Cystatin SN (CST1) Is a Poor Independent Prognostic Biomarker for Gastrointestinal Diffuse Large B-Cell Lymphoma

Introduction

Emerging evidence is showing a notable correlation between protease activity and unfavorable prognosis of cancers [1,2]. Therapeutically, the mechanism of the protease-involved cancer progression is regarded to be degradation of the extracellular matrix [3]. Cysteine proteases, such as cathepsins, papain, and calpains, are kinds of proteolytic enzymes that are widely distributed in tissues [4]. Among the cysteine proteases family, cysteine cathepsins play a variety of functions, including antigen presentation, bone modeling, and epidermal homeostasis [5]. Cathepsin B (CatB) is a member of the cysteine cathepsin family that acts in cellular processes, including tumor development and invasion. Extracellular CatB was demonstrated to facilitate the degradation of extracellular matrix proteins, including collagen, laminin, and fibronectin, thus promoting tumor metastasis through the remodeling of the extracellular environment [6]. In several cancer types, such as breast cancer, colorectal cancer, or endometrial cancer, CatB can promote tumor proliferation, invasion, and metastasis [7]. Cystatin SN (CST1), belongs to a member of the type 2 cystatin family, and is a specific inhibitor of CatB. However, CST1 was proven to promote tumorigenesis, increase cancer cell proliferation and invasion and tumor recurrence in previous studies [8–10], which is enigmatic as a CatB inhibitor, because CST1 is supposed to be tumor suppressor, considering the definite role of CatB in cancer progression.

Lymphoma is a group of highly heterogeneous tumors originating from lymphocytes in the blood and lymph, mainly including non-Hodgkin lymphoma (NHL) and Hodgkin lymphoma (HL) [11]. NHL and HL account for approximate 90% and 10% of all lymphomas, respectively [12]. The 5-year overall survival (OS) rates in the United States for all NHL and HL are 69% and 8%, respectively [13,14]. B-cell lymphoma is the dominant subtype of NHL, accounting for 85% to 90% of all NHL cases, while T-cell and natural killer (NK)-cell neoplasms account for only 10% to 15% [15]. Diffuse large B-cell lymphoma (DLBCL) and the follicular lymphoma are the 2 main histological types of B-cell lymphoma, with an incidence rate as high as 65% for all NHL [16].

DLBCL is characterized by its clinical and molecular heterogeneity. The current classification of DLBCL includes the International Prognostic Index (IPI) and the cell-of-origin classification [16]. The gastrointestinal tract is the most common site of extra-nodal NHL, accounting for 30% to 40%, and the stomach is the most common site of primary gastrointestinal NHL, accounting for 60% to 70%, followed by the small intestine, ileum, cecum, colon, and rectum [17]. Moreover, gastrointestinal tract diffuse large B-cell lymphoma (GI-DLBCL) is the most common histological subtype of NHL, followed by mucosa-associated lymphoid tissue lymphoma [18,19]. Compared with the study of DLBCL, that of GI-DLBCL is insufficient, and there are no well-accepted risk factors, prognostic biomarkers, or staging systems for GI-DLBCL. An effective prognostic biomarker and biomarker-based staging system is essential to stratify high-risk patients and guide individual treatment, especially for tumors that lack a well-acknowledged treatment guideline. Unfortunately, studies of GI-DLBCL biomarkers are rare, partially because of the difficulty in establishing a post-surgical cohort.

To investigate the potential prognostic biomarkers of GI-DLBCL, we detected CST1 expression and evaluated its clinical significance by analyzing the relationship between CST1 expression, clinicopathological factors, and outcome. In detail, we established a GI-DLBCL cohort, enrolling 84 patients with GI-DLBCL. The expression of CST1 was detected with immunohistochemistry (IHC) in these patients, and the correlations between CST1 expression and clinicopathological factors were detected with the chi-square test. The prognostic significance of CST1 was evaluated by univariate analysis with the Kaplan-Meier method, and independent prognostic risk was identified by multivariate analyses with the Cox regression model.

Material and Methods

PATIENTS AND ETHICS:

From 2005 to 2020 in the Second Affiliated Hospital of Shandong First Medical University, a total of 84 patients underwent surgical resection because of a gastrointestinal tumor, and the routine pathology analysis confirmed the diagnosis as GI-DLBCL. These patients did not receive pre-operational treatment, such as chemotherapy. The follow-up was obtained by telephone enquiry. The clinicopathological factors were retrieved from the patients’ medical records, and the specimens were obtained from the Pathological Department after the approval of patients. The cohort consisted of 30 male patients and 54 female patients, with an average age of 55 years. The enrolling criteria included that (1) patients underwent radical surgical resection, and there was enough specimen for IHC, and (2) there was available follow-up and medical records. The excluding criteria included (1) post-operational survival less than 3 months, and (2) patients had other malignancies. The overall survival time was defined from the time of surgery to patients’ death or censoring. In our study, the follow-up time was from 6 to 93 months, with an average time of 37.2 months. All patients were staged according to the Lugano staging system for gastrointestinal non-Hodgkin lymphoma [20]. The study was approved by the Institutional Review Board of Shandong First Medical University, and informed consent was obtained from the patients or their legal guardians. CHOP (cytoxan+adriamycin+vincristine+prednisone) strategy was applied for the patients who required chemotherapy.

IMMUNOHISTOCHEMISTRY:

The diagnosis of GI-DLBCL was identified by hematoxylin and eosin staining and the IHC staining of DLBCL biomarkers, such as CD19 (1: 100, Abcam, ab134114), CD20 (1: 100,Abcam, ab78237), CD5 (1: 50, Santa Cruz,sc-1180), Ki67 (1: 100, Abcam, ab15580), and PAX5 (1: 50,Santa Cruz, sc-13146). Expression of CST1 was detected with IHC, and the results were evaluated with a senior pathologist for semi-quantification. IHC of CST1 was performed with the streptavidin peroxidase complex method. Briefly, the paraffin-embedded specimens were de-paraffinizated and rehydrated with xylene and graded alcohol first. The slides were immersed in citrate buffer (pH=6.0), which was boiled by microwave for 30 min for optimal antigen retrieval. After that, the specimens were incubated in 3% hydrogen peroxide to block the endogenous peroxidase. The primary antibody of CST1 (1: 100, Abcam, ab124281) and the corresponding secondary antibody were used to incubate the specimen in order. Final visualization of CST1 was achieved by incubation with 3,3′-diaminobenzidine solution.

SCORING SYSTEM OF IHC RESULTS:

Results of the IHC scores were evaluated by 2 senior pathologists unaware of the clinical data. The staining intensity and percentage of stained cells were scored independently, and the semi-qualified IHC scores were the multiplied product of the staining intensity and the percentage of stained cells. The scores of staining intensity were identified as score 0 for negative staining, 1 for weak, 2 for moderate, and 3 for strong staining. The scores of positive tumor cells were defined as score 1 for <25% of positively stained cells; 2 for 25% to 50% of positively stained cells, 3 for 50% to 75% of positively stained cells, and 4 for more than 75% positively stained cells. As the product of the multiplication of these 2 scores, the final IHC score ranged from 0 to 12. The IHC cut-off was defined by the receiver operating characteristic (ROC) curve. The point with the highest sensitivity plus specificity was set as the cut-off. In our study, the cut-off of CST1 was 5.0 in ROC curve, representing that a score higher than 5.0 was set as the CST1^high subset.

REAL-TIME QUANTITATIVE PCR (RT-QPCR):

Total RNAs of the fresh GI-DLBCL tissues and adjacent normal GI tissues were extracted by TRIzol reagent (Thermo Fisher), according to its manual. After detection of OD260/280 to confirm RNA purity, reserve transcription and quantitative amplification were performed with the RT-qPCR RT kit (Qiagen, Venlo, Netherlands) and SYBR Green Master Mix (Thermo Fisher). The 2−ΔΔCT calculation method was used for qualification with GAPDH as an internal control. ΔΔCT=(CT,Target-CT,GADPH)Time × -(CT,Target - CT,GADPH)Time 0, according to a previous study [21]. Time x is any time point and Time 0 represents the 1×expression of the target gene normalized to GADPH. Each qPCR experiment consisted of at least 8 replicates. The primers were as follows:

STATISTICAL ANALYSIS:

All the statistical significance was analyzed by SPSS 22.0 (IBM Corp, Armonk, NY, USA). The correlations between CST1 and the clinicopathological factors of GI-DLBCL were analyzed by the chi-square test, and the effect size was analyzed by the Phi coefficient. The Kaplan-Meier method was used to estimate the OS rate, and the log-rank test was applied to calculate the statistical significance. After the validation of proportional-hazards assumption, the Cox regression hazards model was used to identify the independent prognostic factors. A P value <0.05 was regarded as statistically significant.

Results

RETROSPECTIVE COHORT OF GI-DLBCL:

A retrospective cohort including 84 GI-DLBCL patients was established, consisting of 54 male and 30 female patients. The ratio of male to female patients was 1.8/1(54/30). The average age of these patients was 55.0 years. This cohort included 47 patients with gastric DLBCL and 37 patients with intestinal DLBCL. Data of the clinicopathological factors, including tumor diameter, macroscopic type, lesion number, extranodal involvement, tumor site, chemotherapy, and Lugano stage, were all retrieved from the patients’ medical records (Table 1).

EXPRESSION OF CST1 IN GI-DLBCL:

Moreover, CST1 expression in the 84 patients with GI-DLBCL was investigated with IHC, which divided the cohort into subsets with low and high CST1 expression (Figure 1A). The percentages of subsets with low and high CST1 expression were 72.62% and 27.38%, respectively (Table 1). The intracellular localization of CST1 in GI-DLBCL is mainly in cytoplasm. Moreover, in 7 pairs of GI-DLBCL tissues and tumor-adjacent tissues, expression of CST1 mRNA was detected with RT-qPCR (Figure 1B). Interestingly, levels of CST1 mRNA in GI-DLBCL tissues were substantially higher than that in adjacent tissues.

CORRELATION BETWEEN CST1 EXPRESSION AND CLINICOPATHOLOGICAL FACTORS:

The correlations between CST1 expression and clinicopathological factors were analyzed with the chi-square test (Table 2). Among the clinicopathological factors, including the sex and age of patients, tumor diameter, macroscopic type, lesion number, tumor site, chemotherapy, Lugano stage, serum lactate dehydrogenase (LDH), performance status (PS), and IPI, IPI was positively correlated with CST1 expression (P=0.007), showing that patients with high CST1 expression seemed to have higher IPI scores (Table 2). Moreover, serum LDH tended to be correlated with CST1 expression (P=0.053).

PROGNOSTIC SIGNIFICANCE OF CST1 IN GI-DLBCL:

The prognostic significance of CST1 and the clinicopathological factors of GI-DLBCL were estimated with univariate analysis. The statistical significance between different subtypes in the univariate analysis was calculated with the log-rank test. In our study, CST1 expression was a prognostic biomarker of GI-DLBCL (P=0.012; Figure 2A). The 5-year OS rates of patients with low and high CST1 expression were 73.9% and 54.5%, respectively (Table 3). Moreover, we divided the GI-DLBCL patients into gastric DLBCL (gDLBCL) and intestinal DLBCL (iDLBCL) and analyzed the prognostic significance of CST1 in both subtypes. As the result, high CST1 expression indicated poor outcomes of both gDLBCL and iDLBCL (P=0.007 and 0.039, respectively; Figure 2B, 2C).

Additionally, high serum LDH levels (P=0.002), high PS (P=0.003), advanced Lugano stage (P=0.002), and high IPI (P=0.001) were also indicators for poor prognosis of GI-DLBCL (Figure 2D–2G). Other clinicopathological factors, including the sex and age of patients, tumor diameter, macroscopic type, lesion number, and tumor site had no significant correlation with OS (Figure 3A–3F). Interestingly, post-operational chemotherapy after the radical surgery had no strong correlation with the OS rate in our study, which may suggest that not all patients after surgery would benefit from post-operational chemotherapy (Figure 3G). However, this result certainly needs multi-centered and prospective clinical study for further validation.

To test the independent prognostic significance of CST1 in GI-DLBCL, we performed multivariate analysis with the Cox regression hazard model (Table 2). The significant prognostic factors in univariate analysis, including serum LDH, PS, Lugano stage, and CST1 expression, were included in the multivariate analysis for validation of independent prognostic significance. IPI was excluded from the Cox regression hazard model because it was the comprehensive score of the above variables. In our study, CST1 was identified as an independent prognostic biomarker of GI-DLBCL (P=0.043). The hazard ratio (HR) of high CST1 expression was 3.07, compared with that of low CST1 expression. In addition, PS (P=0.007, HR=4.21) and Lugano stage (P=0.018, HR=3.63) were the independent prognostic factors, predicting the poor outcomes.

Discussion

The understanding of the biology, genetics, and diagnostic methods of NHL has been significantly improved in recent decades. However, the understanding of GI-DLBCL, the most frequent extra-nodal NHL, remains stagnant. The treatment of GI-DLBCL includes surgery, radiotherapy, and chemical immunotherapy, and the prognosis is usually favorable after appropriate treatment. The current prognostic factors include disease stage, histopathological classification, and immunophenotype, but there are still no well-accepted biomarkers of GI-DLBCL. Having no available specifically prognostic biomarker and risk stratification system of GI-DLBCL limits the individualized therapy and precision treatment of GL-DLBCL. The most widely-used stage system is the IPI system, which uses superficial variables such as serum LDH, performance status, Lugano stage, and extranodal invasion. Effective biomarkers and molecular classification are urgently needed. Studies on the DLBCL biomarkers are common, but studies on GI-DLBCL prognostic biomarkers are rare. Coexpression of MYC and Bcl-2 has prognostic implications in DLBCL, but its clinical significance in GI-DLBCL is unclear [22]. In a previous study, programmed cell death ligand 1 expression on microenvironment immune cells had a favorable impact on the outcome [23]. Moreover, β2 microglobulin and lesion length after chemotherapy were independent predictors of gastric DLBCL [24]. Recent advances classified DLBCL into non-germinal center B-cell-like lymphoma (GCB) and non-GCB or activated B-cell-like (ABC) subtypes based on the cell of origin profile, and ABC lymphomas show a poorer prognosis than do GCB lymphomas [25]. Fortunately, more attention was paid to lymphoma originating from GI tract [26]. Many new treatment strategies of GI lymphoma have emerged, including antibody-based therapy, such as monoclonal antibody and bispecific T-cell binding antibody, immunomodulatory drugs, molecular targeted therapies, and cell-based therapies [26,27]. Although this large number of new drugs has been created and added into the regimen combination of DLBCL, a randomized controlled trial with a large sample size is absent, and there is still no definitive conclusion.

Rituximab is the most widely used monoclonal antibody for GI-DLBCL treatment, and rituximab plus CHOP (R-CHOP) therapy has been shown to have a better outcome than CHOP [25]. R-CHOP chemotherapy plus immunotherapy is the standard treatment method for DLBCL. As for GI-DLBCL, there are quite more treatment options, and current option is that a multidisciplinary approach should be applied, including chemotherapy, surgery, radiotherapy, or a combination of these modalities. However, the best combination of the above options has not received full consensus. Continuous efforts keep being made to determine the optimal management of GI-DLBCL, but the studies are retrospective and heterogeneous [28–30].

Lymphoma is extremely heterogeneous, and each histological types has distinct biological characters. Although GI-DLBCL is the main type of extra-nodal non-Hodgkin lymphoma, the incidence of GI-DLBCL is very low, compared with that of common malignancy [31].

Here, we constructed a retrospective cohort consisting of 84 patients, which is a relatively large cohort in a GI-DLBCL study. Based on this cohort, we identified CST1 as a new and effective biomarker of GI-DLBCL, indicating poor prognosis. The sample size of our cohort is large for GI-DLBCL, but is not large enough for a common tumor such as lung cancer, because insufficient sample size would increase the potential biases. Therefore, a prospective and multi-center cohort is certainly needed to further confirm the prognostic significance of CST1 in GI-DLBCL. However, building a prospective study of GI-DLBCL requires long-time effort because the incidence of GI-DLBCL is low, most cases of GI-DLBCL have an indolent course, and the follow-up should be very long (normally more than 5 years). Moreover, the treatment options of GI-DLBCL varies, making the homogeneous cohort even more difficult. Our results can help develop the molecular classification of GI-DLBCL and may promote precision in the treatment of GI-DLBCL. We hope that our conclusions, which uncover the clinical significance of CST1 in GI-DLBCL, trigger the interest of researchers and promote the establishment of a prospective and multi-center cohort.

Cystatins are essential to cells and tissues because they can protect them from inappropriate proteolysis. The CST1, CST2, and CST4 genes encode the S-type (salivary) cystatins SN, SA, and S, respectively [32]. CST1 protein is mainly distributed in the submandibular gland, gallbladder, and uterus, but it is also highly expressed in malignancies. To date, the mechanism of how CST1 induces cancer progression has not been well studied. The existing studies provide several possible explanations but there is no consensus. For example, CST1 was reported to promote tumorigenesis of colorectal cancer via recovering CTSB activity by competition with CST3 [33]. Moreover, CST1 is considered to suppress cathepsin and promotes progression of gastric cancer by promoting the Wnt signaling pathway [34]. The substrates of CST1 include papain, ficin, cathepsin C (CTSC), and HSV-1 [35]. The molecular mechanism of the CST1 pro-tumor role has not received full consensus. In gastric cancer, CST1 can promotes gastric cancer progression by inhibiting ferroptosis via regulating GPX4 protein stability or by activating the Wnt pathway [36,37]. In lung cancer, CST1 can promote epithelial-mesenchymal transition and serves as a prognostic biomarker [38]. However, our present study did not investigate the underlying mechanism of CST1-involved poor prognosis of GI-DLBCL, partially because there is no proper cell or animal model of GI-DLBCL.

Conclusions

In summary, we established a GI-DLBCL cohort of 84 patients and identified CST1 as a prognostic biomarker of GI-DLBCL for the first time. CST1 is highly expressed in GI-DLBCL, compared with in the normal GI tissues. High expression of CST1 is an independent prognostic biomarker indicating an unfavorable prognosis. Our results suggest that CST1 detection could be a promising method to stratify high-risk patients and guide individual treatment. Targeting CST1 may be a potential approach for future treatment of GI-DLBCL.