Blood and lymphatic vessels contribute to the impact of the immune microenvironment on clinical outcome in non-small-cell lung cancer†

Abstract

OBJECTIVES

Lymphangiogenesis plays a critical role in the immune response, tumour progression and therapy effectiveness. The aim of this study was to determine whether the interplay between the lymphatic and the blood microvasculature, tumour-infiltrating lymphocytes and the programmed death 1 (PD-1)/programmed death ligand 1 (PD-L1) immune checkpoint constitutes an immune microenvironment affecting the clinical outcome of patients with non-small-cell lung cancer.

METHODS

Samples from 50 squamous cell carcinomas and 42 adenocarcinomas were subjected to immunofluorescence to detect blood and lymphatic vessels. CD3^pos, CD8^pos and PD-1^pos tumour-infiltrating lymphocytes and tumour PD-L1 expression were assessed by immunohistochemical analysis.

RESULTS

Quantification of vascular structures documented a peak of lymphatics at the invasive margin together with a decreasing gradient of blood and lymphatic vessels from the peritumour area throughout the neoplastic core. Nodal involvement and pathological stage were strongly associated with vascularization, and an increased density of vessels was detected in samples with a higher incidence of tumour-infiltrating lymphocytes and a lower expression of PD-L1. Patients with a high PD-L1 to PD-1 ratio and vascular rarefaction had a gain of 10 months in overall survival compared to those with a low ratio and prominent vascularity.

CONCLUSIONS

Microvessels are an essential component of the cancer immune microenvironment. The clinical impact of the PD-1/PD-L1-based immune contexture may be implemented by the assessment of microvascular density to potentially identify patients with non-small-cell lung cancer who could benefit from immunotherapy and antiangiogenic treatment.

INTRODUCTION

The role of lymphangiogenic structures in the composition of the microenvironment of a tumour is dual and extremely complex. On one hand, they promote tumour progression by actively contributing to metastatic dissemination [1]; on the other hand, the antitumour immune response can be initiated only with preserved vascular function to control immune cell trafficking, adhesion and homing [2]. Blood and lymphatic vessels determine a fine and continuous interaction between tumour cells and tumour-infiltrating lymphocytes (TILs) so that a delicate but dynamic balance is established, by which an anti- or pro-tumourigenic contexture is defined in the process of ‘immunoediting’ [3], thereby dictating disease outcome [4, 5].

Metastasis formation implies an angiogenic switch mediated by the levels of vascular endothelial growth factor and the development of new vascular structures. Tumour-associated vessels are usually characterized by aberrant morphological structure that hinder their immune functions [6] by limiting extravasation and the antitumour action of cytotoxic T-lymphocytes, ultimately shaping an immunosuppressive contexture [7]. Angiogenesis and immunosuppression are hallmarks of cancer [8], exploited by cancer cells for dissemination, tumour escape and lymphocyte anergy [9].

Although angiogenesis promotes tumour progression, vascular endothelial growth factor and its receptor inhibitors have shown limited efficacy in patients with metastatic disease in terms of overall survival (OS) and durable response [10, 11]. This finding may be due to several mechanisms, including compensatory pro-angiogenic signalling, vessel co-option, vasculogenic mimicry and endothelial cell precursors or cancer stem cell-mediated vasculogenesis [12–14].

Given the mutual influence of TILs, vessels and tumour cells in shaping the immune contexture [15] and considering the current success of immunotherapy, the association of antiangiogenic drugs with immune checkpoint inhibitors [16] represents an ongoing attractive strategy. Interestingly, the combination of antiangiogenic factors with programmed death-1 (PD-1)/programmed death ligand 1 (PD-L1) checkpoint inhibitors displays synergistic activity. Suitable doses of antiangiogenic drugs may lead to vessel normalization, thereby improving immune cell trafficking and engraftment and increasing the concentration of immune checkpoint inhibitors at tissue level and activity to counteract the immunosuppressive state [17, 18].

The actual mechanism of the interaction between immune checkpoint inhibitors and antiangiogenic drugs remains to be defined, although these preliminary observations support studies designed to improve our understanding of the role of lymphangiogenic structures within the cancer immune contexture.

The aim of the present study was to determine whether lymphangiogenic structures actively participate in the tissue composition of the non-small-cell lung cancer (NSCLC) microenvironment, by assessing their incidence and distribution in relation to TILs as well as the expression of the PD-1/PD-L1 immune checkpoint. Importantly, the potential impact of lymphatic vessels on the clinical outcome of patients with NSCLC was investigated. Our results suggest that a more comprehensive analysis of the immune contexture may provide prognostic and predictive tools for the response to immunotherapy and its combination with antiangiogenic treatments.

MATERIALS AND METHODS

Patient population

The study was performed on 92 consecutive patients with NSCLC [50 squamous cell carcinoma (SCC), 42 adenocarcinoma (ADC), Stage I/II 72%], who were undergoing lung resections at the Unit of Thoracic Surgery, University Hospital of Parma from 2011 to 2015. This research was approved by the institutional review board for human studies (ethical committee) of the University Hospital of Parma, in accord with principles listed in the Helsinki declaration. Patients were enrolled after they gave informed consent to the use of biological samples for research purposes. The patient population comprised 68 men and 24 women, aged 45–84 years. Smoking status, histological status and pathological staging of the disease are listed in Table 1. Patients lacking a complete clinical history and/or inadequate tissue samples for histochemical studies were excluded from the statistical analysis.

Immunohistochemical analysis

Blood and lymphatic vessels

Five-micrometre-thick sections were cut from formalin-fixed, paraffin-embedded NSCLC samples and processed for immunohistochemical analysis. To document the presence of lymphatic and blood vessels, sections were exposed to double immunofluorescence using, respectively, anti-D2–40 and anti-von Willebrand Factor (vWF) antibodies. Sections were then incubated with fluorescein isothiocyanate- and tetramethylrhodamine-conjugated specific secondary antibodies. Nuclei were counterstained with 4′,6-diamidine-2′-phenylindole and coverslips were mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA). The antibodies used for immunohistochemical analysis are listed in Table 2.

The density and distribution of the vascular structures were computed by sampling a tissue area ranging from a minimum of 9.39 mm² to a maximum of 47.10 mm², using a fluorescence microscope (Olympus BX60) at ×200 final magnification. Microvascular density was evaluated in the spared lung (SL), the invasive margin (IM) and the tumour core (T) according to the illustration shown in Fig. 1A.

Tumour-infiltrating lymphocytes

TILs were characterized by the immunohistochemical detection of CD3, CD8 and PD-1. Sections were incubated with primary anti-CD3, anti-CD8 or anti PD-1 antibodies. The reaction was revealed by immunoperoxidase with 3,3′-diaminobenzidine staining; nuclei were counterstained by haematoxylin. The antibodies used for characterization of TILs are listed in Table 2.

The density (n/mm²) of the TILs was obtained by computing a tissue area ranging from a minimum of 2.97 mm² to a maximum of 18.77 mm² in SCCs and from a minimum of 2.51 mm² to a maximum of 17.17 mm² in ADCs. The incidence of TILs was evaluated from their proximal (cell–cell contact with cancer cells or localized within 20 µm linear distance from cancer cells) or distal (>20 µm from cancer cells) location.

Programmed death-1/programmed death ligand-1 immune checkpoint

The PD-1/PD-L1 immune checkpoint was investigated by the immunohistochemical estimation of PD-L1 expression in neoplastic cells and the number and localization of PD-1^pos TILs.

PD-L1 expression was assessed after a comparative evaluation between immunoperoxidase and immunofluorescence on control tissues (human placenta) and on serial sections from the same cases using different anti-PD-L1 antibodies. Following stringent criteria of specificity and reproducibility, the anti PD-L1 antibody clone 28–8 was selected for further analysis. PD-L1 fluorescent images were captured with a digital camera (QICAM) connected to a fluorescent microscope (Olympus BX60) at a final magnification of ×200. All images were acquired with precalibrated gain and exposure time. A specific fluorescence was carried out by merging the emission signals from different excitation lengths on the same microscopic field. Neoplastic and stromal expressions of PD-L1 were ascertained by the simultaneous detection of the immunofluorescence of cytokeratin. Double immunofluorescence was also performed to concomitantly evaluate the expression of PD-1^pos TILs and PD-L1. The list of antibodies used for this analysis is provided in Table 2.

On each sample, a tissue surface from a minimum of 1.23 mm² to a maximum of 46.95 mm² was analysed, involving the counting of a minimum of 28 000 to a maximum of 600 000 tumour cells. The fractional area occupied by the fluorescent signal and the intensity of that signal, expressed as integrated optical density per unit area, were evaluated using software for image analysis (Image Pro Plus 4.0). PD-L1 levels were expressed as integrated optical density in each case, and cut-off values were established for statistical purposes.

In addition, to provide a rough estimation of the activation of the immune checkpoint, the number per square millimetre of both neoplastic cells labelled by PD-L1 and PD-1^pos TILs was morphometrically determined. Subsequently, the numerical ratio of PD-L1^pos tumour cells to PD-1^pos lymphocytes was computed (PD-L1 to PD-1 ratio). The magnitude of sampling of each investigated phenotype is reported in Supplementary Material, Table S1.

Statistical methods

Statistical analyses of the data were performed using SigmaPlot Software (Systat Software, San Jose, CA, USA) and SPSS Software (Released 2016. SPSS Statistics for Windows, Version 24.0; IBM Corp., Armonk, NY, USA). Median and quartiles were used for the analysis and graphical illustrations. Given the non-normal distribution of the data (Kolmogorov–Smirnov test), the Kruskall–Wallis test was used to compare the different groups. If significant differences were detected, pairwise multiple comparisons were then performed with Dunn’s post hoc test. The association between 2 variables was measured by the Pearson correlation coefficient due to the normality of the residuals, despite the non-normality of the variables. For repeated measures, the Friedman test followed by Dunn’s post hoc test was performed. A multivariate Cox regression was performed to define potential independent factors influencing the OS. The log-rank test was used to analyse the diversity in survival between the different groups. OS was considered as the time from the date of lung resection surgery to the date of death. Patients without an event were censored at the last time point at which they were known to be alive. Statistical significance was set at P < 0.050.

RESULTS

Vascular structures

Vascular structures were assessed by the morphometric measurements of their density and distribution in the different portions of the NSCLC samples: the SL, the IM and the T (Fig. 1A). Lymphatic vessels were identified by the expression of podoplanin (D2–40), whereas blood vessels were detected through labelling of the vWF. Vascular profiles expressing both D2–40 and vWF were also observed (Fig. 1B).

The quantitative measurement documented an increase in both lymphatic and blood vessels in ADC compared to SCC samples, whereas in both NSCLC histotypes a similar gradient from the distal area (SL) to the intratumour site (T) was observed. Compared to the SL, the IM areas showed a higher density of lymphatics and a lower incidence of blood vessels (P < 0.001 for ADC); however, a significant decreasing gradient from IM throughout the T in D2–40^pos lymphatics and vWF^pos blood vessels was documented in both ADC and SCC samples (P < 0.010). Thus, the density of the vessels in the T were greatly reduced compared to those in the distal unaffected lung (SL); specifically, a significant decrease was documented in vWF^pos vessels in both ADC (P < 0.001) and SCC (P = 0.002). D2–40 and vWF double-positive structures showed a trend similar to that of the lymphatics, with a significant rise from the SL to the IM areas (P = 0.008) (Fig. 1C).

Vascular structures and tumour, node and metastasis staging

Lymphangiogenesis and blood supply are essential for tumour growth and development. Thus, we determined whether vascular structures had an impact on NSCLC staging.

At variance with the T factor (tumour extension) that was not associated with changes in lymphatic and blood vessels, the N factor (nodal involvement) correlated with a lymphangiogenic response both in spared and neoplastic portions of the lung. In particular, when compared to N0 and N1 cases, NSCLC samples from N2 patients showed a significant 2-fold increase (P = 0.050) in lymphatics located at the IM and a 3.1-fold increase (P = 0.021) in the blood vessels detected in the SL parenchyma. Importantly, a statistically significant direct correlation was documented between N status and vWF^pos blood vessels located in the SL (r = 0.5, P = 0.007, Pearson correlation). All these parameters were emphasized when the pathological stage was considered, documenting in patients with Stage III disease a significant difference in SL blood vessels, which were 3-fold higher compared to those in patients with Stage I (P = 0.041) and 2.2-fold higher compared to patients with Stage II disease (Fig. 2).

Vascular structures and immune contexture

The implication of lymphangiogenesis in shaping the cancer immune contexture was then assessed. Therefore, the significance of microvascular density was implemented by its correlation with other relevant phenotypes participating in the constitution of the cancer microenvironment. To this end, an extensive analysis of the incidence and distribution of CD3^pos, CD8^pos and PD-1^pos TILs, together with the expression of PD-L1 by neoplastic cells, was performed.

As expected, a substantial variability in the number of TILs was documented among patients with NSCLC due to different genetic, clinical and biological backgrounds. However, our data suggested that the magnitude of TILs within the tumour was strongly correlated with the density and distribution of the vessels. Indeed, samples with CD3^pos cells above the median had a higher density of both lymphatic and blood vessels. Specifically, at the IM, D2–40^pos and vWF^pos vessels were, respectively, 2-fold (n.s.) and 2.4-fold (P = 0.050) higher than samples with low CD3^pos TILs (Fig. 3A and B).

The importance of different lymphocyte phenotypes in the tumour immune response prompted us to analyse the impact of the CD8^pos cytotoxic T-cell subpopulation on the vascular structures. In SCC, in agreement with our previous findings of CD3 lymphocytes, a high number of intratumour CD8^pos cells was associated with increased vessel density. However, when the relative proportion of cytotoxic cells over the entire population of CD3^pos TILs was considered, a higher microvascular density in the T and at the IM was detected in samples from both histotypes displaying a low CD8^pos/CD3^pos ratio (data not shown).

The PD-1/PD-L1 pathway is an essential checkpoint of the tumour immune contexture, conditioning a suppression of the immune response [19, 20]. The correlation between vessel density and PD-1^pos TILs was investigated. We focused on proximal PD-1^pos lymphocytes located within a 20-µm linear distance from cancer cells that most likely interact with the PD-L1 ligand. A low number of proximal PD-1^pos TILs correlated with a decreased microvascular density (Fig. 3C and D), reaching statistical significance for vWF^pos vessels in the SL (−27%, P = 0.050) and the T (−40%, P = 0.032) (Fig. 3D).

Although PD-L1 expression significantly varied among and within patients with NSCLC, a lower expression of PD-L1 was coupled with an increase in lymphatics (2.2-fold higher, ns) and blood vessels (+25%, ns) located at the IM (Fig. 3E and F).

Microvascular density, programmed death ligand-1 to programmed death-1 ratio and clinical outcome

The immune contexture critically interacts with cancer cells and establishes the immune response, thereby affecting patient outcome. We first attempted to determine the presence of factors with an independent prognostic value. However, a multivariate analysis of the most relevant clinicopathological and immunohistochemical features did not show any variable with statistical impact on OS (data not shown).

Thus, patients with NSCLC within the same pathological stage were stratified according to low or high OS, in the attempt to detect microenvironmental clues with clinical impact. Indeed, several distinct settings of the immune contexture were documented.

Thus, within each stage, patients with low OS displayed a higher incidence of PD-1, lower expression of PD-L1 and increased vascularity (data not shown). Interestingly, among the several investigated parameters, the PD-L1 to PD-1 ratio had the highest clinical impact. Specifically, lower OS was associated with a significant 79% decrease (P = 0.009) in the PD-L1 to PD-1 ratio (Fig. 4A). This more accurate parameter of PD-1/PD-L1 pathway activation was calculated by computing the density of the PD-L1^pos cells and the proximal PD-1^pos lymphocytes (Fig. 4B). Moreover, in the overall population of NSCLC, a striking impact of the PD-L1 to PD-1 ratio on OS was documented by the Kaplan–Meier curve. Indeed, patients with a high ratio (n = 14) had a statistically higher OS compared to those with a low ratio (n = 20) [hazard ratio 0.238, 95% confidence interval (CI) 0.052–1.089; P = 0.044] (Fig. 4C). Interestingly, when patients were stratified according to a PD-L1 to PD-1 numerical ratio and the incidence of peritumour and tumour vessels, a gain of 10 months in OS was documented in patients with a high ratio and vascular rarefaction, although this finding did not reach statistical significance (data not shown).

DISCUSSION

Although angiogenesis has been repeatedly implicated in cancer progression and treatment, the contribution of lymphangiogenic structures to the composition and clinical impact of the immune contexture has not been fully elucidated. An extended analysis of the tumour microenvironment and vascular architecture becomes particularly relevant in light of the success of immunotherapy and the promising results from its combination with antiangiogenic strategies.

In our cohort of patients with NSCLC, the distribution of the vessels documented a peak of lymphatic vessels at the IM and a decreasing gradient of both lymphatic and blood vessels throughout the tumour core. The presence of a similar gradient of double-positive vascular profiles suggested inflammatory processes leading to phenotypic changes in tumour vasculature. The fading of density of the vessels from the IM to the neoplastic core could be ascribed to several factors. Intratumour lymphangiogenic structures may be compressed and/or obstructed by the increased interstitial pressure exerted by growing cancer cells. In addition, the intense remodelling of the extracellular matrix may lead to failure of vascular responsiveness; vascular cells located in the tumour core may express altered and unrecognizable antigens, whereas cancer cells acquire the ability to withstand the lack of oxygen and blood supply [21]. Peritumoural vessels have been shown to be fundamental for tumour progression, representing the preferred route for tumour invasion and metastatization to lymph nodes and distant organs [22]. In line with these contentions, our observations strongly suggest that vessels located in the SL and at the IM are determinant for tumour dissemination. Indeed, a more extended nodal involvement and a higher stage of the disease were associated with increased lymphatics at the IM and blood vessels in the distal spared parenchyma, underlining that the immune shaping that occurs at the periphery may condition tumour growth and progression. Although the density and distribution of the vessels have been indicated as significant prognostic factors of recurrence and survival, they are still not routinely assessed in the staging evaluation [23]. In our cohort of patients with NSCLC, in the presence of an equal magnitude of lymph node involvement, different tumour size did not show a predictable correlation with microvessels. Conversely, in the presence of equal T status, lymph node involvement strongly influenced vessel density, showing a direct correlation. More knowledge about the observed correlations could be critical to include new parameters for the pathological assessment and to define a proper adjuvant post-surgical treatment in patients affected by early-stage NSCLC, specifically in the presence of considerable vascularity.

Cancer development and progression are dependent on cellular alterations that are the result of the deep and continuous interaction between tumour cells and the tissue microenvironment. Vessels represent the network responsible for the constitution of such a microenvironment and play a fundamental role in the control of immune cell trafficking and homing to neoplastic tissues. TILs constitute the natural immune response to cancer and their incidence has been proved to be a predictive factor of OS and prognosis in several types of cancer, including NSCLC, breast cancer, ovarian carcinoma, colorectal cancer, melanoma and non-Hodgkin lymphoma [24]. However, our tissue analysis showed that a higher incidence of CD3^pos cells was correlated with increased microvascular density, particularly at the IM, which in turn was associated with a poorer prognosis. This finding could be explained by the notion that the quantification of lymphocytes based on morphology or CD3 labelling is not sufficient to cover the intrinsic variability of TILs phenotypes within the tumour. Indeed, the immune microenvironment is an intensely composite context, in which vessels may carry not only cytotoxic cells with antitumour properties but also immunosuppressive phenotypes, such as T_reg cells, myeloid-derived suppressor cells and M2-like tumour-associated macrophages [25, 26].

This contention is strengthened by the analysis of the fraction of CD8^pos cells among the total CD3^pos TILs and its impact on microvascular density. In both ADC and SCC cases with a low CD8^pos/CD3^pos ratio, a higher frequency of blood and lymphatic microvessels was detected in each portion of the samples, especially at intratumour and IM sites (data not shown).

Despite the curative intent of complete resection in patients with early-stage NSCLC, the recurrence rate remains high and even patients within the same stage show important disparity in terms of risk of relapses and clinical outcome [27]. The tumour microenvironment may be implicated in such variability. Thus, in the attempt to identify potential factors involved in this phenomenon, we stratified cases on the basis of the OS rate, documenting relevant differences in the characterization of the tumour contexture. This investigation may be crucial for the definition of reliable biomarkers with predictive value and capable of addressing suitable adjuvant and/or neoadjuvant therapeutic approaches. In particular, the PD-1/PD-L1 immune checkpoint, a pathway successfully targeted by multiple immunotherapies in the treatment of advanced NSCLC [28,–,30], was documented to be strongly implicated in patient outcomes. The computation of the PD-L1 to PD-1 numerical ratio, an indirect index of cancer immune escape and lymphocyte anergy, allowed us to speculate on the actual activity of the pathway. At the tissue level, a high ratio, given by high PD-L1 expression and low PD-1 incidence, and the ensuing lower activation of the immunosuppressive status, was correlated with low vessel density and prolonged OS. Importantly, a gain of 10 months in OS was seen in patients whose sample displayed 2 main characteristics of NSCLC immune contexture at tissue level: a high PD-1 to PD-L1 ratio and decreased vascular density.

Limitations

The lack of investigations about lymphocyte subpopulations constitutes a limitation of our study, which requires a wider assessment of TILs to better define the clinical impact of the immune contexture. In light of these observations, it is conceivable that within the cancer contexture, the balance between the different players may lead to different prognosis, clinical outcome and response to therapy.

The actual mechanisms of the interplay among the different microenvironmental factors are far from being fully clarified. Further biological and functional analyses are required to overcome the limitations of the present study. Thus, the involvement of different types of molecules, such as angiogenic factors, cytokines and the chemokine network and their implication in cancer lymphangiogenesis remain to be determined.

Lymphatic and blood vascular structures represent an active and essential component of the tumour immune contexture, interacting with TILs and tumour cells and affecting NSCLC prognosis. Thus, a more comprehensive analysis of the cancer contexture and revisiting tumour lymphangiogenesis functional to TILs recruitment are essential to define immune settings associated with different prognosis, risk of recurrences or progression, in the attempt to achieve true personalized therapies.

CONCLUSIONS

The antitumour immune response is strongly dependent on lymphangiogenic structures, lymphocyte subpopulations and immune checkpoint status. Changes in their quantitative and functional arrangement may be present within the cancer microenvironment, resulting in tumour growth or regression.

Our study highlights the importance of lymphatic and blood vessels as essential components of the tumour immune contexture that significantly impacts the biology and clinical outcome of NSCLC. In particular, we documented their influential role in the main pathological characteristics of the tumour, such as nodal involvement and stage. Remarkably, our analysis provides evidence of interplay between lymphoangiogenic structures and tumour cells, suggesting a reciprocal influence on PD-L1 expression.

The evaluation of vessels may add value to the assessment of the NSCLC immune contexture and its potential prognostic and predictive role for the identification of patients who could benefit from immunotherapy, antiangiogenic approaches and combinatory regimens.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

ACKNOWLEDGMENTS

We gratefully thank Nicoletta Campanini for her invaluable technical assistance.

Funding

This work was supported by the Assiociazione Italiana per la Ricerca sul Cancro (AIRC) [grant number IG-2016-Id19026 to F.Q.]

Conflict of interest: none declared.

REFERENCES

Author notes