Integrator is a Global Promoter-Proximal Termination Complex

BLURB

Integrator broadly drives promoter-proximal termination of RNAPII through phosphatase and RNA endonuclease activities. Here, we summarize recent work illuminating the molecular underpinnings of Integrator activity. Further, we describe the critical role Integrator plays in human health and disease, through its regulation of gene expression and noncoding RNA synthesis.

SUMMARY

Integrator is a metazoan-specific protein complex capable of inducing termination at all RNAPII-transcribed loci. Integrator recognizes paused, promoter-proximal RNAPII and drives premature termination using dual enzymatic activities: an endonuclease that cleaves nascent RNA and a protein phosphatase that removes stimulatory phosphorylation associated with RNAPII pause release and productive elongation. Recent breakthroughs in structural biology have revealed the overall architecture of Integrator and provided insights into how multiple Integrator modules are coordinated to elicit termination effectively. Further, functional genomics and biochemical studies have unraveled how Integrator-mediated termination impacts protein-coding and noncoding loci. Here, we review the current knowledge about the assembly and activity of Integrator and describe the role of Integrator in gene regulation, highlighting the importance of this complex for human health.

INTRODUCTION

In Metazoans, the regulated pausing of RNA polymerase II (RNAPII) and its controlled release into productive elongation are major points of gene regulation^1,2. After synthesizing 20–50 nt of RNA, RNAPII is bound by the elongation factor SPT5 and the NELF complex, which promote the stable pausing of RNAPII^3–7. During pausing, RNAPII remains active and engaged on the DNA template while awaiting further signals for productive elongation. Recruitment of the kinase P-TEFb allows for the release of paused RNAPII into the gene body, in large part because phosphorylation of SPT5 and the carboxy-terminal domain (CTD) of the largest subunit of RNAPII triggers the dissociation of NELF and binding of elongation factors that stimulate transcription elongation^1–3,8.

As RNAPII transcribes across the gene body, the CTD is further phosphorylated, and elongation continues processively until RNAPII reaches the polyadenylation sequence (PAS) at the gene 3′ end⁹. The PAS sequence designates the appropriate location for pre-mRNA cleavage by the Cleavage and Polyadenylation (CPA) machinery, which is coupled with the dephosphorylation of SPT5 and the RNAPII CTD¹⁰. Together, dephosphorylation of the elongation complex and RNA cleavage by the CPA machinery slow elongation and facilitate transcription termination, wherein RNAPII releases both the nascent RNA and DNA template.

Importantly, not all promoter paused RNAPII is destined to transcribe a full-length RNA, and increasing evidence supports a model where pause release is balanced with an alternate fate of promoter-proximally paused RNAPII, namely premature termination (Figure 1). Accordingly, interest in premature termination as a gene regulatory strategy has grown substantially, as has the appreciation that much of this is carried out by Integrator, a termination complex that is highly implicated in both development and disease^11,12. Here, we describe recent progress towards understanding premature termination driven by Integrator and highlight the conceptual and functional similarities with the CPA machinery. More details on canonical transcription termination at gene 3′ ends are provided in an accompanying review by Passmore and colleagues.

Figure 1. Integrator contains modules with both endonuclease and phosphatase activities.

Shown are schematics depicting the balance between pause release by P-TEFb vs. Integrator-mediated termination, depicting the cleavage activity of the Integrator endonuclease, and the phosphatase activity of Integrator-associated PP2A.

Discovery of Integrator and overall architecture of the complex

Integrator was initially purified as a complex associated with the RNAPII CTD¹³, comprised of 12 subunits that were numbered by descending size (INTS1-INTS12)¹³. Integrator was found to participate in 3′-end formation of the U-rich small nuclear RNAs (snRNAs) that form central components of the spliceosome¹³, implicating this novel complex in RNA processing and transcription termination.

Two additional Integrator subunits, INTS13 and INTS14, were subsequently identified through a genome-wide RNAi screen for factors required for snRNA biogenesis¹⁴, and validated as subunits of Integrator using immunoprecipitation and mass spectrometry^15,16. More recently, multiple studies have provided evidence for the existence of INTS15 using systems biology and biochemical approaches^17–20.

Several additional factors have been found to play central roles within Integrator. Most notably, subunits of the PP2A phosphatase demonstrate a biochemically stable association with Integrator and cryo-EM structures of the Integrator-PP2A complex reveal intimate interactions of the PP2A-A scaffold subunit and PP2A-C enzymatic subunit with multiple surfaces on Integrator^21–23. Incorporating a phosphatase within Integrator has profound implications since the P-TEFb-mediated phosphorylation of the paused elongation complex is critical for RNAPII progression into productive elongation²⁴. Indeed, the Integrator-associated PP2A phosphatase was recently shown to antagonize transcriptional kinases to suppress pause release and transcription elongation (Figure 1)^21–23. Beyond PP2A, mass spectrometry studies have identified a collection of weak interactions with proteins involved in various cellular processes^15,16,25.

Upon the identification of Integrator, primary sequence inspection of its subunits yielded little insight into the function of the complex, with the key exception of INTS9 and INTS11, which are members of β-CASP/metallo-β-lactamase (MβL) family of DNA/RNA endonucleases (Figure 2)^26,27. These two Integrator subunits are paralogous to the cleavage and polyadenylation specificity factors CPSF100 and CPSF73, respectively²⁸. These observations provided critical clues that Integrator could cleave nascent RNA. The parallels between Integrator and the CPA machinery extend further, as INTS9 and INTS11 interact in a manner reminiscent of the CSPF100/73 heterodimer²⁹, and INTS11, like CPSF73, possesses catalytic activity, whereas INTS9, like CPSF100, lacks several critical amino acids thought to be required for activity. Notably, cleavage of nascent RNA by Integrator or the CPA machinery releases an RNA with a protective 5’ cap and leaves RNAPII associated with a short, uncapped (5′-monophosphate) RNA (Figure 1). This cleavage event can facilitate termination of the elongating RNAPII because the uncapped RNA 5’-end provides an entry point for exonucleases such as XRN2 and/or helicases that destabilize the elongation complex^30–32. However, the direct connections between Integrator endonuclease activity and transcription termination remain to be fully elucidated.

Figure 2. Domain organizations of Integrator and PP2A subunits.

Domains are indicated as boxes while vertical lines indicate boundaries between neighboring domains. The size of each subunit is provided as number of amino acids in the human ortholog. Flexible segments in the subunits are shown in gray. Abbreviations are: N-terminal domain (NTD); Middle domain (MD); C-terminal domain (CTD); N-terminal cap (Ncap); Plant Homeodomain Finger (PHD); Metallo-β-Lactamase (MβL); Metallo-β-Lactamase-associated CPSF73, Artemis, SNM, and PSO (β-CASP); von Willebrand factor type A (vWA); Huntington, Elongation factor 3, Protein Phosphatase A, Tor1 (HEAT). Domains containing α+β or only β secondary structure elements are labeled. Domains in INTS9 and INTS11 contain α+β elements. All other domains contain only α helices.

STRUCTURAL CHARACTERIZATION OF INTEGRATOR AND INTERACTIONS WITH PAUSED RNAPII

Recent breakthroughs have yielded structures of Integrator associated with PP2A-A and PP2A-C²³, as well as Integrator-PP2A bound to paused RNAPII^33,34, which have provided insight into the overall physical organization of the complex and its mechanism of activation. These and other recent structural studies^35–41 demonstrate that Integrator is assembled from a ‘core’ constructed of backbone and shoulder modules (Figures 2 and 3A) which are bound by discrete endonuclease, phosphatase, and auxiliary modules. Notably, while the entire Integrator complex was included in the sample for structural studies, the auxiliary module was not observed, nor were INTS3 or INTS12, likely because they are flexibly tethered in this state of Integrator. Likewise, many segments of individual subunits are not present in the atomic model due to flexibility. Nevertheless, the general architecture and, most notably, the modularity of the complex is apparent.

Figure 3. The overall architecture of Integrator in an inactive state.

(A). (Left) Schematic of the structure of Integrator-PP2A complex in an inactive state³³, with subunits colored as in Fig. 2. IP₆ observed in the structure of Drosophila ICM³⁶ is shown in a sphere model. The metal ions in the active site of INTS11 and manganese ions in the active site of PP2A-C are shown as spheres. (Right) Structure shown at left after 90° rotation around the vertical axis. (B). Structures of the individual Integrator modules. Module structures are derived from the structure shown in panel 3A with the exception of the INTS13/INTS14 structure, which is from³⁹. Structure figures were produced with PyMOL (www.pymol.org).

The Integrator core: backbone and shoulder modules

The Integrator backbone module consists of INTS1, INTS2, and INTS7 (Figures 2 and 3B). Not surprisingly, INTS1 makes extensive contacts with other members of Integrator, consistent with it being the largest subunit of the complex. There are direct contacts between the C-terminal domains of INTS1 and INTS2 (Figure 3B). The INTS7 N-terminal domain (NTD) adopts a crescent-shaped structure and interacts with INTS1 and INTS2, while the INTS7 middle domain (MD) interacts with INTS2 Ncap and helical repeat R1 (Figure 3B). Although not seen in any Integrator structures, INTS12 likely associates with the Integrator backbone module³⁴. Biochemical, yeast two-hybrid, and cellular studies demonstrate that INTS12 utilizes a ‘microdomain’ to interact with the N-terminal region of INTS1⁴².

The Integrator shoulder module contains the heterodimer of INTS5 and INTS8³⁷, with INTS5 wrapping around INTS8, generating a relatively inflexible structure (Figure 3B). The shoulder module has intimate contacts with the C-terminal repeats of INTS1 and INTS2 and is arranged perpendicular to the backbone module, with the two modules forming a cruciform shape. Altogether, the backbone and shoulder modules function as a scaffold for interactions with the other modules.

The Integrator phosphatase module

Early purifications of Integrator revealed an association with PP2A^15,16,25 while independent identification of factors associated with PP2A yielded Integrator subunits^43,44. The critical role of PP2A-mediated dephosphorylation of the transcription machinery for Integrator function was recently demonstrated, in both human cells and Drosophila, thus revealing the importance and conservation of the Integrator-PP2A interaction²¹. The phosphatase module consists of PP2A-C, PP2A-A, and INTS6 (Figure 3B)^22,23. This assembly contrasts all previously described PP2A complexes, which include a PP2A-B regulatory subunit⁴⁵. Given the established role of B regulatory subunits in guiding the recognition of PP2A substrates, we and others have proposed that Integrator serves a similar regulatory purpose, directing PP2A activity towards RNAPII and elongation factors^21–23. Notably, while INTS3 was not observed in the structure of Integrator, it binds INTS6³⁵ and crosslinking-mass spectrometry supports its proximity to the phosphatase module³⁴. INTS3, however, is not uniformly associated with Integrator, and is also in the sensor of single-stranded DNA (SOSS) complex, which is important for DNA double-strand break repair³⁸.

The stable association of the phosphatase module is strongly dependent on interactions with the shoulder module (Figure 3A). PP2A-C and INTS6 associate with INTS2, INTS5, and INTS8 to form a critical interface between the phosphatase module and the Integrator core. Indeed, excluding INTS5 or INTS8 from recombinant Integrator complexes results in loss of the phosphatase module, whereas removal of INTS11 has no effect²³. Further, a highly conserved WFEFLL motif within INTS8 directly contacts PP2A-A. Yeast two-hybrid studies show that INTS8 and PP2A-A can interact in the absence of other Integrator subunits, and their association is dependent on the WFEFLL motif²¹. Moreover, mutation of WFEFL residues causes a loss of PP2A from both human and fly Integrator²¹, resulting in a dramatic increase in phosphorylation of the RNAPII elongation complex and pause release.

The Integrator cleavage module (ICM)

As suggested by homology with the CPSF100/73 heterodimer within the canonical CPA machinery, dimerization of INTS9/11 proteins is required for endonuclease activity²⁹. Further, genetic and biochemical studies demonstrated a requirement for INTS4 for Integrator-mediated RNA cleavage^46,47. Subsequent structural studies have shed light on the molecular basis of ICM assembly and architecture^36,37,40. The ICM is located on one side of the Integrator core (Figure 3A), making direct contacts with the backbone module through INTS4. In comparison, INTS9 and INTS11 have no reported contacts with other Integrator subunits, suggesting that INTS4 anchors the ICM to the Integrator core. The catalytic segment of INTS11 (MβL and β-CASP domains) forms a pseudo-dimer with the equivalent segment of INTS9, which is likely stabilized by the NTD of INTS4 acting as a scaffold (Figure 3B). The organization of this pseudo-dimer is similar to that of CPSF100/73 observed in the active human U7 snRNP⁴⁸. The INTS11 CTD1 and CTD2 have tight interactions with the equivalent regions of INTS9 (Figures 2 and 3B), and the CTD1 of INTS11 is crucial for recruiting INTS4³⁷. However, in the structure of Integrator-PP2A complex, INTS11 is in an inactive, closed state and there is no room to accommodate the RNA substrate.

The overall structures of the isolated human and Drosophila ICM^36,37 are essentially the same as the human ICM within the Integrator complex²³, suggesting that there are no conformational changes in ICM when incorporated into Integrator. Unexpectedly, an inositol hexakisphosphate (IP₆) molecule was found in the structure of Drosophila ICM (Figure 2B)³⁶, and EM density consistent with IP₆ is present in human ICM³⁷. The binding site is located at the interface of all three subunits of ICM, with IP₆ having ionic interactions with several highly conserved residues (Figure 3B) in an electrostatically positive pocket. Although mutations of residues interacting with IP₆ do not abolish Integrator assembly, they disrupt Integrator function in Drosophila and human cells³⁶. The binding site is 55 Å away from the active site of INTS11, suggesting an allosteric regulation of activity.

The Integrator auxiliary module

The least understood Integrator module consists of INTS10, INTS13, INTS14, and INTS15. Loss of these subunits gives rise to only modest levels of snRNA misprocessing and minimal changes to the transcriptome, suggesting that this module is not critical to Integrator’s broad termination function^14,20,39,49. The structure of the INTS13-INTS14 complex shows that the two molecules are highly intertwined (Figure 3B), with an extensive interface³⁹. The domain organizations and structures of INTS13 and INTS14 are similar to Ku70 and Ku80, which are required for DNA double-strand break repair (Figure 2)⁵⁰. However, INTS13-INTS14 is expected to have a distinct nucleic-acid binding mode, as the DNA bound to Ku70-Ku80 clashes with INTS13-INTS14. Accordingly, INTS13-INTS14 is suggested to bind RNA rather than DNA³⁹, although the functional consequence of this interaction remains unclear. Curiously, the domain organization of INTS6 is also similar to that of INTS13 and INTS14 (Figure 2), and its β-barrel domain shares a similarity with Ku70-Ku80³³. The auxiliary module also has interactions with the ICM^34,37,39,51, through a segment in the C-terminal region of INTS13 ³⁹. Although INTS10 primarily contacts INTS14 in this module^14,39, recent studies have revealed interactions between INTS10 and INTS15 (CG5274 in Drosophila and C7orf26 in human)¹⁹, unveiling INTS15 as another subunit of the auxiliary module (Figure 2).

The function of the auxiliary module is enigmatic. INTS13 has been reported to associate with EGR1 and NAB2 to promote enhancer activation, but it is unclear if this function requires the entire auxiliary module⁴⁹. Notably, the association of the INTS13 C-terminus with the ICM suggests a potential role in regulating Integrator cleavage activity, which is supported by mutations in this domain⁵¹.

Integrator adopts an active conformation when associated with paused RNAPII

Structures of the isolated ICM or Integrator-PP2A capture INTS11 in an inactive conformation that would not accommodate RNA^23,36,37. However, the active conformation of INTS11 was observed in structures of Integrator-PP2A associated with a paused elongation complex (PEC) which includes SPT5 and NELF (Figures 4A and 4B)^33,34. In this complex, the PEC is embraced by ‘arms’ from Integrator, and the N-terminal repeats of INTS1 become ordered, enabling Integrator to contact the RPB2 subunit of RNAPII ~70 Å away from the body of Integrator (Figure 4A).

Figure 4. Overall architecture of Integrator in an active state.

(A). (Left) Schematic of the structure of Integrator-PP2A-RNAPII PEC complex in an active state ³⁴. Integrator subunits are colored according to Fig. 2, with RNAPII in gray, NELF and SPT5 in marine. (Right) Structure shown at left after 90° rotation around the vertical axis. (B). Nascent RNA in the active site of INTS11³³. The RNA is shown in orange, and DNA in olive. The nascent RNA exits RNAPII, and SPT5 helps to direct it to the active site of INTS11. (C) Zoom-in of CTD peptides associated with Integrator subunits with a numbered projections towards the active site of PP2A-C. The black sphere represents residue 1487 of RPB1 and is the last amino acid modeled in the structure. INTS1, INTS6, and INST11 are omitted for clarity.

There is a change in the position of ICM in the complex with PEC compared to Integrator-PP2A alone, which facilitates the interaction with SPT5 and the activation of INTS11³⁴. The catalytic segment of INTS11 is in direct contact with the KOWx-4 domains of SPT5, the only connection between INTS11 and the PEC (Figures 4A and 4B). This contact likely helps to bring INTS11 into an active conformation, with a 17° rotation of its β-CASP domain that opens the active site. This active state of INTS11 is similar to that of active CPSF73 observed in the histone 3′-end processing machinery⁴⁸. In addition, the SPT5 KOWx-4 domains located near the RNA exit site of RNAPII appear to direct the RNA towards the INTS11 active site (Figure 4B). The structures indicate a distance of ~22 nucleotides between the active sites of the polymerase and INTS11, which agrees with cell-based measurements of Integrator-mediated RNA cleavage^52,53.

The interactions observed between Integrator-PP2A and the PEC provide a compelling explanation for the enrichment of Integrator with paused, promoter-proximal RNAPII^21,52–57. First, the three-helix bundle in the N-terminal region of INTS6 (Figure 2) interacts with NELF-B (Figure 4A), which is uniquely present in paused RNAPII^33,34. Second, the structure of the PEC complex indicates that Integrator would sterically clash with transcription initiation factors and Mediator, implying that Integrator would not associate with a pre-initiation complex^34,58. Third, several Integrator binding sites on RNAPII are occluded upon association of SPT6 and PAF1 during the conversion of the paused RNAPII to a productive elongation complex, suggesting that association of Integrator with RNAPII during productive elongation would require significant structural rearrangements or would exhibit a lower binding affinity^8,34. Finally, the mode of RNAPII CTD interaction and catalytic activity of the Integrator phosphatase module is most consistent with action on a paused polymerase^33,34. The CTD repeats interact with several Integrator subunits (Figure 4C), and in the presence of NELF and DSIF, Integrator exhibits no preference for CTD phosphorylation status³³. Critically, CTD interactions with Integrator appear to form a path radiating to the active site of PP2A-C³³, suggesting that while Integrator can interact with RNAPII harboring a phosphorylated CTD, PP2A activity will lead to dephosphorylation (Figure 4C). Accordingly, Integrator-PP2A removes phosphates from the RNAPII CTD as well as SPT5²¹. The consequence of this phosphatase activity is to prevent the transition of paused RNAPII to productive RNA synthesis, and to reduce RNAPII elongation rate. Notably, slower elongation could facilitate RNA cleavage by the ICM, analogous to the role of PNUTS-PP1 phosphatase within the CPA machinery¹⁰. Importantly, pausing is a general feature of all RNAPII transcription, with evidence of paused elongation complexes at mRNAs, upstream antisense RNAs (uaRNAs), long noncoding RNAs (lncRNAs) and enhancer RNAs (eRNAs)^59,60. Thus, Integrator can broadly associate with PECs at coding and noncoding loci by recognizing specific features of paused RNAPII.

INTEGRATOR LOCALIZATION, SPECIFICITY AND FUNCTION

Despite a widespread convergence of data indicating that Integrator is a termination complex acting on RNAPII paused in early elongation, many questions remain about the specificity of Integrator activity and the impact of termination on transcription levels.

Integrator is globally enriched near transcription start sites

Given the intimate contacts of Integrator subunits with paused RNAPII it is not surprising that all Integrator subunits analyzed to date by ChIP-seq display enrichment just downstream of TSSs^21,52–57, and Integrator occupancy closely tracks with levels of promoter-associated RNAPII. Consistently, most Integrator-mediated termination occurs on RNAPII very early in elongation^21,52,53,55. Despite this promoter enrichment, Integrator subunits can remain associated with RNAPII as it enters the gene body, and Integrator has been implicated in termination at some canonical mRNA 3′ ends^54,61. Although questions remain about the prevalence of Integrator within gene bodies and which factor(s) might stabilize Integrator-RNAPII interactions once NELF dissociates, intriguing data indicate a role for Integrator in mRNA 3′-end formation under stress conditions^54,61. Specifically, cellular challenges such as osmotic stress or viral infection can cause failures in mRNA cleavage and 3′-end formation, resulting in RNAPII elongation >10kb beyond the typical site of transcription termination^62,63. This Downstream of Gene (DoG) transcription⁶⁴ represents a fundamental defect in the termination process and can allow RNAPII readthrough into neighboring genes, raising the specter of transcriptional interference⁶⁵. Notably, the mRNA genes that generate DoGs during stress partially overlap with genes that show evidence of readthrough past the 3′-end when INTS11 is depleted. Further, hyperosmotic stress was found to reduce Integrator association with RNAPII⁶¹, suggesting that a subset of protein-coding genes deploy Integrator as a backup to the CPA machinery under conditions of stress or immune challenge. However, in normal cellular conditions, INTS11 depletion affects canonical 3′-end formation at a limited number of transcripts^54,66. Long-term depletion of INTS11 was found to alter the expression of CPSF73⁵⁴ suggesting an interesting level of feedback among 3′-end processing machineries that could confound long-term depletion studies. Indeed, a recent study used a fast-acting degron to deplete INTS11 in mouse ES cells found no significant role for Integrator in canonical mRNA 3′-end formation under normal growth conditions⁶⁷.

Specificity of Integrator activity

Early data suggested that Integrator functioned uniquely at snRNA genes, with Integrator directly recruited to these promoters by interactions with the transcription factor snRNA activating protein complex (SNAPc) and with RNA cleavage directed by the ‘3′ box motif’ (Baillat and Wagner, 2015; Baillat et al., 2005). However, since this time, Integrator has been found to act at nearly every species of noncoding RNA (ncRNA), including lncRNA^68,69, PIWI-interacting RNAs⁷⁰, telomerase RNA⁷¹ uaRNAs^66,72 and eRNAs^66,73. Moreover, Integrator targets RNAPII at mRNA TSSs, regulating protein-coding gene activity^{52,55–57,74}. This broad spectrum of targets makes it difficult to envision models involving selective promoter recruitment by TFs, and indeed Integrator occupancy broadly correlates with RNAPII levels rather than specific TF motifs or protein factors^56,57. This widespread association of Integrator with paused RNAPII raises questions about regulation of the INTS11 endonuclease. Given data from snRNAs, an appealing model was that motifs in RNA modulate INTS11 activity. However, in contrast to the CPA machinery, no Integrator subunit contains a sequence-specific RNA binding domain, and sequences resembling the 3′ box were not observed near most Integrator target genes^66,75. Thus, it remains an open and intriguing question how Integrator-mediated cleavage might be controlled.

Integrator’s functional roles

Below, we highlight current models and remaining questions about the consequences of Integrator-mediated termination, using enhancers and protein-coding genes as examples.

Integrator at enhancers

RNAs generated at enhancers are typically short (<300 nt) and display heterogeneous 3′ ends, some of which are generated by Integrator activity. Integrator efficiently terminates paused RNAPII at enhancers, driving a rapid turnover of early elongation complexes and promoting the synthesis of short RNA species^52,60,73. Integrator loss delays eRNA 3′-end formation, with cleavage and termination carried out farther downstream by alternative termination complexes such as the CPA machinery. Accordingly, depletion of Integrator subunits results in the formation of eRNAs that are longer, yet less abundant^{49,52,66,67,73}. In the absence of a clear model for eRNA function⁷⁶ however, the consequences of Integrator activity at enhancers remain unclear. One study reported that Integrator facilitates enhancer-promoter looping at several stimulus-dependent genes⁷³ but this remains to be investigated more broadly. We propose that Integrator-mediated recycling of RNAPII at enhancers could maintain RNAPII dynamically engaged at the locus so that it is rapidly available for transfer to the promoter during gene activation. Moreover, if early termination by Integrator enables rapid re-initiation of transcription at enhancers, this could promote the synthesis of a short-lived ‘cloud’ of eRNAs around the enhancer that serve as binding surfaces for transcription factors or co-activators⁷⁷. Conversely, increased production of extended eRNAs with longer retention times on chromatin in the absence of Integrator might promote RNA-protein interactions. Future studies of eRNA function and a more detailed analysis of Integrator action at enhancers are thus warranted.

Premature termination at protein-coding genes

Long term depletion of Integrator (e.g., using 48–96 h RNAi treatment) in mammalian or Drosophila cells consistently reveals up- and down-regulation of hundreds of protein coding genes^{52,53,56,78,79}. The differential effects of Integrator loss on gene activity have suggested that Integrator could be repressive, stimulatory, or inconsequential for mRNA expression depending on the gene and the context. Fundamentally, either repressive or stimulatory effects could be envisioned for Integrator, depending on the status of the RNAPII complexes that are targeted for termination. For example, if Integrator terminates transiently paused elongation complexes that would otherwise produce a mature RNA, then Integrator activity would be repressive for transcription^21,52,66. Indeed, models wherein premature termination attenuates gene activity are well established in bacteria, yeast and metazoan systems⁸⁰. If instead, Integrator terminates inactive RNAPII that has stalled and is obstructing the DNA template, then Integrator would serve an activating role^53,57,74. Critically, these models are not mutually exclusive, and either scenario could dominate depending on the cellular conditions. For example, under normal growth conditions, Integrator might primarily serve to attenuate expression of stress-responsive genes, but upon activation of stress- or DNA-damaging pathways, Integrator could become critical for removal of stalled RNAPII and gene induction.

Genes within stress- and signal-responsive pathways are recurrently affected by Integrator loss across cell types and species, with a particular enrichment of immediate early genes such as Jun and Fos^{21,22,52,53,55–57}. These findings suggest a common set of targets or pathways, despite a lack of evidence for gene-specific Integrator recruitment^12,21,57.

Recent work using rapid, degron-mediated depletion of INTS11 in mouse embryonic stem cells sheds light on the function and specificity of Integrator. Acute degradation of INTS11 causes universal increases in RNAPII complexes released from promoter regions into genes⁶⁷, suggesting that Integrator broadly limits RNAPII release into elongation. However, loss of INTS11 did not significantly increase the expression of most genes, due to elongation defects in RNAPII. Investigation of these defects revealed that rapid loss of INTS11 did not dissociate other Integrator subunits from RNAPII, with evidence that the phosphatase module remained active on elongating polymerase. Consequently, phosphorylation of RNAPII and SPT5 was impaired, impacting the rate and processivity of elongation. As a result, only short mRNA genes were upregulated, along with a repertoire of inherently short ncRNAs. Of note, rapidly inducible and stress-responsive factors, including Jun and Fos, tend to be encoded by short transcripts with short or no introns⁸¹. These results suggest that the consistent activation of specific immediate-early genes encoding TFs, kinases and signaling regulators reflects the length of these genes rather than specific activities of Integrator.

INTEGRATOR IMPORTANCE IN PHYSIOLOGY AND DISEASE

The broad presence of Integrator would suggest fundamental importance to cellular function and organismal development. Indeed, depletion of Integrator subunits disrupts an array of cellular processes and differentiation pathways^71,82–86. Consistently, homozygous loss of most Integrator subunits causes lethality in multiple model organisms^{46,82,87–89}. The one exception to this trend is INTS6, which was initially identified as DICE1 (Deleted in Cancer 1) based on its frequent deletion in cancers⁹⁰. However, the lack of essentiality of INTS6 likely reflects that it is the only Integrator subunit with a paralog within the human genome (INTS6-like).

Studies have revealed particular importance for Integrator in developing and differentiating neuronal cell types. In mice, the Integrator core has been found to interact with Cohesin subunit Nipbl and ZFP609 and modulate the expression of genes important for neuronal migration during development⁷⁹. In Drosophila, depletion of backbone or shoulder module subunits leads to increased type II neuroblasts, and thus these subunits are required to prevent de-differentiation of intermediate neural progenitor cells⁹¹. In humans, mutation of INTS1 or INTS8 has been found to cause severe neurodevelopmental defects, including profound intellectual disability, epilepsy, and structural brain abnormalities^92,93. More recently, BRAT1 has been found to interact with the INTS9/11 heterodimer and mutations in BRAT1 are associated with numerous neurodevelopmental disorders⁹⁴.

Deleterious human Integrator mutations have also been informative on how Integrator modules interact with each other. In the case of INTS8, patients presenting neurological dysfunction are hypomorphic and predominantly express a form containing a three amino acid deletion of a conserved EVL motif near the C-terminal region⁹². The INTS8-ΔEVL was found to associate with the rest of the complex poorly, and recent structural findings indicate that the EVL motif lies within a region of INTS8 that is likely critical to maintain tight interaction between the shoulder and phosphatase modules²³. Similarly, two distinct mutations near the C-terminus of INTS13 that are causative of a specific ciliopathy disease⁵¹ are predicted to disrupt the cleavage-module binding motif^34,39 underscoring the importance of ICM interactions with the auxiliary module⁵¹. These examples highlight the value of characterizing diseased states caused by disrupted Integrator interactions.

FUTURE PERSPECTIVES

While the first 15 years of Integrator research have provided significant insight into its function, many questions remain unanswered. From a structural perspective, it is not yet known if Integrator only exists as a full complex or whether individual modules are separable. Importantly, whereas long term depletion of Integrator subunits often destabilizes the entire complex, short term degradation strategies are now allowing more surgical removal of specific subunits and modules. Such approaches will elucidate whether the two catalytic activities of Integrator are independent or coordinated. Additionally, a clear understanding of where the auxiliary module associates with the rest of the complex and how it contributes to Integrator function is still lacking. Finally, provocative biochemical and structural connections between Integrator and DNA damage sensing machinery^35,38,95 have been observed, but whether these interactions represent a novel function for Integrator or repurposing of its established termination activity is unknown.

The selectivity and regulation of Integrator activity also remains to be defined, with broad Integrator occupancy raising the question of how cells govern the balance between premature termination and pause release. We propose that mechanisms exist to deactivate Integrator or evict it from the paused elongation complex to enable gene induction. Possible candidates for this are factors that recruit P-TEFb, which could destabilize Integrator association with RNAPII by phosphorylation of SPT5 and the CTD, and dissociation of NELF. However, how P-TEFb activity might be coordinated to specifically overcome the Integrator-associated PP2A phosphatase function remains an active area of research. Further, how Integrator determines where to cleave nascent RNA is not clear. While sequence elements appear to govern snRNA 3′-end formation, these sequences are not found at other Integrator targets⁷⁵. We therefore propose that INTS11-mediated cleavage activity is modulated by protein factors associated with the paused RNAPII, potentially SPT5 or NELF.

Finally, while mutations within Integrator subunits can have dire consequences on human development and health, the specific gene sets most sensitive to these mutations are only beginning to be understood⁶⁷. Moreover, it isn’t clear why specific tissue types or developmental stages are asymmetrically impacted by reduced Integrator integrity. Regardless of these unknowns, our understanding of Integrator constituency and function in gene expression has undergone a remarkable evolution, and progress is expected to continue at a rapid pace.