Loss of Neuroligin 4X Induces An Intrinsic Innate Immune Response in TNBC

Background Immune checkpoint blockade therapies, which act on T cell inhibitory receptors, including CTLA-4 and PD-1, induce durable responses across diverse cancers. However, most patients do not respond to these therapies, and initially responsive cancers may relapse. Identifying molecular mechanisms that influence therapeutic responses and resistance is critical to realize the full therapeutic potential of immune checkpoint inhibitors. The presence of immune infiltrates in the tumor microenvironment is associated with positive outcomes in breast cancer, specifically in triple-negative breast cancer (TNBC). The underlying mechanisms driving this response are unclear. We have previously identified Neuroligin 4X (NLGN4X) as a protein expressed in TNBC. Methods Bioinformatic analysis was used for pathway analysis of TCGA TNBC patient dataset. Immunohistochemistry was performed on breast cancer tissue microarray for NLGN4X protein expression. RNA-seq was performed on MDA-MB-231 breast cancer cells for differential gene expression upon gene knockdown. Cytokine array, western blot, cell adhesion array and Nanostring was performed to determine the role of NLGN4X in TNBC. Conclusion Loss of NLGN4X leads to innate immune activation in breast cancer and coincides with an aggressive phenotype of cancer. This study identifies the role of NLGN4X in regulating interferon signaling and immune microenvironment in TNBC.


Introduction
Clinical evidence has shown that some breast tumors, including the subset of TNBC, elicit a robust antitumor immune response [1,31]. TNBC tumor have high infiltration of immune cells including CD8+T, CD4, macrophages, B cells and neutrophils. Patients with high intratumoral infiltrations of CD8+ cells responds well to immune checkpoint inhibitors (ICIs) and has better disease-free survival demonstrating a better disease-free survival [4,22,32]. The attraction of these immune cells is due to chemokines which mediate immune cell trafficking towards the tumor [30]. Among these chemokines, elevated levels of the C-C motif chemokine ligand 5 (CCL5), C-X-C motif chemokine ligand 9 and 10 (CXCL9 & CXCL10), are frequently associated with increased recruitment of CD8+ T lymphocytes to tumor sites [23]. Cancer cell intrinsic mechanism play an important role in shaping tumor immune microenvironment. Studies have indicated that activation of DNA sensor, cGAS-STING [7,10] and RNA sensor, retinoic acid inducible gene I (RIG-I) (encoded by DDX58) [9,11] in tumor cells contributes to regulating antitumor immunity and responsiveness to checkpoint blockade. Tumor subtype, aberrant expression of oncogenes and mutational load highly impact the tumor microenvironment. For example, a recent study identified a DNA damage response-deficient (DDRD) molecular subtype in breast cancer. This subtype was characterized by upregulation of Interferon Stimulated Genes (ISG) and lymphocyte infiltration in part due to inherent defect in DNA repair mechanism and activation of S phase STING [26]. Similarly, another study demonstrated that loss of DDX3X leads to activation of antiviral innate immune response by aberrant cytosolic accumulation of endogenous dsRNAs in the breast cancer cells, which triggers intrinsic type I IFN production via activation of cytoplasmic dsRNA sensing pathway [6]. Activation of these sensors can enhance the immunogenicity of otherwise poorly immunogenic tumors. A recent study showed that RIG-I activation in breast cancer cells resulted in tumor cell-intrinsic expression of inflammatory cytokines, leukocyte-recruiting chemokines, and increased expression of major histocompatibility (MHC)-I components [12]. Grusso et al., identified defined distinct tumor immune microenvironment (TIME) by integrating spatial resolution of immune cells with laser capture microdissection gene expression and characterized fully inflamed (FI) tumors by low cholesterol/high IFN signaling signature. Patients with FI TNBC have the best outcomes with standard-of-care chemotherapies and immune checkpoint blockade (ICB) [17]. However, molecular mechanism regulating the recruitment of these immune effectors to TME as well as in the tumor epithelium are not well understood.
Neuroligins constitute a family of neuronal transmembrane synaptic proteins whose structural and biochemical characteristics are indicative of a role in heterotypic cell adhesion. The neuroligin gene family consists of five members (NLGN1 at 3q26, NLGN2 at 17p13, NLGN3 at Xq13, NLGN4 at Xp22, and NLGN4Y at Yq11) [18,39]. Their large extracellular N-terminal domain is homologous to serine esterases.
They are of great importance in mediating synapse formation in the central nervous system, and they interact with neurexins from the opposite side (in trans) of the synaptic cleft in a calcium-dependent manner [24]. Neuroligins have also been implicated in vascular remodeling during angiogenesis [16,29]. Recent studies have shown a mitogenic effect of NLGN3 in glioma proliferation and progression [36].
Our previous the first study to link the expression of neuronal cell adhesion molecules, neuroligins, in breast cancer [18]. NLGN4X plays a vital role in enhancing tumorigenesis and aggressiveness of breast cancer with MCPH1 loss/mutation [35]. Here we describe for the first time that loss of NLGN4X activates MAVS-IRF signaling that induces IFN signaling and cytokine secretion from TNBC. NLGN4X expression negatively correlates with ISG and T-cell attractant cytokines in the patient dataset. Nanostring and immunohistochemistry (IHC) analysis of tissue microarrays from breast cancer patients shows high immune infiltration in NLGN4X-low breast cancer samples. This suggest that NLGN4X may constitute a novel biomarker of "fully inflamed" tumors.

Cell lines and Reagents
Fetal calf serum (FCS) and cell culture media (Dulbecco's modified Eagle's medium, DMEM)

Western Blot
Cells were lysed in ice-cold complete 1x RIPA buffer (PMSF solution, sodium orthovanadate solution, protease inhibitor cocktail solution, and 1x lysis buffer) (Santa Cruz Biotechnology, Santa Cruz, Ca). The proteins were then quantified using the BCA Protein Assay Kit (Pierce Biotechnology, Rockford, IL). 25 μg of protein from each sample was separated by a 4-12% SDS-PAGE gel and then transferred to a 0.2 μm polyvinylidene difluoride (PVDF) membrane. Membranes were first blocked with 5% nonfat dry milk in TBS-T and incubated overnight with primary antibodies (Rig 1, MDA5, MAVS, Anti IRF 7, IFIT3, Slug, P-SMAD2/3, Zeb1, Zeb2, Snail1) at a dilution of 1:1000). After washing, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibody for 1 hr. Membranes were washed and blots were visualized using enhanced chemiluminescence. To verify that equal amounts of protein was loaded from each sample, the membrane was stripped with mild stripping buffer and reprobed with GAPDH (Abcam). Finally, the antibody signals were detected by the ECL system on Amersham imager 680.

Cell adhesion assay
Assay for ECM cell adhesion array was carried out as specified by the manufacturers (ECM540-Millipore).
The plate was rehydrated and incubated for 10 minutes at RT. 1 X 10 6 cells/mL were added to the plate and incubated for 2 h at 37 o C in a CO2 incubator. After extensive washing with assay buffer, cells were stained with cell stain solution and incubated for 5 minutes at RT. 100 µL of extraction buffer was added and incubated for 10 minutes at RT by gently rotated on an orbital shaker after repeated wash with deionized water. Finally, the absorbance was read at 540 nm on a microplate reader.

Cytokine antibody arrays
Assay for cytokine antibody arrays was carried out as specified by the manufacturers (ab133997-Abcam).
Briefly, conditioned media prepared from MDA-MB-231 and MDA-MB-468 SCR and NLGN4X-KD cells was collected and incubated with the detection antibody cocktail. The sample/antibody mixture was then added onto the blocked membrane, containing 42 different capture antibodies. After washing, the membrane was incubated with diluted Streptavidin-HRP and Chemi Reagent Mix was added. The cytokines signals were detected by the ECL system on Amersham imager 680.

IHC staining
The breast cancer tissue microarrays (TMA) were obtained from Novus Biological (Littleton, CO) and stained as described before [18]. These include 40 infiltrating ductal carcinoma breast cancer, 10 metastatic lymph node and 9 adjacent normal breast tissues cores. The use of these tissues was approved by the Institutional Review Board of Tuskegee University. Briefly, tissues were de-paraffinized and for antigen retrieval of NLGN4X the slides were pressure-cooked for 10 minutes. After blocking in 3% goat serum for 1 hour in humidity chambers with NLGN4X antibody (1:100) (Gentex, CA), the slides were incubated with HRP conjugated goat anti-mouse/anti-rabbit secondary antibody (Jackson Immunoresearch Laboratories Inc, West Grove, PA) for 40 minutes. The antigen-antibody reaction was visualized after treating with diaminobenzidine (Sigma-Aldrich, MO). The slides were counterstained with hematoxylin (Sigma-Aldrich, MO) for 1 minute.

TCGA data collection and analysis
All raw data with attached clinical data in various format (txt, fastq, or bam) of each selected samples from TCGA was downloaded and parse through QC pipeline for low quality before normalization and statistics.
For the fastq format, data were processed by removing adaptor sequences using Trimmomatic tool [3], alignment with HISAT2 [19] algorithm, and removal of PCR replicates using SAMtools [21] before expression level quantification. The resulting BAM or SAM files, or directly downloaded txt files were analyzed through a workflow of Partek Genomics Suite (PGS, St Louis, MO, USA). Reads per kilobase of transcript per million mapped reads (RPKM) normalization [8] was performed. The resulting gene list was used for functional analysis such as signaling pathways by using Ingenuity Pathway Analysis (IPA, Redwood City, CA USA). Tumor samples were grouped into high, medium and low groups based on NLGN4X gene expression levels, then these groups were compared by using ANOVA.

Nanostring analysis
The DSP assay was performed on breast cancer TMA slide sourced from Superbiochips Laboratories through the Nanostring Technology Access Program.  Table I). Each selected ROI measured 600µm and oligonucleotides released from the assay antibodies through UV-light exposure. Oligonucleotides released from individual ROI were aspirated, deposited into single wells of 96-well plates and hybridized to barcode-labelled probes [14]. Hybridized probes were processed and counted using the nCounter Prep Station and nCounter Digital Analyzer [2].

Statistics
The significance of difference between two variables was assessed by the Student's t test. The difference was considered significant if the p value was <0.05. Data from all experiments are expressed as mean ± standard error mean (SEM). All statistical calculations were performed using Microsoft Excel.

Neuroligin 4X expression decreases in higher grades primary tumor
Bioinformatic analysis of TCGA dataset showed that shallow deletion of NLGN4X with increasing grade of tumor as shown in Fig. 1A. IHC analysis of the NLGN4X antibody in a TMA all stages of breast cancer with lymph node metastasis and normal adjacent breast tissue) also revealed its decreased expression in tumors with increasing stage and nodal metastasis. More specifically, in Fig. 1B, NLGN4X expression was higher in T2N0M0 as compared to T2N1M0 or T2N3aM0. Interestingly, when we compared NLGN4X expression in the primary tumor and its matched lymph node metastasis, the loss NLGN4X in the highergrade primary tumors, was re-expressed in the lymph node metastatic tissues as show in. Thus, the loss of NLGN4X in expression in higher grade primary tumor coincides with aggression and spread of cancer and its expression in lymph node metastasis suggests that its re-expression is required for subsequent cell adhesion and tumor progression.

Neuroligin 4x knockdown decreases adhesion and induces EMT
To determine if NLGN4X loss does promote dissemination, we used siRNA to silence NLGN4X in two aggressive TNBC cell lines, MDA-MB-231(EMT-type) and MDA-MB-468 (epithelial-type) ( Fig. 2A) and analyzed the ability of the NLGN4X-siRNA treated cells to adhere to ECM as compared to NLGN4X-SCR using ECM adhesion assay kits. As shown in  2C). TGFβ has been shown to induce EMT in breast cancer [38]. To determine if induction of EMTmediating genes is via the TGFβ signaling pathway, smad2/3 phosphorylation was measured in the NLGN4X-KD or SCR treated cells. Smad2/3 was highly phosphorylated in MDA-MB-468-NLGN4X-KD cells compared to SCR while there was no change in MDA-MB-231 cells.

Loss of Neuroligin 4X increases the expression of genes involved antiviral innate immune response in TNBC
To assess the impact of NLGN4X on the overall gene expression in breast cancer cells, the transcriptome of MDA-MB-231 NLGN4X-KD cells was analyzed using RNA sequencing. There were more than 500

Inhibition of NLGN4X activates MAVS-IRF7 signaling axis
Next, we investigated which signaling pathway is responsible for type I IFN production in the NLGN4X-KD breast cancer cells. Activation of pattern recognition receptors, Toll8 like receptor family proteins (TLR3, TLR7), melanoma differentiation associated gene 5 (MDA5) (encoded by IFIH1), RIG-1 have been shown to induce type I IFN production [25,27]. We examined the expression RIG-1, MDA5 and their downstream effector, MAVS in the MDA-MB-231 and MDA-MB-468 SCR and KD cells. The protein expression of MDA5 and RIG-I were significantly increased in the MDA-MB-231 NLGN4X-KD cells as compared to SCR treated cells (Fig. 4, A). There was no change in the protein expression of RIG-1 and MDA5 in MDA-MB-468 NLGN4X-KD cells (Fig. 4, B). However, the proteins levels of MAVS were significantly increased in MDA-MB-231 NLGN4X-KD and MDA-MB-468 NLGN4X-KD as compared to SCR (Fig. 4). Next, we investigated if there were changes in the transcription factors (IRF3, IRF7, NFκB, TBK1) responsible for IFN signaling in NLGN4X-KD cells. There was no change in the phosphorylation of TBK1, NFκB or IRF3 in NLGN4X-KD cancer cells (data not shown) but there was a significant increase in IRF7 levels in both MDA-MB-231 NLGN4X KD and MDA-MB-468 NLGN4X KD cells as compared to SCR. (Fig. 4). Consistent with the role of MAVS-IRF7 signaling in type I IFN induction [5,20], as mentioned above the protein levels of IFIT3 (Fig. 4)

Low NLGN4X level is associated with chronic activation of the innate antiviral immune response in TNBC patients.
We next assessed the transcriptomes of TCGA TNBC patients with varying NLGN4X expression and analyzed 2000 differentially expressed genes in NLGN4X hi versus NLGN4X low groups (adjusted P < 0.05) (Fig. 6A), and found that NLGN4X mRNA was negatively associated with genes controlling IFNγ signaling, cytokine signaling, antigen presentation (Fig. 6B). Pearson correlation of NLGN4X negatively correlated with the ISG (IFITM1, IRF, DDX58) and chemokines (CCL2, CCL7, CCL5, CXCL11) (Fig. 6 C). NLGN4X negative correlation with ISG and chemokine expression in TCGA TNBC patient dataset is consistent with the above described invitro data where NLGN4X KD induces IFN signaling and chemokine secretion.

Low NLGN4X expressions corresponds to immune infiltration in breast cancer
To validate the above invitro and patient dataset findings, we explored the correlation of NLGN4X expression with the immune markers in breast cancer tissue on tissue microarray. For this, we employed nanostring DSP technology [33,37] using 40-panel markers that included markers for T cells, B cells, macrophages, Treg cells, PI3K-Akt signaling to interrogate differential protein expression in tumor epithelial and stromal compartments between patients who had high NLGN4X expression vs low NLGN4X expression. Each core was divided into epithelium (panCK positive) and stroma (panCK negative). Cores with low/absent NLGN4X expression had significant more infiltration of immune population in the stroma (panCK negative) as well as in the tumor epithelium (panCK positive) (Fig. 6 C, D) as compared to NLGN4X-high cores (Fig. 6 A, B). The volcano map of these cores shows differential expression of the target genes. The genes that were significantly upregulated in the stroma of the NLGN4X low core were B Borderline significant genes were myeloid/macrophage (CD68, CD68b, CD11c, ARG1), B cell (CD20), memory T cell (CD45RO), pan-RAS and EpCAM while in epithelium of NLGN4X high core were ERalpha and INPP4B (Fig. 7F). This data suggests that TNBC with low NLGN4X expression recruit large number of immune-regulatory cells in the stroma as well as in the tumor epithelium and becomes "fully inflamed".

Discussion
As the cancer progresses to a higher stage, it becomes more mesenchymal, aggressive, invasive and metastatic. While a plethora of factors has been implicated in the induction of EMT and increase in invasiveness, the role of NLGN4X in this context in TNBC is unknown. Here, we show that NLGN4X The findings in this study provides insight into the intrinsic mechanism in TNBC that regulates IFN signaling and can shape the TIME. Studies have shown the role of DNA damage and cGAS-STING mechanism in the induction of IFN signaling in the TNBC [10]. Other studies have also shown the role of dsRNA sensors include TLR3, TLR7, RIG-I, MDA5, and protein kinase R (PKR) in antiviral innate immune response [34]. The activation of these DNA/RNA sensors has been shown to enhance anti-tumor immunity via induction of cancer-derived type I IFN. Tumor cell-intrinsic type I IFN response has been achieved through the intracellular dsRNA transcription by using a CDK4/6 inhibitor [15] or DNA methylation inhibitors [28]. Our data shows that reducing the expression of NLGN4X enhances cancer cell-intrinsic type IFN response. GSEA analysis of DEGs from the RNA-seq analysis data of NLGN4X- population as compared to the low. This was also supported by the nanostring DSP analysis using a panel of 40 antibodies that included T cell, B cell, macrophages, Treg's and neutrophils showed that low levels of NLGN4X in metastatic breast cancer cores coincides with high immune infiltration not only in the stroma but also in the tumor epithelium, enabling the tumor as "fully inflamed" while the NLGN4X positive tumors had negligible immune infiltration. The data also suggest that while there was CD8+ infiltration, there were myeloid suppressor cells, Treg's and checkpoints that can be targeted in NLGN4X-low population to increase the effectiveness of ICB. Furthermore, the tumor epithelium was also enriched in myeloid/macrophage markers (CD163, CD14, CD68, ARG1) suggesting "myeloid mimicry" of tumors which has been shown to evade immune invasion in some tumors [13]. With this data, we are tempted to speculate that NLGN4X loss enhances cell-intrinsic innate antiviral immunity that leads to IFN signaling and cytokine secretion and recruitment of immune cells to the tumor site.

Conclusion
In the summary, we have probed into the molecular mechanism and significance of NLGN4X on the immune microenvironment and breast cancer progressiveness. For the first time, we have showed the role of a neuronal cell-adhesion molecule in regulating IFN signaling in cancer cells. Loss of NLGN4X induces an anti-viral state by activating MAVS-IRF7 signaling pathway in TNBC cells. This data is supported by bioinformatic analysis of TNBC patient dataset and Nanostring analysis of breast cancer TMA using 40 panel of immune markers, suggesting that Neuroligin 4X is a negative regulator of IFN signaling in TNBC and its high expression creates an immunosuppressive environment. For further work, we will perform invivo studies to determine how loss of NLGN4X affect immune signaling, immunotherapy response and metastasis and invitro experimentation to identify the molecular mechanism by which loss of NLGN4X induces MAVS-IRF7 signaling. There is an unmet demand for the development of biomarkers that can predict immunotherapy response or enable enhance the current immunotherapy regimen. The absence/loss of NLGN4X can be studied further as a biomarker of "fully inflamed tumors" that can benefit from immunotherapy by ICB.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.            NLGN4X KD induces cytokine secretion in breast cancer cells. Human cytokine antibody array was performed following the manufacturer's protocol to detect secreted chemokines from MDA-MB-231 cells that were a. treated with scramble or, b. NLGN4X-siRNA , and MDA-MB-468 cells, c. treated with scramble or, d. NLGN4X-siRNA. Red boxes designate the cytokines whose expression was increased by NLGN4X-siRNA treatment as compared to SCR. The experiments were performed three times. MCP1-CCL2, MCP2-CCL8, MCP3-CCL7.