NMR microsystem for label-free characterization of 3D nanoliter microtissues

Abstract

Performing chemical analysis at the nanoliter (nL) scale is of paramount importance for medicine, drug development, toxicology, and research. Despite the numerous methodologies available, a tool for obtaining chemical information non-invasively is still missing at this scale. Observer effects, sample destruction and complex preparatory procedures remain a necessary compromise. Among non-invasive spectroscopic techniques, one able to provide holistic and highly resolved chemical information in-vivo is nuclear magnetic resonance (NMR). For its renowned informative power and ability to foster discoveries and life-saving applications, efficient NMR at microscopic scales is highly sought after, but so far technical limitations could not match the stringent necessities of microbiology, such as biocompatible handling, ease of use, and high throughput. Here we introduce a novel microsystem, which combines CMOS technology with 3D microfabrication, enabling nL NMR as a platform tool for non-invasive spectroscopy of organoids, 3D cell cultures, and early stage embryos. In this study we show its application to microlivers models simulating non-alcoholic fatty liver disease, demonstrating detection of lipid metabolism dynamics in a time frame of 14 days based on 117 measurements of single 3D human liver microtissues.

Figure 1

Samples and experimental design. Representative confocal images of human liver microtissues generated in-vitro after incubation in medium containing bovine serum albumin (BSA) (a) or oleic acid coupled to BSA (OA/BSA) (b) for 7 days. MTs were fixed and stained with Hoechst (nuclei, blue), CellMask (cell membrane, red) and Nile Red (lipids, green). The images clearly distinguish between a lean and fatty liver phenotype reminiscent of microvesicular steatosis. Scalebar = 100 μm. (c) Scheme depicting culture conditions of liver microtissues, starting after spheroid formation (day -3) and incubation until day 14 days with different media compositions, i.e. fasting medium (low insulin, low glucose) and diabetic medium (high insulin, high glucose) with fractionated plasma LDL. The three resulting experimental groups are lean livers (LEAN), steatotic livers “on diet” (SOD), steatotic livers “on fat” (SOF). (d) Read-out and medium refreshing time points.

Figure 2

CMOS-based NMR probes for nL 3D MTs. (a) Photographs of our CMOS-based NMR sensor (6 × 2.5 cm) designed for in-vivo/in-vitro experiments at the nL scale. Left: sensor closed and ready for being inserted in the magnetic field. Right: sensor with open lid. (b) Photograph of the micro-system used to perform NMR spectroscopy of MTs. (c) Photographs of the CMOS microchip and its microcoil. The coil region appears with a different color due to the absence of metal density filling in order to minimize parasitic effects. (d) 3D rendering at scale of the micro-system and sample. The sample is approximated by a sphere having a diameter of 240 μm. In the indicated reference frame, the static field B₀ and gravity are along the x axis. (e) Cross-section schematics of the micro-system, defined as crossing transversally the chip at the microcoil center. The curved shape of the sample-positioning micro-structure enables placement of MTs of different sizes in the most sensitive area of the microcoil. (f) Maps of sensitivity of the integrated microcoil in experimental conditions (τ = 20 μs, i = 11 mA) at y = 0 and x = 0 cross-sections. The static B₀ field is oriented along the x axis, while the z axis is perpendicular to the coil surface. The octagonal loops of the coil are approximated with circular ones. The unitary field B_u, defined as the field produced by a current of 1 A in the coil loops, is computed via Biot-Savart. The component orthogonal to B₀ (i.e. B_uzy) is then considered to compute the nutation angle θ. Red dashed circular shapes indicate the sample as in (d-e). The excitation current i = 11 mA is deduced by matching maximum signal intensity from experiments and sensitivity maps computation. This is in line with electronic simulations of the CMOS circuit in Cadence.

Figure 3

NMR spectroscopy of 3D microtissue cultures. (a) 1D ¹H NMR spectrum obtained from a steaotic MT exposed to fasting medium (SOD) at Day10 over a measurement time of 1 h, i.e. averaging 1800 scans. (b) 1D ¹H NMR spectrum obtained from the same SOD MT as in (a) over a measurement time of 10 min, i.e. averaging 300 scans. (c) The complete series of measurements of individual MTs at Day14, where 7 measurements were collected for SOD, LEAN, and SOF MTs. Each measurement is obtained in 10 min time (i.e., averaging 300 scans). (d, e) NMR spectra averaged by experimental group at Day1 (d) and Day14 (e).

Figure 4

Label-free detection and evolution of lipids. Data are computed from spectra obtained by averaging 300 scans (i.e., 10 min measurement time). At each time point biomarkers of MTs are grouped as LEAN, SOD, SOF. At Day1, SOD and SOF coincide. 7 data points are acquired for SOD, except 5 data points at Day3. 7 data points are acquired for SOF, except 5 data points at Day14. For LEAN, 7 data points are acquired at Day1 and Day14, 6 data points at Day7 and Day10, 5 data points at Day3 and Day5. In total, 117 experiments are performed on single MTs. At Day1, significance is computed with a student t-test. From Day3, one-way ANOVA (p value ‘p_A’ is indicated) and Tukey–Kramer tests are applied (biomarkers distributions are shown in S11). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. L₁: integral area from 1.18 to 1.46 ppm. L₂: integral area from 1.8 to 2.2 ppm. L₃: integral area from 2.5 to 3 ppm. PLC: L₁/L_{1_LEAN_Day1}. PLC values are corrected for volume variation to represent concentration values (see “Methods”). Direct comparison at Day1 and Day14 of: (a) L₃/L₁ (b) L₂/L₁ (c) L₃/L₂ (d) PLC. Complete time evolution of: (e) L₃/L₁ (f) PLC.