Structure and functional mapping of the KRAB-KAP1 repressor complex

Abstract

Transposable elements are a genetic reservoir from which new genes and regulatory elements can emerge. However, expression of transposable elements can be pathogenic and is therefore tightly controlled. KRAB domain-containing zinc finger proteins (KRAB-ZFPs) recruit the co-repressor KRAB-associated protein 1 (KAP1/TRIM28) to regulate many transposable elements, but how KRAB-ZFPs and KAP1 interact remains unclear. Here, we report the crystal structure of the KAP1 tripartite motif (TRIM) in complex with the KRAB domain from a human KRAB-ZFP, ZNF93. Structure-guided mutations in the KAP1-KRAB binding interface abolished repressive activity in an epigenetic transcriptional silencing assay. Deposition of H3K9me3 over thousands of loci is lost genome-wide in cells expressing a KAP1 variant with mutations that abolish KRAB binding. Our work identifies and functionally validates the KRAB-KAP1 molecular interface, which is critical for a central transcriptional control axis in vertebrates. In addition, the structure-based prediction of KAP1 recruitment efficiency will enable optimization of KRABs used in CRISPRi.

Keywords: CRISPRi; H3K9me3; Krüppel-associated box; Transposable element; heterochromatin.

Figure 1. Structure determination of the ZNF93‐KAP1 core complex. See also Fig EV1

1. Domain organization of the crystallized complex. T4L, T4 lysozyme; B1, B‐box 1; B2, B‐box 2; CC, coiled‐coil.

2. 2F _o ‐ F _c electron density map for one asymmetric unit. The map is contoured at 1 σ and colored by domain as in (A), except the second CC, which is in red.

3. Superposition of the crystal structure of ZNF93 KRAB‐A (green) on the solution NMR structure of the KRAB‐A box from a mouse KRAB‐ZFP (UniProt A0A087WRJ1; orange; Saito et al, 2003). A sequence alignment of the two domains is shown below.

4. Superposition of the ZNF93 KRAB‐A crystal structure (green) and the AlphaFold2 prediction (Jumper et al, ; Varadi et al, 2022) of ZNF93 KRAB (colored by confidence score, pLDDT). A sequence alignment of the two domains is shown below.

5. Close‐up of the electron density (2F _o ‐ F _c map) for the wild‐type ZNF93 KRAB‐A domain, contoured at 1.0 σ. Residues chosen for mutation to methionine are highlighted in cyan (I11), pink (C20), orange (L28) or yellow (L40). The selenium sites located for each of these variants in anomalous Fourier maps are shown as spheres.

Figure EV1. Anomalous Fourier maps of selenomethionine derivatives of KAP1 RBCC in complex with four different ZNF93 methionine‐insertion mutants, related to Fig 1

1. A–D

   X‐ray diffraction data were collected at the selenium K absorption edge. Anomalous Fourier maps were contoured at 3.5 σ. The Fourier maps show the positions of selenium atoms of the selenium‐substituted methionine residues the KRAB and KAP1 RBCC domains. KAP1 RBCC in complex with ZNF93 KRAB I11M, (A), ZNF93 KRAB C20M, (B), ZNF93 KRAB L28M, (C), ZNF93 KRAB L40M, (D).

Figure 2. Structure of the ZNF93 KRAB domain of bound to KAP1 RBCC

1. Domain organization of the crystallized complex (as in Fig 1A).

2. Overall structure of the KRAB‐KAP1 complex. Domains are colored as in (A). Zn atoms are shown as gray spheres.

3. Structure of ZNF93 KRAB:KAP1 RBCCΔB1 complex predicted by AlphaFold‐Multimer (Evans et al, ; Mirdita et al, 2022).

Figure 3. Molecular details of the KRAB‐KAP1 interface. See also Fig EV2

1. The KRAB‐KAP1 complex is shown in surface representation, with the KRAB domain in its natural orientation (top) or rotated by 180° to reveal the interaction surface contacting KAP1 (bottom). Residues in the interface are colored according to their atomic properties using the YRB scheme (Hagemans et al, 2015).

2. KRAB domain Hidden Markov Model (HMM) logo (Mistry et al, 2021). Residue numbers below the logo refer to the KRAB‐A consensus sequence (to convert to ZNF93 residue numbers, add 2).

3. Closeups of key residues in the KRAB‐KAP1 interface. The corresponding positions in the KRAB consensus sequence are shown in the lower panels (logo residue numbers refer to the consensus sequence; add 2 to obtain ZNF93 sequence). 12 KRAB residues required for KAP1 binding (Tycko et al, 2020) are highlighted in the logos in cyan. Aliphatic residues forming part of the hydrophobic core of the KRAB‐KAP1 interface are highlighted in yellow.

Figure EV2. Amino acid sequence alignments of KRAB domains, related to Fig 3

1. The KRAB domains from ZNF93, ZNF10, and ZIM3. Fused to catalytically inactive Cas9, the KRAB domains of ZNF10 and ZIM3 are used for potent gene repression that can be programmed in a sequence‐specific manner via the CRISPRi approach (Gilbert et al, ; Thakore et al, ; Alerasool et al, 2020).

2. Alignment of unusual KRAB domains. Variant KRAB domains (vKRAB, as classified by (Helleboid et al, 2019), and two standard KRAB domains (sKRAB) which were non‐repressive in a high‐throughput screen (Tycko et al, 2020), are aligned to the KRAB domain of ZNF93. ZNF93 residues directly contacting KAP1 in the crystal structure are in bold; residues not contacting KAP1 are in gray. KAP1‐contacting residues that deviate from the ZNF93 sequence are highlighted in green, orange, or red, depending on whether the mutation is expected to be tolerated, moderately deleterious or highly deleterious, respectively. The right‐hand column shows the expectation of whether each KRAB domain will interact with KAP1, based on our structural analysis (green, likely to bind tightly; orange, may still bind; red, unlikely to bind). The repressive sKRAB domains of ZNF91 and ZIM3 are shown for reference.

Figure 4. Effects of point mutations in the KRAB binding site and KAP1 binding and silencing. See also Fig EV3

1. Surface plasmon resonance (SPR) KAP1‐KRAB binding assay. MBP‐KRAB was immobilized on the chip. WT or R311E KAP1 RBCC were flowed over the chip. Data points are shown in dark red or blue; fits are shown as light red or blue lines. See Fig EV3 for binding kinetics constants.

2. Position of the mutations in the KRAB‐KAP1 interface.

3. LINE‐1 reporter repression with single point mutants and a previously described KRAB binding‐deficient KAP1 variant containing four mutations in the CC domain (CCmut; V293S/K296A/M297A/L300S).

4. SVA reporter repression with the same set of mutants as in (B). Data were normalized to KAP1 KO cells transfected with an empty vector (EV).

Data information: In (C) and (D), data are presented as fold‐repression of reporter luciferase luminescence. Error bars represent standard error of the mean between measurements (n = 3). Data are representative of at least three independent (biological replicate) experiments. Lower panels: Western blots of cell lysates from KAP1 KO HEK293T cells transfected with each of the variants or empty vector. Uncropped blots available in Source Data. Source data are available online for this figure.

Figure EV3. Surface plasmon resonance (SPR) KAP1‐KRAB binding assay titration curves, related to Fig 4

Upper panel: SPR sensorgrams for KAP1 binding to immobilized MBP KRAB. The fits for the association and dissociation kinetics are shown in red. Lower panel: Data were fitted using a biphasic kinetic model with PRISM 9 (GraphPad) to determine rate constants (k _on, k _off) and binding affinities (K _d).

Figure EV4. Expression levels and genomic distribution of KAP1 variants in HEK239T cells, related to Fig 5

1. Heatmaps and summary plots illustrating KAP1 ChIP‐seq enrichment over H3K9me3 peaks (left) and protein‐coding gene promoters (right) genome wide, in KAP1 KO cells and complemented cell lines expressing WT KAP1 or KRAB binding‐deficient KAP1 variant K296S/M297S/L300S/V293S (CCmut). TSS, transcriptional start site. The summary plots illustrate mean coverage values (RPKM) for each sample minus the signal of the pooled input sample. Only reads mapping uniquely (MAPQ > 10) were retained. ChIP‐seq experiments were run in duplicate.

2. Example genome browser snapshots of KAP1 enrichment in KAP1 KO cells complemented cell lines expressing WT or CCmut KAP1. Top, a gene with KRAB‐dependent KAP1 binding at the 3′ end and KRAB‐independent KAP1 binding at the promoter. Bottom, an LTR transposon bound by KAP1 in a KRAB‐dependent manner. The KAP1 KO cells and a pooled input track are shown as controls. Only reads mapping uniquely (MAPQ > 10) were retained. Scales are in RPKM, reads per kilobase per million.

3. Western blot of WT, KAP1 KO, and KAP1‐complemented KAP1 KO HEK293T cells. These cells were used for CUT&RUN genomic profiling (Figs 5 and EV5).

4. Confocal immunofluorescence microscopy of the HEK293T cells used for CUT&RUN genomic profiling stained with anti‐KAP1 antibody and DAPI nuclear stain. Scale bars, 10 μm.

5. Co‐immunoprecipitation of KAP1 and SETDB1. KAP1 was immunoprecipitated from HEK293T KAP1 KO cells stably expressing 3xFLAG‐KAP1 (WT or CCmut) using ANTI‐FLAG M2 beads. The experiment was performed either in the presence or absence of the SUMO protease inhibitor N‐ethylmaleimide (NEM). Uncropped blots available in Source Data.

Source data are available online for this figure.

Figure 5. Genome‐wide CUT&RUN analysis of H3K9me3 distribution in cells expressing wild‐type and KRAB binding‐deficient variants of KAP1. See also Figs EV4 and EV5

1. Example genome browser snapshots of H3K9me3 distribution over the hg38 reference in the presence of different KAP1 variants. H3K9me3 distribution is shown at a HERVK9‐int element (upper) and a cluster of LINE‐1 elements (lower). A control IgG track from parent HEK293T cells is shown for comparison. WT, wild‐type; CCmut, KRAB binding‐deficient KAP1 variant. Only reads mapping uniquely (MAPQ > 10) were retained. Scales are in RPKM, reads per kilobase per million. Experiments were run in duplicate (biological replicates) with similar results.

2. Heatmaps and summary plots illustrating H3K9me3 levels over H3K9me3 peaks genome wide, in cells expressing different KAP1 variants. Peaks were called on Control cells using SEACR in stringent mode (Meers et al, 2019) against the IgG control, and only regions longer than 1 kb retained. The summary plots illustrate mean values for each sample. Only reads mapping uniquely (MAPQ > 10) were retained. Experiments were run in duplicate with similar results.

3. Pairwise quantifications of H3K9me3 CUT&RUN counts for KAP1‐complemented cells (n = 2) versus KAP1 KO cells (n = 2) over RepeatMasker retrotransposons (hg38) generated with DESeq2 (Love et al, 2014). Only reads mapping uniquely (MAPQ > 10) were retained. Red, H3K9me3 enriched in complemented cells; blue, H3K9me3 enriched in KAP1 KO cells. Enrichment cutoff: P < 0.05; log‐fold change ±3.

4. Heatmaps showing H3K9me3 CUT&RUN signal enrichment over full‐length (> 6 kb) LINE‐1 subfamilies (left, red), reference SVAs > 1 kb (center, green) and THE1 LTR elements > 1.5 kb (right, orange) in cells expressing different KAP1 variants. H3K9me3 is rescued upon complementation with WT KAP1 but not CCmut KAP1. Only reads mapping uniquely (MAPQ > 10) were retained; note that only the flanking regions of SVAs are mappable even with a 2 × 150‐bp paired end sequencing strategy. Experiments were run in duplicate with similar results.

Figure EV5. Genome‐wide analysis of H3K9me3 distribution in cells expressing wild‐type and KRAB binding‐deficient variants of KAP1, Related to Fig 5

1. Pairwise quantifications of H3K9me3 CUT&RUN counts for KAP1‐complemented cells (n = 2) versus KAP1 KO cells (n = 2) over reference H3K9me3 peaks called in the parental control cell line. Only reads mapping uniquely (MAPQ > 10) were retained. Gray indicates significant enrichment of signal (cut‐off: P < 0.0001; fold‐change > 3) in the complemented line.

2. Representative H3K9me3 distribution (over the hg38 reference) in the presence of different KAP1 variants at three different types of KAP1‐independent loci: an intronic LINE‐1 (L1PA) element bound by the HUSH complex (upper); a HUSH‐bound long exon (middle); and a centromeric region (lower). A control IgG track from parent HEK293T cells is shown for comparison. WT, wild‐type; CCmut, KRAB binding‐deficient KAP1 variant. Only reads mapping uniquely (MAPQ > 10) were retained. Scales are in RPKM, reads per kilobase per million. Experiments were run in duplicate with similar results.

3. Heatmaps and summary plots illustrating H3K9me3 levels over H3K9me3 peaks at HUSH‐bound loci (Douse et al, ; Seczynska et al, 2022), in cells expressing different KAP1 variants. Peaks were called on control cells using SEACR in stringent mode (Meers et al, 2019). The summary plots illustrate mean values for each sample. Only reads mapping uniquely (MAPQ > 10) were retained.

4. Quantitative comparison of H3K9me3 changes upon KAP1 knockout over loci regulated by HUSH (Douse et al, ; Seczynska et al, 2022) versus those regulated by KAP1 (as defined in this study). The central bands denote the medians. Boxes represent the interquartile range (IQR). Whiskers extend 1.5x IQR beyond the box. Statistical test: Wilcoxon rank sum test with continuity correction.

5. Bar plot illustrating the distribution of loci where H3K9me3 was not reduced in KAP1 KO cells compared with the parental control cell line (KAP1‐independent loci).