Relationship between SUVmax on 18F-FDG PET and PD-L1 expression in hepatocellular carcinoma

Our study was to investigate the correlation between 18F-FDG uptake in HCC and tumor PD-L1 expression in HCC, and assess the value of 18F-FDG PET/CT imaging for predicting PD-L1 expression in HCC. A total of 102 patients with confirmed HCC were included in this retrospective study. The PD-L1 expression and immune cell infiltrating of tumors were determined through immunohistochemistry staining. The SUVmax of HCC lesions were assessed using 18F-FDG PET/CT. The correlation between PD-L1 expression and the clinicopathological were evaluated by the Cox proportional hazards model and the Kaplan-Meier survival analysis. The SUVmax of HCC primary tumors was higher in patients with poorly differentiated HCC, large tumor size, portal vein tumor thrombus, lymph node and distant metastases, and death. The SUVmax of HCC are correlated with the PD-L1 expression and the number of cytotoxic T cells and M2 macrophage infiltration. PD-L1 expression was significantly correlated with tumor SUVmax, tumor differentiation, tumor size, portal vein tumor thrombosis, and patient survival status and infiltrating M2 macrophages. Further, our results confirmed that SUVmax, portal vein tumor thrombosis, and the number of infiltrating M2 macrophages were closely related to PD-L1 expression and were independent risk factors by multivariate analysis. The combined assessment of SUVmax values and the presence of portal vein tumor thrombosis by 18F-FDG PET/CT imaging can help determine PD-L1 expression in HCC. FDG uptake in HCC was positively correlated with the PD-L1 expression and the number of cytotoxic T cells and M2 macrophage infiltration. The combined use of SUVmax and portal vein tumor thrombosis by PET/CT imaging assess the PD-L1 expression better in HCC. These findings also provide a basis for clinical studies to assess the immune status of tumors by PET/CT.


Introduction
Approximately 840,000 new cases of hepatocellular carcinoma (HCC) and 780,000 deaths attributable to this disease are reported worldwide each year [1]. Patients with HCC in China account for nearly half of t the world's patients with HCC, and 300,000 to 400,000 Chinese die of HCC every year [2,3]. Many patients with HCC are diagnosed at the advanced stage of disease, precluding the possibility of surgical treatment. Traditional radiotherapy, chemotherapy, and molecular targeted therapy often yield unsatisfactory outcomes for patients with advanced liver cancer [4]. However, the recent emergence of anti-programmed death 1 (PD-1)/programmed death ligand 1 (PD-L1) immunotherapy has provided a new and effective treatment option for patients with advanced liver cancer [5].
PD-1 and PD-L1 are closely associated with cellular and humoral immunity. PD-1 is an immunosuppressive receptor, while PD-L1 is the primary ligand for PD-1 [6]. Some malignant tumors express PD-L1, including liver cancer and urological tumors, lung cancer, and melanoma. For tumors showing high PD-L1 expression, anti-PD1/PD-L1 therapy has better efficacy [7,8]. However, since not all tumor cells express PD-L1, assessment of the PD-L1 expression level in tumor cells by immunohistochemical analyses or other Xiang Zhou and Yongquan Hu contributed equally to this work.
This article is part of the Topical Collection on Oncology -General.
* Jianjun Liu jianjun_liu2018@163.com 1 3 pathological methods is essential before administration of PD-L1 inhibitor therapy [9]. Although pathological diagnosis is the gold standard for evaluating PD-L1 expression levels, assessment of each lesion with multiple biopsies when multiple lesions are present or when PD-L1 expression needs to be assessed dynamically is not feasible. Positron emission tomography/computed tomography (PET/CT) is a valuable technique for predicting tumor phenotypes, and several studies have shown its potential for predicting tumor phenotypes and indicating PD-L1 status in lung cancer and head and neck cancer [10][11][12]. However, the correlation between PD-L1 expression and the [ 18 F]fluorodeoxyglucose ( 18 F-FDG) standardized uptake value (SUV) in HCC is still rare, and the approach for predicting PD-L1 expression in HCC by using PET has not been reported.
In the present study, we retrospectively analyzed 102 patients with HCC who underwent 18 F-FDG PET/CT. PD-L1 and inflammatory cell infiltration in the tumor microenvironment were evaluated by immunohistochemical analyses. The purpose of this study was to investigate the correlation between 18 F-FDG uptake in HCC and tumor PD-L1 expression as well as inflammatory cell infiltration in HCC, and also to assess the value of 18 F-FDG PET/CT imaging for predicting PD-L1 expression in HCC.

Participants
Among the 102 patients who had HCC confirmed by pathological assessments and had undergone 18 F-FDG PET/CT at Renji Hospital of Shanghai Jiao Tong University and the First Affiliated Hospital of Bengbu Medical College from January 2020 to December 2021. Eligibility criteria for recruitment were as follows: (1) liver lesions pathologically confirmed as hepatocellular carcinoma; (2) Specimens of liver lesions were available for immunohistochemical analysis. (3) PET/CT examination and surgical pathological confirmation with an interval of ≤ 1 month. Most patients with hepatocellular carcinoma underwent biopsy, but a small percentage underwent FDG PET/CT scanning in our hospitals. All 102 patients met these criteria (Fig. 1) and provided informed consent for this study. The study was approved by the Ethics Committee of the Renji hospital and First Affiliated Hospital of Bengbu Medical College, and complied with the principles of the Declaration of Helsinki as revised in 2013.

F-FDG PET/CT examination
Biograph 64 PET/CT equipment (Siemens, Germany) was used for the imaging examinations in both of hospitals. Patients fasted for 4-6 h before the examination and had blood glucose levels below 6.3 mmol/L. The patients received an intravenous injection of 18 F-FDG (radiochemical purity, > 95%) at 5.55 MBq/kg body weight, and PET/CT was performed 60 min after the injection. The patients were then placed in the supine position. The CT scan parameters were as follows: voltage, 120 kV; current, 140 mA. The PET scan parameters were as follows: 3D model, 2 min/bed, scan bed height adjusted according to patient height (adult normal: 6), and a 128*128 matrix. PET images were acquired after CT and PET attenuation correction using ordered subset expectation maximization (OSEM: 2 iterations, 28 subsets) reconstruction to obtain tomographic images. The PET/CT fusion images were acquired on a Siemens post-processing workstation.

Image and data analysis
The images were interpreted by two experienced nuclear medicine physicians. All data were imported into IntelliSpace Workstation (IntelliSpace Portal v7.0, Philips Healthcare, The Netherlands) for processing. Lesion boundaries were automatically outlined by the TUMOR TRACE software, and lesion SUVmax was automatically calculated.

Immunohistochemical staining
Paraffin-embedded HCC tissue specimens with a section thickness of 4 μm were obtained and stained for histochemical analysis. Sections (4-mm slices) were obtained using a microtome and then stained using a Nexes automated immunostainer (Ventana Medical Systems, USA). Primary antibodies against PD-L1, CD8, FOXP3, and CD206 were obtained from Abcam (1:400). The staining intensity scores were interpreted and scored as follows: 0, 0% of cells stained; 1, 1%-9% of cells stained; 2, 10%-49% of cells stained; and 3, 50%-100% of cells stained. Immunohistochemistry (IHC) scores of 0-9 were obtained by determining the product of staining intensity and staining area. IHC scores ≥ 4 were considered to indicate high expression, and IHC scores < 4 were considered as low expression. All IHC results were evaluated by two experienced observers who were blinded to the patient's condition. In the event of disagreement between the two readers, a final assessment was arrived at by consensus.

Tumor microenvironment immune cell counts
CD8, FOXP3, and CD206 are specific markers for cytotoxic T cells, regulatory T cells, and M2 macrophages, respectively [13]. After immunohistochemical staining for the relevant cellular markers, the number of immune cells infiltrating the tumor tissue was counted by Visiopharm software (Visiopharm, Horsholm, Denmark) analysis. During microscopic evaluation of the biopsy specimens, the mean number of cells in three fields of view was calculated.

Statistical analysis
The data were analyzed using SPSS 20.0 software. All data were analyzed using non-parametric statistical methods, and P < 0.05 was considered to indicate statistical significance. The correlation between SUVmax and tumor biological indices was determined by Spearman analyses. Multivariate analysis of independent factors associated with PD-L1 expression was performed, and receiver operating characteristic (ROC) curve analysis of each index was used to predict the value of PD-L1 expression.

Correlation between clinicopathological features and 18 F-FDG uptake in PET/CT scans in patients with HCC
The 102 HCC patients (age, 31-78 years) included 71 male and 31 female patients. The patients had undergone 18 F-FDG PET/CT scans preoperatively or before biopsy.
We compared the correlations of SUVmax values in primary HCC with the pathological characteristics of HCC. Among the HCC patients, SUVmax did not differ significantly between patients under and over 60 years of age and showed no significant correlation with sex (P > 0.05). However, SUVmax was significantly correlated with the degree of differentiation, tumor size, portal vein embolization, lymph node metastasis, distant organ metastasis, and patient survival status (P < 0.05). The SUVmax of HCC primary tumors was higher in patients with poorly differentiated HCC, large tumor size, portal vein tumor thrombus, lymph node and distant metastases, and death (Table 1).
In the immunohistochemical analyses, we evaluated the tumor PD-L1 expression and the number of cytotoxic T cells, regulatory T cells, and M2 macrophages infiltrating the tumor microenvironment (Fig. 2). We further analyzed the correlation between SUVmax and tumor PD-L1 expression as well as immune cell infiltration in HCC. SUVmax was positively correlated with PD-L1 expression in HCC (r = 0.421, p < 0.001), and HCC tissues with high PD-L1 expression were more likely to show high FDG uptake (Fig. 3A, Table 2). Cytotoxic T cells, regulatory T cells, and M2 macrophages are important inflammatory cells associated with tumor immune. Our results showed no significant correlation between the SUVmax and the number of infiltrating regulatory T cells in HCC, but the number of cytotoxic T cells and infiltrating M2 macrophages was positively correlated with SUVmax (p < 0.05), and the corresponding correlation values were 0.020 and 0.001, respectively (Table 2).

Correlation between PD-L1 expression and clinicopathological features in primary HCC
We analyzed the correlation between PD-L1 expression levels in the primary lesions and the other clinicopathological features of HCC. We assessed the association of PD-L1 expression with patient age, sex, tumor differentiation, tumor size, degree of differentiation, portal vein tumor thrombosis, lymph node and distant metastasis, patient survival status, and SUVmax of the primary lesions. The results showed that PD-L1 expression was not significantly correlated with patient sex, age, and lymph node and distant metastasis. However, PD-L1 expression was significantly correlated with tumor SUVmax, tumor differentiation, tumor size, portal vein tumor thrombosis, and patient survival status. In particular, the SUVmax of primary tumors with high PD-L1 expression was significantly higher than that of primary tumors with low PD-L1 expression (P < 0.001) (Fig. 3B, Table 3). Figure 4 shows the typical images of PD-L1 IHC staining and 18 F-FDG PET/CT scans in patients with high or low PD-L1 expression. Similarly, we further analyzed the relationship between PD-L1 expression and immune cell infiltration in HCC. The  Table 4).

Multivariate analysis of PD-L1 expression level in primary HCC
Assessment of PD-L1 levels in HCC is valuable for evaluating patient prognosis and predicting the effectiveness of anti-PD1/PD-L1 therapy. Our study showed that PD-L1 expression in primary HCC was correlated with the SUVmax, degree of tumor differentiation, tumor size, portal vein tumor thrombosis, patient survival status, and the number of infiltrating M2 macrophages. Therefore, we further investigated the independent risk factors affecting PD-L1 expression by multivariate analysis. Our results confirmed that SUVmax, portal vein tumor thrombosis, and the number of infiltrating M2 macrophages were closely related to PD-L1 expression and were independent risk factors ( Table 5).

Predictors of PD-L1 expression in HCC by 18 F-FDG PET/CT imaging
SUVmax, portal vein tumor thrombosis, and the number of infiltrating M2 macrophages were independent risk factors closely related to PD-L1 expression in HCC. PD-L1 expression changes dynamically during tumor progression and treatment. Moreover assessment of macrophage infiltration by biopsy or detection of PD-L1 expression directly in metastatic tumors is difficult. Therefore, we tried to indirectly measure PD-L1 expression by 18 F-FDG PET/CT imaging. We selected SUVmax and portal vein tumor thrombus, which are two indicators that can be dynamically observed by imaging. First, we analyzed the ability of the SUVmax value alone and in combination with portal vein tumor thrombus for predicting PD-L1 expression by ROC curve analysis (Fig. 5). The sensitivity and specificity of a SUVmax cut-off value of 4.55 for predicting PD-L1 expression were 63.6% and 82.6%, respectively. When the SUVmax was combined with portal vein tumor thrombus to predict PD-L1 expression, the area under the ROC line was larger; the sensitivity for predicting PD-L1 expression was 63.6%; and the specificity increased to 87.1%. (Table 6). We also divided HCC patients into a high-probability group (SUVmax ≥ 4.55, portal vein tumor thrombosis+), medium-probability group (SUVmax ≥ 4.55, portal vein tumor thrombosis-, or SUVmax < 4.55, portal vein tumor thrombosis+), and low-probability group (SUVmax < 4.55, portal vein tumor thrombosis-) on the basis of the HCC SUVmax and the presence of portal vein cancer thrombosis. Our results showed that the probabilities of high PD-L1 expression in the high-, medium-, and low-probability groups were 87.5%, 48.7%, and 12.7%, respectively (Table 7).

Discussion
The Warburg effect is a hallmark of tumors [14]. Tumor glycolysis is regulated by various oncogenes, cancer suppressor genes, and the tumor microenvironment, so tumor glycolysis  is closely related to the biological properties of tumors [15][16][17]. Some studies have reported that the 18 F-FDG uptake in HCC correlates with tumor microvascular invasion (MVI), and SUVmax > 5.85 is the best cut-off value to differentiate MVI-positive HCC in 18F-FDG PET/CT imaging [18]. Yoh et al. demonstrated that the FDG uptake capacity of HCC was strongly correlated with survival prognosis of patients [19]. In the present study, we first evaluated the correlation between FDG uptake, based on the SUVmax value, and the pathological characteristics of patients with HCC. Our results showed that the SUVmax value was significantly correlated with the degree of differentiation, tumor size, portal vein embolism formation, lymph node metastasis, distant organ metastasis, and patient survival status in HCC (P < 0.05), consistent with the findings of many current studies.
Tumor immune escape and immunotherapy sensitivity are current research hotspots in oncology research. Tumor immune escape and immunotherapy efficacy are closely related to PD-L1 expression in tumor cells and inflammatory cell infiltration in the tumor microenvironment [20,21]. Therefore, in this study, we further compared the relationship between SUVmax and PD-L1 expression as well as inflammatory cell infiltration in HCC to explore the relationship between tumor immunity through HCC FDG uptake and tumor immunity. Cytotoxic T cells, regulator T cells, and M2 macrophages are important cells that regulate tumor immunity in the tumor microenvironment, and these inflammatory cells can take up glucose themselves, thereby increasing the tumor FDG uptake value and the SUVmax [22]. Shinji Itoh et al. found that FDG uptake in  HCC was positively correlated with CD8+ T-cell counts and CD68+ macrophage infiltration in HCC [23]. Young-Sil An demonstrated that FDG uptake in breast cancer brain metastases is correlated with macrophage infiltration [24]. Our study showed that many HCC tissues show infiltration of cytotoxic T cells and M2 macrophage cells, and the number of infiltrating cells was positively correlated with FDG uptake, but the number of regulatory T cells infiltrating HCC tissues was relatively small and did not correlate with FDG uptake. Thus, 18 F-FDG PET/CT may be useful to evaluate the status of inflammatory cell, especially cytotoxic T-cell and M2 macrophage, infiltration in HCC. PD-L1 is an important immune checkpoint and immunotherapy target, and prediction of PD-L1 expression can help guide immunotherapy. Several studies have explored the prediction of tumor PD-L1 expression status on the basis FDG uptake. Ruohua Chen et al. [25] showed that high 18 F-FDG uptake by bladder cancer is associated with elevated PD-1/ PD-L1 expression and that SUVmax = 22.7 was the best cut-off value to predict high PD-L1 expression in bladder cancer. In this study, our results also showed that the SUVmax in HCC was closely correlated with PD-L1 expression. HCC with high PD-L1 expression was usually accompanied by higher SUVmax values, suggesting that high PD-L1 expression was a malignant feature of high glucose metabolism in HCC. It may be useful to assess PD-L1 expression in HCC patients by 18 F-FDG PET/CT, and thereby guide clinical immunotherapy. However, the mechanism underlying the association between FDG uptake and PD-L1 protein expression has not been fully elucidated. One possible mechanism is that lactate-induced TAZ-dependent upregulation of PD-L1 expression in tumor tissues [26], in addition to the key genes regulating glucose metabolism, PKM2, and hypoxia factor, regulates PD-L1 expression [27,28].
The tumor PD-L1 expression status is closely related to the efficacy of immunotherapy. In order to better predict PD-L1 expression, we further analyzed the correlation between PD-L1 expression and different clinical characteristics and found that PD-L1 expression was also significantly correlated with tumor differentiation, size, portal vein tumor thrombosis, and patient survival status. Multifactorial analysis further confirmed that SUVmax, the number of infiltrating M2 cells, and portal vein tumor thrombosis were independent predictors of PD-L1 expression. In fact, several studies have demonstrated that M2 macrophages are associated with PD-L1 expression, and the mechanism may be related to the direct induction of PD-L1 expression by M2 macrophages secreting transforming growth factor (TGF)-β1 in tumor cells [29]. The molecular mechanisms underlying the correlation between portal vein tumor thrombosis and PD-L1 expression are unclear. We speculate that HCC with high PD-L1 expression may secrete more Cytokines [30,31], such as interleukins leading to targeted migration of HCC cells to the portal vein [32,33]. Nevertheless, the specific molecular mechanisms need to be further explored.
Although SUVmax, the number of infiltrating M2 cells, and portal vein thrombosis were closely correlated with PD-L1 expression in HCC, direct assessment of the number of infiltrating inflammatory cells with PET/CT or currently available imaging tools is difficult. Therefore, we selected two parameters, SUVmax and the presence of portal vein tumor thrombosis, and attempted to predict PD-L1 expression on the basis of 18 F-FDG PET/CT images alone. In the ROC analysis, the highest sensitivity  and specificity for predicting PD-L1 was observed when the SUVmax cut-off value for HCC was 4.55. When the SUVmax and portal vein tumor thrombus were combined to predict PD-L1 expression, the area under the ROC line was larger than that obtained with SUVmax alone, indicating that the combination of SUVmax and portal vein tumor thrombus could better predict PD-L1 expression in HCC than a single index. On the basis of our findings, patients with HCC could be divided into a high-probability group, medium-probability group, and low-probability group for PD-L1 expression. Our results showed that the probability of high PD-L1 expression in the high-probability group was 87.5%, while the probability of high PD-L1 expression in the low-probability group was only 12.7%. These results further suggest that the combined assessment of SUVmax values and the presence of portal vein tumor thrombosis by 18 F-FDG PET/CT imaging can help determine PD-L1 expression in HCC. In our study, we explored the correlation between FDG uptake and immune markers in HCC, and suggested how to judge PD-L1 expression in HCC by 18 F-FDG PET/CT. Nevertheless, the study had some limitations. First, due to the limited number of patients, some validation cohorts are missing, necessitating a prospective validation study to confirm the current findings. Second, FDG PET/CT examination is often recommended for hepatocellular cancer patients with suspected lymph nodes and distant metastases. The enrolled patients may have a higher percentage of advanced cancer in our study.

Conclusions
This study showed that SUVmax and portal vein tumor thrombosis were independent predictors of PD-L1 expression in HCC. Both parameters could be evaluated by PET/CT imaging, allowing non-invasive assessment of PD-L1 expression in HCC. In addition, our findings also indicated that FDG uptake in HCC was positively correlated with the number of cytotoxic T cells and M2 macrophage infiltration. These findings also provide a basis for clinical studies to assess the immune status of tumors by PET/CT.