Abstract Background Patients with human epidermal growth factor receptor 2 positive (HER2+) or triple-negative breast cancer are at higher risk than those with other breast cancer subtypes for developing brain metastases, including leptomeningeal metastases. Despite advances in the treatment of primary breast tumors, the prognosis of breast cancer patients with brain metastases remains poor. Frequent and sensitive monitoring of response to treatments is crucial to optimize treatments for brain metastases. Multimodal molecular analyses of cells in cerebrospinal fluid (CSF) may provide a less invasive approach than biopsy for monitoring disease in the central nervous system. Methods A 41-year-old female with refractory leptomeningeal metastases from HER2 + breast cancer had serial sampling of cerebrospinal fluid (CSF) from an Ommaya reservoir prior to, and after, the addition of sacituzumab govitecan to her treatment regimen. Cytopathologic clinical analysis, circulating tumor cell enumeration, single-cell RNA sequencing, and cytokine profiling assays were performed on collected samples. Results A reduction in CSF tumor cells and absence of HER2-expressing cells was observed following treatment, with MRI confirming decreased tumor size. Single-cell RNA sequencing revealed diverse tumor and immune cell populations in CSF, highlighting gene expression signatures associated with breast cancer and triple negative disease and dynamic changes in cellular composition throughout the course of treatment. Cytokine profiling identified increased pro-inflammatory cytokines post-treatment. Conclusion These findings highlight the potential utility of molecular profiling of CSF as an early indicator of treatment response and an approach for identifying resistance mechanisms and optimal combination therapies. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-025-06671-4. Keywords: Cerebrospinal fluid, Leptomeningeal disease, Breast cancer, Sacituzumab govitecan, Single-cell sequencing, Circulating tumor cells Introduction Breast cancer is the most common cancer among females and the second cause of cancer- related mortality in women worldwide. The American Cancer Society estimated 313,510 new cases for 2024 in the United States alone, with almost 43,000 deaths [[34]1]. Women with triple negative breast cancer (TNBC) have the highest risk of metastasis, including metastasis to the brain. The development of brain metastases significantly limits quality of life through impaired cognitive and sensory functions and is associated with short overall survival [[35]2]. Approximately 30–50% of women with TNBC or human epidermal growth factor receptor 2 positive (HER2+) breast cancer ultimately develop and succumb to brain metastases [[36]3, [37]4]. Treating brain metastases has been particularly challenging due to the unique anatomical and functional features of the brain and genetic differences between primary brain tumors and tumors that metastasize to the brain [[38]5]. Brain metastases often manifest years after the primary tumor. Since metastases and primary tumors develop in two different locations, it is expected that the tumor microenvironment (TME) is different as well as the treatments [[39]6]. Additionally, resistance to therapy often occurs and monitoring tumor biology over the course of the treatment is essential to improve therapeutic response. Diagnosis and management of patients with breast cancer brain metastasis (BCBM) currently rely on magnetic resonance imaging (MRI), cytologic evaluation of tumor cells in the cerebrospinal fluid (CSF), and brain biopsies. However, imaging of brain metastases lacks specificity, CSF-cytology has extremely poor sensitivity, and repeated invasive procedures such as brain biopsies can pose significant risk to the patient [[40]7, [41]8]. Therefore, there is a critical need for more specific and less invasive methods for diagnosing and monitoring patients with BCBMs that can inform aspects of tumor biology that influence treatment decisions. In the era of targeted therapies and molecularly driven clinical decision making, molecular characterization of tumors is fundamental. Liquid biopsies that collect and analyze tumor components in body fluids, including circulating tumor cells (CTCs), cell-free tumor DNA (ctDNA), circulating tumor RNAs (ctRNA), and exosomes, represent a potential solution when a conventional tissue biopsy is not feasible [[42]9]. The use of blood for assessing central nervous system (CNS) malignancies has been challenging due to the low prevalence of CTCs [[43]10]. However, CTCs may be found in higher concentrations in the CSF and can provide fundamental information about the genomic characteristics of brain tumors, representing a relatively non-invasive liquid biopsy [[44]11]. Moreover, it is now demonstrated that CSF is a relevant source to characterize immune cell populations present in the TME of brain lesions and can be used to monitor the evolution of the cancer immune response [[45]12]. We report here multimodal single cell RNA sequencing (scRNA-Seq) profiling, CTC enumeration, and proteomic analysis of cells in the CSF as tools for monitoring a patient with BCBM. CSF represents a potential source to predict and represent tumor progression, with the goal of developing accurate BCBM diagnosis and monitoring of molecular changes during the treatment with limited invasive procedures. This not only could offer more therapeutic options at onset, but could also allows physicians to adapt their treatment regimen quickly to maximize therapeutic response. Results Case description and sample collection A 41-year-old female with recurrent HER2 + metastatic breast cancer presented with leptomeningeal involvement. She was originally diagnosed with ductal carcinoma, HER2+, progesterone receptor negative (PR-), and estrogen receptor negative (ER-). She received four cycles of neoadjuvant adriamycin/cyclophosphamide chemotherapy and weekly paclitaxel, trastuzumab, and pertuzumab. Her disease was stable until two years later when she had new onset headaches and vomiting. A brain MRI revealed three central nervous system (CNS) lesions and she was treated with cyberknife radiotherapy with 19 Gy single fractions to each lesion. Two months later, the patient had worsening neurological symptoms and a brain MRI revealed previously known right posterior fossa and right medial parietal lesions decreased in size, but new leptomeningeal enhancement in left frontal region indicative of leptomeningeal disease involvement. A lumbar puncture showed malignant tumor cells consistent with breast cancer. She initiated treatment with capecitabine, trastuzumab, and tucatinib. An Ommaya intraventricular shunt was placed for disease monitoring and delivery of intrathecal methotrexate and Ara/C. Serial samples of CSF were collected before and after initiating systemic therapy with sacituzumab govitecan and sent for clinical testing as well as for research analysis (Fig. [46]1A, B). Fig. 1. [47]Fig. 1 [48]Open in a new tab Treatment history, CSF tumor cells, and MRI scan demonstrating tumor response. (A). Overview of the treatment history, including all prior resections, standard therapies, and off-label treatments. (B). Number of cerebrospinal fluid (CSF) tumor cells (left y-axis) and percentage of HER2 positive cells (right y-axis) over time. Purple line represents CSF tumor cells, and light blue line represents HER2 positive cells. Red arrows mark sample collection dates. Both CSF tumor cells and HER2 positive cells decrease following treatment. (C). MRI scans showing the largest tumor diameter before and one month after treatment with Sacituzumab govitecan. Arrows indicate the tumor region, highlighting the reduction in tumor size Cytopathologic reduction in CSF tumor cells and phenotypic characterization CSF was collected approximately 2 and 6 weeks after sacituzumab govitecan was initiated with concurrent intrathecal therapy. Cytopathologic analysis and genomic characterization with CNSide™ revealed a significant decrease in the number of CTCs as well as the absence of cells expressing HER2 at both time points compared to baseline (Table [49]1; Fig. [50]1B). An MRI scan performed approximately 4 weeks after initiating sacituzumab govitecan demonstrated a notable decrease in tumor size (Fig. [51]1C), providing radiographic confirmation reduced tumor burden. Table 1. Patient-level results of cerebrospinal fluid analysis by standard cytopathologic analysis and cnside™, and phenotypic and genomic characterization of CSF-TCs using cnside™ Sample Pre- treatment After 2 weeks After 6 weeks CSF Volume (mL) 9 11 17 TC Enum Result DETECTED DETECTED DETECTED Total TC/mL 1625.67 1.45 1.18 Total TC Count 14,631 16 20 CK + Count 11,527 7 19 CK- Count 3104 9 1 HER2 Result DETECTED NOT DETECTED NOT DETECTED ER Result NOT DETECTED NOT DETECTED NOT DETECTED PD-L1 Result NOT DETECTED NOT DETECTED NOT DETECTED PR Result NOT DETECTED NOT DETECTED NOT DETECTED [52]Open in a new tab ScRNA-seq identified major tumor and immune cell populations in CSF Using single-cell RNA sequencing (scRNA-seq), we embarked on a comprehensive exploration of the cellular landscape within the CSF milieu. Through the unsupervised clustering algorithms followed by UMAP, we characterized distinct cell populations, highlighting the heterogeneity of tumor and immune cell types within the CSF. Notably, our analyses revealed the presence of CD8 and CD4 T cells, B cells, and myeloid cells alongside CTCs, and a distinctive gene expression profile reflective of breast cancer (BC). These BC-associated CTCs exhibited a prominent expression signature encompassing a suite of genes including PTPRF (LAR), KRT8, KRT19, and KRT18, underscoring their malignant potential and implicating them in disease progression (Fig. [53]2A, B, C). Using integrated clustering analysis, we identified distinct cellular populations within the CSF samples, as depicted in Fig. [54]2D, the integrated UMAP plot revealed clusters representing diverse cell types, suggesting heterogeneity within the CSF cell population. Fig. 2. [55]Fig. 2 [56]Open in a new tab Temporal dynamics of CSF single-cell transcriptomes and integrated UMAP visualization. UMAP plots representing single-cell RNA profiles obtained from (A) pre-treatment, (B) mid-treatment or (C) post-treatment CSF samples. Each point on the plot represents an individual cell, with colors indicating sample origin. The plots visualize the transcriptional heterogeneity among cells and the effects of treatment on gene expression profiles. UMAP plots depicting single-cell RNA expression data for cancer genes that were significantly expressed (p-value < 0.05, log2(foldchange) ≥ 1). These plots highlight the expression patterns of cancer-related genes across different treatment timelines, providing insights into the molecular changes associated with treatment response. (D) Integrated UMAP plot displaying clusters of CSF cells identified in three samples. Each cluster represents distinct cellular populations identified through unsupervised clustering analysis Subsequently, we examined the expression levels of genes associated with breast cancer (PTPRF, KRT18, KRT19, KRT8) and TNBC signature genes (CRABP2, TM4SF1, KRT7) during treatment. The expression levels of genes associated with breast cancer markers and the TNBC signature varied across samples. Violin plots depicting these expression dynamics provide insights into the variability of gene expression levels, potentially indicating differential treatment responses or disease subtypes (Fig. [57]3A). Furthermore, we investigated the expression patterns of genes associated with cell death and the cell cycle. As shown in Fig. [58]3B, violin plot demonstrates the distribution of expression levels for cell death-associated genes (PCP4, MORN2, FHIT), while Fig. [59]3C depicts the distribution of expression levels for cell cycle-associated genes (TOP2A, MKI67, CCND1, CDT1) during treatment. The dot plot in Fig. [60]3D illustrates the distribution of gene expression levels across samples from the three time points. This visualization provides a detailed depiction of gene expression dynamics, potentially identifying critical biomarkers or therapeutic targets for further investigation. Fig. 3. [61]Fig. 3 [62]Open in a new tab Treatment-induced gene expression changes. Violin plots depicting the expression levels of (A) breast cancer marker genes (PTPRF, KRT18, KRT19, KRT8), the TNBC signature gene (CRABP2, TM4SF1, KRT7), (B) cell death-associated genes (PCP4, MORN2, FHIT) and (C) cell cycle-associated genes (TOP2A, MKI67, CCND1, CDT1) during treatment. The violin plots illustrate the distribution of gene expression levels across samples, with the width representing the density of expression values. (D) Dot plot illustrates the distribution of gene expression levels across three samples. Each dot represents the expression level of a specific gene in a particular sample. Dots are color-coded to indicate different expression levels After creating heatmaps of differentially expressed genes (DEGs) in each sample (data not shown), the first 20 genes with more than 2-fold change expression levels were analyzed to identify the relevant biological functions and pathways. Ridge plots reveal the expression patterns of the top 20 DEGs across samples (Fig. [63]4A). Each ridge represents the expression profile of an individual gene, with the width indicating the density of expression values. This visualization provides a nuanced understanding of the diverse expression patterns exhibited by the identified DEGs, potentially highlighting regulatory roles in the studied conditions. The functional context in which the identified DEGs operate was also analyzed, underscoring potential biological pathways and biological processes enriched by the top 20 DEGs affected by differential gene expression (Fig. [64]4B, C). Moreover, our interrogation of the scRNA-seq data facilitated the precise classification of distinct cell populations based on their cellular origins, enabling the examination of dynamic changes in cellular composition throughout the course of treatment. Fig. 4. [65]Fig. 4 [66]Open in a new tab The top 20 differentially expressed genes (DEGs), profiling pathway distribution and biological processes enrichment. (A) Ridge plots representing the expression patterns of the top 20 Differentially Expressed Genes (DEGs) across samples. Each ridge represents the expression profile of a single gene, with the width of the ridge corresponding to the density of expression values. (B) This pie chart visualizes the distribution of pathways associated with the first 20 Differentially Expressed Genes (DEGs). Each sector represents a pathway, with the size proportional to the number of DEGs involved. (C) The biological processes enriched by the first 20 Differentially Expressed Genes (DEGs). Each segment represents a biological process, with the area indicating the relative contribution of DEGs to that process. The chart offers an overview of the key biological processes impacted by the analyzed genes Interestingly, analyzing the data without applying any filtration revealed a small yet remarkably proliferative cell population with high expression levels of breast cancer markers and HER2 expression which persisted post-treatment, possibly suggesting the presence of treatment-resistant tumor cells (Supplementary data, Fig. [67]S1). However, because we only studied a small number of samples, there might be missing information. To clarify the accuracy of this finding and how it could impact treatment decisions, future studies with larger datasets are needed. Inflammatory cytokines identified in CSF Our cytokine profiling showed an increase of IL-1β, IL2, IL4, IL5, IL6, IL8, RANTES and epidermal growth factor (EGF) two weeks after treatment (Fig. [68]5). Expression of some of these pro-inflammatory cytokines listed genes (PCP4, MORN2, FHIT), while Fig. [69]3C depicts the distribution of expression levels for cell cycle-associated genes (TOP2A, MKI67, CCND1, CDT1) during treatment. The dot plot in Fig. [70]3D illustrates the persistent expression of monocyte chemotactic protein 1 (MCP-1) and macrophage migration inhibitory factor (MIF) cytokines in all samples. Fig. 5. [71]Fig. 5 [72]Open in a new tab Dynamics of inflammatory cytokine concentrations in CSF pre and post treatment. This figure illustrates the concentration of inflammatory cytokines in cerebrospinal fluid (CSF) supernatants before and after treatment, as determined by IsoPlexis Human Cancer Signaling Secretome Panel analysis. Graph presents the quantitative analysis of various inflammatory cytokines, highlighting changes in their levels following treatment Discussion Identification of the serial CSF tumor and immune contexture allows evaluation of the patient’s response to treatments such as immunotherapy or chemotherapy. This early identification gives clinicians the opportunity to choose the best treatment regimen for the patient at the earliest time possible and identify optimal combination [[73]12]. Consistent with findings from other groups, our results indicate that CSF can provide fundamental information about the genomic characteristics of brain tumors like BCBM and could be used as a reasonably non-invasive liquid biopsy with further validation [[74]13]. Molecular profiling of CSF is at the forefront of transforming CNS tumor diagnostics leading the way to personalized therapeutic approaches [[75]14]. We performed scRNA-Seq on CSF cells longitudinally collected from a patient with dural/leptomeningeal metastases from TNBC. Single-cell profiling allowed us to chart major cell types and transient tumor-specific states present in the circulating cells in CSF. Significantly among the identified cell subsets were myeloid-derived suppressor cells (MDSCs), an immunosuppressive myeloid lineage population implicated in the establishment of an immunosuppressive microenvironment conducive to tumor growth and therapeutic resistance (Supplementary data, Fig. [76]S2). We also observed that pro-inflammatory cytokine levels in CSF may corroborate the progression/regression results achieved by single cell sequencing results. Imaging data and symptoms were aligned with our single cell sequencing data and cytokine content of CSF. The TNBC signature expression changed considerably during treatment with sacituzumab govitecan. Highly expressed markers at baseline were CRABP2, KRT7, KRT8, KRT18, KRT19, PTPRF and TM4SF1, suggestive of metastases and tumor growth. Previous studies revealed overexpression of CRABP2 and KRT7 in ER + and ER − mammary cancer cells induce tumorigenesis and breast cancer lung metastasis. KRT7 also enhances drug resistance and induces apoptosis resistance of tumor cells [[77]15, [78]16]. The suppression of KRT19 has been shown to enhance the proliferation, migration, and formation of spheres in breast cancer cells in vitro through communication mediated by NUMB within the Wnt/Notch signaling pathway [[79]17], along with the activation of AKT signaling [[80]18]. Additionally, transcription of KRT19 is increased by the activation of the HER2/ERK/SP1 signaling pathway, leading to the translocation of KRT19 to the HER2 receptor which thereby stabilizes HER2 activation in both breast and lung cancers [[81]19]. PTPRF is known to be associated with cell invasion, migration, and metastasis in breast cancer [[82]20]. TM4SF1 expression is also positively correlated with cell migration, playing a major role in metastatic reactivation, and promoting relapse in TNBC [[83]21]. The decrease in expression of these markers post-treatment suggests a potential reduction in tumor growth and migration, which was consistent with imaging data showing tumor shrinkage. These molecular changes reflect a favorable treatment response, indicating the efficacy of the therapeutic intervention in inhibiting TNBC progression. Our cytokine analysis revealed persistent expression of MIF and MCP-1 at all time points with no significant change during the treatment course. MIF expression often correlates with tumor aggressiveness and poor patient outcomes [[84]22], and MCP-1 is an important mediator of pro-tumorigenic effects of epithelial cells and induces tumor cells proliferation [[85]23]. Two weeks after treatment, there was a noticeable surge in the secretion of growth factors and inflammatory cytokines within the CSF. Among these, epidermal growth factor (EGF) stands out for its role in promoting the motility and invasion of tumor cells. Studies have shown that EGF can significantly enhance the ability of cancer cells to migrate and infiltrate surrounding tissues, contributing to disease progression [[86]24]. Additionally, interleukin-1 beta (IL-1β) emerges as a key player in driving the inflammatory response associated with cancer progression. IL-1β is known to elevate the expression of several chemokines, including CXCL8, CCL2, and CCL5. These chemokines play critical roles in orchestrating the recruitment of immune cells and promoting tumor cell migration and invasion. Within the context of TNBC, these pro-metastatic activities are integral components of the complex interplay between the tumor, the surrounding stromal cells, and the inflammatory milieu [[87]25]. Furthermore, interleukin-6 (IL-6) adds another layer of complexity to the tumor microenvironment. Beyond its well-established role in inflammation, IL-6 has been implicated in driving metastasis and conferring resistance to therapy, particularly in estrogen receptor-positive (ER+) breast cancer cells. The activation of IL-6 signaling pathways can enhance the ability of cancer cells to disseminate and colonize distant sites, contributing to disease recurrence and treatment failure [[88]26]. At this time point, tumor size was bigger, and the percentage of tumor cells significantly increased in CSF. Interestingly, these tumor cell populations stopped expressing those genes, instead we observed highly expressing PCP4 and FHIT genes. According to other studies, PCP4 actively prevents apoptosis in human breast cancer cells [[89]27] and FHIT is an independent survival indicator for breast cancer patients, reduced FHIT expression was seen more often in tumors with metastases [[90]28]. As anticipated based on these molecular indicators, the response to treatment becomes evident by the fourth week. While there were signs of tumor shrinkage, indicating an initial therapeutic effect, there remains a portion of the tumor tissue that persists despite treatment. This persistence underscores the challenges posed by tumor heterogeneity and the development of resistance mechanisms, highlighting the need for targeted therapeutic strategies tailored to the specific molecular characteristics of the tumor. Validation of the observed molecular changes and their correlation with treatment response in larger patient cohorts would provide robust evidence for their clinical utility as predictive biomarkers. Additionally, further exploration of combination therapies targeting specific molecular pathways holds promise for enhancing treatment efficacy and overcoming resistance mechanisms. Moreover, the development of non-invasive biomarkers, including cytokine profiles and gene expression signatures, could facilitate real-time monitoring of treatment response and early detection of disease progression. Continued efforts in preclinical research utilizing advanced in vitro and in vivo models are essential for elucidating the functional significance of identified molecular alterations and guiding the development of novel therapeutic strategies tailored to the molecular characteristics of individual tumors. These future directions have the potential to significantly impact clinical outcomes and pave the way towards more personalized and effective treatment approaches for TNBC with CNS metastases. Conclusion Through a longitudinal analysis of CTCs, single-cell profiling, and cytokine changes in CSF, alongside clinical imaging data, we have demonstrated the potential utility of CSF as a valuable resource for monitoring treatment response and disease progression in patients with breast cancer brain metastases (BCBM). Integrating single cell sequencing and cytokine analysis of CSF cells and supernatant provides a more comprehensive understanding of the tumor microenvironment and its impact on treatment responsiveness. This approach could offer clinicians actionable insights into the dynamic response to therapy and aid in BCBM prognostication. Methods Sample collection Samples were collected after consent to participate and publish was obtained from the study participant. Approximately 10 mL of CSF was collected from the patient’s Ommaya reservoir prior to treatment, 2 and 4 weeks after treatment initiation. Specimens received in sterile containers were processed immediately upon arrival. Each CSF sample was separated from its cellular components using centrifugation (10 min at 1,300 × g at 22 °C). Cell pellets were re-suspended in freezing media (FBS 90%, DMSO 10%). Both supernatant and cell suspension were separately transferred to cryovials and kept frozen at − 80 °C. After the cell suspension vials were frozen, they were transferred to liquid nitrogen to prolong the preservation of the live cells. Clinical data collection The patient’s clinical information was collected retrospectively under protocol JWCI-19-1101 approved by the PSJH IRB. CNSide CTC enumeration The critical processing steps of the cell capture assay were performed as described [[91]29]. All steps were performed at ambient temperature. Briefly, contrived samples were created and immediately stored in CEE- Sure™ CSF collection tubes (Biocept, Inc., CA) containing 0.6 mL of diazolidinyl urea (DU; Sigma, St. Louis, MO) for no less than 48 h to recreate clinical sample transportation time. Samples were centrifuged at 400 x g for 5 min. Supernatant was aspirated and cell pellets were re-suspended and incubated with primary capture antibody cocktail PN1986. Unbound capture antibodies were removed through multiple wash- centrifugations at 400 x g for 5 min followed by aspiration. Samples were then incubated with a biotinylated goat-anti-mouse secondary Fab (Jackson Immuno Research, West Grove, PA). After additional wash cycles with Phosphate-Buffered Saline (PBS; Corning, Glendale, AZ), samples were flowed through a proprietary streptavidin-coated microfluidic device (Biocept, Inc., San Diego, CA). Cells labeled with the primary antibody capture cocktail PN1986 were immobilized via biotin streptavidin affinity in the microfluidic device. Microfluidic devices loaded with samples underwent several rounds of washing with PBS (Corning, Glendale, AZ) followed by methanol to fix and permeabilize cells. Fluorescent Immunocytochemistry (ICC) of the immobilized cells was performed in parallel using the following antibody conjugates: Streptavidin, Alexa Fluor™ 647 conjugate (Thermofisher Scientific, Carlsbad, CA), CD45, Alexa Fluor™ 594 conjugate (Biolegend, San Diego, CA), and Cytokeratins, Alexa Fluor™ 488 conjugate (ThermoFisher Scientific, Carlsbad, CA). Cells underwent subsequent staining with 4′,6-diamidino-2-phenylindole (DAPI; ThermoFisher Scientific, Carlsbad, CA). Microfluidic devices were scanned on an automated fluorescence imaging scanner Bioview^® (Billerica, MA) and evaluated by clinical laboratory scientists. Cells that were streptavidin positive (SA+), cytokeratin positive (CK+) and CD45 negative (CD45-) were considered tumor cells, whereas cells that were streptavidin negative (SA-) or CD45 positive (CD45+) were considered not to be tumor cells. For accuracy studies, cytokeratin was evaluated for CD45- and SA + cells. Cells were classified as CK + and CK- tumor cells depending on the presence and absence of the cytokeratin. Single cell RNA sequencing Single-cell 3’ RNA-seq libraries were generated from CSF samples using the Chromium Next GEM Single Cell 3’ Kit v3.1 (10X Genomics, Pleasanton, CA, USA) following the manufacturer’s protocol. Briefly, cells were encapsulated with barcoded Gel Bead in a single partition using the Chromium Controller and a 10X barcoded library constructed. Libraries were then sequenced on Illumina NextSeq 550 instrument (San Diego, CA, USA). Bases were called using the Illumina RTA3 method and gene expression matrices were generated using Cell Ranger 5.0.0 software. Cell type annotations Corrected and filtered gene expression matrices were SCTransformed with Seurat 3.6.3 on a per sample basis and then integrated through harmonizing ‘anchors’ as recommended for cell type identification in Seurat documentation. The number of reads, number of features, and percent of mitochondrial reads were regressed out in the data scaling step of SCTransform, and the top 3000 most variable features were used. Principal component analysis (PCA) was then run on the integrated assay. The first ten principal components (PCs) were then used to generate a shared nearest-neighbor graph which was then clustered under the Louvain algorithm with a resolution of 0.5. Uniform manifold approximation and projection (UMAP) was then created using the first 10 PCs and 30 nearest neighbors. Canonical cell type markers were used to identify expected cell types (Supplementary data, Table [92]S1). Cell type clusters were grouped into three categories (tumor and mixed group, T cell group, and monocyte group) then clustering procedure was repeated to differentiate cell types more specifically within the categories. Quality control and data analysis Data analysis was performed using both Loupe Browser and Seurat package in R v3.6.3. Known cell markers used to classify clusters included cell cycle, breast cancer, drug target, nerve, T cell, B cell, NK cell, plasma, monocyte, myeloid derived suppressor cells (MDSC), tumor-associated macrophages and microglia. Any cells with fewer than 200 mapped features were eliminated, as well as any features present in fewer than three cells. Exclusion criteria were: 1) < 500 expressed genes, 2) > 6000 expressed genes, and 3) > 25% UMIs of mitochondria genes. Based on the chromosomal position data of each gene by ShinyGO analysis, all input genes used in DE analysis were mapped on chromosomal position in ManhattanPlot (number of genes = 2,741, which are from max. 3,000 genes that were detected as highly variable genes in Seurat.integration analysis). Pathway enrichment analysis was conducted to investigate cell-specific functional pathways. CSF supernatant cytokine analysis Cytokine concentrations in CSF supernatants were quantified by loading supernatants onto the CodePlex Secretome Human Cancer Signaling Panel chip (IsoPlexis # CODEPLEX-2L04), and the chip was loaded into the IsoSpark reader (IsoPlexis, Branford, CT). Automated analysis of raw data was performed using IsoSpeak software (IsoPlexis). Electronic supplementary material Below is the link to the electronic supplementary material. [93]Supplementary Material 1^ (1.6MB, docx) Acknowledgements