The clinical significance of tertiary lymphoid structure and its relationship with peripheral blood characteristics in patients with surgically resected non-small cell lung cancer: a single-center, retrospective study

Abstract

Introduction

The presence of tertiary lymphoid structure (TLS) in tumor tissues has been reported to be a factor associated with a good prognosis in several types of cancers. However, the relationship between TLS formation and peripheral blood findings remains unclear. The purposes of the study were to evaluate the effect of the presence of TLS on survival and determine the peripheral blood characteristics associated with TLS formation in non-small cell lung cancer (NSCLC) patients.

Methods

A total of 147 consecutive NSCLC patients who underwent lung resection at Fukushima Medical University Hospital between 2013 and 2017 were enrolled. TLS expression was evaluated, and the relationships between clinical parameters and outcomes were analyzed. Peripheral blood mononuclear cells (PBMCs) were further analyzed by mass cytometry to characterize the TLS-positive microenvironment.

Results

Forty-six patients had high TLS expression, and the remaining 101 patients had low TLS expression. In stage II to IV patients (n = 35), disease-free survival was longer in the high TLS expression group (p = 0.027). A low neutrophil to lymphocyte ratio (NLR) < 2.75 in the peripheral blood was associated with high TLS expression (p = 0.003). Citrus analysis after mass cytometry assay showed that the number of cells expressing HLA-DR and CD9 in PBMCs was lower in the high TLS expression group.

Conclusion

High TLS expression is associated with a good prognosis after surgery in stage II and III NSCLC patients. In the peripheral blood, a low NLR and few antigen-presenting cells indicate the presence of TLS in the tumor microenvironment.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-021-03067-3.

Introduction

Lung cancer remains the leading cause of cancer death, with an estimated 1.8 million deaths (18%) worldwide [1]. According to a survey in Japan, the number of surgically resected cases of lung cancer is increasing, with 44,140 cases of lung cancer surgically resected in 2017 [2]. The long-term outcomes of resected cases are improving, and the five-year survival rate has improved from 51.9% in 1994 to 74.7% in 2010 [3]. Immune checkpoint inhibitors (ICIs) now have a place in standard therapy for the treatment of advanced or recurrent non-small cell lung cancer (NSCLC), in addition to chemotherapy and molecular-targeted therapy. Whereas ICIs show marked therapeutic effects in some patients, there are still unclear points about the mechanisms and biomarkers that can predict their effectiveness. Currently known predictors are the expression of programmed cell death-ligand 1 (PD-L1) in tumor tissue [4, 5], tumor mutation burden [6, 7], tumor-infiltrating lymphocytes (TILs) [8, 9], gut microbiota [10–12], and tertiary lymphoid structure (TLS) [13–15]. Studies evaluating detailed components such as myeloid cells and other immune cells in the tumor microenvironment are increasing [16]. We have been studying TILs and TLS in the tumor microenvironment.

TLS is an ectopic lymphoid structure found in inflammation, infection, and tumor tissue [17]. It is composed of mature dendritic cells, T cells, B cells forming germinal centers, and high endothelial venules (HEVs) [18]. It is thought that TLS plays a role in priming and activating T cells in the tumor microenvironment. Rakaee et al. reported that TLS in the tumor microenvironment was associated with a good prognosis in NSCLC [19]. After neoadjuvant ICI therapy, TLS was found in the tumor regression bed as an immune-related pathologic response [20]. These reports suggest that TLS may support an effective antitumor effect [13]. However, the associations between TLS and the clinical characteristics of patients with NSCLC are still not well-known.

In particular, the characteristics of peripheral blood with TLS in the tumor microenvironment are largely unknown. Zhu et al. reported that greater CD4 and CD8 T cell clonality in tumor and peripheral components was correlated with high TLS density in NSCLC [21]. It is supposed that the tumor microenvironment, including TLS, is associated with peripheral blood characteristics, but there is no consensus about their associations. This is because a large tumor sample is needed to evaluate TLS histologically.

This study was performed to examine the clinical significance of TLS in surgically resected NSCLC. Disease-free survival (DFS) and clinical characteristics were analyzed. Additionally, the correlations of characteristics of peripheral blood lymphocytes with TLS were also examined; the characteristics of peripheral blood lymphocytes were compared between a TLS high-expression group and a TLS low-expression group in patients with NSCLC.

Materials and methods

Patients. A total of 147 patients with primary NSCLC who underwent lung resection at Fukushima Medical University from 2013 to 2016 were enrolled. Preoperative peripheral blood samples were collected and preserved in all cases. None of the patients underwent preoperative chemotherapy, including ICIs. Staging was evaluated pathologically based on the International Union Against Cancer TNM classification, 7th edition.

Evaluation of TLS. TLS was defined as the presence of lymphocyte collections and HEVs [17]. An HEV was identified by immunohistochemical staining of paraffin-embedded tissue using rat anti-mouse antibody against peripheral node addressin (PNAd) (1:100; cat. no. 553863; BD Pharmingen, Inc., San Diego, CA, USA). The primary antibody was detected using biotinylated secondary anti-mouse IgG (1:400; cat. no. E0413; Dako; Agilent Technologies, Inc., Santa Clara, CA, USA). Counterstaining with Mayer’s hematoxylin (Muto Pure Chemicals Co., Ltd., Tokyo, Japan) was done. TLS was measured with a weak magnification (10 × 4) and 10 fields of view. Micrographs were judged by two physicians without knowledge of the patients’ background characteristics. In 10 fields of view, specimens without TLS were defined as -, those with 1 to 5 TLSs were defined as + −, those with 6 to 10 TLSs were defined as + , and those with 10 or more TLSs were defined as +  + . Analysis was performed comparing low-expression (-, + -) with high-expression (+ , + +). Figure 1 shows an example of each.

Fig. 1.

Micrographs of typical TLS high-expression and low-expression cases. a TLS high-expression case. b TLS low-expression case. TLS is defined as the confirmation of lymphocyte collection and HEV. HEV is identified by immunohistochemical staining using an antibody against peripheral node addressin (PNAd). The sections were counterstained by hematoxylin

Statistical analysis. DFS was analyzed as a main endpoint and was defined as relapse of the primary tumor. DFS curves were estimated by Kaplan–Meier methods. DFS was compared by the log-rank test between groups of patients who expressed high and low TLS levels. The χ² test or Fisher’s exact test was used for univariate analysis, as appropriate. P values less than 0.05 were considered significant. The Cox proportional hazards analysis using the forward stepwise likelihood ratio method was carried out with baseline factors. Baseline factors were TLS level, age, sex, histology, smoking history, tumor diameter, pathological T status, pathological N status, pathological stage, epidermal growth factor receptor (EGFR) mutation, and the neutrophil to lymphocyte ratio (NLR). Statistical analysis was performed using SPSS version 24 (IBM, Armonk, NY, USA).

PBMC collection and staining for CyTOF. Preoperative blood was drawn at the time of admission. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll-Paque and then cryopreserved at −80 °C until assayed. Five typical cases were chosen from the TLS low-expression and high-expression groups, respectively. Their samples were thawed and washed twice in phosphate-buffered saline (PBS). The cells were resuspended to 3 × 10⁶/mL in PBS, and Cell-ID Cisplatin-¹⁹⁸Pt (Fluidigm, South San Francisco, CA, USA) was added. Samples were mixed well and incubated at room temperature for five minutes. The staining was quenched twice with 5 mL of Maxpar Cell Staining Buffer (Fluidigm), and the cells were centrifuged. The cells were then resuspended to 5 mL of Maxpar Cell Staining Buffer. All cell surface antibody cocktails were added and incubated at room temperature for 15 min (Supplementary Table 1). Cells were washed twice in 2 mL of Maxpar Cell Staining Buffer (Fluidigm). The cells were then fixed in 2% formaldehyde. The data were analyzed on a Helios mass cytometer (Fluidigm).

viSNE and citrus analysis. All of the obtained CyTOF.fcs files were uploaded to Cytobank. Before viSNE and citrus analysis, single and live cells were gated on Cytobank. ¹⁹¹Ir and ¹⁹³Ir DNA intercalator and ¹⁹⁸Pt were used to distinguish live intact singlets from debris and cell aggregates. viSNE is a dimensionality reduction algorithm available in Cytobank [22]. Thirty-four markers (Fluidigm), including CCR7, CD11a, CD127, CD16, CD25, CD27, CD3, CD4, CD44, CD45, CD45RA, CD45RO, CD57, CD69, CD8, HLA-DR, CCR4, CCR5, CD161, CD2, CD28, CD49D, CD5, CD7, CD9, CXCR3, CD137, CD95, CTLA-4, ICOS, LAG3, OX40, PD-1, and TIM-3, were used to build the viSNE map. The expression level of each marker was visualized by color on the maps.

Citrus (cluster identification, characterization, and regression) is an automated analysis to detect statistically distinct features in subpopulations [23]. Files were categorized into the TLS low-expression group and the TLS high-expression group, and the 34 markers described above were used as clustering parameters. A Nearest Shrunken Centroid (PAMR) model was used as the predictive algorithm. The False Discovery Rate (FDR) threshold confirmed that the model did not include too many false-positive features.

Results

Associations between TLS and clinical characteristics. Table 1 shows the clinical features of a total of 147 cases. The median age was 70 years, and 66.7% were male. Adenocarcinoma was seen in 72.7% of all cases. A total of 112 cases (76.1%) were stage I, with Stage IA in 77 cases (52.3%) and IB in 35 cases (23.8%). One case was stage IV because pleural dissemination was found on pathological examination after surgery. The median observation period was 35 months, and recurrence was observed in 31 patients (21.1%). The preoperative NLR was 0.71–9.22, with a median of 2.75. Of the 147 cases, TLS expression was low (−and + −) in 101 cases (68.7%) and high (+ and + +) in 46 cases (31.3%). Table 2 shows a comparison of the degree of TLS expression for each clinical background. The median NLR, 2.75, was adopted as the cutoff. There were no significant differences in age, sex, histology, smoking history, tumor diameter, or stage between the groups.

Table 1.

Characteristics of all patients

Characteristic	n	%
Median age at surgery, y (range)	70 (40–87)
Median follow-up after surgery in months (range)	35 (1–61)
Sex
Male	98	66.7
Female	49	33.3
Histology
Adenocarcinoma	107	72.7
Squamous cell carcinoma	39	26.5
Others	1	0.7
Smoking, median pack-years (range)	32 (0–129)
Median tumor diameter (cm) (range)	2.6 (0.8–11.0)
Pathological T status
T1a	47	32.0
T1b	35	23.8
T2a	47	32.0
T2b	4	2.7
T3	10	6.8
T4	4	2.7
Pathological N status
N0	122	83.0
N1	11	7.4
N2	13	8.8
N3	1	0.7
Pathological stage
IA	77	52.3
IB	35	23.8
IIA	8	5.4
IIB	9	6.1
IIIA	14	9.5
IIIB	3	2.0
IV	1	0.7
EGFR mutation
Positive	27	18.4
Negative	53	36.1
Not available	67	45.6
PD-L1
High	9	6.1
Low	9	6.1
Not available	129	87.8
Median NLR (range)	2.75 (0.71–9.22)
Recurrence
Yes	31	21.1
No	116	78.9
PNAd
−	61	41.5
+ −	40	27.2
+	33	22.4
+  +	13	8.8

NLR Neutrophil to lymphocyte ratio; PNAd Peripheral node addressin

Table 2.

Differences in characteristics between TLS low-expression and high-expression patients

Characteristic	TLS		P
	Low	High
Age, y (median 70)			0.286
< 70	46	26
≥ 70	55	20
Sex			0.529
Male	69	29
Female	32	17
Histology			0.790
Adenocarcinoma	73	34
Squamous cell carcinoma	27	12
Others	1	0
Smoking, pack-years			0.443
< 30	48	25
≥ 30	53	21
Tumor diameter (cm)			0.867
< 2.6	49	23
≥ 2.6	52	23
Pathological T status			0.432
T1a	35	12
T1b	22	13
T2a	30	17
T2b	4	0
T3	8	2
T4	2	2
Pathological N status			0.274
N0	83	39
N1	7	4
N2	11	2
N3	0	1
Pathological stage			0.951
IA	54	23
IB	22	13
IIA	5	3
IIB	7	2
IIIA	10	4
IIIB	2	1
IV	1	0
EGFR mutation			0.629
Positive	18	9
Negative	39	14
Not available	44	23
PD-L1			0.403
High	8	1
Low	6	3
none	87	42
NLR (median 2.75)			0.003
< 2.75	57	38
≥ 2.75	44	8
Recurrence			0.281
Yes	24	7
No	77	39

TLS Tertiary lymphoid structure, NLR Neutrophil to lymphocyte ratio

TLS and survival. In all patients, there was no significant association between recurrence and the expression of TLS (p = 0.105). The Cox proportional hazards analysis showed that stage was the only independent factor associated with DFS for all patients (p < 0.001, HR 0.126, 95% CI 0.060–0.267). Stage I patients accounted for 76.1% of all cases, and each patient in stage II-IV accounted for a small percentage. In order to examine the effect of TLS expression on survival, the patients were divided into an early (stage I) group and an advanced (stage II-IV) group. The association between recurrence and TLS expression was then evaluated (Table 3, Fig. 2). Although there were no significant associations between recurrence and the expression of TLS in the early stage (p = 0.58) and in the advanced stage (p = 0.074), there was a tendency for the low TLS expression group to have more recurrence in advanced stage lung cancer. The DFS curve showed no significant difference in TLS expression in the stage I group, but there was a marked difference (p = 0.027) in the stage II-IV group, and the prognosis was good in the TLS high-expression group. On multivariate analysis, TLS was the only independent factor associated with DFS in the stage II-IV group (p = 0.040, HR 0.272, 95% CI 0.078–0.945).

Table 3.

Association between recurrence and TLS

Event	TLS		P
	Low	High
Stage I			0.58
Recurrence −	68	32
Recurrence +	8	4
Stage II-IV			0.074
Recurrence −	9	7
Recurrence +	16	3

TLS Tertiary lymphoid structure

Fig. 2.

Disease-free survival and TLS expression. Postoperative disease-free survival does not differ by TLS expression in stage I NSCLC (a, p = 0.651), but it is significantly better in the TLS high-expression group in stage II-IV NSCLC (b, p = 0.027)

viSNE and citrus analysis to identify the distinct features of PBMCs between TLS low-expression and high-expression groups. Live intact single cells in PBMCs could be clearly divided into subsets on the viSNE map (Fig. 3). Citrus analysis was then used to identify distinct subpopulations between the TLS low-expression and high-expression groups. A predictive model was produced using all PBMC events (Fig. 4). A Nearest Shrunken Centroid (PAMR) model found a distinguishing cluster between the TLS low-expression and high-expression groups (Fig. 4b). Abundance of the cluster was lower in the TLS high-expression group than in the TLS low-expression group (Fig. 4c). The location of the cells in this cluster was drawn in color on the viSNE map (Fig. 4d). This cluster was found to include CD3⁻, CD4⁻, CD8a⁻, CD25⁻, CD9⁺, and HLA-DR⁺ cells (Fig. 4e).

Fig. 3.

viSNE maps in distinct immune subsets of peripheral blood mononuclear cells. CyTOF viSNE plots of major immune cell populations and expressions of several proteins are shown in a characteristic patient

Fig. 4.

Citrus analysis to identify the cell cluster that correlates with TLS expression. a Citrus clustering of immune subsets from CyTOF data of peripheral blood mononuclear cells is shown. The heat spectrum in each graph indicates the expression of each marker in a cluster. b A Nearest Shrunken Centroid (PAMR) model shows the cluster to distinguish patients with high TLS expression from those with low TLS expression. The distinguishing cluster is highlighted in red. c The abundance of cells within the identified distinguishing cluster is lower in patients with high TLS expression. d CyTOF viSNE plots of the identified distinguishing cell cluster are drawn in color. e Metal signal intensities of indicated markers are shown in the histogram. Intensities of the identified distinguishing cluster are drawn in red, and those of the background are drawn in blue. CD3, CD4, and CD8a are negative, and CD9 and HLA-DR are positive in the identified distinguishing cluster

Discussion

This study examined the associations between TLS and not only survival, but also patients’ background characteristics, including peripheral blood features, in patients with NSCLC. Three points were identified for the first time in this study. The first is that high TLS expression was associated with a good prognosis in stage II-IV NSCLC. The second is that the expression of TLS was significantly lower in the group with high NLR than in the group with low NLR in preoperative peripheral blood samples. Third, the abundance of the cell cluster expressing CD9 and HLA-DR without expression of CD3, CD4, and CD8 was lower in PBMCs in cases with high TLS expression.

It has been reported that the presence of TLS can be associated with a good prognosis in various malignant tumors such as breast cancer [24], colorectal cancer [25], lung cancer [19, 26], malignant melanoma [27, 28], and liver cancer [29]. Rakaee et al. reported that the number of TLSs was lower in stage III patients than in stage II patients with NSCLC, but they concluded that TLS is an independent factor associated with a good prognosis [19]. In the present study, TLS was not associated with NSCLC stage, but it was associated with a good prognosis in stage II-IV NSCLC. This result was consistent with previous reports. A larger study cohort may be needed to analyze recurrence-free survival in stage I NSCLC, because the five-year recurrence-free survival rate was over 80%.

On the other hand, the correlated clinical factors, especially peripheral blood characteristics, have not been elucidated in previous reports. The present study showed the peripheral blood characteristics correlated with TLS expression for the first time. These were NLR and the cell cluster expressing CD9 and HLA-DR. With regard to NLR, it may be possible to assume the TLS status in the tumor microenvironment by evaluating the leukocyte fraction using peripheral blood samples. It has already been shown that high NLR is associated with a poor prognosis in NSCLC [30, 31]. This study linked the NLR and TLS, which suggests that these prognostic factors may be due to a series of mechanisms. As for the cell cluster that was identified, we should focus on CD9 and HLA-DR. CD9 is expressed on monocytes and macrophages, although it is attracting attention as a marker of extracellular vesicles. CD9 was first identified on the human lymphohematopoietic progenitor cell [32]. CD9 expression was found on various cells such as megakaryocytes, platelets, B and T lymphocytes, dendritic cells, mast cells, eosinophils, and basophils [33–35]. HLA-DR is an isotype of HLA-class II molecules expressed on antigen-presenting cells including dendritic cells, B cells, monocytes, and macrophages. The cell cluster was not T cells based on the present results. Therefore, the cell cluster is assumed to consist of antigen-presenting cells. This result suggests that antigen-presenting cells in peripheral blood may be recruited into the tumor microenvironment in patients with high TLS expression.

There have also been attempts to induce TLS for treatment [36]. It has been reported that the combined use of anti-angiogenic therapy and anti-PD-L1 therapy increased HEV formation and TLS formation in experimental models of breast cancer and neuroendocrine pancreatic tumors [37]. It is already known that cholesterol clefts, foreign body-reactive giant cells, and necrosis, which are characteristic of cell death, can be seen in excised specimens of patients undergoing conventional preoperative chemotherapy [38]. Clinical trials of surgical therapy after ICI combination therapy as preoperative treatment are underway, and these studies have performed pathological analyses after ICI administration [20]. Cottrel et al. reported that TLS in the tumor region was characteristically observed along with cell death in cases showing a good response to preoperative treatment with ICIs. On the other hand, nonspecific collection of TILs unrelated to the treatment response was observed [20]. In other words, it is suggested that TLS, not TILs, may be a better predictor of the antitumor effect.

Although TLS is composed of diverse elements, the presence or absence of mature dendritic cells and HEV is important in differentiating TLS from TILs. There are some reports of immunohistochemical staining and identification of mature dendritic cells [25–27, 39, 40] and reports of HEV staining and identification using PNAd [24]. In the present study, TLS was identified based on the presence of HEVs, which are morphologically easy to recognize. However, a large sample of tumor tissue is needed for TLS evaluation. It is thought that it would be of great clinical benefit if we could predict the status of TLS from peripheral blood samples.

The limitations of the study are the low number of cases and the retrospective evaluation in a single-center. Since TLS was identified based on the presence of HEVs in this study, a detailed evaluation of the constituent cells may show further trends. Another limitation is that the panel used in CyTOF analysis was focused on T cells. When this study was being designed, we hypothesized that the T cell subpopulation in peripheral blood may differ according to TLS expression in the tumor microenvironment. However, the results showed that the distinguishing cluster in peripheral blood did not consist of T cells. Further study is needed to validate the present result and to identify the detailed features of the cell cluster.

In conclusion, high expression of TLS was associated with a good prognosis in surgically resected stage II to IV NSCLC patients. In the peripheral blood, low NLR was correlated with the presence of TLS in the tumor microenvironment. Antigen-presenting cells in the peripheral blood may be recruited to the tumor microenvironment in NSCLC patients with TLS. It may be possible to infer the expression of TLS from peripheral blood samples even in cases whose tissue is difficult to obtain.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

HEV

High endothelial venule

ICI

Immune checkpoint inhibitor

NLR

Neutrophil to lymphocyte ratio

NSCLC

Non-small cell lung cancer

PD-L1

Programmed cell death-ligand 1

PNAd

Peripheral node addressin

TIL

Tumor-infiltrating lymphocyte

TLS

Tertiary lymphoid structure

Author contributions

MF, SM and HS contributed to the study conception and design. Material preparation and data collection were performed by MF, SM, SI, HY, HM, HT, YO, MW, TI, TY, NO, YM, TH, JO, MH, MH and YS. Data analysis was performed by MF and SM. The first draft of the manuscript was written by MF. Review and editing were performed by SM and HS. All authors read and approved the final manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or non-for-profit sectors.

Declarations

Conflict of interest

The authors declare that they have no competing interests.

Consent to participate

Written, informed consent was obtained from all participants.

Ethical approval

This study was approved by the Human Ethics Committee at Fukushima Medical University (approval no. 30161).