Fibroblast Growth Factor–2 Regulates the Stability of Nuclear Bodies

Fibroblast growth factor–2 regulates the stability of nuclear bodies

Alexander-Francisco Brunsa, Jeroen van Bergeijka, Christina Lorbeera, Anna No¨ llea,b, Julia Jungnickela, Claudia Grothea,b, and Peter Clausa,b,1

aHannover Medical School, Department of Neuroanatomy, Hannover, Germany; and bCenter for Systems Neuroscience Hannover, Hannover, Germany

Edited by Charles A. Dinarello, University of Colorado Health Sciences Center, Denver, CO, and approved June 3, 2009 (received for review January 7, 2009) Nuclear bodies are distinct subnuclear structures. The survival of SMN-positive foci are more frequent in adult than in fetal stages. motoneuron (SMN) gene is mutated or deleted in patients with the The number of SMN-positive foci is correlated with SMA neurodegenerative disease spinal muscular atrophy (SMA). The disease severity in fibroblasts from SMA patients (19). It is still gene product SMN is a marker protein for one class of nuclear unclear whether the number of gems is directly responsible for bodies denoted as nuclear gems. SMN has also been found in Cajal disease pathology or is merely indicative of decreased SMN bodies, which co-localize with gems in many cell types. Interest- expression levels in SMA patients. Functionally, gems have been ingly, SMA patients display a reduced number of gems. Little is suggested to be part of the snRNP recycling machinery (20). known about the regulation of nuclear body formation and sta- We have previously shown that a nuclear isoform of fibroblast bilization. We have previously shown that a nuclear isoform of the growth factor–2 (FGF-223) directly interacts and co-localizes ﬁbroblast growth factor–2 (FGF-223) binds directly to SMN. In this with SMN (21, 22). FGF-2 belongs to the expanding family of study, we analyzed the consequences of FGF-223 binding to SMN growth factors with intranuclear and intracellular localization in with regard to nuclear body formation. On a molecular level, we addition to their role as extracellular signaling molecules (23). showed that FGF-223 competed with Gemin2 (a component of the This growth factor is expressed as different isoforms that are SMN complex that is necessary for gem stabilization) for binding to translated from one mRNA species. The 18 kDa isoform (FGF- SMN. Down-regulation of Gemin2 by siRNA caused destabilization 218) is generated after translation initiation at a conventional of SMN-positive nuclear bodies. This process is reﬂected in both AUG start codon, whereas upstream alternative CUG start cellular and in vivo systems by a negative regulatory function of codons are used for expression of the high-molecular weight FGF-2 in nuclear body formation: in HEK293 cells, FGF-223 decreased isoforms (hmw) 21 and 23 kDa FGF-2 (rat FGF-221 and FGF- the number of SMN-positive nuclear bodies. The same effect could 223). FGF-2 isoforms are differentially regulated during devel- be observed in motoneurons of FGF-2 transgenic mice. This study opment, after pharmacological treatment, and after nerve injury demonstrates the functional role of a growth factor in the regu- (24), indicating different biological roles for the individual lation of structural entities of the nucleus. isoforms. In this study, we looked at the molecular and functional 23 nuclear organization ͉ nuclear structure ͉ spinal muscular atrophy ͉ consequences of FGF-2 /SMN complex formation with regard survival of motoneuron protein to Gemin2 binding and nuclear body formation. On a molecular level, we showed that FGF-223 competes with Gemin2 in a he nucleus of a cell is organized into several distinct sub- concentration-dependent manner for binding to SMN. Because Tstructures, including chromatin, nucleoli, speckles, and dif- interaction between Gemin2 and SMN is important for the ferent types of nuclear bodies (1, 2). Larger molecular aggregates stabilization and integrity of gems (25, 26), we analyzed the formation of SMN-positive nuclear foci after overexpression of in the nucleus are denoted as nuclear bodies (1). Coiled bodies 23 were discovered as accessory bodies of nucleoli (3) and were FGF-2 . In this study, we refer to the nuclear structures labeled later renamed Cajal bodies (4, 5). Cajal bodies contain the by the anti-SMN antibody as ‘‘SMN-positive nuclear bodies/foci’’ to highlight the fact that both Cajal bodies and gems are labeled. marker protein coilin (6). Formation of Cajal bodies occurs de On a cellular level, the number of gems decreased in HEK293 novo in a non-hierarchical process (7). Newly assembled small cells when the intracellular concentration of the FGF-223 iso- nuclear ribonucleoprotein (snRNP) particles accumulate in Ca- form was increased, whereas Cajal bodies were not affected. jal bodies after their nuclear import (8). Another type of nuclear Interestingly, in transgenic mice that overexpress FGF-2, the body was described as ‘‘gemini of coiled bodies’’ or gems. These number of gems in motoneurons was similarly decreased. These nuclear foci contain the survival of motoneuron (SMN) protein results demonstrate a function for a nuclear growth factor as a and its interacting protein Gemin2 (SIP1) (9). SMN is mutated direct modulator of the structural integrity of gems. in patients with spinal muscular atrophy (10). At least in the cytoplasm, SMN is part of a larger complex containing spliceo- Results somal Sm proteins and a group of proteins designated Gemins FGF-223 Competes with Gemin2 for SMN Binding. The binding of 2–8 (11). SMN and coilin directly interact in Cajal bodies (12). Gemin2 to SMN has been mapped to amino acid residues 13–44, Many factors determine the number and morphology of which are encoded by exons 1 and 2a (27–29). Other studies SMN-containing nuclear bodies. Formation of Cajal bodies is indicate binding to residues 52–91 of human SMN (28). More- dependent on SmB, as up-regulation of this snRNP component over, exon 2 of Smn codes for a domain (residues 28–51) able to

leads to the appearance of this nuclear substructure (13). In a CELL BIOLOGY straightforward analysis, it has been shown that phosphorylation of critical residues in the coilin C terminus regulates the number Author contributions: A.-F.B., J.v.B., and P.C. designed research; A.-F.B., J.v.B., C.L., A.N., of Cajal bodies (14). The number of Cajal bodies increases in and P.C. performed research; J.J. contributed new reagents/analytic tools; A.-F.B., J.v.B., transformed cells (15). Also, Cajal bodies and gems display a A.N., C.G., and P.C. analyzed data; and P.C. wrote the paper. higher degree of co-localization in later developmental stages The authors declare no conﬂict of interest. (16). SMN contains a motif encoded by exon 2b that is respon- This article is a PNAS Direct Submission. sible for targeting to Cajal bodies (17). In most cell lines, Cajal 1To whom correspondence should be addressed. E-mail: [email protected]. bodies and gems co-localize (5), but gems represent a distinct This article contains supporting information online at www.pnas.org/cgi/content/full/ class of nuclear substructures (11, 16, 18). In motoneurons, 0900122106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900122106 PNAS ͉ August 4, 2009 ͉ vol. 106 ͉ no. 31 ͉ 12747–12752 Downloaded by guest on October 6, 2021 demonstrate competition between FGF-223 and Gemin2 for binding to SMN.

Knock-Down of Gemin2 Expression Causes Destabilization of Gems, but Not Cajal Bodies. The functional consequences of Gemin2 loss with regard to formation of nuclear bodies is unknown. Here, we down-regulated Gemin2 by siRNA transfection of HEK293 cells. Briefly, cells were transfected with a Gemin2-specific siRNA oligonucleotide or a scrambled siRNA that had a sequence with the same purine/pyrimidine base composition. Co-transfections with fluorescein-labeled dsRNA oligomers were performed to ensure that only transfected cells were analyzed in the subse- quent counts of nuclear bodies. Three days after transfection, 57.3% Ϯ 8.3% of the cells (n ϭ 3 experiments, n ϭ 2659 cells) still showed fluorescence of the control oligomers. This approach excludes the possibility that the knock-down-efficiency was very high in only a few cells. Down-regulation of Gemin2 protein levels was observed after 48 hours by Western blot (Fig. 2A). After 72 hours, cells displayed a 44.8% knock-down (Ϯ4.9%, n ϭ 3) of the starting level, confirming successful and Fig. 1. FGF-2 competes with Gemin2 for SMN binding. (A) Gemin2 was significant down-regulation as determined by densitometry (Fig. expressed in an in vitro transcription/translation reticulocyte system, and 2B). To quantify the number of SMN-positive nuclear bodies, equal amounts were immobilized on recombinantly expressed GST-SMN cou- cells were fixed and processed for immunocytochemistry with an pled to glutathione beads. The complexes were incubated with increasing 23 anti-SMN antibody. The number of SMN-positive nuclear bodies amounts of FGF-2 in the presence of BSA (lanes 5–11). Substitution of low Ͼ amounts of FGF-223 with BSA to a total amount of 20 ␮g FGF-223/BSA revealed per cell was counted (n 800 cells analyzed) and a frequency the same results (data not shown). Beads were subsequently washed, and distribution was calculated (Fig. 2C). Few cells displayed three bound proteins were analyzed by Western blotting with antibodies against or more SMN-positive nuclear bodies, with most cells showing FGF-223 and Gemin2. The experimental controls indicated an absence of no or one nuclear body. For statistical evaluation, we compared non-specific binding by Gemin2 and FGF-223 (lanes 3 and 4, respectively). Lanes the overall frequency distributions between the different con- 1 and 2 confirm that Gemin2 and FGF-223 bind to SMN. A reduction of Gemin2 ditions (␹2 test). At 72 hours after transfection with Gemin2- binding was seen after the addition of 10 ␮g FGF-223, reflecting an estimated specific siRNA or scrambled siRNA, a significant change in the 23 stoichiometry of FGF-2 /Gemin2 of 2:1, given that most of the active FGF-2 frequency distribution between both conditions could be ob- exists as a dimer. (B) Competition was confirmed by co-immunoprecipitation served (Fig. 2C). More nuclei lacked SMN-positive nuclear with anti-SMN antibody in nuclear lysates. HEK293 cells were transfected with pFGF-223-DsRed2 only (lanes 3 and 6), equal amounts of pFGF-223-DsRed2 and structures when treated with Gemin2 siRNA. These data indi- pDsRed2 (lanes 2 and 5), and pDsRed2 as a control (lanes 1 and 4), so that cells cate that Gemin2 regulates the number of SMN-positive nuclear expressed an increasing amount of FGF-223-DsRed2. This increase in expression bodies. Its loss at least partially destabilizes these nuclear was confirmed by Western blot (WB) analysis with anti-FGF-2 antibody. Im- structures. In contrast, Cajal bodies stained for coilin as a marker munoprecipitation (IP) was carried out with anti-SMN antibody and analyzed protein were not affected by down-regulation of Gemin2 (Fig. by SDS–PAGE/Western blot analysis with anti-SMN and anti-Gemin2 antibod- 2C). To determine the fraction of nuclear bodies that are both ies. Although FGF-2 concentrations in the lysates increased, the Gemin2 signal coilin positive and SMN positive in HEK293 cells, we performed decreased (lanes 4–6), which indicates competition between the two mole- 23 double staining experiments with the appropriate antibodies cules for binding to SMN. At the highest FGF-2 expression level, a low (Fig. 2D). Most of the stained endogenous structures were coilin amount of Gemin2 was immunoprecipitated (lane 6). This finding agrees well with quantitative data for SMN-positive nuclear bodies, as some nuclear positive, but only half of the Cajal bodies were SMN positive; that bodies were still present, even at high FGF-2 concentrations in assays quanti- is, 50.3% of the structures were SMN positive and 49.7% of these fying the numbers of SMN-positive structures (Fig. 3B). hc, heavy chains of the were Cajal bodies. SMN-antibody used for IP. Overexpression of FGF-223 in HEK293 Cells Decreases the Number of SMN-Positive Nuclear Bodies. Because a stable Gemin2/SMN com- bind to nucleic acids in vitro (30). Previously, we have shown that plex is a prerequisite for gem formation and/or stabilization (25), binding of U2 small nuclear RNA (snRNA) is not affected by we tested the consequences of FGF-223 overexpression on the FGF-223 binding to SMN (22). Therefore, we asked whether number of SMN-positive nuclear bodies. HEK293 cells were FGF-223 is able to compete with Gemin2 for binding to SMN. transfected with a plasmid encoding full-length SMN fused to the Full-length human SMN was recombinantly expressed and im- green fluorescent protein (EGFP) and co-transfected with con- 23 18 mobilized on glutathione–agarose beads by a GST tag at its C trol-, FGF-2 - or FGF-2 -DsRed constructs. This approach terminus. Gemin2 was expressed in a coupled in vitro transcrip- allowed the identification of co-transfected cells. The number of tion/translation reticulocyte lysate system. As expected, Gemin2 SMN-positive nuclear bodies per cell, indicated by EGFP signals (Fig. 3A), was counted and plotted as a frequency distribution for bound to SMN, as shown by a Western blot of the extensively all analyzed cells (Fig. 3B). No change in the frequency distri- washed beads (Fig. 1A). Incubation of the complex with increas- 18 23 bution between SMN/empty vector and SMN/FGF-2 could be ing amounts of recombinantly expressed and purified FGF-2 observed, as these proteins did not interact. Importantly, co- led to a decrease in the amount of Gemin2 bound to SMN (Fig. transfection with FGF-223 revealed a significant left shift be- ␮ 1A). A clear reduction of Gemin2 binding (2.5 g in each lane) tween the distributions (Fig. 3B). Cells transfected with FGF-223 ␮ 23 could be seen with the addition of 10 g of FGF-2 . In addition, displayed significantly less SMN-positive nuclear bodies than the immunoprecipitation experiments were performed with an anti- empty vector control cells. As an additional control, we trans- SMN antibody in nuclear extracts from cells expressing different fected cells with a SMN deletion construct lacking the first 27 aa amounts of FGF-223 (Fig. 1B). An increased concentration of residues (SMN28–294), which contains part of the binding site FGF-223 resulted in immunoprecipitation of less Gemin2, which for FGF-223 (22) and the binding site for Gemin2 (29). There- corroborates the result of the pull-down experiments. These data fore, the lack of this sequence in SMN precludes the binding of

12748 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900122106 Bruns et al. Downloaded by guest on October 6, 2021 Fig. 3. Overexpression of FGF-223 in HEK293 cells decreases the number of SMN-positive nuclear bodies. HEK293 cells were transfected with full-length SMN fused to EGFP and co-transfected with empty vector, FGF-218- or FGF- 223-DsRed constructs for the detection and evaluation of co-transfected cells only. (A) Representative examples of HEK293 cells displaying different numbers of SMN-positive nuclear foci after transfection with pSMN1–294-EGFP. Scale bar, 20 ␮m. (B) Numbers of SMN foci in the nuclei were counted and plotted as frequency distributions after transfection with the respective constructs. SMN1–294/FGF-223 showed a significant difference of the frequency distribution compared with the SMN1–294/empty vector control, whereas co-transfection with SMN1–294 and the FGF-218 isoform did not change the Fig. 2. Knock-down of Gemin2 causes destabilization of gems. Gemin2 frequency distribution. As additional controls, cells were transfected with protein levels were down-regulated by siRNA and the number of SMN-positive pSMN28–294-EGFP lacking the FGF-223 interaction domain in the N terminus. nuclear foci was evaluated in HEK293 cells. (A) Western blots of cell lysates No difference from the SMN28–294/empty vector control was detected. Three obtained from cells treated with Gemin2-specific siRNA and scrambled siRNA, independent experiments were performed per condition with the following respectively, at 48 and 72 hours after transfection. (B) Densitometric measure- numbers of measured cells: SMN1–294/empty vector n ϭ 619; SMN1–294/FGF- ment of Gemin2 bands on Western blots from three independent experi- 218, n ϭ 377; SMN1–294/FGF-223, n ϭ 636; SMN28–294/empty vector, n ϭ 557; ments. Expression levels of Gemin2 were normalized and calculated relative to SMN28–294/FGF-218, n ϭ 501; and SMN28–294/FGF-223, n ϭ 502. Error bars ␤-tubulin levels. Bars represent arithmetic means with standard errors. *P Ͻ represent standard errors of the means. ***P Ͻ 0.001 (␹2 test). Overexpression 0.05 (Mann-Whitney test). (C) HEK293 cells were transfected with Gemin2- of FGF-223 was estimated to be at least 10-fold more than that of endogenous specific or scrambled siRNA. SMN-positive nuclear foci were counted after FGF-2 level (data not shown). (C) HEK293 cells were transfected with FGF-223- immunocytochemistry with anti-SMN antibody and fluorophore-coupled sec- EGFP or empty vector, and the number of coilin-positive nuclear bodies (Cajal ondary IgG. Cajal bodies were similarly labeled with an anti-coilin antibody bodies) was counted in 209 FGF-223-expressing cells and 244 control cells. No and counted. To ensure counting of foci in transfected cells only, co- significant difference was detected between the frequency distributions transfection with a fluorescein-labeled dsRNA oligomer was performed. For (␹2 test). control siRNA transfectants, 809 (SMN experiment) and 492 (coilin experiment) cells were analyzed (three independent experiments); for Gemin2 siRNA transfectants, nuclear bodies in 856 (SMN experiment) and 492 cells ence (Fig. 3B), demonstrating the importance of the SMN N (coilin experiment) were counted in three independent experiments. Num- terminus for the interaction between both proteins. A compar- bers of foci were plotted as frequency distributions. Error bars represent standard errors of the means. The frequency distributions of scrambled and ison between full-length SMN and the SMN28–294 deletion Gemin2-specific siRNA-treated cells are significantly different (***P Ͻ 0.001, mutants revealed a significant difference, indicating that the ␹2 test). For coilin, no significant difference was detected between the fre- missing part of the Gemin2 binding site is responsible for CELL BIOLOGY quency distributions (␹2 test). (D) HEK293 cells (n ϭ 362) were stained with destabilization of nuclear bodies. However, SMN-positive nu- anti-SMN and anti-coilin antibodies and the co-localizations of the signals clear bodies can still be found, probably because of dimerization ϩ were analyzed by counting SMN- and coilin-positive nuclear bodies (Coi of endogenous SMN with the overexpressed deletion mutant. SMN ϩ), coilin-negative and SMN-positive nuclear bodies (Coi Ϫ SMN ϩ), as well as coilin-positive and SMN-negative (Coi ϩ SMN Ϫ) nuclear structures. Dimerization of SMN with the SMN28–294 mutant is not affected because the first self-association domain is encoded by exon 2b (amino acid residues 52–91) (28) and the second domain both proteins (22, 29). The comparison between the frequency is in the YG-box region encoded by exon 7 (31). Concomitant distribution of cells transfected with the deletion construct/FGF- with the overexpression of SMN, the average number of SMN- 223 and the empty vector control revealed no significant differ- positive structures increased from the endogenous level of 1.8

Bruns et al. PNAS ͉ August 4, 2009 ͉ vol. 106 ͉ no. 31 ͉ 12749 Downloaded by guest on October 6, 2021 of SMN between the cytoplasm and the nucleus. Analysis of the same Western blots with anti-FLAG antibody showed co- immunoprecipitation of FGF-223. However, most FGF-223 was found in the nuclear extract fraction (Fig. 4, lanes 6, 7, 14, and 15). Co-expression of SMN-EGFP led to a higher amount of SMN/FGF-223 complexes in the cytoplasm (Fig. 4, lane 6) compared with non–pSMN-EGFP–transfected cells (Fig. 4, lane 7). This did not occur in the nuclear fraction (Fig. 4, lanes 14 and 15), indicating a rate-limiting mechanism for SMN transport into the nucleus, which is independent of SMN expression levels. FGF-223 was in part proteolytically degraded to a previously described 16-kDa fragment (32) found in the core sequence that is common to all FGF-2 isoforms (Fig. 4, lanes 2, 3, 10, and 11). This fragment did not interact with SMN (Fig. 4, lanes 6, 7, 14, and 15), as it lacks the arginine-rich N terminus of the FGF- 223-isoform.

Transgenic FGF-2 Mice Display a Reduced Number of SMN-Positive Nuclear Bodies. Overexpression of FGF-2 in transgenic mice (TgFGF-2) should result in a decrease of the number of SMN- positive nuclear bodies, but not Cajal bodies in vivo. To test this hypothesis, we quantified nuclear bodies in motoneurons of spinal cords from TgFGF-2 mice by anti-SMN and anti-coilin immunohistochemistry and compared them with motoneurons from wild-type littermates. These mice overexpress all human isoforms of FGF-2 (18, 22, 23, and 24 kDa) in all analyzed tissues (33). Lumbar spinal cords were dissected, rapidly fixed, and processed for immunohistochemistry with anti-SMN antibody. Motoneurons were identified by their topography in the ventral horn of segments L2–L4 and morphologically by their large, multipolar cell bodies (Fig. 5A). The numbers of SMN-positive Fig. 4. FGF-223 and SMN co-immunoprecipitation in cytoplasmic and nuclear extracts. HEK293 cells were transfected with pFGF-223-3XFLAG and full-length nuclear bodies were counted in wild-type mice as well as pSMN-EGFP plasmids to determine the site of SMN/FGF-223 interaction. Nu- TgFGF-2 mice. The comparison of the frequency distributions clear and cytoplasmic extracts were prepared and SMN immunoprecipitations revealed a significant difference between the two, with fewer (IP) were performed. The complexes were analyzed by Western blotting with SMN-positive nuclear bodies in the TgFGF-2 animals (Fig. 5B); antibodies against FLAG-tagged FGF-223 (␣-FLAGM2), SMN (␣-SMN), and Hi- however, Cajal bodies were not affected (Fig. 5B). These data stone H3 (␣-Histone H3), which served as a control for successful nuclear confirm the hypothesis that an increased intracellular amount of extract preparation. Most FGF-223 was found in the nuclear extract fraction FGF-2 destabilizes nuclear body formation of specific sub- (lanes 14, 15) compared with the cytoplasmic fraction (lanes 6, 7), indicating nuclear structures. an interaction between SMN and FGF-223 mainly in the nucleus. FGF-223 was partly proteolytically degraded to a known 16-kDa fragment (32), which can Discussion be seen in the input (lanes 2, 3, 10, and 11), but did not interact with SMN 23 because of the lack of the SMN-interaction domain. lc, light chains of of the In this study, we showed that intracellular FGF-2 negatively SMN-antibody used for IP; hc, heavy chains of the SMN-antibody used for IP. regulates the number of SMN-positive nuclear bodies in vitro in HEK293 cells as well as in vivo in adult mouse motoneurons. The results in the cellular and in vivo systems are supported at the per cell to 2.6 per cell, as calculated from the control experi- molecular level by the finding that FGF-223 competes with ments. Whereas a significant change in the number of SMN- Gemin2. We have previously shown that FGF-223 is in a common positive foci in FGF-223-overexpressing cells occurred, no such complex with SMN and directly interacts with its N terminus, change was found for Cajal bodies (Fig. 3C). To further char- whereas FGF-218 is not co-immunoprecipitated by SMN, prob- acterize the nuclear bodies that form in the presence of FGF-223, ably because of the lack of the RG-rich N terminus present in the we performed high-resolution confocal microscopy [supporting higher molecular weight FGF-2 isoform (21). FGF-223 is able to information (SI) Fig. S1]. In agreement with our quantitative bind directly to a sequence within the first 90 aa residues of SMN data (Fig. 3B), we showed that overexpressed FGF-223 localized (21, 22). This sequence encoded by exons 1–3 of the SMN gene SMN to nuclear bodies that were not coilin positive (Fig. S1A, f–h). has been described as a binding site for nucleic acids and Gemin2. We previously demonstrated that U2 small nuclear Co-Immunoprecipitation of FGF-223 and SMN from Cytoplasmic and RNA and FGF-223 do not compete for binding to SMN (22). Nuclear Extracts. We performed co-immunoprecipitation exper- Therefore, we explored whether FGF-223 competes with Gemin2 iments to analyze the relative protein levels in nuclear and for binding to SMN. As predicted by the overlapping binding cytoplasmic extracts in the presence and absence of FLAG- domains of FGF-223 and Gemin2 on SMN, increasing amounts tagged FGF-223. HEK293 cells were transiently transfected with of FGF-223 displaced Gemin2 as shown in an in vitro binding FGF-223-3XFLAG or empty control vector. Input as well as assay and by co-immunoprecipitation. This finding confirms a immunoprecipitated proteins were analyzed by Western blot competitive mode of interaction with SMN. Formation of gems analysis using the anti-SMN antibody (Fig. 4). The same blot was is dependent on co-expression of SMN and Gemin2 (34). incubated with anti-Histone H3 antibody to demonstrate effi- Moreover, depletion of SMN by siRNA leads to a concomitant cient isolation of nuclear proteins. No change in endogenous or down-regulation of Gemin2 and a complete destabilization of overexpressed SMN levels was detected between the nuclear and gems in HeLa cells (25). However, a reciprocal effect on SMN cytoplasmic extracts when FGF-223 was present (Fig. 4, lanes 6, expression by down-regulation of Gemin2 was not observed (25). 8, 14, and 16). Therefore, FGF-223 did not change the partition With regard to coilin, depletion of SMN by siRNA causes coilin

12750 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900122106 Bruns et al. Downloaded by guest on October 6, 2021 integrity of gems may not be directly relevant to SMA disease progression. Instead, other nuclear or axonal SMN functions could be responsible for motor neuron degeneration (38–44). Furthermore, only a subset of SMN molecules is bound by FGF-223 (21, 22). This finding does not exclude the possibility that FGF-223 modulates splicing and snRNP formation by binding to SMN, thereby defining a subclass of snRNPs with distinct properties. A FGF-223-dependent decrease of gems could influ- ence a putative function of SMN in snRNP-recycling. In this scenario, two possibilities for FGF-2 activity exist: first, FGF-2 is translated and immediately transported to the nucleus; second, exogenous FGF-2 is internalized upon binding to the extracellular portion of FGF-tyrosine kinase receptors and is transported to the nucleus. In this way, FGF-223 interacts with two proteins in the splicing pathway, namely SMN and the 66-kDa subunit of splicing factor SF3a (21, 45), which is a component of U2 snRNPs (46). The growth factor FGF-2 may act as an intracellular modulator of splicing, allowing cell-signaling– dependent regulation of splicing events. Interestingly, FGF– receptor 1 (FGFR1) is also a nuclear protein (47) and the central molecule of a cascade named integrative nuclear FGFR1 signaling (INFS) (48). This pathway leads to shuttling of FGFR1 to the nucleus and feeding forward of signals to the transcriptional coactivator CREB-binding protein, thereby up-regulating gene activities associated, for example, with differentiation and development of neuronal cells (48). Moreover, nuclear FGFR1 localizes to nuclear speckles containing snRNPs (49) and the splicing factor SC35 (50). FGF-223—but, surprisingly, not FGF- 218—binds to nuclear FGFR1 and regulates its nuclear mobility Fig. 5. SMN-positive nuclear bodies in FGF-2 transgenic mice. Motoneurons (51). These data argue for the possibility that FGFR1 could act from lumbar spinal cords of FGF-2 transgenic mice (TgFGF-2), which overex- in RNA metabolism. It is an interesting question as to whether press all isoforms, and their wild-type littermates were analyzed for differ- FGF-223 could be a structural or functional link between nuclear ences in the number of SMN-positive foci. Overexpression of FGF-223 was FGFR1 signaling and SMN activities: SMN could be a regulator estimated to be at least ﬁvefold greater than that of the endogenous FGF-2 of the INFS pathway, thereby exerting effects on, for example, level (data not shown). (A) Representative sections of lumbar spinal cords neuronal differentiation. showing motoneurons with (arrows) and without SMN-positive nuclear bod- In conclusion, we have demonstrated that a nuclear growth ies. Cells were stained with mouse anti-SMN/Alexa Fluor 555 goat anti-mouse factor is able to regulate the number of SMN-positive nuclear antibody as well as with Hoechst 33258. Overlays of stains are shown. Scale bar, 20 ␮m. (B) SMN-positive nuclear bodies (upper panel) were counted in foci. Besides gems and Cajal bodies, several types of nuclear spinal cord motoneurons of wild-type mice (n ϭ 7 animals, n ϭ 229 cells) and bodies exist in cells; however, only a few factors are known to be transgenic FGF-2 mice (n ϭ 8 animals, n ϭ 273 cells). The same experiment was important for their structure, formation, or number. These performed with anti-coilin antibody to detect Cajal bodies (lower panel; results further our knowledge about the regulatory processes wild-type, n ϭ 3 animals, n ϭ 288 cells; transgenic FGF-2 mice, n ϭ 3 animals, that determine the number of SMN-containing nuclear bodies. n ϭ 382 cells). The frequencies of cells with a certain number of nuclear bodies were plotted. Error bars represent standard errors of the means. **P Ͻ 0.01 (␹2 Methods test); n.s., non-signiﬁcant. Plasmids. Full-length human SMN pSMN1–294-EGFP and the mutants pSMN1– 239-EGFP and pSMN235–294-EGFP were described before (43). For expression of Gemin2, a complete cDNA contained in pCMV-SPORT6 was used. For to localize in multiple nuclear foci instead of in Cajal bodies (35, co-expression experiments, untagged and DsRed2 (Invitrogen)–tagged FGF- 36) and to partly disperse into the nucleolus (36). In this study, 223 and FGF-218 were used as described previously (21). FLAG-tagged FGF-223 we performed a knock-down of Gemin2 in HEK293 cells and was constructed by cloning FGF-223 cDNA into the EcoRI/HindIII sites of analyzed the formation of SMN-positive foci. A quantitative p3XFLAG-CMV-14 (Sigma), resulting in clone pFGF-223-3XFLAG. The SMN- evaluation revealed a significant decrease in the average number deletion mutant pSMN119–294-EGFP, comprising amino acid residues 119– of SMN-positive foci; however, complete destabilization was not 294, was cloned by polymerase chain reaction (PCR) into the EcoRl/Sall sites of observed, in agreement with previously published data (25). pEGFP-N2 (Clontech). The cDNA of coilin was reverse transcription–PCR cloned in frame within the NheI/HindIII sites of pEGFP-N1 (Clontech), resulting in the Gemin2 knock-down did not affect the number of Cajal bodies; pCoilin-EGFP plasmid. The plasmid GFP-mCoilin⌬RG (12) was a generous gift whereas knock-down of SMN caused a significant decrease in the of G. Matera (University of North Carolina at Chapel Hill, NC). number of Cajal bodies (37). Gemin2 had a more subtle, but clearly visible, effect on nuclear body formation upon down- Competition Assay. Recombinant SMN protein was expressed in E. coli using

regulation. In agreement with these data, overexpression of the pET System (Novagene), as described previously, and then purified and CELL BIOLOGY FGF-223 in vitro and in vivo led to a decrease in gems but not immobilized on glutathione agarose (22). Gemin2 was synthesized in vitro by Cajal bodies. This effect was not observed with the non- use of the SP6-TnT-Quick (Promega) coupled Transcription/Translation Sys- interacting isoform FGF-218, which confirms the specific effect tem. Proteins were analyzed and quantified by Coomassie staining after of FGF-223. sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). FGF-2 The functional context of FGF-223-binding to SMN was also was purified as described earlier (22), but with the modification of using the E. coli strain BL21 DE3-trx. A constant amount of Ϸ1 ␮g SMN protein bound evaluated. Although the number of gems is significantly reduced to glutathione beads was used for the competition experiments. Beads were in motoneurons of FGF-2 transgenic mice, no SMA-like phe- blocked for 30 minutes at 4 °C in 1 ml IP-buffer 300 (20 mmol/l Tris/HCl, pH 7.5, notype developed in the 3-month-old animals used for these 300 mmol/l NaCl, 2 mmol/l EDTA, 25 mmol/l ␤-glycerophosphate, 1 mmol/l analyses. These data indicate that the number or structural sodium orthovanadate, 1% Triton X-100, 1% sodium deoxycholate, and 1ϫ

Bruns et al. PNAS ͉ August 4, 2009 ͉ vol. 106 ͉ no. 31 ͉ 12751 Downloaded by guest on October 6, 2021 Complete protease inhibitor (Roche) supplemented with 1% [wt/vol] bovine Additional methods are described in SI Text. serum albumin [BSA]). Gemin2 protein (2.5 ␮g) was added to each competition 23 reaction as well as increasing amounts of recombinant FGF-2 . IP-buffer 300 was ACKNOWLEDGMENTS. We thank Hella Brinkmann, Kerstin Kuhlemann, Na- added to a total reaction volume of 300 ␮l. The reactions were incubated for 2 tascha Heidrich, and Hildegard Streich for expert technical help. The anti- hours at 4 °C and subsequently washed three times with 1 ml IP-buffer 300 and coilin antibody was a generous gift of Angus Lamond. We are also very two times with 1 ml phosphate-buffered saline (PBS) followed by Western blot grateful for the gift of the GFP-mCoilin⌬RG plasmid from Greg Matera. This analysis as previously described (21). The primary antibodies anti-FGF-2 (BD study was supported by the Fritz Thyssen Foundation, Ko¨ln, Germany and the Biosciences) and anti-Gemin2 (Abcam) were used for Western blotting. Ro¨chling Foundation, Mannheim, Germany.

1. Dundr M, Misteli T (2001) Functional architecture in the cell nucleus. Biochem J 28. Young PJ, et al. (2000) The exon 2b region of the spinal muscular atrophy protein, SMN, 356:297–310. is involved in self-association and SIP1 binding. Hum Mol Genet 9:2869–2877. 2. Handwerger KE, Gall JG (2006) Subnuclear organelles: New insights into form and 29. Liu Q, Fischer U, Wang F, Dreyfuss G (1997) The spinal muscular atrophy disease gene function. Trends Cell Biol 16:19–26. product, SMN, and its associated protein SIP1 are in a complex with spliceosomal snRNP 3. Ramoń y Cajal SR (1903) A simple method for selective staining of the protoplasmic proteins. Cell 90:1013–1021. reticulum and its effects on various neuronal organs from vertebrates and inverte- 30. Lorson CL, Androphy EJ (1998) The domain encoded by exon 2 of the survival motor brates (Translated from Spanish). Trab Lab Inverst Biol 2:129–221. neuron protein mediates nucleic acid binding. Hum Mol Genet 7:1269–1275. 4. Gall JG (2000) Cajal bodies: The first 100 years. Annu Rev Cell Dev Biol 16:273–300. 31. Lorson CL, et al. (1998) SMN oligomerization defect correlates with spinal muscular 5. Cioce M, Lamond AI (2005) Cajal bodies: A long history of discovery. Annu Rev Cell Dev atrophy severity. Nat Genet 19:63–66. Biol 21:105–131. 32. Malecki J, Wesche J, Skjerpen CS, Wiedlocha A, Olsnes S (2004) Translocation of FGF-1 6. Andrade LE, et al. (1991) Human autoantibody to a novel protein of the nuclear coiled and FGF-2 across vesicular membranes occurs during G1-phase by a common mecha- body: Immunological characterization and cDNA cloning of p80-coilin. J Exp Med nism. Mol Biol Cell 15:801–814. 173:1407–1419. 33. Coffin JD, et al. (1995) Abnormal bone growth and selective translational regulation 7. Kaiser TE, Intine RV, Dundr M (2008) De novo formation of a subnuclear body. Science in basic fibroblast growth factor (FGF-2) transgenic mice. Mol Biol Cell 6:1861–1873. 322:1713–1717. 34. Paushkin S, Gubitz AK, Massenet S, Dreyfuss G (2002) The SMN complex, an assem- 8. Jady BE, et al. (2003) Modification of Sm small nuclear RNAs occurs in the nucleoplasmic blyosome of ribonucleoproteins. Curr Opin Cell Biol 14:305–312. Cajal body following import from the cytoplasm. EMBO J 22:1878–1888. 35. Girard C, Neel H, Bertrand E, Bordonne R (2006) Depletion of SMN by RNA interference 9. Liu Q, Dreyfuss G (1996) A novel nuclear structure containing the survival of motor in HeLa cells induces defects in Cajal body formation. Nucleic Acids Res 34:2925–2932. neurons protein. EMBO J 15:3555–3565. 10. Lefebvre S, et al. (1995) Identification and characterization of a spinal muscular 36. Lemm I, et al. (2006) Ongoing U snRNP biogenesis is required for the integrity of Cajal atrophy-determining gene. Cell 80:155–165. bodies. Mol Biol Cell 17:3221–3231. 11. Gubitz AK, Feng W, Dreyfuss G (2004) The SMN complex. Exp Cell Res 296:51–56. 37. Shpargel KB, Matera AG (2005) Gemin proteins are required for efficient assembly of 12. Hebert MD, Szymczyk PW, Shpargel KB, Matera AG (2001) Coilin forms the bridge Sm-class ribonucleoproteins. Proc Natl Acad Sci USA 102:17372–17377. between Cajal bodies and SMN, the spinal muscular atrophy protein. Genes Dev 38. Briese M, Esmaeili B, Sattelle DB (2005) Is spinal muscular atrophy the result of defects 15:2720–2709. in motor neuron processes? Bioessays 27:946–957. 13. Sleeman JE, Ajuh P, Lamond AI (2001) snRNP protein expression enhances the forma- 39. Jablonka S, Rossoll W, Schrank B, Sendtner M (2000) The role of SMN in spinal muscular tion of Cajal bodies containing p80-coilin and SMN. J Cell Sci 114:4407–4419. atrophy. J Neurol 24(Suppl 1):I37–I42. 14. Shpargel KB, Ospina JK, Tucker KE, Matera AG, Hebert MD (2003) Control of Cajal body 40. Jablonka S, Sendtner M (2003) Molecular and cellular basis of spinal muscular atrophy. number is mediated by the coilin C-terminus. J Cell Sci 116:303–312. Amyotroph Lateral Scler Other Motor Neuron Disord 4:144–149. 15. Spector DL, Lark G, Huang S (1992) Differences in snRNP localization between trans- 41. Monani UR (2005) Spinal muscular atrophy: A deficiency in a ubiquitous protein; a formed and nontransformed cells. Mol Biol Cell 3:555–569. motor neuron-specific disease. Neuron 48:885–896. 16. Young PJ, et al. (2001) Nuclear gems and Cajal (coiled) bodies in fetal tissues: Nucleolar 42. Zhang HL, et al. (2003) Active transport of the survival motor neuron protein and the distribution of the spinal muscular atrophy protein, SMN. Exp Cell Res 265:252–261. role of exon-7 in cytoplasmic localization. J Neurosci 23:6627–6637. 17. Morse R, Shaw DJ, Todd AG, Young PJ (2007) Targeting of SMN to Cajal bodies is 43. van Bergeijk J, Rydel-Konecke K, Grothe C, Claus P (2007) The spinal muscular atrophy mediated by self-association. Hum Mol Genet 16:2349–2358. gene product regulates neurite outgrowth: Importance of the C terminus. FASEB J 18. Navascues J, Berciano MT, Tucker KE, Lafarga M, Matera AG (2004) Targeting SMN to 21:1492–1502. Cajal bodies and nuclear gems during neuritogenesis. Chromosoma 112:398–409. 44. Zhang H, Xing L, Singer RH, Bassell GJ (2007) QNQKE targeting motif for the SMN- 19. Coovert DD, et al. (1997) The survival motor neuron protein in spinal muscular atrophy. Gemin multiprotein complexin neurons. J Neurosci Res 85:2657–2667. Hum Mol Genet 6:1205–1214. 45. Gringel S, van Bergeijk J, Haastert K, Grothe C, Claus P (2004) Nuclear fibroblast growth 20. Pellizzoni L, Kataoka N, Charroux B, Dreyfuss G (1998) A novel function for SMN, the factor-2 interacts specifically with splicing factor SF3a66. Biol Chem 385:1203–1208. spinal muscular atrophy disease gene product, in pre-mRNA splicing. Cell 95:615–624. 46. Brosi R, Groning K, Behrens SE, Luhrmann R, Kramer A (1993) Interaction of mamma- 21. Claus P, et al. (2003) Differential intranuclear localization of fibroblast growth factor-2 lian splicing factor SF3a with U2 snRNP and relation of its 60-kD subunit to yeast PRP9. isoforms and specific interaction with the survival of motoneuron protein. J Biol Chem Science 262:102–105. 278:479–485. 47. Stachowiak MK, Maher PA, Joy A, Mordechai E, Stachowiak EK (1996) Nuclear local- 22. Claus P, Bruns AF, Grothe C (2004) Fibroblast growth factor-2(23) binds directly to the ization of functional FGF receptor 1 in human astrocytes suggests a novel mechanism survival of motoneuron protein and is associated with small nuclear RNAs. Biochem J for growth factor action. Brain Res Mol Brain Res 38:161–165. 384:559–565. 48. Stachowiak MK, Maher PA, Stachowiak EK (2007) Integrative nuclear signaling in cell 23. Sorensen V, Nilsen T, Wiedlocha A (2006) Functional diversity of FGF-2 isoforms by intracellular sorting. Bioessays 28:504–514. development—a role for FGF receptor-1. DNA Cell Biol 26:811–826. 24. Grothe C, Meisinger C, Claus P (2001) In vivo expression and localization of the 49. Peng H, et al. (2001) Novel nuclear signaling pathway mediates activation of fibroblast fibroblast growth factor system in the intact and lesioned rat peripheral nerve and growth factor-2 gene by type 1 and type 2 angiotensin II receptors. Mol Biol Cell spinal ganglia. J Comp Neurol 434:342–357. 12:449–462. 25. Feng W, et al. (2005) Gemins modulate the expression and activity of the SMN complex. 50. Stachowiak EK, Fang X, Myers J, Dunham S, Stachowiak MK (2003) cAMP-induced Hum Mol Genet 14:1605–1611. differentiation of human neuronal progenitor cells is mediated by nuclear fibroblast 26. Ogawa C, et al. (2007) Gemin2 plays an important role in stabilizing the survival of growth factor receptor-1 (FGFR1). J Neurochem 84:1296–1312. motor neuron complex. J Biol Chem 282:11122–11134. 51. Dunham-Ems SM, et al. (2009) Fibroblast growth factor receptor-1 (FGFR1) nuclear 27. Wang J, Dreyfuss G (2001) Characterization of functional domains of the SMN protein dynamics reveal a novel mechanism in transcription control. Mol Biol Cell 20:2401– in vivo. J Biol Chem 276:45387–45393. 2412.

12752 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900122106 Bruns et al. Downloaded by guest on October 6, 2021