Nuclear GRP75 Binds Retinoic Acid Receptors to Promote Neuronal Differentiation of Neuroblastoma

Retinoic acid (RA) has been approved for the differentiation therapy of neuroblastoma (NB). Previous work revealed a correlation between glucose-regulated protein 75 (GRP75) and the RA-elicited neuronal differentiation of NB cells. The present study further demonstrated that GRP75 translocates into the nucleus and physically interacts with retinoid receptors (RARα and RXRα) to augment RA-elicited neuronal differentiation. GRP75 was required for RARα/RXRα-mediated transcriptional regulation and was shown to reduce the proteasome-mediated degradation of RARα/RXRαin a RA-dependent manner. More intriguingly, the level of GRP75/RARα/RXRα tripartite complexes was tightly associated with the RA-induced suppression of tumor growth in animals and the histological grade of differentiation in human NB tumors. The formation of GRP75/RARα/RXRα complexes was intimately correlated with a normal MYCN copy number of NB tumors, possibly implicating a favorable prognosis of NB tumors. The present findings reveal a novel function of nucleus-localized GRP75 in actively promoting neuronal differentiation, delineating the mode of action for the differentiation therapy of NB by RA.


Introduction
Neuroblastoma (NB) is the most common and deadly cancer in patients who are identified during the first year of life, and are often diagnosed as an aggressive and metastatic disease that leads to high mortality [1]. Despite the noted improvement in the overall outcome in patients with NB, the 5-year survival rates among children with high-risk NB have only improved slightly. This result is mainly attributed to the fact that key molecular pathways controlling NB tumorigenesis remain elusive.
Current treatments for NB include a combination of chemotherapy, surgery, radiation, bone marrow transplantation, immunotherapy, and differentiating agents [2]. NB is the only pediatric cancer treated with differentiation reagents as the first-line of defense [2]. One of the most potent differentiation inducers for NB is retinoic acid (RA) [3,4]. The results of independent trials have consistently shown that the administration of RA significantly improves the overall survival rates after bone-marrow transplantation [5,6,7]. Consistent with these findings, NB cells treated with all-trans RA at doses used clinically display evident neuronal differentiation and reduced proliferation [8]. Therefore, the elucidation of molecular mechanisms underlying RA-induced neuronal differentiation in NB could pave the way for the development of novel therapeutic strategies for NB.
RA-elicited signaling is primarily mediated by retinoid receptors [9]. The retinoid receptors can be categorized into two subfamilies: retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Upon binding RA, the ligand-bound RAR/ RXR heterodimers undergo a conformational change, causing their translocation to the nucleus, where they act as transcription factors targeting retinoic acid responsive elements (RARE) within the promoters of genes involved in differentiation and growth arrest [10,11]. Recent studies show that the activity of RA receptors can be regulated by posttranslational modifications, such as phosphorylation and ubiquitination [12]. Moreover, the ligandinduced transactivation of RAR/RXR heterodimers could also be modulated by various adaptor proteins in the nucleus [13].
Glucose-regulated protein 75 (GRP75) was first identified as a member of the Hsp70 family that could function in multiple subcellular compartments [14]. Accumulated evidence has demonstrated the versatility of GRP75 in regulating cellular stress responses, mitochondrial homeostasis, intracellular trafficking, antigen presenting, cell proliferation, differentiation, and tumorigenesis [15]. Previous work from our group demonstrated that GRP75 is upregulated in RA-treated NB cells and in NB patients with favorable prognostic outcome [16]. The present study further investigates the possible role of altered GRP75 expression in the regulation of RA-elicited neuronal differentiation in NB.

Results
Nuclear translocation of GRP75 is significantly enhanced upon retinoic acid-induced neuronal differentiation in neuroblastoma cells We previously demonstrated that GRP75 expression is significantly increased in retinoic acid-treated NB cells and is associated with a favorable prognosis in NB patients [16]. In the present study, immunofluorescence confocal microscopy revealed that GRP75 was highly enriched in the nuclei of RA-treated NB cells compared to untreated cells. The cross-sectional views of stacked images unambiguously revealed that the nuclear translocation of GRP75 was significantly enhanced for an approximately 4-fold increase in RA-treated cells compared to DMSO-treated controls (Figure S1A-C). To further confirm the RA-induced nuclear localization of GRP75, the levels of GRP75 in both nuclear and cytoplasmic fractions prepared from RA-and DMSO-treated NB cells were determined by Western blotting, and the results showed a dramatic increase in nuclear GRP75 in RA-treated cells compared to control cells, while the cytosolic pools of GRP75 remained unchanged in response to RA treatment ( Figure S1D). We also found that RA can induce the nuclear translocation of GRP75 in a dose-dependent manner ( Figure S13). This RA-induced nuclear translocation of GRP75 was independently confirmed in two separate NB cell lines (SK-N-DZ and SK-N-SH NB, Figure S12). Together, the present findings provide the first direct evidence that GRP75 can be localized to the nucleus, where it potentially plays a critical role in the RA-elicited neuronal differentiation of NB cells.
The interaction between nuclear GRP75 with retinoic acid receptors is significantly enhanced in differentiated NB cells RA primarily binds to RARa/RXRa heterodimers to transduce its downstream signaling [17]. It is therefore possible that the nucleus-localized GRP75 could be actively involved in the RAinduced neuronal differentiation of NB cells by interacting with RA receptors and modulating their activities. To investigate this possible mechanism, the potential association between GRP75 and RARa/RXRa in the nucleus in response to RA-induced neuronal differentiation was investigated. Using co-immunoprecipitation, a physical interaction between endogenous GRP75 and RARa and RXRa in the nucleus was demonstrated. This novel interaction between GRP75 and RARa/RXRa was significantly enhanced in the nucleus in RA-treated NB cells compared to controls, but it was trivial and irresponsive to RA treatment in the cytosol ( Figure 1A, B). Co-immunoprecipitation with a non-specific mouse IgG did not reveal any detectable interaction between GRP75 and RA receptors ( Figure S11). Moreover, the co-localization of GRP75 with RARa/RXRa heterodimers was vividly observed in the nucleus of RA-treated SH-SY5Y cells and increased upon RA stimulation ( Figure 1C-F, Figure S2), suggesting that RA could selectively and significantly enhance the binding of GRP75 to RARa/RXRa. The interaction between GRP75 and RARa/ RXRa was independently validated in two MYCN-nonamplified (SK-N-SH and SK-N-MC) and three MYCN-amplified NB cell lines (SK-N-DZ, SK-N-BE, and IMR-32) ( Figure S3), suggesting a critical role of GRP75 in RA signaling and NB differentiation. Together, these results strongly suggest that GRP75 participates in RA-elicited neuronal differentiation of NB cells through its direct interaction with nuclear RARa/RXRa heterodimers.
Down-regulation of GRP75 inhibits RA-elicited activation of RARa/RXRa receptors To determine whether the binding of GRP75 to the RARa/ RXRa receptor complex in the nucleus is necessary for the RAinduced neuronal differentiation of NB cells, the expression of RA target genes was examined in GRP75-deficient SH-SY5Y cells in the presence or absence of RA. Using an RARa/RXRa-driven luciferase reporter gene construct (RARE-Luc), RA-induced reporter gene expression was determined to be abrogated by the down-regulation of GRP75, suggesting that GRP75 is required for RA-triggered transcriptional regulation (Figure 2A). This finding was further corroborated by data showing that GRP75-deficient SH-SY5Y cells exhibit a significant reduction in the RA-elicited activation of the RARb promoter, a known RARa/RXRa downstream target gene [18] ( Figure 2B). The knockdown efficiency of both GRP75-targeting shRNAs was approximately 75%, as seen in the mRNA transcript and protein levels ( Figure 2E and Figure S4). Using reporter gene constructs derived from MYCN and NEDD9 promoters [19,20], the depletion of GRP75 in SH-SY5Y cells was further demonstrated to effectively inhibit the RA-elicited stimulation of NEDD9 promoter activity (a neuronal differentiation marker) and abolish the RA-triggered suppression of MYCN promoter activity (a proliferation marker) ( Figures 2C, D). The present data clearly suggest that GRP75 could be a novel transcriptional coactivator of RARa/RXRa and therefore modulate RA-elicited neuronal differentiation of NB cells.
To further substantiate the role of GRP75 in RARa/RXRamediated transcriptional regulation, the mRNA transcript levels of selected RA target genes were examined by real-time quantitative PCR. Several RA target genes, such as CLMN, CRABP2, HOXD10, RARb, NAV2, NEDD9, RET, TH, and TrkA, have been shown to promote neuronal differentiation, whereas others, including MYCN, Survivin, and p34, function in cell proliferation [19,21,22,23,24]. In the present study, while the expression of nine RA target genes involved in neuronal differentiation was significantly elevated in RA-treated NB cells compared to controls, knockdown of GRP75 dramatically compromised the RA-induced transcriptional up-regulation of these pro-differentiation target genes in NB cells ( Figure 2G). Concomitant with defective neuronal differentiation, down-regulation of GRP75 could favor cell growth by diminishing the RA-elicited transcriptional suppression of pro-proliferation target genes (MYCN, Survivin, and p34) ( Figure 2F).
Consistent with the essential role of GRP75 in neuronal differentiation, the ectopic expression of GRP75 in SH-SY5Y cells significantly enhanced RA-triggered transactivation of RARa/RXRa ( Figure 3A-C). The GRP75-dependent transcriptional regulation of RA target genes was further explored through the ectopic expression of GRP75 in SH-SY5Y cells. Overexpression of GRP75 significantly potentiated the expression of differentiation-promoting genes and suppressed the expression of pro-proliferation genes in NB cells with or without RA ( Figure 3E, F). Together, the regulation of RA-elicited neuronal differentiation by GRP75 was independently confirmed in separate albeit related NB cell lines SH-SY5Y and SK-N-SH ( Figure 2, Figure 3, Figure  Figure 1. Nuclear GRP75 physically interacts with RARa/RXRa in differentiated NB cells. (A and B) Nuclear and cytoplasmic extracts were prepared from SH-SY5Y cells treated with RA (10 mM) or vehicle alone (0.1% DMSO) for various intervals. Clarified lysates of treated cells were subjected to immunoprecipitation using anti-GRP75 antibodies, followed by Western blot analysis with anti-RARa or RXRa. The levels of immunoprecipitated nuclear RARa and RXRa were normalized with those of nuclear GRP75 from the same immunopreciptate. The ratio of RARa or RXRa to GRP75 in DMSO-treated cells at a specific interval was referred to as 1 fold of relative interaction. Quantitative results are shown as the mean interactions of GRP75 with RARa or RXRa (6SEM) of three individual experiments and were analyzed by Student's t test. *p,0.05. (C-F) Immunofluorescence staining representing the colocalization between GRP75 and RARa/RXRa. SH-SY5Y cells were treated with RA (10 mM) or vehicle alone (0.1% DMSO) for 3 d and processed for immunofluorescence staining with anti-MAP2, anti-GRP75, and either anti-RARa or RXRa. Nuclei were visualized by DAPI counterstaining. Three-dimensional analysis of the co-localization of GRP75 and either RARa or RXRa by z-stack images was denoted by intersecting lines in the x, y, and z axes. Scale bar = 20 mm. Overlapped pixels (yellow) corresponding to GRP75 (green) and either RARa (C, red) or RXRa (E, red) were defined as the co-localization of GRP75 with either RARa or RXRa. The co-localization of GRP75 with either RARa (D) or RXRa (F) was determined as the mean (6SEM) percentages of nuclear GRP75-specific pixels overlapped with either RARa-or RXRa-specific pixels from at least three different viewing areas per experiment in three independent experiments and analyzed by Student's t test. *p,0.05. doi:10.1371/journal.pone.0026236.g001 shRNA-infected SH-SY5Y cells treated with RA or DMSO as described above were harvested and processed for total RNA isolation by TRIzol reagent. The levels of GRP75 mRNA transcript as well as GAPDH (internal control) were determined by real-time RT-PCR. The normalized level of GRP75 transcript in Control-sh-infected DMSO-treated cells was referred to as 1 fold of relative GRP75 expression. Quantitative data were calculated as the mean (6SEM) relative GRP75 expression of triplicate measurements from three independent experiments and analyzed by Student's t test. (F-G) Total RNA transcripts of shRNA-infected SH-SY5Y cells treated with RA or DMSO were analyzed by real-time RT-PCR for the expression of RA target genes essential for cell proliferation (F) and neuronal differentiation (G). All quantitative data were calculated as the mean (6SEM) from three independent experiments and analyzed by Student's t test. *p,0.05. doi:10.1371/journal.pone.0026236.g002  Figure S10). The present results clearly demonstrate for the first time that GRP75 can be actively involved in RARa/ RXRa-mediated transcriptional regulation in RA-triggered neuronal differentiation of NB cells.

GRP75/RARa/RXRa complexes cooperatively bind to the retinoic acid response element (RARE) within the promoters of RA-responsive genes
To further demonstrate the functional role of the GRP75-bound RARa/RXRa receptor complex in RA-elicited transcriptional regulation, the RARa/RXRa-mediated binding to DR5 RARE within the promoter of RARb, a direct target gene downstream of RA-elicited signaling [18], was examined using chromatin immunoprecipitation (ChIP) in SH-SY5Y cells infected with lentiviral particles expressing shRNAs targeting GFP (Control-sh) or GRP75 (GRP75-sh-1 and GRP75-sh-2) in the presence or absence of RA. The RNAi-mediated down-regulation of GRP75 significantly diminished the recruitment of RARa and RXRa receptors to RARE-containing promoter regions in RA-treated NB cells ( Figure 4A), suggesting that GRP75 can be recruited to RARE consensus sequence within the promoter of RARb and indispensable for the binding of RARa/RXRa receptor complexes to the promoters of RA target genes and that it critically modulates the RA-elicited expression of pro-differentiation genes. Since GRP75, RARa, and RXRa are all indispensable for cell viability, the depletion of GRP75 in SH-SY5Y cells in the absence of RA could compromise the basal viability of NB cells [9,25,26]. It is thus possible that the recruitment of RARa/RXRa to the RARE region could be enhanced to compensate the loss in GRP75-dependent maintenance of cell viability in response to GRP75 knockdown. These findings thus favor a model in which the binding of GRP75 is a prerequisite for the efficient association of RARa/RXRa complexes with RARE-containing promoter regions to modulate the expression of RA target genes in the regulation of neuronal differentiation.

Down-regulation of GRP75 promotes RARa/RXRa degradation through a proteasome-mediated pathway
Accumulated evidence also suggests that the ubiquitin proteasome system (UPS)-mediated degradation and the binding to molecular chaperones can modulate the activity of steroid receptors, including RAR/RXR retinoid receptor family [27,28,29,30]. We thus demonstrated that RNAi-mediated down-regulation of GRP75 in RA-treated SH-SY5Y cells resulted in a significant increase in the ubiquitination and UPS-mediated degradation of ligand-bound RARa/RXRa, compared to mock-infected cells ( Figure 5). Consistent with its protective role, the overexpression of GRP75 resulted in a significant attenuation in the ubiquitination of both RARa and RXRa ( Figure S5). Together, these findings suggest that GRP75dependent increase in the stability of ligand-bound RARa/RXRa heterodimers could selectively potentiate ligand-bound RARa/ RXRa-mediated transcriptional regulation, synergistically augmenting RA-elicited neuronal differentiation.

An enhanced interaction between GRP75 and RARa/ RXRa heterodimers is associated with favorable outcomes in an in vivo xenograft NB mouse model
To validate the significance of the interaction between GRP75 and RARa/RXRa heterodimers in the modulation of RA-elicited neuronal differentiation in vivo, the correlation of these novel protein-protein interactions with tumor progression was examined in a xenograft NB mouse model [31]. We found that RA treatment significantly inhibits tumor growth beginning on the fourth treatment day until the completion of the treatment period in comparison to vehicle-treated control animals based on tumor size measurements. The assessment of harvested tumor xenografts at the conclusion of treatment revealed that both tumor size and tumor weight were significantly decreased in RA-treated mice (Tables S1), indicating that RA therapy can effectively inhibit the progression of NB. The correlation between the level of GRP75/ RARa/RXRa tripartite complexes and tumor growth was further analyzed in the harvested NB xenografts. Our data showed that the association of GRP75 with either RARa or RXRa is significantly increased in RA-treated xenografts compared to controls ( Figure 6, Table S1). Linear correlation analysis demonstrated that the formation of either GRP75/RARa or GRP75/RXRa complexes was markedly elevated in tumors with reduced volume or weight (RA-treated xenografts), in stark contrast to the control tumors with larger volume or weight ( Figure S6). These results clearly demonstrate that the complex formation of tripartite GRP75/RARa/RXRa is inversely correlated with the progression of NB.
The interaction between GRP75 and RXRa/RARais correlated with a higher grade of histological differentiation and a normal MYCN copy number in human NB tumors To corroborate the relationship between the RA-induced formation of nucleus-localized GRP75/RARa/RXRa tripartite complexes with the differentiation of human NB, the interaction between GRP75 and RARa/RXRa heterodimers was examined in NB tumors with various grades of histological differentiation. Using co-immunoprecipitation with an anti-GRP75 antibody and lysates derived from 30 human NB tumors, including 13 ganglioneuroblastomas (GNBs), 9 differentiating NBs (DNBs), and 8 undifferentiated NBs (UNBs), the interaction between GRP75 and RARa/RXRa was found to be significantly stronger in tumors with higher grades of histological differentiation (GNB and DNB) than in those with an undifferentiated histology (UNB) ( Figure 7A, B and Figure S8). Furthermore, the interaction RARb-Luc) and a GRP75-expressing construct at various concentrations for 24 h, followed by treatment with 10 mM RA for 24 h. Luciferase signals derived from reporter gene constructs were determined by Steady-Glo luciferase assay reagents and normalized by protein concentration. The normalized luciferase signal in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative luciferase activity. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments and were analyzed by Student's t test. *p,0.05. (D) The levels of GRP75 mRNA transcripts in SH-SY5Y cells transiently transfected with a GRP75-expressing vector were determined by quantitative real-time RT-PCR. The normalized level of GRP75 transcripts in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative expression. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments. (E and F) SH-SY5Y cells were transfected with an empty vector or a GRP75-expressing vector for 48 h, followed by treatment with or without RA for 24 h. The transcript levels of various RA-responsive genes in transfected cells were determined by quantitative realtime RT-PCR. The normalized transcript level in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative expression. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments and were analyzed by Student's t test. doi:10.1371/journal.pone.0026236.g003 between GRP75 and RARa/RXRa was higher in tumors with a normal MYCN copy number compared with those with MYCN amplification which carry a very unfavorable prognosis ( Figure 7C, D). These pieces of evidence were consistent with our finding that GRP75 is required for RA-elicited down-regulation of MYCN expression ( Figure 2D, F). The present data thus unequivocally support a model in which the level of tripartite GRP75/RARa/ RXRa complexes in NB tumors is tightly associated with the histological grade of differentiation and, possibly, a favorable prognosis of NB tumors. Our data thus favors a model in which, upon RA-induced neuronal differentiation, GRP75 could be recruited to the ligand-bound RARa/RXRa heterodimers to cooperatively regulate the expression of RA downstream genes and avert UPS-mediated degradation of RA-bound RARa/ RXRa ( Figure S7A). Given that the molecular structure of human GRP75 has not been resolved, we then based on the structure of Escherichia coli HSP70 chaperone chain A (PubMed accession P0A6Y8, Protein Data Bank code 2KHO_A) that exhibits the highest amino acid sequence identity to human GRP75 to predict the three-dimensional model of GRP75. We then further simulated the possible docking of modeled GRP75 structure to known RARa or RXRa structure. This molecular simulation predicted that GRP75 could bind to the ligand-binding domain or the DNA-binding domain of RARa/RXRa heterodimers ( Figure  S7B, C). In conclusion, the present findings further strengthen the prognostic value of the expression level of GRP75 in NB [16], and delineate the mechanism underlying the GRP75-dependent regulation of RA-elicited RARa/RXRa-dependent neuronal differentiation.

Discussion
GRP75 has been shown to bind multiple partner proteins to govern diverse cellular functions [14,32]. The alteration in the cellular distribution of GRP75 could also correlate with the status of cellular immortality [33,34]. Consistent with this notion, our data clearly show that the association of GRP75 with RARa and RXRa is remarkably increased in the nucleus and coincides with the RA-elicited growth arrest, concomitant with a tight correlation between RA-induced nuclear translocation of GRP75 and RAtriggered neuronal differentiation. These data strongly favor a model in which nuclear GRP75 could stably form complexes with RARa/RXRa heterodimers and actively participate in RAtriggered neuronal differentiation of NB cells through the persistent modulation of RARa/RXRa activity ( Figure S7).
The present data provide the first direct evidence that nucleuslocalized GRP75 is essential for RARa/RXRa-mediated transcriptional regulation and that the GRP75/RARa/RXRa tripartite complexes physically bind to RARE to modulate the expression of RA target genes for neuronal differentiation (Figures 2, 3, 4). These results also support previous findings showing that RA-elicited down-regulation of MYCN expression is a prerequisite for the neuronal differentiation of NB cells, while constitutive overexpression of MYCN can counteract RA-induced neuronal differentiation [35,36]. GRP75 could thus play a central role in coordinating the transcriptional regulation of MYCN expression through the stimulation of RARa/RXRa activity, modulating a positive auto-regulatory loop for MYCN in NB cells [20]. GRP75 could also act as a transcriptional co-activator to potentiate RARa/RXRa-mediated transactivation of pro-differentiation genes, constituting a positive feedback loop that can potentiate RA-induced neuronal differentiation of NB cells. Consistent with the essential role of GRP75 in neuronal function [37], our results thus strengthen the notion that the functions of retinoid receptors, like other nuclear receptors, could be regulated by molecular chaperones, such as GRP75, in an evolutionarily conserved fashion [38,39,40].
The activity of nuclear receptors could be regulated by posttranslational modifications, including phosphorylation and ubiquitination [41]. Previous studies have shown that phosphorylation of RARs within their TFIIH binding sites is essential for RAmediated embryonic development and underlies the pathogenesis of xeroderma pigmentosum [42]. RARs have also been shown to interact with SUG-1 to induce its ubiquitination and degradation by the proteasome upon RA stimulation [43]. A recent study revealed that Hsp27 can form a complex with androgen receptor (AR) and co-migrate into the nucleus to modulate the transcriptional activity of AR through an alteration in the UPS-mediated degradation of AR [28]. The activity of RA receptors was also shown to be modulated by interacting proteins, such as calreticulin [44]. Consistent with the functional roles of chaperones [45], the present results showing that GRP75 interacts with RARa and RXRa and is involved in the UPS-mediated degradation of these receptors further substantiate the idea that the magnitude and function of retinoid-elicited signaling could depend on the efficiency of the UPS-mediated processing of ligand-bound retinoid receptors ( Figure 5 and Figure S5). A recent study showed that the molecular chaperone HSP90 binds to the E3 ubiquitin ligase CHIP and prevents CHIP-mediated degradation of leucine-rich repeat kinase 2 (LRRK2) [46]. The present findings that GRP75 could stabilize the RARa/RXRa complex by modulating the UPS pathway to potentiate the RA-elicited  neuronal differentiation of NB cells further support a critical role of the UPS pathway in the regulation of the activity of RA receptors [47].
The association between GRP75 and RARa/RXRa heterodimers is of clinical significance. In a xenograft mouse model of NB and a cohort of patient samples, the formation of tripartite GRP75/RARa/RXRa complexes was inversely correlated with tumor progression and consistently predicted a favorable outcome ( Figure 6, Figure 7, Figure S6, and Table S1). Given that the Nterminal ATP-binding domain and the C-terminal substratebinding domain of GRP75 are highly identical to those of human Hsp70 [48,49], GRP75 could act like Hsp70 and modulate the conformations of RARa and RXRa. Based on the molecular modeling for the interaction between GRP75 and either RARa or RXRa ( Figure S7), GRP75 is likely to interact with either the ligand-binding domain or the DNA-binding domain of RARa/ RXRa heterodimers to synergistically stabilize the RA-bound GRP75/RARa/RXRa complexes and sustain their transcriptional activation. Synthetic molecules designed to either induce the de novo formation of GRP75/RARa/RXRa complexes or stabilize the pre-existent ones could thus have profound therapeutic implication for NB.
In summary, the present study identifies a novel function of GRP75 in regulating RA-elicited neuronal differentiation through direct interaction with RARa/RXRa heterodimers in the nucleus. The elucidation of the molecular mechanism involved in the formation of tripartite GRP75/RARa/RXRa complexes provides the basis for the development of a novel therapeutic strategy for NB that could be combined with other existing differentiating regimens to improve the overall outcomes of NB patients [50,51]. Our data could serve as the foundation for the generation of molecules that could simultaneously prevent UPS-mediated degradation of RARa/RXRa and extend the pro-differentiation effect of GRP75/RARa/RXRa complexes.

Ethics Statement
The Institutional Review Board of National Taiwan University Hospital approved the complete follow-up protocols and this study. We obtained the written informed consent from all participants involved in this study. The animal study protocol used in this study was approved by the Institutional Animal Care and Use Committee of Academia Sinica (Approval Protocol ID Immunofluorescence confocal microscopy SH-SY5Y cells were grown on coverslips and treated with 10 mM RA in DMSO or 0.1% DMSO alone for the indicated time. Immunofluorescence staining was performed as described previously [16].

Reporter gene constructs and lentiviral shRNAs
The DR5-TK and RARb-luciferase reporter constructs were kindly provided by Dr. Jonathan Kurie (The University of Texas M.D. Anderson Cancer Center) [52]. The Nedd9 promoter constructs were kindly provided by Dr. Margaret Clagett-Dame (University of Wisconsin-Madison) [19]. The shRNA lentiviral vectors targeting human GRP75 (GRP75-sh-1 and GRP75-sh-2) and a GFP-targeting control lentiviral vector (Control-sh) were purchased from the National RNAi Core Facility, Academia Sinica, Taiwan.

Co-immunoprecipitation and Western blotting
The nuclear extracts and cytosolic pools of treated cells were isolated using the Nuclei EZ Prep Kit (Sigma) as described in the manufacturer's instructions. For co-immunoprecipitation, 20 mg of goat anti-GRP75 antibody was incubated with 200 ml of immobilized protein A agarose in PBS for 2 h at 4uC. Immunoprecipitated proteins were eluted by the addition of 4X sample loading buffer and boiling at 100uC for 10 min, followed by SDS-PAGE and Western blotting.

Luciferase reporter assay
shRNA-transduced cells were transfected with a promoterspecific luciferase reporter gene construct (either RARE-luc, RARb-luc, NEDD9-luc, or MYCN-luc) using Lipofectamine 2000 at 37uC overnight. Promoter-driven luminescence in the clarified lysates of transfected cells was determined using the Steady-Glo luciferase assay reagent (Promega) and normalized to the protein content of the lysates.

RNA isolation, reverse transcription, and quantitative real-time PCR
Total RNA was isolated from transduced SH-SY5Y cells using the TRIzol Reagent (Invitrogen). Purified RNAs (2 mg) were reverse-transcribed to the first-strand cDNA using the SuperScript III First-strand cDNA Synthesis Kit (Invitrogen). Equivalent amounts of cDNA were used in quantitative realtime PCR using the SYBR Green I Master reagent on the LightCycler 480 Real-Time PCR System (Roche) with genespecific primer pairs.

Chromatin immunoprecipitation (ChIP)
Treated cells were fixed with 1% formaldehyde for 10 min at 37uC, and the reaction was terminated by the addition of 1.25 M glycine (1:9, v/v). Fixed cells were lysed and sonicated in buffer A (1% SDS, 10 mM EDTA, 50 mM Tris, pH 8.0) to shear DNA. Clarified lysates were subject to immunoprecipitation using antibodies (10 mg) against either GRP75, RARa, RXRa, or mouse IgG (as control) at 4uC overnight with agitation, followed by incubation with protein A-conjugated agarose at 4uC for 1 h.

Protein stability
Protein synthesis in infected cells was inhibited with 50 mg/ml cycloheximide for 3 h, followed by the addition of 10 mM RA or vehicle alone (0.1% DMSO) and incubation at 37uC for various intervals.

Ubiquitination assay for RARa and RXRa
SH-SY5Y cells were infected with GRP75-sh-1, GRP75-sh-2, or control-sh at 37uC for 2 d, followed by the removal of infection mixture and incubation with fresh medium containing 10 mM RA or 0.1% DMSO for 16 h in the presence or absence of 5 mM MG132.

Patients and sample preparation
A cohort of 30 histologically confirmed NB patients with complete follow-up protocols approved by the Institutional Review Board of National Taiwan University Hospital, Taipei, Taiwan, were enrolled in this study. The Institutional Review Board of National Taiwan University Hospital also approved this study, and we obtained informed written consent from all participants involved in this study. Tumor samples were obtained during surgery and immediately frozen in liquid nitrogen. The categorization of tumor biopsies was based on the International Neuroblastoma Pathology Classification scheme [53].
Additional information regarding details of individual experimental procedures can be found in the Supplemental Methods S1 submitted along with the main manuscript. Figure S1 GRP75 is translocated into the nucleus of differentiated neuroblastoma cells. (A) Immunofluorescence microscopy analysis of GRP75 in the nuclei of NB cells. SH-SY5Y cells were treated with or without RA (10 mM) for 3 d and processed for immunofluorescence staining with an anti-GRP75 antibody (green). Nuclei were counterstained with DAPI (blue). Insets, two-fold magnification of highlighted cells (arrow). Scale bar = 20 mm. (B) Three-dimensional analysis of individual cells by z-stack confocal images at specific sites marked by intersecting lines in the x, y, and z axes. (C) Quantitative analysis of the intensity of cells double labeled for GRP75 and DAPI (nuclear DNA). Data are expressed as the average percentage (6SEM) of nuclear GRP75 co-localized with RARa from three independent experiments. *p,0.05. (D) The nuclear extracts of SH-SY5Y cells treated with or without RA for various intervals were resolved by SDS-PAGE and analyzed by immunoblotting with the indicated antibodies. Histone H1 and Lamin A/C were markers for nuclear extracts, while GAPDH was included as a protein loading control for cytosolic pools. The prolonged exposure for GAPDH blot (Nucleus) revealed no contamination of cytosolic proteins in the isolated nuclear extracts. Similarly, overexposure of histone H1and lamin A/C-labeled blots (Cytosol) showed that isolated cytosolic pools were free from contamination of nuclear proteins. (TIF) Figure S2 The interaction between GRP75 and RARa/ RXRa is increased in RA-treated SH-SY5Y cells. (A) SH-SY5Y cells were grown on coverslips and treated with 10 mM RA for various intervals. Cells treated with vehicle alone (0.1% DMSO) were included as controls. Treated cells were fixed with 4% paraformaldehyde and subjected to immunofluorescence staining using goat anti-GRP75 (green), mouse anti-MAP2 (red), and rabbit anti-RARa (red). Nuclei were counterstained with DAPI. The inset shows the magnification of the highlighted region Following treatment with RA (10 mM) or vehicle alone (0.1% DMSO) for 24 h at 37uC, the luciferase signals in clarified lysates of treated cells were determined and normalized with protein concentration. Normalized luciferase signal of DMSO-treated Control-sh-infected cells were referred to as one fold of relative luciferase activity. (F-G) Infected SK-N-SH cells treated with RA or DMSO as described above were harvested and processed for total RNA isolation by TRIzol reagent. Total RNA transcripts of shRNA-infected SK-N-SH cells treated with RA or DMSO were analyzed by real-time RT-PCR for the expression of RA target genes essential for cell proliferation (F) and neuronal differentiation (G). The normalized level of GRP75 transcript in Control-shinfected DMSO-treated cells was referred to as 1 fold of relative expression. All quantitative data were calculated as the mean (6SEM) from three independent experiments and analyzed by Student's t test. *p,0.05. (TIF) Figure S10 Overexpression of GRP75 strengthens RAelicited activation of RA receptors in SK-N-SH cells. (A) SK-N-SH cells were transiently transfected with an empty vector or a GRP75-expression vector for 24 h. Ectopic expression of GFP-GRP75 in RA-treated transfected cells was analyzed by Western blot analysis with anti-GFP antibody (upper panel, for GRP75). GAPDH (lower panel) was used as a protein loading control. (B and C) SK-N-SH cells were transiently co-transfected with a RA-responsive reporter gene construct (2 mg of RARE-Luc or RARb-Luc) and a GRP75-expressing construct at various concentrations for 24 h, followed by treatment with 10 mM RA for 24 h. Luciferase signals derived from reporter gene constructs were determined by Steady-Glo luciferase assay reagents and normalized by protein concentration. The normalized luciferase signal in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative luciferase activity. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments and were analyzed by Student's t test. *p,0.05. (D) The levels of GRP75 mRNA transcripts in SK-N-SH cells transiently transfected with a GRP75-expressing vector were determined by quantitative real-time RT-PCR. The normalized level of GRP75 transcripts in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative expression. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments. (E and F) SK-N-SH cells were transfected with an empty vector or a GRP75-expressing vector for 48 h, followed by treatment with or without RA for 24 h. The transcript levels of various RAresponsive genes in transfected cells were determined by quantitative real-time RT-PCR. The normalized transcript level in DMSO-treated cells transfected with an empty vector alone was referred to as 1 fold of relative expression. Quantitative results are presented as the mean (6SEM) of triplicate measurements from three independent experiments and were analyzed by Student's t test.

Supporting Information
(TIF) Figure S11 The specificity of the mouse anti-GRP75 antibody for co-immunoprecipitation is validated in cultured cells, tumor xenograft, and human primary NB tumors. The cellular lysates (nuclear and cytosolic fractions, A), homogenates derived from xenografted tumors of mice (B), and clarified extracts derived from human primary NB tumors with different histological grades of differentiation (C) were immunoprecipitated with a mouse anti-GRP75 or a mouse control IgG. The GRP75-bound proteins were analyzed by Western blotting with a goat anti-GRP75, rabbit anti-RARa or rabbit anti-RXRa antibody, respectively. In (C), G, ganglioneuroblastoma; D, differentiated neuroblastoma; U, undifferentiated neuroblastoma. (TIF) Figure S12 RA induces nuclear translocation of GRP75 in SK-N-BE and SK-N-SH cells. SK-N-DZ (A) and SK-N-SH (B) cells were treated with 10 mM RA for 24 h, the nuclear lysates were subject to Western blot analysis. The levels of nuclear GRP75 were normalized with those of histone H1. The normalized level of GRP75 in cells without RA treatment was referred to as one fold of relative nuclear translocation. All quantitative data were calculated as the mean (6SEM) from three independent experiments. (TIF) Figure S13 RA treatments induce the nuclear translocation of GRP75 in a dose-dependent manner. SH-SY5Y cells were treated with various concentrations of RA (0.1, 1, and 10 mM) for 24 h, and the nuclear extracts derived from treated cells were analyzed by immunoblotting with an anti-GRP75 or anti-histone H1 (protein loading control of nuclear fraction) antibody. The levels of nuclear GRP75 were normalized with those of histone H1. The normalized level of GRP75 in cells treated with vehicle alone (0.1% DMSO) was referred to as one fold of nuclear GRP75. All quantitative data were calculated as the mean (6SEM) from three independent experiments. (TIF)

Table S1
The levels of GRP75-bound RARa and RXRa in xenografts in comparison to tumor volume and tumor weight in mice treated with RA or vehicle.

(DOC)
Methods S1 Additional information regarding details of individual experimental procedures. (DOC)