Nuclear export factor 3 regulates localization of small nucleolar RNAs

Small nucleolar RNAs (snoRNAs) guide chemical modifications of ribosomal and small nuclear RNAs, functions that are carried out in the nucleus. Although most snoRNAs reside in the nucleolus, a growing body of evidence indicates that snoRNAs are also present in the cytoplasm and that snoRNAs move between the nucleus and cytoplasm by a mechanism that is regulated by lipotoxic and oxidative stress. Here, in a genome-wide shRNA-based screen, we identified nuclear export factor 3 (NXF3) as a transporter that alters the nucleocytoplasmic distribution of box C/D snoRNAs from the ribosomal protein L13a (Rpl13a) locus. Using RNA-sequencing analysis, we show that NXF3 associates not only with Rpl13a snoRNAs, but also with a broad range of box C/D and box H/ACA snoRNAs. Under homeostatic conditions, gain- or loss-of-function of NXF3, but not related family member NXF1, decreases or increases cytosolic Rpl13a snoRNAs, respectively. Furthermore, treatment with the adenylyl cyclase activator forskolin diminishes cytosolic localization of the Rpl13a snoRNAs through a mechanism that is dependent on NXF3 but not NXF1. Our results provide evidence of a new role for NXF3 in regulating the distribution of snoRNAs between the nuclear and cytoplasmic compartments.

Small nucleolar RNAs (snoRNAs) 2 are a class of non-coding RNAs that range from 60 to 250 nucleotides in length. In mammalian cells, the majority of snoRNAs are encoded within introns and processed from lariats during splicing. snoRNAs assemble into small nucleolar ribonucleoprotein (snoRNP) complexes with RNA-modifying enzymes and traffic to the nucleolus, where their canonical function is to guide sitespecific nucleotide modifications of rRNA and snRNA. Two classes of snoRNAs are defined by conserved sequence motifs, associated proteins, and snoRNA-directed modifications. Box C/D snoRNAs complex with fibrillarin and guide 2Ј-O-methylation of targets, whereas box H/ACA snoRNAs complex with dyskerin and guide pseudouridylation of targets (1)(2)(3). In addition to these canonical roles, there is growing evidence for non-canonical molecular functions of snoRNAs, including miRNA-like post-transcriptional regulation, pseudouridylation of mRNA targets, modification of pre-mRNAs to direct splicing, and regulation of RNA editing (4 -7).
Our laboratory previously identified that box C/D snoRNAs encoded within the introns of the ribosomal protein L13a (Rpl13a) gene are critical mediators of the cellular response to palmitate-induced lipotoxic stress and play a physiological role in tissue responses to oxidative stress (8 -10). Although Rpl13a snoRNAs localize primarily in the nucleus, these and other snoRNAs are readily detected in the cytoplasm under homeostatic conditions (11,12), and they accumulate outside the nucleus during lipotoxic conditions, oxidative stress, and serum starvation (8,10,13). In addition to the pathways for nuclear export of snoRNAs implicated by these findings, studies in Xenopus oocytes provide evidence for an endogenous pathway that transports snoRNAs from the cytoplasm to the nucleus following microinjection (14). Moreover, U8 snoRNA associates with the import factor snurportin 1 in mammalian cells, and knockdown of this transport adaptor is associated with increased cytoplasmic U8 (11). Whether snurportin 1 serves a broader role in transport of the overall class of snoRNAs and whether other transport proteins function to regulate the distribution of snoRNAs between the nucleus and cytoplasm is not known.
We identified nuclear export factor 3 (NXF3) through an shRNA loss-of-function genetic screen in human umbilical vein endothelial cells to identify genes required for the cytotoxic response to metabolic stress. Based on sequence similarity, NXF3 is considered to be a member of a nuclear RNA export factor family that includes the known mRNA exporter, nuclear export factor 1 (NXF1). Like other family members, NXF3 shuttles between the nucleus and cytoplasm and can associate with and transport poly(A) ϩ RNA (15). Given that our genetic screens implicated both snoRNAs and NXF3 in lipotoxic response pathways, we hypothesized that NXF3 may also serve as a transporter for snoRNAs. Herein, we demonstrate that NXF3 regulates distribution of the Rpl13a snoRNAs between the nucleus and the cytosol under homeostatic conditions. This work was supported by National Institutes of Health Grants R01 DK064989, P30 DK020579, and T32 HL007275 and by American Heart Association Grant 15PRE25220014. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Sequence data have been deposited in NCBI's Gene Expression Omnibus (accession number GSE106301). 1  NXF3 associates with a broad array of box C/D and box H/ACA snoRNAs, including the Rpl13a snoRNAs. Treatment of cells with the adenylyl cyclase activator forskolin decreases cytosolic snoRNAs through a mechanism that is dependent on NXF3 expression. Our data provide new insights into the molecular regulation of snoRNA localization.

Knockdown of NXF3 protects against lipotoxic cell death in endothelial cells
To identify genes critical for the cell death response to lipotoxicity, we performed a genome-wide loss-of-function shRNA screen in immortalized human umbilical vein endothelial cells (HUVECs), which are sensitive to lipotoxicity. Cells were transduced with seven pools of lentivirus, each containing ϳ10,000 unique shRNAs targeting expressed human genes and a puromycin selectable marker. As a control, cells were transduced in parallel with a non-silencing shRNA virus that has no homology to known mammalian genes. HUVECs that survived puromycin selection were then treated for 48 h in media supplemented with pathophysiological levels of palmitic acid to model lipotoxic conditions. Compared with cells transduced with the non-silencing control shRNA, cells transduced with an shRNA targeting NXF3 demonstrated less induction of markers of endoplasmic reticulum stress, an early feature of the lipotoxic response (Fig. 1A). Furthermore, cells with the NXF3-targeting shRNA had improved viability under lipotoxic conditions (Fig.  1B). The shRNA targeting NXF3 led to a 72% decrease in NXF3 mRNA levels (Fig. 1C) and a 31% decrease in NXF3 protein expression (Fig. 1D), consistent with knockdown of its gene target in these lipotoxicity-sensitive human cells. NXF3 is a member of the NXF family that includes the general mRNA export receptor, NXF1. Our shRNA library included shRNAs that target NXF1, but cells transduced with NXF1-targeting shRNA or shRNA targeting other known RNA transporters did not survive the palmitate selection. Together these observations implicate a critical role for NXF3, but not NXF1, in the lipotoxic response pathway.

NXF3 knockdown endothelial cells have aberrant Rpl13a snoRNA localization
Although NXF3 can associate with mRNAs, other potential cargo such as non-coding RNAs has not been investigated (16). Given the important role of Rpl13a snoRNAs U32a, U33, U34, and U35a in lipotoxic and oxidative stress-induced cell death and the observation that these snoRNAs accumulate in the cytoplasm under metabolic stress conditions (8,10), we hypothesized that NXF3 knockdown might affect localization of snoRNAs. We assessed Rpl13a snoRNA localization in cytosolic fractions of control-and NXF3-shRNA-transduced cells by isolating cytosolic and non-cytosolic fractions using digitonin extraction. Under normal homeostatic growth conditions, levels of cytosolic Rpl13a snoRNAs in NXF3 shRNA-transduced cells were significantly increased compared with control shRNA-transduced cells, even though total cellular abundance for these snoRNAs was unchanged ( Fig. 2A). Western blottings demonstrated that both nuclear (nucleophosmin) and cytosolic (HSP90, ␣-tubulin) markers segregated as expected in both control and NXF3 shRNA-transduced cells (Fig. 2B). Consistent with prior observations in other cell types that oxidative stress stimulates cytosolic accumulation of Rpl13a snoRNAs (8,17), U32a, U33, and U34 snoRNAs accumulate in the cytoplasm of control shRNA-transduced HUVECs in response to treatment with hydrogen peroxide (Fig. 2C). Hydrogen peroxide had no detectable effect on localization of U35a, which is the least abundant of the Rpl13a snoRNAs and shows less robust cytosolic accumulation in response to other metabolic stress inducers (8). In NXF3 shRNA-transduced cells, hydrogen peroxide did not cause further increase in the cytoplasmic levels of Rpl13a snoRNAs. Elevated cytoplasmic levels of the snoRNAs under homeostatic conditions in NXF3 shRNA-transduced cells are unlikely to be the result of increased basal oxidative stress, because intracellular hydrogen peroxide levels as assessed by dihydroethidium staining were unchanged compared with control cells (Fig. 2D). Nucleophosmin staining showed preservation of discrete nucleolar foci in NXF3 shRNA-transduced cells, in contrast to the diffuse nucleophosmin staining characteristic of nucleolar stress in hydrogen peroxide-treated cells (Fig. 2E), indicating that NXF3 knockdown does not cause nucleolar stress (18). Our findings that loss-of-function of NXF3 perturbs Rpl13a snoRNA subcellular localization indicates that NXF3 functions under homeostatic conditions to maintain the distribution of these non-coding RNAs between the nucleus and cytoplasm. Although NXF3 has been proposed HUVECs were transduced with non-targeting (ctrl) or NXF3-targeting shRNA. A, quantification of ER stress markers (CHOP and ATF4) by RT-qPCR relative to Rplp0 (relative units, RU) following 4 h of incubation in media containing 500 M palmitate. n ϭ 3 independent experiments. B, viability following 48 h of incubation in media containing 500 M palmitate. n ϭ 4. C, RT-qPCR quantification of NXF3 mRNA relative to Rplp0. n ϭ 3. D, representative Western blot analysis of NXF3 and HSP90 expression (top panels) and corresponding Ponceau-stained gel (bottom panel). Graphs report quantification of NXF3 expression relative to HSP90 or total protein from n ϭ 3. Means Ϯ S.D. *, p Ͻ 0.05; **, p Ͻ 0.01 for NXF3 knockdown versus control.

NXF3 regulates localization of snoRNAs
as an mRNA export factor, accumulation of Rpl13a snoRNAs in the cytosol in the setting of NXF3 knockdown implicates an additional role for this protein in nuclear import of these small non-coding RNAs.

NXF3 and NXF1 associate with snoRNAs
Because NXF3 is a member of a nuclear RNA transporter family, and genetic manipulation of NXF3 impacted Rpl13a snoRNA localization, we hypothesized that NXF3 may serve as a snoRNA transporter. To further elucidate the effects of NXF3 on snoRNA localization, we used NIH3T3 murine fibroblasts and H9c2 rat cardiomyoblasts, cell types that are also sensitive to lipotoxocity but grow more rapidly in cell culture and transfect with greater efficiency than HUVECs (8,10,19). To test whether NXF3 physically interacts with the Rpl13a snoRNAs, we transiently transfected NIH3T3 fibroblasts with NXF3-GFP or GFP alone as a negative control. We also transfected cells with NXF1-GFP, another NXF family member with related domain structure (Fig. 3A). We performed immunoprecipitations using a GFP antibody or non-immune IgG as a control. Antibody directed against GFP efficiently and specifically pulled down NXF3-GFP, NXF1-GFP, or GFP, whereas no NXF3-GFP, NXF1-GFP, or GFP was recovered in control immunoprecipitations (Fig. 3B). Real-time PCR quantification revealed a 9 -12-fold increase in Rpl13a snoRNA association with immunoprecipitated NXF3-GFP compared with immunoprecipitated GFP (Fig. 3C). Unexpectedly, the Rpl13a snoR-NAs also associated with NXF1-GFP. These data indicate that under mild detergent conditions chosen to preserve RNAprotein interactions, both NXF3-GFP and NXF1-GFP associate with the Rpl13a snoRNAs. This may reflect an ability of these transporters to bind a broad range of RNA species.
We used RNA-sequencing to determine whether NXF3 associates only with Rpl13a snoRNAs or with snoRNAs more broadly. We collected RNA from cell lysates before immunoprecipitation (input) and RNA from immunoprecipitates (pulldown) of NIH3T3 murine fibroblasts transfected with NXF3-GFP. Because most snoRNAs lack poly(A) tails and do not efficiently prime with random hexamers, they are not well-represented in libraries prepared using approaches suitable for larger RNAs (20). To achieve broad coverage of snoRNAs from both input and pulldown samples, we size-selected RNAs from 30 to 375 nucleotides, a range that would exclude most miRNAs and include most snoRNAs. Libraries were prepared using an Illumina small RNA library preparation kit, sequenced, and aligned to mm9 Refseq Transcripts. We focused our analyses on the non-rRNA sequences, which accounted on average for 55% reads in both input and pulldown samples (Table 1). Greater than 98% of non-rRNA reads in the input and pulldown

NXF3 regulates localization of snoRNAs
samples aligned to snoRNAs, as expected ( Table 2). In addition to the Rpl13a snoRNAs, NXF3 pulldown recovered many other snoRNAs ( Table 3), most of which belonged to the box C/D class of snoRNAs, known to be most abundant (21)(22)(23). Approximately 6% of reads in both input and pulldown aligned to box H/ACA snoRNAs. Because of the methods required to capture snoRNA sequences in library preparation, it was not possible to determine whether NXF3 specifically enriches for snoRNAs over other classes of RNAs. Nonetheless, the lack of significant enrichment of specific snoRNA species in pulldown relative to input snoRNAs indicates that NXF3 broadly associates with this class of small RNAs and does not have specificity for the Rpl13a snoRNAs or other particular species (Table 3).
To validate association of NXF3 with additional box C/D and box H/ACA snoRNAs and to test whether NXF1 also interacted , sequence with similarity to nuclear transport factor 2 (NTF2-like (p15 binding)), ubiquitin-associated-like domain (UBA), nuclear localizing sequence (NLS), nuclear export sequence (NES), and CRM1-binding region. B and C, NIH3T3 fibroblasts were transfected with control plasmid (GFP) or plasmid encoding NXF3-GFP or NXF1-GFP. Immunoprecipitation with ␣-GFP or rabbit IgG (control) was analyzed by Western blot analysis using ␣-GFP (B). Coimmunoprecipitating RNA was quantified by RT-qPCR for Rpl13a snoRNAs and selected box C/D and box H/ACA snoRNAs. Graphs (C) report enrichment in NXF3-GFP-or NXF1-GFP-transfected cells over GFP-transfected cells. Means Ϯ S.D. for n ϭ 3. *, p Ͻ 0.05 for NXF3-GFP-or NXF1-GFP-transfected versus GFP-transfected.

NXF3 regulates localization of snoRNAs
with these snoRNAs, we transfected NIH3T3 fibroblasts with NXF3-GFP, NXF1-GFP, or GFP as control, immunoprecipitated with antibody to GFP, and we performed RT-qPCR on recovered RNA for abundant snoRNAs for which stem-loop primers were successfully designed. Each of the snoRNAs tested showed enrichment in NXF3-GFP pulldown over GFP pulldown (Fig. 3B). Most of these snoRNAs also associated with NXF1-GFP. Together, our data indicate that both NXF3 and NXF1 are capable of associating with a broad range of snoRNAs.

Knockdown and overexpression of NXF3 alters cytosolic Rpl13a snoRNAs in cardiomyoblasts
Although both NXF3 and NXF1 can associate with snoR-NAs, only NXF3 conferred enhanced viability in our lipotoxicity screen in HUVECs. To compare the effects of NXF3 and NXF1 on snoRNA localization, we transfected H9c2 rat cardiomyoblasts with siRNAs to knock down either NXF3 or NXF1. Compared with control non-targeting siRNA, two independent siRNAs directed against NXF3 led to a 93 and 84% decrease in NXF3 mRNA levels, respectively, in cardiomyoblasts. NXF3 siRNAs were specific and did not significantly change levels of family member NXF1 (Fig. 4A, left graph). We found that commercially available antibodies directed against rodent NXF3 are not specific for this family member (data not shown), and our attempts to generate a specific antibody that recognizes the rat or murine protein were unsuccessful, thus precluding confirmation of the extent of knockdown of the endogenous NXF3 protein. Nonetheless, knockdown of NXF3 mRNA led to a 50 -100% increase in cytosolic levels of the Rpl13a snoRNAs (Fig. 4A, right graph) without perturbing efficiency of fractionation (Fig. 4B). In contrast, siRNA knockdown of NXF1 had no effect on cytosolic levels of Rpl13a snoRNAs (Fig. 4, C and D). Thus, accumulation of Rpl13a snoRNAs in the cytosol can result from either acute or chronic knockdown of NXF3. This effect is consistent across different cell types (endothelial cells and cardiomyoblasts) and is not observed with knockdown of NXF1.
Our data are consistent with a model in which NXF3 function is required to maintain low cytoplasmic levels of the

NXF3 regulates localization of snoRNAs
Rpl13a snoRNAs. To test whether NXF3 promotes movement of snoRNAs between the cytosol and nucleus, we overexpressed murine NXF3 by transiently transfecting NIH3T3 murine fibroblasts with a GFP-tagged NXF3 construct (NXF3-GFP), a GFP-tagged NXF1 construct (NXF1-GFP), or a plasmid expressing GFP alone as control. We consistently observed more efficient transfection with GFP (76%) compared with NXF3-GFP (46%) or NXF1-GFP (40%), suggesting that endogenous mechanisms serve to restrict the overall expression of NXF3 and NXF1 (Fig. 5A). Nonetheless, compared with expression of GFP alone, expression of NXF3-GFP decreased cytosolic levels of the Rpl13a snoRNAs by 16 -34%, whereas NXF1-GFP did not significantly change cytosolic levels (Fig. 5, B and C). Together with findings from knockdown of NXF3, our data suggest a model in which NXF3 specifically promotes movement of snoRNAs from the cytoplasm to the nucleus and determines the distribution of the snoRNAs between these cellular compartments.

Regulation of snoRNA trafficking and association with NXF3
Nucleocytoplasmic transport can serve as a point of regulation of subcellular localization for proteins as well as RNAs (24). Cyclic AMP is a potent regulator of protein trafficking, but it is not known whether this second messenger impacts NXF3 or snoRNA localization. We found that treatment of fibroblasts for 1 h with forskolin (FSK), an adenylyl cyclase activator, which increases cyclic AMP levels, significantly decreased cytosolic Rpl13a snoRNA levels compared with DMSO-treated controls (Fig. 6, A and B). To test whether FSK affects NXF3 localization or its association with snoRNAs, we transfected fibroblasts with a plasmid for expression of NXF3-GFP or NXF1-GFP. Similar to our findings with untransfected cells, cells overexpressing NXF3-GFP or NXF1-GFP had decreased cytosolic snoRNA abundance following 1 h of treatment with FSK (Fig. 6, C-F). Moreover, FSK increased association of NXF3-GFP with the Rpl13a snoRNAs (Fig. 7A). Concomitantly, immunofluorescence microscopy revealed increased nuclear localization of NXF3-GFP (Fig. 7B). By contrast FSK did not affect snoRNA association with NXF1-GFP or the cellular distribution of NXF1-GFP, which is predominantly nuclear (Fig. 7, C and D). Together, these observations are consistent with a model of NXF3-mediated trafficking of snoRNAs from the nucleus to the cytoplasm that is regulated by FSK. Moreover, these findings lend further specificity to the function of NXF3 in snoRNA trafficking.
To establish whether the FSK-induced changes in localization of Rpl13a snoRNAs are dependent on NXF3, cardiomyoblasts were transiently transfected with siRNA to knock down NXF3 (Fig. 8A) and then treated with FSK. Although FSK decreased cytosolic Rpl13a snoRNAs as expected in control siRNA-transfected cells, this effect was abrogated when NXF3 was knocked down with either of two independent siRNAs (Fig.  8B). In contrast, NXF1 siRNA knockdown had no effect on the FSK-snoRNA phenotype (Fig. 8B). Taken together, our data are most consistent with a model in which NXF3 plays a critical role in return of snoRNAs from the cytoplasm to the nucleus, a pathway that can be activated by FSK.

Discussion
The endogenous substrates for NXF3, a member of a family of RNA transport proteins, are not well-characterized. Additionally, relatively little is known regarding the regulation of distribution of snoRNAs between the nucleus and cytoplasm. Herein, we demonstrate that gain-and loss-of-function of

NXF3 regulates localization of snoRNAs
NXF3 alters subcellular distribution of snoRNAs encoded within the Rpl13a locus. Furthermore, we show that NXF3 associates with many box C/D and box H/ACA snoRNAs and regulates distribution of these small non-coding RNAs as a class. Our data are most consistent with a model in which snoR-NAs cycle between the nucleus and cytoplasm and NXF3 efficiently imports snoRNAs from the cytoplasm to the nucleus, a function that helps to maintain low levels of snoRNAs in the cytoplasm under basal conditions (Fig. 9). Furthermore, we show that regulation of snoRNA distribution by FSK requires NXF3.
NXF family members have been identified based on sequence similarity and domain architecture that for most includes an amino-terminal containing RNA-binding protein interaction domain, a p15/NXT-heterodimerization domain, and a carboxyl-terminal domain involved in nuclear pore binding (25,26). NXF1, NXF2, NXF3, and NXF5 have been shown to function in nuclear export of mRNAs, whereas NXF7 has been proposed to function in the cytoplasmic trafficking of mRNAs (15,(27)(28)(29)(30)(31). NXF3, which lacks carboxyl-terminal domains that function in nuclear export and nucleoporin binding, has a CRM1-dependent nuclear export sequence that allows it to partner with this transport receptor for its mRNA export function (15). Our study provides evidence that NXF3, but not NXF1, is also involved in transporting snoRNA cargos, expanding the functional repertoire of NXF proteins. Moreover, our data support a model in which NXF3 has a role in returning snoRNAs to the nucleus. Functional domains of NXF3 required for nuclear export of mRNA have been characterized (15,26). However, nuclear localization sequences can be highly variable, and examination of the NXF3 sequence did not reveal motifs predicted with high likelihood to function in nuclear import, raising the possibility that NXF3 could function in a complex with other proteins. Our data provide evidence of a role for NXF3 in snoRNA transport. However, interaction of NXF3 with snoRNAs could be direct or indirect. In contrast to other NXF family members, RNA-binding domains of NXF3 are pre-

NXF3 regulates localization of snoRNAs
dicted but have not been functionally confirmed (32). In future studies, experimental delineation of RNA-binding residues in NXF3 will facilitate structure-function studies to determine whether this protein interacts directly with snoRNAs. In addition, future development of an in vitro nuclear transport assay for snoRNAs will help to test our model.

NXF3 regulates localization of snoRNAs
Following transcription, snoRNAs that are processed from the introns of pre-mRNA or that are generated from independently transcribed units first move to Cajal bodies, where they assemble in to mature snoRNPs, and subsequently to the nucleolus, where they encounter nascent rRNA substrates. PHAX, a known adaptor for RNA export, is required for movement to the Cajal body, and CRM1, a nuclear export receptor, is required for movement to the nucleolus (33). A growing body of evidence indicates that snoRNAs traffic to the cytoplasm and extracellular space (8, 10 -13, 34). Whether snoRNAs undergo processing in the cytoplasm analogous to snRNAs, function outside the nucleus, or are secreted in a regulated fashion remains to be determined. Nonetheless, cells maintain a gradient of snoRNAs such that the vast majority of endogenously expressed snoRNAs remain localized within the nucleus, and when exogenous snoRNAs are introduced into the cytoplasm, they are trafficked into the nucleus through the nuclear pore complex (14,35). To date, PHAX and Snurportin 1 have been shown to function in nucleocytoplasmic trafficking of U8, an independently transcribed, capped box C/D snoRNA (11), but it is not known whether these proteins regulate the cytosolic trafficking of other snoRNAs. Our study provides the first demonstration of a transport protein that regulates cytoplasmic localization of snoRNAs as a class, including both box C/D and box H/ACA species, and it is likely functioning to return cytoplasmic snoRNAs to the nucleus. Inhibition of CRM1 function with leptomycin B decreased, rather than increased, cytosolic snoRNA abundance, 3 suggesting that in contrast to the mRNA export function of NXF3, NXF3-mediated return of snoRNAs to the nucleus may be independent of this transport co-receptor.
Transport of RNAs between the nucleus and cytoplasm is a highly regulated process, and our studies provide new evidence that FSK regulates the association of snoRNAs and the transport protein, NXF3, as well as the trafficking of both snoRNAs and NXF3. Although it is possible that FSK-induced translocation of NXF3 leads to increased association with snoRNAs that are concentrated in the nucleus, our data establish that NXF3 is required for acute regulation of nuclear snoRNA import by FSK. FSK is a natural compound that activates cyclic AMP signaling through both PKA-dependent and PKA-independent mechanisms (36 -39). This could lead to altered post-translational modifications, such as phosphorylation of NXF3 or other proteins that function in snoRNA transport. Although examination of the NXF3 primary sequence suggests a number of potential phosphorylation sites, FSK did not change NXF3 phosphorylation, as determined using a panel of ␣-phosphotyrosine, ␣-phosphoserine, and ␣-phosphothreonine antibodies. 3 It is possible that the available antibodies lack the appropriate specificity or that NXF3 is not a direct target of FSK. Future structure-function and mass spectrometry studies will be necessary to address these possibilities.
Knockdown of NXF3 confers resistance to lipotoxicity, a process dependent on snoRNAs from the Rpl13a locus. These noncoding RNAs accumulate in the cytoplasm early during lipotoxicity (10,17,19). Loss of NXF3 also alters distribution of these snoRNAs at baseline and prevents the acute increase in cytosolic levels with lipotoxicity. The abundance of cytoplasmic Rpl13a snoRNAs in NXF3-knockdown cells is comparable with the levels achieved in control cells following acute lipotoxic exposure. This suggests that the dynamic change in cytosolic snoRNAs, rather than the absolute cytoplasmic levels of these non-coding RNAs, is a key contributor to lipotoxic injury.
Recent genetic studies and identification of diseases associated with dysregulation of snoRNAs have expanded the biological repertoire for snoRNAs beyond their well-appreciated canonical role in modification of nascent rRNAs (9,40,41). Although it is certainly possible that dynamic changes in snoRNA-directed rRNA modifications contribute to these intriguing phenotypes, localization of snoRNAs in the cytoplasm also raises the possibility that non-canonical snoRNA functions are mediated through effects on novel targets outside the nucleolus. A complete understanding of this new snoRNA biology requires further dissection of snoRNA trafficking and its regulation. Our identification of a role for NXF3 in regulated snoRNA trafficking represents a first step in this important direction.

Materials
Digitonin, FSK, phenylmethanesulfonyl fluoride (PMSF), and dimethyl sulfoxide (DMSO) were from Sigma. Protein A Dynabeads and TURBO DNase were from Thermo Fisher Scientific. Doxorubicin (DOX) was from Cayman Chemical. TRIzol LS and SUPERase-IN RNase inhibitor was from Life Technologies, Inc. DTT was from Calbiochem, and ␤-mercaptoethanol was from Sigma. Laemmli Sample Buffer was from Bio-Rad.

Plasmids
The NXF3 cDNA was obtained from the Sun lab (42) and cloned into pEGFP-C1 (Clontech) at BamHI and SalI sites. The full-length cDNA for mouse NXF1 was obtained by PCR from the total RNA of NIH3T3 mouse fibroblasts using primers containing BamHI and SalI restrictions sites. NXF1 cDNA was 3 M. W. Li, and J. E. Schaffer, unpublished data.

NXF3 regulates localization of snoRNAs
cloned into pEGFP-C1 (Clontech) at BamHI and SalI sites. PCR-derived sequences were confirmed by DNA sequencing.

shRNA screen
The shRNA screen was performed using Decode Pooled Lentiviral shRNA Libraries by Thermo Fisher Scientific. Immortalized HUVECs were transduced with lentiviral shRNA pools at a multiplicity of infection of 0.1 and then selected using puromycin. Puromycin-resistant cells were then treated with 500 M palmitate for 48 h. After removing palmitate, cells were grown to form colonies in normal media. Colonies were hand-picked and re-tested for viability following lipotoxic exposure. For this, cells were plated in a 96-well plate in triplicate (1000 cells per well) and treated with 500 M palmitate for 24 h. Cells transduced with non-targeting shRNA viral particles (designed to not target mRNAs in the mammalian genome) were used as negative control. Colonies with significantly greater viability than control cells were selected for further analysis. Genomic DNA was isolated from palmitate-resistant cells, and PCR was used to amplify barcodes corresponding to specific shR-NAs for sequencing. Sequences were compared with the Decode shRNA libraries or analyzed using BLAST to identify the knocked down genes.

Subcellular fractionation
For fractionation by detergent extraction, cell pellets were incubated in digitonin buffer (150 mM NaCl, 50 mM HEPES, pH 7.4, 100 g/ml digitonin, 5 mM EDTA, and 0.1 unit/l SUPERase-In RNase inhibitor) for 10 min and centrifuged for 10 min at 2000 ϫ g to yield a cytoplasmic supernatant and non-cytoplasmic pellet containing membranes and nuclei (44).

RNA isolation and quantitative real-time PCR
Total RNA was isolated from cytoplasmic or non-cytoplasmic fractions and from immunoprecipitates using TRIzol LS according to manufacturer's protocol (Life Technologies, Inc.). cDNA was synthesized with SuperScript III first-strand synthesis system (Invitrogen) using oligo(dT) priming for mRNAs and target-specific stem-loop priming snoRNAs (10). qPCR was performed using PerfeCTa SYBR Green SuperMix (Quantabio). Relative quantitation of target transcript expression was calculated using the ddCT method using Rplp0 as an endogenous control on an ABI 7500 fast real-time PCR system.

Immunoprecipitation
Cells were sonicated in 50 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA, 0.5% Nonidet P-40 containing 1ϫ cOmplete protease inhibitor mixture (Sigma), 1 mM PMSF, and 0.1 unit/l RNase inhibitor SUPERase-In. After treatment with 0.3 units/l TURBO DNase, insoluble material was removed by centrifugation at 20,000 ϫ g for 10 min at 4°C. Lysate was incubated with ␣-GFP (2.5 mg/ml) and protein A Dynabeads for 4 h at 4°C. Complexes were washed four times with lysis buffer. For Western blot analysis, protein was eluted by incubation in Laemmli Sample Buffer, and RNA was eluted by incubation in TRIzol LS.

RNA sequencing
Input and immunoprecipitated RNA from three independent experiments was isolated and concentrated using RNA Clean and Concentrator-5 (Zymo Research). Indexed RNA-sequencing libraries were made using the Illumina TruSeq small RNA kit. Libraries were amplified with 14 cycles of PCR, pooled, and separated by PAGE. Products with sizes 30 -375 nucleotides were excised and collected from the gel. Sequencing was performed using an Illumina HiSeq 2500. Demultiplexed data were analyzed using Partek Flow (build version 6.0.17.0305 Copyright ©; 2017 Partek Inc., St. Louis, MO). Sequences were evaluated for quality and trimmed to remove adaptor sequence. A custom Linux script was used to remove murine ribosomal sequences according to the Illumina iGenome murine genome annotation. Remaining sequences were aligned to the murine genome (mm9 Refseq Transcript August 1, 2016, Bowtie for Illumina, default settings), and aligned reads were quantified, normalized, and annotated. Analysis was limited to genes with Ͼ100 aligned reads. A step-up false discovery rate of 0.05 was considered significant. Sequence data have been deposited in NCBI's Gene Expression Omnibus (accession number GSE106301).

NXF3 regulates localization of snoRNAs Immunofluorescence
Cells were grown on glass coverslips coated with 0.8% gelatin, fixed with 4% paraformaldehyde, permeabilized with Nonidet P-40, and blocked with 200 g/ml ChromPure IgG (Jackson ImmunoResearch) corresponding to secondary antibody species. Detection for NPM used ␣-NPM (Life Technologies, Inc., 325200; 1:1000) with secondary Alexa Fluor 350 donkey anti-mouse IgG (Life Technologies, Inc., A10035; 40 g/ml). GFP transfected cells were grown on glass coverslips and fixed with 4% paraformaldehyde. Coverslips were mounted on microscope slides using SlowFade Antifade reagent (Life Technologies, Inc., S2828). Slides were imaged on a Zeiss Axioskop 2 mot plus microscope with a Zeiss AxioCam MRm camera with a ϫ40 oil immersion objective at room temperature. Images were acquired using AxioVision Rel. 4.8 software. Images were processed identically using the ImageJ software package.

Statistics
Biochemical results are presented as mean Ϯ S.D. for a minimum of three independent experiments. Statistical significance was assessed by two-tailed unpaired t test. p value Ͻ 0.05 was considered significant.