Main

The HUSH transcriptional silencing complex contains three core subunits, the transcription activation suppressor (TASOR), M-phase phosphoprotein 8 (MPP8) and periphilin 1 (PPHLN1), as well as the associated subunits that are transiently present. The complex binds to RNA at target loci, where it recruits the H3K9me3-specific histone methyltransferase SETDB1 and MORC2, which maintains epigenetic silencing through chromatin compaction. The core complex subunits were discovered by the Lehner laboratory in a screen for the epigenetic regulation of variegated expression levels of green fluorescent protein (GFP)1. Both TASOR, previously known as FAM208A, and PPHLN1 had earlier been proposed as being in complexes with direct H3K9me3-binders, among them MPP8 (ref. 2). The Lehner laboratory showed that the three proteins assemble into a stable protein complex, named HUSH, which silences reporter constructs integrated into the genome in proximity to heterochromatin (Table 1 and Fig. 1)1. The phenomenon of abnormal juxtaposition of a euchromatic gene with heterochromatin to cause its transcriptional silencing in a stochastic manner, termed position effect variegation (PEV), was identified in 1930 (ref. 3). However, although other earlier identified suppressors of PEV, such as the H3K9 histone methyltransferase SU(VAR)3–9 and its interactor and H3K9me2/3 reader heterochromatin associated protein 1 (HP1), are highly conserved across diverse eukaryotes, including humans, several of the core components of the HUSH complex are absent in lower eukaryotic model organisms, a potential reason why the complex was not identified before 2015. Like its human homolog, murine Fam208a (also known as D14Abb1e or MommeD6), had earlier been suggested to play a role in transgene silencing4. MPHOSPH8 (also known as MPP8), initially discovered as a highly phosphorylated protein during M phase5, was known to bind H3K9me3 in vitro6 and in vivo7. Given that H3K9me3 is a main feature of heterochromatin, a potential role as an epigenetic repressor seemed natural. Originally, PPHLN1 was identified as an insoluble nuclear protein present in keratinocytes8, and it was suggested to be a transcriptional co-repressor involved in cell-cycle regulation9.

Table 1 HUSH core complex and associated members
Full size table
Fig. 1: Schematic of the architecture of the HUSH complex.
figure 1

The HUSH core complex consists of the chromodomain-containing protein MPP8, the RNA-binding protein PPHLN1 and TASOR, which bridges the interaction between MPP8 and PPHLN1. Depletion of either core complex subunit destabilizes the two other proteins. By contrast, removal of associated members does not affect the stability of the core complex subunits. The core complex partners interact with H3K9 tri-methyltransferase SETDB1 and its associated protein ATF7IP, and with the ATPase domain containing chromatin remodeler MORC2 to repress its viral and genomic targets. Assembly of the HUSH core complex depends on the interaction between the C terminus of MPP8 and the DomI domain within TASOR, as well as between TASOR residues 1014–1095 and the C terminus of PPHLN1 (HUSH assembly-dependent regions are highlighted in red). In addition to these regions, assembly-independent regions contribute to HUSH function, as the N terminus of PPHLN1 and different domains within the TASOR N terminus (highlighted in yellow). DomI, domain I.

Full size image

Since its original discovery as a regulator of PEV, a body of work has highlighted a crucial role for the HUSH complex in transcriptional silencing. Specifically, transcribed LINE1s (long interspersed nuclear elements class 1) have emerged as the major HUSH targets. LINE1 is a family of transposable elements that comprise 17% of the human genome10. Although the majority are inactive, several have retained their ability to retrotranspose, which can lead to the instability and evolution of genomes. Thus, the finding that HUSH silences LINE1 elements suggests that the complex has a crucial role in maintaining genome integrity. In the following sections, we will discuss the composition and molecular architecture of HUSH, its genomic targets composed of endogenous retroelements, integrated viral DNA, the mechanisms by which HUSH is targeted to these genomic regions and how it silences transcription. Finally, we will discuss the role of HUSH complex members in mammalian development, antiretroviral immunity, and diseases such as cancer.

Composition and architecture of the HUSH complex

Core complex members

MPP8 is the structurally best-characterized HUSH core subunit and has been shown to exist as two isoforms of 860 and 867 amino acids, respectively. They differ by 41 amino acids in their C terminus but are otherwise identical. Isoform 1, described here (Fig. 2), contains an N-terminal chromatin organization modifier (chromo) domain (residues 59–118) and four C-terminal ankyrin repeats (amino acids 600–728). Similarly to the chromodomain-containing proteins polycomb (Pc) and HP1 that can bind to histone H3 peptides trimethylated at lysine 27 (H3K27me3)11 and lysine 9 (H3K9me3)12,13, respectively, the chromodomain of MPP8 has been shown to bind H3K9me3 peptides via a hydrophobic pocket formed by three aromatic residues (amino acids F59, W80 and Y83)14,15. Alanine substitution of the tryptophane residue has also been shown to reduce the MPP8-heterochromatin interaction in vivo7.

Fig. 2: Domain structures of the HUSH core complex and associated catalytic subunits.
figure 2

Schematic representation of the domain structures of H. sapiens HUSH core subunits MPP8, PPHLN1 and TASOR and associated catalytic subunits SETDB1 and MORC2. For each protein, the specific isoform shown and its UniProt identifier are indicated in brackets. The x axis indicates the amino acid position within the respective protein sequence. Below the x axis, the annotated or predicted domains of each protein, including the suggested binding sites for interaction partners and functionally critical regions for transgene silencing, are highlighted. The y axis shows the ‘predictor of natural disordered regions’ (PONDR, VSL2) score, which indicates the probability for a disordered region. S5, transducer S5-like domain in which CC1 is embedded; ATPase domain, formed by the GHKL ATP binding domain and the transducer S5-like domain; TCD, Tudor-chromodomain. The red dashed line indicates a disordered region prediction score of 0.5. Crystal structures were downloaded from the Protein Data Bank (PDB): 3LWE (MPP8 chromodomain), 6TL1 (TASOR DUF3715), 6SWG (TASOR–periphilin core complex), 3DLM (SETDB1 Tudor domain) and 5OF9 (MORC2 residues 1–603). Square brackets indicate the region for which a certain function has been shown. Double-headed arrows indicate interaction partners.

Full size image

The N-terminal region of MPP8 containing the chromodomain was further implicated in binding H3K9 mono- and di-methyltransferase G9A-like protein (GLP) and DNA methyltransferase DNMT3A7,15, raising the possibility of an interplay between H3K9 and DNA methylation at the genomic regions to which MPP8 associates. Interestingly, although HP1 dimerizes through its C-terminal chromoshadow domain16, in vitro studies have suggested that the first β-strand within the chromodomain of MPP8 can mediate the homodimerization of two MPP8 proteins14,15, leading to the hypothesis that an MPP8 dimer constitutes the center of a DNMT3A–MPP8–MPP8–GLP silencing complex. However, MPP8 dimerization and its functional relevance for DNMT3A–GLP recruitment have yet to be addressed in vivo.

Previous studies have also suggested a role of the MPP8 chromodomain17,18 for its interaction with the HUSH-associated proteins SETDB1 and ATF7IP (Fig. 1). Although a detailed map of the MPP8–SETDB1 interaction sites remains to be established, the chromodomain of MPP8 has been shown to recognize an H3K9-like sequence in ATF7IP methylated by G9A/GLP19.

The chromodomain and structured C-terminal region of MPP8 are connected by an intrinsically disordered region (IDR; amino acids 110–504). Immediately downstream of the chromodomain, a small region (amino acids 112–263) has been shown to bind the catalytic core of histone deacetylase SIRT1 (residues 254–489)20. The MPP8–SIRT1 interaction was suggested to promote target gene silencing in a reciprocal manner involving SIRT1-dependent antagonization of PCAF1-mediated MPP8 acetylation at lysine 439 (ref. 20). However, the importance of this interaction in HUSH silencing is not clear.

Ankyrin repeat domains (ARDs) are generally defined as repeating motifs of a β-hairpin-helix-loop-helix (β2α2) structure and are mostly known for mediating protein–protein interactions21. However, the three-dimensional (3D) structure of this region in MPP8 has not been determined. Other than a potential contribution to TASOR binding, as suggested by in vitro reconstitution experiments22, no activity has so far been reported for the ARD. Interestingly, recent findings provide compelling evidence that a structured region without domain annotation downstream of the ARD (amino acids 729–860) contributes to TASOR binding18,22 and fulfills critical roles in transgene silencing22 and the maintenance of stem-cell self-renewal18. Additionally, a small, structured region upstream of the ARD (residues 500–560) seems to be necessary but not sufficient for TASOR binding and transgene silencing22. Future studies focusing on structural and biochemical properties of these C-terminal regions will be crucial to shed light on mechanisms mediated by MPP8.

TASOR is the largest subunit of the HUSH complex, with the longest of at least four isoforms being 1,670 amino acids long (Fig. 2). Recent biochemistry studies have placed TASOR at the center of the HUSH core complex18,22. The ordered yet structurally uncharacterized region, termed DomI (residues 525–633), mediates the TASOR–MPP8 interaction22, and TASOR residues 1014–1095 contribute to PPHLN1 binding23. Supporting a crucial role for these TASOR regions in assembling the HUSH core complex is the fact that both regions are essential for transgene silencing22. The N-terminal half of TASOR also contains a domain of unknown function (DUF3715, amino acids 106–332) and a Spen paralog and ortholog C-terminal (SPOC) domain (amino acids 350–505), both of which, interestingly, are involved in transgene silencing in a HUSH assembly-independent manner22. From its crystal structure, it has become clear that DUF3715 is a non-catalytic PARP domain implicated in transgene silencing, endogenous LINE1 repression and H3K9me3 deposition, through its extended β8–β9 loop (residues 303–316, particularly Y305)22. Another residue in DUF3715, L130 (corresponding to amino acid L135 in Homo sapiens), has been shown to be essential for PEV and early embryonic development in mice24. In its C terminus, bioinformatics analysis predicted two domains, which, based on their structural homology to yeast RNA-induced initiation of transcriptional silencing (RITS) complex subunit Chp1, were named DomII/PIN (residues 1233–1456)22. Being dispensable for HUSH-dependent transgene repression, the functional importance of the TASOR C terminus remains unclear.

The PPHLN1 gene encodes at least eight isoforms, with isoform 2 (374 amino acids) being the best characterized and a component of the HUSH complex (Fig. 2). Interestingly, insertions of retroviral elements seem to constitute evolutionary drivers for the transcript diversity observed for PPHLN1 (ref. 25). PPHLN1 was originally identified as an insoluble protein localized in the nucleus, containing a putative nuclear localization signal (residues 110–116), as well as in the cell periphery of differentiated keratinocytes, co-localizing with the cornified envelope protein periplakin via amino acids 8–31 (ref. 8). Protein insolubility was suggested to be conferred by the disordered N-terminal sequence (residues 1–127), which, intriguingly, was found to contribute to HUSH-dependent silencing of transgenes and H3K9me3 methylation of endogenous loci23. In the context of cell-cycle regulation, PPHLN1 is believed to play a role in S-phase progression in a SIN3A/HDAC1-dependent manner9,26. An N-terminal region of PPHLN1 (residues 32–86) has been suggested to mediate the interaction with the SIN3A/HDAC1 co-repressor complex9. Whether the function of PPHNL1 in cell-cycle progression involves direct transcriptional regulation and depends on its association with the HUSH complex remains to be determined. The C-terminal half of PPHLN1 has been predicted to be composed of helical heptad repeats and proposed to be involved in homodimerization8. More recently, this region was identified to mediate TASOR binding and transgene silencing23. The crystal structure of the minimal PPHLN1–TASOR core complex (PPHLN1 amino acids 292–367 and TASOR residues 1014–1095) has shed more light on the roles of the C-terminal region in PPHLN1 (ref. 23). Three hydrophobic residues (L326, L333 and I337) are required for homodimerization of PPHLN1 in vitro, packing two PPHLN1 helical hairpins against each other. The importance of PPHLN1 dimer formation in vivo, however, remains to be clarified. At the PPHLN1–TASOR interface, residue L356 of each PPHLN1 molecule is essential to contact a distinct region in TASOR (residues 1014–1052 and 1072–1093, respectively). Alanine substitution of L356 abolishes transgene repression, arguing that the PPHLN1–TASOR interaction is crucial for HUSH function.

As described above, PPHLN1 was recently shown to bind endogenous RNA27, and it has been suggested that the disordered N terminus of PPHLN1 targets the HUSH complex to transcriptionally permissive regions via non-specific RNA binding23. However, a more precise definition of how the N-terminal region of PPHLN1 is involved in mediating sequence-specific binding remains to be elucidated. Importantly, PPHLN1 recognizes its RNA targets in the absence of SETDB1 and before deposition of H3K9me3, supporting a mechanism for HUSH target recognition that does not, or at least does not entirely, depend on the interaction between the MPP8 chromodomain and H3K9me3 (ref. 27). In addition to the direct PPHLN1–RNA interaction, TASOR may bind RNA indirectly via its association with proteins of the RNA-binding machinery22.

HUSH-associated proteins

In addition to the core components, HUSH associates with additional proteins (Fig. 1). Unlike knockout of TASOR, MPHOSPH8 and PPHLN1, knockout of associated members SETDB1 (ref. 1) and MORC2 (ref. 28) does not influence the core protein levels and vice versa (for detailed reviews on the general roles of SETDB1 and MORC2, see refs. 29,30).

The histone H3K9 tri-methyltransferase SET domain bifurcated histone lysine methyltransferase 1 (SETDB1) was found associated with the three core subunits1. SETDB1 is one of five mammalian histone methyltransferases catalyzing the methylation of H3K9 (refs. 31,32,33,34,35) (Fig. 2). Although no homologs in Drosophila melanogaster and Caenorhabditis elegans have been characterized for HUSH core complex subunits, the SETDB1 homologs, Egg and MET-2, respectively, are well described36,37 (Table 1). The SETDB1 N terminus consists of a triple Tudor domain (TTD) and a putative methyl-CpG binding domain (MBD), which, however, lacks methylated DNA-binding capacity in vitro38. Although Tudor domains are known to recognize methylated lysine and arginine residues39,40, and recent evidence supports the binding of SETDB1 TTD to H3K9me3 tails additionally acetylated at lysine 14 (refs. 41,42), it is yet to be determined whether this interaction contributes to HUSH function. The catalytic activity is mediated by the bifurcated SET domain, containing an evolutionary conserved insertion of 347 amino acids, and the adjacent cysteine-rich regions (pre-SET and post-SET)31.

The chromatin remodeler MORC family CW-type zinc-finger 2 (MORC2), identified in a CRISPR/Cas9 screen for human PEV repressors, was shown to associate with the HUSH core complex28. MORC2 is expressed as two isoforms, with isoform 2 (970 amino acids) only differing from isoform 1 (1,032 amino acids, described here) by missing the first 62 amino acids (Fig. 2). The microrchidia (MORC) protein family comprises four members in human cells—MORC1/2/3/4—and contains an ATPase module consisting of a gyrase, a histidine kinase and MutL (GHKL) domain and an S5 domain43. In addition, members of the MORC family contain a CW domain and coiled-coil motifs. Genetic complementation experiments have demonstrated the requirement of the catalytic activity, the CW domain and all three coiled-coil domains for MORC2 transgene silencing activity28,44. In contrast, a predicted Tudor-chromodomain, which is absent in other MORC family members, has been shown to be dispensable for this function28,44. The second coiled-coil region (CC2, residues 548–603) is thought to mediate the interaction to TASOR and MPP8 (ref. 28), while the other coiled-coil regions and the CW domain may contribute to transgene repression in an assembly-independent manner. A recent X-ray crystal structure of the first 551 amino acids of MORC2 (ref. 45) showed that the first coiled-coil domain (CC1, residues 282–361) is embedded in the S5 domain of the ATPase module and that it can bind dsDNA via its positively charged arginine residues (amino acids R326, R329 and R333) in vitro. Notably, dimerization of the ATPase module was found to contribute to HUSH-mediated repression in an ATPase activity-independent manner. Accordingly, a mutation preventing dimer formation (Y18A) abolishes HUSH-mediated repression of transgenes, while a constitutive dimerization mutation (S87L) leads to increased repression. Hence, ATP-dependent dimerization of two DNA-bound MORC2 molecules and potential subsequent DNA loop formation and chromatin compaction may regulate HUSH-dependent transgene silencing. Unlike the CW domain in MORC3, which has been shown to bind H3K4me3 peptides46, the CW domain of MORC2 contains a degenerate aromatic cage45, preventing it from binding methylated lysines. Instead, perturbation of the CW–ATPase interface results in HUSH hyperactivation, pointing towards a function in regulating MORC2 ATPase activity rather than promoting chromatin binding. However, it is still discussed whether the MORC2 CW domain has residual histone-binding activity in vivo, because the purified mouse MORC2A CW domain was found to bind H3 peptide independently of its K9 methylation status44. The molecular basis of the contribution of the third coiled-coil (CC3) domain to HUSH-mediated repression is unknown.

Based on current knowledge, a model emerges where different regions and domains of the proteins in the HUSH complex contribute to its roles in the repression of transgenes and endogenous loci (Fig. 1). However, to obtain a better understanding of the contributions of the HUSH core and its associated subunits would require the determination of the structures of critical regions, such as the N-terminal region in PPHLN1 and TASOR and the C-terminal part of MPP8, and to investigate their contribution to processes mediated by the HUSH complex in vivo.

Molecular insights into HUSH-mediated transcriptional silencing

Targets of the HUSH complex

LINE1s and other transposable elements

HUSH silences the transcription of LINE1s and thereby also their retrotransposition47,48. Transposable elements make up almost half of the human genome49. Earlier studies have linked the HUSH-associated proteins, SETDB1 and MORC family members, to transposon silencing. SETDB1 is well known for its repressive effect on endogenous retroviruses (ERVs) in mouse embryonic stem (mES) cells50,51, and mouse and Arabidopsis thaliana MORC proteins silence transposable elements in the murine male germline and plants, respectively52,53. LINE1 elements are the only autonomously active retrotransposons in the human genome, with an estimated 80–100 copies retaining the potential to mobilize (reviewed in ref. 10). LINE1 retrotransposition requires two proteins, open reading frame 1 (ORF1) and ORF2. These are only expressed by transcriptionally active full-length LINE1s. Notably, although LINE1s account for ~17% of the human genome, HUSH is only observed at a small fraction of these, primarily targeting active full-length LINE1 classes (as L1PA1–L1PA5)47. Indeed, ablation of the HUSH subunits in human embryonic stem (hES) cells or K562 human leukemia cells leads to an increase in the transcription of evolutionary young L1PA families and ORF1 protein production, suggesting that the HUSH complex selectively represses repetitive elements with retained mobilization potential47. Consistent with this, HUSH also silences active LINE1 and ERV subfamilies in mice, targeting the evolutionary youngest LINE1 classes in mES cells, such as L1Md-A and L1Md-T, as well as intracisternal A-particle (IAP) elements18,44,47,48. Notably, a recent study, using a diverse set of reporter constructs, demonstrated that HUSH can silence a broad range of intronless genes independent of their unique sequence27. Together, these results suggest that HUSH distinguishes endogenous (mostly intron-containing) genes from foreign intronless DNA and ‘host genome-integrated cDNAs’, such as the LINE1 elements.

Zinc-finger protein (ZNF/ZFP) genes

In addition to transposable elements, ZNF genes have also been identified as endogenous targets for HUSH-mediated transcriptional repression1,28,47 (Fig. 3a). Because they contain introns, the defining feature for ZNF genes to be preferentially targeted by HUSH has yet to be understood. The ZNF genes bound by HUSH are predominantly genes coding for KRAB-ZNF (KZNP) family members that can be actively transcribed47. Interestingly, chromatin immunoprecipitation followed high-throughput sequencing (ChIP–seq) profiling demonstrated specific HUSH enrichment at their 3′ ends22,47. The increased length of the 3′ exons of ZNF genes may therefore contribute to PPHLN1-dependent HUSH recruitment (for details, see the HUSH recruitment to unintegrated viral DNA section)27. HUSH seems to exert a direct transcriptional effect on KZFP genes, with HUSH removal resulting in increased expression levels, concomitant with decreased levels of H3K9me3 (refs. 1,28).

Fig. 3: Models for HUSH complex recruitment and HUSH-mediated transcriptional repression.
figure 3

a, Foreign and genomic DNA subject to HUSH-mediated silencing. HUSH transcriptionally represses both unintegrated DNA as well as the integrated provirus. Endogenous genomic regions targeted by the HUSH complex constitute transposable elements (TEs), particularly evolutionary young LINE1 elements, as well as ZNF genes and rDNA. L1PA elements are broadly bound by HUSH subunits, with MORC2 and MPP8 binding being skewed toward the 5′ end and 3′ end, respectively, and ZNF genes being bound at their 3′ end. The ChIP–seq profiles were adapted from ref. 47. b, Factors involved in promoting HUSH chromatin association/targeting. PPHLN1 binding to transcripts mediates HUSH recruitment to LINE1s. SETDB1-catalyzed H3K9me3 interacting with the chromodomain of MPP8 recruits the HUSH complex in the context of PEV. ZNF protein NP220 mediates direct HUSH recruitment to unintegrated DNA. c, Model for HUSH-mediated silencing. Epigenetic silencing is mediated by MORC2-dependent chromatin compaction, SETDB1-catalyzed H3K9 trimethylation. ATF7IP promotes nuclear accumulation and stabilization of SETDB1. G9A or GLP histone methyltransferases contribute to the interaction between MPP8 and ATF7IP. RNA decay is catalyzed by the NEXT complex and CNOT1. At unintegrated murine viral DNA, NP220 mediates the recruitment of the HUSH complex and a role for histone deacetylation has been implicated in this context. H3ac indicates histone H3 acetylation as detected by a pan-acetyl H3 antibody. The single-headed arrow indicates placement of a histone modification, and the double-headed arrow indicates interaction partners. The red arrow indicates an inhibitory effect and the brown arrow the inhibition of transcription.

Full size image

Ribosomal DNA

HUSH has also been suggested to be involved in the cohesin-dependent transcriptional repression of ribosomal genes (Fig. 3a) upon double-strand break (DSB)-induced damage of ribosomal DNA (rDNA)54. In this context, depletion of MPP8 or cohesin complex member SMC1, which was found to interact with PPHLN1 in response to DSB damage, were shown to lead to decreased levels of SUV39H1/H2-dependent H3K9me3 on rDNA. However, it remains to be determined whether rDNA is directly repressed by HUSH in a cohesin-dependent manner and whether the observed increase in H3K9me3 levels upon damage is a direct consequence of HUSH-mediated silencing.

Integrated and unintegrated viral DNA

Experimental data have suggested that certain viruses are silenced by a HUSH-dependent silencing mechanism following their integration into host DNA (Fig. 3a). For example, HUSH has been reported to prevent the expression of integrated HIV-1 reporter viruses in Jurkat T cells, a cellular model for HIV-1 latency1, and to restrict the spread of replication-competent HIV-1 viruses55,56. Moreover, the HUSH complex has been shown to silence unintegrated murine retroviral DNA by a mechanism involving its recruitment by the zinc-finger protein NP220 (ref. 57).

In summary, recent studies have expanded the repertoire of target regions that are silenced by the HUSH complex (Fig. 3a). HUSH is recruited to LINE1s and other long, transcribed, adenine-rich, intronless DNA regions, and these are the main molecular targets for HUSH-mediated silencing. In addition, HUSH is recruited to both integrated and unintegrated viral DNA, ZNF genes and rDNA, leading to their silencing.

Regulation of HUSH chromatin association/targeting

HUSH targeting to LINE1s, transgenes and endogenous genes

Several studies have supported a role for RNA-guided HUSH recruitment (Fig. 3b), including transcription-dependent recruitment of HUSH subunits to a LINE1 transgene47. Moreover, both PPHLN1 and TASOR have been identified in the mRNA interactome in human cell lines58,59, and studies have demonstrated PPHLN1 binding to specific RNA species, with full-length, evolutionary young LINE1 transcripts showing the highest enrichment27. Notably, PPHLN1 was also found to preferentially bind other transcripts derived from long, intronless, adenine-rich protein-coding genes and non-LINE1 overlapping, processed pseudogenes27. Hence, PPHLN1 binding specificity may enable HUSH to discriminate intron-containing coding genes from invading intronless mobile elements, such as LINE1s, hence distinguishing ‘self’ from ‘non-self’. Yet, the discrimination is not absolute, as PPHLN1 also binds transcripts from some intron-containing protein-coding genes, including ZNF genes. Consistent with HUSH repression correlating with transgene length27, the above-average length of the last exon may determine PPHLN1 binding to ZNF genes.

HUSH recruitment to unintegrated viral DNA

In the context of unintegrated viral DNA repression, the zinc-finger protein NP220 (also known as ZNF638 in human and ZFML in mouse) was demonstrated to mediate the recruitment of the HUSH complex and is required for transcriptional silencing57. Because NP220 was also identified as a TASOR interactor in HeLa cells22 and an MPP8 interactor in mES cells18, NP220 may also be involved in regulating HUSH binding to specific sites of the genome; however, so far, this suggestion has not been supported by published results.

HUSH targeting in the context of PEV

The chromodomain of MPP8 was originally suggested to mediate the genomic localization of the HUSH complex in PEV1. This H3K9me3-dependent recruitment mechanism was based on the observations that the expression of chromodomain mutant MPP8 prevented the initial establishment of reporter repression in HeLa cells, and that silenced reporter integrations were enriched in proximity to H3K9me3-rich regions1.

However, although an interaction between H3K9me3 and the MPP8 chromodomain could be a plausible explanation for HUSH recruitment to heterochromatin-proximal areas, this is not supported for HUSH recruitment to its physiological targets. These targets are located within open euchromatin, and the binding of HUSH is regulated by transcription of intronless DNA following the recruitment to PPHLN1 binding to nascent RNA, or by the recruitment of NPP20 to unintegrated viral DNA.

How does the HUSH complex silence transcription?

Although the original models suggested that HUSH simply represses its targets via cycles of MPP8-dependent reading and SETDB1-mediated writing of H3K9me3, later studies have shown that the mechanism by which HUSH silences transcription is more complex (Fig. 3c).

MORC2 has been identified as a key downstream mediator of HUSH activity28. Knockout of MORC2 phenocopies the ablation of HUSH core subunits and triggers the upregulation of evolutionary young LINE1 elements, ZNF genes and other elements28,44,47. Moreover, MORC2 was found to bind LINE1 elements selectively at their 5′ end44,47. Importantly, MPP8/TASOR/PPHLN1 triple-knockout cells show impaired binding of MORC2 to a subset of its target sites overlapping SETDB1-mediated H3K9me3, suggesting HUSH-dependent recruitment of MORC2 to these regions28. Together with the known activity of MORC2, this suggests that HUSH-dependent silencing is mediated by the ATPase domain-dependent chromatin remodeling activity of MORC2 (refs. 28,44).

Although the interaction between MPP8 chromodomain and the H3K9me3 mark is required for the initial establishment of HUSH repression of integrated vectors in PEV1, the importance of this interaction for the maintenance of repression seems to be context-dependent. In fact, complementation of MPP8 knockout cells with chromodomain mutant MPP8 partially restores HUSH-mediated reporter silencing1. Moreover, N-terminally truncated mouse MPP8 (amino acids 112–858) was shown to mediate LINE1 repression and sustain HUSH-dependent mES cell survival without detectable binding to chromatin18. Similarly, a chromodomain-truncated human MPP8 was found to be sufficient for GFP reporter silencing22. Therefore, the current available results do not support a role for H3K9me3 in maintaining HUSH-mediated repression. Despite mediating target repression for mouse LINE1 elements, MPP8 (amino acids 112–858) fails to maintain H3K9me3 levels, indicating that once silenced H3K9me3 is dispensable for the maintenance of HUSH-dependent silencing18. Hence, although H3K9me3 appears to be required for establishing silencing, MPP8 chromodomain-independent interactions with chromatin and/or RNA may be sufficient to maintain silencing of HUSH targets.

The well-known SETDB1 interacting protein, activating transcription factor 7 interacting protein (ATF7IP)60,61, has also been shown to be essential for HUSH-mediated repression as ATF7IP knockout cells fail to silence HUSH-repressed reporters in PEV and phenocopy endogenous transcriptional changes, including derepression of ZNF genes and the loss of H3K9me3 seen in SETDB1 knockout cells62. The ATF7IP–SETDB1 interaction may enable HUSH-mediated functions by promoting SETDB1 retention inside the nucleus63 and preventing proteasomal degradation of nuclear SETDB1 (ref. 62) (Fig. 3c). Moreover, the G9A/GLP-dependent MPP8/ATF7IP interaction was shown to contribute to efficient transgene silencing19.

For unintegrated retroviral DNA, HDACs are recruited by NP220, independently of the HUSH complex, and have been shown to contribute to transcriptional silencing57. Although the putative role of MORC2 as a transcriptional repressor in gastric cancer potentially requires the activity of its interaction partners HDAC1 (ref. 64) or HDAC4 (ref. 65), a general involvement of HDACs in HUSH-mediated transcriptional repression has not been shown.

Recent results have also suggested that HUSH cooperates with RNA decay components to restrict transposable element expression. In one study, MPP8 was shown to interact with the zinc-finger protein ZCCHC8 (ref. 66), a component of the nuclear exosome targeting (NEXT) complex, which primarily degrades short, non-polyadenylated transcripts67,68,69. This binding led to the recruitment of NEXT to HUSH-bound regions, resulting in the degradation of non-full-length transcripts arising from premature transcription termination of transposable elements that otherwise would evade repression by HUSH66. These results suggest that transcriptional and post-transcriptional silencing cooperate to provide a fail-safe mechanism for the efficient silencing of HUSH-mediated repression of transposable elements (Fig. 3c).

Another study showed that TASOR interacts with several RNA degradation factors, among them CNOT1, which catalyzes the shortening of the poly(A)-tail at the mRNA 3′ end, as well as NEXT complex component MTR4 (ref. 70). Functionally, simultaneous depletion of TASOR and CNOT1 by short interfering RNAs (siRNAs) led to increased levels of RNA as measured from a HeLa-integrated HIV-1 reporter, while only having limited effects on nascent transcription. These results suggest that HUSH and CNOT1 may cooperate to repress HIV at a post-transcriptional level; however, the importance of the HUSH–CNOT1 interaction for destabilizing endogenous HUSH target RNA remains to be established.

In summary, epigenetic silencing by HUSH is broadly mediated by its associated members, comprising SETDB1-catalyzed H3K9 trimethylation and MORC2-dependent chromatin compaction. The involvement of HDAC-dependent histone deacetylation and RNA decay catalyzed by the NEXT complex and CNOT1 appear context-dependent, and their general role in HUSH-mediated transcriptional silencing needs further investigation.

HUSH complex members in mammalian development, antiretroviral immunity and disease

The different HUSH subunits have critical roles during mouse development and in antiretroviral immunity. Moreover, their overexpression and mutation have linked them to disease, including cancer. Here we will summarize the different roles that have been proposed for HUSH (Fig. 4 and Table 2).

Fig. 4: Biological roles of the HUSH complex.
figure 4

a, HUSH represses transcription from invading viral DNA, such as latent-stage HIV. In the replicating stage, HUSH-mediated repression was found to be counteracted by provirus-encoded Vpx protein, which mediates the proteasomal degradation of the HUSH complex. This results in replication of the provirus. b, HUSH subunits are essential for the development of the mouse embryo and maintenance of self-renewal of mouse embryonic stem cells (Table 2). c, MPP8 was found to be overexpressed in different types of cancer, with MPP8 knockout resulting in decreased migration and invasion of colon and gastric cancer cells as well as metastasis of melanoma cells. FGFR2–PPHLN1 fusion events are detected in 16% of intrahepatic cholangiocarcinoma samples. An oncogenic role was suggested for the TASOR(I374V) variant in patients with rectal adenocarcinoma. The HUSH complex was suggested as a therapeutic target in AML, with MPP8 depletion increasing mouse survival. ICM, inner cell mass; Ub, ubiquitin.

Full size image
Table 2 Reported roles of HUSH complex members in development and disease
Full size table

Role in embryonic development and pluripotent stem cells

Mouse genetic studies have demonstrated essential roles for HUSH complex subunits in early embryonic development. Mice deficient for Pphln1 do not survive beyond embryonic day 7.5 (E7.5), which suggests an essential role before or during gastrulation71. Similarly, inactivation of Tasor results in peri-gastrulation lethality, showing defects in primitive streak elongation and the progression of the epithelial-to-mesenchymal transition24,72. Interestingly, this phenotype is partially rescued in Trp53 heterozygous mice72, suggesting that at least part of the phenotype in Tasor knockout mice is due to a DNA damage response.

Recently, we showed that MPP8 is essential for sustaining the self-renewal of ground-state pluripotent mES cells18. Inactivation of MPP8 in mES cells cultured in 2i/LIF results in cell-cycle arrest in G1-phase and spontaneous differentiation. Equally, Pphln1 and Tasor are vital for ground-state mES cells. Indeed, maintenance of the proliferative state depends on the integrity of the HUSH core complex and correlates with the efficient repression of evolutionary young LINE1 elements. Hence, although DNA methylation is crucial to safeguard committed cells from aberrant LINE1 expression73, the HUSH complex may fulfill this role in ground-state pluripotent stem cells, which are considered to be DNA-hypomethylated. In agreement with this hypothesis, metastable mES cells, in which LINE1 elements display substantially higher levels of DNA methylation73,74, are less affected by the absence of MPP8, but are highly dependent on MPP8 when Dnmt1/3a/3a are deleted18. Notably, maintenance of self-renewal by MPP8 is independent of its chromodomain binding to H3K9me3 and SETDB1-dependent maintenance of H3K9me3, indicating that the HUSH complex maintains mES cell self-renewal in a H3K9me3-independent manner. MPP8 knockout mice are born at lower ratio (8% instead of the expected 25%) and with reduced body size and weight75, and in a different study were characterized to show complete embryonic lethality before E11.5 (ref. 76), suggesting an important function for MPP8 in mouse development.

Consistent with a key role in peri-implantation development, Setdb1-deficient mice die at E3.5–E5.5, and SETDB1 is required for mES cell growth77. Studies using conditional Setdb1−/− mES cells have linked this defect to SETDB1 negatively regulating the expression of endogenous retroviruses by catalyzing H3K9 methylation50. Indeed, 69 different long terminal repeat (LTR) retrotransposons, primarily belonging to the ERV1 and ERVK families, were transcriptionally de-repressed in Setdb1 knockout mES cells78. Interestingly, LINE1 elements were not strongly de-repressed in Setdb1−/− ES cells, supporting the observation that the HUSH complex can silence LINE1 elements independently of H3K9me3. The effect on LTR retrotransposons targeted by SETDB1 might explain why Setdb1−/− mice lethality temporally precedes that of Pphln1- and Tasor-depleted mouse embryos.

SETDB1 is expressed in primordial germ cells (PGCs) and is required for germline development79. Although high protein expression levels of TASOR were observed in testis24, it is not clear whether the HUSH complex contributes to germline formation. Like mES cells, germ cells display low levels of DNA methylation80. Therefore, in addition to Piwi-interacting RNAs (piRNAs), epigenetic complexes capable of repressing transposable elements may have a particularly important role in germ cells, preventing the formation of new transposable element insertions in the germline that can be directly transmitted to the offspring81. However, because inactivation of Setdb1 in early germ cells has little impact on LINE1 expression, it suggests that a potential role for HUSH in regulating LINEs in the germline would not be dependent on SETDB1.

Morc2a knockout mice die during embryogenesis at E13.5, and MORC2A was suggested to suppress the repetitive-like protocadherin gene cluster in a HUSH-dependent manner, thereby contributing to brain development76.

In summary, genetic studies have shown critical functions for the HUSH complex in pluripotent stem cells and mouse embryogenesis. Interestingly, Pphln1 and Tasor knockout mice do not develop past E7.5, whereas Mphosph8 mice develop until E11.5, suggesting a different requirement for the core complex members in HUSH-dependent silencing. Additional studies will be required to understand the molecular mechanisms by which the different subunits contribute to normal development, and to clarify which of these mechanisms are HUSH-dependent.

Role of HUSH in antiretroviral and innate immunity

To defend against infections, host cells have evolved diverse mechanisms. TASOR was identified in an siRNA screen to identify proteins restricting HIV replication, pointing towards a potential role in antiretroviral immunity82. As described above, TASOR was recently proposed to mediate its antiviral activity by directly repressing transcription from invading viral DNA through HUSH-mediated PEV1. Hence, HUSH might provide a cell-autonomous epigenetic mechanism to restrict provirus activation. Strikingly, provirus-encoded proteins were shown to counteract HUSH repression by mediating its proteasomal degradation, consistent with an evolutionary arms race between the virus and its host55,56,83.

Can the antiviral activity of HUSH be harnessed to treat HIV (for a more detailed discussion see ref. 84)? By suppressing HIV replication in activated CD4+ T cells, antiretroviral combination therapy effectively reduces the viral load to below the detection limit. This has markedly lowered mortality from AIDS and improved the quality of life of patients85. Yet, complete viral eradication remains difficult to achieve, as HIV can persist in a latent form in resting CD4+ long-lived memory T cells86. Aiming to make latently infected cells ‘visible’ for antiretroviral therapies, reactivation of HIV expression constitutes an attractive therapeutic strategy87. Therefore, inhibition of HUSH might represent a route for targeting latent HIV infections. Experiments addressing the potency of HUSH inactivation in patient samples and potential side effects arising from derepression of endogenous HUSH targets, such as LINE1s and KZFPs, are vital before further consideration of such an approach.

Moreover, in addition to its ability to repress exogenous genetic elements, recent studies have linked the HUSH complex to regulation of the human innate immune system88. In the absence of viral infection, HUSH depletion can induce a type I interferon response in primary human fibroblasts that depends on dsRNA sensors MDA5 and RIG-I.

Role of HUSH and associated proteins in disease

Several of the core HUSH complex subunits and associated proteins (Table 2) have been identified as putative oncogenes in different types of cancer. MPP8 is overexpressed in breast cancer7, gastric cancer89, osteosarcoma90, melanoma91 and non-small cell lung cancer (NSCLC)92, and other data, including a study in colon cancer93, suggests that knockdown of MPP8 decreases cell proliferation in cancer cell lines89,93. Mechanistically, studies have suggested that MPP8 overexpression inhibits apoptosis through involvement of the p53/BCL2 signaling pathway89,90,93, and in osteosarcoma and NSCLC, MPP8 overexpression has been linked to the suppression of the HOXA5 signaling pathway90,92. MPP8 has also been suggested to promote tumor growth in breast cancer via the transcriptional repression of epithelial genes (for example, Cdh1), resulting in epithelial-to-mesenchymal transition and therefore increased cancer cell migration and invasiveness7. In support of such a role for MPP8, knockdown has been shown to affect the migratory and invasion potential of colon and gastric cancer cells by downregulating N-cadherin89,93 and where MPP8 knockdown inhibits lung metastasis of melanoma cells in vivo91.

PPHLN1 and TASOR have also been found to be associated with cancer. Transcription analysis discovered FGFR2–PPHLN1 fusion events in 16% of intrahepatic cholangiocarcinoma samples94. The FGFR2–PPHLN1 fusion protein, containing FGFR2 residues 1–768 at its N terminus and PPHLN1 residues 24–367 at its C terminus, has been associated with transforming potential in vitro, which could be rescued by FGFR2 inhibition. These data show that the activity of FGFR2 is required; however, although the entire PPHLN1 protein is part of the fusion protein, and it therefore has the potential to be in a complex with the remaining members of the HUSH complex, it is not known whether the HUSH complex is required for the transforming activity. The TASOR(I374V) variant has been identified in ~6% of patients investigated for rectal adenocarcinoma and is associated with decreased five-year disease-free and overall survival95. The functional role of this conservative amino acid change is currently unknown.

Although these studies suggest a potential role for MPP8, PPHLN1 and TASOR in the development and maintenance of cancer, there is very limited knowledge regarding how they contribute. Moreover, it is not clear which, if any, of the described pro-oncogenic functions are mediated as part of the HUSH complex. A recent publication, studying acute myeloid leukemia (AML), has provided some insight into how MPP8 can contribute to cancer75. The authors showed that MPP8 contributes to both the development and maintenance of AML, and importantly that the protein is not required for normal hematopoiesis. Moreover, by investigating the expression of LINE1 elements, the authors showed that endogenous LINE1 elements were induced in the AML cells in which MPP8 was depleted. The cyclin-dependent kinase inhibitor p21 was also found to be induced in the MPP8-depleted AML cells, presumably due to a LINE1-induced DNA-damage response, and this induction was found to be required for the increased survival of the AML mice. Excitingly, this study suggests that MPP8 could be an interesting therapeutic target for the treatment of AML.

Conclusions and perspectives

Since the discovery of the HUSH complex in 2015, tremendous progress has been made in determining the structure and cellular functions of its initially largely uncharacterized core subunits MPP8, TASOR and PPHLN1 (summarized in Figs. 1 and 2); the genetic elements subjected to HUSH-dependent silencing and how these may be specifically selected and targeted (summarized in Fig. 3a,b); and the potential silencing mechanism(s) mediated by HUSH (summarized in Fig. 3c).

We have realized the following:

  • The HUSH complex is a major player in regulating the activity of unintegrated and integrated viral DNA, and transposable elements, such as LINEs.

  • The intricate mechanism by which HUSH regulates transcription is context-dependent, involving H3K9me3-dependent recruitment of the HUSH complex in PEV, PPHLN1 RNA-mediated recruitment to LINEs, and SETDB1, MORC2 and potentially HDACs in gene silencing. Moreover, recent results have also shown that HUSH cooperates with RNA decay components in gene silencing.

  • The HUSH core and associated complex components are involved in disease. Mechanistic insights into this involvement have been gained through studies showing that the HUSH complex silences retroviral elements, and that the derepression of these elements may lead to a DNA-damage response, which could be used in therapeutic approaches.

To gain further insights into the biological and pathological roles of the HUSH complex it will be essential to understand whether additional target genes/genetic elements are regulated by the HUSH complex. This could be helped by new technologies such as long-read sequencing DNA technologies allowing locus-specific resolution of repetitive elements96. Moreover, it will be important to further establish the mechanisms by which the HUSH complex silences genetic elements and contributes to disease, and to provide more insights into the potential of therapeutically targeting the complex. Ultimately, this knowledge may help to select the patients who could potentially benefit from HUSH inhibitors and design regimes for combination therapies. Indeed, recent results have demonstrated a role for the HUSH complex in suppressing tumor immunogenicity97, suggesting that HUSH inhibitors should be explored in combination with the use of checkpoint inhibitors98 or DNA hypomethylating agents99. These are indeed exciting times for this young and expanding research field.