Prognostic Impact of the Number of Peritumoral Alveolar Macrophages in Patients With Stage I Lung Adenocarcinoma

Osamu Noritake National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Keiju Aokage National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Ayako Suzuki Tokyo Daigaku Kashiwa Campus Kenta Tane National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Tomohiro Miyoshi National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Joji Samejima National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Toyohumi Yoshikawa Nagoya University Graduate School of Medicine Faculty of Medicine: Nagoya Daigaku Daigakuin Igakukei Kenkyuka Igakubu Tokiko Nakai National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Masahiro Tsuboi National Cancer Center-Hospital East: Kokuritsu Gan Center Higashi Byoin Genichiro Ishii (  gishii@east.ncc.go.jp ) National Cancer Center Hospital East https://orcid.org/0000-0003-4158-0286

Additionally, some studies have reported that the peritumoral microenvironment in uences tumor progression. In pancreatic cancer, the expression of secreted protein acidic and rich in cysteine (SPARC) by peritumoral broblasts indicates a poorer prognosis for patients (Infante et al. 2007). Moreover, hyaluronan in peritumoral stroma is associated with cancer cell spreading in breast cancer (Auvinen et al. 2000). Peritumoral macrophages promote tissue remodeling and proangiogenic pathways in hepatocellular carcinoma, by inducing Th17-mediated in ammation in peritumoral stroma (Kuang et al. 2010). For small cell lung cancer, Iriki reported that the macrophages that accumulated near the peripheral areas of tumor nests are important for tumor progression via STAT3 activation, suggesting that peritumoral macrophages may be deeply involved in tumor progression (Iriki et al. 2017). Based on these ndings, we hypothesized that peritumoral AMs are associated with the prognosis of lung cancer. Thus, this study aims to investigate the prognostic impact of the number of peritumoral AMs in patients with stage I lung adenocarcinoma.

Patients
Patients with pathological stage I adenocarcinoma who underwent lobectomy or pneumonectomy with systemic lymph node dissection at the National Cancer Center East between January 2011 and December 2015 were enrolled in this study ( Supplementary Fig. 1). We excluded patients who received preoperative therapy and those with adenocarcinoma in situ, undetectable tumors in surgical specimens, and synchronous lung cancer. Finally, 514 patients were enrolled in this study. The study was approved by the institutional review board of the National Cancer Center Hospital East (IRB approval number 2020 − 239).
The clinical characteristics of patients were retrospectively retrieved from the Thoracic Surgical Database of the National Cancer Center Hospital East.

Pathologic evaluation
All surgical specimens were xed in 10% formalin, and embedded in para n. All tumors were cut at 5-mm intervals, and 4µm thick-sections were stained using the hematoxylin and eosin (HE) method. The Verhoeff-van-Gieson (VVG) method was used to visualize elastic bers to identify lymphovascular and pleural invasion. Histological diagnosis was based on the fourth edition of the World Health Organization series, and the disease stages were categorized according to the guidelines of the 8th edition of the TNM Classi cation of Malignant Tumors.

Calculation of the number of alveolar macrophages
We selected the largest cross-sectional slide for each patient to calculate the number of AMs (Figs. 1A-D). If the tumor was on multiple slides in the largest cross-section, we counted all these slides. The HE slides were scanned and captured using a digital slide scanner (Aperio VERSA SL200; Leica, Biosystems, Nußloch, Germany) and were reviewed by two pathologists (O.N and G.I) who were blinded to the clinicopathologic information of each slide. The peritumoral alveolar space was de ned as the air space outside the tumor within three alveoli, and the number of macrophages per alveolar space was counted. AMs were distinguished from spread through air spaces (STAS) of tumor cells using the following methods. Macrophages in smokers typically have cytoplasm containing faint brown pigment and black carbon granules, whereas in nonsmokers the pigment is lacking and the cytoplasm is sometimes foamy. Nuclei are small, uniform, and regular, without atypia. Nuclear folds are frequent, and nucleoli are inconspicuous or absent. Contrastingly, STAS generally lack cytoplasmic pigment or foamy cytoplasm. They often grow in cohesive clusters and nuclei are atypical with hyperchromasia and frequent nucleoli. The number of peritumoral AMs was determined as the average of the top ve AMs in each patient.
We investigated the correlation between the number of peritumoral AMs in HE-stained and peritumoral CD68-positive AMs. Forty patients were examined for CD68 expression in the peritumoral alveolar space Figs. 2A, B). The distribution of peritumoral AMs on HE slides and CD68-positive macrophages in each patient strongly correlated (p < 0.01, r² = 0.91) (Fig. 1E). Therefore, we assessed that HE slides were e cient for counting the number of peritumoral AMs, and later the number of peritumoral AMs was evaluated using HE slides only.
Gene expression signature analysis on the Cancer Genome Atlas data sets We analyzed the difference in gene signatures between the number of AMs in lung adenocarcinoma, using the Cancer Genome Atlas-Lung Adenocarcinoma (TCGA-LUAD) datasets. All analyses were performed using R software (version 4.0.2) (https://www.r-project.org/; The R Foundation for Statistical Computing, Vienna, Austria). We downloaded the data of 155 patients with stage I lung adenocarcinoma in June 2020 from the Genomic Data Common (GDC) Data Portal (https://portal.gdc.cancer.gov) using the R/Bioconductor package TCGA biolinks. Using representative digital slides of each patient that can be viewed on the site, we counted and calculated peritumoral AMs and divided them into two groups in the manner described above. Twenty-ve patients were excluded because of di culty in evaluating peritumoral AMs. Differentially expressed genes between the high and low AM groups were identi ed using the DESeq2 package (version 1.28.1) [Love MI et al]. Genes were extracted using adjusted p-values (padj) of < 0.05 and |log2 fold changes (log2FC)| of > 1. We performed hierarchical clustering of genes that differed in expression between the high and low AM groups (|log2FC|>1), and heatmap analysis was

Statistical analysis
Associations between the clinicopathologic variables and the number of peritumoral AMs were analyzed using Fisher's exact test (for categorical variables) and the Mann-Whitney U test (for continuous variables). The number of peritumoral AMs in each predominant subtype group was compared using the Mann-Whitney U test. Pearson correlation coe cient was used to evaluate the correlation between the number of peritumoral AMs on HE slides and the number of CD68-positive peritumoral AMs. To evaluate prognosis, receiver operating characteristic curves of the number of peritumoral AMs were used. Overall survival (OS) was de ned as the interval between the date of surgery and the date of death from any cause or the last follow-up. Disease-free survival (DFS) was de ned as the time between the date of surgery and the date of recurrence detection, death from any cause, or the last follow-up. Survival curves were estimated using the Kaplan-Meier method, and differences in OS and DFS were compared by using the log-rank test. Hazard ratios (HRs) were estimated using the Cox proportional hazards model. The date of data cut-off was May 2019 at our institution. These data were analyzed using EZR version 1.51, a graphical user interface for R program.

Correlation between clinicopathological factors and the number of peritumoral AMs
The median number of peritumoral AMs was 15.5 per alveolar space, and we divided patients into two groups based on this value. Patients with a higher number of peritumoral AMs were men, with smoking history; high-grade pathologic T stage; larger pathologic total and invasive tumor sizes; the presence of pleural invasion, lymph vessel invasion and vascular invasion; and a higher proportion of invasive predominant subtypes (Table 1). Correlation between the numbers of peritumoral AMs and the predominant subtype The relationships between the number of peritumoral AMs and predominant subtypes are presented in Fig. 2, wherein the micropapillary-predominant subtype was excluded because of one case. The median number of peritumoral AMs for the predominant subtype was 9.6 in lepidic, 20.4 in papillary, 20.2 in acinar, 27.2 in solid subtypes. The number of peritumoral AMs in the predominantly solid subtype was signi cantly higher than that in the acinar subtype (p < 0.01), papillary subtype (p < 0.01), and lepidic subtype (p < 0.01). There were no signi cant differences between the number of peritumoral AMs in the predominantly acinar and papillary subtypes (p = 0.78), and the number of peritumoral AMs in these two subtypes was signi cantly higher than that in the lepidic subtype (both p < 0.01). In the univariate analysis, a larger invasive tumor size; the presence of pleural invasion, lymphatic permeation, presence of vascular invasion; and a higher number of peritumoral AMs were signi cantly associated with shorter DFS (Table 2). In the multivariate analysis, a higher number of peritumoral AMs was an independent prognostic factor for DFS (HR 1.85, 95% con dence interval (CI): 1.11-3.09, p = 0.02), in addition to invasive tumor size and vascular invasion.

Evaluation of the survival impact of the number of peritumoral AMs
Patients were strati ed into two groups according to the different cut-off values of the number of peritumoral AMs. Multivariate analyses showed that although the cut-off value was 10, there was no signi cant difference between patients with high and low peritumoral AMs. Both the cut-off values of 20 and 30 were independent factors for DFS (HR 2.01, 95% con dence interval: 1.28-3.15, p < 0.01; HR 3.27, 95% CI: 2.10-5.08, p < 0.01, respectively), and the hazard ratio gradually increased as the cut-off value increased (Table 3). This result suggested the possibility that the higher the number of peritumoral AMs, the more likely the tumor would recur. Survival and gene ontology analysis by the number of peritumoral AMs in TCGA database Of the 130 patients, the median number of peritumoral AMs that we calculated on TCGA digital slides was 11.3, and we divided them into two groups using this value. Patients with a higher number of peritumoral AMs showed signi cantly shorter DFS (p = 0.04) (Fig. 4A), whereas there was no signi cant difference in OS (data not shown).
We extracted 212 differentially expressed genes (DEGs) between patients with high and low peritumoral AMs. A heatmap of 212 DEGs is presented in Fig. 4B. The Metascape pathway enrichment analysis was performed, which identi ed the top 20 biological processes in the DEGs, including chemotaxis (GO: 0006935) and epithelial proliferation (GO: 0050673) (Fig. 4C).

Discussion
In our study, clinicopathological factors showed that patients with a high peritumoral AM content had more aggressive tumors, such as with larger pathologic invasive tumor size; the presence of pleural invasion, lymphatic permeation, and a higher proportion of more invasive predominant subtype, than patients with a low peritumoral AM content. The multivariate analysis revealed that the number of peritumoral AMs was an independent prognostic factor. These results suggest that the number of peritumoral AMs is a strong factor for poor prognosis in patients with stage I lung adenocarcinoma. To the best of our knowledge, this is the rst report showing that the number of AMs is associated with the prognosis of stage I lung adenocarcinoma.
In the analysis using the TCGA dataset, patients with a high peritumoral AM content showed shorter DFS than those with a low peritumoral AM content, indicating that the results using our hospital cohort are reproducible. The signi cant difference in some GO pathways, such as chemotaxis and epithelial proliferation, might be one of the reasons why the prognosis was poorer in the group with a high peritumoral AM content. We suggest two possible reasons why the number of peritumoral AMs differs depending on the patient: (1) Cancer cells themselves mobilize the peritumoral AMs, and the ability to attract AMs varies depending on the characteristics of cancer cells in each case. (2) AMs that are originally present in alveoli before tumorigenesis and the characteristics of AMs in each case may in uence the development of adenocarcinoma. To verify our theory, we made the following considerations: We de ned nonadjacent alveolar space as the air space 2 mm away from the tumor edge on the largest cross-sectional slide, to avoid overlapping with the peritumoral alveolar space, and counted macrophages per alveolar space ( Supplementary Figs. 3A, B). There was a strong correlation between peritumoral and nonadjacent AMs (r² = 0.723; Supplementary Fig. 3C). The median number of nonadjacent AMs was 8.6 per alveolar space.
In the multivariate analysis, there were no signi cant differences between patients with a high and low non-adjacent AM content with the cut-off values of 8.6 and 10 (p = 0.54 and 0.50, respectively) (Supplementary Tables 1, 2

Declarations
Funding: This study was supported in part by the National Cancer Center Research and Development Fund (31-A-6).

Con icts of interest:
The authors declare that they have no con ict of interest.
Availability of data and material: There are no availability of data and material.