Adult diffuse gliomas account for about 80% of primary malignant tumors in brain. Glioblastoma, which is also referred to as WHO grade IV glioma, is one of the most lethal cancers and is associated with only 15 months median survival (1-3). According to the recent 2021 WHO classification, adult diffuse gliomas can be classified into three subtypes: isocitrate dehydrogenase (IDH)-mutant (typically, astrocytoma), IDH mutant with chromosome 1p and 19q co-deletion (typically, oligodendroglioma), and IDH-wildtype (typically, glioblastoma) (4). These three subtypes exhibit distinct genomic features and clinical outcomes. Glioma was at the forefront of the WHO genomic characterization and is one of the most extensively studied cancers (5-7). Yet despite extensive studies on gliomas, new therapeutic strategies are needed to improve clinical outcomes. Exploring as yet under-studied potential drug targets may be key in providing therapeutic options for this disease.
The family of G protein-coupled receptors (GPCRs) is the largest one of transmembrane proteins. They are the largest protein family of druggable targets for approved drugs; about 500-700 drugs that target GPCRs are approved for clinical use (8, 9). Repurposing approved GPCR-targeting drugs is an efficient way of developing new treatment options, as it reduces the time and effort involved in this process. To develop new treatment options for cancer, it is essential to carry out a systematic and intensive analysis of the roles of GPCRs in cancer. Among GPCRs, chemokine receptors, such as C-X-C motif chemokine receptor 4 (CXCR4), have been widely studied in gliomas (10, 11). However, in non-olfactory tissues, the role of odorant receptors (ORs), which are the largest subfamily of GPCRs and account for about 60% of all human GPCRs, remains unknown (12-14). In particular, the role of ORs in cancers such as glioma is largely unexplored, despite the enormous number of OR genes (13, 15). A well-organized review of non-olfactory GPCRs as therapeutic targets in glioblastoma by Byrne
Human GPCR families have been assigned 1414 unique gene symbols, including pseudogenes, by the HUGO (Human Genome Organization) Gene Nomenclature Committee (17). GPCRs are transmembrane receptors that bind specific ligands and activate intracellular signaling, thereby mediating senses such as vision, smell, and taste (18, 19). There are diverse subgroups within GPCR superfamilies. Human GPCRs can be largely classified into five families based on phylogenetic grouping, namely glutamate, rhodopsin, adhesion, frizzled, and secretin receptors; this grouping is known as the GRAFS classification (20, 21). In particular, the rhodopsin-like class comprises 93% (1315/1414) of human GPCRs, and ORs make up the largest portion, 61.7% (873/1414), of rhodopsin-like GPCRs according to the HUGO Gene Nomenclature Committee (17).
A previous transcriptome analysis of GPCRs revealed that the expression levels of GPCRs vary by tissue depending on the physiological function of the GPCR (22). ORs are expressed in olfactory epithelium where they conduct their major function of sensing molecules related to smell. In addition, expression of ORs is detected outside of the nose; these receptors are called “ectopic OR” (13, 14, 23). In particular, expression of the limited number of ORs in brain was detected using PCR and in situ hybridization techniques (24-27), and recently, systematic analysis on OR expression using microarray or RNA-sequencing revealed that various ORs such as OR51E1/Olfr558 and OR51E2/Olfr78 were expressed in brain region including cerebellum (28-30). In addition to expression of intact OR gene, Flegel
Next-generation sequencing technology has been widely used in cancer research, including systematic genomic and transcriptomic studies of GPCRs to identify new potential therapeutic targets in cancer (41-44). Until recently, GPCR-based genomic studies focused on non-olfactory receptors, but newer systematic studies highlight the role of ORs as potential contributors to tumorigenesis (45-48).
Given that ORs may have potential oncogenic roles in solid tumors, and that ORs expressed in normal brain and during neurodegenerative disease have functions distinct from odor perception, we hypothesized that ectopic ORs would perform tumor-related functions in brain. As further evidence for the role of ectopic ORs in tumor-related functions in brain, we previously reported that Olfr78 (human homolog, OR51E2) induced M2 polarization, thereby promoting tumor progression and metastasis (49). M2-like tumor-associated macrophages are abundant in the glioma microenvironment, and mediate increased resistance to radiotherapy and immunotherapy (50-52). Glioma is the most common brain malignant tumor. Therefore, a large number of studies have investigated, including the role of GPCRs in glioma. Our current analysis of PubMed literature showed that the oncogenic or tumor-suppressive roles of non-olfactory GPCRs, including CXCR4, SMO, and DRD2, have been reported in over 1,000 publications (Fig. 1A). However, only three publications describing the glioma-related functions of ORs were found: 1) Glioblastoma patients harboring OR4Q3 (mouse Olfr735) mutations have a poor prognosis (53); 2) OR51F2/Olfr568 is a potential mediator of the 4-gene signature predicting temozolomide response in lower-grade glioma (LGG) patients (54); 3) OR7E156P acts as a long non-coding RNA that contributes to tumor growth and invasion via the OR7E156P/miR–143/HIF1A axis in glioma cells (55). Also, recent single-cell tran-scriptome study of ORs in pan-cancer revealed that 65.9% and 55.2% of glioblastoma and astrocytoma cells express OR genes, obeying “one neuron-one receptor” rule (i.e., a single neuron express one OR gene (56, 57) although this study was not searchable in PubMed by the query “(OR gene name [Text Word]) AND (glioma[Text Word]) OR (glioblastoma[Text Word])” (45).
To investigate the overall expression pattern of GPCRs in glioma, we used the integrated RNA-seq-based expression profiles of LGG and glioblastoma multiforme (GBM) from the Cancer Genome Atlas (TCGA; downloaded from https://gdac.broadinstitute.org/) (6). A total of 656 primary LGG and GBM samples were available, of which 229, 258, and 169 samples were IDH wild-type (IDHwt), IDH mutant without chr1p/chr19q co-deletion (IDHmut-non-codel), and IDH mutant with chr1p/chr19q co-deletion (IDHmut-codel), respectively. In total, 721 GPCR genes were mapped on the integrated LGG and GBM expression profiles, and 375 of 721 genes were categorized as OR genes, including 11 pseudogenes. RNA expression profiles of glioma revealed that 60% of non-olfactory GPCR genes show ubiquitous expression (expressed (log2(RSEM+1) > 0.1) in > 90% of samples, and 81% of the non-olfactory GPCR genes were expressed in more than half of TCGA gliomas. Three-quarters of ORs were barely expressed, and the remainder of ORs were detected only in a subset of gliomas (Fig. 1B). These observations imply that ORs might exhibit sample- or subtype-specific functions in glioma and the glioma microenvironment. Hence, we investigated the expression and mutational profiles of ORs by the glioma subtype and their clinical relevance.
When we explored the expression profiles of ORs in more detail, 373 of 375 (99.5%) ORs mapped on the integrated LGG and GBM expression profiles were expressed in at least one sample (Fig. 1B). However, 237 ORs (63.5%) were expressed only in a small subset (< 5%) of glioma samples. On the other hand, expression of OR2A9P was detected in all gliomas, and another eight ORs (OR7D2, OR2H2/Olfr90, OR7A5/Olfr57, OR2A7, OR13A1/Olfr211, OR2L13/Olfr166, OR2W3/Olfr322, and OR2C1/Olfr15) were expressed in > 90% of glioma samples.
Because glioma subtypes have subtype-specific genomic and clinical features, we performed unsupervised hierarchical clustering using 43 ORs with highly variable expression between glioma samples (standard deviation ≥ 0.3 and expressed in more than one-third of samples) to identify whether the different subtypes of gliomas displayed distinct OR expression profiles (6). Interestingly, each subtype was clustered together, indicating that enrichment of the different subsets of ORs depended on the glioma subtype (Fig. 1C). High expression levels of OR51E1 (mouse Olfr558) and OR51E2 (mouse Olfr78) were noticeable in IDHwt subtypes; OR4N2 (mouse Olfr733) was expressed in IDHmut-non-codel subtypes; and expression of OR2L13 (mouse Olfr166), OR4N4, and OR4N3P was higher in IDHmut-codel subtypes than in other subtypes (Fig. 1C, D). Similar to our findings, Pappula
Next, we performed Cox regression survival analysis using gene expression levels to investigate the prognostic potential of OR gene expression. Univariate Cox regression analysis revealed that seven genes (OR2AE1, OR2B6/Olfr11, OR10AD1/Olfr287, OR51E1/Olfr558, OR7C1/Olfr1352, OR2A9P, and OR51E2/Olfr78) were associated with unfavorable overall survival (adjusted P-value < 0.05 and hazard ratio > 1). There were 18 genes associated with favorable overall survival, including OR4N2 (mouse Olfr733), OR4N4, and OR4N3P (adjusted P-value < 0.05 and hazard ratio < 1) (Fig. 2A). However, all of the survival-associated genes were filtered out during multivariate analysis with age, IDH status, and histologic grade; these factors are known prognostic markers of gliomas (59). Therefore, we re-conducted a survival analysis for each subtype. OR51E1 (mouse Olfr558, adjusted P-value = 0.03; hazard ratio = 1.15) and OR2C1 (mouse Olfr15, adjusted P-value = 0.03; hazard ratio = 0.68) were associated with potential unfavorable and favorable outcomes, respectively for IDHwt glioma (Fig. 2B-D). OR51E1 (mouse Olfr558) was previously reported as a potential biomarker for prostate cancer and small intestine neuroendocrine carcinoma (60). Four genes (OR13J1/Olfr71, OR2B6/Olfr11, OR4K2/Olfr730, and OR7A5/Olfr57) were associated with poor overall survival in IDHmut-non-codel gliomas, and another four genes (OR2H2/Olfr90, OR5K2/Olfr177: favorable; OR51E1/Olfr588, OR7C1/Olfr1352: unfavorable) were associated with overall survival in IDHmut-codel gliomas. However, no OR genes were associated with a significant prognosis in the IDHmut-non-codel and IDHmut-codel subtypes after P-value adjustment (Fig. 2B). OR2B6 (mouse Olfr11) has also been previously identified as a cancer biomarker (48); thus further analysis of these identified survival-associated OR genes is needed to determine if they are potential biomarkers and/or therapeutic targets in glioma (46, 48).
Single-nucleotide variants (SNV) and small insertion/deletions (INDEL) can cause changes in protein structure or mRNA degradation through nonsense-mediated mRNA decay; thus, SNV and small INDEL can generate proteins with gain or loss of functions. According to previous studies (61, 62), various natural variants of human ORs exist. Thus, in addition to transcriptomic analysis, we explored the mutational status of ORs in gliomas. Among 799 primary glioma samples that were subjected to either whole-exome or whole-genome sequencing from TCGA, 499 samples carried at least one OR non-silent somatic mutation. Overall, 330 OR genes went through 1056 somatic mutations in the 499 samples, and of the 1056 mutations, 964 (91.3%) were missense mutations (Fig. 2F). For IDHwt gliomas, 262 samples (74.4%) possessed at least one OR mutation, while 144 of 265 (54.3%) IDHmut-non-codel samples and 83 of 169 (49.1%) IDHmut-codel samples had mutations in OR genes (Fig. 2G). Figure 2E summarizes the 15 most mutated OR genes in gliomas. OR5D18 (mouse Olfr1155) was the most frequently mutated OR gene and was detected in 10 samples. Of note, OR5AR1 (mouse Olfr1019) mutations were detected in seven IDHwt samples, but OR5AR1 (mouse Olfr1019) mutations were absent in IDHmut gliomas. Taking these findings together, we suggest that functionally validating mutations in OR genes will be key to understanding the biology of gliomas.
To date, the role of ORs in glioma has barely been investigated and is likely to be underestimated. Through a systematic analysis and review of olfactory GPCRs in glioma, we revealed that there is glioma subtype-specific expression of ORs such as OR51E1/Olfr558, OR51E2/Olfr78, OR4N2/Olfr733, OR2L13/Olfr166, OR4N4, and OR4N3P. The existence of subtype-specific OR expression indicates that these ORs likely play roles in tumor formation and in the tumor microenvironment during glioma development or progression. In addition, identifying ORs associated with overall survival rates in glioma patients suggests that the expression level of ORs is a potential biomarker to predict glioma prognosis, and that these ORs should be investigated as therapeutic targets. To evaluate the oncogenic ability of ORs, an in-depth multi-omics analysis should be conducted with patient-derived resources and reliable translational models. Moreover, ORs are expressed in normal brain, and expression is up- and down-regulated in neurodegenerative diseases (32, 34). These observations demonstrate that ORs expressed in brain stromal cells and immune cells might form part of the glioma microenvironment and hence contribute to tumor progression. Therefore, single-cell RNA-sequencing analysis and reliable in vivo and/or in vitro models mimicking the tumor microenvironment are required to screen and functionally validate the cell type-specific expression of ORs.
This work was supported by grants from the National Research Foundation (2021R1A2C1009258 to J. Koo and 2021R1C1C1 004653 to H.J. Cho), the Bio & Medical Technology Development Program (2020M3A9D3038435 to J. Koo), the Korean Mouse Phenotype Center (2019M3A9D5A01102797 to J. Koo).
The authors have no conflicting interests.