Leptomeningeal metastatic cells adopt two phenotypic states

Abstract Background Leptomeningeal metastasis (LM), or spread of cancer cells into the cerebrospinal fluid (CSF), is characterized by a rapid onset of debilitating neurological symptoms and markedly bleak prognosis. The lack of reproducible in vitro and in vivo models has prevented the development of novel, LM‐specific therapies. Although LM allows for longitudinal sampling of floating cancer cells with a spinal tap, attempts to culture patient‐derived leptomeningeal cancer cells have not been successful. Aim We, therefore, employ leptomeningeal derivatives of human breast and lung cancer cell lines that reproduce both floating and adherent phenotypes of human LM in vivo and in vitro. Methods and Results We introduce a trypsin/EDTA‐based fractionation method to reliably separate the two cell subsets and demonstrate that in vitro cultured floating cells have decreased proliferation rate, lower ATP content, and are enriched in distinct metabolic signatures. Long‐term fractionation and transcriptomic analysis suggest high degree plasticity between the two phenotypes in vitro. Floating cells colonize mouse leptomeninges more rapidly and associate with shortened survival. In addition, patients harboring LM diagnosed with CSF disease alone succumbed to the disease earlier than patients with adherent (MRI positive) disease. Conclusion Together, these data support mechanistic evidence of a metabolic adaptation that allows cancer cells to thrive in their natural environment but leads to death in vitro.

malignancies, the prevalence of LM is increasing. 1 Cancer cells may access the space through a variety of means, including passage through Bateson plexus via the venous circulation or choroid plexus via arterial circulation, direct invasion of spinal and cranial nerves, or spread from brain parenchyma through direct penetration of the glia limitans. [2][3][4][5]  In order to reproducibly study the dynamics of these adherent and floating LM subpopulations, we have employed leptomeningeal derivatives 7 of human breast and lung cancer cell lines that reproduce both phenotypes of human LM in vivo and in vitro. To reliably separate the two cell subsets, we introduce a trypsin/EDTA-based fractionation method. In vitro, we find that floating cells display a decreased proliferation rate, lower ATP content, and are enriched in distinct metabolic signatures when compared with their adherent counterparts. Long-term fractionation and transcriptomic analysis suggest high degree plasticity between the two phenotypes in vitro.
Floating cells colonize mouse leptomeninges more rapidly and associate with shortened survival. Remarkably, this finding is mirrored in a retrospective patient dataset: Patients with exclusively cytology (+) disease and MRI (−) disease succumbed more rapidly to LM than their counterparts with cytology (−), MRI (+), or cytology (+), MRI (+) disease. Together, these results support a model whereby cancer cells in LM exist in a plastic equilibrium between adherent and floating states; the floating phenotype represents the lethal variant of these cells.

| Human studies
CSF in excess of that required for clinical decision making was collected from patients undergoing lumbar puncture. Air-dried cytospin CSF preparations were fixed in 4% PFA 5 minutes at room temperature, rinsed in PBS, and stained with Giemsa. Clinical data were obtained under MSKCC Institutional Review Board-approved protocol 13-039 "Gene expression patterns in Leptomeningeal Metastasis." Clinical information, including tumor tissue diagnosis, coulter counter CSF counts, time to LM diagnosis, etc, was abstracted from the medical record and de-identified. The presence of leptomeningeal disease in gadolinium-enhanced MRI brain was evaluated based on hyperintense signal in the leptomeningeal space present on T1 postcontrast sequences and absent on T1 precontrast and susceptibilityweighted sequences. Patients with incomplete pathological annotation (date and result of primary diagnosis, cytology/CTC count and brain and spine MRI, and date of death) and patients with intracranial metastases other than LM were excluded from further analyses. All patients provided informed consent.

| Animal studies
Cell fractions of MDA-MB-231 (MDA-231) LeptoM and PC9 LeptoM cells were prepared as described above. "Mixed" population of floating and adherent cells (mixed in 1:1 ratio prior to injection) was used as a control. The growth rate of unfractionated cells was characterized previously. 7  were introduced intracisternally, using the procedure described previously. 7 Briefly, deeply anesthetized mice were positioned prone over a 15-mL conical tube to place cervical spine in flexion. The occiput was palpated, and a Hamilton syringe fitted with a 31G beveled (cutting) needle was introduced between the occiput and C1 at an angle.  and Fiji/ImageJ (v1.51j, NIH).  (J.R. and X.T.) using qPCR ( Figure S3).

| Floating LM cancer cells exhibit an aggressive phenotype
To assess the phenotypic difference between the floating and adherent LeptoM cells in vivo, we injected equal numbers of floating, adherent, and mixed cells into the cisterna magna of immunodeficient animals ( Figure S4A). In both PC9 lung ( Figure 6A, Figure S4B) and MDA-231 breast cancer LeptoM models ( Figure 6C, Figure  Reasoning that a strong in vivo phenotype in mouse models may reflect human disease, we abstracted clinical information from the charts of 35 patients newly diagnosed with LM from breast cancer (n = 10), lung cancer (n = 23), or both (n = 2) (  Figure 7A). The onset of LM after primary diagnosis did not affect the site of LM ( Figure 7B). However, lung cancer was more likely to present with CSF-only disease ( Figure 7C).

| DISCUSSION
Metastatic colonization of distant body sites is an extremely inefficient biological process with fatal consequences. 11 The hostile environment of secondary sites provides substantial selective pressure, requiring significant transcriptomic and epigenetic adaptations in cancer cells, mirrored by plastic changes in their phenotype. 12 This phenomenon is particularly obvious in the case of LM, when typically adherent, tissue-bound cancer cells spread and become nonadherent within the nutritionally sparse cerebrospinal fluid. 13 Clinical observations and autopsy studies report that leptomeningeal cancer cells exist in two contrast phenotypes-freely floating in the CSF and adhering to leptomeninges and growing in plaques. 14 The success in in vitro expansion of patient-derived LM models is limited; attempts to reproduce these findings have gone unreported. 15 The lack of in vivo and in vitro LM models severely limits the development of novel therapeutic approaches. To overcome this bottleneck, we employ iteratively selected, clinically relevant in vitro xenograft models that reproduce the features of LM in vivo. 7  Funding Acquisition, A.B.