Adoptive cell therapy in paediatric extracranial solid tumours: current approaches and future challenges

Abstract

Immunotherapy has ignited hope to cure paediatric solid tumours that resist traditional therapies. Among the most promising methods is adoptive cell therapy (ACT). Particularly, ACT using T cells equipped with chimeric antigen receptors (CARs) has moved into the spotlight in clinical studies. However, the efficacy of ACT is challenged by ACT-intrinsic factors, like lack of activation or T cell exhaustion, as well as immune evasion strategies of paediatric solid tumours, such as their highly immunosuppressive microenvironment. Novel strategies, including ACT using innate-like lymphocytes, innovative cell engineering techniques, and ACT combination therapies, are being developed and will be crucial to overcome these challenges. Here, we discuss the main classes of ACT for the treatment of paediatric extracranial solid tumours, reflect on the available preclinical and clinical evidence supporting promising strategies, and address the challenges that ACT is still facing. Ultimately, we highlight state-of-the-art developments and opportunities for new therapeutic options, which hold great potential for improving outcomes in this challenging patient population.

Highlights

• •Different paediatric solid tumours share targetable common antigens.
• •Expanded and/or engineered αβ T, γδ T, NK, and NKT cells are explored ACT sources.
• •Pronounced immunosuppression hinders ACT efficacy.
• •New engineering techniques and combination therapies could bypass immunosuppression.
• •A disparity exists between target identification and their clinical implementation.

1. Introduction

Immunotherapy is revolutionising the field of adult oncology, leading to improved survival in some poor-prognosis tumours [1]. It also holds great promises for paediatric solid tumour treatment, where immune cell infiltration is associated with a favourable prognosis [2], [3], [4]. However, a significant clinical benefit of immunotherapy is not yet achieved. The exception is the anti-GD2 monoclonal antibody (mAb) for high-risk neuroblastoma, integrated in the standard of care, which has considerably improved patient outcome [5], [6]. Although antibody-based therapies are the most explored approach for immunotherapy [7], their efficacy is still poor in childhood cancers [8], [9], [10]. Cellular immunotherapies offer a highly personalised therapeutic option with the benefit of a long-term immune protection and may provide an option for patients who still develop disease progression upon conventional care. To enhance tumour recognition and effector function of immune cells, ACT uses patient- or donor-derived cells that are activated, expanded ex vivo, and reinfused into the patient, with or without prior engineering. αβ T cells are the most explored cell subset for paediatric solid tumour ACT. Other, more innate/innate-like subsets, such as natural killer (NK), invariant NKT (iNKT), and γδ T cells, have also been used (Fig. 1).

Fig. 1.

Summary of the ACT strategies currently used for the treatment of paediatric extracranial solid tumours: conventional αβ T cells, NK, iNKT, and γδ T cells are currently evaluated as an adoptive cell strategy for different paediatric solid tumours. Briefly, αβ T cells can be (1) expanded and reinfused in the patient (CTL/TIL); (2) engineered to express a CAR, also in combination with safety switches and cytokine expression; (3) engineered with a tumour-specific TCR; and (4) engineered with a CAR and trained to express NK-cell markers. NK cells can be (1) adoptively transferred in the patients, often in combination with tumour-targeting mAb; (2) modified to express homing receptors; or (3) CAR-like constructs. iNKT cells are either (1) adoptively transferred or (2) engineered with a CAR, also in combination with cytokine production. Lastly, γδ T cells are currently (1) adoptively transferred, engineered (2) with a CAR, or (3) with an αβ TCR, and (4) their γδ TCR could be transduced in αβ T cells to generate the so-called TEG cells. CAR, chimeric antigen receptors; CIK, cytokine-induced killer; CTL, cytotoxic T lymphocytes; iNKT, invariant NKT; NK, natural killer; TIL, tumour-infiltrating lymphocytes.

In paediatric oncology, ACT has proven remarkably successful in treating haematological malignancies [11], [12] but still meets challenges in solid tumours [13], [14]. The limited success is related to (1) an immunosuppressive tumour microenvironment (TME), characterised by suppressive immune cells such as regulatory T cells, tumour-associated macrophages (TAM), myeloid-derived suppressor cells (MDSC), and their production of immunosuppressive mediators [15], [16]; (2) the paucity of tumour neoantigens and/or an impaired tumour antigen presentation [17], [18]; (3) a low mutational burden [19], [20]; and (4) poor trafficking to the tumour site [21], [22].

In this review, we discuss the current progress in ACT for paediatric extracranial solid tumours. We address the different immune cell subsets currently used for ACT, the state-of-the-art approaches used to overcome challenges, and future perspectives to improve ACT efficacy while maintaining safety.

2. Conventional αβ T cells

2.1. αβ T cells isolated from tumour (TILs)

One widely explored ACT uses tumour-infiltrating lymphocytes (TILs) isolated from patients’ tumour. Patients possess T cell clones that could potentially recognise tumour antigens, but their amount is limited, and they fail to be activated because of the immunosuppressive TME [23]. To reactivate them, isolated T cells are expanded ex vivo through rapid expansion protocols, tested for tumour reactivity, and reinfused into patients [23]. However, this strategy holds some limitations, including limited expansion of the cells after reinfusion and the reliance on MHC/HLA expression for antigen recognition, expressed at low levels in many paediatric solid tumours [18].

TIL therapy was pioneered in advanced melanoma where it demonstrated striking efficacy, resulting in long-lasting cure in some patients [23], [24]. It has also demonstrated its efficacy for adult sarcoma (NCT04052334, NCT03725605, NCT03935893) and new studies are recruiting young adults in their patient cohort (NCT03449108). TILs have been efficiently isolated and expanded from neuroblastoma (the most common extracranial solid tumour in children [25]) in the presence of high-dose IL-2 [26], but evidence for TIL therapy efficacy in other paediatric solid tumours is limited. Since TIL efficacy in these tumours may be hampered by the low neoantigen and MHC-1 expression, their efficacy could be improved by transduction with a CAR (CAR-TILs). This strategy could combine the natural tropism of the TILs to the tumour with enhanced tumour reactivity [26].

2.2. Antigen-specific αβ T cells isolated from peripheral blood

Tumour-reactive cytotoxic T lymphocytes (CTLs) can also be isolated from the patients’ peripheral blood. CTL therapy for paediatric extracranial solid tumours has been so far used to target Epstein–Barr virus (EBV)-infected carcinoma cells, cancer/testis antigens (CTAs), and other solid tumour antigens, which we will discuss here.

2.2.1. EBV

CTLs have been used for the treatment of EBV-associated paediatric nasopharyngeal carcinoma and are currently evaluated in different clinical trials (Table 1). Data from two independent EBV-CTL clinical trials (NCT00078546, NCT00609219) suggest a benefit, with 2/7 evaluable paediatric patients achieving complete responses (CR), 2/7 partial responses (PR), 2/7 progressive disease (PD), and 1/7 no response with no/limited toxicity. Another trial (NCT01498484) reported a mean overall response through 12 months of 64%. Of note, these clinical benefits are more pronounced in patients with local disease [27], [28], [29], [30], [31]. Due to their success in trials, EBV-CTLs are already suggested for clinical implementation for refractory/relapsed disease forms [32].

Table 1.

Summary of clinical trials using CTL/TIL for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT00953420	Carboplatin and docetaxel + EBV-CTL	Nasopharyngeal carcinoma	10 years and older	Baylor College of Medicine	Completed	2	Not available
NCT00609219	EBV-CTL	Head and neck cancer	Children, adults, older adults	Baylor College of Medicine	Completed	1	Five out of 10 patients <18 years. Of these five: one CR, two PR, one no response, one NA. No/limited toxicity.
NCT01498484	EBV-CTL	EBV-induced lymphomas, EBV-associated malignancies	Children, adults, older adults	Atara Biotherapeutics	Completed	2	Mixed ORR at 65.3 months depending on trial arm (from 14.3% to 68%). OR through 12 months ∼64% (mean of all arms). Limited serious adverse events.
NCT00078546	EBV-CTL + anti-CD45 Mab	Nasopharyngeal cancer and EBV infections	Children, adults, older adults	Baylor College of Medicine	Completed	1	Three out of eight patients <18 years. Of these three: one CR >24 months and two PD.
NCT01447056	LMP CTL	Nasopharyngeal carcinoma and others	Children, adults, older adults	Baylor College of Medicine	Completed	1	Not available
NCT00516087	LMP1- and LMP2-specific CTLs	Nasopharyngeal carcinoma	Children, adults, older adults	Baylor College of Medicine	Completed	1	Not available
NCT02287311	MABEL (EBV-CTL)	Nasopharyngeal carcinoma and others	Children, adults, older adults	Baylor College of Medicine	Recruiting	1	Not available
NCT02822495	Tabelecleucel (EBV-CTL)	Nasopharyngeal carcinoma, EBV infections, EBV+ associated lymphoma, others	Children, adults, older adults	Atara Biotherapeutics	Temporarily not available	NA	Not available
NCT04554914	Tabelecleucel (EBV-CTL)	EBV-associated diseases	Children, adults, older adults	Atara Biotherapeutics	Recruiting	2	Not available
NCT02065362	TGF-beta resistant EBV-CTL	Nasopharyngeal carcinoma	Children, adults, older adults	Baylor College of Medicine	Active, not recruiting	1	Not available
NCT03449108	TILs (LN-145 or LN-145-S1)	Sarcomas and others	16–70 years	M.D. Anderson Cancer Center	Recruiting	2	Not available
NCT02239861	Tumour-associated antigen (TAA)-CTLs	Rhabdomyosarcoma	2–80 years	Baylor College of Medicine	Completed	1	Not available
NCT02789228	TAA-CTLs	Ewing sarcoma, Wilms tumour, neuroblastoma, rhabdomyosarcoma, soft tissue sarcomas, osteosarcoma, adenocarcinoma, and oesophageal carcinoma and renal cell carcinoma.	6 months–60 years	Children's National Research Institute	Active, not recruiting	1	11 out of 15 therapy responders. No disease progression in six out of 11 responders. No DLT and no infusion-related adverse events.

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

CR, complete responses; CTL, cytotoxic T lymphocytes; DLT, dose limiting toxicity; EBV, Epstein–Barr virus; OR, overall response; ORR, objective response rate; PD, progressive disease; PR, partial responses; TIL, tumour-infiltrating lymphocytes.

2.2.2. CTAs and other solid tumour antigens

CTAs are a large family of antigens that are only expressed during embryonic development and in adult male germ cells but are (re-)expressed by several malignancies, which makes them attractive targets for ACT [33]. PRAME, NY-ESO-1, MAGE, and SSX are CTAs that could be suitable immunotherapy targets for paediatric solid malignancies, especially for neuroblastoma and sarcomas [34], [35], [36], [37]. Additional immunotherapy targetable non-CTA antigens are WT1, a typical antigen for Wilms tumours [38], [39] and survivin, an antiapoptotic protein expressed by many malignancies [25].

PRAME-specific CTL killing activity was increased in vitro with sarcoma and neuroblastoma cell lines, by combining them with 5-aza-2′-deoxycytidine (DAC) administration [40]. DAC leads to DNA demethylation and reactivation of MHC gene expression and, consequentially, to increased antigen presentation on tumour cells [40]. CTL activity against PRAME, WT1, and survivin is being evaluated in an ongoing phase I clinical trial for relapsed/refractory solid tumours (NCT02789228). To date, this trial showed safety and efficacy, with 11/15 patients responding to the therapy but still undergoing PD, and 6/11 responders experiencing no disease progression at a median of 13.9 months [41]. Another clinical trial is currently evaluating the efficacy of PRAME, NY-ESO-1, MAGE-A4, survivin, and SSX-CTL for the treatment of rhabdomyosarcoma (RMS) (NCT02239861), with no data posted yet.

CTL and TIL therapies represent a valuable tool for the treatment of these malignancies, also given the ease of cell manufacturing compared to other T cell engineering techniques and limited side effects. However, the presence of tumour-reactive T cells is limited in paediatric solid tumours [42], due to often low/absent MHC expression [18] and low mutational burden [19]. Strategies to increase tumoral MHC expression by pharmaceutical intervention are currently being explored and could greatly benefit future CTL/TIL therapies [43].

2.3. TCR-engineered T cells (TCR-T)

Both CTL and TIL therapies rely on naturally occurring T cell populations in the cancer patients, which are not always sufficiently present. This can be circumvented by artificially generating T cells with tumour-specific TCRs but will not overcome the problem of absent MHC. Still, with novel strategies to enhance tumoral MHC expression [44] and novel neoantigens being identified [45], [46], TCR-T remain a valuable option. Crucial for TCR-T therapy is the selection of highly tumour-specific antigens to balance maximum tumour killing with minimal toxicities.

2.3.1. Cancer testis antigens

As already mentioned, CTAs are attractive targets for ACT. NY-ESO-1 TCR-T cells have been exploited in clinical trials for the treatment of paediatric and adult synovial cell sarcomas [47]. In one trial, 4/6 adult patients treated with NY-ESO-1-engineered T cells + IL-2, reported significant clinical responses with one patient achieving 18-months PR (NCT00670748) [48]. However, in a trial in paediatric and adult patients with synovial sarcoma, CR was observed only in 1/44 patients and PR in 14/44 patients (NCT01343043). This trial suggested that efficacy was dependent on persistent stem-cell memory (TSCM) and central memory (TCM) T cell subsets in patients (NCT01343043) [49]. TSCM can replenish the TCM pool, which can quickly convert into effector cells upon encountering NY-ESO-1 antigen, favouring rapid tumour cell clearance [50], [51]. Preclinically, T cells targeting NY-ESO-1 have demonstrated their efficacy in neuroblastoma models, controlling both localised and disseminated tumour progression [52].

Other CTAs suitable for neuroblastoma and sarcoma are PRAME and MAGE-A [34], [35], [36], [37]. MAGE-A TCR-T are currently being evaluated only in a clinical trial for adult malignancies, such as synovial sarcoma (NCT03132922).

Despite the attractiveness of CTAs for ACT, current trials targeting NY-ESO-1 for paediatric solid tumours have yet to demonstrate a clear therapeutic benefit (Table 2). Studies exploiting other CTAs or combination therapies with TME-targeting compounds represent some possibilities to increase efficacy.

Table 2.

Summary of clinical trials using TCR-engineered αβT cells for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT03462316	NY-ESO-1 TCR	Sarcomas	14–70 years	Sun Yat-sen University	Active, not recruiting	1	Not available
NCT02775292	NY-ESO-1 TCR + vaccine therapy + nivolumab	Adult and childhood solid neoplasms	16 years and older	Jonsson Comprehensive Cancer Center	Completed	1	Not available
NCT02650986	NY-ESO-1 TCR/dnTGFbetaRII transgenic T cells	Synovial sarcoma and others	12 years and older	Roswell Park Cancer Institute	Active, not recruiting	1, 2	Only one patient <18 years old. Only toxicity results available. High-grade CRS in four out of 15 patients.
NCT01343043	NY-ESO-1-TCR	Neoplasms	4 years and older	GlaxoSmithKline	Completed	1	CR in one out of 44 patients. 14 PR, 16 SD, and three PD. High-grade CRS in five out of 50 patients.
NCT02457650	NY-ESO-1-TCR	Neuroblastoma, synovial sarcoma, others	1 year and older	Shenzhen Second People’s Hospital	Unknown	1	Data available for four adult patients. One PR, one SD, and two NR. Grade 1–3 adverse events
NCT03240861	NY-ESO-1-TCR	Sarcoma, other NY-ESO-1+ malignancies	16 years and older	Jonsson Comprehensive Cancer Center	Recruiting	1	Not available
NCT02869217	NY-ESO-TCR	Synovial sarcoma and others	16 years and older	University Health Network, Toronto	Active, not recruiting	1	Two PR, five SD, one PD, and one pending. CRS 1/2 in five out of nine patients.

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

CR, complete responses; CRS, cytokine release syndrome; NR, non-responder; PD, progressive disease; PR, partial responses; SD, stable disease.

2.3.2. Other antigens

Preclinical data suggest that pappalysin-1 (PAPPA) and six-transmembrane epithelial antigen of prostate-1 (STEAP1) are potential targets for paediatric Ewing sarcomas. Ewing sarcoma xenograft models treated with either PAPPA [53] or STEAP1 TCR-T cells [54] exhibited a significant reduction in tumour growth. For paediatric nasopharyngeal carcinoma, TCR-T can be engineered against the EBV antigens latent membrane protein 1 and 2 (LMP1/2). Such EBV-LMP2 TCR-T lysed LMP2⁺ target cells [55] and controlled tumour growth in mice bearing LMP2⁺ human epithelial tumours [56]. EBV TCR-T are currently evaluated in clinical trials for adult nasopharyngeal carcinomas with no results posted yet (NCT03648697, NCT03925896, NCT04509726).

Overall, TCR-T represent a promising tool for the treatment of paediatric extracranial solid tumours, but further studies are needed to explore more targets and to potentiate current cell products. Furthermore, low-MHC expression limits the usage of TCR-T and contributes to the popularity of CAR-engineered αβ T cells (CAR-T), which are currently the most investigated ACT product for paediatric extracranial solid malignancies.

2.4. CAR-engineered αβ T cells (CAR-T)

CAR-T recognise tumour surface antigens MHC-independently, which circumvents MHC restriction but limits the number of available targets to surface molecules. First-generation CAR-T are engineered with a receptor composed of an extracellular antibody-derived antigen-recognising part (single-chain variable fragment [scFv]), a transmembrane portion, and an intracellular part composed of a CD3ζ domain responsible for signal transduction [57]. They displayed limited persistence because of a lack of costimulation [58]. To improve this, second- and third-generation CAR-T have been designed with one or more costimulatory domains like 4-1BB, CD28, and OX-40 [59]. Lastly, new and advanced design strategies and combinatorial approaches have been implemented to further improve CAR-T efficacy, persistence, proliferation, and safety, such as the transduction with cytokines and the addition of safety switches [60], [61]. These innovative strategies also try to overcome the pronounced immunosuppressive TME present in many paediatric solid tumorous, which can impair CAR-T antitumour activity.

2.4.1. First-generation CAR-T

The first CAR-T tested in a clinical trial for paediatric solid tumours were EBV-specific CTLs expressing a GD2-CAR in neuroblastoma patients (NCT00085930) [62], [63]. GD2 is a diganglioside widely expressed in a variety of paediatric tumours [64], [65], [66]. In this trial, 3/11 patients with active disease at the time of CAR-T infusion achieved CR, 1/11 a PR, and 7/11 exhibited either stable or PD. Most importantly, patient survival was strongly associated with CAR-T persistence in the circulation [62], [63].

First-generation CAR-T against cell-adhesion molecule L1CAM (CD171), associated with the poor prognosis of several solid tumours [67], were evaluated in a pilot study for neuroblastoma patients. In total, 1/6 patients experienced prolonged survival of 4.5 years after the first infusion; 1/6 achieved CR after additional chemotherapy; 1/6 displayed a PR with a survival of 1.5 years after the first infusion; and 3/6 suffered PD [68].

First-generation CAR-T against the foetal acetylcholine receptor, whose expression is increased in residual cells of chemotherapy-treated RMS patients, could lysate antigen-expressing tumour cells in vitro [69], but has not yet been evaluated in vivo or in patients.

In conclusion, first-generation CAR-T demonstrated a proof-of-concept in patients and paved the way for studies that focused on improving CAR-T persistence and efficacy by the addition of costimulatory domains, leading to the advent of second-generation CAR-T.

2.4.2. Second-generation CAR-T

Many of the studies investigating second-generation CAR-T sought to identify the costimulatory domain that was most beneficial for CAR-T activation, persistence, and resistance to exhaustion. Second-generation CAR-T have been generated against a wide range of target proteins on paediatric tumours, including GD2, B7-H3, L1CAM, human epidermal growth factor receptor 2 (HER2), glypican 2 and 3 (GPC2, GPC3). Some of these CAR-T are already implemented in clinical trials (Table 3).

Table 3.

Summary of clinical trials using CAR-engineered αβT cells for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT02919046	GD2 CAR-T	R/R neuroblastoma	1–14 years	Sinobioway Cell Therapy Co., Ltd., China	Unknown	NA	Not available
NCT02765243	4SCAR-GD2	R/R neuroblastoma	1–14 years	Zhujiang Hospital, China	Suspended	1	Tumour regression in 2/34 patients. After 1 year, 38% patients SD, 15% PR, 47% disease progression. No grade 3, 4 CRS.
NCT02992210		Solid tumour	1–65 years	Shenzhen Geno-Immune Medical Institute	Unknown	1,2
NCT03373097	GD2-CART01, containing suicide gene	R/R neuroblastoma, sarcomas, other GD2+ tumours	1–25 years	Bambino Gesù Hospital and Research Institute, Italy	Recruiting	1,2	9/27 CR, 8/27 PR, 5/27 SD, 5/27 no response. 60% 3-year overall survival and 36% event-free survival. 1/27 grade 3 CRS.
NCT04539366	GD2 CAR-T	R/R neuroblastoma, osteosarcoma	0–35 years	National Cancer Institute (NCI), USA	Recruiting	1	Not available
NCT04637503	GD2, PSMA, and CD276 CAR-T cells	Neuroblastoma	1–65 years	Shenzhen Geno-Immune Medical Institute, China	Recruiting	1,2	Not available
NCT02311621	CD171-specific CAR-T cells expressing EGFRt (second-generation T cells)	R/R neuroblastoma	18 months–26 years	Seattle Children's Hospital, USA	Active, not recruiting	1	Not available
	CD171-specific CAR-T cells expressing EGFRt (third-generation T cells)
	CD171-specific CAR-T cells expressing EGFRt (long spacer second-generation T cells)
NCT02761915	1RG-CART	R/R neuroblastoma	1 year and older	Cancer Research UK	Completed	1	No CR. Antitumour activity in three out of six patients associated with grade 1–3 CRS.
NCT04432649	Fourth generation CD276-specific CAR	Solid tumours	1–75 years	Shenzhen Geno-Immune Medical Institute	Recruiting	1, 2	Not available
NCT04897321	B7-H3 CAR-T cells	R/R paediatric solid tumours	0–21 years	St. Jude Children's Research Hospital, USA	Recruiting	1	Not available
NCT03635632	C7R-GD2.CART cells	Neuroblastoma, sarcomas, and others	1–74 years	Baylor College of Medicine, USA	Recruiting	1	Not available
NCT04864821	CD276 CAR-T	Osteosarcoma, neuroblastoma, and others	1–70 years	PersonGen BioTherapeutics (Suzhou) Co., Ltd., China	Unknown	early 1	Not available
NCT00085930	EBV-specific CTLs expressing GD-2 CAR	Neuroblastoma	Up to 21 years	Baylor College of Medicine	Active, not recruiting	1	CR in three out of 11 patients, one PR. Grade 1–3 localised pain in three patients.
NCT01460901	GD2 CAR-modified tri-virus CTL	Neuroblastoma	18 months–17 years	Children's Mercy Hospital Kansas City	Completed	1	Non-complete response and death of disease in all three patients. No short-term toxicity.
NCT02107963	GD2 CAR-T	Sarcomas, neuroblastoma, melanoma	1–35 years	National Cancer Institute (NCI), USA	Completed	1	Not available
NCT02932956	GPC3 CAR-T cells	Liver cancer	1–21 years	Baylor College of Medicine	Active, not recruiting	1	Not available
NCT00902044	HER2 CAR	Sarcoma	Children, adults, older adults	Baylor College of Medicine	Active, not recruiting	1	No CR. OS of 10.3 months. No toxicity.
NCT04995003	HER2 CAR-T cells in combination with checkpoint blockade	Sarcoma	1–25 years	Baylor College of Medicine	Recruiting	1	Not available
NCT03721068	iC9.GD2.CAR.IL-15 T-cells	Neuroblastoma, Osteosarcoma	18 months and older	UNC Lineberger Comprehensive Cancer Center, USA	Recruiting	1	Not available
NCT03721068	iC9.GD2.CAR.IL-15 T-cells	Neuroblastoma and osteosarcoma	18 months and older	UNC Lineberger Comprehensive Cancer Center	Recruiting	1	Not available
NCT01953900	iC9-GD2-CAR-VZV-CTLs	Osteosarcoma, neuroblastoma	Children, adults, older adults	Baylor College of Medicine, USA	Active, not recruiting	1	Not available
NCT01822652	iC9-GD2-CD28-OX40 (iC9-GD2) T cells	Neuroblastoma	Children, adults, older adults	Baylor College of Medicine, USA	Active, not recruiting	1	Out of 11 patients: two CR, one grade 1 CRS, two grade 3/4 fever and neutropenia, low grade neurotoxicity.
NCT04715191	IL-15 and IL-21 GPC3 CAR-T cells	Liver cancer, rhabdomyosarcoma, malignant rhabdoid tumours, and others	1–21 years	Baylor College of Medicine	Not yet recruiting	1	Not available
NCT04377932	IL-15 GPC3 CAR-T cells	Liver cancer, rhabdomyosarcoma, malignant rhabdoid tumours, and others	1–21 years	Baylor College of Medicine	Recruiting	1	Not available
NCT04483778	Second-generation 4-1BBζ B7-H3-EGFRt-DHFR	R/R paediatric solid tumours	0–26 years	Seattle Children’s Hospital, USA	Recruiting	1	Not available
NCT03618381	Second-generation 4-1BBζ EGFR806-EGFRt	R/R paediatric solid tumours	1–30 years	Seattle Children’s Hospital, USA	Recruiting	1	Not available
NCT00889954	TGFBeta resistant HER2/EBV-CTLs	HER2 positive malignancies	3 years and older	Baylor College of Medicine	Completed	1	Not available

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

CAR, chimeric antigen receptors; CR, complete responses; CRS, cytokine release syndrome; CTL, cytotoxic T lymphocytes; EGFRt, epidermal growth factor receptor; GPC3, glypican 3; HER2, human epidermal growth factor receptor 2; OS, overall survival; PR, partial responses; SD, stable disease.

2.4.2.1. GD2

Preclinically, 4-1BB GD2 CAR-T exerted their antitumour activity in vitro and in a xenograft model, where tumour growth was suppressed in 5/6 mice [70]. Clinically, GD2-CD28 CAR-T were implemented in a phase I clinical trial for relapsed/refractory neuroblastoma tumours. In total, 3/6 patients receiving >10⁸/m² CAR-T showed antitumour activity at the primary and at metastatic site, but no CR was reported and 2/3 responding patients died within 5 months after the first CAR-T infusion due to PD [71]. In the other six patients receiving lower cell doses, CAR-T failed to efficiently expand and was below the detection threshold. Tumour reactivity was associated with grade 1–3 cytokine release syndrome (CRS) and toxicity linked to systemic immune activation (NCT02761915) [71].

2.4.2.2. B7-H3

Second-generation CAR-T have also been developed against B7-H3, a ubiquitous immune checkpoint molecule highly expressed by many paediatric solid tumours [72], [73], [74]. B7-H3 CAR could promote tumour regression in in vivo models of osteosarcoma, Ewing sarcoma [75], and neuroblastoma [76]. In the first study, MGA271 antibody-based B7-H3 CAR-T mediated complete regression of osteosarcoma and Ewing sarcoma xenografts [75]. MGA271-based CAR-T cells only respond to high-antigen-density tumour cells and not to low antigen densities, limiting their on-target off-tumour toxicity [75]. In neuroblastoma, mice receiving the highest 376.96 antibody-based CAR-T dose remained in remission for more than 90 days [76]. Second-generation CD28 B7-H3 CAR-T coexpressing 41BBL are currently being assessed in a phase I clinical trial for paediatric solid tumours (NCT04897321) [77]. Importantly, this second-generation CAR +/−41BBL was superior in terms of in vitro effector function and in vivo antitumour activity to its third-generation counterpart [77]. This underlines that there is no simple relationship between the number of endodomains and CAR-T effector function and the importance of individually testing different endodomain combinations for each CAR-T to find the optimal design.

2.4.2.3. L1CAM

4-1BB L1CAM CAR-T displayed a potent antitumour activity in vitro and an immunocompromised mouse model, and acceptable safety profile in non-human primates [78]. These cells are now under evaluation, together with CD28:4-1BB third-generation CAR-T, in a phase I clinical trial for neuroblastoma (NCT02311621) [78]. Of note, preclinical data show that CD28 versus 4-1BB costimulation in L1CAM CAR-T results in better expansion and tumour control in mouse models of neuroblastoma [79]. These results are in line with recent studies that show CD28 costimulation as better suited for low antigen densities, which are more common in solid versus haematological malignancies [80].

2.4.2.4. HER2

HER2 is expressed at low levels by sarcoma tumours and correlates with poor clinical outcome [81]. Despite its low expression, second-generation CD28 HER2 CAR-T efficiently eliminated tumour cells in vitro and in an osteosarcoma xenograft model [82], [83]. Moreover, in a metastatic mouse model, 8/10 animals treated with HER2 CAR-T achieved complete regression of metastases [82]. In a phase I clinical trial, among 19 patients with HER2⁺ sarcoma tumours, CD28 HER2 CAR-T persisted at least 6 weeks in 7/9 patients but with no post-infusion expansion. Treatment was associated with a median overall survival (OS) of 10.3 months without toxicities. However, no CR was observed and 13/19 patients underwent PD, suggesting the need of improving CAR-T expansion (NCT00902044) [84].

2.4.2.5. GPC2 and GPC3

GPC2 and GPC3 are oncofoetal proteins that can be upregulated by solid tumours [85]. GPC2 is highly expressed by neuroblastoma tumours [86]. GPC2 CAR-T were switched from 4-1BB [87] to CD28 costimulation (also in combination with c-Jun overexpression) [80] to tune the cells for low antigen density. Survival of high-tumour burden mice treated with CD28 GPC2 CAR-T was increased by ∼20 days compared to 4-1BB GPC2 CAR-treated mice and by ∼30 days compared to control mice [80]. GPC3 is highly expressed in hepatocellular carcinoma. Second-generation 4-1BB GPC3 CAR-T displayed a high proliferative potential, a Th-1 cytokine profile, and a strong antitumour activity in xenograft models of hepatocellular carcinoma and malignant rhabdoid tumours [88]. These cells are under evaluation in a phase I clinical trial with no results posted yet (NCT02932956). Of note, there is evidence that soluble GPC3 could hamper GPC3 CAR-T activity, representing a possible limitation to the success of this cell therapy [89].

2.4.3. Third-generation CAR-T

Although promising, trials with second-generation CAR-T still fail to achieve optimal tumour control. Studies are therefore moving towards third-generation CAR-T that comprise multiple costimulatory domains to improve CAR-T function.

In a preclinical model, neuroblastoma-targeting CD28:OX-40 GD2 CAR-T were superior in terms of proliferation, cytokine release, and effector functions to their first- and second-generation counterparts [90]. They have been evaluated in a clinical trial for neuroblastoma tumours (NCT01822652), in which only 2/11 patients achieved CR, 4/11 patients were alive with disease at the time of results publication, and 5/11 died of disease [91]. Very recent data of a phase I/II clinical trial (NCT03373097) show extremely encouraging results in treating neuroblastoma patients with CD28:4-1BB GD2 CAR-T. In this trial, 9/27 neuroblastoma patients had a CR after one CAR-T infusion, 8/27 had a PR, 5/27 stable disease (SD), and 5/27 no response with a 3-year OS of 60% and event-free survival (EFS) of 36%. Grade 3 CRS occurred only in one patient, and CAR-T persistence was reported up to 30 months after infusion in 26/27 patients [92].

For other targets, third-generation CD28:4-1BB GPC3 CAR-T have also been developed, and they abolished tumour growth in patient-derived hepatocellular carcinoma xenografts [93].

2.4.4. Next-generation CAR-T

With next-generation CAR-T, we refer to CAR-T designs that go beyond the manipulation of costimulatory domains and make use of complex engineering techniques to enhance T cell persistence, functionality, and infiltration, while reducing therapy-associated toxicities. These innovative strategies are also focusing on the fine-tuning of immunomodulatory stimuli present in the TME to reduce signals, which could dampen the CAR-T response, thereby increasing their antitumour potential.

2.4.4.1. Cytokine-transduced next-generation CAR-T

CAR-T can be transduced with cassettes that lead to the transcription of cytokines upon receptor signalling. These CAR-T, also referred to as T cells redirected for antigen-unrestricted cytokine-initiated killing, can be transduced with cytokines sustaining T cell proliferation. GD2 CAR-T coexpressing IL-15 are currently evaluated in a phase I clinical trial for neuroblastoma and osteosarcoma (NCT03721068). IL-15-transduced GD2 CAR-T displayed a stem-cell-like phenotype, a decreased expression of inhibitory receptors, and better persistence in a metastatic neuroblastoma xenograft model, compared to the classical GD2 CAR-T [94]. Moreover, IL-15 has been expressed in GPC3 CAR-T, together with IL-21, which resulted in better proliferation, reduced apoptosis, increased central and stem-cell memory phenotypes, and improved antitumour activity in hepatocellular carcinoma-bearing mice [95]. These cells are currently under evaluation in two phase I clinical trials for paediatric solid tumours (NCT04377932, NCT04715191). GPC3 CAR-T have also been engineered to coexpress IL-7, achieving an increased CAR-T persistence and antitumour activity [96]. Of note, the GPC3 CAR-T used in this study contained an intracellular CD3ε domain, instead of the classical CD3ζ, resulting in lower production of cytokines, which could be associated with a lower possibility of developing CRS, one of the major toxicities in CAR-T-treated patients [96]. In a preclinical neuroblastoma and melanoma model, GD2 CAR-T were engineered to coexpress IL-7 and CCR2b, the receptor for chemokine CCL2, produced by various tumours, exhibiting a stronger antitumour activity and tumour infiltration compared to the second-generation counterpart [97]. IL-18 and IL-12 are also important cytokines for sustained CAR-T persistence. IL-18-engineered GD2 CAR-T showed enhanced activation and production of pro-inflammatory cytokines when co-cultured with GD2-expressing cells [98]. Interestingly, IL-18 also recruited monocytes and NK cells, contributing to the antitumour activity [98]. IL-18 and IL-12-secreting GD2 CAR-T can be further improved by the addition of a modified NFAT promoter in the transgene, which results in locoregional cytokine release upon CAR signalling and a consequential diminished systemic toxicity. NFAT-modified IL-18+IL-15 GD2 CAR-T display increased cytotoxicity, activation, and monocyte recruitment in vitro [99]. Lastly, CAR-T could also be engineered for indirect expression of cytokine receptors. GD2 CAR-T transduced with p40, a subunit of the IL-23 receptor, showed superior antitumour activity and safety compared to IL-18 or IL-15-engineered cells in a neuroblastoma xenograft model [100].

Overall, these results point out the importance of exploring new approaches of CAR-T engineering to support and improve CAR-T functionality beyond the addition of costimulatory domains, for example, by the implementation of additional cargos.

2.4.4.2. Safety of next-generation CAR-T

One critical issue with CAR-T is treatment-associated toxicities, of which the most reported are CRS, neurotoxicity, and on-target off-tumour toxicities [101]. New designs implementing safety switches to turn off transiently or stably CAR-T activity have been developed. GD2 (NCT01822652, NCT02765243, NCT02992210) and B7-H3 (NCT04432649) CAR-T, in clinical trials for neuroblastoma and B7-H3⁺ solid tumours, respectively, contain an inducible caspase 9 (iCas9), which is activated upon administration of the drug AP1903, leading to CAR-T apoptosis [102], [103]. Overall, results from a trial with iCas9 GD2 CAR-T showed very limited toxicities but modest efficacy independent from the iCas9 cassette (tumour regression in 2/34 treated patients, Table 3) [102], [104]. Alternatively, truncated forms of epidermal growth factor receptor (EGFRt) or HER2t can be recognised by the antibodies cetuximab and trastuzumab leading to T cell apoptosis [105], [106]. These are used in ongoing B7-H3 (NCT04483778) and EGFR (NCT03618381) CAR-T trials for children with recurrent/refractory solid tumours. Inducing CAR-T ablation with these strategies also leads to an abrogation of the antitumour response [107]. Thus, new approaches to transiently turn off CAR-T activity have evolved. Preclinical studies on GD2-CAR-T with transient CAR expression, achieved by electroporation instead of viral transduction, demonstrated reduced CAR-T persistence and toxicities [108], [109]. However, this temporal CAR expression results in poorer tumour infiltration and the need for repetitive CAR-T administrations [108], [109]. Another way to improve CAR-T specificity is represented by split-signalling AND logic-gate CAR-T, which was recently applied to target ALK, the second most common mutated gene in neuroblastoma tumours [110], [111], [112]. T cells were engineered to coexpress a first-generation ALK-targeting CAR and a B7-H3 CAR bearing only the costimulatory receptor but lacking CD3ζ. Only when both receptors were engaged, full T cell activation was induced, granting specificity of the design [112]. SynNotch CAR-T can also mitigate on-target-off-tumour toxicity. In these cells, expression of the CAR is dependent on the engagement of a SynNotch receptor, which extracellularly bears a tumour-specific scFv and intracellularly has a site that can be cleaved, releasing a transcription factor that can activate the CAR transcription [113]. The need for simultaneous recognition of two antigens increases CAR-T tumour specificity. In preclinical studies for neuroblastoma tumours, these cells have been generated targeting B7-H3 upon GD2 binding [114]. Mice treated with three doses of these CAR-T showed an increased 40-day survival compared to control or single-infused mice and no sign of neurotoxicity.

2.4.5. Combination strategies

ACTs have been also combined with small-molecule inhibitors to improve killing and circumvent the resistance that arises with single-agent therapies.

GD2 CAR-T have been combined with MEK inhibitors or doxorubicin, displaying a synergistic effect in an in vivo neuroblastoma model and in vitro with osteosarcoma cell lines [115], [116]. Despite the ability of MEK inhibitors to synergise in vivo with CAR-T, MEK inhibition in vitro resulted in suppression of GD2 CAR-T killing [115]. Combining L1CAM CAR-T with lysine demethylase 1A inhibitors boosted antigen-independent FAS/FAS-ligand-mediated killing and improved therapy efficacy in vitro [117]. Subpharmacological doses of sorafenib, an inhibitor of kinases involved in tumour proliferation and angiogenesis, synergised with GPC3 CAR-T in a hepatocellular carcinoma mouse model [118]. The synergistic effect was explained by a combined apoptotic activity on tumour cells and an indirect effect of sorafenib modulating macrophage activity, leading to increased IL-12 secretion and, consequentially, better CAR-T activation [118].

2.4.5.1. Overcoming the immunosuppressive TME

It has become clear that improving CAR-T efficacy should be coupled with strategies that can manage the immunosuppressive TME of paediatric solid tumours. These tumours exploit many tactics to avoid immune cell recognition, including (1) downregulation of MHC molecules, hindering antigen presentation and abrogating TCR-mediated killing; (2) expression of immune checkpoint molecules, including the well-known PD-L1; (3) recruitment of immunosuppressive cell populations, such as MDSCs, an immature myeloid cell population with the ability to suppress immune responses; (4) secretion of immunosuppressive soluble factors; and (5) displaying an immunosuppressive metabolic profile. Both smarter CAR-T designs and combinatorial strategies are needed to overcome these immunosuppressive features [119], [120], [121].

To address the issue of immune checkpoint molecule expression, GD2 CAR-T therapy has been coupled with programmed cell death protein 1 (PD-1) blockade [122]. In a phase I clinical trial for neuroblastoma tumours treated with GD2 CAR-T + anti-PD-1 either without or after lymphodepletion, only 2/11 patients achieved CR. PD-1 inhibition showed no further enhancement of CAR-T expansion and persistence and no clear clinical benefit (NCT01822652) [91].

MDSCs have been shown to impair GD2 CAR-T efficacy in neuroblastoma and sarcoma [123], [124]. In a sarcoma xenograft model, GD2 CAR-T therapy combined with MDSC depletion using all-trans retinoic acid resulted in an increased antitumour effect and better survival [124].

Arginine is an essential amino acid for T cell proliferation, which is depleted in the TME due to the high consumption by tumour cells. To improve their metabolic fitness in an arginine-low environment, GD2 CAR-T were engineered to express arginine resynthesis pathway enzymes that can restore arginine supply, resulting in increased CAR-T proliferation and enhanced tumour control [125].

2.4.6. Other strategies to improve CAR-T therapy

In addition to next-generation CAR-T and combinatorial strategies, other elaborated designs are currently under evaluation to tailor CAR-T therapy for the paediatric TME, while reducing CAR-T-associated toxicities.

An alternative solution to regular CAR-T therapy is represented by CAR cytokine-induced killer (CAR-CIK) cells. CAR-CIK are derived from polyclonal T cell populations that, upon specific culturing, acquire the expression of several NK cell surface markers but retain the expression of their endogenous TCR [126], [127]. These cells are considered as potent as CAR-T but present limited toxicity due to their slow division rate, higher susceptibility to apoptosis, and persistent IFN-γ expression [128]. They have been explored also in allogeneic settings thanks to their limited graft-versus-host effect. CAR-CIK cells targeting chondroitin sulphate proteoglycan 4 (CSPG4) [129] and erythroblastic oncogene B 2 (ERBB2) [130] have been generated for the treatment of soft tissue sarcomas and RMS, respectively. CSPG4 CAR-CIK-treated mice with low tumour burden experienced a delay in tumour growth up to 11 days compared to control mice [129], while ERBB2 CAR-CIK treatment could increase survival of ∼20 days compared to untreated mice, but with no significant difference compared to WT CIK-treated mice [130].

To broaden the spectrum of antigens for CAR-T cells beyond surface molecules, so-called peptide-centric or TCR-mimicking CAR-T have been generated that can recognise antigens presented on MHC molecules, allowing CAR-T cells to recognise also intracellular oncoproteins. Peptide-centric CAR-T have been generated targeting PHOX2B, a transcription factor expressed at high levels in neuroblastoma [131], achieving complete regression of the tumour in xenograft models [132]. These innovative cells represent a powerful means to boost the efficacy of ACTs for paediatric solid tumours, but they still rely on MHC expression.

Overall, CAR-T therapy is the most studied ACT strategy for paediatric solid tumours because of their MHC independence. However, current αβ T cell-based ACT products have the common disadvantage of not being applicable as off-the-shelf products due to the risk of graft-versus-host disease (GVHD) and can therefore only be generated from autologous sources. To reduce the GVHD risk, genome-edited αβ T cell-based products are currently under development: CAR insertion and simultaneous ablation of T cell receptor alpha constant and β-2 microglobulin may allow to establish ‘off-the-shelf’ CAR-T cells [133].

3. NK cells

3.1. NK-based immunotherapy

NK cells exert a pivotal role in immunosurveillance by killing of malignant cells and by recruitment of other immune cells [134]. In oncological patients, NK cells are often impaired, mainly due to the immunosuppressive TME, and their tumour infiltration is limited [135], [136], [137], [138]. For example, osteosarcoma patients have low circulating NK cell numbers, but functionality and IFN-y production are retained [139].

NK function and cytotoxicity rely on a fine-tuned balance of inhibitory and stimulatory receptor engagement [140]. NK cells can directly kill target cells by antibody-dependent cell cytotoxicity (ADCC), which is mediated by immunoglobulins binding surface antigens on tumours and activating receptors CD16A (FcγRIIIA) or CD32c (FcγRIIC) on NK cells, leading to NK cell degranulation, cytokine secretion, and tumour cell lysis [141]. Another important mechanism of NK cell-target recognition relies on the ‘missing self-response’ [142], where the absence of MHC-I triggers NK cell activation and killing due to reduced binding of MHC-I molecules to the inhibitory killer immunoglobulin-like receptors (KIR). Paediatric tumours express low levels of MHC-I [18] and are therefore ideal targets for NK cell therapy. In addition, the use of haploidentical NK cells benefits of their alloreactivity induced by the mismatch between KIR receptors on donor cells and their ligands on recipient cells, resulting in a significant NK-versus-tumour effect [143]. Their ability to selectively recognise cancerous cells together with the lack of GVHD makes NK cells a promising ACT candidate, even as an allogeneic product [144], [145], [146].

Two ACT strategies exploiting NK cells have been employed in paediatric solid tumours: 1) haploidentical haematopoietic stem-cell transplantation (HSCT) to permanently establish the donor NK cells, and 2) transient expansion and adoptive transfer of NK cells from a haploidentical donor.

3.2. NK-enriched HSCT

Haploidentical T and B cell-depleted transplants offer a unique opportunity for mismatching KIRs on donors’ NK cells against HLA molecules present on the tumour, enhancing the graft versus tumour (GVT) effect [146], [147]. Of note, the risk of GVHD must be taken in consideration and balanced during haploidentical HSCT, as the transplant might be non-fully depleted in T cells. While there are numerous studies observing the benefits of KIR mismatch in leukaemias [144], [148], clinical and preclinical studies in solid tumours have lagged behind. Nevertheless, some studies showed increased susceptibility to allogenic NK-mediated osteosarcoma killing when the KIR-HLA-I incompatibility was maximised [149]. The feasibility of NK-enriched haploidentical HSCT to increase GVT was tested in an early clinical trial for relapsed solid cancers in children and young adults (NCT01804634). The procedure was effective and safe, with an OS rate of 88% at 6 months, 56% at 12 months, and 21% at 2 years. This study also underlined the need to optimise immune reconstitution after transplant to decrease morbidity and mortality in future trials [150]. A phase II trial (NCT02100891) is currently ongoing on solid tumours, including Ewing sarcoma, neuroblastoma, RMS, and osteosarcoma, consisting of a low-intensity chemotherapy and radiation therapy, followed by haploidentical HSCT and donor-derived NK infusions (Table 4).

Table 4.

Summary of clinical trials using NK cells for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT02508038	Haploidentical SCT + zoledronate	Neuroblastoma, osteosarcoma, Ewing sarcoma, rhabdomyosarcoma (RMS), leukaemia, lymphoma	7 months–21 years	University of Wisconsin, Madison	Recruiting	1	Not available
NCT03420963	Cord blood-derived expanded allogeneic NK cells + cyclophosphamide, etoposide	Refractory neoplasms	12 months–40 years	M.D. Anderson Cancer Center	Recruiting	1	Not available
NCT04214730	Donor NK cell infusion and chemotherapy	Solid tumours	10–90 years	Yantai Yuhuangding Hospital	Recruiting	NA	Not available
NCT00640796	Haploidentical NK cells + IL-2+ cyclophosphamide, fludarabine, mesna	Relapsed/refractory haematological malignancies, Ewing sarcoma family of tumours, and RMS	Children, adults	St. Jude Children’s Research Hospital	Completed	1	Not available
NCT00698009	Haploidentical NK cells + Il2+ fludarabine, cyclophosphamide, mesna	Neuroblastoma	Children, adults, older adults	M.D. Anderson Cancer Center	Not active, terminated (slow accrual)	2	Not available
NCT02409576	Haploidentical NK cells infusions + IL2	Ewing sarcoma, RMS	Children, adults, older adults	National University Hospital, Singapore	Unknown	1, 2	Not available
NCT02100891	Haploidentical SCT + donor NK cells infusions	Ewing sarcoma, neuroblastoma, RMS, osteosarcoma, CNS tumours	Children, adults, older adults	Medical College of Wisconsin	Active, not recruiting	2	6-months DCR is 72%, 1 and 2 year OS are 64% and 40%, respectively, and PFS is 29% and 22%. 4/15 patients developed acute or chronic GVHD.
NCT00582816	Haploidentical SCT and donor NK cells infusions	Leukaemia, solid tumours	6 months–25 years	University of Wisconsin, Madison	Terminated (toxicity)	1,2	Not available
NCT01875601	NK cell infusion + rhIL15	Sarcoma, neuroblastoma, others	2–29 years	National Cancer Institute	Completed	1	Not available
NCT00877110	Haploidentical NK cells + anti-GD2 m3F8	Neuroblastoma bone marrow, sympathetic nervous system	Children, adults, older adults	Memorial Sloan Kettering Cancer Center	Completed	1	10 CR/PR (29%), 17 NR (47%), 8 (23%) progressive disease. No GVHD.
NCT01857934	Autologous SCT and haploidentical NK cells + IL2+ anti-GD2 hu14.18K322A	Neuroblastoma	Up to 18 years	St. Jude Children’s Research Hospital	Active, not recruiting	2	60 out of 64 PRs or better, including 40 patients with mCRs. Only 31 patients received additional NK cell infusions.
NCT02130869	Autologous SCT haploidentical NK cells + IL2+ G-CSF + anti-GD2 hu14.18K322A + melphalan + busulfan	Relapsed/refractory neuroblastoma, lymphoma, high-risk solid tumour	Up to 21 years	St. Jude Children’s Research Hospital	Completed	1	Not available
NCT02650648	Haploidentical NK cells + anti-GD2 hu3F8 (naxitamab) + IL-2	High-risk neuroblastoma	Children, adults, older adults	Memorial Sloan Kettering Cancer Center	Active, not recruiting	1	Not available
NCT01576692	Haploidentical NK cells + anti-GD2 hu14.18K322A	Neuroblastoma	Up to 21 years	St. Jude Children’s Research Hospital	Completed	1	Response rate 61.5% (four CR, one very good PR, three PR), five SD. One patient with unacceptable toxicity and four patients discontinued treatment because of adverse events.
NCT02573896	Autologous NK cells + anti-GD2 ch14.18/CHO (dinutuximab) + lenalidomide	Neuroblastoma	Up to 30 years	New Approaches to Neuroblastoma Therapy Consortium	Active, not recruiting	1	Not available
NCT03209869	Haploidentical NK infusions + immunecytokine hu14.18-IL2 + GM-CSF	Neuroblastoma, osteosarcoma	7 months–21 years	University of Wisconsin, Madison	Not active, withdrawn (limited resources due to COVID-19)	1	Not available
NCT03242603	Haploidentical NK cells + anti-GD2 ch14.18/CHO (dinutuximab)	Recurrent neuroblastoma	6 months–25 years	National University Hospital, Singapore	Unknown	1, 2	Not available
NCT04211675	Allogenic TGFβi NK cell + anti-GD2 ch14.18/CHO (dinutuximab) + irinotecan, temozolomide, sargramostim	Neuroblastoma	Up to 29 years	Nationwide Children’s Hospital	Recruiting	1, 2	Not available

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

CNS, central nervous system; CR, complete responses; GVHD, graft versus host disease; mCR, metastatic complete response; NK, natural killer; NR, non-responder; OS, overall survival; PFS, progression-free survival; PR, partial responses; SCT, stem-cell transplantation; SD, stable disease.

3.3. Adoptive NK-cell therapy

Adoptive allogenic NK infusions are a valid option for replacing the patient’s NK cells, often poorly functional, while also providing a strong GVT effect with the KIR-HLA mismatch.

It has proven to be safe and well tolerated mainly due to a lack of GVHD and treatment-related toxicity [145]. The efficacy of allogenic NK infusions was first demonstrated by successful transfer and expansion of haploidentical NK cells in AML patients, where 5/19 patients with poor prognosis achieved CR [145]. Further feasibility proofs were confirmed in other studies on solid childhood cancers [151], [152].

3.3.1. NK cell transfer coupled with mAb

NK cells are major players involved in the efficacy of mAb-based therapies thanks to their ability to induce ADCC [141]. Among paediatric solid tumours, anti-GD2 antibody therapy with dinutuximab (human-mouse chimeric 14.18) and naxitamab (humanised 3F8) has shown improved EFS, OS, and substantial antineoplastic activity against neuroblastoma [153], [154], [155]. To boost the antitumour effect, mAb has been combined with the adoptive transfer of ex vivo-expanded NK cells. Expanded activated NK cells in combination with dinutuximab increased cytotoxicity against neuroblastoma in vitro and in a mouse model, compared with mAb or NK cells adoptive monotherapy [156].

In a phase I trial, 13 patients with recurrent/refractory neuroblastoma were treated with haploidentical NK cells, chemotherapy, GM-CSF, IL-2, and a humanised version of dinutuximab (hu14.18K322A) [157]. Although all patients were heavily pretreated, the response rate was 61.5% (four CR, one very good PR, and three PR), five patients had a SD, and OS was 77% at 1 year [158]. In a phase I trial with haploidentical NK-cells in combination with murine 3F8 in high-risk neuroblastoma patients [159], 10/35 patients (29%) reached CR or PR, OS was 30.7 months, and no GVHD was detected. Since most patients received other treatments after the trial, survival could not be solely attributed to the trial treatment. Moreover, NK cells were undetectable 14 days post-infusion, which leaves it unclear if the effect was NK-mediated. A follow-up phase I trial involving 3F8 with superior ADCC activity (NCT02650648) is currently active.

These findings show that NK cells can be safely combined with anti-GD2 mAb, showing antineuroblastoma effects.

3.3.2. Adoptive transfer of genetically engineered NK cells

Genetic manipulation of NK cells could improve antitumour effects, homing, and in vivo persistence. NK cells migratory capacity can be enhanced by increasing the expression of the chemokine receptor CXCR2 binding to CXCL3, CXCL5, and CXCL8, which are expressed by a variety of solid tumours [160]. Viral transduction of CXCR2 on primary NK cells augmented tumour migration and killing [161]. The expression of IL-15, an important activating cytokine for NK cells, could also be incorporated to enhance the in vivo expansion and persistence of NK cells [162].

NK cells can also be modified to express CARs. The use of NK cells as an engineering platform can yield some advantages over T cells, as they can be produced from an allogenic source, providing the option of an ‘off-the-shelf’ therapy [163]. Moreover, CAR-T-related toxicities can be overcome thanks to the short lifespan of CAR-NK cells [164]. Although there are still few studies in paediatric oncology, the use of CAR-NK cells targeting CD19 has been proven safe and effective in trials in adult patients with CLL [165], and two pilot studies (NCT00995137 and NCT01974479) have been approved for the treatment of ALL.

As previously stated, one of the main challenges for immunotherapy in solid tumours is the highly immunosuppressive microenvironment. NK cells engineered to express activating receptor NKG2D fused with the cytotoxic ζ chain of T cells (NKG2D.ζ-NK cells) showed improved cytotoxicity in vitro and were able to alter the TME in favour of an antitumour response, eliminating suppressive MDSCs in a xenograft TME model [166]. NKG2D.ζ-NK cells could even secrete pro-inflammatory cytokines and chemokines to recruit and activate co-delivered CAR-T that would be otherwise suppressed [167].

In conclusion, while NK cells offer a wide range of advantages and their safety and efficacy in cancer treatment have been proven, additional work is required to enable robust tumour infiltration and improve patient survival.

4. iNKT cells

4.1. iNKT cells in paediatric antitumour immunity

iNKT cells are T cells that lay in-between innate and adaptive immunity, expressing both, classical T and NK cell markers [168], [169], [170]. They owe their name to their semi-invariant TCR [171], [172] recognising lipids presented in the context of the CD1d molecule, a non-polymorphic MHC molecule [173], [174]. Because of their MHC independency, these cells represent a valid alternative to classical αβ T cell therapy for paediatric tumours [18].

High levels of tumour-infiltrating iNKT cells have been associated with good prognosis in a variety of solid tumours, including neuroblastoma [175], [176], where iNKT cell infiltration seems to be negatively regulated by the oncogene MYCN, overexpressed in aggressive tumour forms [177], [178]. Hepatoblastoma, RMS, and neuroblastoma patients display a frequency of circulating iNKT cells similar to those of healthy controls, and the cells retain their ability to expand in vitro upon stimulation with synthetic α-galactosylceramide, the prototypic iNKT cell antigen [179], suggesting their strong potential as ACT against paediatric extracranial solid tumours. However, iNKT cells constitute only around 0.01–0.1% of peripheral blood mononuclear cells, limiting their use as ACT products. Among paediatric extracranial solid tumours, iNKT cells have been so far only used for the treatment of neuroblastoma, leaving a gap in their implementation for other malignancies.

4.1.1. CAR-engineered iNKT cells

For neuroblastoma, GD2 CAR-iNKT cells demonstrated an efficient cytotoxicity in an in vitro model [180]. Upon target recognition, iNKT cells have the peculiar feature of producing both Th1 and Th2 cytokines. Comparing 4-1BB versus CD28 costimulation, the 4-1BB endodomain favoured a Th1 polarisation, leading to increased production of IFN-γ and GM-CSF and reduction of IL-4 and IL-10 [180]. Third-generation 4-1BB:CD28 GD2 CAR-iNKT cells displayed an enhanced persistence in a neuroblastoma xenograft model, where they could control tumour growth without inducing GVHD [180].

iNKT cells have also been coengineered with GD2-CAR + IL-15. IL-15 is (1) a key cytokine for iNKT cell development and homoeostasis [181], [182], (2) protective towards iNKT cell inhibition mediated by TAM [183], and (3) low IL-15 or iNKT cell presence within the neuroblastoma TME is associated with poor patient outcome [184]. CD28 GD2/IL15 CAR iNKT cells displayed enhanced persistence, antitumour activity, and tumour localisation in neuroblastoma mouse xenografts, as well as a reduced exhaustion profile in vitro [185]. These cells are currently evaluated in a phase I trial for relapsed/refractory neuroblastoma tumours (NCT03294954), and preliminary data showed a good safety profile but only 2/12 PR and 1/12 CR, which correlated with high CAR-iNKT cell expansion and regression of the bone metastatic lesions (Table 5) [186], [187].

Table 5.

Summary of clinical trials using iNKT cells for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT03294954	GD2 IL15 CAR-NKT	Neuroblastoma	1–21 years	Baylor College of Medicine	Active, not recruiting	1	No dose-limiting toxicities observed. 1/12 grade 2 CRS. Regression of metastatic lesions and CR in 1/12 patients, PR in 2/12 patients.

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

CAR, chimeric antigen receptors; CR, complete responses; CRS, cytokine release syndrome; NKT, invariant NKT; PR, partial responses.

4.1.2. ACTs with expanded iNKT cells

iNKT cells can also be isolated from the peripheral blood, expanded ex vivo, and reinfused in patients [188]. iNKT cells also provide maturation stimuli to dendritic cells and monocytes [189], [190], leading to IL-12 secretion and NK cell activation [191]. In an in vitro neuroblastoma model, iNKT cell infusion was combined with anti-GD2 antibodies, enhancing NK cell ADCC activity towards GD2⁺ tumour cells, providing optimal tumour killing [192].

In conclusion, the therapeutic potential of iNKT cells for paediatric solid tumours remains widely unexplored, offering several possibilities for the development of new allogeneic ACT strategies to improve patient survival, while decreasing GVHD-related toxicities.

5. γδ T cells

γδ T cells are T lymphocytes that contain a TCRγ and TCRδ chain and constitute around 1–5% of total T cells in peripheral blood. Combinations of the different TCRγ and δ chains form different γδ T cell types [193]. Vδ1-containing T cells are abundant in healthy epithelia and in tumour infiltrates [194], whereas the Vγ9Vδ2 subtype is rich in the peripheral blood [195]. Since mouse γδ TCRs are not homologous to human, preclinical studies are restrained to xenografts in immunodeficient mice or non-human primate models [196].

γδ T cells combine innate-like rapid responses and clonal expansion with pleiotropic effector functions. They can be activated in an HLA-unrestricted fashion through their TCR by non-peptide antigens, such as bisphosphonate isopentenyl-5-pyrophosphate (IPP), that could result from malignant transformation [197]. The synthetic phosphoantigen zoledronic acid was shown to selectively activate TCRVγ9Vδ2⁺ lymphocytes in clinical trials, due to the accumulation of intracellular IPP [198]. When activated, they induce apoptosis by releasing granzymes and perforins [199]. Soon after activation, γδ T cells acquire the ability to process and present antigens on MHC molecules to conventional CD4⁺ and CD8⁺ T cells, helping to initiate peptide-specific αβT cell responses [200].

5.1. γδ T cells in cancer immunotherapy

γδ T cell-based immunotherapy harbours two strategies: in vivo stimulation of γδ T cells through systemic administration of aminobisphosphonates and IL-2, or ex vivo expansion of γδ T cells for adoptive transfer. In vivo expansion of γδ T cells offers a relatively inexpensive and straightforward delivery option, while the ex vivo expansion allows control and optimisation of the effector cell population. Adoptive transfer of Vγ9⁺Vδ2⁺ T cells has been tested clinically in various solid tumours [201], and several groups have reported methods to expand also Vδ1⁺ T cells [202], [203], [204], [205].

In neuroblastoma, Vδ1⁺ as well as Vδ1⁻Vδ2⁻ γδ T cell subsets show consistent higher innate tumour toxicity compared to Vδ2⁺ [206], partly due to the production of soluble NKG2D ligands sMICA, sMICB, and ULPB1-6 by neuroblastoma cells, which can block the NKG2D receptor on γδ T cells [207], [208].

The cytotoxicity of activated and expanded Vγ9Vδ2⁺ γδ T cells was enhanced by combination with mAbs in vitro [209], [210], [211] and in vivo [212]. A phase I trial combining Dinutuximab and ex vivo-expanded allogeneic γδ T cells in relapsed/refractory neuroblastoma (NCT05400603) has recently started (Table 6).

Table 6.

Summary of clinical trials using γδ T cells for paediatric extracranial solid tumours (updated June 2023).

NCT	Compound	Indication	Age group	Sponsor	Status	Phase	Results
NCT05400603	Expanded γδ T cells + anti-GD2 ch14.18/CHO + temozolomide, irinotecan, and zoledronate	Neuroblastoma	12 months–16 years	Emory University	Recruiting	1	Not available

See the ‘Clinical trial search’ section in supplementary methods for more information regarding search terms.

5.2. Combining αβ and γδ TCR engineering

Novel strategies, both by transferring αβTCRs into γδ T cells and by transferring γδ TCRs into αβT, have been explored. The first offers the advantage of redirecting γδ T cell immunity against specific antigen targets and has been tested in vitro against viral [213] and melanoma tumour targets [214]. However, αβTCR gene transfer carries inherent limitations, such as the restriction to particular HLA types [215]. One way to overcome MHC restriction is the transfer of γδ TCRs into αβ T cells, resulting in T cells engineered with defined γδ TCRs (TEGs) [216]. TEG002 consists of αβT cells engineered to express the tumour-specific Vγ9Vδ2 TCR and showed higher cytotoxicity against neuroblastoma organoids than untransduced αβ T cells, while also surviving longer than endogenous γδ T cells [217]. TEG002 was superior to T cells expressing either an αβ-TCR or a γδ-TCR alone, and effectively recognised and killed 50% of the tested organoids [218]. A phase I trial (NCT04688853) investigating the safety of TEG002 in multiple myeloma patients is ongoing and will possibly lead to further clinical investigation in paediatric solid tumours.

5.3. γδ T cells engineering

Γδ T cells have also been engineered to express a CAR to acquire additional antigen-specific cytotoxicity and proliferative response, while retaining their innate anti-cancer properties. First-generation GD2 CAR (14. G2aζ) in Vγ9⁺Vδ2⁺ T cells showed increased antigen-specific tumour reactivity, which was reflected in IFN-γ production and CD69 expression upon co-culture with the GD2⁺ neuroblastoma LAN-1 cells line [219].

γδT-cell responses can be tuned by modulating the level of stimulation. Vδ2⁺ γδ T cells have been engineered to coexpress their endogenous Vγ9Vδ2 TCR and a chimeric costimulatory receptor composed of an extracellular GD2-binding scFv and an intracellular DAP10 costimulatory domain [220]. The intracellular DAP10 domain is derived from NKG2D, an important activating receptor for γδ T cells, which paediatric tumours tend to evade [207], [208]. This provides additional costimulation to the γδ TCR, restoring Vγ9Vδ2⁺ T cell cytotoxicity and cytokine production in the presence of transformed cells [220].

6. Conclusions and future perspectives

In the past years, much effort has gone into the improvement of ACTs for paediatric extracranial solid malignancies. Many of the ongoing clinical trials are providing new insights on how to ameliorate ACTs for better efficacy and lower toxicities, with some of these trials achieving incredibly positive results [92]. ACT strategies currently exploit expanded and/or engineered conventional αβ T cells, NK, γδ T, and iNKT cells. αβ CAR-T are so far the most explored strategy because of their MHC independence, which makes them appealing for low MHC-expressing paediatric extracranial solid tumours. However, they are limited to the recognition of surface antigens. Strategies, such as CTL/TILs and TCR-αβ T cells recognising intracellular antigens, are also being explored and interest is increasing in off-the-shelf products using innate and innate-like cells that could be engineered from allogeneic sources without the risk of GVHD.

It has become clear that paediatric solid tumours are complex entities that cannot be eradicated using single-agent therapies. The pronounced immunosuppressive microenvironment and low-MHC expression are exemplary obstacles to ACT [44], [221], [222]. In this review, we have discussed novel approaches using complex engineering strategies or combination therapies that could outsmart the tumours.

Future studies will have to address the current gap between bench and bedside to aid clinical implementation of newly developed ACTs for paediatric solid malignancies. While the field is expanding for some tumour types, such as neuroblastoma and sarcomas, more effort is needed for the clinical translation to rare paediatric kidney and liver tumours. As many of these cancers share common antigens, this could serve as a means for a broader clinical implementation (Fig. 2). Currently, a big disparity exists between the great preclinical effort of identifying new targets and the subsequent implementation of novel therapies into clinical practice (Fig. 2). Better preclinical models and an agreement on efficacy thresholds for clinical translation will be crucial to make new therapies accessible for paediatric patients. It will become essential to identify optimal patient selection criteria and monitor treatment responses during therapy using sequential tumour biopsies to enable for personalised therapies.

Fig. 2.

Overview of adoptive cell therapy (ACT) targets currently exploited in preclinical (left) and clinical (right) practice for paediatric extracranial solid tumours. Some tumours, such as neuroblastoma and sarcomas, are over-represented in the scenario of ACTs for paediatric solid malignancies. Of note, some targets, such as the CTAs NY-ESO, PRAME, and MAGE, are shared among several of these malignancies, making them ideal for more translational approaches. In bold are represented targets that are currently being investigated both at a preclinical and clinical level. GPC2, glypican 2; fACHr, fetal acetylcholine receptor; ERBB2, erythroblastic oncogene B 2; CSPG, chondroitin sulphate proteoglycan 4; STEAP1, six-transmembrane epithelial antigen of prostate-1; EGFR, epidermal growth factor receptor.

In conclusion, while the field of ACT in paediatric extracranial solid tumours is rapidly evolving, there is a need to develop more advanced strategies tailored to solid tumour environments to improve therapy efficacy and work towards curative interventions.

Funding

This work has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement, No. 956285 (VAGABOND). This work is part of the research programme Vernieuwingsimpuls Vidi (Combining targeted compounds in neuroblastoma tumours; is two better than one?) with project number 91716482 and Veni (Release the beast: Boosting CAR-T cell immunotherapy for neuroblastoma) with project number 09150162010022, which is partly financed by the Dutch Research Council (NWO) and ZonMW. In addition, this project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 716079 Predict, from grant H2020-iPC-826121, and from Villa Joep (Joining forces to activate T cell immunity against High-Risk Neuroblastoma).

CRediT authorship contribution statement

Elisa Zappa: Study conceptualization and design, Data curation and quality control, Manuscript writing and editing. Alice Vitali: Study conceptualization and design, Data curation and quality control, Manuscript writing and editing. Kathleen Anders: Data quality control, Manuscript editing and review. Jan J. Molenaar: Data quality control and manuscript review. Judith Wienke: Study conceptualization, Data quality control, Manuscript editing and review. Annette Künkele: Study conceptualization, Data quality control, Manuscript editing and review.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material

.