A non-catalytic role for inositol 1,3,4,5,6-pentakisphosphate 2-kinase in the synthesis of ribosomal RNA

The inositol phosphate kinases IPK2/IPMK and IP5K comprise an evolutionarily ancient signaling pathway that is conserved from yeast to man (Hatch and York, 2010; Seeds et al., 2007). These two kinases co-operate in the synthesis of inositol hexakisphosphate (InsP₆) from the lipid-generated inositol (1,4,5)-trisphosphate [Ins(1,4,5)P₃] signal that mobilizes Ca²⁺ stores. IPMK acts first in this metabolic sequence, by phosphorylating Ins(1,4,5)P₃ to Ins(1,3,4,5,6)P₅, which IP5K then converts to InsP₆. The latter has several functions, including the regulation of DNA repair (Cheung et al., 2008), and the promotion of mRNA export from the nucleus (Alcázar-Román et al., 2010; York et al., 1999). InsP₆ is also an essential structural cofactor for ADAR2, an mRNA editing enzyme (Macbeth et al., 2005), and InsP₆ is a precursor for another group of cellular regulators, the inositol pyrophosphates (Burton et al., 2009).

In an earlier study we demonstrated that, in mammals, IP5K is predominantly a nuclear enzyme (Brehm et al., 2007). That observation has furthered general interest in understanding the significance of nuclear compartmentalization of inositol phosphate signaling (Martelli et al., 2011; Seeds et al., 2007). This topic is complex, because the nucleus is a heterogeneous organelle with several morphologically and functionally distinct sub-compartments (Dundr and Misteli, 2010). One of these is the nucleolus, the site of ribosomal RNA (rRNA) synthesis, which is the initial step in ribosome biogenesis (Drygin et al., 2010). The transcription machinery that generates rRNA responds to a vast array of information from cellular signaling cascades, so as to regulate the ribosome production that guides cell differentiation, proliferation, and adaptation to stress (Drygin et al., 2010). Furthermore, alterations in rates of rRNA synthesis appear to be one of the most important molecular adaptations of cancer cells (Drygin et al., 2010). Thus, there is great interest in understanding how rRNA production is regulated. However, research in this area is not limited to elucidating the catalytic activities that synthesize rRNA. There are also a number of unresolved questions concerning how nucleolar structure is organized. Nucleoli are not membrane-delimited; they are dynamic, multiprotein complexes. Thus, the regulation of nucleolar assembly and disassembly is heavily reliant upon the structural roles of molecular scaffolds (Hemmerich et al., 2011; Vondriska et al., 2004). The identification of these scaffolding molecules, and the characterization of the molecular interactions that they mediate, are therefore also key to understanding the regulation of rRNA synthesis. Furthermore, the study of scaffold proteins in general is vital to determining how the flow of cellular information is controlled (Good et al., 2011). In the current study we show that IP5K is a scaffold that has structurally and functionally essential associations with several key proteins that exert controlling influences upon rRNA production.

Through confocal immunofluorescence of H1299 cells (Fig. 1A), primary human skin fibroblasts (Fig. 1B) and MCF-7 cells (Fig. 1C), we determined that endogenous IP5K was distributed throughout the nucleus, including in nucleoli. The nucleoli were identified both morphologically and by using fibrillarin (Fig. 1A–C) as a marker (Lam et al., 2005). Western analysis also detected IP5K in nucleoli purified from H1299 cells (supplementary material Fig. S1). A nucleolar pool of IP5K was not detected in earlier studies of nucleolar constituents by mass spectrometry (Ahmad et al., 2009). However, the statistical and algorithmic analysis of such data are an imperfect compromise between minimization of false-positives while reducing false-negatives (Tharakan et al., 2010).

We threaded the amino-acid sequence of IP5K through NoD, an algorithm that identifies putative nucleolar localization motifs (Scott et al., 2011), but no sequence in the protein attained the threshold to be considered a candidate (Fig. 1D). We therefore directly screened for nucleolar IP5K binding partners by subtractive proteomic analysis of anti-GFP immunoprecipitates from lysates which were prepared from cells in which either GFP or an IP5K-GFP fusion protein were expressed. A nucleolar resident, TCOF1 (Valdez et al., 2004), was identified as an IP5K binding protein with high confidence (supplementary material Fig. S2A–H). Moreover, we detected some co-localization of endogenous TCOF1 and IP5K in the nucleoli of intact H1299 cells (Fig. 1E). We next overexpressed TCOF1-GFP in H1299 cells, from which we prepared cell lysates for incubation with immobilized GST-IP5K. Western analysis (Fig. 1F) confirmed that TCOF1 bound to IP5K.

TCOF1 is a crucially important regulator of rRNA synthesis through its interactions with upstream binding factor (UBF) (Dixon et al., 2006; Valdez et al., 2004). It is therefore significant that GST-IP5K also bound to endogenous UBF in H1299 cell lysates (Fig. 1F). We also detected partial co-localization of UBF with IP5K in situ by confocal immunofluorescence analysis of H1299 cells, primary fibroblasts, and MCF-7 cells (supplementary material Fig. S3). The actions of both UBF and TCOF1 in vivo are modulated by protein kinase CK2 (CK2) (Isaac et al., 2000; Lin et al., 2006). Indeed, CK2 is a key player in the overall regulation of rRNA synthesis so that the degree of ribosomal production can dynamically match changes in cell growth and proliferation (Drygin et al., 2010). We found GST-IP5K to associate with endogenous CK2 in H1299 cell lysates (Fig. 1F). Another nucleolar protein, nucleolin, did not associate with IP5K (Fig. 1F); the latter result implies that there is some selectivity to the interactions of IP5K with TCOF1, UBF and CK2.

The spatial dynamics of nucleolar proteins are thought to be directly related to the degree of rRNA synthesis (Hernandez-Verdun et al., 2010). We therefore investigated if the organization of the nucleolar pool of IP5K might be sensitive to the inhibition of rRNA synthesis that is induced by serum starvation (Yuan et al., 2002). We assayed rRNA production in H1299 cells on a cell-by-cell basis by lightly permeabilizing the plasma membrane and then labeling nascent RNA with bromouridine (BrU) in a ‘run-on’ assay (Ko et al., 2000). We ensured that BrU was specifically incorporated into rRNA, by blocking mRNA synthesis with α-amanitin (compare supplementary material Fig. S4A with Fig. S4B); furthermore, no signal was observed in the absence of BrU (compare supplementary material Fig. S4C with Fig. S4D). We next serum-starved the H1299 cells to inhibit rRNA synthesis (supplementary material Fig. S4E), following which we added back serum to reinitiate rRNA production (supplementary material Fig. S4F). Little endogenous IP5K was present in the nucleoli of serum-starved cells (Fig. 1G; supplementary material Fig. S5A). The nucleolar pools of IP5K were restored within 8 hours of serum re-addition (Fig. 1H; supplementary material Fig. S5B).

We also more specifically inhibited rRNA synthesis with a low concentration of Actinomycin D (Act D) whereupon components of the Pol I rRNA transcriptional complex segregate into ‘caps’ (Hernandez-Verdun et al., 2010; Shav-Tal et al., 2005) that circumnavigate the nucleolar periphery. After ActD treatment we observed that some IP5K was also present in nucleolar caps (supplementary material Fig. S6A,B). This same phenomenon was observed when IP5K was overexpressed (supplementary material Fig. S6C). It seems unlikely that the changes in the nature of the IP5K signal, following such different procedures as serum-starvation and Act D treatment, both reflect a similar staining artifact such as epitope masking. Instead, our data (Fig. 1G,H; supplementary material Figs S4–S6) indicate that there are aspects of nucleolar IP5K spatial dynamics that are intimately associated with the degree of rRNA synthesis.

The interactions of IP5K with TCOF1, UBF and CK2 (Fig. 1) raise the possibility of a new role for IP5K as a molecular scaffold that may be functionally important for rRNA synthesis. That idea was strengthened by the observation that, in MCF-7 cells, IP5K co-localized with Pol I (Fig. 1I) and newly synthesized rRNA (detected by BrU incorporation; Fig. 1I,J).

We could not use RNA interference to further test our hypothesis that IP5K is a scaffold. The ‘knock-down’ of IP5K expression would not only ablate any structural contribution from IP5K, but would also eliminate its catalytic function in InsP₆ synthesis; eukaryotic cells cannot survive the loss of InsP₆ (Frederick et al., 2005; Verbsky et al., 2005). Indeed, the treatment of H1299 cells with IP5K siRNA led to a decrease in the size of the nucleus (supplementary material Fig. S7) and loss of cell viability (data not shown).

There is a well-characterized, alternative methodological approach to determine if a protein has a structural role as a molecular scaffold. Unlike a protein's roles in signaling and communication, which are enhanced upon its overexpression, an excess of a molecular scaffold inhibits its function (Fig. 2) (Good et al., 2011). That inhibition [sometimes called a ‘dominant-negative’ effect (Kim et al., 2011)], comes from the experimental perturbation of the stoichiometry between a scaffold and its binding partners (Fig. 2) (Good et al., 2011). Overexpression of a scaffold sequesters individual binding partners away from each other, thereby severing their connection in a cellular communication pathway (Fig. 2). This was the case with IP5K: two of its binding partners, UBF and TCOF1 (Fig. 1F) were normally restricted to nucleoli (Fig. 3A,B, left hand column; Fig. 3C,E) (Valdez et al., 2004). Following IP5K-GFP overexpression in H1299 cells, the UBF and TCOF1 signals became smaller and denser and some of the UBF and TCOF1 were translocated out of the nucleolus, sometimes as far as the nuclear periphery (Fig. 3A,B,D,F). Some IP5K still colocalized with a portion of these translocated proteins (supplementary material Fig. S8). Furthermore, while the UBF and TCOF1 signals overlapped in nucleoli of control cells (Fig. 3E), as previously reported (Valdez et al., 2004), the two signals were almost completely separated when IP5K-GFP was overexpressed (Fig. 3F). That dispersion of TCOF1 also caused it to separate from the nucleolar pool of Pol I (Fig. 3C,D). This disruption of the organization of the multimolecular rRNA transcriptional complex is consistent with the expected outcome of the overexpression of a scaffolding protein (see Fig. 2).

Unlike siRNA of IP5K, which compromises both the catalytic and the scaffolding actions of IP5K (see above), we were able to separate these two functions in the overexpression paradigm; we used a catalytically dead IP5K^C162Y construct (Brehm et al., 2007). This mutant kinase caused the UBF and TCOF1 to relocate out of the nucleolus to the same extent as did the wild-type IP5K (Fig. 3A,B, right-hand column).

We next investigated if the structural perturbations that are induced by IP5K overexpression would impact rRNA synthesis. First, in control experiments, we found rRNA synthesis was not affected by the overexpression of a construct (DsRedNuc; Fig. 4) that accumulated in nucleoli (supplementary material Fig. S9, upper panels) but did not affect the localization of nucleolar UBF (supplementary material Fig. S9, lower panels). We also overexpressed TCOF1, a nucleolar protein that provides regulatory input into rRNA synthesis (Dixon et al., 2006). As expected for a protein that participates in cellular communication, TCOF1 overexpression increased rRNA synthesis (by 35%; Fig. 4). In contrast, overexpression of IP5K-GFP led to a 72% reduction in rRNA synthesis (Fig. 4); such a ‘dominant-negative’ effect of overexpression is diagnostic of a scaffolding role for IP5K (see Fig. 2). Overexpression of the IP5K^C162Y-GFP kinase-dead mutant inhibited rRNA synthesis to the same extent as did IP5K-GFP (Fig. 4), confirming that the effect of the kinase was not catalytic in nature. An important conclusion to be drawn from these experiments is that rRNA synthesis was specifically perturbed by overexpression of a scaffold (IP5K), and not just by the overexpression of proteins in the nucleolus (DsRedNuc) per se.

To further characterize the proposed scaffolding role for IP5K, we searched for a candidate sequence in the kinase that might mediate specific interactions with one of the binding partners described in Fig. 1F. Polybasic residues can mediate protein/protein interactions (Williams, 2003) and we identified one candidate tripeptide, RKK, in IP5K (residues 41–43; Fig. 5A). For this tripeptide to be able to associate with other proteins, it would be expected to be exposed on the protein's surface. However, the human IP5K (hIP5K) structure is unknown. On the other hand, the crystal structure of Arabidopsis thaliana IP5K (AtIP5K) has been solved (González et al., 2010). The sequences of AtIP5K and hIP5K are 40% homologous and the beta sheets are all conserved (González et al., 2010). Immediately C-terminal of the third beta-sheet in hIP5K1 is the RKK tripeptide that aligns with sequence in AtIP5K that lies on the protein surface (Fig. 5A,B). To illustrate that point we have modeled the aligned RKK tripeptide onto the AtIP5K structure (Fig. 5B).

To test the functionality of the RKK sequence, we mutated it to QQQ. In control experiments, we found that this mutant (IP5K^QQQ) retained full catalytic activity (Fig. 5C), indicating that its overall structure was not significantly perturbed. Nevertheless, this mutation almost completely abolished the interaction between IP5K and UBF (Fig. 5D). Moreover, following overexpression of the mutant kinase, UBF remained in the nucleolus (Fig. 5E), in contrast to our finding that UBF translocated out of the nucleolus after overexpression of wild-type IP5K (Fig. 3). Significantly, IP5K^QQQ still bound to TCOF1 and CK2 (Fig. 5D). Furthermore, the QQQ mutation did not prevent overexpression of the IP5K from promoting TCOF1 translocation out of the nucleolus (Fig. 5F). These data directly validate that the RKK tripeptide in IP5K mediates an interaction of the kinase with UBF that is both direct and specific.

We have established that cells contain a nucleolar pool of IP5K that is an essential component of the cellular machinery that produces rRNA. Since rRNA synthesis initiates ribosome biogenesis, which in turn synthesize cellular proteins, this new role for IP5K is fundamental to the life and destiny of every cell. Of further interest is our determination that the mechanism of action of IP5K in rRNA production is independent of its cell-signaling kinase activity, and instead involves its hitherto unknown function as a molecular scaffold. This finding causes us to re-assess the prevailing dogma that the compartmentalization of IP5K is completely subservient to a signaling paradigm that is driven by the localized production of InsP₆ and other inositol-based regulatory molecules (Alcázar-Román and Wente, 2008; Seeds et al., 2007).

Nucleolar organization is a highly dynamic and tightly regulated process. If the architectural function of IP5K is important in regulating aspects of nucleolar disassembly and assembly [for example when the cell enters and exits mitosis (Drygin et al., 2010)], that would be just as important in controlling rRNA production as the catalytic activities that transcribe rDNA. Indeed, our data suggest that there is a close relationship between the spatial dynamics of the nucleolar pool of IP5K and the degree of rRNA synthesis. In addition to their ability to stabilize multimolecular complexes, scaffolds also facilitate information flow through biological networks by enforcing the co-localization of proteins that functionally interact with each other (Fig. 2) (Good et al., 2011). Such a regulatory function is possible for IP5K, which binds CK2, UBF and TCOF1 (Fig. 1F); rRNA synthesis is regulated by interactions between UBF and TCOF1 (Valdez et al., 2004), and also by the phosphorylation of UBF and TCOF1 by CK2 (Isaac et al., 2000; Lin et al., 2006). The facilitation of such communication pathways by IP5K may assist the information flow from cellular signaling cascades, so as to regulate the ribosome production that guides cell differentiation and proliferation (Drygin et al., 2010).

Our study also contributes to understanding the basis of the biological sophistication of higher organisms. The genomic era has helped us recognize that increased organismic complexities have not just arisen from an increase in the number of protein-encoding genes. It is now appreciated that alternate gene splicing has evolved on a large scale to provide multiplicity of protein functions. But there is a more unusual and often unpredictable evolutionary pathway to gene multifunctionality: ‘moonlighting’ (Gómez et al., 2011). This is a phenomenon whereby one protein acquires two roles that are strikingly independent of each other. This is the case for IP5K; prior to our discovery that this protein acts as a scaffold, its sole function (albeit one that is indispensable) was thought to be as a kinase in the pathway that synthesizes InsP₆ and other inositol phosphate signaling molecules (Alcázar-Román and Wente, 2008; Ives et al., 2000; Seeds et al., 2007). The data that we have obtained add a new dimension to inositol phosphate research, by demonstrating that in addition to synthesizing InsP₆, IP5K also has a non-catalytic function as a molecular scaffold.

It is intriguing to consider how multifunctionality of IP5K might have arisen. The catalytic core of the enzyme is well understood (González et al., 2010), and so attention should be directed at the limited number of other regions of the protein that might be available for scaffolding functions. The association of scaffolds with binding partners can arise from some unique properties of structurally disordered regions of proteins. Such domains have the flexibility to alter conformation and secondary-structure so as to make specific protein/protein interactions (Gómez et al., 2011). We do not know if this is the case for hIP5K, as its structure is not known. However, the crystal structure of the Arabidopsis homologue reveals the presence of disordered loops on the kinase's surface (González et al., 2010). In addition, our own molecular modeling and mutagenic studies identified an RKK tripeptide that is necessary for hIP5K to bind UBF. This was a specific interaction; these three residues did not mediate binding of IP5K to TCOF1 or CK2 (Fig. 5D,F). Multiple sequence alignments (Fig. 5A; supplementary material Fig. S10) indicated that the human RKK tripeptide is conserved (as R/KKK) in IP5K homologues in rat, mouse, and the horse, but in contrast, there are homologues in plants, yeast, zebrafish and flies that do not possess this motif. Thus, this interaction between IP5K and UBF may have emerged in higher eukaryotes.

There are human health implications to our determination that the IP5K gene is multifunctional. Pleiotropy is a common property of genes and SNPs that are associated with disease traits (Sivakumaran et al., 2011). This topic has implications for specificity in the identification of molecular targets for drug development. Genetic risk-profiling, and understanding of epistatic contributions to diseases and morphological defects are also impacted by pleiotropy (Sivakumaran et al., 2011). It is therefore particularly pertinent that animal development is critically dependent on spatio-temporal control over rRNA synthesis. The breakdown of these controlling events in utero can be catastrophic; a dysregulation of rRNA synthesis in neural crest cells compromises the biogenesis of ribosomes and so cell migration is perturbed, leading to defective facial morphogenesis, including Treacher-Collins Syndrome (TCS) (Dixon et al., 2007). The genetic basis of these developmental disorders is incompletely understood; mutations in TCOF1 undoubtedly contribute, but a poor correlation between TCOF1 genotype and phenotype has researchers in this field looking for additional genetic factors (Dixon et al., 2007; Splendore et al., 2000). Our work suggests that because of the physical interactions between IP5K and TCOF1, the IP5K gene may have epistatic influences upon TCS. It is therefore intriguing that disruption of embryonic IP5K expression recapitulates one of the phenotypes of TCS, namely, impaired neural crest cell migration (Verbsky et al., 2005). Our studies further indicate that the lethality of IP5K-null phenotype may not arise out of loss of InsP₆ alone, as was previously thought (Verbsky et al., 2005), but may also reflect the importance of the scaffolding role of IP5K to ribosome biogenesis. This participation of IP5K in ribosome biogenesis, a cellular process that is fundamental to the life of every organism, considerably elevates the status of this particular inositol phosphate kinase.

H1299 human lung carcinoma cells (CRL-5803, ATCC), MCF-7 human breast cancer cells (HTB-22, ATCC) and primary human dermal fibroblasts (explant cultures were obtained from skin biopsies by standard methods), were all cultured in DMEM including 10% FBS and 1% Penicillin/Streptomycin at 37°C and 5% CO₂. All media and cell culture additives were purchased from Invitrogen. H1299 nucleoli were prepared as previously described (Vascotto et al., 2009).

Cells were prepared for analysis by confocal immunofluorescence as previously described (Brehm et al., 2007). The source of the anti-IP5K antibody (used at 1∶40 dilution) was as previously described (Brehm et al., 2007). Antibody specificity was validated by the loss of signal after RNAi against IP5K (supplementary material Fig. S7) and by control experiments performed without primary antibody (supplementary material Fig. S11). Sources of other antibodies are as follows (with dilutions used): Anti-fibrillarin, Abcam ab4566 (1∶1000); anti-Pol I, Santa Cruz, sc-21751 (1∶100); anti-TCOF1, Sigma Aldrich, GW22821 (1∶10,000); anti-UBF, sc-13125 Santa Cruz (1∶100); goat anti-rabbit Alexa Fluor 488, 546 or 647, Invitrogen (1∶5000); goat anti-mouse Alexa Fluor 488 or 546, Invitrogen, (1∶5000); anti-chicken Alexa Fluor 546, Invitrogen (1∶12,000); Alexa Fluor 647; anti-BrU antibody, Invitrogen (1∶100). In the latter case, a 633 nm laser line from a helium/neon laser was used with a 560-nm long pass emission filter with the pinhole set to 2.10 Airy unit.

The cDNAs encoding either IP5K-GFP, IP5K^C162Y-GFP (Brehm et al., 2007), IP5K^QQQ-GFP (⁴¹RKK⁴³ changed to QQQ), TCOF1-GFP or DsRedNuc (a red fluorescent protein ‘DsRed’ fused to a nuclear translocation signal; Clontech) were each transfected into H1299 cells 24 hours after seeding using Lipofectamine LTX (Invitrogen) according to the manufacturer's instructions. IP5K^QQQ-GFP was generated by modified QuikChange site-directed mutagenesis as previously described (Wang and Malcolm, 1999) using the following primer pair: forward 5′-CTGAAGTTTCCTCCAAATCAGCAGCAGACCTCGGAAGAGATATTT-3′, reverse: 5′-AAATATCTCTTCCGAGGTCTGCTGCTGATTTGGAGGAAACTTCAG-3′ and vector pEGFP-N1 containing the cDNA of human IP5K. The open reading frame of full-length cDNA of IP5K was cloned into pGEX-4t3 (GE Healthcare) encoding an N-terminal GST-tag using the primer pair: 5′-GATGAATTCCATGGAAGAGGGGAAGATG-3′ and 5′-GATGCGGCCGCTTAGACCTTGTGGAGAAC-3′ introducing restriction sites NotI at the 5′-end and EcoRI at the 3′-end. This vector and the same mutagenic primers described above were used for Quikchange mutagenesis to generate GST-IP5K^QQQ. These vectors were transduced into Escherichia coli Strain BL21Star. Bacteria were grown in LB-medium with 100 µg/µl ampicillin and 0.2% (w/v) glucose until an OD₆₀₀ of approx. 0.6 was reached. The culture was cooled to 12°C and GST-IP5K expression was induced by addition of 0.25 mM isopropyl-β-D-thiogalactopyranoside. After 20 hours bacteria were harvested by centrifugation (4100×g, 10 minutes, 4°C), resuspended in 137 mM NaCl, 2.7 mM KCl, 4.3 mM NaH₂PO₄, pH 7.5) and lysed by freeze/thaw cycles. Triton X-100 (0.5% v/v was added, and samples were sonicated and then centrifuged (14,400 g, 11 minutes, 4°C). GST-IP5K and GST-IP5K^QQQ were then purified using a glutathione-sepharose column (GST-Buster, Amocol) according to the manufacturer's instructions.

Kinetic assays of recombinant GST-IP5K and GST-IP5K^QQQ (3 µg) were performed at 37°C in a final assay volume of 400 µl in medium containing 2.5 µM Ins(1,3,4,5,6)P₅, 20 mM HEPES (pH 7.2), 50 mM KCl, 8 mM MgCl₂, 1 mM EDTA, 5 mM ATP and 1 mM DTT. Inositol phosphates were extracted and analyzed by MDD-HPLC as described previously (Brehm et al., 2007).

For GST pull-down experiments, lysates were prepared from either wild-type H1299 cells, or H1299 cells in which TCOF-GFP was overexpressed, in M-PER Mammalian Protein Extraction Reagent (Pierce) supplemented with Complete Protease Inhibitor cocktail (Roche) according to the manufacturer's instructions. Lysates were incubated for 4 hours at 4°C with either purified GST-IP5K, GST alone, or GST-IP5K^QQQ, each of which were immobilized on glutathione-cellulose beads (GST-buster QF, Amocol). The bound proteins were washed six times for 10 minutes with ice-cold PBS containing 0.01% Tween 20. Bound proteins were then removed from the beads by denaturation in SDS running buffer (10 minutes at 95°C). The beads were pelleted by centrifugation (30 seconds, 2000 rpm, 4°C). The supernatant was removed and proteins were separated by SDS/PAGE and transferred onto a PVDF membrane using standard western blotting techniques. A chicken anti-TCOF1 antibody was used to detect TCOF1-GFP. The anti-UBF antibody was obtained from Santa Cruz (sc-13125). CK2 was detected with a mouse anti-CK2 antibody (BDE Biosciences, Cat. No. 611610). The secondary anti-mouse and anti-chicken antibodies were conjugated with HRP. Antibody signals were developed with the ECL plus kit (GE) and documented with a Kodak imager.

Lysates were prepared (see above) from H1299 cells in which either IP5K-GFP or GFP alone had been expressed for 24 hours. After immunoprecipitation with anti-GFP antibody, the protein G agarose bound protein complexes were denatured for 10 minutes at 77°C and separated by SDS PAGE. The gel was stained with Sypro Ruby according to the manufacturer's instructions. Each lane of the gel was divided into 24 equal pieces and digested with modified porcine trypsin (Promega). Each digest was independently analyzed by nanoLC-ESI-MS/MS using an Agilent 1100 nanoLC system on-line with an Agilent XCT Ultra ion trap mass spectrometer with the Chip Cube Interface as described previously (Choi et al., 2007). The resulting 24 data sets were then combined and peak lists were generated using the Data Extractor feature of the SpectrumMill software from Agilent (Choi et al., 2007). The resulting peaklists were then searched using the MS/MS search function of the Spectrum Mill suite using the mammals limited NCBInr database with allowances made for two missed tryptic cleavages and mass tolerances of 1.5 Da for the precursor ions and 0.8 Da for fragment ions. Variable methionine oxidation was also allowed. The data that were obtained from the NanoLC-ESI-MS/MS analyses were also processed using the Mascot Distiller software (MatrixScience) using essentially factory settings. The resulting peak lists were interrogated via the Mascot search engine using the MS/MS mode and the same search settings described above for the Spectrum Mill searches.

To monitor rRNA synthesis, BrU was incorporated into nascent rRNA in a “run-on” assay; the incorporated BrU was then detected by confocal immunofluorescence microscopy as previously described (Ko et al., 2000). H1299 cells were cultured in 3.5 cm plates with a glass insert (NUNC). Cells were washed twice with PBS (pH 7.4), and then with buffer containing 20 mM Tris-HCl, pH 7.4, 5 mM MgCl₂, 0.5 mM EGTA and 0.5 mM PMSF. The same buffer (supplemented with 0.05% Triton X-100) was used to lightly permeabilize the cells for 5 minutes at room temperature. The synthesis of rRNA was then initiated by adding 50 mM Tris-HCl, pH 7.4, 100 mM KCl, 5 mM MgCl₂, 0.5 mM EGTA, 25 U/ml RNasin, 1 mM PMSF, 0.5 mM of ATP, CTP, and GTP, and 0.2 mM Br-UTP (Sigma) in the presence of α-amanitin (1 µg/ml). The reactions were continued for 30 minutes at 37°C. The cells were then washed (PBS containing 25 U/ml RNasin; room temperature), fixed (PBS containing 3% paraformaldehyde, 25 U/ml RNasin, and 0.1% BSA, 10 minutes, 37°C), and fully permeabilized with PBS containing 0.3% Triton X-100 and 0.1% BSA for 3 minutes at 37°C. Cells were washed twice with 0.1% BSA in PBS for 5 minutes and stained overnight at 4°C with an Alexa Fluor 647-labeled antibody that detects BrU (1∶100 in PBS containing 0.5% BSA). Finally, cells were washed three times with PBS containing 0.1% BSA for 10 minutes.

In randomly selected fields of view, the BrU fluorescence intensities (FI) were estimated using ImageJ for all nuclei that were visible in the DIC. In every picture, the mean FI of the transfected cells was calculated as a percentage of the mean FI in non-transfected cells. Background correction was not performed generally, since its values were too low to affect the data. For statistical evaluation of data, unpaired Student's t-test was performed using GraphPad InStat version 3.06 (GraphPad Software).

We thank Dr Benigno Valdez (Baylor College of Medicine, Houston, TX) for the cDNA of TCOF1-GFP. We also thank Jeff C. Tucker and Holly Rutledge for help with confocal microscopy, Julia Giese, Alien Kruse, Gesa König and Stefan Alschausky for performing GST-pulldowns and western blot analysis, Huanchen Wang for the molecular modeling, Michelle Scott for extracting the NoLS data, Katina Johnson for mass spectrometry support, and Eliane Uebele and Hongying Lin for assaying kinase activities.