Extranodal natural killer/T-cell lymphoma (ENKTL) is an aggressive and unusual subtype of non-Hodgkin lymphoma (NHL) that is associated with the Epstein-Barr virus (EBV) (1, 2). The incidence of this disease is relatively high in East Asia and South America, while it is low in the Western world. The nasal cavity is the most frequently affected site, often involving the nasopharynx and non-nasal areas such as the skin, gastrointestinal tract, lung, liver, salivary gland, and testis (3). Fine needle aspiration or core needle biopsy is widely used in ENKTL diagnosis. Necrosis is often seen in diagnostic biopsies and this may result in a delay in diagnosis. In a study with a median follow-up of 44.9 months, 26% of all stage I or II patients and 71% of stage III or IV patients died (4). Concurrent chemoradiation, sequential chemoradiation, and sandwich chemoradiation are recommended to treat stage I or II patients by the National Comprehensive Cancer Network. Stage III or IV patients were suggested to undergo these treatments followed by radiotherapy (5–7). Disease progression was still the most important factor leading to death, while pathogenesis of ENKTL is still unknown. Recently, increasing numbers of biomarkers related to disease progression and prognosis have been described. The overexpression of enhancer of zeste homolog 2 (EZH2) and H3K27-specific histone methyltransferase in ENKTL patients' tissue is associated with disease progression and prognosis (8). The atypical cells found in the cerebrospinal fluid (CSF) of patients with ENKTL are positive for CD3, CD43, and CD56, and some studies use these markers to diagnose ENKTL (9, 10).

The CSF is formed primarily in the ventricular choroid plexus and exists within the ventricular system and subarachnoid space. The proteins in the CSF originate from three different sources: cells within the CSF, plasma, and neural tissue (11). The CSF is considered an ideal source of high-concentration, disease-related proteins and cells. The proteome of normal human CSF was detected using immunoaffinity depletion and 2D liquid chromatography tandem mass spectrometry with over 2,500 proteins being identified (12, 13). The proteome is widely used in the detection of protein in the CSF of patients. The CSF samples of patients with Alzheimer's disease were analyzed and the levels of 750 proteins changed, identifying four proteins that had levels different from those in healthy individuals (14). Approximately 19 proteins have been identified in eight peer-reviewed articles as proteins with differential expression proteins in the CSF of glioma patients (15). Biomarkers in the CSF to predict the conversion process from clinically isolated syndrome (CIS) to multiple sclerosis (MS) have been identified to reach a correct diagnosis (16).

To identify changes in proteomic markers of the CSF in ENKTL patients, we used label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS) to process the samples. We found that histidine-rich glycoprotein (HRG), tissue inhibitor of matrix metalloproteinase-1 (TIMP-1), serpin family A member 3 (SERPINA3), fibrinogen alpha chain (FGA), fibrinogen gamma chain (FGG), transferrin (TF), fibrinogen beta chain (FGB), amyloid beta precursor protein (APP), and angiotensinogen (AGT) significantly differed between the two patient groups. These proteins may be potential markers of ENKTL and may be related to disease progression.

To identify the differences in the proteome between ENKTL patients and non-ENKTL controls, we collected CSF samples from 12 patients in West China hospital, six of whom had ENKTL and the rest were tumor-free patients suspected of having meningitis (patient information is listed in Supplementary Table 1). The CSF samples (Supplementary Table 2) were processed and analyzed using LC-MS/MS following the process outlined in Figure 1. We adopted a strict criterion to restrict the protein content to that based on the lowest special peptide after applying a 1% FDR standard according to the value of the peptide.

In the aggregate, 421 non-excrescent and functional proteins were identified among the samples. Of these proteins, 45 proteins were quantified matching the mentioned criteria. We then performed enrichment analysis for GO based on human proteins. We discovered that the response to platelet degranulation, innate immunity, cell adhesion, complement activation, negative regulation of endopeptidase activity, and extracellular matrix organization were significantly enriched (P < 0.001) in the biological procedural GO analysis (Figure 2A). The top seven, in terms of proportion of the quantified proteins in the GO molecular function analysis (Figure 2B), were antigen binding, serine-type endopeptidase activity, serine-type endopeptidase inhibitor activity, immunoglobin receptor binding, carbohydrate binding, receptor binding, and structural molecule activity. The cellular component analysis of proteins mainly indicated that they were highly correlated to the extracellular region (Figure 2C). The analysis of the KEGG-pathway indicated that the pathways containing complement and coagulation cascades, platelet activation, and cell adhesion molecules were enriched (Figure 2D).

In the hierarchical cluster analysis of the CSF proteome in ENKTL and control patients, only the differential expression proteins were taken into consideration (Figure 3). Twelve proteins were downregulated and 33 proteins were upregulated. We then used stings to analyze protein-to-protein interaction (Figure 4). We identified nine core proteins that were closely connected to each other and changed drastically between ENKTL and control patients. HRG, TIMP-1, SERPINA3, FGA, FGG, TF, FGB, APP, and AGT were all higher in ENKTL patients (Supplementary Figure 1).

To identify the differences between ENKTL patients and control patients, we performed LC-MS/MS analysis. In our study, we identified 421 proteins, of which 45 proteins were significantly different between groups. HRG, TIMP-1, SERPINA3, FGA, FGG, TF, FGB, APP, and AGT were significantly upregulated in ENKTL patients.

APP, A2M, and SERPINA 3 are acute phase proteins. APP is a single transmembrane protein expressed in a wide variety of cells and has a receptor-like structure. Several pathophysiological functions of APP have been discovered. APP is engaged in neuronal growth cone adhesion and plays an independent role as an operating cell adhesion molecule to bind to extracellular matrices (19). It was reported that the expression of APP is increased in both malignant breast cancer cell lines and breast cancer tissues (20). APP is also overexpressed in the papillary thyroid carcinoma and colorectal carcinoma (17, 21). Furthermore, the overexpression of APP is related to the prognosis. The function of A2M, which is a pan-proteinase inhibitors of the A2M family, is different from that of APP (22). A2M acts as a carrier protein and regulator for growth factors, polypeptide hormones, and cytokines. It is widely accepted that A2M is upregulated in human hepatocellular carcinoma (23). SERPINA3 is a protein of the serpin superfamily and is also known as α1-antichymotrypsin. SERPINA3 restricts the activation of several serine proteases, including chymotrypsin and cathepsin G. The overexpression of SERPINA3 leads to loss of adherence and delay of apoptosis (24). The upregulation of SERPINA3 has been reported in some cancer types, including endometrial carcinoma, lung carcinoma, colon carcinoma, breast carcinoma, and melanoma (25–30).

FGA, FGB, and FGG were all upregulated in the ENKTL patients. Two sets of α, β, and γ chains compose a large plasma protein called fibrinogen. These three chains are encoded by three independent genes. There are a number of reasons for fibrinogen upregulation, of which malignant tumor accounts for 20% based on a Chinese article. Previous studies indicate that the serum level of FGA is upregulated in many tumors, such as gastric cancer, breast cancer, nasopharyngeal carcinoma, acute lymphocytic leukemia, esophageal squamous cell carcinoma, and renal cell cancer (31–35). The abnormal expression of FGG was also reported in hepatocellular carcinoma and ovarian cancer (36). Furthermore, it was reported that the upregulation of FGB raised the expression of the other two. These three chains are interactively regulated in human cancer cells (37). The mechanism of the overexpression of these three proteins needs to be further explored.

TIMPs are specific inhibitors of matrixins and take part in controlling the local activities of matrix metalloproteinases (MMPs) in tissues (38–40). There are four known inhibitor subtypes referred to as TIMP-1, TIMP-2, TIMP-3, and TIMP-4. TIMP-1 inhibits MMP-9 with a high affinity. It has been proved that TIMPs promote cell growth and have anti-apoptotic activity (41). Disruption of the MMP-TIMP balance can lead to a number of pathogenic processes including tumor invasion, metastasis, and angiogenesis (42). According to a previous study, TIMP-1 levels are upregulated in non-small cell lung cancer and advanced breast carcinoma (43, 44). Additionally, it was reported that higher expression of TIMP-1 occurs in adenocarcinoma cells related to the higher stromal intensity of squamous cell carcinoma and non-small cell lung cancer (43). It was reported that the expression of TIMP-1 upregulated and promoted the development of EBV-associated malignancies in EBV-infected cells (45).

Previously, studies have shown that upregulation of SERPINA3 expression was an index of worse prognosis in ovarian cancer than low SERPINA3 level (46). A high TIMP-1 level was also associated with a worse prognosis (47, 48). The relationship of APP to prognosis are yet to be explored in tumors. In a further study, we will follow patients to observe the progression-free survival and overall survival and elucidate the relationship between proteins and survival.

There are some limitations of our study. First, the number of patients included in this research was limited. Second, there are various subtypes of NHL. ENKTL, an aggressive type of NHL, is special and difficult to distinguish. Therefore, more variable morphologic and immunophenotypic features are required when narrowing the differential diagnosis. The control group was also limited by the number of involved patients. In further studies, more control groups can be used, including diffuse large B cell lymphoma, Burkitt's lymphoma, or other types of lymphoma.

We performed a proteomic analysis of the CSF in ENKTL patients, using a label-free method to compare proteomic marker variation between tumor patients and non-tumor patients. We discovered that HRG, TIMP-1, SERPINA3, FGA, FGG, TF, FGB, APP, and AGT were significantly upregulated and are potential markers for diagnosing ENKTL, either individually or in combination with each other. It is possible that these proteins play an important role in the occurrence of ENKTL lymphoma. The proteomic changes are a potential method of precisely identifying the pathogenesis of ENKTL.

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

This study was approved by Review Board of Sichuan University. Patients Informed Consent was obtained from all participating patients or guardians in accordance with a University of Sichuan Review Board-approved protocol. This study was carried out in accordance with the recommendations of Sichuan University with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Sichuan University.

QL and XM mainly generated with the idea and collected the data. QL wrote the manuscript. HZ and YZ helped to correct the grammar and does some statistical work. YG did the trials.