Multi-wavelength photoacoustic imaging of inducible tyrosinase reporter gene expression in xenograft tumors

Photoacoustic imaging is an emerging hybrid imaging technology capable of breaking through resolution limits of pure optical imaging technologies imposed by optical-scattering to provide fine-resolution optical contrast information in deep tissues. We demonstrate the ability of multi-wavelength photoacoustic imaging to estimate relative gene expression distributions using an inducible expression system and co-register images with hemoglobin oxygen saturation estimates and micro-ultrasound data. Tyrosinase, the rate-limiting enzyme in melanin production, is used as a reporter gene owing to its strong optical absorption and enzymatic amplification mechanism. Tetracycline-inducible melanin expression is turned on via doxycycline treatment in vivo. Serial multi-wavelength imaging reveals very low estimated melanin expression in tumors prior to doxycycline treatment or in tumors with no tyrosinase gene present, but strong signals after melanin induction in tumors tagged with the tyrosinase reporter. The combination of new inducible reporters and high-resolution photoacoustic and micro-ultrasound technology is poised to bring a new dimension to the study of gene expression in vivo.

Multispectral photoacoustic image of a small +TYR tumor after DOX-induction of melanin production. The scale bar represents 1 mm.

Supplementary Figure 4
Quantitative T1 relaxation times (in ms) derived from MR images of tubes containing pelleted +TYR cells. Cells were prepared with or without DOX (columns) and with varying concentrations of ferric citrate (rows).

Supplementary Figure 5
In vivo MRI imaging of T1 relaxation times. (a) T1 map (in ms) of a mouse bearing -TYR and +TYR tumors after DOX and ferric citrate treatment. (b) Mean reduction in T1 values for +TYR tumors in three mice.

Supplementary explanation of multispectral unmixing algorithm
Data from multispectral photoacoustic images were de-mixed using a constrained least-squares algorithm described here. The algorithm permits positivity constraints as well as the ability to require estimated oxygen saturation levels to be above a threshold.
Define a molar extinction matrix such that is the molar extinction coefficient of species = { , , 2} at wavelength . The vector of relative concentrations to be estimated at each pixel location is ( ) = [ ( ) ( ) 2 ( )] while the estimated photoacoustic initial pressure spectra is ̂( ) , where the th element ̂( ) is the estimated photoacoustic signal at wavelength normalized by estimates of the laser fluence. We measured mean incident laser power as a function of wavelength, I 0 ( ). The photoacoustic initial pressure at location , is modelled as ( ) = Γ ( , )Φ( , ) where Φ( , ) ∝ I 0 ( ) is the unknown wavelength dependent fluence, ( , ) is the local optical absorption coefficient and Γ is the Gruneisen parameter, assumed to be spatially constant. We assume that ( , ) = ( ) is a linear combination of concentrations of the dominant components (Hb, HbO2 and melanin) weighted by the molar extinction coefficients, where is the j th column of . Then given experimental photoacoustic images at wavelengths , = 1,2, … , , and measured laser powers I 0 ( ), the task is to solve the system of equations ̂= I 0 ( ) + for the estimated concentrations ̂( ) where is noise. Here ( ) is an unknown position-dependent coefficient encompassing wavelength-dependent fluence variations and is modeled as a wavelengthindependent constant modified by a wavelength dependent perturbation = (1 + δ ). In this way the system to be solved can be written as ̂= + where are the relative concentrations (with unknown scaling constant common to all concentrations), ̂ has elements ̂=̂/ I 0 ( ) and are the laser-power-normalized photoacoustic signals, and now encompasses both stochastic noise and bias due to perturbations δ . In the manuscript we refer to ̂ as just for simplicity. We use constrained least squares to estimate for each image pixel. We solve ̂= argmin {| −̂| } subject to constraint ̂≤ . In choosing the constraint conditions, we impose positivity of the reconstructed relative concentrations ̂≥ , and require the estimated oxygen saturation 2 ≡ 2 2 + to be greater than some threshold : 2 ≥ to be physiological. This requires 2 (1 − ) − ≥ 0. With these constraints it is simple to show that the constraint matrix should be chosen as while is taken as = [ ] . The function lsqlin in MATLAB (Mathworks Inc.) was used to implement the constrained linear least-squares unmixing. When = 0 the function lsqnonneg may be used instead. In order to obtain accurate spectral de-mixing, we acquire photoacoustic images using a large number of sequential wavelengths so that the system ̂= + is highly over-determined, based on the hypothesis that the more non-degenerate information used, the more accurate the relative concentration estimates will be.
When displaying de-mixed images, each pixel was first classified as blood or melanin. Pixels with total hemoglobin levels greater than a threshold fraction of the maximum value (we used 50%) were classified as blood and displayed on a red-to-blue colormap representative of the estimated oxygen saturation ( 2 ). The remaining pixels were assigned to a green-to-black colormap representing estimated relative melanin concentration. This estimation procedure is not rigorously quantitative, but presents a simple way to visualize approximate 2 and melanin distributions and observe inducible reporter gene expression in vivo.