LILRB3 Suppresses Immunity in Glioma and is Associated with Poor Prognosis.
Qiyuan Zhuang,Ying Liu,Hanze Wang,Ziyang Lin,Li Sun,Yue Liu,Yingying Lyu,Liang Chen,Hui Yang,Ying Mao
DOI: https://doi.org/10.1002/ctm2.1396
IF: 8.554
2023-01-01
Clinical and Translational Medicine
Abstract:Dear Editors, Gliomas are the most prevalent type of primary brain tumour. Within the glioma tumour microenvironments (TMEs), tumour-associated macrophages (TAMs) that hold high expression of co-inhibitory receptors constitute a significant portion of the immune landscape.1, 2 The effects analysis of co-stimulatory/inhibitory receptors mRNA levels on overall survival and multivariable Cox regression identified LILRB3 as an independent negative prognostic factor among various grades in glioma (Figures S1A–C). In the Huashan glioblastoma (GBM) cohort, we observed that patients with dense LILRB3 had significantly worse survival outcomes (n = 78, Figure 1A,B), independent of treatment strategy and tumour location (Figure S1D). These findings were validated in the CGGA_325 cohort, CGGA_693 cohort, GSE83300 and GSE68848 (Figure S1E–H). Single-cell RNA-sequencing (scRNA-seq) of TAMs from GBM patients3 showed LILRB3 was predominantly expressed on monocyte-derived TAMs rather than microglia-derived TAMs (Mg-TAMs) (Figure 1C). We found a significant colocalization of CD206 and LILRB3 in the cell membrane (Figure 1D), and increased infiltration of LILRB3+ CD206+ cells in the tumour region (Figure 1E). The detailed analysis or experiment methods are described in the Supporting Information. The reciprocal interactions between GBM cellular states and TAMs modulate the tumour immune environment.4 By spatial 10x transcriptome previously reported by Ravi et al.,5 we found that LILRB3+ Mo-TAMs had close proximity to mesenchymal (MES-like) states, rather than other cellular states (Figure 1F,H). Besides, LILRB3+ Mo-TAMs, highly expressed in the microvascular region, are closely associated with exhausted T cells and MES-like features in the Ivy GAP cohort (Figure 1G). Other cellular signatures proposed by Neftel et al.6 exhibited negative or no correlation with LILRB3 expression. Higher correlation coefficients of LILRB3 with MES-like scores were shown by locally weighted regression (Figure 1H,I and Figure S2A,B). Macrophages cocultured with MES28 cells (MES-like glioma stem cells) performed a significantly higher mean LILRB3 MFI level (Figure 2C–E) and higher gene expression (Figure 2F). Since high MES-like expression was associated with a worse prognosis in both IDH mutant and IDH wild-type glioma patients (Figure S3A,B), we found most of the mesenchymal genes were upregulated in temozolomide (TMZ)-resistant U251. Moreover, concurrent TMZ treatment increased the infiltration percentage of IFNγ+ CD4+ and IFNγ+ CD8+ in the GL261 mouse model and upregulated mesenchymal gene expression (Figure S3C,E–M). These findings are consistent with the higher LILRB3 expression of recurrent GBM (Figure 2A,B) in the GLASS cohort (primary and recurrent paired samples, n = 86) and in TAM clusters of both primary and recurrent samples.3 Notably, 19q13.42, where the LILRB3 is located, was the most frequently deleted locus in TCGA-LGG (54%) (Figure S2E,F). LILRB3 expression was lower in the 1p/19q codeletion group (Figure S2C,D,H, I), and significantly affected by the deletion status of 19q13.42 by marker-based scores analysis (Figure S2E). Apolipoprotein E (APOE) has been reported to be a major lipid carrier, transporting lipids to cells and tissues.7 Using 7-tesla-field-strength magnetic resonance spectroscopy, we found Lipid to Creatine ratio in high-grade gliomas is around 1.30 (Figure S2J), higher than the ratio in normal brain tissue around 0.59.8 Notably, LILRB3 and APOE were expressed higher in the central region (Figure 2G).9 Abundant expression of APOE in LILRB3+ myeloid cells was observed compared to LILRB3− cells (Figure 2H,I). We then used proximity ligation assay to examine the interaction of APOE and LILRB3 after coculture in situ. Coculture caused increased binding signals in HMC3 compared with the control group (Figure 2J,K). Gene Set Enrichment Analysis of HALLMARK gene sets revealed high enrichment scores for IL6/JAK/STAT3 signaling in the high-LILRB3-expressed GBM (Figure 2L,M). Higher LILRB3 and p-SHP1/SHP1 in high-grade glioma tissues were observed compared to lower-grade glioma or non-neoplastic brain by immunoblotting (Figure 2N–P). This finding is in line with previous reports demonstrating that LILRB3 can interact with SHP-1 in 293T cells.10 Moreover, high expression of APOE shows a lower T cell activation expression pattern in GBM (Figure 2Q). Notably, a low expression level of PD-1 and PD-L1 was detected (Figure S4A–E) in TAMs. ScRNA-seq and immunofluorescence staining showed almost no co-labeled PD-1 and LILRB3 (Figure 3A–C). Based on immunohistochemistry (IHC) staining, only 5% of glioma tissues with low LILRB3 expression exhibited high PD-1 expression, while both low and high PD-1 expression was observed in high-LILRB3-expression samples (Figure 3D,E and Figure S7A). Although mediating effects analysis showed that TMZ treatment was protective for patients with high LILRB3 expression (indirect effect of 0.73, ratio of 47.3%) (Figure S3D), these findings indicate that high LILRB3 may hold an extra immunosuppressive effect besides PD-1. Subgroups are divided by LILRB3 and PDCD1 classifier genes in both TCGA and CGGA datasets: double-high, median (high expression of either LILRB3 or PDCD1), and double-low (Figure 3F). The optimal cluster number was determined using NMF and k-means unsupervised clustering (Figures S5 and S6). The double-high groups had significantly higher Treg and myeloid-derived suppressor cell signature scores compared to the median or double-low groups (Figure 3G and Figure S7B–G). We then employed multiplex IHC to quantify FOXP3+, IFN-γ+, and GZMB+ T cells in the median (n = 9) and double-high (n = 8) groups of GBM patients (Figures 3H, I). GFAP+ cells did not differ significantly between the two groups, indicating similar absolute numbers of tumour cells (Figure 3L). Compared to the median group, the double-high group exhibited an increased proportion of FOXP3+ T cells and decreased proportions of IFN-γ+ and GZMB+ T cells in GBM samples (Figure 3J–N). In conclusion, we identified LILRB3, which is highly expressed in monocyte-derived TAMs, as an independent negative prognostic factor among various grades of gliomas. LILRB3 expression was positively correlated with the mesenchymal-like cellular state in GBM and upregulated after cocultured with MES-like glioma cells. In glioma TME, LILRB3 was bound with APOE and was accordant with the activation of SHP-1. PD-1 and LILRB3 exhibited consistent but different distribution patterns which could be used to identify three glioma immune subgroups. Double-high group (LILRB3 high and PD-1 high) exhibited a ‘colder’ TME compared with other groups, indicating LILRB3 as a potential target of PD-1 antagonist combination therapies. This work is supported by the National Key Research and Development Program of China (No. 2020YFA0804200), NSFC grant (No. 82073166, 82273203), Science and Technology Commission of Shanghai Municipality (20Z11900100) and Shanghai Municipal Science and Technology Major Project (No. 2018SHZDZX01). Hui Yang is supported by the Program for Professors of Special Appointment (Eastern Scholar) at the Shanghai Institutions of Higher Learning (Grant SSF151005). The authors declare no conflict of interest. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.