B7-H3 and GD2 overexpression as immunotherapeutic targets in retinoblastoma

Abstract

Retinoblastoma (RB) is the most prevalent malignant eye tumor in children. Chimeric antigen receptor T-cell therapy showed significant promise in treating blood cancers and could potentially be effective for RB. A critical step toward advancing this therapy involves identifying surface antigens on RB cells. This study focuses on evaluating the expression of tumor antigens and immune checkpoint proteins in primary and secondary enucleated RB samples and analyzing their relationship with clinical and pathological data. We performed immunohistochemical analysis on 94 formalin-fixed, paraffin-embedded RB tissue samples from Thai patients to assess the expression of B7-H3, GD2, CD171, and PD-L1. B7-H3 was the most frequently expressed antigen, appearing in 51.06% of samples (48/94) with mild to moderate membranous intensity. GD2 was present in 36.17% of samples (34/94) with moderate staining intensity. CD171 was detected in 9.57% of samples (9/94), showing weak to mild membranous intensity. PD-L1 was the least expressed, found in only 4.26% of samples (4/94) with mild to moderate membrane staining. There were no significant correlations between the expression of these antigens and patient characteristics. Our study emphasizes the variability in tumor antigen expression in RB. The notable expression of B7-H3 and GD2 suggests they could be promising targets for immunotherapy. These findings encourage further investigation into multi-specific or combination immunotherapies for treating retinoblastoma.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-34331-6.

Introduction

Retinoblastoma (RB) is the most common intraocular malignant tumor in children¹. The primary goals in managing RB are to save the patient’s life and prevent metastasis, followed by preserving the affected eye and, if possible, restoring vision. Although advance treatments for retinoblastoma include intra-ophthalmic artery and intravitreal chemotherapy were used^2,3, patients in developing countries often present with metastatic or recurrent disease, leading to poor outcomes. New treatment strategies are needed to address chemo-resistant or advanced retinoblastoma, aiming to preserve the eye and minimize the side effects of high-dose systemic chemotherapy, particularly the risk of secondary cancers. Targeted immunotherapy, employing both antibodies and engineered T-cells, has achieved notable clinical success, especially in the treatment of hematological cancers^4–6. Chimeric antigen receptor (CAR) T-cell therapy has revolutionized cancer treatment⁷, achieving highly effective and well-tolerated clinical results in a range of cancers^4–6. Numerous immunotherapies have been investigated to target tumor-associated antigens, including disialoganglioside (GD2)^8–11, L1 cell adhesion molecule (L1CAM) or CD171^10,12, programed death-ligand 1 (PD-L1)^13,14, B7-H3^15,16, and EpCAM¹⁷. Discovering these tumor-specific surface antigens is an essential initial step in designing CAR T-cells for immunotherapy. Disialoganglioside (GD2) is tumor-associated antigens (TAAs) known to be highly expressed in RB^8,10. High expression of GD2 is reported in various neuroectodermal tumors^9,18 and is present at low levels in normal tissues such as peripheral nerves, skin melanocytes, brain cells, and osteoprogenitors¹⁹. GD2 shows promise as a target for immunotherapy in RB and as a potential biomarker because it is more highly expressed in the serum of RB patients than in healthy individuals^20,21, as well as in the bone marrow of patients with extraocular metastasis²². Another surface antigen that demonstrated significant expression in RB is CD171^10,23. It has been linked to increased resistance to apoptosis and chemotherapy in several cancers, including retinoblastoma, neuroblastoma, ovarian cancer, pancreatic cancer, and glioblastoma^{10,12,23–28}. Targeting immune checkpoint proteins like PD-L1 and B7-H3 presents promising strategies for immunotherapy against solid tumors. Programmed death-ligand 1 (PD-L1) is mainly expressed in hematopoietic cells but is also found in non-hematopoietic cells, such as endothelial and epithelial cells, as well as various tumor cells^13,32. Previously, we reported that Y79 RB cells resisted GD2-CAR T-cell killing by upregulating PD-L1 expression on tumor cells⁸. Elevated PD-L1 expressions might play a role in developing resistance to immunotherapy in solid tumors including RB. Additionally, B7-H3 has been observed to be significantly overexpressed in various types of tumors²⁹, and its notable overexpression in primary RB tumor samples¹⁶. However, the prevalence of these tumor-associated antigens (TAAs) and immune checkpoint proteins in human RB tumor tissues within the Thai cohort remains unknown.

This research aims to investigate the expression of proteins via immunohistology analysis of tumor surface antigens and immune checkpoint proteins, such as GD2, CD171, PD-L1, and B7-H3, in human retinoblastoma (RB) cell lines, as well as in primary and secondary enucleated formalin-fixed paraffin-embedded (FFPE) tissue samples from Thai RB patients. Our study emphasizes the variability in tumor antigen expression in RB. Our results support advancing CAR T-cell and antibody therapies into multi-specific or combination immunotherapy approaches for the treatment of retinoblastoma.

Results

Expression of B7-H3, GD2, CD171, and PD-L1 in human retinoblastoma cell lines

The success of targeted immunotherapy for retinoblastoma (RB) hinges on the detailed analysis of tumor antigen expression in RB tissues. However, performing this analysis is often challenging, as tumor biopsies are typically avoided due to the risk of spreading the tumor. To address this challenge, we first investigated the feasibility of targeting retinoblastoma (RB) by assessing the expression of tumor-associated antigens and immune checkpoint proteins in human RB cell lines (Y79 and WERI-Rb-1) compared with a normal retinal pigment epithelial cell line (ARPE-19). Flow cytometric analysis revealed markedly elevated surface expression of the disialoganglioside GD2 in both RB cell lines, whereas GD2 expression was undetectable in ARPE-19 cells (Fig. 1a and Fig. S1). GD2 expression was further validated using both fluorophore-conjugated and unconjugated anti-GD2 monoclonal antibody (clone 14g2a, 1:25 dilution) in independent staining assays, confirming robust GD2 expression in RB cells but not in normal retinal epithelial cells (Supplementary Information).

Fig. 1.

Surface antigen and immune checkpoint molecule expression on two human retinoblastoma (RB) cell lines, Y79 and WERI-Rb-1, and normal retinal pigmented epithelial cell line (ARPE-19). (a) GD2 surface expression on Y79, WERI-Rb-1, and ARPE-19 cell lines. (b) CD171 surface expression on Y79, WERI-Rb-1, and ARPE-19 cell lines. (c) B7-H3 surface expression on Y79, WERI-Rb-1, and ARPE-19 cell lines. (d) PD-L1 surface expression on Y79, WERI-Rb-1, and ARPE-19 cell lines.

Likewise, CD171 was detected in nearly all RB cell lines, showing low-level expression in the ARPE-19 cell line (Fig. 1b and Fig. S1). B7-H3 was highly expressed in WERI-Rb-1 but absent in Y79 cells, while also being present in ARPE-19 cells (Fig. 1c and Fig. S1). PD-L1 was expressed at low levels in both RB cell lines but was detected in nearly 100% of ARPE-19 (Fig. 1d). These findings indicate that GD2 and CD171 are uniquely expressed in RB cell lines, whereas the expression of immune checkpoint proteins B7-H3 and PD-L1 varies among cells and can also be detected in normal retinal epithelial cells. Notably, all four antigens were also identified in RB tumor samples, highlighting their relevance as potential targets for cancer immunotherapy.

Clinicopathological characteristics of enucleated RB tumor samples

To validate target antigens in retinoblastoma (RB) tumors, we conducted a retrospective study using formalin-fixed paraffin-embedded (FFPE) blocks from 92 Thai patients. We analyzed a total of 94 FFPE samples collected from 2016 to 2023 through immunohistochemistry (IHC). Detailed demographic data and clinical staging information are provided in Table 1. The study included 57 males (61.96%) and 35 females (38.04%), with a median age of 2.2 years. Bilateral disease was observed in 34 patients (36.96%), while 58 patients (63.04%) exhibited unilateral retinoblastoma. Only two patients with bilateral disease underwent bilateral enucleation, totaling 94 enucleated eyes—43 from the left eye and 51 from the right. Most patients (90; 97.83%) had no family history of RB. A significant majority of eyes (83; 88.30%) were enucleated at an advanced clinical stage (Group E). Sixty-seven eyes underwent primary enucleation, while 27 received at least one cycle of systemic chemotherapy before secondary enucleation. In retinoblastoma cases subjected to secondary enucleation, treatment modalities consist of systemic chemotherapy, intra-arterial chemotherapy, and intravitreal chemotherapy. The pathological characteristics of the enucleated specimens are summarized in Table 2. More than half of the tumors exhibited an endophytic growth pattern (54.25%). Additionally, most specimens showed no choroidal invasion (61.7%) and no scleral invasion (94.68%). Other high-risk pathological features included optic nerve head involvement, which was pre-laminar in 43.62% of samples, absent in 39.36%, and post-laminar in 17.02%. Most enucleated samples showed no tumor presence in the anterior chamber (82.98%) and no vitreous seeding (69.15%).

Table 1.

Demographic data of Thai retinoblastoma patients.

Characteristics	Number (%)
Gender
Male	57 (61.96)
Female	35 (38.04)
Laterality
Unilateral	58 (63.04)
Bilateral	34 (36.96)
Median age of presentation	2.2 years
Family history of RB
Yes	2 (2.17)
No	90 (97.83)
Clinical staging
Intraocular	83 (88.30)
A: very low risk	0
B: low risk	0
C: moderate risk	0
D: high risk	0
E: very high risk	83 (88.30)
Extraocular	11 (11.70)
Enucleation
Primary	67 (71.28)
Secondary	27 (28.72)
Eye enucleated
Left eye	43 (45.74)
Right eye	51 (54.26)

Table 2.

Pathological features of the enucleated specimen.

Parameters	Number of samples (%)
Growth pattern
Endophytic	51 (54.25)
Exophytic	3 (3.19)
Mixed	14 (14.90)
No data	26 (27.66)
Tumour in the anterior chamber
Yes	16 (17.02)
No	78 (82.98)
Vitreous seeding
Yes	29 (30.85)
No	65 (69.15)
Choroidal invasion
Massive	20 (21.28)
Focal	16 (17.02)
No	58 (61.70)
Scleral invasion
Yes	5 (5.32)
No	89 (94.68)
Optic nerve head involvement
Pre-laminar	40 (42.55)
Post-laminar	17 (18.09)
No	37 (39.36)
Surgical margin of optic nerve involvement
Yes	3 (3.19)
No	91 (96.81)

Higher levels of the immune checkpoint molecule B7-H3 compared to PD-L1 in human retinoblastoma tissues

We examined the expression of tumor antigens GD2 and CD171, along with immune checkpoint proteins B7-H3 and PD-L1, in enucleated RB tumor tissues, with the results detailed in Table 3. The immunohistochemical (IHC) analysis of immune checkpoint molecules, including B7-H3 and PD-L1, was performed using either manual optimization techniques or the automated DAKO platform for detecting PD-L1. The staining for B7-H3 showed the highest expression levels in retinoblastoma (RB) tumor tissues (Table 3), being present in 51.06% of the samples (48 out of 94). Neuroblastoma tissue was used as a positive control for B7-H3 staining (Fig. 2a). While 46 of the 94 RB samples showed no staining for B7-H3 (Fig. 2b), those with positive B7-H3 expression in RB tumor cells exhibited mild to moderate staining on the cell membrane (Fig. 2c,d). However, B7-H3 positivity was not observed in the normal retina region adjacent to the tumor sample.

Table 3.

Summary of immunohistochemical (IHC) staining results for B7-H3, GD2, CD171, and PD-L1 in 94 enucleated retinoblastoma tumor samples.

IHC results and interpretation			Tumor antigens
		B7-H3	GD2	CD171	PD-L1
Negative (n)	–	46	60	85	90
Positive (n)
Distribution	Intensities
Few < 25	Trace	25	15	2	0
25–50	Mild (1 +)	16	15	7	4
50–75	Moderate (2 +)	6	3	0	0
Diffuse 75–100	Moderate (2 +)	1	1	0	0
Total positive number (n)		48	34	9	4
% Positive		51.06	36.17	9.57	4.26

Fig. 2.

Immunohistochemical analysis of B7-H3 expression in FFPE retinoblastoma (RB) tissue sections. (a) Positive control showing B7-H3 staining in neuroblastoma tissue. (b) Negative B7-H3 staining in RB tissue. (c) Positive B7-H3 staining in RB with mild intensity (1 +) and distribution in less than 25% of the tissue. (d) Positive B7-H3 staining in RB with moderate intensity (2 +) and distribution in 50–75% of the tissue. (Magnification: 0.6X, 40X, 63X, and 100X).

For the PD-L1 analysis, we utilized two monoclonal antibodies: the clone 22C3 antibody, applied through an automated DAKO method, and manual staining using the clone E1L3N antibody. The automated detection with the 22C3 clone indicated that all 94 samples were negative PD-L1. To confirm these results, we conducted manual staining using another clone of anti-PD-L1 antibody (E1L3N)¹³. Placental tissue sections served as positive controls for PD-L1 E1L3N immunoreactivity (Fig. 3a). Negative PD-L1 staining was observed in 90 of the 94 retinoblastoma samples (Fig. 3b). In the 4 samples (4.26%) that tested positive for PD-L1, the staining intensity ranged from trace to mild (Fig. 3c,d).

Fig. 3.

Expression of PD-L1 in human retinoblastoma (RB) tissues. (a) Positive control showing PD-L1 staining in placental tissue. (b) Negative PD-L1 staining in RB tissue. (c) Positive PD-L1 staining in RB with 5–10% distribution. (d) Positive PD-L1 staining in RB with mild intensity (1 +) and distribution in 20% of the tissue. (Magnification: 0.6X, 40X, 63X, and 100X).

Higher levels of the tumor-associated antigens GD2 compared to CD171 in human retinoblastoma tissues

In earlier findings, we noted a high expression of GD2 in retinoblastoma tumor samples (8 samples)⁸. To confirm our previous results, we evaluated GD2 expression through IHC analysis in an expanded cohort of Thai RB patients. Ninety-four RB tumor tissue sections were stained using the anti-GD2 clone 14g2a. GD2 IHC revealed cytoplasmic and membranous staining, indicating overall GD2 expression throughout the tumor cells without subdivision into subcellular locations. Neuroblastoma tissue was used as a positive control for GD2 staining (Fig. 4a). Sixty retinoblastoma samples did not show GD2-positve (Fig. 4b), whereas GD2 was positively expressed in 34 of the 94 samples, accounting for 36.17%. The staining intensity varied from trace, mild (1 +), to moderate (2 +) (Fig. 4c,d), with most samples showing trace or mild GD2 positivity (Fig. 4c). Only four samples exhibited moderate (2 +) intensity (Fig. 4d). We did not observe GD2 positivity in the normal retina area next to the tumor sample.

Fig. 4.

Immunohistochemical analysis of GD2 expression in FFPE retinoblastoma (RB) tissue sections. (a) Positive control showing GD2 staining in neuroblastoma tissue, (b) Negative GD2 staining in RB tissue. (c) Positive GD2 staining in RB with mild intensity (1 +) and distribution in 25–50% of the tissue. (d) Positive GD2 staining in RB with moderate intensity (2 +) and distribution in 75–100% of the tissue. (Magnification: 0.6X, 40X, 63X, and 100X).

For CD171 staining, a neuroblastoma tumor sample was used as a positive control for CD171 staining (Fig. 5a). The majority of RB samples lacked CD171 expression (Fig. 5b), with only 9 samples (9.57%) testing positive for CD171 (Fig. 5c and Table 3). Positive CD171 expression was identified by weak to mild intensity of membranous staining (Fig. 5c). Furthermore, we evaluated the presence of these four antigens within the same tissue sample. The individual positivity rates for B7-H3 and GD2 are 24.47% and 14.8%, respectively (Fig. 6). Positive expression for both B7-H3 and GD2 antigens was observed in 14.8% of the samples. The incidence of double positive staining for combinations such as B7-H3/CD171, B7-H3/PD-L1, or GD2/PD-L1, as well as triple positives like B7-H3/CD171/PD-L1 or B7-H3/GD2/PD-L1, was uncommon, occurring at rates under 5% (Fig. 6). A portion of the RB samples (4.25%) exhibited positive expression for all four antigens. Notably, 32.98% of the retinoblastoma samples lacked expression for any of the four antigens examined. These data imply that there may be unidentified tumor antigens in RB tumor samples.

Fig. 5.

Immunohistochemical analysis of CD171 expression in FFPE retinoblastoma (RB) tissue sections. (a) Positive control showing CD171 staining in neuroblastoma tissue, (b) Negative CD171 staining in RB tissue. (c) Positive CD171 staining in RB with mild intensity (1 +) and distribution in 25–50% of the tissue. (Magnification: 0.6X, 40X, 63X, 100X).

Fig. 6.

A bar graph illustrates the percentage distribution of immunohistochemical staining results for four different antigens in human retinoblastoma tissues.

Furthermore, we assess the expression of tumor antigens GD2 and CD171, along with immune checkpoint proteins B7-H3 and PD-L1, in both the adjacent normal retina and enucleated RB tumor tissues. The immunostaining for GD2, CD171, and B7-H3 showed negative results in the adjacent normal retina in all cases (Fig. S2).

This study investigates potential links between the expression of specific surface antigens and the clinicopathological characteristics of patients with RB. Our findings indicated that there was no association between the positive staining of these antigens and variables such as patient age, gender, diagnosis, laterality, family history, whether the enucleation was primary or secondary, the stage of retinoblastoma, or the pathological status (Table 4). Survival analysis showed no statistically significant difference between antigen-positive and antigen-negative groups for B7-H3, GD2, and CD171 (log-rank p > 0.05) (Fig. 7a–c). However, patients with B7-H3–positive or CD171-positive tumors exhibited a trend toward better survival compared with their antigen-negative counterparts (Fig. 7a and c).

Table 4.

Correlations between B7-H3, GD2, CD171, and PD-L1 antigen expression and clinicopathological data.

Characteristics	B7-H3			GD2			CD171			PD-L1
	Positive (n = 48)	Negative (n = 46)	P value*	Positive (n = 34)	Negative (n = 60)	P value*	Positive (n = 9)	Negative (n = 85)	P value*	Positive (n = 4)	Negative (n = 90)	P value*
Gender
Male	30 (62.5%)	26 (56.5%)	0.555	20 (58.8%)	36 (60.0%)	0.911	7 (77.8%)	49 (57.6%)	0.304	0 (0.0%)	56 (62.2%)	0.024
Female	18 (37.5%)	20 (43.5%)		14 (41.2%)	24 (40.0%)		2 (22.2%)	36 (42.4%)		4 (100.0%)	34 (37.8%)
Laterality
Unilateral	27 (56.3%)	31 (67.4%)	0.267	18 (52.9%)	40 (68.7%)	0.188	7 (77.8%)	51 (60.0%)	0.457	2 (50.0%)	56 (62.2%)	0.636
Bilateral	21 (43.7%)	15 (32.6%)		16 (47.1%)	20 (33.3%)		2 (22.2%)	34 (40.0%)		2 (50.0%)	34 (37.8%)
Family history of Retinoblastoma	0 (0.0%)	2 (4.3%)	0.336	0 (0.0%)	2 (3.3%)	0.429	0 (0.0%)	2 (2.4%)	0.601	0 (0.0%)	2 (2.2%)	0.351
Clinical staging
Group E	43 (89.6%)	40 (87.0%)	0.692	30 (88.2%)	53 (88.3%)	1.000	8 (88.9%)	75 (88.2%)	1.000	4 (100.0%)	79 (87.8%)	0.705
Extraocular	5 (10.4%)	6 (13.0%)		4 (11.8%)	7 (11.7%)		1 (11.1%)	10 (11.8%)		0 (0.0%)	11 (12.4%)
Pathological data
Tumor in anterior chamber	11 (22.9%)	5 (10.9%)	0.120	4 (11.8%)	12 (20.0%)	0.307	4 (44.4%)	12 (14.1%)	0.043	1 (25.0%)	15 (16.9%)	0.824
Vitreous seeding	17 (35.4%)	12 (26.1%)	0.328	9 (26.5%)	20 (33.3%)	0.489	4 (44.4%)	25 (29.4%)	0.451	2 (50.0%)	27 (30.3%)	0.564
Choroidal invasion
Focal	7 (14.6%)	9 (19.6%)	0.158	5 (14.7%)	11 (18.3%)	0.863	1 (11.1%)	15 (17.6%)	0.883	0 (0.0%)	16 (17.8%)	0.223
Massive	14 (29.2%)	6 (13.0%)		8 (23.5%)	12 (20.0%)		2 (22.2%)	18 (21.2%)		1 (25.0%)	19 (21.3%)
Scleral invasion	3 (6.3%)	2 (4.3%)	1.000	1 (2.9%)	4 (6.7%)	0.650	0 (0.0%)	5 (5.9%)	1.000	0 (0.0%)	5 (5.6%)	0.862
Optic nerve involvement
Prelaminar	21 (44.7%)	19 (42.2%)	0.931	13 (40.6%)	27 (45.0%)	0.695	6 (66.7%)	34 (41.0%)	0.336	1 (25.0%)	39 (44.8%)	0.362
Postlaminar	9 (19.1%)	8 (17.8%)		5 (15.6%)	12 (20.0%)		1 (11.1%)	16 (19.3%)		0 (0.0%)	17 (19.5%)
Surgical margin of optic nerve involvement	1 (2.1%)	2 (4.3%)	0.613	1 (2.9%)	2 (3.3%)	1.000	0 (0.0%)	3 (3.5%)	1.000	0 (0.0%)	3 (3.4%)	0.917

*Chi-Square tests and Fisher’s Exact tests.

Fig. 7.

Kaplan–Meier survival analysis of B7-H3, GD2, and CD171. Kaplan–Meier curves showing overall survival (OS) stratified by negative vs. positive expression of B7-H3 (a), GD2 (b), and CD171 (c) in retinoblastoma tumor samples. Survival differences were compared using the log-rank test.

Discussion

Retinoblastoma (RB) is categorized as a "cold tumor," marked by minimal tumor-infiltrating lymphocytes, limited presence of CD8 + T-cells, absence of PD-L1 expression within the tumor microenvironment¹⁴, and low HLA expression. CAR T-cell therapy can circumvent the need for MHC-restricted T-cell recognition, thereby activating immune cells and effectively destroying tumor cells⁷. Numerous studies have tried to develop CAR T-cell therapy for RB^8,10,11. This therapy presents an alternative for treating RB that is resistant to chemotherapy, with the potential benefits of preserving the eye and minimizing both the side effects associated with high-dose systemic chemotherapy and the risk of developing secondary cancers. The initial step in CAR-T cell therapy or immune checkpoint blockade therapy involves identifying surface antigens and/or immune checkpoint molecules in patient RB tumor samples. The ideal target antigen for CAR T-cell therapy should be a tumor antigen located on the extracellular surface of the tumor cell membrane, exhibiting high expression levels on tumor cells while having limited or low-level expression in normal cells. We used two human retinoblastoma cell lines, Y79 and WERI-Rb-1, as well as a normal retinal pigment epithelial cell line, ARPE-19, for surface staining of all four antigens. Both human RB cell lines showed high expression levels of GD2 and CD171, whereas their expression was low in normal retinal epithelial cells, suggesting that GD2 and CD171 could serve as target antigens for CAR T-cell therapy. Additionally, we investigated the expression of immune checkpoint molecules PD-L1 and B7-H3, as specific antibodies are available to block these proteins, which have been approved for use in other types of tumors³⁰. Interestingly, we discovered that high levels of B7-H3 expression on the cell surface were present exclusively in WERI-Rb-1 cells, with no expression in the Y79 cells. The varying levels of B7-H3 expression in the two RB cell lines are still uncertain. Other studies showed that B7-H3 inhibits T-cell activation and functions²⁹. In comparison, the two RB cell lines showed slightly lower levels of PD-L1 on their surfaces. However, we observed varying expressions of B7-H3 and high level of PD-L1 in ARPE-19 normal retinal epithelial cell line. The expression PD-L1 is variable as it often responds to immune cell activation or cytokines present in the tumor microenvironment²⁷. These results suggest that GD2, CD171, B7-H3, and PD-L1 may be present in human RB tumor tissues.

The increased expression of immune checkpoint molecules such as PD-L1 and B7-H3 is known to inhibit T-cell antitumor activity, potentially diminishing the effectiveness of T-cell and CAR T-cell therapies^8,28. B7-H3 is an immune checkpoint molecule associated with negative clinical outcomes such as tumor growth, metastasis, treatment resistance²⁹. Ganesan et al., demonstrated significant increased expression of B7-H3 in RB tumors compared to normal retinal tissue by immunoblotting¹⁷. All RB tumor samples (n = 35) had B7-H3 expression at varying intensities between the tumor lobules and blood vessels¹⁷. On the contrary, our immunohistochemical analysis showed that more than half (51.06%; 48/94) of the RB tissue samples expressed B7-H3 but did not observe in the blood vessels. The variation in expression levels observed is likely due to differences in the primary anti-B7-H3 antibody used for IHC staining between their study and ours. Nonetheless, this information encouraged us to confirm B7-H3 as a potential target for RB immunotherapy.

Targeting immune checkpoint molecules, especially PD-L1, has shown potential in cancer treatment with immune checkpoint blockade antibodies³⁰. We are interested in PD-L1 expressions in RB tumors because our previous research demonstrated that the RB cell line can increase PD-L1 expression to suppress the anti-tumor activity of GD2-CAR T-cells⁸. However, recent findings showed that PD-L1 was either undetectable or present at low levels (4.26%) in RB tumor tissues when using the anti-PD-L1 clones 22C3 and E1L3N, respectively. In contrast to the study by Sing L et al., they found that PD-L1 was expressed in 46 out of 144 (31.94%) patients with primary retinoblastoma and in 22 out of 188 (18.64%) patients with chemo-reduced retinoblastoma¹³. Although both their study and ours utilized the same anti-PD-L1 clone E1L3N, we proposed that the varying stages of tumor sample collection and the extensive chemotherapy administered to RB patients before collecting the enucleated tumor samples could potentially impact PD-L1 expression. Given the limited existing literature on PD-L1 expressions in RB, additional research is necessary to clarify the unclear findings and enhance our understanding of their role in the disease. Nonetheless, we propose that employing immune checkpoint inhibitors targeting B7-H3 and/or PD-L1 could be advantageous for treating RB.

The glycolipid antigen GD2 is prominently expressed in cancers such as neuroblastoma, glioblastoma, osteosarcoma, and retinoblastoma^{8,10,11,15,16,18}. Several studies confirmed the presence of GD2 gangliosides in tumor tissues^8,10,15,18 and the serum of RB patients²¹. GD2 synthase mRNA has been identified in the blood and cerebrospinal fluid of patients with extraocular metastasis of RB²². Our recent data showed that 36.17% (34 out of 94) of tumor samples showed positive GD2 staining. We utilized the anti-GD2 monoclonal antibody clone 14G2.a, which is commonly employed for surface staining in flow cytometry and immunohistochemistry⁸. Nonetheless, our findings offer valuable insights into the further development of CAR-T cells targeting this antigen. CD171, also known as L1CAM, is a surface antigen associated with neurogenesis, neuronal migration, axon guidance, and synapse formation, and plays a significant role in promoting tumor cell invasion^12,23–28,. It was found to be expressed at a low level (9.57%) in our study. This result contrasts with Jo et al.'s findings, which reported CD171 expression in 26 out of 30 enucleated RB samples²³. Their research indicates that CD171 promotes RB cell proliferation and chemoresistance through adhesion-mediated mechanisms in both in vitro and in vivo settings. Conversely, another study found CD171 to be positive in 15 out of 30 primary RB tissue (50%) samples, showing variability in frequency. These differences may stem from variations in immunostaining techniques and the specific antibody clones utilized. Despite these inconsistencies, the overall findings suggest that targeting CD171 could be a viable strategy for treating chemotherapy-resistant RB, emphasizing the need to understand tumor heterogeneity and its therapeutic implications.

The study identified varying expression levels of B7-H3, GD2, CD171, and PD-L1 in human retinoblastoma using enucleated samples. Consistent with other studies^15,16, the data demonstrated significant expression of B7-H3 in human retinoblastoma tumors, indicating its potential as a promising target for CAR T-cell therapy in retinoblastoma. We suggest two approaches to target B7-H3: employing either CAR T-cells specifically designed for B7-H3 or utilizing B7-H3 bispecific T-cell engagers^33,34. Nonetheless, there is currently no documentation of B7-H3-targeted immunotherapy being used for RB. Our unpublished findings have shown that B7-H3 T-cell engagers exhibit strong anti-tumor effects against RB tumor cell lines in vitro. For RB therapy, we propose that an immunotherapy strategy targeting a single antigen, using either antibodies or CAR T-cells, might not be sufficient to achieve clinical effectiveness. Numerous reports have underscored various strategies to tackle these challenges, including the dual targeting of B7-H3 and GD2 using bispecific antibodies and CAR T-cells in other types of tumors^35,36. The hypothesis suggests that using a combination of dual-targeting bispecific antibodies and/or CAR T-cells could enable the targeting of tumor cells that express either B7-H3 or GD2 individually, which represents 75% of RB tumors. Nonetheless, this study has several limitations. For instance, manual optimization and staining are labor-intensive processes that require specialized expertise, and the outcomes might be dependent on antibodies. Additionally, individual antigen staining does not allow for the co-expression analysis of more than two target antigens simultaneously. To investigate this matter, we recommend using multiplex immunofluorescence staining followed by analysis with a high-resolution imaging tool. Additionally, choosing candidate target antigens may miss other potentially significant proteins present in the tumors. This is evidenced by this study’s finding that nearly a third of the RB samples did not express any of the four targeted antigens. This indicates the presence of tumor antigen heterogeneity in retinoblastoma. We recommend further exploration to identify alternative targets, validating them through immunohistochemistry or employing other approaches like membrane proteome analyses^37–39.

Importantly, our findings provide strong evidence of tumor antigen heterogeneity in retinoblastoma. We suggest that employing treatment strategies involving multi-specific CAR T-cell therapies or combining CAR T-cell therapy with immune checkpoint blockade antibodies, could enhance RB immunotherapy. This research offers valuable insight into the potential of CAR T-cell therapy to preserve the eye, treat extraocular or metastatic RB, and reduce the side effects of high-dose systemic chemotherapy, particularly the risk of secondary cancers.

Materials and methods

Retinoblastoma samples and ethical statement

A retrospective study was performed on retinoblastoma (RB) tumor samples and medical records from 92 children diagnosed with retinoblastoma between 2006 to 2023, with a total of 94 eyes enucleated. Formalin-fixed paraffin-embedded (FFPE) specimens were collected and stored by the Department of Pathology at the Faculty of Medicine Siriraj Hospital, Mahidol University. All FFPE specimens were reposited at the Department of Pathology, with access restricted to authorized people. Demographic and clinical data of the patients were retrieved and recorded using clinical record forms. Data analysis commenced following the completion of immunohistochemistry and clinicopathological data collection.

All methods were carried out in accordance with all relevant guidelines and regulations. Prior to data collection, approval was obtained from both the Medical Records Office and the Institutional Review Board of the Faculty of Medicine Siriraj Hospital (IRB approval COA no. Si189/2023), as authorized by the Siriraj Ethical Committee, Faculty of Medicine Siriraj Hospital, Mahidol University. The study was conducted in accordance with the principles of the Declaration of Helsinki for biomedical research involving human subjects. Due to the retrospective nature of the study, the Siriraj Ethical Committee, Faculty of Medicine Siriraj Hospital, Mahidol University, waived the need to obtaining informed consent.

Retinoblastoma cell lines and culture

The Y79 (ATCC Cat# HTB-18, RRID: CVCL_1893), and WERI-Rb-1 (ATCC Cat#HTB-169, RRID: CVCL_1792), are human retinoblastoma cell lines. These cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco Life Technologies), supplemented with 10% heat-inactivated fetal bovine serum (FBS, Hyclone), 100 U/mL penicillin, and 100 mg/mL streptomycin (Invitrogen). ARPE-19 is a normal retinal pigmented epithelia cell (ATCC Cat#CRL-2302, RRID: CVCL_0145) cultured in DMEM/F12 medium supplemented with 10% FBS and penicillin/streptomycin.

Antibodies for flow cytometry

Phycoerythrin (PE)-conjugated antibodies (Abs) to GD2 (clone 14.G2a) and PE-conjugated anti-CD171 antibodies (clone L1-OV198.5) were purchased from BioLegend (San Diego, CA, USA). Allophycocyanin APC conjugated anti-B7-H3 (clone MIH42), APC-conjugated Abs to CD274 (PD-L1) (clone 29E.2.A.3), were purchased from BioLegend (San Diego, CA, USA).

Retinoblastoma tissue immunohistochemistry

Immunohistochemistry on the studied specimens was conducted in the Laboratory of the Department of Pathology at the Faculty of Medicine Siriraj Hospital, Mahidol University. Two ocular pathologists independently evaluated the immunostaining results, categorizing them as either positive or negative. The immunohistochemistry protocols for each surface antigen were strictly adhered to. To prepare the formalin-fixed paraffin-embedded (FFPE) human RB tissue sections, the samples were deparaffinized through a series of washes with xylene, followed by 100%, 90%, and 70% ethanol, each wash lasting for 5 min. After deparaffinization, the sections were rehydrated in phosphate-buffered saline (PBS). Antigen retrieval was performed in each section using various buffers and methods, as detailed in the Supplementary Table S1. The immunohistochemistry (IHC) was conducted manually, following the institute’s standards and the manufacturer’s guidelines. The primary antibodies included the anti-GD2 monoclonal antibody (clone 14.G2a; BD Pharmingen), anti-CD171 monoclonal antibody (mAb clone UJ127.11; Invitrogen), anti-B7-H3 mAb (clone F-11; Santa Cruz Biotechnology), anti-PD-L1 mAb (clone E1L3N; Cell Signaling Technology), and anti-PD-L1 mAb (clone 22C3; Dako). After the primary antibody incubation, the antigen–antibody complexes were detected using Dako EnVision anti-rabbit/mouse horseradish peroxidase (Dako). The sections were developed with diaminobenzidine (DAB) at room temperature for 5 min and subsequently counterstained with hematoxylin (Dako). They were then dehydrated in 70%, 90%, and 100% ethanol before being mounted with a coverslip. Positive control tissue samples were included for each antigen staining, as detailed in the Supplementary methods.

Assessment of immunohistochemical results

Tissue sections were independently examined by two ocular pathologists (K.J. and M.U.), who were unaware of each other’s findings. All immunostaining is evaluated in viable, non-necrotic areas. Immunopositivity of anti-GD2, anti-CD171, and anti-B7-H3 staining are reported as positive if the membrane staining of tumor cells is identified. Distribution of positively stained cells is grouped as < 25%, 26–50%, 51–75% and 76–100% and intensity of positively stained cells is categorized as trace (very faint, incomplete membrane staining), mild/1 + (mild incomplete membrane staining), moderate/2 + (moderate incomplete staining), marked/3 + (marked complete membrane staining). Expression of anti-PD-L1 (clone E1L3N) and anti-PD-L1 (clone 22C3) is reported as positive if any membrane staining of tumor cells is identified. Distribution of positively stained cells is shown as percentage of all tumor cells. The intensity of staining is identified as trace (very faint, incomplete membrane staining), 1 + (mild incomplete membrane staining), 2 + (moderate incomplete staining), 3 + (marked complete membrane staining).

Sample size calculation

Based on a previous study by Jo DH et al. CD171 was expressed in 26 out of 30 samples (87%)²³. The following formula was used to calculate the required sample size: The confidence level was set at 95% (1-alpha), corresponding to a Z score of 1.96. The rate of proportion (P) was 0.87, and the allowable error (d) was 0.1 (d). As a result, a total sample size of 44 was determined to be necessary²³.

Statistical analysis

The Kolmogorov–Smirnov test was employed to assess the normality of the numerical dataset. For the analysis of quantitative data, percentage expression was used to describe the expression of RB surface antigens in both primary and secondary enucleated specimens. The relationship between the expression of RB surface antigens and clinicopathological data was analyzed using either the Chi-Square test or Fisher’s exact test for qualitative variables. The parameters used for clinical and histopathological high-risk analysis include gender, laterality, family history of retinoblastoma, clinical staging, and pathological data including tumor in anterior chamber, vitreous seeding, choroidal invasion, scleral invasion, optic nerve involvement, and surgical margin of optic nerve involvement. For quantitative variables, including age and surface antigen expression, the Mann–Whitney U test was used. A P value of less than 0.005 (after Bonferroni correction) was considered statistically significant. All data analyses were conducted using SPSS version 29 (IBM).

Abbreviations

B7-H3

B7 Homolog 3

CAR

Chimeric antigen receptor

CD171

Cluster of Differentiation 171

FFPE

Formalin-fixed paraffin-embedded

GD2

Disialoganglioside

IHC

Immunohistochemistry

L1CAM

L1 cell adhesion molecule

MHC

Major histocompatibility complex

PD-l

Programmed death-1

PD-L1

Programmed death ligand-1

RB

Retinoblastoma

Author contributions

W.N. and J.S. contributed to the design of the experiments, analyzed the original data, and wrote the original draft. K.K. performed all immunohistochemical (IHC) staining. K. J. and M.U. validated and interpreted the IHC data. K. C. and S. S. collected and validated the clinical data. S.U. performed statistical analysis. M.J. optimized IHC staining and provided resources. P.Y. and L.A. contributed to the overall design and supervision of this study. All authors reviewed the manuscript.

Funding

This work was financially supported by the National Research Council of Thailand (NRCT), Mahidol University (grant number N42A650364), and the Siriraj Foundation (D1671). JS received support from the Mid-Career Grant, Fiscal year 2565 (grant number N42A650364). KK was supported by a Siriraj Research Grant from the Faculty of Medicine Siriraj Hospital, Mahidol University. Additional support was provided by the Siriraj Research Funds (grant number R016634006 and R016737004) from the Faculty of Medicine Siriraj Hospital, Mahidol University. JS, KK, MJ, and PY also received funding for research and innovation activities by National Research Council of Thailand (NRCT) (grant no. N34E670096), channeled through the Thailand Hub of Talents in Cancer Immunotherapy (TTCI).

Data availability

Data is available on request due to privacy/ethical restrictions but is available from La-ongsri Atchaneeyasakul (email: atchanee@hotmail.com) on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.