A decision process for drug discovery in retinoblastoma

Abstract

Intraocular retinoblastoma treatment has changed radically over the last decade, leading to a notable improvement in ocular survival. However, eyes that relapse remain difficult to treat, as few alternative active drugs are available. More challenging is the scenario of central nervous system (CNS) metastasis, in which almost no advancements have been made. Both clinical scenarios represent an urgent need for new drugs. Using an integrated multidisciplinary approach, we developed a decision process for prioritizing drug selection for local (intravitreal [IVi], intrathecal/intraventricular [IT/IVt]), systemic, or intra-arterial chemotherapy (IAC) treatment by means of high-throughput pharmacological screening of primary cells from two patients with intraocular tumor and CNS metastasis and a thorough database search to identify clinical and biopharmaceutical data. This process identified 169 compounds to be cytotoxic; only 8 are FDA-approved, lack serious toxicities and available for IVi administration. Four of these agents could also be delivered by IT/IVt. Twelve FDA-approved drugs were identified for systemic delivery as they are able to cross the blood-brain barrier and lack serious adverse events; four drugs are of oral usage and six compounds that lack vesicant or neurotoxicity could be delivered by IAC. We also identified promising compounds in preliminary phases of drug development including inhibitors of survivin, antiapoptotic Bcl-2 family proteins, methyltransferase, and kinesin proteins. This systematic approach may be applied more broadly to prioritize drugs to be repurposed or to identify novel hits for use in retinoblastoma treatment.

Introduction

Retinoblastoma is the most common primary neoplasm of the eye in childhood[1]. The current use of novel routes to more selectively deliver chemotherapy to the eye, such as intraarterial chemotherapy (IAC) or intravitreal administration (IVi), has resulted in major improvements in eye preservation while reducing systemic drug exposure[1–4]. However, for eyes that relapse after conventional therapy or have refractory tumors, treatment options are few, and many eyes must be enucleated. Even less progress has been witnessed in the development of treatments for metastatic retinoblastoma, a stage of the disease that accounts for up to 50% of cases in low- and middle-income countries[1, 5–7]. Up to 60% these patients can be cured if treated with high-dose chemotherapy and autologous hematopoietic stem-cell rescue but still face a poor prognosis with an invariably fatal outcome in cases of central nervous system (CNS) involvement[8, 9]. Thus, there is a clear need to develop novel treatment strategies using more effective and less toxic chemotherapy.

The most commonly used drugs for retinoblastoma are carboplatin, etoposide, and vincristine, delivered via intravenous infusion¹; melphalan, topotecan, and carboplatin intraarterially[3, 4, 10] (IAC), and melphalan or topotecan intravitreally (IVi)[1, 2]. The first three have been clinically implemented based on results in other pediatric malignancies and their pharmacological profile was only studied after their clinical use. Topotecan and melphalan activity was assessed in preclinical models before their clinical implementation but a complete pharmacokinetic profile for local administration was available only after clinical use [11–13]. Comprehensive pharmacological characterization involves the use of patient-derived cell lines to mimic the biological characteristics of intraocular and metastatic disease[14]. Such approaches are often used to identify new drugs for retinoblastoma, especially in the context of high-throughput screening (HTS) for active compounds among libraries of thousands[15–17]. Nonetheless, antitumor activity should not be the only factor to prioritize a candidate drug for further development; factors such as penetration across the ocular and brain barriers, toxicity profile, and the availability of adequate formulations are important when considering clinical utility.

We hereby propose a model for prioritizing potential drug candidates that most deserve further characterization, adding a selection process to assign the most suitable compounds according to intent of treatment, biopharmaceutical properties, and previous knowledge of their pharmacology.

Materials and Methods

Patient-derived tumor cells

Two patient-derived primary cell cultures, one established from an intraocular tumor of an upfront enucleated eye (HPG-RBT-12L) and the other from cerebrospinal fluid (CSF) tumor relapse after initial chemotherapy treatment (HPG-CSF-1), were used in the present study. Cell line origin, authentication, and culture conditions were previously described[18–20]. The commercial cell line Y79 (ATCC HTB-18, Manassas, VA) was used as a control for HTS and cultured as well described elsewhere[14, 21].

Patient tumor samples were obtained after protocol and informed consent approval by the Institutional Review Board at Hospital de Pediatria JP Garrahan (protocol #838) and written informed consent was obtained from parents or guardians. All procedures involving human subjects were performed in accordance with the Declaration of Helsinki.

High-throughput screening

Optimal HTS assay conditions were established based on previous work published elsewhere[15, 22–24] and adapted to our system conditions described in Online Resource 1. Cells were exposed to each of the 2,700 medicinally active and structurally diverse compounds contained in the Selleck Bioactive and Selleck FDA-approved libraries (SelleckChem, Houston, TX). Cells were automatically counted, seeded in 384-well plates, and cultured for 24 h. Next, compounds were diluted in 0.5% DMSO (v/v) and transferred into the well plates using an Agilent Bravo Liquid Handler, resulting in a final concentration of 4.8 μM. All plates also included positive (melphalan) and negative (vehicle) controls; melphalan was used based on its potent and reproducible in vitro cytotoxicity[21, 25]. After 72 h of drug exposure, cell viability was determined[15, 24]. Compounds that showed > 70% cytotoxic activity at 4.8 μM were considered pharmacologically active against retinoblastoma.

Decision process

A classification process was built for FDA-approved drugs as potential agents to be repurposed or new hits to be prioritized for various treatment scenarios of retinoblastoma. The decision process was based on a series of questions with binary answers (yes = 1, no = 0) identified by consensus from a multidisciplinary collaborative team of clinical experts.

The starting point was defined as pharmacologically active compounds against HPG-RBT-12L and HPG-CSF-1 cells. From the group of active drugs, we filtered for those that were FDA-approved for clinical use or were tested in clinical trials in adults and have been or are being evaluated in Phase I/II pediatric trials. The next question was intended to describe the clinical scenario for their potential use and thus, compounds were divided according to administration routeusing the following criteria:

Local treatment

• Intravitreal (for the treatment of intraocular tumors): injectable solution, devoid of vesicant or irritant effects, and devoid of neurologic or ocular toxicity.

• Intrathecal or intraventricular (for the treatment of CSF dissemination): same as for IVi injection, with evidence of clinical use of IT/IVt delivery[26–30].

Intraarterial or intravenous administration

• Intravenous (for chemoreduction of intraocular tumors, systemic treatment of extraocular tumors): injectable solution, able to penetrate the blood-brain barrier (BBB) or blood-ocular barrier (BOB), and not associated with serious adverse events or serious neurologic or ocular toxicity.

• Oral: same purpose and characteristics as for i.v. administration and pharmaceutically formulated for oral administration.

• Intraarterial chemotherapy (local treatment delivered into blood vessels): same criteria as for i.v. and does not require in vivo activation or prodrugs.

In all cases, injectable solution referred to the availability of an aqueous and sterile solution as the drug must be dissolved in a biocompatible solvent for IVi, IAC, or IT/ IVt administration. For i.v. or IAC treatment, candidate drugs must penetrate the BOB to target the tumor localized in the retina and/or vitreous humor. Thus, we searched for reports that the drugs could penetrate the BOB, and if not available, data on penetration across the BBB was considered a surrogate based on the barriers’ anatomical similarities[29, 31–33]. We also searched for published reports on pharmacokinetics in preclinical models that favored drug distribution in the brain or clinical endpoints supporting CNS activity and therefore, justifying drug penetration across the BBB. Serious adverse events were defined as those that may result in death, a life-threatening adverse event, or hospitalization or prolongation of hospitalization except for hematological toxicities or other clinically manageable toxicities (grade III/IV adverse events by the Common Terminology Criteria for Adverse Events (CTCAE) v5)[34].

For each compound, we searched for answers to these detailed questions in several databases as described in Online Resource 1[35–37]. To avoid missing compounds of interest, we built a parallel decision process for drug discovery to identify new hits that have been excluded because of lack of clinical use or adult or pediatric clinical studies but that still show cytotoxic activity against retinoblastoma cells and have an adequate safety profile according to clinical or preclinical studies.

Statistical analysis for sensitivity and precision assessment could not be performed because selected features consisted of questions relevant to but also mandatory for the clinical implementation of a novel treatment. Thus, the resulting features intrinsically provided the best separation of the dataset or largest information gain.

Results

In total, 169 compounds were highly active against both HPG-RBT-12L and HPG-CSF-1 cells as indicated by > 70% cytotoxic activity (Table 1). Among these, we selected 113 drugs evaluated in clinical trials in adults, of which 59 have been investigated in Phase I/II studies in pediatric patients or clinically used in children. For rapid translation into the clinic, we selected 30 FDA-approved drugs as potential candidates to be repurposed, which were then classified into two categories according to likely administration route, one for local conservative treatment (IVi) or CSF dissemination (IT/IVt) and the other including IAC, oral, or i.v. chemotherapy for conservative therapy or adjuvant or neoadjuvant therapy for extraocular disease.

Table 1. Compounds active in patient-derived retinoblastoma cells.

Compounds found to be active (> 70% cell growth inhibition) in HPG-RBT-12L (derived from intraocular tumor) and HPG-CSF1 cells (cells disseminated into the cerebrospinal fluid of a relapsed heavily pretreated patient). Name, supplier catalog number, and mechanism of action are detailed.

Compound (n = 169)	Selleck code	Mechanism of action
(−)-Parthenolide	S2341	NF-κB /HDAC dual inhibitor
(+)-JQ1	S7110	BET inhibitor
Hydroxycamptothecin	S2423	Topoisomerase I inhibitor
5-Iodotubercidin	S8314	Adenosine kinase inhibitor
Abexinostat	S1090	HDAC inhibitor
ABT-737	S1002	BH3 mimetic inhibitor
Alvespimycin	S1142	HSP90 inhibitor
AR-42	S2244	HDAC inhibitor
ARQ 621	S7355	KSP inhibitor
AT13387	S1163	HSP90 inhibitor
AT9283	S1134	Aurora A,B / JAK2/3 dual inhibitor
AUY922	S1069	HSP90 inhibitor
AZ5104	S7298	EGFR inhibitor
AZ6102	S7767	TNKS inhibitor
AZD1480	S2162	JAK1/2 inhibitor
AZD8055	S1555	mTOR inhibitor
AZD9291	S7297	EGFR inhibitor
Bardoxolone methyl	S8078	Keap1/ Nrf2 activator
BAY 11–7082	S2913	NF-κBα (IκBα) inhibitor
Bay 11–7085	S7352	NF-κB α (IκBα) inhibitor
Belinostat	S1085	HDAC inhibitor
Benzethonium	S4162	Antiseptic/ tensioactive
BGT226	S2749	PI3K/ mTOR dual inhibitor
BI 2536	S1109	PLK1/ BRD4 inhibitor
BIIB021	S1175	HSP90 inhibitor
BIX 01294	S8006	Histone methyltransferase G9a inhibitor
BMS-754807	S1124	IGF-1R/ IR dual inhibitor
Bortezomib	S1013	Proteasome inhibitor
Brefeldin A	S7046	Antimicrobial (ATPase inhibitor)
Buparlisib	S2247	PI3K inhibitor
BV-6	S7597	cIAP / XIAP dual inhibitor
Caffeic acid phenethyl ester	S7414	NF-κB inhibitor
Camptothecin	S1288	Topoisomerase I inhibitor
Carfilzomib	S2853	Proteasome inhibitor
CB-5083	S8101	p97 inhibitor
Celastrol	S1290	Proteasome inhibitor
Cepharanthine	S4238	Anti-inflammatory
Ceritinib	S7083	ALK inhibitor
Cetrimonium	S4242	Antiseptic (quaternary ammonium)
Cetylpyridinium	S4172	Antiseptic (quaternary ammonium)
CH5138303	S7340	HSP90 inhibitor
Ciclopirox	S2528	Antifungal/ iron chelator
Cladribine	S1199	Adenosine deaminase inhibitor
CPI-203	S7304	BET inhibitor
Crystal Violet	S1917	Antiseptic (antimitotic)
Cucurbitacin B	S8165	Phytochemical (mitosis inhibitor)
CUDC-101	S1194	HDAC/ EGFR / HER2 triple inhibitor
CUDC-907	S2759	PI3K/ HDAC dual inhibitor
CX-6258	S7041	Pim kinase inhibitor
Dacinostat	S1095	HDAC inhibitor
Daunorubicin	S3035	Anthracycline-topoisomerase II inhibitor
Delanzomib	S1157	Proteasome inhibitor
Deltarasin	S7224	PDEδ/Kras dual inhibitor
Digoxin	S4290	Cardiac glycoside (Na/K ATPase inhibitor)
Domiphen	S4186	Antiseptic (quaternary ammonium)
Dovitinib Dilactic Acid	S2769	RTK inhibitor
Dp44Mt	S7909	Iron chelator
Entinostat	S1053	HDAC inhibitor
Epoxomicin	S7038	Proteasome inhibitor
Ethacridine lactate	S4196	Antiseptic
Etoposide	S1225	Epipodophyllotoxin (topoisomerase II inhibitor)
Fedratinib	S2736	JAK-2 inhibitor
Gambogic Acid	S2448	Apoptosis inducer
Ganetespib	S1159	HSP90 inhibitor
GDC-0980	S2696	PI3K/ mTOR dual inhibitor
Geldanamycin	S2713	HSP90 inhibitor
Gemcitabine	S1714	Antimetabolite
Givinostat	S2170	HDAC inhibitor
GMX1778	S8117	NAMPT inhibitor
GSK J4	S7070	JMJD3/ UTX dual inhibitor
GSK1070916	S2740	Aurora B/C inhibitor
GSK1324726A	S7620	BET inhibitor
GSK2126458	S2658	PI3K inhibitor
Hesperadin	S1529	Aurora B kinase inhibitor
HSP990	S7097	HSP90 inhibitor
ICG-001	S2662	Wnt pathway inhibitor
Idarubicin	S1228	Anthracycline (topoisomerase II inhibitor)
IKK-16	S2882	IKK inhibitor
Irinotecan	S2217	Topoisomerase I inhibitor
Ispinesib	S1452	KSP inhibitor
Ixazomib	S2180	Proteasome inhibitor
JTC-801	S2722	NOP antagonist
KPT-185	S7125	CRM1 inhibitor
KPT-276	S7251	CRM1 inhibitor
KW-2478	S2685	HSP90 inhibitor
Linsitinib	S1091	IGF-1R/ IR dual inhibitor
LMK-235	S7569	HDAC inhibitor
LMTX™	S7762	Tau protein aggregation inhibitor
LY3023414	S8322	PI3K/ mTOR / DNA-PK triple inhibitor
M344	S2779	HDAC inhibitor
MCB-613	S7913	SRC stimulator
Melphalan	S8266	Nitrogen mustard (DNA alkylator)
Methylene Blue	S4535	Oxidation-reduction agent
MG-101	S7386	Cysteine protease inhibitor
MG-132	S2619	Proteasome inhibitor
MI-773	S7649	MDM2 inhibitor
Milciclib	S2751	CDK/ TRK dual inhibitor
Mitomycin C	S8146	Antibiotic
Mitoxantrone	S1889	Topoisomerase II inhibitor
MK-1775	S1525	Wee1 inhibitor
MLN9708	S2181	Proteasome inhibitor
Mocetinostat	S1122	HDAC inhibitor
Navitoclax	S1001	BCL-2 inhibitor
NH125	S7436	eEF-2 inhibitor
NMS-E973	S7282	HSP90 inhibitor
NSC 319726	S7149	p53 reactivator
NVP-ADW742	S1088	IGF-1R inhibitor
NVP-AEW541	S1034	IGF-1R inhibitor
Obatoclax	S1057	BCL-2 inhibitor
Oligomycin A	S1478	Antibiotic (ATPase inhibitor)
Omaveloxolone	S7672	NRF2 activator
ONX-0914	S7172	Proteasome inhibitor
Oprozomib	S7049	Proteasome inhibitor
Oridonin	S2335	PI3K inhibitor
OTX015	S7360	BET inhibitor
Ouabain	S4016	Cardiac glycoside (Na/K ATPase inhibitor)
Pacritinib	S8057	JAK2/ FLT3 dual inhibitor
Panobinostat	S1030	HDAC inhibitor
PD0166285	S8148	Wee1 /Myt1 dual inhibitor
PF-04929113	S2656	HSP90 inhibitor
PF-477736	S2904	Chk1 inhibitor
Piperlongumine	S7551	ROS inducer
Pirarubicin	S1393	Antibiotic (anthracycline)
Pracinostat	S1515	HDAC inhibitor
PRI-724	S8262	Wnt signaling pathway inhibitor in cancer stem cell
PU-H71	S8039	HSP90 inhibitor
Puromycin	S7417	Antimicrobial/ antibiotic
Quisinostat	S1096	HDAC inhibitor
RAF265	S2161	RAF/ VEGFR2 dual inhibitor
Resminostat	S2693	HDAC inhibitor
Ricolinostat	S8001	HDAC inhibitor
Rigosertib	S1362	PLK1 inhibitor
Romidepsin	S3020	HDAC inhibitor
SB743921	S2182	KSP inhibitor
SBI-0640756	S8181	EIF4G1 inhibitor
Scriptaid	S8043	HDAC inhibitor
Selinexor	S7252	CRM1 inhibitor
Sepantronium	S1130	Survivin inhibitor
SGI-1776	S2198	Pim kinase inhibitor
SGI-7079	S7847	Axl inhibitor
SN-38	S4908	Topoisomerase I inhibitor
SNX-2112	S2639	HSP90 inhibitor
SRT1720	S1129	SIRT1 agonist
Stattic	S7024	STAT3 inhibitor
Staurosporine	S1421	Antifungal/ protein kinase C inhibitor
STF-118804	S7316	NAMPT inhibitor
TAE226	S2820	FAK/ IGF-1R dual inhibitor
Tanespimycin	S1141	HSP90 inhibitor
Teniposide	S1787	Epipodophyllotoxin (topoisomerase II inhibitor)
Tenovin-6	S4900	HDAC inhibitor
Terfenadine	S4353	Antihistamine
TG101209	S2692	JAK-2 inhibitor
TIC10	S7963	TRAIL inducer
Topotecan	S1231	Topoisomerase I inhibitor
Trichostatin A (TSA)	S1045	HDAC inhibitor
Tubercidin	S8095	Antibiotic (polymerase inhibitor)
UM171	S7608	HSC self-renewal agonist
UNC0631	S7610	Histone methyltransferase G9a inhibitor
UNC-2025	S7576	Mer/ FLT3 inhibitor
VER-49009	S7458	HSP90 inhibitor
VER-50589	S7459	HSP90 inhibitor
Verdinexor	S7707	CRM1 inhibitor
Vinblastine	S4505	Vinca alkaloid-mitotic inhibitor
Vincristine	S1241	Vinca alkaloid-mitotic inhibitor
Vinflunine	S2209	Vinca alkaloid-mitotic inhibitor
Vinorelbine	S4269	Vinca alkaloid-mitotic inhibitor
Vorinostat	S1047	HDAC inhibitor
VS-5584	S7016	PI3K/ mTOR dual inhibitor
XL888	S7122	HSP90 inhibitor

ALK, anaplastic lymphoma kinase inhibitor; BCL-2, B-cell lymphoma 2; BET, bromodomain and extraterminal domain; BRD4, bromodomain-containing protein 4; CDK, cyclin-dependent kinase; Chk1, checkpoint kinase 1; cIAP, cellular inhibitor of apoptosis protein-1; CRM1, chromosome region maintenance 1; DNA-PK, DNA-dependent protein kinase; Eef-2, eukaryotic elongation factor 2; EGFR, epidermal growth factor receptor; EIF4G1, eukaryotic translation initiation factor 4 gamma 1; FAK, focal adhesion kinase; FLT3, Fms like tyrosine kinase 3; HDAC, histone deacetylase inhibitors; HER2, human epidermal growth factor receptor type 2; HSC, hematopoietic stem cell; HSP90, heat shock protein 90; IGF-1R, insulin-like growth factor 1 receptor; IKK, inhibitor of nuclear factor kappa-B kinase; IR, insulin receptor; JK2/3, Janus kinase 2,3; JMJD3, Jumanji domain-containing protein D3; Keap1, Kelch-like ECH-associated protein 1; Kras, Kirsten rat sarcoma viral oncogene homolog; KSP, kinesin spindle protein; MDM2, mouse double minute 2 homolog; mTOR, mammalian target of rapamycin; NAMPT, nicotinamide phosphoribosyltransferase; NFR2, nuclear factor erythroid 2-related factor; NF-Κb, nuclear factor kappa-light-chain-enhancer of activated B cells; NOP, nociceptin Receptor; Nrf2, nuclear factor (erythroid-derived 2)-like 2; PDEδ, photoreceptor cGMP phosphodiesterase δ subunit; PI3K, phosphatidylinositol 3-kinase; PLK1, Polo-like kinase 1; RKT, receptor tyrosine kinases; SIRT1, Sirtuin 1; SRC, steroid receptor coactivator; STAT3, signal transducer and activator of transcription 3; TNKS, Tankyrase; TRAIL, TNF-related apoptosis-inducing ligand; TRK, tropomyosin receptor kinase; UTX, ubiquitously transcribed X-chromosome tetratricopeptide repeat protein; VEGFR2, vascular endothelial growth factor receptor 2; XIAP, X-linked inhibitor of apoptosis protein.

Only compounds that do not require metabolic activation and for which an injectable solution is available may be administered IVi or IT/IVt and thus were considered for local delivery. Subsequently, 13 compounds were excluded due to reports of vesicant, neurological, or ocular toxicities. Thus, 8 drugs, accounting for 5% of the total number of active compounds (n = 169), met all selection criteria for local IVi administration (Figure 1). This group of agents consisted of a histone deacetylase inhibitor (HDACi, romidepsin), one dye that also acts as an oxidation-reduction agent (methylene blue), one cardiac glycoside (digoxin), one antimetabolite (gemcitabine), and four agents already used in retinoblastoma including a topoisomerase I inhibitor (topotecan), a nitrogen mustard (melphalan), and two epidophyllotoxins (etoposide and teniposide).

Figure 1. Decision sequence to prioritize novel drugs for local and systemic administration in retinoblastoma.

Serious adverse events were those classified as grade III/IV adverse events by the Common Terminology Criteria for Adverse Events (CTCAE) v5. Hematological toxicities or other clinically manageable toxicities (named as serious adverse events) were not included.

^a Compounds in Phase I/II pediatric trials or that are used in routine clinical management.

^b Compounds in Phase I/II pediatric trials or that are used in routine clinical management by the indicated route of administration.

Abbreviations: IT, intrathecal; IVt, intraventricular; IVi, intravitreal injection; IAC, intraarterial chemotherapy; i.v, intravenous infusion, BBB, blood-brain barrier; BOB, blood-ocular barrier.

IT or IVt administration may provide an alternative route of local CNS drug delivery for drugs that do not cross the BBB but display cytotoxicity against retinoblastoma cells. Of the 8 drugs considered candidates for local treatment (topotecan, etoposide, teniposide, and romidepsin), 4 were only eligible for IT/IVt administration after ruling out 4 active compounds with reported neurologic toxicity or serious adverse events after IT or IVt administration (i.e. digoxin, gemcitabine) [26, 38, 39].

Twelve of the 30 drugs that fulfilled the criteria for drug repurposing do cross the BBB or BOB and are not associated with serious adverse events and thus could be applied for systemic (i.v., oral, or IAC) treatment. These compounds included five agents that are already clinically used to treat metastatic retinoblastoma: two topoisomerase I inhibitors (topotecan, irinotecan), two topoisomerase II inhibitors (etoposide, teniposide), and an anthracycline (idarubicin). The other agents identified included two antimetabolites (gemcitabine, cladribine), and additional repurposing drugs consisting of one dye (methylene blue), two HDACi (vorinostat, panobinostat), an ALK inhibitor (ceritinib), and an exportin 1 (XPO1) inhibitor (selinexor) (Figure 1). Eight of these drugs are available in an aqueous formulation and could be considered for i.v. treatment. Of this set, we excluded compounds with vesicant effects (idarubicin[40]) or that require metabolic activation (irinotecan and cladribine), and thus, 5 compounds remained prioritized for IAC administration. Melphalan was retained despite its myelosuppressive effects because of its high cytotoxicity; dose reductions have been proven effective for intraocular treatment by IAC administration and eliminate such risks[3, 11, 41].

The candidates that emerged as new hits in the discovery branch included 61 agents that inhibited primary cell growth by ≥ 90% and have not been clinically tested in adults or children. Eighteen agents were eligible for systemic administration, as they are able to cross the BOB or BBB, of which four lack severe toxicity, including JTC-801, brefeldin A, parthenolide, and omaveloxolone (Figure 1, right flowchart; mechanism of action summarized in Table 2).

Table 2.

Selected agents prioritized for drug discovery.

Systemic administrationa
Compound	Mechanism of action	Known and mechanism-related effects
Brefeldin-A	Antimicrobial; ATPase inhibitor	• Macrocyclic lactone • Exerts transcytosis inhibition resulting in impaired transcription • Induces caspase-mediated cell death independent of p53 status in several human tumor cell lines; also inhibits VEGF secretion[88, 89] • Safe in cultured rat photoreceptors[90]
JTC-801	Nociceptin receptor (NOP) antagonist	• Cytotoxic against NCI-60 human cancer cell panel and panel of human xenograft including ortothopic and metastatic tumors[91] • Effects mediated by pH-dependent cell death[91] • Induces apoptosis mediated by PI3K-AKT[92] • Inhibits invasion and migration[92]
Omaveloxolone	Nuclear factor erythroid 2-related factor (Nrf2) activator	• Antioxidant inflammation modulator (AIM); reduces oxidative stress and inflammation • Antitumor activity may also involve inhibition of NF-κB, decreased cyclin D1 levels, increased p21, and JNK phosphorylation[93] • Inhibits growth of several tumor cell lines • Well tolerated in adult patients with advanced solid tumors after oral administration[93] • Protects against oxidative stress-induced damage in retinal pigmented epithelium (RPE) cells[94]
(−)Parthenolide	NF-κB/ histone deacetylase dual inhibitor	• Sesquiterpene • Causes downstream PI3K/AKT pathway inhibition, reducing cell proliferation and inducing apoptosis in glioblastoma as well as several adult tumor cells[95] • Apoptosis induction may be mediated by generation of reactive oxygen species (ROS)[95] • Combination with the plant-derived compound okadaic acid increased oxidative stress that led to apoptosis in Y79 retinoblastoma cells mediated by PTEN[96]
Local administration b
YM155	Survivin inhibitor	• Inhibits transcription and expression of survivin, a member of the inhibitor of apoptosis family that also regulates mitosis, DNA damage repair and autophagy[97, 98] • Survivin overexpression is associated with poor prognosis and chemoresistance in other cancers, in which its inhibition slows tumor growth in preclinical models[97, 98] • Survivin is overexpressed in retinoblastoma tumors and retinoblastoma cell cultures[83, 99] • Suppressing survivin expression reduces Y79 cell growth and migration potential, and slows tumor growth in a xenograft model[100] • Low toxicity, no ocular toxicity[101, 102] • Poorly penetrates the BBB[103] •Antitumor effects are time dependent, requiring careful attention to treatment schedule [102]
ABT-737	BH3 mimetic inhibitor	• Inhibits anti-apoptotic B-cell lymphoma-2 (Bcl-2) and Bcl-xL proteins, resulting in apoptosis[104, 105] Promotes apoptosis in commercial WERIRB1 cells; relatively nontoxic to normal human photoreceptor cells[84] • Ineffective against Y79 cells, which do not express Bax and thus, not likely to be proapoptotic inall retinoblastomas •Navitoclax, a related compound, induces apoptosis in Y79 cells; the triple combination of this agent with topotecan and RAD51 inhibitor showed synergistic effects in vitro and in vivo[106] • Related Bcl-2 inhibitor obatoclax caused neurologic symptoms in patients in clinical trials, calling for evaluation of these adverse events after ocular administration of this agent[107, 108] • Local administration recommended because of severe thrombocytopenia reported in humans with navitoclax, probably resulting from on-target BclxL inhibition[109]
BIX01294 and UNC0631	G9a/ G9a-like protein (GLP)	• G9a and GLP are histone methyltransferases and specifically methylate the non-histone target p53 at lysine 373, leading to its inactivation[110, 111]
	inhibitor	• G9a is aberrantly regulated in multiple tumors[110] • No evidence of methyltransferase dysregulation in retinoblastoma to date
Ispinesib, ARQ621, and SB743921	Kinesin spindle protein (KSP) inhibitor	• Part of a promising family of antimitotic compounds that may circumvent current limitations of the antimitotic taxanes such as neurotoxicity and resistance • Overexpression of kinesin spindle (KIF) proteins, in particular KIF14, has been associated with tumor growth and poor prognosis in several cancers, but its role in tumor development and progression is still unclear[112] • These agents also inhibit Eg5, have shown cytotoxic activity against preclinical models, and have been evaluated human Phase I/II clinical studies in other cancers[113, 114] • KIF14 overexpression is among the most frequent aberrations in retinoblastoma, observed in almost 50% of cases[115]. • Overexpression of KIF14 in a transgenic model of retinoblastoma promoted tumor formation and enhanced tumor burden[116]
NH125	Eukaryotic elongation factor 2 (eEF-2) inhibitor	•Eukaryotic elongation factor-2 kinase (eEF2K) inactivates eEF2, limiting protein synthesis under stress[117] • Activation of eEF2K has been detected in cells under stress such as nutrient deficiency and promotes cell survival by inhibiting apoptosis[117] • eEF2K overexpression in some adult tumors has been associated with tumor growth and poor patient survival[117, 118] • This agent has cytotoxic activity and causes apoptosis in several adult tumor cell lines[119]
NSC 319726	p53 reactivator	•Cytotoxic in cells and mouse tumor models carrying the TP53R175H mutation, one of the most frequent TP53 mutations that causes gain-of-function by impairing Zn binding dependence for DNA binding [120] • Functions as a zinc ionophore, delivering the metal ion to the mutant p53 and restoring wildtype conformation[120] • Increases intracellular reactive oxygen species by this transcriptional activation of p53-R175H[121] • Other mechanism for cytotoxicity in wild-type p53 cells involve binding to copper and boost of intracellular reactive oxygen species resulting in cell growth arrest[122].
BAY 11–7085	NF-κB (nuclear factor κB) α (IκBα) inhibitor	• NF-kB transcription factors are associated with tumor growth and survival in addition to key function in innate immunity[123] • In retinoblastoma, NF-kB is overexpressed compared with normal retin[124] • Inhibition of NF-kB induces in vitro retinoblastoma cell apoptosis and sensitizes cells to doxorubicin [125]
SBI-0640756	Eukaryotic translation initiation factor 4 gamma 1 (EIF4G1) inhibitor	• EIF4G1 is a key component of the eIF4F protein complex that mediates initiation of mRNA translation • Several tumor cell lines overexpress EIF4G1 mRNA[126] • Exerts anti-proliferative effects in human melanoma cell lines[127]
Stattic	Signal transducer and activator of transcription 3 (STAT3) inhibitor	• STAT3 is a transcription factor that regulates cell cycle progression, survival, and inflammation[128]. STAT3 is constitutively activated in several human cancers and hyperphosphorylation has been associated with aggressiveness and immune evasion[129] •STAT3 is upregulated in retinoblastoma; STAT3 knockdown inhibits Y79 cell proliferation and suppresses tumor growth in vivo[130] • Non-toxic in RPE cells[131]

^aAble to cross the blood-brain or blood-ocular barriers and not associated with serious adverse events

^bCannot penetrate the blood-brain or blood-ocular barriers, or evidence that they can is not available; not associated with serious adverse events.

The remaining 43 agents were considered for local administration (IVi), from which we selected 12 compounds without vesicant or neurotoxic effects (reported for the compounds themselves or the pharmacological family to which they belong), and for which antitumor activity has been reported in models of adult or pediatric cancer. The final selection included inhibitors of survivin, antiapoptotic Bcl-2 family proteins, methyltransferase, and kinesin proteins, among others (Figure 1, right flowchart; full list with mechanisms of action in Table 2). Most of these hits’ mechanisms of action have not yet been determined in retinoblastoma, but the variety of cellular processes implicated suggests that at least some may improve tumor control.

Discussion

We report a drug selection process based on a rational, unbiased design to identify and prioritize potentially active drugs adapted to the common clinical scenarios for the treatment of retinoblastoma. This process, based on a strategy used to prioritize novel treatment combination for brain tumors [17, 29, 42, 43] combines pharmacological screening results with available data on clinical use and phase I/II trials, safety, and biopharmaceutical characteristics. Identified candidate drugs are then categorized according to their feasibility for use in various clinical scenarios (i.e., administration routes). Our screen and literature review identified 31 drugs worthy of further investigation in retinoblastoma, of which 15 have been clinically characterized and 16 that deserve further analysis.

Of note, three agents used as standard-of-care for retinoblastoma treatment were excluded during the decision process, implying that our classification prioritized drugs that demonstrated greater potency even compared with clinically used agents. Specifically, carboplatin and doxorubicin did not reach the cytotoxicity threshold of inhibiting cell growth by > 70% at the evaluated concentration, in line with previous reports[18, 44]. Vincristine was omitted due to its poor penetrance across the BBB and evidence of retinal toxicity after intravitreal injection in primates[45–48].

Eight compounds fulfilled the decision criteria for consideration as IVi-administered drugs. Three of these drugs are already used as standard therapy for retinoblastoma. Two others, teniposide and gemcitabine are used to treat pediatric tumors by the i.v. route[26, 49]. A newer drug with potential for IVi administration to treat retinoblastoma was the HDACi romidepsin. HDACis are a promising class of anticancer agents because their cytotoxic effects are mediated by epigenetic regulation, leading to selective activity against cells with deregulated E2F1 activity, as occurs in retinoblastoma due to loss of pRb[50, 51]. In fact, the cytotoxic effects of HDACis were previously reported in patient-derived and murine retinoblastoma cell lines[51]. The remaining two IVi candidates were digoxin and methylene blue. The preclinical profile of digoxin was previously reported by our group, showing a favorable local disposition after intravitreal injection in a rabbit model, but local retinal toxicity warrants further study prior to clinical translation[22, 52]. For methylene blue, a dye used for sentinel lymph node mapping for breast cancer and management of drug-induced methemoglobinemia, there is ample preclinical evidence of antitumor effects that may be potentiated by photodynamic therapy[53, 54]. Its affinity for pigments such as melanin and potentially for the retinal pigmented epithelium (RPE) may add activity against the tumor but also unwanted retinal toxicity[55].

The other local administration route considered in our prioritization system is IT/IVt administration, a means of bypassing the blood-brain and blood-cerebrospinal fluid barriers [26, 27, 29, 30, 56–58] to target leptomeningeal dissemination, a common and usually fatal dissemination site for retinoblastoma, which is challenging given the limited ability of chemotherapeutics that cross the BBB[1, 5, 8, 9]. Our selection process for IT/IVt administration, among the candidates listed above for the IVi route, excluded four drugs for which toxicity after IT/IVt administration has been reported[26, 38, 59]. Of the remaining four candidates, etoposide and topotecan have been given via the IT/IVt route to treat pediatric brain tumors. IVt etoposide has been used in children with metastatic brain tumors or neoplastic meningitis with good tolerability and favorable pharmacokinetics in CSF despite poor penetration into the parenchyma in some patients[26, 60, 61]. IT topotecan is currently used in retinoblastoma patients with histopathological features associated with high risk for CNS relapse[62, 63]. The remaining candidates, teniposide and romidepsin, have not been given IT/IVt to humans, though the IT safety of the HDACi panobinostat, in the same family as romidepsin, is supported by a study in nonhuman primates[64].

Most of the 12 drugs identified for systemic treatment of metastatic retinoblastoma (i.e. that can penetrate the BBB/BOB) have long been used for pediatric solid tumors (e.g. gemcitabine, irinotecan) or even for retinoblastoma (e.g. etoposide, topotecan), except for the two HDACi (vorinostat and panobinostat)[51, 65–68]. Romidepsin was excluded as a candidate for systemic treatment based on its cardiac effects, association with electrolyte imbalance, and poor CSF penetration [69–71].

Two new hits identified for the systemic route are noteworthy for the novelty of their mechanisms. Selinexor is an XPO1 inhibitor that alters nuclear-cytoplasmic export of cargo proteins, such as Rb, E2F1, and p53, leading to mislocalization and promoting cell survival and tumor progression[72, 73]. Preclinical evaluations of this drug by the Pediatric Preclinical Testing Program against a panel of cell lines and in vivo xenografts derived from solid and hematological malignancies showed promising responses[73]. This hit remained in the selection process despite reports of neurotoxicity in treated children at the maximum studied dose as this did not occur at a lower dose of 50 mg/m² and even lower doses could be used to avoid other grade 3 toxicities[74]. The other hit, ceritinib, is an FDA-approved ALK inhibitor able to cross the BBB, making it a promising candidate for targeting brain tumors[75, 76]. Ceritinib also inhibits several other targets, including the type 1 insulin-like growth factor receptor (IGF1R) and insulin receptor (IR)[77]. The IGF pathway has long been implicated in tumor cell proliferation and survival and its activation has been reported in pediatric brain tumors, likely relating to the high expression of IGF1R and its ligands during embryonic development[76, 78]. Y79 retinoblastoma cells produce IGF and display autocrine IGF1R-mediated growth, supporting the likely efficacy of ceritinib in this cancer[79]. Nonetheless, aberrations in ALK in retinoblastoma should be further explored.

Among the candidates for systemic administration, 6 have an aqueous formulation, lack vesicant effects, and do not require metabolic activation and were thus prioritized for IAC. Most of these compounds are already used in the clinic to treat intraocular retinoblastoma except for methylene blue.

Novel agents in the early stages of drug development may be a source of alternative compounds. We identified 16 drugs that showed activity against our primary cell cultures and are not associated with severe toxicity. Of these, four are able to cross the BBB/BOB and thus could be further studied for their applicability in the systemic treatment of retinoblastoma[80–82]. These agents display a variety of mechanisms of action including transcytosis, NF-kB/HDAC inhibition, and NRF2 activation (Table 2). The 12 hits that poorly penetrate the BOB/BBB and could be delivered by IVi or IT/IVt have similarly varied mechanisms, such as inhibition of survivin, histone methyltransferases, and kinesin spindle and antiapoptotic proteins, and p53 reactivation (Table 2). Among these compounds, the survivin inhibitor YM155 and the antiapoptotic protein inhibitor ABT737 have reported activity in retinoblastoma preclinical models[83, 84]. Other compounds have not been studied in retinoblastoma but show antitumor activity in pediatric and adult solid tumor models.

The present results may not be definitive, as a larger number of primary cell lines should be tested to characterize potential diversity in pharmacological response based on inter-individual variability in genomic features and drug sensitivity, but unfortunately retinoblastoma cells are difficult to grow in culture. Also, non-tumor cell lines should be screened to evaluate potential non-specific side effects of the prioritized compounds. Another limitation is that we based cytotoxicity on cell growth inhibition at a single concentration; further studies should characterize the complete dose-response profile for the intended candidates. Finally, we may have excluded some potentially active candidates because of toxicity at the reported concentrations; some of these may be effective at potentially safer and lower doses. One such drug was carfilzomib, a potent proteasome inhibitor that presents a lower incidence of peripheral neuropathy than bortezomib [85, 86]. Other compounds were ruled out due to scarce available information, especially regarding safety for IT/IVt use. Belinostat, an FDA-approved HDACi, was excluded because of potential cardiotoxicity that prevents i.v. administration and the incidence of severe neurologic toxicity that prohibits local administration[36, 87]. Nonetheless, our stringent criteria enhance our confidence that we selected compounds with the most promising features for development for use in retinoblastoma.

In conclusion, we developed a system for drug prioritization for retinoblastoma based on decision trees incorporating information from large-scale phenotypic screening of in vitro activity in patient-derived tumor cell lines, published clinical and preclinical evidence, and biopharmaceutical characteristics. This system identified and prioritized 36 compounds worthy of further study to determine safety and efficacy in this rare pediatric cancer.