1.Establishment of leptomeningeal metastasis TNBC cells and xenograft animal model
Approximately 10–30% of patients with metastatic BC develop CNS metastases. CNS metastasis is clinically classified into two major metastatic phenotypes: parenchymal metastasis [18, 19] and LM. In parenchymal metastasis, tumor cells enter the parenchyma through the endothelium surrounding the vasculature and the connected astrocytes, thereby causing perivascular proliferation [20–22]. In LM, tumor cells enter the CSF compartment and adhere to the pia mater surrounding the brain and spinal cord [2, 8, 23–25]. As revealed by histopathological analysis, our xenograft model demonstrated two major subtypes of CNS metastatic cells: (1) tumor cells adhered to the blood–brain barrier and (2) tumor cells colonized the pia mater in the entire brain section (Supplementary Fig. S1A). Analysis using a IVIS Spectrum In Vivo Imaging System (IVIS) combined with microcomputed tomography (microCT) revealed MDA-MB-231 cells undergoing obvious colonization in the lung, brain, and spinal cord in the seventh week after IC injection (Fig. 1A); this supported that our xenograft mouse model had two CNS metastatic subtypes. To investigate the mechanism of leptomeningeal-specific metastasis in TNBC, we established leptomeningeal-tropic TNBC cell lines (LeptoM1, LeptoM2, and LeptoM3) derived from MDA-MB-231 through a series of in vivo selections (Fig. 1B). The photon flux signals represented isolated LeptoM2 and LeptoM3 cells that primarily colonized in the brain and spinal cord but not in the lung or the liver, unlike LeptoM1 cells in vivo and ex vivo (Supplementary Fig. S1B). We detected a 100% specificity of the spinal cord and brain colonization in mice injected with LeptoM3 cells unlike in mice with brain-tropic (BrM3) and MDA-MB-231; however, metastasis was not detected for any other organs (Fig. 1C). Combined with microCT analysis, IVIS was used to detect the specific location of the photon flux signal in mice. The results indicated that LeptoM3 specifically colonized the spinal cord tissues but did not cause spinal bone erosion (Fig. 1D). Further, the BrM3 cells specifically colonized the brain in mice (Supplementary Fig. S1C). Further, histopathology was used to confirm the LeptoM3 cells infiltrating location in the brain and spinal cord. After 7 weeks of IC injection, the LeptoM3 cells colonized the pia mater in brain sections and adhered to and invaded the nerve tissues in the leptomeningeal space of spinal sections (Fig. 1E). Moreover, immunohistochemical (IHC) staining was used to detect the circulating LeptoM3 cells in the leptomeningeal space of the lumbar spine and those adhered to the nerve tissues (Fig. 1F). Early LM was detected among mice in the third week of injection with LeptoM3 cells, whereas the parental and BrM3 cell signals were undetectable till the fourth or fifth week after injection (Fig. 1G). Compared with the MDA-MB-231 and BrM3 groups, the survival rate of mice injected with LeptoM3 cells decreased significantly (Fig. 1H; p = 0.0124). These results indicated that isolated LeptoM3 cells caused specifically early LM and poor survival in vivo.
2. Correlation of high endogenous ICAM2 in TNBC with LM in vitro and in vivo
LM primarily occurs through two main infiltration routes, namely the rapid accumulation of tumor cells in the choroid plexus after hematogenous dissemination [2] and the growth of cancer cells in arachnoid mater and adhered to the pia mater [26]. To investigate the accumulation of tumor cells in the choroid plexus, we constructed an artificial BCB in vitro and tested permeability using fluorescence dye to exclude the cell migration through barrier disruption (Supplementary Fig. S2A-B). The luciferase activities of tumor cells were assessed, and cell numbers were determined in vitro (Supplementary Fig. S2C). The cell numbers of migrated trans-BCB cells increased 4-fold or 1.3-fold in LeptoM3 cells relative to MDA-MB-231 or BrM3, respectively (Fig. 2A). Furthermore, the number of invasive cells significantly increased 8.5-fold in LeptoM3 cells compared with in MDA-MB-231 cells under low serum attracted conditions, mimicking the physiological conditions of CSF (Supplementary Fig. S3A). The isolated LeptoM3 cells notably caused a time-dependent increase in BCB adhesion abilities compared with MDA-MB-231 and BrM3 cells (Fig. 2B). In addition, tumor cells adhered to the pia mater were evaluated using a brain slice adhesion assay[27]. The number of adhesive tumor cells increased 3.4-fold or 3.3-fold in the LeptoM3 cells compared with MDA-MB-231 or BrM3 cells, respectively. The LeptoM3 cells primarily colonized the pia mater in the brain; however, MDA-MB-231 and BrM3 cells primarily colonized the parenchyma of the brain (Fig. 2C and Supplementary Fig. S3B). In LM progression, the isolated LeptoM3 cells enhanced BCB adhesion and trans-BCB migration abilities in vitro. Moreover, we proposed that specific plasma membrane proteins expressed in LeptoM3 cells may cause LM as a result of tumor cell colonization in the choroid plexus.
To identify the specific plasma membrane proteins in LM, proteomic analysis was performed for parental (MDA-MB-231), brain-tropic (BrM3), and leptomeningeal-tropic (LeptoM3) TNBC cells. We then compared the potential plasma membrane proteins and ranked the protein profiles on the basis of two filtering criteria: (1) the proteins should be plasma membrane proteins and (2) the proteins should be detectable only in LeptoM3 cells but not in BrM3 or in MDA-MB-231 cells. As illustrated in Fig. 2D, AGRV1, OR8S1, VTN, ITGA2, and ICAM2 were specifically expressed in the LeptoM3 cells. Moreover, another regulating cell adhesion ICAM1 that belongs to members of the immunoglobulin and integrin supergene families notably decreased in CNS metastatic cells. The results of qRT-PCR indicated that only ICAM2 increased notably in LeptoM3 cells. The changes in mRNA levels of AGRV1, OR8S1, VTN, and ITGA2 were not notable in LeptoM3 cells, and the corresponding molecules were excluded from the potential molecules (listed in Fig. 2E). The results of western blotting revealed that ICAM2 notably increased by 15.0-fold or 25.2-fold in total lysates or the membrane fraction of LeptoM3 compared with in BrM3 and MDA-MB-231 cells, respectively. By contrast, ICAM1 notably decreased in LeptoM3 compared with in BrM3 and MDA-MB-231 cells (Fig. 2F and Supplementary Fig. S4A). Furthermore, in-gel digestion proteomic analysis revealed that ICAM2 is present in the unique band of LeptoM3 cells compared with MDA-MB-231 and BrM3 by silver staining (Supplementary Fig. S4B). The data from IF also indicated that ICAM2 was highly expressed on the surface of LeptoM3 cells but was not detected on the MDA-MB-231 and BrM3 cells (Fig. 2G and Supplementary Fig. S4C). However, ICAM1 was detected on the cell surface of MDA-MB-231 cells (Supplementary Fig. S4C). To investigate whether tumor cells accumulated on the choroid plexus, we determined the expression of the ICAM2 and ICAM1 proteins. The BCB-adhered MDA-MB-231 cells were isolated using an in vitro artificial BCB assay (Supplementary Fig. S4D). The BCB-adhered MDA-MB-231 cells exhibited higher levels of ICAM2 and lower levels of ICAM1 than parental MDA-MB-231 cells; this pattern was similar to that observed for in vivo isolated LeptoM3 cells (Fig. 2H). To investigate whether high levels of ICAM2 and low levels of ICAM1 have organotropic metastasis properties, different CNS metastasis tropic cells including BrM-831, BrM3, and LeptoM3 cells were subjected to Western blotting analysis. High levels of ICAM2 and low levels of ICAM1 were observed in leptomeningeal tropic TNBC cells but not in brain tropic and parental cells (Fig. 2I). To investigate whether high levels of ICAM2 could promote LM, independent clones of ICAM2 positive MDA-MB-231 cells were sorted through flow cytometry[28]. The protein levels of ICAM2 were increased by 1.8–3.7-fold in ICAM2-positive MDA-MB-231 cells compared with parental cells (Supplementary Fig. S4E). Moreover, the protein levels of ICAM1 decreased to 30–50% in ICAM2-positive MDA-MB-231 cells compared with parental cells. The IF results also indicated that ICAM2 was highly expressed on the cell surface of ICAM2-positive MDA-MB-231 cells (Supplementary Fig. S4F-G). Furthermore, the relative levels of ICAM2 were inversely correlated with the levels of ICAM1 in parental, LeptoM3, and ICAM2-sorting TNBC cells (R2 = 0.6982; Fig. 2J). The ICAM2-sorting MDA-MB-231 cells not only notably increased the BCB adhesion ability in vitro (Supplementary Fig. S4H) but also promoted the spinal cord colonization in vivo (Fig. 2K). IHC revealed that tumor cells infiltrating into the leptomeningeal space exhibited high levels of ICAM2, but ICAM1 was not detected (Fig. 2L). Chi-square analysis showed that high levels of ICAM2 MDA-MB-231 cells significantly increased spinal metastasis compared with low levels of ICAM2 MDA-MB-231 cells (p = 0.0143) but did not affect other organs in vivo (Fig. 2M). These results demonstrated that high levels of ICAM2 in TNBC cells was correlated with the incidence of LM in vitro and in vivo.
3.Two-way demonstration of ICAM2 promoting the LM in vitro and in vivo
To characterize the functional roles of ICAM2 in LM among patients with TNBC, we analyzed ICAM2-overexpressing MDA-MB-231 cells and two independent ICAM2-downregulating LeptoM3 cells. The IF images showed that the plasma membrane location of ICAM2 can be detected in LeptoM3 cells but not in the ICAM2-downregulating LeptoM3 cells (Fig. 3A). The BCB adhesion ability significantly decreased in the ICAM2-downregulating LeptoM3 cells compared with scramble cells (Fig. 3B, p < 0.05). The trans-BCB ability significantly decreased in ICAM2-downregulating LeptoM3 cells compared with scramble cells (Fig. 3C, p < 0.001). The metastatic lesions of the spinal cord were detected in the scramble group but not in the ICAM2-downregulating group in vivo and ex vivo (Fig. 3D-E). As confirmed by histopathological data, tumor cells suspended in the CSF compartment of the lumbar spine were detected in the scramble group but not in the ICAM2-downregulating group (Fig. 3F). Moreover, LM was first detected at 4 weeks in LeptoM3 cells-injected mice; however, it was undetected until the seventh week in mice in the ICAM2-downregulating group (Fig. 3G). Furthermore, the IF results showed that the membrane location of ICAM2 was detected in ICAM2-overexpressing MDA-MB-231 cells but not detected in the vector control (Fig. 3H). The results indicated that BCB adhesion ability significantly increased by 3.2-fold (p < 0.001) in ICAM2-overexpressing MDA-MB-231 cells compared with the vector control (Fig. 3I). The trans-BCB ability significantly increased by 1.5-fold (p < 0.001) in ICAM2-overexpressing MDA-MB-231 cells compared with vector control (Fig. 3J). Most importantly, after 7 weeks of IC injection, the metastatic lesions of the spinal cord were detected in the ICAM2-overexpressing group but not in the vector control group (Fig. 3K-L). Histopathological data demonstrated that tumor cells can infiltrate into the leptomeningeal space that were detected in the ICAM2-overexpressing group but not in the vector control group (Fig. 3M). After 4 weeks of IC injection, LM was detected in 66.6% mice injected with ICAM2-overexpressing MDA-MB-231, whereas it was not detected in the control group until the eleventh week (Fig. 3N). These results indicated that ICAM2 promoted early LM in vivo and increased the BCB adhesion and trans-BCB migration in vitro.
4. Enhancement of cancer stem cell properties due to high level ICAM2
In tumor metastasis progression, the stemness and tumor initiation abilities are crucial for tumor cells to survive environments with poor protein, glucose, and cytokine content and to form secondary tumors [29]. To investigate whether the thousands of high ICAM2 of BCB adhesive and migrated cells from circulation can promote secondary tumor forming in the CSF compartment, two-way demonstrations were performed to determine whether ICAM2 levels could regulate stemness and tumor-initiating abilities. First, the number of spheres significantly increased in ICAM2-overexpressing cells compared with vector control (Fig. 4A, sphere numbers: vector control, 28 ± 1.3; ICAM2-overexpressing cells, 53 ± 3.4, p < 0.01). Second, the protein levels of CD44, CD133, CD133high, and CD44high TNBC are indicators of cancer stem cell characteristics that promote metastasis [30–32]; the protein levels increased in the ICAM2-overexpressing cells compared with vector control (Supplementary Fig. S5A). Third, the mRNA levels of stemness-related genes including SOX2, NANOG, NOTCH1, and PROM1 significantly increased in the ICAM2-overexpressing cells compared with vector control (Supplementary Fig. S5B). Fourth, the sphere numbers of ICAM2-downregulated LeptoM3 cells significantly decreased from 32 ± 3.3 to 14 ± 4.8 and 13 ± 1.5 (Fig. 4B, p < 0.001). Fifth, the protein levels of CD44 and CD133 decreased in the two independent ICAM2-downregulated LeptoM3 cells compared with scramble cells (Supplementary Fig. S5C). A tumor-initiation ability of 1×103 ICAM2-overexpressing groups was detected in the second week after injection, whereas the vector control groups were undetectable till the third week after injection. Six weeks after mammary fat pad injection in the 100 tumor cells, 25% of mice in the ICAM2-overexpressing group reported tumor formation, whereas no tumor was detected in the vector control group until the tenth week (Fig. 4C). Finally, tumor initiation was evaluated in mice injected with ICAM2-overexpressing cells and vector control cells with mammary fat pad injection; the following tumor initiation ability was detected using IVIS every week. After 10 weeks of mammary fat pad injection, 1×103 and 1×102 ICAM2-overexpressing groups exhibited 100% and 50% of tumor-initiation ability, respectively, compared with control groups, which already had 50% and 0% tumor-initiation ability (Fig. 4D). The qRT-PCR results indicated that the mRNA levels of stemness-related genes including OCT4, NANOG, NES, PROM1 (CD133), and CD44 significantly increased in LeptoM3 cells compared with those in BrM3 cells, indicating that leptomeningeal tropic TNBC cells acquired CSCs phenotypes (Fig. 4E). CD44 and CD133 proteins significantly increased in the LeptoM3 and BCB adhesion cells, which had high levels of the ICAM2 protein, but not in BBB adhesion cells, which have low ICAM2 protein expression (Fig. 4F and Supplementary Fig. S5D). Furthermore, the percentages of CD44- or CD133-positive subpopulations were evaluated for MDA-MB-231, BCB adhesion, and LeptoM3 cells using flow cytometry. The CD44 positive subpopulation increased from 12% in MDA-MB-231 to 93% and 98% in BCB adhesion cells and two independent LeptoM3 cells, respectively (Fig. 4G). The CD133 positive subpopulation increased from 0.5% in MDA-MB-231 to 1.1%, 2.7%, and 3.1% in BCB adhesion cells and two independent LeptoM3 cells, respectively (Fig. 4H). Together, these results indicated that the LeptoM3 and BCB adhesion cells expressed higher stemness markers CD44 and CD133, and the overexpression of ICAM2 promoted sphere formation and tumor initiation.
5. ICAM2 interacts with endogenous ICAM1 of choroid plexus cells
To investigate the mechanism of ICAM2-mediated LM in TNBC, we evaluated the physically interacting ICAM2 proteins in the choroid plexus through bioinformatic analysis. Based on the (1) physically interacting partners of ICAM2 predicted from STRING datasets (Fig. 5A, https://string-db.org/) and (2) physically interacting partners of ICAM2 expressed in the choroid plexus derived from the database [33]. Four potential targeting molecules were identified, namely ICAM1, ITGAM, ITGAL, and EZR (Fig. 5B). Through the KEGG pathway and molecular function analysis of nine predicted ICAM2 interacting proteins, the results indicated that cell adhesion and transendothelial migration are crucial pathways for the physical interaction of ICAM2 proteins (Fig. 5C and Supplementary Table S3 and S4). The Human Protein Atlas datasets (https://www.proteinatlas.org/) show that ICAM1, ITGAM, and ITGAL have tissue specificity. However, the EZR had no tissue specificity expression and was thus excluded from our candidates. Further, Western blot was used to investigate the endogenous protein levels of ICAM1 and ITGAM in choroid plexus epithelial cells, HEK293T cells, and HUVEC cells. The ICAM1 exhibited a higher expression in choroid plexus epithelial cells compared with HEK293T and HUVEC cells (Fig. 5D and Supplementary Fig. S6A). The protein levels of ITGAM can’t be detected in choroid plexus epithelial cells; ITGAM was excluded from the candidates. ICAM1 was used for further investigation in the current study. We performed cellular component and subcellular location analysis of nine predicted ICAM2 interacting proteins; the results indicated that the physically interacting ICAM2 proteins were primarily distributed in the plasma membrane (Supplementary Fig. S6B). The results of IF indicated that ICAM1 was primarily detected at the plasma membrane in the choroid plexus epithelial cells (Supplementary Fig. S6C). To investigate the membrane form of ICAM2 and whether it interacts with ICAM1 of choroid plexus cells, the ICAM2 containing a His tag on the C-terminal region was transiently transfected into the HEK293T or MDA-MB-231 and the membrane ICAM2 was examined by IF assay (Supplementary Fig. S6D). The membrane form of ICAM2 proteins, with a molecular weight of 66 KDa, was purified by Ni-beads and imidazole elution (Supplementary Fig. S6E-F). The results of pull-down assay using His-ICAM2 showed that purified His-ICAM2 can physically interact with endogenous ICAM1 proteins in the choroid plexus epithelial cells (Fig. 5E). Further, the number of LeptoM3 cells interacting with choroid plexus epithelial cells has increased in a time-dependent manner compared with MDA-MB-231 cells in the coculture system (Fig. 5F). The IF assay revealed that ICAM2 colocalized with ICAM1 in the choroid plexus epithelial cells at the interface (Fig. 5G). These results indicated the possibility of LeptoM3 cells interacting with the choroid plexus through ICAM2–ICAM1 interactions.
6. Neutralizing ICAM2 abolishes the leptomeningeal metastasis in vitro and in vivo
Blocking the ICAM2-ICAM1 interaction was proposed to abolish the ICAM2-mediated LM. To investigate whether blockade of ICAM2 and interruption of ICAM2 and ICAM1 interaction can act as potential therapeutic and preventive approaches for LM, anti-ICAM2 or anti-ICAM1 neutralizing antibodies were used to treat LeptoM3 cells or choroid plexus epithelial cells in vitro. The cell numbers of BCB adhesion significantly decreased in LeptoM3 cells treated with 2.5 µg/mL and 5.0 µg/mL anti-ICAM2 neutralizing antibodies compared with the IgG control-treated group in a dose-dependent manner (Fig. 6A). The cell numbers of trans-BCB migration notably decreased in the anti-ICAM2 neutralizing antibody-treated group compared with the IgG control group in a dose-dependent manner (Fig. 6B). The sphere numbers significantly decreased in the anti-ICAM2 neutralizing antibody group compared with the IgG control group (Fig. 6C and Supplementary Fig. S7A). The invasive cell numbers significantly decreased in the anti-ICAM2 neutralizing antibody group compared with the IgG control group in a dose-dependent manner (Supplementary Fig. S7B). Furthermore, the cell numbers of BCB adhesion in the choroid plexus monolayer treatment with 1.2, 6.0, and 12.0 µg/mL anti-ICAM1 neutralizing antibodies significantly decreased compared with the IgG control-treated group in a dose-dependent manner (Fig. 6D). BCB pretreatment with an anti-ICAM1 neutralizing antibody significantly decreased the cell numbers of trans-BCB compared with the IgG control-treated group (Fig. 6E). Combining anti-ICAM2 with anti-ICAM1 neutralizing antibody treatment synergistically reduced the BCB adhesion ability of LeptoM3 cells compared with treatment with only anti-ICAM2 or anti-ICAM1 neutralizing antibodies (Fig. 6F). These results indicated that the blockade of the interaction of ICAM2 and ICAM1 attenuated the progression of LM, in addition to attenuating BCB adhesion, trans-BCB, and stemness abilities in vitro.
To investigate the efficacy of treatment with anti-ICAM2 neutralizing antibodies as a therapeutic approach for LM, the proposed murine model mimics patients with BC who have not undergone surgery. The mice were IC injected with 1 × 105 LeptoM3 cells, which were preincubated with 5 µg/mL IgG control or ICAM2 antibodies. Then, antibodies were IC administered every 2 days. The LeptoM3 cells treated with 5 µg/mL IgG control or ICAM2 antibody could not change the protein levels of ICAM2 (Supplementary Fig. S7C). After 7 weeks of IC injection, the results indicated that IgG control-treated group had mice with the colonization of the spinal cord. However, the ICAM2 antibody-treated group attenuates the colonization of the spinal cord in vivo and ex vivo (Fig. 6G). Early progression of LM in the mice injected with LeptoM3 cells and treated with IgG control were detected in the third week after injection, whereas progression in the ICAM2 antibody-treated group was not detectable till the sixth week after injection (Fig. 6H). Compared with the IgG control-treated group, the survival rate of the mice injected with LeptoM3 cells following treatment with ICAM2 antibody was significantly decreased (Fig. 6I, p = 0.0114). We investigated whether pretreatment with anti-ICAM2 neutralizing antibodies in 1 × 103 LeptoM3 cells to mimic blocking the circulating tumor cells after surgical resection could be applied as preventive care for LM in TNBC. The mice were IC injected with 1 × 103 LeptoM3 cells, which were preincubated with 5 µg/mL IgG control or ICAM2 antibody. After 10 weeks of IC injection, 83.3% mice in the IgG control-treated group had LM; however, the ICAM2 antibody-pretreated group attenuated colonization in the spinal cord ex vivo and in vivo (Fig. 6J-K). Figure 6L illustrates that pretreated ICAM2 antibody significantly prolonged the survival rate of the mice compared with IgG control treatment (p = 0.031). These results demonstrated that the blockade of ICAM2 in TNBC cells prevented early LM and indeed can act as a potential therapeutic and preventive target for LM in vivo.