Adaptor proteins mediate CXCR4 and PI4KA crosstalk in prostate cancer cells and the significance of PI4KA in bone tumor growth

The chemokine receptor, CXCR4 signaling regulates cell growth, invasion, and metastasis to the bone-marrow niche in prostate cancer (PCa). Previously, we established that CXCR4 interacts with phosphatidylinositol 4-kinase IIIα (PI4KIIIα encoded by PI4KA) through its adaptor proteins and PI4KA overexpressed in the PCa metastasis. To further characterize how the CXCR4-PI4KIIIα axis promotes PCa metastasis, here we identify CXCR4 binds to PI4KIIIα adaptor proteins TTC7 and this interaction induce plasma membrane PI4P production in prostate cancer cells. Inhibiting PI4KIIIα or TTC7 reduces plasma membrane PI4P production, cellular invasion, and bone tumor growth. Using metastatic biopsy sequencing, we found PI4KA expression in tumors correlated with overall survival and contributes to immunosuppressive bone tumor microenvironment through preferentially enriching non-activated and immunosuppressive macrophage populations. Altogether we have characterized the chemokine signaling axis through CXCR4-PI4KIIIα interaction contributing to the growth of prostate cancer bone metastasis.


Page 3/13
TTC7B and PI4KIIIα are inhibited upon TTC7B knockdown (Fig. 3A, B, C). We have previously shown that PI4KIIIα activity is localized to the invasive protrusions in prostate cancer cells as measured by PI4P immunostaining and regulates the multiple chemokines induced prostate cancer cell invasion as the receptors for chemokines are shown to interact with PI4KIIIα adaptor protein TTC7B [21]. As expected, both chemokines CXCL12 and CXCL8 induced chemoinvasion in scr siRNA transfected cells (Fig. 3D). In TTC7B siRNA transfected cells, both basal as well as chemokine-induced chemoinvasion is inhibited ( Fig. 3E, F), suggesting that TTC7B regulates chemokine receptor-induced cancer cell invasion.
PI4KA knockdown inhibits bone tumor growth. CXCL12/CXCR4 signaling promote initial establishment of prostate cancer colonization in bone microenvironment by actively competing with osteoblastic niche and these interactions along with cross talk with bone microenvironment factors promoting bone tumor growth [32]. Thus, to determine the role of PI4KA in CXCL12/CXCR4 signaling contributing to bone tumor growth, we have developed stable knockdown of PI4KA in PC3-CXCR4 and C4-2B cell with two independent PI4KA shRNA (#27 and #30) through lentiviral transduction. PI4KA expression and kinase activity is inhibited in both cell lines upon shRNA mediated stable knockdown (Fig. 4A). Matrigel invasion of both PI4KA shRNA knockdown cell lines is inhibited compared to Src shRNA cells in basal as well as CXCL12 induced cell invasion ( Fig. 4B and C). To determine the signi cance of PI4KA knockdown on bone tumor growth, Scr and PI4KA knock down PC3-CXCR4 cells implanted in the mice tibiae. Luciferase imaging of bone tumors show that tumor growth is signi cantly inhibited (p < 0.05 for #27 and p < 0.01 for #30 PI4KA shRNA) in both PI4KA shRNA knock down cells over scr shRNA infected cells ( Fig. 4D and E). Taken together, these ndings show that the PI4KA signaling contributes to the growth of cancer cells in bone.
PI4KA expression associates with PCa metastasis and tumor proliferation. Among the PI4P kinases, we found that PI4KA highly expressed in the metastatic prostate cancer compared to the matched primary tumors of PC patients [21]. We evaluated multiple NCBI datasets and found that PI4KA is signi cantly associated with metastatic disease, and in therapy resistance disease its expression is associated with neuroendocrine phenotype (Supplementary Fig. 3). To determine the impact of PI4KA expression on disease outcome in metastatic hormone sensitive PCa (mHSPC), we performed RNA-sequence analysis on 50 metastatic tumor biopsies, out of which 39 are bone biopsies, 10 lymph node and 1 liver biopsy. Clustering analysis of RNA-seq showed clear separation of gene expression pro les among mHSPC tumor biopsies ( Supplementary Fig. 4). The bone met samples were categorized based on median high and low PI4KA expression and heatmaps of this analysis shows a set of genes commonly regulated across the tumor biopsy samples (Fig. 5A). The gene-set enrichment analysis (GSEA) of bone met biopsies showed enrichment of several leading-edge pathways involved in cell-proliferation in biopsies with high PI4KA expression ( Fig. 5B and Supplementary Fig. 5). The top-most signi cant enrichment plots of some of the gene sets with FDR < 0.05 include the P53 pathway, Myc targets, PI3K-AKT-MTOR signaling, and Wnt-Beta-Catenin signaling. This data suggests that high PI4KA expression in bone mets is associated with cell proliferation, and this data is in line with our cell and animal bone xenograft experiment ( Fig. 4B-D) that knock-down on PI4KA in PCa cells leads to inhibition of bone tumor growth.
PI4KA expression correlates with poor overall survival (OS) along with an immunosuppressive phenotype in bone mHSPC. As prior data show that PI4KA expression is signi cantly higher in mHSPC samples, and its expression contributes to intraosseous growth in intratibial bone tumor model probably through the activation of tumor cell proliferative pathways in mHSPC. To determine the impact of PI4KA expression on clinical progression of the PCa, overall survival of the patients was evaluated based on tumor expression levels of PI4KA. The Kaplan-Meier graphs show poor OS in cohorts of bone biopsies, with high PI4KA and its upstream regulator CXCR4 expression (Fig. 6A), and this is also signi cant with the PI4KA adaptor proteins participating in the membrane localizations such as TTC7, EFR3 and FAM126 proteins (Fig. 6B). Among all the PI4P producing enzymes high tumor PI4KA and PI4K2B expression is associated with poor overall survival in mHSPC patients, whereas the expression of other two PI4P producing enzymes PI4KB and PI4K2A do not have statistical signi cance. This is not the case in soft-tissue biopsies, as no signi cant changes in OS was noticed with the expression of these genes ( Supplementary Fig. 6).
We further characterized the immune landscape of bone and soft tissues using differential gene expression analysis platform (CIBERSORTX) [33,34] through their LM22 signature. Analysis showed that metastatic bone biopsies showed a more immunosuppressive phenotype with the macrophage pro le of a higher presence of non-activated macrophages (M0) and immunosuppressive macrophages (M2), while soft-tissues showed a more immune-active phenotype with an increased presence of immune-active macrophages (M1). CD4 + T-cells where shown to be of higher expression in their unprimed, naïve state in high PI4KA expressing bone biopsies, along with T-cells gamma delta ( Supplementary Fig. 7). The latter cell type gamma-delta is known to have both pro-tumor and antitumor effects depending on the subset of gamma-delta [35], but this CIBERSORTX platform does not allow us to differentiate between the subsets. The overall trend shows bone biopsies with a higher presence of naïve and immunosuppressive cells. Analysis of low and high PI4KA expressing metastatic bone biopsies show that high PI4KA expressing bone biopsies showed a more non-activated, immunosuppressive phenotype, with increased expression of M0 and M2 macrophages. Although the expression of M1 immune-active macrophages were similar between low and high PI4KA biopsies, the presence of M0 and M2 makes the environment more irresponsive to any anti-tumor immune activity (Fig. 6C). Interestingly other immune cells of signi cance in the pro ling, such as mast cells, natural killer cells (NK), and T-cells CD4 memory cells, were of higher expression in their resting state in high PI4KA expressing biopsies (Fig. 6C).
The selective presence of immunosuppressive cells in high PI4KA-expressing tumors implies that PI4KA activity contributes to tumor growth, and anti-PI4KA therapies have the potential to reverse suppressive phenotype in metastatic prostate cancers.

Discussion
Here we characterized PI4KIIIα and its adaptor proteins as interactors of CXCR4 mediating CXCR4-CXCL12 induced cellular invasion and proliferation, conferring poor clinical outcomes in mHSPC patients. We hypothesized that PI4KIIIα binds to CXCR4 through its adaptor protein, and promote downstream signaling, thus playing a role in CXCR4-CXC12 mediated tumorigenicity through invasion and proliferation.
The initial proof of concept that, PI4KIIIα interacts with CXCR4 through its adaptor proteins EFR3B and TTC7B was supported by transfection followed by coimmunoprecipitation studies in cos-7 cells using transient expression in our previous study [21]. Here, we asked whether this interaction is present in prostate cancer cells and drives chemokine signaling. The data support this concept that in prostate cancer cells, endogenous interaction is present between CXCR4 and PI4KIIIα adaptor protein TTC7B. This interaction is observed with three different PCa cell lines with distinct genetic makeup with respect to AR and TMPRSS2-ERG fusion status. Thus, this endogenous interaction appears to be common among prostate cancer cells. Based on the immune precipitation data, we show that TTC7B binds to PI4KIIIα in the cytosol and upon CXCR4 activation this complex is recruited to the plasma membrane and binds to CXCR4.
Presence of complex formation between CXCR4 and TTC7B in intact cells also determined with proximity ligation assay. Our data also show that EFR3B also binds with CXCR4 (Ref 21 and Supplementary Fig. 1), as EFR3 recruitment of PI4KIIIα is established earlier; together our data support the model that TTC7B and PI4KIIIα complex recruited to the CXCR4 and EFR3B complex on the plasma membrane, thus, CXCR4-EFR3B serves as a docking place for TTC7B-PI4KIIIα complex.
Next, we investigated functional signi cance of CXCR4-TTC7B interaction in promoting PI4KIIIα activity in PM. We used a novel PI4P probe by Hammond et al, that utilizes the P4M domain biosensor to detect the PI4P lipids, from L.pneumophila SidM [31]. This probe was characterized for its high speci city and e ciency in localizing to PI4P in live cells, compared to prior probes, and was shown to detect localization in organelles such as PM, Golgi and late endosomes. Another advantage of this probe for use in our study is that the probes are already validated [31] to have just the appropriate amount of a nity to PI4P, in order to detect the uctuating PI4P abundance on the PM. In our study the CXCL12 activated CXCR4 is widely studied to be on the PM, so we focus on the uorescence uctuations on the PM when studying this CXCR4-PI4KIIIα crosstalk. Chemokine CXCL12 induced a signi cant increase in the PI4P levels in PM presumably through TTC7B-P4KIIIα localization to PM. This increased localization to PI4P is abrogated in the presences of CXCR4 inhibitor and as expected speci c PI4KIIIα inhibitors also abrogated SidM binding to PI4P suggesting chemokine signaling selectively activating PI4KIIIα activity on PM.
Our study contributes a new insight into the signi cance of CXCR4 induced activation of PI4KIIIα. CXCR4 overexpression enhanced the PI4P levels in the invasive projections in prostate cancer cells and abrogation of production through knockdown of PI4KIIIα leads to inhibition of chemokine induced PC cell invasion. Several members of chemokine receptors were shown to bind with PI4KIIIα adaptor proteins and translocate PI4KIIIα to PM to regulate its activity at PM. The common theme for chemokine receptor activation of PI4KIIIα is the binding to its adaptor protein for local production of PI4P for cellular invasion [21]. In support of this concept, we show that knockdown of TTC7B leads to inhibition of PI4KIIIα interaction with CXCR4 in plasma membrane and subsequent inhibition of its recruitment to PM. Similar to PI4KIIIα knockdown, TTC7B knockdown also signi cantly inhibited cellular invasion of prostate cancer cells suggesting that TTC7B adaptor mediated recruitment of PI4KIIIα to CXCR4 is critical for chemokine induced PC cell invasion. In addition, other chemokine CXCL8 induced cell invasion also inhibited in TTC7 knockdown cells suggesting TTC7B adaptor function regulating PI4KIIIα activity at membrane is common to other chemokine signaling. Previous studies including ours demonstrate that CXCR4 signaling contributes to prostate cancer cell colonization to bone tumor microenvironment and interact with osteoblastic niche for establishment of PC bone metastasis, herein, our data also demonstrate that PI4KIIIα knockdown signi cantly inhibited bone tumor growth suggesting CXCR4 downstream signaling through PI4KIIIα activation in PCa cells contributes to bone tumor growth. These observations are inline with critical cellular function of PI4KIIIα signaling in plasma membrane in maintaining PI(4,5)P 2 synthesis, localizing certain PM associated proteins and maintaining PM cholesterol content[26].
Here we provided the clinical signi cance of PI4KA expression in prostate cancer. Our previous study demonstrates that PI4KA expression is higher in metastasis compared to primary prostate cancer and in addition we showed that PI4KA expression is higher in metastasis with matched primary cancer tissue from patients [21]. Here in with analysis of two other data sets (GSE6919 and Ref. 41) we show that the level of PI4KA expression correlates with PCa metastatic tumor and neuroendocrine differentiation phenotype. Altogether, the tumor expression data suggest PI4KA is associate with aggressive forms of PCa. These ndings support our DEG analysis that identi ed metastatic bone biopsies expressing high PI4KA, to be enriched in cell proliferative pathways, potentially contributing to the metastatic tumor growth. The relevance of PI4KA is also evident from metastatic samples from hormone sensitive PCa patients, that exhibit poor OS and PSA progression in bone biopsies with higher expression levels. The COX regression analysis also showed signi cance in genes associated with PI4KA to be implicated in poor OS in bone biopsies. PCa also has a highly immuno-suppressive environment due to a multi-factorial regulatory, pro-tumorogenic and immunosuppressive environment, contributing to broad mechanisms of resistance [19,[36][37][38][39][40] C42B and PC3 stable, lentiviral generated cell lines were maintained in RPMI-1640 (Gibco-Invitrogen-Life Technologies) supplemented with 10% heatinactivated FBS (Hyclone, Fisher Scienti c), 1% P/S (50 units/ml penicillin, 50 ug/ml streptomycin, Gibco) and appropriate selection antibiotics (40ug/ml blasticidin S for PC3-RFP and PC3-CXCR4 overexpressing cells; puromycin at 2ug/ml for PC3 scrambled-shRNA or 24ug/ml for PC3-CXCR4 shRNA knockdown cells; 40ug/ml blasticidin S and 0.35ug/ml puromycin for PC3-CXCR4 overexpressing and PI4KA or scrambled shRNA knockdown cells). Cell cultures were performed at 37°C with 5% CO2. All cell lines were authenticated with STR analysis (Genomics core at Michigan State University, East Lansing, MI) and shown to have markers respective for each cell line as established by ATCC, and were tested for mycoplasma contamination with Venor-GeM mycoplasma detection kit (Sigma Biochemicals, St. Louis, MO).
Lentiviral generation of stable cell-lines. Stably transduced PC3-CXCR4 and C4-2B cells with a knocked-down (GIPZ shRNA-PI4KA lentiviral construct) of PI4KA gene were produced using a Trans-Lentiviral Packaging Kit (Thermo-Fisher Scienti c) according to manufacturer's protocols. Brie y, GIPZshRNA-PI4KA lentiviral construct targeting the two independent sequences (#27 TAG ATC TCC AGT TGG CCA C (NM_058004 : 4660-4678) and #30 TCA CTA ACT CCA CAT CGC T (NM_058004 : 5516-5534)) of PI4KA mRNA (NCBI Reference Sequence: NM_058004.3) were obtained through GE Dharmacon (Lafayette, CO 80026)/Wayne State University Biobank Core Facility and used in a similar manner to transduce cells with infectious, replication incompetent lentiviral particles to generate stable PI4KA-knockdown cells using puromycin for selection of stable clones. Two clones were further characterized for PI4KA and knockdown and used in subsequent experiments.
Western Blot analysis. Total cellular proteins were extracted using RIPA buffer with 1x Protease inhibitor cocktail (Roche, Indianapolis, IN). Protein was quanti ed using BCA protein assay (Pierce Biotechnology, Rockford, Il). Western blotting was performed using SDS-PAGE with gel transfer to a nitrocellulose membrane. Membranes were blocked in 5% BSA, probed with primary antibody in 5% BSA, and with secondary antibody linked with horseradish peroxidase, in 5% BSA. Enhanced chemiluminescence (ECL) substrate and autoradiography lm was used to detect proteins. Densitometry was performed using image J software.
Immunoprecipitation. Prostate cancer cells were grown in their respective complete media till 70% con uency. Transient transfections were performed if showing native interactions with over-expressions using 15ug of plasmid and 15ul Lipofectamine2000 Transfection Reagent (Life Technologies, Invitrogen) in 6mls Opti-MEM media (Gibco). Cells were serum starved overnight, washed with PBS and treated with ligand CXCL12 (Peprotech, nal [200ng/ml]) for 10 minutes or left untreated. Cells were then lysed with 500ul/100mm-plate RIPA lysis buffer. Lysates were rotated in 4C for 15 minutes, and centrifuged at 15,000rpm for 15 minutes. The supernatants were used to determine protein concentration using BCA protein assay kit. (ThermoFisher Scienti c). 400-600ug of protein lysate were rotated with 4ug of antibody (CXCR4 AB1846 Millipore) at 4C overnight, followed by rotation with 40ul of Pierce Protein A/G agarose (ThermoScienti c) next day for 2 hours at 4C. The samples were centrifuged at 5000rpm for 30secs, beads were washed 3 times with RIPA wash-buffer and resuspended in denaturing sample buffer. The input samples along with the immunoprecipitation samples were heated at 100C for 5 minutes and immunoblotted as per Western blot analysis protocol.
PI4KIIIα lipid kinase assay. In vitro PI4KIIIα lipid kinase assays were performed as described earlier (22). Post kinase assay the chloroform-extracted PI(4)P product was separated by thin-layer chromatography (TLC) in n-propanol-2M acetic acid (65:35 v/v). PtdIns was visualized with I 2 vapor following PI(4)P detection through autoradiography. PI4KIIIα activity was set as one-fold in control PC3 cells (PC3 scr and PC3-RFP) and compared with CXCR4 manipulated cells.
Cell proliferation and invasion assays. For cell invasion 24-well 8um transwells (Falcon) were coated with 37.5ug Matrigel per insert, cells were seeded on the top of the chamber in serum-free media, along with chemo -attractants in the lower chamber in serum-free media. After 24 hours cells were stained with 0.9% crystal violet and imaged for quantitation.
Fluorescence microscopy. Cells were plated on coverslips coated with poly-L-lysine (Sigma) in a 6-well plate, and transfected with 2.5ug of respective plasmids with lipofectamine. Cells were serum starved overnight; treated with either 2uM GSK-F1 or 4ug/ml AMD-3100 for 2 hours and then ligand-induced with CXCL12 (200mg/ml, Peprotech). After treatment, cells were xed with 4% PFA with 0.2% glutaraldehyde at room temperature for 15 minutes. After aspiration, cells were further incubated with 50mM NH4Cl in PBS and washed for 10 minutes, 3 times. Then cells were thoroughly washed with water, and mounted on slides using Vectashield with DAPI. Cells were imaged using a Leica DMi3000 B uorescence microscope.
Proximity Ligation Assay. Cells were plated on chamber slides and serum starved overnight. After ligand induction with CXCL12 (200ng/ml), cells were xed with 4% PFA and permeabilized with mild buffer 0.01% Tween-20 in PBS. The cells were further treated as per the Sigma Duolink DUO92101-1 kit. CXCR4mouse and TTC7B-rabbit antibodies were used for primary incubation and observed for the presence of co-localization of CXCR4 and TTC7B using a uorescence microscope.
Intra-tibial bone tumor growth assessed by luciferase imaging.
For in-vivo studies PC3-CXCR4 Scrambled shRNA Control and PC3-CXCR4 PI4KA shRNA (#27 and #30) cells were infected with Luciferase-2 lentiviral particles for in-vivo bioluminescence images of tumors. Castrated four to ve-week old male Nude mice (Taconic Farms, Germantown, NY) were used in this study. All the procedures including animal housing, surgery, imaging, the methods of anesthesia and euthanasia prior to tumor tissue analysis were performed as per the institutional animal care and use committee approved protocol. For intratibial implantation, 1.0 x 10 6 cells in a 10 µl volume were injected per bone.
Animals were imaged periodically to measure bone tumors. Luciferase imaging of tumors were performed with Carestream Invivo Xtreme system.
Patient and Clinical data. Pre-biopsy samples from patients with metastatic hormone-sensitive prostate cancer (mHSPC) were used for the clinical analysis in these studies. These patients participated in a multi-center trial conducted in 4 different centers in the US. An informed consent was obtained from all the patient prior to their participation in the trial. These men have no history of seizures, and have adequate marrow, renal and liver function, with a median age of 65. Clinical outcome data of the PSA response rate and Overall Survival statistics were obtained from this study for our analysis (Clinical trial: Identi er: NCT02058706). Total RNA was extracted using the RNeasy midi kits (Qiagen) along with ON-Column DNase digestion (Qiagen), as described below, and sequenced at University of Michigan Genomics core. A total of 39 bone biopsies and 23 soft-tissue biopsies were used in this study after RNA quality was con rmed, and RNA-sequencing (RNA-seq) was performed. The GSEA analysis was performed on the resulting RNA-seq TPM data. Cibersortx machinelearning tool was also utilized to identify immune expression pro le using the LM22 signature.
Gene Set Enrichment Analysis. Pathway enrichment was performed using GSEA software (version 4.1). The Hallmark pathway dataset was downloaded from the MsigDB database of the GSEA website. The median high vs low-PI4KA expression pro le of biopsies from mHSPC patients, and the attribute les were enriched and analyzed by default weighted enrichment statistics. The number of permutations was set to 1000.
Gene Expression Omnibus Database. GSE6919 from the public database of GEO-NCBI was used to analyze expression pro les from tissue biopsies that represent prostate cancer progression from normal tissue free of any alteration, normal tissue adjacent to primary tumor, primary tumor to metastatic prostate tumor samples. The total 504 samples from these different states were extracted, by downloading the platform and matrix les from the GSE6919 database. Data in the les were analyzed using GraphPad Prism 6.
Statistical Analysis. GraphPad Prism 6 and R were used to assess statistical signi cance. Distributions of continuous outcomes were checked for normality and, if needed, non-parametric approaches were used. Comparisons between two groups were performed using unpaired t-test or Mann-Whitney test. For three or more groups, one-or two-way ANOVA was used, followed by Tukey's post-hoc pairwise comparisons. Survival outcomes were graphically summarized using Kaplan-Meier curves, and a log-rank test and Cox regression analysis were used to compare between groups. If a P value was less than 0.5, it was considered statistically signi cant. *: P value < 0.05. ** P value < 0.01. *** P value < 0.001. NS stands for "not signi cant". Data availability:

Declarations
The data sets generated during this study are in the process of depositing into the NCBI/GEO resource. Once this process is completed the unique geo identi er will be provided to the journal. The other data sets are available from the corresponding author on reasonable request.   Adaptor protein TTC7B interacts with CXCR4 and also regulates invasion in PC3-CXCR4 prostate cancer cells. A) TTC7B knock-down using siRNA (100nM) in inputs and pull-down of CXCR4 in co-immunoprecipitation assays, showing immunoblots of A) TTC7B B) PI4KIIIα and C) CXCR4. D) Matrigel invasion assays after treatment with TTC7B siRNA or Scramble siRNA (10x; upper and lower panels), under basal and ligand-induced CXCL12 (200ng/ml), CXCL8 (50ng/ml) conditions. (E,F) Graphs represent mean and standard error of triplicates. Comparisons were performed usingtwo-way ANOVA followed by Tukey's post-hoc comparisons (ns, p≥0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001, in comparison to no ligand control).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformation.docx