Non-invasive imaging using reporter genes altering cellular water permeability

Abstract

Non-invasive imaging of gene expression in live, optically opaque animals is important for multiple applications, including monitoring of genetic circuits and tracking of cell-based therapeutics. Magnetic resonance imaging (MRI) could enable such monitoring with high spatiotemporal resolution. However, existing MRI reporter genes based on metalloproteins or chemical exchange probes are limited by their reliance on metals or relatively low sensitivity. Here we introduce a new class of MRI reporters based on the human water channel aquaporin 1. We show that aquaporin overexpression produces contrast in diffusion-weighted MRI by increasing tissue water diffusivity without affecting viability. Low aquaporin levels or mixed populations comprising as few as 10% aquaporin-expressing cells are sufficient to produce MRI contrast. We characterize this new contrast mechanism through experiments and simulations, and demonstrate its utility in vivo by imaging gene expression in tumours. Our results establish an alternative class of sensitive, metal-free reporter genes for non-invasive imaging.

Figure 1. AQP1 functions as a genetically encoded reporter for diffusion-weighted MRI.

(a) Illustration of the impact of aquaporin expression on water diffusion across the cell membrane and the resulting decrease in diffusion-weighted signal intensity. (b) Diffusion-weighted images of CHO, U87 and Neuro 2a cell pellets expressing AQP1 or GFP, acquired using a b-value of ∼1,000 s mm⁻². Scale bars, 3 mm. (c) ADC of water in CHO, U87 and Neuro 2a cells expressing AQP1 relative to GFP controls, measured at Δ_eff=398 ms. Transgene expression in CHO cells was induced with 1 μg ml⁻¹ doxycycline, whereas U87 and Neuro 2a cells express AQP1 from a constitutive promoter. n=4 (U87, Neuro 2a) and 5 (CHO) biological replicates. (d) Longitudinal (T₁) and (e) transverse (T₂) relaxation rates in cells expressing AQP1 or GFP. n=3 (Neuro 2a, CHO) or 4 (U87) biological replicates. (f) Cell viability on AQP1 or GFP expression. n=12 (resazurin assay), 6 (ATP content), 4 (LDH release) and 3 (ethidium staining) biological replicates. Error bars±s.e.m. (g) Phase-contrast images of CHO, U87 and Neuro 2a cells expressing AQP1 or GFP. Scale bars, 10 μm.

Figure 2. AQP1 reports gene expression over a large dynamic range.

(a) Diffusion-weighted images (acquired at Δ_eff=398 ms, b=2,089 s mm⁻²) of CHO cells expressing AQP1 or GFP (control) and treated with varying doses of doxycycline to induce transgene expression. Scale bar, 3 mm. (b) Percent change in ADC of water in AQP1-expressing CHO cells (relative to control cells expressing GFP) as a function of doxycycline concentration, measured at different diffusion times. n=4 biological replicates. Error bars±s.e.m. (c) Levels of AQP1 expression on the membrane of CHO cells at different levels of doxycycline induction, estimated based on quantitative western blotting and relative expression of a co-transcribed GFP reporter, compared with values calculated from Monte Carlo simulations based on the ADC results in b. n≥3 biological replicates. Error bars±s.e.m.

Figure 3. AQP1 expression is observable in mixed cell populations.

(a) Illustration of the effect of an increasing fraction of AQP1-labelled cells in a tissue on the overall diffusivity of water. (b) Monte Carlo simulation predictions of change in ADC as a function of the fraction of cells expressing AQP1 in a mixed cellular lattice. (c) Top: diffusion-weighted MRI (acquired at Δ_eff=198 ms, b=2,334 s mm⁻²) of cells comprising AQP1-labelled cells mixed with GFP-labelled control cells in varying proportions. Bottom: image of mixed populations containing 0, 5 and 10% AQP1-expressing cells acquired using Δ_eff=398 ms, b=8,000 s mm⁻², to maximize contrast for the low AQP1 fraction scenario, smoothed with a Gaussian filter (radius 1.5 pixels). Scale bars, 3 mm. (d) Percent change in ADC in mixed AQP1/GFP cell pellets as a function of the fraction of AQP1-expressing cells. N=4 biological replicates. Error bars±s.e.m.

Figure 4. AQP1 enables the imaging of gene expression in intracranial tumour xenografts.

(a) Experimental approach to establishing bilateral tumours in the striatum, inducing transgene expression, and performing diffusion-weighted MRI. (b) Representative diffusion-weighted image of a horizontal brain slice with bilateral tumour xenografts, 48 h after doxycycline injection. Inset shows a diffusion-weighted image of the same mouse acquired before doxycycline injection. Images were acquired at Δ_eff=98 ms and b-value=1,000 s mm⁻². Dashed lines indicate the tumour ROI(s). Scale bar, 2 mm. (c) Average diffusion-weighted image intensity of AQP1-expressing tumours relative to contralateral GFP-expressing tumours before and after doxycycline induction. n=5 biological replicates. Error bars±s.e.m. (d) Confocal fluorescence image of a representative 100 μm section of a mouse brain implanted with GFP and AQP1 tumours. The AQP1 tumour appears dimmer due to diminished GFP translation from the IRES sequence. Cell nuclei are counterstained using TO-PRO iodide (red). Scale bar, 2 mm. (e) Low ( × 10) and (f) high ( × 30) magnification images of 5 μm haematoxylin–eosin-stained sections of intracranial tumour xenografts expressing AQP1 and GFP. Scale bars, 30 and 10 μm, respectively. (g) Longitudinal measurements of tumour growth in bilateral subcutaneous xenografts induced using doxycycline to express AQP1 or GFP 11 days following tumour inoculation. (h) Mean end-point tumour mass and (i) images of AQP1- and GFP-expressing subcutaneous tumours harvested 9 days after doxycycline induction of gene expression. n=4 biological replicates. Scale bar, 1 cm. Error bars±s.e.m.