TgBAC8 mouse lines were generated and described previously as transgenic mice carrying 200 kb of sequence from Xic including Xist (Sun et al., 2015). C57BL/6J-TgBAC8-2087 line was used in this study, for which Tg2087/+ mice were crossed to wildtype C57B1/6J mice to obtain offspring that contain littermates of transgenic (Tg2087/+) and wildtype (+/+) animals. Tail tip fibroblasts (TTF) were isolated from tail tip tissues of transgenic and wildtype mice from the same litter and grown in DMEM medium containing 10% FBS supplemented with MEM non-essential amino acids, 25 mM HEPES pH 7.2–7.5, 0.1 mM 2-mercaptoethanol, penicillin, and streptomycin. TTFs were then immortalized with Simian virus 40 large T cDNA following the protocol as previously described (Brown et al., 1986). Transgenic samples and wildtype (control) samples were processed in parallel following the same procedure.

Total RNA was extracted from cells grown on a 15 cm tissue culture plate using Trizol (Thermo Fisher Scientific), followed by DNase treatment using TURBO DNA-free^TM Kit (Thermo Fisher Scientific). Polyadenylated mRNA was isolated from total RNA using Dynabeads^TM mRNA Purification Kit (Thermo Fisher Scientific). RNA integrity was checked by Bioanalyzer (Agilent 2100), followed by fragmentation with Mg2+ -catalyzed hydrolysis.

cDNA library was constructed as previously described (Mortazavi et al., 2008). In brief, the first-strand cDNA was reverse-transcribed with random primers, followed by deoxyadenosine (dA) addition and adaptor ligation. Then, an RNA gel-purified was performed for size selection of 300 bp, followed by UNG treatment for strand specificity and PCR amplification.

cDNA sequencing was performed as previously reported (Mortazavi et al., 2008). DNA quality and size were checked by Bioanalyzer (Agilent 2100), quantified by KAPA Library quantification Kit (Illumina), followed by sequencing on NextSeq500 (Illumina) using paired-end mode. The sequencing data have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series Accession number GSE 156393.

Analysis of RNA-seq data was performed with the Bioconductor package Rsubread (index building, alignment with genome mm10). The count table was generated with the feature count function in R and the counts were converted into FKPM by using the R package “count to FPKM.” For differential gene expression, we used the package DESeq.

DNA and RNA FISH were performed with probes generated using BAC clones as DNA templates (Sun et al., 2013). Mouse chromosome 1 BAC clones were obtained from the CHORI BACPAC Resources Center¹ with the following ID numbers: RP24-402L7 (Segment A); RP24-159O10 (Segment B); RP24-232L18 (Segment C); RP23-2C9 (Segment E). Fluorescent Cyanine and Fluorescein probes were then generated using a Nick Translation Kit (Roche) supplemented with either Fluorescein-12-dUTP or Cyanine-3-dUTP (Enzo) according to the manufacturer’s protocol. Unincorporated nucleotides were removed using G-50 columns (GE). FISH probes were ethanol precipitated and resuspended in 50% formamide, 2xSSC pH7.4, 10% dextran sulfate, 1 mg/ml BSA, and 100 ng/μl mouse Cot-1 DNA. Transgenic and wildtype TTFs were grown on gelatin treated 10-well microscope slides (Fisher Scientific). Cells were permeabilized in the CSKT buffer at 4°C for 5 min and then fixed in 4% paraformaldehyde at room temperature for 10 min. For RNA FISH, slides were not denatured. For DNA FISH, slides were treated in 400 μg/ml RNaseA at 37°C for 30 min, followed by a PBS wash and a treatment with 0.2 N HCl 0.5% Tween-20 at room temperature for 10 min. The slides were denatured in 70% formamide, 2xSSC pH7.4 at 80°C for 10 min, followed by sequential dehydration in 70, 80, and 100% ethanol at 4°C. FISH probes of 10 ng/μl were denatured at 80°C for 10 min and then kept at 37°C for more than 15 min, and then loaded directly onto cell samples on the slides. With coverslips placed on top of the cells on the slides, hybridizations were carried out in a humidified chamber at 37°C for 16 h. Afterward, slides were washed with 50% formamide, 2xSSC pH7.4 at 37°C for 3 times, and then with 0.1xSSC pH7.4 at 37°C for 3 times. Slides were air dried and mounted in VECTASHIELD Mounting Media (Vecta Labs) with DAPI 1.5 μg/ml. Slides were visualized and imaged with Zeiss LSM780 and LSM700 confocal microscopes, and the images were processed with ZEN Software (Zeiss) and Volocity software (Perkin Elmer).

For dual-color DNA FISH, chromosome paints were purchased as XCyting Mouse Chromosome Painting Probes (MetaSystems): XMP X Orange D-1420-050-OR (Chr X); XMP Y Green D-1421-050-FI (Chr Y); XMP 1 Green D-1401-050-FI (Chr 1); XMP 4 Green D-1404-050-FI (Chr 4); XMP 9 Green D-1409-050-FI (Chr 9). For chromosome and Xic localization analysis, Z-section images were captured and analyzed within each individual section across all sections of the Z stack. In particular, co-localization events between the chromosome paint signals and the Xic signals were detected by examining the section images individually in Volocity software. Measurements between chromosomal segments (on chromosome 1) and the transgene were performed with 3D Image Analysis Tools in Volocity software. The BAC probes yield discrete pinpoint spots by DNA FISH and distances are measured by drawing a straight line between the center of each spot. Nuclei with more than two resolvable signals were scored. Cells that displayed two or more signals and the two shortest measurements between the chromosomal segment and the transgene were recorded and the average value was taken for each cell. The normalized distance was obtained by subtracting the median distance measurements across all WT cells from each averaged measurement for each TG cell. Statistical analysis was performed in R Studio.

Using transgenic mouse models, our previous work had defined a 200-kb Xic sequence including five lncRNAs (Figure 1A), which are sufficient to induce and maintain transgenic Xist expression from an autosome in both male and female mouse embryos (Sun et al., 2015). In this study, we analyzed the transcriptional consequence of the ectopic Xist expression in the transgenic mouse line Tg2087, which carries two copies of the Xic sequence integrated in a single autosomal locus. Hemizygous Tg2087/+ males are fertile, whereas homozygous Tg2087 males are sterile with defective sperm cells (Sun et al., 2015). We isolated tail-tip fibroblasts (TTFs) from the progeny of hemizygous Tg2087/+ crossed with wildtype, which produced equal frequencies of Tg2087/+ and +/+ offspring. A comparative gene expression analysis between transgenic and wildtype cells was performed using immortalized TTFs from Tg2087/+ and +/+ male littermates. The immortalized TTF cells were tetraploid due to SV40 Large T-transformation. As shown in Figure 1B visualizing Xist expression by RNA FISH, wildtype (WT) male cells showed no Xist signal, whereas the tetraploid Tg2087/ + male cells showed two Xist clouds. This confirms a stable transgenic Xist expression.

To investigate how ectopic Xist expression alters transcription of other genes, we performed total RNA sequencing on transgenic (TG) and wildtype (WT) samples in duplicate. Analysis of the RNA-seq results validated the Xist expression induced in transgenic male cells as compared to the wildtype controls. As shown in Figures 1C,D, WT male fibroblasts had no or only baseline expression of Xist. By contrast, TG male fibroblasts exhibited a significant increase of Xist transcripts. Comparative gene expression analysis identified differentially expressed coding and non-coding genes in TG as compared to WT. 130 genes were down-regulated and 83 genes were upregulated in TG (with adjusted p < 0.01 and Abs(logFC) ≥ 1). In particular, chromosomes 1, 2, and 9 were associated with relatively higher numbers of down-regulated genes with 14, 12, and 11 down-regulated genes, respectively (Figure 1E and Supplementary Table 1). Further analysis showed that Chr 1 carried the most number of down-regulated genes that are physically linked: the heatmap which displays any down-regulated gene that is less than 5 Mb apart from another down-regulated gene within the same chromosome showed seven genes were clustered in Chr 1, and five and four genes clustered in Chr 2 and Chr 9, respectively (Figure 1F). These results indicate that the autosomal Xic transgene in Tg2087/ + maintained active transcription of Xist, which affected genome-wide expression with specific silenced genes mostly clustered in Chr 1, Chr 2, and Chr 9.

Transcriptomic changes between TG and WT cells could be the result of both primary and secondary effects of transgene integration. To determine which chromosome was most affected by Xist transgene silencing, we applied three criteria: (1) the number of genes down-regulated within the chromosome, (2) the percentage of genes down-regulated out of all the genes within the chromosome, and (3) the number of down-regulated genes that are located in clusters within the chromosome. Using these three criteria, Chr 1 and Chr 9 were clearly the top candidates most affected by transgene silencing. Note that the transcriptional silencing on Chr 2 appeared much less significant once the total number of genes on the chromosome was taken into account. Specifically, 0.7% of the genes on Chr 2 were down-regulated as compared to 1.2 and 1.0% of genes down-regulated on Chr 1 and Chr 9, respectively (Supplementary Table 2). Additionally, we also considered a less stringent cutoff threshold of Abs(logFC) = 0.5) for RNA-seq analysis. This led to slightly higher numbers of genes shown as down-regulated across most of the chromosomes, and moreover, chromosomes 1, 8, 9 were ranked together at the top with the highest percentages of down-regulated genes (Supplementary Table 2). However, Chr 8 was excluded for further analysis since the down-regulated genes on Chr 8 had only three of them being clustered within the chromosome. In all, we concluded that chromosomes 1 and 9 were most affected by the Xist transgene silencing in the TG male fibroblasts.

Given the high occurrence of down-regulated genes clustered in Chr 1 and Chr 9 of Tg2087/+ cells, we examined whether the Xic transgene is integrated in either of the two chromosomes in Tg2087. To test the association of the Xic transgene with a specific chromosome, two-color DNA FISH was performed on Tg2087/+ cells using a combination of chromosome painting and Xic sequence-specific probes. Since the cells were tetraploids, probes for the Xic locus showed up as two spots in wildtype males and four spots in transgenic males (Figure 2A). Using chromosome X (Chr X) paint as a positive control, the dual-color DNA FISH showed the co-localization of two Chr X domains with two Xic sequence (Xic) spots in both WT and TG male cells: two additional Xic spots separated from Chr X were clearly present only in the TG cell (Figure 2B). By contrast, a negative control using chromosome Y (Chr Y) paint in combination with the Xic probes showed distinct localizations of two Chr Y domains in both WT and TG male cells: a total of four Xic spots were seen only present in the TG cell (Figure 2C). These control experiments validated the dual-color DNA FISH in not only detecting specific chromosomes but also determining their association with Xic in WT and TG cells.

To test whether the Xic transgene is associated with Chr 1 or Chr 9, we used chromosome paints specific for Chr 1 and Chr 9, which were each compatible with Xic probes for the dual-color DNA FISH on the Tg2087 TTF cells (Figures 2D,F). In addition, we used the chromosome paint specific for Chr 4 as a reference and control for any possible influence of chromosome size in resolving the DNA FISH signals combining Xic probes and chromosome painting (Figure 2E). Mouse chromosome 4 is 156.86 Mb, which represents the average size between mouse chromosome 1 (195.15 Mb) and mouse chromosome 9 (124.36 Mb) according to Mus musculus GRCm39 (Genome Reference Consortium Mouse Reference 39). The mouse chromosome paints (MetaSystems) specific for Chr 1, Chr 4, and Chr 9 were each compatible for hybridization together with the Xic probes in a dual-color DNA FISH process (Figures 2D–F). In WT cells, two Xic pinpoint signals corresponding to the two endogenous Xic loci on the two X chromosomes showed no obvious linkage to Chr 1, Chr 4, or Chr 9 chromosome paint signals. In TG cells, two out of the four copies of Xic pinpoint signals corresponded to the endogenous Xic and the other two Xic pinpoints corresponded to the transgenic Xic. As shown in Figure 2D, distinct associations between Xic pinpoints and chromosome paints were observed for two Xic signals and Chr 1 signals, consistent with the model illustrating Xic transgene in the autosome (Figure 2A). In comparison, associations between Xic pinpoints and the chromosome paints were much less obvious for Chr 4 and Chr 9 signals (Figures 2E,F), suggesting that the co-localization between Xic and chromosome paint signals is chromosome-specific and likely based on genetic linkage independently of the chromosome size.

An illustration of all co-localization patterns and quantitation of events observed are summarized in Figure 2G. The co-localization of three copies of Xic signals with chromosome paints observed in a few situations was likely due to transient proximity contacts between Xic and genetically unlinked chromosomal regions. Co-localization of two copies of Xic with chromosomes was observed in 77% of TG cells for Chr 1, very close to the co-localization observed for Xic with chromosome X in 79% of TG cells (p = 0.62). The distribution patterns observed thus suggest an association of Xic with Chr 1 very similar to the linkage of Xic with Chr X in the TG cells. As controls, in WT cells, the co-localization pattern of Xic with Chr 1 was the same as the pattern of Xic with chromosome Y (p = 1.000) and was significantly different from the pattern of Xic with chromosome X (p < 0.001). These data were consistent with the observation of endogenous Xic (on Chr X) separate from Chr 1 and Chr Y. Therefore, the strong co-localization of Xic with Chr 1 in TG cells indicates a linkage between the transgenic Xic and chromosome 1. By contrast, for Chr 4 and Chr 9, their co-localization with two copies of Xic appeared in only 4 and 27% of TG cells, respectively, both significantly less than the occurrence of co-localization between chromosome X and Xic (p < 0.001).

We noted that transient proximity contacts between a chromosome and the Xic sequence (endogenous or transgenic) would contribute to the observed colocalization events. In WT cells with the endogenous Xic, any co-localization between Xic and a non-X chromosome would indicate a transient proximity contact. As shown in Figure 2G, Xic co-localized with Chr Y in 11% of WT cells. In comparison, the frequency of such co-localization was very similar at 14% for Chr 1 (p = 1.00) and 7% for Chr 4 (p = 1.00) but appeared higher at 27% for Chr 9 (p = 0.14). Using Chr 4 as a reference for an autosome with its chromosome size ranked between Chr 1 and Chr 9, which both showed a higher frequency of co-localization with Xic in WT cells as compared to Chr 4, we noted that such co-localization events observed as expected for Chr 1 (one-tailed p = 0.35) but significantly more frequent for Chr 9 (one-tailed p = 0.03). These data suggest that transient proximity contacts occurred between the endogenous Xic and autosomes; while Chr 1 and Chr 4 exhibited similar frequencies at close distances with the endogenous Xic, Chr 9 had more frequent proximity contacts with Xic.

In TG cells with the Xic probes detecting both endogenous and the transgenic Xic, the co-localization between Xic and chromosome paint signals reflects both the genetic linkages and the transient proximity contacts. The co-localization frequency of Xic/Chr 1 in TG cells was comparable to that of Xic/Chr X, both of which were significantly higher than that of Xic/Chr 4 or Xic/Chr 9. Our overall observations thus support a genetic linkage between Xic and chromosome 1 equivalent to Xic and chromosome X. A high frequency of Xic/Chr 9 co-localizations observed in TG cells, as compared to that of Xic/Chr 4, was likely contributed by the frequent proximity contact between chromosome 9 and the Xic locus.

Gene expression analysis from RNA-seq showed two major clusters of down-regulated genes from Chr 1 in TG cells when compared with WT controls (Figure 3A). This suggested to us that the Xic transgene may be integrated proximally to one of these two clusters in Chr 1. To test this, we designed DNA FISH probes targeting the sequence segments (170–210 kb) within the two transcriptionally repressed clusters in Chr 1 (Figure 3B): Segments B and E are 800 kb apart within the 1.5 Mb Cluster 1; Segments A and C are 250 kb apart within the 1.1 Mb Cluster 2. The chromosomal distance between Cluster 1 and Cluster 2 is 60 Mb. Among the transcriptionally repressed genes in TG cells, the Insulin-like growth factor binding protein genes Igfbp2 and Igfbp5 are located in Segment B; the actin-binding Tensin protein gene Tns1 is in Segment E; the actin filament pointed end-capping protein Leiomodin 1 gene Lmod1 is in Segment A; the cardiac muscle troponin T (cTnT) protein gene Tnnt2 is in Segment C (Figure 3B). Among these five genes, Igfbp2 and Igfbp5 were most affected by the transgene, with an expression reduction of >95% in TG cells compared to WT controls; Lmod1 expression was reduced by 69%, with Tns1 expression reduced by 66% and Tnnt2 expression reduced by 58%, in TG cells (Figure 3C and Supplementary Table 1). By contrast, the DNA primase large subunit gene Prims2 is located 60 Mb distal of Cluster 1 with no expression change in the TG cells compared to WT controls. Similarly, the denticleless E3 ubiquitin protein ligase homolog gene Dtl is located 60 Mb distal of Cluster 2 and its expression was not affected by the transgene (Figure 3C).

We used the dual-color DNA-FISH to map the nuclear positions of the sequence segments relative to Xic inside the cell (Figure 4A). In the WT cell, two copies of Xic signals corresponded to the endogenous Xic sequence from two X chromosomes in the tetraploid male cell, together with four copies of the “Segment” signals from four Chr 1’s. In the TG cell, four copies of Xic signals were present, two from the endogenous genes and two from the transgenes, with four copies of the “Segment” sequence signals. As illustrated in Figure 4B, we measured the spatial distance between an Xic signal spot and a “Segment” signal spot within the three-dimensional nucleus, and recorded the two shortest straight-line distances among all pairwise measurements for each cell. In the WT control cell, the shortest distances between “Segment” and Xic represent the spatial proximity of the Chr 1 segment sequences to the endogenous Xic, when they become or are closely localized in the cell nucleus. In the TG cell, the shortest distances between “Segment” and Xic represent the spatial proximity of the Chr 1 segment to both the endogenous and the transgenic Xic. To compare the nuclear positions of four Chr 1 segments relative to Xic in the TG cell, we normalized the distance measurements in TG cells against the same measurements in the WT control cell. In doing so, we eliminated the bias due to any close transient associations between Chr 1 segment sequences and the endogenous Xic.

Analysis of the normalized measurements showed that Xic had the closest spatial proximity to Chr 1 Segment A in the TG cell (Figure 4C). The comparisons of distance distributions illustrated that the spatial proximity between Segment A and Xic was significantly closer than the spatial distance between Segment C and Xic, even though Segment A and Segment C are both within the same Cluster 1. In contrast, Segment E is located in Cluster 2 and separated from Segment C with a large chromosomal distance of >60 Mb in between, however, displayed comparable spatial distance to Xic in the TG cell as Segment C. We noted that Segment B located in Cluster 2 showed the most variable spatial distance from Xic, with a median value close to the value of Segment A from Xic, even though Segment B is >60 Mb chromosomal distance away from Segment A. These results illustrate a model that Xist expressed from the Xic transgene on Chr 1 is capable of accessing various locations on Chr 1 for silencing, and that Xist RNA induces autosomal gene silencing by searching in three dimensions for sequences of close proximity (Figure 5).