An optimal genetically-encoded probe for photoacoustic (PA) imaging should exhibit high optical absorption, low fluorescence quantum yield, and an absorption maxima within the near-infrared (NIR) window. One promising candidate is a newly engineered chromoprotein (CP), designated dark small ultra-red fluorescent protein (dsmURFP), which is based on a cyanobacterial phycobiliprotein. To optimize dsmURFP characteristics for PA imaging, we have developed a directed evolution method to iteratively screen libraries of protein variants with three different screening systems. Firstly, we took inspiration from dark-acceptor (also known as dark-quencher)-based Förster resonance energy transfer (FRET) constructs, and used dsmURFP as a dark acceptor from a mCardinal fluorescent donor. The rationale for this design was that the higher the extinction coefficient of the dsmURFP, the more the emission of the donor would be quenched. In addition, more energy transferred to the dark acceptor would lead to more thermoelastic expansion and a stronger PA signal. Three rounds of evolution using this first strategy resulted in dsmURFP1.3 that quenched the emission of mCardinal ~2-fold more efficiently than dsmURFP. Secondly, an absorption-based screening based on visual inspection of plates led to identification of the variant dsmURFP1.4, which exhibited a 2-fold higher absorbance and a 5 nm red shift. Thirdly, we developed a colony-based photoacoustic screening method. To demonstrate the utility of our optimized variants, we used photoacoustic imaging to visualize dsmURFP and its variants in phantom and in vivo experiments using chicken embryo models and murine bacterial bladder infection models.
Previously we described the potential for multiple illumination photoacoustic tomography to provide quantitative reconstructions, however this work used only simulated data. We have developed a custom photoacoustic-ultrasound tomography system capable of multiple illuminations and parallel acquisition from a 256 element 5 MHz transducer ring array with 8-cm diameter. The multiple illumination scheme uses a free-space light delivery geometry where a rotational stage scans a pulsed laser beam onto different incident locations around the sample. For each illumination location a photoacoustic image is reconstructed using a modified backprojection algorithm. Images from different source locations have the potential to be combined to form an improved deep-tissue image using our previously developed iterative algorithms. We complement the photoacoustic imaging data with unique ultrasound imaging data. Most previous ultrasound tomography methods have used migration algorithms, iterative ray-based analysis, wave-equation modeling, or frequency-based algorithms that all demand large amounts of data and computational power. We propose a new UST method that offers isotropic resolution, provides scattering contrast, as well as the potential for measuring ultrasound scattering anisotropy and decoupling density and compressibility contributions. The imaging system is driven by a Verasonics scan engine and programmed for both ultrasound and photoacoustic imaging modes. Resolution has been measured to be 150 μm for ultrasound and 200 μm for photoacoustic images. Imaging capabilities are demonstrated on phantoms with custom-tailored ultrasound scattering and optical properties, as well as in murine models.
To understand the pathogenic processes for infectious bacteria, appropriate research tools are required for replicating and characterizing infections. Fluorescence and bioluminescence imaging have primarily been used to image infections in animal models, but optical scattering in tissue significantly limits imaging depth and resolution. Photoacoustic imaging, which has improved depth-to-resolution ratio compared to conventional optical imaging, could be useful for visualizing melA-expressing bacteria since melA is a bacterial tyrosinase homologue which produces melanin. Escherichia coli-expressing melA was visibly dark in liquid culture. When melA-expressing bacteria in tubes were imaged with a VisualSonics Vevo LAZR system, the signal-to-noise ratio of a 9× dilution sample was 55, suggesting that ∼20 bacteria cells could be detected with our system. Multispectral (680, 700, 750, 800, 850, and 900 nm) analysis of the photoacoustic signal allowed unmixing of melA-expressing bacteria from blood. To compare photoacoustic reporter gene melA (using Vevo system) with luminescent and fluorescent reporter gene Nano-lantern (using Bruker Xtreme In-Vivo system), tubes of bacteria expressing melA or Nano-lantern were submerged 10 mm in 1% Intralipid, spaced between <1 and 20 mm apart from each other, and imaged with the appropriate imaging modality. Photoacoustic imaging could resolve the two tubes of melA-expressing bacteria even when the tubes were less than 1 mm from each other, while bioluminescence and fluorescence imaging could not resolve the two tubes of Nano-lantern-expressing bacteria even when the tubes were spaced 10 mm from each other. After injecting 100-μL of melA-expressing bacteria in the back flank of a chicken embryo, photoacoustic imaging allowed visualization of melA-expressing bacteria up to 10-mm deep into the embryo. Photoacoustic signal from melA could also be separated from deoxy- and oxy-hemoglobin signal observed within the embryo and chorioallantoic membrane. Our results suggest that melA is a useful photoacoustic reporter gene for visualizing bacteria, and further work incorporating photoacoustic reporters into infectious bacterial strains is warranted.
Antibiotic drug resistance is a major worldwide issue. Development of new therapies against pathogenic bacteria requires appropriate research tools for replicating and characterizing infections. Previously fluorescence and bioluminescence modalities have been used to image infectious burden in animal models but scattering significantly limits imaging depth and resolution. We hypothesize that photoacoustic imaging, which has improved depth-toresolution ratio, could be useful for visualizing MelA-expressing bacteria since MelA is a bacterial tyrosinase homologue involved in melanin production. Using an inducible expression system, E. coli expressing MelA were visibly black in liquid culture. Phosphate buffered saline (PBS), MelA-expressing bacteria (at different dilutions in PBS), and chicken embryo blood were injected in plastic tubes which were imaged using a VisualSonics Vevo LAZR system. Photoacoustic imaging at 6 different wavelengths (680, 700, 750, 800, 850 and 900nm) enabled spectral de-mixing to distinguish melanin signals from blood. The signal to noise ratio of 9x diluted MelA bacteria was 55, suggesting that ~20 bacteria cells could be detected with our system. When MelA bacteria were injected as a 100 μL bolus into a chicken embryo, photoacoustic signals from deoxy- and oxy- hemoglobin as well as MelA-expressing bacteria could be separated and overlaid on an ultrasound image, allowing visualization of the bacterial location. Photoacoustic imaging may be a useful tool for visualizing bacterial infections and further work incorporating photoacoustic reporters into infectious bacterial strains is warranted.
Proc. SPIE. 9323, Photons Plus Ultrasound: Imaging and Sensing 2015
KEYWORDS: Signal to noise ratio, Tissues, Scattering, Laser scattering, Ultrasonics, Monte Carlo methods, Transducers, Photoacoustic tomography, Acquisition tracking and pointing, Photoacoustic spectroscopy
We compare scanned-mosaicking and blanket illumination schemes for wide-field photoacoustic tomography with potential applications to breast imaging. For each illumination, a locally high-SNR image patch is reconstructed then mosaicked with image patches from other illuminations. Because the beam is not diffused over the entire area, the fluence of the beam can be maximized, therefore maximizing the signal generated. Moreover, the imaging can potentially still be done fast enough within a breath-hold. A Monte Carlo simulation as a function of beam-spot size and depth is performed to quantify this signal gain. We experimentally test both schemes using a 256-element Imasonic ring array on a tissue-mimicking phantom. We were able to verify the simulated signal gain of 2.9x under 0.5 cm of tissue with the experimental data, and measured the signal gain decrease expected when imaging deeper into the tissue. We also measured the effectiveness of averaging the diffused beam versus the scanned-mosaicking approach, and observed that for the same scan times and limited laser power output, scanned-mosaicking was able to produce a higher SNR than the blanket illumination approach. We have shown that this technique will allow wide-area PAT to utilize the maximum SNR available from any system while minimizing the number of acquisitions to reach this SNR.