Advancing liver cancer treatment with dual-targeting CAR-T therapy | Journal of Nanobiotechnology

Expression analysis of AFP and GPC3 in liver cancer

To develop effective and safe CAR-T cell therapy for liver cancer, we investigated the expression of AFP and GPC3 in hepatocellular carcinoma (HCC) and normal tissues. This analysis helps predict potential therapeutic effects and adverse events associated with CAR-T products targeting these antigens.

We utilized the GEPIA2 database ( which includes gene expression data from The Cancer Genome Atlas (TCGA). Analysis of pan-cancer samples revealed significant expression of AFP and GPC3 primarily in liver cancer tissues, with minimal expression in normal tissues and other cancer types (Supplementary Fig. 1A, B and Supplementary Tables 1, 2).

Furthermore, we analyzed a cohort of 368 tumor tissues and 160 normal liver tissues from patients with HCC. This analysis confirmed significant differences in the expression levels of AFP and GPC3 between liver cancer tissue and adjacent non-cancerous tissues (Supplementary Fig. 1C and D).

To strengthen the clinical relevance of our findings, we conducted a retrospective analysis on 65 liver cancer patients diagnosed between 2014 and 2021 at Zhuhai People’s Hospital. Immunohistochemical staining of their pathological specimens revealed that 50.77% of the patients were positive for AFP, 70.77% were positive for GPC3, and 38.46% expressed both antigens (Supplementary Fig. 1E, F, G and Supplementary Table 3).

These findings strongly suggest that CAR-T cells targeting AFP and GPC3, liver cancer-specific antigens, hold promise for improved treatment efficacy and enhanced safety.

Development of high-affinity AFP-specific antibody

Building on the success of our second-generation GPC3 CAR-T with high anti-tumor efficacy, we are initiating the development of an AFP-targeting TCR mimic antibody to explore potential synergistic effects in combination therapy.

Given the high prevalence of the HLA-A*02:01 allele (8.15% globally, 12.04% in China), we focused on constructing the hAFP_{158 − 166}/HLA-A*02:01 complex for optimal antigen presentation (Fig. 1A Up). Following purification of recombinantβ2m and HLA-A*02:01 proteins, we co-complexed them with the hAFP_{158 − 166} peptide at a defined ratio. Utilizing Capto Q ImpRes, we purified the complexes to achieve 99.09% purity (Fig. 1A Down).

To identify high-affinity antibodies targeting AFP presented by HLA-A*02:01, we employed a biopanning strategy on an artificially synthesized phage display antibody library. After five rounds of liquid-phase screening, we identified five candidate VHH antibodies (Fig. 1B, C, D). Monoclonal ELISA was used to select five candidate antibodies exhibiting low background and significant signal-to-noise ratios. These candidates were sequenced and designated as 1B3, 1C11, 1C4, 1D12, and 2F9. Notably, the Ab61 antibody, encoded by the gene described in patent CN107106671A, was synthesized as a positive control. Protein A affinity chromatography purified this positive control. Subsequent one-step Protein A affinity chromatography and SEC-HPLC analysis confirmed high purity for all VHH antibodies (Supplementary Fig. 2A, B and Supplementary Table 4).

Finally, affinity assays were performed using a ForteBio AHC sensor-conjugated antibody loaded with each candidate antibody and the hAFP_{158 − 166}/HLA-A*02:01 complex. All candidate antibodies, except 1D12, demonstrated superior dissociation constants compared to the Ab61 positive control. Notably, the 1B3 antibody exhibited an affinity constant approximately five-fold higher than Ab61 (Fig. 1E and Supplementary Table 5).

These results highlight the successful development of a high-affinity, specific antibody (1B3) targeting hAFP presented by HLA-A*02:01, a crucial component for our dual-targeting CAR-T cell therapy strategy.

(A) Up, Construction process of hAFP_158-166/HLA-A*02:01 complex. Down, Purity analysis of hAFP_158-166/HLA-A*02:01 complex by SEC-HPLC. (B) Antibody screening process: ① Synthesis of artificial nanobody phage display library. ② Screening for humanized VHH antibodies. ③ Elution of phages. ④ Analysis of target phages. ⑤ Phage amplification and acquisition of candidate VHH plasmid sequences. (C) The acquisition and screening process of antibodies. (D) Five rounds of liquid phase elution antibody enrichment effect (In each round of elution, each column represents a concentration gradient of phage stock solution, which is diluted into a gradient at a 3-fold ratio from left to right). (E) Results of affinity assay of different antibodies binding to peptide hAFP_158-166/HLA-A*02:01: the affinity of AFP Ab (1B3) antibody reached 0.504nM; the affinity of AFP Ab (1C11) antibody was 1.21nM; the affinity of AFP Ab (1D12) antibody was 3.02nM; the affinity of AFP Ab (1C4) was 0.865nM; the affinity of AFP Ab (2F9) was 2.06nM; the affinity of Bench marker (Ab61) was 2.34nM. The affinity of the three antibodies (1C11, 1D12 and 2F9) was comparable to the Bench marker (Ab61) (The six curves in Figure E are the fitted curves for the first six antigen concentrations)

In vitro functional evaluation of the VHH antibodies for AFP and the associated CAR-T

To assess the specificity of the VHH antibodies, we performed binding assays using T2 cells (The T2 Cell is a human lymphocyte hybridoma cell that faithfully presents the target antigenic peptide to the cell surface, thereby targeting the T cell corresponding to the target antibody.), loaded with various peptide antigens. The results demonstrated that the antibodies exhibited high specificity for the hAFP_{158 − 166}/HLA-A*02:01 complex, with minimal binding to T2 cells loaded with irrelevant peptides (Fig. 2A). Notably, the 1B3 antibody demonstrated strong binding affinity not only to the HLA-A*02:01 complex but also to the HLA-A*02:07 complex, suggesting potential applicability in a broader patient population (Supplementary Fig. 2C and D).

To evaluate the cytotoxic potential of the VHH antibodies, we conducted in vitro ADCC assays. The target cells (HepG2) were engineered to express the hAFP158-166/HLA-A*02:01 complex on their surface (HepG2-MiniG). We co-cultured various tumor cell lines HepG2 cell (hAFP⁺/HLA-A*02:01⁻), HepG2-MiniG cell (hAFP_{158 − 166}⁺/HLA-A*02:01⁺), SK-HEP-1 cell (hAFP⁻/HLA-A*02:01⁺), and Raji cell (hAFP⁻/HLA-A*02:01⁻) and Daudi cell (hAFP⁻/HLA-A*02:01⁻) with PBMCs and different VHH antibodies. The results showed that all five antibodies exhibited specific cytotoxicity against hAFP_{158 − 166}⁺/HLA-A*02:01⁺ tumor cells, with minimal non-specific killing of other cell lines (Supplementary Fig. 3A, B, C, D and E). It is particularly noteworthy that the AFP Ab (1B3) and AFP Ab (1C11) antibodies did not exhibit non-specific cytotoxicity against SK-HEP-1 cells at any concentration (Supplementary Fig. 3D).

To further investigate the therapeutic potential of these VHH antibodies, we engineered CAR-T cells incorporating the VHH sequences. The target cells (SK-HEP-1 and HepG2) were engineered to express the hAFP_{158 − 166}/HLA-A*02:01 complex on their surface (named SK-HEP-1-MiniG and HepG2-MiniG), enabling recognition and elimination by CAR-T cells (Fig. 2B).

In vitro repeated stimulation experiments demonstrated that 1B3 CAR-T cells exhibited robust expansion and specific cytotoxicity against target cells, comparable to the positive control Ab61 CAR-T cells. Moreover, 1B3 CAR-T cells displayed lower non-specific cytotoxicity against hAFP⁻/HLA-A*02:01⁻ cells, suggesting improved safety. Cytokine release assays revealed that 1B3 CAR-T cells produced similar levels of cytotoxic cytokines as Ab61 CAR-T cells, but with lower baseline cytokine secretion in the absence of target cells. This suggests that 1B3 CAR-T cells may offer a favorable balance of efficacy and safety (Fig. 2C, D, E, and F).

These findings collectively demonstrate the promising potential of the AFP Ab (1B3) antibody and its corresponding CAR-T cells for the targeted treatment of liver cancer, particularly in patients expressing HLA-A*02:01 or HLA-A*02:07.

(A) All antibodies showed binding activity to T2 cells loaded with hAFP_158-166 peptide, but had non-specific binding to T2 cells loaded with NY-ESO and 19 mixed peptides. (B) Illustration of hAFP_158-166/HLA-A*02:01 TCR-like CAR-T binding to tumor cells and killing them. (C) In vitro cytotoxicity assay of CAR-T cells constructed with various antibodies against different cancer cells (SK-HEP-1, HepG2, SK-HEP-1-MiniG, HepG2-MiniG). (D) The fold expansion of each CAR-T cell in the in vitro repeated stimulation experiment (target cells: SK-HEP-1-MiniG). (E) The secretion levels of IFN-γ by CAR-T cells in the cytotoxicity assay at an effector-to-target ratio of 1:1. (F) The secretion levels of IL-2 by CAR-T cells in the cytotoxicity experiment at an effector-to-target ratio of 1:1

In vivo efficacy of 1B3 CAR-T cells

To evaluate the in vivo efficacy of 1B3 CAR-T cells, we established a subcutaneous tumor model using SK-HEP-1-MiniG cells in mice. Treatment with 1B3 CAR-T cells significantly inhibited tumor growth compared to the control group (Fig. 3A, B, and C). Notably, intratumoral administration of CAR-T cells demonstrated enhanced efficacy compared to intravenous administration. Importantly, 1B3 CAR-T cells were well-tolerated, with no significant changes in body weight observed (Fig. 3D).

To further assess the durability of the anti-tumor response, we established a secondary tumor challenge model (Fig. 4A and Supplementary Table 6). Mice were treated with 1B3 CAR-T cells and, after complete tumor regression, were re-challenged with SK-HEP-1-MiniG cells on the contralateral flank. Remarkably, even at lower doses, 1B3 CAR-T cells effectively inhibited the growth of the secondary tumor, demonstrating potent long-lasting anti-tumor activity (Fig. 4B).

To investigate the mechanisms underlying the therapeutic efficacy of 1B3 CAR-T cells, we analyzed the persistence and cytokine release profiles of these cells in vivo. Flow cytometry analysis revealed that 1B3 CAR-T cells persisted in the tumor microenvironment for an extended period, contributing to sustained tumor control (Fig. 4C). Additionally, 1B3 CAR-T cells exhibited a balanced cytokine release profile, with elevated levels of IFN-γ and IL-2, crucial for effective tumor cell killing and T cell proliferation, respectively (Fig. 4D, E).

The safety profile of 1B3 CAR-T cells was assessed by monitoring body weight and peripheral cytokine levels. No significant weight loss was observed in any of the treatment groups, indicating good tolerability (Supplementary Fig. 4A). Notably, 1B3 CAR-T cells induced lower levels of IL-6, a cytokine associated with cytokine release syndrome (CRS), compared to the Ab61 control (Supplementary Fig. 4B). These findings suggest that 1B3 CAR-T cells may have a favorable safety profile, with a reduced risk of cytokine release syndrome.

(A) Flow chart of the first round animal experiments (In this study, the first CAR-T cell therapy injection was administered on Day 0, SK-HEP-1-MiniG were inoculated in mice on Day − 9, and the second CAR-T cell therapy injection was administered on Day 14. Throughout the experimental period, tumor size was measured three times per week. The proportion of human T cells and cytokine levels were assessed through peripheral blood testing). (B) Administration methods and dosages of CAR-T in each group, tumor volume of mice at endpoint, and tumor inhibition rate. (C) Tumor growth curve of mice in all groups (each group: n = 5). (D) Changes in body weight of mice in each treatment group (each group: n = 5)

(A) Flow chart of the second round animal experiments (In this study, SK-HEP-1-MiniG cells were inoculated subcutaneously in the right flank of mice on Day − 31. When the average tumor volume reached 151 mm³, CAR-T cells were administered via intratumoral injection. Due to significant tumor regression, tumor cells were re-inoculated subcutaneously in the left flank on Day 43 to simulate a tumor recurrence model, aiming to evaluate the sustained anti-tumor efficacy of CAR-T cells). (B) Tumor growth curve of mice in all groups, rechallenge: mice with tumor regression were rechallenged with tumor cells on the contralateral side to observe the long-lasting tumor-suppressive effects of CAR-T (each group: n = 5). (C) CAR-positive rate of CAR-T cells in peripheral blood of mice on day 7, day 14 and day 21 (each group: n = 5). (D) Secretion level of IFN-γ in the peripheral blood serum of mice on day 7, day 14 and day 21 (each group: n = 5). (E) Secretion level of IL-2 in the peripheral blood serum of mice on day 7, day 14 and day 21 (each group: n = 5)

Low-dose 1B3 CAR-T therapy: a promising approach for cancer treatment

To further investigate the therapeutic potential of 1B3 CAR-T cells, we conducted a third round of in vivo experiments, focusing on low-dose therapy. Mice bearing subcutaneous SK-HEP-1- MiniG tumors were treated with intratumoral injections of 1 × 10⁶ or 3 × 10⁵ 1B3 or Ab61 CAR-T cells (Fig. 5A). Remarkably, even at the low dose of 3 × 10⁵ cells, 1B3 CAR-T cells demonstrated superior tumor suppression compared to Ab61 CAR-T cells (Fig. 5B and Supplementary Table 7). To assess the long-term efficacy, mice were re-challenged with tumor cells on the contralateral flank. Both 1B3 and Ab61 CAR-T cells effectively suppressed tumor recurrence, highlighting the durability of the anti-tumor response (Fig. 5B).

In contrast to the previous experiments with higher initial tumor burden and CAR-T cell doses, the peak in CAR-T cell expansion and cytokine release occurred later in the low-dose setting. Specifically, a peak in IFN-γ and TNF-α levels was observed on day 14, while IL-2 levels remained elevated throughout the study period (Fig. 5C, D, E and Supplementary Fig. 4D).

Regarding safety, no significant weight loss was observed in any treatment group (Supplementary Fig. 4C). The transient elevation in IL-6 levels, a cytokine associated with cytokine release syndrome (CRS), was minimal and did not persist. These findings suggest that low-dose 1B3 CAR-T therapy offers a favorable safety profile with minimal adverse effects.

Overall, these results demonstrate the potent anti-tumor activity and favorable safety profile of low-dose 1B3 CAR-T therapy. This approach holds significant promise for the treatment of liver cancer with 1B3 antibody targeting AFP.

(A) Flow chart of the third round animal experiments (In this experiment, to challenge CAR-T cells with larger tumors, we inoculated SK-HEP-1-MiniG cells subcutaneously on the right flank of mice on Day − 33. When the tumor volume reached 182 mm³, CAR-T cells were administered via intratumoral injection. Subsequently, rapid tumor regression was observed. On Day 59, mice were re-challenged with SK-HEP-1-MiniG cells subcutaneously on the left flank to assess the long-term antitumor efficacy of CAR-T cells). (B) Tumor growth curve of mice in all groups, rechallenge: mice with tumor regression were rechallenged with tumor cells on the contralateral side to observe the long-lasting tumor-suppressive effects of CAR-T (each group: n = 5). (C) CAR-positive rate of CAR-T cells in peripheral blood of mice on day 7, day 14 and day 21 (each group: n = 5). (D) Secretion level of IFN-γ in the peripheral blood serum of mice on day 7, day 14 and day 21 (each group: n = 5). (E) Secretion level of IL-2 in the peripheral blood serum of mice on day 7, day 14 and day 21 (each group: n = 5)

GPC3 CAR-T cells with integrated AFP-CD3 bite: enhanced anti-tumor activity

While our previous generation of GPC3 CAR-T cells demonstrated promising anti-tumor activity, we sought to further optimize their efficacy by incorporating additional targeting strategies. The integration of an optimized AFP-CD3 BiTE into the CAR-T cell design provides a dual-targeting approach, potentially enhancing tumor cell killing and overcoming potential resistance mechanisms.

We constructed second-generation GPC3 CAR-T cells incorporating 4-1BB as a co-stimulatory domain, integrated an optimized, secreted AFP-CD3 BiTE into the CAR-T cell design. The BiTE molecule, upon secretion, can bind to both AFP-expressing tumor cells and T cells, both endogenous and CAR-T cells, forming an immune synapse and activating bystander T cells (Fig. 6A, B).

To evaluate the functional activity of these CAR-T cells, we established a target cell line (Sk-hep-1-GPC3-AFP) overexpressing both GPC3 and the hAFP_{158 − 166}/HLA-A*02:01 complex (Fig. 6C). In vitro cytotoxicity assays and repeated stimulation experiments demonstrated that GPC3 CAR-T cells with the integrated AFP-CD3 BiTE (GPC3 CAR-T-AFP-CD3) exhibited superior anti-tumor activity compared to single-target GPC3 CAR-T cells (Fig. 6F, G). Furthermore, phenotypic analysis revealed that GPC3 CAR-T-AFP-CD3 cells displayed a mixed phenotype of central memory T cells (T_CM) and effector memory T cells (T_EM) (Fig. 6D, E). This suggests that GPC3 CAR-T-AFP-CD3 cells may possess enhanced persistence and long-term anti-tumor activity. Notably, GPC3 CAR-T-AFP-CD3 cells consistently demonstrated superior performance in both cytotoxicity assays and IFN-γ secretion assays (Fig. 6F, G).

These findings highlight the potential of our dual-targeting strategy to improve the efficacy of CAR-T cell therapy for liver cancer. By combining direct tumor cell killing with bystander T cell activation, this approach may overcome the challenges associated with tumor heterogeneity and immune suppression.

(A) Designed and optimized second-generation GPC3 CAR-T cells that incorporate 4-1BB as a co-stimulatory domain and are capable of secreting BiTE molecules that bind to AFP-expressing tumor cells and T cells. (B) lllustration of GPC3 CAR-T/AFP-BiTE construction. (C) Confirmation of GPC3 and AFP expression on target liver cancer cell line (SK-HEP-1-GPC3-AFP) by flow cytometry. (D) Flow cytometry analysis was performed to detect the memory/effector phenotype of CAR T cells. (E) Flow cytometry analysis was performed to detect the memory/effector phenotype of CAR T cells after repeated stimulation experiment in vitro. (F) Cytotoxicity of CAR-T Cells in vitro cytotoxicity assay. (G) The secretion levels of IFN-γ in both cytotoxicity assays and autocrine secretion of various CAR-T cells

Mechanistic insights into enhanced efficacy of GPC3 CAR-T-AFP-CD3 cells

To understand the mechanisms underlying the superior anti-tumor activity of GPC3 CAR-T-AFP-CD3 cells, we performed transcriptome sequencing on these cells after repeated stimulation. Principal component analysis revealed distinct gene expression profiles between GPC3 CAR-T-AFP-CD3 cells and control groups (Fig. 7A and B). Volcano plots and Venn diagrams identified differentially expressed genes (DEGs) across groups (Figs. 7C-F and Supplementary Tables 8, 9, 10 and 11). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses revealed enrichment of pathways associated with enhanced T cell function in GPC3 CAR-T-AFP-CD3 cells compared to controls (Supplementary Fig. 5A, B, C, D, E and F). Notably, genes like TNF, CD40LG, and CSF2 emerged as key players in the functional distinction between GPC3 CAR-T-AFP-CD3 and other CAR-T cell variants (Fig. 7G, H and I). These findings suggest that the incorporation of the AFP-CD3 BiTE into GPC3 CAR-T cells (GPC3 CAR-T-AFP-CD3) modulates gene expression, potentially contributing to their superior anti-tumor activity. Further investigation of these key genes may provide valuable insights for optimizing CAR-T cell therapy for liver cancer.

(A) The box plot of gene expression level. (B) The principal component analysis of the gene expression of samples in each group. (C) The volcano plot of the overall distribution of differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T group; (D) The volcano plot of the overall distribution of differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T-AFP group; (E) The volcano plot of the overall distribution of differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T-CD3 group. (F) The common and unique differentially expressed genes among different comparison groups. (G) The PPI network constructed by differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T group. (H) The PPI network constructed by differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T-AFP group. (I) The PPI network constructed by differentially expressed genes when comparing the GPC3 CAR-T-AFP-CD3 group with the GPC3 CAR-T-CD3 group

In vivo efficacy of GPC3 CAR-T with AFP-CD3 bite

To assess the in vivo efficacy of GPC3 CAR-T cells with the integrated AFP-CD3 BiTE (GPC3 CAR-T-AFP-CD3), we established a subcutaneous tumor model in mice (Fig. 8A). When tumor volumes reached 80–120 mm³, mice were intravenously administered with 1 × 10⁶ or 3 × 10⁶ CAR-T cells. The group treated with 3 × 10⁶ GPC3 CAR-T-AFP-CD3 cells exhibited significant tumor growth inhibition and prolonged survival compared to the control groups (Fig. 8B and Supplementary Table 12). Importantly, no significant weight loss was observed in any treatment group, indicating a favorable safety profile (Fig. 8C). Furthermore, increased levels of IFN-γ in the peripheral blood and expansion of CAR-positive T cells were observed in mice treated with GPC3 CAR-T-AFP-CD3 cells (Fig. 8D, E). These findings suggest that the BiTE-based strategy enhances the anti-tumor activity of CAR-T cells, with the potential recruiting and activating endogenous T cells.

In conclusion, our study demonstrates the potential of GPC3 CAR-T cells with the integrated AFP-CD3 BiTE as a promising therapeutic approach for liver cancer. This dual-targeting strategy offers improved anti-tumor efficacy and a favorable safety profile, making it a promising candidate for future clinical translation.

(A) Schematic of the liver cancer xenograft model used to investicate the in vivo activitof CAR-T cells (In this experiment, SK-HEP-1-GPC3-AFP cells were inoculated subcutaneously on Day − 15. When tumor volume reached 80–120 mm³, mice were randomly grouped. The first intravenous administration of CAR-T cells was performed on Day 0, followed by a second intravenous administration on Day 14). NSG mice were subcutaneously injected with 3 × 10⁶ target cells, and CAR-T cells were injected at specified time points. Each group consisted of n = 5 mice. (B) Tumor growth was assessed three times per week following CAR-T cell injection. (C) ln vivo toxicity was evaluated by monitoring the mice‘s body weights throughout the treatment. (D–E) Peripheral blood samples were collected weekly to analyze the levels of CD3⁺ T cells (D) and IFN-γ (E)