Genomic Regulation of Transcription and RNA Processing by the Multitasking Integrator Complex

Abstract

In higher eukaryotes, fine-tuned activation of protein-coding genes and many non-coding RNAs (ncRNAs) pivots around the regulated activity of RNA polymerase II (Pol II). The Integrator complex is the only Pol II-associated large multi-protein complex that is metazoan-specific and has therefore been understudied for years. Integrator comprises at least 14 subunits, which are grouped into distinct functional modules. The phosphodiesterase activity of the core catalytic module is co-transcriptionally directed against several RNA species, including long non-coding RNAs (lncRNAs), U small nuclear RNAs (U snRNAs), PIWI-interacting RNAs (piRNAs), enhancer RNAs (eRNAs), and nascent pre-mRNAs. Processing of ncRNAs by Integrator is essential for their biogenesis, and at protein-coding genes Integrator is a key modulator of Pol II promoter-proximal pausing and transcript elongation. Recent studies have identified an Integrator-specific protein phosphatase 2A (PP2A) module, which targets Pol II and other components of the basal transcription machinery. In this Review, we discuss how the activity of Integrator regulates transcription, RNA processing, chromatin landscape, and DNA repair. We also discuss the diverse roles of Integrator in development and tumorigenesis.

Introduction

The eukaryotic basal transcription machinery revolves around the enzymatic activity of three distinct RNA polymerases¹. Shortly after the initial isolation of RNA polymerase II (Pol II) holoenzyme comprising its 12 core subunits², it became apparent that efficient transcription initiation and transcript elongation required a host of nuclear accessory factors³. In fact, the assembly of the Pol II holoenzyme into a pre-initiation complex (PIC) in vivo and in vitro requires six multi-subunit general transcription factors (GTFs)^4–6, and the co-activator Mediator complex (26 subunits in mammals)^7–9. The ensuing dynamics of protein allosteric changes, post-translational modifications and dissociation of PIC components that occur during the earliest steps of transcription is termed transcription initiation and constitutes the main rate-limiting step of protein-coding gene transcription in lower eukaryotes, such as the yeasts Saccharomyces cerevisiae and Schizosaccharomyces pombe.

Additional mechanisms and protein complexes that regulate gene expression post-initiation have evolved in multicellular organisms. In metazoans, an arrest of Pol II activity occurs at most protein coding genes (and some non-coding RNA (ncRNA) loci) immediately downstream to the transcription initiation site. This so-called ‘pausing’ of Pol II is a reversible state observed in most metazoans, which deadlocks Pol II while it is engaged in the initial transcription of a nascent, capped pre-mRNA^10–13. The paused holoenzyme derives from the PIC¹⁴, yet is distinct from the initiating holoenzyme: while parting ways with Mediator and several GTFs, Pol II associates with DRB-sensitivity inducing factor (DSIF) and negative elongation factor (NELF)^15,16. DSIF and NELF support pausing in multiple ways, for instance by preventing the association of core Pol II subunits with elongation factors such as the PAF complex and SPT6, and blocking the ribonucleotide entry funnel^17,18. Cyclin-dependent kinase 9 (CDK9), the catalytic subunit of positive transcription elongation factor b (pTEFb), targets several components of the paused Pol II to promote release of the holoenzyme from the transcription start site (TSS) region and elicit productive transcript elongation^15,19,20. Most notably, CDK9 phosphorylates NELF subunits (thereby promoting its dissociation), DSIF (triggering allosteric changes), and the C-terminal domain [G] (CTD) of Pol II at its Ser2 residues (Box 1).

Box 1. The C-terminal domain of RNA polymerase II.

RPB1, the catalytic subunit of RNA polymerase II (Pol II) has a large, disordered region known as the C-terminal domain (CTD)¹⁸². The CTD is a low-complexity domain of 378 aa, essentially composed of tandem heptad repeats of Y-S-P-T-S-P-S. Although the heptad is highly conserved from yeast to humans, the number of repeats varies from 26 in Saccharomyces Cerevisiae to 52 in human. The Pol II CTD is crucial for transcription regulation, through phosphorylation of its heptad repeats by multiple kinases from the PIC assembly steps to termination. The combination of phosphorylated residues is thought to function as a biochemical code that enables Pol II to recruit a broad range of protein complexes, including Mediator, Integrator, the capping machinery, the spliceosome and cleavage and polyadenylation specificity factor¹⁸². All the heptads can be phosphorylated in vivo, each carrying one or two modified residues on average¹⁸³. The most targeted residues are Ser2 and Ser5 and their phosphorylation is associated with Pol II pause release into productive elongation and with transcription initiation, respectively¹⁸³. Ser7 phosphorylation is commonly found at Pol II molecules across the bodies of coding and non-coding loci and it may have a role in 3’-end formation of U small nuclear RNAs⁶¹. As a disordered, low-complexity protein domain, the CTD has not been structurally characterized, with the exception of few repeats that have been modeled when bound to the Mediator and the Integrator complexes^32,36. Notably, Integrator has robust affinity for the CTD in both its unphosphorylated and phosphorylated forms^21,62 and the Int–PP2A module is capable of directly dephosphorylating Ser residues^31,33. Recent structural data also suggest that phosphorylated Tyr1 residues may enhance the affinity of Integrator for the CTD³².

Integrator was serendipitously discovered in 2005 as a new multi-subunit complex in human cells, capable of binding the Pol II CTD (Box 1)²¹. Mass spectrometry revealed peptides from 12 uncharacterized open reading frames associated to several Pol II subunits (Fig. 1). Orthologs for all subunits were identified throughout metazoans, but not in yeast, suggesting that the complex is unique to multicellular eukaryotes. Sequence homology-based annotation tools revealed a MBL/β-CASP [G] region within Integrator complex subunit 9 (INTS9) and INTS11, which are highly homologous to cleavage and polyadenylation specificity factor subunit 73 (CPSF73; also known as CPSF3) and CPSF100 (also known as CPSF2), respectively²², providing the first hint that INTS9 and INTS11 may be endowed with RNA endonuclease activity. In fact, depletion of either the largest subunit (INTS1) or the putative catalytic core (INTS11) of Integrator resulted in specific accumulation of unprocessed, precursor U small nuclear RNAs [G] (U snRNAs)²¹. U snRNA processing is known to occur co-transcriptionally, thereby connecting Integrator to Pol II activity. Thus, the complex responsible for cleaving U snRNA transcripts in metazoans was found at last.

Figure 1. Subunits of the mammalian Integrator complex.

The 15 known subunits of the mammalian Integrator complex are depicted to scale. Both INTS6 and its mutually exclusive paralog INTS6L are shown. All annotated protein motifs and domains are named and boxed. The availability of published structural data (cryo-EM or X-ray crystallography) is marked by an underline. Colored bars on top of each subunit diagram denotes a mapped interaction surface with another member of Integrator, protein phosphatase 2A (PP2A) or sensor of ssDNA (SOSS) complex, according to the specific color code. Lastly, a symbol on the left end indicates whether the corresponding subunit is a component of the endonucleolytic, Integrator–PP2A (Int-PP2A) or enhancer modules, or whether it is found in the SOSS DNA repair complex.

The crucial role of Integrator in U snRNA processing is conserved in many organisms, as first shown in 2011 in Drosophila melanogaster²³, followed by studies conducted in the nematode Caenorhabditis elegans²⁴ and the highly regenerative planarian Schmidtea mediterranea²⁵. Orthologs of INTS9 and INTS11 have also been identified in plants²⁶, but not in fungi, further corroborating the hypothesis that the Integrator complex is the focal point of a transcription regulation network that supports multicellularity. Beyond the original core of 12 subunits, Asunder (INTS13) and the von Willebrand factor A domain containing 9 (INTS14)²⁷ were identified as putative components of the Integrator complex using gene reporter assays, and later confirmed to be functionally and biochemically associated to Integrator^28–30. More recently, serine/threonine-protein phosphatase 2A (PP2A) has been found to stably associate with the Integrator complex, revealing new functions for Integrator in modulating transcription of protein coding genes through dephosphorylation of the Pol II CTD and of Pol II associated complexes such as DSIF^31–34.

From the initial focus on a handful of U snRNA genes, a decade of research has unveiled how Integrator is a genome-wide orchestrator of transcription regulation and RNA processing, being recruited to most protein coding genes and ncRNA loci. In this Review, we discuss the catalytic and non-catalytic functions of the Integrator complex, how they contribute to genome regulation, and how Integrator’s malfunction contributes to tumorigenesis in somatic cells and leads to a range of developmental disorders in germinal cells. Biochemical and structural evidence reveal Integrator as a modular, versatile molecular machine that can operate in either Pol II-dependent or Pol II-independent manners. The versatility of Integrator is based on a shared handle and adaptions to different tasks in multiple biological processes.

Structural insights into Integrator

A deeper understanding of the intricate biochemical processes that contribute to gene expression requires structural information on the many involved protein complexes. For instance, cryo-EM analysis of the transcription initiation machinery helped deconvolute how the Mediator complex operates during the earlier steps of transcription^35,36. Although the architecture of Integrator is far less characterized than that of Mediator, the recent influx of crystallography and cryo-EM data is depicting an outline of this large and elusive complex^28,29,32,33 (Fig. 2).

Figure 2. Structural features and assembly of Integrator.

A) The functional modules of Integrator are biochemically and structurally distinct. The enhancer module comprises INTS13 and INTS14, which stabilize one another when forming a heterodimer. INTS10 biochemically associates with the INTS13–INTS14 heterodimer, although structural information on INTS10 has not yet been obtained. The Int– serine/threonine-protein phosphatase 2A (PP2A) module comprises the PP2A heterodimer, which includes the scaffold subunit PP2A-A the smaller catalytic subunit PP2A-C, assembled on INTS5 and INTS8. INTS6 further bridges the PP2A dimer to INTS5–INTS8A. The endonuclease module can be reconstituted as a stable trimer of INTS11 (catalytic subunit), INTS9 (catalytic, inactive) and INTS4 (scaffold subunit). The sensor of ssDNA (SOSS) complex, which is implicated in DNA repair, is assembled around INTS3. Although a structure of the complete SOSS complex is not available, the N-terminal half of INTS3 has been crystallized with the small subunits of SOSS, nucleic acid-binding protein 1 (NABP1) or NABP2, and INTS3 and NABP interacting protein (INIP). The remaining C-terminal moiety of INTS3 has been independently crystallized as a homodimer, in association with a small C-terminal tail of INTS6.

B) The structure of the fully assembled Integrator complex with PP2A (INTAC) reveals a central backbone (INTS1, INTS2 and INTS7) accommodating the two catalytic modules (endonuclease and phosphatase) on opposite sides.

Protein Data Bank entries with structures of Integrator: 7CUN, 7PKS, 7BFQ, 7BFP, 6SN1, 7BV7, 6WLG, 4OWW.

Endonucleolytic core and backbone components

The sequence similarity between INTS11–INTS9 and CPSF2–CPSF3 led to speculation of close structural analogies between these two endonucleolytic modules recruited by Pol II^21,22,37,38. In fact, the catalytic core of Integrator adopts a conformation strikingly similar to that of CSPF: the MBL/β-CASP domains at the N-termini of INTS11 and INTS9 form two pseudo-symmetric lobes with outward-facing catalytic pockets (the catalytic site of INTS9 is inactive as it lacks zinc-binding ability)^28,33. Unlike existing CSPF structures, INTS11 was captured in an apparent inactive conformation, as the cleft leading to the active site is too narrow to accommodate RNA^28,33. However, there is evidence that binding of Integrator to the paused Pol II elongation complex containing NELF and DSIF, may elicit opening of the catalytic cleft³².

INTS4, the structural homologue of symplekin in the CPSF complex, folds around the catalytic lobes with its N-terminal HEAT repeats [G] and stabilizes their interaction (Fig. 2a). In fact, INTS4, INTS9 and INTS11 form a stable trimeric complex when co-expressed in insect cells^28,39. The C-termini of INTS11 and INTS9 comprise a robust dimerization domain with multiple contact points. In addition to stabilizing the heterodimer, INTS4 anchors the endonucleolytic module to the rest of the Integrator complex through two separate domains: its N-terminal HEAT repeats contact the largest subunit, INTS1, while a C-terminal helix bundle makes extensive contact with INTS7 (Ref.^32,33) (Fig. 1; Fig. 2b).

INTS1, INTS7 and INTS2 form the central backbone of the Integrator complex, making extensive contact with one another. During the biogenesis of Integrator, INTS1, INTS2 and INTS7 may form a nucleating core (Fig. 2a), onto which additional modular components assemble (Fig. 2b). Whereas the INTS2 and INTS7 structures were nearly fully recovered, only about a third of the large 220 kDa INTS1, comprising its C-terminus, was initially solved³³. Ensuing analyses of Integrator bound to Pol II yielded the structure of an additional central domain of INTS1, which binds INTS2, and captured the mobile N-terminal region of INTS1 through crosslinking mass spectrometry³². When bound to the paused Pol II elongation complex, Integrator makes extensive contacts with core Pol II subunits (RPB2 is contacted by an INTS1 N-terminus region, and RPB3 and RPB11 are engaged by INTS7 and INTS2, respectively), with DSIF (INTS11 interacts with SPT5) and with NELF (INTS6 interacts with NELF-B)³². Overall, Integrator wraps around the reconstituted paused Pol II through multiple contact points, including at the POL II CTD of RPB1 (Ref.³²). The Integrator endonuclease module is positioned close to the RNA exit site of Pol II, ready to accommodate and cleave nascent RNA about 20 nucleotides away from the active site³².

The Int–PP2A module and the INTAC complex

The isolation of the Integrator–PP2A complex (INTAC), from nuclear extracts of human 293T cells, revealed that subunits INTS5 and INTS8 are assembled on top of INTS1, INTS2 and INTS7 to form a large, T-shaped scaffold, thereby anchoring the newly identified phosphatase module of Integrator (Int–PP2A)³³. INTS5 and, particularly, INTS8 are crucial for tethering components of the phosphatase PP2A to Integrator^32,33 (Fig. 2b). PP2A is one of the most abundant serine/threonine phosphatases in mammalian cells and is generally found in two different compositions. PP2A dimers are assembled from a scaffold subunit (PP2A-A) comprised of a series of HEAT repeats arranged into a horseshoe-like alpha-solenoid shape, and a small globular catalytic subunit (PP2A-C)⁴⁰. Although they are stable in vitro, the heterodimers are not considered the functional phosphatase conformation in vivo. Instead, PP2A trimers incorporate an additional regulatory subunit (PP2A-B)^41,42, which confers target specificity and ultimately licenses PP2A to dephosphorylate a broad array of protein targets, from mitotic spindle components to intracellular signaling mediators⁴³.

The structure of INTAC, instead, reveals a PP2A heterodimer tethered to Integrator primarily by INTS8 (and INTS5), which makes extensive contacts with the N-terminal HEAT repeats of PP2A-A^32,33 (Fig. 2b). INTS6 is another central component of Int–PP2A: it binds PP2A (catalytic and scaffold subunits) and helps anchor the phosphatase to the Integrator backbone (Fig. 2b) while simultaneously contacting NELF³². Additionally, the INTS6 C-terminal domain may bind INTS3 (Fig. 2a; Fig. 1), although the INTS3 structure has not been solved yet with the fully assembled complex. Deletion of INTS6 results in loss of PP2A components from Integrator^31,33, whereas deletion of INTS8 results in loss of both PP2A components and INTS6 from Integrator³¹. Importantly, integrity of the remaining complex is preserved following disruption of the Int–PP2A module³¹. Surprisingly, Int–PP2A is devoid of the PP2A-B regulatory subunit, and structural models suggest that INTS6 and INTS8 would clash with its binding³³. Canonical PP2A complexes are deemed functional only in association with PP2A-B and it is unclear whether Integrator would be sufficient to confer target specificity. Furthermore, the inherent conformational flexibility of PP2A (especially of PP2A-A) and the lack of structural data on most PP2A-B proteins leaves the door open for a role of PP2A-B within Integrator.

The INTS13–INTS14–INTS10 ‘enhancer’ module

INTS13 and INTS14 were not initially recovered as stable Integrator subunits²¹. Subsequent studies identified both proteins as putative Integrator components based on a functional screening in D. melanogaster²⁷. Human INTS13 associates with the full Integrator–Pol II complex and is also found in a low molecular weight module in vivo alongside INTS14 and INTS10 (Ref.³⁰) (Fig. 2a). These three subunits can be, in fact, reconstituted in vitro^28,29 and, remarkably, INTS13 and INTS14 are found physically interlinked in a heterodimeric conformation that stabilizes both proteins and suggests they are co-translated²⁹. INTS10 binds the heterodimer at the N-terminal VWA domain [G] of INTS14 (Fig. 1), but its inherent conformational flexibility has precluded further structural analysis. There are structural analogies between a reconstituted INTS13–INTS14 heterodimer and the DNA repair Ku70–Ku80 heterodimer, which binds DNA double-strand break ends²⁹. In vitro data further suggest that INTS13, INTS14 and INTS10 have some ability to bind single-stranded RNA (ssRNA), double-stranded RNA and ssDNA²⁹. Functionally, the INTS13–14–10 module is implicated in transcription regulation as it binds cis-regulatory enhancer elements genome-wide³⁰, but seems largely dispensable for the endonucleolytic activity of Integrator at U snRNAs^29,30. Nonetheless, INTS13 appears to be connected to the cleavage module, which it binds through a conserved C-terminal stretch²⁹.

INTS3 and the SOSS complex

Following INTS1 and INTS2, the third largest subunit of Integrator, INTS3, is a 118kDa protein that is found also in the sensor of ssDNA (SOSS) complex with the small ssDNA-binding proteins nucleic acid-binding protein 1 (NABP1) or NABP2 and with the uncharacterized factor INTS3 and NABP interacting protein (INIP) (Fig. 2a). The SOSS complex is believed to sense ssDNA from DNA damage sites and a partial crystal structure of the complex has been obtained⁴⁴. The N-terminal half of INTS3 acts as a scaffold for the assembly of the ternary complex. Specifically, INTS3 is structured as two alpha helix domains, whose interface generates the C-shaped cavity that docks NABP1 or NABP2. INIP binds to a groove located at the opposite end of INTS3. NABP1 and NABP2, as predicted by the presence of an OB-fold domain [G]⁴⁵, can associate with short DNA oligos while assembled within the SOSS complex⁴⁴. The SOSS complex also associates with INTS6 or with its paralog INTS6L; these interactions seem to occur through the disordered C-terminal tail of INTS3⁴⁶ (Fig. 1; Fig. 2a). The C-terminal half of INTS3, devoid of the disordered tail, was independently expressed in bacteria and crystallized, revealing an elongated HEAT-repeat structure^47,48 (Fig. 1). In addition, there is evidence for homo-dimerization of the INTS3 C-terminal moiety (Fig. 2a). The INTS3 dimer can accommodate the 87 amino acid C-terminal tail of INTS6 (Ref.⁴⁷), which was previously identified as strictly required for INTS6 binding to SOSS. Additionally, the dimer generates a positively charged groove that has been proposed to bind ssRNA and ssDNA⁴⁷, supporting the intrinsic DNA binding ability of NABP1 and NABP2. It remains unclear whether SOSS exists in vivo as an independent complex or is a functional appendix of the fully assembled Integrator complex.

Taken together, the structural data suggest that Integrator is organized in distinct functional modules (Fig. 2). The two known enzymatic activities, endonuclease and phosphatase, latch onto a common scaffold and are physically separated by at least 150 Å. Future studies will elucidate the extent of crosstalk between these two modules. In addition to the catalytic cores, INTS13, INTS14 and INTS10 form a discrete subcomplex (the enhancer module) and INTS3 assembles with two small proteins implicated in DNA repair to form the SOSS complex. INTS6 is an essential component of the Int-PP2A phosphatase module and has also been reported as a SOSS component. Intriguingly, density maps of INTS3 assembled with the full Integrator could not be obtained, suggesting a high degree of INTS3 conformational flexibility. Additional efforts are needed to understand the conformation of this elusive subunit and its precise involvement in other Integrator modules.

Additional experiments are also needed to clarify how the enhancer module is recruited to the larger Integrator complex and its influence on the endonucleolytic module. Lastly, structural information is lacking for INTS12, a low molecular weight Integrator subunit that contains a PHD finger and thus may endow the complex with the ability to read histone methylation patterns.

Processing of non-coding RNAs

RNA cleavage is a key function of the Integrator complex. Like the CPSF complex, Integrator targets nascent RNAs and releases them from the transcription bubble. In this section, we discuss how the activity of Integrator is directed to a broad range of long and short non-coding RNAs⁴⁹ (Table 1).

Table 1.

The functions of Integrator at different non-coding RNAs

RNA SPECIES	FUNCTION of INTEGRATOR	CELL TYPE, ORGANISM	REFS
Short RNAs (<0.2kb)
U snRNAs	3’-box-mediated cleavage	All cell types, various metazoans	21,23,25,59,63
Viral miRNAs	Transcript release and maturation	Infected lymphocytes, marmoset	83,84
piRNAs	3’ cleavage (unknown motif)	Germ cells, Caenorhabditis elegans	87,88
Long RNAs (>0.2kb)
eRNAs	3’ cleavage (at unknown motif)	Multiple cell types, human	58,81,82
TERC	3’ cleavage (at unknown motif)	Human cell lines	92
NEAT1	Supporting early transcription termination	Human cell lines	94
Other lncRNAs	Supporting early transcription termination	Human cell lines	82

eRNA, enhancer RNAs; lncRNAs, long non-coding RNAs; miRNAs, microRNAs; NEAT1, nuclear paraspeckle assembly transcript 1; piRNAs, PIWI-interacting RNAs; snRNAs, small nuclear RNAs; TERC, telomerase RNA template component.

The U snRNA termination machinery

U snRNAs are short (100–200 bp) uridine-rich non-coding RNAs that have essential roles in pre-mRNA splicing. The U snRNAs U1, U2, U4 and U5 are embedded in the spliceosome, whereas U3 and U7 are implicated in rRNA processing and in transcription termination of histone mRNAs, respectively^50–52. Integrator has emerged as the key U snRNA transcription termination machinery, mirroring the role of the CPSF at protein-coding genes. In fact, the endonuclease module of Integrator (Fig. 2a) has significant homology with CPSF²¹. Elongating Pol II approaching the 3’-end of a gene, encounters an adenine-rich poly(A)-tail signal, which triggers CTD-dependent recruitment of CPSF along with the cleavage stimulation factor (CstF) complex. The CPSF2–CPSF3 heterodimer is endowed with RNA endonucleolytic activity⁵³: the zinc-dependent MBL/β-CASP domain of CPSF3 releases the capped nascent mRNA for immediate polyadenylation³⁷. The active MBL/β-CASP domain of Integrator within INTS11 shares an overall 40% identity and 60% similarity with that of CPSF3. Initial experiments showed that Integrator is recruited at U1 and U2 loci (both at the promoter and around the cleavage site), and that depletion of INTS11 and INTS1 resulted in accumulation of unprocessed, precursor U1 and U2 transcripts²¹. U snRNA processing is dependent on the catalytic site of INTS11 (Ref.²¹). Although lacking catalytic activity on its own, INTS9 forms a tight heterodimer with INTS11^28,38 and is essential for proper transcription termination of U snRNAs³⁸. Disrupting the interaction surface of the INTS9–INTS11 heterodimer results in defective endonucleolytic activity³⁸. The scaffold subunits INTS4^28,39 (Fig. 2a) is also functionally required for U snRNA termination (Fig. 3a). Unlike CPSF⁵³, the catalytic activity of Integrator has yet to be translated into a standardized biochemical assay, limiting our ability to dissect the contribution of additional subunits to the reaction.

Figure 3. Integrator terminates transcription of non-coding RNAs.

A) U small nuclear RNA (U snRNA) loci have distinctive promoter elements termed distal sequence element (DSE) and proximal sequence element (PSE), which bind the OCT1 (POU2F1) transcription factor and the small nuclear RNA-activating protein complex (SNAPc). OCT1 and SNAPc contribute to the recruitment of a transcription initiation-competent RNA polymerase II (Pol II) holoenzyme including DRB-sensitivity inducing factor (DSIF), negative elongation factor (NELF) and the Integrator complex. Shortly after transcribing through the 3’ box, which is a highly conserved motif at the termination site of all U snRNAs, Integrator cleaves the nascent small RNA, triggered by phosphorylation (P) of Ser7 of the C-terminal domain (CTD) of Pol II’s largest subunit (RBP1). A 3’ stem-loop in the precursor U snRNA and recognition of the ensuing 3’ box RNA sequence by a set of Integrator accessory subunits (INTS?) may support an efficient cleavage process.

B) Enhancer loci are activated by sequence-specific transcription factors that recruit the co-activator Mediator complex. Upon Mediator recruitment and assembly of the transcription pre-initiation complex (not shown), bi-directional transcription of the enhancer locus occurs, producing long (>200bp) sense and antisense transcripts that are called enhancer RNAs (eRNAs). Both sense-transcribing and antisense-transcribing Pol II holoenzymes recruit the Integrator complex to terminate transcription and release eRNAs without eliciting their polyadenylation.

C) Caenorhabditis elegans PIWI-interacting RNAs (piRNAs) are generally encoded by mini-genes grouped into large clusters. A 20–40bp piRNA precursor is cleaved co-transcriptionally by the Integrator complex, which is recruited by the Pol II CTD and by the Pol II subunit RPB9.

U snRNAs have a peculiar locus architecture, which includes unique DNA motifs that are required for recruiting Pol II: a distal sequence element (DSE) and a proximal sequence element (PSE) relative to the transcription initiation site⁵⁴ (Fig. 3a). DSE and PSE are also necessary for transcription termination, suggesting that Integrator must be loaded with Pol II at the initiation site^55,56 (Fig. 3a). Furthermore, an AT-rich sequence of about 14–16 nucleotides, termed the 3’ box, is consistently found in human snRNA loci⁵⁷. The 3’ box is essential for U1 termination and is positioned immediately downstream of the processed 3’ end⁵². Upon depletion of either INTS11 or INTS1, lack of cleavage activity leads to accumulation of unprocessed transcripts spanning the 3’ box²¹. The cleavage defects observed following loss of Integrator include Pol II downstream-readthrough and aberrant accumulation of Pol II on chromatin up to few kilobases past the 3’ end of U snRNA loci⁵⁸. In organisms with high gene density, such as C. elegans, readthrough of U snRNA genes can lead to aberrant transcription of downstream protein-coding genes⁵⁹ and can generally be regarded as a signal of environmental stress, leading to heat shock response²⁴. Although Integrator is the preeminent machinery of U snRNA transcription termination, CSPF complexes may be able to partially compensate for its absence. Furthermore, stochastic cleavage and termination of U snRNA may occur at some loci in a sequence-dependent manner⁶⁰.

Genome-wide data suggest that Integrator is recruited to U snRNA loci co-transcriptionally (at initiation or shortly after transcription initiation)^30,58. Integrator recruitment occurs, likely, through the Pol II CTD (Box 1) and may be favored by CTD Ser7 phosphorylation⁶¹, although unbiased proteomic studies show that Integrator has comparable affinity for both phosphorylated and unphosphorylated CTD⁶². Overall, it remains unclear how Integrator binds different CTD isoforms — recent data suggest that INTS2, INTS4 and INTS7 may all provide some affinity to the Pol II CTD, especially in the presence of phosphorylated Tyr1 CTD residues³² (whereas modelling of phosphorylated Ser7 residues shows decreased binding affinity). Other proteins, including DSIF, which is a universal regulator of Pol II pausing at protein-coding genes, may also be important for the recruitment of Integrator to Pol II⁶³. Once recruited to transcribed U snRNA loci, it is unclear whether the INTS11 catalytic subunit is present in its active conformation. An intriguing hypothesis is that recognition of the 3’ box triggers INTS11 endonucleolytic activity through allosteric mechanisms. Recent reports suggest that the enhancer module INTS13–INTS14–INTS10 may bind to ssRNA containing the U1 3’ box²⁹, which raises the possibility that a dedicated module of Integrator surveys the nascent U snRNA transcript and recognizes the 3’ box to further engage the endonucleolytic module. Alternatively, the co-transcriptional formation of the U1 stem-loop structure may elicit binding of INTS13 (Ref.²⁹). However, depletion of INTS13 in vivo has little or no effect on U snRNA processing^29,30,64. Whether Integrator specifically recognizes the 3’ box sequence has not been formally demonstrated. Additional factors may also contribute to the 3’-end processing activity of Integrator, including NELF^32,63,65.

U snRNAs are encoded in hundreds of copies scattered across the mammalian genome, though only a subset (~40) are transcribed in any given cell⁶⁶. Nearly all loci are transcribed by Pol II, with the exception of U6 and U6atac, which are Pol III genes⁶⁷. Although ChIP-seq data for Integrator subunits reveal recruitment of the complex across the whole spectrum of Pol II and Pol III U snRNA loci, most studies on Integrator have focused on Pol II. Integrator dynamics and recruitment at U6 sites remains to be elucidated, but it must be noted that Pol II is also found at most Pol III genes⁶⁸.

Biogenesis of enhancer RNAs

Developmental processes require temporally-coordinated transcriptome changes that are orchestrated by thousands of enhancer elements^69–72. Enhancers promote transcription of specific genes by engaging their proximal promoter through the formation of regulatory chromatin loops, which are stabilized by genome architecture modulators such as the cohesin complex and CTCF^73,74. Enhancer and promoter elements are structurally similar in that their activation requires an accessible, nucleosome-free region, to which sequence-specific transcription factors and the Mediator complex can be recruited^9,69 (Fig. 3b). Similar to promoters, a PIC assembles at enhancers to allow transcription initiation by Pol II⁵. In fact, nearly all active enhancer elements are sites of active Pol II transcription, which is largely bi-directional^75,76 (Fig. 3b). Transcription at enhancer regions produces short-lived non-coding enhancer RNAs (eRNAs), which are relatively short (less than 1kb⁷⁵) and accumulate at chromatin. There are functional implications to eRNAs: they may be required for enhancer-mediated activation of their cognate protein-coding genes⁷⁷. In fact, eRNAs transcription in mammals largely precedes activation of their cognate protein-coding genes⁷².

A few eRNAs are polyadenylated by the CPSF complex, which efficiently couples polyadenylation to 3’-end cleavage of the nascent transcript^78,79. However, the vast majority of eRNAs, which lack a poly(A) tail, is cleaved by the Integrator complex. In fact, Integrator is loaded at active enhancers^30,58,80,81 (Fig. 3b) and its depletion elicits transcriptional readthrough at both 3’ ends of bi-directional enhancer loci^58,81,82. The readthrough is accompanied by accumulation of Pol II downstream of the transcription termination site, similar to what has been observed at U snRNA loci upon Integrator depletion⁵⁸. Furthermore, depletion of Integrator results in increased aberrant polyadenylation of eRNAs, underscoring Integrator being the eRNA termination machinery of choice in physiological conditions⁵⁸. Impaired termination of eRNAs prevents their release from the transcription bubble and may thus impair promoter activation^58,81. It is unclear whether specific sequences at enhancer loci dictate cleavage. Unlike U snRNAs, enhancer loci are highly diverse in sequence, poorly conserved across species and they lack any motif similar to the 3’ box. Nonetheless, the ability of Integrator to generate and maintain a pool of eRNAs necessary to enforce enhancer–promoter interactions appears fundamental for the execution of stimulus-dependent cellular processes such as differentiation^58,81.

A growing pool of RNA targets

The scope of Integrator’s endonucleolytic activity extends beyond U snRNAs and eRNAs (Table 1). In addition to the function of INTS11 at pre-mRNAs and its biological consequences, which are discussed in the next section, many ncRNA species have recently been identified as targets of the Integrator complex in different organisms. Particularly, Integrator has been implicated in post-transcriptional gene silencing mediated by small RNAs such as microRNAs (miRNAs) and PIWI-interacting RNAs (piRNAs). Although miRNAs are best studied in multicellular organisms, they are also encoded by certain viral genomes. For instance, the DNA genome of herpesvirus Saimiri (HVS), a gamma-herpesvirus originally isolated from squirrel monkeys, encodes a handful of protein-coding genes and seven small ncRNAs, whose architecture is similar to human snRNAs⁸³. Three out of the seven viral snRNAs contain a stem-loop-forming sequence at the 3’ end, well past the 3’ box that dictates their termination site. In HVS-infected primate lymphocytes, the host Integrator complex is recruited co-transcriptionally to cleave and release the pre-snRNAs as well as, separately, a downstream stem-loop that functions as an effective primary miRNA (pri-miRNA) transcript⁸³. Integrator is then capable of cleaving the 3’ end of the pri-miRNA stem-loop, thereby generating a viral precursor-miRNA that is exported to the cytosol to undergo maturation by the host miRNA maturation pathways⁸⁴. Pri-miRNA cleavage by INTS11 appears to be dependent on a distinct, slightly degenerate, A-rich 3’ box⁸⁴. Intriguingly, in human cells, depletion of INTS11 (and other Integrator subunits) leads to a global reduction of the pool of endogenous mature miRNAs, but does not affect transcription or processing of the pri-miRNA⁸⁵. Integrator may be required for stabilizing mature miRNA by facilitating their loading onto Argonaute proteins thus possibly favoring their targeting and cleavage of target mRNAs⁸⁵.

Similar to miRNAs, piRNAs function in post-transcriptional silencing in association with PIWI (Argonaute family) proteins. Their role is largely restricted to germinal cells, where they surveil transgenerational inheritance by maintaining genome integrity through silencing of transposable elements⁸⁶. In C. elegans, piRNAs arise predominantly from two large gene clusters and are transcribed as individual units⁸⁶ (Fig. 3c). The 20–40 nucleotides piRNA precursors are transcribed by Pol II; depletion of Integrator’s catalytic subunit causes a marked reduction of piRNA abundance and an increase in precursor size⁸⁷. The abundance of piRNA is controlled directly by the catalytic activity of Integrator, rather than by modulation of Pol II elongation⁸⁷ (Fig. 3c). Whereas Integrator is generally recruited co-transcriptionally by the Pol II CTD, in nematodes the RPB9 subunit of Pol II may have an additional role in promoting localization of Integrator at piRNA loci in vivo⁸⁸ (Fig. 3c). Genetic screens implicate Integrator in piRNA biogenesis of D. melanogaster^89,90, suggesting that transgenerational silencing of transposable elements may rely on Integrator in all metazoans.

Whereas eRNAs, microRNAs and piRNAs comprise large and diverse classes of transcripts, there are single-copy ncRNAs that function as highly specialized regulators of biological processes. For instance, maintenance of chromosome ends hinges on the activity of telomerase, a reverse transcriptase capable of synthesizing new DNA ends using as template the long non-coding RNA (lncRNA) telomerase RNA template component (TERC)⁹¹. Transcription termination of TERC by 3’-end cleavage relies on Integrator and is specified by TERC’s promoter, similar to U snRNAs⁹². The ~20kb lncRNA nuclear paraspeckle assembly transcript 1 (NEAT1) is an essential component of paraspeckles, which are stress-related nuclear bodies⁹³. In physiological conditions, the endonucleolytic module of Integrator promotes early termination of NEAT1, thereby generating a ~4kb isoform that does not efficiently nucleates paraspeckles⁹⁴. Interestingly, NEAT1 cleavage by Integrator may be coordinated with, rather than antithetical to, recruitment of the CPSF complex, since the short NEAT1 isoform is known to be polyadenylated⁹⁵. Further crosstalk between Integrator and CPSF may occur at the 3’end of select protein-coding genes, as we discuss in the ensuing section. Lastly, the role of Integrator in overseeing transcription termination of ncRNAs⁴⁹ might be more extensive that anticipated. In fact, promoter upstream transcripts (PROMPTs) that originate antisense to a large set of protein coding genes appear to be terminated primarily by Integrator⁸². In this regard, the transcription elongation factor SPT6 seems critical to ensure proper recruitment of Integrator at many ncRNA loci, including eRNAs⁸².

In summary, an increasing body of work demonstrates that Integrator operates as a 3’-end RNA processing factor for a broad range of long and short ncRNAs in several species (Table 1). The endonucleolytic activity of Integrator is co-transcriptional and maintains homeostatic levels of U snRNAs, TERC, piRNAs, and certain promoter antisense transcripts.

Transcription of protein-coding genes

RNA cleavage by Integrator is not restricted to ncRNAs. In fact, nascent mRNAs are targeted by Integrator, primarily during Pol II pausing and early elongation. Cleavage may also occur during late elongation and around the termination site. Thus, the scope of Integrator’s activity at coding genes encompasses the entire transcription process and, in addition to the cleavage module, implicates the newly discovered phosphatase module as a tuner of Pol II processivity.

Co-transcriptional targeting of nascent mRNAs by the endonucleolytic module

The tight control of Pol II activity at protein-coding genes is crucial for executing coordinated gene expression programs that vary across cells, tissues and developmental stages. Following transcription initiation, and before the onset of productive transcript elongation, Pol II undergoes a regulatory step of pausing, which occurs after polymerization of a 30–50nt nascent, capped mRNA^{10,11,96–101}. This promoter-proximal Pol II pausing is one of the most conserved mechanisms of gene regulation and co-evolved in multicellular organisms as a rheostat for transcriptional gene regulation during development^102,103. Nonetheless, the role of Pol II pausing is not limited to development, as adult tissues rely on promoter-proximal Pol II pausing to be able to induce a rapid and coordinated transcriptional response to a variety of extracellular cues^104–108.

Several reports have implicated Integrator in the regulation of Pol II pausing and elongation, with the proposed mechanism varying according to the model system and the Integrator subunits examined. The first genome-wide analysis of Integrator chromatin occupancy revealed widespread association of INTS11 and Pol II at active genes¹⁰⁸. In human cells, Integrator is highly enriched at Pol II initiation and pausing sites and remains co-transcriptionally associated with Pol II throughout the gene body and the 3’-end of highly active genes^{30,31,108,109}. Similarly, INTS3 broadly associates with promoter-proximal Pol II pausing sites, in correlation with NELF and DSIF occupancy¹¹⁰. Additional data from D. melanogaster cell lines suggest that Integrator may be most enriched at genes with ‘unstable’ promoter-proximal Pol II pausing, which are marked by fast turnover of paused Pol II and characterized by lower levels of histone H3 Lys4 trimethylation compared to genes with more stable pausing¹¹¹.

Integrator appears to have distinct functions in Pol II regulation at different sets of protein-coding genes. For instance, the depletion of INTS11 or INTS1 blunts stimulus-dependent activation of immediate early genes (IEGs)¹⁰⁸. IEGs accumulate promoter-proximal Pol II and are primarily regulated by a pause-release mechanism. Integrator depletion prevents Pol II release at IEGs upon stimulation^108,109, in a catalytic-dependent manner¹⁰⁹. An Integrator-dependent pause-release mechanism extends well beyond stimulus-dependent transcription and is shared by a large fraction of genes, while transcribed at steady-state levels¹⁰⁹. In fact, depletion of Integrator in HeLa cells hampers productive elongation and correlates with Pol II stalling upon reaching the +1 nucleosome, where a second, NELF-independent barrier to elongation is mounted¹¹². Notch-dependent transcription in tumor cells also appears to be relying on INTS11, the catalytic subunit of Integrator¹¹³.

By contrast, in lower eukaryotes, Integrator seems particularly prone to suppression of transcription. In fact, depletion of the catalytic core in D. melanogaster leads to de-repression of a set of protein-coding genes, whereas a far smaller number of genes are downregulated¹¹². In addition to repressing steady-state transcription, stimulus-dependent transcription is also curtailed by Integrator at D. Melanogaster copper-induced promoters (where new Pol II initiation occurs)⁶⁴, whereas the activation of heat shock genes, which is dependent on pause-release, is enforced by Integrator similar to human IEGs¹⁰⁸.

There are multiple lines of evidence that Integrator directs its catalytic activity against nascent transcripts early during their elongation^64,109,111, albeit with alternative outcomes (Fig. 4). According to one model of transcription activation, mRNA cleavage by the Integrator after 21-nucleotides is conducive to release of paused Pol II, followed by loading of an elongation-competent Pol II complex¹⁰⁹ (Fig. 4a). According to an alternative model of elongation repression, Integrator is preferentially recruited to a subset of genes under promoter-proximal Pol II pausing, where its endonucleolytic activity results in premature transcription termination [G] (PTT), therefore preventing paused Pol II complexes from becoming fully licensed for elongation¹¹¹ (Fig. 4b). A similar role of ‘premature termination’ has been described for the CSPF factor PCF11 (Ref.¹¹⁴), and PTT has long been considered an important mechanism of transcription regulation¹¹⁵.

Figure 4. Cleavage of nascent pre-mRNA at RNA polymerase II pausing sites.

Integrator regulates the release and stability of RNA polymerase II (Pol II) pausing at promoter-proximal sites through its cleavage of the nascent mRNA. Conflicting models may explain the role of Integrator in transcriptional gene regulation.

A) Cleavage of nascent RNAs has been proposed to elicit gene activation by clearing away long-paused Pol II complexes that are not competent for productive elongation. Elongation-competent Pol II holoenzymes can be effectively recruited only after cleavage of the nascent pre-RNA by Integrator and premature termination of the lingering polymerase. Elongation-competent Pol II are successfully licensed for productive transcript elongation past the pausing site, without any further Integrator-mediated cleavage occurring.

B) Alternatively, Integrator has been proposed to preferentially target a subset of genes, enforcing post-initiation transcriptional repression. Genes recruiting high levels of Integrator maintain lower levels of promoter-proximal Pol II by continuously dislodging paused polymerase and therefore curtailing productive elongation (left). Conversely, genes recruiting less Integrator maintain stable levels of paused Pol II and can support productive elongation (right). In all models, the cleavage activity of Integrator at the pausing site of protein-coding genes is believed to occur following capping (m⁷G) of the nascent pre-mRNA.

Beyond the Pol II pausing region, the endonucleolytic activity of Integrator may be further used at protein-coding genes for broad-range termination of non-productive or stalled transcription, as far as few kilobases from the TSS¹¹⁶. This activity occurs especially at lowly expressed loci, in a way similar to how Integrator restricts the accumulation of certain lncRNAs⁸². Thus, in mammalian cells, Integrator may be a tool to tame pervasive transcription. According to multiple ChIP-seq datasets obtained from human cells, Integrator is tied to transcribing Pol II from initiation through termination. Integrator has been proposed to regulate transcription termination and Pol II positioning at protein-coding genes that have a motif similar to the 3’ box of U snRNAs¹¹⁰ and to mediate termination of mRNAs of replication-dependent histones¹¹⁷. Recent data suggest that, although most protein-coding genes are exclusively dependent on the CPA machinery for transcription termination and polyadenylation, a subset of genes enriched in proximal alternative polyadenylation sites require the activity of INTS11 to ensure selection of the proximal site¹¹⁸. More generally, the activity of Integrator at transcription termination sites is crucial during cellular stress conditions (i.e. hyperosmotic shock), when transcription downstream of canonical polyadenylation sites become pervasive and is partly attenuated by Integrator activity¹¹⁹.

The phosphatase module of Integrator

Over the past two years, multiple reports have uncovered a novel mechanism through which Integrator regulates Pol II transcription in association with PP2A. PP2A is a highly conserved, ubiquitously expressed Ser/Thr phosphatase^40,43, which accounts for the majority of phosphatase activity in any given cell^40,120. In vivo, PP2A works primarily as a trimeric complex of a scaffold (PP2A-A), a catalytic (PP2A-C) and a regulatory (PP2A-B) subunit⁴⁰. Regulatory subunits are encoded by at least 15 different genes in human and confer target specificity to the A/C catalytic core¹²¹ (Fig. 5a). The first evidence for an association between the Integrator and PP2A came from comprehensive proteomics studies suggesting the existence of robust protein–protein interactions between submits of the two complexes^122,123. A decade later, a landmark paper reported the cryo-EM structure of nine subunits of the Integrator complex bound to a PP2A holoenzyme: the catalytic subunit PP2A-C and the scaffold subunit PP2A-A (together termed INTAC)³³. While striving to obtain a first comprehensive structure of Integrator through overexpression of several tagged subunits followed by affinity purification, the authors also purified stoichiometric amounts of endogenous PP2A, suggesting that most endogenous Integrator may in fact carry the phosphatase module (Fig. 2; Fig. 5a) and the resulting INTAC could be the prevailing Integrator-complex variant in mammalian cells. Interestingly, no PP2A-B subunits were reported to be stably associated with Integrator, and structural data from PP2A heterotrimeric complexes suggests a steric clash between PP2A-B and INTS6 or INTS8 (Ref.³³). Two additional studies functionally characterized the association of Integrator and PP2A and its biological significance, broadening the scope of action of Integrator in the Pol II transcription process^31,34.

Figure 5. Integrator with the phosphatase module inhibits pause-release of RNA polymerase II.

A) Canonical serine/threonine-protein phosphatase 2A (PP2A) complexes are composed of PP2A-A (scaffold subunit), PP2A-C (catalytic subunit) and PP2A-B (regulatory subunit) (Top). Although PP2A-A–PP2A-C heterodimers are active in vitro, their efficient targeting in vivo requires assembly with PP2A-B. The Integrator complex is frequently found associated with PP2A (Int–PP2A), together also known as INTAC. In the Int–PP2A complex, PP2A-B is replaced by INTS6, while INTS8 contacts the N-terminus of PP2A-A (bottom). All other known Integrator subunits are stably associated with Int–PP2A.

B) The phosphatase module of Integrator inhibits promoter-proximal pause-release of RNA polymerase II (Pol II) by antagonizing phosphorylation of several targets of the kinase positive transcription elongation factor b (pTEFb; cyclin-dependent kinase 9–cyclin T1). Recruitment of pTEFb to Pol II is thought to enable pause-release and productive transcript elongation at protein-coding genes. Int–PP2A and pTEFb compete for a number of target residues, including at the Pol II C-terminal domain (CTD; most notably Ser2), and Ser666 and Thr806 of SPT5 (subunit of DRB-sensitivity inducing factor (DSIF)). In addition to Int–PP2A, the phosphatases PP4 and PP1–PNUTS also dephosphorylates SPT5. Notably, the PP1 phosphatase was found biochemically to be associated with Integrator.

PNUTS, serine/threonine-protein phosphatase 1 regulatory subunit 10; SEC, super elongation complex.

As discussed above, transcription by Pol II is regulated by dynamic phosphorylation of the highly conserved Pol II CTD (Box 1) and of other Pol II-associated factors such as NELF and DSIF¹²⁴. Whereas the kinases involved in this process have been extensively studied^125,126, an understanding of the relevant phosphatases, especially in higher eukaryotes, has lagged behind largely due to the increased complexity and apparent promiscuity of phosphatase complexes in higher eukaryotes¹²⁰. Progression through the distinct stages of Pol II transcription requires the activity of specific CDKs such as CDK7, CDK9, CDK12 and CDK13 (Ref.¹²⁷). In particular, paused Pol II is targeted by CDK9–cyclin T1 (the p-TEFb complex), which is recruited through the super elongation complex¹²⁸ to phosphorylate the Pol II CTD, DSIF, NELF and other Pol II associated factors (i.e. PAF1 and SPT6) at the transcription bubble (Fig. 5b). Recruitment of pTEFb is paramount to release of paused Pol II and transition into a productive elongation phase. Consistent with recruitment of Integrator to active genes, PP2A is also found near the pausing site and at the gene bodies of most (if not all) expressed genes, in multiple cell types^31,33. Depletion of INTS8 or INTS6 prevents efficient PP2A recruitment to chromatin^31,33 and is broadly associated with increased transcriptional output measured as both nascent RNA and steady state RNA levels^31,33,34. Phosphorylated DSIF (SPT4–SPT5) is conducive to productive elongation¹⁸, and loss of Int–PP2A leads to increased phosphorylation of SPT5 at Ser666 (Ref.^34,129) and at Thr806 (Ref.^31,129), suggesting that Pol II pausing downstream the TSS is the result of a balancing activity between phosphatases and kinases (Fig. 5b).

More specifically, Int–PP2A appears to functionally oppose pTEFb, and INTS6 depletion releases pausing induced by highly specific CDK9 inhibitors (CDK9i)³¹. Furthermore, CDK9i and allosteric activation of PP2A synergize to secure Pol II in a paused deadlock, effectively killing elongation-addicted acute myeloid leukemia cells³¹. Even in the absence of functionality of the elongation factor SPT5, loss of Int–PP2A promotes elongation¹²⁹, underscoring its all-around role in enforcing Pol II pausing. In fact, phosphorylated Ser2 CTD residues appear to be another key substrate of Int–PP2A³¹ (Fig. 5b). Phosphorylated Ser5 and phosphorylated Ser7 residues may also be targeted by Integrator^31,33,34,129 (Box 1), and ChIP-seq data of PP2A-A and PP2A-C indicate they diffusely interact with chromatin across the body and 3’-end of most genes³¹, suggesting that Int–PP2A modulates Pol II activity even beyond pausing. PP2A is not active at Tyr residues and, intriguingly, phosphorylated Tyr1 CTD may be necessary for Integrator binding³². In addition to PP2A, PP1–PNUTS and PP4 have been also linked to Pol II regulation, opposing CDK9 (p-TEFb) activity at specific residues throughout the transcription process^130,131. Remarkably, Integrator and PP1 were found to be biochemically associated^31,122 (Fig. 5b), albeit at a lower stoichiometry compared with PP2A, and the combined PP1 and PP2A inhibition improves rescue of CDK9i-induced pausing over PP2A inhibition alone³¹. The activity of PP1 and PP2A may be interdependent and largely coordinated; the interplay with other phosphatases operating on Pol II (SSU72, RPAP2, FCP1) will need to be further explored¹²⁰.

Non-catalytic roles of Integrator

Beyond the endonucleolytic core and the Int–PP2A phosphatase module, Integrator subunits have been shown to assemble into two additional modules with distinct functions: the enhancer module and the SOSS complex.

The INTS13, INTS14 and INTS10 subunits of Integrator can assemble into the smaller, biochemically independent^28–30 enhancer module (Fig 2a). This module was shown to form in vivo in hematopoietic progenitor cells and to bind developmental enhancer elements independently of the main Integrator complex and in the absence of Pol II³⁰. None of these three stably associated subunits contains predicted DNA-binding motifs or chromatin-binding domains (Fig. 1). Intriguingly, INTS13 and INTS14 are capable of binding nucleic acids to some extent, thereby providing a putative mechanism for recruitment that is independent of active transcription²⁹. Additionally, the transcription factor early growth response protein 1 (EGR1) and its co-factor NAB2 interact with the enhancer module and are required for its recruitment to poised enhancers³⁰. Enhancer activation by EGR1–NAB2–INTS13 axis is necessary for monocyte commitment of myeloid progenitor cells, highlighting a new Integrator function at enhancers beyond transcription termination. The enhancer module may effect transcription of developmental genes beyond myeloid cells, as suggested by a report of INTS13 mutations in a subtype of an oral-facial-digital developmental syndrome¹³². Additional work will need to clarify how the enhancer module interlaces with the endonucleolytic module and the rest of Integrator^29,132.

The SOSS complex comprises INTS3 (SOSS-A), NABP1 (SOSS-B1) or NABP2 (SOSS-B2) and the uncharacterized protein INIP (SOSS-C)^133–135. The three proteins form a stable complex in vivo and in vitro, in which INTS3 acts as a scaffold^47,48; the SOSS complex may also include INTS6 or INTS6L^46,48 (Fig. 2a). NABP1 and NABP2 were discovered and studied for their ability to bind ssDNA through their OB fold, similar to the RPA complex that is a well-studied player in DNA replication and DNA repair^136,137. Generation of ssDNA ends at sites of double–strand DNA break (DSBs) is paramount for DNA repair and genome integrity. During homologous recombination (HR), 3’ overhangs are generated at DSBs and promptly recognized by ssDNA binding proteins, which protect them from exonucleases and initiate the HR process. INTS3 is necessary for recruitment of NABP1 and NABP2 to DSBs^133–135 and supports efficient HR-mediated repair^134,135. Correspondingly, depletion of any SOSS component sensitizes cells to ionizing radiation (which produces DNA breaks)^44,134,135. DSBs repaired by HR are first recognized by the MRE11–RAD50–NBS1 (MRN) complex, which recruits and activates the kinase ATM to ultimately coordinate checkpoint activation and recruitment of additional repair proteins. SOSS components are required for recruitment of MRN and for ATM activation^138,139. INTS3, NABP1 and NABP2, and INIP frequently co-precipitate with various Integrator subunits (Fig. 1); furthermore, INTS6 can associate and work with SOSS and is also assembled in the phosphatase module of Integrator. At this stage, there is no conclusive evidence whether SOSS is an integral part of the main Integrator complex and additional work will have to elucidate the dual role of INTS6 as a recruiter and stabilizer of Int–PP2A and as a recruiter of SOSS to sites of DNA damage.

Integrator in development and disease

As a complex that ubiquitously regulates transcription and RNA processing in metazoans, Integrator is deemed essential across all tissue types and developmental stages. Integrator’s composite architectural and functional modularity originates a nuclear regulatory hub utilized by Pol II, transcription factors and DNA damage response proteins. The pathological implications of Integrator’s malfunction vary according to the subunit and module disrupted, from intellectual disability¹⁴⁰ and leukemogenesis¹⁴¹ to chronic obstructive pulmonary disease¹⁴². In this section we discuss the role of Integrator in a subset of inheritable neurodevelopmental disorders, and in multiple tumor types.

Roles of Integrator in development and developmental disorders

Integrator function is required during early development and tissue morphogenesis, as well as cell differentiation in the adult organism. In developing mouse embryos, deletion of Ints1 results in early lethality, likely by destabilization of the entire complex¹⁴³. Similarly, germ-line mutations of D. melanogaster Integrator core components result in mid-to-late larval lethality²³. In zebrafish embryos, mutation of Ints6 disrupts progression of gastrulation owing to de-repression of dorsal organizer genes, resulting in severe dorsalization¹⁴⁴. This phenotype is consistent with INTS6-mediated enforcing of Pol II pausing³¹. Integrator is also crucially required in early embryonic stages in Artemia sinica (brine shrimp)¹⁴⁵. Furthermore, a fully functional Integrator complex is deemed essential to coordinate transcriptional programs in later stages of development, as well as during differentiation of adult stem cells. For example, Integrator malfunction compromises hematopoiesis at multiple levels: by downregulating SMAD–BMP signaling¹⁴⁶, failing to activate EGR1-dependent developmental enhancers³⁰, and de-repressing Polycomb target genes¹⁴⁷. In the stem cell-rich planarian flatworm, loss of core Integrator subunits disrupts stem-cell maintenance and tissue regeneration²⁵. In a mouse model of adipogenesis, differentiation depends on increased expression of Integrator subunits¹⁴⁸. The role of Integrator in regulating developmental gene expression programs has been further investigated in neuronal development. In mouse neuronal progenitor cells, Integrator coordinates the expression of a set of genes responsible for proper migration of newly formed neurons across the developing cerebral cortex, in coordination with the zinc finger protein Zfp609 and the cohesin loading factor Nibpl⁸⁰. During D. melanogaster development, Integrator prevents dedifferentiation of intermediate neural progenitor cells, thereby facilitating terminal cell fate commitment¹⁴⁹. All the above mechanisms may contribute to a handful of developmental cognitive syndromes that have been recently described (Table 2). Recessive mutations in INTS8 cause a severe neurodevelopmental disorder characterized by brain malformations, facial dysmorphism and severe cognitive delay¹⁴⁰. Recessive mutations in INTS1 have a similar clinical phenotype^140,150,151, with additional characteristic traits such as juvenile cataract (Table 2). Interestingly, all affected individuals show severely impaired or absent speech, perhaps suggesting the inability to sufficiently establish the neural circuits governed by the FOXP1–FOXP2 transcription factors axis¹⁵².

Table 2.

Integrator mutations in human development

SUBUNIT	MOLECULAR FUNCTION	MAIN PHENOTYPE	REFS
INTS1 (biallelic)	Central backbone, provides scaffolding to catalytic modules	Cognitive delay, absence of speech, cataracts and/or glaucoma, facial dysmorphism	150,151
INTS8 (biallelic)	Phosphatase module, required for PP2A recruitment	Cognitive delay, absence of speech, motor impairment	140
INTS13	Enhancer module, not required for cleavage or phosphatase activity	Oral-facial-digital anomalies, speech abnormality	132

PP2A, serine/threonine-protein phosphatase 2A

The phenotypes of INTS8 mutations may shed a light on the role of the Int–PP2A module and Pol II pausing in human development. Interestingly, point mutations in PP2A-A are associated with a wide spectrum of neurodevelopmental disorders^153–157. We speculate that transcription impairment in developing embryos carrying these PP2A-A mutations may be a major pathogenic driver. Future work on these developmental disorders will have to assess the precise contributions of disrupting canonical PP2A ternary complexes versus Integrator-bound PP2A. In line with the role of INTS1 and INTS8 in human development, indirect disruption of Integrator’s activity may underlie another neurodevelopmental syndrome, termed Galloway–Mowat¹⁵⁸, which is genetically determined by loss-of-function mutations in WDR73 (reported as a novel interactor of Integrator’s endonucleolytic module). Lastly, a component of the enhancer module of Integrator, INTS13, has been recently associated with a developmental ciliopathy, resulting in several oro-facial and digital anomalies¹³² (Table 2). Primary cilia are microtule-based sensory organelles that protrude from the surface of most cells and transduce and regulate several signaling pathways during development¹⁵⁹. Integrator may affect centriole localization and ciliogenesis by ensuring correct processing of ncRNAs crucial to the process^160,161 or by maintaining the proper dosage of ciliary gene transcripts through regulation of their promoter-proximal Pol II pausing¹³².

Roles of Integrator in tumorigenesis

Ubiquitously-expressed transcription co-activator complexes such as the nucleosome remodeler SWI/SNF have recently emerged as hotbeds of tumor-driving mutations¹⁶². By comparison, Integrator is less frequently mutated in human cancers and its role in tumorigenesis was not immediately recognized. Analysis of all 15 Integrator subunits (including INTS6L) across a comprehensive panel of primary tumor samples from the Cancer Genome Atlas collection revealed that nonsynonymous mutations in Integrator subunits occur in up to 10% of patients¹⁶³. The highest rate of mutations in all tumor types combined is found in INTS1, INTS2 and INTS8, whereas INTS3 and INTS7 are frequently mutated in diffuse large B-cell lymphomas and pancreatic adenocarcinomas, respectively¹⁶³. Intriguingly, a pan-tumor analysis highlighted INTS10 as the gene most subjected to purifying selection during tumor evolution¹⁶⁴. In essence, tumors bearing one copy of wild-type INTS10 (hemizygous), preserve this wild type allele by selecting against cells that accumulate missense mutations, underscoring the importance of INTS10 for tumor progression and suggesting Integrator is a tumor vulnerability¹⁶⁴. Similarly, the Integrator catalytic core INTS9–INTS11 is rarely mutated in tumors and there are different nuances to its potential role as a disease driver, depending on the tumor stage and origin. On one side, the catalytic module of Integrator promotes expression of a large fraction of genes in mammalian cells, including cell cycle activators and anti-apoptotic proteins^108–110. Particularly, Integrator acts downstream of MAP kinase signaling by enabling transcriptional responsiveness to MAPK-activating mutations that are frequently found in different tumor types such as melanomas¹⁶⁵. Inhibition of INTS11 could contribute to curbing tumor cell growth especially in this subset of tumors¹⁶⁵. On the other side, tumor initiation may benefit from reduced INTS11 activity. The recently identified role of INTS11 in preventing paraspeckles formation thorough suppression of the full-length NEAT1 isoform, suggests that the pro-oncogenic role of these nuclear bodies may also depend on hampered Integrator activity⁹⁴. This role could explain how reduced expression of Integrator components correlates with poor response to chemotherapy and with dismal survival rates in ovarian cancerpatients⁹⁴.

Several Integrator subunits are overexpressed in esophageal adenocarcinomas compared to normal epithelial cells; in these tumors, Integrator supports aberrant activation of Notch signalling and is essential for cell growth and tumorigenesis¹¹³. INTS7, INTS8 and INTS13 are frequently found upregulated in RNA-seq datasets of primary tumors matched to healthy tissue samples¹⁶³. Upregulation of INTS8 is also specifically associated with epithelial-to-mesenchymal transition and increased metastatic potential in hepatocellular carcinoma¹⁶⁶. Interestingly, these subunits carry very different roles within the complex. Expression of another subunit, INTS3, is amplified in some hepatocellular carcinomas¹⁶⁷, but repressed in a subset of acute myeloid leukemias through an aberrant mechanism of intron inclusion, which is brought about by misregulation of serine/arginine-rich splicing factor 2 (SRSF2)-dependent splicing¹⁴¹ (Table 2). Loss of INTS3 in hematopoietic stem and progenitor cells is crucial for boosting clonal expansion and proliferation, a hallmark of leukemogenesis¹⁴¹.

Discovery of the Integrator phosphatase module³¹ has major implications on tumorigenesis and suggests that part of Integrator may operate universally as a tumor suppressor. Int–PP2A restrains growth of MLL-rearranged leukemias and of solid tumors by enforcing Pol II pausing and opposing CDK9-dependent transcript elongation³¹. INTS6 is a key component of Int–PP2A and was originally named deleted in cancer 1 (DICE1), after its propensity to undergo deletion in cancer cells¹⁶⁸. In fact, loss or downregulation of INTS6 correlates with poor survival and is common in many tumor types³¹, including prostate cancer, hepatocellular carcinoma and hematopoietic malignancies^169–171. PP2A has long been regarded to as a tumor suppressor: activity of its various complexes is reduced in multiple tumor types by way of epigenetic silencing, somatic mutations and upregulation of endogenous inhibitors^172–178. Small molecule activators of PP2A have been recently developed and have demonstrated broad antitumoral activity^179–181. Certain class of PP2A activators are effective at increasing Int–PP2A recruitment to chromatin and act synergistically with CDK9 inhibitors to deadlocks a paused Pol II conformation, resulting in reduced tumor burden and increased survival in different mouse models³¹.

Conclusions and future directions

Fifteen years after its discovery, Integrator has now come into prominence as a keystone of transcription and co-transcriptional RNA processing in all higher eukaryotes. Universally recruited to Pol II-transcribed loci, likely at the transcription initiation–pausing transition, Integrator surveils the early as well as the late steps of the transcription process. Beyond controlling expression of Pol II genes, Integrator’s tasks in genome regulation further span from enhancer activation to sensing DNA damage sites. The biochemical, structural and genetic dissection of the Integrator complex is advancing our overall understanding of biological processes such as ncRNA biogenesis and Pol II pausing. Integrator studies have revealed how RNA cleavage solves polymerase pausing, for better or worse. Additional mechanistic insights will further clarify whether gene activation and repression mediated by INTS11 are mutually exclusive, and whether evolutionary pressure has reshaped the global roles of Integrator. Future studies should ideally develop and use physiologically more relevant experimental systems to model Integrator activity during promoter-proximal Pol II pausing, such as primary human and mouse embryonic-like cells and differentiated cells.

The recent, exciting discovery of Int–PP2A revealed that Integrator is endowed with a second catalytic activity, in addition to being an RNA phosphodiesterase (similar to the CPA complex). Notably, Int–PP2A is the first non-canonical PP2A complex described so far, raising intriguing questions on how canonical PP2A modulators (such as PP2A-B subunits, endogenous inhibitors, paralogous subunits) intertwine with Integrator and the transcription process. The RNA cleavage and the dephosphorylation activities of PP2A are likely coordinated during transcription, characterization and modelling of which will require extensive structural biology and biochemical investigations. Recent breakthroughs in structural biology have unveiled how Integrator is folded around a paused Pol II holoenzyme, also revealing structural clash points with elongation factors such as PAF and SPT6. Functional and biochemical evidence, however, suggest that Integrator is intimately associated to Pol II well beyond the promoter-proximal pausing site. Future studies of this inherently flexible complex will have to determine how Integrator adapts its conformation to escort polymerase through elongation and termination. Furthermore, the full scope of Integrator’s phosphatase activity may extend beyond Pol II (and associated factors) and target transcription factors, chromatin modifiers or even nucleosomes.

Disruptions of Integrator’s activity through copy number alterations or point mutations of its subunits are being systematically reported. Integrator is also emerging as a nexus of transcriptional responses to various mitogenic signaling pathways and cellular stress. In this capacity, Integrator may have an essential role in tumor development. Given the broad sphere of action of Integrator at chromatin, there is need for rigorous pathophysiological studies that will pinpoint the specific activity and module of the complex that drive tumorigenesis and developmental disorders. By clarifying the molecular basis of Integrator malfunction, we will better appreciate its potential as a pharmacological target.

Glossary

U small nuclear RNA

Uridine-rich small nuclear RNAs transcribed by Pol II that have essential roles in pre-mRNA splicing

MBL/β-CASP domain

The metallo-β-lactamase (MBL) fold is shared by a diverse set of enzymes, including a large family of bacterially-derived antibiotic hydrolases. In a subset of beta-lactamase enzymes operating on nucleic acid substrates, the MBL fold further extends into a beta-CASP globular domain to form an active nuclease site

C-terminal domain

The unstructured and highly repetitive C-terminal domain of the largest subunit of Pol II is intricately phosphorylated to regulate Pol II function

HEAT repeats

A protein structural motif composed of tandem repeats of 2 α-helices linked by a short loop

OB-fold domain

The oligonucleotide or oligosaccharide binding fold is an evolutionarily ancient protein domain capable of binding nucleic acids

VWA domain

The von Willebrand factor type A domain is an alternating sequence of α-helices and β-strands, generally mediating protein–protein interactions

Premature transcription termination (PTT)

generally occurs when elongating Pol II is arrested at any point after the TSS and before reaching a canonical termination site, usually resulting in release of unstable transcripts