Dynamic Peripheral T-Cell Analysis Identifies On-Treatment Prognostic Biomarkers of Atezolizumab plus Bevacizumab in Hepatocellular Carcinoma

Introduction

Hepatocellular carcinoma (HCC) is the sixth most prevalent cancer worldwide and the third leading cause of cancer-related mortality [1]. Although surgical and locoregional treatment approaches are feasible in certain instances, systemic therapies are the preferred option for approximately 50–60% of patients [2]. Multi-targeted tyrosine kinase inhibitors, such as sorafenib and lenvatinib, targeting a variety of molecular pathways, have been applied as first-line treatments [3]; however, recent breakthroughs have initiated a new era in systemic therapies, with immune-checkpoint inhibitors (ICIs) showing effectiveness in treating patients with HCC [4]. The IMbrave150 trial, which demonstrated the superiority of a combination of atezolizumab and bevacizumab (AB) over sorafenib in terms of progression-free survival (PFS), overall survival (OS), and objective response rate (ORR), introduced a milestone in the systemic treatment of HCC [5].

However, some questions remain unanswered because a notable proportion of patients do not show a response to AB: the regimen has an ORR of 27.3% and a median PFS of 6.6 months, which require improvement [5]. This emphasizes the urgent need to identify biomarkers that can inform both the initial treatment strategy and ongoing treatment decisions for patients with HCC undergoing AB treatment. Although several predictive or prognostic biomarkers of the AB regimen, including the etiology of chronic liver disease, tumor markers/stages, and liver function, have been reported [6], they are not applicable in clinical practice.

Biomarkers can also be identified by investigating tumor or blood samples because the AB regimen targets programmed cell death-ligand 1 (PD-L1) and vascular endothelial growth factor (VEGF), which are important molecules for T-cell exhaustion and angiogenesis in the HCC tumor microenvironment (TME). Use of blood samples rather than tissue samples for biomarker identification might be beneficial because liver biopsy is not feasible for a considerable proportion of HCC patients due to the possibility of procedure-related complications. Additionally, use of blood samples is more feasible in terms of repetitive sampling, allowing serial analyses to be performed in experimental, clinical, and translational studies.

T cells expressing programmed cell death receptor-1 (PD-1) are the major targets of ICI treatment [7]. T cells communicate with immunosuppressive cell populations, such as tumor-associated macrophages or regulatory T cells (Tregs) in the TME [8], suggesting that they play a critical role in anti-tumor immune responses and the efficacy of ICI treatment. Immunohistochemistry and transcriptome analyses using HCC tumor tissue samples revealed that higher frequencies of total or CD8⁺ T cells within tumor tissues may be associated with favorable ORR and PFS in AB treatment of HCC [9].

Dynamic biomarkers evaluated during the early stages of treatment have recently been proposed to have superior predictive capabilities as compared with that of static biomarkers, which are measured only at the beginning of ICI treatment [10]. In fact, a recent study clearly showed that early peripheral T-cell proliferation predicts tumor T-cell infiltration and the efficacy of anti-PD-L1 treatment [11]; however, data on the baseline status and dynamic changes in the peripheral T-cell population following ICI treatment in HCC are lacking.

In the present study, we performed T-cell analysis using peripheral blood mononuclear cells (PBMCs), prospectively collected from AB-treated patients with HCC at baseline and immediately before the second dose (week 3). By scrutinizing dynamic T-cell changes, we identified biomarkers predicting clinical outcomes, including immune-related adverse events (irAEs). This study provides initial evidence demonstrating how T cells respond to AB treatment and their potential utility as biomarkers of response to AB treatment in patients with HCC.

Materials and Methods

Study Population

We prospectively enrolled 65 consecutive patients with HCC who underwent AB treatment at four medical centers in Korea. This study followed the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Seoul St. Mary’s Hospital (approval number: KC22EASI0342), Incheon St. Mary’s Hospital (approval number: OC22TISI0090), Eunpyeong St. Mary’s Hospital (approval number: PC19OESI0017), and Uijeongbu St. Mary’s Hospital (approval number: UC23TASI0008). Informed consents from all study patients were acquired.

Serial PBMCs were collected before treatment and at 3 weeks post-initial AB dose, prior to administration of the second dose. Peripheral PBMCs were isolated using Ficoll-Paque (GE Healthcare, Chicago, IL, USA) and preserved in a mixture of fetal bovine serum and 10% dimethyl sulfoxide (Sigma-Aldrich, St Louis, MO, USA).

Flow Cytometry

Stored PBMCs were thawed and labeled with surface marker antibodies for 10 min at room temperature. For intracellular staining, cells were fixed and permeabilized using the FoxP3 Staining Buffer Kit (Invitrogen, Waltham, MA, USA) according to the manufacturer’s protocol, and were then stained with antibodies for intracellular markers for 30 min at 4°C. Antibodies used for flow cytometry were the following: BV421-conjugated anti-human CD14 (M5E2, BD Biosciences, Franklin Lakes, NJ, USA), BV421-conjugated anti-human CD19 (HIB19, BioLegend, San Diego, CA, USA), BV510-conjugated anti-human CD25 (BC96, BioLegend), PerCP-Cy5.5-conjugated anti-human CD3 (UCHT-1, BD Biosciences), PE-Cy7-conjugated anti-human CD45RA (HI100, BD Biosciences), APC-H7-conjugated anti-human CD4 (SK3, BD Biosciences), APC-conjugated anti-human PD-1 (EH12.2H7, BioLegend), V500-conjugated anti-human CD3 (UCHT-1, BD Biosciences), PerCP-Cy5.5-conjugated anti-human CD8 (SK1, BD Biosciences), PE-conjugated anti-human TIGIT (A15153G, BioLegend), PE-Cy7-conjugated anti-human CD4 (SK3, BD Biosciences), PE-Texas Red-conjugated anti-human CD14 (61D3, Invitrogen), PE-Texas Red-conjugated anti-human CD19 (HIB19, Invitrogen), PE-conjugated anti-human CD38 (HB-7, BioLegend), Alexa Fluor 700-conjugated anti-human CD4 (RPA-T4, BioLegend), PE-Cy7-conjugated anti-human Ki-67 (B56, BD Biosciences), V450-conjugated anti-human Granzyme B (GB11, BD Biosciences), FITC-conjugated anti-human Perforin (δG9, BD Biosciences), and PE-conjugated anti-human FoxP3 (PCH101, Invitrogen), PE-conjugated anti-human TCF-1 (7F11A10, BioLegend), BV421-conjugated anti-human CCR7 (M5E2, BioLegend), and APC-H7-conjugated anti-human HLA-DR (G46-6, BD Biosciences). Dead cells were excluded using the Live/Dead™ Fixable Violet Dead Cell Stain Kit or the Live/Dead™ Fixable Red Dead Cell Stain Kit (Invitrogen). Multicolor flow cytometry was performed using an LSRFortessa flow cytometer (BD Biosciences). Data analysis was conducted using FlowJo software (TreeStar Inc., Ashland, OR, USA).

Assessment of Clinical Outcomes

Radiological and laboratory data were collected during enrollment. Regular imaging was conducted every 3–4 cycles of AB to monitor treatment effectiveness using the modified Response Evaluation Criteria in Solid Tumors [12]. OS and PFS were tracked from the start of treatment to death, last follow-up, or disease progression. The ORR combined “complete” and “partial” responses, while the disease control rate included complete response, partial response, and stable disease.

Statistical Analysis

Continuous variables are represented by mean ± standard deviation, while categorical variables are shown as number (%). Student’s t tests and χ² tests were used to compare continuous and categorical variables, respectively, and Bonferroni correction was also performed as a p value adjustment. Cox regression analysis was used to identify the factors related to OS or PFS. Variables with a p value <0.01 in univariate analyses were included in multivariate analyses. The hazard ratio (HR) and 95% confidence intervals (CIs) were calculated using Cox regression analysis. Kaplan-Meier curves were used to assess OS and PFS across the different groups. Statistical significance was set at p < 0.05. The optimal cutoffs were determined using the area under the receiver operating characteristic curves and Youden’s index. All analyses were conducted using R statistical software version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) or GraphPad Prism version 8 (GraphPad Inc., La Jolla, CA, USA).

Results

CD8⁺ T Cells Show a Significant Response to AB Treatment at 3 Weeks after the Initial Dose

The baseline characteristics of the 65 patients enrolled in this study are presented in online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000541181). The mean age was 64.5 years, and 10.8% of patients exhibited Child-Pugh class B liver function. Portal vein invasion was present in 49.2% of patients, while extrahepatic metastases were observed in 61.5% of patients. An objective response (OR) was achieved in 24.6% of patients, while 61.5% demonstrated disease control (DC). Additionally, 23.1% of the patients experienced irAEs.

We first examined changes in the relative frequencies of CD4⁺ T cells, CD8⁺ T cells, and Treg cells. The gating strategies for these populations are shown in online supplementary Figure S1. We found a significant increase in the relative frequency of CD8⁺ T cells, but no change was noted in the frequencies of total T cells, CD4⁺ T cells, or Tregs (Fig. 1a–c). Additionally, the activated Treg population, which is primarily responsible for the suppressive function of FoxP3⁺CD4⁺ T cells [13], remained unchanged (Fig. 1c). Further analysis of the phenotypic changes in CD4⁺ and CD8⁺ T cells post-AB treatment revealed no alterations in the frequency of CD38⁺ or HLA-DR⁺ cells among CD4⁺ T cells (Fig. 1d). In contrast, these populations were significantly increased within CD8⁺ T cells (Fig. 1e, f), indicating substantial activation of CD8⁺ T cells by AB treatment, in contrast to CD4⁺ T cells. Moreover, the expression of immune-checkpoint molecules, including PD-1 and T-cell immunoreceptors with Ig and ITIM domains (TIGIT), was significantly increased in CD8⁺ T cells after AB treatment, whereas no such increase was observed in CD4⁺ T cells (Fig. 1d–f). These results suggested that CD8⁺ T cells are the cell population that significantly responds to AB treatment, unlike CD4⁺ T cells or Tregs.

Fig. 1.

Dynamic changes in frequencies and phenotypes of T-cell populations. PBMCs, collected from 65 patients before and week 3, were analyzed by flow cytometry and compared. a Frequencies of total CD3⁺, CD4⁺, and CD8⁺ T cells, (b) representative plots of CD4⁺ and CD8⁺ T cells, (c) frequencies of regulatory T cells (Tregs) and activated Tregs were compared between pretreatment and week 3. d–f CD38⁺, HLA-DR⁺, PD-1⁺, and TIGIT⁺ cell populations among CD4⁺ T cells (d) and CD8⁺ T cells (e) were compared between the two time-points, and the representative plots of CD8⁺ T cells are presented (f). n.s., not significant; *p < 0.05; **p < 0.01; ****p < 0.0001.

CD8⁺ T Cells Show a Proliferation along with an Activation following AB Treatment

To investigate the dynamic changes in the functional aspects of T cells after AB treatment, we examined the expression of the cytotoxic molecules granzyme B and perforin and the proliferation marker Ki-67 in CD4⁺ and CD8⁺ T cells. Consistent with the changes in phenotype and relative frequencies, CD4⁺ T cells did not show any changes in the frequencies of granzyme B⁺, perforin⁺, or Ki-67⁺ cells after AB treatment (Fig. 2a). Furthermore, granzyme B⁺ and perforin⁺ cell populations within CD8⁺ T cells were not significantly changed (Fig. 2b), suggesting that the cytotoxic capacity of T-cell populations at week 3 might not be affected by AB treatment.

Fig. 2.

Dynamic changes in cytotoxic and proliferative capacities of T-cell populations. PBMCs, collected from 65 patients before and week 3, were analyzed by flow cytometry and compared. a–c Granzyme B⁺, perforin⁺, and Ki-67⁺ cell populations were compared between two time-points in CD4⁺ T cells (a) and CD8⁺ T cells (b), and the representative plots for Ki-67⁺ cells among CD8⁺ T cells are presented (c). Correlations between the dynamic changes of the frequency of CD8⁺ T cells (d), or CD38⁺/HLA-DR⁺/PD-1⁺/TIGIT⁺ cells among CD8⁺ T cells (e) and Ki-67⁺ cells among CD8⁺ T cells were analyzed. n.s., not significant; ****p < 0.0001.

We then examined the proliferation marker Ki-67 on CD8⁺ T cells at each time-point because the relative frequency of CD8⁺ T cells was increased at week 3, as shown in Figure 1. The frequencies of Ki-67⁺ cells within CD8⁺ T cells had increased (Fig. 2b and c). This increase in the Ki-67⁺ population correlated with the relative increase in CD8⁺ T cell frequency (Fig. 2d), suggesting that the increase in the relative frequency of CD8⁺ T cells might be a result of CD8⁺ T cell proliferation after AB treatment.

Additionally, changes in Ki-67⁺ cells within CD8⁺ T cells were also associated with changes in CD38⁺/HLA-DR⁺/PD-1⁺/TIGIT⁺ cell populations of CD8⁺ T cells (Fig. 2e), indicating that the CD8⁺ T-cell proliferation might be accompanied by T-cell activation, and that the PD-1 and TIGIT upregulation noted at week 3 might be associated with T-cell activation, as previously described [14]. We further performed correlation analyses between Ki-67⁺CD8⁺ T cells and CD8⁺ T cells with the negativity of CD38, HLA-DR, PD-1, and TIGIT. At baseline, only CD38 and HLA-DR negativities inversely correlated with the Ki-67⁺CD8⁺ T cells suggesting these activation markers, but not exhaustion markers, are associated with the pretreatment proliferation of CD8⁺ T cells. However, at week 3, PD-1 negativity was also showed inverse correlation, suggesting CD8⁺ T-cell proliferation might be occurred along with the PD-1 upregulation, which might be associated with the T-cell activation (online suppl. Fig. S2). When we performed Bonferroni corrections, all parameters representing dynamic changes of CD8⁺ T-cell frequency and its phenotypes also showed statistical significance (online suppl. Table S2).

Within CD8⁺ T Cells, the PD-1⁺ Population Show a Selective Response to AB Treatment, Compared to the PD-1⁻Population

Next, we hypothesized that PD-1⁺CD8⁺ cells, rather than PD-1⁻CD8⁺ cells, might react to AB treatment because atezolizumab mainly targets the PD-1–PD-L1 pathway to restore anti-tumor CD8⁺ T-cell responses [7]. When we investigated the CD38-and HLA-DR-positivity of PD-1⁺ and PD-1⁻populations among CD8⁺ T cells, both populations showed a similar increase in CD38⁺ and HLA-DR⁺ cells after AB treatment (Fig. 3a, b). However, when we investigated the Ki-67- and TIGIT-positivity on PD-1⁺ and PD-1⁻ populations among CD8⁺ T cells, the frequencies of Ki-67⁺ and TIGIT⁺ cells within the PD-1⁺ population were significantly increased after AB treatment, whereas the frequencies of these cells did not increase within the PD-1⁻ population (Fig. 3c–e). These findings suggest that the upregulation of Ki-67 and TIGIT may represent the selective activation of PD-1⁺CD8⁺ T cells by AB treatment. On the other hand, we further analyzed the correlations between Ki-67⁺PD-1⁺CD8⁺ T cells and TIGIT⁺PD-1⁺CD8⁺ T cells at baseline and week 3, but no significant correlations were observed (online suppl. Fig. S3).

Fig. 3.

Comparison of dynamic changes between PD-1⁻CD8⁺ T cells and PD-1⁺CD8⁺ T cells. a–d Frequencies of CD38⁺ cells (a), HLA-DR⁺ cells (b), Ki-67⁺ cells (c), and TIGIT⁺ cells (d) were measured and compared before treatment and at week 3 after the initial AB treatment among PD-1⁺ (left) and PD-1⁻ (middle) subpopulations. Dynamic changes in those phenotypes were compared between changes between PD-1⁻CD8⁺ T cells and PD-1⁺CD8⁺ T cells (right). e Representative plots for Ki-67⁺ and TIGIT⁺ cells among PD-1⁻CD8⁺ T cells and PD-1⁺CD8⁺ T cells at pretreatment and week 3 are presented. n.s., not significant; *p < 0.05; **p < 0.01; ****p < 0.0001.

Baseline Differentiation Status of PD-1⁺CD8⁺ T Cells Is Associated with CD8⁺ T-Cell Proliferation and TIGIT Upregulation after AB Treatment

We next hypothesized that the baseline differentiation status of PD-1⁺CD8⁺ T cells might be associated with proliferation induced by AB treatment. Therefore, we investigated the central memory cell (CM), effector memory cell (EM), and EM-expressing CD45RA (EMRA) subpopulations, as well as T-bet or TCF-1 expression among PD-1⁺CD8⁺ T cells. We first examined frequencies of PD-1 positivity in CM, EM, and EMRA populations, and EM population showed higher PD-1 expression than CM or EMRA, although the difference between CM and EMRA was not significant (Fig. 4a). We next observed that the baseline frequency of the CM subpopulation of PD-1⁺CD8⁺ T cells correlated significantly with the dynamic increase in Ki-67⁺ cells on CD8⁺ T cells after AB treatment (Fig. 4b and online suppl. Table S3).

Fig. 4.

Differentiation status of PD-1⁺CD8⁺ T cells at pretreatment and its association with dynamic changes of CD8⁺ T cells. a PD-1 expression on CD8⁺ central memory (CM), effector memory (EM), and effector memory expressing CD45RA (EMRA) cells at baseline. b Correlations between pretreatment CM cells and dynamic changes of Ki-67⁺ cells among PD-1⁺CD8⁺ T cells were analyzed (n = 14). c Representative plots for baseline memory T-cell populations among PD-1⁺CD8⁺ T cells according to CCR7 and CD45RA expression in expanders (n = 9) and non-expanders (n = 5), as well as a comparative graph are presented. d Correlations between CM cells among PD-1⁺CD8⁺ T cells and pretreatment PD-1⁺ cells among CD8⁺ T cells (left) or TCF-1 expression on PD-1⁺CD8⁺ T cells (right). e Correlations between pretreatment EMRA cells among PD-1⁺CD8⁺ T cells and dynamic changes in TIGIT⁺ cells among CD8⁺ T cells were analyzed (n = 14). f Representative plots for baseline memory T-cell populations among PD-1⁺CD8⁺ T cells in ΔTIGIT <6.4 (n = 5) and ΔTIGIT ≥6.4 (n = 9) groups, as well as a comparative graph are presented. **p < 0.01.

We next divided patients into non-expander (≤3) and expander (>3) groups according to the change in Ki-67⁺ frequency within PD-1⁺CD8⁺ T cells. Interestingly, the baseline CM frequency within PD-1⁺CD8⁺ T cells was significantly higher in the expander group than that in the non-expander group (Fig. 4c). We further found a negative correlation between baseline CM cells within PD-1⁺CD8⁺ T cells and PD-1⁺ cells within CD8⁺ T cells, as well as a positive correlation between baseline CM cells within PD-1⁺CD8⁺ T cells and TCF-1 expression of PD-1⁺CD8⁺ T cells (Fig. 4d). These findings suggested a link between T-cell exhaustion and progenitor cell populations in exhausted T cells.

Additionally, we questioned whether the dynamic change in TIGIT⁺ cells within CD8⁺ T cells correlated with the baseline differentiation status of PD-1⁺CD8⁺ T cells. We found a negative correlation between the frequency of the EMRA population within PD-1⁺CD8⁺ T cells and the dynamic changes in TIGIT⁺CD8⁺ T cells (Fig. 4e and online suppl. Table S3). Cox regression analysis identified that the cutoff point of ΔTIGIT was 6.4, and that point was only significant factor among various cutoff points (online suppl. Table S4). When we divided patients into ΔTIGIT <6.4 and ΔTIGIT ≥6.4 groups, the frequency of EMRA cells among PD-1⁺CD8⁺ T cells was significantly higher in the ΔTIGIT <6.4 group than in the ΔTIGIT ≥6.4 group (Fig. 4f). Collectively, these findings suggested that the baseline differentiation status of PD-1⁺CD8⁺ T cells is associated with CD8⁺ T-cell proliferation and TIGIT upregulation after AB treatment.

Dynamic Changes in Ki-67⁺/PD-1⁺CD8⁺ T Cells and TIGIT⁺/CD8⁺ T Cells after AB Treatment Predict Clinical Outcomes

To determine the clinical relevance of our observations, we first investigated whether there were differences in each clinical parameter between the non-expander and expander groups. We found that patients with Child-Pugh class 5A were more frequently observed in the expander group (online suppl. Table S5). This finding suggests that T-cell proliferation, induced by AB treatment, can be influenced by baseline liver function. There were not correlations between dynamic changes of peripheral between Ki-67⁺PD-1⁺CD8⁺ T cells or TIGIT⁺CD8⁺ T cells and bevacizumab dosage (online suppl. Fig. S4).

Importantly, the expander showed higher ORR compared to the non-expander (15/43, 34.9% vs. 1/22, 4.5%, p = 0.017, online suppl. Table S5; Fig. 5a), whereas the ΔTIGIT ≥6.4 group showed the higher percentage of future development of irAEs (11/27, 40.7% vs. 6/38, 15.8%, p = 0.049, online suppl. Table S5). When we compared the baseline frequency or phenotypes of CD8⁺ T cells and their dynamic changes according to clinical responses, we did not find any differences between the non-OR and OR groups or between the non-DC and DC groups (online suppl. Table S6). However, when we divided patients into non-irAE and irAE groups, we found a greater increase in TIGIT⁺ cells among CD8⁺ T cells in the irAE group, suggesting that TIGIT⁺CD8⁺ cells might play a role in prevention or development of irAEs, although further investigation is required.

Fig. 5.

Differences in clinical outcomes according to the dynamic changes in CD8⁺ T cells. a Objective response rate (ORR), Kaplan-Meier curves comparing progression-free survival (PFS), and overall survival (OS) between expanders (n = 26) and non-expanders (n = 10) are presented. b Kaplan-Meier curves comparing PFS and OS between ΔTIGIT <6.4 (n = 23) and ΔTIGIT ≥6.4 (n = 13) groups are presented.

We then evaluated whether there were differences in PFS and OS between the expander and non-expander groups. PFS and OS were significantly longer in the expander group than in the non-expander group (HR 3.07 and 4.03) (Fig. 5a). In multivariate analysis, the increase in Ki-67⁺ cells within PD-1⁺CD8⁺ T cells week 3 AB treatment, together with the alpha-fetoprotein (AFP) level and Eastern Cooperative Oncology Group (ECOG) was an independent factor for PFS (Table 1). Furthermore, when we investigated whether PFS and OS differed between ΔTIGIT <6.4 and ΔTIGIT ≥6.4 groups, we found that PFS and OS was tended to be longer in the ΔTIGIT ≥6.4 than in the ΔTIGIT<6.4 group (HR 2.51 and 3.01, Fig. 5b). In the multivariate analysis, the increase in TIGIT⁺ cells within CD8⁺ T cells was also an independent factor, together with the Child-Pugh score and ECOG, for OS (Table 1). We performed univariate Cox regression analyses using baseline T-cell frequencies and phenotypes among CD4⁺ and CD8⁺ T cells, but there was no factor associated with the OS and PFS (online suppl. Table S7). Baseline PD-1⁺CD8⁺ CM or EMRA cells they also were not statistically significant. When we investigated baseline and dynamic changes of myeloid populations, they were not associated with any clinical outcomes (online suppl. Table S8).

Table 1.

Univariate and multivariate analyses regarding factors associated with the PFS and OS

PFS	Univariate			Multivariate
	HR	95% CI	p value	HR	95% CI	p value
Gender	1.195	0.357–3.999	0.772			n.s.
Age	0.969	0.935–1.005	0.095			n.s.
Viral etiology	1.903	0.83–4.364	0.129			n.s.
AFP	1.000	1.000–1.000	0.004	1.000	1.000–1.000	0.006
PIVKA-II	1.000	1.000–1.000	0.762			n.s.
Child-Pugh score	1.828	1.102–3.033	0.020			n.s.
Largest intrahepatic tumor size	1.029	0.966–1.097	0.371			n.s.
Multiple intrahepatic tumor	1.799	0.762–4.245	0.180			n.s.
Portal vein invasion	0.809	0.378–1.730	0.585			n.s.
ECOG	1.926	1.057–3.508	0.032	3.376	1.653-6.895	0.001
Extrahepatic spread	1.288	0.593–2.797	0.522			n.s.
irAEs	0.427	0.161–1.134	0.088			n.s.
Increase of Ki-67+/PD-1+CD8+	0.310	0.145–0.666	0.003	0.390	0.171-0.892	0.026
Increase of TIGIT+ / CD8+	0.394	0.167–0.930	0.033			n.s.

OS	Univariate			Multivariate
	HR	95% CI	p value	HR	95% CI	p value
Gender	2.056	0.269–15.731	0.487			n.s.
Age	0.978	0.926–1.033	0.428			n.s.
Viral etiology	1.682	0.527–5.367	0.380			n.s.
AFP	1.000	1.000–1.000	0.097			n.s.
PIVKA-II	1.000	1.000–1.000	0.779			n.s.
Child-Pugh score	3.398	1.664–6.939	0.001	3.573	1.636-7.804	0.001
Largest intrahepatic tumor size	1.032	0.945–1.127	0.484			n.s.
Multiple intrahepatic tumor	1.301	0.408–4.151	0.657			n.s.
Portal vein invasion	0.252	0.07–0.907	0.035			n.s.
ECOG	2.265	1.03–4.984	0.042	2.716	1.087–6.785	0.032
Extrahepatic spread	1.913	0.598–6.122	0.274			n.s.
irAEs	0.457	0.171–1.226	0.120			n.s.
Increase of Ki-67+/PD-1+CD8+	0.244	0.081–0.735	0.012			n.s.
Increase of TIGIT+/CD8+	0.331	0.092–0.988	0.040	0.193	0.049-0.754	0.018

AFP, alpha-fetoprotein; PIVKA, protein induced by vitamin K antagonist-II; ECOG, Eastern Cooperative Oncology Group; TIGIT, T-cell immunoglobulin and ITIM domain.

Finally, we aimed to validate whether these markers are predictive of OS and PFS across various subgroups (online suppl. Table S9). As a result, expanders showed better OS in younger patients, both viral and non-viral etiologies, ECOG 1, absence of vascular invasion, presence of EHS, and large intrahepatic tumors. They also demonstrated better PFS in the younger age group under the same conditions, with or without vascular invasion and EHS, and regardless of tumor size. Patients with ΔTIGIT ≥6.4 exhibited favorable OS in older age, absence of vascular invasion, and presence of EHS. Additionally, they showed better PFS in subgroups with poorer Child-Pugh scores, ECOG 0, and no vascular invasion. These findings suggest that dynamic changes in T-cell proliferation and TIGIT expression can serve as biomarkers for predicting OS and PFS among various subgroups. Taken together, these findings clearly indicate that dynamic peripheral blood CD8⁺ T-cell changes can be used as on-treatment prognostic biomarkers of AB treatment in patients with HCC, which is in line with the study regarding ICI treatment with other solid tumors [15, 16].

Discussion

In the present study, we focused on the role of dynamic T-cell changes measured by flow cytometry as predictive or prognostic biomarkers, using blood samples obtained from a multicenter, prospective cohort of patients with HCC who received their first dose of AB treatment. We demonstrated that dynamic changes in CD8⁺ T cells can be used as an on-treatment prognostic biomarker of patients with HCC undergoing AB treatment.

A previous study showed that pretreatment assessment of CD8⁺ T cells in tissues, assessed by immunohistochemistry, and effector T-cell signatures, derived by transcriptome analyses, may indicate the ORR or PFS in patients undergoing AB treatment [9]. Nevertheless, the dynamic changes in the T-cell population and their potential roles as biomarkers remained unclear. Furthermore, mechanistic insights regarding T-cell responses to AB treatment also required clarification, considering that T cells are the main responders and effectors in ICI treatment of HCC [17]. By analyzing serial peripheral PBMCs, we found that activation and proliferation occurred mainly in the CD8⁺ T-cell population. We also identified that, among the CD8⁺ T-cell population, PD-1⁺ cells, rather than PD-1⁻ cells, were selectively responsive to AB treatment. Moreover, the baseline differentiation status of PD-1⁺CD8⁺ T cells was associated with dynamic changes in the CD8⁺ T-cell population after AB treatment. Importantly, our analysis highlights the potential of dynamic changes in Ki-67 and TIGIT expression in CD8⁺ T cells as on-treatment prognostic biomarkers of response to AB treatment in HCC. This approach may lead to more tailored and effective treatment strategies for HCC.

Use of dynamic biomarkers which were previously analyzed in serial tumor biopsies from ICI-treated patients presents challenges due to the invasiveness of tissue biopsies [18]. Blood samples, which are less invasive and easier to collect [16], offer an opportunity for serial sampling, resulting in a more practical approach for predicting responses to ICI treatment. While tumor tissues offer direct insights into changes within the TME and among various immune cells, the proliferation of peripheral T cells can also serve as a predictor of tumor infiltration and the clinical response to ICI treatments [11]. Recent studies suggested that dynamic biomarkers assessed serially in the initial phases of treatment may offer enhanced predictive accuracy over static biomarkers, which are evaluated solely at baseline during ICI treatment [10]. In this respect, our observation that dynamic changes in Ki-67 and TIGIT expression among CD8⁺ T cells correlate with PFS and OS is interesting. This suggests that monitoring these markers could provide real-time insights into the efficacy of AB treatment for HCC. These dynamic biomarkers might offer a better understanding of treatment response than static biomarkers, which are only assessed at baseline. Furthermore, our observations were made using flow cytometry, which is a more expeditious and streamlined method than the transcriptome analyses that have been investigated previously [9, 19].

Our study indicated marked CD8⁺ T-cell proliferation and activation after AB treatment. This contrasts with the lack of significant changes in CD4⁺ T cells and Tregs, underscoring that CD8⁺ T cells are the main targets of ICI treatment in patients with HCC. This finding aligns with the mechanism of action of ICIs, which predominantly aim to enhance anti-tumor T-cell responses by reinvigorating exhausted CD8⁺ T cells [20]. Interestingly, we found an association between the baseline differentiation status of PD-1⁺CD8⁺ T cells and the proliferation of or upregulation of TIGIT in CD8⁺ T cells after AB treatment. Notably, the CM population within PD-1⁺ T cells is associated with better immunotherapy outcomes in other cancers [21]. This population can be derived from stem-like or progenitor cell populations among exhausted T cells [22], which is consistent with our findings. We found that the CM population among PD-1⁺CD8⁺ T cells significantly correlated with CD8⁺ T cell proliferation after AB treatment. This population was linked to expression of TCF-1, a marker of progenitor cells associated with the proliferative potential of exhausted T cells [23]. This suggests that the pretreatment differentiation status of patients can influence the efficacy of ICIs, offering a potential avenue for personalized treatment strategies in HCC immunotherapy. Although the mechanism behind this observation needs to be investigated in future studies, antigenicity of CD8⁺ T cells might be also associated with their proliferation in HCC patients undergoing AB treatment, as previously described [9].

In addition, the CD8⁺ T-cell proliferation after AB treatment was associated with liver function in our cohort. Liver cirrhosis and associated systemic inflammation can affect CD8⁺ T-cell phenotypes and functions [24], but their association with the baseline differentiation status of CD8⁺ T cells and their relationship with the clinical outcomes of AB treatment for HCC need to be clarified in future studies.

Another interesting finding was the potential predictive role of T-cell changes in irAEs, which is important for patient management and treatment optimization. Data regarding markers associated with irAEs in HCC immunotherapy are lacking. A recent report showed that peripheral blood CXCR3⁺CD8⁺ T_EM cells evaluated using single-cell RNA sequencing were associated with irAEs [19]. We found that the dynamic increase in TIGIT⁺ cells within the CD8⁺ T-cell population was more significant in patients who developed irAEs, and that a lower baseline EMRA population among PD-1⁺CD8⁺ T cells, which can be less immunogenic upon ICI treatment, was associated with this dynamic change. A previous study using transcriptome analysis showed that TIGIT expression marks T-cell exhaustion [25]. However, it can also be upregulated upon T-cell activation to regulate hyperfunction [26]. Considering these fundamental explanations, TIGIT upregulation in CD8⁺ T cells might play a role in restricting the hyperfunction of T cells in patients who develop irAEs; however, it remains possible that TIGIT⁺CD8⁺ T cells can contribute to the development of irAEs, which should be elucidated in future studies. Nevertheless, an association between TIGIT upregulation in CD8⁺ T cells and improved OS was apparent in the present study. This may be linked to irAEs because the development of irAEs is associated with OS and PFS in patients with HCC undergoing AB treatment [27]. On the other hand, TIGIT expression on CD8⁺ T cells was predictive of response to anti-PD-1 therapy in human solid cancers [28]. Therefore, Future mechanistic studies on the association between TIGIT⁺CD8⁺ T cells and irAEs or clinical responses should be performed.

This study has some limitations because there is possibility that PD-1⁻CD8⁺ T cells can also be activated and subsequently become PD-1⁺CD8⁺ T cells, expressing Ki-67 or TIGIT, but in vivo tracking of these cells is limited in researches using human samples. Nonetheless, concurrent activations of both PD-1⁺ and PD-1⁻CD8⁺ T cells might represent a overall activation of the CD8⁺ T-cell population induced by ICI therapy, encompassing possible anti-tumor immune responses. Therefore, both mechanistic studies using animal models and clinical studies with larger study number need to be performed.

In summary, dynamic changes in peripheral blood CD8⁺ T cells measured using flow cytometry can be used as on-treatment biomarkers related to PFS and OS. Mechanistically, this dynamic change may be linked to the baseline differentiation status. Although our findings should be verified in future, our results significantly advance the understanding of the role of dynamic biomarkers in predicting the response to AB treatment in patients with HCC. Our findings highlight the potential of assessment of peripheral blood T cells by flow cytometry as less invasive and practical alternative to tumor tissue biopsies and transcriptome analyses.

Statement of Ethics

This study followed the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Seoul St. Mary’s Hospital (Approval No.: KC22EASI0342), Incheon St. Mary’s Hospital (Approval No.: OC22TISI0090), Eunpyeong St. Mary’s Hospital (Approval No.: PC19OESI0017), and Uijeongbu St. Mary’s Hospital (Approval No.: UC23TASI0008). Written informed consents from all study patients were acquired.

Conflict of Interest Statement

The authors have no conflicting financial interests to declare.

Funding Sources

This research was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT (Grant No. RS-2024-00337298 to P.S.S.). This research was also supported by the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (Grant No. RS-2024-00406716 to J.W.H.).

Author Contributions

J.W.H. and P.S.S. designed the study. J.W.H., P.S.S., S.K.L., J.H.K. (Ji Hoon Kim), J.S.Y., H.S.C., S.W.N, H.Y., S.H.B., J.H.K. (Jung Hyun Kwon), J.W.J., J.Y.C., and S.K.Y. provided clinical data and samples. J.W.H., M.W.K., D.H.S., and E.J.J. performed the experiments. J.W.H. and M.W.K. collected and analyzed data. P.S.S. supervised the analyses and manuscript preparation. J.W.H., M.W.K., and P.S.S. wrote and edited the manuscript.

References

1. Llovet JM, Castet F, Heikenwalder M, Maini MK, Mazzaferro V, Pinato DJ,. Immunotherapies for hepatocellular carcinoma. Nat Rev Clin Oncol. 2022;19(3): 151–72.

2. Korean Liver Cancer Association KLCA and National Cancer Center NCC Korea. 2022 KLCA-NCC Korea practice guidelines for the management of hepatocellular carcinoma. Clin Mol Hepatol. 2022;28(4): 583–705.

3. Chan LL, Chan SL. The evolving role of lenvatinib at the new era of first-line hepatocellular carcinoma treatment. Clin Mol Hepatol. 2023;29(4): 909–23.

4. Han JW, Park SH. Advances in immune checkpoint inhibitors for hepatocellular carcinoma.JLiver Cancer. 2021;21(2): 139–45.

5. Finn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim T-Y,. Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma.NEnglJMed. 2020;382(20): 1894–905.

6. Han JW, Jang JW,. Predicting outcomes of atezolizumab and bevacizumab treatment in patients with hepatocellular carcinoma. IntJMol Sci. 2023;24(14): 12068.

7. Han JW, Yoon SK. Immune responses following locoregional treatment for hepatocellular carcinoma: possible roles of adjuvant immunotherapy. Pharmaceutics. 2021;13(9): 1387.

8. Sung PS. Crosstalk between tumor-associated macrophages and neighboring cells in hepatocellular carcinoma. Clin Mol Hepatol. 2022;28(3): 333–50.

9. Zhu AX, Abbas AR, de Galarreta MR, Guan Y, Lu S, Koeppen H,. Molecular correlates of clinical response and resistance to atezolizumab in combination with bevacizumab in advanced hepatocellular carcinoma. Nat Med. 2022;28(8): 1599–611.

10. Lesterhuis WJ, Bosco A, Millward MJ, Small M, Nowak AK, Lake RA. Dynamic versus static biomarkers in cancer immune checkpoint blockade: unravelling complexity. Nat Rev Drug Discov. 2017;16(4): 264–72.

11. Wu TD, Madireddi S, de Almeida PE, Banchereau R, Chen YJ, Chitre AS,. PeripheralTcell expansion predicts tumour infiltration and clinical response. Nature. 2020;579(7798): 274–8.

12. Llovet JM, Lencioni R. mRECIST for HCC: performance and novel refinements.JHepatol. 2020;72(2): 288–306.

13. Han JW, Sung PS, Hong SH, Lee H, Koh JY, Lee H,. IFNL3-adjuvanted HCV DNA vaccine reduces regulatoryTcell frequency and increases virus-specificTcell responses.JHepatol. 2020;73(1): 72–83.

14. Banta KL, Xu X, Chitre AS, Au-Yeung A, Takahashi C, O'Gorman WE,. Mechanistic convergence of the TIGIT and PD-1 inhibitory pathways necessitates co-blockade to optimize anti-tumor CD8(+)Tcell responses. Immunity. 2022;55(3): 512–26.e9.

15. Kamphorst AO, Pillai RN, Yang S, Nasti TH, Akondy RS, Wieland A,. Proliferation of PD-1+ CD8Tcells in peripheral blood after PD-1-targeted therapy in lung cancer patients. Proc Natl Acad SciUS A. 2017;114(19): 4993–8.

16. Kim KH, Cho J, Ku BM, Koh J, Sun JM, Lee SH,. The first-week proliferative response of peripheral blood PD-1(+)cd8(+)Tcells predicts the response to anti-PD-1 therapy in solid tumors. Clin Cancer Res. 2019;25(7): 2144–54.

17. Sankar K, Gong J, Osipov A, Miles SA, Kosari K, Nissen NN,. Recent advances in the management of hepatocellular carcinoma. Clin Mol Hepatol. 2024;30(1):1– 15.

18. Chen PL, Roh W, Reuben A, Cooper ZA, Spencer CN, Prieto PA,. Analysis of immune signatures in longitudinal tumor samples yields insight into biomarkers of response and mechanisms of resistance to immune checkpoint blockade. Cancer Discov. 2016;6(8): 827–37.

19. Chuah S, Lee J, Song Y, Kim HD, Wasser M, Kaya NA,. Uncoupling immune trajectories of response and adverse events from anti-PD-1 immunotherapy in hepatocellular carcinoma.JHepatol. 2022;77(3): 683–94.

20. Kim HD, Song GW, Park S, Jung MK, Kim MH, Kang HJ,. Association between expression level of PD1 by tumor-infiltrating CD8(+)Tcells and features of hepatocellular carcinoma. Gastroenterology. 2018;155(6): 1936–50.e17.

21. Sakuma M, Mimura K, Nakajima S, Kaneta A, Kikuchi T, Nirei A,.Apotential biomarker of dynamic change in peripheral CD45RA(-)CD27(+)CD127(+) central memoryTcells for anti-PD-1 therapy in patients with esophageal squamous cell carcinoma. Cancers. 2023;15(14): 3641.

22. Pais Ferreira D, Silva JG, Wyss T, Fuertes Marraco SA, Scarpellino L, Charmoy M,. Central memory CD8(+)Tcells derive from stem-like Tcf7(hi) effector cells in the absence of cytotoxic differentiation. Immunity. 2020;53(5): 985–1000 e11.

23. Kim KH, Kim HK, Kim HD, Kim CG, Lee H, Han JW,. PD-1 blockade-unresponsive human tumor-infiltrating CD8(+)Tcells are marked by loss of CD28 expression and rescued by IL-15. Cell Mol Immunol. 2021;18(2): 385–97.

24. Albillos A, Lario M, Alvarez-Mon M. Cirrhosis-associated immune dysfunction: distinctive features and clinical relevance.JHepatol. 2014;61(6): 1385–96.

25. Ostroumov D, Duong S, Wingerath J, Woller N, Manns MP, Timrott K,. Transcriptome profiling identifies TIGIT asamarker of T-cell exhaustion in liver cancer. Hepatology. 2021;73(4): 1399–418.

26. Zheng Q, Xu J, Gu X, Wu F, Deng J, Cai X,. Immune checkpoint targeting TIGIT in hepatocellular carcinoma. AmJTransl Res. 2020;12(7): 3212–24.

27. Fukushima T, Morimoto M, Kobayashi S, Ueno M, Uojima H, Hidaka H,. Association between immune-related adverse events and survival in patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab. Oncologist. 2023;28(7): e526–33.

28. Simon S, Voillet V, Vignard V, Wu Z, Dabrowski C, Jouand N,. PD-1 and TIGIT coexpression identifiesacirculating CD8Tcell subset predictive of response to anti-PD-1 therapy.JImmunother Cancer. 2020;8(2): e001631.