DACIT device for axon cancer cell interaction testing in 2D and 3D

Summary

Peripheral innervation is increasingly recognized as a critical regulator of tumor progression, yet in vitro models that enable controlled study of axon-cancer cell interactions remain limited. Here, we present the Device for Axon-Cancer cell Interaction Testing in 2D and 3D (DACIT), a microfluidic platform that spatially separates neuronal somas from axons and cancer cells. This configuration supports experimental designs where compartments can be exposed to either identical or distinct conditions. Moreover, the channel height allows the incorporation and monitoring of tumor spheroids, enabling quantification of tumor growth and 3D invasion. We demonstrate DACIT compatibility with common cellular assays, including immunofluorescence, invadopodia assays, pharmacological perturbations, live-cell imaging, and 3D spheroid invasion. Together, these features establish DACIT as a versatile tool to interrogate how peripheral axons influence cancer cell behavior.

Graphical abstract

Highlights

• •Understanding cancer cells interactions with peripheral axons enables cancer research
• •DACIT enables controlled 2D and 3D study of cancer cell interactions with axons
• •DACIT provides cellular and media compartmentalization
• •DACIT is optimized for high-resolution imaging in fixed and live-cell studies

Biological sciences; Neuroscience; Cell biology; Cancer

Introduction

While the nerve infiltration into tumors was identified more than 100 years ago, it was not until recently that neoaxonogenesis has been recognized as one of the hallmarks of cancer.^1,2 Even now, cancer cell-neuron interactions are mostly studied in central nervous system, while the interactions between neurons and cancer cells in the peripheral organs remain generally understudied. Few recent studies have started demonstrating the impact of the autonomic or somatic peripheral nervous system in tumor growth and metastasis, including pancreatic, melanoma, and breast cancer.^3,4,5,6,7 These observations generated an increased interest in understanding the molecular and cellular mechanisms that drive interactions between peripheral neurons and cancer cells and developing appropriate technical solutions for such studies.

Several technical reasons can explain this knowledge gap. First, innervation in mammals has a unique anatomy where the peripheral neurons’ somas (cell bodies) are located in ganglia distributed along the body: close to the cranial nerve, in the vicinity of the vertebral column, or the abdominal cavity. Hence, tumor innervation and possible physical contact with tumor cells is achieved via axons, axon bundles (fascicles). In humans, distance between the ganglion and the tumor that the ganglion innervates can be longer than one meter. Mimicking such cellular compartmentalization in tissue culture-based experiments is important to maintain anatomical similarity and avoid experimental artifacts that can occur in simple co-culture settings. Neurotoxicity of the glutamate, which appears at high concentrations in the fetal bovine serum used for culturing most cancer cell lines is one classical example of such experimental differences.⁸ Second, currently available models of peripheral neurons do not reflect the complexity and diversity of the normal peripheral nervous system.^9,10,11 As a result, most studies use embryonic or adult primary culture from rats or mice to study neuron behavior. However, the use of primary culture limits the number of neurons that can be obtained at a given time and also introduces the issue of the presence of non-neuronal cells, such as satellite glial cells, immune cells, fibroblasts, endothelial cells, or Schwann cells,¹² which affect cancer cell behavior^13,14 but can be difficult to eliminate without affecting the health of neurons. While a few recent studies have reported successful differentiation of induced pluripotent stem cells into sensory neurons,^15,16,17 this approach is not yet widely utilized.

To provide the compartmentalization between soma and axons, Campenot developed a chamber consisting of a collagen-coated Petri dish with scratches to guide axon growth.¹⁸ Recently, principles of microfluidics were added to the Campenot chamber, with several variations. The progress of microfabrication and the inspiration of the original concept of the Campenot chamber triggered the development of a series of compartmentalized microfluidic devices to analyze anything from axonal regeneration^19,20 to cellular and biochemical assays or drug screening²¹ in 2D cultures.

We focused on developing a microfluidic chamber which can allow analysis of both 2D and 3D cell cultures. The 3D culture represents a physiologically relevant alternative to animal models, as it better recapitulates in vivo cell behaviors, including proliferation, morphology, adhesion, motility, etc.^22,23 Here, we present a novel Device for Axon Cancer cell Interaction Testing in 2D and 3D (DACIT). DACIT allows monitoring of the interactions between axons and cancer cells growing on thin layers of extracellular matrix (ECM) in 2D, or spheroids embedded in the 3D ECM. DACIT can be used to expose neuronal soma and axons to different, or similar, media conditions, using compartmentalization or diffusion/mixing of media, respectively. Moreover, DACIT allows genetic and chemical treatments and high-resolution, time-lapse imaging.

Results

DACIT layer design and the fabrication of SU-8 master

Using hard and soft lithography, several microfluidic devices have been developed to isolate axons from neuronal soma (Table 1) and to investigate diverse processes such as axonal gene expression,²⁰ neuronal activity under axonal stimulation,²⁴ synaptic network dynamics,²⁵ and axon interactions with other cell types including oligodendrocytes,²⁶ Schwann cells²⁷ or cancer cells.²⁸ Building on the two-compartment device originally developed by Taylor et al.,²⁰ we designed a microfluidic platform to study axons interacting with cancer cells growing in both two-dimensional cultures and three-dimensional tumor spheroids embedded in ECM (DACIT).

Table 1.

Comparison between published microfluidic devices and DACIT

Microfluidic device type	Material/Fabrication	Main biological application	Interaction	Key feature device	Axon isolation	Fluidic isolation	Cellular isolation	Specialized equipment required	Channel height	Reference
Two-compartment device	PDMS soft-lithography	Axon isolation	Neurons only	Axon-only compartment culture	Yes	Yes	ND	Photolithography and soft lithography	100 μm	Taylor et al.29
Three-compartment device	PDMS soft-lithography	Neuronal differentiation on the chip	Neurons with Schwann cells	Microfluidic delivery of cells	ND	ND	ND	Photolithography and soft lithography. Microfluidic pump	100–250 μm	De Vitis et al.30
Two-compartment device	PDMS soft-lithography	Neuronal activity	Neurons only	Axon isolation for drug testing	Yes	Yes	ND	Photolithography and soft lithography	ND, ND	Atmaramani et al.31
Three-compartment device	PDMS soft-lithography	Analyze neuronal networks	Neurons only	Design enables synaptic network reconstruction	Yes	Yes	Yes	Photolithography and soft lithography	ND	Virlogeux et al.25
Two-compartment device	PDMS soft-lithography	Migration of cancer cells along neurites	Neurons with cancer cells	Studies of cancer cells migration along neurites	Yes	ND	Yes	Photolithography and soft lithography	150 μm	Lei et al.32
Two-compartment device	PDMS soft-lithography	Recapitulation of myelin formation	Neurons with oligodendrocytes	Open neuronal compartment	Yes	ND	Yes	Photolithography and soft lithography	100 μm	Kerman et al.26
Three-compartment device	PDMS soft-lithography	Myelination process	Neurons with Schwann cells	Channel promotes axon myelination in 3D, enabling long-term culture (40 days).	Yes	ND	Yes	Photolithography and soft lithography	100 μm	Hyung et al.27
Two-compartment device	Fused silica MPS patterned with micrometer-scale features and attached to an MEA substrate	Neuron activity	Neurons with other cells	Suited to analyze neuronal activity in presence of other cells for MEAs in 3D	Yes	ND	Yes	Selective laser etching (SLE), photolithography, adhesive bonding via fineplacer and multi-electrode array (MEA) system	300 μm	van der Moolen et a.24
Two-compartment device	PDMS soft-lithography	Axon-cancer cell interaction in 2D/3D	Neurons + cancer cells	Channel height accommodates 3D assays; optimized for time-lapse imaging of neurons and cancer cells	Yes	Yes	Yes	Photolithography and soft lithography	up to 700 μm	DACIT, this paper

ND, no determined.

DACIT consists of two parallel channels (1.5 mm wide), each with 6 mm punched wells at both ends for loading media and cells (Figure 1A). The channels are interconnected by microgrooves (300 μm in length, 5 μm in width, and 3 μm in height) that permit axons to cross while restricting neuronal soma and cancer cells to their compartments (Figure 1A, middle and right).

Figure 1.

Design and production of DACIT

(A) DACIT design in 3D view (left), top view (middle), and side view (right). Neuronal (yellow) and axonal (magenta) compartments each contain two wells (6 mm diameter), where solutions can be added or removed.

(B) Microgroove photomask (left) with blue alignment marks on each side and macrochannel photomask with red alignment marks. Overlay shows the alignment of microgrooves (blue) and macrochannels (red). (Bi) Close-up of the microgrooves connecting the left and right macrochannels.

(C) Main steps in SU-8 master preparation. First, a layer of SU-8 is spin-coated and baked on a silicon wafer. The microgroove photomask (green) is aligned on the baked SU-8, followed by UV exposure, development, and soft and hard baking to obtain the microgroove layer. Next, the macrochannel layer is produced by two rounds of spin-coating and baking at 95°C before UV exposure with the macrochannel photomask (orange). Finally, the wafer is developed and baked to obtain the master.

(D) Final SU-8 master. Scale bars, 2 cm.

(E) Top view of the DACIT. Scale bars, 1 cm.

(F) Brightfield cross-section image of DACIT showing a channel height of 700 μm. The dotted black line indicates the coverslip position. Scale bars, 200 μm.

(G) 3D-printed holders designed to mount and image either a single DACIT device (right) or six devices simultaneously (left). Scale bars, 2 cm.

(H) Aluminum holder for time-lapse imaging of six DACIT devices. Scale bars, 2 cm.

Because 3D culture more closely mimics in vivo tumor behavior,^33,34 DACIT was designed to allow only axonal extension through the 3-μm high microgrooves, while accommodating spheroid heights of approximately 400 μm in the microchannel.

Achieving this required channel heights substantially larger than in standards devices, where typical height is 50–100 μm. To this end, we separated the master design into two layers: one photomask containing the microgrooves (Figure 1B, blue) and another containing the channels (Figure 1B, red). Master fabrication was achieved by multilayer photolithography on a silicon wafer coated with SU-8 3000 epoxy resin³⁵ (Figure 1C). First, the microgroove layer was patterned by spin-coating, baking, and UV exposure through the microgroove photomask. Subsequently, two spin-bake cycles were performed to build the 700-μm thick channel layer, followed by alignment and exposure with the channel photomask. The resulting master contained four identical DACIT patterns (Figure 1D) and could be replicated by smooth-on molding to generate up to 20 devices simultaneously.

Using this master with standard polydimethylsiloxane (PDMS) soft lithography, punching out the wells and plasma bonding PDMS to glass coverslip we produced final DACITs (Figure 1E). Cross-sectional imaging confirmed a channel height of 700 μm (Figure 1F).

Due to the small size of DACIT device (24 × 24 mm), mounting it on the microscope stage require an adaptor. For this purpose, we designed custom 3D-printed holders for one, or six, DACITs (Figure 1G), which can be made of variety of materials, including polypropylene or polycarbonate, which can be autoclaved to maintain sterility. In addition, for use in time-lapse imaging, we adapted our designs to CNC-machine holders out of aluminum (Figure 2E). High thermal conductivity of aluminum facilitates maintaining DACIT media temperature at 37°C in any standard environmental chambers.

Figure 2.

DACIT enables media compartmentalization or diffusion and supports cellular compartmentalization

(A) Fluidic compartmentalization (left: 200 μL of media loaded in the neuronal compartment and 150 μL in the axonal compartment) or free diffusion (right, 200 μL loaded in each compartment) were monitored 0–48 h, via 3 kDa fluorescent dextran signal. Dextran diffuses from the axonal compartment in a retrograde fashion, moving across the microgrooves and into the neuronal compartment. Pseudocolor 0–700 μm maximum projections (top, legend on the right) show dextran distribution in microchannels and microgrooves at 24 h. Note that the signal from the microgrooves (3 μm in height), compared to the signal from the macrochannels (700 μm in height), is expected to be approximately 700/3 = 233-fold higher, appearing similar to the background level. Bottom panels show dextran distribution over the entire course of 0–48 h (1 h interval). Readings from the axonal compartment are shown as red dotted lines and from the neuronal compartment as black dotted lines. Scale bars, 200 μm.

(B) PC12-GFP or PGP9.5-labeled primary DRG cultures after 3 days in culture. Fluorescence was pseudocolored black for clarity. Left: close-up of axons extending across the microgrooves. Scale bars, 100 μm. Right: overview showing axons populating the axonal compartment. Scale bars, 200 μm.

(C) Quantification of axonal length in PC12 and primary DRG cultures from five fields of view.

(D) Inverted immunofluorescence images of monocultures of BT-549 breast cancer cells (F-actin, top) or sensory neurons (PGP9.5, bottom) fixed on day 3 post-loading in DACIT. Scale bars, 200 μm.

(E) Immunofluorescence showing single channels in grayscale (left) and fluorescent overlay image (right). Neurons are labeled with PGP9.5 (white), Schwann cells (SC) with Sox-10 (yellow), BT-549 cancer cells (CC) with CK8/18 (magenta), and nuclei with DAPI (cyan). Neuronal soma is pointed with yellow arrows (left). Scale bars, 200 μm.

DACIT allows the cellular and fluidic compartmentalization

Previous studies using microfluidic devices have shown that different loading volumes in macrochannels generate fluidic isolation through differences of hydrostatic pressure across the microgrooves.²⁰ To confirm this in DACIT channels, which have macrochannels taller than conventional devices, fluidic isolation was evaluated by loading the axonal compartment with 3 kDa fluorescent dextran and monitoring its distribution over 48 h, under the loading conditions supporting either fluidic compartmentalization or free diffusion (Figure 2A).

Dextran was loaded into the 6-mm diameter wells by a 200 μL pipette. In the fluidic compartmentalization mode, where 200 μL of DRG media was added to the neuronal compartment and 150 μL of cancer cell media was added to the axonal compartment, dextran permanently remained confined to the original compartment (Figure 2A, left). In contrast, in the diffusion mode, where 200 μL of media was added to each of the compartments, dextran flowed across the microgrooves and reached equilibrium within 6 h (Figure 2A, right). These results demonstrate that DACIT maintains fluidic isolation for molecules as small as 3 kDa when loaded with different media volumes.

DACIT’s ability to separate neuronal soma from axons was assessed using three commonly employed neuronal models: murine PC12 cell line, and either adult or embryonic primary murine DRG neurons. Across all models, soma remained confined to the neuronal compartment, while axons extended into the axonal compartment (Figure 2B). Depending on the type of neurons used, density and rate of axonal extension varied. While immortal and easy to passage, PC12 cells³⁶ produced relatively short neurites that did not consistently reach the axonal compartment on day 3 (Figure 2B, top; Figure 2C). At the same time, axons from primary DRG cultures robustly extended into the axonal compartment (Figure 2B, middle and bottom; Figure 2C). As expected, embryonic DRG axons extended faster, formed thicker bundles that filled the microgroove space, and displayed greater length and branching compared to axons from adult DRGs.³⁷

To test cellular compartmentalization, we loaded two devices with monoculture of DRG neurons or cancer cells. In one device, the neuronal compartment was loaded with primary DRG cells and in the other, axonal compartment was loaded with cancer cells. On day 3, we verified that axons from the neuronal compartment have successfully crossed the microgrooves and populated axonal compartment. In both devices, neuronal soma and cancer cells remained confined to their respective compartments (Figure 2D).

We next tested compartmentalization with cells present in both compartments. We plated primary DRG cells in the neuronal compartment and, on day 3, added cancer cells to the axonal compartment. After 24 h post-addition of cancer cells (96 h total), the culture was fixed and stained for nuclei (DAPI), neurons (PGP9.5), Schwann cells (Sox10), and cancer cells (CK8/18) (Figure 2E). Similar to the results of the DRG monoculture, the neuronal compartment contained neuronal soma (PGP9.5⁺, yellow arrows) and extended axons across the microgrooves (PGP9.5+). In addition, Schwann cells (Sox10⁺) and other non-neuronal DRG cells (satellite glia, endothelial cells, fibroblasts, etc.) still present in the primary culture can be visible in DAPI channel. Importantly, cancer cells were not present in the neuronal compartment (CK8/18, magenta). In contrast, the axonal compartment contained PGP9.5⁺ axons and CK8/18⁺ cancer cells, but no Sox10⁺ Schwann cells or neuronal soma.

Notably, this cellular separation was maintained throughout the 96-h experiment. In comparison with a previously reported device, where cancer cells were able to migrate across the microgrooves,²⁸ DACIT provides more reliable long-term cellular separation, making it particularly well suited for studying sustained cancer cell-axon interactions.

Examples of 2D assays assessing neuronal and cancer cell behaviors in DACIT

To test how DACIT performs in assays measuring neuronal responses to stimuli, we chose to monitor intracellular calcium fluctuations in neurons. Calcium spikes are known to initiate processes such as neurotransmitter release and are widely used as a readout of neuronal activity.³⁸ To monitor calcium spikes in DACIT, we labeled the primary sensory neurons with the fluorescent calcium indicator GCaMP6f and performed time-lapse imaging under basal conditions (Figure 3A and Video S1), or with additional stimulating treatments applied to the axonal compartment (Figure 3B). While intracellular calcium influx is expected to increase in both axons and soma upon stimulation, GCaMP6f changes are typically measured in the soma, where the signal is stronger. Under basal conditions (0–2 min), few neurons exhibited calcium spikes with peaks above the threshold value (Figure 3B, left, dotted line). The subsequent stimulations with capsaicin, agonist of sensory neurons (2–4 min), or with KCl, which induces a global calcium influx (4–6 min), demonstrated a significant increase in the number of responsive neurons (Figure 3B, right).

Figure 3.

Two-dimensional (2D) assays probing neuronal or cancer cell behavior, done in DACIT

(A) Stills from a time-lapse recording of GCaMP6f-expressing adult DRG neurons plated in the neuronal compartment of DACIT. Scale bars, 50 μm.

(B) GCaMP6f signal over time, as measured in neuronal soma without additional stimulation (gray background) and after stimulation with capsaicin (Cap, 1 μM, orange background) and KCl (50 mM, blue background). The violin plot (right) shows the mean and quartiles of the relative number of responsive sensory neurons (ΔF/F₀ > 3) from 10 fields of view in two independent experiments, done with different primary cultures. Statistical analysis was performed using the Mann-Whitney test; ∗p < 0.05.

(C) Standard experimental timeline of 2D assays in DACIT. Cancer cells were plated once axons populated the axonal compartment. (D) 4T1 breast cancer cells (magenta, F-actin; green, E-cadherin) plated on 2D gelatin interact with axons (white, anti-PGP9.5) in the axonal compartment of DACIT. Scale bars, 25 μm.

(E) Schematic of the gelatin degradation assay. Cancer cells (magenta) are plated on a fluorescent gelatin layer (green). Invadopodia assembly and maturation is followed by local degradation of the gelatin, visible as black spots in a fluorescent gelatin layer.

(F) Gelatin degradation puncta (black) in human breast cancer BT-549 cells (magenta, F-actin) cultured in the axonal compartment of DACIT. The yellow dashed box outlines the zoomed-in area (right). F-actin-positive invadopodia (yellow arrows, upper) colocalize with gelatin degradation puncta (magenta arrows, lower). Scale bars, 25 μm.

(G) Degradation areas per cell are similar in DACIT and in 48-well plate. Measurements were done in breast cancer cells MDA-MB-231. Data from 30 fields of view per condition are shown. Two experiments were analyzed for DACIT and three experiments for 48-well plate. Statistical test: Mann-Whitney test shows no significance, ns.

Video S1. Calcium dynamics measured in the neuronal compartment of DACIT

Time-lapse recording of GCamMP6f-expressing sensory neurons monitored at high-speed (14 Hz, 0–105 s long recording). Scale bars, 50 μm.

Next, we tested DACIT’s ability to visualize axon-cancer cell interactions. Primary sensory neurons were seeded into the neuronal compartment and allowed to extend axons into the axonal compartment for 4 days before the addition of cancer cells (Figure 3C). Using this strategy, we were able to visualize axon-cancer cell contacts and the expression of specific proteins during their interaction. In Figure 3D, axons labeled with the neuronal marker PGP9.5 (white) are seen interacting with cancer cells labeled for F-actin, E-cadherin, and DAPI (nuclei). In addition, to monitor cancer cell attachment and motility in the presence of axons, we have performed a 20-h time-lapse (Video S2). While axons are visible by brightfield microscopy, the use of cell trackers or retrograde tracers can further facilitate axon tracing.³⁹

Video S2. Cancer cells and axons interacting in the axonal compartment of DACIT

Time-lapse recording of 4T1 cancer cells spreading in the presence of axons, low-speed recording (4 images/hour, 20 h). Scale bars, 50 μm.

To evaluate performance of DACIT in 2D assays requiring ECM, we measured invadopodia-mediated gelatin degradation. This assay assesses the ability of invadopodia protrusions assembled by cancer cells to degrade ECM. Fluorescent labeling of gelatin facilitates visualization and quantification of high-contrast dark spots, limited to the locations of invadopodia degradation⁴⁰ (Figure 3E). We coated the axonal compartment with fluorescent gelatin, followed by poly-L-lysine (PLL) and laminin, to enable axonal attachment. After loading with cancer cells followed by 24-h incubation, we observed F-actin puncta characteristic of invadopodia (Figure 3F, right, yellow arrows), often colocalized with degraded gelatin (Figure 3F, left, magenta arrows). To validate that the level of degradation was not affected by confinement in the small DACIT compartment, we ran the assay in a glass-bottom plate in parallel (Figure 3E). Comparison of degradation areas confirmed that gelatin degradation levels were similar.

3D spheroid invasion assay in DACIT

Finally, we tested the 3D tumor spheroid invasion in DACIT. Cancer spheroid invasion assays account for parameters such as cell-cell, cell-ECM interactions and hypoxia, mimicking in vivo tumor conditions better than 2D cultures.^41,42

We generated 50-cell spheroids using the hanging-drop technique,⁴³ embedded in collagen/matrigel ECM, and loaded the mixture into the axonal compartment of DACIT (Figure 4A). To validate the rate of spheroid invasion in DACIT and the appropriate response to matrix metalloproteinase (MMP) inhibitors,⁴⁴ we treated spheroids with GM6001 and collected images at 2 h (Day 0), 24 h (Day 1), and 48 h (Day 2) after embedding into DACIT.

Figure 4.

3D Spheroid invasion assay in DACIT

(A) Experimental timeline of 3D spheroid invasion assays in DACIT. Tumor spheroids were added after axons reached the axonal compartment.

(B) Representative images of 4T1 spheroids embedded in a collagen: Matrigel mix. Spheroid invasion was monitored in the presence of DMSO (control) or GM6001 (25 μM) at 2 h post-embedding (Day 0), 24 h (Day 1), and 48 h (Day 2). Scale bars, 200 μm. Graphs show the mean ± SEM relative spheroid area (C) and minimum-to-maximum circularity (D) of nine spheroids per treatment from two biologically independent experiments. Statistical differences (∗p < 0.05 compared with control) were determined using an unpaired, two-tailed Mann-Whitney test.

(E) Top (left) and side (right) views of tumor spheroids (4T1-mScarlet, magenta) embedded in a 3D ECM mix of collagen: Matrigel (blue) in the presence of neurons (yellow), fixed 24 h after invasion. Scale bars, 200 μm.

(F) Brightfield image of a spheroid embedded in 3D ECM, 24 h post-embedding. Microgrooves are visible on the left side of the image. Scale bars, 200 μm.

(G) Maximum projection of invasive strands from a 4T1-mScarlet 3D spheroid (magenta), 24 h post-embedding, interacting with axons (yellow, PGP9.5). Scale bars, 100 μm.

(H) 3D Imaris reconstruction and (I) single z-slice of an invasive strand in the spheroid (magenta) interacting with axons (yellow). Scale bars, 50 μm.

Consistent with results from standard 3D invasion assays,⁴⁴ spheroids in DACIT initiated invasion after 24 h and formed invasive strands by 48 h (Figures 4B–4D). As expected, spheroids treated with GM6001 extended slower than untreated ones (Figure 4C) and maintained their circularity compared with untreated spheroids (Figure 4D).

Using fluorescently labeled collagen, we confirmed that as the invasion proceeds on Day 2, spheroids remain fully embedded in the 3D ECM, within the axonal compartment of DACIT (Figure 4E). The location of the spheroid allowed for free invasion in all dimensions, without limits posed by microgroove edges or top and bottom of the compartment (Figure 4E). Multiple axons were contacting cancer cells within invasive strands, as seen in the maximum projection image (Figures 4F and 4G). By analyzing 3D reconstruction (Figure 4H) and single z-slice through the spheroid (Figure 4I), we certified the presence of direct physical interactions between axons and cancer cells.

Overall, DACIT enables monitoring 2D and 3D culture of cancer cells and neurons in the absence or presence of cancer cell-axon interactions. DACIT allows both cellular and fluidic compartmentalization, enabling different culture media or exposure to different treatments in the neuronal and axonal compartment.

Discussion

Recent studies have shown that the peripheral neurons interact with cancer cells in tumors and influences tumor progression.^7,45,46 However, studying cancer cell-axon interactions in a controlled system remains challenging. In the lack of appropriate peripheral neuron models, most current in vitro studies currently rely on primary cultures. Previous approaches have involved co-culturing cancer cells with either dissociated neurons⁵ or intact ganglia.^47,48 However, primary cultures contain non-neuronal cells such as Schwann cells, which can influence cancer cell behavior.^13,14 These cells are difficult to eliminate, complicating co-culture studies.

Microfluidic devices help overcome some of these challenges by isolating axons and enabling media compartmentalization. For example, Taylor et al. used hard- and soft-lithography to design a two-compartment device separating neuronal soma from axons, allowing axon treatment and recovery for analysis.²⁹ Commercial versions and later modifications have yielded platforms for studying neuronal activity under different treatments,^24,31 analyzing synaptic networks,²⁵ and examining axon interactions with other cells, such as during myelination^27,30 or in cancer contexts.²⁸ However, none of the mentioned devices allow studies of axon-cell interactions within a complex 3D model. Although some allow cell embedding in 3D ECM, they are limited to heights <200 μm. This limitation makes them unsuitable for monitoring invasion and growth of tumor spheroids or organoids, which require heights >300-μm high microchannels to freely invade in 3D over multiple days.⁴⁹

Here, we present a microfluidic device for axon-cancer cell 2D and 3D interaction testing (DACIT). Compared to the original two-compartment device,²⁹ DACIT incorporates two important modifications: (1) our SU-8 master production strategy results in 300–700 μm high macrochannels that can accommodate expanding tumor spheroids, and (2) PDMS and coverslip bonding is achieved by oxygen plasma treatment, which provides a durable seal, allowing multi-day monitoring or time-lapse imaging. These modifications enable analysis of neuronal and cancer cell activities in 2D or 3D over extended periods.

Fabricating an SU-8 master mold with a high aspect ratio (3 μm microgrooves and >300 μm channels) is a challenging task in lithography due to the potential structural collapse.⁵⁰ Several approaches, such as e-beam, X-ray, and two-photon lithography, as well as optimizing post-exposure and soft-baking temperatures, have been developed to address this problem. However, these approaches are often associated with drawbacks such as high costs, time consumption, and low temporal stability.^{51,52,53,54,55} Using a multilayer strategy with the SU-8 photoresist 3000 resin, we fabricated channels up to 700 μm in height with costs and processing times comparable to those of conventional SU-8 masters.

To study axon-cancer cell interactions over multiple days, we used oxygen plasma treatment to bond PDMS devices to glass coverslips. Many PDMS devices use bonding with PLL, which facilitates recovery of axons or molecular material.^19,20 However, this approach poses limitations as it allows PDMS detachment. Even a smallest detached area of PDMS allows for highly motile cells such as cancer cells, or Schwann cells, to migrate along axons, through the microgrooves and reach the neuronal compartment. Plasma bonding, in contrast, permanently seals the compartments, preventing cell crossover. While this complicates molecular recovery, it enables long-term monitoring of axon-cancer cell interactions.

We validated DACIT for imaging axon-cancer cell interactions in both fixed and live samples. Since time-lapse imaging of unsealed microfluidic devices can be lead to evaporation, to maintain optimal humidity, we designed custom 3D-printed holders compatible with lids from microplates, or 35-mm culture dishes. For long-term imaging, we also developed an aluminum holder, which has a thermal conductivity ∼400 times higher than standard 3D-printing polymers (235 vs. 0.5 W/mK). Additional measures, such as topping culture media with mineral oil, or surrounding the DACIT with PBS-filled containers inside the environmental chamber, further reduce evaporation.

Notably, despite the high density of the 3D ECM and the presence of primary culture debris in the soma compartment, channel obstruction—commonly reported in microfluidic devices—was not an issue in DACIT. Likewise, we did not observe any negative impact on cell survival when cultured in DACIT (Video S2). Previous studies have suggested that increased channel height in microfluidic devices improves cell survival, likely by allowing greater medium volume and reducing shear stress.³⁰

Similar to other microfluidic devices,^20,56,57 fluidic isolation in DACIT depends on the differential volumes within macrochannels, allowing either compartmentalization or free diffusion. This feature has previously been used to test axon-regenerating molecules^19,20 in one compartment, or to culture different cell types in the two compartments.^26,27 Taking advantage of this property, we used DACIT to retrograde-trace sensory neurons.³⁹

We demonstrated DACIT’s utility by seeding either dissociated embryonic or adult murine DRGs in the neuronal compartment and murine cancer cell lines, either in 2D, or as 3D spheroids embedded in ECM, in the axonal compartment. The device’s versatility, however, is not limited to murine models. The neuronal compartment can also be seeded with intact 3D DRGs (murine or patient-derived), or with 2D cells or 3D neurospheres of human neurons derived from iPSCs. On the other hand, the axonal compartment can host 3D patient-derived cancer organoids or be complexified by addition of non-cancer cells, such as immune cells or fibroblasts. This flexibility opens opportunities for translational applications, including drug efficacy testing.

While axon bundles and even single axons are sometimes visible using brightfield microscopy (Video S2), axon labeling facilitates visualization and assures all axons will be appropriately segmented and quantified. To image live axons in DACIT, retrograde tracer dyes can be directly added to the axonal compartment, following successful axonogenesis.^39,58 Additionally, adeno-associated viruses (AAV) expressing fluorescent proteins under neuron-specific or ubiquitous promoters can be added to the primary culture few days prior to plating in DACIT.⁵⁹ Timeline of the expression of fluorescent proteins via AAVs is both culture- and virus-specific and has to be carefully measured, as the primary cultures can only be viable up to a week.

Other microfluidic devices have been developed to analyze neuronal activity in the presence of other cell types. For instance, van der Moolen et al.²⁴ integrated microelectrode arrays (MEAs) into a microfluidic system, enabling high temporal resolution of neuronal activity. However, this required specialized equipment and high cost of fabrication. In DACIT, we demonstrated neuronal activity assessment using fluorescent calcium indicators. Compared to MEAs, calcium imaging provides slightly lower temporal resolution but can provide spatial resolution at single-cell level.

To demonstrate DACIT usefulness in studying cell-axon interactions in 2D and 3D ECM context, we performed invadopodia-mediated gelatin degradation assay in 2D followed by 3D spheroid invasion assay. Due to differential expression of integrins, matrix composition affects neuron and cancer cell behaviors differently.⁶⁰ Neuronal attachment is commonly achieved by use of laminin, while cancer cell invasion typically requires rigid, fibrillar ECM.⁴² To promote axonogenesis while maintaining cancer cell invasiveness, we slightly modified standard invasion assays. In gelatin degradation assays, standard fluorescent gelatin was overlaid with PLL and laminin. Similarly, in 3D spheroid invasion, standard ECM consisting solely of collagen I was mixed with laminin-containing matrigel, to encourage axonal extension and interaction with invading strands in spheroids.

In summary, we demonstrate that DACIT enables visualization and experimental manipulation of axon-cancer cell interactions in a biologically relevant ECM context. DACIT anatomically mimics axonogenesis and interactions between peripheral neurons and cells in peripheral organs and it can be used to study interactions between a wide range of peripheral neurons and cancer cells in both 2D and 3D. Furthermore, DACIT allows assessment of neuronal effects on cancer cell adhesion, morphology, migration, and invasion, as well as cancer cell effects on axon outgrowth and neuronal activity. DACIT can be further adapted to test drug effects on neuron-cancer cell interactions or to explore interactions between neurons and other cell types, including immune, endothelial, or muscle cells. Recent advances in differentiating hiPSCs into functional peripheral neurons^16,61,62 provide opportunities to utilize human neurons with patient-derived organoids in DACIT, creating a powerful platform for studying human neuron-cancer interactions.

Limitations of the study

The DACIT device can accommodate spheroids or organoids up to 700 μm in diameter. While increasing the channel height is theoretically possible by adding more coating cycles during master fabrication, greater heights complicate alignment with the microgroove layer and may compromise master stability. In addition, while permanent bonding ensures stable compartmentalization for several days, it reduces the efficiency of recovering cells, proteins, or RNA for downstream analyses. The number of neurons that can be stimulated through the axonal compartment is also limited to those whose axons extend into this region, representing only a fraction of the total neuronal population. Finally, if 4D imaging is required for periods longer than 2 days, a syringe pump is necessary to prevent the DACIT from drying out.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Bojana Gligorijevic (bojana.gligorijevic@temple.edu).

Materials availability

All materials will be shared by the lead contact upon request.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request
• •This paper does not report original code.
• •Other items will be shared by the lead contact upon request.

Acknowledgments

We thank the following funding sources for their support: NIH NCI R01 CA230777; American Cancer Society Research Scholar grant 134415-RSG-20-34-01-CSM, DOD BCRP Breakthrough Award BC230197, PA-CURE, and W. W. Smith Charitable Trust awards to B.G., METAvivor Early Career Research Award to I.V.Q., and 5R01 NS094402 from NIH NINDS to G.M.T. This work was carried out in part at the Singh Center for Nanotechnology, part of the National Nanotechnology Coordinated Infrastructure Program, which is supported by the National Science Foundation grant NNCI-2025608. We also thank Dr. Eric Johnston from Nanotechnology Singh Center for his help with fabrication of the SU8 master, Fox Chase Cancer Center Machine shop, and Peter Lelkes lab for the kind gift of PC12-GFP cells.

Author contributions

Conceptualization, I.V.-Q., V.A., E.T., and B.G.; sample preparation, I.V.-Q., K.A., V.A., E.B., N.H., X.Z., and S.K.; data acquisition, analysis, and interpretation, I.V.-Q., K.A., V.A., E.B., S.K., and B.G.; writing, I.V.-Q., K.A., V.A., E.B., E.T., N.H., X.X., G.T., and B.G.; supervision, E.T., G.T., and B.G.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-PGP 9.5 (mouse monoclonal). Working dilution 1:100	Bio SB	Clone: BSB-46; RRID:AB_3720319
Anti-PGP9.5 antibody (rabbit polyclonal). Working dilution 1:200	Abcam	Cat# ab15503; RRID:AB_301912
Anti-Keratin K8/K18 (guinea pig polyclonal, serum) Working dilution 1:200	PROGEN	Cat# GP11; RRID:AB_2904125
Anti-Sox-10. Working solution 1:100	Abcam	Cat# ab180862; RRID:AB_2721184
Anti-E-cadherin. Working solution 1:100	Abcam	Cat# ab40772; RRID:AB_731493
Chemicals, peptides, and recombinant proteins
100 mm Silicon Wafer	Waferpro, USA	Product #: C04203
SU-8 (2005, 2100, 2150, 3050, 3005)	Kayaku Advanced Materials, Inc., USA	NA
SU-8 developer	Kayaku Advanced Materials, Inc., USA	Cat# NC9901158
Disposable aluminum dishes	KichenDance.com	Cat#80063
Smooth-cast 310	www.smooth-on.com	NA
Sylgard™ 184 Silicone Elastomer, Electron Microscopy Science, Sylgard™ 184, Flash Point = 121°C (250°F)	Electron Microscopy Sciences	Cat#24236-10
Acetone	Fisher Scientific	Cat#A18-4
Isopropanol	Fisher Scientific	Cat#S25371A
Poly-L-lysine solution,0.1% (w/v) in H2O	Sigma-Aldrich	Cat#P8920
Matrigel® Basement Membrane Matrix	Corning	Cat#356234
HBSS (+): HBSS with Ca2+, Mg2+ (without phenol red)	Gibco	Cat#14025076
Dextran, Alexa Fluor™ 680; 3,000 MW, Anionic	Invitrogen	Cat#D34681
HBSS (−): HBSS without Ca2+, Mg2+ (without phenol red)	Gibco	Cat#14170161
HEPES 1M	Gibco	Cat#15630-080
Collagenase P	Sigma-Aldrich	Cat#11249002001
Trypsin - 0.05% EDTA	Life Technologies	Cat#25-300-120
Horse serum. Heat inactivated	Atlanta	Cat#S12150
Deoxyribonuclease I [15000 units in 6 mL HBSS]	Worthington	Cat#LS002139
Cell strainer mesh 40 μm	Corning	Cat#431750
Glass Pasteur pipets	Fisherbrand	Cat#13-678-20C
Neurobasal™-A Medium	Gibco	Cat#21103049
B-27™ Supplement (50X), serum free	Gibco	Cat#17504044
Glutamax	Gibco	Cat#35050061
Penicillin-Streptomycin (10,000 U/mL)	Gibco	Cat#15140122
FdU	Sigma-Aldrich	Cat#PHR2589
Neural Growth factor (Ngf 2.5S Native Mouse Protein)	Invitrogen	Cat#13257019
Molecular probes CellTracker; Red CMTPX	Invitrogen	Cat#C34552
DMEM (Dulbecco’s Modified Eagle’s Medium)	Corning	Cat#10-013-CV
Fetal Bovine Serum	Atlanta Biologicals	Cat#S11550
Nutragen® Type I Collagen Solution, 6 mg/mL (Bovine)	Advanced Biomatrix	Cat#5010-D
Methyl cellulose	Sigma	Cat#M6385
Lyophilized CF® Dye SE or Biotin SE (3 vials)	Biotium	Cat#92211A
Corning™ Collagen I, High Concentration, Rat Tail	Corning	Cat#354249
DPBS (10×), no calcium, no magnesium	Corning	Cat#14200166
DMEM (Dulbecco’s Modified Eagle’s Medium)	Corning	Cat#10-013-CV
Fetal Bovine Serum	Atlanta Biologicals	Cat#S11550
Paraformaldehyde, 16% w/v aq. soln., methanol free	Alpha Aesar	Cat#43368
Bovine Serum Albumin	Sigma-Aldrich	Cat#A4503
Phalloidin dye	Life Technologies	Cat# A22283
Glutaraldehyde	Sigma	Cat#G5882
Glycine	ThermoScientific	Cat#J62407.22
Laminin	Sigma	Cat#L2020
GM6001 (Item No. 14533)	Cayman Chemical company	Cat# 14533
AAV5-syn-GCaMP6f (pAAV.Syn.GCaMP6f.WPRE.SV400)	Addgene	Cat#100837-AAV5
Experimental models: Cell lines
4T1-mScarlet cells	Laboratory of Bojana Gligorijevic	NA
Primary Mouse Sensory Neurons (PMSNs)	Balb/CJ female 6-8 week-old mice	The Jackson Laboratory
Rat embryos (E16)	NA	NA
Other
Mask aligner	ABM USA	ABM3000HR
Spin coater	Brewer Science	CEE 200× Precision Spin Coater
Resist	Microchem	NA
Oven	Fisher Scientific	NA
Biosafety cabinet	NuAire	Labgard Class II Type A2
Biopsy punches (6 mm)	Miltix	MIL3336
Coverslips (24mm) 24x24mm No.1	Globe Scientific Inc	Model#1405-10
Sonicator	Fisher Scientific	FS-6
Plasma machine	PLASMA ETCH	PE-25 Series Plasma System
Hot plate	Fisher Scientific	Isotemp 11-100-49SH
Stereo microscope for DRG recovery	Leica	A60
Centrifuge for cancer cell culture	Eppendorf	Centrifuge 5430R
Centrifuge for PMSN dissociation	Eppendorf	Centrifuge 5804R 15amp version
10 cm culture dishes	Corning	Product number: 353003
3.5 cm culture dishes	Corning	Product number: 353001
CO2 Incubator	ThermoScientific	Napco Series 8000 DH
Confocal Microscope	Olympus	Fluoview1000-MPE
Widefield Microscope	Nikon	Eclipse Ti2
Environmental chamber	Tokai Hit, Japan	Model STXG-WELSX-SET

Experimental model and study participant details

Authenticated, mycoplasma-free BT549 cells were obtained from the cell bank at Fox Chase Cancer Center. Authenticated, mycoplasma-free 4T1 cells were obtained from ATCC. Cell authentication included genetic evaluation using a human 9-marker STR profile and an interspecies contamination test. All cell lines were tested for mycoplasma annually and whenever changes in cell proliferation rates were detected. See “Cell culture” section for details of maintenance.

Primary murine DRG cells were obtaining from Balb/CJ female 6-8 week-old mice (The Jackson Laboratory, stock number 000651). Primary rat DRG embryonic cells were collected from E16 embryos from pregnant Crl:CD(SD) rats (Charles River, strain code 001). Mice and rats were sacrificed by isoflurane overdose, followed by cervical dislocation and DRG cells collected from adult mice or rat embryos as described at the “Sensory neuron dissection and dissociation” section.

To match with the gender of the 99% of the breast cancer cases, only female mice were utilized. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Temple University (Protocol # 5072).

Method details

Design and fabrication of DACIT mask and master

DACIT was designed utilizing LayoutEditor and consists of two layers. The first layer contains microgrooves which connect left and right channels; the second layer was used for generating the channels, which house neurons on the left, and axons and cancer cells on the right side. Microgrooves and channel chrome masks (5″ × 5″ × 0.090″ Quartz Plate, QZ) were produced by Compugraphics (Waterbury, CT).

Microgroove layer fabrication using SU-8 photoresist

To fabricate a very thin layer with strong attachment properties, we used SU8-3000 series photoresist (KAM-SU-8-3000, Kayaku Advanced Materials Inc., Waterborough, MA). SU8-3050 stock was diluted to the viscosity of SU8-3005 using the formula m1v1=m1v2 and spin-coated. The wafer was soft-baked for 2 min at 95°C and exposed to 100 mJ/cm² of UV light in the mask aligner (ABM 3000HR Mask Aligner, ABM-USA, San Jose, CA). The exposed layer went through two post-exposure bakes: one for 1 min at 65°C and one for 1 min at 95°C. The layer was developed after cooling down, using SU-8 developer (Kayaku Advanced Materials Inc., Waterborough, MA).

Channel layer fabrication using SU-8 photoresist

A chrome mask (5″ × 5″ × 0.090″ QZ) or a transparency 10 000 dpi film (ArtNet Pro, San Jose, CA) were used. To fabricate a thick layer, SU8-2100 stock was spin-coated on the wafer without dilution, with the alignment marks covered using duct tape. Duct tape was removed before the 2^nd soft bake, which was done at 95°C for 10 min. Next, a second layer of SU8-2100 was added and spin-coated on top of the first, to increase the channel height. The 3^rd soft bake at 95°C for 30 min was done, and finally the wafer with a thick layer of baked SU8-2100 was placed in mask aligner (ABM 3000HR Mask Aligner, ABM-USA, San Jose, CA). The two alignment marks on photomask and wafer were carefully aligned utilizing x, y, and δ knobs, and the layer was exposed to 300 mJ/cm² of UV light. The exposed layer went through post-baking for 5 min at 65°C and 20 min at 95°C. After cooling down, the wafer with SU8 was developed for 30 min with constant swirling in SU8 developer. Finally, the master was washed with acetone, isopropanol, and water in order and dried with air flow. The cleaned master was hard-baked at 150°C for 10 min for thermal stability.

DACIT fabrication

DACITs are made from polydimethylsiloxane (PDMS) using the silicone elastomer (Sylgard 184, Sigma-Aldrich, Burlington, MA). Briefly, the elastomer base and the curing agent (Electron Microscopy Sciences, Cat#24236-10) were mixed at a 1:10 ratio, poured on the SU-8 master, and cured overnight at 65°C. The next day, the cured PDMS was peeled off, individual DACITs were cut out using scalpel and wells were punched out using 6 mm diameter tissue punches (Robbins instruments, Cat#RBP-20). Individual DACIT devices were plasma bonded to clean 24 x 24 mm coverslips (Globe Scientific Inc, Cat #1405-10). DACITs were sterilized, first by a wash with 70% ethanol, and then by a 15 min exposure to the biosafety cabinet UV light. Next, DACITs were coated with 50 μg/ml Poly-L-lysine overnight, followed by 30 min incubation with 1:10 Matrigel® Basement Membrane Matrix (Corning, Cat#356234) diluted in PBS.

Sensory neuron dissection and dissociation

In adult mice, DRG dissection and dissociation were performed as described previously.⁶³ Briefly, female mice were sacrificed, the spinal cord was removed, and maximum number of DRGs (∼30) per mouse were recovered. Recovered DRGs were digested with 5 mg/ml Collagenase P (Sigma-Aldrich, Cat#11249002001) for 45 min at 37°C, followed by a 5 min digestion by trypsin/EDTA at 37°C. Trypsin was inactivated with 10% horse serum (Atlanta, Cat#S12150) in a Neurobasal medium (Gibco, Cat# 21103049), and cells were disaggregated using a polished pipet. Cells suspension was filtered through a 40 μm cell mesh strainer (Corning, Cat#431750) to remove cell clusters and centrifugated through a 10% bovine serum albumin (BSA, Sigma-Aldrich, Cat# Cat#A4503) layer, to remove debris and non-neuronal cells. Cells were concentrated by centrifugation, resuspended at 125,000 cells/ml in adult DRG media and plated in the DACITs.

In rat embryos, DRG dissection and dissociation was done at E16. Female and male embryos were removed and DRGs were collected and placed on cold HBSS. Embryonic rat DRGs were first digested with Collagenase (10kU/ml) for 15 min at 37°C. Collagenase was then inactivated with fetal bovine serum (FBS), and digestion with trypsin was performed for 15 min at 37°C. Trypsin was inactivated with FBS and washed out by centrifugation. Dissociated DRGs were triturated on neurobasal media using pipet tips (P1000 and then P200) before quantification. Cells were resuspended in embryonic DRG media and loaded into DACIT as described.

Cell culture

All cells were maintained at 37°C with 5% CO₂ in a medium with 100 U/ml of Penicillin-Streptomycin (Gibco, Cat#15140122). Human breast cancer cell line BT549 and murine mammary cancer cell line 4T1 were grown in cancer media, consisting of DMEM (Gibco, Cat#10-013-CV) with the addition of 10% fetal bovine serum (FBS, Atlanta Biologicals, Cat#S11550). Neuronal PC12 cell line was maintained in DMEM medium with 7.5% FBS, and 7.5% horse serum and differentiated with adding 25 ng/mL of Neural Growth factor (NGF, Invitrogen, Cat#13257019). The primary culture of adult neurons was maintained in adult DRG media, consisting of neurobasal medium (Gibco, Cat# 21103049) complemented with B27 (Gibco, Cat# 17504044), and Penicillin-Streptomycin. Primary culture of embryonic sensory neurons was maintained in embryonic DRG media, consisting of neurobasal media supplemented with B27 (Gibco, Cat#17504044), Glutamax (Gibco, Cat#35050061), FdU (Sigma-Aldrich, Cat#PHR2589) and 25 ng/ml of NGF. Using the fluidic compartmentalization feature of DACIT, neurons are cultured in the neuronal compartment loaded with DRG media, while cancer cells are cultured in the axonal compartment, in DMEM supplemented with 10% serum. To ensure the device does not dry out, media levels were replenished every 48 hours. Under these conditions, both cancer cells and axons are viable over the length of the experiments, 5+ days.

Cancer cell spheroids were prepared using the hanging drop method as described previously.^43,44 Briefly, 1250 4T1 cells were placed in 1 ml of complete DMEM media (Gibco, Cat# 11965118) containing 4.8 mg/ml methylcellulose (Sigma-Aldrich, Cat# M6385) and 20 μg/ml Nutragen (Advanced Biomatrix, Cat#5010-D). Drops containing 50 cancer cells were placed on the lid of a cell culture dish and incubated 24-36 h at 37°C with 5% CO₂. Properly formed spheroids were selected by morphology and loaded in DACIT.

DACIT coating with ECM and loading with cells

ECM coating for the 2D invadopodia-mediated gelatin degradation assays

Prior to loading with neurons and cancer cells, the axonal compartment was coated with 0.2% fluorescently labeled gelatin, as described in a previous study.⁴⁴ Briefly, the DACIT's axonal compartment was treated sequentially with 1N HCl (Fisher Scientific, Cat#SA48-1), 50 μg/ml poly-L-lysine (Sigma, Cat#P8920), and 0.2% labeled gelatin. Coated devices were cross-linked with 0.2% glutaraldehyde (Sigma, Cat#G5882) and quenched with 50mM Glycine (ThermoScientific, Cat#J62407.22). To promote axonal growth, gelatin layer was further coated with 50 μg/ml poly-L-lysine (Sigma, Cat#P8920) for 20 min at room temperature, washed twice with PBS, and incubated with 0.15 mg/ml laminin (Sigma, Cat#L2020) for 20 min at room temperature.

Loading with neurons

10-20,000 DRG cells resuspended in 10 μl of media were loaded into the neuronal compartment of DACIT. After allowing 20 min for cell attachment, 190 μl or 150 μl of media were added in the neuronal or axonal compartment, respectively. The following day, fresh media containing 50 ng/ml of NGF was added into the axonal compartment, equalizing the volumes of liquid present in the two compartments such that NGF diffuses across the microgrooves. At 3 days in vitro (DIV), media was removed from the axonal compartment, and the compartment was washed twice with HBSS without Ca²⁺, Mg²⁺ (Gibco, Cat#14170161).

Loading with cancer cells

For 2D assays, 7,500 cancer cells in 10 μl of media were loaded into the axonal compartment of DACIT. After 30 min, 50 μl of adult DRG media was added to the neuronal compartment and 140 μl of cancer cell media was added to the axonal compartment. Cells were incubated at 37°C with 5% CO₂ until fixation, replenishing media every 48h to avoid DACIT drying out.

For 3D assays, 50-100 cell spheroids were prepared using 4T1 cells. Spheroids were checked for round morphology using light microscope, washed 3 times with warm DMEM, and embedded in 2 mg/ml of collagen I (Corning, Cat#354249) with 20% of Matrigel (Corning, Cat#356234) on ice to avoid polymerization. Then, spheroids were loaded into the axonal compartment, and DACITs were incubated at 37°C with 5% CO2 until analysis. When required, spheroids were treated with 1% DMSO (control) or with 25μM GM6001 (Cayman Chemical company, Cat# 14533).

Transduction of DRG cells in DACIT

The soma and axonal compartments of DACIT were coated with 50 μg/mL Poly-L-lysine, followed by a 1:20 dilution of Matrigel in PBS, as previously described. Next, 10⁴ DRG cells were plated into the axonal compartment in complete Neurobasal medium. At 2 DIV, AAV5-Syn-GCaMP6f was added to the neuronal compartment at a multiplicity of infection (MOI) of 10⁵ viral genomes per cell. pAAV.Syn.GCaMP6f.WPRE.SV40 was a gift from Douglas Kim & GENIE Project (Addgene viral prep # 100837-AAV5; http://n2t.net/addgene:100837; RRID:Addgene_100837). Expression of GCaMP6f was monitored daily using a Nikon Eclipse Ti2 microscope. At DIV5, GCaMP6f-expressing cells were observed, and 10³ 4T1 breast cancer cells were plated into the axonal compartment of DACIT.

Immunofluorescence

Immunofluorescence of neurons and spheroids in DACIT, was previously described.⁴³ Briefly, media was removed, and the cells were washed with PBS. Cells were incubated with a fixing/permeabilizing solution (4% PFA with 0.5% Triton X-100 in PBS) for 10 min at room temperature. Solution was removed, and cells were incubated with fixing solution (4% PFA in PBS) for 20 min at room temperature. After fixation, cells were washed thrice with washing solution (PBS containing 0.05% Tween-20), and unspecific sites were blocked with blocking solution (1% FBS, 1% BSA in PBS) for 1 hour at room temperature. Blocking solution was removed, and cells were incubated overnight at 4°C with the primary antibodies diluted in the blocking solution. The next day, the primary antibody was removed, and the cells were washed thrice with washing solution for 10 min each at room temperature. Then, cells were incubated for 1 h at room temperature with the secondary antibodies diluted in the blocking solution. After incubation, cells were washed thrice with the washing solution for 10 min each at room temperature. DACITs were stored in a humidity chamber filled with PBS at 4°C.

We used the following primary antibodies: PGP9.5 (BioSB, Cat# BSB 2115. Diluted 1:100 or Abcam, Cat# ab15503. Diluted 1:200); Sox10 (Abcam, Cat# ab180862. Diluted 1:100); CK8/18 (PROGEN, Cat#GP11. Diluted 1:200); E-cadherin (Abcam, Cat# ab40772. Diluted 1:100). Phalloidin dye for F-actin (Life Technologies, Cat# A22283).

Fluidic compartmentalization and free diffusion

Adjusting the media volumes in the soma and axonal compartments allows for the modification of hydrostatic pressure, enabling fluidic isolation in one compartment (compartmentalization).^19,20 The increased hydrostatic pressure resulting from the channel height overcomes the microfluidic resistance in the microgrooves. Consequently, matching the media volumes between compartments results in media diffusion through the microgrooves.

Compartmentalization or diffusion of media were tested by adding 3 KDa of 680-Alexa Fluor Dextran (Invitrogen, Cat#D34681) in the axonal compartment of DACIT, followed by collecting 3D stacks every hour over two days. Time-lapse images were analyzed using Fiji (Imagej2 version 2.14.0/1.54f). Briefly, the intensity decay of maximum intensity projection was corrected for photobleaching. Identical regions of interest were selected in the neuronal and the axonal compartments and the mean intensity signals were plotted over time.

Imaging, quantification and analyses

To mount and image DACITs, custom-designed holders for individual DACITs or 6 DACITs were 3D printed using ABS or CNC-machined using aluminum. Imaging was done using the confocal microscope (Olympus FV12000MPE), with UPLSAPO10X2 and UPLSAPO30XS objectives. For live imaging, DACITs were placed on sterilized aluminum holders and kept at 37°C with 5% CO₂ and 95% humidity using an environmental chamber (STXG-WELSX-SET, Tokai Hit, Japan). To avoid axon damage, imaging was performed using low-laser power.

The length of the axons was tracked using maximum intensity projection of z-stacks in NeuronJ plugin⁶⁴ in Fiji, ImageJ2 2.9.0/1.53t. Axons were measured from the microgrooves to the axonal terminals, or the longest length observed in the image. The 3D spheroid surface was reconstructed using Imaris (Imaris x64 v8.3.1).

GCaMP6f imaging was performed on a resonant scanning confocal microscope (FV3000, Olympus) equipped with a 30× oil-immersion objective (UPLSAPO30XS, 1.05 NA, Olympus) at a frame rate of 5-14 Hz. In some experiments, GCaMP6f⁺ neurons plated in the neuronal compartment were imaged while capsaicin (1 μM) or KCl (50 mM) were added by pipetting into the axonal compartment. GCaMP6f fluorescence dynamics were analyzed in Fiji (ImageJ2, version 2.9.0/1.53t). Fluorescence changes were calculated using the formula ΔF/F₀ = (F(t) − F₀) / F₀. The relative number of responding neurons (ΔF/F₀ > 3) was calculated as the fraction of neurons that showed a response either during baseline imaging or after addition of capsaicin or KCl.

3D spheroids were analyzed in Fiji. Spheroid shape was determined using brightfield and fluorescent channels by magic wand or threshold mask. Area and shape parameters were quantified in Fiji. Statistical analyses were performed using GraphPad Prism 10 (version 10.6.0, build 796).

Quantification and statistical analysis

Statistical analyses and graphs were performed in GraphPad Prism software (Version 10.6.0). Figure 3B: The relative number of responsive neurons after adding capsaicin (+Caps) or KCl (+KCl) was evaluated on three independent biological replicates. Statistical analysis using an unpaired, two-tailed Mann–Whitney test comparing baseline versus +Caps, P value = 0.0217 or baseline versus +KCl, P value = 0.0401, is reported. Figure 3F: The gelatin degradation was analyzed on 30 fields of view per condition. All data points from two (DACIT) or three (48-well plate) replicates were pooled and subjected to a two-tailed Mann-Whitney test. P value = 0.0699 (not significant). Figures 4C and 4D: Spheroid growth from two biological replicates (n≥9 per condition) was monitored. Figure 4C shows the mean ± SEM relative spheroid area. Statistical differences were determined using an unpaired, two-tailed Mann–Whitney test on day 2. P value = 0.0159 (∗). Figure 4D shows the minimum-to-maximum circularity of these spheroids on day 0 and day 2. Statistical differences comparing with control were determined using an unpaired, two-tailed Mann–Whitney test. P value = 0.0217 (∗).

Published: December 26, 2025