Protocol for intracardiac delivery and live imaging of NK-tumor interactions in an ex ovo chick embryo model

Summary

We present a protocol to evaluate natural killer (NK)-tumor interactions using an ex ovo chick embryo system. We describe procedures to culture GFP-labeled SH-SY5Y neuroblastoma cells, expand and Hoechst-label human NK cells, and perform sequential intracardiac injections. We use fluorescence imaging to visualize NK and tumor cell distribution in vivo and live imaging to assess interactions in vitro. This approach provides a rapid and accessible platform to quantify NK cell activity in a living vertebrate model.

Graphical abstract

Highlights

• •Protocol for preclinical evaluation of NK cell cytotoxicity in an ex ovo model
• •Workflow for ex ovo culture and intracardiac co-injection of tumor and NK cells
• •Instructions on NK cell and neuroblastoma cell preparation
• •Quantitative image analysis using Cellpose and Otsu-based segmentation methods

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

We present a protocol to evaluate natural killer (NK)-tumor interactions using an ex ovo chick embryo system. We describe procedures to culture GFP-labeled SH-SY5Y neuroblastoma cells, expand and Hoechst-label human NK cells, and perform sequential intracardiac injections. We use fluorescence imaging to visualize NK and tumor cell distribution in vivo and live imaging to assess interactions in vitro. This approach provides a rapid and accessible platform to quantify NK cell activity in a living vertebrate model.

Before you begin

Investigating NK cell–tumor interactions is central to therapeutic development, yet murine xenograft systems routinely used for this purpose are expensive, slow, technically demanding, and require ethical approval. In line with the 3Rs (Replacement, Reduction, and Refinement), alternative vertebrate models are needed for early-stage functional testing before committing to animal trials. Chick embryos provide an attractive solution because their immune system is permissive to xenografts and experimentation before day 14 occurs before nociception.^1,2,3 The adaptive immune system of the chick embryo matures around embryonic day 18, at which point xenograft rejection begins.⁴ Ex ovo culture—where embryos are transferred from the eggshell onto an open support membrane—further enhances accessibility by allowing direct visualization and repeated intervention under physiologically relevant conditions.

Chick embryo models, including chorioallantoic membrane (CAM) assays, have long been used to investigate tumor growth, angiogenesis, metastasis, and drug responses across diverse cancer types, including neuroblastoma.^5,6 More recently, a CAM-based platform has been used to assess the in vivo efficacy of human CAR T cells against solid tumors, supporting the feasibility of chick embryo systems for preclinical testing of adoptive cellular immunotherapies.⁷ GD2-targeted CAR T cell therapies have demonstrated enhanced cytotoxic potency against neuroblastoma in clinical settings.⁸ Tumor xenografts and T cell–based therapies have been established in CAM model, but dynamic visualization of NK cell effector function—such as NK cell persistence, localization, and cytotoxic interaction with tumor targets—has not been demonstrated. This protocol addresses that gap by establishing a reproducible workflow for visualizing GD2-targeted NK cell-mediated tumor interaction in living embryos (ex ovo method) and complementing it with high-resolution in vitro assays.

Here, we describe the culture of cryopreserved SH-SY5Y neuroblastoma cells and human NK cells, incubation of fertilized eggs and transfer of embryos to an ex ovo culture system, intracardiac xenografting of tumor cells, and subsequent administration of NK cells. We outline fluorescence imaging steps for monitoring tumor persistence and NK localization over time, as well as a complementary co-culture assay for direct visualization of effector–target engagement. Guidance for basic quantification and comparison across embryos is provided.

Before beginning, investigators should obtain fertilized eggs, validated cell stocks, and the specialized materials listed in the key resources table. Proper preparation of culture media, imaging dyes, and sterile injection tools is essential for the successful execution of the protocol. Although this workflow is optimized for SH-SY5Y neuroblastoma cells and human NK cells, it can be adapted to other tumor models or immune effector populations that are permissive in the chick embryo system.

Innovation

This protocol introduces methodological advances that extend the utility of the chick embryo system for studying human immune–tumor interactions. First, we refine ex ovo embryo culture to provide unrestricted optical access and repeated vascular entry, enabling reliable sequential intracardiac delivery of GFP-labelled neuroblastoma cells followed by Hoechst-labeled human NK cells. This novel configuration improves visualization and procedural flexibility relative to conventional in ovo chorioallantoic membrane approaches.

Second, we establish conditions that support persistence and detection of human NK cells in the embryo by activating NK cells in vitro with human IL-15 before injecting them to ex ovo system. This modification enhances NK cell viability in ex ovo and overcomes a key limitation that has previously hindered adoptive immune cell studies in avian embryo models.

Third, we integrate whole-embryo fluorescence imaging with a streamlined live co-culture assay to capture NK–tumor engagement under defined conditions. This pairing enables cross-validation of effector–target interactions observed in vivo without requiring intravital microscopy or specialized imaging infrastructure.

Finally, we introduce optional automated analysis tools incorporating Cellpose-assisted segmentation, classical image processing (including Otsu thresholding), and Python-based proximity quantification.⁹ These computational components improve object detection, standardize readouts, and facilitate reproducible assessment of NK–tumor dynamics across experiments.

Together, these innovations deliver a versatile and ethically favorable workflow for rapid functional evaluation of NK cell behavior and other immune effector responses in a living vertebrate embryo, while providing complementary in vitro readouts and automated image-analysis pipelines.

Institutional permissions

This study was conducted under the ethical approval of the HUS Regional Committee on Medical Research Ethics (HUS/6934/2025) for the use of human-derived materials. All human samples are processed in accordance with applicable national legislation and institutional guidelines. Primary human NK cells were isolated from buffy coats obtained from blood donations at the Finnish Red Cross Blood Service under institutional permit FRCBS 178/6/2023. Ex ovo chick embryo experiments are performed before embryonic day 14, before the onset of nociception, and therefore do not require separate ethical approval under national regulations.

All laboratory procedures are carried out in compliance with institutional guidelines and the TENK Responsible Conduct of Research framework. Researchers seeking to implement this protocol must obtain all necessary permissions and approvals required by their own institutions and national regulatory bodies.

Preparation of SH-SY5Y neuroblastoma cell culture medium

Timing: 1 h

• 1.
  Prepare SH-Sy5y cell culture medium.

  • a.
    Add 22 mL EMEM medium and 22 mL F12 medium to a sterile 50 mL Falcon tube.
  • b.
    Add 5 mL of inactivated Fetal Bovine Serum, 0.5 mL of Penicillin-Streptomycin (10,000 U/ml) and 0.5 mL of L-Glutamine (200 mM).
  • c.
    Mix well. Total volume is 50 mL.
• 2.SH-Sy5y cell culture medium must be stored at 2°C–8°C and used within 2 weeks.

Preparation of NK cell MACS expansion medium

Timing: 1 h

• 3.
  Prepare NK cell MACS Expansion medium.

  • a.
    Add 46.158 mL of NK MACS Basal medium to a sterile 50 mL Falcon tube.
  • b.
    Add 2.5 mL of Human AB Serum, 0.5 mL of Penicillin-Streptomycin (10,000 U/mL) and 0.5 mL of MACS Supplement.
  • c.
    Add 167 μL Human IL-2 (stock 150,000 IU/mL) to 500 IU/ml final concentration.
  • d.
    Add 175 μL Human IL-15 (stock 40,000 IU/mL) to 140 IU/ml final concentration.
  • e.
    Mix well. Total volume is 50 mL.
• 4.NK cell MACS Expansion medium must be stored at 2°C–8°C and used within 2 weeks.

Note: Human IL-2 and IL-15 cytokine stock concentrations are calculated based on each specific lot’s concentration, which varies from lot to lot.

Preparation of fast green

Timing: 5 min

• 5.Add 1 mL PBS to a tube.
• 6.Weigh 1 mg of Fast Green.
• 7.Add the Fast Green with PBS and mix carefully. The final concentration of the solution is 1 mg/mL.

Note: Use the Fast Green solution to stain the cell suspension before intracardiac injection into the chick embryo for better visualization of the injections.

Preparation of the injection pipette

Timing: 1 h

In this section, a glass injection pipette for intracardiac microinjections is prepared (Figure 1).

• 8.Use glass pipette tip(s), with ∼1 μm tip diameter.

Note: These can be made by stretching the pipette with, e.g. automated micropipette puller.

• 9.Color the sharp and thin tip of the pipette with a permanent marker to improve visual guidance during injection and enable precise positioning of the pipette tip within the target vessel (Figure 1A).
• 10.Cut the tip shorter with a diamond glass pen (Figure 1B).

Note: The tip should be wide enough for the cells to travel through.

• 11.Using a gas burner, slightly polish the opposite side of the pipette to prevent damaging of the pipette holder.
• 12.Connect the glass pipette with the manual microinjection needle set composed of a 5 mL syringe, tubing and connector (Figures 1C and 1D). Test the tip by aspirating PBS.
• 13.Prior to using the pipette, wash it by aspirating sterile PBS in and out slowly several times.
• 14.Make volume level markings into the glass pipette tip e.g., 1 μL, 2 μL and 3 μL volumes by aspirating e.g., 1 μL of PBS and marking the tip at appropriate locations.

CRITICAL: During intracardiac cell injections, this will help in monitoring the volume of cells injected, e.g., to inject 3 μL, one needs to aspirate 4 μL and make sure the last 1 μL stays inside the pipette tip (to avoid air bubbles going in).

Figure 1.

Preparation of the injection pipette

(A) A glass pipette tip (∼1 μm diameter) is colored at the sharp end with a permanent marker for visibility during injection.

(B) The tip is cut using a diamond glass pen to achieve an optimal diameter of the tip (wide enough for cell passage).

(C) The wide end of the glass pipette is slightly polished by the gas flame to reduce the damage to the silicone sleeve of the pipette holder.

(D) The prepared pipette is connected to a manual microinjection set consisting of a 5 mL syringe, tubing, and a pipette holder.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Human peripheral Natural Killer cells	N/A	N/A
Chemicals, peptides, and recombinant proteins
EMEM	ATCC	30–2003
Ham’s F12 Nutrient Mix	Gibco	11765–054
Fetal Bovine Serum	Gibco	10082–147
Penicillin-Streptomycin (10,000 U/mL)	Gibco	15140–122
L-Glutamine (200 mM)	Gibco	25030–081
NK MACS Basal Medium	Miltenyi Biotec	130-112-968
MACS Supplement (100×)	Miltenyi Biotec	130-113-102
Human AB Serum	Sigma	H5667
Human IL-2, premium quality	Miltenyi Biotec	130-097-746
Human IL-15, premium quality	Miltenyi Biotec	130-095-765
Trypan Blue	Gibco	15250061
CTS™ TrypLE Select Enzyme	Gibco	A1285901
CTS™ DPBS, without calcium chloride, without magnesium chloride	Gibco	A12856-01
PBS, 10×	Corning	20-031-CV
Fast Green FCF, powder	Sigma-Aldrich	F7258-25G
Hoechst 33342 (20 nM)	Thermo Fisher Scientific	62249
Brilliant Green 1% alcohol, 25 ml	SamaraMedProm	EAN 4603933001418
Virkon Rely+ON desinfiointi tablets 50×5g	Lanxess	4470676
Ultrapure water	N/A	N/A
Deposited data
https://doi.org/10.5281/zenodo.19111461 (GitHub code)	N/A	N/A
Experimental models: Cell lines
Neuroblastoma cell line SH-Sy5y-eGFPluc	ATCC and iCell, Blood Service	Ref.10
Experimental models: Organisms/strains
Fertilized chicken eggs	Haaviston Siitoskanala, Panelia, Finland	N/A
Software and algorithms
Zen Blue 3.3 software	ZEISS	N/A
Zen Lite 3.13 software	ZEISS	N/A
MATLAB R2023b (23.2.0.2859533)	MathWorks	N/A
Python 3.13.0	Python Software Foundation	https://www.python.org
Cellpose (2.2.0)	HHMI Janelia Research Campus, USA	https://www.cellpose.org
Other
TC-20 Cell Counter	BioRad	N/A
TC-20 Counting Slides	BioRad	1450016
100 mm dish, Nunc™ Cell Culture/Petri Dishes	Thermo Fisher Scientific	150350
6 well, Nunc™ Non-Treated Multidishes	Thermo Fisher Scientific	150239
μ-Slide 8 Well ibiTreat	Ibidi	80826
Egg incubator, Comact S 84	Grumbach	N/A
Automatic Eggturner, AT-42 (universal eggtrays), EU 220-230VAC (adapter), 12VAC/300mA/50-60 Hz (input).	Campo24	ASIN B004163DTM
Cell incubator	Sanyo	10070347
Olympus SZX12 -microscope	Olympus	N/A
Olympus U-RFL-T fluorescence light source	Olympus	N/A
Intralux 4000-1 light source	Intralux	N/A
Olympus Tokai Hit thermal plate	Olympus	N/A
Olympus OM-D E-M1II camera with four thirds adapter MMF-1 objective	Olympus	N/A
ZEISS, AXIO ZOOM.V16 microscope	ZEISS	N/A
PlanNeoFluar Z 1.0× objective	ZEISS	N/A
PlanNeoFluar Z 1.0× light source	ZEISS	N/A
0.63× Camera Adapter	ZEISS	N/A
ZEISS Axiocam 305 color camera	ZEISS	N/A
ZEISS Axio Observer.Z1 microscope	ZEISS	N/A
Plan-Apochromat 20×/0.8 objective	ZEISS	N/A
HXP 120 C light source	ZEISS	N/A
Hamamatsu sCMOS imaging device	Hamamatsu Photonics	N/A
TempModule S	ZEISS	N/A
CO2 Module S	ZEISS	N/A
XLmulti S1 incubator	ZEISS	N/A
Deep glass bowl	N/A	N/A
Parafilm ® M	Sigma-Aldrich	P7793
15cm petri dishes	Corning	430597
Food-grade plastic wrapping	Fredman	6727
Pasteur pipette, Transfer pipette 3,5 mL	Sarstedt	86.1171
Scalpel	Swann-Morton	0506
27G needle, HSW Fine-Ject 27GX3/4″ Nro 20, 0,4×10mm RW/LB	Henke-Sass, Wolf GmbH	4710004020
Glass Pasteur pipette	Fischer Scientific	15202699
Microcapillary puller, Model P-97, flaming/brown micropipette puller	Sutter Instrument Co	N/A
Permanent marker, AccuLiner	MonAmi	14801
Diamond glass pen	Glaswarenfabrik Karl Hecht	41182
Gas burner, small	Biltema	17–680
Gas beholder (90% butane, 10% propane)	Biltema	17–975
Tokai Hit Thermo Plate	Olympus	N/A
50 mL falcon tubes	Corning	430829
15 mL falcon tubes	Fischer Scientific	339657
1.5 mL Eppendorf tubes	Fischer Scientific	FB74031

Materials and equipment

SH-Sy5y cell culture medium

REAGENT	Final concentration	Amount
EMEM	N/A	22 mL
F12	N/A	22 mL
Fetal Bovine Serum	10%	5 mL
Penicillin-Streptomycin (10,000 U/ml)	100 IU/mL	0.5 mL
L-Glutamine (200 mM)	2 mM	0.5 mL
Total	N/A	50 mL

Store at 2-8 °C and use within 2 weeks.

NK cell MACS Expansion medium

REAGENT	Final concentration	Amount
NK MACS Basal Medium	N/A	46.158 mL
MACS Supplement (100×)	1×	0.5 mL
Human AB Serum	5%	2.5 mL
Penicillin-Streptomycin (10,000 U/mL)	100 U/mL	0.5 mL
Human IL-2 (150,000 IU/mL)	500 IU/mL	167 μL
Human IL-15 (40,000 IU/mL)	140 IU/mL	175 μl
Total	N/A	50 mL

Store at 2–8 °C and use within 2 weeks.

Step-by-step method details

Day −5–3: Neuroblastoma cell culture

Timing: 1 h (per day)

In this step, the neuroblastoma SH-Sy5y GFP-positive cells are thawed and cultured to ensure they are in optimal condition for xenografting in chick embryos on day 3.

Note: Culturing should start around 1 week before (e.g. Day −5) the cells are xenografted on day 3.

• 1.Warm SH-Sy5y cell culture medium in the water bath to 37°C.
• 2.
  Thaw vial(s) of neuroblastoma cells expressing enhanced green fluorescent protein (GFP) e.g., SH-Sy5y-eGFPluc.

  • a.
    Add 9 mL warm SH-Sy5y cell culture medium to a 15 mL Falcon tube.
  • b.
    Thaw the cell vial(s) in a 37 °C water bath until only a small block of ice remains.
  • c.
    Gently transfer cells from the vial to warm medium in a Falcon tube.
  • d.
    Centrifuge cells 300 × g for 5 min at RT.
  • e.
    Discard supernatant and re-suspend cells in 1 mL of SH-Sy5y cell culture medium.
• 3.
  Count cells with BioRad TC-20 cell counter.

  • a.
    Suspend cells.
  • b.
    Take 20 μL cell sample(s) to 1.5 mL eppendorf tube(s).
  • c.
    Add 20 μL Trypan Blue and mix.
  • d.
    Count cells using BioRad TC-20 cell counter.

CRITICAL: Trypan Blue is harmful and potentially carcinogenic. Avoid ingestion, inhalation, and skin/eye contact. Use appropriate personal protective equipment (PPE).

• 4.
  Plate thawed cells in 2×10⁴ cells/mL recovery density on surface-treated dishes.

  • a.
    Add SH-Sy5y cell culture medium to cells so that the final density is 2×10⁴ cells/mL.
  • b.
    Plate 10 mL of cells on 100 mm cell culture surface treated dish e.g., Nunc™ Cell Culture/Petri Dishes (Thermo Fisher Scientific).
  • c.
    Place cells in incubator at 37 °C and 5% CO₂.
• 5.
  Subculture cells every 2 to 3 days, or when confluency reaches 80%.

  Note: SH-Sy5y cells are adherent cells and require detachment via trypsinization before they can be sub-cultured.

  • a.
    Aspirate old media. Wash cells by adding 2 mL DPBS. Aspirate DPBS.
  • b.
    To detach cells, add 1 mL of TrypLE per 100 mm dish. Incubate 5 min at 37 °C.
  • c.
    Add 2 mL of media to stop TrypLE. Split e.g., 1:3 to multiple dishes.

    Note: Optionally, count cells as previously. Add medium to a subculture density of 1 × 10⁴ cells/mL. Plate 10 mL cells per 100 mm dish.

  • d.
    Place cells in incubator at 37 °C and 5% CO₂.

Day 0: Incubation of fertilized eggs

Timing: 15 min

In this step, fertilized chicken eggs are placed in an incubator to begin their development.

CRITICAL: Check the eggshell for possible cracks during transportation. Try to use eggs freshly as storage decreases viability. If egg incubation doesn’t start upon receiving the eggs, you may keep the eggs at 13–18°C (55–65 °F). The viability of eggs decreases over time; use them preferably within 7–14 days, keep them in moderate humidity and turn them periodically during storage.

• 6.Set the egg incubator at 38°C with 60%–65% humidity.
• 7.Place eggs in an upright position on the egg turner into an egg incubator.
• 8.Connect the egg turner so that it starts rotating.
• 9.Incubate eggs until day 3.

Note: Because embryo development can vary between eggs, some embryos may not reach the appropriate developmental stage for intracardiac injection. Based on our experience, we recommend starting with approximately twice the number of fertilized eggs required for the experiment to compensate for unfertilized eggs, contamination, or embryos that do not reach the desired developmental stage.

Day 2: Preparation of waterbeds

Timing: 5 min (per waterbed)

In this step, waterbeds are prepared for the ex ovo embryos using materials such as deep glass bowls, ultrapure water, sterile food-grade plastic wrapping, and 15 cm petri dishes (Figure 2).

• 10.Take a clean, deep glass bowl and put it on a 15 cm petri dish (Figure 2A).
• 11.Pour ultrapure water into the bowl so that it fills from 1/2 to 2/3 of the bowl (Figure 2B).

Note: Too much water will overflow, too little will not make enough soft surface for the egg.

• 12.Take clean food-grade plastic wrapping and place it on top of the bowl so that it is fully touching the water, making a waterbed for the embryo (Figure 2C).
• 13.Finally, place the petri dish lid on top of the bowl.
• 14.Place the ready-made waterbeds to warm up in the incubator at 37°C.

Note: If needed, the waterbeds can be prepared on day 3 and used when they are ready, but it is better to use them prewarmed.

Figure 2.

Preparation of waterbeds

(A) Schematic representation of the waterbed setup used to support chick embryos ex ovo. A glass bowl is filled with ultrapure water and sealed with food-grade plastic wrap to create a suspended membrane surface. The wrapped bowl is positioned inside a Petri dish, which provides stability and a sterile handling platform; a Petri dish lid can be placed over the system to reduce evaporation.

(B) Photograph showing an appropriate water fill level within the glass bowl, illustrating the required convex water profile for membrane support.

(C) Photograph showing correct placement of the plastic wrap fully contacting the water surface, forming a smooth, tensioned support layer suitable for embryo transfer. This configuration maintains hydration, mechanical stability, and optical access for embryo handling, culturing, and imaging.

Day 3 (morning): Setting up ex ovo culture

Timing: 10 min (per embryo)

In this step, fertilized eggs are opened and the embryos are transferred from the eggshell onto the waterbeds (Figure 3, Methods Video S1). If any embryos need to be discarded, see Day 5: Disposal of embryos (step 38–40).

• 15.
  Clean the eggs and mark the embryo’s position.

  • a.
    Disconnect the egg turner.
  • b.
    Take the eggs out of the egg incubator (Figure 3A).
  • c.
    Wipe the eggs clean using tissue paper dipped in antiseptic liquid, e.g., Brilliant Green (Figures 3B and 3C).

    CRITICAL: Avoid using 70% Ethanol for antiseptic treatment of the eggs. It will affect the development of chicken embryos.

    Note: Use of Brilliant Green will also give contrast to eggshell, revealing subtle cracks. Cracked eggs can be discarded.

  • d.
    Turn the eggs sideways back into the incubator and mark the top highest point of the egg’s side with “X” using a pencil (Figure 3D).

    CRITICAL: Do not turn on the egg turner. Eggs need to be still.

  • e.
    Incubate for 15–30 min to allow the embryos to orient to the highest point (marked with X).
  • f.
    After the embryo has settled, take them out of the incubator.

    CRITICAL: Ensure that the X marking remains on top.

• 16.
  Opening fertilized eggs and transferring the embryos onto waterbeds for ex ovo culture.

  • a.
    Carefully yet firmly, place your palm on top of the egg so that the X is covered, and lift the egg (Figure 3E). Do not rotate the egg in any way.

    Note: Cover as much surface area of the egg as possible. Your palm supports the eggshell and helps prevent uncontrolled cracking during shell opening.

  • b.
    Take a disposable scalpel in your other hand.

    • i.
      Keep the scalpel blade between your fingers so, that only 2 to 3mm of the blade’s tip is visible (Figure 3F).
    • ii.
      With the scalpel tip, drill a small hole to the bottom of the egg – opposite side of the X, which marks the position of the embryo (Figure 3G).
    • iii.
      Using the scalpel, saw a small cut of 2 to 3mm through the shell. The cut should be oriented circumferentially so that it will direct the eggshell’s cracking around the egg.

      Note: A little albumin from the egg might start leaking out.

  • c.
    To crack the eggshell, place the scalpel into the cut and twist the blade. This will cause the eggshell to crack circularly around the egg in the cut’s direction.
    Take the egg very close to the waterbed and gently crack the egg open, letting the embryo fall on the waterbed plastic (Figure 3H).

    CRITICAL: The eggshell halves must separate widely to prevent the yolk sac from being squeezed or damaged by sharp shell edges. To ensure this, gently widen the circular crack with the scalpel and slowly open the shell so the embryo can slide onto the waterbed without compression.

    CRITICAL: Check the viability and positioning of the embryos. The embryo should be positioned at the top for future experimental procedures.

  • d.
    Remove air bubbles surrounding the embryo using a Pasteur pipette as much as possible without disturbing the yolk. Do not touch the yolk or the embryo.
  • e.
    Cover the waterbed with the lid and place the embryos in the incubator at 37°C and 5% CO₂ (Figure 3I).

    CRITICAL: Embryos are injected and imaged on their waterbed. Do not remove embryos from the water beds.

Figure 3.

Setting up ex ovo culture

(A) Remove eggs from the incubator and prepare them for egg opening.

(B) Dip tissue paper in antiseptic solution (e.g., Brilliant Green).

(C) Wipe eggs clean with an antiseptic solution.

(D) Place eggs sideways in the incubator and mark at the highest point with an “X” to indicate embryo position. The egg turner is disconnected to keep the eggs still.

(E) Hold the egg with your whole hand for optimal support during egg cracking. Keep the X marking against your palm while working. See Methods Video S1.

(F) Take a disposable scalpel and keep it between your fingers so that 2-3 mm of the tip is visible.

(G) Using the disposable scalpel, drill a small hole and make a 2–3 mm cut on the opposite side of the “X” to initiate cracking.

(H) The eggshell is cracked circularly, and the embryo is gently released onto the waterbed without damaging the yolk or embryo. Check embryos for viability and correct positioning. In the correct position, the embryo is in the middle of the yolk, facing up. Some bubbles in the albumin are visible.

(I) Remove air bubbles with a Pasteur pipette, cover waterbeds and placed in an incubator at 37 °C and 5% CO₂.

Day 3 (afternoon): Initiate NK cell cultures

Timing: 1 h

In this step, NK cells are thawed and cultured for 24 h as human primary NK cells should be cultured for a minimum of one day before use in assays to prevent loss of viability immediately after thawing.

Note: In this particular experimental setup, NK cells have been transduced with lentiviral vectors to express anti-GD2 CAR receptor or with an empty lentiviral vector (mock). After this, the transduced NK cells have been expanded with K562-mbIL21-CD48-41BBL feeder cells in NK cell Expansion medium for a total of 13 days. Then they were cryo-preserved and used in this ex ovo experiment. Anti-GD2 CAR and mock NK cell treatments have been chosen to address the research-specific questions about the functionality of anti-GD2 CARs. Use of this assay is not restricted to these transduced NK cells; NK cell treatments can be adapted according to individual experimental requirements.

• 17.Warm NK cell MACS Expansion medium in the water bath to 37 °C.
• 18.
  Thaw vial(s) of NK cells (10×10⁶ cells in 50×10⁶ cells/ml density).

  • a.
    Add 2 mL warm NK cell MACS Expansion medium to a 15 mL Falcon tube.
  • b.
    Thaw the cell vial(s) in a 37°C water bath until only a small block of ice remains.

    Note: Alternatively, thaw vial(s) by warming them up in hand. This is best for vial(s) with a small cell volume.

  • c.
    Gently transfer cells from vial(s) to warm medium in a Falcon tube.

    Note: NK cells are sensitive to freeze-thaw cycles. To avoid stressing them, it is preferable not to centrifuge the NK cells, and they should be frozen at a higher density whenever possible.

• 19.
  Count cells with BioRad TC-20 cell counter.

  • a.
    Suspend cells.
  • b.
    Take 20 μL cell sample(s) to 1.5 mL Eppendorf tube(s).
  • c.
    Add 20 μL Trypan Blue and mix.
  • d.
    Count cells using BioRad TC-20 cell counter.

CRITICAL: Trypan Blue is harmful and potentially carcinogenic. Avoid ingestion, inhalation, and skin/eye contact. Use appropriate PPE.

• 20.
  Plate thawed cells in 2×10⁶ to 4×10⁶ cells/mL recovery density on non-surface-treated dishes.

  • a.
    Add NK cell MACS Expansion medium so that the final density is a minimum of 2×10⁶ cells/mL.
  • b.
    Plate 3–6 mL of cell per non-treated 6-well cell culture plate, e.g., Nunc™ non-treated dish (Thermo Fisher Scientific).
  • c.
    Place cells for 24 h in the incubator at 37°C and 5% CO₂.

Note: If cultured longer than 1 day, count NK cells every 2 to 3 days and add fresh media to 1 to 2×10⁶ cells/mL subculture density.

Day 3 (evening): Xenografting via intracardiac injection

Timing: 10 min (per embryo)

In this step, embryos are injected with GFP-positive neuroblastoma SH-Sy5y cells using intracardiac injections (Figure 4, Methods Video S2). If any embryos need to be discarded, see Day 5: Disposal of embryos (step 38–40).

• 21.Inspect the embryos and discard dead and possibly contaminated embryos.

CRITICAL: Optimally intracardiac injection is performed at Hamburger-Hamilton (HH) stage 15-18, which roughly corresponds to 72 – 96 h of incubation. To ensure successful intracardiac injection, inspect the embryo development. Injections can begin once the heart has developed sufficiently. Embryos develop quickly; an undeveloped embryo in the morning can be ready by evening.

• 22.
  Divide embryos into randomized groups to be injected with or without neuroblastoma cells. Example cell division could be:

  • a.
    No neuroblastoma cells -group, 5 embryos.
  • b.
    Neuroblastoma cells only -group, 5 embryos.
  • c.
    Neuroblastoma cells with mock NK cells -group, 5 embryos.
  • d.
    Neuroblastoma cells with anti-GD2 CAR NK cells -group, 5 embryos.
• 23.
  Prepare neuroblastoma cells.

  • a.
    Aspirate old media. Wash cells by adding 2 mL DPBS. Aspirate DPBS.
  • b.
    To detach cells, add 1 mL of TrypLE per 100 mm dish.
  • c.
    Incubate 5 min at 37°C.
  • d.
    Collect cells into a Falcon tube with 9 mL of PBS.
  • e.
    Centrifuge cells 300 × g for 5 min at RT.
  • f.
    Discard supernatant and resuspend pellet in 1 mL of PBS.
  • g.
    Suspend cells.
  • h.
    Take 20 μL cell sample(s) to 1.5 mL Eppendorf tube(s).
  • i.
    Add 20 μL Trypan Blue and mix.
  • j.
    Count cells using BioRad TC-20 cell counter.
  • k.
    Add PBS to 25,000 cells/3 μL injection density.

CRITICAL: Trypan Blue is harmful and potentially carcinogenic. Avoid ingestion, inhalation, and skin/eye contact. Use appropriate PPE.

• 24.
  Prepare the embryos one embryo at a time.

  • a.
    Take the embryos out under the upright microscope (Figure 4A).
  • b.
    Pipette a little bit of PBS on top of the embryo.

    CRITICAL: Handle embryos individually, as they are sensitive to temperature changes. Monitor the embryo's heartbeat closely, and if you notice any decline in its vitals, return the embryo to the incubator immediately. To maintain warmth, use a thermal plate.

  • c.
    Using a 27G needle, break the extra embryonic membrane surrounding the embryo.

    Note: To avoid hitting the embryo with the needle, try to catch the membrane on the tip of the needle and pull upwards to puncture the membrane.

• 25.
  Perform intracardiac injections on the embryo.

  • a.
    Take a parchment of Parafilm or clean petri dish on the table or under a microscope.
  • b.
    Take the glass pipette microinjection set and wash it by aspirating PBS in and out on parchment of Parafilm/dish (Figure 4B).
  • c.
    On top of the Parafilm/dish, combine 3 μL of neuroblastoma cells with 0.5 μL of Fast Green and mix with the glass pipette microinjection set.

    Note: This will color the cell suspension, helping to visualize the injection success better.

  • d.
    Carefully and slowly inject 3 μL of the cell suspension into the thickest part of the “heart” (Figures 4C–4F).

    Note: Use the markings in the tip to ensure that the last 0.5 μL is not injected into the embryo to avoid injection of the air accidentally. Some of the cell suspension and blood may leak during injection. These embryos should be discarded as this leads to inaccuracy in downstream analysis.

  • e.
    Wash the glass pipette tip by aspirating PBS several times after every injection.
• 26.After injection, cover the embryos’ waterbed with the lid and place them back in the incubator at 37°C with 5% CO₂.
• 27.Proceed with the next embryo.

Note: The success of the injection can be confirmed by using a fluorescent microscope with GFP settings to detect neuroblastoma.

Figure 4.

Perform intracardiac injections on the embryo

(A) Position the embryo under an upright microscope and add a small amount of PBS to keep it hydrated and warm.

(B) Wash the glass pipette microinjection set by aspirating PBS on a Parafilm sheet under the microscope.

(C) Insert the pipette into the beating cardiac coil and inject the cell suspension slowly.

(D) Monitor injection stages. Start with a full injection volume visible in the pipette. See also Methods Video S2.

(E) Inject slowly until Fast Green dye appears in major blood vessels.

(F) Complete the injection when the dye spreads widely through the bloodstream. Do not inject all cells to avoid air bubbles going into blood stream.

Methods Video S3. In vitro fluorescence imaging of GFP+ neuroblastoma cells, related to step 46b

The field of view for the video is 17.37 mm × 16.68 mm.

Day 4: Intracardiac injection of NK cells

Timing: 10 min (per embryo)

In this step, NK cells are stained with Hoechst and resuspended in PBS supplemented with IL-15 (280 IU/mL) immediately before injection. IL-15 is not used as a prolonged pre-activation step but is included in the injection medium to support NK cell survival after intracardiac delivery, with a single administration sufficient to maintain NK cell persistence for up to 72 h (Figure 7, Methods Video S2). If any embryos need to be discarded, see Day 5: Disposal of embryos (step 38–40).

• 28.Inspect embryos from the previous day (Day 3) and discard dead and any contaminated embryos.
• 29.
  Prepare NK cells.

  • a.
    Collect cells to a Falcon tube and fill up with PBS.
  • b.
    Centrifuge cells 300 × g for 5 min at RT.
  • c.
    Discard supernatant and resuspend pellet in 1 mL of PBS.
• 30.
  Stain NK cells.

  • a.
    Add 2 μL of Hoechst per 1 mL of NK cells.
  • b.
    Incubate for 10 min at 37°C.
  • c.
    Wash the stained NK cells four times by centrifugation at 300 × g for 5 min at RT, discarding the supernatant and resuspending the cell pellet with PBS.
• 31.
  Activate NK cells with IL-15.

  • a.
    Pipette 7 μL of human IL-15 (40,000 IU/mL stock) per 1 mL of PBS to create 280 IU/mL IL-15 in PBS solution.
  • b.
    Re-suspend NK cells in 280 IU/mL IL-15 in PBS.

    Note: This step will boost NK cell survival ex ovo.

  • c.
    Suspend cells.
  • d.
    Take 20 μL cell sample(s) to 1.5 mL Eppendorf tube(s).
  • e.
    Add 20 μL Trypan Blue and mix.
  • f.
    Count cells using BioRad TC-20 cell counter.

    CRITICAL: Trypan Blue is harmful and potentially carcinogenic. Avoid ingestion, inhalation, and skin/eye contact. Use appropriate PPE.

  • g.
    Adjust to 100,000 cells/3 μL density with 280 IU/ml IL-15 in PBS.
• 32.
  Prepare the embryos one embryo at a time.

  • a.
    Take the embryos out under the upright microscope (Figure 4A).
  • b.
    Pipette little bit of PBS on top of the embryo.

    CRITICAL: Work with embryos one at a time. Embryos are sensitive to temperature changes, so keep them outside the incubator only briefly. Monitor the embryo’s heart rate under the microscope. If the heart rate visibly slows or becomes irregular, immediately return the embryo to the incubator to recover. Maintain warmth during handling using a thermal plate. Embryos typically recover normal cardiac activity within 2–5 min after being returned to the incubator.

  • c.
    Using a 27G needle, break the extra embryonic membrane surrounding the embryo.

    Note: To avoid hitting the embryo with the needle, try to catch the membrane on the tip of the needle and pull upwards to puncture the membrane.

• 33.
  Perform intracardiac injections on the embryo.

  • a.
    Take a parchment of Parafilm or a clean petri dish to the table or under a microscope.
  • b.
    Take the glass pipette microinjection set and wash it by aspirating PBS in and out on parchment of Parafilm/dish (Figure 4B).
  • c.
    On top of the Parafilm/dish, combine 3 μL of NK cells with 0.5 μL of Fast Green and mix with the glass pipette microinjection set.

    Note: This will color the cell suspension, making it easier to visualize the success of the injection.

  • d.
    Carefully and slowly inject 3 μL of the cell suspension into the thickest part of the “heart” (Figures 4C–4F).

    Note: Use the markings in the tip to ensure, that last 0.5 μL is not injected into the embryo to avoid injection of the air accidentally. Some of the cell suspension and blood may leak during injection. These embryos should be discarded as this leads to inaccuracy in downstream analysis.

  • e.
    Wash the glass pipette tip by aspirating PBS several times after every injection. Change the glass pipette tip between mock and anti-GD2 CAR NK cell injections.
• 34.After injection, cover the embryos waterbed with lid and place back in the incubator at 37 °C with 5% CO₂.
• 35.Proceed with the next embryo.

Note: The injection success can be confirmed under fluorescent stereo microscope using GFP settings to detect neuroblastoma and DAPI settings to detect NK cells.

Figure 7.

Automated quantification and persistence of intracardially injected NK cells in the chick embryo model

Quantification of NK cell persistence across 24, 48, and 72 hours following intracardiac injection of NK cells either supplemented with 140 IU IL-15 or injected without IL-15.

Each line represents an individual embryo plotted on a logarithmic scale, demonstrating embryo-to-embryo variation but consistently greater NK cell persistence when IL-15 is administered. n=4 each group, each line representing an individual sample.

Day 4 and 5: Fluorescent microscopy imaging ex ovo

Timing: 30 min (per embryo)

In this step, the success of SH-Sy5y GFP+ neuroblastoma and NK cell injections is visualized and imaged using an Axio Zoom.V16 fluorescent stereo microscope, PlanNeoFluar Z 1.0× objective with excitation at 450–490 and emission at 500–550 for GFP and excitation at 335–383 and emission at 420–477 wavelengths for DAPI settings, respectively. Images are taken with an Axiocam 305 camera and processed with ZEN Blue software (see Table S1).

• 36.Take embryos one by one to the fluorescent stereo microscope.

Note: To maintain warmth for the embryo, place a thermal plate beneath it during upright microscopy.

• 37.Detect the neuroblastoma and NK cells in ex ovo with a fluorescent stereo microscope using GFP settings to detect neuroblastoma cells and DAPI settings to detect NK cells.

Note: Investigate the interactions between neuroblastoma and NK cells in the yolk vasculature, the developing chorioallantoic membrane (CAM), and the embryo. Capture images around the embryo. For quantitative downstream analysis, image cells in a standardized way using e.g. same objective and zoom for every image and embryo. Additionally, capture images from predefined, standardized anatomical locations relative to the embryo e.g. observe the area of the yolk vasculature to the left of the embryo.

Day 5: Disposal of embryos

Timing: 24 h

Ex ovo culture is sensitive to contamination, therefore, contaminated or non-viable embryos should be removed promptly to prevent microbial spread. Cultures should be inspected daily. Disposal of contaminated embryos and termination at the experimental endpoint are performed using the same procedure.

CRITICAL: Follow institutional biosafety and ethical guidelines for the handling and disposal of early-stage chick embryos.

• 38.Remove embryos from the incubator.

Note: If embryos remain viable at the experimental endpoint (typically embryonic day 5–6), incubation is discontinued, and embryos are allowed to die following removal from the controlled incubation environment. At this developmental stage, chick embryos are generally considered pre-nociceptive in many research guidelines.

• 39.Apply 1–2% Virkon™ solution to the culture and allow it to act for at least 15 min to 18 hours to ensure decontamination.
• 40.Dispose of embryos and associated materials according to institutional biohazard waste procedures.

Day 6: In vitro imaging

Timing: 6–24 h

In this additional optional step, interactions between SH-SY5Y GFP+ neuroblastoma cells and NK cells are visualized in vitro as a complementary approach to support ex ovo fluorescence microscopy data (Methods Videos S4 and S5). We acquired images using an Axio Observer Z1 spinning disc confocal microscope, Plan-Apochromat 20×/0.8 objective with excitation at 450–490 and emission at 500–550 for GFP and excitation at 335–383 and emission at 420–477 wavelengths for DAPI settings, respectively. Images are taken with a Hamamatsu sCMOS and processed with ZEN Blue software (see Table S2).

• 41.
  Collect neuroblastoma cells.

  • a.
    Aspirate old media. Wash cells by adding 2 mL DPBS. Aspirate DPBS.
  • b.
    To detach cells, add 1 mL of TrypLE per 100 mm dish.
  • c.
    Incubate 5 min at 37°C.
  • d.
    Add 2 mL of SH-Sy5y cell culture medium to deactivate TrypLE.
  • e.
    Collect cells into the Falcon tube.
• 42.
  Collect NK cells.

  • a.
    Collect cells into a Falcon tube and fill up with PBS.
  • b.
    Centrifuge cells 300 × g for 5 min at RT.
  • c.
    Discard supernatant and resuspend pellet in 1 mL of PBS.
• 43.
  Stain NK cells.

  • a.
    Add 2 μL of Hoechst per 1 mL of NK cells.
  • b.
    Incubate for 10 min at 37°C.
  • c.
    Wash stained NK cells four times by centrifugation at 300 × g for 5 min at RT, discarding the supernatant and re-suspending the cell pellet with PBS.
  • d.
    After final wash, re-suspend NK cells to SH-Sy5y cell culture medium.
• 44.
  Count neuroblastoma and NK cells.

  • a.
    Suspend cells.
  • b.
    Take 20 μL cell sample(s) to 1.5 mL eppendorf tube(s).
  • c.
    Add 20 μL Trypan Blue and mix.
  • d.
    Count cells using BioRad TC-20 cell counter.
  • e.
    Adjust neuroblastoma cells to 5,000 cells/100 μL and NK cells to 2,500 cells/100 μL in SH-Sy5y cell culture medium.

CRITICAL: Trypan Blue is harmful and potentially carcinogenic. Avoid ingestion, inhalation, and skin/eye contact. Use appropriate PPE.

• 45.
  Starting neuroblastoma and NK cell co-cultures.

  • a.
    Pipette 5,000 neuroblastoma cells (100 μL) per well in the 8-well ibiTreat plate(s).
  • b.
    Incubate neuroblastoma cells until they have attached to the bottom and spread (typically 2–6 h).
  • c.
    Pipette wells with or without 2,500 NK cells (100 μL) per well in the 8-well ibiTreat plate. Total volume is then 200 μl per well.
• 46.
  Imaging of neuroblastoma and NK cell interaction.

  • a.
    Image the ibiTreat plate using an epifluorescent spinning disc confocal microscope to ensure high image acquisition speed.
  • b.
    Take images from different parts of the well using 5 min imaging interval in first 2 – 5 h of imaging to visualize cell interactions between neuroblastoma and NK cells, and 20 min imaging interval to visualize neuroblastoma cell death in 2D environment.

Note: 4 h is generally sufficient time to observe targeting NK cells to neuroblastoma cells and cell-cell contact. Use shorter intervals (5 min or less) at the beginning of experiment.

Methods Video S4. In vitro co-culture of Hoechst labeled mock-transduced NK cells with GFP+ neuroblastoma cells, related to step 46b

The field of view for the video is 17.37 mm × 16.68 mm.

Methods Video S5. In vitro co-culture of Hoechst labeled GD2-CAR NK cells with GFP+ neuroblastoma cells, related to step 46b

The field of view for the video is 17.37 mm × 16.68 mm.

Expected outcomes

Successful execution of this protocol yields longitudinal fluorescence images and complementary datasets describing NK cell persistence and NK cell–tumor spatial interactions in ex ovo chick embryos, together with matched in vitro assays (Figures 5A–5F). Throughout the experimental timeline, embryos remain viable and optically accessible, enabling intracardiac xenografting of neuroblastoma cells on Day 3, subsequent intracardiac administration of NK cells on Day 4, and repeated imaging on Days 4–5 before the experimental endpoint on Day 6 (Figure 6).

Figure 5.

Fluorescent microscopy imaging ex ovo

(A–B) Representative fluorescent stereo microscopy images of embryos injected with LucGFP-SH-SY5Y neuroblastoma cells alone (controls). Green GFP signal marks tumor cell distribution along the yolk vasculature.

(C–D) Embryos were injected with neuroblastoma cells, followed by intracardiac delivery of mock NK cells. Hoechst-labeled NK cells (blue) are detectable along vessels, but the GFP signal remains widely distributed, indicating limited tumor clearance.

(E–F) Embryos injected with neuroblastoma cells followed by GD2-CAR NK cells. Hoechst-positive NK cells (blue) appear clustered within tumor-dense regions, and GFP signal is qualitatively reduced relative to controls and mock treatments, suggesting enhanced CAR NK–mediated tumor clearance. All images acquired using fluorescence stereomicroscopy with GFP and DAPI channels.

Each panel represents an independent embryo imaged on Day 4–5. Scale bars = 1 000 μm. See also Tables S1 and S2.

Figure 6.

Representative developmental progression of ex ovo embryos across the experimental timeline

Day 3: Following transfer to ex ovo culture, embryos display a well-defined vascular network. Tumor xenografting is performed at this stage via intracardiac delivery.

Day 4: Embryos continue normal development, and the circulatory system becomes fully accessible for a second intracardiac injection. Human NK cells are administered at this time point.

Day 5: Embryos remain viable and transparent, allowing visualization of intact vasculature and tissue structures. Brightfield and fluorescence imaging are performed to monitor graft distribution, NK persistence, and overall developmental stage.

Day 6: Embryos exhibit continued growth and vascular expansion. This marks the experimental endpoint, at which tissue imaging is finalized and embryos are terminated according to protocol requirements.

Across all timepoints, embryos maintain characteristic morphology and vascular clarity, supporting imaging access and procedural manipulation throughout the assay period.

Following intracardiac xenografting, GFP-positive neuroblastoma cells are typically visible within blood, vascular structures or nearby soft tissue compartments on the day of injection and progressively decline over several days. Hoechst-labeled NK cells with or without IL-15 activation appear as discrete signals throughout the embryo. This method can be used to study NK cell behavior in ex ovo embryo models. IL-15 pre-activation reproducibly improves NK cell detectability over time by supporting enhanced survival within the ex ovo environment. These kinetics can be visualized as embryo-specific decline curves over 24–72 hours (Figure 7).

Using this protocol, differences in neuroblastoma cell persistence can be visualized across experimental conditions following intracardiac co-injection of tumor cells and NK cells. Quantitative analysis of GFP-positive neuroblastoma cells per unit area reveals variable tumor burdens across individual embryos; however, embryos receiving GD2-CAR NK cells frequently display reduced detectable tumor signal by Day 4–5 compared with embryos receiving mock-transduced NK cells, in which measurable neuroblastoma fluorescence is more commonly retained (Figures 8A and 8B). Inter-embryo variability is expected in this model, but relative differences between treatment conditions can be consistently resolved when injection, imaging, and analysis parameters are standardized. Based on our experience, we recommend starting with approximately twice the number of embryos required per experimental condition to compensate for losses due to developmental variability, contamination, or injection failure.

Figure 8.

Quantification of tumor burden and NK–tumor proximity in the ex ovo neuroblastoma model

(A and B) Day 4 and 5 quantifications of GFP-positive neuroblastoma cells/mm2 across embryos injected with tumor cells alone, tumor plus mock-transduced NK cells, or tumor plus anti-GD2-CAR NK cells.

Each dot represents embryos injected with tumor cells alone, tumor plus mock-transduced NK cells, or tumor plus anti-GD2-CAR NK cells. Each dot represents one of nine independently sampled fields, collected from 3–4 embryos per condition.

(C and D) Day 4 and 5 proximity analysis showing the number of NK cells detected within 75 μm radius of neuroblastoma cells. NK-containing conditions exhibit measurable localization toward tumor cells, with the GD2-CAR NK group displaying a tendency toward increased clustering around tumor cells relative to mock NK controls.

Each dot represents one of nine independently sampled fields, collected from 3 and 4 embryos per condition. The dot value is the mean cell number retreat from all neuroblastoma cells in the fields (range of 3–47 neuroblastoma cells per field).

(E and F) Schematic representation and Cellpose/Otsu image examples (i,ii) of the proximity analysis strategy. Hoechst-labeled NK cells within a defined 75 μm radius of a GFP-positive neuroblastoma cell were classified as proximal and included in downstream quantification; cells outside this threshold were excluded. Across all panels, bars represent group means and dots denote individual fields of view.

These findings demonstrate measurable tumor persistence and NK colocalization patterns in embryos over time and suggest enhanced tumor clearance and engagement by GD2-CAR NK cells.

N=8 in each group, each dot represents an individual sample and column their average, Statistical analysis (A–D) was done with two-sample t test. p < 0.05 (∗), p < 0.01 (∗∗), p < 0.001 (∗∗∗).

Spatial interaction analysis provides complementary insight into effector–target behavior by quantifying NK cells located within a defined proximity (e.g., 75 μm) of individual neuroblastoma cells (Figures 8C and 8D). This analysis typically identifies NK–tumor contacts in both mock and GD2-CAR NK conditions. In mock NK datasets, multiple NK cells often localize near tumor targets without complete loss of tumor fluorescence, whereas GD2-CAR NK–treated embryos more frequently exhibit reduced tumor signal despite fewer proximal NK cells per target. Together, these measurements illustrate how combined tumor burden and proximity metrics can be used to compare interaction patterns across differently modified NK cell populations, rather than to establish definitive measures of therapeutic efficacy.

The in vitro arm of this workflow recapitulates these ex ovo interaction profiles under controlled settings (Figures 9A–9E, Methods Videos S3, S4, and S5). Time-lapse microscopy panels serve as a visual reference for expected interaction patterns, while ex ovo quantitative outputs describe mean tumor counts and NK–target proximity metrics. In vitro imaging can be used to visualize how tumor-only cultures maintain stable fluorescence, how mock NK co-cultures retain viable tumor cells with multicellular contacts, and how GD2-CAR NK co-cultures demonstrate progressive target elimination over time.

Figure 9.

Live-cell fluorescence imaging of neuroblastoma–NK cell co-cultures illustrate distinct cytotoxic interaction patterns across treatment groups

Time-lapse fluorescent microscopy images (50 μm scale bar) show sequential frames (i–vi) captured over the course of co-culture. Time intervals between consecutive frames are indicated in the figure. The figure displays the most significant morphological changes of the cells.

(A) SH-SY5Y neuroblastoma monoculture, exhibiting clustered GFP signal and normal morphology over time.

(B and C) Co-cultures with mock-transduced NK cells, in which frequent NK–target contacts are observed, often involving multiple NK cells engaging a single neuroblastoma cell.

NK accumulation around targets is visible, but GFP signal persists, indicating limited killing efficiency.

(D and E) Co-cultures with GD2-CAR NK cells, showing fewer NK–target contacts yet progressive loss of GFP fluorescence, consistent with more effective neuroblastoma elimination.

CAR NK cells frequently associate singly with targets, suggesting higher per-cell cytotoxic potency.

Across multiple recorded fields (see scene IDs), GD2-CAR NK cultures consistently exhibited reduced neuroblastoma fluorescence intensity compared with mock NK or tumor-only wells, mirroring clearance patterns observed in the ex ovo system.

This assay therefore provides a complementary, high-resolution visualization of NK–tumor dynamics under controlled conditions. See also Methods Videos S3, S4 and S5.

Together, these ex ovo and in vitro outcomes provide reproducible readouts—including tumor burden, NK persistence, and NK–tumor proximity—that users can expect when applying this protocol. Although absolute magnitudes may differ across experiments, consistent directional trends are typically observed: tumor-only and mock NK conditions retain detectable neuroblastoma signal, while anti-GD2 CAR NK treatment leads to more pronounced reduction or loss of target cells. These patterns, which emerge in both ex ovo and in vitro assays, serve as practical benchmarks for users to assess successful implementation of the workflow.

Quantification and statistical analysis

Fluorescence datasets were analyzed using a combined classical and deep-learning–assisted image analysis workflow. Raw CZI-format images acquired from the microscope were first processed using ZEN 3.12 (ZEN Lite, ZEISS) to perform background and illumination correction, including rolling background subtraction and contrast enhancement. The images were then exported as JPEG/TIFF files for downstream image segmentation and quantitative analysis.

Segmentation of Hoechst-labeled NK cells and GFP-positive neuroblastoma cells was performed using different approaches depending on signal quality and imaging time point. For Day 4 images, where the fluorescence signal was strong and cell boundaries were well defined, segmentation was performed using Cellpose (pre-trained nuclei model). For Day 5 images, where overall Hoechst fluorescence intensity was reduced and Cellpose performance was compromised, segmentation was performed using an Otsu-based adaptive thresholding approach. In both cases, binary masks were refined using watershed-based separation to resolve touching objects, followed by morphological opening and minimum-area filtering to remove noise and exclude objects below a biologically plausible size. Key adjustable parameters for both Cellpose (Figure S1)- and Otsu-based (Figure S2) workflows, including model settings, size thresholds, and classification criteria, are summarized in Figure S1 and S2 to enable reproducibility and method adaptation.

Cell counts and object features were exported to Python scripts as tables containing unique object IDs, centroid coordinates, and per-image totals (Figures 10A–10C). https://doi.org/10.5281/zenodo.19111461 (GitHub code).

Figure 10.

Workflow for automated fluorescence image analysis, combining Cellpose segmentation, threshold-based filtering, and proximity quantification, size-to-pixel conversion, and MATLAB cell segmentation and counting

(A) Raw fluorescence image acquired ex ovo, showing GFP+ neuroblastoma cells (green) and Hoechst-labeled NK cells (blue). Scale bar: 200 μm.

(B) The same image following background removal and contrast enhancement using classical image processing approaches (including Otsu thresholding for intensity-based segmentation), followed by CellPose-assisted segmentation to refine object boundaries and improve detection accuracy. See also Figure S1.

(C) Output of spatial proximity analysis in which Cellpose-segmented tumor cells were circled and NK cells located within a 75 μm radial distance were counted. Proximity quantification was implemented using Python-based image processing scripts integrating threshold-derived masks and Cellpose labeling. See also Figure S2.

(D) Screen capture of the raw image file metadata showing the Scaling (per Pixel) values and Image Size.

These metadata parameters are used to convert pixel distances into real-world spatial units (μm).

The per-pixel scaling factor is applied to compute a 75 μm radius around each GFP+ neuroblastoma cells, required for downstream Cellpose-based proximity analysis (as shown in panel C).

(E) Automated MATLAB code-based quantification and persistence of intracardially injected NK cells in the chick embryo model. Representative fluorescence image showing automated detection of Hoechst-labeled NK cells following intracardiac delivery into ex ovo chick embryos.

Detected cellular objects are displayed as red boundaries overlaid on the original image to illustrate segmentation accuracy.

Because fluorescence images were acquired at slightly different magnifications, the biological proximity threshold of 75 μm corresponded to different pixel distances across image sets (Figure 10D). For each image batch, pixel-to-micron conversion factors were extracted from microscope metadata and applied as shown in Equation 1.

Conversion from micrometers to pixels, where X is the per-pixel scaling value extracted from microscope metadata.

$$\frac{75\mu m\left(radius\right)}{X\mu m/pixel}$$ (Equation 1)

NK–tumor proximity was calculated by computing the Euclidean distance between tumor cell centroids (gx,gy) and NK cell centroids (bx, by) in pixel space, and applying an image-specific pixel radius corresponding to 75 μm (typically 52–150 pixels). This calibration ensured that proximity measurements reflected a constant biological distance independent of imaging resolution.

Cell counts were normalized to image area by calculating the scaled field size (mm²) from image metadata (Image Size, Scaled) and reporting values as cells/mm². All quantitative analyses were performed on images acquired from three biologically independent embryos per condition, unless otherwise stated.

For NK cell persistence analysis, each embryo was treated as an independent biological replicate (n = 4 embryos per condition). Hoechst-labeled NK cells were automatically detected and counted from ex ovo fluorescence images using a MATLAB-based image analysis workflow, as illustrated in Figure 10E. NK cell counts were extracted at 24, 48, and 72 h following intracardiac injection and plotted on a logarithmic scale for each embryo to visualize temporal changes in cell numbers. Because intracardiac delivery efficiency varies between embryos, analyses focused on within-embryo changes over time rather than pooled absolute cell counts. Persistence profiles were evaluated descriptively without formal hypothesis testing due to limited sample size and non-normal data distribution.

The MATLAB code for cell segmentation and cell counting is as follows:

img = imread('Snap-7330.jpg');

% Extract the blue channel (where cells are brightest)

blue_channel = img(:,:,3);

% Enhance contrast for better object separation

blue_contrast = imadjust(blue_channel);

% Calculate Otsu's threshold as starting point

level_otsu = graythresh(blue_contrast);

% Use a higher threshold to retain only the brightest cells

thresholdOffset = 0.30; % Try values between 0.12 to 0.20 for best result

level = min(level_otsu + thresholdOffset, 1);

% Binarize the image using the adjusted threshold

bw = imbinarize(blue_contrast, level);

% Remove small objects: set an increased minimum expected cell size

minCellSize = 20; % Increase this further if needed!

bw_clean = bwareaopen(bw, minCellSize);

% Label components (likely cells)

[labeled_cells, num_cells] = bwlabel(bw_clean);

% Display the result

figure;

imshow(img); hold on; visboundaries(bw_clean, 'Color', 'r');

title(['Number of detected cells: ', num2str(num_cells)]);

hold off;

% Display cell count in the command window

disp(['Number of detected cells: ', num2str(num_cells)])

For spatial interaction assays, NK–tumor proximity was quantified by identifying NK cell centroids located within the calibrated 75 μm radius surrounding tumor cell centroids. Nine fields per embryo were analyzed, with per-embryo averages serving as experimental units. In vitro co-culture datasets were used as a qualitative reference only to visualize mock and GD2-CAR NK cell engagement patterns and were not subjected to quantitative image analysis.

Graphs were prepared using Excel or GraphPad Prism and displayed as scatter distributions or embryo-specific trajectory plots with means. Embryos excluded prior to imaging due to contamination, injection failure, or developmental arrest were removed based on predefined technical criteria. Image-level exclusions occurred only when predefined technical artefacts were observed (e.g., non-yolk imaging regions, blur-induced object merging, or background misclassification) and were applied before numerical data generation. No post hoc exclusion based on experimental outcome was performed. Example segmentation masks, centroid outputs, and processed data tables are provided via the referenced GitHub repository.

Limitations

The ex ovo chick embryo system provides a rapid and accessible platform for visualizing early interactions between neuroblastoma cells and human NK cells; however, several practical and biological constraints define the scope of the method. Embryos can only be used up to the developmental stage considered pre-nociceptive under local regulations, and all procedures must be performed within this permitted window. A further limitation is that intracardiac injection is restricted to a narrow developmental window (embryonic day 3–4), as earlier stages lack sufficient cardiac development for reliable targeting, whereas later stages present increased tissue thickness makes injection technically challenging.

Neuroblastoma cells introduced into the circulation survive for a limited period because they do not form long-term intravascular engraftments in this model. If extended tumor persistence is required, intracelomic injection may be used as an alternative route, as tumor cells remain longer in the extraembryonic cavity than in the bloodstream. Similarly, human NK cells require supplemental IL-15 for sustained viability, and very high NK cell numbers may exceed the capacity of the embryonic circulatory system. The overall experimental throughput is also constrained by the number of embryos that can be reliably injected, imaged, and tracked by a single operator within the relevant developmental timeframe.

Fluorescence stereomicroscopy allows convenient, repeated imaging of live embryos, but it is largely restricted to superficial structures such as the yolk and CAM vasculature. Deeper embryonic regions cannot be visualized with this modality, and processes occurring within the embryo body remain inaccessible. Complementary imaging approaches—including bioluminescence imaging (BLI) or, where appropriate, confocal imaging or dual photon microscopy with respective 200 μm and 2 to 3mm depths—may help assess deeper tumor burden or NK–tumor interactions.

Finally, image acquisition and downstream computational analysis can introduce variability if field selection, illumination, or image quality differ between embryos. Although standardized imaging settings and segmentation quality-control steps reduce these influences, some degree of user-dependent variability remains inherent to the system. Overall, these limitations do not reduce the utility of the ex ovo model but instead define the experimental questions that it is best suited to answer.

Troubleshooting

Problem 1

Problems with embryo sensitivity and developmental variability. Embryos are highly sensitive to environmental conditions and developmental stage. Moisture loss, cooling, or use of embryos outside the optimal developmental window can lead to injection failure or embryo death. These problems are related to Day 0 (Incubation of fertilized eggs), Day 3 (Xenografting via intracardiac injection), Day 4 (Intracardiac injection of NK cells), and Days 4 and 5 (Fluorescent microscope imaging ex ovo).

Potential solution

• •When embryos are outside the incubator, maintain moisture by applying PBS drops on top of the embryos. Limit time outside the incubator for injections or imaging.
• •Keep embryos warm using a thermal plate.
• •Monitor heartbeat and return embryos immediately to incubator if the heartbeat slows down.
• •Use embryos at approximately day 3 to 4 of development (HH embryo 15–18 days). Only perform injections when cardiac structures are clearly visible, well filled with blood and beating, indicating sufficient vascular diameter for reliable intracardiac access, suitable for intracardiac injection.
• •Incubate twice the number of eggs to have excess embryos to compensate for unusable samples.

Problem 2

Problems with intracardiac injection accuracy and embryo positioning (Figures 11A–11E). Incorrect injection angle, membrane displacement, or suboptimal embryo orientation can result in vascular damage, failed intracardiac access, or poor imaging conditions. These problems are related to Day 0 (Incubation of fertilized eggs), Day 3 (Xenografting via intracardiac injection) and Day 4 (Intracardiac injection of NK cells).

Figure 11.

Intracardiac injection accuracy, embryo position, contamination and debris

(A) The injection without Fast Green makes it difficult to evaluate the success of intracardiac injection.

(B) The injection without Fast Green makes it difficult to visualize whether cells go into the bloodstream or leak to the surrounding tissues.

(C) Injection with Fast Green makes it easy to see cells in the needle and blood stream.

(D) Injection with Fast Green makes it easy to visualize when it is time to stop cell injection (leaving 0.5 μl of cells still inside the needle).

(E) Position of the embryo determines if the embryo can be intracardially injected. Here, the embryo is on the side, but the heart could still be injected. Imaging, however, will be difficult since the embryo and the whole yolk vasculature are not visible.

(F) Contaminated embryos showing white growth.

(G) Trash separated from the injection pipettes markings and Fast green-stained cells leaking to the surrounding tissues.

Potential solution

• •Insert the glass pipette slowly at a shallow angle to avoid vascular tearing.
• •Practice injections on non-experimental embryos.
• •Add Fast Green dye to confirm successful injection and avoid repeated punctures (Figures 11A–11D).
• •Ensure embryos are positioned at the top, not on the side, of the yolk/waterbed for reliable access and imaging (Figure 11E).

Problem 3

Problems with cell dose and sterility-related issues (Figures 11F–11G). Excessive NK cell numbers may overload embryonic circulation, while inadequate aseptic technique can lead to contamination or imaging artefacts. These problems are related to Day 3 (Xenografting via intracardiac injection) and Day 4 (Intracardiac injection of NK cells).

Potential solution

• •Titrate NK cell numbers and begin with lower effector doses to avoid overloading the embryonic circulation and to maintain embryo viability.
• •Maintain sterile techniques during injection and imaging to avoid contamination (Figure 11F).
• •Avoid excessive marking ink and rinse glass pipettes gently to minimize black marking debris falling off from the pipette during injection (Figure 11G).
• •Replace glass pipette tips every 4–5 embryos.

Problem 4

Problems with imaging acquisition artefacts and variability (Figures 12, 13, and 14). Fluorescence imaging may be affected by blur, autofluorescence, uneven illumination, or inconsistent magnification, which can impair accurate detection of NK and tumor cells. These problems are related to Day 4 and 5 (Fluorescent microscope imaging ex ovo).

Figure 12.

Common segmentation artefacts in fluorescence image analysis

(A–C) Examples of inaccurate cell detection using simple HSV-based or intensity-only segmentation, resulting in merged tumor objects, merged NK cells, or false-positive background detection.

(D) Combined Cellpose- or Otsu-based segmentation followed by watershed separation improves discrimination of individual cells and background. Circles indicate the calibrated 75 μm NK–tumor proximity radius. Scale bar = 200 μm.

Figure 13.

Optical challenges during fluorescence imaging of ex ovo chick embryos

(A–C) Representative fields showing partial embryo views, uneven surface curvature, or embryonic structures within the field of view, leading to shading, depth-dependent signal attenuation, and focal plane variability.

(D–E) Fields containing both embryo and chorioallantoic membrane (CAM), illustrating heterogeneous optical properties and variable background fluorescence.

(F) Membrane curvature causes focal plane inconsistencies and reduced image sharpness across the field.

Ex ovo embryos exhibit heterogeneous optical properties and variable surface topography that can impair the detection of GFP-labeled tumor cells and Hoechst-stained NK cells.

To minimize artefacts, prioritize superficial regions with stable optical access, adjust illumination, apply flat-field correction or background subtraction, and exclude fields where uneven illumination or depth-related signal loss prevents reliable segmentation. Scale bar = 1 000 μm.

Figure 14.

Impact of inconsistent image magnification on downstream quantitative analysis

(A) Zooming in on cells increases image clarity but reduces the represented field size. When calculating proximity thresholds (e.g., 75 μm radius around neuroblastoma cells), larger pixel radii must be applied to compensate for the higher magnification.

(B) Zooming out captures a larger embryo area but decreases cell resolution, making segmentation more challenging.

In this case, smaller pixel radii must be applied to maintain the correct 75 μm proximity threshold.

Inconsistent zoom introduces variability in field size, which affects cell density calculations (cells/mm²) and proximity measurements. To ensure accurate analysis, convert pixel-based distances to micrometer equivalents for each image individually.

Potential solution

• •Exclude blurred images, as they can result in single cell identification problems during segmentation in analysis (Figures 12A and 12B). Refocus to ensure clear cell boundaries (Figure 12D).
• •Apply background subtraction or flat-field correction to reduce autofluorescence, which can result in a greenish background falsely identified as green cells (Figure 12C).
• •Prefer superficial imaging regions with stable optical conditions like flat yolk vasculature. Avoid imaging the embryo or CAM, which are more three-dimensional (Figures 13A–13F).
• •Maintain consistent magnification and record imaging metadata. Uneven zooming leads to complications in downstream processing when the mm-to-pixel scaling factor varies across images (Figures 14A and 14B).
• •Exclude fields with severe illumination gradients or distortion.

Problem 5

Problems with quantification and segmentation errors (Figures 12 and 14; Table 1). Automated image analysis may introduce errors due to clustering, background signal, blur, or pixel-scale variability, affecting accurate cell counting and proximity measurements. These problems are related to quantification and statistical analysis.

Table 1.

Quantification and segmentation errors

Artefact/failure mode	Where shown	Why it happens	Impact	Recommended handling
Clustered tumor cells segmented as one object	Figure 12A	Touching GFP+ tumor cells form contiguous regions that cannot be separated by basic masking	Underestimates tumor count; inflates object area	Use Cellpose to detect objects, then apply watershed separation to split touching cells; exclude clusters that remain inseparable
Multiple NK cells merged into one mask	Figure 12B	Blur/defocus reduces boundary contrast; segmentation merges nearby nuclei	Underestimates NK counts; distorts proximity metrics	Exclude blurred fields before analysis; only segment images with visually resolvable NK contours
Green background counted as tumor “cells”	Figure 12C	Autofluorescence/illumination scatter produces a diffuse GFP-like signal; global masks convert haze into many small objects	False-positive tumor objects; inflates tumor burden	Apply background subtraction, then Otsu thresholding to suppress diffuse signal, followed by minimum-area filtering; exclude fields with uncorrectable haze
Background separated successfully after preprocessing	Figure 12D	Combined preprocessing improves cell–background contrast and splits objects	Restores usable masks for quantification	Use the combined strategy (background subtraction → Otsu → Cellpose → watershed, plus size filtering) and validate with mask previews before batch runs
Pixel-scale variation between images	Figures 14A and 14B	Magnification/field size changes alter pixel-to-μm scaling and image area	Incorrect 75 μm radius and non-comparable densities	Extract pixel-to-μm scale per image from metadata; convert 75 μm to image-specific pixel radius; normalize counts to cells/mm2

Potential solution

• •Use combined segmentation approaches (Cellpose + Otsu + watershed) (Figure 12D).
• •Apply minimum-area filtering and validate segmentation masks visually.
• •Exclude images with unresolved clustering or background artefacts (Figures 12A–12C).
• •During quantification, normalize area measurements using image-specific pixel-to-micron scaling (Figures 14A and 14B), normalize cell counts to e.g., cell/mm² (Figures 8A and 8B), and convert biological distances (e.g., 75 μm proximity radius) into image-specific pixel units (Figure 10D).
• •Refer to the troubleshooting table for detailed artefact-specific handling.

Together, these measures ensure that segmentation and proximity quantification reflect consistent biological thresholds, thereby reducing bias from background, clustering, blur, and scale variability.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Farhana Jahan (farhana.jahan@bloodservice.fi).

Technical contact

Technical questions related to execution, imaging, ex ovo intracardiac injection, and other ex ovo related enquiries should be directed to and will be answered by the technical contact, Ilya Skovorodkin (ilya.skovorodkin@oulu.fi).

Materials availability

This protocol did not generate new unique reagents. All cell lines and reagents used are commercially available or were obtained from previously published sources, as detailed in the key resources table section. Any materials generated as part of this study are available from the lead contact upon reasonable request and subject to standard material transfer agreements.

Data and code availability

• •All datasets generated or analyzed during this study are included in the section on expected outcomes, quantification, and analysis. Example processed images, segmentation masks, and analysis scripts used for fluorescence image quantification are publicly available at Zenodo: https://doi.org/10.5281/zenodo.19111461.
• •No additional unpublished datasets were generated in this study.

Acknowledgments

This work was supported by funding from the Finnish Cultural Foundation, the Cancer Foundation Finland, the Blood Service Research Foundation, and the Academy of Finland GeneCellNano Flagship Program (grant no. 374293), as well as Business Finland (Dnro 4172/31/2024, Data-Inductor). This research was conducted at the Finnish Red Cross Blood Service and the University of Oulu. We thank Lotta Andersson for technical assistance. We also acknowledge the Biocentre Oulu Imaging Core Facility, University of Oulu, for expert imaging support. We are especially grateful to Anne Martikainen and Maarit Mähönen from the National Virus Vector Laboratory, A.I. Virtanen Institute, University of Eastern Finland, for producing and titrating the lentiviral vectors used in this study. We further thank Jan Koski for his valuable assistance with the construction design.

Author contributions

M.S. performed experiments, acquired imaging data, analyzed images and datasets, prepared figures, and wrote the manuscript. E.J. performed experiments and acquired imaging data. T.A. developed and implemented the image analysis workflows and assisted with quantitative data processing. H.S. performed experiments and performed MATLAB-based data analyses. I.S. performed intracardiac injections and embryo imaging and contributed to optimization of the ex ovo workflow. H.G., M.K., S.V., D.S., and S.Y.-H. contributed to manuscript review and provided general scientific input. F.J. conceived and designed the study, performed experiments and data analysis, and supervised and wrote the manuscript. All authors reviewed the manuscript and contributed to its revision.

Declaration of interests

The authors declare no competing interests.