife. 2019 Nov 18;8:e47214. doi: 10.7554/eLife.47214.019

Author response

Summary:

The study generates single or double deletions of Firre and Dxz4 in mice, and show that despite the repeats being conserved in mammals, these mutants are viable, fertile and show no defect in random or imprinted XCI. Instead, the lack of Firre and Dxz4 results in dysregulated genes on autosomes in an organ-specific manner. Although the authors present mostly negative data, the manuscript would be of interest to the field and is significant because (1) it analyzes the phenotype in vivo in a mouse model, contrasting with previous papers that performed experiments only in cell culture. (2) The current manuscript also sheds light on a previous claim by Giorgetti et al. (2016) that Dxz4 is required for genes to escape from XCI. (3) It is revealed that the loci also do not play a role in imprinted XCI. All three reviewers believe that the scope of the work is in principle in line with eLife. However, for publication in eLife, we would require several major revisions.

1) Better controls to rule out strain background or off-target artifacts, when looking at changes in autosomal gene expression. (See reviewer 3 notes).

We agree with the reviewer comments that the relatively low number of backcrossing (fewer than 5) of the founder mice with the C57BL/6J strain could result in strain background differences between wild type and KO strains. To control for potential strain background or off-target effects artifacts, we performed all necessary control experiments in our initial submission which we will highlight and discuss in more detail below.

a) All the wildtype controls samples that we used for RNA-seq originate from backcrossing of the Firre, Dxz4 and double deletion founder mice with the C57BL/6J strain in order to ensure a matching genetic background to the KO strains and, thus, to allow for proper comparison between wildtype and KO strains. To clarify this point, we added the following sentence to the “Materials and methods” section (Mouse strains): “In order to control for strain background, all wildtype control mice were obtained from backcrossing of founder mice with C57BL/6J to match the KO strain background.”

b) To minimize potential strain background or off-target effects, for the Firre-Dxz4 double KO bodymap we used four biological replicates per group, collected from different litters.

c) The dysregulation of autosomal genes that was observed in the DKO mice, was confirmed in independently generated SKO mice, strongly arguing for a specific effect of the deletions, rather than strain background or off-target effects. For example, the largest spleen-specific transcriptional effect on autosomes detected in the DKO bodymap was also detected in independently-generated Firre SKO but not in the Dxz4 SKO. Thus, the observed autosomal genes dysregulation can be confidently attributed to the specific deletion of the Firre locus rather than to potential background strain differences or off-target effects.

2) More backcrosses and RNA-seq of placenta, as it is the only tissue with non-random XCI, and therefore the only tissue where the authors can distinguish between effects stemming from Firre/Dxz4's roles on the inactive vs. the active X. (See Reviewer 2)

In addition to the above examples that strongly suggest background is not influencing our conclusions, it is important to note that the suggested backcrosses would take approximately a year. At the very least this would require redoing the entire sample collection and sequencing again.

In the placenta we had not observe dysregulation of X-linked genes or gene escape in the presence of the deletions on either Xa or Xi. We have now included one additional placenta biological replicate of the DKO (deletion on Xa and Xi) and confirmed that the deletion of these loci does not have an effect on the X chromosome (new Figure 3). To further investigate whether the deletions have any impact on DNA methylation levels of the inactive X chromosome, we have now also performed reduced representation bisulfite sequencing (RRBS) on placentas lacking both Firre and Dxz4 on Xa or Xi. We found similar methylation levels as in wildtype, providing further evidence that deletion of these loci does not affect X chromosome biology (new Figure 3—figure supplement 1).

The reviewer also commented that the placenta was the only tissue in our dataset undergoing non-random XCI and we agree that it is worth examining another tissue with non-random XCI. Thus, to further explore the role of non-random XCI in Firre and Dxz4 mutants, we collected the visceral yolk sac, an extra-embryonic tissue that, similarly to the placenta, undergoes imprinted XCI. We performed RNA-seq analysis of the visceral yolk sac and as observed in the placenta, we did not detect dysregulation of X-linked genes or gene escape in the presence of the deletions either on Xa or Xi (new Figure 3) a further evidence that the lack of the Firre lncRNA or the mega-structures does not affect imprinted X chromosome inactivation.

3) Direct versus indirect effects of Dxz4/Firre on autosomal expression. If the authors would like to claim a direct effect, CHART analyses would be required. Otherwise, the statements should be toned down. (See reviewer 1 notes).

We agree with the reviewer that the effect could be either direct or indirect. Indeed, we had acknowledged that further studies are needed to establish whether the effect of the megastructures on autosomal genes regulation is direct or indirect. With regard to the autosomal targets of Firre RNA that we reported previously (Slc25a12, Ypel4, Eef1a1, Atf4 and Ppp1r10 in male mouse embryonic stem cells, PMID: 24463464), we did find that Ypel4 is downregulated in the spleen of both DKO and Firre SKO mice. We have now highlighted this finding in the revised manuscript. While this finding would suggest a direct role of Firre RNA in at least some of the expression changes, we agree that our experimental approach does not permit to discriminate between direct and indirect effects. To clarify this point, we have revised the text in the Discussion section as follow: “The major function of these dysregulated genes appears to be in ontologies associated with chromosome structure and segregation, which is in line with the known role for Firre lncRNA in nuclear organization. This relationship points to an RNA dependent role, which may be either direct or indirect, on autosomal gene regulation.”

4) The 2 main figures should be broken up. (See reviewer 1 notes).

To reduce complexity, we broke up the two main figures into 4 figures. As suggested by the reviewers, new Figure 3 is now dedicated to the analysis of imprinted XCI and gene escape, and includes the new results from the visceral yolk sac.

5) Revision of statistical analyses. (See reviewer 2 notes).

Thank you very much for the observation, we now reported the correct p-value in new Figure 1B. We also added the sex ratio, and provided additional details regarding how calculations were performed in the figure legend.

6) Finally, to the extent possible, please try to address the additional concerns of reviewers 2 and 3.

See point by point reviewer response.

Reviewer #1:

The paper from Andergassen et al., is a succinct report made up of mostly negative findings and confirms the negative results of several past publications. Although the knockouts rule out a role in XCI, the report is significant in that (1) it analyzes the phenotype in vivo in a mouse model, contrasting with previous papers that performed experiments only in cell culture; In the majority of past reports, it was shown that the conserved X chromosome "mega-structures" controlled by the Firre and Dxz4 alleles are not required for XCI in cell lines. (2) The current manuscript also sheds light on a previous claim by Giorgetti et al., (2016) that Dxz4 is required for genes to escape from XCI. No effect on gene escape is seen by Andergassen et al., in vivo. (3) The loci also do not play a role in imprinted XCI.

The authors generate mice carrying a single or double deletion of Firre and Dxz4, and show that despite the repeats being conserved in mammals, these mutants are viable, fertile and show no defect in random or imprinted XCI. One positive finding is that the lack of Firre and Dxz4 results in dysregulated genes on autosomes in an organ-specific manner. By comparing the dysregulated genes between the single and double deletion, they categorize superloop, megadomain, and Firre locus-dependent gene sets and see that Firre deletion has the greatest effect on autosomal expression signatures. The manuscript is within the scope of papers published in eLife and would be of interest to the field. Overall, the analysese are done well and the conclusions are mostly supported by the data. However, I recommend several revisions:

1) While I accept that there are changes in autosomal gene expression, I am less convinced that the effects are direct. I am also not convinced that X-chromosomal superstructures directly affect autosomal gene expression. If the authors would like to claim a direct role, they must perform additional analyses, including – for example – CHART analysis to correlate changes at specific loci with binding of the RNA, or 3C or Hi-C to determine if the loop domains interact with autosomal gene targets. Do these autosomal targets include the loci that the authors previously showed to interact with Firre?

2) If the authors do not wish to include additional work, they should tone down the conclusions regarding autosomal changes and acknowledge the possibility of indirect effects, per point 1.

We thank the reviewer for the comments. Indeed, we had acknowledged that the effect of the megastructures on autosomal genes regulation could be either direct or indirect (Discussion section of the first submission: “Whether these changes are directly regulated by the Barr body structure remains to be investigated”). With regard to the autosomal targets of Firre RNA that we reported previously (Slc25a12, Ypel4, Eef1a1, Atf4 and Ppp1r10 in male mouse embryonic stem cells, PMID: 24463464), we found that Ypel4 is downregulated in the spleen of both DKO and Firre SKO mice. We have now highlighted this finding in the Results section of the revised manuscript, as follow: (“The Firre locus dependent gene set—the largest group—contains only downregulated genes, including the gene Ypel4 that was previously described toform interchromosomal interactions with the Firre locus.”). While this finding would suggest a direct role of Firre RNA in at least some of the expression changes, we agree that our experimental approach does not permit to discriminate between direct and indirect effects. We have now clarified this point as indicated in “Essential revision” point 3.

3) The figures are very dense, difficult to read (panels are small), and should be broken up into additional figures. eLife allows more than 2 display panels. At least one figure should be devoted to a more careful and deeper analysis of XCI gene escape, since the lack of effect on escape is a major point of the paper. And another figure should likewise be devoted to a deeper analysis of imprinted XCI, since this is another major point of the paper.

Please refer to “Essential revision” point 4.

Reviewer #2:

The manuscript by Andergassen, et al., describes the generation of mouse single and double knockouts of two X-linked macrosatellites, Firre and Dxz4. The authors analyze sex ratio, litter size and expression phenotypes of these three knockouts relative to wild-type, i.e. F1 hybrids of musculus x castaneous crosses. Whether and how these macrosatellites may participate in X chromosome inactivation or escape from XCI is an important question, due to previously described roles of Firre (not referenced PMID: 26048247 and PMID:25887447) and Dxz4 (not referenced PMID:26248554) in gene expression and inactive X chromosome conformation, respectively. However, as presented, in vivo defects were non-comprehensively analyzed, and gene expression experiments improperly controlled and somewhat over-interpreted:

1) The authors main conclusion, that there is no sex ratio distortion in the double ko, is partially undermined by a miscalculated p-value (should be 0.2712 under a p=0.5 cumulative binomial) and only six reported litters. This table should also include the outcomes of the wildtype control and single homozygous ko matings.

Thank you very much for catching this error. We have now corrected it as indicated in point 5 of “Essential revision”.

Given that the male-female ratio of the DKO progeny was not significantly altered, we reasoned that testing for sex ratio skewing within the WT control group was not required. With regard to the SKO strains, the sex ratio of the Firre SKO, and corresponding WT, was evaluated as part of another recent study from our group (Lewandowski et al., 2019), and it was found to be normal. Based on the finding in the DKO and Firre SKO strains, we reasoned that a sex ratio skewing in the Dxz4 SKO strain would also be unlikely.

2) Differences in the genetic background of wild-type (not specified), single knockouts (after only two backcrosses: ~12.5% BALB/C genome for Firre, and ~6.25% DBA for Dxz4), and the double ko (variable ~1.6% BALB/C and ~6.25% DBA), are not taken into account in the gene expression analysis. This would be a non-issue with more backcrosses to C57BL/6J, at least five (see Silver, mouse genetics book).

We agree with the reviewer of the importance to control for potential strain background differences. We addressed this point in “Essential revision” point 1. In addition, we would like to clarify that the Firre SKO mouse was generated by inserting LoxP sites flanking the Firre gene body into JM8A ESCs (PMID: 19525957) which are of Bl6 background and not 129xBl6 as we indicated in the original submission. The ESCs were then injected into blastocysts to obtain chimeras that were then backcrossed to Bl6 to obtain the founder mouse. The founder mouse was then crossed with CMV-CRE mice B6.C-Tg(CMV-cre)1Cgn/J, JAX Stock no: 006054. While the CRE JAX strain was generated from BALB/c-I ESCs, stock 006054 was backcrossed to the Bl6 background for 10 generations. Thus, the background strain is Bl6 and not BALB/c as we indicated in the original submission. We apologize for the confusion. The resulting Firre SKO mouse was subsequently backcrossed three times with Bl6, more details in Lewandowski et al. bioXriv 2019. Based on this consideration the background of the Firre SKO is almost 100% Bl6 (Figure 1—figure supplement 1A). The Dxz4 single deletion strainwas generated by co-injecting Cas9 mRNA together with two guide RNA’s that span the Dxz4 locus into pronuclear stage 3 (PN3) zygotes isolated after mating super ovulated B6D2F1 (50% Bl6 50% DBA) female mice with Bl6 males resulting in a founder mouse with 75% Bl6 background. The Founder mouse was then backcrossed two times with Bl6, and can thus be considered as 93.75% Bl6 background (Figure 1—figure supplement 1a). The Firre-Dxz4 double deletion strain, was generated by piezo-assisted Intracytoplasmic sperm injection of Firre SKO sperm (Bl6 background) into B6D2F1 (50% Bl6 50%DBA) oocytes (at PN3, Cas9 mRNA was co-injected with the two Dxz4 locus spanning gRNAs.) The resulting DKO founder mouse was then backcrossed two times with Bl6, and can thus be considered as 93.75% Bl6 (Figure 1—figure supplement 1A). All the wildtype controls samples that we used for RNA-seq originate from backcrossing of the Firre, Dxz4 and DKO founder mice with the Bl6 strain in order to ensure a matching genetic background to the KO strains and, thus, to allow for proper comparison between wildtype and KO strains.

We have now provided a more detailed description of the generation of all strains, as well as how the corresponding WT mice were obtained, in the Materials and methods section of the revised manuscript.

3) Low and inconsistent numbers of replicates in the RNA-seq that reduce statistical power. This is especially problematic for attributing differential expression to loss of megadomains and superloop, because this analysis depended on comparing across multiple differential gene lists, which shrink with low statistical power (see only two Dxz4 ko replicates). This could be addressed by additional replicates from the single ko's, to get at least four across the board for liver, spleen and placenta.

We agree with the reviewer that having additional replicates from the SKO would increase the statistical power of the analysis in assigning differential expression to loss of Firre and megadomains and superloop. However, we only used the SKO to verify the direction of dysregulation observed in the DKO bodymap, where we had enough statistical power given that we used four biological replicates per group for this analysis.

4) Homozygous animals also lack Firre RNA (and possibly Dxz4-associated transcripts) expressed from the active X. The analysis here assumes that there are no confounding interactions between differential genes attributed to one group or another. A deeper analysis of the placental data in heterozygous F1 hybrid animals, where such interactions can be excluded, would be preferable and likely yield more insight towards identifying megadomain- or superloop-specific gene expression, if any.

Thank you very much for this suggestion. To address this point, we have performed an additional analysis to test whether the DKO deletion on either Xa or Xi might result in autosomal gene regulation in the placenta, a tissue where we can disentangle the functional role of Firre RNA on Xa from that of megadomain and superloop structures that only exist on the Xi. We detected only a few autosomal genes dysregulated in the DKO deletion on either Xa or Xi that were not changed in the SKO, suggesting that the mega-structures and the lncRNA Firre have no impact on autosomal gene regulation in the placenta (Supplementary file 1 sheet A-B). We have now highlighted this finding in the Results section of the revised manuscript, as follow: (“Notably, by using the same criteria we detected only a few autosomal genes dysregulated in the DKO that were not changed in the SKO, suggesting that the mega-structures and the lncRNA Firre have no impact on autosomal gene regulation in the placenta. (Supplementary file 1 sheet A-B).”)

In conclusion, while the authors present a worthwhile question and set of experiments, more careful analysis and interpretation is necessary to address functions of Firre and Dxz4 in XCI and development. At minimum, the placental RNA-seq should be repeated and fully analyzed after more than 5 backcrosses, with at least 4 replicates for each of the 4 genotypes presented.

We have now included a fourth biological replicate of the placenta DKO as indicated in point 2 of “Essential revision”. For the strain background/backcrossing issue please refer to “Essential revision” point 1.

Additional novelty could come from three interesting observations reported here: (1) enrichment for differential genes implicated in faithful chromosome segregation in mice lacking Firre expression, (2) Xist upregulation by 30% in Dxz4 single and double ko's (Figure 4—figure supplement 2B), and (3) shift in allelic ratio in the spleen of Dxz4 ko's (Figure 3—figure supplement 2B). Are there specific genes driving this or is the deletion skewing XCI?

a) It would be very interesting to follow up the Firre dependent enrichment of differential genes involved in chromosome structure and segregation, however we believe that this would be outside the scope of this study and will instead be followed up by a separate study.

b) Xist upregulation by 30% in Dxz4 single and DKO is an interesting observation because it suggests that the Dxz4 locus or the resulting megastructure is regulating Xist expression. Given that we don’t detect dysregulated genes on the X chromosome in the absence of Dxz4, our interpretation of this finding is that Xist upregulation occurs only from the inactive X chromosome, suggesting a role for maintaining X chromosome silencing. It would be very interesting to follow Xist expression in Dxz4 deletions to test whether Xist upregulation correlates with aging, however, that is outside the scope of this current study.

c) In Figure 3—figure supplement 2B we show the allelic ratio of the E12.5 F1 brains (not the spleen) and found the expected XCI skewing ratios, a well-documented effect in female cells from crosses between Bl6 and CAST that results in the predominant inactivation of the Bl6 X chromosome. As pointed out by reviewer 2 and 3, the XCI skewing was even stronger in the Dxz4 and DKO brains. To test whether the additional skewing is significant we performed a t-test on the Xist allelic ratios between WT (n=17) and each of the KO strains (n=6) and observed significant skewing (t-test, adjusted p-value = 0.0108) in the presence of the DKO deletion (Figure 3—figure supplement 2A right), suggesting that the Bl6 chromosome is further biased towards silencing in mice lacking both Firre and Dxz4. We have now highlighted this finding in the Results section of the revised manuscript, as follow: (“However, our DKO animals show significant skewing of the Xist allelic ratios (t-test, adjusted p-value = 0.0108) (Figure 3—figure supplement 2A right panel), suggesting that the Bl6 chromosome is further biased towards silencing in mice lacking both Firre and Dxz4.”)

Reviewer #3:

1) The authors should perform Southern blotting and/or other assays to confirm that no autosomal off-target mutations were introduced when targeting the Firre and Dxz4 loci. PCR-screening alone as done at the moment (Figure 1—figure supplement 1) is not sufficient to rule that out especially considering the low number of back-crossings and the variability between replicates when looking at changes in autosomal gene expression (Figure 2D). This would be important especially as a main message of the paper is that there are Firre/Dxz4-dependent changes in autosomal gene expression.

Please refer to point 1 of “Essential revision”.

2) Figure 2: The different knockout alleles have been generated in mixed strain backgrounds (129, BALB/C, DBA) and only relatively few backcrosses seem to have been performed to pure B6 background (>2). Where did the wildtype mice come from, to which autosomal gene expression levels were compared in the adult expression bodymap? Were they of equivalent mixed strain background, or were wildtype controls of pure strain background used? It is important that controls of matched strain background were used throughout the study, in order to avoid differences in autosomal gene expression due to strain-background effects.

We agreed with the reviewer comment and have addressed this in “Essential revision” point 1-2 and in the response to reviewer 2 point 2.

3) Figure 3—figure supplement 2B and Figure 4—figure supplement 4C: Xist RNA seems to be upregulated dependent on Megadomain deletion (Figure 4—figure supplement 4C) and the allelic-ratio seems to skew random X-inactivation in the brain more towards the M Musculus (Dxz4-deleted) allele in Dxz4 and Firre/Dxz4 KO mice. The authors should check, if Xist skewing and allelic ratios of X-linked genes are significantly changed during random XCI in the Dxz4 and Firre/Dxz4 KOs.

Thank you very much for this observation; we have addressed this in reviewer 2 point 4c.

4) Figure 2D: The megadomain-specific upregulation of urinary cluster genes in liver seems to be very variable between replicates. It would be good to add a third replicate for delta Dxz4 as at the moment one replicate shows strong upregulation, while the second one doesn't. Also, in the Firre/Dxz4 double-mutants half of the replicates show an effect while others don't. How do the authors explain this discrepancy between replicates?

As noted by reviewer 2, the expression of major urinary protein genes was shown to be highly variable across individuals (PMID: 26973837) which explains the discrepancy across replicates.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Reviewer 1 is satisfied with the revisions (no comments attached). Reviewer 3 also feels that the paper is much improved but remains concerned about strain background. We know that it may be difficult to completely rule out strain background issues, and thus do not consider this to be an impediment to publication. However, reviewer 2 has a number of remaining concerns. In particular:

1) Please include a supplementary table listing all RNA-seq samples as Rev2 suggested.

We listed all the RNA-seq samples in the supplementary table (sheet: J sample information) and updated the table legend.

2) Please present a Venn diagram listing the CAST/BL6 -common and specific escaping genes in placental RNA-seq.

We included a Venn diagram listing common and strain specific escaper genes in the placenta in Figure 3—figure supplement 2A.

3) Please add a p-value for that difference in median X-linked expression and Xist differences.

We have added the FDR-adjusted p-values for the X-linked genes.

4) Please report Dxz4 SKO & DKO sex ratios in Figure 1B.

As pointed out in the previous round of revision, the male-female ratio of the DKO progeny was not significantly altered (Figure 1B, 6 litters, 53 pups p=0.2712), suggesting that sex ratio skewing in the Dxz4 SKO strain would be unlikely. For this reason, we kept Dxz4 SKO line to a minimum and mainly used the strain to generate crosses with CAST for the allele-specific placenta analysis. The sex-ratio of the Firre SKO progeny is also normal (Lewandowski et al., in press) and, since the Firre locus is the main driver of the observed expression changes, we feel that assessing sex-ratio in the Dxz4 SKO is outside the scope of this study.

The main biological question of this study was to investigate the in vivo role of the two Xlinked loci Firre and Dxz4, that provide the platform of the largest conserved chromatin structures in female mammals. We are confident that we addressed this question in the current manuscript and believe that the addition of the male-female ratios of the Dxz4 SKO progeny will not change the conclusion of this study, bearing in mind that the ratios were not altered in the DKO.

5) Reviewer 2 also requested that you please re-analyze genome-wide differential expression without log2 FC cutoff in the placental RNA-seq data to distinguish the Firre locus from superloop and megadomain differentially expressed genes (if any) for Figure 4.

As suggested, we have re-analyzed genome-wide placental RNA-seq differential expression without log2FC cutoff. By comparing differentially expressed genes between the DKO and each of the SKO we find 3 overlapping genes between the DKO and Firre SKO on Xa (Firre RNA-specific), 45 genes between the DKO and Firre SKO on Xi (superloop-specific) and none between the DKO and Dxz4 SKO on Xi (megadomainspecific). Of note, in all cases (with the exception of the deleted Firre gene), the mean log2FC between the DKO and SKO’s is close to zero (see Author response image 1). This result indicates that the mega-structures and the lncRNA Firre have no impact in gene regulation in the placenta as we observed in our original analysis (log2FC ≥1).

Author response image 1.

Author response image 1.

Open in a new tab

Given that these results are consistent with those from our original analysis, we did not include the new analysis in the revised manuscript.

6) Resolve the incongruence between the genes listed as commonly differentially expressed in Figure 4 and the gene lists in the supplementary file. Please see comments below for more details.

The genes listed in the supplementary file (Supplementary file 1 sheet G-H) and reported in the heatmap in Figure 4D (and Figure 4—figure supplement 2A) have been identified by differential expression analysis and represent genes dysregulated in at least one of the three KO strains (FDR ≤ 0.1) where the direction of the expression change in either one of the single KO agrees with that of the DKO (|log2FC| ≥ 1). The genes in Figure 4E (MA-plot and Venn diagram), instead, are the genes within the top 5 enriched dysregulated gene sets identified by the GSEA analysis in the DKO and Firre SKO spleen (Figure 4—figure supplement 2C).

To make clear that these are two different analyses we have added the following header to Figure 4E “GSEA identified chromosome structure and segregation gene sets”. We have also provided additional information in the method section and included the enriched GSEA gene sets and the corresponding DESeq2 output in Supplementary file 1 sheet I (genes ranked by lfcMLE) to allow for replication of the analysis. Given that all the information of figure 4E is now provided in the supplementary table, we removed the Venn diagram and instead added the spearman correlation of the fold changes.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Andergassen D, Meissner A, Rinn JL. 2019. In vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation. NCBI Gene Expression Omnibus. GSE127554 [DOI] [PMC free article] [PubMed]

    Supplementary Materials

    Supplementary file 1. RNA and reduced representation bisulfite sequencing analysis.

    (A) Log2FC (lfcMLE) and adjusted p-values (padj) from DEseq2 differential expression analysis, computed by comparing female WT placentas of the forward cross (n = 8) with each of the deletion strains (deletion on Xa, n = 3–4) and female WT placentas of the reverse cross (n = 9) with each of the deletion strains (deletion on Xi, n = 3–4). (B) Log2FC (lfcMLE) and adjusted p-values (padj) from DEseq2 differential expression analysis, computed by comparing male WT placentas of the forward cross (n = 9) with each of the KO strains (n = 3). (C) Methylation levels as measured by reduced representation bisulfite sequencing (RRBS) for WT placenta and DKO deletion on Xa or Xi. (D) Imprinted ratios of X-linked placenta genes for WT and each of the deletion strains (deletions on Xa or Xi, 0.5 = 100% maternal, −0.5 = 100% paternal). The allelic ratios for each replicate per genotype was combined by using the median. (E) Imprinted ratios of X-linked visceral yolk sac genes for WT and DKO (deletions on Xa or Xi, 0.5 = 100% maternal, −0.5 = 100% paternal). The allelic ratios for each replicate per genotype was combined by using the median. (F) Differentially expressed genes (DEseq2 differential expression analysis) in all tissues (bodymap) of DKO female and liver and spleen of SKO female. (G-H) Log2FC of megadomain, superloop and Firre locus dependent gene sets in the liver and spleen. (I) Top enriched gene sets identified in the DKO and Firre SKO spleen. (J) Information of every analyzed sample in this study.

    elife-47214-supp1.xlsx (2.8MB, xlsx)
    DOI: 10.7554/eLife.47214.013
    Transparent reporting form
    elife-47214-transrepform.pdf (318.6KB, pdf)
    DOI: 10.7554/eLife.47214.014

    Data Availability Statement

    Sequence data and alignments have been submitted to the Gene Expression Omnibus (GEO) database under accession code GSE127554.

    The following dataset was generated:

    Andergassen D, Meissner A, Rinn JL. 2019. In vivo Firre and Dxz4 deletion elucidates roles for autosomal gene regulation. NCBI Gene Expression Omnibus. GSE127554

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES