Enhanced expression of FCER1G predicts positive prognosis in multiple myeloma

Background: Multiple myeloma (MM) is the second most common hematologic malignancy worldwide and does not have sufficient prognostic indicators. FCER1G (Fc fragment Of IgE receptor Ig) is located on chromosome 1q23.3 and is involved in the innate immunity. Early studies have shown that FCER1G participates in many immune-related pathways encompassing multiple cell types. Meanwhile, it is associated with many malignancies. However, the relationship between MM and FCER1G has not been studied. Methods: In this study, we integrated nine independent gene expression omnibus (GEO) datasets and analyzed the associations of FCER1G expression and myeloma progression, ISS stage, 1q21 amplification and survival in 2296 myeloma patients and 48 healthy donors. Results: The expression of FCER1G showed a decreasing trend with the advance of myeloma. As ISS stage and 1q21 amplification level increased, the expression of FCER1G decreased (P = 0.0012 and 0.0036, respectively). MM patients with high FCER1G expression consistently had longer EFS and OS across three large sample datasets (EFS: P = 0.0057, 0.0049, OS: P = 0.0014, 0.00065, 0.0019 and 0.0029, respectively). Meanwhile, univariate and multivariate analysis indicated that high FCER1G expression was an independent favorable prognostic factor for EFS and OS in MM patients (EFS: P = 0.006, 0.027, OS: P =0.002,0.025, respectively). Conclusions: The expression level of FCER1G negatively correlated with myeloma progression, and high FCER1G expression may be applied as a favorable biomarker in MM patients.


Background
Multiple myeloma (MM) is a hematologic malignancy characterized by the monoclonal expansion of bone marrow plasma cells (BMPCs) [1][2][3]. The International Staging System (ISS) divides MM into three categories based on the levels of β2-microglobulin and albumin at diagnosis, which are Ivyspring International Publisher surrogate markers of tumor burden. Additionally, 1q21 amplification is considered a high-risk genetic feature, which is the most common chromosomal aberration in MM [4,5]. In recent years, genetic biomarkers are starting to play an increasingly important role in the prognosis of myeloma [6,7]. Therefore, it is necessary to investigate novel biomarkers to predict the prognosis of MM, so as to help improve the prognostication and treatment of MM.
FCER1G is a protein coding gene located on chromosome 1q23.3 [8]. It has been reported that FCER1G interacts with other factors and participates in various nuclear pathways [9]. Specifically, FCER1G is a constitutive component of the high-affinity immunoglobulin E (IgE) receptor and interleukin-3 receptor complex. It is mainly involved in mediating the allergic inflammatory signaling of mast cells, selectively mediating the production of interleukin 4 (IL4) by basophils, and initiating the transfer from T-cells to the effector T-helper 2 subset [10,11]. It also forms a functional signaling complex together with the pattern recognition receptors CLEC4D and CLEC4E in myeloid cells. Previous studies have shown that FCER1G is an innate immunity gene and may be involved in the development of eczema, meningioma and childhood leukemia [12][13][14]. FCER1G is associated with the progression of clear cell renal cell carcinoma (ccRCC) and may improve prognosis by affecting the immune-related pathways. In addition, FCER1G is underexpressed in acute myeloid leukemia [15]. Moreover, FCER1G is a critical molecule in signaling pathways that are widely involved in a variety of immune responses and cell types [16]. However, the prognostic role of FCER1G in MM remains largely unknown.
Here, we explored the relationship between FCER1G expression and myeloma progression, ISS stage, 1q21 amplification, and survival, using the gene expression data of 2296 MM patients and 48 healthy donors. We were able to demonstrate that high expression of FCER1G was a good indicator of MM and was related to positive outcomes.

Data source
In this study, we selected 2296 myeloma patients and 48 healthy donors from the Gene Expression Omnibus database (GEO). In order to assess the relationship between FCER1G expression and the prognosis of MM patients, the sample was divided into two cohorts. In the first cohort, there were six independent microarray datasets (GSE39754, GSE5900, GSE2113, GSE6477, GSE47552, GSE13591). This cohort included 48 healthy donors and 640 MM patients in different stages of monoclonal gammopathy (104 monoclonal gammopathy of undetermined significance (MGUS), 69 smoldering myeloma (SMM), 452 multiple myeloma (MM) and 15 plasma cell leukaemia (PCL)). This cohort was used for microarray expression analysis.
The second cohort consisted of three big independent microarray datasets of MM patients, GSE2658, GSE4204 and GSE24080. In GSE2658, the gene expression data of 559 MM patients was evaluated by the Affymetrix Human Genome U133 Plus 2.0 Array. Samples in GSE4204 were pre-treatment bone marrow aspirates from 538 MM patients. In GSE24080, the gene expression profiling of highly purified bone marrow plasma cells was performed in 559 newly diagnosed MM patients. This cohort was mainly used for survival analysis, and the expression of FCER1G in different 1q21 amplification levels and different ISS stages was also described.
All the samples were classified according to the International Myeloma Working Group criteria [17]. The diagnosis of MM (ICD-10 C90.0) was established in accordance with the World Health Organization guidelines [18]. The diagnosis of MGUS require more than 10% plasma cell infiltration in the bone marrow, while the levels of monoclonal protein could not exceed 30 g/L and there would be no evidence of related organ or tissue impairment (ROTI) defined as hypercalcemia, renal impairment, anemia, or bone lesions attributed to plasma-cell proliferation. SMM was defined with bone marrow plasmacytosis exceeding 10%, monoclonal protein level greater than 30 g/L, in the absence of ROTI [19]. The diagnostic definition of PCL is based on Kyle's criteria, where peripheral blood plasma cell absolute count greater than 2 × 10 9 /L or percentage of the while blood cells more than 20% [20,21].
In GSE39754, the DNA microarray data of CD138+ myeloma cells from 170 newly diagnosed MM patients, and plasma cells (PCs) from 6 normal donors, were quality controlled and normalized with the aroma Affymetrix package. The gene expression level was estimated with a probe level model (PLM) [22]. In GSE5900, International Myeloma Working Group criteria were used to classify patients as having MGUS, SMM, or symptomatic MM [19]. In GSE6477, Bone marrow aspirate samples were obtained and enriched for CD138+ cells. In GSE64552, bone marrow samples were obtained from 20 patients with MGUS, 33 with high-risk SMM and 41 with MM. All samples corresponded to newly diagnosed untreated patients [22]. In GSE2113, the gene expression profiles of purified plasma cells (PCs) were purified from bone marrow Series, after red blood cell lysis with 0.86% ammonium chloride, using CD138 immunomagnetic microbeads [22]. In GSE13591, pathological bone marrow specimens from 41 MM and 4 plasma cell leukemia (PCL) patients at diagnosis (27 males; median age 67 years, range 46 -85) were obtained. The plasma cells of the samples were purified ( ≥ 90%) from the bone marrow samples. Samples in GSE2658 and GSE4204 were pre-treatment bone marrow aspirates from multiple myeloma patients [23,24]. The GSE24080 dataset was contributed by the Myeloma Institute for Research and Therapy at the University of Arkansas for Medical Sciences (UAMS, Little Rock, AR, USA). Gene expression profiling of highly purified bone marrow plasma cells was performed in newly diagnosed patients with MM. Plasma cells were enriched by anti-CD138 immunomagnetic bead selection of mononuclear cell fractions of bone marrow aspirates in a central laboratory [25].
All clinical and molecular information and microarray datasets of these patients were publicly accessible at the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo). All experiment design, quality control, and data normalization were in line with the standard Affymetrix protocols. The research was conducted in accordance with the International Conference and the Declaration of Helsinki.

Microarray analysis
All microarray data were identified in GEO, and we employed statistical analysis to investigate significantly abnormally expressed genes on every microarray dataset. Briefly, gene expression data were obtained by using Affymetrix human Genome 133 plus 2.0. All designs and quality control of the microarray experiment and data normalization were in line with the standard Affymetrix protocols. Patients with FCER1G expression values above the median for all MM patients were classified as FCER1G high , and the others were considered to be FCER1G low . P-value < 0.05 in unpaired t-test analysis and fold change (FC, log2) > 0.5 or < -0.5 was utilized to determine the differential expression of genes (DEGs).

Statistical analysis
All statistical analysis was performed by R software 3.5.0. Each dataset was first evaluated for normality of distribution by the Kolmogorov-Smirnov test to decide whether a non-parametric rank-based analysis or a parametric analysis should be used. The Fisher exact and Wilcoxon rank-sum tests were used to test hypotheses in categorical and continuous variables, respectively. The samples in the second cohort were divided into two groups (FCER1G high , n = 280, FCER1G low , n = 279) based on the median expression values of FCER1G. Different gene expression analysis was performed by the limma package [26]. The Kaplan-Meier method and Cox regression multivariate analysis were used to estimate the survival analysis, with group comparisons made by using the log-rank test. Clusterprofiler package was used to identify GO enrichment terms and KEGG pathways [27]. For all statistical analysis, P-value< 0.05 was considered significant.

The expression level of FCER1G decreased with the progression of multiple myeloma
In order to understand the expression of FCER1G in MM patients and other different myeloma stages, we employed six datasets to analyze the expression level of it. We observed that the expression level of FCER1G decreased with the progression of myeloma. Remarkably lower expression of FCER1G was found in 170 MM patients than in 6 normal donors (P = 0.0096, Fig. 1A). In GSE5900, a significant decrease of  Fig. 1D). The same trend was also found in GSE2113 dataset among MGUS (n = 7), MM (n = 39), and PCL (n = 6) (P = 0.0059, 0.19, 0.012, Fig. 1E), as well as in the GSE13591 dataset including normal donor (n = 5), MGUS (n = 11), MM (n = 133) and PCL (n = 9) (Fig. 1F). In summary, the expression of FCER1G decreased with the evolution of monoclonal gammopathy, suggesting that FCER1G might be involved in the malignant progression of myeloma.

The expression of FCER1G in MM patients between different ISS stages
To further investigate the value of FCER1G expression, we compared the expression level of FCER1G at different ISS stages in 559 MM patients. A trend of decreasing FCER1G expression level in stages I, II and III ( Fig. 2A, P = 0.19, 0.035, 0.00031). We also compared the expression of FCER1G in different serotypes of different ISS stages. In serum immunoglobulin A (IgA) group and serum immuno-globulin G (IgG) group, the expression of FCER1G in stage I, II and III decreased gradually. However, there was no statistical significance in the serum free light chain (FLC) group (Fig. 2B, FLC: P = 0.41, IgA: P = 0.0085, IgG: P = 0.014, Kruskal-Wallis test). These results indicated that low expression of FCER1G correlated with the severity of MM.

Differences in clinical and other classic prognostic biomarkers in MM between FCER1G high and FCER1G low groups
Using the GSE24080 dataset of 559 MM patients, we also analyzed the baseline characteristics between high and low FCER1G expression groups. We divided the samples into two groups based on the median value of FCER1G expression: FCER1G low (n = 279) and FCER1G high (n = 280). Between the two groups, there are no significant differences in the demographic factors, such as age, gender and race. However, FCER1G was more likely to be associated with isotype (P = 0.019), cytogenetic abnormality (P = 0.021) and different therapy options (P = 0.013). Additionally, the MM patients with low FCER1G expression were more likely to have a higher beta-2 microglobulin (B2M), creatinine (CREAT), aspirate plasma cells (ASPC) ,bone marrow biopsy plasma cells (BMPC) and lower

Prognostic value of FCER1G expression in MM
By using the Cox regression model, we computed multivariate hazard ratios for different variables of 559 MM patients. Univariate analysis results showed that FCER1G and albumin (ALB), beta-2 microglobulin (B2M), bone marrow biopsy plasma cells (BMPC), hemoglobin (HGB), number of magnetic resonance imaging (MRI) were all closely related to EFS and OS with significant P values (Table  2). Furthermore, in multivariate analysis for EFS, the hazard ratio of hemoglobin was 0.66 (P = 0.023), while the hazard ratio of FCER1G expression was 0.7 (P = 0.024). These two factors were significantly related to the EFS in MM patients. For OS, the hazard ratio of beta-2 microglobulin was 1.66 (P = 0.007) and the hazard ratio of FCER1G expression was 0.69 (P = 0.02), indicating that both had a close association with OS. The hazard ratio of albumin and number of magnetic resonance imaging were 0.58 and 1.93 (P = 0.001, <0.001). FCER1G expression value was a stable factor affecting the survival level of MM patients (Table 3).

The expression of FCER1G in different amplification levels of 1q21
1q21 amplification is associated with poor prognosis, and FCER1G is located on chromosome 1q23.3. We compared FCER1G expression level under the different amplification of 1q21. There was a statistically significant difference of the expression levels between different levels of 1q21 amplification. The expression of FCER1G was decreased with the amplification of 1q21 (Fig. 3A, P = 0.0036, Kruskal-Wallis test).

FCER1G predicts the survival level in MM
From all the results above, we could assume that the low expression of FCER1G was related to adverse outcomes of MM. Thus, we further analyzed the survival level in the second cohort. We found that the FCER1G low group had significantly shorter OS compared to the FCER1G high group in two independent datasets of GSE2658 and GSE4204 ( Fig.  4A and B, P = 0.0014, 0.00065, respectively). The same prognostic value of FCER1G in MM was also found in GSE24080 ( Fig. 4C and D, OS, P = 0.0019, EFS, P = 0.0057). Likewise, the survival level retains similar results at the milestone points of Year 2008 ( Fig. 4E and F, OS, P = 0.0029, EFS, P = 0.0049).

Different expression and pathway analysis for DEGs of FCER1G high versus FCER1G low
In order to find the genes associating with FCER1G, we analyzed the differential expression values of FCER1G high versus FCER1G low . As many as 709 genes were up-regulated and 14 genes were down-regulated (P< 0.05, (FC, log2)> 0.5 or < -0.5, Fig.  5A). Heatmap showed the top 15 up-regulated genes and 14 down-regulated genes (Fig. 5B). By using the DEGs, we analyzed the enriched GO terms and KEGG pathways. Among the biological process terms of GO, most of DEGs were enriched in leukocyte migration (GO:0050900), cell chemotaxis (GO:0060326), humoral immune response (GO:0006959), and regulation of inflammatory response (GO:0050727) (Fig. 5C). In the KEGG analysis results, Staphylococcus aureus infection (hsa05150), Systemic lupus erythematosus (hsa05322) and complement and coagulation cascades (hsa04610) were the most enriched pathways (Fig.  5D).

Module screening from the PPI network
Finally, all the top 29 DEGs of FCER1G high versus FCER1G low were used to calculate the correlativity between those genes (Fig. 6A). We also screened the protein-protein interaction (PPI) network in the String database by using the top 29 DEGs [28]. Most of the up-regulated genes and two down-regulated genes (MYC and HIST1H2AD) were interactional in the PPI network (Fig. 6B). Then we discovered two subnetworks by using MCODE in Cytoscape (Fig. 6C, D). In the PPI network, C1QB, C1QA, C1QC, CD163, CD14, S100A8, S100A9, LTF, LYZ and FCGR3A were all reported to be associated with MM in early research. FCER1G acts as a core gene in both the general network and two subnetworks.

Discussion
Our research demonstrated that the expression level of FCER1G showed a decreasing trend in the deterioration of plasma cell malignancy. Higher expression of FCER1G in MM patients was associated with favorable prognosis. Likewise, the GO and KEGG pathways mainly enriched in defense response, immune response, and inflammatory response. PPI network also revealed that many cancer-associated genes interacted with FCER1G. All of these results show that FCER1G may be a tumor suppressor gene in myeloma.
Early studies found that FCER1G transduced activation signals from various immunoreceptors [10,29]. It was functionally linked to mediate neutrophil activation and was also involved in platelet activation. Associated diseases included Bleeding Disorder, Platelet-Type, 11(BDPLT11) and Mitochondrial Complex I Deficiency. FCER1G also engaged in many immune responses and played a tumor-promoting role in many kinds of tumors, such as meningioma, Clear cell renal cell carcinoma (ccRCC), childhood leukemia and Acute Myeloid Leukemia(AML) [12,14,15,30]. It was also reported that the demethylation of FCER1G was induced by IL15 in the NKp30+CD8+ T cell population exhibiting high natural killer-like antitumor potential [31]. FCER1G inhibits the expression of certain Alzheimer's disease susceptibility genes by participating in Herpes simplex (HSV-1) escape strategy [9]. Interestingly, the abundant expression of FCER1G was found in the circulating tumor cells of a prostate cancer patient who was sensitive to docetaxel chemotherapeutic reagent [32]. In the PPI network, many important genes that were associated with MM had been screened. S100A9 was reported significantly down-regulated in MM patients and further support MM survival by stimulating angiogenesis and cytokine secretion [33,34]. S100A9 was directly implicated in promoting Myeloid-derived suppressor cells (MDSC), which plays a critical role in the MM progression and can be considered as a therapeutic target in this disease [35]. C1QB, C1QA, and C1QC were all the complement c1q chains. Early reports showed that complement c1q acts in the tumor micro environment as a cancer-promoting factor independently of complement activation [36]. LTF has identified as a Cereblon (CRBN) binding protein and established relevance to MM biology [37]. LYZ was found as an element of the 9-genes prognostic signature and might be an independent prognostic factor in patients with multiple myeloma [38]. HLA-DRA plays an important role in bone lesions common in MM patients by participating in the immune response activation pathway [39]. CD163 is a tumor-associated macrophage marker, the high level of Monocyte/ macrophage-derived soluble CD163 was associated with higher stage according to the ISS and with other known prognostic factors in multiple myeloma [40,41]. MYC activation is associated with hyperdiploid MM and shorter survival, and also plays a causal role in the progression of monoclonal gammopathy to multiple myeloma. MYC protein overexpression is a feature of progression and adverse prognosis in multiple myeloma [42][43][44][45]. In recent research, it was also proved that sialyltransferase inhibition leads to inhibition of tumor cell interactions with VCAM1, and improves survival in a human multiple myeloma mouse model [46]. FCGR3A was proved to be associated with anti-tumor response. [47]. The polymorphisms of FCGR3A play an important role in First-Relapsed ovarian cancer, metastatic breast cancer, and metastatic colorectal cancer [48][49][50][51]. Early research also found that FCGR3A was associated with infections of MM patients [52].
The PPI results showed that FCER1G was a hub gene in the network and directly interact with many MM associated genes. As we demonstrated previously, downregulation of FCER1G expression was closely related to the deterioration of myeloma. Combined with GO and KEGG analysis results above, FCER1G might interact with other MM associated genes and was mainly involved in leukocyte migration, cell chemotaxis, and immune and inflammatory response pathway, and therefore exerted an anti-cancer effect in multiple myeloma.

Conclusions
To sum up, we have clearly demonstrated that high FCER1G expression was a good prognostic factor in MM patients. The expression level of FCER1G decreased with the progression of myeloma. Moreover, GO term enrichment, KEGG pathways, and PPI networks involved in MM provided insights into the pathogenesis processes associated with varying FCER1G expression. The underexpression of FCER1G could serve as a promising therapeutic target for MM patients.
However, in this research, the exact pathophysiologic role of FCER1G in myeloma cells was not been fully demonstrated. Further studies including the molecular mechanism and deeper genomic research of FCER1G in myeloma deterioration will be urgently required. and led the whole project. All the authors revised and approved the final version of the manuscript.

Availability of data and materials
The datasets of this report were generated by GEO.

Ethical approval and consent to participate
This study was approved by the Helsinki declaration and its subsequent amendments.