Autocrine GMCSF Signaling Contributes to Growth of HER2⁺ Breast Leptomeningeal Carcinomatosis

Among patients with metastatic breast tumors, 10% to 30% develop central nervous system (CNS) metastases (1, 2). Several factors positively correlate with a greater risk of brain metastases, among them: poorly differentiated tumors; HER2-enriched, luminal HER2, basal-like, and triple-negative breast cancer subtypes; and having four or more metastatic lymph-nodes (3, 4). HER2⁺ breast leptomeningeal carcinomatosis (HER2⁺ LC), which occurs when HER2⁺ breast tumor cells spread to the cerebrospinal fluid (CSF)-containing leptomeninges surrounding the brain and spinal cord (5–8), is an ominous complication of breast cancer with a dire prognosis (6–8). Once established, HER2⁺ LC can invade the parenchyma to produce focal neurologic damage (9). Any malignancy can spread to the leptomeninges; however, given the high incidence of breast cancer (and particularly the HER2⁺ subtype) worldwide, breast cancer is the most common origin (10). Although significant progress has been made in developing breast cancer treatments that target systemic disease, efficacy in the CNS remains a challenge, thus leading to an increase in the incidence of HER2⁺ LC (11). Indeed, HER2⁺ LC typically develops while the systemic tumor burden is well-managed (12–15), and 30% of HER2⁺ LC cases are diagnosed as the first manifestation of cancer after a substantial disease-free interval (14, 16–18).

HER2⁺ LC remains incurable, with few treatment options and response rates often less than 20% (19–25). The current standard-of-care for HER2⁺ LC management is multidisciplinary, including radiotherapy (RT) and intrathecal chemotherapy (ITC; refs. 26–29). Methotrexate (MTX), a DNA alkylating drug, is frequently used as palliative ITC for HER2⁺ LC (9, 30–33). However, this approach has limited success and causes serious side effects (26–29). Furthermore, patients with HER2⁺ LC are excluded from clinical trials due to poor prognosis and to minimize results that are not reproducible (6, 34–36). Therefore, our goal was to identify novel therapeutic targets to improve the management of this intractable disease.

Little is known about how HER2⁺ breast cancer cells proliferate in the leptomeninges, which are acellular and poor in protein, glucose, and cytokine content (5–8). Thus, in this study, we used primary HER2⁺ LC patient-derived (“Lepto”) cell lines (37) to identify the molecular mechanisms that promote HER2⁺ LC development in this unique context. We found that oligodendrocyte progenitor cells (OPC), which are abundant in white matter, inhibit HER2⁺ LC growth in vitro (using Lepto cell lines) and in vivo (in HER2⁺ LC xenograft models in NOD/SCID mice), limiting the spread of HER2⁺ LC beyond the leptomeninges. We also conducted cytokine array-based analyses of media conditioned by Lepto cells and various CNS cell types, which revealed that GMCSF is an oncogenic autocrine driver of HER2⁺ LC that is significantly overexpressed by HER2⁺ LC patient-derived tissues and cell lines. In addition, using LC/MS-MS-based analyses, we demonstrated that the OPC-derived factor TPP1 proteolytically degrades GMCSF and can thus suppress HER2⁺ LC growth. Finally, we showed that combined treatment with anti-GMCSF neutralizing antibodies plus a pan-Aurora kinase inhibitor (CCT137690) synergistically inactivates GMCSF signaling and reduces HER2⁺ LC growth in vitro and in vivo. Collectively, these findings indicate that GMCSF overexpression confers a survival advantage to HER2⁺ LC cells, suggesting that GMCSF inhibition could be an effective therapeutic approach to treat patients with HER2⁺ LC.

Use of human specimens was approved by the City of Hope (COH) Institutional Review Board (IRB; protocols #07047 and #16015; refs. 38–40). Written informed consent was obtained from all patients under protocols #07047 and #16015, and the study was conducted in accordance with the Declaration of Helsinki, institutional guidelines, and all local, state, and federal regulations. All mouse studies were approved by the COH Institutional Animal Care and Use Committee (protocol #10044). We only used female NOD/SCID mice for all the in vivo experiments because the HER2⁺ Leptomeningeal carcinomatosis (HER2⁺ LC) occurs predominantly in females.

All chemical compounds/drugs, antibodies, culture media, its supplements, and the analyses of software used in this manuscript are listed and described in the Supplementary Tables S1 and S2, along with their source information and research resource identifier numbers.

Derivation of Lepto lines from HER2⁺ LC patient-derived tumors is described in ref. 37. Briefly, nodular HER2⁺ LC tumors from HER2⁺ LC patients who underwent surgeries to acquire biopsies for pathologic confirmation of HER2⁺ LC or to decompress localized symptomatic lesions (IRB protocols #07047 and #16015). Each specimen was mechanically dissociated, and CD44⁺/CD24⁻/EpCAM+/CD49f⁻ cells (displaying epithelial and cancer stem cell phenotypes) were FACS sorted and maintained at 37°C and 5% CO₂. Low-passage Lepto lines were cryobanked using STEM-CELLBANKER Cryopreservation Media (AMSBIO) and banked in liquid nitrogen at −180°C. The Lepto cells were cultured in hCSF-supplemented advanced Dulbecco's Modified Eagle Medium/Ham's F-12 Nutrient Mixture (DMEM:F-12; Life Technologies) with various supplements (Supplementary Table S2) on collagen-coated T-75 flasks, as described previously (37, 39).

Cell lines (HEK293T, MDA-MB-231, BT-474, and T47D) were obtained from ATCC (details listed in Supplementary Table S1). All the lines were biweekly tested for mycoplasma contamination. The cell lines were either grown in RPMI+ 10% FBS + Pen-Strep (1×), IMDM +10% FBS + Penn-Strep (1×), or RPMI +20% FBS + Pen-Strep (1X) according to guidance from ATCC. HEK293T were cultured in DMEM +10% FBS+ Penn-Strep (1×). Cell line authentication was done by short tandem repeat profiling at the IDEXX Bioanalytic Laboratories Inc., and tested as Mycoplasma negative by PCR (Agilent Mycosenser Mycoplasma Assay Kit) as recent as 15 days prior to last experiments.

Human iPSC-derived multipotent NPCs were obtained from EMD Millipore (catalog no. SCC035) and propagated using ENStem-A neural expansion medium (catalog no. SCM004). Cells were terminally differentiated to neurons or oligodendrocytes using ENStem-A Neuronal Differentiation Medium (catalog no. SCM017) or human OPC Expansion Media (catalog no. SCM107; basal medium with PDGF-AA, NT3, FGF2, T3, and retinoic acid), respectively, following the supplier's recommendations. Differentiated microglia, oligodendrocytes, neurons, OPCs, and reactive astrocytes were purified by immunostaining with anti-CD45, anti-GALC, anti-CD90, anti-NG2, and anti-HepaCAM antibodies, respectively, sorted by FACS as described in ref. 41, and propagated in supplier-recommended media. To recapitulate the in vivo microenvironment, cells were grown in hCSF for various durations for the in vitro experiments. Cell morphology and differentiation status were monitored using immunofluorescence (IF) staining with anti-CD45, anti-GALC, anti-CD90, anti-NG2, and anti-HepaCAM antibodies.

Cell viability was assessed using a Cell Titer Glo Luminescent Cell Viability Assay Kit, according to the manufacturer's protocol. Apoptosis was measured using Annexin V-FITC or staining of phycoerythrin PE-conjugated CD326 (EpCAM-PE). Annexin V-FITC binding was analyzed by flow cytometry using an FITC signal detector, and EpCAM-PE staining was analyzed using a PE emission signal detector. Adherent Lepto cells were trypsinized and washed once with FBS-containing media before incubation with Annexin V-FITC or EpCAM-PE.

For HER2⁺ LC-derived Lepto lines based xenografts model development, 4- to 6-week-old female NOD/SCID mice were used that were maintained under pathogen-free conditions in accordance with guidelines and therapeutic interventions approved by COH Institutional Animal Care and Use Committee (IACUC 10044). As HER2⁺ leptomeningeal carcinomatosis (HER2⁺ LC) occurs predominantly in females (Gender), only female NOD/SCID mice were used for all the in vivo experiments. For the overall survival analyses, tumor progression and tumor seeding studies, NOD/SCID 4- to 6-week-old female mice were utilized. Animals were housed under standard conditions in the ARCH facility at COH. All animal experiments were carried out under approved IACUC protocols and followed COH's animal care procedures.

To evaluate the effects of treatment on tumor growth, overall survival, and tumor seeding analyses in vivo, control, and variously transduced mCherry and firefly luciferase (mCherry: LUC, Addgene_29783) Lepto lines were injected at 100K density or various other densities in 20 μL PBS buffer via cisterna magna puncture into cohorts of female NOD/SCID mice. At 7- and 14-day postimplantation of Lepto cells, mice were intrathecally injected with either OPCs or OPCs-shGFP or OPCs-shTPP1 (100K or 200K in 20 μL PBS), CCT137690 (50 mg/kg; Fig. 5G–I), TPP1 (50–150 ng/mL), anti-GMCSF antibodies in PBS (8 μg/g; Fig. 2H–J or 4 μg/g; Fig. 5G–I), or vehicle (PBS alone). Tumor growth was monitored weekly by BLI on a Xenogen Imaging System (Xenogen Corp.). Mice were injected with 100 mg/kg D-luciferin, and two sets of in vivo BLI images in one projection were acquired, resulting in a collection of eight images. Mice were then euthanized, and their brains collected, fixed in formalin (Thermo Fisher Scientific), and subjected to Western blot analysis or hematoxylin and eosin (H&E) or IHC analysis. On each in vivo BLI image, a region of interest (ROI) encompassing the entire mouse except the tail was placed, and the total signal in the ROI was quantified using Living Image software (version 2.50; Xenogen). The total signals of all images obtained in a single imaging session were averaged to determine the whole-body signal intensity, which was used as a marker of whole-body tumor burden. As per IACUC (#10044), the experimental endpoint of the animals were death and/or reaching biologically humane endpoint based on tumor burden, weight loss, mobility, food/water refusal. Once the tumor bearing NOD/SCID mice reached biological endpoint, the animals were subjected to final BLI for tumor burden before euthanization.

Data shown in figures are mean values ± SE, using data generated from n = 3 biological replicates with n = 2 technical replicates present in each biological replicate. Statistical significance between groups was determined using one- or two-way ANOVA, followed by multiple comparisons with Bonferroni multiple comparisons correction. The level of significance used was α:0.05. Other statistical evaluations were performed using the Student t test. The software used for the above-mentioned analyses was GraphPad Prism 8.4.1. Kaplan–Meier curve was used to model overall survival. P value of <0.05 was considered statistically significant. Significance in statistical analyses in the figures is represented by *, P < 0.05, or **, P < 0.05, or ***, P < 0.05. All additional methods are described in the Supplemental Materials and Methods Section.

The RNA sequencing data generated in this study has been deposited in Gene Expression Omnibus (GEO) under accession number GSE179280.

To determine whether host glial cells in the CNS impact HER2⁺ LC growth and development, we developed HER2⁺ LC patient-derived lines that we call “Lepto” lines, which demonstrated the unique spinal cord migration functionality in vivo, as do HER2⁺ LC tumor cells (37). HER2⁺ LC patient-derived lines differed transcriptomically different from other HER2± metastatic breast cancer cell lines (Supplementary Fig. S1A). We then immuno-panned CNS cell types from human induced pluripotent stem cell (iPSC)-derived neural progenitor cells (NPCs). Specifically, we used anti-CD45, anti-GALC, anti-CD90, anti-NG2, and anti-HepaCAM antibodies to sort microglia, oligodendrocytes, neurons, OPCs, and astrocytes, respectively, by FACS (41). Cells were maintained in human CSF (hCSF) for various time periods (4–5 days), during which, all cell types maintained typical morphology and marker expression patterns (Fig. 1A). We then cocultured each cell type in Boyden chambers with primary HER2⁺ LC patient-derived Lepto1 or Lepto2 cells and assessed their effects on Lepto cell viability (37). Coculture of both Lepto lines with astrocytes increased their proliferation, whereas coculture with OPCs induced Lepto cell apoptosis (Annexin V-based FACS staining; Fig. 1B) and reduced their viability (CellTiter-Glo Luminescent Cell Viability Assay; Fig. 1C). IF imaging of mCherry: LUC-labeled Lepto cells cocultured 48 hours with or without OPC-conditioned media indicated more robust Annexin V (green) and procaspase-3 (magenta) staining, indicative of increased apoptosis, in cells grown in conditioned media (Fig. 1D; Supplementary Figs. S1B and S2A). To characterize these effects in vivo, we injected mCherry:LUC-labeled Lepto cells (100K) into the cisternae magna of adult NOD/SCID mice (on day 0), followed by OPCs co-implantation (100K) on days 7 and/or 14 (Scheme; Fig. 1E). Then, from days 14 to 50, we monitored tumor growth via bioluminescence imaging (BLI). Mice co-implanted with Lepto cells (on day 0) and OPCs (on day 7 ± day 14) showed significantly decreased tumor growth based on BLI relative to non-OPC injected control animals bearing Lepto tumors and prolonged survival (Fig. 1F, left and right, respectively; Supplementary Table S3). Furthermore, histopathologic analyses of H&E-stained axial sections of spinal cord and sagittal sections brain of Lepto bearing NOD/SCID mice injected with no OPCs or OPCs (D7+14) revealed marked reduction in levels of Lepto-derived tumors (Fig. 1G; Supplementary Fig. S2B). These findings confirm the inhibitory effects of OPCs observed in vitro and in vivo.

To identify factors that initiate and drive growth of HER2⁺ LC tumors, we cocultured primary patient-derived Lepto cells with OPCs in a Boyden chamber for 72 hours, for comparison with OPCs or Lepto cell controls grown as monolayers. When we analyzed growth medium from samples using a Cytokine XL array (Fig. 2A; Supplementary Fig. S3A and Supplementary Tables S4 and S5), we observed significantly higher GMCSF concentrations in media of monocultured Lepto cells relative to cocultured OPC/Lepto cells or monocultured OPCs (Fig. 2A and B). In addition, primary Lepto cells expressed higher levels of GMCSF transcripts than did various iPSC-derived CNS cell types (Fig. 2C). Moreover, nodular patient derived HER2⁺ LC tissues expressed higher levels of GMCSF transcripts compared with other primary and metastatic patient derived tumors or normal human breast and brain tissues (Fig. 2D). IHC analyses of Lepto bearing mouse brain sections revealed significantly higher levels of the phospho-GMCSF receptor α subunit (GMCSFRα) in HER2⁺ LC lepto derivd tumor tissues relative to surrounding brain tissues (Fig. 2E and quantification on right panel). Furthermore, Western blot analysis of Lepto cells cultured with or without OPCs revealed decreased phosphorylation of GMCSFRα and its downstream targets, STAT5, AKT, and ERK1/2 in the presence of OPCs compared with Lepto cells cultured without OPCs [Fig. 2F (Western Blots) and quantification of Western blots on right panel], suggesting that OPCs inhibit GMCSF secretion from Lepto cells.

We next assessed disruption of GMCSF signaling in vitro. To do so first we assessed apoptotic markers in Lepto cells grown in either hCSF-supplemented media (controls), OPC-conditioned hCSF-supplemented media, or hCSF containing anti-GMCSF neutralizing antibodies, which have been used clinically in various cancer treatments (42, 43). Quantification of surface Annexin V via flow cytometry indicated that Lepto cells grown in OPC-conditioned hCSF or hCSF containing anti-GMCSF antibodies were significantly more apoptotic (∼90%) than control (vehicle treated) Lepto cells (Fig. 2G). Next, we assessed potential antitumor effects of anti-GMCSF neutralizing antibodies on tumor growth in vivo by administering those antibodies intrathecally to xenograft mouse models of HER2⁺ LC and monitoring tumor formation by BLI (Fig. 2H). Relative to vehicle-treated controls, antibody-treated mice showed reduced tumor progression (based on BLI counts; Fig. 2J, left) and increased overall survival (Fig. 2J, right; Supplementary Table S3). These findings were supported by IHC analysis of sagittal brain sections from lepto bearing NOD/SCID mice, which demonstrated that the anti-GMCSF neutralizing antibodies suppressed tumor growth in the brain stem regions (Fig. 2I, Sagittal brain sections stained with HER2 antibody). Taken together, these results confirm the contributory role of GMCSF signaling in HER2⁺ LC tumor progression and suggest that treatment with anti-GMCSF neutralizing antibodies could serve as a potential strategy to target HER2⁺ LC.

To further assess GMCSF effects, we established Lepto cells conditionally overexpressing GMCSF by cloning the GMCSF open reading frame (ORF) downstream of ZsGreen1-IRES in a Tet-On 3G inducible expression system (Vector design; Fig. 3A). Addition of doxycycline (DOX; 5 μg/mL) to the culture media of Lepto cells transduced with this construct significantly increased GMCSF protein levels [in Lepto cell lysates; Fig. 3B, left (Western blot) and right (quantification of the Western blots)], with a concomitant increase in ZsGreen1 levels as detected by FACS and IF imaging (Fig. 3C and D). DOX-induced GMCSF expression in Lepto cells increased their viability in the presence of OPCs relative to that seen in the absence of DOX, based on analysis using Cell Titer-Glo assays (Fig. 3E). Moreover, in the presence of OPCs DOX-exposed GMCSF overexpressing Lepto cells showed decreased apoptosis as determined by Annexin V flow cytometry-based analysis than did comparably cocultured Lepto cells without DOX induction (Fig. 3F).

Next, to assess consequences of GMCSF loss of function, we inserted GMCSF shRNA (shGMCSF) upstream of ZsGreen1 and employed the same Tet-On 3G inducible expression system to conditionally knockdown GMCSF in Lepto cells (Fig. 3G). DOX (5 μg/mL) treatment of transduced Lepto cells significantly reduced GMCSF protein levels relative to “No DOX” controls [Fig. 3H, left (Lepto1) and right (Lepto2) and I (quantification of Western blots)]. ELISA analysis confirmed decreased GMCSF secretion from DOX-treated Lepto cells transduced with shGMCSF (Fig. 3J). DOX-treated, shGMCSF-transduced Lepto cells also showed decreased proliferation in vitro relative to “No DOX” Lepto cell controls (Fig. 3K).

We next compared the ability of control and GMCSF knockdown Lepto cells with seed tumors in mice. Briefly, 8 NOD/SCID mice per group were implanted with varying density of Lepto cells with or without shGMCSF transduction and found that the number of injected cells required to generate tumors in at least 1 of 8 mice was 10³-fold less for control Lepto cells (100) than for GMCSF knockdown cells (10⁵ to 10⁶; Fig. 3L). To further confirm that GMCSF drives Lepto cell growth in vivo, we first established constitutive GMCSF-overexpressing Lepto cells and transduced them with the inducible GMCSF-shRNA vector (Supplementary Figs. S4A and S4B). ELISA analysis confirmed abrogation of GMCSF secretion following DOX treatment (Supplementary Fig. S4C). Then using the DOX induction system, we compared the ability of GMCSF-overexpressing versus GMCSF-depleted Lepto cells to seed tumors in NOD/SCID mice. The number of injected cells required to generate tumors in at least 1 of 4 mice was 10³-fold less for GMCSF-overexpressing (100) as compared with GMCSF-depleted (10⁵ to 10⁶) Lepto cells (Fig. 3N). Histopathologic analyses of H&E-stained sagittal brain sections demonstrated that GMCSF overexpressing Lepto cells were able to form tumor in the various regions of the brain including the brain stem, whereas shGMCSF overexpression or coexpression with GMCSF (ORF) led to significantly decreased tumor growth (Fig. 3M).

We next asked if OPCs secrete factor(s) that inhibit GMCSF signaling and potentially induce Lepto cell apoptosis. To do so, we analyzed the secretomes of human astrocytes and OPCs, cultured alone or with Lepto cells, using LC/MS-MS. We identified 38 unique proteins present in OPC-conditioned hCSF whose levels remained unchanged in OPC-Lepto cocultures (Fig. 4A; Supplementary Table S5). Of the 38 candidate proteins, only CGREF1, ENO, PTPRZ1, SPARC, and TPP1 were known secreted proteins located extracellularly (Supplementary Table S5). PROSPER-based predictions of GMCSF protease cleavage sites (44) revealed multiple sites for various serine proteases (Supplementary Table S6). Among the five candidates, only TPP1, a serine protease in the sedolisin family, acts as a nonspecific lysosomal peptidase that cleaves N-terminal tripeptides (45). When we examined media conditioned by monocultured OPCs or (?) OPCs cocultured with Lepto cells, we observed higher levels of TPP1 than seen in media of monocultured Lepto cells or in astrocyte-conditioned media (Fig. 4B). To assess whether OPC-derived TPP1 proteolytically degrades GMCSF, we cultured Lepto cells 24 hours with 50 or 100 ng/mL of recombinant TPP1 protein in media supplemented with hCSF and measured GMCSF levels by ELISA. Both TPP1 concentrations significantly reduced GMCSF levels in the media relative to vehicle-treated control Lepto cells (Fig. 4C). Furthermore, TPP1 treatment also suppressed Lepto cell viability via usage of CellTiter-Glo Luminescent Cell Viability Assay (Fig. 4D). To determine if recombinant TPP1 or OPC-secreted TPP1 can degrade Lepto-secreted GMCSF, we treated culture media of Lepto cells conditionally overexpressing GMCSF with 50 ng/mL TPP1 and, separately cocultured GMCSF-overexpressing Lepto cells with OPCs. Both treatment with exogenous TPP1 and and coculture with OPCs decreased concentrations of secreted GMCSF protein in Lepto cell media, as determined by ELISA (Fig. 4E). Furthermore, both conditions induced Lepto cell apoptosis (based on flow cytometry-based surface Annexin V staining; Supplementary Fig. S5A).

To determine if TPP1 loss of function in OPCs would alter concentrations of secreted GMCSF in Lepto line cocultures, we transiently transfected iPSC-derived OPCs with siTPP1 and the following day cocultured them (or control OPCs transfected with siGFP or siLUC) with Lepto cells conditionally overexpressing GMCSF (Supplementary Fig. S5B). One day later, siTPP1-transfected OPCs showed significantly decreased levels of TPP1 transcripts relative to control OPCs, as measured by RT-qPCR (Supplementary Fig. S5C). Interestingly, GMCSF protein levels were significantly higher in media from Lepto cells cocultured with TPP1-depleted OPCs than in media from Lepto cells cocultured with control OPCs (Supplementary Fig. S5D).

To assess effects of OPC-derived TPP1 on GMCSF signaling in Lepto cells in vivo, we co-implanted Lepto cells (100K) with or without OPCs (100K or 200K) into the cisternae magna of NOD/SCID mice. As anticipated, we observed OPC density-dependent suppression of HER2⁺ LC tumor progression, as indicated by BLI counts (Fig. 4F and G) and histopathologic analyses using H&E staining (Fig. 4I). Moreover, co-implantation of OPCs with Lepto cells increased animal survival relative to mice implanted with Lepto cells only (Fig. 4H; Supplementary Table S3). Analysis of sera from these mice revealed that TPP1 protein levels increased with OPC density (Fig. 4J), an effect that corresponded to decreased GMCSF levels (Fig. 4K). Interestingly, OPCs co-implanted with Lepto cells did not exhibit substantial changes in viability between days 5 and 20 postimplantation (Fig. 4L).

Next, to determine whether OPC-secreted TPP1 inhibits Lepto cell growth in vitro or in vivo, we transduced iPS derived OPCs with either TPP1 shRNA or control GFP shRNA. Analyses of TPP1 protein levels via Western blot analysis demonstrated that the protein levels of TPP1 were significantly reduced in OPCs transduced with shTPP1 relative to OPCs transduced with shGFP [Supplementary Figs. S5E and S5F (quantification of Western blots in SF5E)]. Then, coculture of GFP depleted OPCs with Lepto cells significantly reduced GMCSF secretion from Lepto cells relative to Lepto cells exposed to shTPP1 transduced OPCs (Supplementary Fig. S5G). Next, we co-implanted control or TPP1 knockdwon OPCs (100K or 200K) into the cisternae magna of NOD/SCID mice bearing Lepto tumors. We observed decreased growth of Lepto-derived tumors in mice implanted with control OPCs compared with mice implanted with TPP1 knockdown OPCs, based on BLI quantification from days 10 to 26 (Fig. 4M, left). Accordingly, animal survival was significantly decreased in mice implanted with OPC-shTPP1 relative to control OPCs (Fig. 4M, right; Supplementary Table S3). Analysis of sera from these animals revealed higher TPP1protein levels in OPC-shGFP- compared with OPC-shTPP1-implanted mice (Fig. 4N, left). Also GMCSF levels in sera were also relatively higher in OPC-shTPP1-implanted mice (Fig. 4N, right). Finally, H&E staining of coronal (Fig. 4O) and sagittal (Fig. 4P) brain sections showed reduced tumor growth in shGFP-OPC co-implanted Lepto tumor bearing NOD/SCID mice (n = 3) compared with shTPP1-OPC co-impanted Lepto tumor bearing mice (n = 3). Analysis of horizontal brain sections revealed comparable effects (Supplementary Fig. S5H). Overall, in vitro (Fig. 4B–E) and in vivo (Fig. 4F–P) analyses suggest that OPC-secreted TPP1 degrades GMCSF and suppresses GMCSF signaling, decreasing Lepto cell viability and tumor progression.

Current treatment of HER2⁺ LC tumors relies on cytotoxic ITC, which indiscriminately kills rapidly dividing cells. Given our findings that GMCSF signaling can drive HER2⁺ LC growth, we searched for drugs that could synergistically target and inhibit this signaling pathway. To do so, we performed a chemical genetics screen using Lepto cells treated with compounds from the LOPAC-1280 library (Scheme; Supplementary Fig. S6A). Specifically, GFP-labeled Lepto cells were seeded in 384-well plates at 1,000 cells per well, and 1 day later, 0.1% DMSO (control) or one of three concentrations of each LOPAC-1280 compound (100, 200, or 500 nmol/L) was added to cells. After 72 hours, we analyzed cell viability using MitoTracker staining. Of all compounds tested, the pan-Aurora kinase inhibitor CCT137690 had the strongest effects, inhibiting Lepto cell viability by ∼95% (Supplementary Figs. S6B and S6C; Supplementary Table S7). Dose–response analysis showed Lepto lines to be sensitive to CCT137690 at all concentrations tested with an IC₅₀ value ∼18 nmol/L (Fig. 5B). CCT137690 (100 nmol/L) treatment of cultured Lepto cells also significantly induced cell apoptosis based on surface Annexin V staining (Fig. 5C; Supplementary Figs. S6D and S6E).

CCT137690 inhibits Aurora A, B, and C kinases. Aurora A participates in crosstalk with GMCSF signaling to regulate effectors such as STAT5, AKT, and mTOR (Supplementary Fig. S7A), which reportedly promote Lepto cell proliferation and viability (46–48). We observed elevated Aurora A expression levels (mRNA) in nodular HER2⁺ LC tissues relative to other primary and metastatic tumors and normal breast and brain tissues (Fig. 5A). Thus, we asked whether combining anti-GMCSF neutralizing antibodies with CCT137690 (anti-GMCSF+CCT137690) would antagonize Lepto tumor initiation, growth, and/or relapse. To evaluate effects on Lepto tumor initiation, we treated cultured Lepto cells for 24 hours with DMSO (control), anti-GMCSF neutralizing antibodies, CCT137690, or anti-GMCSF+CCT137690 and then cultured them 7 days in conditions favoring tumorsphere formation. After 7 days, control cells developed numerous round tumorspheres, whereas cells treated with anti-GMCSF antibodies, CCT137690, or anti-GMCSF+CCT137690 formed fewer and smaller tumorspheres (Fig. 5D). CCK assays of the same tumorspheres showed that, relative to DMSO, all treatments reduced the proportion of sphere-initiating cells (Fig. 5E). Notably, combining anti-GMCSF antibodies with CCT137690 reduced the proportion of live, tumorsphere-initiating cells by ∼80%, an effect significantly greater than any single reagent. To assess effects on Lepto tumorsphere growth, we allowed untreated Lepto cells to form tumorspheres for 5 days and then treated them 2 days with DMSO (control), anti-GMCSF antibodies, CCT137690, or anti-GMCSF+CCT137690. CCK assays confirmed that, relative to DMSO, all treatments—and most significantly the combination treatment—reduced primary tumorsphere cell viability (Fig. 5F).

To assess treatment effects on relapse, we developed secondary tumorspheres from the primary tumorspheres assessed in Fig. 5F. Briefly, after treatment of primary tumorspheres with DMSO (control), anti-GMCSF antibodies, CCT137690, or anti-GMCSF+CCT137690, surviving cells were dissociated and allowed to form secondary tumorspheres for 12 days in standard stem cell medium only (Fig. 5D). We then dissociated the secondary tumorspheres and subjected the cells isolated from secondary tumorspheres to CCK assays and found significantly fewer viable cells in secondary tumorspheres pretreated with anti-GMCSF+CCT137690 relative to DMSO-treated controls. Tumorspheres treated with CCT137690, and anti-GMCSF antibodies also generated fewer viable cells than DMSO-treated controls, but the effect was less robust than that observed for anti-GMCSF+CCT137690 (Fig. 5D).

Finally, to evaluate these effects on HER2⁺ LC growth in vivo, we administered anti-GMCSF+CCT137690 (as well as anti-GMCSF and CCT137690 alone and vehicle control) to NOD/SCID mice on days 5, 10, and 15 after Lepto cell implantation (Fig. 5G). Compared with vehicle control or individual treatments (anti-GMCSF and CCT137690), the anti-GMCSF+CCT137690 combination treatment significantly reduced tumor progression as indicated by BLI analysis (Fig. 5H, left) and increased overall survival (Fig. 5H, right; Supplementary Table S3). Subsequent H&E staining of the sagittal brain sections from variously treated Lepto bearing NOD/SCID mice demonstrated that relative to vehicle treated Lepto bearing mice, anti-GMCSF+CCT137690 treated as well as anti-GMCSF and CCT137690 treated mice demonstrated decreased Lepto tumors in the brain stem as well as in other regions of the brain indicated by red arrows (Fig. 5I).

CNS metastases from breast cancer occasionally spread to the parenchymal brain or leptomeninges (49–52). HER2⁺ breast cancer is the most common solid tumor origin of leptomeningeal metastasis (53–60). Once tumor cells reach leptomeninges, they may spread via the CSF (61). Thus, diagnoses can be made via positive cytology of aspirated CSF samples. However, in some cases, adherent nodular deposits develop on the surface of the brain, spinal cord, and spinal roots, allowing diagnosis based on MRI alone (15, 22, 61, 62). The presence of nodular deposits is associated with the greatest suffering from headaches and intractable pain due to cranial and spinal nerve invasion (22, 61, 62).

Approximately 84% of breast cancers reportedly contain at least one genomic alteration that could be exploited as a treatment target (63), and genetic screens have identified promising therapeutic targets in breast cancer (64). However, only a few targets, including phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), AKT1, and ERBB2 (4, 64) have been validated in clinical studies, and the success rate of these strategies is surprising low. Therefore, there is a need to identify additional targets and therapies that effectively target HER2⁺ LC tumors that metastasize to the leptomeninges. Furthermore, studies have shown that brain metastasis leads to astrocyte-mediated alterations in brain tissue around the tumor, which limit and can negatively impact intratumoral chemotherapeutic delivery (65). Thus, identification of drugs that can penetrate through the brain tissue surrounding the HER2⁺ LC tumor to effectively target growing HER2⁺ LC is urgently required, which in turn requires a better understanding of molecular mechanisms that govern migration of HER2⁺ LC tumor cells from the brain stem to the acellular leptomeningeal environment in the spinal cord and support HER2⁺ LC growth in such a acellular environment. To identify such therapies, we established and analyzed primary HER2⁺ LC patient-derived Lepto cells (37), which led to the discovery that OPCs found primarily in white matter inhibit HER2⁺ LC cell viability in vitro and in vivo. Furthermore, we go on to show that GMCSF serves as an autocrine driver contributing to Lepto cell growth in vitro and in vivo. It is also important to note that although the effects of anti-GMCSF neutralizing antibodies are significant, but the inhibition of GMCSF signaling pathway is incomplete (Fig. 2F and H–J), suggesting that GMCSF is not the sole driver HER2⁺ LC and there must be other signaling pathways enabling the growth of HER2⁺ LC growth in vivo in the leptomeninges. As evidence, we report that conditional GMCSF overexpression partially blocks OPC-induced Lepto cell apoptosis in vitro, and that comparable effects seen in vivo are reversible by DOX-induced GMCSF knockdown in implanted Lepto cells.

LC-MS-based analyses reported here identified the protease TPP1 as a candidate regulator of GMCSF-mediated signaling in HER2⁺ LC. Interestingly, TPP1 reportedly functions as a lysosomal serine protease that serves as a nonspecific lysosomal peptidase (66, 67). TPP1 deficiency is associated with various fatal neurodegenerative diseases (68–72), such as neuronal ceroid lipofuscinoses; however, its role in inhibiting HER2⁺ LC tumor development in white matter has not been explored. In this study, we observed that extracellular GMCSF levels dropped significantly when Lepto cells were either cultured with OPCs or treated with recombinant TPP1. In addition, quantification of GMCSF levels in Lepto cell culture media and in serum derived from mice bearing xenograft Lepto tumors indicated that TPP1 secreted from OPCs degrades GMCSF. Co-implantation of TPP1 depleted OPCs in Lepto-derived tumor bearing NOD/SCID mice reversed the OPC-mediated inhibition of Lepto cell growth (relative to co-implantation with normal OPCs or shGFP-OPCs), suggesting that TPP1 derived from OPCs may degrade GMCSF and antagonize growth of HER2⁺ LC tumors in leptomeningeal regions. We propose that intrathecal administration of recombinant TPP1 and/or inhibition of GMCSF signaling may be a viable therapeutic option to target HER2⁺ LC growth. That idea was supported by our finding that administration of anti-GMCSF neutralizing antibodies suppresses GMCSF-mediated signaling and significantly impairs HER2⁺ LC development in vivo.

To identify additional drugs that suppress Lepto cell growth, we performed a chemical genetics screen using the LOPAC-1280 compound library. The strongest inhibitor of Lepto cell viability was CCT137690, a highly selective pan-Aurora kinase inhibitor. Interestingly, Aurora kinases regulate mitotic activities, such as centrosome maturation, spindle assembly, chromosome segregation, and cytokinesis (73–78), and their inhibitors have been extensively studied as novel anti-mitotic drug targets (79, 80). Aurora A kinase overexpression was observed in HER2⁺ LC patient derived tissues and cell lines. Interestingly, our unbiased chemical screen identified Aurora inhibitor I as well as targeted analyses identified MK5108, which are inhibitors of Aurora A kinase. MK-5108 (VX-689), which has entered clinical trials in the United States, in patients with advanced and/or refractory solid tumors. Cell-Titre-Glo based dose titrations and comparison of IC₅₀ values of CCT137690, Aurora inhibitor I and MK5108 (Fig. 5B; Supplementary Fig. S7B) showed that CCT137690 was more effective in Lepto cells. Hence, CCT137690 was further pursued for in vitro and in vivo combinatorial analyses along with anti-GMCSF antibodies in Fig. 5. Analysis presented in Fig. 5D–I indicates that CCT137690 and anti-GMCSF neutralizing antibodies synergize to inactivate GMCSF effectors and strongly inhibit primary and secondary Lepto tumorsphere initiation, growth, and relapse in vitro. Moreover, in xenograft mouse models, combination treatment with CCT137690 and anti-GMCSF neutralizing antibodies antagonized Lepto tumor growth and augmented overall animal survival more potently than single treatment with CCT137690, TPP1, or anti-GMCSF antibody, suggesting that comparable strategies could be used to target HER2⁺ LC tumors in patients. Future research is warranted to optimize ITC with TPP1, CCT137690, and/or anti-GMCSF for application in the clinic.

In summary, we have identified and characterized neural niche-specific crosstalk between HER2⁺ LC tumors and OPCs residing predominantly in white matter. We showed that GMCSF acts as an autocrine oncogenic driver of HER2⁺ LC growth in vivo and report that intrathecal administration of the protease TPP1, the selective pan-Aurora kinase inhibitor CCT137690, and/or anti-GMCSF antibodies may be an potential strategy to treat HER2⁺ LC in the clinic.

No disclosures were reported.

K.I. Ansari: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. A. Bhan: Software, formal analysis, validation, methodology, writing–original draft, writing–review and editing. M. Saotome: Resources. A. Tyagi: Resources. B. De Kumar: Resources, writing–original draft, writing–review and editing. C. Chen: Resources, writing–original draft, writing–review and editing. M. Takaku: Resources. R. Jandial: Conceptualization, resources, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The authors express their gratitude to the COH Analytical Cytometry Core Facility. This work was made possible by the generous support of the COH Department of Surgery and a grant from the United States Department of Defense Breast Cancer Research Program (W81XWH-19-1-0310). The RNA-seq data analyses were supported by the UND Genomics core. M. Takaku was supported by the University of North Dakota Start-up and P20GM104360 from the National Institutes of Health.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.