All‐trans retinoic acid (ATRA)‐induced TFEB expression is required for myeloid differentiation in acute promyelocytic leukemia (APL)

In acute promyelocytic leukemia (APL), normal retinoid signaling is disrupted by an abnormal PML‐RARα fusion oncoprotein, leading to a block in cell differentiation. Therapeutic concentrations of all‐trans‐retinoic acid (ATRA) can restore retinoid‐induced transcription and promote degradation of the PML‐RARα protein. Autophagy is a catabolic pathway that utilizes lysosomal machinery to degrade intracellular material and facilitate cellular re‐modeling. Recent studies have identified autophagy as an integral component of ATRA‐induced myeloid differentiation.


| INTRODUC TI ON
Retinoids are signaling molecules related to Vitamin A (retinol) that have well-established roles in development and differentiation. 1 Dietary retinoids are metabolized within cells to a bioactive form, all-trans-retinoic acid (ATRA), which exerts its effects by mediating gene transcription. 1 ATRA binds retinoic acid receptors (RARs) and rexinoid receptors (RXRs). These ligand-dependent transcription factors form RXR-RAR heterodimers that preferentially bind to specific DNA sequence motifs, termed retinoic acid response elements (RAREs), in the promoters or enhancers of target genes. 1 Through interactions with diverse, co-regulatory proteins, the RXR-RAR complexes regulate transcription. 2 Acute promyelocytic leukemia (APL), a distinct subtype of acute myeloid leukemia (AML), is defined by the clonal proliferation of granulocyte precursors halted at the promyelocyte stage of development. 3 Cytogenetically, APL is distinguished by a chromosomal translocation affecting the RARα gene locus on chromosome 17q21 in malignant clones, 4 most commonly generating the functional PML-RARα fusion oncogene. 4 The PML-RARα oncoprotein binds to RAREs on retinoid target gene promoters. Because of its high affinity for co-repressor protein complexes, PML-RARα causes repressive epigenetic changes that inhibit the transcription of RARα target genes, many of which are involved in myeloid differentiation. [5][6][7] While physiologic concentrations of ATRA (~1 nmol/L) are unable to overcome this transcriptional repression, therapeutic concentrations (~1 μmol/L) restore granulocytic differentiation in leukemic cells, allowing them to complete a finite life cycle. 6 The incorporation of ATRA into clinical regimens has transformed outcomes for APL patients, with cure rates of >80% now observed. 8 Recent data show that ATRA has a dual therapeutic effect on the PML-RARα oncoprotein: (a) it induces a conformational change in the oncoprotein that results in the dissociation of co-repressor proteins and allows the preferential binding of transcriptional co-activators, and (b) it promotes the degradation of the PML-RARα oncoprotein through co-operating proteolytic mechanisms. 4,9 This oncoprotein elimination, particularly in leukemia-initiating cells, results in sustained clinical responses. 9 Recently, ATRA-induced differentiation has been shown to involve the induction of autophagy. [10][11][12][13] Autophagy is a constitutive pathway involved in the turnover of damaged or redundant proteins. 14,15 The core machinery of autophagy is largely comprised of "AuTophaGy-related" (ATG) proteins, which through a series of conjugation reactions are sequentially recruited to regulate autophagosome formation and fusion. 16 While initially described as a non-specific degradation pathway, substrate selectivity in autophagy is achieved via adaptor proteins such as sequestosome1 (SQSTM1/p62) and NBR1. 17,18 The regulation of autophagy is complex and occurs at cytoplasmic and transcriptional levels. The serine/threonine kinase mammalian target of rapamycin (mTOR) acts as a master cytoplasmic regulator of autophagy. 18 Further cytoplasmic regulation is provided by adenosine monophosphate-activated protein kinase (AMPK) and myo-inositol-1,4,5-triphosphate (IP 3 ) signaling. 19 The basic helix-loop-helix leucine zipper transcription factor EB (TFEB) promotes the transcription of a gene network that co-ordinates both autophagy and lysosomal biogenesis. 20,21 Under basal conditions, TFEB is phosphorylated by mTOR and retained at the lysosome. 22 Following autophagy activation, TFEB is dephosphorylated and enters the nucleus to exert its transcriptional effects. 21 Notable autophagy-related TFEB targets include LC3B, SQSTM1/p62, and ATG9B, a critical transmembrane protein in autophagosome membrane assembly. 21 It is known that ATRA induces autophagy in human APL cells. [10][11][12][13] This ATRA-induced autophagy is important for the successful granulocytic differentiation of leukemic blasts. 11,12,23 Indeed, autophagy is likely involved in the degradation of the highly aggregate-prone PML-RARα oncoprotein, a process dependent on the SQSTM1/p62 adaptor protein. 12,13 However, as differentiation involves significant protein remodeling, autophagy may also drive leukemic differentiation independently of oncoprotein elimination, as it has does in other hematopoietic cell types. 4,11 Thus, the molecular communication between retinoid signaling pathways and autophagy remains poorly understood and is the focus of this study.
Using genome-wide RNA sequencing technology, we examined the changes that occur in autophagy-related genes during ATRAinduced APL differentiation. We show for the first time that TFEB mRNA is strikingly increased by ATRA. Furthermore, shRNA-mediated knockdown of TFEB in APL cells impeded myeloid differentiation of leukemic cells. Collectively, our data implicate TFEB as an important mediator of ATRA-induced differentiation in APL.

| Cell culture & drug treatments
Cells were cultured as described in Faria et al 24 Valproic acid (1 mmol/L, Sigma-P4543) and arsenic trioxide (0.5 μmol/L, Sigma-A1010) were used. Brefeldin A (Sigma-B7651) was diluted from a 5 mg/mL stock in DMSO to a 10 umol/L working concentration. BFA was added 24 hours prior to analysis. All-trans-retinoic acid (ATRA) (Sigma-R2625) was diluted from a 1 mmol/L stock in 100% ethanol -to 1 µmol/L working concentration. The NB4 and ATRAresistant NB4 lines were described previously. 24 Primers:  G G TCG TG AC T TCC TG AG AC A AT-3′, R :5′-TTGCAAGCCGAATCTGGACTGT-3′; Gene expression amplicons were validated by sequencing. The transcript levels in biological replicates (n = 6) were normalized to hPRT levels, and relative differences calculated. 28 Graphical displays and measurements of statistical significance were performed on GraphPad Prism software.

| Creation of an autophagy/myeloid database
A composite list of 522 autophagy-related human genes was generated from three publicly available autophagy databases: (www.amigo.geneo ntolo gy.org) (Table S1A). We also generated a list of 217 genes associated with myeloid differentiation comprising known hematopoietic lineage markers such as ITGAM and key hematopoiesis-associated genes sourced from the GO:0030099 gene ontology (Table S1B).

cells in WIG format
(the summary output post alignment and MACS peak calling) from the NCBI-GEO database. We visualized RARα and RNA polymerase II binding within the TFEB and p62/sqstm1 loci using Integrated Genome Viewer software (https ://www.broad insti tute.org/igv/home).

| Autophagy
Cells were stained with the Cyto-ID autophagy detection kit (Enzo Life Sciences ENZ-51031-K200) according to the manufacturer's in- structions. An increase in the number of autophagic vesicles is detected as an increase in fluorescence in the 488-2 channel.
All data were collected on the BD LSR II flow cytometer with BD FACS Diva acquisition software. Gating and overlay histograms were generated using Flowjo data analysis software.

| RNA sequencing of ATRA-treated APL cells confirms activation of a myeloid differentiation program
We and others have recently shown that autophagy plays a role in the NB4 model of APL differentiation. 11 To understand the transcriptional signaling pathways involved, we used genome-wide RNA-seq to identify gene expression changes following exposure to therapeutic concentrations of ATRA (1 μmol/L) for 72 hours. As has previously been shown, we morphologically confirmed that ATRA induces differentiation. Untreated NB4 cells have large round nuclei with minimal cytoplasm ( Figure 1A). ATRA-treated cells display increased nuclear lobulation and a decreased nuclear to cytoplasmic (N:C) ratio characteristic of maturing granulocytes ( Figure 1A). We  (Table S2A). Enrichment analysis showed that many differentially expressed genes grouped to gene ontologies comprising genes in the lysosome and immune response (Table S2B).
We examined the transcript changes of 217 genes involved in myeloid differentiation (Table S2C). Cufflinks reads per kilobase of exon per million fragments mapped (RPKM) outputs from our duplicate samples were clustered using Cluster 3.0 software, and graphical outputs visualized as heatmaps using TreeView. A magnified section of our heatmap depicts a cluster of key myeloid genes, including CD11b/ITGAM, CEBPα, CEBPε, and GATA2, with similar patterns of expression ( Figure 1B). Overall, we observed a >2-fold increase in the expression of 36 (bold print) and decreased expression of 12 (grayprint) myeloid differentiation genes (Table S2C).
Among the ATRA-upregulated genes were those encoding the critical myeloid transcription factors PU.1 (SP1 gene) and CEBPε ( Figure 1B). 32 The GATA2 transcription factor preserves a stemlike state in hematopoietic cells and is downregulated during differentiation. 33 We found that ATRA-treated NB4 cells showed decreased GATA2 mRNA levels and we observed a similar decrease in transcripts for MYC, another transcription factor associated with an undifferentiated state ( Figure 1B). 34 We detected a decrease in the transcripts of CEBPα, a gene encoding a transcription factor involved in early myeloid differentiation but downregulated during the terminal phase of the process ( Figure 1C). 35 The mRNA levels of the ITGAM gene encoding CD11b, a frequently used surface marker of myeloid differentiation, were greatly increased ( Figure 1B). This finding was validated by quantitative PCR of ITGAM transcripts in NB4 cells treated with ATRA or vehicle for 72 hours ( Figure 1C).

| RNA sequencing data indicates ATRAmediated autophagy gene transcription
We examined ATRA-regulated expression changes in autophagy-related genes in NB4 cells by cross-referencing our autophagy gene list with those genes that displayed significant (q < .05), >2-fold in mRNA levels in our RNA-seq. This identified 2514 genes differentially expressed genes (DEGs) with ≥2-fold change following ATRA treatment in NB4 cells. Of the 522 members of the composite autophagy-related genes, 84 were also differentially regulated in ATRA-treated NB4 cells (Table S2D), consistent with reports linking autophagy and retinoid-induced differentiation. 11 Among the mRNAs increased was transcription factor EB (TFEB) (+11.5-fold change, q = .00022), a master transcriptional coordinator of autophagy and lysosomal biogenesis. 36 To assess whether TFEB, a critical regulator of autophagy, is itself under direct retinoid receptor regulation, we interrogated two independent, publicly available RARα AML ChIP-seq datasets Furthermore, SQSTM1/p62, a known transcriptional target of TFEB, was also induced by 6-fold (P < .0001; Figure 2B,i,ii). 37 This was comparable to a 5.9-fold induction calculated by RNA-seq analysis ( Table   F I (Table S2A), and we measured a comparable 2.5-fold induction by qPCR (P < .0001; Figure 2B,iii).
We then examined whether these genes were induced by ATRA in a differentiation-resistant derivative of NB4 APL cells, NB4R. We observed an ATRA induction of TFEB, but to a lesser degree than that seen in the ATRA-sensitive NB4 cell line (2.5fold increase) ( Figure 2C,i). Similarly, the induction of SQSTM1/ p62 and SIK3 was impaired in the differentiation-resistant line ( Figure 2C,ii,iii), suggesting a role for these genes in leukemic cell differentiation.
We then assessed whether elevation of these autophagy-associated genes could be detected in APL patient cells, treated ex vivo with ATRA. Mononuclear cells separated from the bone marrow of a newly diagnosed APL patient were cultured ex vivo ± ATRA for 72 hours.
TFEB, SQSTM1/p62, and SIK3 were induced to levels comparable to those in NB4 ( Figure 2D,i-iii). This induction coincided with increased mRNA levels of CD11b and PIK3CD ( Figure 2D,iv-v), two genes induced during myeloid differentiation of APL cells (as in Figure 1C,E), confirming ex vivo differentiation of these primary patient leukemic cells.
We also analyzed expression of these autophagy-associated genes in cDNA from peripheral blood mononuclear cells (PBMCs) isolated from patients with solid urological tumors that were treated with liposomal ATRA as part of a phase 1/2 clinical trial in 2007. 38 We detected a 2-fold induction of TFEB by ATRA in PBMCs from this non-leukemic patient cohort ( Figure S1).

ATRA-mediated gene expression in NB4
We  Figure 3B, Table S3C). Another way to state these results is that the comparison of the shScr ± ATRA provides the number of genes affected by ATRA, 3916, and the comparison of the shScr ± ATRA and the shTFEB + ATRA shows that 1200 genes, a large number, are regulated by ATRA in the shScr but not in the shTFEB cells, and thus the expression of these 1200 genes either directly or indirectly requires expression of TFEB and includes no-

| TFEB regulates CLEAR network gene expression during ATRA-induced differentiation of APL cells
TFEB is a master regulator of a gene network that controls both autophagy and lysosomal biogenesis, known as the Coordinated Lysosomal Expression and Regulation (CLEAR) network. 39 This includes a network of 471 genes reported to be direct transcriptional targets of TFEB. 21 We examined TFEB expression and the TFEB-regulated CLEAR genes in our dataset, comparing vehicle and ATRA-treated shScr and shTFEB cells (Table S4A). Using hierarchical clustering, we visualized the expression of a subset of CLEAR network genes in ATRAtreated shScr and ATRA-treated shTFEB cells ( Figure 3C). This analysis of the gene cluster, which included TFEB, identified a sub-cluster of genes, including TFEB, GABARAP, and p62/SQSTM1. That expression of p62/SQSTM1, GABARAP, and TFEB clusters together is significant and suggests potential co-regulation of these genes ( Figure 3D). We identi- differentiation. 12 Other known ATRA-regulated autophagy genes include granulin, GRN, a gene encoding a glycoprotein with a role in inflammation that is highly expressed in myeloid cells ( Figure 3C, highlighted). 40

| TFEB depletion reduces ATRA-mediated myeloid differentiation of APL cells
We next evaluated whether TFEB expression was required for ATRA-induced the expression of the myeloid differentiation signature in APL cells by comparing vehicle and ATRA-treated control and TFEB-depleted cells (Table S4B). Cufflinks RPKM values for these genes in shScr-Ctrl, shScr-ATRA, shTFEB-Ctrl, and shTFEB-ATRA cells were visualized using TreeView heatmaps ( Figure 4A- Table S4B). From our panel of 217 myeloid genes, we identified clustered subnetworks of genes, suggesting potential co-regulation that displayed ≥1.5-fold increases in transcript levels during ATRA treatment of non-transfected NB4 cells (Table S2C). We used a 1.5fold change cutoff for Figure 4A-B to detect statistically significant We also validated expression of these differentiation mark-  Table S4B). Conversely, the ATRAassociated reduction in GATA2 was greater in shScr (−10-fold) compared to the shTFEB cells (−5.8-fold) ( Figure 4B, Table S4B). 33 Depletion of TFEB also impaired ATRA-induced reduction of

| Effect of TFEB knockdown on autophagy
Our data thus far have suggested that ATRA-induced TFEB plays an important role in leukemic cell differentiation. We next evaluated whether the effects of TFEB were related to its regulation of autophagy. To do this, we initially compared autophagic flux in the shScr and shTFEB clones. Autophagic flux refers to the entire autophagy process, from sequestration to degradation. Autophagosome accumulation or an increase in autophagy markers may be a consequence of either increased autophagy initiation or a block in autophagosome turnover. Therefore, to differentiate between induction of autophagosomes and failure of turnover, cells were pretreated with chloroquine to block lysosome function and autophagosome turnover. Any autophagosome accumulation beyond that observed with chloroquine alone is then attributed to enhanced autophagy initiation. We

| Evidence for alternative autophagy
The previous LC3II blot indicated higher basal levels of LC3 II protein in the shTFEB clone ( Figure 5). Therefore, despite TFEB knockdown, there are still autophagosomes present, which also accumulate when the cells are treated with chloroquine. TFEB is only one of several transcription factors known to influence autophagy and alternative mechanisms of autophagy have been described, which do not need all of the components of autophagy initiation complexes. We therefore investigated whether there might be a TFEB-independent/ alternative autophagy present in these cells. One of these pathways is thought to originate at the Golgi and is disrupted by Brefeldin A (BFA), a golgi inhibitor. 42 We therefore examined whether BFAsensitive autophagy might contribute to the autophagy present during differentiation.

| Effect of BFA on differentiation of shScr and shTFEB cells
To assess the involvement of BFA-dependent autophagy in differentiation, we examined the effect of BFA on CD11b expression.
Triplicate data were combined as mean fluorescence intensities, with each control normalized to one ( Figure S4,iii). These data again highlight the effect of TFEB silencing on ATRA-induced differentiation, an effect that is further enhanced by BFA treatment.
In this experiment, we cannot distinguish between a reduction in surface CD11b due to inhibition of alternative autophagy or due to impaired Golgi trafficking. Future studies may wish to address this as soon as more selective inhibitors of alternative autophagy pathways become available.

| D ISCUSS I ON
We and others have reported that ATRA induces autophagy in APL cells and that this autophagy contributes to granulocytic differentiation. [11][12][13]23 The mechanisms through which ATRA activates autophagy, however, have not been well characterized. Autophagy is mTOR-dependent, as mTOR inhibition with rapamycin increases autophagy and promotes PML-RAR degradation. 13 PU.1 binds at the promoter of the microtubule-associated protein 1S (MAP1S) gene, a positive regulator of autophagy, and MAP1S inhibition interferes with APL cell differentiation. 43 As PU.1 expression is suppressed in APL and restored by therapeutic levels of ATRA, this is a potential indirect mechanism of autophagy induction. 44 ATRA may also have a role in autophagosome maturation through redistribution of a cation-dependent mannose-6-phosphate receptor to the developing autophagosome, leading to vesicle acidification. 45 Building on our previous work, we assessed the transcriptional effects of ATRA on autophagy-related genes (Table S3B).
We found altered transcript levels of >80 autophagy-related genes following ATRA addition, opening avenues for further study of the cross talk between retinoid signaling and autophagy. From this gene list, we focused on the master autophagy regulator TFEB as a potential critical communicator between retinoid signaling and autophagy. Although previous studies have used microarray technology (eg, GSE19201) to assess gene expression changes in APL cells following ATRA treatment, this is the first study to use unbiased, genome-wide approaches to examine the role of TFEB in ATRA-induced differentiation.
Two independent, published ChIP-seq analyses 7,31 show RARα and RNApol II recruitment to the TFEB locus, indicating that TFEB may be a direct retinoid receptor target (Figure 2A). We show that TFEB expression increases in NB4 cells treated with ATRA ( Figure 2B). We also show TFEB induction in cultured primary human APL cells and in non-myeloid primary cells treated with ATRA ( Figure 2D). TFEB expression is induced in murine hepatocytes by overexpression of cyclic AMP response element-binding protein (CREB) and its partner, CRTC2. 46  impaired ATRA induction of the key autophagy-associated genes p62/SQSTM1, GABARAP, and ATG16L1 ( Figure 3D,E,F). Our data suggest that TFEB is a key mediator of ATRA-induced autophagy processes during myeloid differentiation.
The large fold increase in TFEB mRNA during ATRA-mediated APL differentiation, along with our data that this induction is at- Indeed, we confirm that TFEB knockdown alters the expression of downstream CLEAR network genes ( Figure 3). As a protein degradation pathway, autophagy has been implicated in the degradation of the aggregate-prone PML-RARα oncoprotein, a process necessary for sustained therapeutic remissions in APL. 9,12,13 We previously showed that promoting autophagy can potentiate ATRA-mediated differentiation of the HL60-Diff-R cell line, a human AML line that does not harbor the PML-RARα oncoprotein and does not differentiate in response to ATRA treatment alone. 11 This finding suggests that autophagy plays a role in the cellular remodeling that occurs during differentiation and is not solely involved in oncoprotein elimination. Thus, TFEB induction may be critical in managing the considerable proteostatic stress of a differentiating cell.
Our analysis of LC3 II expression and autophagosome accumulation indicates an impairment in autophagic flux in shTFEB cells.
This impairment in autophagy may be a key factor in the impairment of differentiation in shTFEB cells. The importance of autophagy in these cells is underscored by the fact that shTFEB cells can utilize an additional TFEB-independent pathway that can be inhibited by brefeldin A. This pathway is not just upregulated as a compensation in the shTFEB cells-it is also present in the shScr control cells-suggesting co-existence of autophagy pathways. This alternative pathway has previously been described in the context of leukemia cells. Wang et al reported an alternative autophagic mechanism, in canonical autophagy-defective leukemia cells (Atg7-

deleted K562 cells)-which was reversed by brefeldin A. 50
This is the first study to indicate a potential role for alternative autophagy in differentiation. It is imperative to better understand the contribution of various autophagy pathways in order to effectively modulate it for the purposes of enhancing differentiation.
It is important to acknowledge that TFEB is a transcription factor that targets E-box (CANNTG) sequences on DNA and regulates the transcription of genes outside the CLEAR network. TFEB downstream-target genes involved in diverse cellular processes, including inflammation, metabolism, and cell death, have been reported. 39 The effects of TFEB on differentiation could also be mediated by these genes, and indeed, a wider range of TFEB target genes may be identified in different cell contexts.
In conclusion, this study explores the communication between retinoid signaling and autophagy during APL cell differentiation. We identify TFEB, a master transcriptional regulator of autophagy and lysosomal biogenesis, as a direct retinoid target that is induced during ATRAmediated differentiation. Importantly, we show that TFEB and autophagy play a critical role in the myeloid differentiation process. Induction of TFEB may be of therapeutic benefit in broadening the use of differentiation therapy in ATRA-resistant, acute leukemias. The prospect of stimulating TFEB expression for therapeutic gain is an enticing one.

ACK N OWLED G EM ENTS
We thank the McKenna, Gudas, and Mongan laboratories for sug-

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests.