Development and validation of a controlled heating apparatus for long-term MRI of 3D microfluidic tumor models

Abstract

Conventional testing of novel contrast agents for magnetic resonance imaging (MRI) involves cell and animal studies. However, 2D cultures lack dynamic flow and in vivo MRI is limited by regulatory approval of long-term anesthesia use. Microfluidic tumor models (MTMs) offer a cost-effective, reproducible, and high throughput platform for bridging cell and animal models. Yet, MRI of microfluidic devices is challenging, due to small fluid volumes generating low sensitivity. For the first time, an MRI of MTMs was performed at low field strength (1 T) using conventional imaging equipment without microcoils. To enable longitudinal MRI, we developed (1) CHAMP-3 (controlled heating apparatus for microfluidics and portability) which heats MTMs during MRI scans and (2) an MRI-compatible temperature monitoring system. CHAMP-3 maintained chip surface temperature at ~37°C and the media inside at ~35.5°C. Enhanced T₁-weighted MRI contrast was achieved in 3D MTMs with free manganese (Mn²⁺) solutions and Mn²⁺ labeled tumor cells.

Graphical Abstract

1 |. INTRODUCTION

Magnetic resonance imaging (MRI) is widely used for cancer diagnostics due to superior soft tissue contrast, nonionizing radiation, and three-dimensional (3D) imaging capability. To visualize malignant masses on MRI, contrast agents are injected intravascularly and circulate within the body to the target site. Clinically used gadolinium-based contrast agents (GBCAs) interact with surrounding water molecules, producing a bright signal on T₁-weighted MRI. Despite widespread use, GBCAs possess several limitations: (1) Many standard GBCAs have low relaxivities of ~4 to 5 mM⁻¹ s⁻¹ at 1.5 to 4.7 T.¹ Lower relaxivity results in a weaker signal, making early detection of small lesions challenging. (2) GBCAs generate constant bright contrast in well-vascularized areas and their lack of targeting causes nonspecific accumulation in benign and malignant tumors.^2–4 (3) GBCAs are retained in the brain and bone of some healthy patients, which raises safety concerns.^5–9

Metal oxide nanoparticles (NPs) such as manganese oxides (MnO, MnO₂, Mn₂O₃, and Mn₃O₄) and iron oxides (Fe₂O₃ and Fe₃O₄) are being explored as alternative contrast agents due to their stronger MRI signal, biocompatibility, biodegradability, and surface customization.^10,11 Novel MRI contrast agents are typically tested in two-dimensional (2D) monolayer cell culture before animal models. In vitro cell culture often does not accurately predict the outcomes from in vivo studies due to a simplistic setup, lacking flow, multicellular interactions, extracellular matrix (ECM), and the endothelial barrier.^12,13 Costly preclinical investigations require trained staff, long study durations, and regulatory approval. In vivo tests provide a dynamic and complex environment, but results show high variability and limit strict parametric testing.

3D microfluidic tumor models (3D MTMs) bridge the gap between cell culture and animal models, creating a platform that is cost-effective, reproducible, customizable, and has high throughput with prolonged imaging time. 3D MTMs mimic in vivo barriers to NP transport, including (1) physiological flow that incorporates drag force and binding affinity of NPs with cells,^14–18 (2) a complete vascular lumen exhibiting the enhanced permeability and retention effect of tumors in vivo,^13,17–20 (3) tumor ECM by surrounding malignant cells in Matrigel^®, and (4) tumor cell membrane uptake. Parameters can be tightly controlled, and chips are ready in a few days after cell seeding, reducing time and labor. Previous studies have utilized fluorescence microscopy of 3D MTMs to measure diffusion, cell uptake, and cell viability after administration of NPs with varying size, charge, surface chemistry, and targeting.^{13–15,18,20,21} NP uptake in 3D MTMs more accurately recapitulated the in vivo outcome, whereas static monolayer cultures and 3D tumor spheroids did not.¹³ Utilizing 3D MTMs for testing novel MRI contrast agents would prove beneficial due to improved prediction of in vivo NP performance.

To our knowledge, 3D MTMs have not been explored as a high throughput platform for screening MRI contrast agents. MRI of microfluidic devices is challenging, due to the small volumes of fluid present which generate low sensitivity. Several studies increased sensitivity using remote MRI of microfluidic chips by measuring signals at the outlet to investigate flow mixing and velocity distribution without live cells.^22–25 Recently, a few groups have measured the magnetic properties of cells in microfluidic chips using MRI or nuclear magnetic resonance (NMR), but none examined 3D MTMs.^26–31 Two groups used MRI contrast agents, but cells were labeled outside the devices with iron oxide NPs^29–31 or Gd-chelates.²⁸

Increasing MRI or NMR sensitivity of cell detection within microfluidic devices required various strategies such as specialized microcoils,^27–31 high field strengths of 11.7T²⁶ to 14T,²⁷ customized acquisition or reconstruction algorithms,^26,28 and >2 mm channel widths.^28–31 We demonstrate for the first time that it is possible to achieve MRI of microfluidic devices at low field strength (1 T) without microcoils. Our approach employs a standard commercially available mouse whole-body volumetric radiofrequency (RF) coil with conventional acquisition and reconstruction algorithms, expanding the utility of microfluidic MRI. Only standard imaging equipment is needed. Furthermore, channel diameters we imaged were 0.2 mm to 1.8 mm, which is smaller compared to previous literature at low field strength (<2.5 T).^28–31

Careful temperature control provides optimal conditions for maintaining cell homeostasis and survival during long-term imaging. To enable long-term MRI of 3D MTMs, we designed a three-part device called controlled heating apparatus for microfluidics and portability (CHAMP-3) which contains: (1) a 3D printed chip holder, (2) a water perfusion apparatus to maintain cell temperature at 37°C, and (3) a self-sustained power source for portability between different image modalities such as MRI and fluorescence microscopy. The chip holder secures the 3D MTM in a horizontal orientation during MRI and contains sufficient reference water signal to improve image quality, while the heating apparatus promotes cell survival and consistency of MRI readings, as the T₁ signal is temperature sensitive.³² The portable power source will be incorporated in future work. We describe in detail the design and application of (1) CHAMP-3 for temperature maintenance and MRI of 3D MTMs and (2) an MRI-compatible temperature monitoring system for measuring the fluid temperature within MTMs, enabling replication of the setup by other researchers. Commercially available 3D MTMs from SynVivo were tested which contained vascular channels connected to a central tumor chamber by a porous interface.

CHAMP-3 successfully maintained the surface temperature of the chip at ~37°C and the fluid inside the chip at ~35.5°C after 1 h of operation. While running CHAMP-3 within the MRI machine, increasing manganese contrast (Mn²⁺) concentrations resulted in a 2-fold to 4.25-fold enhancement in T₁-weighted MRI contrast in 3D MTMs compared to water alone. Future work aims to test MRI NP contrast agent signal generation within tumor and endothelial cells seeded in 3D MTMs. By incorporating 3D MTMs into testing platforms for MRI contrast agents, NP testing can be expedited before preclinical studies, reducing animal burden, saving time, and decreasing cost.

2 |. MATERIALS AND METHODS

2.1 |. Materials

For the complete list of materials for CHAMP-3, the temperature measurement device, and consumables, please see the Supplementary materials Excel file.

2.2 |. SynVivo 3D MTMs

Commercially available idealized co-culture microfluidic chips obtained from SynVivo, Inc. were composed of poly(dimethylsiloxane) (PDMS) plasma bonded to a standard glass slide (see Figure 1A). The device has two outer vascular channels 200 μm in width by 100 μm in height, mimicking small arteries. The 1.8 mm diameter tumor chamber (100 μm in height) mimics small tumors and was selected to maximize the number of MRI voxels. A porous membrane connects the vascular channels to the tumor chamber, enabling contrast agent delivery to the tumor chamber from the vasculature.

FIGURE 1.

3D microfluidic tumor model (MTM) and chip holder setup for magnetic resonance imaging (MRI). (A) Commercialized SynVivo MTM containing a circular central tumor chamber surrounded by connected vascular channels. (B) The chip holder snugly secures the MTM to the top platform via extruded side arms. Tygon tubing connected to the MTM enters and exits the chip holder via small holes below the side arms. (C) The chip holder’s bottom chamber holds the 5 mm diameter reference water tube surrounded by the coiled warm water tubing of CHAMP-3. (D) The whole-body mouse radiofrequency coil tightly surrounds the MTM in the chip holder, reducing chip movement during MRI bed insertion and removal.

2.3 |. MTM chip holder

The 3D printed chip holder was designed to enhance the image quality of acquired MRI scans and the functionality of CHAMP-3. The holder’s top platform is flat to reproducibly position the chip horizontally during MRI scanning. To prevent any movement of the chip during insertion and removal of the MRI bed, the chip was tightly secured to the holder via side arms (see Figure 1B). The chip holder’s bottom chamber was designed to carry a static reference NMR water tube to provide a sufficient ¹H proton signal for a successful scout image. Furthermore, the bottom chamber fully encloses heated Tygon tubing wrapped around the NMR water tube. The wrapped water tubing is in physical contact with the chip platform to improve heat transfer to the MTM and minimize heat loss to the surroundings (see Figure 1C). The chip holder snugly fits inside the mouse whole-body volumetric RF coil to further prevent any chip movement (see Figure 1D).

2.4 |. Design and 3D printing of the MTM chip holder

The chip holder has iterated over several prototypes via SolidWorks and Autodesk Tinkercad software (see Figure 2). The original prototype (see Figure 2A) secured the MTM on the top platform by extruded side arms and housed the reference water tube underneath. Rectangular holes at the top of the design accommodated small Tygon tubing carrying solutions into and out of the chip. The final prototype eased the set-up of CHAMP-3 elements by splitting the holder into two separate interlocking 3D pieces (see Figure 2B). By removing the top platform, the heated Tygon tubing of CHAMP-3 could be easily wrapped around the reference water tube and placed into the bottom chamber. The openings in the side walls of the bottom chamber were closed to promote greater temperature insulation, increasing the heating efficiency of CHAMP-3.

FIGURE 2.

3D models of the microfluidic chip holder for magnetic resonance imaging (MRI). (A) Initial design with extruded side arms held the microfluidic tumor model in place on the top platform. The holder did fit within the MRI coil but raised the complexity of preparing CHAMP-3 and resulted in greater temperature loss from open sides. (B) The final design eased the installation of CHAMP-3 with two interlocking 3D-printed pieces. Warm water tubing was wrapped around the reference nuclear magnetic resonance water tube and secured in the base of the holder which allowed for better temperature insulation through its closed walls. Dimensions indicate the length, width, and height for both designs.

The final chip holder design was created using Autodesk Tinkercad software. The stereolithography (STL) file was sliced using Bambu Studio version 1.7.7.89 and additively manufactured using a Bambu Lab X1 Carbon 3D printer with the largest flat surfaces (chamber top and bottom) on a 55°C heated textured PEI build plate. The holder was made using 13.01 g (4.29 m) of Bambu PLA Basic filament at 220°C using suggested supports. Printer settings included a 0.2 mm layer height, aligned seam position, and a maximum print speed of 258 mm/s. Total print time was 31 min 24 s at a cost of $0.33.

2.5 |. Thermocouple temperature measurement circuit

To validate CHAMP-3’s ability to heat the fluid in the MTM to physiological temperature, a temperature measurement system was constructed that consisted of four components: (1) main thermocouple sensor, (2) reference junction, (3) signal conditioner, and (4) data acquisition (DAQ) system. A T-type thermocouple was utilized in the design with a reference junction compensation.

2.6 |. Thermocouple sensor

Both a K-type and a T-type thermocouple (Minnesota Measurement Instruments) were compared to evaluate their MRI-compatibility when inserted into the fluid outlet of the MTMs during the operation of CHAMP-3 in the MRI machine. As shown in Figure S1, the K-type produced a large, dark magnetic susceptibility artifact that impairs visibility of the vascular channels on the left side of the MTM. The artifact was not present when the T-type thermocouple was used, which is due to its lower magnetic nickel content (T-type: copper/constantan vs. K-type: chromel/alumel). MRI compatibility of T-type thermocouples has been previously demonstrated in gel phantoms.³³ Our study is the first to apply the T-type thermocouple for measuring temperature of small fluid volumes inside microfluidic devices within the MRI environment.

In our final design, a T-type thermocouple was used to measure the temperature of fluid exiting the MTMs in the MRI machine, while the chip was heated to 37°C by CHAMP-3. The thermocouple wires, 0.25 mm in diameter, were covered with perfluoroalkoxy (PFA) insulation. After stripping the insulation from the ends of both wires, the dissimilar metals were joined together at the sensor tip location. The thermocouple functioned by creating a voltage proportional to the applied temperature based upon the Seebeck effect.³⁴ The wires were placed within a 1.575 mm diameter alumina ceramic shield to secure the thermocouple as it passed through the RF coil in the MRI handling bed. The diamagnetic ceramic insulator surrounding the T-type thermocouple wires was also hypothesized to prevent any magnetic field distortions that could be created by the low nickel content (~45%)³⁵ in the constantan wire, as diamagnetic and paramagnetic materials alter the magnetic field in opposite directions³⁶; Figure S1 shows the lack of such distortions when the T-type was used. Implementing an unsheathed thermocouple sensor directly exposed to the fluid provided a fast temperature response with high accuracy. Upon exiting the MRI machine, the thermocouple wires were wrapped several times around a ferrite ring, which removed electromagnetic interference that could be created from the application of RF waves during MRI scan acquisition. The ferrite ring’s purpose was to prevent damage to the temperature measurement circuits and decrease any RF-induced noise in the temperature measurements. Due to the well-contained fringe fields in benchtop MRI systems, the ferrite ring was placed just outside the MRI machine (<15 cm) to avoid magnetization. Different MRI systems and field strengths may require additional distance between the ferrite ring and the MRI machine. The end of the thermocouple wire was secured within a mechanical assembly that was additively manufactured out of PLA and fastened using heat-set inserts and 3 mm bolts. Mechanical components of the thermocouple system implemented a “castle-mount” bracket configuration (see Figure S2) and were designed to provide the structure housing for the thermocouple and accompanying DAQ system.

2.7 |. Reference junction

As the voltage generated from the thermocouple was directly proportional to the temperature difference between the sensor tip and its connection junction, it was necessary to also have an accurate measurement of the reference junction to apply the appropriate offset compensation for high-accuracy temperature measurements. For this reference junction compensation, an isothermal reference junction was connected to the other end of the thermocouple sensor and was in the upper part of the mounting assembly (see Figure S2). To measure the temperature of the isothermal reference junction, a Texas Instrument LM35 linear analog temperature sensor was used to output a voltage linearly proportional to the surrounding temperature. The LM35 was used to measure the ambient temperature which served as the reference junction temperature.

2.8 |. Signal conditioner

The signal conditioning circuit appropriately amplified the inherent low-amplitude input signal from the thermocouple and LM35 temperature sensor to an appropriate 0–5 V across the desired temperature measurement range (i.e., 0–50°C) to be captured by the DAQ system (see Figure 3). To perform this amplification, two AD620ANZ operational amplifiers were implemented for both the T-type thermocouple and LM35 sensor, which are DC-coupled electronic operational amplifiers with a differential input. Each operational amplifier has a different gain for the two components to appropriately scale the measured voltage between 0 and 5 V. In particular, the T-type thermocouple and LM35 amplifiers have a set gain of G_T = 2471 (R_g,T = 20.0 Ω) and G_Ref = 10.158 (R_g,Ref = 5.39 kΩ), respectively. These amplifiers were selected as they offer low noise and input bias current, desirable for use in this system due to the high gain requirement for the thermocouple signal. The resistor type used for setting gain was chosen to reduce resistor uncertainty (i.e., ±1% maximum). To power both AD620ANZ amplifiers, a balanced input voltage of ±12 V was required, necessitating an ICL7662CPA+ inverter. The ICL7662CPA+ is a monolithic charge pump inverter that produces a corresponding negative voltage for a positive voltage signal, that is, the output of −12 VDC from the +12 VDC input signal for the AD620ANZ amplifiers (see Figure 3).

FIGURE 3.

Wiring diagram for the thermocouple signal conditioning and data acquisition system. Power connected to the barrel jack provides 12 V of DC (VDC) voltage for operating the components of the thermocouple measurement system. The ICL7662CPA+ inverts the +12 VDC to −12 VDC before entering the AD620ANZ operational amplifiers. The thermocouple measures fluid temperature exiting the microfluidic tumor model while the LM35 measures the ambient room temperature. Both signals are separately amplified to the 0 to 5 V range by the two AD620ANZ operational amplifiers and measured by the Arduino Nano. The 5 V linear voltage regulator converts the +12 VDC from the barrel jack to +5 VDC to power the Arduino Nano.

2.9 |. DAQ system

The thermocouple DAQ system consisted of an Arduino Nano that could measure the analog voltages from the amplified thermocouple and reference junction signals. To provide the final temperature measured at the thermocouple sensor location, the DAQ processing code provided the necessary reference junction compensation. Temperature conversion from the voltage generated by the thermocouple followed the NIST ITS-90 scale,³⁷ which implements a reference junction at T = 0°C. As the reference junction temperature for this temperature measurement system was taken at ambient temperature (T ≈ 25°C), this reference offset must be considered. For this compensation, the amplified voltages from the thermocouple and LM35, that is, V_T,amp and V_ref,amp, respectively, were first transformed back into their raw voltage values using Equations 1 and 2 below:

$$V_T=\frac{V_{\text{T},\text{amp}}}{G_{\text{T}}}$$ (1)

$$V_{\text{ref}}=\frac{V_{\text{ref,amp}}}{G_{\text{ref}}}$$ (2)

where V_T is the raw voltage and G_T is the gain of the thermocouple and V_ref is the raw voltage and G_ref is the gain of the LM35 temperature sensor. The raw voltages were then added together to become the compensated thermocouple voltage (V_T,comp) using Equation 3 below:

$$V_{\text{T},\text{comp}}=V_{\text{T}}+V_{\text{ref}}$$ (3)

V_T,comp was converted to the corresponding temperature using the NIST T-type thermocouple conversion table.³⁸ The DAQ system and signal conditioning boards were housed in an additively manufactured enclosure made from PLA (see Figure S3). A wire harness was constructed to connect the daughter board from the thermocouple and reference junction to the main board.

2.10 |. 3D printing of thermocouple mount and circuitry enclosure

The thermocouple mount and accompanying circuitry case were designed using AutoDesk Fusion 360, exported into STL file format, and sliced using the PrusaSlicer 2.3.0 program, creating the geometric code for additive manufacturing. The print settings implemented a gyroid in-fill pattern, 0.15 mm layer height and 15% material in-fill. Parts were manufactured using a Prusa iMK3S 3D printer using PLA filament with the temperature of the hot-end extruder being set to 215°C and the hardened steel sheet adhesion bed being pre-heated to 60°C. Combined, the parts took 20 h and 33 min and used 184 g of filament, with a cost of $4.60.

2.11 |. Thermocouple validation

First, the T-type thermocouple was validated against an MRI-compatible thermistor (SA Instruments, Inc) in a heated water bath across a temperature range between T ≈ 30 to 50°C, which was within the expected regime of the fluid during the microfluidic experiment. For this validation, the thermistor and the thermocouple were first both suspended in a beaker containing hot water (starting at T ≈ 49°C), which slowly cooled over time. Temperature measurements were taken continuously using the CoolTerm software (thermocouple) and SA Instruments’ Small Animal Monitoring and Gating System software (thermistor) and manually recorded every 5 min until the water reached the lowest value of T ≈ 30°C. Throughout the test, it was ensured that the thermistor and thermocouple only contacted the water and were not in contact with the beaker walls to ensure measurement uniformity. The values from the thermistor and thermocouple were then compared to one another to create a linear fit (see Figure S4). The linear fit of the two independent measurement sets corresponded to R² values of 0.9981 and 0.9999, respectively, which denoted the expected high linearity from this temperature measurement system compared to the thermistor.

Next, the T-type thermocouple was validated against the MRI-compatible thermistor to measure the surface temperature of MTMs during MRI operation. MTMs were primed with cell culture media and heated to 37°C through CHAMP-3 application as described below. Both the thermocouple and thermistor wires were secured to the chip surface with tape (ensuring no tape contacted the probe tips). Temperature measurements were recorded continuously as above. Once the MTM surface temperature stabilized after insertion into the MRI machine, temperature measurements were recorded manually once every minute over 30 min to assess the accuracy and precision of the thermocouple versus thermistor during MRI acquisition. Percent accuracy and precision were calculated according to Equations (4) and (5) below:

$$\text{\%Accuracy}=100-\left(\frac{\left|{\overline{T}}_{\text{thermocouple}}-{\overline{T}}_{\text{thermistor}}\right|}{{\overline{T}}_{\text{thermistor}}}\right)\times 100$$ (4)

$$\text{\%Precision}=100-\left(\frac{{\sigma }_{\text{T},\text{thermocouple}}}{{\overline{T}}_{\text{thermocouple}}}\right)\times 100$$ (5)

where ${\bar{T}}_{\text{thermocouple}}$ and ${\bar{T}}_{\text{thermistor}}$ are the average temperatures recorded from the thermocouple and the thermistor over 30 min, respectively, and σ_{T,thermocouple} is the standard deviation of the thermocouple temperature measurements over the same time period.

2.12 |. Priming and cell seeding of 3D MTMs

For CHAMP-3 temperature validation studies, MTMs were primed with RPMI cell culture media. For MRI scanning, MTMs were primed with deionized (DI) water for baseline/unenhanced MRI acquisition or various concentrations of Mn²⁺ (12.5–3200 μM) to enhance MRI signals via reducing T₁ relaxation time. A total of eight Tygon tubes 100–150 mm long (0.508 mm inner diameter, 1.524 mm outer diameter) were attached to the inlets and the outlets of the MTMs. A 1 mL luer-lock syringe filled with media, DI water, or Mn²⁺ was attached to a 24-gauge blunt-tip needle and inserted into an inlet tube. The solution was manually perfused through the MTMs, and the corresponding outlet was clasped with a slide clamp after a few drops exited the outlet tube. This procedure was repeated for all channels across the MTMs. A 30-min additional priming of the MTMs via a pneumatic primer (SynVivo cat# 205001) connected to a research-grade nitrogen gas cylinder was performed according to the manufacturer’s instructions to prevent bubble formation in the devices. The primed MTMs were incubated overnight at 37°C prior to temperature validation, mimicking the conditions for cell seeding and growth within the MTMs.

Human luminal A T47D breast cancer cells were donated by Dr. Elena Pugacheva. Cells were cultured in RPMI-1640 medium supplemented with 10% FBS, 1% penicillin–streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, and 4500 mg/L glucose and grown inside a humidified incubator at 37°C, 5% CO₂. T47D cells were unlabeled (negative control) or labeled with 100 μM Mn²⁺ in RPMI medium overnight. Prior to cell seeding, cells were washed in PBS. MTMs were primed with PBS and incubated at 4°C for 30 min. While on ice, a previously prepared 1:5 Matrigel-RPMI serum-free media mixture was used to coat the tumor chamber of the MTMs with a gel-like matrix to encourage cell colonization and entrapment in the tumor chamber. MTMs were buried in ice for a minimum of 1 h to allow for Matrigel coating. MTMs were manually perfused with serum containing RPMI medium, and two small Eppendorf tubes for the waste collection were installed to the outlets of the vascular and the tumor chamber. Under an inverted microscope, 25 × 10⁶ cells/mL of unlabeled or labeled T47D cells were seeded into the tumor chamber with a flow rate of 8 μL/min. After seeding, the inlet and outlet tubing was clamped to prevent cell leakage from the tumor chamber. MTMs were moved to a 37°C incubator for 4 h prior to MRI scanning to promote cell attachment.

2.13 |. CHAMP-3 design for temperature maintenance of 3D MTMs

As discussed previously, the CHAMP-3 system is comprised of three elements: a chip holder, a heated water perfusion device to maintain the MTM at physiological temperature, and a portable power source (see Figure 4). The portable power source is a goal of future work; however, the current CHAMP-3 system is already mobile through placement on a pushcart. A Fisherbrand^™ Isotemp^™ Heated Bath Circulator served as the main heating source and circulated warm water through the CHAMP-3 setup on the fastest flow setting at 55.5–57.5°C (see Figure 4A). Large Tygon tubing (12.7 mm inner diameter, 15.9 mm outer diameter) was connected to the inlet and outlet of the circulating water bath and was downsized to smaller Tygon tubing (1.59 mm inner diameter, 3.18 mm outer diameter) prior to entering the MRI handling bed. The Tygon tubes circulating warm water were wrapped around a 5 mm reference water NMR tube placed at the base of the chip holder (see Figure 4B). The wrapped Tygon tubing was in physical contact with the platform of the chip holder where the MTM was placed, which promoted heat transfer through the PLA, the glass slide, and the PDMS containing the microchannels. Prior to temperature measurements and introduction of the 3D MTM, the CHAMP-3 system was preheated from room temperature to 55.5–57.5°C, which took ~18 min.

FIGURE 4.

Schematic of CHAMP-3 design. (A) A heated circulating water bath perfuses warm water at 55.5–57.5°C into (B) smaller Tygon tubing (blue) wrapped around the reference water nuclear magnetic resonance tube at the base of the chip holder (yellow). (C) The coiled warm water tubing heats the upper platform of the chip holder, maintaining the 3D microfluidic tumor model (MTM) at 37°C. (D) The syringe pump perfuses the imaging solution (pink) at 1 μL/min through the two vascular inlets of the 3D MTM to mimic physiologic flow. (E) A waste tube collects unwanted fluid after flowing through the MTM device. (F) The thermocouple device monitors the temperature of the fluid leaving the MTM through insertion into a vascular channel outlet. (G) The thermocouple wires enter a 3D-printed housing assembly which also measures the ambient room temperature through a separate LM35 temperature sensor (reference junction). (H) The thermocouple and reference temperature signals are amplified and then processed by an Arduino unit that converts the voltage measurements to temperature on a portable computer. (I) The portable computer system utilizes Arduino IDE and CoolTerm software to record the temperature measurements. Figure created with BioRender.com.

After CHAMP-3 preheating, a previously primed MTM was moved from the 37°C incubator to the upper platform of the chip holder and secured through the extruded side arms (see Figure 4C). Two syringes filled with RPMI media were attached to the vascular inlets of the MTM, and the solution was perfused via a Standard Infuse/Withdraw PHD Ultra Syringe Pump (Harvard Apparatus) with a constant physiologic flow rate of 1 μL/min (see Figure 4D). This flow rate was chosen, as it corresponds to approximately 1 dyne/cm² physiologic shear stress.¹⁵ Subsequently, the fluid needed to be collected after flowing through the device into a waste tube (see Figure 4E); therefore, a 1.5 mL Eppendorf tube was taped to the end of an MTM outlet tube and placed in the free space of the MRI bed. The MRI-compatible thermistor taped on top of the MTM monitored the surface temperature of the device. MTM surface temperature measurements were recorded continuously as described above and sampled manually at 5-min intervals for a total of 20 time points. Concurrently, the T-type thermocouple (see Figure 4F) was inserted into the Tygon tubing exiting one of the vascular channel outlets to monitor the temperature of the fluid as it left the MTM. The other end of the thermocouple wires entered through the 3D printed assembly (see Figure 4G) which also housed the room temperature measurement sensor (reference junction). The Arduino unit (see Figure 4H) operated through a portable computer (see Figure 4I) via Arduino IDE software. CoolTerm software accessed the data produced from the Arduino unit and automatically recorded the fluid temperature exiting the MTM continuously for a 1-h 35-min period. Fluid MTM temperatures were also manually recorded every 5 min. Temperature readings of the MTM’s surface and fluid inside were monitored inside the MRI machine during scanning. The final assembled working prototype of CHAMP-3 is shown in Figure 5.

FIGURE 5.

Real images of CHAMP-3 components showing the temperature validation setup in the magnetic resonance imaging (MRI) machine. (A) Heated circulating water bath. (B) Small warm water Tygon tubing enters and exits the MRI handling bed through the catheter port. (C) 3D microfluidic tumor model (MTM) in the chip holder secured inside the MRI handling bed. (D) Syringe pump with two 1 mL syringes to perfuse fluid to the 3D MTM. During MRI scanning, the Tygon tubing connecting the syringes to the MTM is fished through a second port into the MRI machine (marked by *). (E) Thermistor (orange) secured to 3D MTM surface. (F) Thermocouple wire inserted into 3D MTM outlet; a ceramic insulator covers the thermocouple wire as it exits the Tygon tubing within the radiofrequency coil imaging volume. The T-type thermocouple wire (blue) exits the MRI bed through the second port (*), wraps around a ferrite ring (red circle), and enters (G) the 3D-printed housing assembly that also contains the reference temperature junction. (H) Data acquisition and signal conditioning systems. (I) Laptop running Arduino IDE and CoolTerm software to capture thermocouple fluid temperature. The nearby laptop (left) runs SA Instruments’ Small Animal Monitoring and Gating System software to capture the thermistor chip surface temperature.

2.14 |. MRI acquisition of 3D MTMs

3D MTMs were prepared for MRI acquisition by priming different chips with free Mn²⁺ solutions at concentrations of 0, 12.5, 50, 200, 400, 800, 1600, and 3200 μM in DI water. The MTM was primed the same day as acquisition to prevent air bubble formation and secured in the chip holder on the MRI handling bed after the CHAMP-3 system reached equilibrium at 55.5–55.7°C. An unmodified, commercially available mouse whole body volume RF coil (30 mm inner diameter, 65 mm length) standard with the 1.0 T Bruker ICON MRI machine was used. The RF volume coil was fitted over the MTM and the MRI handling bed was inserted into the MRI machine. After the MTM was situated in the MRI, a flow of 1 μL/min of the desired solution was passed through the vascular channels of the MTM using the syringe pump. Since the vascular channels and the tumor chamber are connected via a porous interface, the solutions easily penetrated the tumor chamber during dynamic flow. To enable sufficient water signal for successful acquisition of scout images of the 3D MTM, two additional sources of water were used: (1) As mentioned previously, a water-filled 5 mm diameter NMR tube was placed in the bottom chamber of the chip holder for CHAMP-3. (2) A 2 mm diameter electron paramagnetic resonance (EPR) tube containing the same solution as that in the 3D MTM was placed at approximately the same coronal slice level as the tumor and vascular channels; to achieve this, the small EPR tube was placed next to the PDMS block on the glass slide containing the 3D MTM.

The contrast-to-noise ratio (CNR) of increasing Mn²⁺ solutions was measured from MRI scans acquired using a 2D FLASH sequence: echo time (TE) = 3.713 ms, repetition time (TR) = 300 and 600 ms, field of view (FOV) = 30 mm × 30 mm, slice thickness = 0.8 mm, matrix size = 128 × 128, in-plane resolution = 234 μm × 234 μm, flip angle = 35°, number of excitations (NEX) = 20. Image acquisition time was ~12.5 min for TR = 300 ms and ~25 min for TR = 600 ms. Circular regions of interest (ROI) were defined for the tumor chamber, PDMS and background noise. Images were analyzed with MicroDicom and CNR was calculated using Equation (6) below:

$$\text{CNR}=\left(\frac{{\text{Signal}}_{\text{tumor}\text{chamber}}-{\text{Signal}}_{\text{PDMS}}}{{\sigma }_{\text{background}\text{noise}}}\right)\times 0.66$$ (6)

where σ is the standard deviation of the background noise and 0.66 is the Rayleigh correction factor applied to account for the pixel distribution of the background noise following the Rayleigh distribution.^39,40 Equation (7) was applied to calculate the fold change increase in CNR of Mn²⁺ solutions in the tumor chamber with respect to DI water:

$$\text{CNR}\text{fold}\text{change}=\frac{{\text{CNR}}_{\text{Mn}2+}}{{\text{CNR}}_{\text{water}}}$$ (7)

MTMs seeded with unlabeled or labeled T47D cells were scanned at 1 T using the same MRI parameters and CNR calculation described above. During scanning, 1 μL/min of RMPI media was perfused through the vascular channels of the MTM using the syringe pump.

2.15 |. Statistics

In order to provide evidence that temperature readings of the T-type thermocouple are as accurate as those of the MRI-compatible thermistor, MTM surface temperature readings around 37°C of both thermistor and T-type thermocouple underwent an equivalence test using JMP analysis software. Data was stacked based on temperature probe for N = 3 experiments (n = 30 data points per run). A fit least squares model was created and all pairwise comparisons equivalence tests were run to determine the effect of different runs or probe types on the temperature reading of the microfluidic chip surface. A difference of Δ = 0.2°C between runs and a difference of Δ = 0.120°C between probes were set to be considered zero. Both tests were two-sided and α was set to 0.05.

A Multivariate test was conducted in JMP Pro 17.2.0 for the MTM surface temperature and RPMI media fluid temperature (n = 20 time points each) to examine the correlation between the two temperature measurements in each experiment. Normal distribution analysis and goodness of fit were generated in the same software. Because the data was not normally distributed, nonparametric testing was performed via Wilcoxon/Kruskal–Wallis tests after stacking the data to prove the reproducibility of temperature measurements between experiments using χ² and the p-value. CHAMP-3 temperature measurements were plotted in GraphPad Prism.

A two-tailed t-test with Welch’s correction was performed in GraphPad Prism to evaluate the significance between MRI CNR data at TR = 300 ms versus TR = 600 ms. CNR values at TR = 300 ms and TR = 600 ms were analyzed for MRI acquisitions of MTMs prepared with different Mn²⁺ concentrations versus water. CNR data of N = 3 experiments were compared by variance analysis (one-way analysis of variance with Dunnett multiple comparisons) using GraphPad analysis software. CNR values of MTMs prepared by water served as the control for all group comparisons. α was set to 0.05. All experiments for temperature validation and MRI CNR analysis of Mn²⁺ solutions were run in triplicate. CNR values of MRI acquisition of Mn²⁺ labeled and unlabeled T47D MTMs were evaluated by GraphPad software using an unpaired standard T-test with no further correction in N = 2 labeled and N = 2 unlabeled T47D seeded MTMs. α was set to 0.05.

3 |. RESULTS AND DISCUSSION

3.1 |. Thermocouple validation

To validate the performance of the T-type thermocouple compared to the MRI-compatible thermistor, the MTM surface temperature was measured by both probes over 30 min once the readings stabilized after insertion into the MRI machine. During MRI scanning, the T-type thermocouple achieved 99.7% accuracy and 99.8% precision (37.2 ± 0.08°C) compared to the MRI-compatible thermistor. As the T-type thermocouple generates a continuous voltage signal of 43 μV/°C, the sensor’s temperature resolution depends on the analog to digital sampling of the Arduino DAQ system, which was determined to be 0.05°C, based on the DAQ voltage resolution and corresponding temperature-per-voltage output from the thermocouple.

The equivalence test between N = 3 runs revealed that the performance of the thermistor and T-type thermocouple are reproducible over different experiments with a maximum mean difference of 0.177 between N = 1 and N = 3, and a minimal mean difference of 0.023 between N = 2 and N = 3. The equivalence test between the thermistor and the thermocouple illustrated that both measurements are equivalent with a maximum p-value of 0.013. These metrics provided confidence in the application of the T-type thermocouple for CHAMP-3 temperature validation studies.

3.2 |. Validation of temperature maintenance of 3D MTMs by CHAMP-3

Figure 6 illustrates the validation of MTM surface temperature and fluid temperature exiting the MTM over time after applying CHAMP-3. Before the MTM was placed on the heated chip holder, the CHAMP-3 system reached equilibrium at 55.5–57.5°C over 18 min. For temperature validation testing, the system was operational for a minimum of 2 h for one test per day. Tests were conducted independently three times on different days. Only temperature values at 5-min intervals are shown. A commercially available MRI-compatible thermistor measured the MTM surface temperature, while a T-type thermocouple-based fluid temperature sensor measured the RPMI media temperature inside the chip during MRI scanning. As depicted in Figure 6, CHAMP-3 elevated the MTM surface temperature to ~37°C after 25 min on the heated chip holder. This physiological temperature creates an optimal environment for cell function and survival in the MTM during long-term MRI scanning.

FIGURE 6.

CHAMP-3 temperature validation of microfluidic tumor models (MTMs) inside the magnetic resonance imaging machine. Average MTM surface temperature (red) and average media temperature within the MTM (black) are shown over time after applying CHAMP-3. The chip surface stabilized at ~37°C whereas media inside the chip stabilized at ~35.5°C. Mean temperatures ± standard deviation for N = 3 experiments.

On the other hand, the media temperature inside the MTM after 25 min on the heated chip holder reached 35.5 ± 0.04°C, with minor fluctuations over time. As the media was dynamically flowing through the MTM, it entered the inlets at room temperature and was heated from the surrounding PDMS encasing the microchannels; this explains why the temperature of the fluid exiting the MTM was less than the surface temperature of the chip. Since the fluid temperature is slightly less than body temperature, cell survival will still be promoted while avoiding exposure to hyperthermia conditions that could occur if the temperature approaches 40°C.⁴¹

Datasets of the MTM surface temperatures and the MTM media exiting temperatures of three experiments underwent multivariate testing. The media temperature exhibited a positive correlation of 94.45% with the MTM surface as it increased and maintained a temperature of ~37°C. The normal distribution and goodness of fit model failed to prove that the data was normally distributed. Wilcoxon/Kruskal–Wallis tests revealed that temperature readings of MTM surface temperature and the media temperature remained consistent between the three temperature validation runs with a χ² value of 1.48 and a p-value of 0.477. χ² < 5.991 (DF = 2) suggested that the medians of the three experiments are equal and the temperature range of the MTM surface and media is reproducible in the CHAMP-3 designed system.

3.3 |. MRI of 3D MTMs using CHAMP-3

During the application of CHAMP-3, T₁-weighted MRI scans of 3D MTMs were acquired to estimate how the CNR changed after administration of Mn²⁺ at 12.5, 50, 200, 400, 800, 1600, and 3200 μM over multiple TRs. The dynamic flow of DI water or Mn²⁺ at 1 μL/min within the chip did not negatively impact image quality compared to static imaging of the same solutions (see Figure S5). As shown in Figure 7, the signal within the central tumor chamber increased with higher Mn²⁺ concentrations to enhance CNR, thereby reducing the appearance of background noise. Both the 1.8 mm wide tumor chamber and the 0.2 mm wide vascular channels were visible after addition of Mn²⁺ contrast. However, the highest Mn²⁺ concentration of 3200 μM Mn²⁺ significantly suppressed T₁-weighted MRI signal due to T₂ shortening effects.⁴²

FIGURE 7.

Representative T₁-weighted magnetic resonance imaging (MRI) scans of 3D microfluidic tumor models (MTMs) primed with deionized (DI) water and increasing Mn²⁺ concentrations at repetition time = 600 ms. (A) Schematic of SynVivo’s idealized co-culture microfluidic chips with a central tumor chamber (red) surrounded by vascular channels (blue). T₁-weighted 2D FLASH MRI scans of 3D MTMs perfused with (B) DI water (C) 12.5 μM Mn²⁺ (D) 50 μM Mn²⁺ (E) 200 μM Mn²⁺ (F) 400 μM Mn²⁺ (G) 800 μM Mn²⁺ (H) 1600 μM Mn²⁺ and (I) 3200 μM Mn²⁺. Increased Mn²⁺ concentrations enhanced signal-to-noise ratio (SNR) of the tumor chamber and vascular channels; however, 3200 μM Mn²⁺ significantly suppressed T₁-weighted MRI signal due to T₂ shortening effects. The vertical bright line in some images is due to the standard electron paramagnetic resonance tube; it is present in adjacent slices for the other MTMs.

The CNR of the tumor chamber compared to the surrounding PDMS followed a nonlinear trend (see Figure 8) at increasing concentrations of Mn²⁺, like previously reported CNR studies with MRI contrast agents.^42–44 CNR increased with higher Mn²⁺ concentrations up to 200 μM, plateaued until 400 μM, and then decreased until 3200 μM. The CNR at TR = 600 ms was not significantly greater than CNR at TR = 300 ms. All Mn²⁺ concentrations other than the lowest (12.5 μM) and highest (3200 μM) concentrations significantly enhanced the CNR of the tumor chamber compared to DI water. The CNR at 200 μM Mn²⁺ was ~4.25-fold higher than that of DI water and ~2.4-fold higher than the lower Mn²⁺ concentrations of 12.5 and 50 μM.

FIGURE 8.

Contrast-to-noise ratio (CNR) of Mn²⁺ in water at repetition time (TR) = 300 ms (red) and TR = 600 ms (blue). Average CNR of the tumor chamber versus the surrounding poly(dimethylsiloxane) was measured for deionized (DI) water (0 μM Mn²⁺) and increasing concentrations of Mn²⁺ (12.5, 50, 200, 400, 800, 1600, and 3200 μM) within the 3D microfluidic tumor models. CNR increased with higher Mn²⁺ concentrations up to 200 μM, plateaued until 400 μM, and then decreased until 3200 μM. Mean CNR values ± standard deviation are shown from triplicate experiments. One-way analysis of variance with Dunnett multiple comparisons was performed relative to DI water. For TR = 300 ms: † p < 0.05, ††† p < 0.001. For TR = 600 ms: **p < 0.01, ***p < 0.001.

T₁-weighted MRI was also performed of MTMs seeded with T47D breast cancer cells. As shown in Figures 9 and 10, CNR was enhanced for MTMs containing T47D cells labeled with 100 μM Mn²⁺ versus unlabeled cells. The CNR for MTMs with labeled T47D cells was 1.6-fold higher for TR = 600 ms and 2.0-fold higher for TR = 300 ms compared to MTMs with unlabeled cells; however, the contrast increase was only significant at the shorter TR. Since there was a 4-h delay in between cell labeling with Mn²⁺ and MRI scanning to allow for cell attachment, the labeled cells could have expelled some Mn²⁺ before imaging to reduce the overall observed CNR. Future studies with NPs will involve labeling tumor cells in real-time within the MTMs and tracking contrast increase longitudinally with MRI. As previously mentioned, T₁-weighted MRI is sensitive to temperature, where higher temperatures increase T₁ relaxation times³² to further decrease MRI sensitivity. Our results demonstrate it is possible to successfully capture T₁-weighted MRI images of 3D MTMs at physiological temperature at low field strength without microcoils or specialized equipment, sequences, or reconstruction algorithms.

FIGURE 9.

Representative T₁-weighted magnetic resonance imaging (MRI) scans of 3D microfluidic tumor models (MTMs) seeded with T47D tumor cells. T₁-weighted 2D FLASH MRI scans of 3D MTMs containing (A, B) unlabeled T47D cells and (C, D) 100 μM Mn²⁺ labeled T47D cells with (A, C) repetition time (TR) = 300 ms and (B, D) TR = 600 ms. MRI of Mn²⁺ labeled T47D cells generated enhanced contrast versus unlabeled cells.

FIGURE 10.

Contrast-to-noise ratio (CNR) of 3D microfluidic tumor models (MTMs) seeded with T47D tumor cells at repetition time (TR) = 300 ms and TR = 600 ms. CNR increased for MTMs containing T47D cells labeled with 100 μM Mn²⁺ compared to MTMs containing unlabeled cells. *p < 0.05.

3.4 |. Limitations

This study encountered several limitations that warrant consideration for the interpretation and generalization of the findings. First, T₁ mapping could not be performed due to magnetic field gradient frequency drift at long TRs, which prevented comprehensive characterization of Mn²⁺ within the tumor chamber. As r₁ (relaxivity) of Mn²⁺ in 3D MTMs could not be evaluated, CNR was analyzed instead. In future studies, CNR will be the preferred approach as it is fast enough to provide longitudinal intracellular contrast dynamics over time, with each scan being 12.5 min (TR = 300 ms) or 25 min (TR = 600 ms). Second, older equipment such as the ICON MRI, dating back to the early 2000s, may produce lower-quality images compared to newer systems. In future work, we plan to adapt CHAMP-3 for imaging on the 1.0 T Aspect M7 MRI to compare the image quality. Third, the current setup does not account for the 5% CO₂ necessary for proper pH buffering that occurs inside an incubator. Future studies will explore supplementing cell media with HEPES buffer to control pH in seeded 3D MTMs within the MRI machine, as it is CO₂-independent. Finally, the CHAMP-3 water bath operating temperature will need to be recalibrated following changes to the tubing diameter or length, which highlights a procedural limitation that requires attention to maintain accurate temperature control throughout experiments.

4 |. CONCLUSIONS

In summary, we have developed CHAMP-3, an easy-to-use portable heating system for enabling long-term MRI of 3D MTMs. The MTM’s surface temperature and fluid inside stabilized at ~37°C and ~35.5°C, respectively, when the heated chip holder was secured in the handling bed inside the MRI machine. When compared to a commercial MRI-compatible thermistor, our T-type thermocouple temperature measurement system performed similarly in terms of accuracy and precision, while being more affordable. Further, we successfully acquired MRI images of the tumor chamber and vascular channels of the 3D MTM containing free Mn²⁺ at low field strength (1.0 T) using an unmodified, standard volume RF coil and conventional sequences and reconstruction algorithms. In addition, an MRI of MTMs seeded with T47D tumor cells revealed an increased CNR for Mn²⁺ labeled cells versus unlabeled cells. In conclusion, CHAMP-3 and its future improvements will lead to 3D MTMs as high throughput long-term testing platforms for MRI contrast agents that will better predict in vivo performance, reducing animal burden, saving time, and decreasing cost.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in the supplementary material of this article.