Identification of Resf1 as a potential metastasis susceptibility gene

Mouse outcrosses with various strains of inbred and wild-derived strains demonstrated that the inherited genome significantly affected tumor and metastatic burden outcomes [7]. One of the most significantly different tumorigenic burdens among the inbred mice used in the outcross was MOLF/EiJ, a Japanese wild-derived strain (Fig 1A). Compared with FVB/NJ control mice, MOLF/EiJ mice developed tumors with significantly extended latency and exhibited reduced tumor growth and pulmonary metastases (Fig 1B) [8] [9]. To investigate the genetic basis of this observation, a MOLF/EiJ x FVB/MMTV-PyMT cross was performed followed by a backcross to FVB/NJ to segregate the genome during meiosis, resulting in N = 171 backcross (N₂) animals (Fig 1C). To identify regions of the MOLF/EiJ genome associated with changes in tumor phenotypes, these N₂ animals were genotyped using the Illumina Mouse Medium Density Linkage Panel. Logarithm of odds (LOD) scores displayed a significant correlation in all 3 tumor phenotypes (tumor burden, tumor latency, and metastatic burden) at the distal portion of chromosome 6 (Fig 1D). Thus, there may be metastasis modifier genes at distal chromosome 6 that affect all 3 tumor phenotypes based on the genetic association between the traits mapped to the locus.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mouse strains with differing metastatic capacity identify potential metastasis susceptibility genes on chromosome 6.

Fig 1 is adapted from Ha et al.[8]. (A) Kaplan-Meier survival plot of FVB/NJ x (MOLF/EiJ x MMTV-PyMT) mice. The p-value is based on a log rank test. Comparisons are of both MOLF/EiJ and FVB/NJ F1 mice after crossing with MMTV/PyMT. (B) Pulmonary surface metastases were counted after MMTV-PyMT cross with each strain. P-value is calculated by Mann-Whitney test. (C) A schematic of our breeding scheme to segregate chromosomes during meiosis. (D) Genetic association analysis based LOD scores (y-axis) observed for metastasis, tumor latency, and tumor burden, with the x-axis representing chromosomes lined up from head to tail. The upper horizontal dashed line represents a p-value significance threshold of 0.05 after permutation testing. (E) Zoomed-in region of distal chromosome 6 containing candidate genes for study. (F) GOBO analysis of the 12 gene signature identified from gene expression-tumor phenotype correlation analysis. DMFS (distant metastasis-free survival) plotted as a Kaplan-Meier survival curve shows a significant reduction in survival in ER+ breast cancer and (G) ER- breast cancer with lower expression. High (blue), intermediate (red), and low (gray) levels of gene expression-tumor phenotype correlation gene signature.

https://doi.org/10.1371/journal.pgen.1011236.g001

To identify candidate genes in this region of chromosome 6, gene expression-tumor phenotype correlation analysis was performed on 134 tumors from N₂ animals, as previously described [8]. Briefly, each of the 134 tumors was screened for genes at the peak of the distal chromosome 6 peak that was associated with distant metastasis-free survival (DMFS) in this population using Affymetrix ST 1.0 array analysis [8]. Using a significance threshold of p = 0.05, 10 genes associated with metastatic outcome were identified (Fig 1E).

To further validate the potential role of these 10 genes in human breast cancer, the candidate genes were used to generate a gene signature that could be applied to human breast cancer gene expression data. The human orthologs for each gene were identified and a human gene signature was generated by assigning weights and directionality based on the hazard ratios determined in the mouse Affymetrix data [8]. The GOBO (Gene Expression-Based Outcome for Breast Cancer Online) web tool [10] was used to stratify the disease outcome of these 10 genes in human breast cancer as well as determine their associations with DMFS. A significant association with DMFS was observed in ER- breast cancer (Fig 1F), but significance was not observed in ER-positive (ER+) breast cancer (Fig 1G). This is consistent with one or more previously studied genes within this signature playing a significant role in the progression of this highly malignant subtype of human breast cancer.

Examination of the location of the 10 candidate genes on the chromosome revealed that 5 genes (Arntl2, 2810474O19Rik/Resf1, 4833442J19Rik/Etfbkmt, Fgfr1op2, Slco1a5) were located at the peak of the LOD score for all 3 tumor phenotypes [8]. Examination of the Kaplan-Meier (KM) plots for DMFS in the mouse data revealed the greatest stratification for survival for Arntl2, Resf1, and Etfbkmt (S1 Fig), which were selected for further analysis. The validation and characterization of the role of Arntl2 in breast cancer metastasis was previously described by our group [8]. The potential role of Resf1 in breast cancer metastasis is discussed here.

Polymorphisms in the Resf1 promoter alter gene expression.

The gene expression-tumor phenotype correlation analysis indicated that variation in Resf1 expression might contribute to metastatic progression, consistent with the hypothesis that most inherited susceptibility is mediated by changes in gene expression rather than coding mutations [11,12]. Indeed, KM curves of mouse tumor expression data demonstrated that animals with lower Resf1 had worse survival (Fig 2A). Furthermore, RNA-sequencing (RNA-seq) analysis comparing FVB/NJ PyMT tumors versus MOLF/EiJ x FVB/NJ F1 tumors showed that the more highly metastatic homozygous FVB/NJ tumors had lower Resf1 expression (Fig 2B). The RNA expression data for human breast cancer available through the METABRIC consortium were consistent with this observation, as patients with ER- tumors and lower expression of RESF1 had worse outcomes than those with intermediate or high levels (Fig 2C). Furthermore, published DNase hypersensitivity (DHS) mapping data from the 3134 mouse mammary tumor cell line was mapped onto the Resf1 sequence and revealed two DHS sites, proximal to and overlapping the 5’ UTR, suggesting a potential promoter function in this region (Fig 2D). Examination of the sequence across this region revealed the presence of 28 single nucleotide polymorphisms (SNPs) between FVB/NJ and MOLF/EiJ within the DHS sites upstream of and overlapping the 5’ UTR (S2A Fig). Promoter reporter assays of the 5’ UTR containing both DHS sites revealed that the MOLF/EiJ allele had significantly higher transcriptional activity, confirming a functional consequence of the inherited promoter variants (Figs 2E and S2B). These results suggest that inherited variants in the transcriptional control elements of the Resf1 gene may drive differential expression from the FVB/NJ and MOLF/EiJ alleles that alter metastatic capacity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. RESF1 expression levels corroborate human data and are associated with upstream SNPs.

(A) From the N₂ backcross cohort in Fig 1C, 131 mice were categorized into Resf1 over and under-expressed groups and plotted as Kaplan-Meier survival curves showing worse DMFS survival in animals under-expressing Resf1. (B) RNA-seq of paired spontaneous mammary tumors and lung metastases from FVB/NJ and MOLF/EiJ mice that were crossed with MMTV-PyMT were analyzed for Resf1 expression levels and shows higher expression levels in the less metastatic MOLF/EiJ strain. (C) Query of the METABRIC human breast cancer tumor database shows significant reduction in survival at lower expression levels of RESF1. High (red), intermediate (gray), low (blue). (D) Query of the UCSC BLAT Genome Browser for the 5’ UTR and upstream region of Resf1 displays DHS peaks (green) in the highlighted yellow area. Included are locations of primers used for PCR and cloning of the promoter enhancer region. (E) Upstream regions of Resf1 in FVB/NJ and MOLF/EiJ were cloned into the luciferase pGL4.23 reporter plasmid, and dually transfected with the Renilla hRluc plasmid into HEK293T cells, with empty vector as a negative control. Values show ratio of Firefly luciferase to Renilla in empty vector control, FVB/NJ upstream region, and MOLF/EiJ upstream region. P-value based on unpaired t-test.

https://doi.org/10.1371/journal.pgen.1011236.g002

Decreased Resf1 expression is associated with increased autochthonous metastasis

To validate Resf1 as a metastasis modifier and to determine its disease phenotype in vivo, a Resf1 genetically engineered mouse model (GEMM) generated by an enhancer/gene trap insertion was used. The C57BL/6-Et(EGFP/Cre)^16255Rdav mouse model (hereafter, the gene trap model) was obtained from the Mutant Mouse Resource & Research Centers. This model was generated by lentiviral gene trap construct integration into the first exon of Resf1, resulting in the expression of an EGFP-cre fusion and the truncation of the Resf1 message (www.mmrrc.org/catalog/sds.php?mmrrc_id=34574). qRT-PCR analysis of mammary tumor tissue from wildtype and gene trap x FVB/MMTV-PyMT littermates confirmed a reduction of Resf1 mRNA (Fig 3A). Tumor weight and lung metastases of the gene trap x PyMT F1 mice were collected and assessed at the humane endpoint. The Resf1 gene trap mice developed significantly larger primary tumors and significantly more lung metastases than wildtype littermate control mice, even when normalizing metastasis counts to account for the larger tumors (Fig 3B–3D). Furthermore, the incidence of metastases in the gene trap mice was significantly higher than in the control mice (67% vs. 36%, p = 0.05, Fisher’s exact test; Fig 3E), consistent with the poor survival observed in human patients with lower RESF1 expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Lower stromal levels of Resf1 increase metastatic burden and incidence.

(A) qRT-PCR analysis of Resf1 in primary tumors from WT and genetrap hypomorph crossed with MMTV/PyMT mice (n = 5 per group). Control and hypomorph mice were crossed with MMTV-PyMT mice to induce spontaneous mammary tumors and pulmonary metastases (B-E). (B) Primary mammary fatpad tumors were collected and weighed for WT (n = 27) and hypomorph (n = 38) and resulted in significantly larger tumors in hypomorph mice, p-value calculated by Mann-Whitney test. (C) Surface lung metastases were counted, resulting in more metastases in hypomorph mice, p-value calculated by Mann-Whitney test. (D) Normalization of lung metastases per gram tumor to account for larger tumor size remained significantly higher in hypomorph mice, p-value calculated by Mann-Whitney test. (E) Metastatic incidence was higher in hypomorph mice compared to control.

https://doi.org/10.1371/journal.pgen.1011236.g003

Decreased Resf1 expression in cell lines paradoxically reduces the metastatic capacity

The role of RESF1 in metastatic disease was further explored by using orthotopic transplants for spontaneous metastasis assays. We attempted to generate cell lines with stable Resf1 overexpression but were unsuccessful, suggesting that there is a threshold beyond which Resf1 expression negatively impacts cell viability. Therefore, subsequent efforts focused on shRNA-mediated knockdown (KD) of Resf1.

Resf1 was knocked down via stable shRNA transduction of 4 constructs into 6DT1 and Mvt1, both mouse cell lines syngeneic to FVB/NJ [13] (Fig 4A and 4E). Resf1 KD was confirmed by qRT-PCR (Fig 4A and 4E). The H4 and H6 shRNAs displayed consistent KD in both cell lines and were used for further in vivo experiments. shScramble control and Resf1 shRNA H4 and H6 cells were injected orthotopically into the 4th mammary fat pad of syngeneic FVB/NJ mice. Primary mammary tumors and lung metastases were collected and assessed after 28 days. Inconsistent primary tumor growth changes upon shRNA-mediated KD were observed for the 6DT1 and Mvt1 cell lines, indicating cell line-specific effects on tumor cell growth (Fig 4B and 4F). Surprisingly, both KD cell lines showed a reduction in lung metastases compared to the control (Fig 4C and 4G), in contrast to our expectations based on the human data and mouse gene trap x PyMT model, which formed more lung metastases (Fig 3B–3D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Decreased Resf1 in cell lines paradoxically decreases metastatic capacity.

(A) qRT-PCR analysis of 6DT1 shRNA-mediated Resf1 stable KD cells. (B) Weight of primary tumors from 6DT1 Control (scramble), H4, and H6 cells orthotopically injected into the 4^th mammary fatpad of syngeneic FVB/NJ mice, n = 15 mice per group. (C) Surface pulmonary metastasis in mice from (B). (D) Pulmonary metastases normalized per gram tumor from (B). (E) qRT-PCR analysis of Mvt1 shRNA-mediated Resf1 stable KD cells. (F) Weight of primary tumors from Mvt1 Control (scramble), H4, and H6 cells orthotopically injected into the 4^th mammary fatpad of syngeneic FVB/NJ mice, n = 10 mice per group. (G) Surface pulmonary metastasis in mice from (F). (H) Pulmonary metastases normalized per gram tumor from (F). (I) qRT-PCR analysis of MDA-MB-231 shRNA-mediated RESF1 stable KD cells. (J) Weight of primary tumors from MDA-MB-231 Control (scramble) and G5 KD cells orthotopically injected into the 4th mammary fatpad of nu/nu mice, n = 10 mice per group. (K) Surface pulmonary metastasis in mice from (J). (L) Pulmonary metastases normalized per gram tumor from (J). P-values in B-D and F-H calculated by ordinary one-way ANOVA with Dunnett’s multiple comparison test. P-values in J-L calculated by Mann-Whitney test.

https://doi.org/10.1371/journal.pgen.1011236.g004

Although previous studies have demonstrated that the MOLF/EiJ x PyMT backcross population approximates the diversity of human breast cancer subtypes at the transcriptional level [8] [14,15], the MMTV-Myc tumors from which the Mvt1 and 6DT1 cell lines derived are thought to most closely resemble the human luminal subclasses [16]. To determine whether the opposite metastatic capacity observed in these cell lines was due to either a species-specific effect or a subtype-specific effect, RESF1 was stably knocked down (Fig 4I) in the highly malignant triple-negative MDA-MB-231 human cell line and orthotopically implanted into immunocompromised NU/J mice. KD of RESF1 did not significantly alter primary tumor growth and significantly reduced lung metastasis, which is consistent with the findings from our 6DT1 and Mvt1 allografts (Fig 4J–4K). The difference between the cell line data and human patient or endogenous mouse data may therefore be the result of an artifact introduced into cells during the establishment of ex vivo cell lines. Alternatively, Resf1 may have opposing functions at the primary and secondary site, similar to Tgfβ which has previously been shown to have tumor suppressive functions at the primary site but pro-tumorigeneic effects for tumor invasion and metatasis [17].

To further address this question and to control for potential off-target effects in the gene trap mouse model, a second Resf1 GEMM from the KOMP project (C57BL/6NCrl-2810474O19Rik^{em1(IMPC)Mbp/Mmucd}; hereafter referred to as the Resf1 knockout (KO) mouse) was incorporated into the study. This GEMM carries a CRISPR-Cas9 mediated deletion of exon 4, which contains almost the entirety of the Resf1 coding sequence, and results in a 50% reduction of Resf1 mRNA levels. The Resf1 KO mouse was bred to the PyMT model to compare the tumor phenotypes of PyMT⁺/Resf1^wt/wt and PyMT⁺/Resf1^wt/KO compound heterozygous animals. Consistent with the gene trap model, the reduction of Resf1 by CRISPR-mediated deletion was associated with increased tumor burden, more rapid tumor growth, and an increase in the number of pulmonary metastases (S3A–S3D Fig).

To test for a potential role for Resf1 in the tumor stroma, orthotopic transplants of 6DT1 cells into the mammary fat pads of (FVB/NJ x Resf1 KO) F₁ or wildtype FVB littermate animals were performed. However, no differences were observed between the two genotypes, suggesting Resf1 functions in a tumor-autonomous fashion (S3E–S3H Fig). Moreover, genotyping analysis of matched tail and primary tumor samples from PyMT⁺/Resf1^wt/wt and PyMT⁺/Resf1^wt/KO F₁ mice showed the loss of the wildtype allele in many of the animals (S3I Fig), supporting the role of Resf1 as a tumor suppressor. This is consistent with previous studies that suggest that Resf1 is a rare tumor suppressor gene [18]. In contrast, RNA-seq analysis of matched FVB-PyMT tumors and metastases demonstrated higher Resf1 expression in metastases, which suggests a potential pro-metastatic role at the secondary site (S3J Fig). Combined, these data suggest a complex dual role of Resf1 in tumor progression analogous to TGFβ, with functions as a tumor suppressor in the primary tumor but metastatic promoting abilities in the secondary site.

RESF1 is a nucleoplasmic protein

The Human Protein Atlas revealed RESF1’s primary localization in nucleoli, aligning with its potential role in ribosomal biogenesis. Our immunofluorescence analysis using the same antibodies confirmed this localization (S4A and S4B Fig). However, when epitope-tagged mouse or human RESF1 was transiently transfected, Resf1 was excluded from nucleoli and instead localized to the nucleoplasm (S4C Fig), calling into question the specificity of the antibodies used for immunofluorescence staining. To address this, we generated rabbit monoclonal antibodies against the mouse RESF1 ortholog, validated by western blot in transiently transfected cells and Resf1 CRISPR-KO mouse cell lines (S4D–S4E Fig). In contrast to human antibodies (15622 and 13816), the recombinant anti-mouse antibodies (15A4 and 186C7) stained nuclear speckles (S4F Fig). However, none of these antibodies, including the human 13816 antibody, exhibited signal loss CRISPR-KO cell lines, suggesting off-target reactivity in immunofluorescence conditions.

To resolve the nuclear localization of RESF1, subcellular fractionation was performed. Fractionation of 4T1 and 6DT1 mouse cancer cells revealed that endogenous RESF1 was predominantly located in the nuclear fraction (S4G Fig). Next, we used CRISPR to add dTAG-2xHA to the C-terminus of endogenous RESF1. Immunofluorescence analysis of the HA tag showed RESF1-dTAG-2xHA staining in the nucleoplasm (S4H Fig). Nucleolar and nuclear speckle staining was also observed in a subset of cells, suggesting that RESF1 can occupy these structures. This also raises the possibility that the anti-RESF1 antibodies may be recognizing a RESF1-associated structure in addition to the RESF1 protein itself, though further work would be required to address this possibility.

RESF1 is associated with compound G4 quadruplexes

To further dissect the role of RESF1 in the nucleus, chromatin association analysis was performed. Fortuitously, a search of the Sequence Read Archive (SRA) public database revealed a proximity labeling BioTAP XL [19] study of the human ortholog of RESF1 deposited by the Elledge laboratory (PRJNA509912). This experiment consisted of a tetracycline-inducible RESF1 construct fused to an endogenous biotinylation targeting signal to biotinylate and isolate RESF1 chromatin. Mapping this data against the T2Tv2 human genome revealed 15882 sites significantly associated with RESF1 after an FDR cut-off of 0.05. Examination of the sites revealed significant associations of RESF1 across highly expressed 7S RNAs (Fig 5A), exons of protein-coding genes (Fig 5B), and ribosomal RNA repeats (Fig 5C). The BioTAP XL data was then remapped on hg38 and analyzed with the GREAT tool [20]. 14,929 BioTAP XL sites were mapped to the hg38 genome and were associated with 61% (11,546/18,777) of the annotated genes in this genome build. The majority of the BioTAP XL sites were within 0–5 kb downstream of the transcription start site (TSS), suggesting a potential association with transcribed sequences (Fig 5D). Gene Ontology analysis revealed a significant association of RESF1 with genes involved with translation, noncoding RNA, and ribosomal biogenesis (S5A and S5B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. RESF1 is associated with many regions across the genome.

BioTAP XL data was mapped against the T2Tv2 human genome and showed significant associations between Resf1 and (A) 7S RNAs, (B) exons of protein-coding genes, and (C) ribosomal RNA repeats. (D) Histogram graph showing the majority of the BioTAP XL sites within 0–5 kb downstream of the TSS. (E) Comparisons of the top 1000 RESF1 BioTAP XL binding sites versus the bottom 1000 sites revealed a significant association between RESF1 and GC-enriched sequences. (F) Identification of the de novo motif with the potential to form G4s within the BioTAP XL sites.

https://doi.org/10.1371/journal.pgen.1011236.g005

Further examination of some of the most significantly associated RESF1 sites revealed highly GC-enriched sequences. Comparisons of the top 1000 versus the bottom 1000 demonstrated that the most significant RESF1-associated sites had approximately 70% GC content and 70.3% (11179/15822) overlapped CpG islands in the T2Tv2 genome (Fig 5E). De novo motif detection identified GGVGGCNGVGGHDGS (Fig 5F) as an enriched potential motif within the BioTAP XL sites. This motif resembled sequences that have the potential to form DNA G4 quadruplexes, a nucleic secondary structure formed by stacking two or more Hoogsteen base-paired guanine tetrads to form a DNA “knot”. DNA G4 quadruplexes have been implicated in a number of nuclear activities, including transcriptional control, DNA stability, and RNA splicing [21,22,23,24]. DNA G4 predictions were therefore calculated for the T2Tv2 transcriptome using the R package pqsfinder and compared to the BioTAP XL data. 93.4% of the BioTAP XL-associated sites (14842/15882) overlapped predicted DNA G4 quadruplexes in the transcriptome, supporting the possibility that RESF1 is primarily associated with quadruplexes in transcribed sequences rather than DNA G4 quadruplexes associated with transcriptional control elements.

To better understand the association of RESF1 with transcribed sequences, the ribosomal RPL27 gene was selected for further study. RPL27 showed a strong RESF1-associated peak over exon 3 (Fig 6A). Moreover, the promoter and 5’ UTR of RPL27 is associated with DNA G4 quadruplex as well as iMotif structures [25] but is not associated with RESF1, consistent with a non-promoter/enhancer role for RESF1 in transcription. Exon 3 of RPL27 was reanalyzed for potential G4 quadruplexes spanning up to 45 base pairs with the web tool QGRS Mapper. QGRS Mapper predicted tandem DNA G4 quadruplexes on the template strand of exon 3 and a single predicted quadruplex on the non-template strand (Fig 6A). Other RPL27 exons contained only a single predicted DNA G4 quadruplex on the non-template strand. Electrophoretic gel shift assays of an oligo spanning this predicted DNA G4 were not consistent with the formation of a quadruplex in vitro. However, DNA secondary structure prediction indicated this sequence had a strong predilection to form a DNA hairpin, which was supported by circular dichroism analysis. Previous studies mapping G4 quadruplexes in cytoplasmic RNA (SRA accession number PRJNA673726) [26] did reveal the presence of a RNA G4 quadruplex in exon 3 of the RPL27 mRNA, suggesting the potential for quadruplex formation in more physiological conditions (S5C Fig). Circular dichroism and DNA G4 gel shift assays using oligos spanning each of the predicted exon 3 quadruplex sequences on the template strand demonstrated a shift in the presence of potassium ions, but not sodium or lithium, consistent with the formation of DNA G4 quadruplexes (Fig 6B–6D). Together, these data suggest that RESF1 may be associated with transcribed sequences containing one or more DNA G4 quadruplexes on both strands.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. RESF1 is associated with G4 quadruplexes on the template and non-template strands.

(A) BioTAP XL analysis showing a significant association between RESF1 and RPL27 exon 3 containing G4s on both strands. (B) Gel mobility shift assays with RPL27 compound G4s (G4.1 and G4.2) on the template strand. (C-D) Circular dichroism analysis of RPL27 compound G4s on the template strand with the presence of lithium, sodium, or potassium ions.

https://doi.org/10.1371/journal.pgen.1011236.g006

To further examine this possibility, DNA G4 quadruplex analysis was performed on the top 1000 RESF1 binding sites with QGRS Mapper. 96.6% (966/1000) of the RESF1-associated sites contained QGRS Mapper-predicted quadruplexes on both DNA strands, with 95.4% of sites containing 2 or more predicted quadruplexes on at least one strand. The DNA G4 quadruplexes predicted were highly enriched for quadruplexes containing only two guanine tetrads, which are thought to be less stable than quadruplexes with 3 or more tetrads. Combined, these data suggest that RESF1 is associated predominately with transcribed sequences containing one or more potential duplex DNA G4 quadruplex structures on both DNA strands.

RESF1 metastatic phenotype is not associated with global alterations in translation

RESF1 has been previously implicated in mRNA transport and protein production [27]. To determine if this activity accounted for the Resf1 metastasis phenotype, Resf1 CRISPR knockouts (KO) were generated in the 4T1 and 6DT1 cell lines (S4E Fig). O-propargyl-puromycin (OPP) assays were performed to examine potential global changes in protein synthesis. However, in contrast with published results, no difference in OPP signal between 4T1 or 6DT1 Resf1 KO cells compared to the control was observed (Fig 7A). To assess nascent RNA production, an ethynyl uridine (EU) pulse assay to label actively transcribed RNAs was performed. In contrast to Ritter et al. [27], no increase in the labeling of cytoplasmic RNA was observed in Resf1 KO or KD cells (Figs 7B and S6). PolyA FISH analysis also showed no significant changes between control and Resf1 KO cell lines, suggesting the loss of RESF1 did not alter global mRNA production or transport (Fig 7C). Furthermore, mTOR antibody arrays to investigate alterations in translational efficiency (S7 Fig) and western blot analysis of critical canonical unfolded protein response (UPR) targets showed no significant changes between control and KD cell lines grown on glass or plastic culture surface with and without prior tunicamycin (TM) treatment for UPR activation (S8 Fig). Thus, RESF1 depletion does not appear to alter global protein translation in metastatic cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Resf1 is not associated with its known functions in metastasis.

Confocal fluorescence microscopy images of (A) a 30-minute OPP (o-propargyl-puromycin) pulse, (B) a 30-minute EU (ethinyl uridine) pulse, and (C) PolyA mRNA FISH, (D) 5.8S, 18S, and 28S rRNA FISH of 4T1 CRISPR KO cell lines. (E) Western blot analysis showing shRNA-mediated KDs of Resf1 in 6DT1, Met1, and MDA-MB-231 cell lines. (F) Tumorsphere assays with 4T1 and 6DT1 CRISPR KO cells, 6DT1, Met1, and MDA-MB-231 KD cells. P-values were calculated by ordinary one-way ANOVA with Dunnett’s multiple comparison test.

https://doi.org/10.1371/journal.pgen.1011236.g007

To determine whether ribosomal RNA is altered in Resf1 KO cell lines, rRNA subunit RNA FISH was performed. We observed no significant differences in 5.8S, 18S, or 28S levels between Resf1 control and KO cell lines (Fig 7D). Moreover, GSEA analysis of cell line RNA-seq data indicated significant suppression of ribosomal subunit transcripts in the KD cell lines (S9A–S9C Fig). Western blot analysis, however, revealed no significant difference in several ribosomal proteins except for a decrease in the RPL22 subunit protein in Resf1 shRNA KD cells (S9D Fig). However, metastasis assays revealed that reduced RPL22 in 6DT1 cells did not significantly alter metastasis (S9E–S9H Fig). Consistent with ribosomal RNA FISH data, KD of Ubf (S10C Fig), a transcription factor for 45S ribosomal precursor RNA, did not significantly reduce the expression of all 4 rRNA subunits (S10A Fig). Ubf KD also did not change the ability of tumor cells to form lung metastases (S10D–S10F Fig). Taken together, these results suggest that the Resf1-mediated metastasis phenotype is not associated with large alterations in ribosomal biogenesis or translational efficiency.

RESF1 metastatic phenotype is not associated with epigenetic silencing of endogenous retroviruses

RESF1 was recently found to interact with the histone lysine methyltransferase SETDB1 to epigenetically repress endogenous retroviral elements in embryonic stem cells [28]. To determine whether suppression of Resf1 altered the expression of endogenous retroviral elements in breast cancer cell lines, qRT-PCR analysis was performed. No differences in the expression of any of the endogenous elements were observed (S11A Fig). Furthermore, global levels of H3K9me3 were unchanged (S11B Fig). To more directly test whether the RESF1 metastatic phenotype was mediated by SETDB1, metastasis assays were performed with Setdb1 shRNA KD 6DT1 cells (S11C Fig). Although a significant reduction of metastasis was observed in the sh975 cell lines (S11D Fig), this result was not observed for the other two shRNAs utilized (S11E Fig). Together, these data suggest that the metastatic phenotype observed upon Resf1 KD is not related to the interaction of RESF1 with SETDB1 or the reactivation of endogenous retroviruses.

RESF1 depletion suppresses epithelial-to-mesenchymal transcriptional programs

RESF1 has also been implicated in mouse embryonic stem cell self-renewal and germ cell specification [29]. Cellular plasticity and stem cells are linked to epithelial-to-mesenchymal transition (EMT) [30], which is a process implicated in metastasis where cells lose expression of epithelial markers and gain mesenchymal and stem-like phenotypes. To investigate the potential role of cellular plasticity, tumor sphere analysis, which is thought to measure stem-like capacity in cell lines, was performed. However, growth in 3D culture was not significantly and consistently different for both mouse and human breast cancer KD and KO cell lines (4T1, 6DT1, Met1, and MDA-MB-231; Fig 7E and 7F), arguing against an increase in stem-like properties. However, given that opposite results were observed between the allograft and autochthonous GEM model metastasis assays and that previous work from our laboratory has demonstrated that tumor cells rapidly acquire permanent mesenchymal-like transcriptional programs after in vitro culture [31], GSEA analysis of the autochthonous primary tumors was examined. In contrast to the in vitro models, GSEA analysis of both the KO x PyMT and Genetrap x PyMT tumors demonstrated a highly significant suppression (FDR < 0.001) of EMT-associated transcriptional programs in the autochthonous setting (Fig 8A and 8B). Mesenchymal breast cancer cells are thought to be non-proliferative. Thus, the in vivo suppression of the EMT transcriptional program in the Resf1 depleted autochthonous tumors is consistent with the increased tumor growth and decreased tumor latency. The precise mechanism for this phenomenon, however, has yet to be determined.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Resf1 depletion enhances the expression of Myc targets.

(A-D) GSEA analysis snapshots of MYC-target pathways that are upregulated in 6DT1 shRNA KD, 4T1 CRISPR KO, HEK293T siRNA, and MDA-MB-231 shRNA KD. (E) BioTAP XL analysis of the association between RESF1 and MYC exons.

https://doi.org/10.1371/journal.pgen.1011236.g008

In summary, here we present evidence that RESF1 is a chromatin-associated nucleoplasmic protein that functions as a tumor suppressor in ER- breast cancers. Depletion of RESF1 results in increased tumor growth and metastasis, potentially through suppression of the EMT pathway. RESF1 is enriched on exons of actively transcribed genes that contain 1 or more predicted DNA G4 quadruplex on both the template and non-template strand, implying an important role in transcriptional elongation. However, the precise mechanism by which RESF1 alters metastatic potential remains unclear at present. Further investigations will be required to elucidate the function of this large, unstructured protein and its role in cancer.

Ethics statement

All animal studies were performed under Animal Study Protocols LPG-002 and LCBG-004, approved by the Bethesda NCI Animal Care and Use Committee.

MOLF/EiJ backcross and identification of candidate genes

The FVB/NJ (RRID:IMSR_JAX:001800) x [MOLF/EiJ (RRID:MGI:3487124) x MMTV-PyMT/FVB (RRID:MGI:3032640)] cross resulted in 171 N2 backcross animals as previously described in [55]. N2 backcross animals were genotyped using the Center for Inherited Disease Research (www.cidr.jhmi.edu). QTL analysis mapping was conducted by utilizing the R/QTL software with the J/QTL interface [56]. QTL peaks were deemed significant when their p-values fell below 0.05 following correction for genome-wide significance through 10,000 permutations [8]. To identify potential candidate metastasis susceptibility genes, Affymetrix analysis was performed on tumors from the MOLF backcross as previously described [8] and candidate genes prioritized for analysis based on proximity to the LOD score peak.

Cell lines and culture methods

Mouse and human cell lines were cultured for several experiments. Mouse 6DT1 (RRID:CVCL_C8PZ), 4T1 (RRID:CVCL_0125), and Mvt1 (RRID:CVCL_C8Q2), MDA-MB-231 (RRID:CVCL_0062) cells (generously provided by Dr. Lalage Wakefield, NCI) and human HEK293T (RRID:CVCL_0045) cells were cultured in full DMEM (Gibco CAT#: 10313039, 10% FBS, 1% penicillin-streptomycin (P/S), 1% glutamine). Cells were incubated at 37°C with 5% CO₂ in tissue culture flasks or dishes. Stable KD cell lines were cultured with various selection antibiotics described below.

Virus production for lentiviral stable knockdown

TLC lentiviral shRNA constructs against Resf1 were obtained from Dharmacon. shRNAs targeting the mouse Resf1 gene, sh-Resf1 H4 (5’-TTTCGATTTGTAACATAGGTC-3’) and sh-Resf1 H6 (5’-AAACAACCAAGTTACCTCAGG-3’) were used for Resf1 KD in mouse cell lines. For RESF1 KD in human cell lines, shRNAs targeting the human RESF1 gene, sh-RESF1 565 (5’-GAATACAAACTCTGTGGAAGA-3’) and sh-RESF1 G5 (5’-GATGGATTTGAGATGCTACAA-3’) were used.

For lentivirus production, 1 μg of shRNA-containing plasmids and 1 μg of viral packaging plasmids, pMD2.G (250 ng) (Addgene plasmid #12259) and psPAX2 (750 ng) (Addgene plasmid #12260) (both were obtained as a gift from the Trono lab), were transfected into the human 293T cell line (Thermo Fisher Scientific) using X-tremeGene 9 DNA transfection reagent (Roche). 48 hours post-transfection, virus-containing supernatant was passed through a 45 μm filter to obtain viral particles, which were transferred to human or mouse breast cancer cell lines. 24 hours post-transduction, the viral media was replaced with fresh 10% DMEM. Finally, after 24 hours, the cells were selected with 10 μg/ml puromycin (Sigma) or 5 μg/ml blasticidin (Gibco) in complete DMEM.

Orthotopic mouse injections

Female virgin FVB/NJ mice were obtained from The Jackson Laboratory at ~6 weeks of age and housed in the NCI Animal Facility for 2 weeks to acclimate before injections at ~8–9 weeks of age. Virgin female athymic NU/J (RRID:IMSR_JAX:002019) mice were obtained from The Jackson Laboratory at the same age but were not injected until ~18 weeks of age.

For FVB/NJ mice, 6DT1 and Mvt1 cells were plated in P/S and selective antibiotic-free media two days before injections. For NU/J mice, MDA-MB-231 cells were plated in the same manner. On the second day, cells were lifted with 0.05% trypsin, washed 3 times in PBS, counted, and resuspended in room temperature PBS to a concentration of 1 x 10⁵ cells per 100 μl. For spontaneous metastasis assays, 100 μl of solution containing 1 x 10⁵ cells were surgically injected into the 4th mammary fat pad of the mice. Mice were euthanized after 28–30 days using an intraperitoneal administration of Avertin followed by cervical dislocation. Primary tumors were removed and weighed, lungs were removed, and surface metastases were counted then placed in formalin for 24 hours followed by 70% ethanol. All singular metastasis assay dissections were conducted by a single investigator after being blinded.

FVB/NJ x C57BL/6J Resf1 knockout mice

For semi-tumor non-autonomous investigation of Resf1, male Resf1 heterozygous C57BL/6J (C57BL/6NCrl-2810474O19Rik^{em1(IMPC)Mbp/Mmucd}: RRID:MMRRC_048846-UCD) mice carrying a CRISPR-Cas9 mediated deletion of exon 4 were bred with 11-week-old WT female FVB/NJ mice. To genotype the F1 generation, we followed the protocol provided by the Knockout Mouse Project at UC Davis, where the C57BL/6J Resf1 KO parental mice were originally obtained. Spontaneous metastasis assays were performed on heterozygous F1 females as previously described. Primary tumors and lungs were removed and analyzed as previously described.

For the gene trap hypomorph mouse experiment, the mouse strain used for this research project B6(129S)-Et(EGFP/cre)^{16255Rdav/Mmmh}, RRID:MMRRC_034574-MU, was obtained from the Mutant Mouse Resource and Research Center (MMRRC) at University of Missouri, an NIH-funded strain repository, and was donated to the MMRRC by Ronald L. Davis, Ph.D., The Scripps Research Institute and Paul Overbeek, Ph.D., Baylor College of Medicine.

MMTV-PyMT/FVB x C57BL/6J Resf1 knockout mice

For a true tumor non-autonomous investigation of Resf1, female Resf1 heterozygous mice were crossed with male MMTV-PyMT/FVB mice. All mice in the F1 generation were genotyped for both the PyMT transgene and Resf1 zygosity. Females that were WT and heterozygous for Resf1 and containing the PyMT transgene were maintained until 120 days when they were euthanized. Dissections were conducted by a single investigator after being blinded to the genotype of the mice. Primary tumors were weighed, and surface pulmonary metastases were counted.

CRISPR knockout cell lines

Single-guided RNA (sgRNA) against the Resf1 coding region was designed using the CRISPick program [57] (https://portals.broadinstitute.org/gppx/crispick/public). Three sgRNAs targeting the different regions of the largest exon (exon 3) were selected: sgRNA1 5’-TATCTCATAATCCCGACGGA-3’, sgRNA 2 5’-TCGGGAAACTGATATGTTCA-3’, and sgRNA 3 5’-TGTGACAAGTTGGCGGAACC-3’. sgRNAs were annealed and ligated into the LentiCRISPRV2-GFP vector (obtained as a gift from Dr. Ji Luo Lab, NCI). To create stable CRISPR KO cell lines, 1 x 10⁶ 293T cells were plated in 6 cm dishes in P/S-free 10% FBS DMEM media 24 hours before transfection. Cells were transfected with 1 μg of sgRNA and 1 μg of viral packaging plasmids (250 ng pMD2.G and 750 ng psPAX2), as described above, using 6 μl of Xtreme Gene 9 transfection reagent (Roche). After 24 hours of transfection, media was replaced with fresh 10% DMEM, supplemented with 1% P/S and 1% Glutamine, for another 24 hours. Virus-containing supernatant was then passed through a 45 μm filter to obtain viral particles, which were transferred to 50,000 4T1 and 6DT1 cells. 24 hours post-transduction, the viral media was replaced with fresh 10% DMEM. Heterogeneous 4T1 and 6DT1 KO cells were FACS-sorted by GFP fluorescence. A total of 1000 cells were sorted and single cell clones were manually plated into 96-well plates. Clones were allowed to grow until they reached confluence. CRISPR KO of each clone was confirmed by genomic DNA PCR and Sanger sequencing. KO was further validated by western blot using a custom antibody generated against mouse RESF1 protein (GeneScript) at 1:000 dilution.

RESF1-dTAG-HA construct

RESF1-dTAG knockin vector was generated using the previously described protocol[58]. C-terminal targeting guide RNA was designed using the CRISPick webpage, and sgRNA oligonucleotides were obtained from IDT, annealed, and ligated in the BbsI restriction site of pX330A-1×2 (Addgene, #58766) plasmid to create the pX330A-1x2-cRESF1 vector. The PITCh sgRNA from the pX330S-2-PITCh (Addgene #63670) vector was then inserted into the pX330A-1x2-cRESF1 vector using Golden Gate Assembly (New England Biolab). To generate the donor vector containing dTAG-2xHA sequences, pCRIS-PITChv2-dTAG-Puro (RESF1) (Addgene, #91796) vector was utilized, serving as a template for PCR amplification of the dTAG-2xHA sequence. The PCR product, which included homology arms flanking the sgRNA region and replaced the RESF1 homology arm, was Gibson cloned (NEBuilder Hifi DNA assembly kit, NEB) into the same vector after digestion with MluI restriction enzyme.

To generate the RESF1 dTAG knockin cell line, human 293FT cells were grown on 10cm dishes and transfected with sgRNA-containing vector (pX330A-1x2-cRESF1/PITCh) and donor vector using Xtremegene 9 transfection reagent (Roche). After 72 hours, positive clones with the c-terminal RESF1 dTAG knockin were selected by treating with 1 μg/ml puromycin for 7 days, followed by single-cell cloning. PCR genotyping confirmed the presence of the RESF1 knockin.

RNA isolation, qRT-PCR, and RNA-seq

For qRT-PCR, cells were seeded at equal density on tissue culture dishes 24 hours before RNA isolation. After 24 hours, media was aspirated and TriPure isolation reagent (Roche) was added to the dish to collect cells before purification. The concentration of pure RNA was measured using the DeNovix DS-11 Spectrophotometer before being reverse transcribed at equal concentration using iScript (BioRad), following the manufacturer’s recommended protocol. Quantitative real-time PCR (qRT-PCR) was performed using Veriquest SYBR green PCR master mix (Applied Biosystems). mRNA expression was measured by the threshold cycle. Normalization of target mRNA counts was normalized to Peptidylprolyl isomerase (Ppib) for both human and mouse. The Ct of the normalization gene was subtracted from the Ct of the target genes (ΔCt). Expression levels of the target genes were calculated using the equation: Expression = 2^-(ΔCt)*1000 and compared to the control. Controls were normalized to 1 and experimental targets were relatively compared. Primer sequences are provided in the supplementary information (S1 Table).

For RNA-seq, cells were seeded at equal density in triplicate per condition. After 24 hours, media was aspirated and TriPure isolation reagent was added to the dish to collect cells before centrifugation to separate RNA and organic phases. After separation, the RNA phase was transferred to a new tube and isopropanol was added to the RNA to precipitate. The RNA isopropanol precipitation sample mixture was added to a Qiagen RNeasy Spin column for purification using the Qiagen RNeasy kit. Samples were analyzed by the Agilent 2200 TapeStation electrophoresis system. All samples used for RNA-seq had an RNA integrity number (RIN) score greater than 7 and were sent to the Illumina Sequencing Core at the Frederick National Labs for sequencing.

Protein extraction

For protein isolation, cells were seeded at equal densities on tissue culture dishes. After 24–48 hours, media was aspirated from the dish and cells were washed once with PBS. For cell lysis, Golden Lysis Buffer (20 mM Tris pH 8.0, 400 mM NaCl, 5 mM EDTA, 1 mM EGTA, 10 mM NaF, 1 mM Na pyrophosphate, 1% Triton X-100, 10% glycerol, Complete EDTA-free protease inhibitor cocktail (Roche), and phosphatase inhibitor (Sigma)) was added to the dish, which was incubated on ice for 20 minutes, after which time cells were scraped off and added to an Eppendorf tube. Proteins were extracted by centrifugation. Protein concentration was measured using the Pierce BCA Protein Assay Kit and measured at 560 nm wavelength on a Versamax spectrophotometer.

Western blotting

Equal amounts of protein lysate previously measured to 20 μg had proportional amounts of NuPage Reducing agent and NuPage LDS Sample Buffer (Invitrogen) added to each sample. Proteins were denatured by boiling for 5 minutes at 95°C before being loaded onto NuPage Bis-Tris or Tris-Acetate gels with appropriate buffer (MOPS buffer for Bis-Tris gels or Tris-Acetate buffer for Tris-Acetate gels) to separate the proteins in the gel during the run cycle. Gels were then removed from the cassettes and transferred to PVDF membranes (Millipore) using a wet transfer method. Membranes were stained with Ponceau, then blocked in 5% milk in TBST (TBS + Tween) for one hour before adding a primary antibody diluted in fresh 5% milk for overnight incubation at 4°C. Membranes were then washed 3 times with TBST for 10 minutes each before adding a secondary antibody diluted in 5% milk for 1 hour at room temperature. Membranes were washed 3 times again with TBST. Before imaging, ECL Prime Western Blotting Detection Reagent (GE Healthcare) was added for 5 minutes for activation. Membranes were imaged with ImageQuant 800 OD (Amersham). Densitometry analysis was performed using ImageJ.

The following primary antibodies were used: mouse anti-Actin (1:10,000; Abcam), rabbit anti-RESF1 (1:1000; custom), and rabbit anti-KIAA1551 15622 (1:1000; Sigma). The following secondary antibodies were used: goat anti-rabbit (1:5000; Santa Cruz) and goat anti-mouse (1:10,000; GE Healthcare).

Nuclear co-immunoprecipitation

Co-immunoprecipitations were performed using the Active Motif Nuclear Complex Co-IP Kit following the manufacturer’s recommended nuclear fractionation protocol. 24–48 hours before nuclear isolation, cells were seeded at equal densities on tissue culture dishes. Nuclear lysates were added to a 2 mL low retention tube at equal concentrations along with the appropriate kit buffer and 2 μg of the appropriate antibody, then incubated overnight rotating at 4°C. The next day, Protein G Dynabeads (Invitrogen) capture beads were washed 2 times with the kit wash buffer, then 50 μl of beads were added to each sample and placed on the rotator for 1 hour rotating at 4°C. After incubation, samples with beads were washed 3 times with kit buffer. 2X NuPage buffer and reducing agent buffer were added to each sample at an appropriate amount before boiling at 95°C for 5 minutes. Samples were then loaded onto PAGE gels for western blot as described.

Ethinyl uridine (EU) pulse

Cells were seeded at equal densities on 12-mm glass coverslips (Electron Microscopy Sciences, catalog #71887–04) that were placed on a plate. After 24–48 hours, cells were incubated for 30 minutes with EU from a Click-iT RNA Imaging Kit (Invitrogen). After incubation, cells were fixed and processed through the kit protocol. After Click chemistry was utilized to add Alexa Fluor 594 to EU, DAPI was added, and then cells were washed twice with PBS. Cells were mounted on slides with ProLong Glass Antifade Mountant (Invitrogen). Confocal images were taken using a Zeiss LSM 780 in the NCI Microscopy Core Facility.

O-propargyl-puromycin (OPP) pulse

Cells were seeded at equal densities on 12-mm glass coverslips (Electron Microscopy Sciences, catalog #71887–04) that were placed on a plate. After 24–48 hours, cells were incubated for 30 minutes with OPP from a Click-iT Protein Imaging Kit (Invitrogen). After incubation, cells were fixed and processed through the kit protocol. After Click chemistry was utilized to add Alexa Fluor 594 to EU, DAPI was added, and then cells were washed twice with PBS. Cells were mounted on slides with ProLong Glass Antifade Mountant (Invitrogen). Confocal images were taken using a Zeiss LSM 780.

Immunofluorescence staining

Cells were seeded at equal densities on 12-mm glass coverslips (Electron Microscopy Sciences, catalog #71887–04) that were placed on a 12-well plate. After 24–48 hours, media was aspirated from the well, washed 3 times with PBS then fixed with 4% paraformaldehyde (PFA) for 20 minutes. PFA was removed, and then coverslips were washed twice with PBS. After washing, 0.5% Triton X-100 in PBS was added for 5 minutes at room temperature to permeabilize the cells. After permeabilization, cells were washed 3 times with PBS, and 5% BSA in PBS was added for 1 hour for blocking. After blocking, the primary antibody diluted in 5% BSA was added to the coverslips and incubated overnight at 4°C. After incubation, coverslips were washed 3 times with PBS before adding diluted secondary antibody in 5% BSA in PBS for 1 hour at room temperature followed by 3 washes with PBS. After washing, DAPI diluted in PBS (1 μg/ml final concentration) was added to the coverslips and incubated for 10 minutes at room temperature. The coverslips were again washed 3 times with PBS before mounting on slides with ProLong Glass Antifade Mountant (Invitrogen). Confocal images were taken using a Zeiss LSM 780.

The following primary antibodies were used: C12orf35 13816 (1:500; Sigma HPA013816), KIAA1551 15622 (1:500; Sigma HPA015622), 186C7 (1:500; GenScript custom antibody), Fibrillarin (1:500; Abcam), NPM1 (1:500; Abcam), SC35 (1:1000; Sigma) and HA-tag (1:200; Cell Signalling). Secondary antibodies used were mouse Alexa Fluor 488 (1:200; Invitrogen), rabbit Alexa Fluor 568 (1:200; Invitrogen), and rabbit Alexa Fluor 647 (1:200, Invitrogen)

PolyA mRNA FISH

Cells were seeded at equal densities on 12-mm glass coverslips (Electron Microscopy Sciences, catalog #71887–04) that were placed on a 12-well plate. After 24–48 hours, media was aspirated from the well, washed once with PBS then fixed with 4% paraformaldehyde in PBS for 10 minutes at room temperature. Paraformaldehyde was removed, and 100% ice-cold methanol was added to the coverslips to permeabilize the nucleus for 10 minutes. Methanol was removed, and 70% ethanol was added for at least 10 minutes to rehydrate the cells after the fixation steps. 1 M Tris pH 8.0 was then added for 5 minutes, removed, and hybridization buffer (1 mg/ml yeast tRNA, 0.005% BSA, 10% dextran sulfate, 25% deionized formamide, 2X SSC, and 1 ng/μl fluorescent 5’-Cy3-Oligo d(T)50 probe (Gene Link, catalog #26-4322-02)) was added to the coverslips and incubated at 37°C overnight sealed in a plastic bag with wet kimwipes to maintain humidity. After hybridization, coverslips were washed once with 4X SSC for 5 minutes, and again with 2X SSC for 5 minutes. After washing, coverslips were incubated with 1 μg/ml DAPI in 2X SSC and 0.1% Triton X-100 for 15 minutes and washed twice with 2X SSC for 5 minutes each. Coverslips were mounted on glass slides with ProLong Glass Antifade Mountant (Invitrogen). Confocal images were taken using a Zeiss LSM 780.

RNA FISH

Cells were seeded at equal densities on 12-mm glass coverslips (Electron Microscopy Sciences, catalog #71887–04) that were placed on a 12-well plate. Coverslips were washed once with PBS and fixed in 4% paraformaldehyde in PBS for 10 minutes at room temperature and washed once again with PBS. 70% ethanol was added to the coverslips and incubated at 4°C for 1 hour. Coverslips were washed with 2X SSC with 10% formamide for 5 minutes, then permeabilized with 0.5% Triton X-100 in 2X SSC for 10 minutes at room temperature. Hybridization buffer (10% dextran sulfate, 1 mg/mL yeast tRNA, 10% formamide, 2X SSC, and RNase-free water) with an RNA FISH oligo labeled with a Cy3 fluorophore (1 μl per 100 μl hybridization reaction) was added to each sample and incubated overnight at 30°C sealed in a plastic bag with a few wet kimwipes. After overnight incubation, coverslips were rinsed once and washed with 2X SSC with 10% formamide once for 15 minutes, then washed again for 30 minutes. Coverslips were stained with 5 μg/ml DAPI in 2X SSC and 10% formamide for 30 minutes at room temperature. After DAPI staining, the coverslips were mounted on glass slides with ProLong Glass Antifade Mountant (Invitrogen). Confocal images were taken using a Zeiss LSM 780. Oligo sequences are provided in the supplementary information (S1 Table).

Tumorsphere assay

Cells were seeded at equal densities on ultra-low attachment 24-well plates (Corning) in MethoCult H4100 mixture containing MammoCult (Cat #05621), MammoCult Proliferation Supplement (Cat #05622), hydrocortisone, and heparin. The plates were mixed for 15 minutes at 4°C rocking to lodge cells into the MethoCult. The plates were then incubated at 37°C undisturbed for 7–12 days. After incubation, spheres were quantified and measured in diameter using Celigo.

Gel mobility shift assay and circular dichroism

Oligomers (~4.5 μM) forming G4s were incubated in 100 mM potassium chloride, sodium chloride, or lithium chloride in water at 95°C for 5 minutes before cooling to room temperature slowly (~2 h). The reactions were loaded onto 3% agarose gels containing 1X TAE buffer at 100V at room temperature. The signal was detected using BioRad ChemiDoc Imaging System.

Circular dichroism (CD) spectra were recorded on a spectrophotometer using a 1-mm quartz cuvette from Hellma Analytics with a reaction volume of 200 μl. Samples were prepared the same for gel mobility shift assays. An average of three scans were taken for each sample, and the buffer spectrum was subtracted. For equilibrium CD measurements, wavelengths between 230 nm and 315 nm were scanned.

BioTAP analysis

FASTQ files from the Short Read Archive (PRJNA509912) were downloaded into Partek Flow and based on Phred scores <20 trimmed from the 3’ end. The reads were then aligned to the T2Tv2a or hg38 human genome build using BWA 0.7.17 and peaks called by comparing input and BioTAP XL samples using MACS v 3.0.0a7. Peaks were then quantified and annotated to the T2Tv2a genome using the PartekFlow default settings. Statistical analysis was performed using ANOVA and peaks with an FDR ≤ 0.05 were considered significant. Overlap analysis of the BioTAP XL T2Tv2a peaks was performed by uploading BED files to the UCSC Genome Browser and performing intersection analysis using the UCSC Genome Browser TableBrowser tool. Analysis using the GREAT tool was performed using BED files aligned to the hg38 genome build.

Analysis of the RNA-seq data was performed in Partek Flow by aligning to the T2Tv2a human genome using STAR 2.7.8a using default settings. Reads were normalized by median ratio and differentially expressed transcripts were identified using DESeq2. Transcripts were considered significantly different at an FDR ≤ 0.05.

Nuclear fractionation analysis

Subnuclear fractionation assays were performed with a Thermo Scientific Subcellular Fractionation Kit for Cultured Cells (cat. No. 78840), following the manufacturers suggested protocol.

mTOR pathway analysis

mTOR pathway analysis was performed with the Full Moon Biosystems mTOR Phospho Antibody array (cat. No. PMT138) using shScramble and shG5 MDA-MB-231 cell lysates, following the manufacturers recommended protocol.