Nuclear speckles – a driving force in gene expression

Nuclear speckles are prominent RNA–protein granules or nuclear bodies found in the nucleoplasm of eukaryotic cells (Fig. 1). They are enriched in RNA-binding proteins (RBPs) that belong mainly to the different subsets of pre-mRNA splicing factors (Perraud et al., 1979; Spector et al., 1991, 1983). From an historical perspective, the first mention of these structures is attributed to Santiago Ramón y Cajal, who identified them using light microscopy (Cajal, 1910). Later, they were mostly known as inter-chromatin granule clusters due to their granular appearance and non-random distribution between chromatin regions, as detected by electron microscopy (Swift, 1959; Thiry, 1995a,b). The ‘speckle’ term was coined when nuclei were stained using immunofluorescence with serum from autoimmune disease patients, and a somewhat speckled pattern was observed, but the antigen producing the pattern was unknown at the time (Beck, 1961). Currently, antibodies to splicing factors are used to endogenously observe nuclear speckle structures in fixed cells. A common nuclear speckle marker is SC35, a splicing factor detectable by a monoclonal antibody that was generated using purified spliceosomes and was later determined as identifying the serine–arginine-rich (SR) protein SRSF2 (also referred to as SC35; Fu and Maniatis, 1990, 1992). However, a recent study has demonstrated that this antibody mostly recognizes SRRM2, which is a spliceosome-associated protein. Moreover, SRRM2 and another nuclear speckle component, SON, are the two main components necessary for upholding the nuclear speckle structure (Ilik et al., 2020). These two proteins are the most enriched components within nuclear speckles (Dopie et al., 2020). Nuclear speckles harbor other molecules, such as phosphatidylinositol phosphates (Osborne et al., 2001; Sobol et al., 2018) and RNAs, the latter typically detected using RNA fluorescence in situ hybridization (FISH) to identify polyA tails, but the exact nature of all these RNAs is not completely clear (Calado and Carmo-Fonseca, 2000; Carter et al., 1991; Fay et al., 1997; Huang and Spector, 1991; Puvion and Puvion-Dutilleul, 1996; Smith et al., 2020; Visa et al., 1993). Metastasis associated lung adenocarcinoma transcript 1 (MALAT1) is a long non-coding RNA (lncRNA) distinctly found in nuclear speckles, although it is not required for upholding the nuclear speckle structure (Hutchinson et al., 2007; Nakagawa et al., 2012). Indeed, super-resolution microscopy has highlighted the substructure of the nuclear speckle, showing a core containing the proteins SON, SRRM2 and a transcriptional corepressor called REST corepressor 2 (RCOR2), which is involved in neuronal differentiation, while some non-coding RNAs were found to be located at the periphery of the nuclear speckle (Fei et al., 2017; Rivera et al., 2021).

Several fundamental questions regarding the general understanding of nuclear body structure and function, and specifically nuclear speckle biology, have driven much research in the past decades, putting some issues to rest and raising new ones. The main issues and questions addressed herein are as follows. (1) Dynamics: imaging experiments have shown that the nuclear speckle structure has flexibility and can undergo morphological changes – does this reflect on their possible physiological roles, and what do these changes mean with regard to the dynamics of the components? (2) Composition: which molecules reside within nuclear speckles (proteins and RNAs) and could knowledge of this hint to their biological roles under normal conditions and perturbed states? (3) Biophysical properties: how are these structures maintained as separate entities in the nucleoplasm if they do not have a surrounding membrane? (4) Assembly: how do nuclear speckles form, and is a core scaffold molecule required to initiate nucleation? (5) Function: nuclear speckles are usually neighbors of active genes – does this have a biological meaning? (6) Disease: is there evidence for a connection between nuclear speckles and diseased states? (7) Viral infection: why do nuclear speckles reorganize during different types of viral infections? In this Review, we will summarize the current knowledge on nuclear speckle structure and function with respect to genome organization.

Nuclear speckles are of irregular shape. There can be several tens of them, of varying sizes, in the nucleus at a given time. Numbers and sizes can differ between cell types (Spector and Lamond, 2011). They are positioned in the central area of the nucleus and are not found in the nuclear periphery (Carter et al., 1993; Chen et al., 2018b). During mitosis nuclear speckles disassemble, although several mitotic inter-chromatin granules (MIGs) remain in the cytoplasm to pass their splicing factor content back into the reassembling nucleus (reviewed in Spector and Lamond, 2011). With the advent of fluorescent labeling of proteins in living cells, it became possible to examine nuclear speckles in a live cellular context in order to understand their dynamic behavior with relation to gene expression activity (see question 1 above). Viewing nuclear speckles in real time during interphase within nuclei of living cells by use of a fluorescent splicing factor showed dynamic movements of small particles associating and disassociating at the edges of the nuclear speckles, suggesting that there is continuous entry and exit of splicing factors to and from nuclear speckles (Misteli et al., 1997). Measuring the association and dissociation rates of splicing factors with nuclear speckles was accomplished using photobleaching approaches, demonstrating a constant flux of proteins moving in and out of the nuclear speckles. Most intriguingly, the proteins involved were found to exhibit high mobility properties; namely, exchange rates were quite rapid, in the range of ∼30 s, and the residence times of splicing factors within the nuclear speckle were less than a minute (Kruhlak et al., 2000; Phair and Misteli, 2000). When transcription or splicing are inhibited in the cell, the nuclear speckle structure transitions from an irregular undefined shape to a fully rounded and enlarged body filled with splicing factors (Antoniou et al., 1993; Carmo-Fonseca et al., 1991; Ferreira et al., 1994; Melcak et al., 2000; O'Keefe et al., 1994; Shav-Tal et al., 2001; Spector et al., 1991, 1983); however, splicing factors continue to shuttle between the nuclear speckles and the nucleoplasm (Rino et al., 2007). Mathematical simulations have shown that this enlargement of nuclear speckles is due to higher chances of binding and association of the splicing factors with nuclear speckles, since fewer binding sites are available in the nucleoplasm (Rino et al., 2007). Taken together, studies in living cells have shown that the components of nuclear speckles are in constant flux in and out of the structures and that the physiological state of the cell can affect the internal dynamics of nuclear speckles and transform the concentration of components within.

Despite the dynamic properties of the nuclear speckle structure and continuous exchange of components with the nucleoplasm, it is possible to purify the bodies from cells (Mintz et al., 1999; Saitoh et al., 2004). Subsequent mass spectrometry analysis in these purification studies has helped to discover the protein composition of nuclear speckles, determining which protein families tend to reside and possibly function within (see question 2 above). The above studies using biochemical approaches, as well as microscopy studies (Fong et al., 2013), have detected over 350 proteins. However, the degree of protein loss from the nuclear speckles during these procedures is unclear, and more recently, enzyme-based proximity labeling (in the form of tyramide signal amplification, TSA) has been used to label nuclear speckle proteins within the cells. This approach eliminates loss of components during purification and has enabled the identification of new ones (Dopie et al., 2020). Taken together, these studies have shown that the factors passing through and associating with nuclear speckles are players in the gene expression pathway. In addition to proteins involved in splicing, one can find factors associated with transcription, mRNA processing and modification, mRNA export, and messenger ribonucleoprotein (mRNP) packaging (Fig. 2).

Intriguingly, nuclear bodies including nuclear speckles are not surrounded by membranes yet are retained as separate and distinct entities within the nucleoplasm, even though many of their components are also diffuse in the nucleoplasm. How is this possible (see question 3 above)? The mystery behind this phenomenon has been resolved in recent years with the understanding that these types of condensates, in the nucleus and in the cytoplasm, form by a process termed phase separation (Banani et al., 2017; Strom and Brangwynne, 2019). Phase-separated membraneless bodies are held together by biophysical forces based on weak interactions between the molecules within (Box 1). Nuclear speckles are membraneless condensates of RBPs and RNAs, and are considered to have the properties of liquid–liquid phase-separation (LLPS) bodies (Dopie et al., 2020; Ilik et al., 2020; Kim et al., 2019; Marzahn et al., 2016; Zhang et al., 2016), although additional biophysical models should also be considered (Belmont, 2021; Razin and Gavrilov, 2020). For instance, one specific factor, the deubiquitylase USP42, has been identified as a protein that creates phase-separated droplets and localizes to nuclear speckles by protein–protein interactions based on positive charges in its C terminus, most likely interacting with negatively charged proteins in the nuclear speckle (Liu et al., 2021). USP42 is responsible for integration of the PLRG1 spliceosome factor into the nuclear speckle. Downregulation of USP42 impairs mRNA splicing and attenuates cell proliferation. Interestingly, depletion of different nuclear speckle factors can modify the overall structure. Reduction of USP42 levels disrupts phase separation and nuclear speckle formation altogether (Liu et al., 2021), whereas reductions in the level of SON causes the formation of doughnut-shaped nuclear speckles (Sharma et al., 2010), and depletion of MFAP1 increases nuclear speckle size (Dopie et al., 2020). Other nuclear condensates can influence nuclear speckles. mRNA export factors are found in nuclear speckles and can govern the export of certain mRNAs (Dias et al., 2010; Fan et al., 2018; Mor et al., 2016; Schmidt et al., 2006; Wang et al., 2018). ZFC3H1 is part of the nuclear exosome complex that prevents the export of defective mRNAs, sending them for degradation (Meola et al., 2016). It has been demonstrated that when the nuclear exosome complex is inactivated, ZFC3H1 forms nuclear condensates that trap mRNAs, denying them the possibility to pass through nuclear speckles, and subsequently retaining them in the nucleus (Wang et al., 2021).

Weak interactions can be formed by stretches of amino acids that have negative or positive charge (for example, the RS domain in splicing factors) or by low complexity intrinsically disordered regions (IDRs) that appear in many RBPs (see Box 1). This understanding has helped define the initiation steps required for the assembly of nuclear bodies in general, and nuclear speckles specifically (see question 4 above). Indeed, it has been suggested that the initiation of the formation of the nuclear speckle structure occurs by the stochastic self-assembly of various RBPs, where RNA can serve as an initiating driving force for assembly but where no particular RBP starts the set of events (Bhat et al., 2021; Shevtsov and Dundr, 2011; Tripathi et al., 2012). Still, as mentioned above, there is a degree of order within the structure. Microscopy has shown that MALAT1 lncRNA, as well as spliceosomes containing small nuclear RNAs (snRNAs), concentrate at the peripheral section of the nuclear speckle, whereas some splicing factors localize in the central core (Fei et al., 2017; Girard et al., 2012; Rovira-Clave et al., 2021). Interestingly, the heterogenous nuclear RNP (hnRNP) family of RBPs, which is a general name for many RBPs that bind to RNAs but are not stable components of RNP complexes (Dreyfuss et al., 2002), are not found in nuclear speckles. These findings on substructure and phase separation, together with the fact that many active genes are located near nuclear speckles, have led to development of a model proposing that the pre-mRNA is positioned in a sequence-dependent manner along the speckle–nucleoplasm interface, a region that is defined by phase separation (Liao and Regev, 2021). In this model, exons bound to splicing factors preferentially localize to the nuclear speckle periphery, while introns are prevented from binding due to their association with nucleoplasmic hnRNP proteins. Furthermore, the exon–intron boundaries are primed in position at the interface where spliceosomes are poised to splice. This model helps explain why genes are found in proximity to phase-separated condensates filled with splicing factors, whereas most splicing does not occur inside the condensates. LLPS represents a fascinating and crucial element in the biophysics of the nucleus that has only just begun to be fully understood. The future will bring new insights into nuclear speckle dynamics and formation, and perhaps even LLPS-targeted therapies.

There has been much conjecture on the functional roles that nuclear speckles might play in the nucleus, beginning with the intuitive speculation that, because they contain splicing factors, nuclear speckles are the actual sites of pre-mRNA splicing (Huang and Spector, 1992). However, this suggestion has not panned out; rather, nuclear speckles are found to be associated with sites of high gene activity. It therefore has been imperative to unravel the biological aspects of these neighborhood relationships (see question 5 above). Some studies have suggested that nuclear speckles function as rest stops where mRNAs can attain certain RBPs, such as export factors, as they diffuse out of the nucleus. Alternatively, they can serve as depots of splicing factors, which can be recycled, and control their release to active genes that are some distance away (Hochberg-Laufer et al., 2019a; Misteli et al., 1998). These ideas have been discussed previously in a number of reviews (Chen and Belmont, 2019; Galganski et al., 2017; Hall et al., 2006; Hasenson and Shav-Tal, 2020; Lamond and Spector, 2003); however, no clear-cut function has been determined. Here, we will focus on an evolving theme of examining the association between nuclear speckles and specific genomic regions to determine whether the proximity between nuclear speckles and genes reflects the actual activity levels a gene might reach. Indeed, the many technical innovations and improvements made to genome mapping and imaging approaches in recent years (Table 1) have yielded many surprising insights as to the diverse biology of nuclear speckles.

It became clear quite early on that nuclear speckles do not contain DNA. RNAs have been observed in close association with nuclear speckles, particularly in the vicinity of the periphery of the structures (Bahar Halpern et al., 2015; Hu et al., 2010; Huang and Spector, 1991; Johnson et al., 2000; Jolly et al., 1999; Lampel et al., 1997; Melcak et al., 2001; Puvion and Puvion-Dutilleul, 1996; Smith et al., 2007; Szczerbal and Bridger, 2010; Xing et al., 1993; Xing et al., 1995; Yang et al., 2011; Zhang et al., 1994). Splicing factor assemblies have been observed to be recruited to active intron-containing genes (Brody et al., 2011; Hochberg-Laufer et al., 2019a; Hu et al., 2009; Huang and Spector, 1996; Janicki et al., 2004; Misteli et al., 1997). However, most of these studies examined only small sets of transcript types using microcopy methods. With the advent of molecular approaches such as chromosome conformation capture (3C) techniques, which allow the three-dimensional (3D) high-resolution designation of genome regions in the context of nuclear architecture (Kempfer and Pombo, 2020), it has become possible to examine the rules governing associations between genes and nuclear speckles on a genome-wide basis. One of the major discoveries of the 3C approaches, in particular Hi-C, is the identification of A and B compartments of genome organization (Lieberman-Aiden et al., 2009), with the A compartment relating to active chromatin regions and the B compartment having repressive characteristics. These two compartments have been further partitioned into six subcompartments (A1, A2 and B1–B4; Rao et al., 2014). The association of these compartments with subnuclear structures has been tested; the more repressive parts of the genome have been shown to be in contact with the nuclear lamina and nucleoli (Guelen et al., 2008; Nemeth et al., 2010), whereas nuclear speckles have been found to associate with the active compartments. A study using proximity labeling (TSA-seq) to determine distances between genes and nuclear speckles revealed a surprisingly conserved and deterministic genome organization pattern (Chen et al., 2018b). The distances were found to be highly predictive of gene expression levels, such that proximity correlates with high gene expression and vice versa, strengthening the idea that there is a link between nuclear speckles and regulation of gene expression (Belmont, 2021). TSA-seq analysis found that a fifth of genomic areas are proximal to nuclear speckles and belong to the A1 subcompartment, while A2 subgenomic regions are more distant, and the B subcompartments are even further away (Chen et al., 2018b). These findings are further strengthened by the demonstration that disruption of nuclear speckles alters the expression of 1282 different genes (Hu et al., 2019). Specifically, intra-chromatin interactions in active type-A compartments decrease, while interactions in repressive type-B compartments increase. Although this does not indicate a direct cause and effect relationship, it is clear that nuclear speckles are potent regulators of gene expression.

A different approach called MARGI, which maps interactions between RNA and the genome, used the lncRNAs in nuclear speckles, which tend to localize at the nuclear speckle periphery, as an anchor to determine the genomic sequences interacting with nuclear speckles (Chen et al., 2018a; Sridhar et al., 2017). Use of this approach suggests that the relative positions between nuclear speckles and genomic regions are rather conserved. A technique termed SPRITE, which measures the frequencies of interactions between DNA and RNA in the same region, including nuclear bodies, has shown higher-order spatial contacts of active regions that are enriched for U1 snRNA and MALAT1 lncRNA, both of which are constituents of nuclear speckles (Quinodoz et al., 2018). By focusing on the interactions between RNA and DNA (RD-SPRITE), snRNAs have been found to be spatially enriched near genes transcribed by RNA polymerase II (RNA Pol II) and nuclear speckles (Quinodoz et al., 2021). The power of SPRITE has been further applied in thousands of single mouse cells (scSPRITE) to find that genomic regions that interact with nuclear speckles are more likely to interact with each other (Arrastia et al., 2022). Furthermore, DNA seqFISH+ (sequential DNA FISH coupled with super-resolution microscopy) performed by imaging several thousand gene loci in mouse cells revealed active marks of transcription (including phosphorylation of RNA Pol II at serine 5 and histone acetylation) surrounding nuclear speckles (Takei et al., 2021). Interestingly, FISH to introns of several genes marking transcription sites did not show a high correlation between expression and positioning relative to nuclear speckles. The authors suggest that these genes are not dynamically positioned into transcriptionally active zones, rather, they might be pre-positioned in these zones by means such as epigenetics and genome organization (Takei et al., 2021).

Using multiplexed error-robust FISH (MERFISH), an advanced genomic imaging method, to study genes and RNA at multiple loci, it was observed that the A:B chromatin compartment ratio is significantly higher for actively transcribing genes (Su et al., 2020). Imaging at high-resolution also confirmed previous studies showing an increase in the association of loci of the A compartment with nuclear speckles, and this was also correlated with increased transcriptional activity. Inhibition of transcriptional activity reduced nuclear speckle association rates (Su et al., 2020). In another study, the ubiquitous retrotransposons L1 (also known as LIN1) and B1 (also known as Alu) were found to split the nucleus into exclusive domains, with L1-containing compartments localizing to the lamina and nucleolus, while B1-rich domains associated with the nuclear interior, namely the nuclear speckle (Lu et al., 2021). This suggests a genetically encoded model of chromatin structure, wherein chromatin structure is stabilized by genome attachment to nuclear structures via a phase separation mechanism, enabling the formation of heterochromatin compartments (Lu et al., 2021).

The mechanism that emerges connecting nuclear speckles with gene expression levels is that proximity to a nuclear speckle is achieved by the dynamic properties the 3D structures of genomic loci. The closer a gene is to the nuclear speckle, the higher its transcription and splicing levels, which is probably the result of high concentrations of splicing factors in nuclear speckles – specifically the basal splicing machinery, which is located in the periphery of the nuclear speckle (Brown et al., 2006; Ding and Elowitz, 2019; Guo et al., 2019; Kim et al., 2020; Zhang et al., 2020). This could also drive efficient co-transcriptional splicing (Brody et al., 2011). Besides tethering of genes to nuclear speckle regions via splicing factors and RNAs, it is unclear whether there are other mechanisms that govern association. Strikingly, new evidence shows that nuclear speckles may regulate genes through the direct binding of transcription factors. Indeed, it has been found that the transcription factor p53 (also known as TP53) targets a functionally distinct subset of p53 target genes to nuclear speckles, thereby enhancing their transcription (Alexander et al., 2021). These nuclear speckle-associated genes were found to have higher expression levels than non-speckle-associated p53 gene targets. The association was found to be mediated through the DNA-binding domain of p53 and not through its transactivation domain. One of the genes activated by p53 is p21 (also known as CDKN1A), and p21 expression was found to be significantly reduced when the levels of the nuclear speckle proteins SON or SRSF1 were knocked down, demonstrating the need for the structural and compositional integrity of nuclear speckles to influence gene expression (Alexander et al., 2021). Another study has found that CTCF, a transcription and genome-organizing factor, can move to nuclear speckles upon stress, and this has led to the proposal that CTCF can regulate the movements of certain genomic regions to the periphery of the nuclear speckle as part of cellular programs that control differentiation (Lehman et al., 2021).

Nuclear speckles are found in the nuclear center and not close to the lamina. Two recent studies (Barutcu et al., 2022; Tammer et al., 2022) have examined the organization of the genome focusing on the GC nucleotide content of different regions. GC content is an important parameter in predictions of chromosome positions within the nucleus (Girelli et al., 2020), and GC-rich genomic regions have been detected near nuclear speckles (Chen et al., 2018b). Such studies have suggested that the genome is organized along a nuclear speckle-to-lamina axis, which for instance includes a gradient of active chromatin increasing in concentration towards the center of the nucleus and around nuclear speckles (Crosetto and Bienko, 2020). The two studies have shown that the peripheral and central regions of the nucleus harbor genes with two distinct exon–intron GC-content architectures, and these properties correlate with nuclear speckle association and the coordination of different splicing patterns in the two compartments (Barutcu et al., 2022; Tammer et al., 2022). Genes with exons flanked by long introns, both with low GC content, were found to reside close to the lamina, whereas genes with high-GC-content exons and short flanking introns were found to be more central. It has been suggested that nuclear speckles are responsible for the positioning of genes with high GC content in the central part of the nucleus; in this way, they exert an important role in defining the increased GC content along the nuclear speckle-to-lamina axis, thereby driving high-order genome organization (Tammer et al., 2022). Moreover, nuclear-speckle-associated RNAs have been found to be primarily derived from open chromatin regions in their proximity. These RNAs were found to have specific splicing patterns and to contain retained short introns with high GC content (Lu et al., 2021). In summary, it appears that nuclear speckles act as gene-expression-inducing hubs that contain splicing factors that preferentially interact with specific gene sets that have unique sequence and splicing characteristics.

Possible connections between disease-linked genetic variants and specific nuclear architectures have emerged, implicating changes in 3D genome architecture in a range of pathologies from cancer to limb deformity (Mishra and Hawkins, 2017). This raises the question of whether spatial connections between defective loci and nuclear speckles could account for deficiencies in transcription and splicing, and whether nuclear speckles have a role in disease pathogenesis (see question 6 above) (Galganski et al., 2017). For instance, a potential connection between schizophrenia, which is known to be linked to splicing defects, and a specific nuclear architecture has been found (Fig. 3) (Ahanger et al., 2021). Speckle-associated domains (SPADs) have been found to be highly enriched in schizophrenia-risk-associated loci, as compared to other organizational domains, and even other disease-linked loci, such as those linked to autism spectrum disorder or type 2 diabetes (Ahanger et al., 2021). A different connection has been found in another neurodegenerative pathology, Alzheimer's disease. Tau aggregates, the hallmark pathology of the disease, which form in the cytoplasm and nucleus and contain RNAs, are specifically enriched for snRNAs and small nucleolar RNAs (snoRNAs). Nuclear tau has been found to accumulate in nuclear speckles, altering their dynamics and organization (Lester et al., 2021). Moreover, cytoplasmic tau recruits the nuclear speckle SRRM2 factor, causing its mislocalization, which has been confirmed in brain cells from mouse models and human tissue (Lester et al., 2021; McMillan et al., 2021). Other factors also localize to the aggregates, but no effect on nuclear SON or SR proteins has been observed. Overall, tau aggregates change the balance between cytoplasmic and nuclear RBPs, altering the composition of nuclear speckles, as also seen during different types of cellular stress that lead to formation of stress granule condensates in the cytoplasm (Hochberg-Laufer et al., 2019b; Youn et al., 2018; Zhang et al., 2018). These changes are sufficient to impair pre-mRNA splicing, thus implicating nuclear speckles as a crucial component of tau-mediated pathology.

Aggregations are also known to be mediated by RNAs in trinucleotide repeat expansion disorders (Jain and Vale, 2017). Foci of CUG repeats in myotonic dystrophy type 1 have been shown to accumulate at the periphery of nuclear speckles (Holt et al., 2007; Smith et al., 2007; Taneja et al., 1995). These have been linked to interactions with the alternative splicing factor MBNL1, influencing alternative splicing patterns. Similar observations have been made for Huntington's disease, where foci of CAG repeats have been found to be detained within the nuclear speckle and dependent on its integrity (Urbanek et al., 2016). CGG-repeat foci that form in fragile X syndrome cells have also been associated with nuclear speckles (Sellier et al., 2010). Some repeat expansion diseases yield foci that do not directly interact with the nuclear speckle, such as amyotrophic lateral sclerosis (ALS; Lee et al., 2013), likely owing to the repeats being in excised introns rather than spliced mRNA transcripts. Therefore, repeat foci present a target for prognosis and diagnosis, and disrupting these RNA–RNA foci might be a valuable strategy in treating the underlying disorders. For instance, oligonucleotide-based strategies applied to Huntington's disease cells cause a reduction of RNA foci number, and this readout could potentially be used to assess the end results of these treatments (Urbanek et al., 2017).

Another connection between nuclear speckles and disease is mediated by MALAT1, which is linked to a number of cancers (Goyal et al., 2021). Overexpression of MALAT1 in cancer cells is associated with higher proliferation, invasion and metastasis of tumors, as well as severe impairment of the host immune response (Arun et al., 2020). A large variety of alternatively spliced variants of MALAT1 have been detected, for example in breast cancer (Meseure et al., 2016), suggesting that expression patterns can be used as a predictive biomarker or as a therapeutic target. This supports previous reports that other nuclear speckle factors are upregulated in cancers and that their overexpression is sufficient to increase proliferation and metastasis (Anczukow et al., 2012; Karni et al., 2007). MALAT1 has been found to interact with a variety of fusion proteins that are generated by chromosomal translocations (Chen et al., 2020). These interactions take place in nuclear speckles and result in enhanced export of the chimeric mRNAs and their subsequent translation. Thus, targeting of MALAT1 could potentially be applied to the treatment of cancers that contain driver fusions. Interestingly, the chemotherapeutic drug cisplatin becomes enriched within nuclear speckles and interferes with splicing in treated cells, suggesting that it might be crosslinked to RNAs in the periphery of the nuclear speckle (Rovira-Clave et al., 2021). The clinical relevance of nuclear speckles is still emerging, but it is apparent that they might become valuable in the future for screening, diagnostics and prognostics.

Nuclear architecture undergoes dramatic changes during lytic viral infection, including changes in nuclear bodies. DNA viruses induce chromatin marginalization due to the creation of viral replication compartments (VRCs) in the host cell nucleus (Aho et al., 2017, 2021; Monier et al., 2000; Schmid et al., 2014). How does chromatin reorganization affect nuclear speckles (see question 7 above)? Early studies noticed that nuclear speckles redistribute at the nuclear periphery and are found adjacent to VRCs during herpes simplex virus (HSV-1), Kaposi's sarcoma-associated herpesvirus (KSHV), human immunodeficiency virus (HIV), adenovirus and influenza infections (Alkalay et al., 2020; Francis et al., 2020; Jimenez-Garcia and Spector, 1993; Mor et al., 2016) (Fig. 4). Furthermore, human cytomegalovirus immediate-early 1 (HCMV-IE, also known as UL123) mRNA has been detected close to nuclear speckles, as have spliced exons of adenovirus late RNAs and PAN lncRNA foci of KSHV, whereas Epstein–Barr virus (EBV) genes and human papilloma virus 18 (HPV18) do not associate with nuclear speckles (Bridge et al., 1996; Dirks et al., 1997; Lampel et al., 1997; Vallery et al., 2018). HSV-1 infection leads to the redistribution of nuclear speckles as well as paraspeckles; this has been found to induce the association of SRSF2 with different paraspeckle factors and influence histone modifications located near viral genes (Li and Wang, 2021).

Why do nuclear speckles reorganize upon infection (see question 7 above)? This might be an indirect effect caused by inhibition of either cellular transcription or host RNA splicing (as described above, with rounding up of nuclear speckles during inhibition), which in turn would contribute to the inhibition of cellular gene expression. This would correlate with a view that nuclear speckles do not have a particular function and are simply by-products of gene activity. However, live permeabilized cells still contain nuclear speckles even though gene activity would be perturbed under such conditions (Misteli and Spector, 1996; Spector et al., 1992). Another possibility is that nuclear speckle reorganization is an active process that is at least in part caused by the expansion and movement of VRCs. Live-cell imaging has revealed that VRCs fuse at nuclear speckles, implying that this structural reorganization allows efficient incorporation of available RNA processing factors and/or translocation of the viral RNA to the nuclear speckles, resulting in an optimized hub for processing and export of viral mRNA (Chang et al., 2011). With respect to the proximity of active genomic regions to the periphery of nuclear speckles under regular conditions, viral infection could modify nuclear architecture to distance host genes from nuclear speckles by driving genome marginalization to the nuclear rim, leaving most of the nuclear speckle periphery available to engage with the viral genome, thus enhancing viral transcription. Interestingly, genomic domains associated with nuclear speckles have been found to be preferred integration sites for HIV type 1 (HIV-1). Moreover, HIV-1 VRCs traffic and accumulate within nuclear speckles, which are the preferred destination for HIV-1 VRCs across multiple cell types. The paraspeckle component CPSF6, which is recruited to nuclear speckles under HIV-1 infection, is required for the transportation of HIV-1 to nuclear speckles and also for stabilizing the association of VRCs with nuclear speckles (Francis et al., 2020). Using LLPS, the virus also creates a new membraneless organelle that is distinct from nuclear speckles, containing the viral RNA genome, SRSF2–CPSF6 complexes and reverse transcriptase (Rensen et al., 2021). Interestingly, these bodies contain newly transcribed viral DNA, which challenges the currently held view that reverse transcription takes place in the cytoplasm.

Nuclear speckles are essential for viral RNA transcription, splicing and export (Fig. 4), and nuclear speckle proteins and RNA Pol II are found at sites of viral transcription (Boe et al., 1998; Jimenez-Garcia and Spector, 1993). Some nuclear speckle proteins interact with specific viral promoters and transcripts for transcription and splicing (Cerasuolo et al., 2020; McFarlane et al., 2015; Wang et al., 2016). For example, the M1 transcript of influenza virus associates with nuclear speckles for post-transcriptional splicing, and SON has been found to be crucial for viral replication (Mor et al., 2016). The influenza protein NS1 requires nuclear speckles for host transcription shutoff (Nacken et al., 2021). Interestingly, the cytoplasmic reovirus alters host mRNA splicing of genes responsible for post-transcriptional modifications – the µ2 protein of this virus has been found in nuclear speckles where it interacts with SRSF2, impairing its subnuclear localization (Rivera-Serrano et al., 2017).

Viruses may alter the balance between the concentration of splicing factors in the nucleoplasm versus that in nuclear speckles, which is governed by phosphorylation cycles of factors that determine their release into the nucleoplasm so they can act in splicing (Hochberg-Laufer et al., 2019a; Lamond and Spector, 2003; Misteli et al., 1998). The HSV-1 key regulator protein ICP27 (also known as UL54) blocks the activation of cytoplasmic serine/arginine-protein kinase 1 (SRPK1), which phosphorylates SR proteins, and relocalizes it to the nucleus. The presence of ICP27 in the nucleus leads to hypo-phosphorylation of SR proteins (Sciabica et al., 2003; Tunnicliffe et al., 2019). Furthermore, the human papilloma virus has been found to stimulate SRPK1, resulting in increased levels of phosphorylated nuclear and cytoplasmic SRSF1 (Mole et al., 2020). As for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), SRPK1/2 phosphorylates the N protein, and this event can be inhibited by the FDA-approved kinase inhibitor alectinib to reduce SARS-CoV-2 replication (Yaron et al., 2020 preprint). Moreover, 13 SARS-CoV-2 viral proteins can interact with 51 host RBPs, including SRSFs and hnRNPs (Srivastava et al., 2020). Finally, the function of the snRNPs U1 and U2, core components of the spliceosome, is impaired by the binding of the NSP16 viral protein to suppress host splicing (Banerjee et al., 2020).

With regard to nuclear export, the association and movement of VRCs to nuclear speckles, where they coalesce at the nuclear speckle surface, enhances the export of specific viral transcripts (Chang et al., 2011) (Fig. 4). SRSF3 and SRSF7 are essential for the export of HSV-1 viral RNAs, and knockdown of each factor significantly decreases virus yields, leading to the accumulation of polyadenylated RNA in the nucleus (Escudero-Paunetto et al., 2010). Many viral RNAs are intronless, and some have been found associated with nuclear speckles, presumably to gain nuclear export factors (Wang et al., 2018). For instance, the unspliced M1 influenza transcript is directed to nuclear speckles where it can be spliced to produce the M2 transcript. When M1 localization to nuclear speckles is impaired, the transcript accumulates in the nucleus, implying that its association with nuclear speckles functions in preparing for export (Mor et al., 2016).

The roles of nuclear speckles have recently begun to emerge. Ingenious and novel techniques are being developed at a rapid pace, and each affords an ever-clearer picture of the unique role that nuclear speckles play in gene expression and RNA processing. While the proximity of genes to nuclear speckles clearly enhances transcription and processing of these transcripts, the mechanisms through which specific genes and nuclear speckles find each other and are held together are still unclear. Indeed, nuclear speckles are dynamic entities. Their many components, which are mainly RBPs belonging to the mRNA splicing and export factor families, are in continuous flux with the nucleoplasm. Nuclear speckles can move towards or form in proximity to active genes. Once assembled, they are maintained by LLPS-based mechanisms governed by IDRs that are found in their many RBP components. Viruses might help in this deciphering of the molecular pathways, as many of them use host nuclear speckles for processes that are vital for viral propagation, such as transcription, splicing and export,. The variety of these interactions highlight the complexity of the viral infection pathways, and this is true also for diseased states that involve nuclear speckles. Further examination of these associations will help to unravel not only basic biological pathways underlying gene expression, but also their interplay with biomolecular condensates, pointing to the potential in using nuclear speckles for diagnostics and even therapeutics.

Schemes in the figures were created with BioRender.com.