Relationships of ¹⁸F-FDG PET with tumor microenvironment immunotypes, especially PD-L1 and CD15 expression, and prognosis in oral squamous cell carcinoma

Abstract

Background

The relationship between 2-[18F]-fluoro-2-deoxy-d-glucose–positron emission tomography (FDG-PET) findings and programmed death ligand-1 (PD-L1) expression has been reported in several cancers. We investigated the correlations of FDG uptake with immune cell counts, including myeloid-derived suppressor cells (MDSCs), and PD-L1 expression in the tumor microenvironment. We examined 72 patients with oral squamous cell carcinoma (OSCC) with immunohistochemistry data for PD-L1, CD8, S100A8, CD15, and CD33. We used the maximum standardized uptake value (SUVmax) to reflect FDG uptake in each patient.

Results

High SUVmax and high MDSC counts were associated with poor prognosis. Significantly higher SUVmax was found in patients with high PD-L1 expression and in those with a high CD15⁺ cell density (P = 0.03 and P = 0.02, respectively). In multiple regression analysis, the tumor size had the greatest effect on SUVmax (P < 0.001), followed by PD-L1 (P = 0.014), and when the tumor size was excluded, CD15 (P = 0.02) was included in the prediction equation. FDG uptake in some cold tumor subgroups, low PD-L1 expression, and a low CD8⁺ cell density was linked to significantly lower SUVmax than the other variables. High SUVmax was clearly associated with high PD-L1 expression and/or a high CD15⁺ cell density.

Conclusions

FDG uptake was affected by PD-L1 expression and the density of CD15⁺ cells in cancer tissue. FDG-PET may illuminate the tumor microenvironment immunotypes before biopsy or resection.

Background

Human cancers contain myeloid-derived suppressor cells (MDSCs), and the diverse populations of these cells trigger antitumor immunity [1]. In various cancer, it has been reported that MDSCs infiltrate peripheral blood and tumor tissues. In patients with oral squamous cell carcinoma (OSCC), increased MDSC counts within or around the tumor were obviously associated with poor prognosis [2]. In the microenvironment of oral cancer, programmed death ligand-1 (PD-L1) expression and the presence of immune cells have recently been reported [3, 4]. We also examined the number of immune cells, including PD-L1-positive cells and MDSCs in the tumor microenvironment in patients with OSCC using CD15 as a surrogate marker and found that MDSC counts were significantly lower in patients who showed tumor shrinkage [3].

According to PD-L1 status and the number of tumor-infiltrating lymphocytes (TILs), it is proposed to divide the tumor microenvironment into four groups: types I (PD-L1+ with TILs driving adaptive immune resistance), II (PD-L1 with no TILs, indicating immune ignorance), III (PD-L1+ with no TILs, indicating intrinsic induction), and IV (PD-L1 with TILs, indicating the function of other suppressors in promoting immunological tolerance) [5]

Tumors with a high density of infiltrating immune cells are called "hot" tumors and are thought to be more responsive to checkpoint inhibitor therapy than "cold" tumors with little or no local infiltration of immune cells [6]. MDSCs are among the factors that activate regulatory T cells (Tregs), suppress immune responses, and play an important role in the maintenance of the tumor's cold state in the tumor microenvironment [7].

2-[18F]-fluoro-2-deoxy-d-glucose–positron emission tomography (FDG-PET) is a non-invasive imaging modality for the differential diagnosis, characterization, and staging of malignant tumors [8, 9]. FDG-PET can be especially useful in the diagnosis of head and neck tumors, where tumor staging is an important prognostic factor that largely determines the therapeutic regimen [8, 9].

To date, many studies have reported PD-L1 expression and infiltrating immune cell counts in the tumor microenvironment [10,11,12,13,14]. Recently, the relationship between PD-L1 expression and FDG uptake was confirmed. Such reports in gastric [10], esophageal [15], breast [16], and lung cancers [17,18,19,20], suggested that high PD-L1 expression is correlated with high FDG uptake. Furthermore, a hot tumor immune status, which indicates high PD-L1 expression and the presence of CD8⁺ lymphocytes, tends to portend a poor prognosis [15, 17]. Only one study investigating the relationship between FDG uptake and PD-L1 expression in oral cancer has been reported [21], and certain conclusions in OSCC have not yet been reached. Furthermore, no reports examined the effects of the tumor microenvironment immunotypes divided into the four aforementioned groups on the relationship between FDG uptake and MDSC counts in lesions including cancers of other organs.

In this study, we revealed the correlations of FDG uptake with PD-L1 and CD8 expression in OSCC. Furthermore, we investigated the correlation between FDG uptake and the counts of immune cells including MDSCs in the tumor microenvironment in patients with OSCC via immunohistochemistry using CD15, CD33, or S100A8 as the surrogate marker. We also identified the factor most strongly associated with FDG uptake and evaluated the relationship with prognosis.

Methods

We examined 72 patients with OSCC treated at our hospital between April 2018 and December 2019. Cases in which pathology specimens were not available, such as biopsies performed at other hospitals, and cases with a history of another oral cancer were excluded from the study because recurrence could not be ruled out. Patients in this study were followed-up for 2 years 6 months (median 18 months). Relapse cases consisted of local recurrences and lymph node or distant metastasis. There were 15 relapses and three relapse deaths. The clinicopathological details of the patients are presented in Table 1. We measured tumor size through clinical inspection at first medical examination with reference to imaging data such as those acquired using CT, MRI, and ultrasonography. All patients underwent 18F-FDG-PET for preoperative assessment. Seventy cases underwent FDG-PET after biopsy. Forty-two patients received preoperative S-1 chemotherapy, and the remaining 30 patients did not receive chemotherapy before surgery. Fundamentally, S-1 was orally administered before surgery at a dose of 80–120 mg/day according to the preoperative waiting period (mean, 11.1 days; median, 14 days; range, 1–28 days). The internal dose was dependent on the patient’s body surface area. All patients finally underwent local tumor excision. Patient follow-up was performed through clinical observation, CT, MRI, and FDG-PET imaging. Diagnoses of local recurrent lesions were performed through biopsy or resected lymph nodes. Distant metastases were diagnosed according to a combination of imaging findings such as those acquired using CT, MRI, and FDG-PET. The study was approved by the Institutional Review Board for Clinical Research.

Table 1 Univariate and multivariate analysis of clinical and pathological variables for relapse

PET

Prior to the PET scan, all patients fasted for at least 6 h. The scan was performed using a PET/CT scanner with a 700-mm field of view (Discovery STE; GE Healthcare, Fairfield, CT, USA). After that, we simulated the capture of three-dimensional data 50 min after intravenously administering 5 MBq/kg 18F-FDG. We acquired four to 10 bed positions (3-min acquisition per bed position) according to the range of the imaging. Attenuation-corrected transverse images obtained with 18F-FDG were reconstructed with the ordered-subsets expectation maximization algorithm into 128 × 128 matrices with a slice thickness of 3.27 mm. All 18F-FDG images were interpreted by two experienced nuclear physicians. The physicians were unaware of the patient’s clinical history and pathological data. Any discrepancies between the physicians were resolved by consensus. Functional images of the standardized uptake value (SUV) were produced using attenuation-corrected trans-axial images, the injected doses of 18F-FDG, the patients’ body weight, and the cross-calibration factor between PET and the dose calibrator. We defined the SUV as follows: SUV = radioactive concentration in the region of interest (ROI) (MBq/g)/injected dose (MBq)/patient’s body weight (g). The ROI was manually drawn over the primary tumor on SUV images. When the tumor was > 1 cm in diameter or if the shape of the tumor was irregular or multifocal, a ROI of approximately 1 cm in diameter was drawn over the area corresponding to the maximal tracer uptake. ROI analysis was performed by a nuclear physician with the aid of the corresponding CT scans. The maximal SUV in the ROI (SUVmax) was used as a representative uptake value for evaluating 18F-FDG uptake in each OSCC tumor. The average SUVmax in OSCC is 9.7 according to a previous report from our institution [9]. The cutoff for low or high SUVmax was 9.5 in this study.

Variables of histopathology and immunohistochemistry

All histopathological and immunohistochemical variables were evaluated in biopsy specimens as some cases were affected by preoperative chemotherapy when evaluating resected specimens. According to the World Health Organization criteria, tumors < 2 cm in the greatest dimension were defined as T1, whereas those > 2 cm were given a grade of T2 or higher. The depth of invasion (mm) was measured from the surface of the tumor to the deepest point of an invasive tumor. Immunostaining was performed according to a routine protocol using mouse monoclonal antibodies against PD-L1 (28–8; 1:400 dilution; Abcam, Cambridge, UK), CD8 (C8/144B; 1:50 dilution; Nichirei, Tokyo, Japan), CD15 (Carb-3; 1:80 dilution; Dako), S100A8 (ab92331; 1:2000 dilution; Abcam), and CD33 (PWS44; 1:50 dilution; Leica). Antigen retrieval was performed by boiling samples for 30 min in 0.01 mol/L sodium citrate buffer (pH 6.0).

Immunohistochemistry evaluation

Immunohistochemistry for PD-L1, CD8, S100A8, CD15, and CD33 was performed using biopsy specimens from all patients. In this study, S100A8, CD33, and CD15 were used as immunohistochemical markers for MDSCs. PD-L1 expression was classified into 10% increments according to the positivity rate in the tumor cell membrane. In immunohistochemistry for CD8, S100A8, CD15, and CD33, the evaluation of positive cells was performed under light microscopy at the stroma at the invasive tumor front [3]. CD8⁺ , S100A8⁺ , CD15⁺ and CD33⁺ cells were counted in five fields at the invasive front at a magnification of × 400 (an area of approximately 228,906 µm²), and the average number of cells was calculated for each sample. All immunohistochemical specimens were evaluated twice by one pathologist (M.S.S.) to exclude intra-observer variability.

Subgroups

Based on the result of receiver operator characteristic curves with variable cutoffs for a patient’s prognosis with regards to PD-L1 expression and the densities of CD8⁺, CD15⁺, S100A8⁺, and CD33⁺ cells, cutoffs of 25% for PD-L1 expression, 111 for CD8⁺ cell counts, 17 for CD15⁺ cell counts, 51 for S100A8⁺ cell counts, and 10 for CD33⁺ cell counts were employed (Fig. 1). Tumors were categorized into four groups based on PD-L1 expression in tumor cells and CD8⁺ cell counts at the tumor invasive front as presented in Fig. 2A [5]. Groups I (high PD-L1 expression and high CD8⁺ cell density) and IV (low PD-L1 expression and high CD8⁺ cell density) represented hot tumors, and Groups II (low PD-L1 expression and low CD8⁺ cell density) and III (high PD-L1 expression and low CD8⁺ cell density) represented cold tumors [5]. Tumors were additionally classified into four groups based on PD-L1 expression in tumor cells and CD15⁺ cell counts at the tumor invasive front (A–D) as presented in Fig. 3A, namely Groups A (high PD-L1 expression and low CD15⁺ cell density), B (high PD-L1 expression and high CD15⁺ cell density), C (low PD-L1 expression and low CD15⁺ cell density), and D (low PD-L1 expression and high CD15⁺ cell density).

Fig. 1

Fluoro-d-glucose–positron emission tomography (FDG PET-CT) and immunohistochemical analysis of programmed death ligand-1 (PD-L1) expression, CD8⁺ cells, and CD15⁺ cells. Row A (a 76-year-old female patient with tongue cancer, cT4a), representing group I or group A (Fig. 3A), reveals high FDG uptake, high PD-L1 expression, high CD8⁺ cell counts, and low CD15⁺ cell counts. Row B (a 23-year-old female patient with tongue cancer, cT2), representing group IV or group D, presents high FDG uptake, low PD-L1 expression, and high CD8⁺ and CD15⁺ cell counts. Row C (a 32-year-old female patient with tongue cancer, cT1), belonging to group II or group C, reveals low FDG uptake, low PD-L1 expression, and low CD8⁺ and CD15⁺ cell counts

Fig. 2

A Patients were classified into four groups (I–IV) based on programmed death ligand-1 (PD-L1) expression in tumor cells and CD8⁺ cell counts at the tumor invasive front. B Kaplan–Meier plots of relapse-free survival in the subgroups. The differences in survival between groups I (high PD-L1 expression and high density of CD8⁺ cells) and IV (low PD-L1 expression and high density of CD8⁺ cells) were statistically significant (P = 0.015). C Box-and-whisker diagrams present fluoro-d-glucose (FDG) uptake in the four subgroups. Group II (low PD-L1 expression and low density of CD8⁺ cells) had significantly lower FDG uptake (indicated by the maximum standardized uptake value (SUVmax)] than the other groups (P = 0.043). D The density of CD15⁺ cells at the tumor invasive front was not significantly different among the four groups, but comparing groups I and IV, both of which had high CD8⁺ cell counts despite having high and low PD-L1 expression, respectively, group IV tended to have a higher density of CD15⁺ cells than group I (D)

Fig. 3

A Patients were classified into four groups (A–D) based on programmed death ligand-1 (PD-L1) expression in tumor cells and CD15⁺ cell counts at the tumor invasive front. B Box-and-whisker diagrams present fluoro-d-glucose (FDG) uptake in the four subgroups. Group C (low PD-L1 expression and low density of CD15⁺ cells) had significantly lower FDG uptake (indicated by the maximum standardized uptake value [SUVmax]) than group B (P = 0.04). C Kaplan–Meier plots of relapse-free survival in the subgroups. Group C had significantly better survival than groups A and D (P = 0.005 and P = 0.009, respectively) (high PD-L1 expression and high density of CD15⁺ cells)

Statistical analysis

The χ² test was used to analyze the clinicopathological variables according to the event. In univariate and multivariate analyses, P < 0.05 denoted statistical significance. The Student t test was used to compare SUVmax in the high and low groups and to compare two of each of the four subgroups (I–IV, A–D). Multiple regression was used to analyze the relationships between SUVmax and several independent variables. Relapse-free survival by group (I–IV, A–D) was determined using the Kaplan–Meier method and log-rank test. All statistical analyses were performed using IBM SPSS version 21 software (IBM Corp., Armonk, NY, USA).

Results

Clinicopathological results and prognosis

In the clinicopathological examination of patients with and without relapse, the numbers of S100A8⁺ and CD15⁺ cells at the invasive front and SUVmax were significantly associated with the risk of relapse in univariate analysis (P = 0.019, Table 1). Specifically, patients with high CD15⁺ expression, high S100A8⁺ cell counts, and high SUVmax (≥ 9) had a significantly higher relapse rate. Multivariate analysis revealed that only CD15⁺ cell counts were associated with the risk of relapse. In Kaplan–Meier plots of relapse-free survival (Additional file 1: Figure S1), high SUVmax and high MDSC cell counts (CD15⁺ and S100A8⁺) were associated with poor prognosis. High SUVmax was also linked to poor overall survival. The counts of S100A8⁺ and CD15⁺ cells were not associated with overall survival.

Relationship between FDG-PET and clinicopathological data

The relations between clinical variables and SUVmax values of FDG-PET are shown in Table 2, and these SUVmax values were significantly associated with cT and cStage. FDG uptake tended to be higher in the high S100A8, and CD33 expression groups, whereas CD8⁺ cell counts tended to be lower in the high expression groups. Significantly higher SUVmax values were found in patients with higher PD-L1 expression and in those with higher CD15⁺ cell counts (P = 0.03 and P = 0.02, respectively) (Fig. 4). In correlation analysis, SUVmax was significantly associated with tumor size, CD15⁺ and S100A8⁺ cell counts, and PD-L1 expression. There was no significant correlation of SUVmax with CD33⁺ or CD8⁺ cell counts (Additional file 3: Table S1 and Additional file 2: Figure S2). In multiple regression analysis including all variables, tumor size had the strongest effect on SUVmax, followed by PD-L1, and when the tumor size was excluded, CD15 was included in the prediction equation (Table 3).

Table 2 Univariate analysis of clinical variables for SUVmax of FDG-PET

Fig. 4

Box-and-whisker diagrams show the distribution of the maximum standardized uptake value (SUVmax) in the high and low groups of programmed death ligand-1 (PD-L1) expression and CD8⁺, CD15⁺, S100A8⁺ and CD33⁺ cell counts. Significantly higher SUVmax, denoting higher fluoro-d-glucose uptake, was found in patients with PD-L1 expression group and in those with high CD15⁺ cell counts (P = 0.03 and P = 0.02, respectively). There was no significant difference in SUVmax with respect to the number of CD8⁺, S100A8⁺, or CD33⁺ cells

Table 3 Results of multiple regression analysis for FDG PET

Relationships between PD-L1, CD8, and FDG-PET

In Kaplan–Meier plots of relapse-free survival in subgroups fractionated on the basis of PD-L1 and CD8 (Fig. 2A), the differences in survival between groups I and IV were statistically significant (P = 0.015, Fig. 2B). Group I (high PD-L1 expression and high CD8⁺ cell density, Fig. 1A) was linked to poor prognosis, whereas Group IV (low PD-L1 expression and high CD8⁺ cell density, Fig. 1B) was linked to better prognosis. In box-and-whisker diagrams, there was no significant difference when two among the I–IV groups were compared, although Group II (low PD-L1 expression and low CD8⁺ cell density, Fig. 1C), which was a cold tumor group, had significantly lower FDG uptake than the other groups (I + III + IV) (P = 0.043, Fig. 2C). Group IV tended to have higher CD15⁺ cell counts than Group I, although statistical significance was not achieved (P = 0.059, Fig. 2D). The results shown in Fig. 2 indicate that FDG accumulation correlated with PD-L1 expression but not with that of CD8.

Relationships between PD-L1, CD15, and FDG-PET

In subgroups fractionated on the basis of PD-L1 and CD15 (Fig. 3A), high PD-L1 expression or a high CD15⁺ cell density tended to be associated with high SUVmax and poor prognosis (Fig. 3B, 3C). In particular, Group C (low PD-L1 expression and low CD15⁺ cell density, Fig. 1C) had significantly lower SUVmax than group B (P = 0.04, Fig. 3B). Group C had significantly better prognosis than groups A and D (P = 0.005 and P = 0.009, respectively) (high PD-L1 expression and high CD15⁺ cell density, Fig. 1A, 1B).

Discussion

In recent years, the number of reports related to MDSCs and the cancer microenvironment has increased, including reports in head and neck [22] and oral cancers [2, 23, 24]. In addition, some reports used CD15 [3, 25], CD33 [26], and S100A8 [27] as surrogate markers for MDSCs in immunohistochemistry. In our previous study in OSCC [3], the degree of CD15⁺ cell infiltration in the biopsy specimen was a good marker of recurrence and prognosis. Using the degree of CD15⁺ cell infiltration around cancer cells, which indicates MDSC infiltration, we found that patients with a high CD15⁺ cell density had a significantly higher recurrence rate [3]. The result represents evidence of immunosuppression in the cancer microenvironment by MDSCs.

We investigated the correlation betwen FDG uptake and PD-L1 expression and immune cell counts, including MDSCs. CD15, CD33, and S100A8 were used as surrogate markers in immunohistochemistry for MDSCs. CD15 was used as a marker of granulocytic MDSCs (G-MDSCs) in this study, although some neutrophils were also detected [25]. S100A8 is an inflammatory mediator released by neutrophils and monocytes, and it was used as surrogate marker of MDSCs including both G-MDSCs and monocytic MDSCs (M-MDSCs) [28]. In this study, the numbers of CD15⁺ and S100A8⁺ cells in the same samples were correlated, and more cells were positive for S100A8 than for CD15. Given the results that patients with a high CD15⁺ or S100A8⁺ cell density had poor prognoses, CD15 and S100A8 were considered useful markers for evaluating MDSCs in OSCC. In addition to PD-L1, the number of CD15⁺ cells in cancer tissue was well correlated with SUVmax. Concerning prognosis, CD15, S100A8, and FDG uptake were significantly related to high recurrence rates, but only CD15 was significantly associated with recurrence in multivariate analysis.

Cancer cells use glucose as an energy source, and tumors with high proliferative potential are more likely to accumulate FDG [29]. SUV, a semi-quantitative indicator of FDG accumulation, reflects the malignancy of cancer cells, such as their survival, proliferation, apoptosis, p53 overexpression, and hypoxia [30]. SUV is reported to be a predictive factor for prognosis and therapeutic response in primary head and neck cancers [31,32,33]. In multiple regression analysis in this study, tumor size was most strongly related to SUVmax, followed by PD-L1. Excluding tumor size, it was revealed that CD15 was the next most relevant factor. In oral cancer, it has been reported that SUVmax is correlated with tumor size [34]. Studies investigating the relationship between FDG uptake and PD-L1 expression in lung [17,18,19,20], gastric [10], esophageal [15], breast [16], and oral cancers have been recently reported [21]. In lung cancer, high SUVmax on FDG-PET was correlated with high PD-L1 expression [17,18,19,20]. Kuriyama et al. [15]. reported that high FDG uptake in esophageal squamous cell carcinoma was significantly associated with cancer aggressiveness, high PD-L1 expression and high CD8⁺ cell counts, denoting the so-called hot tumor immune status. In breast cancer [16], patients with high FDG uptake also displayed high PD-L1 expression and high CD8⁺ cell infiltration. These findings of a positive correlation between PD-L1 and FDG uptake aligned with our results in OSCC, although no correlation between FDG uptake and CD8 expression was found in OSCC. It is unclear why increased PD-L1 expression leads to increased FDG uptake. Intrinsically, PD-L1–expressing tumors tend to be high-grade tumors, which may result in increased FDG uptake. The uptake of FDG by tumor cells is closely related to glucose metabolism and hypoxia, and the expression of glucose transporter 1 (GLUT1), which is responsible for glucose metabolism, and HIF-1α, which responds to hypoxia, have been found to play important roles in FDG accumulation [35]. In patients with lung or kidney cancer, PD-L1 expression has been reported to be associated with GLUT1 and HIF-1α expression [36, 37]. Regarding the mechanism of FDG uptake, HIF-1α has been found to be an essential factor associated with the upregulation of GLUT1. Recent studies indicated that increased HIF-1α expression is associated with increased PD-L1 expression [38]. It has also been reported that HIF-1α binds directly to the hypoxia response element in the proximal promoter of PD-L1 and regulates its expression under hypoxia [39]. These results suggest that PD-L1 may be an alternative target of HIF-1α. Interestingly, HIF-1α has also been revealed to promote the accumulation of MDSCs, which is consistent with our findings [40]. The relationship between PD-L1 expression and FDG uptake needs to be further investigated.

In oral cancer, Togo et al. [21]. reported that cold tumors tend to exhibit high FDG uptake, contradicting our findings. In addition to differences of the evaluators, our study also included more recent cases and a slightly higher number of cases than their report. It was possible that our specimens were better preserved and that the immunohistochemistry result was more accurate because the cases were recent. Our results in this study were in good agreement with those in cancers in other organs.

No previous report examined the relationship between MDSCs in cancer tissue (including cancers of other organs) and FDG uptake. MDSCs are myeloid cells or neutrophils like cells and which means histologically high inflammatory cells infiltration in the tumor. It is speculated that the FDG uptake of the lesion in inflammatory states may be higher than the activity or proliferative potential of the tumor cells themselves. It has been reported that intra-abdominal lymph nodes that heavily infiltrated with MDSCs using S100A8/A9 as an indicator can be mistakenly detected as metastatic lymph nodes of cancer by FDG-PET [41]. In other words, tissues infiltrated by a large number of MDSCs may have higher FDG uptake. Considering that S-1, which suppresses MDSCs, had a strong antitumor effect in our previous study of OSCC, the impact of MDSCs on the tumor microenvironment in OSCC may be higher than observed in other cancers.

In subgroups fractionated on the basis of PD-L1 and CD8, Group II (low PD-L1 expression and low CD8⁺ cell density), representing cold tumors, had significantly lower SUVmax. Group I (high PD-L1 expression and high CD8⁺ cell density), representing hot tumors, had the worst poor prognosis, and this result is in good agreement with many previous studies in oral cancer, namely that PD-L1–overexpressing tumors have a poor prognosis and that PD-L1 expression is correlated with high CD8⁺ cell infiltration [42]. In Group IV (low PD-L1 expression and high CD8⁺ cell density), representing hot tumors, the reason for the low PD-L1 expression despite the high CD8⁺ cell density was considered immunosuppression by cells such as MDSCs. In support of this hypothesis, Group IV tended to have higher CD15⁺ cell counts in this study (Fig. 3D). Group IV had a significantly better prognosis than Group I in this study; however, this may be attributable to the active use of S-1, which has an inhibitory effect on MDSCs, preoperatively at our institution. We have created new subgroups fractionated on the basis of PD-L1 and CD15, the two factors in the immune microenvironment that affect FDG uptake. Patients with higher SUVmax values clearly displayed higher PD-L1 expression and/or CD15⁺ cell counts. This suggests that clinically measured FDG-PET SUVmax values in OSCC may reflect the PD-L1 levels of tumor cells and the density of MDSCs in the tumor microenvironment, in addition to the biological growth potential of the tumor. In other words, OSCC with high FDG uptake may be considered high PD-L1 and/or high density MDSCs and may be an indication for anti-PD-L1 or anti-MDSC therapies even before biopsy or resection.

Limitations

Although the number of cases in this study is considered adequate, more than half of the patients were treated with preoperative chemotherapy, and histopathologic evaluation was performed on biopsy specimens. Future evaluations of resected specimens that include a larger area or deeper tumors are desirable.

Conclusions

In OSCC, FDG uptake was affected by tumor size, tumor PD-L1 expression, and the density of CD15⁺ cells in cancer tissue. Patients with high FDG uptake may be good candidates for PD-L1 blockade immunotherapy or MDSC-suppressing therapies such as S-1, and it may be possible to assess their suitability by FDG-PET before biopsy or resection.