Nfib Promotes Metastasis through a Widespread Increase in Chromatin Accessibility

Abstract

Metastases are the main cause of cancer deaths, but the mechanisms underlying metastatic progression remain poorly understood. We isolated pure populations of cancer cells from primary tumors and metastases from a genetically engineered mouse model of human small cell lung cancer (SCLC) to investigate the mechanisms that drive the metastatic spread of this lethal cancer. Genome-wide characterization of chromatin accessibility revealed the opening of large numbers of distal regulatory elements across the genome during metastatic progression. These changes correlate with copy number amplification of the Nfib locus, and differentially accessible sites were highly enriched for Nfib transcription factor binding sites. Nfib is necessary and sufficient to increase chromatin accessibility at a large subset of the intergenic regions. Nfib promotes pro-metastatic neuronal gene expression programs and drives the metastatic ability of SCLC cells. The identification of widespread chromatin changes during SCLC progression reveals an unexpected global reprogramming during metastatic progression.

Figure 1. Small cell lung cancer exists in two distinct chromatin accessibility states

(A) Genetically engineered mouse model of SCLC. (B) GFP^positive primary tumors and metastases within Tomato (Tom)^positive normal tissues (top). GFP and Uchl1 IHC and H&E are shown. Top scale bars = 5 mm. Bottom scale bars = 100 μm. (C) Representative FACS plot of a dissociated tumor. FSC/SSC gated, Lineage^negative, viable (DAPI^negative) cells are shown. GFP^postiveTom^negative SCLC cells are indicated. (D) Correlation of chromatin accessibility of 8 samples by ATAC-seq analysis. T = tumor, L = liver metastasis. (E) Differential accessibility (log₂ fold change in reads per accessible region) plotted against the mean reads per region. FDR is false discovery rate that the absolute value of log₂ fold change is >1. (F) Insertion tracks of samples at an example locus on chromosome 7. Differentially open regions are marked with arrows. (G) Fraction of total regions that are differentially accessible in multiple tissue and cell type comparisons. lfc = log₂ fold change. See also Figure S1.

Figure 2. Changes in chromatin accessibility are gene distal, late replicating, evolutionarily conserved, and enriched for regulatory regions in neuronal tissue

(A) Distance to closest transcription start sites (TSSs) of all accessible regions and differentially open regions. (B) Insertion tracks of merged hypo- and hyper-accessible samples. (Bottom) Reads in distal peaks are binned into 2 Mb windows. Examining differential read counts in windows identifies domains that are enriched for differentially open regions (black bars) and depleted for differentially open regions (grey bars, lower). Replication timing (ES-TT2 cells) and number of TSSs per window is also shown. (C) Number of TSSs per 2 Mb window in domains either enriched or depleted for differentially open regions in SCLC. Enriched domains have fewer TSSs (p < 1×10⁻¹⁰⁰ by Mann-Whitney U test). (D) Average replication timing per 2 Mb window (ES-TT2 cells) in domains either enriched or depleted for differentially open regions in SCLC. Enriched domains have significantly later (more negative) replication timing (p < 1×10⁻¹⁰⁰ by Mann-Whitney U test). (E) Average sequence conservation (phyloP) in differentially open intergenic regions, constitutively open intergenic regions, and closed regions (5 kb downstream of any accessible region, same window size). Differentially open, intergenic regions are twice as likely to have higher average sequence conservation (>0.2) than constitutively open, intergenic regions (p = 1×10⁻¹¹⁴ by Fisher's exact test). (F) Overlap of differentially accessible regions with DNase hypersensitive sites (DHS) from other cell types. Mean overlap with DHS peak calls are shown. Bars represent 95% confidence intervals. See also Figure S2.

Figure 3. Increased Nfib expression in invasive SCLC leads to increased Nfib binding in differentially open regions and occupancy of less canonical sites

(A, B) Motif enrichment in newly open regions compared to other accessible regions. (A) Top known motif enrichments and (B) de novo motif enrichments. (C) Copy number amplification of the Nfib locus inferred from ATAC-seq. (D) Representative IHC for Nfib on tumors at different stages of SCLC progression from TKO-mTmG mice. Lung^Early are neuroendocrine hyperplasias. Scale bars = 500 μm. (E) Nfib expression at different stages of SCLC progression. Number of tumors in each group is indicated. The percent of Nfib^high tumors in early hyperplasias and lung tumors versus lymph node (LN) and liver metastases is significantly different (p < 1×10⁻¹⁶ by Fisher's exact test). (F) Immunofluorescence for Nfib on FACS-isolated GFP^pos cancer cells from primary tumors, disseminated tumor cells (DTCs), and metastases. A representative SCLC cell from a primary tumor and DTCs from the pleural cavity are shown. The percent of Nfib^high DTCs versus Nfib^high lung tumor cells is significantly different (p < 0.0001). (G) (Top) ATAC-seq footprint at the NFI full site. Insertions per site are normalized to have the same average number of insertions 200-500 bp away from motif. (Bottom) Modeled insertion bias of Tn5 around NFI full sites. (H) Occupancy of NFI full sites in the merged hyper- and hypo-accessible samples in bins of motif score (log odds similarity to the consensus motif). Error bars represent 95% confidence intervals. (I) Nucleosome occupancy around NFI full sites in differentially open and constitutively open regions in hypo- and hyper-accessible samples. Shaded areas represent 95% confidence intervals. See also Figure S3 and Table S1.

Figure 4. Cell lines confirm relationship between Nfib expression, Nfib binding by ChIP, and chromatin accessibility state

(A) Immunoblot analysis of Nfib expression in 6 murine SCLC cell lines. Hsp90 shows loading. (B) Copy number amplifications on chromosome 4 from ATAC-seq. (C) Correlation of chromatin accessibility across cell lines and technical replicates. Nfib^high and Nfib^low cell lines are indicated in red and blue, respectively. (D) ATAC-seq insertion tracks of cell lines. Differentially accessible regions are indicated with arrows. (E) Differential accessibility (log₂ fold change in reads per accessible region) between Nfib^high and Nfib^low cell lines, plotted against the mean reads per region. FDR is false discovery rate that the absolute value of log₂ fold change is >0.5. Brackets indicate number of significantly changed peaks. (F) Correlation of differential accessibility between Nfib^high and Nfib^low cell lines and hypo- and hyper-accessible ex vivo samples (r = 0.72). (G) Fold enrichment above input of Nfib ChIP around NFI motif sites for three cell lines. (H) Log₂ fold change in Nfib ChIP-seq reads per merged ChIP peak versus the mean number of reads per peak. FDR is for absolute value of log₂ fold change > 0.5. (I) Distribution of motif scores of sites within Nfib ChIP peaks, either those that gave more signal in the Nfib^high than the Nfib^low cell lines, or those that were not significantly different (constitutive). The maximum scoring NFI full site within each ChIP peak was used. Constitutive peaks have higher motif score than differential peaks (Mann-Whitney p < 1×10⁻¹⁰⁰). (J) Correlation of differential ChIP signal in accessible regions (from ATAC-seq) and differential accessibility of Nfib^high and Nfib^low cell lines (r = 0.35). See also Figure S4.

Figure 5. Nfib maintains chromatin accessibility at a subset of regions and is required for metastatic ability

(A) Immunoblot for Nfib in two Nfib^high SCLC cell lines +/− Nfib knockdown (shNfib). Hsp90 shows loading. shCon = shControl. (B) Differential accessibility with shNfib#1 and control in combined 16T and KP1 cell lines. (C) Correlation of differential accessibility with Nfib knockdown and differential accessibility of hyper- and hypo-accessible ex vivo samples (r = −0.55). (D) Nucleosome occupancy around NFI sites in shNfib and control cells. Shaded regions are 95% confidence intervals. (E) Subcutaneous (SubQ) tumor weight of 16T cells +/− Nfib knockdown 4 weeks after transplantation. Each dot represents a tumor and the line indicates the mean. ns = not significant. (F) Number of liver metastases from SubQ tumors of 16T cells +/− Nfib knockdown. Scale bars = 500 μm. Each dot represents a mouse and the line indicates the mean. *p < 0.025. (G) Percent of SubQ tumor area that expresses Nfib (Nfib^pos) and percent of liver metastases (Met) that are Nfib^pos. Paired analysis of Nfib^pos areas in SubQ and Met tumors is significant (p < 0.0025). (H-I) Representative H&E images and quantification of liver metastases 3 weeks after intravenous transplantation. Scale bars = 500 μm. Each dot represents a mouse and the bar is the mean. ***p < 0.001. (J) Percent of Nfib-positive liver metastases (assessed by IHC) after intravenous transplantation of 16TshNfib cells. (K) Anchorage-independent growth of Nfib^high SCLC cell lines +/− Nfib knockdown. Mean +/− SD is shown. ****p < 0.0001, ***p < 0.001. (L) Matrigel migration assay of Nfib^high SCLC cell lines +/− Nfib knockdown. Mean +/− SD is shown. **p < 0.01 *p < 0.02. See also Figure S5.

Figure 6. Nfib is sufficient to open a subset of sensitive regions and to drive metastatic ability

(A) Expression of Nfib by immunoblot in an Nfib^low cell line (KP22) with doxycycline (Dox)-inducible expression of Nfib. Two Nfib^high cell lines 16T and KP1 are shown. Hsp90 shows loading. (B) Differential accessibility in Nfib-expressing and control samples. (C) Correlation of differential accessibility with Nfib expression and differential accessibility of hyper- and hypo-accessible ex vivo samples (r = 0.26). (D) (Top) ATAC-seq insertions and (Bottom) Nfib ChIP enrichment above input at an example locus. Green arrows highlight regions that close with Nfib knockdown, red arrows highlight regions that open with Nfib expression, and blue arrows show regions with differential ChIP signal. (E) Distribution of accessibility (log2 read count) in 1000 bp windows around NFI motif sites +/− Nfib induction. Panels show motif sites that are specifically more open in hyper-accessible ex vivo samples or that are sensitive to Nfib overexpression in KP22 cell line. The dashed line indicates the average number of reads around motif sites that were not accessible. (F) Colony formation of KP22 cells with induced (Nfib+Dox) and constitutive Nfib expression in anchorage-independent conditions. Mean +/− SD is shown. ***p < 0.0005 **p < 0.001. (G) Light (left) and H&E (right) image of liver metastases of KP22 +/− Nfib expression after intravenous transplantation. Scale bars = 5 mm. (H,I) Number of surface liver metastasis (H) and liver metastases quantified by histology (I). **p < 0.001 See also Figure S6.

Figure 7. Gene expression changes in response to Nfib promote neuronal state

(A) Log₂ fold change of all genes that significantly change (absolute value of log2 fold change > 0.5 at FDR < 0.1) with both knockdown and overexpression by RNA-seq. (B) GO terms (merged into categories) and log₂ fold change in expression of a subset of genes. (C) Overlap of Nfib-sensitive regions with DNase hypersensitive sites (DHS) in other tissues. Dots represent multiple ES cell lines plus technical replicates, or adult and embryonic brain plus technical replicates. (D) Fraction of genes that have at least one distal, Nfib-sensitive region within 100 kb of transcription start site. oe=overexpression, kd=knockdown. Error bars are 95% confidence intervals. ***p < 1×10⁻¹⁰ by Fisher's exact test. (E) Average change in expression with Nfib overexpression of 4 gene categories. Promoter binding = Genes with promoter-proximal ChIP peak within 1 kb of TSS. Nfib-sens. distal region = Genes with a distal, Nfib-sensitive region within 100 kb of the TSS that opens with Nfib overexpression. Error bars are 95% confidence intervals. ns=not significant, **p < 1×10⁻⁷ by Mann-Whitney U test. (F) Example locus showing Nfib ChIP (top two panels) and ATAC-seq data. Black squares below each set of tracks represent significantly changed regions. (G) Model for Nfib-dependent metastatic progression. See also Figure S7 and Table S2.