Neutrophil Extracellular Traps Induced by IL8 Promote Diffuse Large B-cell Lymphoma Progression via the TLR9 Signaling

Diffuse large B-cell lymphoma (DLBCL) is the most common lymphoid neoplasm in adults, accounting for 30% of diagnosed non-Hodgkin lymphoma every year (1). Although the outcome has been greatly improved with current immunochemotherapy, more than 30% of patients still do not respond to these regimens or suffer from relapse (2, 3). Thus, the recognition of dysregulation processes that are critical to the survival of lymphoma cells is key to the development of effective therapies for DLBCL.

Neutrophils account for a significant portion of the inflammatory cells in the microenvironment of various malignancies (4). Although the normal function of neutrophils is to kill microorganisms, accumulating evidence has demonstrated the critical role of tumor-associated neutrophils (TANs) in tumor immune surveillance, metastasis, and proliferation (5–7). Moreover, Karin and colleagues have demonstrated that TANs can promote survival of DLBCL cells (8); and Bertrand and colleagues have reported that TANs correlate with poor prognosis in DLBCL (9). Therefore, TANs can be used as potential targets for designing new antilymphoma therapies. However, the ablation of neutrophils may lead to life-threatening immunosuppression, chemotherapy delay, dose reduction, or early termination of chemotherapy, which may increase the risk for recurrence and death from lymphoma (10). Thus, targeting this cell type may not be an ideal strategy for lymphoma treatment.

Apart from degranulation and phagocytosis, neutrophils also exert their functions via neutrophil extracellular traps (NETs) (11). NETosis describes the process by which neutrophils produce and release NETs (12). NETs are extracellular strands of decondensed DNA complexed with citrullinated histones (H3Cit) and neutrophil granule proteins such as myeloperoxidase (MPO) and neutrophil elastase (NE; ref. 13). These structures are usually generated in response to infectious or artificial stimuli, and have been initially described as a mechanism of antimicrobial defense (14). But recent evidence has linked NETs to cancers metastases and tumor-associated thrombosis (15, 16). Moreover, a previous study demonstrated that the interactions between NETs and CD5⁺ B cells favored the transition from autoimmunity to lymphoma (17). However, it remains unclear whether NETs play a role in DLBCL progression or can be potential therapeutic targets for this disease.

Thus, we sought to determine whether NETs play a role in the growth and dissemination of DLBCL cells and elucidate the underlying mechanism. Here, we provide evidence that the level of NETs was significantly higher in patients with advanced-stage DLBCL and had a poor prognosis. NETs, induced by lymphoma-derived IL8, could regulate important traits in DLBCL including growth and dissemination both in vitro and in vivo via the activation of TLR9 signaling and its downstream pathways. Importantly, disruption of NETs, blocking IL8–CXCR2 axis or inhibiting TLR9 could retard tumor progression in preclinical models. Thus, understanding this cross-talk between lymphoma cells and neutrophils has potential clinical implications in the treatment of DLBCL.

For details, see Supplementary Material and Methods.

Human DLBCL cell lines SU-DHL2, SU-DHL4, and SU-DHL6 and mouse DLBCL cell line A20 were purchased from ATCC. Human DLBCL cell line RC-K8 was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen. The cells were maintained in RPMI1640 (Gibco) supplemented with 10% heat-inactivated FBS. Human neutrophils were isolated from the peripheral blood of healthy donors using Ficoll density gradient centrifugation (18). Mouse bone marrow neutrophils were isolated from femurs and tibias, as described previously (19). The obtained cells were resuspended in DMEM supplemented with 10% FBS.

We collected paraffin-embedded tissue sections (training cohort: n = 123; validation cohort: n = 110) and plasma samples (n = 93) from patients with DLBCL diagnosed at the Sun Yat-sen University Cancer Center (Guangdong, P.R. China) between 2000 and 2015. The clinical study protocol was approved by the ethics committee of Sun Yat-sen University Cancer Center (Guangdong, P.R. China), and all procedures performed in this study were in accordance with the ethical standards of the institutional research committee and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all participants before recruitment.

Approximately 5 × 10⁵ neutrophils were seeded in a 24-well plate. After treatment, cells were fixed, permeabilized, and probed with antibodies against H3Cit (1:100, Abcam; ab5103) and MPO (1:50, R&D Systems; AF3667). Secondary antibodies with Alexa488 (1:200, Life Technologies) and Alexa555 (1:200, Life Technologies) were added and the slides were mounted in Fluoromount-G (Yeasen Biotechnology) with 1 μg/mL DAPI (Beyotime Biotechnology).

To figure out how SU-DHL2 cells induce NETs, we compared the cytokine profile of media from SU-DHL2, SU-DHL4, and SU-DHL6. The level of IL8 was further confirmed using ELISA. Both antibody arrays and ELISA were further described in Supplementary Methods and Materials.

Cell migration assays were performed as described previously (20). Briefly, 24-well chamber plates with 8-μm PET membranes (Costar) were used for migration assays. The bottom compartment contained 600 μL DMEM with 10% BSA. Tumor cells were added to the top compartment and allowed to migrate for 12 hours. The medium (500 μL) was collected from each bottom chamber and analyzed via flow cytometry for 1 minute with a constant flow rate. The migration of non-pretreated cells was set at 100%. The data represent the mean ± SEM of 3 independent experiments.

Tumor cells (5 × 10⁴/mL, 2 mL/well) were seeded in 12-well plates. After treatment, the cells were counted using a hemocytometer at the indicated timepoints (24, 48, 72, and 96 hours). The data represent the mean ± SEM of 3 independent experiments.

Six- to 8-week-old female BALB/c mice (Guangdong Medical Laboratory Animal Center) were housed in a specific pathogen-free mouse facility. A20 cells (2 × 10⁶/100 μL) were injected subcutaneously into the flank of each BALB/c mouse near the inguinal lymph node (21). All animal experiments were approved by the Institutional Animal Care and Use Committee of Sun Yat-Sen University. Mice were sacrificed when tumors reached 2,000 mm³.

Statistical analyses, as defined in each figure legend, were performed using SPSS for Windows version 20.0 (SPSS, Inc.) or GraphPad Prism 5.

To determine whether NETs can be detected in DLBCL, we performed immunofluorescence (IF) staining on a total of 233 patients with DLBCL, including 123 patients as the training cohort and 110 patients as the validation cohort. Different types of NETs including early stage of NETosis, the bona fide NETs, and extracellular fibrous NETs were observed in tumor tissues (Supplementary Fig. S1A). In both cohorts, patients with advanced-stage DLBCL presented with significantly higher levels of NETs than those with early-stage DLBCL [median numbers of NETs per field, training cohort: 3 (0–15) vs. 2 (0–6), P = 0.035; validation cohort: 3.5 (0–20) vs. 1 (0–7), P = 0.000; Fig. 1A–C]. The clinical characteristics of these 2 cohorts at initial diagnosis are presented in Supplementary Table S1.

Using X-tile, we determined 3 as the optimal cut-off value of NETs for assessing overall survival (OS; NETs ≤ 3/field, n = 79; NETs > 3/field, n = 44) in the training set, and we also used it as a uniform cut-off point for OS in the validation cohort. The clinicopathologic characteristics of high- and low-NET groups in both cohorts are listed in Supplementary Table S2.

In the training set, a higher level of NETs was significantly associated with more B symptoms (fever, sweat, and weight loss; P = 0.033) and elevated lactate dehydrogenase level (P = 0.003), and the formation of NETs showed no significant difference between germinal center B-cell like (GCB) and non-GCB patients (P = 0.663). In addition, univariate analysis revealed that a higher level of NETs was associated with remarkably worse progression-free survival [PFS; HR, 3.058; 95% confidence interval (CI): 1.450–6.449; P < 0.01; Fig. 1D] and OS (HR, 3.250; 95% CI, 1.451–7.278; P < 0.01; Fig. 1E). In multivariate analysis, a higher NETs level was found to be an independent predictor for both PFS (HR, 2.840; 95% CI, 1.329–6.068; P = 0.007; Supplementary Table S3) and OS (HR, 3.169; 95% CI, 1.396–7.194; P = 0.006; Supplementary Table S4).

In the validation group, consistent with the training set, no significant differences in the level of NETs were identified between patients with DLBCL classified according to GCB/non-GCB subtype (P = 0.499), and patients with higher NETs level presented with poorer PFS (HR, 4.438; 95% CI, 2.198–8.961; P = 0.000; Fig. 1F) and OS (HR, 6.061; 95% CI, 2.845–12.915; P = 0.000; Fig. 1G). In addition, Cox regression multivariate analysis showed that the level of NETs was an independent prognostic factor for patients PFS (HR, 3.731; 95% CI, 1.814–7.673; P = 0.000; Supplementary Table S3) and OS (HR, 4.745; 95% CI, 2.138–10.533; P = 0.000; Supplementary Table S4) in the validation cohort. And in GCB and non-GCB subgroup from both cohorts, we also observed that higher NETs group had poorer PFS and OS (Supplementary Fig. S1B–S1E).

Because the chromatin in NETs is most likely degraded to soluble nucleosomes leading to the release of fragments comprising DNA and granule proteins like MPO in the plasma (22), we used MPO-specific antibodies to capture circulating NETs (MPO–DNA complexes) and detected them with antibodies specific to DNA in patients' plasma (n = 93). The clinicopathologic characteristics of these patients are listed in Supplementary Table S5, and 60 plasma samples were from validation cohort. We found higher levels of MPO–DNA in patients with advanced-stage DLBCL than those with early-stage DLBCL (OD₄₀₅: 0.99 ± 0.04 vs. 0.65 ± 0.06; P < 0.001; Fig. 1H; Supplementary Table S6). Moreover, patients with DLBCL with higher level of MPO–DNA exhibited inferior PFS (HR, 3.614; 95% CI, 1.363–9.585; P < 0.01; Fig. 1I) and OS (HR, 3.366; 95% CI, 1.687–6.713; P < 0.001; Supplementary Fig. S1F). Further multivariate analysis revealed that the plasma level of MPO–DNA was an independent predictor for both PFS and OS (PFS, HR: 3.121, 95% CI: 1.060–9.187, P = 0.039; OS, HR: 2.609, 95% CI: 1.249–5.450, P = 0.011; Supplementary Tables S8 and S9). Because 60 patients in our study had available NETs staining and plasma MPO–DNA data, we investigated whether the level of NETs in tumor tissues was correlated with soluble NETs (MPO–DNA) in the plasma. Our data showed that the count of NETs was positively correlated with the plasma MPO–DNA level (r = 0.7754; P < 0.001; Fig. 1J).

Together, these data revealed that NETs were detected in clinical tissue samples and plasma of DLBCL, especially in advanced-stage disease. In addition, a high level of NETs was correlated with poor OS and PFS.

Having shown that a high level of NETs in clinical samples was correlated with poor survival, we sought to investigate whether NETs affected the proliferation and migration of DLBCL. We treated human DLBCL cell lines, SU-DHL2, SU-DHL4, and SU-DHL6 and mouse DLBCL cell line, A20 cells, with species-matched PMA-stimulated neutrophil media (PMA was used to stimulate NETs formation) for 96 hours and quantified the cell numbers each day using a hemocytometer, and we observed that this treatment increased tumor proliferation at 96 hours in all cell lines (Fig. 2A–D). Conversely, degradation of NETs using DNase I (1,000 U/mL) abrogated this effect; and NE inhibitor (GW311616A, NEi, 5 μmol/L), which can inhibit NETosis by prohibiting NE nuclear translocation and chromatin decondensation, also acquired similar results (Fig. 2A–D; ref. 23). As expected, treatment with NETs promoted migration of DLBCL cells in vitro, which was attenuated by the administration of DNase I (1,000 U/mL) or pretreatment of neutrophils with NEi (5 μmol/L; Fig. 2E–H). It was known that neutrophils can constitutively secret a proliferation inducing ligand (APRIL), which was further upregulated by stimuli like PMA and also can effectively induce tumor cell growth (24). This raises questions of the protumor ability of NETs in our in vitro model. Interestingly, we found that 6 hours treatment of 50 ng/mL PMA did induce APRIL secretion and could not be interrupted by DNase I or NEi (Supplementary Fig. S2A), but 4 hours treatment of 100 ng/mL PMA (our experiment condition) cannot effectively increase APRIL secretion in neutrophils (Supplementary Fig. S2B), which indicated the collected PMA-stimulated neutrophil media contained limited APRIL. Furthermore, APRIL had no effect on the formation of NETs (Supplementary Fig. S2C). These above data suggested that NETs can independently promote DLBCL proliferation and migration in vitro.

In vivo, we inoculated A20 cells into the inguinal lymph nodal region of BALB/c mice and treated the mice daily with DNase I (2.5 mg/kg i.p.) or NEi treatment (GW311616A, 2.2 mg/kg by gavage; refs. 21, 23). We aimed to observe heterotopic tumor growth and examine whether tumor cells could spread into the axillary lymph nodes that drained the inguinal region (21). Twenty-two days after inoculation, we observed that the DNase I- or NEi-treated mice had fewer NETs in tumor tissues and lower plasma MPO–DNA levels than the control group (Fig. 2I–K). The inhibition of NETosis or the clearance of NETs resulted in significant growth retardation of tumors compared with control group (n = 6 mice/group, 1-way ANOVA; P < 0.001; Fig. 2L; Supplementary Fig. S2D). In addition, both microCT imaging and gross morphologic analysis showed that DNase I and NEi treatment significantly inhibited axillary metastasis (1-way ANOVA, P < 0.001; Fig. 2M and N). Histologic examination and FACS study demonstrated that the enlarged axillary lymph nodes resulted from disseminated A20 cells (Supplementary Fig. S2E–S2G).

The above data indicated protumorigenic effects of NETs in DLBCL, and argued against a cell-specific phenomenon.

Next, we wanted to determine whether DLBCL could induce NETosis by transwell chamber assays and the mechanism underlying this process. We observed that only the neutrophils cocultured with SU-DHL2 cells generated extensive NETs (Fig. 3A and B). Extensive NETs were also produced when neutrophils isolated from murine bone marrows were cocultured with mouse A20 cells (Supplementary Fig. S3A). To figure out how SU-DHL2 induced NETosis, the cytokine profile of the media from SU-DHL2, SU-DHL4, and SU-DHL6 cells were analyzed using RayBio human cytokine antibody array C5. Five cytokines were significantly increased in the supernatant of SU-DHL2 cells (Fig. 3C), and we found only IL8 significantly induced NET formation to levels comparable with the coculture with SU-DHL2 cells (Fig. 3D and E), and blocking IL8 significantly reduced the ability of SU-DHL2 cells to induce NETosis (Fig. 3E; Supplementary Fig. S3B). ELISA further confirmed that the IL8 level in SU-DHL2 cells was highest among these 3 cell lines, which was approximately 10.63- and 8.57-fold higher than those in SU-DHL4 and SU-DHL6 cells, respectively (Supplementary Fig. S3C). To confirm that our results were not isolated to the SU-DHL2 cell line, we next tested in another high IL8-secreting cell line, RC-K8 (24), and primary human DLBCL cells. Consistent with SU-DHL2 result, neutrophils cocultured with RC-K8 or primary malignant cells which could highly secrete IL8 extensively generated NETs and this effect was reduced by blocking IL8 (Fig. 3F; Supplementary Fig. S3D).

Furthermore, we sought to investigate whether patients' plasma could stimulate NET formation (25). We selected 5 patients from early stage and advanced stage, respectively. The clinicopathologic characteristics of these patients are listed in Supplementary Table S10. We found plasma of advanced stage could significantly promote NETosis as compared with early stage (Student t test, P < 0.001; Fig. 3G and H). As we previously proved IL8 derived from lymphoma cells could promote NETosis, which led us to test the plasma IL8 levels in patients with DLBCL, and we did find higher levels of IL8 in patients with advanced-stage DLBCL (Fig. 3I; advanced stage vs. early stage, 946.00 ± 67.17 pg/mL vs. 485.60 ± 50.70 pg/mL; P < 0.001). Furthermore, NETosis induced by plasma could be weakened by the addition of IL8-blocking antibody (Fig. 3J; Supplementary Fig. S4A). Interestingly, we observed plasma IL8 was positively correlated with MPO–DNA level in patients with DLBCL (r = 0.570; P = 0.000; Supplementary Fig. S4B) and patients with high plasma IL8 had significantly shorter PFS and OS (Supplementary Fig. S4C and S4D). All above data indicated IL8 played an important role in NETosis in patients with DLBCL.

Mice lack a direct homolog of IL8, and murine C-X-C motif ligand 1 (CXCL1/KC), C-X-C motif ligand 2 (CXCL2/MIP-2), and LIX are regarded as functional homologs of IL8 (26). Thus, we tested the levels of these cytokines in A20 cells and found that A20 secreted high levels of CXCL1 (1308.00 ± 214.10 pg/mL) and CXCL2 (1154.00 ± 150.80 pg/mL), but not LIX (84.00 ± 19.29 pg/mL). The ability of A20 cells to induce NETosis was abolished by murine CXCL1-neutralizing antibody or/and murine CXCL2-neutralizing antibody (Fig. 3K; Supplementary Fig. S5A).

Therefore, both IL8 and its murine homologs secreted by lymphoma cells induce NETosis.

Previous studies have demonstrated that IL8 binds to the G-protein–coupled chemokine receptors CXCR1 and CXCR2, which led us to hypothesize that CXCR1 or CXCR2 might be involved in NETosis (27). In vitro, we used antibody against FcγRIIA as irrelevant antibody control (28), and found that the antibody specific to CXCR2 abolished IL8-induced NETosis (Fig. 4A; Supplementary Fig. S5B), but had limited effect on known CXCR1-mediated activation in neutrophils (Supplementary Fig. S5C; ref. 29), which suggested that CXCR2 might play a role in IL8-induced NETosis. Next, we investigated the downstream signaling pathways of CXCR2 that mediated NETosis. On the basis of known CXCR2 pathways in neutrophils, we tested the contribution of Src family kinases, MAPKs, and PI3K family kinases in NETosis (30). We found that IL8–CXCR2 axis depended on the Src and MAPK pathways rather than the PI3K pathway to induce NETosis (Fig. 4B). Similarly, the pharmacologic inhibition of Src, p38, and ERK but not of PI3K also led to the abrogation of CXCL1/CXCL2-stimulated NETosis in murine neutrophils (Supplementary Fig. S5D).

Consistent with in vitro results, systemic treatment with CXCR2-neutralizing antibodies led to a significant decrease in the MPO–DNA level in the plasma and NET formation in tumor tissues (n = 5 mice/group; Fig. 4C–E). Our above data showed that the plasma of patients with advanced-stage disease could induce NETosis. We therefore investigated whether the plasma from lymphoma-bearing BALB/c mice had the same effect and the role of CXCR2 in this process. Ex vivo study showed that the effect of plasma from lymphoma-bearing mice was also promotional, and could be significantly abrogated by the pretreatment of CXCR2 inhibitor (SB225002, 100 nmol/L), which indicated CXCR2 did mediate NETosis (Fig. 4F).

More importantly, in vivo, we observed that anti-CXCR2 therapy led to significant retardation of tumors growth and axillary lymph node dissemination (Fig. 4G–I; Supplementary Fig. S6A). Besides, in vitro, anti-CXCR2 treatment had no direct effect on the proliferation of lymphoma cells (Supplementary Fig. S6B and S6C). These data suggested that CXCR2 inhibition in vivo could reduce NETosis, which led to tumor growth retardation.

As shown in previous studies, the initiation of NETosis results in the release of damage-associated molecular patterns, which trigger an innate immune and inflammatory cascade (31). TLRs are important sensors of products of damaged or inflamed self-tissue, and mounting evidence has linked TLRs activation with the pathogenesis of cancer (32), but it is unknown whether NETs require TLRs to exert their functions in DLBCL. Thus, we screened TLRs as possible signaling receptors by culturing SU-DHL2 cells with NETs. Only the mRNA of TLR9, but not other TLRs was significantly increased (Fig. 5A). We therefore evaluated the role of TLR9 in NETs' protumorigenic effects. When we cultured SU-DHL2 cells in vitro with PMA-stimulated neutrophil media, NET-driven proliferation and migration were similar to the effect of a known TLR9 agonist (5 μg/mL ODN2006; InvivoGen) and were abolished by the addition of TLR9 antagonist (2.5 μmol/L ODN-TTAGGG; InvivoGen) to the conditional neutrophil media (Fig. 5B and C). Furthermore, both NETs and TLR9 agonist could only significantly drive the proliferation and migration of wild-type A20 cells and A20 cells transfected with shGFP in vitro, but not TLR9-knockdown A20 cells (Fig. 5D and E; Supplementary Fig. S7A and S7B). These results paralleled the response for the activation of TLR9-dependent pathways (33, 34). NETs significantly enhanced TLR9 expression together with an increase in the phosphorylation of NFκB, MAPK/p38, and STAT3 in SU-DHL2 cells (Fig. 5F). However, after adding DNase I or TLR9 antagonist to the culture system, the NFκB, MAPK/p38, and STAT3 pathways were significantly inhibited accompanied by a decrease in TLR9 expression (Fig. 5F).

To test the contribution of NFκB, MAPK/p38, and STAT3 in NETs-mediated effect, we treated SU-DHL2 and A20 cells with NFκB inhibitor (JSH23, 10 μmol/L), MAPK/p38 (SB202190, 5 μmol/L), and STAT3 (Stattic, 1 μmol/L), respectively. These inhibitors can effectively inhibit cell proliferation and migration in both cell lines (Fig. 5G and H). However, when culturing with PMA-stimulating neutrophil media, the inhibition effects of these 3 inhibitors were weakened, and only combined pharmacologic inhibition significantly decreased proliferation and migration (Fig. 5G and H). These data suggested that NFκB, MAPK/p38, and STAT3 pathway did contribute to the protumorigenic effect of NETs.

To investigate whether the knockdown of TLR9 could abolish NETs' protumorigenic effect in vivo, vector or TLR9-knockdown A20 cells were inoculated subcutaneously into BALB/C mice. There was no difference in the levels of plasma CXCL1 and CXCL2 among groups (Fig. 6A), and the levels of plasma MPO–DNA also showed no difference (Fig. 6B). However, mice inoculated with TLR9-knockdown A20 cells displayed a significant reduction in tumor growth (Fig. 6C and D) and metastasis to the axillary lymph nodes (Fig. 6E and F). IHC analysis of tumor sections revealed a weaker expression of Ki-67, together with p-STAT3, p-p65, and p-p38, in TLR9-knockdown A20 cell–derived tumors (Fig. 6G–K). Consistent with the IHC data, Western blot analysis showed that significant activation of the MAPK, NFκB, and STAT3 pathways were observed in wild-type A20 cell–derived tumor tissues, and this effect was abolished in the TLR9-knockdown groups (Fig. 6L).

Here, we found that higher level of NETs was correlated with dismal outcome in retrospective cohorts of DLBCL. We showed that lymphoma cells can induce NETosis, and the neutrophils' ability to eradicate pathogens through formation of NETs instead promotes tumor progression. Thus, our findings, together with the results of a previous study showing that NET-like structures in lymphoid tissues lead to the transformation from autoimmunity to lymphoma (17), suggest exciting possibilities for targeting NETs to prevent tumor progression.

In this study, increasing NET formation, as measured by IF and circulating MPO–DNA levels in clinical samples, was associated with poor survival in patients with DLBCL. This is the largest cohort and the first study to show evidence of NET formation and its clinical value in DLBCL. Our results are consistent with Zychilinsky who assessed tumor tissues from 8 patients with Ewing sarcoma. Two patients with NETs deposition in their tumors developed early recurrence (35), and the presence of NETs was also detected in tumor tissues and metastatic lymph nodes from 10 patients with colon cancer (36). However, these studies did not figure out the underlying mechanism of NETosis in tumor microenvironment. On the basis of the screening results of cytokine array, our study revealed that NET formation could be induced by lymphoma-derived IL8 by using mice model and patients' plasma samples. IL8 production by lymphoma cells has been reported and it is believed that IL8 promotes neutrophil infiltration in tumor (24), and we further proved that IL8 promoted NETosis via CXCR2 and its downstream pathways but not CXCR1. Although these 2 receptors have similar amino acid sequence within the transmembrane domains and connecting loops and share many common functions, they completely differ in their NH₂- and COOH- terminal regions, which may explain why they possess some distinct ligand-binding properties and signal differently (29, 37).

TLR9 was mainly expressed in immune cells including B cells, dendritic cells, and macrophages, and it acts as a bridge between innate and adaptive immunity in the antitumor responses in solid tumors (38). But in DLBCL, tumor cells can express TLR9 and directly response to the activation of TLR9 (32), and increasing evidence has shown that TLR9 expression on DLBCL exhibits high heterogeneity and the activation of TLR9 in normal or tumor B cells elicits the opposite biological effects (34). James and colleagues found that inhibiting TLR9 in normal B cells could increase accumulation of MYD88^L265P plasmablasts and may impair B-cell malignancies treatment (39). However, Staudt and colleagues identified a multiprotein supercomplex containing MYD88, TLR9, and BCR, which drives prosurvival NFκB signaling in DLBCL (40), and a TLR9 antagonist has shown promising results in the suppression of growth of B-cell lymphoma harboring MYD88^L265P mutation (41). In our study, we found that the stimulation of NETs could increase the expression of TLR9 in tumor cells. Unlike previous studies, in which NETs lead to metastases via sequestering circulating tumor cells or increasing the permeability of vascular structures (23, 42), we proved that the binding of TLR9 led to the activation of multiple pathways including NFκB, MAPK, and STAT3 pathway in tumor cells. The activation of NFκB can lead to myeloma cells survival and IL6 production (43). The MAPK pathway controls fundamental functions like cell proliferation and cell migration (44), and STAT3 activation, which is highly interconnected with NFκB, is associated with poor prognosis in R-CHOP–treated DLBCL (45). More importantly, the inhibition of TLR9 could effectively block these pathways and prevent lymphoma cells migrating to axillary lymph nodes, which provided a promising preclinical indicator for new therapy in DLBCL.

Because NETs consist of neutrophil-derived DNA studded with proteins (14), and TLR9 is an important DNA sensor (32), it makes sense that TLR9 can orchestrate tumor progression in response to NETs. But proteins like NE released during NETosis may also play a role in tumor progression. Previously, protumorigenic effect of NE has been extensively studied in some solid tumors albeit independent of NETs (46), and NE is required for DNA extrusion during NETosis, we further revealed that inhibiting NE not only blocked tumor cell–induced NET formation, but also retarded tumor growth and metastases. But given the protease's role in eliminating pathogens, targeting NE may have adverse consequences on a person's ability to fight infections (47). Because a previous study has shown that the disruption of NETs has little consequence on the degranulation and phagocytosis of neutrophils (48), it may be relatively safe to develop approaches that directly disrupt NETs. Treatment with DNase I was reported to reduce lung metastases in breast cancer (15), and DNase I has been approved for the treatment of systemic autoimmune diseases such as cystic fibrosis without significant adverse events (49). Our results of DNase I serve as a proof of principle that NETs are targets to reduce DLBCL progression. Another strategy may be to prevent NETs from forming altogether by blocking IL8–CXCR2 axis. A CXCR2 inhibitor (AZD5069) has been tested in phase Ib/II trials for advanced solid malignancies, and preliminary data showed good safety and tolerability (50).

Although DLBCL has been regarded as a potentially curable disease, patients are still threatened by recurrence. We aim to provide a better understanding of the dysregulation mechanism, which is vital to the survival of lymphoma cells, and ultimately improve long-term outcomes. In this study, we firstly demonstrated the value of NETs to predict DLBCL inferior survival and to promote DLBCL progression, and provided clear evidence that the TLR9 pathway was critical to tumor progression. We also proved tumor-derived IL8 induced NETs formation via its CXCR2 receptor. Most importantly, our findings provide preliminary evidence that the use of DNase I, NEi, or blocking CXCR2 or TLR9 can inhibit DLBCL progression and will hopefully pave the way for future clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: M. Nie, L. Yang, Y. Xia, W. Jiang

Development of methodology: M. Nie, L. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Nie, L. Yang, X. Bi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Nie, L. Yang, X. Bi, Y. Xia, W. Jiang

Writing, review, and/or revision of the manuscript: M. Nie

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Jiang

Study supervision: Y. Wang, P. Sun, H. Yang, P. Liu, Z. Li, Y. Xia, W. Jiang

This work is supported by the Fundamental Research Funds for the Central Universities (No. 17ykpy77) and National Natural Science Foundation of China (No. 81700196).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.