Serum CXCL10 levels at the start of the second course of atezolizumab plus bevacizumab therapy predict therapeutic efficacy in patients with advanced BCLC stage C hepatocellular carcinoma: A multicenter analysis

Abstract Background & Aims Relationships of serum C‐C motif chemokine ligand 5 (CCL5) and C‐X‐C motif chemokine ligand 10 (CXCL10) levels with hot immune features have been reported in patients with hepatocellular carcinoma (HCC). Therefore, we examined the utility of their levels for predicting the efficacy of atezolizumab plus bevacizumab (Atez/Bev) in patients with HCC. Design In total, 98 patients with HCC treated with Atez/Bev were enrolled, and their initial responses were evaluated at least once via dynamic computed tomography or magnetic resonance imaging. Serum CCL5 and CXCL10 levels were assessed by enzyme‐linked immunosorbent assay before treatment and at the start of the second course of Atez/Bev therapy, and their relationships with treatment efficacy were determined. Results No analyzed factor was associated with the initial therapeutic response. Among the 56 patients with Barcelona Clinic Liver Cancer (BCLC) stage C, serum CXCL10 levels at the beginning of course two (CXCL10‐2c) tended to be higher in responders than in non‐responders in the initial evaluation, and its optimal cutoff level of 690 pg/mL could be used to stratify patients regarding overall survival (OS; high vs. low: not reached vs. 17.6 months, p = 0.034) and progression‐free survival (high vs. low: 13.6 vs. 5.1 months, p = 0.014). In multivariate analysis, high CXCL10 levels and neutrophil‐to‐lymphocyte ratios at the start of course two and Child–Pugh stage A at baseline were independent predictive factors of improved OS. Conclusions Serum CXCL10‐2c levels were predictive of Atez/Bev efficacy in patients with BCLC stage C HCC.


| INTRODUCTION
Hepatocellular carcinoma (HCC) is the sixth most common cancer globally. 1Patients with HCC have poor prognoses because of concomitant chronic liver disease, late diagnoses, and frequent recurrence or progression after treatment. 2,35][6][7][8] The combination of the immune checkpoint inhibitor (ICI) atezolizumab and the anti-vascular endothelial growth factor drug bevacizumab (Atez/Bev) was authorized in 2020 as a first-line treatment for HCC. 9 In the IMbrave150 trial, Atez/Bev therapy displayed greater efficacy than sorafenib in patients with unresectable HCC. 9,10However, prognostic biomarkers have not been fully established for patients treated with Atez/Bev for unresectable HCC.
][13] In prior research, a T cell-inflamed tumor microenvironment (TME) was linked to better responses to ICIs, whereas immune-excluded tumors were prone to ICI resistance. 14HCC with a T cell-inflamed TME comprises approximately 35% of all cases, and the profile of the inflamed class of HCC includes higher expression of C-C motif chemokine ligand 5 (CCL5), CCL4, and other cytokines involved in lymphocyte chemotaxis, such as C-X-C motif chemokine ligand 9 (CXCL9), CXCL10, and CXCL11. 15Interestingly, a previous study detected elevated plasma CXCL9 and CXCL10 levels in mice that responded to ICI therapy, and similar results were recorded in patients with melanoma. 16In addition, Hosoda et al. revealed that low serum CXCL9 levels at baseline predicted early disease progression in patients treated with Atez/Bev for unresectable HCC. 17 These results suggest that the plasma levels of CXCR3 ligands, such as CXCL9 and CXCL10, could be early biomarkers of ICI responsiveness.Furthermore, recent studies identified an association of high circulating interleukin-6 levels with poor clinical outcomes and impaired T-cell function in patients treated with Atez/Bev for unresectable HCC. 18,19 this background, we hypothesized that the levels of cytokines/chemokines related to the inflamed class of HCC could be altered by therapy and that they better reflect the immunological status of patients, highlighting their potential utility as predictors of ICI effectiveness.Among various cytokines and chemokines, serum CCL5 and CXCL10 have been linked to hot immune features, and patients with advanced HCC and high CCL5 and CXCL10 levels could experience greater benefit from ICI therapy. 15Therefore, we retrospectively investigated whether serum CCL5 and CXCL10 levels could predict Atez/Bev efficacy in patients with advanced HCC.

| Diagnosis and treatment of HCC
HCC was diagnosed on the basis of increases in αfetoprotein levels and the results of dynamic CT, MRI, and/or pathology.Patients positive for hepatitis B virus (HBV) surface antigen or hepatitis C virus (HCV) antibody and those with a history of alcohol abuse (≥60 g/day) were considered to have HCC attributable to HBV, HCV, and alcohol, respectively.The Barcelona Clinic Liver Cancer (BCLC) criteria were used for HCC staging. 20he treatment regimen consisted of 1200 mg of atezolizumab and 15 mg/kg body weight bevacizumab every 3 weeks. 9Atez/Bev therapy was discontinued following unacceptable or serious adverse events or clinical tumor progression.In certain BCLC stage A patients, Atez/Bev therapy was initiated when cardiopulmonary function prevented hepatic resection, percutaneous ablation therapy was difficult because of intrahepatic vascular effects or organs near the liver, or transarterial chemoembolization (TACE) was difficult in patients allergic to iodinated contrast agent or those with an HCC type unsuitable for TACE.

| Laboratory tests and evaluation of liver function
Hematologic and blood chemistry tests were performed using standard assays.The neutrophil-to-lymphocyte ratio (NLR) was calculated using the absolute neutrophil and lymphocyte counts in peripheral blood.We assessed liver function using the Child-Pugh classification system and albumin-bilirubin (ALBI) score, which was calculated using serum albumin and total bilirubin levels based on the following formula: ALBI score = (log10 bilirubin [μmol/L] × 0.66) + (albumin [g/L] × −0.085). 21

| Ethical standards
Each patient provided written informed consent prior to enrollment.The study protocol was approved by the Institutional Review Boards of Nagoya City University (acceptance number: 60-21-0065) and Kumamoto University (acceptance number: 2265) and implemented according to the Declaration of Helsinki.

| Chemokine assays
Serum CCL5 and CXCL10 levels were measured using commercial enzyme-linked immunosorbent assay kits per the manufacturer's protocol (R&D Systems, Minneapolis, MN, USA).

| Assessment of responses to therapy
Therapeutic response was determined using the Response Evaluation Criteria in Solid Tumors (RECIST) ver.1.1. 22Therapeutic response was initially assessed using dynamic CT or Gd-EOB-DTPA-MRI (EOB-MRI) approximately 6-9 weeks after treatment initiation and repeated every 6-9 weeks in treatment responders.The objective response rate (ORR) was defined as the sum of the complete response (CR) and partial response (PR) rates, and the disease control rate (DCR) was defined as the sum of the CR, PR, and stable disease (SD) rates.Progression-free survival (PFS) was defined as the time from the start of Atez/Bev therapy to the date of documented progression per RECIST v. 1.1 or death from any cause.If the patient received another systemic chemotherapy or he/she was lost to follow-up before progressive disease (PD) was documented, PFS was censored at the date of the last observation.Overall survival (OS) was calculated as the time from the start of Atez/Bev therapy to death or the last follow-up.

| Statistical analysis
Categorical variables were compared between the groups using Fisher's exact test, and non-categorical variables were analyzed using the Mann-Whitney U-test.Receiver operating characteristic (ROC) curve analysis was conducted, and the area under the curve (AUC) was calculated to identify the optimal serum CXCL10 levels for discriminating responders (CR + PR) and non-responders (SD + PD) using the Youden index (Youden index = sensitivity + specificity -1; range = 0-1).Cumulative OS and PFS were analyzed using the Kaplan-Meier method, and differences were assessed using the log-rank test.Multivariate analysis was performed using the stepwise Cox proportional hazard model to identify factors associated with PFS and OS.Correlation coefficients were calculated using Pearson's correlation test.All reported p-values were two-sided, and p < 0.05 denoted significance.Statistical analysis was performed using EZR (Easy R, Saitama Medical Center, Jichi Medical University, Saitama, Japan), a modified version of R commander (version 1.61). 23

| Patient characteristics stratified by the initial therapeutic response
The characteristics of all patients stratified by the initial therapeutic response are summarized in Table S1.No factors, including serum CXCL10-pre, CXCL10-2c, CCL5pre, and CCL5-2c levels, significantly differed between responders and non-responders.
Next, we compared serum CCL5 and CXCL10 levels by BCLC stage.CXCL10-pre levels did not differ between BCLC stages A/B and BCLC stage C (192 pg/mL vs. 236 pg/ mL, p = 0.278), whereas serum CXCL10-2c levels were numerically higher in patients with BCLC stage C than in those with BCLC stages A/B (308 pg/mL vs. 380 pg/mL, p = 0.143; Figure S7).CCL5-pre levels did not differ between BCLC stages A/B and BCLC stage C (19,950 pg/mL vs. 20,450 pg/ mL, p = 0.331).Meanwhile, serum CCL5-2c levels were higher in patients with BCLC stage C than in those with BCLC stages A/B (21,650 pg/mL vs. 30,500 pg/mL, p = 0.041; Figure S8).Therefore, we focused our analyses on the associations of serum CXCL10-2c and CCL5-2c levels with treatment efficacy in patients with BCLC stage C HCC.The clinical characteristics of the 56 patients with BCLC stage C HCC and their characteristics stratified by the initial therapeutic response are summarized in Table S2 and Table 2.As presented in Table 2, no factors at baseline or at the start of the second course, including serum CCL5-pre and CCL5-2c levels, were associated with the initial therapeutic response (Figure S9), but serum CXCL10-pre (288 pg/mL vs. 228 pg/ mL, p = 0.091) and CXCL10-2c levels (CXCL10-2c: 747 pg/ mL vs. 354 pg/mL, p = 0.103) were higher in responders than in non-responders.Moreover, serum CXCL10 levels greatly increased from baseline to the start of the second course in responders compared with the findings in non-responders (Figure 1).ROC curve analyses were performed to discriminate responders from non-responders at the time of the initial therapeutic response evaluation, and the diagnostic utility of serum CXCL10-pre and CXCL10-2c levels were compared.The AUCs of serum CXCL10-pre and CXCL10-2c levels were 0.667 and 0.661, respectively (Figure S10), and their optimal cutoffs were 231 (positive predictive value [PPV] = 29%, negative predictive value [NPV] = 92%, sensitivity = 0.511, specificity = 0.818) and 690 pg/mL (PPV = 39%, NPV = 89%, sensitivity = 0.756, specificity = 0.636), respectively.When we categorized the patients with BCLC stage C HCC into two groups based on serum CXCL10-2c levels, the CXCL10-2c high group (≥690 pg/mL) included a higher proportion of initial responders than the CXCL10-2c low group (<690 pg/mL, p = 0.027; Table 3).

| Factors associated with PFS and OS
To identify the factors influencing PFS and OS, we conducted univariate and multivariate analyses.Cutoffs for parameters other than serum CXCL10 levels were reported previously. 24Univariate analysis revealed that only serum CXCL10-2c levels were significantly associated with PFS (≥690 pg/mL vs. <690 pg/mL, hazard ratio [HR] = 0.39 p = 0.017; Table S5).Concerning OS, univariate analyses illustrated that NLR, the Child-Pugh score at baseline, and NLR and serum CXCL10 levels at the start of the second course were significantly associated with OS.Considering confounding factors and the results of univariate analysis, multivariate analysis was performed.
The Child-Pugh score at baseline (≥6 vs. 5, HR = 2.73; p = 0.049), NLR at the start of the second course (≥3 vs. <3, HR = 3.26; p = 0.015), and serum CXCL10-2c levels (≥690 pg/mL vs. <690 pg/mL, HR 0.23; p = 0.009) were significantly associated with OS (Table 4).As previously mentioned, serum CCL5 and CXCL10 are associated with hot immune features, and patients with advanced HCC and higher serum CCL5 and CXCL10 levels are more likely to benefit from ICI therapy. 15This study revealed the utility of serum CXCL10 levels after the introduction of Atez/Bev therapy for predicting prognosis among patients with BCLC stage C HCC.
CXCL10 is secreted by multiple cell types including monocytes, endothelial cells, fibroblasts, inflammatory cells, and tumor cells in response to interferon-γ, 25 and increased CXCL10 expression in tumor cells is important for anti-tumor T cell responses. 26The CXCR3 chemokine system, which is associated with CXCL10, plays an important role in CD8+ T cell recruitment to tumors. 27,28In a mouse study, CD8+ T cell infiltration in HCC was induced by CXCL10 expression, 29 and patients with melanoma who responded to therapy had higher plasma CXCL9 and CXCL10 levels after ICI therapy. 16Albeit without significance, serum CXCL10 levels were numerically higher at baseline and at the start of the second course and more strongly increased from baseline to the start of the second course among responders than among non-responders in the initial therapeutic response assessment of patients with BCLC stage C HCC (Figure 1).Based on these results, it is possible that the induction of CXCL10 promoted the infiltration of CD8+ T cells into the tumor and resulted in good therapeutic efficacy.
Our study identified increased serum CXCL10 levels during Atez/Bev therapy as a predictor of therapeutic efficacy in patients with BCLC stage C HCC, but these findings were not replicated in patients with BCLC stage A or B HCC.A previous study of the immune microenvironment in HCC revealed that the immune-high subtype, which is characterized by increased B−/plasma-cell and T-cell infiltration, was associated with poorly differentiated HCC and PD-L1 expression in both tumor and immune cells. 30t has been reported that inflamed (immune-hot) HCC has a better response to ICI therapy. 31It is assumed that BCLC stage C HCC featuring distant metastasis and/or vascular invasion carries a higher risk of high-grade malignancy.From these findings, we speculate that the induction of CXCL10 by ICI therapy differed in BCLC stage C patients, which might have influenced treatment efficacy.
In this study, serum CXCL10-2c levels were independently associated with PFS and OS in patients with BCLC stage C HCC (Table 4 and Table S5).In addition, the Child-Pugh score at baseline and NLR at the start of the second course of therapy were independently associated with OS in this subgroup of patients.Liver function has been linked to the efficacy of Atez/Bev therapy. 324][35] This is the first study, to the best of our knowledge, to reveal an association between serum CXCL10 levels during Atez/Bev therapy and therapeutic efficacy in patients with unresectable HCC.
Contrarily, serum CCL5 levels were not predictive of the efficacy of Atez/Bev.CCR5, the receptor for the chemokine CCL5, has tumor-suppressing and tumor-promoting roles. 36,37Regarding tumor progression, the CCR5/CCL5 interaction is involved in the activation of the Akt pathway, which is related to the development of HCC. 38Conversely, in terms of anti-tumor activity, CCR5 expression on CD8+ T cells is necessary for CD8+ T cell activation and migration to tumor sites. 36CCL5 itself attracts conventional type 1 dendritic cells to the tumor and promotes T cell infiltration into the tumor. 39Interestingly, our study revealed a correlation between baseline serum CCL5 levels and the neutrophil count (Table S3).Neutrophils are involved in the production of ligands that induce tumor cell proliferation and invasion and cytokines that induce angiogenesis. 40eanwhile, serum CCL5-2c levels were correlated with the lymphocyte count (Table S3).Lymphocytes are responsible for the immune function of the host, and decreased lymphocyte counts can impair hosts' anti-tumor immunity and worsen their prognosis. 41Thus, serum CCL5 levels were correlated with both tumor-promoting markers at baseline and anti-tumor markers after the introduction of Atez/Bev therapy.Therefore, we presume that these functions could explain the lack of an association between serum CCL5 levels and treatment efficacy in this study.
Our study had several important limitations.First, as a multicenter, retrospective, observational study, the possibility of selection bias cannot be excluded.Second, the cohort was not large.Additional studies in larger patient populations are needed to confirm the association between serum CXCL10 levels and Atez/Bev efficacy.Third, this study had an insufficient median follow-up period (13.4 months).Fourth, we analyzed patients treated with Atez/Bev without stratification by treatment line.Finally, the efficacy of treatment regimens after Atez/Bev therapy, which can affect OS, was not examined.
In conclusion, this study demonstrated the utility of serum CXCL10 levels after the start of treatment predicting the efficacy of Atez/Bev therapy and patient prognosis in HCC.Future well-designed prospective studies with large numbers of patients are desirable.

2 | METHODS 2 . 1 |
Patients and study design This multicenter, retrospective, observational study included 98 patients with HCC.All patients received Atez/ Bev therapy irrespective of the line of systemic chemotherapy, and their initial responses were evaluated by dynamic computed tomography (CT) or magnetic resonance imaging (MRI) at least once between October 2020 and January 2023 at seven Japanese institutions (Nagoya City University Hospital [n = 20], Gifu Prefectural Tajimi Hospital [n = 8], Toyokawa City Hospital [n = 6], Kasugai Municipal Hospital [n = 5], Nagoya City University West Medical Center [n = 5], Japanese Red Cross Aichi Medical Center Nagoya Daini Hospital [n = 2], Kumamoto University Hospital [n = 52]).Serum samples were obtained at baseline and the beginning of the second course of Atez/Bev therapy and stored at −80°C.

3. 3 |
Appropriate serum CXCL10 for predicting therapeutic response in patients with BCLC stage C HCC

3. 5 |
FigureS11presents the OS and PFS curves of patients with BCLC stage C HCC stratified by the serum CXCL10pre (231 pg/mL) and CXCL10-2c cutoffs (690 pg/ mL).Serum CXCL10-pre levels were not predictive of

F I G U R E 1
Changes of serum CXCL10 levels from baseline to the start of the second course stratified by the initial therapeutic response in patients with BCLC stage C HCC. Changes of serum CXCL10 levels in responders (CR + PR, n = 11) and non-responders (SD + PD, n = 45) from baseline to the start of the second course of Atez/Bev therapy.The dotted lines represent the median serum CXCL10 level of responders with BCLC stage C HCC, whereas solid lines represent the median serum CXCL10 levels of non-responders with BCLC stage C HCC.The horizontal lines represent the interquartile range of the data.p-values were calculated using the Mann-Whitney U-test.Atez/Bev, atezolizumab plus bevacizumab; BCLC, Barcelona Clinic Liver Cancer; CR, complete response; CXCL10, C-X-C motif chemokine ligand 10; HCC, hepatocellular carcinoma; PD, progressive disease; PR, partial response; SD, stable disease.

F
I G U R E 2 OS and PFS stratified by serum CXCL10 levels at the start of the second course of Atez/Bev therapy.(A) OS stratified by serum CXCL10 levels at the start of the second course.The dotted line represents OS among patients with serum CXCL10-2c levels <690 pg/ mL.The solid line represents OS among patients with CXCL10-2c levels ≥690 pg/mL.(B) PFS stratified by serum CXCL10 levels at the start of the second course.The dotted line represents PFS among patients with serum CXCL10-2c levels <690 pg/mL.The solid line represents PFS among patients with serum CXCL10-2c levels ≥690 pg/mL.p-values were calculated using the log-rank test.Atez/Bev, atezolizumab plus bevacizumab; CXCL10, C-X-C motif chemokine ligand 10; CXCL10-2c, C-X-C motif chemokine ligand 10 at the start of the second course; OS, overall survival; PFS, progression-free survival.
Clinical characteristics of the study patients.
Comparison of the characteristics of responders and non-responders at the initial evaluation among patients with Barcelona Clinic Liver Cancer stage C hepatocellular carcinoma.
Comparison of the characteristics of patients with Barcelona Clinic Liver Cancer stage C hepatocellular carcinoma stratified by serum CXCL10 levels at the start of the second course.
T A B L E 3 Note: Hazard ratios were calculated using the Cox proportional hazard method.Abbreviations: AFP, αfetoprotein; CI, confidence interval; CXCL10, C-X-C motif chemokine ligand 10; HR, hazard ratio; NLR, neutrophil-to-lymphocyte ratio; OS, overall survival.*p < 0.05.**p < 0.005.Factors associated with OS in patients with Barcelona Clinic Liver Cancer stage C hepatocellular carcinoma who received atezolizumab plus bevacizumab.
T A B L E 4