Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Apr 9.
Published in final edited form as: Nature. 2014 Aug 3;514(7521):257–261. doi: 10.1038/nature13573

Required Enhancer: Matrin-3 Network Interactions for Pit1 Homeodomain Transcription Programs

Dorota Skowronska-Krawczyk 1,*, Qi Ma 1, Michal Schwartz 1,2, Kathleen Scully 1, Wenbo Li 1, Zhijie Liu 1, Havilah Taylor 1, Jessica Tollkuhn 1, Kenneth A Ohgi 1, Dimple Notani 1, Yoshinori Kohwi 3, Terumi Kohwi-Shigematsu 3, Michael G Rosenfeld 1,*
PMCID: PMC4358797  NIHMSID: NIHMS604603  PMID: 25119036

Abstract

Homeodomain proteins, described 30 years ago1,2, exert essential roles in development as regulators of target gene expression3,4, however the molecular mechanism underlying transcriptional activity of homeodomain factors remains poorly understood. Here, investigation of a developmentally-required POU-homeodomain transcription factor, Pit1/Pou1f1, has revealed that, unexpectedly, binding of Pit1-occupied enhancers5 to a nuclear matrin-3-rich network/architecture6,7 is a key event in effective activation of the Pit1-regulated enhancer/coding gene transcriptional program. Pit1 association with Satb18 and β-catenin is required for this tethering event. A naturally-occurring, dominant negative, point mutation in human Pit1 (R271W), causing combined pituitary hormone deficiency (CPDH)9, results in loss of Pit1 association with β-catenin and Satb1 and therefore the matrin-3-rich network, blocking Pit1-dependent enhancer/coding target gene activation. This defective activation can be rescued by artificial tethering of the mutant R271W Pit1 protein to the matrin-3 network, bypassing the prerequisite association with β-catenin and Satb1 otherwise required. The matrin-3 network-tethered R271W Pit1 mutant, but not the untethered protein, restores Pit1-dependent activation of the enhancers and recruitment of co-activators, exemplified by p300, causing both eRNA transcription and target gene activation. These studies have thus revealed an unanticipated homeodomain factor/β-catenin/Satb1-dependent localization of target gene regulatory enhancer regions to a subnuclear architectural structure that serves as an underlying mechanism by which an enhancer-bound homeodomain factor effectively activates developmental gene transcriptional programs.

During pituitary development, the POU-homeodomain transcription factor, Pit1, is necessary for differentiation of thyrotrope, lactotrope, and somatotrope cell types in both mice and humans10,11. To further understand the molecular basis for Pit1-mediated gene activation, we mapped the genomic localization of Pit1 by ChIP-seq using specific Pit-1 antibody (Fig. S1a, b) in a growth hormone (GH)-expressing rat pituitary cell line (GC). Of 14,466 Pit1 binding sites identified, >80% overlapped with H3K4me2 histone marks but not with transcription start sites, indicative of enhancer elements5 (Fig. 1a, b, Supplementary Table 1).

Figure 1. Co-localization of Pit1 and matrin-3 on enhancers.

Figure 1

Open in a new tab

a, ChIP-Seq analysis of Pit1 and H3K4me2 genome-wide co-localization in GC rat pituitary somatotrope cell line. b, Most DNA sites co-bound by Pit1 and H3K4me2 lie outside gene proximal promoters. c, HA-tagged Pit1 immunoprecipitated from 293-T cells. Co-purifying factors identified with mass spectrometry. d, Co-immunoprecipitation of Pit1 followed by Western blot analysis confirmed Pit1:endogenous matrin 3 and hnRNPU interactions in GC cells. e, Most matrin-3/H3K4me2 sites in GC cells lie outside gene proximal promoters. f, Heat map of ChIP-Seq data on non-promoter genome-wide association of Pit1, H3K4me2, H3K4me1, H3K27Ac, matrin-3 in GC cells centered on Pit1 sites and categorized as transcription termination sites (TTS), intergenic, intronic sites. g, Meta-analysis of matrin-3 ChIP-seq data h, Examples of immuno-FISH experiments showing matrin3 spots (red) colocalized with GH locus spots (green) in GH-expressing GC cells. Chart represents count of percentage of signals exhibiting co-localization, n≥200, +/− SD.

To identify Pit1 interacting factors, we immunoprecipitated HA-tagged Pit1 from 293T cells, and using mass spectrometry, identified matrin-3 and hnRNPU as two highly abundant proteins interacting with Pit1 (Fig. 1c, Supplementary Table 2, 3). The interactions were confirmed by co-immunoprecipitation (co-IP)/Western blot experiments in 293T cells (Fig. S1c) and with endogenous proteins in GC cells (Fig. 1d, Fig. S1d). Both matrin3 and hnRNPU have been previously found in a salt extraction-resistant nuclear fraction, as part of an internal fibrogranular network6,12,13 which we term the nuclear matrin-3-rich network (Fig. S1e). These proteins bind DNA at sites known as matrix/scaffold attachment regions13. ChIP-Seq using a specific α-matrin-3 antibody revealed that >50% of matrin-3 binding sites co-localized with H3K4me2 peaks (Fig. S1f), showing significant enrichment (~16 fold, p-value<1e-100) over the ~3% predicted random co-localization with H3K4me2 peaks, covering ~5% of the rat genome. Of these matrin3/H3K4me2 sites, ~80% were elements distal to transcription start sites (Fig. 1e). This finding provides an initial demonstration of matrin3 association with DNA regulatory sequences in the genome. We noticed that Pit1-associated H3K4me2, H3K4me1 and H3K27Ac-marked enhancers, co-localized with matrin-3 (Fig. 1f–i, Fig. S1g), and confirmed matrin-3 antibody specificity by knockdown of endogenous matrin-3 protein (Fig. S1h, i). To investigate whether regulatory elements bound by Pit1 occupy the same nuclear compartment as matrin-3, we performed immuno-FISH experiments with DNA-FISH probes specific to the growth hormone (GH) locus and α-matrin-3 antibodies, finding that >65% of the FISH signals co-localized with matrin-3 (Fig. 1j, Fig S1j).

Because the pituitary specific homeodomain protein, Prop1, interacts with β-catenin to activate target gene expression14, we investigated β-catenin interacting partners in differentiated pituitary cells by expressing β-catenin fused to the biotin ligase recognition peptide (BLRP), along with biotinylating enzyme BirA, in GC cells. Biotinylated β-catenin was present in both the cytoplasmic and the nuclear compartments (Fig. 2a, Fig. S2a). Streptavidin pulldown, followed by mass spectrometry, revealed that while cytoplasmic β-catenin interacting factors included α-catenin, as previously described15, Pit1 peptides predominated in the nuclear β-catenin fraction (Fig. 2a, Supplementary Table 4–7). This putative Pit1:β-catenin interaction was confirmed between endogenous proteins by co-IP/Western blot experiments using nuclear extracts from GC pituitary cells (Fig. 2b). Using GST-pulldown assays, armadillo repeat 8 of β-catenin proved sufficient to mediate this interaction, exactly within a region that has been previously described to interact with Prop114 as well as Lef/Tcf16 (Fig 2c, Fig S2b).

Figure 2. Pit1 association with LIS resistant nuclear component is β-catenin- and SATB1- dependent.

Figure 2

Open in a new tab

a, BLRP-β-catenin immunoprecipitated from cytoplasmic or nuclear fractions of GC cells. Co-puryifying factors identified by mass spectrometry. b, Co-IP-Western analysis confirmed Pit1:β-catenin interaction and revealed interaction with SATB1. c, GST-pulldown showing β-catenin armadillo repeat 8 (within region previously shown to interact with Prop1 and Lef1) is required for interaction with Pit1. d, Nascent GH transcripts levels after siRNA knockdown of β-catenin and Pit1 analyzed by RT-qPCR. Experiments were repeated 2 times, and p-values calculated using student’s two tailed t-test. (+/− SD; **p<0.01) e, Co-IP of HA-tagged Pit1 protein in 293T cells before and after β-catenin and Satb1 knockdown showing Pit1 interacts simultaneously with both proteins. f, LIS nuclear extraction before and after β-catenin or/and Satb1 knockdown shows that both proteins are needed for Pit1 retention in LIS resistant fraction. g, Co-IP of Pit1 protein in GC cells before and after simultaneous knockdown of β-catenin and Satb1 shows interaction of Pit1 with matrin-3 is dependent on both. h, Examples of immuno-FISH experiments showing GH locus (green) colocalizing with matrin3 (red) with higher frequency in control culture conditions than after siRNA β-catenin and siRNA Satb1 treatment in GC cells. In Pit1 positive, non-GH-expressing MMQ cells, significantly fewer GH loci are associated with matrin3. The chart represents the count of percentage of signals exhibiting co-localization in control vs. siβ-catenin and siSATB1 conditions compared to MMQ cells; n≥200, +/− SD, ***p<0.001

We then examined whether β-catenin plays a role in regulating expression of Pit1 target genes, such as GH. Effective knockdown of β-catenin in GC cells (Fig. S2c) dramatically decreased the level of nascent GH transcripts (Fig. 2d), analogous to what was observed when Pit1 was knocked down (Fig. 2d, Fig. S2c). β-catenin also forms a complex with Satb117, a well-established genome organizer essential for multiple biological processes, including T-cell activation and cancer progression8,18,19. RT-qPCR analysis revealed that the upregulation of Satb1 expression in the mouse pituitary gland coincides with differentiation of Pit1 lineages during pituitary development (Fig. S2d), and immunohistochemical analysis confirmed that SATB1 protein is expressed in somatotropes and other cells of the mouse pituitary gland (Fig. S2e). Co-IP of Pit1 from GC cells revealed an interaction with Satb1 (Fig. 2b) which was reproduced by co-IP of HA-tagged Pit1 in 293T cells (Fig. S2f). Indeed, conditional knock-out of Satb1 in the mouse anterior pituitary significantly decreased expression of GH (Fig. S2g). We therefore investigated whether interaction between Pit1 and Satb1 was dependent on β-catenin, and vice-versa. Exogenously expressed, HA-tagged Pit1 protein was pulled down after knocking down either β-catenin or Satb1 in 293T cells. In the absence of β-catenin, the interaction between Pit1 with Satb1 was diminished, indicating a role for β-catenin in mediating Pit1:Satb1 interaction. Reciprocally, the interaction of Pit1 with β-catenin was partially abolished by Satb1 knockdown (Fig 2e).

Pit1 has been shown previously to co-fractionate with the insoluble, matrin-3 rich, nuclear fraction in in-vitro biochemical studies20. We therefore used lithium 3,5-diidosalicylate (LIS), a chaotropic reagent that extracts a majority of nuclear proteins but preserves the nuclear matrin-3-rich network21, to investigate Pit1 subnuclear locations (Fig. S2h). We observed that, in somatotropes, Pit1 indeed co-fractionated with matrin3 and Satb1 in the LIS resistant insoluble fraction, while β-catenin was partitioned between the insoluble and soluble fractions (Fig. S2i). When both β-catenin and Satb1 proteins were simultaneously depleted (Fig. S2j), the presence of Pit1 in the LIS resistant insoluble fraction was significantly reduced (Fig. 2g). In the double knockdown, total Pit1 protein levels were unchanged and Pit1 protein remained localized to the nucleus (Fig. S2k); however, it was now detectable in the looped-out DNA fraction obtained after DNAase I digestion of nuclear halos (Fig. S2l). Together, these data indicate that localization of Pit1 protein to the insoluble fraction is dependent on interaction with β-catenin and Satb1. When endogenous Pit1 protein was pulled down before and after knocking down both β-catenin and Satb1 in GC cells, the absence of these proteins highly affected interaction between Pit1 and matrin-3 (Fig. 2g).

To further investigate whether association of Pit1-bound enhancers with the matrin-3 fraction is indeed dependent on β-catenin and SATB1, we performed immuno-FISH experiments in GC cells with α-matrin-3 antibodies and DNA-FISH probes specific to the growth hormone (GH) locus. In GH-expressing GC cells, ~65% of the FISH signals co-localized with matrin-3. When both β-catenin and Satb1 were depleted, only ~35% of FISH signals loci co-localized with matrin-3, similar to the overlap of a GH locus in a Pit-1 positive, GH-non-expressing cell line (MMQ) (Fig. 2h, S2m). We thus conclude that association of Pit1 enhancers with the nuclear matrin-3-rich network/structure observed by immuno-FISH (Fig. 1j, Fig. 2h) and biochemical extractions (Fig. 2f Fig. S2h,) is dependent on intact interactions of Pit1 with both β-catenin and Satb1.

Global run-on (GRO-seq) analysis using GC cells before and after simultaneous knockdown of both β-catenin and Satb1 revealed that in total, 1350 coding gene nascent transcripts were downregulated and 916 genes were upregulated under these double-knockdown conditions (Fig. S3a). Out of the 1350 genes positively regulated by both β-catenin and SATB1 (Supplementary Table 8), 991 had Pit1 bound enhancer elements within 200kbp of their start sites (Fig. S3a), and all 991 were significantly downregulated (Fig. 3a). Consistent with their functional importance22 the expression of the eRNAs associated with these 991 Pit1-bound enhancers was significantly affected in the simultaneous knockdown of β-catenin and Satb1 (Fig. 3b). ChIP-seq analysis revealed that Satb1 and β-catenin proteins were located at the center of Pit1-positive enhancers (Fig. S3b–d), strongly supporting the co-binding model of these three proteins. Based on this analysis, we selected several highly down-regulated genes that contain Pit1 enhancers (Fig. S3c) and observed that knockdown of either of the two proteins significantly diminished target gene expression (Fig. S3g). Additionally, knockdown of either Pit1 or β-catenin affected Satb1 association with enhancer elements, as assessed by ChIP-qPCR (Fig. 3c, Fig. S3e). Association of matrin-3 with Pit-1 dependent enhancers was significantly reduced when either β-catenin or SATB1 were downregulated (Fig. S3h). Finally, we confirmed that lack of Pit1 protein inhibited association of selected Pit1-dependent enhancers with matrin-3 (Fig. 3d, Fig. S3f). Together, these data suggest that functional interaction between Pit1-dependent regulatory elements and matrin-3 requires association of the DNA-bound Pit1 with β-catenin and SATB1. Consistent with this suggestion, lack of matrin3 had a negative effect on the expression of target genes (Fig. S3i).

Figure 3. β-catenin and SATB1 influence transcription of genes regulated by Pit1 enhancers.

Figure 3

Open in a new tab

a, Genes containing Pit1 enhancers (991) are significantly downregulated upon knockdown of β-catenin and Satb1. RPKM – reads per kilobase per million b, GRO-seq tag distribution on both strands of Pit1 bound enhancers before and after treatment of GC cells with β-catenin and Satb1 siRNAs. c, ChIP-qPCR analysis of Satb1 association with Pit1 enhancers upon either Pit1 or β-catenin knockdown. Experiments were repeated 3 times, and p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001) d, ChIP-qPCR analysis of matrin-3 association with Pit1 enhancers upon Pit1 knockdown. Experiments repeated 2–4 times, p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001)

Naturally occurring mutations in Pit1 underlie number of combined pituitary hormone deficiency (CPHD) cases in humans, with the R271 residue being the most frequently mutated9. An amino acid 271 maps to the last turn of helix3 in the homeodomain and is located outside the DNA binding surface. While the R271W mutation in Pit1 confers dominant-negative properties (Fig. S4a)9,, it does not affect protein stability or binding to its cognate DNA sites (Fig. S4b–d)9. We therefore investigated whether the R271W mutation affects the composition of the Pit1 complex. Co-IP of recombinant HA-tagged Pit1 proteins followed by Western blot analysis revealed that the R271W mutation disrupted the ability of Pit1 to interact with matrin3 and hnRNPU, apparently due to its inability to bind β-catenin and Satb1 (Fig. 4a). Additionally, biochemical fractionation indicated that the R271W mutant is largely lost from the insoluble fraction (Fig. 4b), although both, WT and R271W proteins were expressed at similar levels and found entirely localized to the nucleus (Fig. S4e). Instead R271W was detected in the ‘looped-out DNA’ fraction (Fig. 4b), suggesting that it had lost its ability to interact with the LIS resistant fraction.

Figure 4. Naturally occurring R271W mutation in human Pit1 affects ability of Pit1 to interact with matrin-3 rich network.

Figure 4

Open in a new tab

a, Western blot analysis showing defective interaction of Pit1R271W mutant with components of nuclear matrin-3 enriched network as well as β-catenin and Satb1. b, Biochemical extraction shows that the amount of R271W mutant protein decreased in the LIS-resistant fraction and associates instead with the looped-out DNA fraction. c, Schematic representation of Pit1 protein with added SAF domain. red “x” – location of amino-acid 271 d, Analysis of effect of overexpression of different forms of Pit1 protein in absence of endogenous Pit1; top panels: ChIP-qPCR analysis of matrin-3 association,; bottom panels: RT-qPCR analysis of transcriptional effect on selected targets. Experiments were repeated 3 or 4 times, and p-values calculated using student’s two tailed t-test. (+/− SD; **p<0.01, ***p<0.001) e, Analysis of effect of ex vivo overexpression of R271W mutant with and without SAF domain in e17.5 snell mutant pituitary glands. GH expression is rescued only when the R271W-SAF protein is overexpressed. Bottom: quantification of percentage of HA-positive GH expressing cells. Two pituitaries were used per condition; p-value calculated using χ2 test. f, Model: Association of Pit1-bound regulatory element with matrin-3 rich network and loss of this interaction based on failure of R271W mutant to interact with β-catenin and SATB1 causing loss of gene activation as in CPHD.

Together, these data suggested that a key function of homeodomain transcription factors such as Pit1 might be to join bound regulatory regions to the matrin-3-rich network in an association required to activate target gene transcription. Because the transcriptionally-inactive R271W mutation interferes with this association, we experimentally tested this hypothesis by artificially tethering the R271W mutant to the matrin-3 network to try to restore target gene transcription. Several protein domains that are responsible for association with the matrin-3-rich “matrix attachment regions” have been previously described and include the SAF/SAP domain (e.g. in hnRNPU 23, SAF-B, PARP proteins24); AT-hooks (HMG proteins25); and the atypical homeodomain, SATB118. We therefore grafted the SAF/SAP domain from rat hnRNPU (AA 4–45), onto HA-tagged wild type and R271W Pit1 (Fig. 4c). As predicted, R271W-SAF hybrid protein localized efficiently in the insoluble nuclear fraction (Fig. S4f). Hybrid WT-SAF and R271W-SAF proteins were expressed in GC cells treated with siRNAs to deplete endogenous Pit1 (Fig. S4g). As expected, matrin3 association with Pit1-dependent enhancers and expression of Pit1-dependent coding genes and eRNAs were lost upon Pit1 knockdown but were restored in the presence of either WT or WT-SAF Pit1 (Fig. 4d, Fig. S4i). In contrast, neither the association of Pit1 enhancers with matrin-3 nor target gene and eRNA expression were rescued in R271W-expressing cells. However, the R271W-SAF hybrid rescued matrin-3 interaction and Pit1 transcriptional activation to levels close to that achieved by wild-type Pit1. These experiments indicate that specific subnuclear localization of Pit1 is a bona fide requirement for its full activity, and that this is the only functional defect in the R271W mutant Pit1 protein.

To investigate which step of enhancer activation is affected by R271W, we tested recruitment of CBP/p300, a known co-factor of Pit126, finding that p300 association with tested enhancer elements was significantly reduced (Fig. S4h). Overexpression of R271W was not able to rescue this recruitment while hybrid R271W-SAF did restore p300 association with Pit1 bound enhancers (Fig. S4h). These data suggest that eRNA transcription and binding of critical co-activators, such as p300/CBP, require recruitment of Pit1 to the matrin-3-rich network.

To further substantiate these observations in intact pituitary, ex vivo rescue approach took advantage of the “Snell” mouse model that harbors a mutation (W261C) in Pit1 that renders it incapable of binding to cognate DNA sites. Overexpressing HA-tagged Pit1-R271W-SAF hybrid, but not Pit1-R271W protein, rescued expression of endogenous GH as visualized by immunohistochemistry (Fig. 4e). Based on these data, we propose a model in which Pit1 association with the matrin-3-rich network is indispensable for its biological activity (Fig. 4f), as demonstrated by the lack of transcriptional activity of the naturally occurring mutant R271W Pit1 (Fig. 4f).

Nuclear organization and genome 3D structure are postulated to play a prominent role in regulation of gene expression28,29. Here, we have presented evidence that a homeodomain transcription factor regulates gene transcriptional programs through interaction with components of subnuclear structure (e.g. matrin-3 rich structure). Other homeodomain proteins (e.g. Lhx3, Prop1, Oct1), as well as different classes of transcription factors30, have been shown to associate with the high salt-extraction resistant fraction, corresponding to the matrin3-rich network. Therefore, our data suggest a model in which this interaction may be required for transcriptional activation for many homeodomain proteins, helping to solve a long-standing gap in our understanding of gene activation programs by this important class of DNA binding transcription factors, and perhaps several other types of DNA binding transcription factors, revealing that the regulated enhancer activation requires localization in a specific subnuclear architectural structure.

Supplementary Material

1. Supplementary Figure 1. Matrin-3 co-localizes with enhancers in GC cells.

a, Validation of Pit1 antibody specificity - qPCR analysis of ChIP signals on two known targets, the GH and NeuroD4 enhancers shows lack of Pit1 signal after Pit1 knockdown. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; ***p<0.001) b. Western-blot confirming Pit1 knockdown efficiency in samples used to assess Pit1 antibody specificity. c. Reciprocal co-immunoprecipitation of matrin3 followed by Western blot analysis confirmed interactions of matrin3 with Pit1. d. Co-immunoprecipitation of HA-tagged, overexpressed Pit1 protein from 293T cells confirms interaction of Pit1 protein with endogenous matrin3 and hnRNP-U proteins (nt, non-transfected). e, Super-resolution image (100x, OMX DeltaVision) of immunostaining with α-matrin3 antibody in GC cells reveals “structure-like” matrin3 network. f. ChIP-Seq analysis reveals that ~50% of matrin3 DNA binding sites co-localize with the enhancer mark, H3K4me2, in GC cells. g, Meta-analysis plot of ChIP-Seq data shows that Pit1 bound enhancers co-localize with matrin3, H3K4me1, H3K4me2 and H3K27Ac, but not with a mark of silent chromatin, H3K27me3. X-axis is the distance from the Pit1 peak in base pairs. h. Western blot confirming matrin3 knockdown efficiency in GC cells used for matrin3 ChIP-Seq. i. Validation of matrin3 antibody specificity - qPCR analysis of ChIP signal on known Pit1 targets shows lack of matrin3 signal after matrin3 knockdown. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; **p<0.01, ***p<0.001) j. Single optical sections of immuno-FISH experiment in GC cells. From left: matrin3, GH-FIS, DAPI, merge - each staining shown separately black and white, and merged as matrin3 antibody staining in red and the GH locus labeled with a DNA probe in green.

2. Supplementary Figure 2. Co-localization and interaction of Pit-1 with β-catenin and Satb1.

a, Western blot showing that biotinylated BLRP-β-catenin is present in the nuclear and cytoplasmic fractions of GC cells. Biotinylated BLRP-β-catenin is visualized with streptavidin-HRP. b, Expression of GST-β-catenin fusion proteins used to map interaction with Pit-1 in GC cell extracts. c, Western blot validation of siRNA knockdown of β-catenin and Pit-1 proteins in GC cell samples used for IP-mass-spectrometry to identify Pit1 and βcatenin- interacting factors. d, RT-qPCR analysis of Satb1 mRNA expression in embryonic and adult mouse pituitary glands showing increased expression that parallels differentiation of Pit1-dependent cell types, n=3, +/− SD. e, Immunohistochemical staining of SATB1 and GH in the anterior lobe of an adult mouse pituitary showing co-localization at the cellular level. f, Western blot showing Co-IP of SATB1 with HA-tagged Pit1 in 293T cells. g, RT-qPCR analysis of GH nascent RNA levels in E16.5 mouse pituitaries harboring a Prop1-CRE conditional deletion of Satb1, +/− SD, *p<0.05, wt=4, cKO=12 h, Schematic representation of the LIS-extraction procedure used to isolate matrin-3-rich network. i, Pit-1 co-localizes with Satb1 and matrin-3 in the LIS-resistant insoluble fraction while β-catenin is present in both LIS-extracted and LIS-resistant insoluble fractions in GC cells. j, siRNA knockdown of β-catenin and Satb1 in GC cells does not affect either the subcellular distribution or the level of Pit1 protein. k, After siRNA knockdown of β-catenin and Satb1 followed by LIS extraction, Pit-1 is detectable in the “looped-out DNA” fraction. l, Immuno-FISH in GC cells before and after siRNA knockdown of β-catenin and Satb1. Single optical section of each experiment is presented. In merge: Matrin3 antibody immunostaining in red and GH loci labeled with a fluorescent green DNA probe.

3. Supplementary Figure 3. β-catenin and SATB1 have transcriptional effects on Pit1 targets.

a, Most β-catenin/Satb1-activated target genes are associated with Pit-1-bound enhancers in GC cells. b, Example picture of ChIP-Seq and GRO-Seq analysis of a Pit-1/Satb1 target gene - the NeuroD4 locus. c, Location of Pit1-dependent enhancers (in yellow) relative to the transcription start sites of target genes analyzed for expression and enhancer association following matrin3, β-catenin, Satb1 and/or Pit-1 siRNA knockdown. d, Heat-map plot of ChIP-seq tag distribution on Pit1 enhancers showing enrichment of Satb1 and β-catenin signal in center of enhancer. e, ChIP-qPCR analysis of Satb1 association with Pit1 enhancers upon either Pit1 or β-catenin knockdown. Experiments repeated 3 times, p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001) f, ChIP-qPCR analysis of matrin-3 association with Pit1 enhancers upon Pit1 knockdown. Experiments repeated 2–4 times, p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001) g, qPCR analysis showing significant change in expression of selected Pit1 target genes upon either Pit1 or Satb1 knockdown. (+/− SD; ***p<0.001). h, ChIP-qPCR analysis of matrin3 association with Pit1-dependent enhancers following siRNA knockdown of β-catenin or Satb1. Experiment repeated 2–3 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001). i, RT-qPCR analysis of relative expression of target gene mRNA following siRNA knockdown of matrin3.

4. Supplementary Figure 4. Pit-1 R271W mutant protein binds cognate DNA sites but does not associate with the nuclear LIS-resistant insoluble fraction.

a, RT-qPCR analysis of mRNA expression of Pit1-dependent target genes in GC cells. Both overexpressed, wild-type and R271W, Pit1 proteins compete for binding to recognition sites with endogenous Pit1. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; *p<0.05) b, Heat map of ChIP-Seq data showing that overexpressed HA-Pit1 WT and HA-Pit1 R271W bind to the same enhancer DNA recognition sites as endogenous Pit1 protein. c, Scattered dot plot of genome-wide peaks consistency analysis between the wild-type and R271W Pit1 to Pit1 enhancers with Pearson correlation coefficient shows a strong resemblance of the data. d, ChIP-Seq analysis comparing normalized tag density on GC samples containing overexpressed wild-type and R271W Pit1 protein. X-axis represents the number of base pairs from the center of the Pit1 peak. e, Both wild-type and R271W Pit1 protein express on similar levels and partition in the nuclear fraction. f, Fusion of the SAF matrix-association domain results in fractionation of R271W-SAF Pit1 protein in LIS-resistant insoluble fraction of GC cells g, Western blot showing equivalent expression of all HA-tagged Pit1 constructs in GC cells. asterisk – product of degradation of overexpressed Pit1 protein h, ChIP-qPCR (top panels) and RT-qPCR (bottom panels) analysis of effect of overexpression of different forms of Pit1 protein on p300 association and enhancer RNA (eRNA) expression, respectively, in the absence of endogenous Pit1. Experiments have been repeated 2–4 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; **p<0.01, ***p<0.001)

Acknowledgments

We thank M. Ghassemian from the University of California, San Diego (UCSD) Biomolecular/Proteomics Mass Spectrometry Facility for assistance with mass spectrometry; C. Nelson for cell culture assistance; J. Hightower for assistance with figure and manuscript preparation; T. Suter for help with images analysis. We acknowledge J. Santini and the UCSD Neuroscience Microscopy Shared Facility (Grant P30 NS047101) for imaging. These studies were supported by grants to M.G.R. from NIH. D.S-K was supported by EMBO Long Term Fellowship, The Swiss National Science Foundation and The San Diego Foundation. M.G.R. is an Investigator with HHMI.

Footnotes

The authors declare no competing financial interests.

Author Contributions: D.S-K and M.G.R. conceived the project. D.S-K performed the majority of the experiments; Q.M. performed the bioinformatic analyses; M.S. contributed 3D-ImmunoFISH. W.L. contributed GRO-Seqs; K.O. assisted in deep-sequencing library preparations and sequencing; Additional experiments/methods were contributed by K.S., J.T., Z.L. and D.N.; H.T. and K.S. assisted in animal based experiments. Y.K. and T.K-S. contributed critical insights and reagents. D.S-K., K.S. and M.G.R. wrote the manuscript. All authors discussed the results and commented on or edited the manuscript.

References

  • 1.McGinnis W, Levine MS, Hafen E, Kuroiwa A, Gehring WJ. A conserved DNA sequence in homoeotic genes of the Drosophila Antennapedia and bithorax complexes. Nature. 1984;308:428–433. doi: 10.1038/308428a0. [DOI] [PubMed] [Google Scholar]
  • 2.Scott MP, Weiner AJ. Structural relationships among genes that control development: sequence homology between the Antennapedia, Ultrabithorax, and fushi tarazu loci of Drosophila. Proc Natl Acad Sci U S A. 1984;81:4115–4119. doi: 10.1073/pnas.81.13.4115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Affolter M, Schier A, Gehring WJ. Homeodomain proteins and the regulation of gene expression. Curr Opin Cell Biol. 1990;2:485–495. doi: 10.1016/0955-0674(90)90132-x. [DOI] [PubMed] [Google Scholar]
  • 4.Prince KL, Walvoord EC, Rhodes SJ. The role of homeodomain transcription factors in heritable pituitary disease. Nat Rev Endocrinol. 2011;7:727–737. doi: 10.1038/nrendo.2011.119. [DOI] [PubMed] [Google Scholar]
  • 5.Heintzman ND, Ren B. Finding distal regulatory elements in the human genome. Curr Opin Genet Dev. 2009;19:541–549. doi: 10.1016/j.gde.2009.09.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Zeitz MJ, Malyavantham KS, Seifert B, Berezney R. Matrin 3: chromosomal distribution and protein interactions. J Cell Biochem. 2009;108:125–133. doi: 10.1002/jcb.22234. [DOI] [PubMed] [Google Scholar]
  • 7.Nickerson J. Experimental observations of a nuclear matrix. J Cell Sci. 2001;114:463–474. doi: 10.1242/jcs.114.3.463. [DOI] [PubMed] [Google Scholar]
  • 8.Kohwi-Shigematsu T, et al. Genome organizing function of SATB1 in tumor progression. Semin Cancer Biol. 2013;23:72–79. doi: 10.1016/j.semcancer.2012.06.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Radovick S, et al. A mutation in the POU-homeodomain of Pit-1 responsible for combined pituitary hormone deficiency. Science. 1992;257:1115–1118. doi: 10.1126/science.257.5073.1115. [DOI] [PubMed] [Google Scholar]
  • 10.Zhu X, Wang J, Ju BG, Rosenfeld MG. Signaling and epigenetic regulation of pituitary development. Curr Opin Cell Biol. 2007;19:605–611. doi: 10.1016/j.ceb.2007.09.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Ingraham HA, et al. A tissue-specific transcription factor containing a homeodomain specifies a pituitary phenotype. Cell. 1988;55:519–529. doi: 10.1016/0092-8674(88)90038-4. [DOI] [PubMed] [Google Scholar]
  • 12.Nakayasu H, Berezney R. Nuclear matrins: identification of the major nuclear matrix proteins. Proc Natl Acad Sci U S A. 1991;88:10312–10316. doi: 10.1073/pnas.88.22.10312. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Romig H, Fackelmayer FO, Renz A, Ramsperger U, Richter A. Characterization of SAF-A, a novel nuclear DNA binding protein from HeLa cells with high affinity for nuclear matrix/scaffold attachment DNA elements. EMBO J. 1992;11:3431–3440. doi: 10.1002/j.1460-2075.1992.tb05422.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Olson LE, et al. Homeodomain-mediated beta-catenin-dependent switching events dictate cell-lineage determination. Cell. 2006;125:593–605. doi: 10.1016/j.cell.2006.02.046. [DOI] [PubMed] [Google Scholar]
  • 15.Su LK, Vogelstein B, Kinzler KW. Association of the APC tumor suppressor protein with catenins. Science. 1993;262:1734–1737. doi: 10.1126/science.8259519. [DOI] [PubMed] [Google Scholar]
  • 16.Graham TA, Weaver C, Mao F, Kimelman D, Xu W. Crystal structure of a beta-catenin/Tcf complex. Cell. 2000;103:885–896. doi: 10.1016/s0092-8674(00)00192-6. [DOI] [PubMed] [Google Scholar]
  • 17.Notani D, et al. Global regulator SATB1 recruits beta-catenin and regulates T(H)2 differentiation in Wnt-dependent manner. PLoS Biol. 2010;8:e1000296. doi: 10.1371/journal.pbio.1000296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Dickinson LA, Dickinson CD, Kohwi-Shigematsu T. An atypical homeodomain in SATB1 promotes specific recognition of the key structural element in a matrix attachment region. J Biol Chem. 1997;272:11463–11470. doi: 10.1074/jbc.272.17.11463. [DOI] [PubMed] [Google Scholar]
  • 19.Cai S, Han HJ, Kohwi-Shigematsu T. Tissue-specific nuclear architecture and gene expression regulated by SATB1. Nat Genet. 2003;34:42–51. doi: 10.1038/ng1146. [DOI] [PubMed] [Google Scholar]
  • 20.Mancini MG, Liu B, Sharp ZD, Mancini MA. Subnuclear partitioning and functional regulation of the Pit-1 transcription factor. J Cell Biochem. 1999;72:322–338. [PubMed] [Google Scholar]
  • 21.Mirkovitch J, Mirault ME, Laemmli UK. Organization of the higher-order chromatin loop: specific DNA attachment sites on nuclear scaffold. Cell. 1984;39:223–232. doi: 10.1016/0092-8674(84)90208-3. [DOI] [PubMed] [Google Scholar]
  • 22.Li W, et al. Functional roles of enhancer RNAs for oestrogen-dependent transcriptional activation. Nature. 2013;498:516–520. doi: 10.1038/nature12210. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Kipp M, et al. SAF-Box, a conserved protein domain that specifically recognizes scaffold attachment region DNA. Mol Cell Biol. 2000;20:7480–7489. doi: 10.1128/mcb.20.20.7480-7489.2000. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Aravind L, Koonin EV. SAP - a putative DNA-binding motif involved in chromosomal organization. Trends Biochem Sci. 2000;25:112–114. doi: 10.1016/s0968-0004(99)01537-6. [DOI] [PubMed] [Google Scholar]
  • 25.Reeves R. Molecular biology of HMGA proteins: hubs of nuclear function. Gene. 2001;277:63–81. doi: 10.1016/s0378-1119(01)00689-8. [DOI] [PubMed] [Google Scholar]
  • 26.Xu L, et al. Signal-specific co-activator domain requirements for Pit-1 activation. Nature. 1998;395:301–306. doi: 10.1038/26270. [DOI] [PubMed] [Google Scholar]
  • 27.Li S, et al. Dwarf locus mutants lacking three pituitary cell types result from mutations in the POU-domain gene pit-1. Nature. 1990;347:528–533. doi: 10.1038/347528a0. [DOI] [PubMed] [Google Scholar]
  • 28.Gibcus JH, Dekker J. The hierarchy of the 3D genome. Mol Cell. 2013;49:773–782. doi: 10.1016/j.molcel.2013.02.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Bickmore WA, van Steensel B. Genome architecture: domain organization of interphase chromosomes. Cell. 2013;152:1270–1284. doi: 10.1016/j.cell.2013.02.001. [DOI] [PubMed] [Google Scholar]
  • 30.S’Iakste NI, S’Iakste TG. Transcription factors and the nuclear matrix. Mol Biol (Mosk) 2001;35:739–749. [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

1. Supplementary Figure 1. Matrin-3 co-localizes with enhancers in GC cells.

a, Validation of Pit1 antibody specificity - qPCR analysis of ChIP signals on two known targets, the GH and NeuroD4 enhancers shows lack of Pit1 signal after Pit1 knockdown. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; ***p<0.001) b. Western-blot confirming Pit1 knockdown efficiency in samples used to assess Pit1 antibody specificity. c. Reciprocal co-immunoprecipitation of matrin3 followed by Western blot analysis confirmed interactions of matrin3 with Pit1. d. Co-immunoprecipitation of HA-tagged, overexpressed Pit1 protein from 293T cells confirms interaction of Pit1 protein with endogenous matrin3 and hnRNP-U proteins (nt, non-transfected). e, Super-resolution image (100x, OMX DeltaVision) of immunostaining with α-matrin3 antibody in GC cells reveals “structure-like” matrin3 network. f. ChIP-Seq analysis reveals that ~50% of matrin3 DNA binding sites co-localize with the enhancer mark, H3K4me2, in GC cells. g, Meta-analysis plot of ChIP-Seq data shows that Pit1 bound enhancers co-localize with matrin3, H3K4me1, H3K4me2 and H3K27Ac, but not with a mark of silent chromatin, H3K27me3. X-axis is the distance from the Pit1 peak in base pairs. h. Western blot confirming matrin3 knockdown efficiency in GC cells used for matrin3 ChIP-Seq. i. Validation of matrin3 antibody specificity - qPCR analysis of ChIP signal on known Pit1 targets shows lack of matrin3 signal after matrin3 knockdown. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; **p<0.01, ***p<0.001) j. Single optical sections of immuno-FISH experiment in GC cells. From left: matrin3, GH-FIS, DAPI, merge - each staining shown separately black and white, and merged as matrin3 antibody staining in red and the GH locus labeled with a DNA probe in green.

NIHMS604603-supplement-1.tif (1.1MB, tif)
2. Supplementary Figure 2. Co-localization and interaction of Pit-1 with β-catenin and Satb1.

a, Western blot showing that biotinylated BLRP-β-catenin is present in the nuclear and cytoplasmic fractions of GC cells. Biotinylated BLRP-β-catenin is visualized with streptavidin-HRP. b, Expression of GST-β-catenin fusion proteins used to map interaction with Pit-1 in GC cell extracts. c, Western blot validation of siRNA knockdown of β-catenin and Pit-1 proteins in GC cell samples used for IP-mass-spectrometry to identify Pit1 and βcatenin- interacting factors. d, RT-qPCR analysis of Satb1 mRNA expression in embryonic and adult mouse pituitary glands showing increased expression that parallels differentiation of Pit1-dependent cell types, n=3, +/− SD. e, Immunohistochemical staining of SATB1 and GH in the anterior lobe of an adult mouse pituitary showing co-localization at the cellular level. f, Western blot showing Co-IP of SATB1 with HA-tagged Pit1 in 293T cells. g, RT-qPCR analysis of GH nascent RNA levels in E16.5 mouse pituitaries harboring a Prop1-CRE conditional deletion of Satb1, +/− SD, *p<0.05, wt=4, cKO=12 h, Schematic representation of the LIS-extraction procedure used to isolate matrin-3-rich network. i, Pit-1 co-localizes with Satb1 and matrin-3 in the LIS-resistant insoluble fraction while β-catenin is present in both LIS-extracted and LIS-resistant insoluble fractions in GC cells. j, siRNA knockdown of β-catenin and Satb1 in GC cells does not affect either the subcellular distribution or the level of Pit1 protein. k, After siRNA knockdown of β-catenin and Satb1 followed by LIS extraction, Pit-1 is detectable in the “looped-out DNA” fraction. l, Immuno-FISH in GC cells before and after siRNA knockdown of β-catenin and Satb1. Single optical section of each experiment is presented. In merge: Matrin3 antibody immunostaining in red and GH loci labeled with a fluorescent green DNA probe.

NIHMS604603-supplement-2.tif (2.5MB, tif)
3. Supplementary Figure 3. β-catenin and SATB1 have transcriptional effects on Pit1 targets.

a, Most β-catenin/Satb1-activated target genes are associated with Pit-1-bound enhancers in GC cells. b, Example picture of ChIP-Seq and GRO-Seq analysis of a Pit-1/Satb1 target gene - the NeuroD4 locus. c, Location of Pit1-dependent enhancers (in yellow) relative to the transcription start sites of target genes analyzed for expression and enhancer association following matrin3, β-catenin, Satb1 and/or Pit-1 siRNA knockdown. d, Heat-map plot of ChIP-seq tag distribution on Pit1 enhancers showing enrichment of Satb1 and β-catenin signal in center of enhancer. e, ChIP-qPCR analysis of Satb1 association with Pit1 enhancers upon either Pit1 or β-catenin knockdown. Experiments repeated 3 times, p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001) f, ChIP-qPCR analysis of matrin-3 association with Pit1 enhancers upon Pit1 knockdown. Experiments repeated 2–4 times, p-values calculated using student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001) g, qPCR analysis showing significant change in expression of selected Pit1 target genes upon either Pit1 or Satb1 knockdown. (+/− SD; ***p<0.001). h, ChIP-qPCR analysis of matrin3 association with Pit1-dependent enhancers following siRNA knockdown of β-catenin or Satb1. Experiment repeated 2–3 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; *p<0.05, **p<0.01, ***p<0.001). i, RT-qPCR analysis of relative expression of target gene mRNA following siRNA knockdown of matrin3.

NIHMS604603-supplement-3.tif (1.1MB, tif)
4. Supplementary Figure 4. Pit-1 R271W mutant protein binds cognate DNA sites but does not associate with the nuclear LIS-resistant insoluble fraction.

a, RT-qPCR analysis of mRNA expression of Pit1-dependent target genes in GC cells. Both overexpressed, wild-type and R271W, Pit1 proteins compete for binding to recognition sites with endogenous Pit1. Experiment repeated 2 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; *p<0.05) b, Heat map of ChIP-Seq data showing that overexpressed HA-Pit1 WT and HA-Pit1 R271W bind to the same enhancer DNA recognition sites as endogenous Pit1 protein. c, Scattered dot plot of genome-wide peaks consistency analysis between the wild-type and R271W Pit1 to Pit1 enhancers with Pearson correlation coefficient shows a strong resemblance of the data. d, ChIP-Seq analysis comparing normalized tag density on GC samples containing overexpressed wild-type and R271W Pit1 protein. X-axis represents the number of base pairs from the center of the Pit1 peak. e, Both wild-type and R271W Pit1 protein express on similar levels and partition in the nuclear fraction. f, Fusion of the SAF matrix-association domain results in fractionation of R271W-SAF Pit1 protein in LIS-resistant insoluble fraction of GC cells g, Western blot showing equivalent expression of all HA-tagged Pit1 constructs in GC cells. asterisk – product of degradation of overexpressed Pit1 protein h, ChIP-qPCR (top panels) and RT-qPCR (bottom panels) analysis of effect of overexpression of different forms of Pit1 protein on p300 association and enhancer RNA (eRNA) expression, respectively, in the absence of endogenous Pit1. Experiments have been repeated 2–4 times, p-values were calculated using the student’s two tailed t-test. (+/− SD; **p<0.01, ***p<0.001)

NIHMS604603-supplement-4.tif (940.3KB, tif)

RESOURCES