Zebrafish tumor xenograft models for drug-screening: a systematic review of methods for treatment assessment

Abstract

Background

Zebrafish xenotransplantation models are increasingly employed as drug-screening platforms in precision oncology, offering the advantages of high replicate numbers and rapid data acquisition. Multiple strategies have been developed to evaluate anticancer therapies in zebrafish xenografts, varying in experimental design, methodology, and analytical approaches. However, a persistent limitation is the complexity of quantifying and automating drug-response measurements, which limits scalability and increases experimental workload.

Material and Methods

We conducted a systematic review to summarize methodologies for evaluating drug efficacy in human cells xenografted into zebrafish larvae. A comprehensive search of PubMed, Scopus, and Web of Science yielded 113 eligible studies after a two-step screening process.

Results

The review focused on two major aspects: xenotransplantation procedures and methods of drug-effect quantification. The analysis describes the injection of approximately 200 tumor cells into the embryos’ yolk sac, followed by a 72-hour treatment, representing the parameters most commonly employed. Drug efficacy was most frequently assessed by measuring tumor fluorescence intensity with a fluorescence stereomicroscope. Notably, 86.75% of studies tested five or fewer conditions per experiment.

Conclusion

In conclusion, there is a clear need to develop standardized protocols that allow testing of a larger number of treatment conditions, thereby supporting the advancement of high-throughput drug screening platforms for personalized medicine.

Background

Cancer remains one of the leading causes of death worldwide, characterized by its complexity and heterogeneity [1]. Despite significant advances in modern therapies, many treatments fail due to drug resistance and patient-to-patient variability, highlighting the limitations of “one-size-fits-all” approaches. As a result, personalized medicine has emerged as a promising strategy to tailor treatments based on the individual characteristics of each patient and each tumor [2].

In oncology research, mammalian models, especially rodents, have long been the cornerstone for evaluating the effects of anticancer therapies [3]. Unlike in vitro systems, animal models offer a higher level of biological complexity, capturing the full physiological context of a living organism [4]. The animals’ price, the complexity of in vivo experiments and the long time needed limit the number of experimental conditions. In this context, researchers have adopted alternative in vivo models to address some of these limitations. Among them, zebrafish (Danio rerio), a teleost fish belonging to the Cyprinidae family, has been increasingly used across various research fields, including cancer drug discovery [5]. The success of zebrafish as a model organism is largely due to its favorable biological traits and simple husbandry requirements. Notably, zebrafish shares the 71.4% of its genome with humans, resulting in significant similarities in biological and physiological processes [6]. Zebrafish can serve as an in vivo model at both embryonic and adult stages; however, the larval stage is mainly employed in oncology research owing to its ease of handling, low maintenance requirements, and several advantageous biological traits. Notably, zebrafish are highly fertile, a single female can produce hundreds of eggs per week, enabling experiments with multiple treatment groups and large sample sizes [7]. Moreover, zebrafish exhibit rapid development compared to other in vivo models, which represents a significant advantage by reducing the time required for data collection and analysis.

All these characteristics make zebrafish particularly suitable for testing drug panels. Zebrafish larvae can absorb small molecules directly from the water, allowing for simple and rapid administration of multiple treatments simultaneously. This approach is widely used to assess the model’s sensitivity to toxicity and to investigate the mechanisms underlying malformations or developmental defects in embryos [8]. The assessment of toxicity effects primarily focuses on the alterations induced in the embryo organism. However, to evaluate the effect of a drug on human cancer cells in vivo, it is required the xenotransplantation of tumor cells. The heterogeneity of cell lines and patient-derived primary cultures poses significant challenges to the development of standardized protocols. Indeed, cell injection procedures vary widely across research groups, introducing variability in both methodological approaches and data analysis workflows. A major challenge in assessing the effects of a treatment on cancer cell viability is accurate quantification. While murine models benefit from well-established and reliable techniques to assess changes in tumor volume, analogous assessments in zebrafish models are conducted using different and non-standardized methodologies. These techniques have to allow for the rapid analysis of drug effects across a high number of embryos simultaneously, within a limited timeframe.

This review wants to summarize the methods used to test the drugs effect on zebrafish cancer xenografts. The objective would be the identification of the best parameters to develop an in vivo robust, comprehensive, and data-driven drug-screening platform to supporting personalized medicine strategies.

Methods

Search strategy

Initially, we retrieved articles reporting experiments on the treatment of zebrafish xenografts in the field of oncology from three databases: PubMed, Scopus, and Web of Science. The search strategies were tailored to the specific characteristics of each database. To ensure comprehensive coverage of the relevant literature, we developed search strings that combined terms related to the population (zebrafish), the research field (oncology), and the main experimental features (xenografts and drug effects). No language restrictions or limit of year was applied.

Screening

The retrieved articles were first screened to remove duplicates using the Rayyan online tool. The second screening step involved excluding articles that did not fall within the scope of the review, based on their titles and abstracts. Studies included in the final systematic review were selected following a full-text screening phase. Both screening stages were independently conducted by two reviewers (OC and GM), who assessed each study individually. A third reviewer (LZ) resolved discrepancies between their decisions.

Eligible studies had to report experiments quantifying the effect of drugs on zebrafish embryo xenografts. During the initial screening (title and abstract), studies were excluded if they met any of the following criteria:

• Use of xenografts to investigate unrelated topics (e.g., angiogenesis, tumor–immune cell interactions, metastasis, etc.);

• Evaluation of nanodrugs or therapeutic vesicles;

• In vivo studies using non-zebrafish models;

• Use of adult zebrafish;

• Reviews, abstracts, editorials, case reports, etc.

The objective of the review was to identify methodologies for in vivo drug screening. In the full-text screening phase, studies were further excluded based on the following criteria:

• Fewer than 15 embryos per condition;

• Experimental endpoint beyond 120 hours post fertilization (hpf);

• Drug administration via injection or in vitro treatment prior to xenotransplantation.

No restrictions were applied regarding the types of drugs tested, as the focus of the review was on the methodologies employed.

Data extraction

Three reviewers (OC, LZ and GM) performed the data extraction from the full-paper text and the disagreements were discussed with a forth investigator (ADV). The data were extract from the “methods and materials” section and from the figures and captions. The following information and data were collected [1]: paper references and publication year [2]; xenograft protocols: cell type, fluorescent dye, number of cells, injection site, drug tested and treatment duration [3]; drug effect quantification methods: approach, instrument and software.

Risk of bias

The risk-of-bias assessment was performed by two independent authors (OC and GM) to evaluate the methodology and quality report of each study included. The disagreements were discussed and resolved between three authors (OC, ADV and GM). The report was conducted using the SYRCLE’s risk-of-bias tool for animal studies [9] with some adaptations. The domains were selected to adapt the risk of bias assessment with the topic of the review in term of animal’s management and experimental design. The methodology features were evaluated for the following quality points: (1) an adequate randomization process for the allocation of animals to experimental groups, (2) comparable baseline characteristics of the animals in each group, (3) allocation to the different groups adequately concealed, (4) blinded outcome assessment by investigators, (5) incomplete or missing data, (6) selective outcome reporting and (7) sample size. All the items were scored as presenting a low, high or unclear risk of bias.

Results

Studies selected

A total of 3201 articles were identified across three independent databases: PubMed (n = 934), Scopus (n = 1137), and Web of Science (n = 1130). After duplicate removal and initial title/abstract screening, 312 articles remained. Following full-text evaluation, 113 studies were included in the final analysis (Fig. 1) and the data extraction is summarized in Table 1.

Fig. 1.

PRISMA flow diagram illustrating the search results and study selection process [10]

Table 1.

List of included studies reporting the experimental parameters of xenotransplantation protocols and methods for data quantification

Study ID	Cell line (cancer type)	Staining method	Cell (n)	Injection site	T (h)	Method	Instruments	Software	Drugs	Conditions (n)
Geiger et al. [11]	U251 (Brain cancer)	RFP-expressing cells	50–200	YS	72	Fluorescence Intensity	Fluorescence microscope	Kodak Molecular Imaging software	Temozolomide + Radiotherapy	5
Zhang et al.[12]	K562 (Leukemia)	CM-Dil	100–200	YS	72	Tumor size	ImageXpressMICRO	MetaXpress software	Imatinib, Dasatinib, Parthenolide, TDZD-8, Arsenic trioxide, Niclosamide, Salinomycin, Thioridazine	9
Yang et al.[13]	U-87 MG (Brain cancer)	RFP-expressing cells	300	YS	48	Fluorescence Intensity	/	ImageJ	dl-NDGA	4
Chiu et al.[14]	H1299 (Lung cancer)	pDsRed-ExpressC1-expressing cells	50	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	/	BPIQ	4
Guo et al.[15]	Mia PaCa-2 (Pancreatic cancer)	/	100–200	PVS	48	Fluorescence Intensity	Fluorescence microscope	/	U0126	2
Tonon et al.[16]	JHH6 (Hepatocellular carcinoma)	CM-Dil	500	YS	24, 48, 72	Human gene expression	PCR machine	ImageJ	Bortezomib	2
Savarese et al.[17]	ACC (hTERT) (Adrenocortica carcinoma)	CM-Dil	240	PVS	72	Tumor size	Fluorescence microscope	Noldus DanioScopeTM	Cisplatin, Doxorubicin, Vorinostat, Everolimus, Olaparib, Palbociclib	8
Huo et al.[18]	A549 (Lung cancer)	CM-Dil	/	YS	72	Fluorescence Intensity	Fluorescence microscope	NIS-Elements D 3.10	Cisplatin, AECS, Cisplatin + AECS1	8
Fan et al. [19]	U-87 MG (Brain Cancer)	/	200	YS	72	Number of cell masses	Fluorescence microscope	/	Honokiol	2
Ikonomopoulou et al. [20]	MM96L (Melanoma)	CM-Dil	200–300	Blood circulation	72	Fluorescence Intensity	Fluorescence microscope	Quantifish	AgGom, HiGom	3
Yu et al.[21]	HSC-3 cells (Head snd Neck cancer)	CM-Dil	50	/	24, 48	Tumor size	Fluorescence microscope	/	Sandensolide	3
Tsering et al.[22]	MGC-803 (Gastric cancer)	CM-Dil	/	PVS	48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Triphala	4
Pudelko et al.[23]	Primary GBM cells and U343-MGA (Brain cancer)	luciferase -expressing cells	100	Blastula	48	Bioluminescence	Hidex Sense Microplate Reader	/	Erlotinib, Crizotinib, Gefitinib, Afatinib	5
Al-Samadi et al.[24]	UT-SCC-24A, UT-SCC-42A, UT-SCC-24B, UT-SCC-42B, HNSCC primary cells (Head and Neck cancer)	CellTrace Far Red	1000	PVS	72	Human gene expression	DigitalTM PCR System, PCR machine	QuantaSoft software	Cisplatin, Erbitux, Afatinib, Erlotinib, Gefitinib, Pimasertib, Temsirolimus, Sirolimus	9
Chou et al.[25]	MCF-7 (Breast cancer)	CM-Dil	200	YS	48	Tumor size	Fluorescence microscope	/	DEHP, Camptothecin, DEHP + Camptothecin	4
Liu et al.[26]	NCI-H23 (Lung cancer)	Fluorescent Cell Linker Kit	850	YS	24, 48	Tumor size	Fluorescence microscope	ImageJ	Paclitaxel	6
Yang et al.[27]	MDA-MB-231 and HCC1806 (Breast cancer)	CM-Dil	200–300	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	/	Gomisin M2	2
Jiao et al.[28]	MCF-7 (Breast cancer)	CM-Dil	200	PVS	24, 48	Tumor size	Fluorescence microscope	/	Betulinic acid	4
Gianoncelli et al.[29]	NCI-H295R (Adrenocortical carcinoma)	CM-Dil	240	PVS	72	Tumor size	Fluorescence microscope	Noldus DanioScopeTM software	Abiraterone Acetate	3
Tomko et al.[30]	MDA-MB-231 (Breast cancer)	CM-Dil	75–150	YS	72	Tumor size	/	/	Abnormal Cannabidiol and Its Analog O-1602	3
Chang et al.[31]	HCT116 and bromelain (Colorectal cancer)	CM-Dil	200	YS	72	Fluorescence Intensity	Fluorescence microscope	/	Bromelain	4
Carter et al.[32]	UM-SCC-81B (Head and Neck cancer)	RFP-expressing cells	/	Pericardial cavity	48	Tumor size	Fluorescence microscope	ImageJ	BH3 mimetic (A-1331852)	3
Usai et al.[33]	HCT 116 (Colorectal cancer) and MIA PaCa-2 (Pancreatic cancer)	CM-Dil	/	YS	72	Fluorescence intensity	Fuorescent stereoscope	ImageJ	Gemcitabin, GEMOX,Gemcitabin/nab-P, GEMCIS, 5-Fluorouracil, FOLFOX, FOLFIRI, FLOT, FOLFOXIRI, ECF	10
Sensi et al.[34]	Glioblastoma primary cells (Brain cancer)	Vybrant® DiI cell-labeling solution	200	Duct of Cuvier	72	Fluorescence intensity	Confocal microscopy	/	5-fluorouracil	3
Wang et al.[35]	A549 (Lung cancer)	DiO	100	PVS	48	Fluorescence intensity	Fluorescence microscope	/	Digoxin, Adriamycin	6
Gauert et al.[36]	SEM, RCH-ACV and 6 primary cultures (Leukemia)	Cell tracker violet	300–500	Pericardium	72	Immunostaining for flow cytometry	Flow cytometer	FlowJo software	Venetoclax	5
Jinendiran et al.[37]	HT-29 (Colorectal cancer)	CM-Dil	/	YS	24, 48, 72	Fuorescence intensity	Fuorescence microscope	/	DKP-1, DKP-2, DKP-3, DKP-4	5
Di Franco et al.[38]	Pieces of pancreatic cancer tissue (Pancreatic cancer)	CM-Dil	Tissue fragments	YS	48	Tumor size or “Response evaluation criteria in solid tumors (RECIST)”	Fluorescence microscope	ImageJ	Gemcitabine, GEMOX, Gemcitabine + nabPaclitaxel, FOLFOXIRI	5
Vicente et al.[39]	MOLM-13 and HL-60 (Leukemia)	CM-Dil	50–75	YS	72	Tumor size	Fluorescence microscope	LasX software	CM-1231	4
Yao et al.[40]	MDA-MB-231 (Breast cancer)	CM-Dil	200	YS	72	Fluorescence Intensity	Fluorescence microscope	Image J	3 HDAC inhibitor	4
Arriazu et al.[41]	HL60 and MOLM-13 cells (Leukemia)	CM-Dil	50–75	YS	72	Fluorescence Intensity	Fluorescence microscope	Leica Application Suite-X software	FTY720, CX-4945, FTY720 + CX-4945	4
Slater et al.[42]	Mel285 and OMM2.5 (Uveal Melanoma)	CM-Dil	200–500	PVS or the vitreous	72	Tumor size	Confocal microscope	/	Guininib, 1–4-dihydroxy quininib or Montelukast	4
Zhang et al.[43]	A375 (Melanoma)	CM-Dil	200	YS	24	Fluorescence Intensity	Fluorescence microscope	/	Theaflavin, Cisplatin	5
Liu et al.[44]	H1299 (Lung cancer)	CM-Dil	50	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	/	4-HPPP	2
Cao et al.[45]	A375 and A2058 (Melanoma)	CM-Dil	/	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Shikonin and Sorafenib	5
Wang et al.[46]	MDA-MB-231 (Breast cancer)	CM-Dil	200	PVS	48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Sanguisorba officinalis L.	3
Delasoie et al.[47]	HCT-116 (Colorectal cancer)	CellTracker™ RedCMTPX	150	YS	72	Tumor size	Fluorescence microscope	ImageJ	Rhenium(I) tricarbonyl-based complexe	4
Abate et al.[48]	MUC-1 and TVBF-7 ACC (Adrenocortical carcinoma)	CM-Dil	250	YS	72	Tumor size	Fluorescence microscope	Noldus DanioScopeTM software	Trabectedin	2
Huang et al.[49]	/	CM-Dil	/	YS	48	Tumor size	Confocal microscope	/	Quinoline Derivative (DFIQ)	3
Li et al.[50]	HELA (Urogynecological cancer)	GFP-expressing cells	/	YS	48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Lipophilic Grape Seed Proanthocyanidin	5
Wu et al.[51]	CNE-1 (Head and Neck cancer)	CM-Dil	100–200	YS	72	Fluorescence Intensity and tumor area	Fluorescence microscope	/	Phenanthroimidazole derivative (, 2,3-dichlorophenyl)	3
Gao et al.[52]	HGC-27 (Gastric cancer)	GFP-expressing cells	300	YS	48	Tumor size	Fluorescence microscope	ImageJ	Actinidia chinensis Planch	3
Song et al.[53]	CM2005.1, CRMM1, and CRMM2 (Melanoma)	CM-Dil	100	PVS	72	Tumor size	Automated microscope system	/	Nutlin-3, Mitomycin C, Nut3 + MMC	4
Sun et al.[54]	MCF-7 (Breast cancer)	Red CMTPX	400	YS	24	Fluorescence Intensity	Fluorescence microscope	/	2-ME2 (2-methoxyestradiol)	5
Rossini et al.[55]	NT2/D1 (Gynecological cancer)	/	250	YS	72	Tumor size	Fluorescence microscope	DanioScope software	Cisplatin, Palbociclib, Cisplatin + Palbociclib	5
Hou et al.[56]	HCT 116 (Colorectal carcinoma)	CM-Dil	300	YS	24	Tumor size	Fluorescence microscope	/	Nannocystin ax, Cisplatin	4
Huang et al.[57]	Hep3B and HepJ5 (Hepatocellular carcinoma)	CM-Dil	200	YS	72	Fluorescence Intensity	Fluorescence microscope	/	Methyl gallate	2
Yang et al.[58]	MCF-7 (Breast cancer)	CM-Dil	100–500	PVS	72	Tumor size	Fluorescence microscope	/	(E)-N-2(5 H)-furanonyl sulfonyl hydrazone derivative	4
Jhou et al.[59]	HSC-3 (Head and Neck cancer)	CM-Dil	50	YS	48	Fluorescence Intensity	Fluorescence microscope	/	Chlorpromazine	3
Wang et al.[60]	MDA-MB-231 (Breast cancer)	CM-Dil	200	PVS	48	Fluorescence Intensity	Fluorescence microscope	/	Ursolic acid	3
Deng et al.[61]	Hep3B (Hepatocellular carcinoma)	CM-Dil	200	YS	72	Tumor size	Fluorescence microscope	/	1β–OH–arenobufagin and sorafenib	5
Li et al.[62]	HepG2 (Hepatocellular carcinoma)	CM-Dil	50	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Casearlucin A and Etoposide	5
Kuo et al.[63]	MCF-7 (Breast cancer)	CM-Dil	300	YS	72	Fluorescence Intensity	Fluorescence microscope	ImageJ	Cisplatin, 5-fluorouracil, cyclophosphamide, Etoposide	8
Lu et al.[64]	CCRF-CEM (Leukemia)	CM-Dil	/	YS	48	Fluorescence Intensity	Fluorescence microscope	NIS-Elements	NOAEL, Cisplatin	5
Huang et al.[65]	HepG2 (Hepatocellular carcinoma)	CM-Dil	200	PVS	72	Tumor size	Fluorescence microscope	ImageJ	Compound Phyllanthus urinaria L. (CP)	3
Ali et al.[66]	Lung primary cultures (Lung cancer)	Fast-DiITM oil	/	PVS	72	Tumor size	Fluorescent stereoscope	HuginMunin software	Erlotinib, Paclitaxel	3
Almstedt et al.[67]	U3013MG, U3054MG, U3180MG (Brain cancer)	GFP- expressing cells	150–200	Brain	90	Fluorescence intensity	Hidex Sense Microplate Reader	/	Marizomib	2
Li et al.[68]	A375 (Melanoma)	CM-Dil	200	YS	24	Fluorescence intensity	Fluorescent microscope	/	TB, 1/10 NOAEL of TB, 1/3 NOAEL of TB, NOAEL of TB, Cisplatin	5
Pietrobono et al.[69]	A375 (Melanoma)	CM-Dil	250	YS	48	Fluorescence intensity	Fluorescent microscope	ImageJ	PLX-4032	2
Moral-Sanz et al.[70]	MM96L (Melanoma)	CM-Dil	/	/	48	Fluorescence intensity	Fluorescence microscope	Quantifish Software	Octpep-1	2
Maradonna et al.[71]	HCT 116 (Colorectal cancer)	Dil Vybrant Red Fluorescent dye	200–400	YS	72	Fluorescence intensity	Fluorescent microscope	Fiji	Anandamide—AEA and AM251	4
Krstic et al.[72]	Panc-1 (Pancreatic cancer)	CellTracker™ RedCMTPX	150	YS	72	Fluorescence Intensity	Fluorescence microscope	Fiji	Coumarin-palladium(II) complex	3
McKeown et al.[73]	MDA-MB-231 (Breast cancer)	CM-Dil	/	YS	48 h	Number of cells after tissues dissociation	Fluorescence microscope	ImageJ	Jadomycin B	4
Sundaramurthi et al.[74]	OMM2.5 (Uveal melanoma)	CM-Dil	200–500	PVS	72	Fluorescence Intensity	Fluorescence microscope	ImageJ	Dacarbazina, Ricolinostat	3
Wang et al.[75]	HepG2 (Hepatocellular carcinoma)	CM-Dil	50	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Se-AQP70-2B, Etoposide	5
Tonon et al.[76]	JHH-6 (Hepatocellular carcinoma)	FAST DiI	500	YS	72	Tumor size	/	/	5-Azacytidine	2
Gamez-Chiachio et al.[77]	HCC1954 LR (Breast cancer)	GFP-expressing cells	/	PVS	48	Fluorescence Intensity	Fluorescence microscope	QuantiFish	Lapatinib, Chloroquine	4
Chen et al.[78]	HSC3 (Head and Neck cancer)	CM-Dil	50	YS	72	Tumor size	Fluorescence microscope	/	Trichodermin	3
Romagnoli et al.[79]	HeLa (Urogynecological cancer)	VybrantTM DiI dye	200	Duct of Couvier	24, 48, 72	Fluorescence Intensity	Fluorescence microscope	/	2-Anilino Triazolopyrimidine (3d)	3
Zheng et al.[80]	A549 (Lung cancer)	CM-Dil	200	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Epimedokoreanin B, Cisplatin	4
Siebert et al.[81]	RMS cells (Soft Tissue Sarcoma)	Dil Vybrant Red Fluorescent dye	100	Blastoderm	24, 72, 120	Tumor size	Fluorescent microscope	Fiji	Vincristine, Dactinomycin, Trametinib	4
Kanellis et al.[82]	MDA-MB-231 (Breast cancer)	luciferase -expressing cells	/	YS	48	Tumor size	Microplate reader	ImageJ	AQ, CuET AQ + CuET	4
Ye et al.[83]	H1299 (Lung cancer)	CM-DiI	200	YS	24	Fluorescence intensity	Fluorescence microscope	NIS Elements D 3.20	QBD, Cisplatin	5
Grissenberger et al.[84]	shSK-E17T, shA673-1c and IC-pPDX-87 (Sarcoma of the Bone)	CellTrace™ Violet	/	PVS	48	Tumor size	Microplate reader (Operetta CLS)	Harmony Software 4.9	Irinotecan, S63845, S64315, A-1331852, YK-4–279, S63845 + YK-4–279, A-1331852 + Irinotecan, S64315 + Irinotecan, A-1331852 + S63845	9
Nasiri Sovari et al.[85]	Panc-1 and -HCT-116 (pancreatic and colorectal cancer)	CellTracker™ RedCMTPX	150	YS	72	Tumor size	Fluorescence microscope	ImageJ	Re(I) tricarbonyl complexes	6
Yang et al.[86]	HepG2 (Hepatocellular carcinoma)	mCherry-labeled	/	PVS	72	Fluorescence Intensity	Fluorescence microscope	ImageJ	CPI-169, Roblitinib	4
Sturtzel et al.[87]	SK-N-MC (skin tumor), 143B (Sarcoma of the Bone), U-87 MG and STA-NB-8 (Brain cancer)	GFP-expressing cells	200–400	PVS	72	Tumor size	Microplate reader (Operetta CLS)	Harmony Software 4.9	YK-4–279, K-975, A-1331852, Temozolomide, Ceritinib, Temozolomide+Ceritinib, 17-DMAG,TSF-15	4
Huang et al.[88]	HeLa (Urogynecological cancer)	CM-Dil	/	YS	48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Cholesterol-based selenocyanate compound (9f)	2
Sundaramurthi et al.[89]	OMM2.5 (Uveal melanoma)	CM-Dil	200–500	PVS	72	Fluorescence Intensity	Fluorescence microscope	ImageJ	Ergolide	2
Xie et al.[90]	HCT 116 (Colorectal cancer)	Vybrant CM-Dil	/	PVS	72	Tumor size	Fluorescence microscope	ImageJ	APR-246 + radiotherapy	4
Sokary et al.[91]	MDA-MB-468 and MDA-MB-231 (Breast cancer)	CM-Dil	/	YS	48	Fluorescence Intensity	Fluorescence microscope	ImageJ	Matcha	3
Tamburello et al.[92]	NCI-H295R, MUC-1, and TVBF-7 cells (Adrenocortical Carcinoma)	CM-Dil	250	YS	72	Tumor size	Confocal microscope	Zen 2.3 Black software	Progesterone	2
Kowald et al.[93]	UM-UC-3 and 6 primary cultures (Gynecological cancer)	Fast-DiI™ dye	300	PVS	72	Tumor size	Fluorescence microscope	ImageJ	Bacillus Calmette-Guérin (BCG) immunotherapy	5
Li et al.[94]	MCF-7 (Breast cancer)	CM-Dil	50	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Chlorahololide D, Etoposide	5
Hou et al.[95]	A549 (Lung cancer)	CM-Dil	50	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Effusanin B, Etoposide	5
Yu et al.[96]	KBvin cells (Head and Neck cancer)	/	/	YS	48	Tumor size	Fluorescence microscope	/	Paclitaxel, 3 arctigenin derivatives	4
Rossini et al.[97]	NT2/D1/-R and NCCIT/-R (Gynecological cancer)	Red fluorescent lipophilic dye	250	YS	72	Tumor size	Fluorescence microscope	Zen 2.3 Black software	Dinaciclib, Cisplatin, Dinaciclib + Cisplatin	4
Korelin et al.[98]	UT-SCC-42A (Head and Neck cancer)	Far Red	1500	PVS	72	Fluorescence intensity	Fluorescence microscope	imageJ	Navitoclax, A-1331852	3
Zhao et al.[99]	HCT116 (Colorectal cancer)	CM-Dil	/	/	48	Area growth rate/Fluorescence intensity	Fluorescence microscope	ImageJ	Penexanthone A, Cisplatin, Penexanthone A + Cisplatin	6
Yen et al.[100]	Kasumi-1 (Leukemia)	CM-Dil	500–1000	YS	24, 48, 72	Huma gene expression	PCR machine	/	Shikonin	2
Zivanovic et al.[101]	MiaPaCa-2 and PANC-1 (Pancreatic cancer)	Red CellTracker	/	YS	72	Tumor size	Fluorescence microscope	ImageJ	Quinoline derivatives 1 and 3	3
Ravichandiran et al.[102]	MCF-7 (Breast cancer)	CM-Dil	300	YS	72	Fluorescence Intensity	Fluorescence microscope	Zeiss hardware	Naphthoquinone derivative	3
Song et al.[103]	Primary culture Ovarian Cancer (Urogynecological cancer)	CM-Dil	500	PVS	48	Tumor size	Fluorescence microscope	ImageJ	Carboplatin	2
van Bree et al.[104]	Murine tumor-isolated neuroepithelial stem cells, UW228-3 (Brain cancer)	Luciferase-expressing cells	100–300	Centre of the cell mass (4hpf)	48	Luciferase activity	Microplate luminometer	/	Sonidegib, 4-hydroxycyclophosphamide (4-HCP)	3
Seiboldt et al.[105]	HD-MB03 (Brain cancer)	CM-Dil	/	YS	72	Tumor size	Confocal microscope	ImageJ	Isotretinoin, Navitoclax	4
Liu et al.[106]	A549 (Lung cancer)	CM-Dil	500	YS	72	Fluorescence Intensity	Confocal microscope	ImageJ	Polysaccharide, Lentinan	5
Kuttikrishnan et al.[107]	K562 cells (Leukemia)	CM-Dil	/	YS	72	Fluorescence Intensity	Fluorescence microscope	ImageJ	Neosetophomone B (NSP-B)	2
van den Bosch et al.[108]	92.1, Mel202 and MP46 (Uveal melanoma)	CM-Dil	300	YS	72	Tumor size	Confocal microscope	Fiji	Ricolinostat, Quisinostad, E7107, withaferin	4
Chen et al.[109]	HSC-3 (Head and Neck cancer)	CM-Dil	200	YS	48	Human gene expression	PCR machine	/	Rhopaloic acid A	4
Dernovsek et al.[110]	SK-N-MC (Ewing sarcoma)	/	/	PVS	72	Tumor size	Microplate reader (Operetta CLS)	/	Triazole-based C-terminal Hsp90 inhibitor	2
Chien et al.[111]	KB/VIN cells (Head and Neck cancer)	CM-Dil	50	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	/	Mosloflavone, Paclitaxel, Mosloflavone + Paclitaxel	6
Kwiatkowska et al.[112]	DLD-1 and HT-29 (Colorectal cancer)	Vybrant Dil	300	YS	48	Human gene expression	PCR machine	/	MM-129, Indoximod, MM-129 + Indoximod	4
Xu et al.[113]	HepG2 (Hepatocellular cancer)	Acridine orange (AO)	/	YS	48	Fluorescence Intensity	Fluorescence microscope	/	Sorbaria Sorbifolia (FDSS), Sorafenib	5
Liu et al.[114]	A549 (Lung cancer)	CM-Dil	50	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	natural diterpene (1b), Etoposide	5
Chou et al.[115]	A549 (Lung cancer)	CM-Dil	50	YS	24, 48	Fluorescence Intensity	Fluorescence microscope	/	TFPA	4
Treis et al.[116]	SK-N-BE [2] (Brain cancer)	GFP-expressing cells	100–300	PVS	72	Tumor size	Automated microscope system	ImageJ	SL-176, GSK-J4, SL-176 + GSK-J4	4
Li et al.[117]	A549 (Lung cancer)	CM-Dil	200	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Garcioligantone J and K, Cisplatin	6
Chien et al.[118]	KB (Head and Neck cancer)	CM-Dil	50	YS	24, 48	Tumor size	Fluorescence microscope	/	DDC, Paclitaxel, DDC + Paclitaxel	6
Jiang et al.[119]	HepG2 (Hepatocellular cancer)	CM-Dil	/	YS	48	Fluorescence Intensity	Confocal microscope	ImageJ	Natural Alkaloid (6-Hydroxymethyldihydronitidine) and Etoposide	5
Wragg et al.[120]	RD, JR1, SMS-CTR, Rh30, Rh4, Rh41 and RMS01 (Soft Tissue Sarcoma)	CM-Dil	250	YS	72	Cell number assessed by tumour dissociation, followed by trypan blue staining and cell counting	Haemocytometer	Fiji	Regorafenib, Infigratinib	2
Chang et al.[121]	MDA-MB-231 (Breast cancer)	CM-Dil	/	/	24, 48	Tumor size	Fluorescence microscope	/	BHM, BHG, Paclitaxel	11
Chen et al.[122]	MDA-MB-231 (Breast cancer)	CM-Dil	/	YS	48	Tumor size	/	/	Honokiol	2
[123]et al.123	MCF-7 and T47D (Breast cancer)	Vybrant DiD	300	YS	72	Human gene expression	PCR machine	/	Selisistat, Paclitaxel, Selisistat + Paclitaxel	4

Abbreviations: YS, Yolk Sac; PVS, Perivitelline Space

The first included study was published in 2013 (Fig. 2). Over time, the number of publications steadily increased, with the exception of 2017, which showed a decline, and 2020, which represented the peak with 20 publications.

Fig. 2.

Number of publications over the years. A quantitative analysis of zebrafish research publications in peer-reviewed scientific journals

The cancers investigated span a wide range of tissue origins (Fig. 3). The kind of cancer most frequently studied is breast cancer, featured in 19 studies (16.5%), and followed by lung (11.3%) and head and neck tumors (10.4%). The most commonly used cell line was the triple-negative breast cancer line MDA-MB-231, employed in 20 of the studies, followed by the colorectal carcinoma line HCT116, the non-small cell lung cancer NSCLC line A549 and the invasive ductal breast carcinoma line MCF-7 each used in 13 studies.

Fig. 3.

Overview of the cancer types included in the article. Number of publications on each tissues

Only eight studies investigated drug responses using patient-derived primary cultures, a critical aspect in the context of personalized medicine. Several studies involving primary culture drug screening were excluded due to an insufficient number of embryos per treatment group.

Xenograft methods and protocols

We reported the main parameters used to perform xenotransplantation of tumor samples in the selected studies (Table 1). The results revealed heterogeneous approaches, particularly regarding the number of injected cells and the dyes used for sample labeling. All studies injected single-cell suspensions, except for one, which transplanted pancreatic cancer tissue fragments into the yolk sac of 48 hpf larvae. While other groups tested treatments on undigested tissue pieces, the study reported here is the only one that included ≥15 embryos per condition.

The staining methods employed can be classified into two categories: fluorescent cell tracker dyes and cell engineering. Among the selected studies, 67.8% used the CM-DiI CellTracker™, making it the most widely used dye. An additional 13.5% used other live-cell trackers emitting in the red/far-red spectrum, while only two studies employed the CellTrace™ Violet Cell Proliferation Kit (blue fluorescence), and another two used the Vybrant™ DiO Cell-Labeling Solution (green fluorescence). Only 10.7% of the studies used a cell engineering approach. In total, ten studies injected cells transfected to express fluorescent proteins, primarily GFP and RFP (6 and 3 studies, respectively). Only two studies quantified the number of transplanted cells via bioluminescence, using cells transfected with the luciferase gene.

Following cell labeling, all studies employed microinjection techniques for xenotransplantation into zebrafish larvae. Due to the permeability of the embryonic body, cells can be transplanted into different anatomical sites, with the assurance that drugs can be absorbed directly from the surrounding water. Nonetheless, 65.8% of the protocols involved xenotransplantation into the yolk sac, while 21.9% targeted the perivitelline space of 48 hpf embryos.

The most heterogeneous experimental parameter observed was the number of cells injected. One of the limitations of the zebrafish xenograft model is the difficulty in determining the exact number of cells delivered per injection. For this reason, many studies reported cell numbers as ranges. As shown in Fig. 4, 50 to 500 of cells per injection represents the most represented range, with a peak of 30 articles injecting approximately 200 cells.

Fig. 4.

Number of cells injected per embryo across the selected studies. For studies reporting a range, each value within the range was counted

The predominant time point for xenotransplantation of cancer cells was 48 hpf. Considering 120 hpf as the experimental endpoint, the maximum treatment duration is 72 hours. Indeed, 52.2% of the studies treated larvae for 72 hours post-injection (hpi), while 43.4% and 3.5% limited the exposure to 48 hpi and 24 hpi, respectively.

In order to develop drug-screening platforms for precision medicine approaches, it is important to leverage the high-throughput potential offered by the zebrafish model. We included only studies that used groups of ≥ 15 embryos. Consequently, 101 papers were excluded because the treatment groups did not meet the selected cut-off. 86.75% of the studies tested five or fewer conditions (including untreated controls), and only two studies tested ten or more conditions (mean = 4.1).

Drugs effect quantification approaches

We collected the main characteristics of the methods used to quantify drug effects on zebrafish tumor xenografts, including the approaches, instruments, and software employed (Table 1). As previously described, the majority of methods utilized fluorescently labeled cells to track and visualize tumor cells within the embryos. The two primary techniques used to assess changes in cell number were fluorescence intensity measurement (50%) and quantification of the fluorescent tumor mass area (39.4%). Two studies employed luminescence-based approaches, using luciferase-expressing cells and microplate readers to quantify emitted light.

A completely different strategy involved the quantification of human RNA following larval digestion. Enzymatic and mechanical dissociation protocols were applied to isolate a mixed population of zebrafish and human cells (5.3%). RNA extraction followed by reverse transcription and PCR analysis was then used to quantify human gene expression relative to zebrafish housekeeping genes, providing an estimate of cancer cell burden.

Measurement of tumor size and fluorescence intensity relies on imaging technologies. Stereomicroscopes were the most commonly used, likely due to their simplicity and rapid image acquisition capabilities. Indeed, 67.2% of the selected articles used stereomicroscopes, while 14.1% employed confocal microscopy. Despite offering higher resolution, confocal imaging requires additional sample preparation and longer acquisition times to reconstruct the three-dimensional structure of fluorescent signals.

Quantification of fluorescence intensity or distribution was performed using software tools. Notably, this information was unreported in 34.5% of the studies. 33.6% used various versions of ImageJ, whereas Fiji was specifically mentioned in only five papers. Interestingly, some research teams adopted specialized software for zebrafish fluorescence quantification, particularly Noldus DanioScope™ (3.5% of studies) and QuantiFish (2.6%).

Risk of bias

The methodological quality of the selected studies was generally considered low in terms of overall risk of bias (Fig. 5). Items 1 and 5 were classified as unclear in all studies, due to the nature of the experimental designs included in this review.

Fig. 5.

Risk of bias assessment of the included studies. The assessment was conducted using the SYRCLE’s risk of bias tool. Each item account for methodological quality and were scored as presenting a high, unclear, or low risk of bias. The classification is presented as the percentage (%) of the assessed studies (n = 113) that received each rating. Seven methodological domains were evaluated: 1) allocation sequence, 2) baseline, 3) allocation concealment, 4) blinding, 5) incomplete data, 6) selective reporting, and 7) sample size. Green bars represent low risk of bias, yellow bars represent unclear risk of bias, and red bars represent high risk of bias

A total of 55.7% of the papers described randomization or screening methods prior to allocation into different treatment groups, while 44.3% did not report any such protocol. Allocation of animals to different experimental conditions was considered to carry a low risk of bias in 100% of cases, as injected and non-injected embryos were consistently placed in the same incubators. The same low risk of bias (100%) was observed for items 4 and 6 (blinding and selective reporting), as no selection of animals was performed at the end of the experiments, and all subjects were included in the outcome analysis.

We also evaluated methodological quality based on sample size. During full-text screening, we excluded studies in which experimental groups consisted of fewer than 15 embryos, unless the sample size was not reported. Studies meeting this criterion were considered to carry a high risk of bias, accounting for 36.6% of the total (41 studies). Despite this, the overall risk of bias was still considered acceptable for item 7.

Discussion

In recent decades, zebrafish has increasingly been used in various scientific fields, including preclinical research. This is due to several advantages, particularly when utilized at the embryonic stage. Specifically, their rapid development, suitability for xenografting, and the ability to absorb small molecules directly from the water make zebrafish an effective model for evaluating the effects of drugs on cancer cells. Additionally, the low number of cells required per xenotransplantation, combined with the species’ high fecundity, which provides a large number of embryos per experiment, further supports the use of zebrafish as a robust platform for drug screening and the assessment of diverse treatments and combination therapies.

In this context, traditional in vivo approaches have long been employed, leading to the development of well-standardized protocols for assessing drug effects. In contrast, owing to the relatively recent adoption of the zebrafish model, methodologies for evaluating drug responses in cancer xenografts remain heterogeneous. In this work, we aimed to review the methodologies reported in the scientific literature to identify a potential protocol suitable for drug-screening procedures in personalized medicine. The ideal system should enable a high number of patient-derived xenografts to be tested against a wide panel of treatments, in order to identify the most effective therapy for each patient.

We screened the entire scientific literature using major databases to identify studies employing the zebrafish model to evaluate drug efficacy on human cancer cells. No restrictions were applied regarding the research field, as the primary focus of this review is methodological. Therefore, exclusion criteria were defined based on experimental design. One of the key criteria concerned the number of larvae used per treatment condition. The published literature reports a wide range of sample sizes considered representative for experimental outcomes. To define a cut-off for inclusion, we selected two recent and relevant publications in the field, Habjan et al. and Bedell et al. [124]^, [125], which reported sample size ranges of 15–20 and 3–15 embryos, respectively. Based on the point of agreement between these studies, we adopted a minimum of 15 embryos per condition as the inclusion threshold. Sample size is inherently linked to experimental objectives, and given the preclinical nature of a drug-screening platform, we opted for a relatively high minimum number. This criterion led to the exclusion of 102 articles. The second major exclusion parameter was the duration of the experiments. Within the European Union, the critical limit beyond which a zebrafish embryo is considered an animal subject to ethical regulation is 5 days post fertilization (dpf), as defined by Directive 2010/63/EU. To propose a universally applicable drug screening method that avoids regulatory constraints, we selected 5 dpf as the experimental endpoint. This choice limits the duration of drug exposure and, consequently, may influence the interpretability of results. Based on this criterion, an additional 70 papers were excluded. As previously explained, the focus of the review is to identify a potential workflow for using zebrafish xenografts to test the efficacy of large drug panels. In this context, treatment administration has to be consistent across all experimental conditions. Drug delivery can be performed using three main methods: directly added to the water, administered via a second injection into the bloodstream, or through pre-treatment of cancer cells prior to injection. A second injection increases both the stress experienced by the larvae and the number of experimental variables, such as the difficulty in assessing injection success. The pre-treatment approach is also challenging to manage and time-consuming, as it requires multiple, independent in vitro treatments followed by individual injections. For these reasons, we opted to include only papers that administer treatments directly into the water. It offers a simple and rapid experimental workflow, two crucial features of a drug-screening system. Its main limitation lies in the types of compounds that can be tested: zebrafish larvae can only absorb small molecules through passive diffusion, rendering the administration of nano and macromolecules unfeasible. Additionally, the zebrafish adaptive immune system takes several weeks to become fully functional, which precludes the testing of immunotherapeutic agents. The absence of active lymphocytes, key players alongside cancer cells in determining the effects of immunotherapy, makes such evaluations impossible. Nevertheless, this limitation is inherent to the zebrafish embryo model itself and is not related to the method of drug administration.

Based on the results of the literature screening and analysis, the most commonly used experimental parameters for performing zebrafish xenografts are: (i) staining of cancer cells with the CM-DiI dye, (ii) injection into the yolk sac, (iii) implantation between 50 to 500 cells per xenograft, and (iv) 72 hours of drug exposure. Regarding the methods used to quantify drug effects on cancer masses, the majority of studies measure the fluorescence intensity of injected cells using stereomicroscopy for imaging and ImageJ software for quantification (Fig. 6).

Fig. 6.

Summary of the parameters and limitations identified for each step of the drug-screening platform. A) schematic representation of the complete workflow for the xenotransplantation of cancer cells into zebrafish embryos; B) most commonly used parameters identified for each step of the protocol; C) limitations associated with each step and potential solutions (indicated by light bulb icons)

This experimental approach is likely the fastest and simplest among those described. Indeed, in vivo cell trackers do not require genetic engineering to induce fluorescence, making it feasible to detect patient-derived primary cells. These samples are often fragile and available in limited quantities, making plasmid transfection particularly challenging. The yolk sac is likely the easiest injection site to access with microneedles due to its size and anatomical location. In contrast, injections into other sites, such as the midbrain or perivitelline space, require greater manual precision due to smaller dimensions and tissue fragility. We consider a broad range of 50–500 cells per injection appropriate to accommodate cell populations of varying sizes. Most studies expose the larvae to treatments for 72 hours, likely because this represents the maximum exposure time before 5 dpf. Overall, this protocol can be applied to the injection of patient-derived primary cultures in a high number of embryos in a relatively short time, an essential feature to test large drug panels. Similar considerations apply to the methods used for quantification. Fluorescence stereomicroscopes are simpler and more practical than confocal microscopes, both in terms of time efficiency and expertise required. Unlike confocal imaging, stereomicroscopy does not require sample fixation, often performed in agarose, and enables rapid, serial image acquisition with sufficient resolution to measure the fluorescence intensity of zebrafish xenografts. Image analysis is relatively quick and can be efficiently performed using commonly available software.

Is it possible to test a high number of drugs and drug combinations using this approach? As previously discussed, the experimental design described in the literature is, in principle, suitable for large-scale drug screening. However, although this was not the primary aim of most of the studies reviewed, only two selected articles tested more than ten treatments. Some excluded studies investigated a higher number of conditions, but did not meet all the criteria for inclusion in the screening analysis. Despite the high fecundity of zebrafish, which ensures the availability of large numbers of embryos, the collected data suggest an experimental limitation in this regard. In particular, managing numerous treatment groups and ensuring consistent and accurate microinjection procedures at scale imposes practical constraints on the number of xenografts that can be handled effectively

To overcome these limitations, the scientific community has devoted considerable effort in recent years to the development of automated approaches aimed at enhancing the potential of zebrafish-based preclinical studies. These innovations specifically target two critical bottlenecks: injection efficiency, and data analysis.

Robotic microinjection platforms have emerged as a transformative solution to address the labor-intensive nature of manual cell injection procedures. These automated systems have been shown to significantly reduce the time required to inject individual embryos while simultaneously increasing the total number of embryos that can be processed compared with manual procedures [126–128]. These improvements in throughput is important not only for expanding the number of experimental conditions and treatments that can be tested within a single study, but also for minimizing the duration that cells spend outside controlled culture conditions. Reducing this exposure time is critical for maintaining cell viability and minimizing stress-induced cellular alterations, which can compromise experimental outcomes. This aspect becomes especially critical when studies involve patient-derived samples, as these primary cells are inherently more fragile and sensitive to environmental perturbations than immortalized cell lines. In this context, minimizing the time spent outside a controlled microenvironment and limiting exposure to mechanical stress during the injection procedure is crucial for improving engraftment success rates and ensuring that the injected cells accurately represent the biological characteristics of the original patient tumor.

Another procedural step that significantly affects the throughput and technical reproducibility of xenograft experiments is the time and skill required to orient zebrafish embryos and larvae into precise positions. For targeting specific anatomical sites such as the brain, perivitelline space, or yolk sac, embryos have to be placed in exact orientations to ensure accurate cell implantation. Manual orientation is not only time-consuming but also introduces variability depending on operator skill and experience. To address this challenge, both commercial companies and academic research groups have developed specialized molds and support systems based on custom-designed wells or microfluidic chips with defined dimensions and shapes that allow embryos to be passively or actively positioned in different orientations with high consistency [126, 127, 129, 130]. These positioning strategies facilitate the injection procedure by reducing the time spent on embryo manipulation and can be seamlessly integrated into automated injection platforms, creating streamlined workflows that minimize manual handling and technical variability across experimental replicates.

Despite significant advances in automated injection methods and embryo positioning systems, time-consuming and subjective data analysis has remained a major bottleneck in the comprehensive assessment of drug efficacy and xenograft outcomes. Manual quantification of experimental endpoints such as tumor growth, metastatic dissemination, and angiogenesis is not only labor-intensive but also prone to inter-observer variability, which can compromise data reproducibility and statistical rigor. To reduce analysis time and improve the objectivity and reproducibility of experimental data, innovative image-based analysis software tools have been developed that advantage computer vision and machine learning algorithms [127, 131]. These computational tools are capable of extracting quantitative information from microscopy images in a significantly shorter time compared with manual operator-based analysis, while applying consistent quantification criteria across all samples. Moreover, these analysis platforms can be directly integrated with automated microinjection and imaging systems, creating end-to-end automated workflows that further reduce the overall time required for experimental execution and data processing [127, 129]. The development of such integrated platforms represents a paradigm shift in zebrafish xenograft methodology, transforming these models from low-throughput proof-of-concept tools into robust, scalable platforms suitable for large-scale preclinical drug screening and personalized medicine applications.

At the end of the literature screening, only eight articles were identified that performed xenotransplantation using patient-derived primary samples. The immediate transplantation of freshly resected primary tissue, without prior in vitro culture, represents the ideal approach. However, primary cultures are difficult to obtain and the maintenance of the number of viable cells recovered after tissue digestion is often low and challenging to handle, especially for microinjection procedures. In many cases, tissue samples have already been exposed to chemotherapy or radiotherapy prior to surgical resection. As a result, the samples may contain necrotic areas and exhibit reduced overall cell viability. Furthermore, tissue digestion protocols can damage delicate cell populations and may leave behind solid clusters or fragments within the suspension. These issues are particularly problematic during microinjection, as they can obstruct the flow of cells through microneedles. These technical challenges likely explain why many otherwise relevant studies did not meet the inclusion criteria, despite their relevance in the scientific community.

Conclusion

This review highlights key limitations in testing drug panels, both in terms of the number and type of therapies. We screened the literature based on globally applicable criteria to ensure data reliability and designed a protocol reflecting the experimental features most commonly used in oncology research. Although the main injection parameters identified do not restrict large-scale drug screening, only a small proportion of studies employed patient-derived cells. As discussed, some features of the injection method and of primary cultures make the xenograft procedure challenging. These aspects may be seen as limitations of the model, potentially hindering its clinical applicability. Nevertheless, co-clinical trials have already been conducted using zebrafish, demonstrating feasibility, albeit with restrictions. Overall, zebrafish xenografts show promise for precision medicine but several aspects still require clarification.

A crucial feature of any drug-screening test for personalized medicine is reproducibility and standardization. While the experimental conditions identified here are relatively simple to perform, they remain operator-dependent, as is typical of most in vivo procedures. To overcome the limitations of manual microinjection protocols, several alternative approaches have been proposed by the zebrafish research community. In recent years, multiple automated and robotic microinjection platforms have been developed to increase the success rate of xenotransplantation and to reduce the time required for experimental procedures. These systems operate not only during the cell injection step but also in data acquisition and analysis, an aspect that is particularly important for standardizing workflows and improving the reproducibility and reliability of in vivo drug-screening platforms designed for personalized medicine.

In conclusion, zebrafish represent an excellent model for drug screening, although important limitations remain. Although challenging, the experimental parameters suggested by the literature analysis appear simple to implement and manageable to automation, making the xenograft of ex vivo cancer samples a virtual clinical approach in the near future.

Abbreviations

YS

Yolk Sac

PVS

Perivitelline Space

Hpf

Hours post fertilizations

Hpi

Hours post injections

Dpf

Days post fertilization

Author contributions

OC and GM designed and conceived the project. OC, LZ, ADV and SV collaborated on data acquisition and analysis. CL and GM supervised the project. OC, LZ and GM drafted the original article.

Funding

This work was partly supported thanks to the contribution of Ricerca Corrente by the Italian Ministry of Health.

Data availability

The datasets used in the current study is available from the corresponding author on reasonable request.

Declarations

Ethics approval and concent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.