Transforming Growth Factor‐β1 as a Biomarker of Malignant Behavior and PD‐L1 Expression in Thymic Epithelial Tumors

ABSTRACT

The clinical significance and molecular mechanisms underlying tumor progression in thymic epithelial tumors (TETs) remain largely unclear. In this retrospective single‐center study, we evaluated the prognostic value of transforming growth factor‐beta 1 (TGF‐β1) and its relationship with programmed death‐ligand 1 (PD‐L1) expression. A total of 92 patients with surgically resected TETs, including 79 thymomas and 13 thymic carcinomas, were included. Immunohistochemical analyses were performed to assess the expression of TGF‐β1, PD‐L1, phosphorylated Smad2 (pSmad2), and pSmad3. Associations between TGF‐β1 expression and clinicopathological features were analyzed, and mechanistic interactions were investigated using two thymic carcinoma cell lines exposed to exogenous TGF‐β1. High TGF‐β1 expression was observed in 28% of patients and was significantly associated with advanced Masaoka stage (III/IV), shorter tumor doubling time (median 328 vs. 713 days, p = 0.042), and lower 5‐year freedom from recurrence (FFR) rates (58.1% vs. 95.1%, p < 0.001, log‐rank test). Coexpression of high TGF‐β1 and PD‐L1 was linked to the poorest prognosis (5‐year FFR: 46.1%) and was identified as an independent predictor of recurrence (adjusted hazard ratio: 7.15; 95% confidence interval: 1.20–42.8). Immunohistochemically, TGF‐β1 expression positively correlated with PD‐L1 and pSmad2/3 expression. In vitro, TGF‐β1 stimulation upregulated PD‐L1 expression in a dose‐dependent manner, accompanied by increased pSmad2/3 activation. These findings indicate that high TGF‐β1 expression demarcates a biologically aggressive TET phenotype and, together with PD‐L1, refines postoperative risk stratification, while its ability to drive PD‐L1 via Smad signaling could support the blockade of these pathways as a potential therapeutic strategy.

Co‐expression of high TGF‐β1 and PD‐L1 was linked to the poorest prognosis (5‐year FFR: 46.1%) and was identified as an independent predictor of recurrence (adjusted hazard ratio: 7.15; 95% confidence interval: 1.20–42.8) in thymic epithelial tumors. In vitro, TGF‐β1 stimulation upregulated PD‐L1 expression in a dose‐dependent manner, accompanied by increased pSmad2/3 activation. These findings indicate that high TGF‐β1 expression demarcates a biologically aggressive TET phenotype and, together with PD‐L1, refines postoperative risk stratification, while its ability to drive PD‐L1 via Smad signaling could support the blockade of these pathways as a potential therapeutic strategy.

Abbreviations

CYLD

cylindromatosis

FFR

freedom from recurrence

GAPDH

glyceraldehyde 3‐phosphate dehydrogenase

HRs

hazard ratios

IFN‐γ

interferon‐gamma

NK

natural killer

NSCLC

non‐small cell lung cancer

PCR

polymerase chain reaction

PD‐L1

programmed cell death‐ligand‐1

PMSF

phenylmethylsulfonyl fluoride

pSmad2

phospho‐Smad2

pSmad3

phospho‐Smad3

ROC

receiver operating curve

TDT

tumor doubling time

TETs

thymic epithelial tumors

TGF‐β1

transforming growth factor‐beta 1

Treg

regulatory CD4+ T‐cell

1. Introduction

Thymic epithelial tumors (TETs) are relatively rare malignant mediastinal tumors. Surgery is the only curative treatment for long‐term survival [1], and patients who undergo complete resection of early‐stage TETs usually show a good prognosis [2]. Conversely, although patients with local invasion or pleural dissemination can also be surgical candidates, they may experience post‐surgical tumor recurrence with pleural dissemination [3, 4]. Thus, multimodal treatments are necessary for these populations. Although previous reports demonstrated poor prognostic factors for TETs, including Masaoka stage and World Health Organization (WHO) histological type [5, 6], given the limited oncogenic mutations and tumor mutation burden in TETs [5], specific biomarkers for tumor progression and therapeutic targets remain unidentified.

Transforming growth factor β (TGF‐β), initially discovered in 1983 [7], is a pleiotropic cytokine related to tumor genesis and progression [8, 9]. TGF‐β acts in the tumor microenvironment, resulting in a complex interplay of responses in cancer and stromal cells within the tumor. TGF‐β can support tumor invasion, epithelial‐mesenchymal transition, neo‐angiogenesis, cell motility, and metastasis [8, 9]. TGF‐β also mediates immune evasion by preventing T‐cell activation and division, decreasing the effector function of T and natural killer (NK) cells, and inducing regulatory CD4+ T‐cell (Treg) differentiation [10, 11, 12]. TGF‐β1 is one isoform of TGF‐β. High TGF‐β1 expression in tumor cells is correlated with poor prognosis in various malignant tumors [13, 14, 15]. However, the significance of TGF‐β1 expression in TETs requires in‐depth investigation.

Programmed cell death‐ligand 1 (PD‐L1) is an immunomodulatory protein expressed on antigen‐presenting cells [16], and a target molecule of immunotherapy. Tumor cells expressing PD‐L1 can reduce antigen‐stimulated T‐cell effector activity and terminate immune responses [17, 18]. Although high PD‐L1 expression is correlated with poor prognosis [19, 20, 21], clinical studies have demonstrated the limited efficacy of immunotherapy in TETs [22, 23, 24].

This study aimed to clarify the significance of TGF‐β1 expression on tumor progression, postoperative prognosis, and the tumor microenvironment related to tumor immune evasion via the TGF‐β/Smad/PD‐L1 signaling pathway.

2. Materials and Methods

2.1. Patients

The study protocol was approved by the Institutional Review Board at Kyoto Prefectural University of Medicine (approval number ERB‐C‐931). This study included patients with thymomas and thymic carcinomas who underwent surgical resection at the University Hospital, Kyoto Prefectural University of Medicine (Kyoto, Japan) between April 2001 and March 2020. The following clinicopathological characteristics from the medical records were evaluated as potential prognostic factors: age, sex, Masaoka stage, WHO classification, presence of myasthenia gravis, tumor size, surgery completeness, adjuvant and neoadjuvant therapies, and recurrence and death. All cases were histologically classified according to the 2015 WHO classification [25].

2.2. Immunohistochemistry (IHC) Examination

This study used formalin‐fixed paraffin‐embedded (FFPE) tissue samples from the enrolled patients. The FFPE blocks with the greatest tumor occupancy were selected as representative tumor samples, and 4‐μm‐thick sections were cut for hematoxylin and eosin (HE) and immunohistochemical staining. Antigens were retrieved by microwaving the slices at 95°C for 20 min in citrate buffer at pH 6.0 to stain for TGF‐β1 and phospho‐Smad2 (pSmad2), or in ethylenediaminetetraacetic acid buffer at pH 9.0 to stain for phospho‐Smad3 (pSmad3). To block endogenous peroxidase activity, the samples were placed in 2% H₂O₂ in phosphate‐buffered saline (TGF‐β1) or methanol (phospho‐Smad2 and phospho‐Smad3) for 30 min. The sections were subsequently incubated with primary antibodies against TGF‐β1 (clone TB21; dilution, 1:100; Santa Cruz Biotechnology, Dallas, TX, US), pSmad2 (dilution, 1:50; Invitrogen, ThermoFisher Scientific, Illkirch, France), or pSmad3 (clone EP823Y; dilution, 1:50; Abcam, Cambridge, UK) at 4°C overnight. Histofine Simple Stain MAX‐PO (R) or MAX‐PO (M) (NITIREI Bioscience, Tokyo, Japan) was used as the secondary antibody, depending on the animal species of the primary antibody. The immune reaction was visualized using 3, 3′‐diaminobenzidine for 10 min.

PD‐L1 immunohistochemical staining was performed as previously described [19] with a primary antibody against PD‐L1 (clone SP263, prediluted; Ventana, Tucson, AZ). The antibodies and chemical compounds are listed in a table as Table S1.

2.3. IHC Evaluation

Positivity was defined as TGF‐β1 immunostaining in the membrane or cytoplasm, PD‐L1 in the membrane, and pSmad2 and pSmad3 in the nucleus of the tumor cells. The immunostaining scores of TGF‐β1, pSmad2, and pSmad3 expression were assessed using a visual grading system based on the staining extent (percentage of positive tumor cells graded on a scale of 0–4: 0 = ≤ 5%, 1 = 5%–25%, 2 = 26%–50%, 3 = 51%–75%, 4 = ≥ 75%) and intensity (graded on a scale of 1–3: 1 = weak staining, 2 = moderate staining, 3 = strong staining) [26]. The percentage grade of positive tumor cells and the staining intensity grade were then multiplied to produce an individual staining score for each sample, ranging from 0 to 12. Based on the receiver operating characteristic (ROC) analysis for predicting tumor recurrence, high and low TGF‐β1 expression were defined as staining scores of ≥ 3 and < 3, respectively. PD‐L1 immunostaining was scored using the tumor proportion score (TPS), with TPS of > 25% and ≤ 25% defined as high and low PD‐L1 expression, respectively [19, 27, 28]. All immunohistochemical results were examined by two independent observers (C.N. and expert pathologist S.T.). The mean staining score was considered for the analyses.

2.4. Tumor Doubling Time (TDT)

The TDT was calculated from preoperative computed tomography (CT) images [29]. The largest diameter on axial CT images of 1‐mm slice thickness was recorded for each tumor at two time points, initially and preoperatively, with an interval of > 20 days. The TDT was calculated for each tumor using the following equation: [30] Td = t × log 2/ [3 × log Dt/D0], where t is the interval between CT sessions and Dt and D0 are the final and initial tumor diameters, respectively. Among 92 patients, we excluded those without at least two serial CT sessions with an interval of > 20 days before surgery (n = 45), those who underwent induction therapy, including steroid therapy (n = 17), and those whose tumors demonstrated spontaneous regression (n = 1). Finally, the association between TDT and clinicopathological features was analyzed in 29 patients.

2.5. Cell Culture and Exposure to TGF‐β1

The human thymic carcinoma cell lines MP57 and 1889c were kindly provided by Giaccone et al. [31] and Ehemann et al. [32] Both cell lines were derived from undifferentiated thymic carcinoma. PD‐L1 expression is enhanced in both cells by interferon‐gamma (IFN‐γ), a strong PD‐L1 inducer [33]. Cells were cultured in RPMI‐1640 medium supplemented with 10% fetal bovine serum, 2 mM L‐glutamine, and 100 U/mL penicillin/streptomycin (Gibco, USA) at 37°C in a 5% CO₂ humidified environment. Protein was isolated using RIPA lysis buffer containing the protease inhibitor cocktail “cOmplete” (Roche, Switzerland), 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1 mM orthovanadate (Sigma‐Aldrich, USA). The protein lysates were quantified using the DC Protein Assay (Bio‐Rad, Germany). The absorbance was measured using a Tecan reader (Tecan, Switzerland).

2.6. Western Blot Analysis

Protein was isolated as described above. A total of 15 μg of protein was separated on precast Mini Protein TGX gels and blotted using the semi‐dry Trans‐Blot Turbo System (Bio‐Rad, Germany). The membranes were incubated with primary antibodies PD‐L1 (clone E1L3N, Cell Signaling, dilution 1:1000), pSmad2/3 (clone D27F4, Cell Signaling, dilution 1:1000), and glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) (Cell Signaling, dilution 1:1000) at 4°C overnight, then washed and incubated with the secondary antibody for 1 h at room temperature. Protein expression was detected using a FUSION Chemiluminescence Imaging System (Paqlab, Germany). ImageJ software version ImageJ (version 1.53a; National Institutes of Health, Bethesda, Maryland, USA) was used to quantify specific signals and calculate relative PD‐L1 expression after GAPDH normalization in biological triplicates.

2.7. Quantitative Real‐Time Polymerase Chain Reaction (PCR)

RNA was isolated using an RNeasy Mini kit (Qiagen, Tokyo, Japan), and cDNA was generated using the UltraScript 2.0 cDNA Synthesis kit (Nippon Genetics, Japan). Gene expression of GAPDH (GAPDH, Hs02758991_g1) and PD‐L1 (CD274, Hs01125301_m1) was detected using TaqMan gene assays (Applied Biosystems, Carlsbad, CA, USA) on a Light Cycler 480 (Roche, Germany) according to the manufacturer's instructions. The relative PD‐L1 mRNA expression levels of biological triplicates were calculated using the delta–delta‐Ct method normalized to GAPDH [34]. The relative PD‐L1 mRNA expression level of the control was adjusted to 1 and was compared with the expression levels of the TGF‐β1‐treated samples.

2.8. Statistical Analyses

Clinicopathological features were compared between the groups using Fisher's exact or chi‐squared tests for categorical variables and the Mann–Whitney U test for continuous variables. Type A, AB, B1, and metaplastic thymoma were categorized as low‐grade malignancies [35, 36], while type B2, B3 thymoma, and thymic carcinoma were categorized as high‐grade malignancies.

Relationships among PD‐L1, TGF‐β1, pSmad2, and pSmad3 expression levels were assessed using chi‐squared or Cochran‐Armitage trend tests. Relative mRNA PD‐L1 expression levels after TGF‐β1 administration were compared among control and treated cells using Dunnett's multiple comparisons test. Freedom from recurrence (FFR) was defined as the time from the date of surgery to that of the first TET recurrence [37]. Survival curves were plotted using the Kaplan–Meier method and compared using a log‐rank test. The hazard ratios (HRs) for FFR were estimated using Cox proportional hazards regression analyses with Firth's correction for TGF‐β1 and PD‐L1 co‐expression [38]. The prognostic impact of TGF‐β1 and PD‐L1 co‐expression was evaluated using multivariable Cox regression analysis to estimate HR after adjusting for all the evaluated variables (age, sex, Masaoka stage, WHO classification, myasthenia gravis, tumor size, resection status) using Firth's correction. A two‐sided p‐value of < 0.05 was considered statistically significant. Statistical analysis was performed using EZR 1.32 (Saitama Medical Center, Jichi Medical University, Japan) and JMP version 13.0 (SAS Institute Inc., Cary, NC, USA).

3. Results

3.1. Patient Clinicopathological Characteristics

The clinical characteristics of the 92 patients with TETs undergoing surgical resection are shown in a table as Table S2. The patients included 48 males and 44 females with a median age of 62 years who were postoperatively monitored for a median of 73 months. No patients received immunotherapy at any stage. The WHO histological classifications included type A (n = 9), AB (n = 27), B1 (n = 10), B2 (n = 22), B3 (n = 9), metaplastic thymomas (n = 2), and thymic carcinomas (n = 13; 11 thymic squamous cell carcinomas and two micronodular thymic carcinomas with lymphoid stroma). TETs classified according to pathologic Masaoka stage included stage I (n = 35), II (n = 37), III (n = 13), and IV (n = 7).

3.2. Correlation Between TGF‐β1 Expression and Patient Characteristics

Representative slides of immunohistochemical staining for TGF‐β1, PD‐L1, pSmad2, and pSmad3 are illustrated in Figure 1A. In addition, representative staining of TGF‐β1 in tumor cells and lymphocytes, along with HE and AE1/AE3 staining, is shown in Figure S1. TGF‐β1 expression was predominantly observed in tumor cells. In addition, a small number of positive lymphocytes were occasionally observed in the peritumoral area. The median (IQR) staining scores for TGF‐β1, pSmad2, and pSmad3 were 0 (0–3), 6.5 (3–9), and 5.5 (2.5–8), respectively, while the median TPS (IQR) of PD‐L1 was 29 (9–71) %. High TGF‐β1 and high PD‐L1 expression were detected in 26 (28.3%) and 51 (55.4%) patients, respectively. High TGF‐β1 expression was detected in 15.4% (2/13) and 30.4% (24/79) cases of thymic carcinoma and thymoma, respectively. The TGF‐β1 expression levels and staining scores according to WHO classification are shown in Figure 1B–E and Figure S2 (multipart image showing the distribution of TGF‐β1, PD‐L1, pSmad2, and pSmad3 immunostaining scores). TGF‐β1 expression was significantly correlated with advanced Masaoka stage (III/IV) and recurrence after surgical resection, whereas no significant correlations with WHO classification, tumor size, or resection status were observed (Table 1).

FIGURE 1.

Representative immunohistochemical staining of TGF‐β1, pSmad2, and pSmad3 and their correlation with PD‐L1 expression levels. Weak, moderate, and strong TGF‐β1, pSmad2 and pSmad3 expression in thymomas. For TGF‐β1, weak, moderate, and strong expression is observed in type B2, AB, and B2 thymomas, respectively. For pSmad2, expression levels correspond to type AB, B3, and B3 thymomas, respectively. For pSmad3, expression levels correspond to type B2, B2, and B3 thymomas, respectively (A). Expression levels of TGF‐β1 (B), PD‐L1 (C), pSmad2 (D), and pSmad3 (E) in each histological subtype of thymic epithelial tumors. Bars: 50 μm. TGF‐β1, transforming growth factor‐β1; pSmad2, phospho‐Smad2; pSmad3, phospho‐Smad3; PD‐L1, programmed death‐ligand 1.

TABLE 1.

Correlations between clinicopathological characteristics and transforming growth factor‐β1 expression in thymic epithelial tumors.

Variables	Low TGF‐β1 (n = 66)	High TGF‐β1 (n = 26)	p
Age (years)	62 (28–72)	61 (22–68)	0.101 a
Age, ≥ 60 years	39 (59)	14 (54)	0.823 b
Sex Male /	32 (48)	16 (62)	0.370 b
Female	/34 (52)	/10 (38)
Masaoka stage			< 0.001 b
I/II	60 (91)	12 (46)
III/IV	6 (9)	14 (54)
WHO classification
Thymoma	55 (83)	24 (92)	0.337 c
Thymic carcinoma	11 (17)	2 (8)
WHO classification			0.622 b
A/AB/B1/metaplastic	36 (55)	12 (46)
B2/B3/carcinoma	30 (45)	14 (54)
Myasthenia gravis	17 (26)	9 (35)	0.658 b
Tumor size (cm)	4.1 (3.2–6.2)	5.5 (4.6–6.0)	0.106 a
Tumor size, > 5 cm	27 (41)	16 (62)	0.120 b
Resection status			0.309 c
R0	60 (90)	21 (81)
R1, R2	7 (10)	5 (19)
Recurrence	5 (7)	10 (38)	< 0.001 c

Note: Data are presented as the median (interquartile range) or number (%).

Abbreviation: TGF‐β1, transforming growth factor‐β1.

^a Wilcoxon rank sum test.

^b Chi‐squared test.

^c Fisher's exact test.

3.3. Correlations Between the TGF‐β1/Smad Signaling Pathway and PD‐L1 Expression

In all TETs, PD‐L1 expression was correlated with TGF‐β1 expression (p < 0.05) (Figure 2A). PD‐L1 expression was also positively correlated with pSmad2 and pSmad3 expression (p < 0.001, respectively) (Figure 2B,C). TGF‐β1 expression was positively correlated with pSmad3 expression (p = 0.024) but not with pSmad2 (p = 0.191) (Figure 2D,E). Among thymomas, similar positive correlations were observed between PD‐L1 expression and TGF‐β1, pSmad2, and pSmad3 expression (all p < 0.01) (Figure S3, a multipart image showing the correlation of TGF‐β1, pSmad2, and pSmad3 expression with PD‐L1 expression and pSmad2 and pSmad3 expression with TGF‐β1 expression in patients with thymoma). TGF‐β1 expression was positively correlated with pSmad3 (p = 0.028) but not with pSmad2 (p = 0.161) (Figure 2D,E).

FIGURE 2.

Correlation of TGF‐β1, pSmad2, and pSmad3 expression with PD‐L1 expression and pSmad2 and pSmad3 expression with TGF‐β1 expression. High PD‐L1 expression is correlated with high TGF‐β1 expression (chi‐squared test) (A) and pSmad2 and pSmad3 (Cochran‐Armitage trend test) (B, C). High TGF‐β1 expression is correlated with high pSmad3 (E). No significant correlation is observed for pSmad2 (p = 0.191) (D). *p < 0.05; ***p < 0.001. TGF‐β1, transforming growth factor‐β1; pSmad2, phospho‐Smad2; pSmad3, phospho‐Smad3; PD‐L1, programmed death‐ligand 1.

3.4. FFR According to TGF‐β1 and PD‐L1 Expression Levels

Among all TETs, high TGF‐β1 was associated with a significantly shorter FFR than low TGF‐β1 (p < 0.001) (Figure 3A). Among early‐stage (Masaoka stage I/II) TETs, high TGF‐β1 was also associated with a shorter FFR than low TGF‐β1 (p = 0.031) (Figure 3B). The group with high TGF‐β1 and PD‐L1 co‐expression was associated with a significantly shorter FFR than the other groups (Figure 3C). Similar results were observed among all thymomas and high‐risk TETs (type B2, B3 thymoma, and thymic carcinoma). Figure S4 shows FFR according to TGF‐β1 and PD‐L1 expression levels in patients with thymoma.

FIGURE 3.

FFR according to TGF‐β1 and PD‐L1 expression levels. FFR according to TGF‐β1 expression levels in 92 patients with TETs (A). FFR according to TGF‐β1 expression levels in 72 patients with Masaoka stage I, II TETs (B). FFR according to the TGF‐β1 and PD‐L1 co‐expression levels in 92 patients with TETs (C). FFR, freedom from recurrence; PD‐L1, programmed death‐ligand 1; TET, thymic epithelial tumors; TGF‐β1, transforming growth factor‐β1.

The univariable analysis showed that high TGF‐β1 expression, high PD‐L1 expression, advanced Masaoka stage (III/IV), WHO classification (type B2 and B3 thymoma and thymic carcinoma), tumor size of > 5 cm, and resection status (R1 and R2) were significantly correlated with a shorter FFR (Table 2). The results of the multivariable analysis revealed that high TGF‐β1 and PD‐L1 co‐expression was a significant independent prognostic factor predicting worse FFR, regardless of age, sex, Masaoka stage, WHO classification, myasthenia gravis, tumor size, and resection status.

TABLE 2.

Univariable and multivariable analyses of freedom from recurrence in 92 patients with thymic epithelial tumors.

Variables	Univariable analysis		Multivariable analysis a
	HR (95% CI)	p	HR (95% CI)	p
TGF‐β1 and PD‐L1 expressions
Low TGF‐β1 and Low PD‐L1	Reference		Reference
Low TGF‐β1 and High PD‐L1	1.39 (0.25–7.83)	0.712	1.22 (0.19–7.72)	0.830
High TGF‐β1 and Low PD‐L1	1.45 (0.06–36.9)	0.823	2.58 (0.09–75.8)	0.582
High TGF‐β1 and High PD‐L1	12.0 (2.71–53.3)	0.001	7.15 (1.20–42.8)	0.031
Age (≥ 60 years vs. < 60 years)	1.33 (0.47–3.77)	0.587	1.01 (0.31–3.32)	0.991
Sex (Male vs. Female)	1.92 (0.65–5.62)	0.236	1.13 (0.32–3.98)	0.852
Masaoka stage (III/IV vs. I/II)	16.4 (5.19–52.0)	< 0.001	2.40 (0.56–10.3)	0.240
WHO classification (B2/B3/carcinoma vs. A/AB/B1/others)	5.18 (1.46–18.4)	0.011	2.10 (0.49–9.02)	0.320
Myasthenia gravis (present vs. absent)	0.55 (0.15–1.95)	0.353	0.42 (0.10–1.67)	0.216
Tumor size (> 5 cm vs. ≤ 5 cm)	18.1 (2.38–138)	0.005	6.79 (1.15–40.0)	0.034
Resection status (R1, R2 vs. R0)	4.51 (1.53–13.3)	0.006	2.05 (0.49–8.63)	0.326

Abbreviations: CI, confidence interval; HR, hazard ratio; PD‐L1, programmed death‐ligand 1; TGF‐β1, transforming growth factor‐β1.

^a The hazard ratios were estimated using Firth's bias correction in Cox regression models [38].

Among all thymomas, similar results were observed in univariable analysis, and high TGF‐β1 and PD‐L1 co‐expression was a significant independent prognostic factor predicting worse FFR in multivariable analysis. Table S3 shows univariable and multivariable analysis results for FFR in 79 patients with thymoma.

3.5. TDT According to TGF‐β1 Expression Levels

We calculated the TDT for 29 patients (16 males and 13 females, median age 69 years). The median initial and preoperative tumor sizes were 39.9 mm and 41.3 mm, respectively. The WHO histological classifications included type A (n = 5), type AB (n = 9), type B1 (n = 2), type B2 (n = 8), type B3 (n = 2), thymic carcinoma (n = 2), and metaplastic thymoma (n = 1). The distribution of Masaoka stages included I (n = 10), II (n = 17), III (n = 2), and IV (n = 1). The median interval between serial CT studies was 54 days.

The median TDT of all tumors was 485 days (range, 96–6876 days; IQR, 313–1157 days). The TDT was significantly shorter in the high TGF‐β1 group than in the low TGF‐β1 group (median 328 vs. 713 days, p = 0.042). Figure S5 image shows the correlation between TDT and TGF‐β1 expression, and Table S4 lists the correlations between TDT and clinicopathological characteristics in TETs. Age, sex, Masaoka stage, WHO classification, myasthenia gravis, and PD‐L1 expression were not associated with TDT.

3.6. TGF‐β1 Enhances PD‐L1 Expression via the Smad Pathway in the 1889c Thymic Carcinoma Cell Line

Evaluation of the protein expression levels of PD‐L1, pSmad2, and pSmad3 in the MP57 and 1889c thymic carcinoma cells revealed weak PD‐L1 expression in 1889c but not in MP57 (Figure 4A). pSmad2 and pSmad3 expressions were minimal in both cell lines. Culture with recombinant TGF‐β1 (0, 5, and 10 ng/mL) for 48 h to assess its effects in these cell lines demonstrated no detectable PD‐L1 expression in MP57 cells, regardless of TGF‐β1 concentration (Figure 4B). However, PD‐L1 expression showed a TGF‐β1 dose‐dependent increase in 1889c cells (Figure 4C,D), similar to reports in non‐small cell lung cancer (NSCLC) cells [39].

FIGURE 4.

Western blot and quantitative PCR analysis of PD‐L1 expression after TGF‐β1 administration in thymic carcinoma cell lines. Baseline PD‐L1, pSmad2, and pSmad3 expression in thymic carcinoma cell lines (A). Protein and mRNA expressions were evaluated after TGF‐β1 administration at the indicated concentrations for 48 h. Western blot analysis showing that PD‐L1 expression does not change in MP57 cells (B), whereas PD‐L1 expression is enhanced along with pSmad2 upregulation in 1889c cells (C). Further analyses of the reactive 1889c cell line. Relative PD‐L1 expression, calculated by normalizing PD‐L1 expression by GAPDH expression in western blot analysis, indicated upregulation of PD‐L1 expression in a TGF‐β1 dose‐dependent manner. The data are based on three biological replicates (D). Relative PD‐L1 mRNA expression showing that the administration of as low as 5 ng/mL TGF‐β1 induced PD‐L1 mRNA expression, an effect that also occurred at 10 ng/mL in 1889c cells. The data are based on three biological replicates with three technical replicates each (E). *p < 0.05. TGF‐β1, transforming growth factor‐β1; pSmad2, phospho‐Smad2; pSmad3, phospho‐Smad3; PD‐L1, programmed death‐ligand 1.

Quantitative PCR, which was used to further determine the effects of TGF‐β1 on transcription, also showed enhanced PD‐L1 mRNA expression in 1889c cells (Figure 4E). To investigate pathways through which TGF‐β1 enhanced PD‐L1 expression, pSmad2 and pSmad3 protein expressions were evaluated. pSmad3 was slightly induced in both cell lines. While pSmad2 expression was significantly upregulated along with PD‐L1 enhancement in the 1889c cell line, it was not upregulated in the MP57 cell line, in which PD‐L1 expression was not affected by TGF‐β1. These results support a potential role of the TGF/Smad pathway in mediating TGF‐β1–induced PD‐L1 expression in TETs.

4. Discussion

In the present study, high TGF‐β1 expression was associated with advanced stages and was a poor prognostic factor in both early and advanced resectable TETs. The results of the clinical preoperative image analysis revealed that high TGF‐β1 expression was related to shorter TDT, namely, rapid growth. These results indicate that TETs with high TGF‐β1 expression have more aggressive features. We also observed a positive relationship between TGF‐β1 and PD‐L1 expression levels in IHC analysis of clinical samples and in vitro, indicating that TGF‐β1 induces PD‐L1 upregulation via the Smad signaling pathway. To the best of our knowledge, this is the first report on the clinical and biological aspects of TGF‐β1 expression in resectable TETs.

Similar to our study, previous studies reported the association of high TGF‐β1 expression with a worse prognosis in various malignant tumors [7, 8, 9]. TGF‐β1 might be associated with tumor progression by stimulating angiogenesis and immunosuppression. However, only one study has described the prognostic significance of TGF‐β in TETs. Duan et al. reported the results of IHC analysis from biopsy specimens in 33 cases of unresectable advanced TETs, in which 65.0% and 15.4% of thymic carcinoma and thymoma cases, respectively, showed high TGF‐β expression [26]. The group of patients with high TGF‐β expression demonstrated marginally lower overall survival compared with those with low TGF‐β expression (median 29.5 months [95% confidence interval [CI]: 18.6–40.4] vs. 62.9 months [95% CI: 15.6–110.1]; p = 0.052). Examination of resectable TETs in the present study revealed high TGF‐β1 expression in 15.4% and 30.4% of thymic carcinomas and thymomas, respectively, showing a lower rate of high expression in thymic carcinoma compared with advanced cases [26]. Moreover, the FFR was significantly shorter in patients with high TGF‐β1 expression among the whole cohort and those with early‐stage disease. Therefore, TGF‐β1 could be a poor prognostic factor in all phases of tumor progression, indicating its critical role in tumor recurrence.

We also observed TGF‐β1 positivity in a small fraction of lymphocytes adjacent to tumor cells. This finding suggests a potential involvement of immune cells in shaping the tumor microenvironment. A more precise characterization of lymphocyte subsets responsible for TGF‐β1 expression will require further studies, and we consider this an important direction for future investigation.

Tumor size is a prognostic factor for recurrence after surgical resection of TETs [4, 40]. Okumura et al. reported a higher incidence of recurrence in patients with thymomas measuring > 5 cm. Since TET tumor size of > 5 cm has a poor prognosis, the 9th edition of the TNM classification includes the new subclassification of the T1 category into T1a (≤ 5 cm) and T1b (> 5 cm) [41]. We also observed in the present study that a tumor size of > 5 cm was a significant independent factor in predicting recurrence. Additionally, we found that the TDT was significantly shorter in patients with TETs with high TGF‐β1 expression than in those with low TGF‐β1 expression. Thus, TGF‐β1 may affect tumor growth in TETs. This biological effect may also be one explanation for the poor prognosis in patients with high TGF‐β1 expression.

In the present study, high TGF‐β1 and PD‐L1 co‐expression were significant independent factors predicting recurrence. In several malignancies, PD‐L1 overexpression is a detrimental factor as it can suppress the antitumor T‐cell immune response, leading to immune evasion [42, 43]. Immune evasion is a new hallmark of cancer and an emerging target of cancer therapy [44]. Previous reports, including ours, demonstrated the association of high PD‐L1 expression in TETs with poor prognosis [19, 20, 21]. Poor prognosis in patients with high PD‐L1 expression is attributed to PD‐L1‐induced T‐cell‐mediated immune suppression in the tumor microenvironment [21].

The IHC results of the present study demonstrated the positive correlations of expression levels between PD‐L1 and TGF‐β1, pSmad2, and pSmad3 (Figure 2). Furthermore, we demonstrated the upregulation of PD‐L1 expression along with pSmad2/3 upregulation in a TGF‐β1 dose‐dependent manner in 1889c cells. In other words, TGF‐β1 upregulated PD‐L1 transcription and translation in TET cells in vitro (Figure 4). TGF‐β1 reportedly upregulates PD‐L1 expression depending upon Smad2 in human lung cancer cells [45], consistent with our findings. However, the response of PD‐L1 to its regulatory molecules differed between the cell lines used in the present study. Similarly, another study analyzing the effect of cylindromatosis (CYLD) on PD‐L1 expression in TET cell lines showed that downregulating CYLD knockdown enhanced IFN‐γ‐mediated PD‐L1 expression in 1889c but not in MP57 [33]. Some differences between these cells are known; for instance, 1889c cells are cytokeratin‐negative and harbor the TP53 mutation, whereas MP57 cells are cytokeratin‐positive and harbor TP53, PIK3R2, and other mutations [31, 32], few other TET cell lines are available, and the reason why these two cell lines responded differently to TGF‐β1 is not clear. However, taken together, the IHC and in vitro data indicated that TGF‐β1 upregulated PD‐L1 by stimulating the Smad pathway in TETs. This pathway may promote PD‐L1‐dependent immune suppression in TETs, which may be one explanation for the poor prognosis in patients with high TGF‐β1 and PD‐L1 co‐expression in the present study.

Nevertheless, it should be noted that not all cases followed this pattern. In our cohort, 30 of 92 cases showed low TGF‐β1 expression despite high PD‐L1 expression. One possible explanation is the temporal variability of TGF‐β1 expression, which may be influenced by microenvironmental conditions; however, as our analysis relied on single time‐point FFPE samples, such temporal dynamics of cytokine expression could not be assessed. Another possibility is that PD‐L1 expression is regulated through TGF‐β1‐independent pathways. In fact, both IFN‐γ signaling and the PI3K/AKT pathway have been reported to upregulate PD‐L1 in thymic carcinoma cells [31, 33]. These alternative mechanisms may account for the TGF‐β1‐low/PD‐L1‐high subgroup observed in the present study.

Similar to our results, a previous study treated NSCLC cells with Bintrafusp Alfa (M7824), a novel clinical‐stage bifunctional agent that targets both PD‐L1 and TGF‐β, which induced significant suppression of tumor growth in tumor xenografts [45]. Moreover, M7824 impeded the functional effects of TGF‐β signaling by enhancing cell proliferation. The authors speculated that M7824 reduced PD‐L1 expression by TGF‐β trapping and might be clinically beneficial to patients as it may alleviate PD‐L1‐dependent immune suppression. Several clinical trials are investigating the utility of M7824 in malignant tumors, including TETs [46, 47, 48]. Given the worst prognosis of patients with high TGF‐β1 and PD‐L1 co‐expression in the present study, suppressing both molecules could be a potential target for drug treatment in aggressive TETs. Therefore, we anticipate the results of a phase II trial of M7824 in TETs [46].

Our study has some limitations. First, few recurrence events occurred during the study period, thus preventing meaningful analyses of the potential of TGF‐β1 expression as a biomarker for FFR in TETs. Although 92 TETs were a relatively large sample size, further studies with even more cases and longer follow‐up times are needed. Second, although the predictive value of TGF‐β1 expression has been determined by IHC scoring in TETs, this scoring system has not yet been established [26]. Further analyses are needed to define TGF‐β1 expression and to understand how TGF‐β1 is expressed biologically in TETs. Third, this study focused on the expression level of TGF‐β1 and its related molecules in tumor cells. As the function of TGF‐β1 is multifocal, fibroblasts, lymphocytes, or other inflammatory cells must also be investigated in future studies to more fully understand the critical role of TGF‐β1 in the tumor microenvironment.

In conclusion, our results suggested that high TGF‐β1 expression in TETs is a poor prognostic factor for patients undergoing surgical treatment for TETs. TDT analysis indicated that TGF‐β1 might affect tumor growth in TETs. Moreover, TGF‐β1 upregulated PD‐L1 expression via the Smad signaling pathway, potentially leading to tumor progression through PD‐L1‐dependent immune suppression in TETs. TGF‐β1 could serve as another therapeutic target for suppressing the PD‐1/PD‐L1 pathway, which is a mainstay of tumor immune evasion.

Author Contributions

Chiaki Nakazono: conceptualization, data curation, formal analysis, investigation, methodology, resources, visualization, writing – original draft. Satoru Okada: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, resources, supervision, visualization, writing – original draft, writing – review and editing. So Tando: investigation, methodology, writing – review and editing. Shunta Ishihara: data curation, methodology, resources, writing – review and editing. Masanori Shimomura: resources, writing – review and editing. Tatsuo Furuya: resources, writing – review and editing. Kenji Kameyama: resources, writing – review and editing. Stefan Küffer: investigation, methodology, resources, writing – review and editing. Denise Müller: investigation, resources, writing – review and editing. Alexander Marx: investigation, writing – review and editing. Satoshi Teramukai: formal analysis, writing – review and editing. Philipp Ströbel: investigation, resources, supervision, writing – review and editing. Kyoko Itoh: supervision, writing – review and editing. Masayoshi Inoue: project administration, resources, supervision, writing – review and editing.

Ethics Statement

The study protocol was approved by the Institutional Review Board at Kyoto Prefectural University of Medicine (approval number ERB‐C‐931).

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Representative immunohistochemical staining of TGF‐β1, AE1/AE3, and HE.

Figure S2: Distribution of TGF‐β1, PD‐L1, pSmad2, and pSmad3 immunostaining scores.

Figure S3: Correlation of TGF‐β1, pSmad2, and pSmad3 expression with PD‐L1 expression and pSmad2 and pSmad3 expression with TGF‐β1 expression in patients with thymoma.

Figure S4: FFR according to TGF‐β1 and PD‐L1 expression levels in patients with thymoma.

Figure S5: TDT according to TGF‐β1 levels in 29 TETs with evaluable CT images.

Table S1: Antibodies and chemical compounds.

Table S2: Clinicopathological characteristics of patients undergoing surgical resection of TETs.

Table S3: Univariable and multivariable analyses for FFR in patients with thymoma.

Table S4: Correlation between TDT and clinicopathological characteristics in TETs.

Data Availability Statement

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.