Bcl-xL is translocated to the nucleus via CtBP2 to epigenetically promote metastasis

Besides its mitochondria-based anti-apoptotic role, Bcl-xL also travels to the nucleus to promote cancer metastasis by upregulating global histone H3 trimethyl Lys4 (H3K4me3) and TGFβ transcription. How Bcl-xL is translocated into the nucleus and how nuclear Bcl-xL regulates H3K4me3 modification are not understood. Here, we report that C-terminal Binding Protein 2 (CtBP2) binds Bcl-xL via its N-terminus and translocates Bcl-xL into the nucleus. Knockdown of CtBP2 by shRNA decreases the nuclear portion of Bcl-xL and reverses Bcl-xL-induced cell migration and metastasis in mouse models. Furthermore, knockout of CtBP2 suppresses Bcl-xL transcription. The binding between Bcl-xL and CtBP2 is required for their interaction with MLL1, a histone H3K4 methyltransferase. Pharmacologic inhibition of MLL1 enzymatic activity reverses Bcl-xL-induced H3K4me3 and TGFβ mRNA upregulation as well as cell invasion. Moreover, cleavage under targets and release using nuclease (CUT&RUN) coupled with next generation sequencing reveals that H3K4me3 modifications are particularly enriched in the promotor region of genes encoding TGFβ and its signaling pathway in the cancer cells overexpressing Bcl-xL. Altogether, the metastatic function of Bcl-xL is mediated by its interaction with CtBP2 and MLL1.


Introduction
Bcl-xL has long been known for its anti-apoptotic function during embryonic development and in pathological conditions (Boise et al, 1993). Bcl-xL executes its anti-apoptotic function in the mitochondrial membrane by binding to and inhibiting the pro-apoptotic activity of Bax/Bak, which are otherwise poised to initiate the apoptotic cell death pathway. Bcl-xL is frequently overexpressed in cancer, whether newly diagnosed or resistant to therapies (Hafezi & Rahmani, 2021). Moreover, the role for Bcl-xL in tumor metastasis has been previously ascribed to its antiapoptotic function, e.g., Bcl-xL increases metastasis by providing a survival advantage to the tumor cells (Fernandez et al, 2000).
We have previously investigated the interdependence of Bcl-xL's metastatic function and its canonical anti-apoptotic function. Using multiple mutants, cell lines, and mouse models, we have discovered that Bcl-xL promotes metastasis independent of its canonical anti-apoptotic function and that the metastatic function requires its nuclear translocation (Choi et al, 2016;Du et al, 2007). We have demonstrated that ABT-737, a prototype Bcl-xL inhibitor, does not affect the migration function of Bcl-xL (Choi et al., 2016). Similar to wild-type (wt) Bcl-xL, antiapoptosis-defective Bcl-xL mutants and an engineered Bcl-xL targeted to the nucleus can still promote tumor cell migration and invasion (Choi et al., 2016). In contrast, engineered Bcl-xLeither targeted to the mitochondria or excluded from the nucleus -does not promote cell migration and invasion even though they protect cells from apoptosis (Choi et al., 2016).
Generation of cell lines expressing inducible shRNAs. The shRNAs for human CtBP2 and MLL1 were designed using the SplashRNA algorithm (Pelossof et al, 2017) and cloned into the LT3RENIR vector, expressing dsRed fluorescence-coupled miR-E shRNAs from an optimized Tet-responsive element promoter (T3G) (Fellmann et al, 2013). was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS, 0.2 mM Lglutamine, 100 units/ml penicillin and 100 µg/ml streptomycin. BON1/TGL/pQCXIP (pQ, vector control) and BON1/TGL/HA-Bcl-xL (wt or mutants) cells have been described (Choi et al., 2016) and maintained in above medium with 0.5 μg/ml puromycin. Cells were cultured at 37 o C in a standard tissue culture humidified chamber.
The following amount of DNA was used in transient transfection into U2OS cells: 1 µg HA-Bcl-xL, 1 µg pQ vector, 1µg full length-CtBP2-V5, 2 µg N-terminal CtBP2-V5, 8 µg C-terminal . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. CtBP2-V5, 2 µg CtBP2-V5-mut1, 2 µg CtBP2-V5-mut2, 4 µg CtBP2-V5-mut3, and 6 µg were used for each transfection. Cells were washed with PBS, then DMEM containing 1% FBS and 0.2 mM L-glutamine (without antibiotics) were added. Transfection complexes were added to cells in a dropwise manner and the dishes were gently swirled. Cells were incubated for 48 hours at 37°C, 5% CO2 before protein lysates were made. All the parental cell lines were authenticated by University of Arizona Genetics Core before use.

Reverse transcription quantitative real-time PCR (RT-qPCR).
Cell lines (MDA-MB-231, MDA-MB-231 with CtBP2 KO, BON1/TGL/pQ and BON1/TGL/HA-Bcl-xL) grown on 6-cm . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 dishes were used to isolate mRNA using RNeasy Plus Mini kit (Qiagen, 74136) containing gDNA Eliminator spin columns. cDNA was generated using the SuperScript™ IV VILO™ Master Mix with ezDNase™ Enzyme (Invitrogen, 11766050), and power SYBR green-based quantitative real-time PCR was performed using primers specific for human BCL2L1 (forward: Animal experiments. NSG mice were purchased from the Jackson Laboratory. All mice were housed in accordance with the institutional guidelines. All procedures involving mice were approved by the institutional animal care and use committee of Weill Cornell Medicine. NSG mice at the age of ~10 weeks were fed with dox-containing diet (Envigo, TD.00426) one day before the injection of cells and injected with one million cells that were treated with . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 doxycycline (0.5 μg/ml for 96 hours) in 100 μl PBS via the left ventricle. For tail vein metastasis assay, 1.5 × 10 6 N134 cells in 150 μl PBS were injected into the tail veins of NSG mice at the age of ~10 weeks. Mice were subjected to bioluminescent imaging using the In Vivo Imaging System Spectrum (PerkinElmer) as described (Choi et al., 2016;Choi et al, 2019;Du et al, 2011;Du et al., 2009).
Fifty thousand cells per cell line were seeded in 1% FBS, 0.4 mg/ml G418, 0.5 µg/ml puromycin, 0.2 mM L-glutamine and 100 units/ml penicillin and 100 µg/ml streptomycin per well of 24-well dish for the proliferation controls. After 24 hours of incubation, cells on the opposite side of the chambers were fixed in 4% paraformaldehyde in PBS for 10 min, washed with PBS-/-for 5 min, stained with 0.1% crystal violet for 30 min and counted in 8 fields under 20X magnification. Subcellular fractionation. Subcellular fractionation was performed based on the method previously described (Choi et al., 2016).
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Caspase-3/-7 activity assay. A total of 1 × 10 4 cells (N134 parental cells and N134 expressing wt Bcl-xL, HA-Bcl-B/xL(BH4-loop) (construct #5), or RCASBP-HA-Bcl-xL/B(BH4-loop) (construct #6)) were seeded in 100 µL culture medium in a 96-well white wall plate. After overnight incubation, cells were treated with DMSO or 10 mM etoposide for 24 hours. Caspase-3 and -7 activities were measured using the Caspase-Glo 3/7 assay kit (Promega, G8090) according to manufacturer's instructions. Luminescence was measured using a Glomax Multi Plus Detection System (Promega, Madison, WI, USA), and caspase 3/7 activities was normalized by the number of cells. Data represented 3 replicates per condition and were expressed as mean of fold change over N134 cells without etoposide ± SEM.
CUT&RUN and next generation sequencing analysis. BON1/TGL/pQ and BON1/TGL/HA-Bcl-xL cells were washed with cold PBS and scraped from cell culture dishes (triplicates). Cells were pelleted at 4 o C, 3,000g for 5 minutes. CUT&RUN was performed from 500,000 cells per . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. condition in triplicates as previously described in (Skene et al, 2018)  was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 NextSeq 2000 sequencer (Illumina). The raw sequencing reads in BCL format were processed through bcl2fastq 2.20 (Illumina) for FASTQ conversion and demultiplexing. Data were processed using Cut&RunTools (https://bitbucket.org/qzhudfci/cutruntools/src/default/) (Zhu et al, 2019) with the default settings. Briefly, reads were adapter trimmed using Trimmomatic (Trimmomatic, RRID: SCR_011848) (Bolger et al, 2014), and an additional trimming step was performed to remove up to 6 bp adapter from each read. Next, reads were aligned to the hg19 genome using bowtie 2 (Langmead & Salzberg, 2012) with the 'dovetail' settings enabled.
Alignments were further divided into ≤ 120-bp and > 120-bp fractions. Alignments from the > 120-bp fractions were used for peak calling with MACS2 (2.1.1) (Zhang et al, 2008), followed by de novo motif searching within the peak regions with MEME suite (4. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;

Results
Nuclear Bcl-xL is detected in breast cancer patients. In healthy cells, most of the Bcl-xL proteins reside in the mitochondria, while some are localized in the cytosolic and ER (Popgeorgiev et al, 2018). However, we have found nuclear Bcl-xL in the liver metastases of pancreatic neuroendocrine tumor (PNET) by immunofluorescent staining, and our studies suggested that nuclear Bcl-xL promotes metastasis (Choi et al., 2016). To broaden the clinical relevance of nuclear Bcl-xL beyond PNET, we sought to determine whether nuclear Bcl-xL can also be found in specimens of breast cancer, the most common malignancy in women. We conducted immunofluorescent analysis using antibodies against Bcl-xL and MTCO1 (a mitochondrial marker) and visualized nuclear DNA with 4′,6-diamidino-2-phenylindole (DAPI).
We detected predominant nuclear Bcl-xL in 3 of 15 cases of breast cancer specimens (Figure 1a, 1b, and data not shown) and weak nuclear Bcl-xL in the other cases ( Figure 1c and data not shown), suggesting that Bcl-xL can also be translocated into the nucleus of the breast cancer cells.
To determine whether mutations of Bcl-xL contribute to the nuclear translocation, we microdissected the breast cancer case that had mostly nuclear Bcl-xL, purified genomic DNA, PCR amplified Bcl-xL, and sequenced the entire Bcl-xL gene. We found no mutation in Bcl-xL (Supplementary Figure S1), suggesting that the change of subcellular localization of Bcl-xL in the cancer cells was not due to mutations in Bcl-xL. This raised a possibility that Bcl-xLinteracting proteins involve in its subcellular localization changes.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10.1101/2023.04.26.538373 doi: bioRxiv preprint

C-terminal Binding Protein 2 (CtBP2) is a novel Bcl-xL interacting protein.
To identify Bcl-xL-interacting proteins that might promote its nuclear translocation, we employed an "ReCLIP" (Reversible Cross-Link Immuno-Precipitation) procedure using cell-permeable, thiol-cleavable crosslinkers to stabilize normally labile interactions in situ prior to isolation (Smith et al, 2011).
We used human PNET BON1/TGL cells overexpressing HA-tagged wild-type (wt) Bcl-xL and the two HA-tagged Bcl-xL mutants defective in anti-apoptosis (mt1 and mt2) (Choi et al., 2016) for ReCLIP with anti-HA magnetic beads. The proteins in ReCLIP were subjected to mass spectrometry analysis to identify Bcl-xL interacting proteins. Among the specific proteins immunoprecipitated by anti-HA magnetic beads from the cells overexpressing HA-tagged wt Bcl-xL and Bcl-xL mutants, but not immunoprecipitated from the parental cells overexpressing the control vector, we focused on the proteins with both a nuclear localization signal (NLS) and a role in metastasis. The top candidate was CtBP2 (Supplementary Table 1) because of the following reasons. First, CtBP2 is a transcriptional regulator containing an NLS (Chinnadurai, 2003(Chinnadurai, , 2009Fang et al, 2006;Paliwal et al, 2012). It can act as a transcriptional repressor or activator. Second, CtBP2 is overexpressed in cancer, including prostate cancer, ovarian cancer, liver cancer, breast cancer, and esophageal squamous cell carcinoma (Cerami et al, 2012;Gao et al, 2013;Takayama et al, 2014;Zhang et al, 2015;Zheng et al, 2015). Third, CtBP2 has been shown to promote epithelial-mesenchymal transition (EMT) and migration of several types of cancer cells (Wang et al, 2013;Zhang et al., 2015;Zheng et al., 2015). Fourth, H3K4me3 is enriched at the CtBP2-binding sites in human embryonic stem cells (Lee et al, 2015) and Bcl-xL overexpression increases global H3K4me3 levels in cancer cells (Choi et al., 2016).
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10. 1101/2023 To verify the interaction between CtBP2 and HA-Bcl-xL identified by IP-mass spectrometry, we performed anti-HA ReCLIP using lysates from BON1/TGL cells overexpressing HA-tagged wt Bcl-xL or the vector, followed by Western blotting using antibodies against CtBP2 and HA. In consistent with the IP-mass spectrometry results, CtBP2 was detected in the anti-HA ReCLIP specifically from cells expressing HA-tagged wt Bcl-xL, but not from the control cells expressing the vector (Figure 2a).  Table S2). This LT3RENIR vector expresses dsRed-coupled miR-E shRNA from an optimized Tet-responsive element promoter (T3G), which strongly reduces the leaky shRNA expression, and its miR-E backbone increases the mature shRNA levels and knockdown efficacy (Fellmann et al., 2013). It also harbors rtTA3 to enable single-vector ("all-in-one") Tet-ON shRNAmir expression (Fellmann et al., 2013). We used the "tet-O-dsRed-shRNA-PGK-rtTA3 (TR-shRNA)" viruses to infect BON1/TGL/HA-Bcl-xL cells (BON1/HA-Bcl-xL).
Infected cells were selected with neomycin to obtain stable clones. We investigated knockdown efficiency of CtBP2 by these different shRNAs in the stable cell lines following 96 hours of doxycycline treatment and Western blotting. Two (#2260 and #2403) shRNAs against CtBP2 successfully reduced the levels of CtBP2 proteins (Figure 2b).
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10. 1101/2023 Co-expression of dsRed and shRNAs by dox induction allows dsRed reporter-based estimation of shRNA expression by a fluorescent microscope. We observed that ~ 50% of cells showed positive dsRed signal after 1 μg/ml doxycycline treatment for 96 hours, indicating a mixed cellular response to doxycycline induction and thus shRNA expression and CtBP2 knockdown (Supplementary Figure S2). This gave us a unique opportunity to determine whether knockdown of CtBP2 affected Bcl-xL nuclear localization by comparing subcellular localization of HA-Bcl-xL in the dsRed-positive cells and the dsRed-negative cells in the same images. We have previously shown that the nuclear Bcl-xL, but not its cytoplasmic counterpart, accounts for its migration/invasion activity (Choi et al., 2016). Using anti-HA immunofluorescent staining, we observed similar nuclear/cytoplasmic ratio of HA-Bcl-xL in the dsRed-positive and dsRednegative BON1/HA-Bcl-xL/TR-shRLuc #713 control cells (Figure 2c and 2d). In contrast, the nuclear/cytoplasmic ratio of HA-Bcl-xL was significantly reduced in the dsRed-positive shCtBP2 cells compared to the dsRed-negative cells in both BON1/HA-Bcl-xL/TR-shCtBP2 #2260 and #2403 cultures (Figure 2c and 2d), suggesting that knockdown of CtBP2 inhibits nuclear translocation of Bcl-xL.

CtBP2 knockdown decreases the ability of Bcl-xL to promote invasion and metastasis.
To examine whether the invasion function of Bcl-xL is mediated by CtBP2, we performed transwell invasion assays using dox-treated BON1/HA-Bcl-xL/TR-shCtBP2 #2260 and #2403 cells and BON1/HA-Bcl-xL/TR-shRLuc #713 control cells. We found that invasion activity was significantly reduced in dox-treated shCtBP2 (#2260 and #2403) cells compared with doxtreated shRLuc (#713) control cells (Figure 2e). On the other hand, there was no significant . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. We next performed biochemical fractionation analysis using these FACS sorted, dsRed-positive cells and found that both soluble nuclear HA-Bcl-xL and chromatin-bound HA-Bcl-xL were greatly reduced in the dox-induced CtBP2 knockdown cells (#2260 and #2403) compared to the dox-induced shRLuc control cells (#713) (Figure 3c). Mek1/2 (cytoplasmic proteins) and histone H3 (a nuclear protein) were used in the Western blotting as the reference for each fractionation respectively (Figure 3c). Consistent with the immunofluorescence results, the parallel biochemical subcellular fractionation assay further validated that CtBP2 mediated the nuclear translocation of Bcl-xL.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 As our data from clinical specimens and invasion assays suggested that CtBP2-driven nuclearlocalization of Bcl-xL may promote metastasis, we performed experimental metastasis assays in mice using FACS sorted cells. We intracardiacally injected 1 x 10 6 dox-treated BON1/HA-Bcl-xL/TR-shRLuc #713 (control cells) and BON1/HA-Bcl-xL/TR-shCtBP2 #2260 cells into NOD/scid-lL2Rgc knockout (NSG) immunodeficient mice. Mice were started on a doxycycline diet one day before tumor injection and kept on doxycycline diet until the end point. The luciferase expression in these cells provided a mean for monitoring the localization and growth of tumor cells through in vivo bioluminescent imaging. Indeed, weekly bioluminescent imaging showed that bioluminescence signals started to increase after 14 days, and metastatic cells were detected at multiple sites on day 28 (Figure 3d and 3e). The increase in bioluminescent signals over time from the CtBP2 knockdown group was significantly less than that from the control shRLuc (#713) group (Figure 3d, GEE analysis, p = 0.0366), suggesting that knockdown of CtBP2 reduced metastatic progression of Bcl-xL-overexpressing BON1 cells in mice.
CtBP2 knockout reduces Bcl-xL transcripts and the nuclear pool of Bcl-xL. In addition to shRNA knockdown of CtBP2 in BON1 PNET cells, we used CRISPR-Cas9 to knock out (KO) CtBP2 in the MDA-MB-231 breast cancer cell line to further test the regulation of nuclear translocation of Bcl-xL by CtBP2. We found that CtBP1, a homolog of CtBP2, did not increase to compensate the loss of CtBP2, but CtBP2 KO reduced endogenous Bcl-xL protein levels ( Figure 4a). Because CtBP2 is a transcription cofactor, we investigated the possibility that CtBP2 regulated Bcl-xL transcription. Our RT-PCR confirmed that CtBP2 mRNA was not detectable in the KO cells ( Figure 4b) and Bcl-xL mRNA was significantly reduced in the CtBP2 KO cells . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 compared to the control cells (Figure 4c). Furthermore, we searched the public research project website, The Encyclopedia of DNA Elements (ENCODE) (Consortium, 2004(Consortium, , 2011, and found that CtBP2 bound near the transcription start site of Bcl-xL in the dataset of CtBP2 ChIP-Seq from human H1-hESC cell line (Supplemental Figure S2). Taken together, CtBP2 also functions as a transcription activator of Bcl-xL.
To determine whether CtBP2 KO also reduced nuclear subcellular localization of endogenous Bcl-xL, we performed biochemical fractionation and immunocytochemical staining in CtBP2 KO cells. The biochemical fractionation analysis showed that the nuclear pool of endogenous Bcl-xL was greatly reduced in the CtBP2 KO cells (Figure 4d). Immunofluorescent staining using antibodies against endogenous Bcl-xL and CtBP2 showed that the Bcl-xL signals were weaker in CtBP2 KO cells than in the control cells and the nuclear/cytoplasmic ratio of Bcl-xL in CtBP2 KO cells was significantly reduced (Figure 4e and 4f). Taken together, CtBP2 plays a dual role of regulating Bcl-xL transcription and translocating Bcl-xL proteins into the nucleus, both of which contribute to promote Bcl-xL's pro-metastatic activity.
The residues 82 to 104 of CtBP2 and the N-terminus of Bcl-xL are required for their interaction. To characterize the Bcl-xL/CtBP2 interaction domains, we used deletion constructs of CtBP2 to define the Bcl-xL interaction domain of CtBP2 (Figure 5a). DNA constructs of 3 different V5-tagged CtBP2 constructs (Full length (445 aa), residues 1 to 321, and residues 322 to 445) (Paliwal et al., 2006) and HA-tagged full-length Bcl-xL were transiently co-expressed in U2OS cells. After 48 hours, cell lysates were prepared for anti-HA ReCLIP. We found that full . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10.1101/2023.04.26.538373 doi: bioRxiv preprint length CtBP2 and N-terminal CtBP2 (residues 1 to 321), but not C-terminal CtBP2 (residues 322 to 445) bound to HA-Bcl-xL (Figure 5b).
To further narrow down the Bcl-xL-binding domain in the N-terminal region of CtBP2, we generated a series of CtBP2 N-terminal deletion constructs with a V5 tag (Figure 5a, mut1: residues 33 to 321, mut2: residues 61 to 321, mut3: residues 82 to 321, and mut4: residues 105 to 321) to teste for their capability to bind Bcl-xL. DNA constructs of HA-tagged Bcl-xL and these V5-tagged CtBP2 deletion constructs were transiently co-expressed in U2OS cells. After 48 hours, cell lysates were prepared for anti-HA ReCLIP. CtBP2 deletion constructs #1-3, but not #4, were detected in the HA-Bcl-xL IP (Figure 5c), suggesting that residues 82 to 104 of CtBP2 were required for its interaction with Bcl-xL.
To determine the CtBP2-binding domain in Bcl-xL, we performed the reciprocal experiments.
We initially generated a series of N-terminal deletion Bcl-xL mutants with HA tag, but they did not express well (data not shown). Because we found that Bcl-B, another Bcl-2 anti-apoptotic family member (Beverly & Varmus, 2009), did not interact with V5-tagged CtBP2 in anti-V5 ReCLIP (Figure 6a and 7b, construct #1), we utilized a set of chimeric Bcl-xL/Bcl-B proteins (Saurabh et al., 2014) to narrow down the CtBP2 binding domain in Bcl-xL. Bcl-xL (#1), Bcl-B (#2), and six chimeric Bcl-xL/Bcl-B constructs (#3-8) were transfected into 293T cells, and all could be transiently expressed to similar levels ( Figure 6b). Anti-V5 (for CtBP2) ReCLIP revealed that WT Bcl-xL, construct #4 that contains Bcl-xL's BH4 domain, BH3 domain, and two N-terminal loop domain and construct #5 that contains Bcl-xL's BH4 domain and the first N-terminal loop domain were detected in the anti-V5 IP, but the other chimeric proteins were not . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10.1101/2023.04.26.538373 doi: bioRxiv preprint detected in the anti-V5 IP (Figure 6b). Therefore, the BH4 domain and the first N-terminal loop domain (residues 1 to 85) of Bcl-xL were needed for its interaction with CtBP2.

The interaction between Bcl-xL and CtBP2 is important for the metastatic function of Bcl-
xL. To investigate the relationship between Bcl-xL's binding ability to CtBP2 and the metastatic function, we performed tail vein experimental metastasis assays. We sub-cloned two chimeric constructs, #5 (which binds to CtBP2) and #6 (which does not bind to CtBP2) into an avian retroviral vector, RCASBP, with an HA tag. We infected the N134 cell line (Du et al., 2007), which is derived from an PNET in a RIP-Tag; RIP-tva mouse, with RCASBP-HA-Bcl-B/xL(BH4-loop) (construct #5) or RCASBP-HA-Bcl-xL/B(BH4-loop) (construct #6). The expression levels of both HA-tagged chimeric proteins in N134 cells were similar by Western blot analysis (Figure 6c). Both chimeric constructs #5 and #6 protected cells from apoptosis as effectively as the wild-type Bcl-xL (Fig. 7d). We injected 1.5 x 10 6 N134 cells expressing chimeric constructs #5 and #6 into the tail vein of NSG mice. After five weeks, organs of the recipient mice were harvested to survey for metastases. We found significantly more liver macrometastases in recipients of construct #5 that binds to CtBP2 than construct #6 that does not bind to CtBP2 (Figure 6e and Supplemental Figure S3), suggesting that the interaction between Bcl-xL and CtBP2 is important for the metastatic function of Bcl-xL.

Mixed lineage leukemia protein-1 (MLL1) interacts with the CtBP2 and Bcl-xL complex.
We previously reported that Bcl-xL increases H3K4me3 and the levels of TGFβ1 through H3K4me3 epigenetic modification at the TGFβ1 promoter (Choi et al., 2016). However, the mechanism by which Bcl-xL increases H3K4me3 was unknown. Because H3K4 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; https://doi.org/10.1101/2023.04.26.538373 doi: bioRxiv preprint methyltransferase, MLL1 (also known as KMT2A), has a putative CtBP2 binding motif (PXDLS), we hypothesized that the Bcl-xL/CtBP2 complex recruits MLL1 to epigenetically regulate transcription of key genes in metastasis. To test this, we investigated whether MLL1 is present in the Bcl-xL/CtBP2 protein complex. We detected the presence of MLL1 in the anti-HA (for Bcl-xL) or anti-V5 (for CtBP2) ReCLIP only when transiently co-expressing the constructs of Bcl-xL and CtBP2 that were able to interact with each other in U2OS cells (Figure 5b, 5c, and 6b). We were not able to detect MLL1 in the ReCLIP when the constructs of Bcl-xL and CtBP2 did not interact with each other in U2OS cells (Figure 5b, 5c, and 6b). Furthermore, we performed anti-HA ReCLIP using lysates from BON1/TGL cells overexpressing HA-Bcl-xL or vector, and we detected MLL1 and CtBP2 in the HA ReCLIP from BON1/TGL cells overexpressing HA-Bcl-xL (Supplemental Figure S4). Altogether, the data suggest that the interaction between CtBP2 and Bcl-xL is required for the binding of MLL1 into this protein complex.
To investigate whether the histone methyltransferase activity of MLL1 is required for Bcl-xL to raise H3K4me3 levels, increase TGFβ1 transcription, and promote invasion, we treated cells with MM-102, a high-affinity, small-molecule peptidomimetic MLL1 inhibitor (Karatas et al, 2013). As previously reported (Choi et al., 2016), overexpression of Bcl-xL increased total H3K4me3 levels ( Figure 7a, lane 3). We found that treatment of MM-102 significantly reduced total H3K4me3 levels (Figure 7a), decreased TGFβ1 transcripts (Figure 7b), and suppressed invasion induced by Bcl-xL in transwell invasion assays (Figure 7c).
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; In addition to pharmacological inhibition of MLL1 activity, we generated dox-inducible shRNA to silence MLL1. We designed six different shRNAs against human MLL1 with the dsRed reporter (Supplementary Table S2) and generated stable BON1/TGL/HA-BcL-xL/tet-O-dsRed-shMLL1-PGK-rtTA3 (BON1/HA-Bcl-xL/TR-shMLL1) cell lines. After 96-hour doxycycline treatment, almost all cells became dsRed-positive (Figure 7d). We examined the knockdown efficacy of MLL1 by Western blot analysis. We found that two shMLL1 constructs (#13408 and #13406) led to the most effective MLL1 knockdown and to a drastic reduction of H3K4me3 levels ( Figure 7e). To determine whether MLL1 mediates the metastatic function of Bcl-xL in vitro, we performed a 3D tumor spheroid invasion assay. We seeded BON1/HA-Bcl-xL/TR-shMLL1 #13408 cells into an ultra-low attachment plate with Matrigel. 96 hours later (labeled as Day 0 in Figure 7f and 7g), we began treatment with 0.5 μg/ml doxycycline. We found that knockdown of MLL1 in the dox group suppressed tumor spheroid invasion of the Bcl-xLoverexpressing cells (Figure 7f). GEE analysis of whole invasive area changes over time showed that the dox group (shMLL1) was significantly less than the no dox group (p<0.0001) ( Figure   7g). On the other hand, knockdown of the control RLuc did not reduce tumor spheroid invasion (Supplemental Figure S5). was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

TGFβ signaling is the top pathway promoted by Bcl-xL-induced
The copyright holder for this preprint (which this version posted April 28, 2023. ; the control BON1/pQ cells (Fig. 9a). Using Ingenuity Pathway Analysis (IPA) for these 1,190 unique H3K4me3 histone modification regions, we found that the top canonical pathway was TGFβ signaling (Fig. 9b). Fig. 9c-9g showed the H3K4me3 and IgG peaks around the transcription start sites of ACVR1, NKX2-5, SMAD3, SMAD5, and ZNF42 that belong to the TGFβ signaling pathway. The CUT&RUN-Seq also revealed differential enrichment of H3K4me3 histone modifications in the TGFβ1 promoter region in the BON1/HA-Bcl-xL cells compared to the control pQ cells (Fig. 9h). RT-qPCR showed that mRNA levels of ACVR1, NKX2-5, SMAD3, SMAD5, and ZNF42 were upregulated in HA-Bcl-xL overexpressing cells compared to control pQ cells (Fig. 9i), suggesting a correlation between the transcription and the increased H3K4me3 marks identified in the CUT&RUN-Seq assay.

Discussion
Metastasis, the dissemination of cancer cells from primary sites to distant organs, accounts for the majority of cancer-associated mortality. Efforts on unraveling the molecular basis of tumorigenesis and metastasis have significantly advanced our understanding of metastasis.
During cancer progression, the ability to override apoptosis is a prominent mechanism that cancer cells usually acquire (Hanahan & Weinberg, 2011). The Bcl-2 protein family plays a central role in regulating the intrinsic apoptosis pathway at the outer mitochondria membrane. At least twelve structure-related members of the Bcl-2 superfamily have been identified which could be categorized to three groups: the pro-apoptotic Bax and Bak, anti-apoptotic members like Bcl-2 and Bcl-xL, and BH-3 only proteins that cooperate with Bax and Bak to promote apoptosis (Youle & Strasser, 2008). Attempts at targeting anti-apoptotic Bcl-xL members in cancers have been primarily focused on inhibiting its anti-apoptotic activity, typically by . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 disrupting its association with pro-apoptotic members and BH3-only proteins to release the sequester of these death-inducing factors and hence to amplify apoptotic signals. These strategies, however, showed limited success in clinical trials (Gandhi et al, 2011;Rudin et al, 2012), suggesting unrecognized alternative mechanisms for Bcl-xL to promote cancer progression.
This study, together with our previous findings (Choi et al., 2016), identifies a novel mechanism that CtBP2 transport overexpressed Bcl-xL into the nucleus and nuclear Bcl-xL drives metastasis of cancer cells independent of its anti-apoptotic activity. Nuclear import of macromolecules is a selective process. Proteins typically rely on the presence of NLS motifs to be recognized by specific transport receptors of the importin/karyopherin family, which mediate the transport of macromolecular cargos across the nuclear pore complex into the nucleus (Freitas & Cunha, 2009). Proteins without NLS sequences, like Bcl-xL, can be recruited to the nucleus through a "piggy-back" mechanism by associating with NLS-containing binding partners. Using a special "ReCLIP" technique to stabilize labile protein-protein interactions in situ, we discovered that CtBP2, a transcriptional regulator, translocates Bcl-xL into the nucleus. CtBP2 contains an KRQR NLS sequence at the N-terminal domain, which is crucial for its accumulation and gene repression functions in the nucleus (Verger et al, 2006;Zhao et al, 2006). In addition, CtBP2 shuttles between the nucleus and cytoplasm, and thus has been implicated in intracellular trafficking (Verger et al., 2006). We demonstrated that knockdown of CtBP2 significantly reduced the nuclear/cytoplasm partitioning of Bcl-xL and invasion and metastasis of the cancer cells induced by Bcl-xL overexpression in vivo.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; The BH1, BH2, and BH3 domains of Bcl-xL create a hydrophobic groove to bind BH3-only pro-apoptotic proteins, inhibit their activity and lead to a pro-survival phenotype (Lee & Fairlie, 2019). The first loop domain of Bcl-xL is unstructured based on NMR or X-ray crystallographic analyses, and it does not contribute to anti-apoptotic activity of Bcl-xL (Muchmore et al, 1996). Here we showed that the BH4 domain and the first N-terminal loop domain (residues 1 to 85) of Bcl-xL are important for its interaction with CtBP2. The result of our domain mapping is in consistent with the finding that ABT-737, a BH-3 mimetic, did not affect the ability of Bcl-xL to promote cell migration (Choi et al., 2016). Because we found that another anti-apoptotic BCL family member, Bcl-B, did not bind to CtBP2, the interaction between Bcl-xL and CtBP2 is not shared among all anti-apoptotic BCL family members. On the other hand, we identified that residues 82 to 104 of CtBP2, which lie within a domain that recognizes PxDLS motif in many transcription factors that bind to CtBP2 (Dcona et al, 2017), are required for CtBP2 binding to Bcl-xL. Although CtBP1 interacts with the epigenetic regulator MLL1 (Xia et al, 2003), it was previously unknown whether CtBP2 interacts with MLL1. Here we showed that MLL1 associated with the Bcl-xL/CtBP2 complex. Despite containing a PxDLS motif, MLL1 only interacts with CtBP2 when Bcl-xL is present. It is likely that Bcl-xL causes exposure of the PxDLS motif in MLL1. We also postulate that MLL1 links to the Bcl-xL/CtBP2 complex to DNA/H3K4me3, and CtBP2 and Bcl-xL change MLL1's H3K4me3 site selection as MLL1 fusion proteins are known to change transcriptional targets (Krivtsov et al, 2017;Xu et al, 2016). Further study will be required to identify the exact components of the Bcl-xL/CtBP2/MLL1 protein complex.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Enhanced TGFβ signaling promotes cancer metastasis (Derynck et al, 2021). We previously demonstrated that overexpression of Bcl-xL induces EMT and increased TGFβ production in a Bax/Bak-independent manner (Choi et al., 2016;Du et al., 2007). We also showed that Bcl-xL increased H3K4me3 at the promoter region of TGFβ and TGFβ-neutralizing antibodies significantly reduced Bcl-xL-mediated invasion (Choi et al., 2016). This work identified a Bcl-xL/CtBP2 complex that interacts with MLL1 to increase H3K4me3 modification on TGFb1 and other genes involved in TGFβ signaling. Using MM102, a high-affinity, small-molecule peptidomimetic inhibitor for MLL1's enzymatic activity, we demonstrated that H3K4 HMT activity of MLL1 is important for Bcl-xL to promote global levels of H3K4me3, TGFβ1 transcripts, and invasion. CtBP2 not only controls Bcl-xL nuclear translocation, but also regulates the transcription of Bcl-xL. It is likely that Bcl-xL drives CtBP2 transcription program as part of the CtBP2 transcription supercomplex to promote metastasis. Our work suggests a novel therapeutic perspective by targeting the interacting interface between CtBP2 and Bcl-xL, and thereby reducing the accumulation of Bcl-xL in the nucleus and attenuating metastasis progression.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023.

Availability of supporting data
The CUT&RUN-Seq datasets of this study are available in Gene Expression Omnibus (accession #GSE221629, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE221629), and the other data are available from the authors upon reasonable request.

Competing interests
The authors declare no competing financial interests.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Smith AL, Friedman DB, Yu H, Carnahan RH, Reynolds AB (2011) ReCLIP (reversible crosslink immuno-precipitation): an efficient method for interrogation of labile protein complexes. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

PLoS
The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  Confocal microscopy images of Bcl-xL (red), mitochondrial marker MTCO1 (green), and DAPI (blue) from two primary human breast cancer (a and c) and lymph node met (b). Scale bar, 20 μm. Original magnification, × 60.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Control
The copyright holder for this preprint (which this version posted April 28, 2023. ;

33
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ; . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted April 28, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023