Nuclear Speckle Biology: At the cross-roads of discovery and functional analysis

Introduction:

Nuclear speckles (NS), typically the second largest membraneless nuclear body, exist in nearly all cells over a wide range of metazoan species, from plants to animals. As reviewed previously, NS correspond to clusters of RNP granules which are also enriched in a wide variety of factors involved in all various steps of gene expression, particularly those involving RNA processing and, most notably, RNA splicing [1–3]. This led early on to suggestions of NS as storage/assembly/modification sites for factors involved in gene expression [2]. Alternatively, early work on NS also described the association of several specific genes with the NS periphery and specific nascent RNAs entering the NS interior, leading to suggestions of NS as a type of gene expression hub for a subset of genes [1]. Over the past 5 years, several orthogonal genome-wide analyses have established the NS periphery as a special nuclear “niche” for thousands of highly expressed genes and specific RNAs. Here we provide a brief overview of nuclear speckle biology and biochemistry and then review the special role NS may play in organizing overall nuclear genome architecture. Finally, we then summarize recent progress made in understanding the functional significance of NS related genome organization in both health and disease.

Overview:

NS were discovered a century ago by Ramon y Cajal using histological staining and then rediscovered twice- first by electron microscopists as interchromatin granule clusters (IGCs) and then through the speckle-like immunostaining of snRNPs and poly-A RNA [4,5]. Here we define NS as the closest light microscopy equivalent to IGCs. This definition is imprecise as regions detected by different NS markers will vary. For example, poly-A RNA [6] and the MALAT1 lncRNA [7] extend several hundred nm beyond NS markers such as the SC35 monoclonal antibody [1] [8], which by immunostaining primarily recognizes phosphorylated SRRM2 [9], and antibodies against SON and SRRM 2, the two proteins found most highly enriched in NS [10]. SON and SRRM2 are considered scaffolding proteins that nucleate NS because their double knockdown disrupts the local concentration of multiple NS marker proteins [9].

A high fraction of NS-enriched proteins contain low complexity, intrinsically disordered regions (IDRs) [3,11,12], and the SRRM2 IDRs are specifically required together with SON for nuclear speckle formation [9]. SON and SRRM2 form separate NS subcompartments, each recruiting different subsets of proteins [13]. Oligomerization of SRRM2 arginine-serine (RS) domains, drives SRRM2 phase separation, facilitated by interactions with RNA [13]. In vivo, NS are dependent on RNA for stability [14]. Proteins with arginine-rich, mixed charge low-complexity domains, including most (~70-80%) RS-domain proteins [15], are recruited to nuclear speckles [12].

Previous work had shown at least four families of kinases phosphorylating splicing factors- SRPKs, CLKs, PRP4K-like, and dual specificity tyrosine phosphorylation-regulated kinase 1A (DIRK1A) [16–18]. A recent study suggests a role for a SRPK1/CLK1 complex in physiologically regulating mobilization of RS-proteins out of nuclear speckles to facilitate splicing [19]. Overexpression of all of these kinases leads to dissolution of NS [17,18,20]. In contrast, the kinase TAO2 maintains NS [21].

Earlier studies suggested that NS functioned either as storage/assembly/modification sites for factors involved in gene expression, especially posttranscriptional RNA processing elsewhere in the nucleus [22,23], and/or as a gene expression hub for a subset of active genes clustered around their periphery [1]. Proximity to NS was proposed to facilitate the gene expression of this gene subset, both through the cycling of factors involved in pre-mRNA processing between nuclear speckles and adjacent active genes and through the boosting of post-transcriptional splicing of nascent RNAs from a subset of these genes. More recently, a third model has been proposed in which NS buffer the nucleoplasmic concentrations of various NS components, thus influencing splicing efficiency and alternative splicing of nascent RNAs throughout the nucleus [24].

All three models build on the observed concentration of many factors involved in different steps of gene expression within nuclear speckles. Proteomic studies, first on isolated IGC RNP granules [11,25] and then exploiting peroxidase-based proximity labeling of whole cells (TSA-MS, APEX) [10,26], confirmed that RNA splicing factors constitute the largest (~50-70%) class of speckle-enriched proteins [10,11]. Importantly, all NS resident proteins are also distributed throughout the nucleoplasm, typically functioning in close molecular proximity to all transcriptionally active genes and their nascent transcripts.

Overall, ~70-90% of all speckle-enriched proteins show GO functional annotations related to varying stages of gene regulation, including transcription (~3-11%) and cleavage, polyadenylation, and nuclear export (5-7%) [10,11]. Transcription-related speckle proteins include multiple proteins involved in pause-release (LARP7, HEXIM, Cyclin-T1, CDK9) and elongation (CDK12, CDK13, Cyclin-K, NELFCD, NELFA). NS proteins include those involved in RNA modifications, such as proteins involved in m6A modification (WTAP, METTL3, KIAA1429, RBM15, ALKBH5, YTHDC1) [10,11,26].

Mapping NS-associated chromosome domains and correlations with gene expression:

The close and reproducible NS association of a significant subset of active chromosomal regions has now been established through several orthogonal genome-wide mapping approaches. These include two sequencing-based methods- SPRITE and TSA-seq- as well as highly multiplexed immuno-FISH [27–33]. Several major conclusions have emerged. First, a set of chromosome regions, termed Speckle Associated Domains (SPADs), show high frequencies of NS contact (e.g. <250 or 300 nm) [31,32] and NS mean distances of several hundred nm [29]. Second, these SPADs are highly conserved among multiple cultured cell lines and brain tissue neuronal cell types [30] [33], despite cell type variation in NS size, numbers, and intranuclear positioning [34]. Third, both large, “Type-1” and small, “Type-2” SON TSA-seq peaks correspond to chromosome regions with locally elevated gene expression and gene density [29]. Although Type-1 and Type-2 peaks are similarly elevated in gene expression, Type-1 peaks show high NS association while Type-2 peaks show low NS association [34]. Thus, nuclear speckle association is not simply a consequence of domain-wide elevated gene expression, as also inferred previously from analysis of transgene arrays [35].

Fourth, comparing different cell types, differences in distance to nuclear speckles show a striking inverse correlation with differences in levels of gene expression [30], unlike differences in contact frequency or distance to the nuclear lamina or nucleoli [34]. Fifth, the type of genes associated versus non-associated with NS differ. GC-content, gene density, and gene expression show an overall inverse correlation with NS distance, while the most highly expressed genes, enhancers, and super-enhancers show a pronounced enrichment very near NS [29]. Furthermore, NS association correlates with increased transcription bursting frequencies [31]. Type-2 peaks are less conserved than Type-1 peaks between cell types, and whereas Type-1 peaks are enriched in housekeeping genes, Type-2 peaks contain many genes with tissue-specific expression [29,30,34]. Additionally, different neuronal cell types shows groups of longer, highly expressed genes that localize away from NS [33,36].

Mapping RNAs associated with NS:

RNAs localized near NS have also been measured using either APEX-seq [26] or, more recently, ARTR-seq, which combines immunostaining with in situ reverse transcription [37]. NS-proximal RNAs share similar properties to NS-proximal genes; they are short, have short introns, and have high GC-content [26,37]. Indeed, 70% of NS-proximal RNAs are transcribed from genes in the SON TSA-seq top two deciles [26]. However, specific NS-proximal RNAs also show increased NS association even away from their sites of transcription, suggesting an inherent NS targeting of these RNAs [26,37].

NS-proximal RNAs show higher intron retention [26,37] and stably NS-enriched RNAs are more enriched in slowly and post-transcriptionally spliced introns and minor splice sites [37]. NS-proximal RNAs with intron retention code for proteins with functions associated with RNA processing, translation, and the cell-cycle [26] while RNAs which either stably or transiently localize in NS code for proteins associated with mRNA metabolism and nuclear localization [37]. Genes close to NS contain more “level” exon-intron junctions and show more intron-skipping events; genes further from NS contain more “differential” exon-intron junctions and show more exon-skipping events [38]. Moreover, distinct classes of splicing factors bind to RNAs produced from these NS-proximal “level” genes versus from distant “differential” genes [26,38]; many of those splicing factors binding NS-proximal RNAs are themselves NS-enriched [10,26]. NS-proximal RNA intron retention varies across cell types and during the cell cycle [26].

A rationale for the observed enrichment of genes with slowly spliced introns near NS [37] would be if NS proximity increased their splicing efficiency. Indeed, using both RD-SPRITE and psoralen cross-linking, a multi-fold higher level of snRNAs was seen over genomic regions closest to NS. Even controlling for the high transcription of these regions, the splicing efficiency of these NS-proximal RNAs was higher [39].

NS as gene expression hubs: moving from correlation towards causality

Moving beyond spatial correlation, live-cell imaging of heat-shock induction of HSPA1 transgenes established a strict temporal correlation of an observed amplification of gene expression of the HSPA1 transgenes always after first NS contact and a decrease in gene expression within minutes after movement away from NS [40,41]. RNA FISH suggested a similar gene expression amplification with NS contact for the endogenous HSPA1 locus, several HSPA1 flanking genes, and HSPH1 [30,41]. A similar gene expression amplification with NS proximity was suggested for the several hundred p53-inducible genes which, like HSPA1 [30,41], are prepositioned near NS prior to induction [42]. Meanwhile, NS dissolution by double SON and SRRM2 KD decreased the expression of ~500 mostly NS-proximal genes (~80% in top SON TSA-seq decile) [43], again linking NS proximity to increased gene expression.

The mechanisms by which NS proximity increase gene expression remain unknown. HSPA1A and HSPA1B are intronless, so splicing is not a factor for these genes. NS are surrounded by a higher density of RNA pol2 Ser2p foci [4], and NS-associated HSPA1 alleles are adjacent to a larger number of more intense RNA pol2 Ser2p foci [41]. In a different gene context, proximity of large, stable RNA pol2 condensates correlated spatially and temporally with the number of nascent RNAs per burst and burst frequency in a super-enhancer and cohesin-dependent manner [44]. Additionally, knockdown of the exosome decreased the relative expression differences of HSPA1B NS associated versus non-associated alleles, suggesting NS contact can protect nascent transcripts from exosome-mediated degradation [41]. Exon splicing enhancers (ESEs) within intronless RNAs like HSPA1A promote their NS targeting and nuclear export [45]. Because RNA export factors compete with the exosome for nascent transcript binding [46], their high NS concentrations might both reduce the degradation and enhance the export of NS-associated gene transcripts [41].

Increased RNA splicing may also increase expression of NS-proximal genes. SON KD increased intron-retention of ~10% of the NS-associated RNAs with intron-retention [26], while a double SON/SRRM2 KD increased intron retention from RNAs stably or transiently interacting with NS [37]. To more directly test the effect of NS-proximity on splicing efficiency, pre-mRNAs from transiently transfected plasmids were tethered to NS using splicing factors concentrated in NS [39]. Increased splicing was observed for NS-tethered RNAs but not for RNAs tethered to splicing factors which do not concentrate in NS.

NS as Hubs for Nuclear Genome Organization:

Beyond the targeting of specific genes to and near NS, the highly reproducible targeting of SPADs to NS has strong implications for overall nuclear genome organization. Computational models of chromosomes as polymer chains using data-driven constraints from Hi-C and DamID data show coalescence of known SPADs in the nuclear interior in these models, suggesting an important contribution of NS in establishing genome radial positioning [47]. Including particle-based interactions of NS and nucleoli in polymer models of chromosomes, and chromosomal interactions with these bodies and the nuclear lamina, more explicitly showed how the differential association of chromosome regions with NS strongly contributes to both genome radial positioning and interior NS localization; here both Hi-C and SPIN states [48], based on Hi-C, DamID, and TSA-seq, were used as input data model constraints [49].

Early immuno-FISH experiments suggested NS might bring specific regions from different chromosomes in close proximity [50,51]. Now both SPRITE and Hi-C demonstrate high frequencies of inter-chromosomal SPAD interactions [27,28,39,52]. More specifically, NS-associated inter-chromosomal Hi-C interactions were significantly enriched in super-enhancers conserved across cell types [52].

How SPADs target to NS is unknown. Proline-rich domains (PRDs) within both p53 [42] and HIF2alpha [53] were required to reposition NS-proximal p53 and NS-proximal HIF2alpha inducible genes even closer to NS after either p53 or HIF2alpha activation, respectively. Sequence comparisons led to identification of similar PRD features across ~1700 proteins related to gene expression and/or NS-enriched [53].

Meanwhile, a different search for NS-targeting factors identified six protein-binding motifs enriched in the NS-hub identified by high interchromosomal contact frequencies [52]. Knockdown of one of these proteins, Myc-associated Zinc finger protein (MAZ), decreased NS-association of several conserved super-enhancers and ~40% of speckle-hub regions and also downregulated ~700 genes which showed decreased speckle-hub association [52]. MAZ and CTCF binding sites frequently colocalize, stabilizing CTCF binding and increasing cohesin interaction; independently of CTCF, MAZ binds and arrests cohesin and shows insulation-activity [54,55].

Interestingly, CTCF binding correlates strongly overall with SON TSA-seq [29] and, more specifically, with SON TSA-seq peaks [56]. CTCF knockdown resulted in a small reduction in SON TSA-seq peak amplitudes; likewise, a slight increase in NS distances of selected NS proximal but not distal genes to NS was observed after both CTCF and cohesin knockdown [56]. Deletion of a PRD, putative speckle-targeting motif (STM) within the RAD21 cohesin subunit also reduced NS association of these same speckle-associated genes [56]. Additionally, both CTCF and cohesin knockdown significantly reduced SON local recruitment over CTCF and other sites contained within SPADs. Thus, RAD21 binding to NS via its STM was proposed to recruit CTCF sites and the DNA between these sites closer to NS, increasing inducibility of speckle-proximal genes [56]. A second example of DNA loop association with NS was suggested by analysis of previously unoccupied CTCF sites that gained CTCF binding after DNA demethylation by DNMT1 inhibition [43]. These “reactivated” CTCF sites map largely to loop anchors interacting with stripe and “highly looped” anchors engaged in multiple loops. Highly looped anchors, CTCF-dependent genes, and reactivated CTCF sites all are enriched near NS as measured by SON TSA-seq, and these reactivated CTCF sites show CTCF-dependent localized recruitment of SON as measured by SON Cut&Tag. This CTCF-associated DNA looping itself is independent of NS [43].

Together, these two experiments suggest a role of CTCF and cohesin in strengthening NS-association of specific gene loci, possibly leading to higher local recruitment of SON and other NS-factors, including to NS-proximal CTCF sites, perhaps through CTCF-bound RNA [57].

NS in the context of physiology and disease:

Broader roles of NS are now emerging across a spectrum of physiological contexts and diseases.

Multiple reports tie NS to stress responses. All heat shock (HS) genes reposition closer to NS after HS, with multiple heat shock genes prepositioned very close to NS (top several percentile) prior to HS [30]. Additionally, some heat shock gene RNAs target to NS through a pathway that expedites their nuclear export [45]. This expedited export contrasts with the suggested retention in NS of many RNAs after HS-induced intron retention [37]. Gene expression amplification after NS contact was demonstrated for several heat shock genes [30,40,41], while elimination of NS through SON/SRRM2 KD reduced cell viability after HS [37]. Meanwhile, CTCF depletion appeared to decrease HS-inducibility specifically of NS-proximal genes [56]. Another physiological stress, hypoxia, reduces NS size through a SRSF6-dependent pathway proposed to have tumor suppressor activity [58].

After another stress response induced by ribotoxins and downstream of p38 MAPK signaling, there is a multi-fold increased targeted excision of Immediate Early Gene (IEG) retained introns accompanied by a noticeable NS-enrichment of some of these IEGs [59]. These IEG retained introns appear to be an example of gene regulation through physiological control of post-transcriptional splicing of “detained introns [60]. Approximately 1/3 of NS-associated retained introns have been annotated as detained introns [26], including a subset whose splicing varies during the cell cycle and which are associated with cell-cycle regulated expression [26]. NS themselves show a 12-hour XBP1-dependent ultradian cycle in morphology changes, possibly driven by changes in NS liquidity, which may be related to a similar 12-hour cycle in SON expression [61]. XBP1 is a key modulator of the unfolded protein response (UPR) and similar changes in SON expression and NS morphology may amplify the UPR and protect against proteome stress [61,62]. In another possible link between changes in NS physical-chemical properties and altered cell physiology, arginine depletion during inflammation decreased ArgRS NS localization, increased SRRM2 mobility, and induced alternative splicing leading to altered cell metabolism and peptide-presentation to immune cells [63].

Recently, ~75% of several hundred proteins identified as NS-enriched by the Human Protein Atlas were found to have known disease connections [64]. Human Phenotype Ontology terms frequently associated with these NS mutated genes, including SON (ZTTK syndrome) and SRRM2, encompass global developmental delay and various categories related to brain developmental defects [64]. Alternative splicing, influenced by many NS-enriched proteins, has been associated with cell fate determination in the developing cerebral cortex [65]. A neuron-specific long isoform of hnRNP D-like (L-DL) localizes in NS and regulates RNA splicing of synaptic genes, which in turn reduces cognitive decline related to aging and Alzheimer’s Disease (AD) [66]. Nuclear aggregates of Tau protein, responsible for several neuropathies, are recruited to NS and alter their structure and composition, while cytoplasmic Tau aggregates mis-localize several NS proteins and small RNAs [67].

Future Challenges and Directions:

The recent progress and renewed interest in NS biology has been fueled by the new ability to map DNA and RNA proximity to NS genome-wide. We anticipate improved throughput of TSA and APEX cytological proximity assays as well as their combined use with molecular proximity assays such as Cut&Tag to help relate NS cytological proximity to changes in local molecular recruitment of NS components. Conceptually, we are still faced with how to interpret results from various manipulations of NS components or understanding whether particular disease-linked mutations involving NS proteins alter actual NS biology or only the action of these proteins outside of NS. We see the need for more sophisticated manipulations of NS that don’t change the functions of NS components at their local recruitment sites outside of NS as well as increased use of live-cell imaging to establish temporal ordering of events such as gene expression and RNA trafficking relative to NS. A new emerging paradox is the unexpected correlation of changes in gene expression with changes in gene positioning across a range of NS distances. We suggest NS may be surrounded by a larger neighborhood of condensates interacting with NS that together influence the nuclear organization of gene expression.

Figure 1. Multifaceted roles of nuclear speckles.

Schematic showing nuclear speckle functional contributions to gene regulation in metazoan nuclei.

Table 1. List of human pathologies related to mutation/altered expression/altered localization of nuclear speckle proteins.

This table, building on the “speckleopathies” described in reference 64 and additional disease connections described here, lists disease/syndrome names, a digital identifier or source reference, the associated nuclear speckle protein, and the causal mechanism. The disease association is based on information available in OMIM (Online Mendelian Inheritance in Man) database and MalaCards: The Human Disease Database.

Disease/syndrome	Digital identifier	Affected protein	Cause
Acrofacial dysostosis 1, Nager type	OMIM#154400	SF3B4	Heterozygous missense mutations
Acute lymphoblastic leukemia (B-ALL)	MCID#LKM062, MCID#PRS040	SRRM1	Upregulation of SRRM1 expression
Autosomal dominant limb-girdle muscular dystrophy-3 (LGMDD3)	OMIM#609115	HNRNPDL	Heterozygous mutations
Clear cell renal cell carcinoma (ccRCC)	Alexander et al, 2023 (Ref. 48)	HIF-2α	Hyperactivation HIF-2α
Congenital heart defects, dysmorphic facial features, and intellectual developmental disorder (CHDFIDD)	OMIM#617360	CDK13	Heterozygous missense mutations
Syndromic neurodevelopmental disorder	PMID: 34050707	DDX23	Missense mutations
Hepatocellular carcinoma	PMC6704583 PMC8900045	WTAP, ZNF207	Upregulation of WTAP expression, ZNF207 as immunotherapy target
Myeloid Malignancies	PMC6966670	LUC7L2, SRSF2, PRPF8, SF3B1, U2AF1	Frequently Mutated
Triple negative breast cancer	PMID:28416606	TRA2A	Overexpression, Promotion of chemotherapy resistance
Pan-cancer	PMC 9465192	RBM39	Overexpression, prognostic marker, multiple cancers
Huntington’s disease	OMIM#612941	PRPF40A/HYPA	Altered localization of HYPA/NAKAP/Huntingtin complex
Intellectual developmental disorder, autosomal dominant 72	OMIM#620439	SRRM2	heterozygous loss-of-function mutations in gene
Intellectual developmental disorder, autosomal recessive 56	OMIM#617125	ZC3H14	Mutated ZC3H14 gene
Intellectual developmental disorder, X-linked syndromic, Hackman-Di Donato type	OMIM# 301039	NKAP	Hemizygous NKAP mutations
Neurodegenerative diseases	Lester et al, 2021 (Ref 62)	SRRM2	Nuclear tau aggregates, relocating SRRM2 from speckles
TAR syndrome	OMIM#274000	RBM8A	Null mutations in RBM8A gene
TARP syndrome	OMIM#311900	RBM10	Mutations in RBM10 gene
Usher syndrome	Likely candidate	LUC7L2	Readthrough mutation
X-linked periventricular heterotopia (PVNH1)	OMIM#300049	FLNA	Hypomorphic or null mutations
ZTTK syndrome	OMIM#617140	SON	Heterozygous mutation in gene