Purification and biochemical characterization of interchromatin granule clusters

Paul J Mintz, Scott D Patterson, Andrew F Neuwald, Chris S Spahr, David L Spector

Author Affiliations

  1. Paul J Mintz1,2,
  2. Scott D Patterson3,
  3. Andrew F Neuwald2,
  4. Chris S Spahr3 and
  5. David L Spector*,2
  1. 1 Department of Molecular Genetics and Microbiology, S.U.N.Y. Stony Brook, Stony Brook, NY, 11794, USA
  2. 2 Cold Spring Harbor Laboratory, 1 Bungtown Road, Cold Spring Harbor, NY, 11724, USA
  3. 3 Amgen Center, MS14‐2‐E, 1 Amgen Center Drive, Thousand Oaks, CA, 91320‐1789, USA
  1. *Corresponding author. E-mail: spector{at}
View Full Text


Components of the pre‐mRNA splicing machinery are localized in interchromatin granule clusters (IGCs) and perichromatin fibrils (PFs). Here we report the biochemical purification of IGCs. Approximately 75 enriched proteins were present in the IGC fraction. Protein identification employing a novel mass spectrometry strategy and peptide microsequencing identified 33 known proteins, many of which have been linked to pre‐mRNA splicing, as well as numerous uncharacterized proteins. Thus far, three new protein constituents of the IGCs have been identified. One of these, a 137 kDa protein, has a striking sequence similarity over its entire length to UV‐damaged DNA‐binding protein, a protein associated with the hereditary disease xeroderma pigmentosum group E, and to the 160 kDa subunit of cleavage polyadenylation specificity factor. Overall, these results provide a key framework that will enable the biological functions associated with the IGCs to be elucidated.


The mammalian cell nucleus is a highly organized structure in which DNA replication, transcription, pre‐mRNA processing, ribosome biogenesis and RNA transport take place (for reviews, see Spector, 1993; Lamond and Earnshaw, 1998). The biochemical constituents and molecular mechanisms of these processes are reasonably well understood. However, less is known about how each of these processes is coordinated spatially and temporally within the structural framework of the nucleus.

At present, only a few nuclear structures have been isolated by biochemical fractionation, i.e. nucleoli, nuclear pore–lamina complex and heterogeneous nuclear ribonucleoprotein particles (for a review, see Singer and Green, 1997). The isolation of these structures has facilitated the identification and functional characterization of their constituents, which enabled their biological roles to be substantiated. However, there are numerous other subnuclear structures including coiled bodies, the perinucleolar compartment (PNC), gems, cleavage bodies, promyelocytic leukemia (PML) bodies and interchromatin granule clusters (IGCs) which are less well characterized at the functional level because thus far it has not been possible to isolate them biochemically (for reviews, see Spector, 1993; Lamond and Earnshaw, 1998).

Pre‐mRNA splicing factors have been localized by immunofluorescence microscopy in a speckled nuclear distribution pattern which corresponds to perichromatin fibrils (PFs) and IGCs at the electron microscopic level (for a review, see Spector, 1993). PFs are fibrillar structures measuring 3–5 nm in diameter and are often found at the periphery of IGCs and distributed throughout the nucleoplasm (Monneron and Bernhard, 1969). In contrast to IGCs, PFs are labeled rapidly after short pulses with [3H]uridine, suggesting that they represent nascent transcripts (for a review, see Fakan, 1994). In support of this view, PFs are sensitive to RNase digestion and are reduced in cells treated with drugs that inhibit transcription by RNA polymerase II (Monneron and Bernhard, 1969; Petrov and Sekeris, 1971).

IGCs, first identified by Swift (1959), represent the major part of the speckled staining pattern. IGCs measure 0.3–1.8 μm along their longest dimension (Thiry, 1995) and are composed of granules measuring 20–25 nm in diameter that appear to be connected in places by a 9–10 nm fiber (for reviews, see Fakan and Puvion, 1980; Spector, 1993). IGCs contain little to no DNA and are highly resistant to nuclease, detergent and high salt extractions (Monneron and Bernhard, 1969; Spector et al., 1983). Bromo‐UTP and [3H]uridine incorporation studies have shown little to no labeling within IGCs, demonstrating that they are not centers of active transcription (for a review, see Spector, 1993). However, labeling has been found at the periphery of IGCs (Fakan and Nobis, 1978; Spector, 1990). Consistent with these findings, several studies have shown highly active genes and their transcripts to be associated with the periphery of speckled domains (for a review, see Huang and Spector, 1997). However, this association does not appear to be essential, as transcription can occur throughout the nucleoplasm (Zhang et al., 1994).

Splicing factors have been shown to be recruited from IGCs to sites of active transcription (Jiménez‐Garcıacute;a and Spector, 1993; Huang and Spector, 1996; Misteli et al., 1997). The molecular mechanism(s) for the recruitment of splicing factors to the sites of transcription is not completely understood. However, it has been shown that phosphorylation of the RS domain of SR proteins is involved in releasing these factors from IGCs and that dephosphorylation is at least part of the signal for the return of these factors to the IGCs (Misteli and Spector, 1996; Misteli et al., 1998). Recently, the C‐terminal domain (CTD) of the largest subunit of RNA polymerase II has been shown to be involved in the intranuclear targeting of splicing factors to transcription sites in vivo (Misteli and Spector, 1999). Truncation of the CTD prevented the accumulation of splicing factors at a newly formed site of transcription, resulting in the inhibition of pre‐mRNA splicing.

While much has been learned about the nuclear distribution of IGCs (for reviews, see Fakan and Puvion, 1980; Spector, 1993; Thiry, 1995), the complete protein composition of these nuclear structures is not known. Elucidating the protein composition of IGCs is central to revealing the functions associated with this nuclear structure. We developed a biochemical nuclear fractionation protocol that enabled us to separate IGCs from chromatin, nuclear lamins, nucleoli and other nuclear bodies. We employed a novel mass spectrometry strategy and peptide microsequencing to identify rapidly many of the 75 proteins that are enriched in this nuclear structure. These data provide a key framework that will enable the biological functions associated with the IGCs to be elucidated.


Biochemical purification and morphological characterization of IGCs

Adult mouse liver nuclei contain 20–30 irregularly shaped IGCs each measuring ∼0.3–1.8 μm along their longest axis (Figure 1A). In order to elucidate the biochemical composition of IGCs, we have developed a cell fractionation procedure to isolate IGCs from mouse liver nuclei (Figure 1B). Nuclei were isolated by sucrose gradient sedimentation (Berezney and Coffey, 1977). This procedure produced clean nuclei with little to no cytoplasmic contamination (Figure 2A).

Figure 1.

(A) Transmission electron micrograph of mouse liver nucleus. Thin section of a nucleus showing interchromatin granule clusters (IGCs) (arrowheads). Bar: 1 μm. (B) Flow chart of the biochemical purification of IGCs.

Figure 2.

Morphological characterization of the IGC fractionation. All samples were post‐stained using the EDTA‐regressive method which preferentially stains ribonucleoproteins (Bernhard, 1969). (A) The nuclear pellet contains well preserved nuclei. The IGCs, nucleoli (N), nuclear lamina (L) and chromatin (C) are readily observed. (B) Thin section of the nuclear pellet after 1% Triton X‐100, DNase I and 0.5 M NaCl treatments. The IGCs are still well preserved, whereas the nucleoli appear partially extracted. Little to no chromatin is present in the extracted nuclei. (C) The cesium sulfate pellet contains nuclear lamina (L), nucleoli (N) and residual nuclear components (asterisk). Bars: 2 μm.

In order to purify IGCs away from other nuclear structures, we proceeded first to remove the nuclear membrane by using a non‐ionic detergent, 1% Triton X‐100 (data not shown). Next, DNA was digested with DNase I and chromatin was released by several sequential extractions with 0.5 M NaCl. After the nuclease digestion and salt extractions, three major types of nuclear structures still remained visible: nuclear lamina, nucleoli and IGCs (Figure 2B). The IGCs remained well preserved whereas the nucleoli appeared to have been extracted after the 0.5 M NaCl treatments. In order to separate the IGCs from other nuclear structures and release them from the nuclear framework, the nuclei were disrupted mechanically in the presence of 5 mM dithiothreitol (DTT) and the resultant homogenate was layered onto a 0.25 M cesium sulfate solution followed by low‐speed centrifugation. The cesium sulfate pellet contained predominantly nuclear lamina, nucleoli and residual nuclear components (Figure 2C). The IGCs, which were enriched in the cesium sulfate supernatant, were retrieved by high‐speed centrifugation (Figure 3A and B). Figure 3A shows a low magnification electron micrograph of a thin section of the IGC pellet revealing a relatively homogenous population of IGCs and little to no apparent contamination by other nuclear structures. The isolated interchromatin granules measured between 18 and 27 nm in diameter (Figure 3B), similar to their in situ counterparts. To confirm that the purified IGCs contained splicing factors, thin sections were immunolabeled with 3C5 monoclonal antibody (Turner and Franchi, 1987), which recognizes a family of SR proteins, followed by colloidal gold‐conjugated secondary antibody. The colloidal gold particles labeled the IGCs (Figure 3C), whereas the control sections, which were not incubated with primary antibody, showed little to no labeling (Figure 3D). These data demonstrate, at a structural level, the high degree of purity of the IGC fraction.

Figure 3.

Transmission electron micrographs of the purified IGC fraction. (A) Low magnification view of the purified IGCs showing a homogenous population Bar: 0.5 μm. (B) High magnification view of the IGCs showing intact and well‐preserved interchromatin granules. Differences in staining intensity reflect the plane of section. Bar: 200 nm. (C) The IGC fraction was labeled with monoclonal antibody 3C5, which recognizes phosphorylated SR proteins, followed by 5 nm colloidal gold‐conjugated secondary antibody Bar: 200 nm; inset: 50 nm. (D) Control section labeled only with the secondary antibody shows little to no immunogold labeling. Bar: 200 nm.

Biochemical composition of IGCs

To investigate the protein composition of each step in the IGC purification, we performed SDS–PAGE using 4–20% gradient gels. Such analyses showed a complex protein composition in all of the fractions examined (Figure 4A). Different proteins were enriched in various fractions. For example, a group of proteins in the 10–15 kDa range were enriched in the Triton pellet, several proteins in the 60–70 kDa range were enriched in the cesium sulfate pellet, and a group of three proteins around 220 kDa were enriched in the IGC pellet. Previous studies have used immunopurification to obtain highly specific macromolecular complexes (e.g. hnRNPs) or organelles (Kvalheim et al., 1987; Piñol‐Roma et al., 1988; Saucan and Palade, 1994). In order to address whether the protein composition of the purified IGC pellet fraction is specific, and not due to other proteins randomly co‐sedimenting with the IGCs, we have purified IGCs further by using magnetic beads coupled with a splicing factor antibody (3C5). The immunopurified IGCs showed a protein composition similar to that of the isolated IGC fraction when examined by SDS–PAGE (Figure 4A and B, lanes 5 and 1, respectively). As shown in the control lanes (Figure 4B, lanes 2 and 3), there was little to no binding of the IGCs to beads that did not contain coupled primary antibody. Since the protein composition of the immunopurified fraction was similar to that of the initial IGC pellet fraction, all further analyses were performed on the IGC pellet fraction.

Figure 4.

One‐ and two‐dimensional gel electrophoresis. (A) All pellet fractions from the IGC purification were separated on a 4–20% SDS gradient gel. Equal amounts (40 μg) of protein were loaded. (B) Magnetic beads coupled with 3C5 monoclonal antibody were used to purify the IGCs further. Immunopurified IGCs (lane 1) show a protein composition similar to that of the biochemically purified IGCs. Control beads (lanes 2 and 3) show no non‐specific proteins. (C and D) The IGC or nuclear pellets were first separated by 2.7% polyacrylamide using pH 3.5–10 ampholites in the first dimension followed by 12.5% SDS–PAGE in the second dimension. The proteins were visualized by silver staining. (C) Nuclear pellet (100 μg). (D) IGC pellet (100 μg). Approximately 75 protein spots were enriched in the IGC fraction as indicated by the circled regions.

In order to identify at greater resolution the constituents of the IGCs, purified IGCs were analyzed by two‐dimensional gel electrophoresis. Approximately 300 protein spots were identified in the IGC fraction (Figure 4D), 75 of which were enriched when compared with the nuclear fraction (Figure 4C). The degree of enrichment for each spot varied. For example, a series of three spots just above the 20 kDa marker were almost undetectable in the two‐dimensional gel of whole nuclei and were significantly enriched in the IGC fraction (Figure 4C and D). Two additional spots, just below the 10 kDa marker, were present in the nuclear fraction and were significantly enriched in the IGCs.

Since IGCs are known to contain pre‐mRNA splicing factors, we were interested next in determining their presence in the purified IGC fraction as a means of further assessing its purity. This was accomplished by immunoblotting the purified IGC fraction. Several pre‐mRNA splicing and non‐splicing factors including SF3a66, SF2/ASF, U2AF65 and U2AF35 were enriched in the IGC fraction (Figure 5A, top row). We found that several other pre‐mRNA processing‐related proteins, U1‐70K, U2‐B″ and PAB II, while present in the IGC fraction, were not enriched (Figure 5A, bottom row). This may be due to their more loose association with the IGCs. Interestingly, the hyperphosphorylated form of the largest subunit of RNA polymerase II is also present in the IGC fraction, but is not enriched. This result was not unexpected since it has been shown that this form of the polymerase is present in nuclear speckles (Bregman et al., 1995). Several other nuclear proteins that have been shown by immunofluorescence to be localized to other nuclear compartments or to be distributed diffusely throughout the nucleoplasm, such as hnRNP A1, lamins A and C, ribosomal protein S6 and the PML protein show little to no immunoreactivity in the IGC fraction (Figure 5B). It must be pointed out here that it is not possible to quantitate the fold enrichment for the IGCs accurately since thus far all of the proteins that have been localized to this structure are also present diffusely throughout the nucleoplasm. The diffuse nuclear pool represents factors that are at transcription sites as well as those that are in transit between IGCs and transcription sites.

Figure 5.

Immunoblot analysis of the IGC fraction. (A) Immunoblots in the upper row show several pre‐mRNA splicing factors that are enriched in the IGC fraction. The lower row shows several other splicing factors and RNA polymerase II to be present, but not enriched, in the IGC fraction. (B) Immunoblots using antibodies that recognize proteins associated with other cellular compartments show little to no immunoreactivity in the IGC fraction. Equal amounts of protein (10 μg) were loaded.

To understand better where in our fractionation scheme proteins associated with different nuclear compartments are enriched preferentially, we immunoblotted all of the pellet fractions from each step in the IGC purification. Figure 6A shows that lamin B1 is enriched predominantly in the cesium sulfate pellet, consistent with our observations of this fraction by transmission electron microscopy (Figure 2C). Interestingly, nucleolar protein B23 (nucleophosmin) (Ochs et al., 1983) was not enriched in the IGC or cesium sulfate pellets (Figure 6B, lanes 5 and 4, respectively). Most of nucleolar protein B23 was removed by the three 0.5 M NaCl extractions (Figure 6B, lane 3). Previous studies have shown that some hnRNPs are also extractable from nuclei at 0.5 M NaCl (Beyer et al., 1977; Piñol‐Roma et al., 1988). We examined the distribution of hnRNP C1/C2 proteins across our fractionation and found that most of these proteins were removed by the nuclease digestion and salt extractions (Figure 6C, lane 3). We have also examined the distribution of SR proteins and found that most of the SR proteins were enriched in the IGC fraction when compared with the nuclear pellet (Figure 6D, lanes 5 and 1, respectively). In addition, there was little to no immunoreactivity in the cesium sulfate pellet. These data, along with the SDS–PAGE analysis, strongly support the view that the IGCs were biochemically purified and demonstrate the complex protein composition of IGCs.

Figure 6.

Further characterization of purified IGCs. Immunoblots were probed with antibodies against (A) lamin B1, (B) nucleolar protein B23, (C) hnRNP C1/C2 or (D) 3C5 antibody to a family of SR proteins. Proteins associated with different nuclear structures were enriched in different fractions.

Mass spectrometry analysis

In recent years, analysis by mass spectrometry has been a powerful approach used to identify the protein components of large macromolecular complexes rapidly by subjecting individual gel bands to analysis (for reviews, see Lamond and Mann, 1997; Patterson, 1998). For example, several recent reports have used this approach to identify the protein constituents of in vitro assembled spliceosomes (Neubauer et al., 1998), yeast U1 small nuclear ribonucleoprotein particles (Neubauer et al., 1997) and the yeast spindle pole (Wigge et al., 1998). Here we have used a novel mass spectrometry approach to identify many of the IGC proteins rapidly. The entire IGC protein mixture was digested enzymatically and subjected to liquid chromatography electrospray ionization tandem mass spectrometry (LC‐MS/MS) in a data‐dependent manner followed by uninterpreted fragment ion searching of non‐redundant and expressed sequence tag databases (dbEST). This is the first time that a large and complex protein mixture has been digested and subjected to mass spectrometry analysis en toto. Purified IGCs (denatured, reduced and alkylated) were first subjected to partial trypsin digestion. The complex peptide mixture was fractionated by HPLC with on‐line ion‐trap electrospray mass spectrometry. In two separate runs, we have thus far identified 33 known proteins (Table I) and ESTs encoding at most 16 proteins (to date) after searching a non‐redundant protein database or dbEST (NCBI and DDBJ/EMBL/GenBank) with the uninterpreted MS/MS spectra. A group of 19 proteins (identified from 71 spectra), previously shown to be localized to IGCs, is composed largely of splicing factors. One member of this group, a previously characterized RNA/DNA‐binding protein (RNPS1) has been identified recently as SF7A, a general activator of pre‐mRNA splicing (A.R.Krainer and A.Mayeda, personal communication). In addition to splicing factors, two proteins related to different types of cancer were identified. A peptide sequence corresponding to a region of the Skip protein was identified. Skip has been shown to be associated with the Ski oncoprotein (Dahl et al., 1998). The Skip protein was also identified when spliceosomes were analyzed by mass spectrometry (Neubauer et al., 1998) and, in addition, it was shown to be localized to nuclear speckles. A second protein, TLS‐associated serine‐arginine (TASR) protein (Yang et al., 1998), which is associated with the TLS (translocated in liposarcoma) protein, was also identified in the IGC fraction. In addition, our analysis identified a subset of proteins (14 identified from 22 MS/MS spectra) that have been characterized in a different biological context and thus far have not been shown to be present in IGCs (Table I). Some of these proteins (e.g. histones and ribosomal proteins) may be contaminants based upon their previous functional associations. However, we cannot rule out the possibility that they may also be minor constituents of the IGCs. Finally, we have identified several new proteins (Table I) and ESTs (data not shown) associated with the IGCs.

View this table:
Table 1. Proteins identified in purified IGC fraction

The identified peptide sequences were used to screen protein databases. Two peptide sequences SLLPNSSQDELMEVEK and LGLIQEDVASSCIPR (matched to Amgen EST sequences) were used to identify a full‐length human cDNA clone, KIAA0324, after searching a non‐redundant protein database (DDBJ/EMBL/GenBank accession No. AB002322). The open reading frame (ORF) of KIAA0324 contains many RS/SR dipeptide repeats and a putative nuclear localization signal (data not shown). We did not find any potential homologs of KIAA0324 in the databases. The predicted molecular mass of this protein is ∼125.6 kDa and it is a highly basic protein (pI = 12). To verify that this protein is a bona fide constituent of the IGCs, a fusion protein was made with yellow fluorescent protein (YFP) and expressed in HeLa cells by transient transfection (Figure 7A). The transfected cells were also immunostained with anti‐SC35 monoclonal antibody which diagnostically identifies nuclear speckles (Figure 7B). The expressed fusion protein co‐localizes with the endogenous SC35 splicing factor (Figure 7C). In addition, the fusion protein was also present in the cytoplasm, suggesting that it may shuttle.

Figure 7.

Characterization of KIAA0324 and KIAA0017. The human full‐length KIAA0324 (A–C) or KIAA0017 (D–F) cDNAs were fused to yellow fluorescent protein (YFP). HeLa (A–C) or BHK (D–F) cells were transiently transfected and immunostained with the anti‐SC35 (B) or anti‐B″ (E) antibodies. The fusion proteins localized in nuclear speckles and diffusely in the cytoplasm. (C) and (F) are the respective merged images of (A) and (B) and (D) and (E). Bars: 10 μm; 5 μm.

A third peptide (AVTIATPATAAPAAVSAATTTSAQEEPAAAPEPR) was matched to a full‐length mouse cDNA clone called Plenty‐of‐Proline (POP) (DDBJ/EMBL/GenBank accession No. AF062655). The POP protein is highly basic (pI = 11.9) and has a predicated molecular mass of ∼101 kDa. The ORF of POP contains many RS/SR dipeptide repeats, several proline‐rich regions and a putative NLS (data not shown). A homolog to POP, SRm160, was identified (DDBJ/EMBL/GenBank accession No. AF048977), which is a human splicing co‐activator previously localized to speckles (Blencowe et al., 1998). Using a sequence alignment program (NCBI, BLAST 2 Sequences), POP shares 90% identity and 92% similarity with SRm160 (data not shown). We have also made a POP–YFP fusion protein and have confirmed that POP is a bona fide component of the IGCs (data not shown). The remaining 16 ESTs currently do not match to any identified full‐length cDNAs.

Microsequence analysis of the 140 kDa protein

In addition to mass spectrometry, peptide microsequencing of gel bands was used to identify new proteins of the IGC fraction. An enriched 140 kDa gel band of the IGC fraction was excised from a Coomassie‐stained SDS–polyacrylamide gradient gel for microsequencing. The peptide sequences obtained (KFVIHPESNNLIIIETD and KNVSEELDRTPPEVSK) were used to search a non‐redundant protein database, and a full‐length cDNA clone, KIAA0017, was identified (DDBJ/EMBL/GenBank accession No. D13642). The protein is acidic (pI = 5.1) and has a predicted molecular mass of ∼136.6 kDa. The ORF of KIAA0017 contains a putative NLS (data not shown); however, no other known motifs were identified.

Database searching using the KIAA0017 sequence has allowed us to identify closely related proteins in other metazoans (Figure 8) (Caenorhabditis elegans and Drosophila melanogaster), in a plant (Arabidopsis thaliana) and in a protozoan (Plasmodium falciparum), suggesting that human protein KIAA0017 performs an important, perhaps essential, biological function that arose early during eukaryotic evolution. Given their high degree of sequence identity to KIAA0017 (43–61%), these proteins are likely to perform the same function in these diverse organisms as KIAA0017 does in humans. Notably, even budding yeast, which lacks IGCs, encodes a member of this family. The function of this yeast protein may be somewhat distinct from that of these other proteins, however, given its lower degree of sequence identity to KIAA0017 (25% identity).

Figure 8.

Alignment of protein sequences related to the human hypothetical protein KIAA0017. The alignment was generated using PSI‐BLAST (Altschul et al., 1997) and conserved residues were highlighted using an automated procedure (Neuwald et al., 1999); minor adjustments in the alignment were made to improve readability. Protein designations are color coded by family using the following scheme: KIAA0017, red; UV‐DDB, blue; CPSF 160 kDa subunit, black. Hydrophobic conserved residues are highlighted in yellow; other conserved residues in red; poorly conserved or unconserved regions in gray. Gaps within the alignment are indicated as integer lengths. DDBJ/EMBL/GenBank identifiers (and organisms) corresponding to sequence designations are: IGC_human, 3540219; IGC_worm, 2804455 (C.elegans); yme9_yeast, 2497090 (S.cerevisiae); uv_arab, 2911067 (A. thaliana); uv_dicdi, 2130171 (D. discoideum); uv_human, 1136228; uv_pombe (S.pombe), 2330717; uv_worm, 3878704 cpsf_human, 1706102; cpsf_pombe, 3738146; cpsf_yeast, 2132494.

A biological function for the KIAA0017 protein family is suggested by weaker, yet clearly significant, sequence similarity to two other protein families: a family of known and putative UV‐damaged DNA‐binding factors (UV‐DDB) (E‐value = 10−27) and a family of RNA cleavage and polyadenylation specificity factors (CPSF, 160 kDa subunit) (E‐value = 10−10). Of these two families, KIAA0017 is more closely related to UV‐DDB proteins which are associated with the hereditary disease xeroderma pigmentosum group E (Keeney et al., 1994).

The genomes of higher eukaryotes appear to encode one member from each of the KIAA0017, UV‐DDB and CPSF (160 kDa subunit) protein families. This is seen, for example, for three metazoans, human, C.elegans and Drosophila, whose genomes are close to being sequenced completely. Thus, all three proteins may be present in most, if not all, metazoans. Similarly, the vascular plant Arabidopsis also encodes all three proteins, although the Arabidopsis proteins closely related to KIAA0017 and CPSF currently are available only as short, partially sequenced fragments. By contrast, budding yeast, whose entire genome has been sequenced, lacks an UV‐DDB‐like protein; the lack of this protein in yeast suggests a potential functional overlap between KIAA0017 and UV‐DDB. Based on available complete genomic sequences, all three families appear to be absent from archaea and eubacteria, which is consistent with a distinctly nuclear role for these proteins.

Remarkably, the similarity between these three families, though characterized by scattered regions of sequence conservation, extends essentially over the entire length of the proteins (Figure 8). This is demonstrated through PSI‐BLAST searches using profiles derived from four distinct, adjacent ∼300 residue regions of the KIAA0017 family alignment; each of these searches easily detects the UV‐DDB and CPSF families (E‐values = 10−6), with the exception of one region that is absent in CPSF proteins. This region, which maps to residues 575–705 of KIAA0017 (Figure 8, bracket), corresponds to an unrelated region within CPSF proteins; surprisingly, even CPSF proteins from different organisms appear unrelated in this region. By contrast, the 575–705 region of KIAA0017 is conserved, albeit weakly (E‐value = 0.0002), in UV‐DDB proteins, again suggesting greater functional similarity between these two families. As a whole, the co‐linear motifs conserved in these proteins, which are very likely to correspond to multiple domains, appear to be present only in these three families and are conspicuously absent from other currently sequenced proteins. This suggests an unusual degree of functional dependency between these conserved domains.

To confirm that the KIAA0017 protein is a constituent of the IGCs, the protein was tagged with YFP and expressed in BHK cells by transient transfection (Figure 7D). The expressed fusion protein (Figure 7D) co‐localizes with the endogenous U2 snRNP B″ protein (Figure 7F), confirming the relevant localization of the expressed fusion protein. In addition to being localized in the nucleus, the fusion protein was also present in the cytoplasm.


Previous studies have used immunofluorescence, in situ hybridization and electron microscopy to assess the biological function of nuclear speckles (IGCs) in the mammalian cell nucleus (for a review, see Misteli and Spector, 1998). These studies have suggested that nuclear speckles play an important role in RNA metabolism based upon the localization of pre‐mRNA splicing factors to this nuclear domain. Recently, nuclear speckles were shown to be dynamic in living cells and to provide splicing factors to sites of transcription which are located on their periphery or some distance from a speckle (Misteli et al., 1997). In order to advance our understanding of nuclear speckles (IGCs) significantly, we have taken a biochemical cell fractionation approach to isolate IGCs from mouse liver nuclei. Here we report the first successful biochemical purification and characterization of IGCs and the identification of three new IGC proteins.

Enormous progress has been made in our understanding of the functional and structural organization of eukaryotic cells. In particular, the highly compartmentalized nature of the cytoplasm and the functions associated with both membrane‐bound (mitochondria, Golgi apparatus, endoplasmic reticulum, microbodies, etc.) and non‐membrane‐bound (ribosomes, microtubules, microfilaments, intermediate filaments, centrioles, etc.) organelles have been identified readily using standard microscopy approaches and fractionation procedures developed early on to purify and characterize these structures (Novikoff and Holtzman, 1970). However, the functional organization of the nucleus remains one of the black boxes in cell biology, primarily because it has been difficult to isolate and characterize biochemically many of the identified nuclear structures such as coiled bodies, gems, cleavage bodies, the PNC, PML bodies, IGCs, PFs, etc. Recently, Singer and Green (1997) argued that nucleoli and the NPC, both of which are non‐membrane‐bound organelles, are bona fide nuclear compartments because they can be identified microscopically and purified biochemically. According to these criteria, IGCs now also qualify as a bona fide nuclear organelle.

Since the identification of interchromatin granule clusters 40 years ago by Swift (1959), a single published attempt has been made to isolate this nuclear structure (Zborek et al., 1990). However, the isolated IGCs could not be fractionated away completely from a large number of nuclear structures including nucleoli, nuclear lamins and other residual nuclear components. We have employed a series of approaches in an effort to characterize purified IGCs morphologically and biochemically. We have isolated IGCs successfully from other nuclear components (lamins, nucleoli and chromatin) without compromising the structural integrity of the granules. Approximately 300 protein spots were identified in the IGC fraction using two‐dimensional gel electrophoresis. We have used a novel mass spectrometry methodology and protein microsequencing to identify the constituents of IGCs. Thus far, 33 known proteins have been identified as well as several uncharacterized ESTs. Of the 33 known proteins, 19 have been shown to be present in the speckles and, as expected, many of them are splicing factors. These data strongly demonstrate that we have purified a highly enriched IGC fraction. Recently, Neubauer et al. (1998) have employed mass spectrometry and have identified 19 new proteins associated with an in vitro assembled spliceosomal complex. Several of the proteins we have identified in the IGC fraction, including Skip, U2AF35, U2AF65, hnRNP C and U2‐snRNP B″, were also found in the purified spliceosome complex. However, members of the SR family that are enriched in the IGCs were not identified in the purified in vitro assembled spliceosome fraction (Neubauer et al., 1998). One interpretation of these data is that the SR proteins are more tightly associated with the IGCs than they are in the in vitro assembled spliceosome.

Most interestingly, analysis of the IGC fraction has allowed us to identify 17 proteins with unknown functions. Thus far, we have investigated three of these unknown proteins (KIAA0324, POP and KIAA0017) and found that they localize in vivo in speckles, in addition to being distributed diffusely throughout the nucleoplasm. Interestingly, the KIAA0324 and KIAA0017 proteins were also localized in the cytoplasm, consistent with the possibility that they may shuttle between the nucleus and cytoplasm. A growing number of nuclear proteins have been shown to shuttle between the nucleus and cytoplasm (for a review, see Nakielny and Dreyfuss, 1997). After searching a non‐redundant protein database, we found that the KIAA0017 protein has sequence similarity throughout its ORF with a UV‐damaged DNA‐binding (UV‐DDB) factor in human, slime mold, C.elegans, Schizosaccharomyces pombe and A.thaliana, suggesting that all of these proteins are related. The UV‐DDB protein (127 kDa) is involved in the damage recognition step of the nucleotide excision repair (NER) pathway (Hirschfeld et al., 1990; Treiber et al., 1992; Otrin et al., 1997). Our finding of an additional cytoplasmic pool of KIAA0017 is consistent with a recent report by Watanabe et al. (1999) showing that the UV‐DDB protein also interacts with the cytoplasmic domain of the Alzheimer's amyloid precursor protein (APP) in co‐immunoprecipitation experiments. At present, the precise biological role of UV‐DDB is unknown. In addition to the UV‐DDB protein, the 160 kDa subunit of CPSF from Saccharomyces cerevisiae, S.pombe and human cells was also found to have conserved sequence similarity to the KIAA0017 ORF. The sequence similarity over their entire length suggests that all three families share common structural/functional constraints that have been evolutionarily conserved. Taken together, this sequence similarity between several distant species suggests that KIAA0017 is involved in one or more important biological functions, possibly associated with DNA repair and RNA processing.

Although the biological relevance of IGCs has been well established during the past 40 years, questions have remained concerning the biological nature of nuclear speckles, giving rise to the suggestion that speckles are the result of exaggerated antibody labeling conditions or artificial threshold effects during digital image acquisition (Fay et al., 1997; Neugebauer and Roth, 1997). Numerous studies have presented conclusive evidence that speckles are bona fide non‐membrane‐bound organelles. First, nuclear speckles (IGCs) can be visualized as a unique structure in the electron microscope without the need for antibody labeling (Swift, 1959; Monneron and Bernard, 1969). Secondly, nuclear speckles can be visualized in living cells using splicing factor–GFP fusion proteins (Misteli et al., 1997). Finally, as we now show, nuclear speckles can be purified biochemically, and new protein members of the isolated speckles are localized to these structures when their cDNAs are expressed in living cells. Our biochemical strategy will enable the elucidation of the functions associated with this nuclear organelle and advance our overall understanding of the functional/structural organization of the nucleus.

Materials and methods

IGC purification

Livers (29–33 g) from 20 five‐week‐old Swiss Webster female mice were washed once in ice‐cold TM5 buffer (10 mM Tris–HCl, pH 7.4 and 5 mM MgCl2, pH 7.4) containing protease inhibitors [1 mM phenylmethylsulfonyl fluoride (PMSF), 5 μg/ml leupeptin and aprotinin (Sigma, St Louis, MO)] and minced. A 6 ml aliquot of 0.25 M sucrose in TM5 buffer was added to 2 ml of minced liver in a 55 ml Dounce homogenizer (Wheaton) and homogenized using the ‘A’ pestle. The homogenate was transferred to a 50 ml tube (Corning, NY) and was filtered twice through two layers of cheesecloth. An equal volume of 2.2 M sucrose in TM5 buffer containing 0.2 mM spermine and 0.5 mM spermidine was added to the homogenate and mixed vigorously. The homogenate was transferred to 30 ml conical centrifuge tubes (Oak Ridge, TN) containing 5 ml each of 2.2 M sucrose (Berezney and Coffey, 1977) and was centrifuged at 39 000 g for 1.5 h at 4°C. The nuclear pellet was washed once with TM5 buffer. On average, 4–6×108 nuclei/ml were recovered. Small samples (3 μl) of the nuclear pellet were taken for examination by phase‐contrast microscopy and transmission electron microscopy.

To isolate IGCs, 4 ml of a nuclear suspension (5×108 nuclei/ml) were treated with Triton X‐100 to a final concentration of 1% in the presence of 2 mM vanadium ribonucleoside complex (Gibco‐BRL, Frederick, MD) for 5 min on ice and centrifuged at 780 g for 5 min. The supernatant (and all subsequent supernatants) was removed and analyzed by SDS–PAGE. The pellet was resuspended in a final volume of 4 ml with TM5 buffer containing protease inhibitors (5 μg/ml leupeptin and aprotinin) and 3 μl of the pellet was set aside for analysis by transmission electron microscopy (Triton pellet). A 880 μl aliquot of the resuspended Triton X‐100 extracted nuclei was placed into four microcentrifuge tubes and 20 μl of RNase‐free DNase I (10 U/μl) (Boehringer Mannheim) was added and incubated for 1 h at 4°C on a slow rotator. The digested chromatin was removed by addition of 5 M NaCl, to a final concentration of 0.5 M, and incubated on ice for 5 min, then centrifuged at 770 g for 5 min at 4°C. Each pellet was resuspended in 500 μl of 0.5 M NaCl in TM5 buffer and treated as above twice. Three microlitres of the final pellet (DNase I/0.5 M NaCl pellet) was analyzed by transmission electron microscopy. Each pellet was resuspended in 450 μl of 0.5 M NaCl in TM5 buffer and divided into two microcentrifuge tubes. The volume was adjusted to 955 μl with 0.5 M NaCl in TM5 buffer, and 5 μl of 1 M DTT was added and incubated for 5 min on ice. The homogenates were passed 10 times through a 27‐gauge needle, pooled and homogenized 20 times using a tight pestle (1 ml Dounce homogenizer, Wheaton). Aliquots (100 μl) of the homogenate were transferred to 16 microcentrifuge tubes containing 500 μl of 0.25 M Cs2SO4 in TM5 buffer and centrifuged at 20 800 g for 2 min, and the supernatant containing the IGCs was transferred to new tubes. Each pellet was resuspended with 20 μl of TM5 buffer and analyzed by transmission electron microscopy (Cs2SO4 pellet). To each supernatant, 200 μl of TM5 buffer was added and divided equally into six 3 ml polycarbonate tubes (Beckman) and centrifuged at 157 000 g in a Beckman Optima TLX ultracentrifuge (TLA rotor 100.3) for 1 h at 4°C. The supernatant was removed and discarded. Each pellet was resuspended in 30 μl of TM5 buffer and transferred to a single microcentrifuge tube. One of the pellets was used for analysis by transmission electron microscopy (IGC pellet). All of the remaining pellets were sonicated for 1–3 min at a constant cycle using a water bath sonicator with a Branson sonifier model 450.


Two‐step purification of antibody 3C5 was done by ammonium sulfate precipitation and DE52 anion exchange chromatography (Harlow and Lane, 1988). The purified 3C5 antibody was coupled to Dynal M‐450 rat anti‐mouse IgM magnetic beads (Dynal, Lake Success, NY) in phosphate‐buffered saline (PBS) (pH 7.4) for 18 h at room temperature, after which coupled beads were washed four times in PBS. In addition, two controls were performed as follows: 190 μg of purified IgMλ and IgMκ was coupled to the magnetic beads or PBS without the primary antibody. Then 250 μg of the IGC fraction was added to the coupled magnetic beads and the final volume was adjusted to 300 μl with PBS (containing 5 μg/ml leupeptin and aprotinin) followed by a 24 h incubation at 4°C on a slow rotator. The beads were washed five times in PBS with protease inhibitors (5 μg/ml leupeptin and aprotinin) and resuspended in 250 μl of PBS. The samples (70 μl) were dissolved in SDS sample buffer (Spector et al, 1998) and the beads were removed by brief centrifugation prior to loading onto a 4–20% SDS gradient gel.

Transmission electron microscopy

All of the pellets from the IGC fractionation were microcentrifuged quickly and fixed with freshly prepared 2% formaldehyde and 0.5% glutaraldehyde in PBS (pH 7.4) (Spector et al., 1998). LR‐White sections (170–200 nm) were cut on a Reichert‐Jung Ultracut E ultramicrotome with a Diatome diamond knife and stained using the EDTA‐regressive method (Bernhard, 1969). Immunogold labeling was performed according to previously published procedures (Spector et al., 1998). The samples were examined at 75 kV in a Hitachi H‐7000 transmission electron microscope.

One‐ and two‐dimensional gel electrophoresis and immunoblotting

Prior to measuring the protein concentration and loading the samples on gels, all pellets were sonicated for 10 s at a constant cycle using a water bath sonicator with a Branson sonifier model 450. The protein concentration was determined by the BCA Protein Assay (Pierce, Rockford, IL), and γ‐globulin (Bio‐Rad, Hercules, CA) was used as a standard. Proteins in the pellet and supernatant fractions from the IGC purification were separated on 4–20% SDS gradient gels (Jule Inc., New Haven, CT) and stained with Coomassie Blue R‐250 or transferred to nitrocellulose as described (Misteli and Spector, 1996). Immunoreactive bands were detected by chemiluminescence using the Chemiluminescence Plus system (NEN, Boston, MA). Equilibrium two‐dimensional gel electrophoresis was performed by the 2‐D Gel Core Facility at Cold Spring Harbor Laboratory.

Mass spectrometry and peptide microsequencing

A 850 μg aliquot of IGC sample was thawed, denatured in 6 M urea/45 mM Tris–HCl, pH 7.0, reduced with 10 mM DTT for 30 min at 37°C, and then alkylated with 20 mM iodoacetamide for 30 min at room temperature. The sample was diluted to give a final concentration of 2 M urea, then 10 μg of sequencing grade trypsin (Boehringer Mannheim, IN) was added and the sample incubated overnight at 37°C. The peptide digest was acidified with 1% formic acid and subjected to solid‐phase extraction using an Oasis cartridge (Waters Corp., MA). The cartridge was washed with 100% MeOH, equilibrated with 0.1% formic acid, and the peptide digest loaded. The cartridge was washed using 0.1% formic acid/7% MeOH, then peptides were eluted using 100% MeOH. The sample was dried, then resuspended in 0.1% formic acid. Aliquots corresponding to ∼220 μg (original estimate) were separated by capillary reverse‐phase HPLC, using a YMC 0.5 mm×150 mm C18 column (Western Analytical, Temecula, CA) with an LCQ ion‐trap mass spectrometer (Finnigan, San Jose, CA) connected on‐line. The mass spectrometer was operated in the data‐dependent manner (triple‐play), whereby the most intense ion in each MS scan was selected for high‐resolution zoom scan analysis to determine the charge state, followed by MS/MS of this selected ion. The dynamic exclusion option was utilized with a mass width of 5 Da for 1.5 min, to reduce the possibility of fragmenting the same ion repeatedly. The mode of operation, instrument settings and data handling are detailed elsewhere (Courchesne et al., 1998). All MS/MS spectra were searched using an uninterpreted fragment ion search program against an in‐house non‐redundant database and a six‐way translation of dbEST (Courchesne et al., 1997).

For microsequencing gel bands, the IGC fraction was separated by 4–20% SDS gradient gels and stained with Coomassie G. The enriched 140 kDa gel band was excised and subjected to Lys‐C endopeptidase digestion, and the resulting peptides were separated by HPLC and automated Edman degradation (Protein Chemistry Core Facility, Cold Spring Harbor Laboratory).

Sequence analysis

The BLAST‐2 and PSI‐BLAST programs (Altschul et al., 1997) were used for searches of the NCBI non‐redundant protein database and the TBLASTN program was used for searches of dbBEST and DDBJ/EMBL/GenBank. The command line version of blastpgp with checkpointing (Altschul et al., 1997) was used to construct PSI‐BLAST profiles of specific protein families. PSI‐BLAST profiles of one family were used to assess the statistical significance of sequence similarity to members of another family.

YFP fusion protein constructs

The ORFs of KIAA0017 and KIAA0324 in pBluescript SK+ were excised with SacII and StuI or with SalI and SacII, respectively, and inserted in‐frame into pEYFP‐C1 (Clontech Laboratories, Inc., Palo Alto, CA). Each fusion protein contained YFP at its C‐terminus.

Cell culture and transient transfection assay

HeLa and BHK cells were grown on 100 mm culture dishes in Dulbecco's modified Eagle's medium (Gibco‐BRL, Gaithersburg, MD) containing 10% fetal bovine serum, 50 U/ml penicillin and 50 U/ml streptomycin at 37°C in 10% CO2. Confluent HeLa cells were collected and the KIAA0324–YFP construct was transfected by electroporation of 5 μg of DNA and 15 μg of salmon sperm DNA at 230 V using a Bio‐Rad Gene Pulser II (Hercules, CA). Cells subsequently were seeded onto glass coverslips in 35 mm Petri dishes and grown for 14 h. For the KIAA0017–YFP construct, BHK cells were transfected and grown for 14 h.

Immunofluorescence and light microscopy

The transfected cells on coverslips were fixed for 15 min at room temperature in freshly prepared 2% formaldehyde in PBS (pH 7.4) according to previously published procedures (Spector et al., 1998). Anti‐SC35 monoclonal antibody was used at a dilution of 1:1000 (Fu and Maniatis, 1990) and anti‐B″ monoclonal antibody was used at a dilution of 1:70 (Habets et al., 1989). Texas red‐conjugated goat anti‐mouse secondary antibody (Cappel Laboratories, Cochranville, PA) was used at a dilution of 1:100. Images were acquired using a Zeiss LSM 510 confocal laser scanning microscope with a Kr/Ar laser using simultaneous dual scans.


We are grateful to Tamara Howard for her excellent electron microscopy assistance. We thank Paula Bubulya, Adrian Krainer, Thoru Pederson and Noriko Saitoh for critical reading of the manuscript, Thoru Pederson and Mona Spector for critical suggestions, and Laura Mintz for editorial assistance. We thank the following individuals for providing antibodies: Gideon Dreyfuss (hnRNP C1/C2, hnRNP A1), Adrian Krainer (SF2/ASF), Tom Maniatis (U2AF65, U2AF35), Angela Kramer (SF3a66), Steve Warren (H5), Elmar Wahle (PAB II), Roel van Driel (PML) and Brian Turner (3C5). We thank Ryuji Kobayashi for peptide microsequencing and Aigoul Nourjanova for two‐dimensional gel electrophoresis. We thank Jacques Camonis for providing the POP cDNA. We are grateful to Takahiro Nagase (Kazusa DNA Research Institute, Japan) for providing the KIAA0017 and KIAA0324 cDNAs. We thank Toshiro Tsukamoto for helpful advice in cloning the YFP fusion constructs. D.L.S. is funded by a grant from the National Institute of General Medical Sciences (GM42694).


View Abstract