Introduction

Lymphoma is a neoplasia of the immune system. Non-Hodgkin lymphoma (NHL) is the main subtype, accounting for around 90%, while the remaining 10% is Hodgkin lymphoma (HL) (Shankland et al. 2012). There are slight geographical variations in the incidence of individual subtypes, with NHL being more common and HL relatively rare in Asia (Saito and Matsuoka 2020). It has been estimated that there will be 85,720 newly diagnosed lymphoma patients and 20,910 deaths in the United States in 2020 (Siegel et al. 2020). Although the striking therapeutic advancements have led to substantial improvements in lymphoma outcomes, the hematological malignancy still threatens human life and impairs the quality of life seriously. Therefore, it is urgent to find more reliable prognostic biomarkers for better control of lymphomas.

Immune cells provide an ideal source for the development of novel prognostic biomarkers due to their ease of detection.The immune components play diverse roles in tumor progression, either inhibiting or promoting tumor growth. The immunosuppressive cells, including myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), are a rare subpopulation that can impair the anti-tumorimmunity via suppressing the activity and functions of T cells and natural killer cells (Ai et al. 2019; Finn 2012; Gabrilovich 2017; Giallongo et al. 2016; Görgün et al. 2013; Lv et al. 2019). For example, we have reported that MDSCs are linked with worse prognosis in a variety of malignant tumors (Ai et al. 2018). Tregs, another immunosuppressive subset, has drawn much attention in recent decades.

Tregs prevail as a specialized sub-class of CD4+ T cells with positive nuclear marker forkhead box P3 (FOXP3) (Hori et al. 2003). Multiple lines of studies have demonstrated that increased tumor-infiltrating Tregs predict poor prognosis of patients in most solid tumors, including breast cancer, ovarian carcinoma, hepatocellular carcinoma, etc. (Bates et al. 2006; Curiel et al. 2004; Sun et al. 2017). These findings could be properly explained by its immunosuppressive property and tumor-promoting function. Intriguingly, the opposite results have also been reported in some other solid tumors, such as colorectal cancer, esophageal carcinoma, and head and neck cancer(Shang et al. 2015; Xu et al. 2017).

Similarly, the prognostic impacts of Tregs in lymphoma patients are somewhat controversial. Some literature has reported that the cell count of tumor-infiltrating Tregs is a positive prognostic factor in lymphoma (Carreras et al. 2006, 2019; El-Dien et al. 2017; Greaves et al. 2013; Kim et al. 2009; Lam et al. 2020; Lee et al. 2008; Nakayama et al. 2017; Peng et al. 2011; Saifi et al. 2010; Wang et al. 2018). But more and more studies have questioned the correlation of Tregs with lymphoma outcome (Blaker et al. 2016; Chao et al. 2015; Chetaille et al. 2009; Gjerdrum et al. 2007; Gomez-Gelvez et al. 2016; Kelley et al. 2007; Lundberg et al. 2019; Muenst et al. 2009; Nam et al. 2014; Richendollar et al. 2011; Tzankov et al. 2008). Thus, it is urgent to clarify the prognostic value of Tregs in lymphoma patients based on the findings from the published studies, which were conducted independently with small sample size. To this end, we conducted the present quantitative meta-analysis and found that increased tumor-infiltrating Tregs indicates a better prognosis for patients with lymphoma.

Methods

Search strategy

This study was completed following the Standard Guide for Meta-Analysis of Tumor Marker Prognosis Research (Sauerbrei et al. 2018). We systematically searched PubMed, Embase, and Web of Science up to May 2020 to obtain studies regarding the role of Tregs in the outcome of lymphoma patients. We used the following keywords: (“Regulatory T cells” or “Tregs” or “FOXP3”) and (“survival” or “outcome” or “prognosis” or “progression” or “mortality” or “recurrence” or “metastasis”). We tried to find more related studies by reviewing the identified reviews, meta-analysis, reports, and other reference lists of related publications. Meanwhile, the “similar articles” function was used to gain more eligible publications.

Selection criteria

Studies that meet the following criteria are included in the present meta-analysis: (i) the prospective or historical cohort studies; (ii) patients diagnosed with lymphoma; (iii) the studies focusing on the association between Tregs and overall survival (OS) or progression-free survival (PFS); (iv) studies providing competent information to calculate hazard ratio (HR) and 95% confidence interval (CI) if HR is not reported directly. In contrast, the exclusion criteria are as follows: (i) case reports, reviews, meeting reports, letters, etc.; (ii) multiple published reports; (iii) studies without enough data to analyze.

Data extraction

Eligible studies were extracted by two researchers independently. Any inconsistency was assessed by the third researcher. The main information of the included studies and patients was collected, including the year of publication, the name of the first author, country of the study, sample size, follow-up time, age range, and survival data.

Each study was evaluated by two reviewers independently, based on the Newcastle–Ottawa Quality Assessment Scale (NOS). If disagreements occur between researchers, a third researcher will solve differences by discussion. NOS uses a star rating system with a total score of 9 points. Studies scored more than seven points were considered as high-quality studies, otherwise as low-quality studies.

Statistical methods

The effect of Tregs on lymphoma prognosis was quantified by the HR. We extracted HR and 95% CI directly from the literature. If they were not reported, we estimated them by using the methods reported by Parmer et al (Parmar et al. 1998). Heterogeneity among the included studies was analyzed using the I2 test and χ2 test (Higgins et al. 2003). A random-effects model was used in the meta-analysis in case of significant heterogeneity (I2 ≥ 50%). Otherwise, the fixed-effects model was used. Significant heterogeneity was analyzed by subgroup analysis and meta-regression. Publishing bias was evaluated by Egger’s test and Begg’s funnel plot (Egger et al. 1997). The level of significance was set at 0.05. All analyses were performed using Stata 12.0 software.

Results

Selection and characteristics of included studies

A total of 1951 related publications were obtained in the initial retrieve. After screening, 23 studies with a total of 2269 patients were finally included (Fig. 1). The basic characteristics of the included studies were shown in Table 1, displaying a wide distribution of 12 countries. There were 6 studies on HL and 17 on NHL, the latter consisting of 12 on B cell lymphomas (BCL), 3 on natural killer/T cell lymphoma (NKTCL), and 2 on T cell lymphomas (TCL). Of the 12 studies on BCL, 5 focused on follicular lymphoma (FL) and 7 on diffuse large B cell lymphoma (DLBCL). All the studies reported OS, while 9 studies, a total of 732 patients, reported PFS. The NOS scores for all the included studies ranged from 5 to 8.

Fig. 1
figure 1

Flow diagram of selecting relevant studies included in the meta-analysis

Full size image
Table 1 Characteristics of studies included in the meta-analysis
Full size table

The prognostic role of FOXP3+ Tregs in the survival of all lymphoma patients

We included 23 studies that reported OS data. Patients with high tumor-infiltrating Tregs density had longer OS with a pooled HR of 0.633 (95% CI 0.528–0.758, P < 0.001) and heterogeneity was significant (χ2 = 86.78, P < 0.001; I2 = 74.6%) (Fig. 2a). Additionally, 9 studies reported PFS. Similarly, elevated Tregs were associated with longer PFS, indicating a prolonged survival (HR = 0.451, 95% CI 0.261–0.779, P = 0.004). Statistical heterogeneity among these 9 studies was also observed (χ2 = 38.47, P < 0.001; I2 = 79.27%) (Fig. 2b).

Fig. 2
figure 2

Forest plots showing the relationship between Tregs and the prognosis of lymphoma. a OS; b PFS

Full size image

Group analysis on the role of FOXP3+ Tregs in the prognosis for different types of lymphomas

To demonstrate the prognostic value of Tregs in diverse subtypes of lymphomas, we conducted a subgroup analysis on individual lymphomas, including HL, FL, DLBCL, NKTCL, and TCL (Fig. 3). High density of Tregs was linked with longer OS in HL (HR = 0.471, 95% CI 0.293–0.760, P = 0.002), DLBCL (HR = 0.523 95% CI 0.334–0.818, P = 0.004), and NKTCL (HR = 0.211, 95% CI 0.112–0.399, P < 0.001), whereas no significant correlation of Tregs with the OS for TCL patients (HR = 1.017, 95% CI 0.809–1.280, P = 0.883) and FL patients (HR = 0.738, 95% CI 0.490–1.110, P = 0.144) were found. Heterogeneity was significant for FL (χ2 = 13.35, I2 = 70.0%), but not significant for HL (χ2 = 8.02, I2 = 37.6%), DLBDL (χ2 = 8.75, I2 = 31.5%), NKTCL (χ2 = 1.01, I2 < 0.1%), and TCL (χ2 = 1.14, I2 = 12.5%).

Fig. 3
figure 3

Subgroup analysis of OS by subtype of lymphoma. HL Hodgkin lymphoma, FL follicular lymphoma, DLBCL diffuse large B cell lymphoma, NKTCL natural killer/T cell lymphoma, TCL T cell lymphoma

Full size image

Risk-adjusted analysis

Among all the included studies, 12 reported the results of a multi-factor analysis of OS (Fig. 4a). The risk-adjusted analysis showed that upregulated Tregs indicated a better OS with a pooled HR of 0.679 (95% CI 0.551–0.837, P < 0.001). But the heterogeneity was significant (χ2 = 50.21, I2 = 78.9%). This finding implies that the high density of tumor-infiltrating Tregs was an independent prognostic factor of the OS in lymphoma patients.

Fig. 4
figure 4

Forest plots of multivariate analysis showing the relationship between Tregs and lymphoma. a OS; b PFS

Full size image

Only five studies provided the multivariate-adjusted analysis of PFS (Fig. 4b). The pooled HR of the multivariate-adjusted analysis showed that the density of Tregs was not significantly associated with PFS (HR = 0.525, 95% CI 0.231–1.195, P = 0.125).

Subgroup analysis and sensitivity analysis

To verify the source of heterogeneity, we performed the meta-regression and sensitivity analysis for OS (Table 2). Studies were grouped according to the region (Eastern or Western), sample size (≥ 100 or < 100), study quality (NOS scores ≥ 7 or < 7), treatment (yes or no), and detection method (whole-tissue sections or tissue microarrays). A similar association of Tregs with OS was acquired in the subgroup analysis, also with significant heterogeneity. However, the heterogeneity was slightly reduced in the subgroup analysis based on the region.

Table 2 Subgroup analysis and meta-regression
Full size table

Further, we conducted a sensitivity analysis. As shown in Fig. 5a, studies by Gjerdrum et al. (Gjerdrum et al. 2007) and Blaker et al. (Blaker et al. 2016) had an evident impact on the results of OS, suggesting that these two studies may be the main source of heterogeneity. After excluding these two studies, the pooled HR was 0.508 (95% CI 0.392–0.657) for OS with lower heterogeneity (χ2 = 36.21, P = 0.005; I2 = 44.8%). Similarly, after dropping the study of Blakeret al, the pooled HR was 0.387 (95% CI 0.257–0.585) for PFS with lower heterogeneity (χ2 = 9.61, P < 0.001; I2 = 27.1%) (Fig. 5b), suggesting that our results were reliable.

Fig. 5
figure 5

Sensitivity analysis of the overall pooled study. a OS; b PFS

Full size image

Intriguingly, there was no change in the relationship between Tregs and the OS in HL, DLBCL, NKTCL, and TCL after excluding the studies by Gjerdrum et al. (2007) and Blaker et al. (2016). However, after excluding the study of Blaker et al. (2016), the subgroup analysis showed that high Tregs were linked with longer OS in FL (HR = 0.613, 95% CI 0.423–0.889, P = 0.010).

Publication bias

To investigate publication bias, we conducted Begg’s test and Egger’s test. Grounded on the results of sensitivity analysis, two studies reported by Gjerdrum et al. and Blaker et al. were excluded. The Funnel plot was symmetric in both OS (Fig. 6a) and PFS (Fig. 6b). The results of Begg’s test (OS: P = 0.131, PFS: P = 0.266) and Egger’s test (OS: P = 0.154, PFS: P = 0.58) indicated that there was no publication bias regarding the included studies.

Fig. 6
figure 6

Publication bias of the present meta-analysis. a OS; b PFS

Full size image

Discussion

For some types of lymphoma, the stage is inadequate to determine patients’ outcome (Matasar and Zelenetz 2008). In this case, other factors, like the international prognostic index (IPI), can be utilized to evaluate the prognosis and classify patients into high-risk or low-risk groups. However, IPI is only applicable to some specific types due to the heterogeneity of lymphomas. Thus, it is urgent to develop more prognostic biomarkers so that hematologists and oncologists can clarify the prognosis of lymphoma patients more accurately. Tregs, an immunosuppressive subpopulation of T cells, play a key role in tumor development by modulating the tumor immune microenvironment. To explore the value of Tregs in the outcome of lymphoma, we conducted this study. To the best of our knowledge, it is the first meta-analysis to analyze the association of Tregs with the prognosis of lymphoma patients.

Our meta-analysis study included 23 publications with a total of 2269 patients. We showed that lymphoma patients with a higher density of tumor-infiltrating Tregs had significantly longer OS and PFS. The analysis of OS presented significant heterogeneity (I2 = 74.6%). To identify the source of heterogeneity, we performed subgroup analysis and mate-regression. Our results in Table 2 show that the region of studies may be one key source of heterogeneity. Moreover, we identified two studies that largely attributed heterogeneity through sensitivity analysis (Blaker et al. 2016; Gjerdrum et al. 2007). After excluding those studies, the association of Tregs with the prognosis was similar to the previous analysis, suggesting that our results are quite reliable.

Our study confirmed that elevated Tregs was associated with better outcomes in lymphoma patients, which is in disagreement with the studies on most types of solid tumors (Shang et al. 2015). The potential underlined mechanisms regarding the positive relationship between Tregs and lymphoma outcomes may be due to the specific immunosuppressive functions of Tregs. First, Tregs may directly induce the death of malignant B cells, which might be supported by the study reported by Lindqvist et al. (2011). In this study, the authors found that CD4+FOXP3+ Tregs isolated from patients with B cell malignancy are able to kill autologous leukemic B cells in vitro. Second, Tregs may suppress the T helper (Th) cells and promote the apoptosis of cytotoxic T cells (CTLs) (Najafi et al. 2019), which might properly explain the reasons for the positive prognostic value of Tregs in NKTCL. Additionally, the lack of functional Th cells may impair B cells (Ueno 2016), which might be an indirect way of Tregs to affect the prognosis of BCL. Third, Tregs can function in suppressing the activity of other immune components, such as dendritic cells (DCs) and NK cells (Ring et al. 2010, 2015).

The molecular mechanisms regarding the immunosuppressive properties of Tregs in hematological malignant diseases have been extensively reported. First, Tregs inhibits the function of conventional T cells (Tcons) and DCs through increasing cyclic adenosine monophosphate (cAMP) expression (Borsellino et al. 2007), leading to inhibition of the proliferation and cytokine production of Tcons (Rueda et al. 2016). Also, with decreased expressions of costimulatory molecules and increased expressions of inhibitory molecules, elevated expression of cAMP can inhibit the antigen-presenting function of DCs (Ring et al. 2010, 2015). Besides, Tregs suppresses immune response through releasing inhibitory immunomodulatory molecules, such as Interleukin-10 (IL-10) and transforming growth factor-β (TGF-β). IL-10 can promote the depletion of T cells in the tumor by regulating the expression of inhibitory receptors (Li et al. 2017; Sawant et al. 2019). TGF-β can impair the activity of natural killer cells (NK cells) and CTLs (Najafi et al. 2019). Additionally, Tregs also can induce NK cells and CD8+ T cell death dependent on Granzyme B and perforin (Cao et al. 2007). Collectively, Tregs exerts immunosuppressive effects through direct contact and release of cytokines.

Lymphomas are a group of hematological malignancies consisting of various subtypes with substantial heterogeneity and diversity of clinical and biological characteristics. Thus, it is reasonable to sub-analyzed the prognostic role of Tregs in different subtypes of lymphoma. Our subgroup analysis showed that a higher density of Tregs was associated with better prognosis in HL, DLBCL, and NKTCL, but with a neutral prognostic claim in TCL and FL. Reed-Sternberg cells (RS cells), characteristic of HL, are maintained by various cytokines. The background of inflammation suppressed by Tregs can inhibit cytokine production. There are many potential pathways as RS cells express various surface receptors (Khan 2006). DLBCL is a subtype of lymphoma with high incidence. Tregs inhibit B cell proliferation directly by inducing its apoptosis (Zhao et al. 2006). Besides, Tregs can inhibit the function of B cells indirectly via suppressing the function of T follicular helper cells (Tfh) (Marinova et al. 2007). About NKTCL, in addition to inhibiting NK cells function and inducing NK cells to death, Tregs highly express CD25, which is a high-affinity receptor of IL-2. IL-2 is necessary for the proliferation and maintenance of NKT cells. Tumor-infiltrating Tregs may compete for IL-2 with tumor cells, resulting in the deprivation of IL-2. So, Tregs improve the clinical prognosis and play the antitumor roles in NKTCL. In TCL, the function of Tregs seems particularly complicated. On one hand, the malignant cells are T lymphocyte, which could be suppressed by Tregs. On the other hand, some studies showed that the cancerous cells could be Tregs-derived (Bonzheim et al. 2008; Roncador et al. 2005). We need more research about the relationship between TCL or FL and Tregs.

Some limitations should be noted in this meta-analysis. First, considering the limited number of studies on NKTCL (n = 3) and TCL (n = 2), the conclusion regarding these subtypes should be interpreted with caution. Second, after excluding two studies that largely attribute heterogeneity, there was still heterogeneity yet, which might be caused by laboratory testing methods and reagents. Third, the literature enrolled did not include unpublished studies. Although the publication bias was tested, the prognostic role of Tregs in lymphoma patients may be overestimated.

Conclusions

Lymphoma patients with a higher density of tumor-infiltrating Tregs have a better prognosis. This is helpful to develop individualized treatment strategies. It is meaningful to conduct further research about Tregs and its subtypes, which can help us find new immunotherapy targets.