HCC accounts for 90% of primary liver cancers and was the sixth most common cancer and the third leading cause of cancer-related death in 2020 [1]. Although a wide range of treatment options exist for HCC, it is important to select an appropriate systemic therapy to prolong the prognosis of patients with HCC when local therapy fails due to recurrence. This study revealed that patients with high blood IL-6 levels were less likely to benefit from Atezo+Bev therapy. This result is consistent with previous studies, and this study contributes significantly as a validation study [5, 6].

IL-6 plays an important role in the development and progression of HCC [12]. Serum IL-6 levels are higher in patients with HCC compared with healthy controls, and elevated serum IL-6 levels are known to be a risk factor for the development of HCC in patients with chronic liver disease, independent of other risk factors such as hepatitis virus infection [16]. Within tumors, IL-6 is known to stimulate tumor growth, survival, angiogenesis, and evasion of immune detection via IL-6/STAT3 signaling, as well as to enhance and propagate oncogenic signals within the TME, thereby promoting tumorigenesis, invasion, and metastasis [12]. However, IL-6 is also known to correlate with the malignancy potential of the tumor itself, such as in patients with advanced stage HCC or high AFP, and has been reported to contribute to shorter OS regardless of treatment [13, 14, 17].

One of the difficulties in biomarker research is that when a marker is associated with prognosis, it is unclear whether it reflects a response to treatment or a difference in the rate of tumor progression regardless of treatment. In addition, cytokines are known to differ significantly not only by tumor stage but also by etiology [16, 18]. To clarify these points, a comparison was made with a background-matched lenvatinib cohort, excluding confounding factors, which is a unique feature of this paper. Unfortunately, due to the small number of patients with high IL-6 levels in the lenvatinib group, a comparison between lenvatinib and Atezo+Bev in patients with high IL-6 levels was not possible. However, there was no difference in PFS between Atezo+Bev and lenvatinib after PSM, but when limited to patients with low IL-6 levels, PFS for Atezo+Bev was longer. These results may suggest that cases with high IL-6 are not merely more aggressive tumors, but the immune state may also weaken the effect of Atezo+Bev.

In our study, both high serum IL-6 and IL-8 levels were significantly associated with shorter PFS in univariate analysis. However, in multivariate analysis, only high IL-6 was identified as an independent factor. There was a positive correlation between serum IL-6 and IL-8 levels (p = 0.0005, r = 0.35, Supplementary Fig. 4c), suggesting that IL-6, being the stronger factor, remained as the independent variable. Therefore, we focused on IL-6 as the only factor related to PFS in the subsequent analyses. However, as the scatter plot shows, there were cases where IL-6 was low but IL-8 was high, or vice versa. Moreover, several reports suggest that patients with high blood IL-8 levels may be less likely to benefit from PD-L1 blockade in other cancers [19, 20]. IL-8 is associated with angiogenesis, cell proliferation, invasion, and migration, and is known to correlate with tumor size and grade [17, 21]. It has also been reported that IL-6 and IL-8 are secreted by neutrophils via HSP90α in the TME of HCC and suppress CD8+ T-cell activation [22]. This indicates that combining IL-6 and IL-8 could potentially provide a more accurate prediction of treatment efficacy, and this warrants further investigation in future studies. Numerous studies have revealed that IL-6 induces an immunosuppressive tumor microenvironment. IL-6 exerts its biological effects by binding to the IL-6 receptor (IL-6R). The IL-6R is a complex composed of IL-6Rα (also known as CD126) and glycoprotein 130 (GP130), which is involved in IL-6 signaling. When IL-6Rα binds to IL-6, GP130 homodimerizes, leading to the rapid activation of JAK1, JAK2, and TYK2, which subsequently phosphorylate tyrosine residues within the cytoplasmic domain of GP130. These phosphorylated sites function as docking areas for members of the STAT family of transcription factors (STAT1, STAT3, STAT5). The STAT3-stimulating activity of IL-6 has recently been associated with the suppression of T cell anti-tumor activities. The IL-6-STAT3 signaling pathway counteracts the TCR-dependent enhancement of GZMB, TNF-α, and IFN-γ expression, thereby inhibiting CTL differentiation. Furthermore, IL-6 released in the tumor microenvironment suppresses the differentiation of IFN-γ-producing TH cells by reducing MHC-II surface expression and IL-12 secretion in dendritic cells of tumor-bearing mice. Conversely, the deletion of IL-6 in tumor-bearing mice enhances the anti-tumor activities of effector T cells and inhibits tumor formation in vivo [23]. In addition, IL-6 promotes the recruitment and differentiation of immunosuppressive myeloid cells, such as MDSCs and TAMs [24]. TAMs can weaken the cytotoxic functions of CD8+ T cells and induce their apoptosis [25]. Thus, IL-6 serves as a key player in suppressing CD8+ T cells—either indirectly through the differentiation of suppressive TAMs or directly—ultimately undermining cancer immunity.

Our study showed that in cases with high serum IL-6 levels, IL-6 uptake into the tumor is similarly enhanced within TME, and the CD8/CD163 cell ratio in the tumor is reduced, which may present an immunosuppressive microenvironment. M2 macrophages are representative key players in TAM [26] and act in a tumor-promoting manner through immunosuppressive, angiogenic, and cancer cell infiltration-promoting effects [26]. M2 macrophages are known to be a poor prognostic factor in HCC [27] and have been suggested to be associated with PD-L1 overexpression [28, 29]. Regarding the relationship between M2 macrophages and IL-6, it is known that IL-6 is involved in the polarization of M0 macrophages to M2 macrophages in cancer cells of other organs [30, 31]. Furthermore, in HCC, M2 macrophages have been shown to contribute to tumor progression via the IL-6/STAT3 signaling pathway [32, 33]. Our study indicates that measurement of serum IL-6 may be able to predict the presence of an immunosuppressive TME. To our knowledge, this is the first paper to comprehensively analyze the relationship between serum IL-6 levels and the immune microenvironment and response to treatment in HCC patients treated with Atezo+Bev.

Markers such as CRP, AFP [34, 35], the CRAFITY score derived from CRP and AFP [36], and the neutrophil to lymphocyte ratio [37] have been reported as promising predictors of treatment benefit from immunotherapy in HCC. In our study as well, CRAFITY was identified as a significant marker for progression-free survival (PFS) in univariate analysis. Further investigation of their relevance, including potential combinations with IL-6, is necessary based on findings from larger multi-institutional studies.

Furthermore, early changes in AFP [38,39,40] and DCP [38, 40] have been reported to be useful in predicting the efficacy of combination immunotherapy. By effectively combining pre-treatment markers with those that assess reactivity after the initiation of treatment, we believe that further improvements in the accuracy of prognosis prediction can be achieved.

In the current study, the IL-6-high group had higher ALBI scores. In the Phase III IMbrave150 study, the response rate to Atezo+Bev tended to decrease with the progression of ALBI grade, with response rates of 32% for ALBI grade 1, 30% for grade 2a, and 25% for grade 2b [41]. Moreover, it has been reported that patients with more advanced ALBI grades prior to treatment have shorter overall survival [42, 43] and PFS [43]. Based on these findings, the poor outcomes observed in cases with high IL-6 levels may not only be due to an immunosuppressive environment but also influenced by reduced hepatic function.

Our analysis of the treatment course after progression (PD) revealed that cases with high IL-6 levels were less likely to reach later-line treatments. Among the cases that started treatment with Atezo+Bev, 10 out of 27 cases with low IL-6 levels were able to reach third-line treatment, while only 2 out of 9 cases with high IL-6 levels were able to do so. During the observation period, among the 70 cases treated with Atezo+Bev as first-line therapy, 28 patients (9 with high IL-6 levels and 19 with low IL-6 levels) died (Supplementary Fig. 7). In the high IL-6 group, the causes of death included 3 patients who died due to the progression of intrahepatic tumors and 3 patients who died from liver failure unrelated to tumor progression (both 10.7%). In contrast, in the low IL-6 group, 9 patients died due to tumor progression (32.1%), and 2 patients died from liver failure unrelated to tumor progression (7.1%), indicating a difference in the causes of death (Supplementary Table 9). These findings suggest that, in addition to the low efficacy of treatment, the decreased reserve capacity may have contributed to the poor prognosis observed in patients with high IL-6 levels.

This study was conducted at a single institution, the number of cases was limited, and sampling bias may exist because it was not possible to examine all tissues in all patients. Furthermore, the cutoff values were evaluated in a limited sample and may change as the number of cases increases. The validity of the cutoff values can be evaluated by increasing the number of cases in the future, but the cutoff values in this study do not deviate significantly compared to previous studies [5, 6]. In this study, there were few cases with high IL-6 levels in the lenvatinib group, and we have not been able to compare the efficacy of Atezo+Bev and lenvatinib in uHCC patients with high IL-6 levels. This is undeniably a limitation when discussing whether Atezo+Bev or lenvatinib is superior in patients with high IL-6 levels. Another limitation of this study is that it was conducted at a single institution. To aim for the generalization of the biomarker, broader validation across different patient populations and institutions is necessary.