KAP1 targets actively transcribed genomic loci to exert pleomorphic effects on RNA polymerase II activity

KAP1 (KRAB-associated protein 1) is best known as a co-repressor responsible for inducing heterochromatin formation, notably at transposable elements. However, it has also been observed to bind the transcription start site of actively expressed genes. To address this paradox, we characterized the protein interactome of KAP1 in the human K562 erythro-leukaemia cell line. We found that the regulator can associate with a wide range of nucleic acid binding proteins, nucleosome remodellers, chromatin modifiers and other transcription modulators. We further determined that KAP1 is recruited at actively transcribed polymerase II promoters, where its depletion resulted in pleomorphic effects, whether expression of these genes was normally constitutive or inducible, consistent with the breadth of possible KAP1 interactors. This article is part of a discussion meeting issue ‘Crossroads between transposons and gene regulation’.


Introduction
KAP1/TRIM28 is a master regulator critical to processes such as stem cell self-renewal and DNA-damage response [1][2][3][4][5][6]. Its best-described function thus far is the binding through sequence-specific targeting Krüppel-associated box domain-containing-zinc finger proteins (KZFPs) to transposable elements (TEs), which leads to the transcriptional silencing of these genetic units [7][8][9][10][11][12]. At these sites, KAP1 acts as a scaffold protein for the formation of a heterochromatin-inducing machinery comprising HP1 (heterochromatin protein 1), the histone methyl-transferase SETDB1, the nucleosome remodelling and deacetylase (NuRD) complex and the histone demethylase KDM1A [9,[13][14][15][16][17]. Yet a growing number of studies additionally describe a diametrically different pattern of genomic recruitment for KAP1, with its accumulation at promoters of actively transcribed genes, independently of KZFPs or any silencing complex [1,13,[18][19][20][21][22]. The mediators and associated effectors, as well as functional consequences of KAP1 recruitment at these loci, are still debated. A model was proposed whereby KAP1 directly bound the DNA fibre and acted as an RNA polymerases II (PolII) stalling factor at gene promoters, where upon appropriate stimuli, it underwent phosphorylation of serine 824 (pS824), which resulted in releasing paused PolII, hence allowing or enhancing transcription of the underlying gene [19]. Another study centred on KAP1 recruitment at gene promoters also found it to correlate with paused PolII genome-wide, and while the molecular mechanisms of its recruitment remained unexplored, there was evidence of KAP1-mediated transcriptional control for inducible genes [21]. At promoters of these genes, KAP1, together with the 7SK and small nuclear ribonucleoprotein (snRNP) complex, appeared to function as a dynamic supplier of inactive positive transcription elongation factor b (P-TEFb), facilitating the transition from the initiation to the elongation phase of PolII-mediated transcription. While some of the findings reported in these two studies coincided, functional data strikingly diverged, since in the former KAP1 removal resulted in increased transcription of KAP1-targeted inducible genes, whereas in the latter, these genes were then repressed, while the expression of most of their housekeeping counterparts was unaltered.
Here, we present proteomic and functional data indicating that KAP1 can associate with a wide range of nucleic acid binding proteins, nucleosome remodellers and other transcriptional modulators, that it can associate with promoters recognized by PolII, and that its impact at these genetic loci is diverse, as predicted from the breadth of its protein interactome.

Results
(a) The interaction network of KAP1 is enriched in nucleic acid binding proteins We previously examined the interactome of KAP1 by immunoprecipitating the endogenous protein in whole-cell extracts of human embryonic stem and K562 erythro-leukaemic cells. While this approach revealed KAP1 association with proteins critical for DNA replication, such as PCNA, RPA1 and subunits of the MCM (MiniChromosome Maintenance) complex, it did not lend itself to the purification of sufficient amounts of material to allow cell fractionation. We thus turned to immunoprecipitation (IP) of a doubly tagged form of KAP1 in K562 cells, extracting the chromatin from the nuclear fraction and subjecting the products to mass spectrometry (MS/MS) analysis. This tandem affinity-purification, where the sequential use of two high-affinity antibodies increased specificity, revealed higher amounts of KAP1 in chromatin extracts, indicating that in the nucleus, this protein is mainly associated with DNA (figure 1a and electronic supplementary material, figure  S1A). Fractionation of precipitated complexes revealed a single elution peak extending over the first seven of 25 fractions of a linear 15 to 35% (v/v) glycerol gradient in both nuclear and chromatin extracts. Although not informative of the number of distinct KAP1-associated complexes, it suggests that these are of similar and relatively small molecular weight (MW), with their migration possibly affected by association with nucleic acids (electronic supplementary material, figure S1B). The very top categories of a gene ontology search based on molecular functions of KAP1 interactors included RNA-binding proteins, in addition to DNA-and chromatin-associated proteins ( figure 1b). Accordingly, a somewhat diffuse band of apparent MW below 20 kDa, detected on the silver stained  Proteins with a total peptide-coverage of 5% or higher, either in the chromatin or in the nuclear fraction, were selected. Lines linking the proteins highlight experimentally determined interactions (http://string-db.org). Disconnected nodes were excluded. Bold labels were used for KAP1 interactors experimentally validated through independent studies (https://thebiogrid.org/). Green labels were used for units of the nucleosome remodelling and deacetylase (NuRD) complex. KAP1 interactors previously detected in unfractionated human embryonic stem cells and K562 cells (threshold p-value < 0.01) [23] are indicated with blue asterisks. gel in the total nuclear but not the nuclease-treated chromatin soluble (CS) fraction, strongly suggested that nucleic acids were co-immunoprecipitated with KAP1 (figure 1a). Known partners of KAP1, such as components of the NuRD complex and HP1 proteins, associated with overexpressed KAP1 in K562 cells (figure 1c). These findings were in agreement with previous MS/MS studies performed by our group with the endogenous protein in human embryonic stem cells (hESC), where we found significant enrichment of the same cofactors [23] (electronic supplementary material, figure S1C,E). In addition, proteins part of the DNA replication and DNA repair machineries, such as components of the MCM protein complex, RuvB-like proteins 1 and 2 (RUVBL1, RUVBL2) and DNA-damage binding protein 1, were co-immunoprecipitated with KAP1 in both experimental settings [23] (electronic supplementary material, figure S1D,E and table S1). Finally, we also detected in these complexes a range of nucleosome remodellers, such as components of SWItch/Sucrose Non-Fermentable (SWI/SNF), and of RNA-binding proteins, namely members of the DEAD-box RNA helicases protein families (electronic supplementary material, figure S1D,E). In particular, both components of the remodelling and spacing factor (RSF), RSF1 and SMARCA5, a complex facilitating activator-dependent PolII transcription initiation [24], were co-purified with KAP1. Moreover, in our experimental settings, we detected a strong enrichment of DDX21, a helicase that functions as PolII transcription elongation factor, by facilitating the release of the P-TEFb from the 7SK snRNP complex [25], as well as genome stability guardian [26]. The breadth and functional diversity of the KAP1 interactome strongly suggests that this master regulator fulfils multifaceted functions (figure 1c).

(b) Extensive recruitment of KAP1 to PolII promoters
In the light of our MS/MS results and recent findings supporting a putatively direct role of KAP1 in PolII-mediated gene transcription, we sought to explore further correlations between KAP1 genomic recruitment and active transcription. For this, we chose the Hepa 1-6 murine hepatoma cell line, because we had previously documented in these cells an extensive overlap between KAP1 binding and histone marks associated with active gene expression [27]. Of note, we found the general features of KAP1 genomic recruitment to be otherwise similar in these cells to those documented in other settings, whether embryonic or differentiated cells from either human or murine origin [8,10,18,28,29]. Analysis of chromatin IP coupled with deep sequencing (ChIP-Seq) data in Hepa 1-6 cells confirmed a significant overlap of sites targeted by KAP1 with regions enriched in the H3K27ac and H3K4me1 active histone marks, and further revealed its co-localization with PolII at many sites (figure 2a). A closer examination indicated that this subset of KAP1-recruiting loci was devoid of the H3K9me3 repressive mark, which otherwise displayed a strong genome-wide correlation with KAP1 peaks at bona fide targets of the KAP1-SETDB1 complex, such as TE-derived sequences and the 3 0 -end of KZFP genes as previously noted in other systems (figure 2a and electronic supplementary material, figure S2A,B) [30]. We examined the distribution of KAP1 binding sites over selected gene features and observed a remarkable enrichment at promoters (approximately 25%), in agreement with the strong correlation with PolII-enriched regions (figure 2b) and consistent with data previously obtained in other cells [8,18,28,29]. We further observed that KAP1 was preferentially binding the promoters of highly transcribed genes, which in turn were characterized by a high PolII pausing index (PI), defined as the ratio between PolII at the transcription start site (TSS) region and over the gene body, in agreement to previously established methods [31] (figure 2c).

(c) Pleomorphic influences of KAP1 on PolII distribution and transcription
We then characterized the functional consequences of KAP1 recruitment at these promoters by comparing the transcriptional profiles of control and of Kap1 knockdown (KD) Hepa 1-6 cells (electronic supplementary material, figure S3A,B).
Overall, KAP1 depletion led to limited perturbations of gene expression, with a totals of only 145 downregulated and 134 upregulated PolII genes (figure 3a, left). However, we could not correlate either of these changes with specific alterations of PolII pausing index (figure 3a, right), determined in both basal and KD conditions. Furthermore, KAP1 binding in the proximity of a TSS was not predictive of the transcriptional deregulation of the corresponding gene in KAP1-depleted cells ( figure 3b, left). Similarly, the PI of KAP1-targeted promoters did not globally change upon KAP1 removal (figure 3b, right and electronic supplementary material, figure S3C). However, when groups of genes with an elevated (≥4) PI at baseline were analysed separately, a significant difference was detected ( figure 3c, left). This category of genes thus followed a previously reported pattern [32]. Nonetheless, when we separated genes within this group based on whether or not KAP1 was recruited over their promoter region, we observed no deviation from the initial trend (figure 3c, right), indicating that for the largest fraction of these genes, PI changes were independent of KAP1 proximal binding.
(d) Induction of early response genes is variably affected by KAP1 mild shifts in PolII distribution (figure 4c and electronic supplementary material, figure S4D).

Discussion
The body of KAP1 interactors reflects its recruitment predominantly to the chromatin fraction of the nucleus. In addition to its known partners in heterochromatin formation and transcriptional silencing, such as HP1 proteins and components of the NURD complex, our analysis of the KAP1 interactome in K562 leukaemic cells identified several factors involved in modulating the structure of euchromatin and the activity of RNA polymerases [24,25,[33][34][35][36][37]. This fits with the detection by several studies including ours of KAP1 at the promoters of numerous PolII-transcribed genes.    KAP1 depletion induced pleomorphic effects on the activity of its PolII promoter targets, and our data at least partly invalidated several recently advanced models [19,21]. Kap1 KD induced a general decrease in PI at genes characterized by a high PI at basal conditions, the vast majority of which was, however, not bound by KAP1 over the promoter region. Moreover, the detected changes in PI were globally of small magnitude and mostly uncoupled from differences in transcriptional levels. In the case of HSP genes, we noted that their epigenetic signature, characterized by strong enrichment in H3K9me3 and lack of active histone modifications, was diametrically different from the one found at other KAP1-bound promoters. Additionally, the high PI of these genes reflected the stalling and inactivity of PolII subunits at their promoters, while the bulk of KAP1-targeted promoters were characterized by high transcriptional activity in addition to high PIs. The expression level of a second group of KAP1-bound inducible genes, comprising transcriptionally active genes that were devoid of the repressive mark H3K9me3, was not shifted by the removal of KAP1 at baseline. Previous investigations illustrating the functional versatility of the regulator, both in its canonical KZFP-mediated TE targeting configuration [38] and in combination with a panel of transcription factors [28,39,40], focused on experimental systems where KAP1 acted either at discrete differentiation stages of selected tissues or in specific signalling pathways. Taken together, these findings indicate that inducible genes represent neither a homogeneous system nor a directly relevant model for studying the roles played by KAP1 at promoters of housekeeping and stably expressed genes, where the regulator must exert influences not detectable in our tissue culture system, or manifested only under exceptional circumstances, as suggested by the viability of mice devoid of KAP1 in the liver or in part of the brain [41,42]. MEFs wild-type (WT) and KO for Kap1 were cultured and generated as previously described [43] (strain C57BL/6 J). Murine hepatoma Hepa 1-6 and K562 cells were cultured using standard methods.

(b) Plasmids and lentiviral vectors
For KAP1 KD experiments, pLKO vector encoding shKAP1 and the empty vector as control were used. Forty-eight hours after transduction, infected cells were selected with 2 µg ml −1 puromycin in growth medium for an additional 72 h. Lentiviral vector production protocols are available at http:// tronolab.epfl.ch and backbones at Addgene (http://www. addgene.org). For the expression of FLAG-HA-KAP1, the human KAP1 partially codon-optimized sequence was cloned in the retroviral pOZ-N vector [44].

(c) Heat shock
Heat shock was performed by placing the culture dish over a 45°C water bath for 5 min, prior to RNA extraction.

(f ) ChIP-qPCR and ChIP-Seq
ChIP and library preparation were done according to [9], with modifications as described in electronic supplementary material, experimental procedures. Sequencing was performed in 100 bp reads run on an Illumina HiSeq 2500. Primer sequences used for ChIP-qPCR are provided in electronic supplementary material, experimental procedures.
(g) LC-MS/MS (liquid chromatography-tandem mass spectrometry) of FLAG-HA-tagged KAP1 KAP1 was purified from Dignam nuclear and chromatin extracts [45] derived from K562 cells stably expressing FLAG-HA-tagged KAP1 (eKAP1), as well as from K562 cells not expressing the construct by two-step affinity chromatography [44]. Nuclear and chromatin-soluble extracts (NE and CS, respectively) were incubated with anti-FLAG-conjugated agarose beads (A2220, Sigma) and the bound polypeptides were eluted with the FLAG peptide (Sigma) under native conditions. The affinity-purified material was then incubated with anti-HA-conjugated agarose beads (HAagarose beads, sc-7392 AC, Santa Cruz) and eluted using HA peptide (Roche) under native conditions. Five per cent of FLAG-HA immunoaffinity-purified eKAP1 or mock immunoprecipitations from 4 l of culture was resolved on SDS-PAGE (NuPage gel, 4-12% gradient, Novex, Life Technologies) and stained with the Silverquest kit (Invitrogen). The remainder of the eluate was precipitated using the ProteoExtract Protein precipitation kit (Calbiochem) and analysed by tandem MS at the Harvard Medical School Taplin Biological Mass Spectrometry facility, Boston, MA, USA. The list of proteins with the relevant MS/MS data is provided in electronic supplementary material, table S1. (i) Bioinformatics analyses and statistics R v. 3.1.2 (http://www.R-project.org) or GraphPad Prism v. 6.0 and 7.0 (http://www.graphpad.com) was used for statistical analyses and graphical representations of the data. H1 hESC MS/MS results [23] were visualized and handled through Scaffold (http://www.proteomesoftware.com/products/scaffold/), using the 'quantitative value' option, selecting hits with a royalsocietypublishing.org/journal/rstb Phil. Trans. R. Soc. B 375: 20190334 p-value < 0.01 and a cut-off of 10 in all replicates. Proteins were further selected by the code GO0005634 for nuclear proteins. The list of proteins with the relative quantitative MS/MS data is provided in electronic supplementary material, table S1.