A NR2E1-Interacting Peptide of LSD1 Inhibits the Proliferation of Brain Tumor Initiating Cells

The recurrence of malignant brain tumor, like glioblastoma, is often due to the existence of brain tumor initiating cells (BTICs) with stem cell properties. As the mortality ratio of brain tumor relapsed patients is very high and lack of ecient therapies, there is an urgent need to develop novel therapeutic methods targeting BTICs. NR2E1 (TLX), an orphan nuclear receptor, is critical for the self-renewal of BTICs. In this study, we found that NR2E1 recruits LSD1, a lysine demethylase, to demethylate mono- and di-methylated histone 3 Lys4 (H3K4me/me2) at the Pten promoters and repress its expression, thereby promoting BTIC proliferation. Using Amide Hydrogen/Deuterium Exchange and Mass Spectrometry (HDX-MS) method, we identied four LSD1 peptides that may interact with NR2E1. One of the peptides that locates at the LSD1 SWIRM domain strongly inhibited BTIC proliferation by promoting Pten expression through interfering NR2E1 and LSD1 function. Furthermore, overexpression of this peptide in human BTIC can inhibit its formation of brain tumor. Hence, this peptide exhibits an interesting potential for therapeutic intervention in malignant brain tumors in future.


Introduction
Malignant brain cancers, like glioblastoma (GBM), are highly heterogeneous and aggressive. They are resistant to chemotherapy and radiation therapy and show a high chance of relapse. The patient survival rate is only about 15 months after diagnosis. This situation has lasted over the past decades although multiple novel therapeutic means have been employed (Houshyari et al., 2015, Mrugala, 2013. It has been suggested that cancer initiating cells (CICs) with stem cell properties underlie the heterogeneity of malignant tumors (Zhu et al., 2014). They are less differentiated and resistant to chemotherapy and radiation treatment. They are thought to be the "root" of tumor occurrence and are responsible for the growth and relapse of tumors. Targeting CICs to treat cancer may help to improve the outcome of clinical therapies (Lin et al., 2013).
Researches have revealed that the recurrence of high-grade gliomas is due to the existence of brain tumor initiating cells (BTICs). BTICs were among the rst CICs derived from a solid tumor (Singh et al., 2003, Singh et al., 2004. BTICs express the neural stem cell surface marker CD133. And as little as 100 BTICs could initiate phenocopies of the original tumors in a NOD.CB17-Prkdcscid/J (NOD SCID) mouse brain (Singh et al., 2004). BTICs and neural stem cells (NSCs) share several similarities, and it has been suggested that BTICs hijack the self-renewal mechanisms of NSCs to support their proliferation. Many studies have shown that the factors important for NSC maintenance also play important roles in brain tumorigenesis. For example, Nestin, which labels NSCs in adult mouse brain, also marks BTICs in glioblastoma and is required for the long-term sustenance of tumor growth . NR2E1(TLX), an orphan nuclear receptor, is essential for the self-renewal of BTICs. NR2E1-positive glioma cells can initiate brain tumors and form spheres in suspension culture (Zhu et al., 2014). Depletion of NR2E1 in mouse primary tumors signi cantly extended animal survival time. Interestingly, GBM patients express a high level of NR2E1 which is correlated with poor survival time (Zhu et al., 2014). NR2E1 may therefore be a valuable target for brain tumor therapy. Just like Nestin, NR2E1 is highly expressed at the hippocampal dentate gyrus and the subventricular zone. It is required for the maintenance and selfrenewal of neural stem cells (Shi et al., 2004a). In NSCs, NR2E1 interacts with LSD1, a histone demethylase, and recruits it to the promoter of Pten. LSD1 then demethylates mono-and di-methylated histone 3 Lys 4 (H3K4) and removes these active epigenetic markers from the regions to silence the expression of Pten, a gene that induces apoptosis, regulates the cell cycle and functions as tumor repressors. Through coordinated repression of Pten, NR2E1 and LSD1 contribute to the proliferation of NSCs and retinoblastoma cells (Shi et al., 2004a, Yokoyama et al., 2008. Pten is an important misregulated tumor suppressor gene in almost all types of cancers. It is now an open and interesting question whether a similar mechanism is also employed in BTICs. Lysine-speci c histone demethylase LSD1 (also named AOF2 or KDM1A or BHC110) is a FAD dependent lysine demethylase. LSD1 can demethylate mono-and di-methylated H3K4 in a complex with CoREST, but shifts its targets to mono-and di-methylated H3K9 when it partners with the androgen receptor (AR) (Lee et al., 2006, Metzger et al., 2005, Shi et al., 2004b, Shi et al., 2005. Thus, by changing partners, LSD1 is involved in both gene activation and gene repression. The N-terminus of LSD1 is a non-structural element and contains a putative nuclear localization signal. Following this region is the Swip3p/Rsc8p/Moira (SWIRM) domain. After the SWIRM domain is an oxidase domain which is involved in demethylating proteins. LSD1 is also linked to the growth of glioblastoma and its inhibition increases the sensitivity of glioblastoma cells to histone deacetylase (HDAC) inhibitor treatment (Singh et al., 2011). Since NR2E1 and LSD1 both play important roles in glioblastoma, we set out to investigate whether BTICs employ the same NR2E1-LSD1 mechanism, as in NSCs, to regulate BTIC proliferation.

NR2E1 and LSD1 are essential for the proliferation of BTICs
To study the role of NR2E1 and LSD1 in BTICs, two BTIC lines, BTIC-1 and BTIC-2 derived from Nestin-TVa mice were employed for the following experiments. These BTICs can e ciently grow mouse brain tumors after transplantation in recipient mice (Zhu et al., 2014). Like neural stem cells (NSCs), BTICs could be maintained in monolayer or non-adherent suspension culture (Figure 1a) (Zhu et al., 2014). The capacity of BTICs to form tumor spheres suggests their stem cell properties. NR2E1 and LSD1 are highly expressed in NSCs and promote the self-renewal of NSCs (Sun et al., 2010). Real-time PCR and western blot assays showed that both NR2E1 and LSD1 were more highly expressed in BTICs than NSCs ( Figure  1b and 1c). We therefore set to examine whether NR2E1 and LSD1 coordinate with each other to regulate the proliferation of BTICs. To knock down Nr2e1 and Lsd1, we designed two different shRNAs for each gene and cloned them into pSuper-puro vector. We transfected these shRNAs in to BTICs respectively by electroporation and selected the transfected cells with puromycin (Sun et al., 2010). Three days after puromycin selection, Nr2e1 and Lsd1 shRNA knockdown led to a signi cantly lower amount of BTICs in culture (Figure 1d and Supplementary Figure 1a). Real-time PCR revealed that the expression of Nr2e1 and Lsd1 was downregulated to about 20% to 40% of their original levels by their respective shRNAs (Figure 1e and Supplementary Figure 1b). We then examined cell viability by the MTT cell proliferation assay, which revealed that the knockdown of Nr2e1 and Lsd1 resulted in reduced cell viability compared to the control knockdown BTICs (Figure 1f and Supplementary Figure 1c). These results suggest that Nr2e1 and Lsd1 are required for the proliferation of BTICs. NR2E1 and LSD1 synergistically repress PTEN to promote BTIC proliferation To examine whether NR2E1 and LSD1 play the same role in BTICs as in NSCs, we performed coimmunoprecipitation assay using whole cell lysate. An anti-NR2E1 antibody could pull down endogenous LSD1, but control IgG could not, suggesting that NR2E1 and LSD1 form a complex in BTICs (Figure 2a). To investigate whether NR2E1 and LSD1 also regulate Pten in BTICs, we knocked down the expression of Nr2e1 and Lsd1 in BTICs by shRNAs. The downregulation of NR2E1 and LSD1 led to the upregulation of PTEN at both the mRNA and protein levels ( Figure 2b  To further examine whether BTICs employed the same NR2E1-LSD1 regulatory mechanism as NSCs, we performed chromatin immunoprecipitation (ChIP) assay to investigate NR2E1 and LSD1 binding pro le. The ChIP assay revealed that NR2E1 and LSD1 both bind to the promoter of Pten (Figure 2g). To examine whether NR2E1 and LSD1 are functional at the promoter of Pten, we performed ChIP assay with antibodies against H3K4me1 and H3K4me2 using chromatin extracted from Nr2e1 or Lsd1 knockdown BTICs. It turned out that both downregulation of NR2E1 and LSD1 led to upregulated enrichment of H3K4me1 and H3K4me2 at the Pten promoter, suggesting that NR2E1 and LSD1 indeed directly repress Pten in BTICs by demethylating H3K4me and H3K4me2 at its promoter ( Figure 2h).

Prediction of Lsd1 peptides involved in the NR2E1-LSD1 interaction
To understand how NR2E1 and LSD1 synergistically function, we employed Amide Hydrogen/Deuterium Exchange and Mass Spectrometry (HDX-MS) to investigate the interaction between NR2E1 and LSD1. A total of 56 pepsin digested fragments covering about 80% of the LSD1 primary sequence were identi ed and analyzed. The difference in deuterium uptake for all the fragments between LSD1 alone and LSD1:NR2E1 complex was measured at 30 seconds, 1, 2, 5 and 10 minutes. As the difference in deuterium exchange was maximum at 1 minute, the deuterium uptake for each peptide in the 1 minute samples was used to monitor the effects of NR2E1 binding with LSD1. A number of regions in LSD1 showed decreased exchange upon interactions with NR2E1, but the maximum difference occurred at the deuterons. Mass spectral isotope envelopes for four peptides, 197-211, 354-377, 481-501 and 537-546, from 1 minute HDX samples showed the most signi cant difference between LSD1 alone and NR2E1-LSD1 complex after deuterium uptake, suggesting these LSD1 peptides may be involved in forming a complex with NR2E1 ( Figure 3b). The difference in deuterium uptake for each peptide was calculated and the results from 1 minute samples were mapped onto the crystal structure of LSD1 (PDB ID: 2Z3Y) (Figure 3c).
We next investigated the function of overexpression of these peptides in BTICs. We cloned the four peptides into the pCAG-puro plasmids and expressed them as Flag-tag peptides. Twelve hours after transfection, puromycin was added to select the transfected cells. At day 3 after puromycin selection, cells were harvested to check protein expression. Western blot with an anti-Flag antibody con rmed that all four Flag-tagged LSD1 peptides were expressed. Interestingly, the level of PTEN was increased in Flag-LSD1-197-211 overexpressed BTICs compared to GFP overexpressed BTICs (Figure 4b). Immunostaining BTICs at this stage with an anti-Ki67 antibody revealed that overexpression of LSD1-197-211 led to fewer Ki67 positive cells (Figure 4c). Further MTT assay con rmed that LSD1-197-211 transfected BTICs showed the most drastic reduction of viable cells, while LSD1-354-377 only slightly decreased BTICs and both LSD1-481-501 and LSD1-537-546 showed almost no effect ( Figure  Although LSD1-197-211 could interfere with the synergistic function of NR2E1 and LSD1 and inhibit the proliferation of BTICs, the speci city of this peptide is unclear. Both NR2E1 and LSD1 are highly expressed in 293T cells (Figure 5a). Knockdown of Nr2e1 by shRNA did not, however, lead to the upregulation of Pten at the mRNA level, suggesting that the NR2E1-LSD1 mechanism is not involved in the proliferation of 293T cells (Figure 5b). To test whether LSD1-197-211 had any effect on the cells that do not rely on the NR2E1-LSD1 based cell proliferation, we overexpressed GFP and LSD1 peptides in 293T cells separately with puromycin selection for three days. Western blot showed that the PTEN protein level was similar in the GFP and LSD1 peptide overexpressed BTICs (Figure 5c). MTT assay was further performed to check the cell viability. Apart from LSD1-354-377, which exhibited slight inhibition of 293T cell growth, other peptides showed no obvious inhibitory effect (Figure 5d). This result suggests that the LSD1-197-211 peptide shows relatively speci c inhibition on BTICs.
To further characterize the speci city of LSD1-197-211, we determined the crystal structure of the human LSD1 SWIRM domain, residues 183-267 (Supp. Table 1 LSD1 197-211 inhibits the brain tumor formation of BTIC Human and mouse NR2E1 protein sequences share more than 97% similarity so does LSD1. In addition, human and mouse LSD1-197-211 peptides are exactly the same. Therefore, we deduced that LSD1-197-211 peptide should be able to repress human BTICs (hBTICs) as well. To test this hypothesis, we derived two hBTIC lines from glioblastoma patients (Supplementary Figure 6a) and generated doxycycline inducible LSD1-197-211-GFP and control GFP transduced hBTICs by abovementioned lentivirus system. Next, we tested the effect of LSD1-197-211 on tumor sphere formation by adding doxcycycline to the culture medium. 3 days after doxycycline induction, LSD1-197-211-GFP transduced hBTICs formed slightly smaller tumor sphere than GFP transduced hBTICs. GFP signal in LSD1-197-211-GFP transduced hBTICs was also weaker than GFP transduced hBTICs. 6 days later, the difference was more drastic (Figure 6a and Supplementary Figure 6b). Not only the number of tumor spheres formed by LSD1-197-211-GFP transduced hBTICs was less than GFP transduced hBTICs, but also the average tumor sphere size of LSD1-197-211-GFP transduced hBTICs was smaller than GFP transduced hBTICs (Figure 6b). These results con rmed that LSD1-197-211 inhibits hBTICs.
To further investigate the function of LSD1-197-211 in vivo, we transplanted the transduced hBTICs into the brains of nude mice. The next day after the transplantation, we started to feed the mice with water containing doxycycline to induce transduced gene expression. Mice transplanted with GFP transduced hBTICs showed obvious brain tumor growth as revealed by GFP positive signals, while almost no GFP positive signals were detected in mice transplanted with LSD1-197-211-GFP transduced hBTICs ( Figure  6c

Discussion
High-grade glioma, including glioblastoma, is the most common primary malignant brain tumor. The general treatment for high-grade glioma includes surgery, radiotherapy and chemotherapy. However, it is virtually impossible to completely resect these in ltrative tumors and concurrent radiotherapy and chemotherapy do not provide any signi cant survival bene t for patients. Five-year survival ratio of patient is still less than 5%. Therefore, novel treatment strategies are desperately needed for this grim disease.
Past clinic research shows that drugs that target epigenetic modi ers yield promising survival bene ts in multiple diseases. However, no therapeutics targeting LSD1 have been developed at present. One of the reasons is that LSD1 is broadly expressed in mammalian tissues, in particular stem cells. Inhibition of LSD1 by AO inhibitors or depletion of LSD1 might therefore cause a signi cant disturbance of its normal physiological function, leading to unwanted side effects. All reported LSD1 inhibitors that bind to the FAD or AO domains are far from ideal, either because of poor selectivity or their polycationic nature (Lee et al., 2006, Schmidt and McCafferty, 2007, Willmann et al., 2012. Potential short peptides that could compete with natural histone H3 substrates or bind at sites beyond the active site of the AO domain by an allosteric mechanism to prevent LSD1 from forming complexes or binding to the nucleosomes are being actively explored (Baron and Vellore, 2012, Shi et al., 2005, Tortorici et al., 2013. Nevertheless, targeting the general LSD1 repressive complex, CoREST/LSD1 or LSD1 binding to the histone tail would still disturb the general function of LSD1 in cells. To achieve more effective and speci c treatment of highgrade glioma, targeting the LSD1 involved speci c cell proliferation mechanism is more appropriate, with minimum side effects. However, since the NR2E1 and LSD1 mechanism is also employed by NSCs for self-renewal, we anticipate that LSD1-197-211 may also inhibit NSC proliferation. If this is the case, neurogenesis disturbance that might result from the use of this peptide for brain glioma treatment would be a concern. It is known that neurogenesis is most active in the fetus and reduces with aging. Elderly people, in whom neurogenesis is very low, have the highest probability of developing high-grade brain glioma. LSD1-197-211 may therefore hold great therapeutic potential for patients of this age group. Besides, research has revealed that the regulatory factors of the self-renewal of NSCs in vitro do not always affect neurogenesis in vivo. For example, mice with mutated inhibitors of DNA binding 1 (Id1), a factor that is required for the self-renewal of NSCs in vitro, show normal neurogenesis and NSC population in vivo (Zhu et al., 2014).
Over all, our study revealed that Lsd1-197-211 may serve as a leading peptide for peptide drug development for glioblastoma. However, further investigation on the peptide delivery, safety and optimization will be needed to bring the discovery closer to application.

Cell culture
Brain tumor initiating cells (BTICs) derived from Nestin-TV-a mice were received as a gift from Dr. Haikun Liu's lab (Zhu et al., 2014). To culture BTICs in monolayer, the cell culture plate was coated with laminin and poly-L-lysine. Cells were grown in DMEM/F12 medium plus 20 ng/ml epidermal growth factor (EGF), 10 ng/ml broblast growth factor (FGF-2), B27 and insulin-transferrin-selenium supplements (ITSS). Cells were digested with accutase for passage. To culture brain tumor initiating cells in sphere, cells were seeded in a low attached cell culture plate (Corning) in the above medium without any coating.

Reverse transcription and real-time PCR
Reverse transcription was performed with 2 µg of total RNA using the PrimeScript RT reagent kit (Takara).
Real-time PCR analysis was performed by using the ABI Prism 7900HT machine (Applied Biosystems) with the SYBR Green mixture (Takara). For each primer, only one correct size band was formed. All experiments were repeated at least three times independently. The nal results were normalized against the expression of β-ACTIN or GAPDH.
Cell growth assay 10 6 BTICs were transfected with 2 ug plasmids of interests by electroporation with Amaxa cell line Nucleofector Kit V using nucleofector II. The transfected cells were seeded on a poly-L-lysine and laminin treated 6-well plate. After 24 hours, the transfected cells were selected with 1ug/ml puromycin. After 3 days, the oating dead cells were washed away and the viability of cells was quantitated using MTT assay or CCK-8 assay kit by following manufacturer's protocol.

Chromatin immunoprecipitation
ChIP assay was carried out as described previously with slight modi cation (Li et al., 2013b). Brie y, cells were xed with 1% (w/v) formaldehyde for 10 minutes at room temperature, and 125 mM glycine was used to inactivate formaldehyde. Chromatins were sonicated to generate average fragment sizes from 200 to 600 bp and immunoprecipitated using the anti-NR2E1 (a gift from Liu's lab), anti-Lsd1 (ab17721), anti-H3K4me1 (ab8895), anti-H3K4me2 (ab7766) antibodies and control IgG or control GFP. The ChIP enriched DNA and input were then decrosslinked and proteins were digested by proteinase K. DNAs were puri ed by phenol:chloroform extractions and followed by ChIP-qPCR analysis using the ABI PRISM 7900 sequence detection system and Kappa SYBR green master mix (Takara). The values of each real-time PCR assay were normalized with its own input value and then compared with the IgG or GFP value to get the enrichment fold. PCR primers were designed to amplify the promoter regions of mouse Pten and control according to previous research (Sun et al., 2010). Each experiment was performed three times independently.

Immuno uorescence staining
The cells were xed were 4% paraformaldehyde (PFA) for 30 minutes at 4 degree, and then washed in cold PBS for 5 minutes 3 times. The nuclei were subsequently permeabilized with PBS with 0.5% Triton X-100 for 30 minutes. Next the cells were blocked with 1% BSA in PBS for 1 hour. The cells were incubated with primary antibody overnight at 4°C. Primary antibodies used were rabbit anti-cleaved caspase 3 antibody (9664S, CST) and rabbit anti-Ki67 (ab15580, Abcam). After wash, the cells were incubated with anti-rabbit secondary antibody conjugated with proper Alexa Fluor label for 1 hour at room temperature in darkness. The nuclei were counterstained with DAPI. The cells were imaged with Olympus IX-73 immuno uorescence microscope.

Co-immunoprecipitation
Immunoprecipitation assays were performed with whole-cell lysates from BTICs or targeted cells transfected with overexpression plasmids. Anti-NR2E1 (ab30942), anti-LSD1(ab17721), anti-FLAG (SIGMA #F 1804) and anti-HA (ab18181) antibodies were used to pull down protein complexes. Immunoprecipitated complexes, bound to the corresponding antibody, were washed extensively with 0.1% Triton X-100 buffer (50 mM Tris-HCl at pH 8, 150 mM NaCl, 1 mM EDTA, 0.1% Triton X-100, 10% glycerol plus Roche protease inhibitor cocktail). The interacting protein bands were resolved with 10% SDS-PAGE gel and transferred to the PVDF membrane, followed by detection with an appropriate primary antibody, an HRP-conjugated second antibody, and an ECL detection reagent.

Animals
The animal experiments were performed under the approval from Animal Experimentation Ethics Committee (AEEC) in the Sixth A liated Hospital of Sun Yat-sen University. 1X10 5 GFP or LSD1-197-211-GFP transduced human BTICs were intracranially transplanted into the frontal lobe of 6 to 8-week-old female nude mice. The cells were suspended in 5 ul PBS and injected into the right frontal lobe at 2 mm lateral and 1 mm anterior to bregma with 2.5 mm depth from the skull base. The mice were fed with water containing fresh 2mg/ml Doxycycline (DOX) daily on day after transplantation. The brain tumor growth was monitored by IVIS Spectrum imaging (PerkinElmer) after transplantation. The mouse brains were harvested for hematoxylin-eosin staining.

Amide Hydrogen/Deuterium Exchange and Mass Spectrometry (HDX-MS)
Human NR2E1 (GenScript) was cloned into pET22b to express protein with C-terminal His tag. The protein was puri ed with Ni-NTA beads and followed by gel ltration. Human LSD1 (ATCC) was cloned into pGEX6 to express protein with N-terminal GST tag. The protein was puri ed with glutathione agarose and the GST tag was removed by precission protease digestion at 4 degree overnight. The eluted protein was further puri ed by gel ltration. To study the peptide of Lsd1 involved in the interaction with NR2E1, 50 μM of full length human LSD1 protein was incubated with 75 μM of human NR2E1 (residues 183-354) in buffer A (25 mM potassium phosphate and 5% glycerol at pH 7.5) for 12 hrs prior to the HDX experiments. 2 μL of LSD1 alone or in complex with NR2E1 was mixed with 18 μL of deuterated buffer A resulting in a nal concentration of 90% D 2 O. Exchange reactions were carried out at 20 ˚C for ve different time points (0.5 to 10 min) and quenched by adding 40 μL of ice cold 0.1% tri uoro acetic acid to get a nal pH of 2.5. 50 μL aliquot of the quenched samples containing 0.83 μM of Lsd1 was injected onto a chilled nano-Ultra Performance Liquid Chromatography sample manager (test version, Waters), specially designed for HDX experiments. Online digestion was carried out using an immobilized pepsin column (Porozyme, ABI) in water containing 0.05% FA at a ow rate of 100 μL/min. The digested sample was desalted in a 2.1 x 5 mm C18 peptide microtrap (ACQUITY BEH C18 VanGuard Pre-column, 1.7 μm, Waters) and eluted using a linear gradient of acetonitrile (8-40%) in 0.1% FA, onto a reverse phase analytical column (ACQUITY UPLC BEH C18 Column, 1.0 × 100 mm, 1.7 μm, Waters) at 40 µL/min. Mass spectra were acquired over the m/z range 200−1700 and continuous instrument calibration was carried out using Glu-Fibrinogen peptide (GFP) at 100 fmol/μl. Peptides were identi ed from MSE data of an undeuterated Lsd1 sample using ProteinLynx Global Server (PLGS 2.4) (test version, Waters) and mapped on to subsequent deuteration experiments using prototype custom software (DynamX, Waters). The average number of deuterons exchanged for each of the pepsin digest fragments was obtained as described before (Mandell et al., 1998). Exchange values were not corrected for the deuterium backexchange, that occurs during sample analysis and so all the results reported in this study are only from the relative deuterium level. The difference in deuterium uptake (subtracting the absolute deuterium level in LSD1 from the NR2E1:LSD1 samples) for 56 pepsin digest fragments at all the time points was plotted using Origin software (Origin Pro v.8.6, OriginLab). The percent difference in deuterium uptake for all of the pepsin digest fragments between LSD1 and NR2E1:LSD1 following 1 min HDX were shown below (Eqn. 1) and mapped onto the crystal structure of LSD1 (PDB ID: 2Z3Y) using PyMOL (PyMOL Molecular Graphics System, Version 1.3, Schrodinger, LLC).

Crystal structure determination
The SWIRM domain of human LSD1 residue 172 to 280, was subcloned into the pET15b vector between the EcoRI and BamH1 restriction sites, and the resulting plasmid was transformed into E. coli strain BL21. Cells were grown at 37 °C in LB medium to an optical density of 0.6 at 600 nm and induced with 0.5 mM IPTG. The collected cells were lysed by French press in a buffer containing 50 mM Tris, pH 7.0, 150 mM NaCl, 1 mM EDTA and 1 mM dithiothreitol (DTT). After removing cell debris by centrifugation at 10,000xg for 30 minutes, the supernatant was mixed with Ni-NTA resin, and then poured into a column. After extensive wash, the SWIRM domain protein was released from the resin by 300 mM imidazole. Selenomethionine (SeMet) protein was expressed following the standard method (Yuan et al., 2003). Native crystals were grown at 20 °C by the hanging drop vapor diffusion method: 2 μl of protein at a concentration of 10 mg ml-1 was mixed with equal amount of reservoir buffer consisting of 0.1 M MgCl 2 , 0.1 M Tris pH 8.5, 25% (w/v) PEG 3350. SeMet derivative crystals were grown in reservoir buffer containing 0.8M K/Na tartrate, 0.1 M MES pH 6, 2.5% (v/v) glycerol. Native SWIRM protein crystallizes in the space groups P2 1 2 1 2 1 and P2 1 2 1 2 whereas SeMet SWIRM protein crystallized in space group I222.
Diffraction data of native crystals and SeMet crystals were collected at beamline A1 and beamline F2 respectively in Cornell High Energy Synchrotron Source. All data were processed with DENZO/SCALEPACK. The structure was determined by multiwavelength anomalous dispersion (MAD) methods using SHELX and SHARP and re ned using CNS (Brunger et al., 1998). Native crystal structures were solved by molecular replacement with PHASER using the MAD structure as an initial model. There are two molecules in asymmetric unit. Pymol was used to calculate the cavity and binding pocket and draw the structure. The human LSD1 SWIRM domain coordinates have been deposited in the Protein Data Bank (Accession code 5IT3).

Declarations Supplementary material
Supplementary information of two tables and six gures is included as part of this submission.

Ethics approval
The animal experiments were performed under the approval from Animal Experimentation Ethics Committee (AEEC) in the Sixth A liated Hospital of Sun Yat-sen University.

Con ict of interest statement
The authors declare that no con icts of interest exist.    Effect of NR2E1 on LSD1 as shown by HDX-MS. a. The difference in absolute deuterium uptake between LSD1 and NR2E1:LSD1. b. Enhanced mass spectra for four pepsin digest fragments of LSD1, 197-211, 354-377, 481-501 and 537-546 which show signi cant differences in exchange upon NR2E1 binding: undeuterated peptide (top), isotopic envelope for the same peptide from LSD1 alone following 1 min deuteration (middle) and isotopic envelope for the same peptide from LSD1 in complex with NR2E1 following 1 min deuteration (bottom). c. Heat map showing the percent decrease in deuterium uptake for NR2E1:LSD1 relative to LSD1 following 1 min of HDX, mapped on the crystal of LSD1 (PDB ID: 2Z3Y). d.