Levels of peripheral blood polymorphonuclear myeloid-derived suppressor cells and selected cytokines are potentially prognostic of disease progression for patients with non-small cell lung cancer

Abstract

Polymorphonuclear-MDSC (PMN-MDSC) have emerged as an independent prognostic factor for survival in NSCLC. Similarly, cytokine profiles have been used to identify subgroups of NSCLC patients with different clinical outcomes. This prospective study investigated whether the percentage of circulating PMN-MDSC, in conjunction with the levels of plasma cytokines, was more informative of disease progression than the analysis of either factor alone. We analyzed the phenotypic and functional profile of peripheral blood T-cell subsets (CD3⁺, CD3⁺CD4⁺ and CD3⁺CD8⁺), neutrophils (CD66b⁺) and polymorphonuclear-MDSC (PMN-MDSC; CD66b⁺CD11b⁺CD15⁺CD14-) as well as the concentration of 14 plasma cytokines (IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12 p70, IL-17A, IL-27, IL-29, IL-31, and IL-33, TNF-α, IFN-γ) in 90 treatment-naïve NSCLC patients and 25 healthy donors (HD). In contrast to HD, NSCLC patients had a higher percentage of PMN-MDSC and neutrophils (P < 0.0001) but a lower percentage of CD3⁺, CD3⁺CD4⁺ and CD3⁺CD8⁺ cells. PMN-MDSC% negatively correlated with the levels of IL1-β, IL-2, IL-27 and IL-29. Two groups of patients were identified according to the percentage of circulating PMN-MDSC. Patients with low PMN-MDSC (≤ 8%) had a better OS (22.1 months [95% CI 4.3–739.7]) than patients with high PMN-MDSC (9.3 months [95% CI 0–18.8]). OS was significantly different among groups of patients stratified by both PMN-MDSC% and cytokine levels. In sum, our findings provide evidence suggesting that PMN-MDSC% in conjunction with the levels IL-1β, IL-27, and IL-29 could be a useful strategy to identify groups of patients with potentially unfavorable prognoses.

Electronic supplementary material

The online version of this article (10.1007/s00262-018-2196-y) contains supplementary material, which is available to authorized users.

Introduction

Lung cancer is the leading cause of cancer-associated deaths worldwide, with approximately 2.5 million new cases per year and an annual estimate of 1.5 million deaths worldwide [1]. NSCLC, the predominant type, accounts for approximately 85% of all newly diagnosed cases [2]. During the past decade, advances in the field of cancer immunology have led to the emergence of ‘cancer immunoediting’ as a unified conceptual framework to explain the dual role that the immune system plays during carcinogenesis. Expressly, the fact that immune cells and their factors can both prevent tumor initiation and progression, or promote tumor proliferation, infiltration, metastasis and resistance [3] in several types of cancer, including NSCLC [3, 4]. For instance, it has been shown that the presence of NK, NKT, CTL, T_h and memory T-cells, correlates with a better prognosis and OS in patients with lung cancer [5]. In contrasts, tumor infiltration by macrophages and Tregs has been shown to promote tumor progression and metastasis [6, 7]. In addition to tumor-infiltrating macrophages and Tregs, the activity of TILs can also be modulated by the recruitment/activation of host suppressor cells such as MDSC, which have been the focus of robust investigation in recent years [8–14].

MDSC are a heterogeneous cell population of immature myeloid cells with highly immunosuppressive activity. MDSC are comprised of: (a) early-stage MDSC (E-MDSC); (b) immature mononuclear cells which are morphologically and phenotypically similar to monocytes (M-MDSC); (c) immature polymorphonuclear cells which are morphologically and phenotypically similar to neutrophils (PMN-MDSC, formerly known as G-MDSC) [15]. In humans, MDSC are found in the mononuclear fraction and are broadly defined as positive for the markers CD33 and CD11b, but negative for the myelomonocytic marker CD15 and HLA-DR. Populations with these characteristics (Lin⁻HLA-DR⁻CD33⁺) represent a group of mixed cells enriched for early-stage myeloid progenitors (E-MDSC). M-MDSC are generally defined as positive for the markers CD14, CD11b, but negative or low for CD15 and HLA-DR (CD14⁺, CD11b⁺, CD15^low/neg, HLA-DR^low/neg). On the other hand, PMN-MDSC are defined as positive for the markers CD11b, CD15, CD33 and CD66b, but negative for the monocytic marker CD14 [9, 16]. In the current study, PMN-MDSC were defined as CD66b⁺CD14⁻CD11b⁺CD15⁺.

Various phenotypes of MDSC have been identified in the peripheral blood of NSCLC and small cell lung cancer patients [9, 17, 18]. Nevertheless, there is considerable disagreement with regards to the predominant type of MDSC in cancer patients, with some studies showing that M-MDSC are responsible for immune suppression [19] and others reporting an increase in the percentage of both M-MDSC and PMN-MDSC in patients with NSCLC [9, 20–22].

Initially, it was reported that a high percentage of MDSC was associated with defective DC function, downregulation in the expression of the CD3ζ chain in T-cells, inhibition of antigen-specific T-cell responses, and altered cytokine levels [23, 24]. Subsequently, it was demonstrated that high frequencies of M-MDSC and PMN-MDSC directly correlate with clinical cancer stage, metastatic burden and poor clinical outcomes in NSCLC patients [17, 20–22, 25, 26]. Since then, it has become evident that MDSC promote tumor growth not only by inhibiting T-cell function and proliferation to create a tolerogenic environment, but also by promoting cancer stemness, angiogenesis, metastasis, stroma deposition and epithelial-to-mesenchymal transition [27, 28].

We have previously reported that cytokine profiles may be used to identify patient subgroups with different prognoses in advanced NSCLC [29]. However, in view of the complex and intertwined roles that immune cells and cytokines play in promoting tumoral growth, the aim of the current study was to determine if there is an association between the percentage of circulating immune cells, in particular PMN-MDSC, and the levels of immunoregulatory cytokines, which, in conjunction, could correlate with different survival outcomes in patients with NSCLC.

Materials and methods

Study design

This prospective observational study was conducted at the Lung Cancer Clinic of the Instituto Nacional de Cancerología (Mexico City, Mexico), between May 2011 and November 2012 with the primary objective of evaluating whether the survival outcomes of NSCLC patients are affected by the levels of plasma cytokines and the frequency of PMN-MDSC.

Patient population

A total of 90 patients with advanced NSCLC were enrolled. Eligible patients were 18 years or older, with a recent diagnosis of pathologically confirmed stage IIIB or IV NSCLC, an Eastern Cooperative Oncology Group (ECOG) performance status score of 2 or less and eligible to receive standard first-line systemic therapy (defined as four or more cycles of platinum doublet chemotherapy), as determined by their treating physician. Key exclusion criteria were: previous first-line therapy (radiation, chemotherapy or immunotherapy), history of autoimmune or chronic viral diseases and recent steroid therapy. Clinicopathological characteristics recorded from patients at the time of diagnosis were as follows: sex, age, smoking history, wood smoke exposure, BMI, number and site of metastases, volume of pleural effusion and comorbidities such as diabetes mellitus and hypertension. Blood samples and clinical information from 25 age- and sex-matched HD were obtained from the Blood Transfusion Center Bank.

MDSC analysis in NSCLC patients

Blood samples were collected and processed prospectively (pretreatment) as follows: 8 ml of blood were drawn into EDTA tubes (BD Biosciences). 1 ml of blood was used for the phenotypic characterization of immune cells, performed soon after blood sample collection. From the remaining 7 ml of blood, plasma was isolated by centrifugation and stored at − 70 °C for subsequent cytokine analysis.

Phenotypic characterization

Phenotypic characterization of PMN-MDSC (CD66b⁺CD11b⁺CD15⁺CD14⁻); neutrophils (CD66b⁺); total lymphocytes (CD3⁺); Th cells (CD3⁺ CD4⁺) and CTL (CD3⁺ CD8⁺) was performed by flow cytometry as previously described [30]. Briefly, 100 µl of fresh whole-blood were incubated in the dark with conjugated anti-human mAbs for 25 min at room temperature. Samples were washed once with RBC lysis buffer (1×) and twice with cell staining buffer. Alternatively, for analysis of intracellular CD3ζ chain expression, cells were first permeabilized and then labelled with an intracellular stain. Data from 100,000 events were acquired on a FACS Aria II Flow Cytometer (BD, Biosciences, San José, Calif., USA) and analyses were performed using FlowJo software 10.1 (TreeStar. Ashland, Ore., USA). The leukocyte population was gated based on morphological parameters on a forward vs. side scatter (FSC/SSC) plot. Fluorescence minus one (FMO) and non-stained samples were used as negative controls. Mean fluorescence intensity (MFI) values were extracted from FlowJo and were used to calculate the expression levels of the CD3ζ chain. The following combination of monoclonal antibodies were used: CD11b/PE Cy7, CD66/PerCP Cy5, CD15/Alexa Fluor 488, CD14/APC Cy7, CD3/PE Cy7, CD4/PerCP APCy5.5, CD8/APC Cy7, CD3ζ/FITC (Biolegend, San Diego, Calif. USA), and VEGFR-1/APC (BD Biosciences, San Jose, Calif., USA).

In vitro functional suppression assay

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Paque (GE Healthcare, Bio-science AB, Uppsala, Sweden) density gradient centrifugation (D = 1.077 g/ml) and washed three times with PBS. The mononuclear cell fraction (enriched with low-density PMN-MDSC, monocytes, lymphocytes and thrombocytes) was sorted using magnetic bead isolation kits (EasySep, StemCell Technologies, Vancouver, Canada) to obtain separate populations of CD66b⁺ cells and CD3⁺CD8⁺cells. The purity of sorted cells was determined by flow cytometry (> 90%). Subsequently, CD3⁺CD8⁺cells were cultured alone, or in combination with CD66b⁺ cells at a ratio of 1:2 (lymphocyte:granulocyte) in 24-well plates previously coated with CD3/CD28 antibodies. Cells were kept in complete RPMI 1640 medium supplemented with 10% FBS for 24 h in a humidified atmosphere of 95% air and 5% CO₂ at 37 °C. To evaluate the functional activity of CD3⁺CD8⁺ cells, the expression of the CD3ζ chain was evaluated by flow cytometry as previously described [20, 31, 32].

Measurement of cytokines and chemokines

Plasma levels of 14 cytokines (IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12 p70, IL-17A, IL-27, IL-29, IL-31, and IL-33, TNF-α, IFN-γ) were quantified using a human Th1/Th2/Th17 cytokine cytometric bead array kit (BD, San Jose, CA, USA) as previously described [29]. Briefly, 50 µl of plasma, or standard solutions, were incubated with 50 µl of mixed capture microbeads for 1.5 h at room temperature. Samples were washed and centrifuged at 200×g for 5 min. The pellet was resuspended in 50 µl of a mixture of PE-conjugated mAbs and incubated for 1.5 h at room temperature and protected from direct light exposure. Samples were then washed and centrifuged once more. Data were acquired using a FACS Aria II Flow Cytometer (BD, Biosciences, Mexico) and analyzed with FCAP Array Software V. 3.0 (Soft Flow, Pecs, Hungary).

Statistical analysis

Data analysis was carried out retrospectively from prospectively collected samples. Patient characteristics are summarized as arithmetic means with SD or as medians with ranges, according to data distribution. Normality was assessed by the Kolmogorov–Smirnov test. Two group comparisons were done using the Student’s t test or Mann–Whitney U. The chi-square (χ²) test was used to compare the distribution of categorical variables between groups. Spearman rank correlation analysis and linear regression analysis were used to determine the association between variables. Immune parameters associated with clinical variables were determined by bivariate analysis. Receiver operating characteristic (ROC) was performed to find the best cut-off point level for each of the cytokines evaluated. Median overall survival (OS) was calculated by the Kaplan–Meier method and survival comparisons between groups were performed using the log-rank or the Breslow test. All tests were two-sided with a P ≤ 0.05 considered as statistically significant. Statistically significant and borderline variables (P ≤ 0.1) were included in the multivariate analyses. All data were analyzed using the SPSS software package version 20 (SPSS, Inc., Chicago, Ill, US) and Prism v6.0 software (GraphPad Software, Inc, La Jolla, Calif., USA).

Results

Patient characteristics

Patients and HD exhibited an even distribution, with no significant differences with regards to age, gender, diabetes or hypertension. In contrast, significant differences were found with regards to smoking history [50 vs. 12%] (P < 0.001) and wood smoke exposure [45.6 vs. 4%] (P < 0.001), which were more commonly reported among NSCLC patients. Approximately 85% of patients had stage IV disease. Seventy-five patients (83.3%) had an Eastern Cooperative Oncology Group (ECOG) performance status of < 1. Central nervous system metastases at the time of diagnosis were found in 26.7% of patients (Supplementary Table 1).

NSCLC patients show alterations in the frequency of immune cell subsets

Compared to HD, NSCLC patients exhibited a significant decrease in the percentage of CD3⁺ cells (46.9% ± 11.2 vs. 28.4% ± 13.7; P < 0.0001), CD3⁺CD4⁺ cells (64.8% [43.7–75.3] vs. 31.3% [11.5–70.4]; P < 0.0001), and CD3⁺CD8⁺ cells (24.5% [13.1–41.7] vs. 21% [10.1–39.7]; P = < 0.006), (Table 1 and Supplementary Fig. 3). In contrast, the myeloid compartment was increased in NSCLC patients, as made evident by a higher frequency of neutrophils in NSCLC patients vs. HD (56.4 ± 19.4 vs. 37.6 ± 7.6; P < 0.0001). Furthermore, the flow cytometric analysis revealed that the percentage of CD66b⁺CD14⁻CD11b⁺CD15⁺ cells (typical PMN-MDSC phenotype) was significantly higher in patients (8.9%) than in HD (2.4%); P < 0.0001 (Table 1 and Supplementary Fig. 1a, b). To determine if cells with this phenotype corresponded to PMN-MDSC, their inhibitory activity was evaluated ex vivo.

Table 1.

Immune cell subsets in NSCLC patients and HDs

Variable	Healthy donors	NSCLC	P
	N = 25	N = 90
Neutrophils (%) Mean ± S.D	37.56 ± 7.6	56.39 ± 19.36	< 0.0001**
PMN-MDSCs (%) Median (range)	2.45	8.9	< 0.0001*
	(1.85–4.67)	(2.3–19.56)
CD3+ (%) Mean ± S.D	46.91 ± 11.24	28.42 ± 13.72	< 0.0001**
CD3+CD4+ (%) Median (range)	64.8	31.3	< 0.0001*
	(43.70–75.30)	(11.50–70.40)
CD3+CD8+ (%) Median (range)	24.5	21	0.006*
	(13.1–41.7)	(10.1–39.7)

Statistically significant are in bold

*Mann Whitney U test

**Student’s t test (two-tailed)

Polymorphonuclear cells have Immunosuppressive activity

To confirm the suppressive effect of CD66b⁺ on autologous CD3⁺CD8⁺ cells, both cells were isolated from the PBMC fraction of NSCLC patients and HD, and co-cultured at a ratio of 2:1 for 24 h, after which the CD3ζ chain expression was evaluated on CD3⁺CD8⁺ cells. Alternatively, CD3⁺CD8⁺ cells from NSCLC patients and HD were isolated and plated in the absence of CD66b⁺ cells. CD66b⁺ cells from NSCLC patients significantly decreased the expression of CD3ζ in CD3⁺CD8⁺ cells (Supplementary Fig. 1c). In contrast, CD66b⁺ cells from HD did not affect the expression of CD3ζ. Similar results were obtained when Jurkat cells were used instead of autologous CD3⁺CD8⁺ cells (data not shown).

The expression of VEGFR (a molecular parameter commonly associated with MDSC) was also evaluated. Approximately 97% of CD66b⁺CD14⁻CD11b⁺CD15⁺ cells isolated from both NSCLC patients and controls were also positive for VEGFR. However, the expression of VEGFR on a per cell basis (quantified as MFI) was significantly higher in PMN-MDSC obtained from NSCLC patients than in PMN-MDSC from HD (P = 0.0001), (Supplementary Fig. 2).

T cells from NSCLC patients exhibit alterations in the expression of the CD3ζ chain

In agreement with the results obtained in the ex vivo co-culture system, it was determined that alterations in T-cell percentages were accompanied by constrains in T-cell function, as demonstrated by changes in CD3ζ chain expression. Compared to controls, NSCLC patients had reduced CD3ζ chain expression in CD3⁺ cells (10,008 [7645–13,465] vs 3096 [1194–7793]; P < 0.0001), CD3⁺CD4⁺ cells (5982 [2257–8421] vs 2289 [956–4523]; P < 0.0001), and CD3⁺ CD8⁺ cells (6990 ± 933 vs. 2996 ± 1137; P < 0.0001), (Table 2 and Supplementary Fig. 4).

Table 2.

CD3ζ chain expression in lymphocytes of NSCLC patients and HDs

Variable	Healthy donors	NSCLC	P
	N = 25	N = 43
MFI CD3ζ of CD3+ Median (range)	10,008 (7645–13,465)	3096 (1194–7793)	< 0.0001*
MFI CD3ζ of CD3+CD4+ Median (range)	5982 (2257–8421)	2289 (956–4523)	< 0.0001*
MFI CD3ζ of CD3+CD8+ Mean ± S.D	6990 ± 933	2991 ± 1137	< 0.0001**

Statistically significant are in bold

*Mann Whitney U test

**Student’s t test (two-tailed)

The frequency of circulating immune cells correlates with plasma cytokine levels

A significant positive association was found between neutrophil % and the levels of IL-17A (P = 0.012). On the other hand, the percentage of PMN-MDSC had an inverse correlation with the plasma levels of IL-1β (P = 0.005), IL-2 (P = 0.01), IL-27 (P = 0.033) and IL-29 (P = 0.039). No significant correlations were found between the percentage of CD3⁺ cells and the levels of any of the cytokines evaluated. However, CD3ζ chain expression in CD3⁺ cells negatively correlated with the levels of IL-1β (P = 0.014), TNF (P = 0.039), and IL-31 (P = 0.031), and positively with the levels of IL-33 (P = 0.028). Similarly, CD3ζ expression in CD3⁺CD4⁺ cells negatively correlated with the plasma levels of IL-31 (P = 0.009). Finally, CD3ζ chain expression in CD3⁺CD8⁺ cells negatively correlated with the levels of IL-10 (P = 0.007), (Table 3).

Table 3.

Correlation analysis of immune cells and cytokine levels in NSCLC patients

Variable	Neutrophils	PMN-MDSCs (%)	CD3 (%)	CD3+CD4+ (%)	CD3+CD8+ (%)	MFI CD3ζ of CD3+	MFI CD3ζ of CD3+CD4+	MFI CD3ζ of CD3+CD8+
IL-17A
SC	0.543	− 0.104	0.046	0.016	− 0.036	− 0.008	0.101	0.084
P	0.007	0.348	0.697	0.895	0.761	0.962	0.534	0.604
IL-1β
SC	0.342	− 0.311	− 0.208	− 0.104	− 0.189	− 0.408	− 0.267	− 0.281
P	0.152	0.005	0.084	0.393	0.116	0.014	0.115	0.096
IL-2
SC	0.261	− 0.283	− 0.168	0.023	− 0.123	− 0.014	0.138	− 0.236
P	0.240	0.010	0.151	0.843	0.297	0.931	0.394	0.143
IL-27
SC	0.048	− 0.261	− 0.026	− 0.197	− 0.184	0.057	0.005	− 0.241
P	0.866	0.033	0.844	0.139	0.168	0.793	0.982	0.258
IL-29
SC	− 0.057	− 0.253	− 0.133	− 0.053	− 0.056	− 0.182	− 0.153	0.013
P	0.839	0.039	0.320	0.691	0.677	0.395	0.474	0.951
IL-33
SC	0.156	− 0.064	0.035	− 0.154	0.127	0.448	0.231	− 0.203
P	0.578	0.605	0.795	0.249	0.343	0.028	0.278	0.341
TNF-α
SC	− 0.299	− 0.093	− 0.145	− 0.033	0.014	− 0.346	− 0.244	− 0.263
P	0.214	0.414	0.231	0.786	0.906	0.039	0.152	0.121
IL-31
SC	− 0.056	0.013	− 0.025	− 0.114	0.137	− 0.440	− 0.521	− 0.295
P	0.844	0.918	0.852	0.393	0.305	0.031	0.009	0.162
IL-10
SC	0.229	− 0.089	− 0.175	− 0.170	− 0.058	− 0.119	− 0.241	− 0.445
P	0.346	0.437	0.148	0.159	0.635	0.488	0.157	0.007
IFN-γ
SC	0.171	− 0.178	− 0.115	− 0.010	− 0.176	0.047	− 0.022	0.102
P	0.446	0.107	0.331	0.935	0.135	0.775	0.895	0.532
IL-4
SC	− 0.027	− 0.061	− 0.141	0.059	− 0.033	− 0.135	0.069	0.150
P	0.904	0.581	0.230	0.616	0.783	0.406	0.674	0.356
IL-6
SC	0.071	− 0.118	− 0.130	− 0.188	− 0.151	0.049	− 0.234	− 0.060
P	0.772	0.301	0.284	0.119	0.211	0.777	0.170	0.728
IL-8
SC	− 0.056	0.145	0.061	− 0.067	0.013	0.073	− 0.245	− 0.191
P	0.821	0.201	0.616	0.579	0.912	0.672	0.150	0.265
IL-12 p70
SC	0.061	0.036	− 0.121	0.140	− 0.101	− 0.133	− 0.232	− 0.250
P	0.803	0.750	0.317	0.247	0.404	0.440	0.173	0.142

Statistically significant are in bold

SC Spearman correlation

Clinical characteristics correlate with CD3ζ chain expression

High CD3ζ chain expression in CD3⁺ cells was more commonly found among older patients (> 60 years 3452 vs. ≤60 years 2332; P = 0.044). Similarly, high CD3ζ expression in CD3⁺CD8⁺ cells was more commonly found among male patients (3654.3 vs. 2569.8: P = 0.016). Finally, a reduced frequency in the number of CD3⁺ cells was associated with advanced stage disease (stage IIIB 34.9 vs. stage IV 27.3; P = 0.006), metastases (35.4 vs. 27.5; P = 0.021) and oligometastases (< 3 33.3 vs. ≥3 27.1; P = 0.041), (Supplementary Table 2).

Clinical characteristics associated with OS

The median follow-up of patients was 18.2 months (range 15–21 months) with an OS of 14.8 months [95% CI 10.3–19.3]. Clinical and demographic characteristics (gender, age > 60, smoking history, wood smoke, hypertension, diabetes mellitus, BMI, metastases, oligometastases, CNS metastases at diagnosis and pleural effusion) did not impact median OS. Disease stage and ECOG status were the two only clinical factors associated with OS. Patients with IIIB stage NSCLC had a better OS than those with stage IV disease (29.5 vs. 11.6 months: P = 0.029). Patients with an ECOG ≤ 1 also had a better OS (16.4 vs. 8.5 months; P = 0.004), (Supplementary Table 3).

Complete blood count parameters associated with OS

An increased number of leukocytes (> 8500/mm³), neutrophils (> 6250/mm³) and platelets (> 338,000/mm³), estimated by complete blood count, were associated with a lower median OS, as were low albumin (< 3.5 gr/dl) and hemoglobin (14.4 gr/dl) levels. Although the total number of lymphocytes was not associated with OS (on its own), a high platelet to lymphocyte ratio (PLR) and a high neutrophil to lymphocyte ratio (NLR) were associated with a worse OS (Supplementary Table 4).

Immune cell characteristics associated with OS in subgroups of NSCLC patients

In view of the fact that no samples from HD had a PMN-MDSC% greater than 8, it was decided to dichotomize patients into two groups: those with above normal range of PMN-MDSC% (> 8%), and those with a PMN-MDSC% within the normal range (≤ 8%). There were no differences in the PFS of patients according to the percentage of PMN-MDSC (Fig. 1a). In contrast, the survival analysis of these groups revealed that PMN-MDSC% correlates with a reduction in OS. Patients with high PMN-MDSC% (N = 53) had an OS of 9.3 months [95% CI 0–18.8] while patients with low PMN-MDSC% (N = 37) had an OS of 22.045 months [4.3–739.7] (Fig. 1b and Supplementary Table 5). Although not statistically significant, there was a tendency for better OS in patients with a higher frequency of total CD3⁺ cells (> 27.4%), and CD3⁺CD8⁺ cells (> 27.8%CD3⁺/CD8⁺). In contrast, there was a tendency for improved OS in patients with a reduced percentage of CD3⁺CD4⁺ cells (≤ 31.3% CD3⁺/CD4⁺). Decreased CD3ζ chain expression in CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ T-cells caused a markedly, albeit not statistically significant, reduction in the OS (Supplementary Table 5).

Fig. 1.

Kaplan–Meier curves of OS by PMN-MDSC% and cytokine levels. a PFS of patients classified by PMN-MDSC%, b OS of patients classified by PMN-MDSC% alone or stratified by PMN-MDSC% in conjunction with the levels of c IL-1 β, d IL-2, e IL-27, f IL-29

Immune cell characteristics in NSCLC patients and HD (high/low PMN-MDSC percentage)

Compared to HD, NSCLC patients (with either high or low PMN-MDSC%) had a higher percentage of neutrophils but a lower percentage of CD3⁺, CD3⁺CD4⁺ and CD3⁺CD8⁺ cells as well as lower CD3ζ chain expression. Interestingly, the only difference between patients with low PMN-MDSC% and those with high PMN-MDSC% was with regards to the frequency of CD3⁺ cells, which was significantly increased (20.8 vs 32.4%; P = 0.0002) in the group of patients with high PMN-MDSC%, (Supplementary Table 6).

PMN-MDSC% and cytokine levels affect OS in NSCLC patients

In an attempt to elucidate the clinical implications of the correlation between PMN-MDSC% and the levels of IL-1β, IL-2, IL-27 and IL-29 previously found, patients were classified into four groups as follows: those with Low PMN-MDSC% (≤ 8%) and those with High PMN-MDSC% (> 8%) and then, further classified as High IL or Low IL. The cut-off point value for the aforementioned cytokines was as follows: IL-1β = 0.2 pg/ml; IL-2 = 1.6 pg/ml; IL-27 = 194.8 pg/ml; and IL-29 = 36.7 pg/ml. Finally, OS was calculated for each group (low IL/low MDSC, low IL/high MDSC, high IL/high MDSC, high IL/low MDSC).

OS in subgroups of patients stratified by PMN-MDSC% and IL-1β levels

Patients with a high percentage of PMN-MDCS (> 8%) and high levels of IL-1β (> 0.2 pg/ml) had a dramatically poor OS (4.9 months). In contrast, OS was increased to 8.5 months in patients with high PMN-MDSC% and low IL-1β (≤ 0.2 pg/ml). In line with this, patients with low PMN-MDSC% and low levels of IL-1β had the best OS (22.1 months). In patients with low PMN-MDSC% and high IL-1β, OS was reduced to 12.9 months, (Fig. 1c; Table 4).

Table 4.

OS by PMN-MDSCs % and cytokine levels

Variable	Median	95% Confidence Interval		P
		Lower Bound	Upper Bound
PMN-MDSCs (%)
≤ 8	22.045	4.355	39.735	0.039
> 8	9.265	0.000	18.81
IL-1β (pg/ml)
≤ 0.2339	14.883	7.563	22.203	0.923
> 0.2339	11.565	3.856	19.273
IL-2
≤ 1.575	12.879	3.380	22.378	0.557
> 1.575	12.320	0.000	24.640
IL-27 (pg/ml)
≤ 194.8	16.427	8.466	24.389	0.645
> 194.8	11.565	8.307	14.822
IL-29 (pg/ml)
≤ 36.74	11.926	5.459	18.394	0.441
> 36.74	16.427	3.477	29.377
IL-1β + PMN-MDSCs
≤ 0.2339 ≤ 8	22.045	5.136	38.954	0.037
≤ 0.2339 > 8	8.509	0.000	19.542
> 0.2339 > 8	4.961	4.792	5.130
> 0.2339 ≤ 8	11.926	9.695	14.157
IL-2 + PMN-MDSCs
≤ 1.575 ≤ 8	36.205	2.357	70.054	0.113
≤ 1.575 > 8	6.505	0.722	12.288
> 1.575 > 8	8.509	0.000	32.529
> 1.575 ≤ 8	12.320	0.000	24.890
IL-27 + PMN-MDSCs
≤ 194.8 ≤ 8	22.045	NR	NR	0.009
≤ 194.8 > 8	15.573	0.000	39.115
> 194.8 > 8	4.895	0.925	8.865
> 194.8 ≤ 8	12.320	0.000	27.206
IL-29 + PMN-MDSCs
≤ 36.74 ≤ 8	22.045	1.425	42.665	0.049
≤ 36.74 > 8	5.092	0.000	11.290
> 36.74 > 8	16.427	0.000	41.046
> 36.74 ≤ 8	36.205	8.252	64.159

Statistically significant are in bold

OS in subgroups of patients stratified by PMN-MDSC% and IL-2 levels

The OS analysis of the four groups of patients classified by PMN-MDSC% and IL-2 levels did not reach statistical significance. Nevertheless, there was a tendency for IL-2 to negatively affect OS in patients with low PMN-MDSC% and to slightly improve it in patients with high PMN-MDSC%. Our results show that patients with low PMN-MDSC% and low levels of IL-2 (≤ 1.6 pg/ml) had the best OS (36.2 months). In patients with low PMN-MDSC% and high IL-2 (> 1.6 pg/ml), OS was decreased to 12.3 months. In contrast, patients with high PMN-MDSC% and low IL-2 had a worse OS (6.5 months) than patients with high PMN-MDSC% and high IL-2 (8.1 months), (Fig. 1d; Table 4).

OS in subgroups of patients stratified by PMN-MDSC% and IL-27 levels

Patients with low PMN-MDSC% and low levels of IL-27 (≤ 194.8 pg/ml) had a higher OS (22.1 months) than those with low PMN-MDSC% and high IL-27 (> 194.8 pg/ml), whose OS was 12.3 months. Similarly, patients with high PMN-MDSC% and low levels of IL-27 had a higher OS (15.6 months) than those with high PMN-MDSC% and high IL-27 (4.9 months), (Fig. 1e; Table 4).

OS in subgroups of patients stratified by PMN-MDSC% and IL-29 levels

Patients with high PMN-MDCS% and low levels of IL-29 (≤ 36.7 pg/ml) had a worse OS (5.1 months) than patients with high PMN-MDSC% and high IL-29 (> 36.7 pg/ml), whose OS was 16.4 months. Similarly, patients with low PMN-MDSC% and high levels of IL-29 had a better OS (36.2 months) than those with low PMN-MDSC% and low IL-29 (22.1 months), (Fig. 1f; Table 4). In a multivariate analysis, ECOG, Albumin (gr/d) and PMN-MDSC (%) remained independent prognostic factors for survival (Table 5).

Table 5.

Analysis of Variance

Variable	Univariate				Multivariate
	Median	95% Confidence Interval		P	HR	95% Confidence Interval		P
		Lower bound	Upper bound			Lower bound	Upper bound
Clinical stage
IIIB	29.47	19.871	39.07	0.029	1.226	0.431	3.125	0.670
IV	11.565	8.191	14.938
ECOG
1	16.427	9.619	23.235	0.004	1.968	1.050	3.688	0.035
2 y 3	8.476	2.836	14.117
Albumin (gr/dL)
≤ 3.5	6.669	1.982	11.357	< 0.0001	0.33	0.180	0.604	< 0.0001
> 3.5	29.47	21.016	37.924
Neutrophils (thousands/mm3)
≤ 6.25	22.177	10.132	34.221	0.033	1.397	0.800	2.440	0.240
> 6.25	11.532	4.306	18.757
PMN-MDSCs (%)
≤ 8	22.045	4.335	39.735	0.039	1.926	1.088	3.408	0.024
> 8	9.265	0.00	18.81

Statistically significant are in bold

HR Hazard ratio

Discussion

Abnormal myelopoiesis and neutrophilic leukocytosis (neutrophilia) have long been observed in NSCLC patient, either at the time of diagnosis or during the course of the disease [33]. However, it was not until the recent discovery of MDSC that the significance of these pathologies began to be elucidated [34].

Our results show that NSCLC patients have a significantly higher percentage of neutrophils than HD. Indeed, an expansion in the size of the myeloid compartment during pathological conditions could be an indicator of an underlying increase in the number of MDSC [15]. The flow cytometry analysis revealed that the percentage of CD66b⁺CD11b⁺CD15⁺CD14⁻ cells (the typical phenotype of PMN-MDSC) was negligible in HD but markedly increased in NSCLC patients, which is in agreement with previous studies [15, 35]. The survival analysis of subgroups of patients stratified according to their neutrophil count (estimated by complete blood count), indicates that OS was significantly reduced among patients with a high neutrophil count. In contrast, when the survival analysis was performed using the phenotypic data (determined by flow cytometry), neutrophil percentage (CD66b⁺ cells) was not associated with changes in the OS of patients. Although the PFS of patients was not affected by the percentage of CD66b⁺CD11b⁺CD15⁺CD14⁻ cells, we found that a high percentage of CD66b⁺CD11b⁺CD15⁺CD14⁻ cells significantly reduced the OS of patients.

Debate continues regarding the distinction between PMN-MDSC from neutrophils due to their morphological similarities and the lack of a unique set of markers to differentiate them (cell-surface signature). Currently, density gradient centrifugation is the most reliable method to separate neutrophils from PMN-MDSC. PMN-MDSC are low-density cells that remain at the interface between the plasma and the Ficoll-Paque layer (commonly referred to as the mononuclear cell layer), in contrast to mature neutrophils, which are high-density cells that migrate through Ficoll media to the granulocytic layer found above the globular fraction.

In contrast to CD66b⁺ cells isolated from the mononuclear cell layer of HD, CD66b⁺ cells from NSCLC patients effectively suppressed T-cell function in vitro, as evidenced by a decrease in their expression of the CD3ζ chain. These results confirm that CD66b⁺CD11b⁺CD15⁺CD14⁻ cells from NSCLC patients have immunosuppressive activity and thus correspond to PMN-MDSC.

Although it has generally been accepted that the defining characteristic of PMN-MDSC is their immunosuppressive and protumor activity, this view has been challenged by studies describing a differential state of neutrophil activation, known as neutrophil polarization, whereby in response to tumor-derived factors, some neutrophils acquire immunosuppressive and protumor functions (referred to as N2 neutrophils) while others prevent tumor development and progression (referred to as N1 neutrophils) [36, 37]. Currently, it is not possible to distinguish PMN-MDSC from N2 neutrophils, as both cell types have overlapping phenotypes and similar functional characteristics [15]. The current study is limited by the fact that only PMN-MDSC were functionally and phenotypically characterized. However, various other subtypes of MDSC have been detected in the peripheral blood of NSCLC patients (M-MDSC, PMN-MDSC, E-MDSC).

Neutrophilia, and the accumulation of immature and immunosuppressive myeloid progenitors, have been associated with tumor secretion of pro-inflammatory cytokines and chemokines [12]. Our group recently published evidence indicating that cytokine profiles define subgroups of NSCLC patients with different prognoses [29]. Building on this previous work, we investigated whether the relationship between plasma cytokine levels and the abundance of specific immune cell subpopulations could be used to generate immune profiles that stratify patients with different survival outcomes. To this end, the abundance of specific immune cells and the concentration of 14 plasma cytokines (IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12 p70, IL-17A, IL-27, IL-29, IL-31, and IL-33, TNF-α, IFN-γ) were determined from peripheral blood samples of NSCLC patients (N = 90) and HD (N = 25).

The cytokine level assessment indicated that neutrophils and PMN-MDSC are associated with different cytokines. For instance, IL-17 levels positively correlated with the percentage of neutrophils but not with the percentage of PMN-MDSC. Conversely, the levels of IL-1β, IL-2, IL-27 and IL-29 correlated with the percentage of PMN-MDSC but not with the percentage of neutrophils.

It has been reported that the frequency of T_h17 cells and the levels of IL-17 are elevated in patients with lung adenocarcinoma [38]. The release of IL-17 stimulates a strong and chronic inflammatory response that leads to an increase in the recruitment of neutrophils [39]. Moreover, there is strong evidence indicating that high levels of IL-17 promote tumor growth, angiogenesis and metastasis, which in turn correlates with a poor prognosis in NSCLC [40, 41].

Previous reports have found that lymphopenia in patients with NSCLC entails a bad prognosis [42, 43]. Recently, we reported NSCLC patients exhibit important alterations in the number and function of peripheral T-cells, especially among patients with a greater nutritional compromise [44]. In accordance with previous reports [17, 20, 25, 26, 44], we have confirmed that compared to HD, NSCLC patients have a lower percentage of CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ cells.

Cancer patients exhibit important alterations not only in the quantity of peripheral and infiltrating T lymphocytes but also in their functional characteristics (underlying their ineffectiveness in containing tumoral growth), as evidenced by the presence of anergic and exhausted T cells [3, 45].

Preclinical studies have demonstrated that macrophages and MDSC cause T-cell anergy by decreasing the expression of the T-cell receptor CD3ζ chain through a mechanism involving increased L-arg catabolism by arginase and iNOS [31, 46]. Indeed, in addition to reduced T-cell proliferation, decreased expression of the T-cell receptor CD3ζ chain has been used as an indicator of MDSC-induced immunosuppression. In line with the above, our results show that, compared to HD, NSCLC patients with both high and low PMN-MDSC% exhibited a dramatic and significant reduction in the percentage of CD3⁺, CD3⁺CD4⁺, and CD3⁺CD8⁺ cells, as well as a significant decrease in the expression of the CD3ζ chain in these immune cell subsets. This finding confirms the association between a high percentage of PMN-MDSC and a reduction in the frequency and function of CD8⁺ [20] and CD4⁺ cells [22].

Among patients with either low or high PMN-MDSC%, those with high levels of IL-1β had the worst OS. These results are consistent with other studies showing that IL-1β activates a pro-inflammatory network that promotes tumor-mediated immune suppression, angiogenesis and invasiveness [47] through several mechanisms including the recruitment, expansion and activation of MDSC [48, 49]. It has been shown that IL-6, increases the number of MDSC, acting either downstream of IL-1β or through a mechanism overlapping that of IL-1β [50]. Furthermore, high IL-1β and IL-6 levels have been associated with a poor prognosis in patients with esophageal squamous-cell carcinoma. In addition, various studies have demonstrated that polymorphisms of the IL1B gene, resulting in increased IL-1β production, are associated with a higher risk of developing NSCLC [51–54]. IL-1β is synthesized as an inactive precursor (pro-IL-1β), requiring proteolytic processing to be secreted in its active form. Stimulation of ‘pattern recognition receptors’ (PRR) on innate immune cells triggers the NF-kB-dependent synthesis of pro-IL-1β and NALP3 (a component of the inflammasome). Further PRR stimulation leads to the activation of the inflammasome and of caspase-1. Caspase-1, in turn, cleaves and hydrolyzes pro-IL-1β into two fragments, one of 26-kDa (with unknown function) and another one of 17-kDa (the mature and active form of IL-1β). Alternatively, during neutrophil-dependent inflammation, IL-1β can be processed by neutrophil-derived proteases such as elastase, cathepsin G and proteinase-3 [56, 57]. Interestingly, a recent study showed that inhibition of IL-1β (for 3 months with canakinumab) was associated with (a) a reduction in the occurrence of fatal and non-fatal lung cancers; (b) a reduction in IL-6 concentration and; (c) a pathologic reduction in the number of neutrophils [57].

A significant negative correlation was found between the levels of IL-2 and PMN-MDSC%, supporting the notion that PMN-MDSC (CD11b^hiLy6G⁺) can inhibit T-cell secretion of IL-2 [58]. Although our study was not sufficiently powered to allow for a meaningful survival analysis, there was a tendency indicating that the effect of IL-2 is dependent on PMN-MDSC%. For instance, we noted that among NSCLC patients with Low PMN-MDSC%, those with high plasma levels of IL-2 had a dramatically shorter OS than patients with low levels of IL-2 (36.205 vs. 12.320 months). On the other hand, among patients with High PMN-MDSC%, those with High IL-2 had a marginally better OS, than those with low IL-2 levels (8.509 vs. 6.505 months). Upon closer examination of our data set, we noticed that a minute shift in the cut-off point level of IL-2 modified the group assignment of one patient, with major consequences in the significance level but without a considerable impact on the aforementioned OS patterns. Therefore, we believe that caution should be exercised in categorically rejecting or accepting the null hypothesis. Particularly, in view of the fact that previous studies have suggested that the presence of immunosuppressive cells may underlie the limited and paradoxical success of IL-2 therapy for the treatment of NSCLC [59, 60].

Our results indicate that high serum levels of IL-27 are associated with poor OS in NSCLC patients, regardless of PMN-MDSC%. For instance, among patients with Low PMN-MDSC%, those with high levels of IL-27 had a worse OS than those with low IL-27 levels (22.045 vs. 12.32 months). Similarly, among patients with High PMN-MDSC%, those with high IL-27 levels had a worse OS than patients with low IL-27 levels (15.573 vs. 4.895 months). Although, these results differ from some preclinical studies indicating that IL-27 exerts a potent antitumor effect [61–65], they are broadly consistent with preclinical and clinical studies reporting that increased serum levels of IL-27 correlate with tumor growth and disease progression in various types of cancer [64–66], including lung cancer [67].

IL-29 has been shown to exert a strong antiproliferative and proapoptotic effect on several NSCL-derived cell lines (Sq-1, Sq-19, LK-1, LK-2, OBA-LK1, 11–18, LK-79, 86-2, Lu99, EBC-1 and A549) as well as a strong antitumor effect on SCID mice previously inoculated with OBA-LK1, LK-1 and A549 cells through a mechanism that involves an increase in the expression of p21 [68]. We now provide clinical evidence suggesting that IL-29 exerts an antitumor effect in NSCLC patients. We show that among patients with low PMN-MDSC%, those with high IL-29 levels exhibited a better OS than patients with low IL-29 levels (36.205 vs. 22.045 months). Similarly, among patients with high PMN-MDSC%, those with high IL-29 levels had a better OS than patients with low IL-29 levels (16.427 vs. 5.092 months).

Taken together, our results provide further evidence to strengthen the notion that determining cytokine profiles and performing an initial assessment of the predominant immune cell subsets present in cancer patients, is an important clinical strategy that can be used to identify high-risk patients more accurately. Although great progress has been made in understanding the function of isolated cytokines under defined experimental conditions, currently there is not a good understanding of how exactly these networks contribute to cancer pathogenesis. This highlights the need for models and tools that can effectively integrate the vast immunological information that studies such as ours have yielded, to translate this information into actionable therapeutic targets.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

BMI

Body mass index

CNS mets

Central nervous system metastases

e-MDSC

Early-stage myeloid-derived suppressor cell(s)

ECOG

Eastern Cooperative Oncology Group

FMO

Fluorescence minus one

G-MDSC

Granulocytic-MDSC(s)

HD

Healthy donor(s)

iMC

Immature myeloid cell(s)

iNOS

Inducible nitric oxide synthase

M-MDSC

Monocytic-MDSC(s)

NLR

Neutrophil to lymphocyte ratio

NR

Not reached

PLR

Platelet to lymphocyte ratio

PMN-MDSC

Polymorphonuclear-MDSC(s)

TAN

Tumor-associated neutrophil(s)

Author contributions

Conception/design: OA, LB. Provision of study material or patients: OA, Interviewed patients and referred them for the study: DF-E, MO-M. Collection and/or assembly of data: EM-S, EM-S, J-MH-M, DM-T, RAM-F. Data analysis and interpretation: LB, EM-S, J-MH-M, OA. Manuscript writing: LB, EM-S, J-MH-M, OA. Final approval of manuscript: LB, EM-S, J-MH-M, MO-M, EM-S, DM-T, RAM-F, DF-E, OA.

Funding

This work was supported by the National Council of Science and Technology of Mexico (CONACYT) [161599 to Lourdes Barrera and 87453 to Oscar Arrieta].

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflicts of interest.

Ethical approval and ethical standards

This study was done in accordance with the provisions of the Declaration of Helsinki and Good Clinical Practice guidelines, under a protocol approved by the Ethics Committee and Institutional Review Board (IRB) of the Instituto Nacional de Cancerología (Mexico City, Mexico), [INCAN (011/018/ICI-CV/683)].

Informed consent

All participants in this study provided written informed consent before any study-related procedures were performed.