Clinicopathological significance of peritumoral alveolar macrophages in patients with resected early-stage lung squamous cell carcinoma

This study aimed to clarify the correlation between the number of AMs and prognosis and to examine the gene expression of AMs in lung squamous cell carcinoma (SqCC). We reviewed 124 stage I lung SqCC cases in our hospital and 139 stage I lung SqCC cases in The Cancer Genome Atlas (TCGA) cohort in this study. We counted the number of AMs in the peritumoral lung field (P-AMs) and in the lung field distant from the tumor (D-AMs). Moreover, we performed a novel ex vivo bronchoalveolar lavage fluid (BALF) analysis to select AMs from surgically resected lung SqCC cases and examined the expression of IL10, CCL2, IL6, TGFβ, and TNFα (n = 3). Patients with high P-AMs had significantly shorter overall survival (OS) (p < 0.01); however, patients with high D-AMs did not have significantly shorter OS. Moreover, in TCGA cohort, patients with high P-AMs had a significantly shorter OS (p < 0.01). In multivariate analysis, a higher number of P-AMs were an independent poor prognostic factor (p = 0.02). Ex vivo BALF analysis revealed that AMs collected from the tumor vicinity showed higher expression of IL10 and CCL2 than AMs from distant lung fields in all 3 cases (IL-10: 2.2-, 3.0-, and 10.0-fold; CCL-2: 3.0-, 3.1-, and 3.2-fold). Moreover, the addition of recombinant CCL2 significantly increased the proliferation of RERF-LC-AI, a lung SqCC cell line. The current results indicated the prognostic impact of the number of peritumoral AMs and suggested the importance of the peritumoral tumor microenvironment in lung SqCC progression.


Introduction
Lung cancer accounts for the highest number of cancerrelated deaths in the world [1]. The standard therapy for stage I lung cancer is surgery [2]. However, some lung cancer patients show recurrence after surgery, and the 5-year survival rate after surgery in stage I non-small cell lung cancer (NSCLC) is 57.1-73.5% [3].
The tumor microenvironment (TME) is one of the key determinants of cancer development, progression, and metastasis in many types of cancer, including lung cancer [4,5]. The TME that focuses on immune cells is called tumor immune microenvironment (TIME) [6]. Within the TIME, tumor-associated macrophages (TAMs) have pro-tumoral functions, such as angiogenesis, immune suppression, cancer cell proliferation, invasion, and metastasis, via several paracrine mechanisms [7][8][9]. For example, chemokine (C-C motif) ligand (CCL) 2-chemokine (C-C motif) receptor (CCR) 2 promotes the recruitment of monocytes to primary and metastatic tumors [9]. Interleukin (IL)-10 polarizes macrophages to M2 phenotype and mediates the expression of programmed cell death ligand-1 in monocytes to inhibit the cytotoxic T cell response and promote immune suppression to mediate regulatory T cells [10]. Hence, TAMs are an important subject of study, and some drugs targeting TAMs are currently undergoing clinical trials [9].
The major role of AMs in the alveoli is to maintain lung homeostasis [11,12]. In a mouse model after NSCLC seeding, AMs in peritumoral areas showed higher expression of several genes, including chemokines, than AMs in the contralateral lung field in early tumors. Moreover, the co-culture of AMs in peritumoral areas and mouse NSCLC spheroids significantly increased cancer cell invasion, suggesting that AMs contribute to tumor progression [13]. Recently, we reported that a higher number of AMs around adenocarcinoma were associated with poor prognosis [14]. In lung metastasis of hepatocellular carcinoma (HCC), AMs are recruited from circulating monocytes by CCL2-CCR2, unlike normal lungs, and AMs themselves can contribute to the formation of TIME to support lung metastasis and progression [15]. AMs are defined as those that present outside the tumor area and self-renew in alveolar, while TAMs are defined as those that present inside the tumor and are recruited from circulating monocyte.
This study aimed to determine whether the number of AMs is a prognostic factor for lung SqCC and to elucidate its molecular mechanism. AMs are present in both peritumoral and distant areas from the tumor. We first counted the number of peritumoral (P-) and distant (D-) AMs and determined whether they were prognostic factors for pathological stage I lung SqCC. Then, we collected AMs from the tumor vicinity bronchoalveolar lavage (BAL) fluid (BALF) (AMs-VBALF) and AMs from distant lung field BALF (AMs-DBALF) using resected lungs and compared the gene expression between them.

Patients
At our hospital, 241 consecutive patients with pathological stage I lung SqCC underwent surgical resection between January 2011 and December 2017. Furthermore, we excluded 117 patients based on the following criteria: (a) central type SqCC (displaced segmental bronchus due to tumor observed on preoperative imaging), (b) underwent sublobar resection, (c) incomplete resection, (d) neoadjuvant therapy, (e) no comprehensive consent, and (f) perioperative death. Finally, 124 patients were enrolled in the study (Supplemental Figure.1A). This study was approved by the institutional review board (IRB approval number; 2020-446).
We validated 495 lung SqCC patients (The Cancer Genome Atlas [TCGA], Firehose Legacy). Furthermore, we included 139 patients based on the following criteria: (a) no stage I lung SqCC, (b) no available virtual slide, and (c) no peritumoral alveolar space on the virtual slide (Supplemental Figure 1B).

Definition and evaluation of P-AMs and D-AMs
Typical AMs in smoker have brown pigment in cytoplasm, sometimes foamy cytoplasm without pigmentation in nonsmoker. Nuclei of AMs are small, uniform, and regular without atypia [16]. Peritumoral alveolar areas were defined within 2 mm of the tumor edges, and we randomly chose 9 square spots by loupe magnification. In the TCGA cohort, we defined peritumoral alveolar areas as adjacent to the tumor and randomly chose 3 square spots by loupe magnification because the virtual slide in TCGA had less peritumoral alveolar area than our hospital slide. We defined the most distant area from the tumor as the distant alveolar area and randomly chose 9 square spots by loupe magnification. The number of P-AMs and D-AMs was defined as the total number of AMs in square spots in peritumoral and distant alveolar areas, respectively, divided by the total area of square spots (/mm 2 ) ( Fig. 1A-F). We investigated correlations between the number of P-AMs in HE-stained, CD68 (KP-1; Roche Diagnostics, Switzerland)-positive AMs, and CD204 (Scavenger receptor class A-E5, Transgenic, Japan)-positive AMs in 10 lung squamous cell carcinoma cases. The results are shown in Supplemental Figure 2. There are very strong positive correlations between HE slides and CD204 slides (R = 0.947, p = 0.001), between HE slides and CD68 slides (R = 0.912, p = 0.001), and between CD68 slides and CD204 slides (R = 0.876, p = 0.001)). Based on these results, we determined that the number of AMs could be counted on HE slides and evaluated AM counts on HE slides only. We defined AMs counts higher than the median as high P-AMs and high D-AMs in the respective cohort.

Evaluation of clinicopathological factors
The clinical characteristics of patients were reviewed from the available medical records. The following clinicopathological factors were investigated retrospectively to assess their impact on patient survival: age, sex, smoking history, interstitial pneumonitis (IP), pathological stage, vessel invasion, lymphatic permeation, and pleural invasion (pl).
Histological diagnosis was based on the 5th edition of the World Health Organization histological classification [18], and the disease stages were categorized according to the guidelines of the 8th edition of the TNM classification [19].

Ex vivo BALF analysis
We performed ex vivo BALF analysis in three lung SqCC patients who underwent lobectomy between December 2020 and January 2021 (Supplemental Table 1). First, we injected saline into the sub-segmental bronchus and collected the saline using a 50-mL syringe; we repeated these procedures until returned volume reached 200 mL for each distant bronchus from the tumor (distant BALF) and the responsible bronchus of the tumor (tumor vicinity BALF) (Supplemental Figure 3A-B). Next, we excluded debris through 40-µm cells strainers ® (Corning Japan, Japan) and separated the pellet from the ex vivo BALF by centrifugation at 300 × g for 15 min. Then, we added phosphate-buffered saline (PBS) to the pellet to make a total of 16 mL fluid, injected the fluid into a BD vacutainer ® CPT (Becton, Dickinson and Company, USA), and separated white blood cell components by centrifugation at 1800 × g for 20 min. Subsequently, we collected the buffy coat in BD vacutainer ® CPT, added PBS to make a total of 40 mL fluid, and separated the pellet by centrifugation at 300 g × g for 15 min. Finally, we added 2 mL BD Pharm Lyse ® (Becton, Dickinson and Company, USA) to the pellet for 15 min at 15-25 °C, added PBS to make a total of 40 mL fluid, and separated the pellet by centrifugation at 200 × g for 5 min (Supplemental Figure 3C).

Real-time reverse transcription-polymerase chain reaction
We extracted RNA from the samples using NucleoSpin RNA Plus ® (Takara Bio Inc., Shiga, Japan). Using Prime-Script RT reagent Kit ® , 0.1 µg of total RNA was reverse transcribed (Takara Bio Inc., Shiga, Japan), and the cDNA generated was amplified using Thermal Cycler Dice ® (Takara Bio Inc., Shiga, Japan). The target gene expression was normalized to the GAPDH expression in the same sample. We used primers for IL10, CCL2, IL6, TGFβ, and TNFα (Supplemental Table 2).

TCGA analysis
We analyzed lung SqCC gene expression signatures using the TCGA dataset. All statistical analyses were performed using R (version 4.0.5) (https:// www.r-proje ct. org/). The R/Bioconductor package TCGA biolinks (version 2.16.1) [21] was used to download the TCGA data from the NCI's Genomic Data Commons, and Summarized Experiment (version 1.18.2) [22] was used to store the TCGA data. We collected TCGA RNA-seq V2 data of cases based on the American Joint Committee on Cancer stage I lung SqCC. We included 198 cases, and 59 cases were excluded because we could not evaluate P-AMs. A total of 139 cases were analyzed in this study (Supplemental Figure 1B). We evaluated P-AMs using the digital histologic slides of each case published in TCGA. Differentially expressed genes (DEG) between cases with and without necrosis were identified using the DESeq2 package (version 1.  [26].
We also evaluated differences in gene expression levels of the receptor and ligand pathways of IL10 and CCL2 between groups with high P-AMs and low P-AMs using the Mann-Whitney U test and showed log2 FC of high/ low P-AMs in a graph.

RERF-LC-AI cells (RIKEN BRC CELL BANK, Japan)
were seeded in a 96-well plate (Life Technologies, Carlsbad, CA) at a concentration of 1000 cells/well for 24 h, and the medium was removed and replaced with MEM alpha (Life Technologies, Carlsbad, CA) containing recombinant CCL2 (ProteinSimple, Inc., Japan) or IL10 (Proteintech Japan, Inc., Japan) (10 ng/mL). Cell proliferation was monitored by analyzing the occupied area (% confluence) of cell images after 72 h using Incu Cyte ZOOM ® (Essen BioScience, Inc., Michigan, USA).

Statistical analysis
Overall survival (OS) and recurrence-free survival (RFS) curves were plotted according to the Kaplan-Meier method and compared using the log-rank test in a univariate analysis. Univariate and multivariate analyses were conducted to identify the predictors of RFS using the Cox proportional hazards model. Two category comparisons were performed using one of the following tests: Chi-square test, Fisher's  exact test, Mann-Whitney U test, or paired t test. Statistical significance was set at p < 0.05.

P-AMs and D-AMs
P-AM and D-AM counts for each patient are shown in Fig. 2A. The median values of P-AMs and D-AMs were 26.7/mm 2 and 11.1/mm 2 , respectively. The number of P-AMs was significantly higher than that of D-AMs (P-AMs/mm 2 vs. D-AMs/mm 2 , p < 0.01, Fig. 2B). There was no correlation between the number of P-AMs and D-AMs (r = 0.31) (Fig. 2C).

Characteristics of patients with high P-AMs
We defined high P-AMs (> 26.7/mm 2 ), low P-AMs (< 26.7/mm 2 ), high D-AMs (> 11.1/mm 2 ), and low D-AMs (< 11.1/mm 2 ). Compared with the low P-AMs group, the high P-AMs group was associated with T2 (p < 0.01), pl (p < 0.01), and vessel invasion (p = 0.04) ( Table 1). There was no significant difference in IP between the high and low P-AMs groups (p = 0.08). On the contrary, the high D-AMs group had significantly more patients with T2 than the low D-AMs group (p = 0.02) (Supplementary Table 3).
We validated the relationship between high P-AMs and lung SqCC prognosis in the TCGA cohort and in 139 stage I lung SqCC patients, and we obtained similar results (5-year OS rate: high P-AMs, 42.2% vs. low P-AMs, 74.6%; HR, 0.40; p = 0.01, Fig. 3E).
In the TCGA cohort of 495 all-stage SqCC patients, the high CCL2 expression group had significantly shorter OS than the low CCL2 expression group, but there was no significant difference between high and low IL10 expression (CCL2: HR, 0.69; p < 0.01; IL10: HR, 0.82; p = 0.15, Supplemental Figure 5A, B).

Discussion
In this study, we found that a higher number of P-AMs was an independent poor prognostic factor for stage I lung SqCC on multivariate analysis. Moreover, CCL2 and IL10 expressions in AMs-VBALF were much higher than those in AMs-DBALF, and CCL2 increased tumor cell proliferation. This is the first study to show that P-AMs may act in a tumorpromoting manner during lung SqCC progression.
Using two independent cohorts, we confirmed that a high number of P-AMs is a poor prognostic factor. In a recent study, the distribution of pSTAT3-positive tumor cells in small cell lung cancer (SCLC) was found to be predominant in the periphery of tumor nests, and IL-6 and CCL4/MIP-1β from macrophages activated STAT3 in SCLC, suggesting that TIME around the tumor is important for tumor progression in SCLC [27]. Recently, Noritake et al. [16] reported that a high number of P-AMs were independently associated with poor prognosis in resected early lung adenocarcinoma. They found that the mRNA expression involved in chemotaxis and epithelial proliferation was enriched in patients with high P-AMs, suggesting that P-AMs may influence the peritumoral microenvironment. Our current results also indicate that P-AMs are a poor prognostic factor in early-stage SqCC, suggesting that a higher number of P-AMs could be an unfavorable prognostic marker in early-stage SqCC as well as in early-stage adenocarcinoma and SCLC.
Recently, María et al. reported that AMs in the peritumoral region of the lung differed in gene expression compared with AMs in the contralateral lung in a mouse model. Fifteen days after NSCLC seeding, AMs in the peritumoral region displayed higher expression of peptidases (Mmp12, Mmp14, and Adamdec1), integrin-binding molecules (Tspan4), major histocompatibility complex class II molecules (H2-M2, H2-AA, H2-AB1, and H2-Q7), and T cell chemoattractants (Ccl17 and Cxcl9) [15]. In the lung metastasis of HCC, CCR2-CCL2 was involved in the recruitment of P-AMs [17]. In the present study, using human samples, CCL2 levels were found to be elevated in AMs-VBALF in lung SqCC. Moreover, recombinant CCL2 significantly increased the proliferation of RERF-LC-AI cells. In addition, using the database of TCGA cohort, we found that the high CCL2 expression group had a significantly poorer prognosis than the low expression group (Supplemental Figure 5A). These findings suggest that AM-derived CCL2 in the peritumoral microenvironment may be involved in tumor cell progression.
We attempted to identify the gene ontology associated with macrophage recruitment between patient groups with high and low P-AMs. However, DEG gene ontology enrichment analysis from Z-score clustering analysis between high and low P-AMs did not show any gene ontology related to AMs (Supplemental Figure 6A, B). Therefore, we specifically investigated the receptors associated with CCL2 and  IL10, which are highly expressed in AMs-VBALF. We found that CCR1, CCR2, CCR3, CCR4, and CCR10 were significantly expressed in the high P-AMs group (Supplemental Figure 6C). The CCL2-CCR2 axis between tumor cells and TAMs reportedly modifies TIME and amplifies tumor cell invasion [28]. These findings suggest that the positive feedback of the CCL2-CCR2 axis might contribute to the formation of the pro-tumoral microenvironment in the peritumoral area, similar to that in the intratumoral area.
In conclusion, we found that a high number of P-AMs was associated with significantly poorer prognosis than a low number of P-AMs in stage I postoperative lung SqCC patients. Moreover, AMs-VBALF-derived CCL2 promoted the proliferation of SqCC cell lines. These findings suggest that P-AMs are involved in tumor progression in early-stage squamous cell lung cancer. In our study, P-AMs may be M2-like macrophage because P-AMs expresses higher levels of CCL2 and IL-10 [29,30]. In recent years, several clinical trials targeting tumor-associated macrophages (TAMs), such as CCR2 and CSF-1R, have been ongoing [31]. There may be a need to target not only TAMs but also P-AMs as a possible future treatment.