Tertiary Lymphoid Structures, Immune Response, and Prognostic Relevance in Non-Small Cell Lung Cancer

Abstract We assessed the presence of ‘tertiary lymphoid structures’ (TLS) in a series of surgically treated non-small cell lung carcinomas (NSCLC). The TLS-density in the tumor periphery (pTLS) ranged from 0 to 1.8 (median 0.45), while in inner tumor areas (iTLS) ranged from 0 to 1.0 (median 0); (p < 0.0001). High pTLS-density was linked with early stage of the disease. Glycolysis-related enzyme expression (MCT1, Hexokinase 2) was linked with high pTLS-density (p < 0.05). High pTLS and iTLS densities were linked with better postoperative prognosis (p = 0.02 and p = 0.01, respectively). Assessment of TLS is a useful prognostic marker in NSCLC.


Background/introduction
The importance of the immune system in controlling cancer development and growth has been documented at the beginning of the 20th century when researchers found that experimental animals can reject tumor transplants (1). Researchers speculated on the existence of a 'natural means' that could destroy the implanted tumors. As the anatomical structure and the histological and cell components of the immune system were gradually revealed, it became clear that this was the most complex organ in the body. The immune system spreads throughout the human body and is the basis and the driver of outstanding healing power.
The modern era of immunotherapy documented the anti-tumor efficacy of the immune system. By understanding how to correct immune pathologies immunotherapy policies may help to the eradication of previously incurable cancers, like advanced non-small cell lung cancer (NSCLC) (2). The development of immune checkpoint inhibitors targeting the PD-1/PD-L1 pathway revolutionized the treatment of patients with metastatic NSCLC so that today 17 and 47% of patients are treated with immunotherapy in first and second line setting, respectively (3). Intense clinical investigation on novel immunotherapy agents targeting pathways beyond PD-1/ PD-L1, like the CTLA-4/CD80, the LAG3 and TIGIT immune inhibitory receptors, the indoleamine-2,3-deoxygenase and others, are expected to further change the landscape of therapy for NSCLC (4).
The immune system is composed of many different tissues. The primary (or central) lymphoid organs, destined to generate lymphocytes from immature progenitor cells, are the thymus that is involved in T-cell lymphopoiesis, and the bone marrow that hosts both B-cell and T-cell precursors. Spleen and lymph nodes are secondary (peripheral) lymphoid organs, where B and T-cells are activated after being exposed to antigens, giving birth to mature immune cells that enter the bloodstream to reach the target tissues. During pathological processes, e.g., chronic inflammation or autoimmune conditions, the 'tertiary lymphoid structures' (TLS) appear. These are abnormal 'lymph node-like' structures that emerge within the affected organs/tissues and are characterized by active germinal centers surrounded by follicular dendritic cells (5). B-cell follicles, T-cell zones are often evident within TLS. Specialized vessels known as high endothelial venules (HEVs) are also present, but TLS do not have afferent lymphatic vessels (6). Within them, both B and T-cell responses can be developed, just like in lymph nodes.
Such TLS are often identified in the periphery or within the body of tumors (7). Several studies support the idea that TLS represent a tumorrelated immune response, with anti-tumor or immuno-regulatory activities, as their abundance has been linked with favorable or ominous prognosis (7). The mechanisms for the organization and maturation of TLS in inflammatory conditions or cancer are under investigation (reviewed in 8,9). Cytokine release, like TNFa, IL-7 and IL-22, are shown essential for this process. The activation of fibroblasts that form a cytokine-and adhesion molecule-reach microenvironment is also important in the formation of TLS. Antigen presentation by dendritic cells seems also to have a key role. Studying the TLS formation and composition in tumors, along with the tumor infiltrating lymphocyte (TIL) density and relevant subpopulations may be important as a prognostic tool after surgery or radical radiotherapy. This may also provide a tool for the characterization of the individual tumor and host immunological status and, eventually, an algorithm to individualize the immunotherapeutic interventions.
In the current study, we assess the presence of TLS in a series of non-small cell lung carcinomas, and we investigate their association with other immunological parameters and prognosis. As anaerobic glucose metabolism and lactate are involved in fibroblast activation (10), a phenomenon that largely characterizes tumors (11), we further examined the hypothesis that tumor glycolytic phenotypes are linked to TLS formation in NSCLC.

Methods
One hundred and three (103) surgically removed non-small cell lung carcinomas (retrieved from Department of Pathology, Democritus University of Thrace) were analyzed. Patients had been treated with lobectomy or pneumonectomy between 2005 and 2010, and none of these patients received adjuvant radiotherapy or chemotherapy. These patients were consecutive according to the date of receipt of the surgical material. Patient and disease characteristics are shown in Table 1. Within this series of patients, 4 had limited metastatic disease and were excluded from survival analysis. Stage 4 cases were treated with further chemotherapy. Three out of 99 patients were lost to follow-up immediately after surgery and were excluded from the survival analysis. Thus 96 patients were included for survival analysis. The study has been approved by the local Ethics and Scientific Committees (ES1/23-01-2019). One representative tissue section containing both tumor and adjacent normal lung was selected from each case, following inspection of hematoxylin-eosin sections. Sixty-one of them were squamous cell carcinomas, 23 adenocarcinomas, and 19 undifferentiated large cell carcinomas. Forty-five were stage 1, twenty-three stage 2, thirty-one stage 3, and four stage 4 carcinomas. The Union for International Cancer Control (UICC) system was used for postoperative staging (https://www. jtcvs.org/article/S0022-5223(17)32136-0/fulltext). Thirteen patients were females and 90 males. The median age of patients was 67 years (range 28-81). The median follow-up of patients was

Assessment of TLSs
The density of TLS was assessed in the tumor periphery and the inner tumor area, in hematoxylin-eosin tissue sections, in all available optical fields. As 'tumor periphery' was defined the area comprised by the first Â200 optical field containing tumor invading front and normal lung adjacent to this front. TLSs were identified in the stroma of the invading tumor and in the normal lung infiltrated by the tumor. Inner tumor areas refer to all available Â200 optical fields of the tumoral tissue section, immediately below the optical filed that defined the tumor periphery. TLSs were identified in the stroma areas. Necrotic areas were excluded. All tertiary lymphoid structures with or without identifiable germinal centers were recorded. The whole tissue section was analyzed at Â200 magnification power. The number of TLS in normal lung across the tumor periphery was recorded and divided by the number of optical fields to provide the peritumoral 'pTLS-score'. A similar procedure was applied within the tumor body, where after counting the TLS number in all Â200 optical fields, we produced the inner tumor area 'iTLSscore', by dividing the total TLS number with the number of optical fields.

Immuno-phenotyping of the TLS
We performed immunohistochemical analysis to examine the specific lymphocyte subtypes residing in the germinal centers and outer areas of TLS. This aimed to the detection of CD4 cells (helper T-cells that regulate innate immune response and are also the main source of regulatory T-cells), CD25/interleukin-2 receptor-a chain and FOXP3/Forkhead-box-P3 (principal markers of regulatory CD4þ T-cells), CD8 (the main source of cytotoxic T-cells), CD20 (B-cells), PD-L1 cells (lymphocytes or monocytes with regulatory activity), CD56 (natural killer NK-cells), CD68 (pan-macrophage marker), PD-1 cells (characterizing mainly cytotoxic lymphocytes vulnerable to PD-L1 expressing lymphocytes, monocytes, and cancer cells), CD47 (dendritic and other immune cells) and SIRPa (macrophage marker involved in the CD48/SIRPa pathway of blockage of cancer cell phagocytosis). Details on the antibodies and immunohistochemical methodology have been reported in a previous study (12,13), and are briefly reported in Supplemental  Table S1.

Other immunohistochemical markers
In the current study, TLS analysis was the original task. Data related to glycolysis (glucose transporter 1 and 2, hexokinase II, monocarboxylate transporter 1 and 2), hypoxia and acidity (hypoxia-inducible factor HIF1a, lactate dehydrogenase LDH5, carbonic anhydrase CA9), and immunological markers (Tumour infiltrating lymphocyte density TILs, PD1þ TIL-density, PDL1þ TIL-density, PD-L1 and CD47 expression by cancer cells), were available from previously published studies on the same series of patients. Methodology and scoring of the above markers can be found in the relevant studies (13)(14)(15).

Statistical analysis
Statistical analysis was performed with the GraphPad Prism 7.0 package and the SPSS (v14.0, SPSS Inc.) program. Comparison between categorical variables or grouped continuous variables we performed with the Fisher's exact t-test and the unpaired two-tailed t-test, respectively. Linear regression analysis tested correlations between continuous variables. The Kaplan-Meier method was used to assess the impact of TLS and other variables on patients' overall survival, and a Cox proportional hazard model was used for multivariate analysis. A p value of <0.05 was used for significance.

Immunohistochemical profile of TLS
The distribution of lymphocytic subtypes in the TLS was immunohistochemically examined ( Figure 1). The density of TLS was assessed in the tumor periphery in all available optical fields (median 10) and the tumor stroma, in hematoxylin-eosin tissue sections in all available optical fields (median 14). T-cells expressing CD8 were exclusively detected in the outer area of TLS, while CD4þ and CD25þ and FOXP3þ cells were identified both in the germinal centers and in the outer TLS area. CD20þ B-cells were noted both in the outer areas and in the germinal centers. CD56þ NK-cells were seldom identifiable, while CD68þ macrophages resided in the TLS areas outside the germinal centers. Lymphocytes expressing PD-L1 and PD-1 were noted in the germinal centers. CD47 was expressed by a cell population in the outer TLS area and not in the germinal centers, while SIRPa was sporadically recorded in the outer TLS area.

TLS-density
The number of optical fields examined ranged from 3 to 19 (median 10) and 4 to 27 (median 14) in the normal lung proximal to the tumor invading front and inner tumor areas, respectively. The number of TLS ranged from 0 to 19 (median 4) and 0 to 18 (median 0) in the invading front and inner tumor areas, respectively. By dividing the number of TLS by the number of optical fields, we calculated the TLS-density. The pTLS-density ranged from 0 to 1.8 (median 0.45), while the iTLS ranged from 0 to 1.0 (median 0). The difference was statistically significant (p < 0.0001); Figure 2(a). Lack of pTLS was recorded in 23/103 (22.3%), and of iTLS in 55/ 103 (53.4%) cases A significant direct association between pTLS and iTLS densities was noted (p ¼ 0.006, r ¼ 0.28). Typical images from hematoxylin-eosin tissue sections highlighting TLS in the peripheral (proximal to invading tumor front) normal lung area and in the tumor stroma are shown in Figure 2(b,c).

TLS-density and histopathological variables
Tumors with stage 1 disease had a significantly higher pTLS-density compared to stage 2 and 3 tumors (median value 0.67 vs. 0.40 vs. 0.23; p < 0.02); Figure 3(a). There was no significant difference for iTLS-density (media value 0.35 vs. 0.15 vs. 0.18; p > 0.11). There was no association between histology types or MIB1 proliferation index and TLS-densities. Table 2 reports the association between pTLS/iTLS density and patient/ histopathological variables.

TLS-density and immune-related parameters
The TLS-densities were not related to the TILdensity, PD1þ TIL-density, or PDL1þ TIL-density, assessed in the invading tumor front or inner tumor areas. A significant inverse association was found between iTLS and FOXP3þ TIL-density (p ¼ 0.05, r ¼ 0. 19). This weak relation was not confirmed when stratifying for stage (lower number of cases). There was no association with CD68þ tumor-associated macrophage TAMdensity. There was no association between TLSdensities and the expression of PD-L1, or CD47 expression by cancer cells.

TLS-density and survival
Survival analysis comprised 96 evaluable cases (56 squamous cell carcinomas, 22 adenocarcinomas and 18 undifferentiated carcinomas). Kaplan-Meier overall (disease-specific) survival analysis showed a significant association of pTLS and iTLS densities with better postoperative prognosis (p ¼ 0.02 and p ¼ 0.01, respectively); Figure 4. Stratifying for histology subtypes, high pTLS was linked with better prognosis in squamous cell and adenocarcinoma subgroups (p ¼ 0.04 and 0.04, respectively); Supplemental Figure S1. High iTLS was linked with significantly better survival in the adenocarcinoma subgroup (p ¼ 0.01). In a multivariate analysis model including stage and TLS-densities, only the stage of disease was an independent prognostic variable (p ¼ 0.0001, HR 1,8), while iTLS approached significance (p ¼ 0.13, HR ¼ 0.3).

Discussion
Lymphocytes are generated by precursor cells residing in the primary lymphoid organs, thus bone marrow and thymus. These migrate to the secondary lymphoid organs, lymph nodes, and spleen, where B-cell and T-cell are activated by antigens and antigen presentation by dendritic cells. In chronically inflamed tissues, ectopic tertiary lymphatic structures(TLS) appear. The genesis of TLS occurs following a partially understood morphogenetic process that involves interactions between mesenchymal precursors (the lymphoid tissue stromal organizer, LTo) and hemopoietic precursors (the lymphoid tissue inducer, LTi) (16). Secretion of inflammatory cytokines like IL-17, IL-22, and IL-7, as well as chemokines, are important in the development of TLS (17,18). TLS often appear in human carcinomas and are more frequently identified in the peritumoral normal tissue, while intratumoral TLS may also exist within the tumor body. Indeed, in our study, one-fourth of tumors did not have TLS in the peritumoral area, while half of the cases did not have TLS in the inner tumor areas. Several studies have focused on the prognostic relevance of TLS in human tumors. In breast cancer, the reports are conflicting. In her-2 positive breast cancer patients, Lee et al. (19) did not find any prognostic relevance, while Sofopoulos et al. (20) found that a high number of TLS in the peritumoral area was linked with poorer disease free and overall survival. However, other studies in breast cancer reported a significant association of TLS with better outcomes (21,22). In colorectal cancer, TLS abundance seems to be linked with a good prognosis (23). TLS assessed in metastatic liver disease predict for better progression-free survival (24).
In the current study, we examined the prognostic role of TLS-density in the peritumoral lung tissue (pTLS) and the tumor body (iTLS) in a series of NSCLCs treated with surgery alone. The TLS-densities had a similar range in all histology subtypes. Low pTLS-density was linked with advanced stage of disease. A study by Rakaee et al. (25) showed a similar lack of association with histology type but lower TLS density in advanced stages of the disease.
Both pTLS-and iTLS-density were directly linked with a better disease-specific postoperative survival. This is in accordance with a study reported by Silina et al. (26), where high TLSdensity, assessed with a mixed peritumoral and inner tumor area scoring system, was linked with better prognosis in a series of squamous cell lung carcinomas. Dieu-Nosjean et al. (27), using as a marker the mature dendritic cells homing the lymphoid tissue, also found a direct association of TLS with a favorable prognosis. Germain et al. (28) proposed that the composition of TLS, and more specifically the presence of follicular B-cells, define protective immunity and favorable prognosis. The limited experience published suggests that TLS have an active role in the anti-tumor immune response in NSCLC. Whether TLS may also define the efficacy of immunotherapy with immune checkpoint inhibitors in lung cancer is unknown. An interesting study, however, supports this hypothesis (29).
Activation of cancer associated fibroblasts is a hallmark of cancer, and metabolic interaction between fibroblasts and cancer cells in the context of direct and inverse Warburg effect have been postulated (30,31). Lactate release during the process of anaerobic glycolytic pathways active in the tumor microenvironment, directly activate fibroblasts (32). As activation of fibroblasts is involved in the formation of TLS, by enriching the tumor microenvironemtn with cytokines and adhesion molecules (8,9,33), it can be postulated that TLS presence in cancer is directly related to tumor metabolism. We, therefore, examined eventual associations between TLSdensity and the metabolic profile of the tumors. An interesting observation was the direct association of TLS-density in the peritumoral area with glucose transporters and the expression of enzymes involved in glycolysis and monocarboxylate trafficking. It is unknown whether glycolysis products extruded out of cancer cells may support the generation of TLS. Examining the finding from another point of view, IL-17 and IL-22 cytokines that are essential for the genesis of TLS, strongly stimulate aerobic glycolysis in cancer cells (30,31). Whether such cytokines provide a link between tumor metabolism and TLS formation demands further investigation.
With regard to other immunological parameters, TLS-density was not related to the density of free tumor-infiltrating lymphocytes (TILs). Although no additional data exist in the literature for NSCLC, TLS-and TIL-density show a direct association in triple negative and her2 ¼ positive breast cancer (19,22). Regarding the association between TIL-density and TLS-density in NSCLC, there is no clear conclusion. Our study showed no correlation between TLS and TIL-densities in the invading tumor front or inner areas. It seems that although TLS may be the source of lymphocytes directed to the tumor, TIL accumulation also occurs despite the presence of TLS. Reports in various cancer types show an association between high TLS density and specific TIL-subpopulations, like B-cells, or between specific TLS composition, like high levels of dendritic cells, and high T-cell presence among TILs (34,35).
In our study, low inner tumor TLS-density was linked to high FOXP3þ regulatory cell density, supporting the suggestion that lack of TLS favors the establishment of immunosuppressive pathways. We found, however, no association between TLS-density and a major immunosppressive pathway driven by PD-L1 expression by cancer cells. In a recent study in solid tumors, TLS-density was directly related to response to anti-PD-L1 immunotherapy independently of the presence of PD-L1 expression (36). Nevertheless, an inversely directed hypothesis seems equally sound; a metabolically immunosuppressive tumor microenvironment favors regression of TLS. This latter is also supported by the marginal association between high LDH5 expression and low iTLSdensity. As acidity suppresses cytotoxic cell proliferation and promotes regulatory cell accumulation (32), it is assumed that it is also involved in TLS disorganization. Lactic acid also suppresses cytokine secretion (33), which may switch off pathways essential for TLS survival in the tumor stroma. Innate lymphoid cells (ILCs) of normal mucosa and epithelia seem to be involved in the formation of TLS (37), and their plasticity regarding their pro-or anti-tumor function is well known (38). A subsequent step to further investigate the role of TLS in the prognosis of lung cancer and other tumors is the identification of specific ILC-subpopulations in TLS that shape the TLS/tumor interactions. The current study brings forward the hypothesis that tumor metabolism can be involved in the TLS shaping toward anti-or pro-tumor activity, which encourages further studies in this direction.

Conclusions
The current study provides evidence that TLS formation is an important component of the anti-tumor immune response in NSCLC, independent of TIL accumulation. High TLS-density, whether in the tumor body or the peritumoral area, is associated with low tumor stage and better postoperative prognosis in NSCLC. Impaired formation or regression of TLS in inner tumor areas seems to associate with regulatory immune cell accumulation. The liaison between TLS and tumor metabolism documented in the current study demands further investigation.

Author contributions
AG: Conception, design, immunohistochemistry analysis, writing of the paper; PC: Data collection, immunohistochemistry, writing of the paper; CAC: Data analysis, interpretation of data, writing of the paper; MIK: Conception, design, data analysis, writing of the paper.

Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.