Improved pentamethine cyanine nanosensors for optoacoustic imaging of pancreatic cancer

Optoacoustic imaging is a new biomedical imaging technology with clear benefits over traditional optical imaging and ultrasound. While the imaging technology has improved since its initial development, the creation of dedicated contrast agents for optoacoustic imaging has been stagnant. Current exploration of contrast agents has been limited to standard commercial dyes that have already been established in optical imaging applications. While some of these compounds have demonstrated utility in optoacoustic imaging, they are far from optimal and there is a need for contrast agents with tailored optoacoustic properties. The synthesis, encapsulation within tumor targeting silica nanoparticles and applications in in vivo tumor imaging of optoacoustic contrast agents are reported.

tissues [32][33][34][35][36][37] . MSOT, on the other hand, is lacking in optimized dyes that can serve as contrast agents. The majority of the available literature utilizes commercially available NIR dyes, which have sub-optimal physicochemical properties and optoacoustic signal for in vivo use [38][39][40][41] . Multiple families of NIR dyes exist, and fine tuning of the physicochemical and biological properties can be achieved through structural manipulation of the dye scaffolds. Cyanine dyes are a versatile class of chromophore. The cyanine scaffold is comprised of two nitrogen containing heterocycles connected by a conjugated polymethine linker. This scaffold is amenable to modification; the heterocycles, the polyene linker, and the substituents can be manipulated to give rise to appealing optical properties and biological interactions. By manipulating discrete molecular characteristics and investigating the optoacoustic properties of the resulting dyes, effective contrast agents for optoacoustic imaging can be developed.
Nanoparticle delivery of contrast agents and therapeutics has gained increased appeal in recent years as novel nanostructures, are developed that display promising properties for clinical use. Mesoporous silica nanoparticles (MSN) have seen use in preclinical imaging studies due to their favorable biocompatibility and modifiable surface topology, which enables them to deliver molecules to tissues of interest 42 . Sequestration of drugs within the pores of silica nanoparticles has been utilized to achieve specific drug delivery of chemotherapeutics, which helps to avoid many deleterious off-target effects associated with those drugs 43 . In a recently published work, CaCO 3 -lentinan and silica nanoparticles were developed with a wormhole pore architecture 44,45 . It was found that the asymmetric nature of the pores gave rise to improved drug release kinetics in vivo. Specifically, the larger surface area allows for increased adsorption of cargo when compared to hexagonally ordered pores 46 , leading to highly controllable cargo retention and release compared to other nanoparticles. Furthermore, the use of silica nanoparticles allows for additional surface functionalization to improve tumor targeting beyond the basic enhanced permeability and retention (EPR) reported for nanoparticles, which only accounts for about 5% of the injected dose reaching the target and about 20-30% improved uptake versus normal organs [47][48][49][50] . With aforementioned benefits including biocompatibility, size, tunability, and high drug loading capacity, MSNs are quickly emerging as clinically viable nanomedicine for delivery of various agents. With that, the work described herein aims to develop an understanding of dye structure in optoacoustic imaging.
From our small library, a lead compound with strong optoacoustic signal has been identified and derivatized in order to produce more effective optoacoustic contrast agents. Throughout the process of preparing derivatives, the structural changes did not destroy the dyes' optoacoustic performances. These dyes were encapsulated within tumor-targeted nanoparticles as potential contrast agents to detect pancreatic tumors in vivo using optoacoustic imaging in mice in real time and in 3D. In this study, we investigate the structural modifications of the dye to promote optimal encapsulation within the nanoparticle and to obtain detectable optoacoustic signal both in vitro and in vivo. Furthermore, our results support the definition of a chemical structure preferred for both in vitro and in vivo optoacoustic imaging studies in a pursuit to translate diagnostic agents to the clinic and address the need for early diagnosis and efficacious treatment of pancreatic cancers.

Results and discussion
As molecular characteristics that give rise to improved optoacoustic signal are poorly defined for small molecule dyes, the development of improved optoacoustic contrast agents was initiated first by screening a large library of pentametine cyanine dyes previously developed in our lab for fluorescence imaging applications (not all compounds shown). This library was prepared starting from commercially available phenylhydrazine derivatives, which undergo Fisher indole synthesis to create substituted indolenine derivatives 1 as outlined in Scheme 1. Alkylation of the indolenine heterocycles 1 with different alkyl halides such as propyl quaternary ammonium, butyl and phenyl propyl halides in toluene under reflux conditions gave the respective the heterocyclic salts 2A-2D. Upon cooling, the reaction mixture to room temperature and subsequent purification via crystallization and co-precipitation with methanol and diethyl ether afforded the heterocyles 2 in good yields. The hygroscopic nature of heterocycles bearing quaternary ammonium moieties allowed us to perform alkylation of compound 2A under nitrogen atmosphere. In addition, precipitation of 2A and 2B, were a tedious, however, careful selection of solvents permitted a successful co-precipitation of these two compounds using methanol and acetone. As shown in Scheme 1, the final dyes 3A-3D are pentamethine dyes and this require the use of the required pentamethine linker to efficiently synthesize them. Though the procedure to the synthesis of these scaffolds seems ubiquitous, development of the require conditions for their synthesis seems challenging, especially small molecules with net positive charges. To develop a facile and efficient approach in the synthesis of the final dyes, compound 3A was obtained via condensation of the respective salt 2A with malonaldehyde bisphenylamine salts in acetic anhydride under nitrogen condition to generate the final dye 3A. UV-vis-NIR spectroscopy and TLC monitored the reaction progress. Upon consumption of the starting materials, the reaction mixture was allowed www.nature.com/scientificreports/ to cool to room temperature and diethyl ether was added to induce precipitation. The precipitate was collected by vacuum filtration. The crude solid was purified by recrystallization from methanol/acetone to yield the final dye 3A. From that initial library, compound 3A (Scheme 1), which contains the quaternary ammonium moieties of the nitrogen atom of the heterocyclic ring and methoxy groups at the 5 position of the benzene ring of the heterocyclic was identified as a promising lead due to its appealing optoacoustic properties. The dyes were loaded into mesoporous silica nanoparticles (MSN). The asymetric wormhole mesoporous silica particles (W-MSN) were synthesized by a Sol-gel method and triethanolamine (TEA) was added to the mixture to result in asymetric wormhole pores [51][52][53] . The wormhole pores can encapsulate more cargo and release cargo at a slower rate than related hexagonal structures synthesized by a similar process 42 . Although a detailed mechanism of the growth formation of wormhole pores of W-MSN is unknown, it is hypothesized that complexing tetramethyl orthosilicate (TMOS) with TEA forms silatrane complexes under anhydrous conditions at elevated temperatures and limits the growth and aggregation of particles 52,54 . In contrast to many wormhole pore-based MSN particles, the present W-MSN have a smaller diameter (~ 33 nm) than the traditional > 50 nm nanoparticle; this should facilitate increased tumor uptake due to the smaller size. In order to achieve a smaller spherical mesoporous silica particle with wormhole pores, the procedure has key features including the use of TMOS instead tetraethyl orthosilicate (TEOS) and tetrapropyl orthosilicate (TPOS) as well as use of the polyalcohol base TEA as a substitute for NaOH or NH 4 OH (Fig. 1). This method creates a monodisperse (PDI < 1) population of MSNs. Following a series of acid and water dialysis, zeta potential of the solution is neutralized and particles are stable at room temperature for several months.
Initial experiments to encapsulate dye 3A in silica nanoparticles revealed a weakness in the compound; while it gave a good optoacoustic signal, the high water solubility of 3A led to leaching from the nanoparticles into the supernatant solution. Despite the appealing properties of the free dye, nanoparticles prepared from the lead compound lacked signal due to the poor retention of the dye. To overcome this issue, another derivatives 3B-3D (Scheme 1) were prepared, which were increasingly hydrophobic. From the initial lead compound, quaternary ammonium groups were replaced with butyl groups to give dye 3B. Similar replacement with a propyl phenyl groups gave compound 3C. Lastly, removal of the heterocyclic methoxy groups of 3C and the addition of a lipophilic meso chlorine atoms produced a new pentamethine cyanine compound 3D.
Compound 3D showed an improved optoacoustic property for the in vivo imaging due its hydrophobicity as well as the addition of chlorine atom to establish the well-established concept of heavy atom effect. As shown in Fig. 2, the prepared compounds share similar optoacoustic spectral responses to initial lead 3A. The MSOT spectra of 3A-D fall between those of the two FDA approved NIR dyes, methylene blue (MB) and indocyanine green (ICG). Unlike MB, which sees a dramatic drop in signal after 700 nm, the signal from our new dyes extends www.nature.com/scientificreports/ much farther into the NIR region. Standard curves of 3A and 3D with concentration corresponding to loaded dye is found in Fig. 1S. Retention of the dyes as measured by the change in absorbance of the supernatant after encapsulation improved as hydrophobicity increased (Table 1). Specifically, following dye encapsulation, the absorbance of the supernatant was measured. Higher signal in the supernatant was indicative of dye in solution and thus, less dye retention. With each modification, the retention of the dye in the nanoparticle improved, with only 9% of compound 3D leaching out of the nanoparticles. Therefore, Compound 3D was selected as our new lead compound, and compound 3A used as the negative control in our study.
Chitosan capping of the nanoparticles is designed to release the nanoparticle cargo in the mildly acidic extracellular matrix of the tumor. To explore the kinetics of pH dependent dye release, the nanoparticles loaded with dye 3A or 3D were washed with water until no absorption corresponding to the dye on UV-vis-NIR was detectable. The absorbance of the supernatant was monitored over time at pH values of 7.4 and 6.6 to mimic www.nature.com/scientificreports/ normal physiological pH and tumor pH respectively. In the release study shown in Fig. 3, compound 3A, with permanently charged trimethylammonium groups on the pendant alkyl chains, showed excessive leaching of the dye from the nanoparticles at pH 6.6. Even at a normal physiological pH of 7.4, compound 3A shows rapid release from the nanoparticles. Dye 3D, the most hydrophobic of the compounds, showed limited release from the nanoparticles, even at a low pH.
To enable efficient trafficking of the nanoparticles to tumors, beyond the limited accumulation achieved by the EPR effect, the dye-loaded CW-MSN were surface functionalized with the V7 pH-low insertion peptide. In normal physiological conditions, the peptide does not interact with cells; however, tumor acidosis leads to protonation and folding of the peptide, forming a transmembrane alpha helix, which anchors it to the tumor cell membrane. Characterization of in vivo compound 3A V7-CW-MSN and compound 3D V7-CW-MSN, tumor accumulation was performed using MSOT imaging on 10 athymic mice bearing orthotopic pancreatic tumors of the cell line S2VP10. Note that S2VP10 cells contain a firefly luciferase and are not red shifted. Red-shifted S2VP10s would have an intrinsic fluorescent signal around 615 nm, adequate for photoacoustic microscopy but not for MSOT imaging 55,56 . Figure 5, presents a frontal view of transverse serial slices of the whole mouse body in 0.5 mm incremented step for mice injected with either compound 3A V7-CW-MSN or compound 3D V7-CW-MSN (juxtaposed). The cross-sectional view optoacoustic imaging of tissue mimicking intra-lipid-agar phantoms containing dye 3A, and 3A encapsulated in CW-MSN (3A CW-MSN), dye 3D, and 3D encapsulated in CW-MSN (3D CW-MSN) is presented in Fig. 4. A more complete comparison of 3A and 3D optoacoustic signal corresponding to OD is shown in Fig. 2S. The two free dye 3A and 3D controls show similar signal. The signal of 3A CW-MSN is very weak compared to 3A only. The dye leaching of 3A observed during preparation of particles explains the absence of optoacoustic signal in 3A CW-MSN in tissue phantoms. The optoacoustic signal of 3D CW-MSN, however, was higher than its dye only analogue and considerably higher than that of 3A CW-MSN. This result substantiates the proposed improved hydrophobicity for nanoparticle loading of 3D over 3A. The improved signal from 3D CW-MSN over the free dye 3D may originate from dye aggregation inside the nanoparticles. While aggregation-based quenching is undesirable for fluorescence applications, it has been shown to improve the optoacoustic performance of similar dye scaffolds 57 .
As shown in Fig. 5, higher mean signal per cross section was used as an indicator of tumor binding. Compound 3A V7-CW-MSN showed low pancreatic tumor accumulation with slightly higher uptake in the liver and spleen. Concentration of optoacoustic signal was observed for 3A V7-CW-MSN in the major blood vessels. Due to this low concentration, deoxy-hemoglobin was used for tumor segregation in compound 3A V7-CW-MSN treated animals. Deoxy-hemoglobin MSOT images are shown in Fig. 3S. Given the early time point of www.nature.com/scientificreports/ imaging (4 h), this low concentration suggests that, in accordance with earlier results for both dye loading and tissue-mimicking phantoms, dye leaching of 3A out of V7-CW-MSN occurred during sample preparation for mouse injection. 3D dye again presents an oppositely situated case; MSOT imaging (4 h post-injection) of mice injected with 3D V7-CW-MSN showed much higher major blood vessel optoacoustic signal than 3A V7-CW-MSN as well as higher signal accumulation in the pancreatic tumor without the need for deoxy-hemoglobin segregation. Whereas off-target accumulation appeared higher, compared to pancreatic tumor signal accumulation for 3A V7-CW-MSN, 3D V7-CW-MSN showed higher signal accumulation in the pancreatic tumor than in the kidney, liver, spleen, and elsewhere. Inclusion of V7 was essential for tumor accumulation regardless of dye used, as shown in Fig. 4S.

Conclusion
As optoacoustic imaging moves into clinical practice, there will be a need for contrast agents with strong signal and good biocompatibility. Organic dyes have revolutionized the potential of fluorescence imaging and are now poised to do the same for optoacoustic imaging; however, the underlying correlations between molecular structure and optoacoustic signal must be clearly defined. The series of pentamethine cyanine dyes discussed herein were prepared with the two-fold goal of having strong optoacoustic signal and sufficient retention in tumor-targeting nanoparticles for in vivo pancreatic tumor imaging. While all four compounds showed strong optoacoustic signal, the first of them, 3A, proved too soluble in aqueous media and leached from the nanoparticles during preparation. With each successive modification to increase hydrophobicity, the retention improved. Hydrophobic dye with chlorine atom at the meso carbon 3D showed the greatest retention in nano-vehicles for imaging and was retained even in acidic media. In both phantom and animal imaging studies, 3D loaded nanoparticles gave superior signal, and, in in vivo imaging of orthotopic pancreatic tumors, clear foci were visible in the animals' abdomens.
From this study, we do acknowledge the limitations of our tested probes 3A-3D concerning their short wavelengths. For our future work, we will use our knowledge we have gained from developing the pentamethine library dyes (Cy5) and focus our efforts on designing the ideal heptamethine dyes (Cy7) for optoacoustic imaging.

Experimental
Chemicals and methods. All chemical reagents and solvents were purchased from Sigma Aldrich (Saint Louis, MO) and were used without further purification. Reaction progress was monitored on silica gel 60 F 254 thin layer chromatography plates (Merck EMD Millipore, Darmstadt, Germany) NMR spectra were recorded on a Bruker Avance 400 MHz spectrophotometer in deuterated solvents purchased from Cambridge Isotope Labs (Andover, MA). All chemical shifts were recorded in parts per million (ppm) relative to the residual solvent. Signals are labelled as follows: s (singlet), d (doublet), t (triplet) m (multiplet) brs (broad signal) and coupling constants (J) are measured in Hertz (Hz). Mass spectra were performed at the Georgia State University Mass Spectrometry Facility using a Waters Q-TOF micro (ESI-Q-TOF) mass spectrometer ora Waters Micromass LCT TOF ES + Premier mass spectrometer. Melting points were recorded on a Mel-Temp apparatus and are not corrected. V7 pHLIP was synthesized by and purchased from CS Bio Company (Menlo Park, CA, USA). All methods were carried out in accordance with relevant guidelines and regulations of the Scientific Report Journal. The new compounds were synthesized and purified as outlined below. The purity of all synthesized compounds 3A-D was confirmed with 1 HNMR, 13 CNMR and ESI-HRMS. These analytical methods used to determine the purity of compounds to be ≥ 95%.
2,3,3-trimethyl-1-(3-phenylpropyl)-3H-indol-1-ium bromide-2d was prepared as previously described 61 . Synthesis of dyes 3A through 3D. The final dyes were prepared as previously described 59 . Briefly, indolium salts 2 (2 mol eq), sodium acetate (2 mol eq) and polymethine linker (malonaldehyde bisphenylamine hydrochloride its chloro derivative) (1 mol eq) were added to a dry, nitrogen flushed round bottom flask equipped with a stir bar. Acetic anhydride (5 ml) was added and the mixture was heated to 60 °C and stirred vigorously. Reaction progress was monitored by UV-vis spectroscopy and TLC. Upon consumption of the starting materials, the reaction flask was allowed to cool to room temperature and diethyl ether was added to induce precipitation. The precipitate was collected by vacuum filtration. The crude solid was purified by recrystallization from methanol/acetone or methanol/ether to give the pure final dyes. Following dye synthesis, 1 mM of each 3A-3D was added into a 1.3%w/w of agar and 6% v/v intralipid tissue-mimicking phantom to acquire spectral signatures with MSOT. 1,3-dien-1-yl)-3,3-dimethyl-1-(3-(trimethylammonio)

Chitosan capping of the W-MSN Core to result in CW-MSN. A gatekeeping component is required
for mesoporous particles to facilitate encapsulation and release of the dye within the mesoporous silica particle. Chitosan (1%) was dissolved in 5% acetic acid (200 mL) using constant stirring for 24 h at 25 °C. The chitosan gatekeeping component was added to the surface of the W-MSN core particles by initially dispersing the W-MSN in ethanol and separating them using ultrasonication (3 min). Next, the solution pH was decreased from pH 3.8 to 3.5 with acetic acid and the 0.1 g of 3-glycidoxypropylmethyldie-thoxysilane (GPTMS) was added to the solution, which was stirred for 3 h to result in a reactive surface of the W-MSN core. The chitosan solution (1%) was added to the reactive W-MSN core at 25 °C with constant stirring for 24 h resulting in CW-MSN. The CW-MSN particles were washed with ethanol (70%) and stored at 4 °C. V7 peptide (V7-CW-MSN). The  Scattering and transmission electron microscopy. The Dynamic Light Scattering (DLS) was conducted using a Zetasizer Nano ZS (Malvern Instruments, UK) with a 5 mW HeNe laser at 633 nm wavelength. The synthesized TMOS-based particles were suspended in Milli-Q water (1.5 mL/L of synthesis). The ultrapure water preserved at room temperature was used in the data analysis with a refractive index of 1.33 and viscosity 0.89 in automatic mode using the 6.0 Zetasizer Nano software. The morphological characterization of the W-MSN was performed with transmission electron microscopy (TEM). Morphological characteristics such as average size and shape were qualitatively analyzed with a Zeiss Supra 35VP electron microscope (Zeiss, Pleasanton, CA, USA). W-MSN were analyzed by placing 10 μL of the sample on a SEM studs and, before viewing, wick dried using filter paper. The average pore size of the W-MSN were calculated based on analysis of 26 CW-MSN Core particles using the Image J software (https ://image j.net/Downl oads).

CW-MSN conjugation with the
Media preparation for the in vitro nanoparticle assays. To evaluate the pH specificity of the nanoparticles in vitro, a phosphate buffered, pH specific media was created that was independent of bicarbonate buffering. A 25 mM phosphate buffer was prepared at pHs 7.4 or 6.6 by dissolving sodium phosphate monobasic and dibasic solution dissolved in distilled water (Sigma-Aldrich, St. Louis, MO, USA). The phosphate buffer was sterilized using an autoclave. This phosphate buffer was the base of both the pH specific media and the washing solution. The pH specific media contained the 25 mM phosphate buffer, 13.6 g of DMEM powder, 10% fetal bovine serum, and 1% L-glutamine and filtered through grade-1 Whatman qualitative filter paper (Sigma-Aldrich, St. Louis, MO, USA). The solution pH was confirmed using a pH meter (Denver Instrument Ultrabasic, Bohemia, NY, USA). The washing solution also contained the same 25 mM pH specific phosphate buffer and included 1% fetal bovine serum. The pH of the solutions was adjusted with sterilized sodium hydroxide (1 M) or hydrochloric acid (1 M) as needed.
In vitro assay to determine the ability of CW-MSN for entrapping either 3D or 3A as cargo for optoacoustic detection.  63 . S2VP10L cells contain traditional firefly luciferase, and are not red shifted. For the surgical procedure, mice were anesthetized with isoflurane at 1.5% in 1 L O 2 and betadine was used to disinfect the skin over the abdomen. An incision into the left upper quadrant allowed for identification of the spleen and pancreas. A solution of 5 × 10 5 S2VP10L cells/30 μL of serum-free DMEM media was injected into the pancreas using a 28G syringe. To prevent leakage of the tumor cells, a sterile cotton tipped applicator was held over the injection site for 30 s. The pancreas and spleen were returned to the abdomen and the wound was closed in a single layer using 5-0 nylon sutures. Mice recovered underneath a warming blanket and were returned to their cages.
Tumor monitoring with bioluminescence imaging. To evaluate potential leakage of cells from orthotopic implantation, bioluminescence imaging was performed immediately following surgery with the advanced molecular imager instrument (AMI) (Spectral Imaging Instruments, Tucson, AZ, USA). Mice were imaged 10 min after IP injection of 2.5 mg luciferin. Any showing signs of peritoneal leakage were excluded from further study. After 5 days, the sutures were removed to prevent artefacts in future imaging studies. Tumor size was followed daily with bioluminescence imaging. At day 7 post-implant, mice bearing S2VP10L were sorted based upon tumor signal so that each group had equal group mean signal based upon bioluminescence signal to evaluate 3D V7-CW-MSN and 3A V7-CW-MSN for in vivo imaging using MSOT.
Evaluation of 3D and 3A V7-CW-MSN accumulation in vivo assessed using multispectral optoacoustic tomography. Detection of 3D V7-CW-MSN and 3A V7-CW-MSN biodistribution within orthotopic pancreatic xenografts was performed using the inVision-512 echo multispectral optoacoustic tomography system containing ViewMSOT version 3.8 software (iThera Medical, Munich, Germany) 45 . Anesthesia for the mice consisted of 1.5% isoflurane inhalant delivered in 0.8 L of medical air and 0.1 L of O 2 . Anesthetic depth was maintained throughout the image acquisitions. Mice were imaged with the ventral side up in the MSOT animal holder. All mice were imaged using MSOT prior to IV injection of 3D V7-CW-MSN and 3A V7-CW-MSN to determine baseline signal. MSOT imaging shows optimal detection at earlier time points 25 www.nature.com/scientificreports/ animal movement. The acquisition time was 10 μs per frame. Pancreatic tumors were identified based upon the deoxy-hemoglobin signal in the live-screen multispectral preview in ViewMSOT 3.8 45 .
MSOT image reconstruction and analysis. Raw data obtained with MSOT were reconstructed with multispectral analysis performed as previously described 20,63 . Spectral analysis was performed at wavelengths corresponding to 3A or 3D dye within the nanoparticles and oxy-and deoxy-hemoglobin. Spectra utilized for spectral unmixing of the nanoparticle are located within Fig. 2 with the spectral shape of 3A in blue and 3D in yellow. Both oxy-and deoxy-hemoglobin spectra were utilized as identified in previous publications 64,65 . Version 3.8 of ViewMSOT software (iThera Medical, Munich, Germany) was used to reconstruct MSOT images from the raw data using a backprojection algorithm at a resolution of 75 μm. Backprojection reconstruction is a very fast, semi-quantitative approach to display the anatomical features of the image, an in-depth description of the algorithm can be found here 66 . The multispectral processing was conducted using linear regression with ViewMSOT 3.8, where known molar absorptivity spectra (e.g., for 3D dye, 3A dye, oxy-, and deoxy-hemoglobin) are used to model the relationship between chromophore concentration and MSOT signal over a range of wavelengths.
The approach assumes knowledge about all absorbers present in the imaged tissue in order to correctly attribute contributions from the different wavelengths to the unmixed component images. In order to ensure comparability among data sets, the reconstruction parameters (field of view, speed of sound, pixel size, and the high/ low pass filters) and spectral unmixing parameters were consistently applied to all data. Spectral unmixing was performed in the absence of correction for fluence heterogeneities and attenuation as a function of tissue depth including spectral coloring. The orthotopic tumors had a similar location and distance from the skin surface from animal to animal, thus, fluence issues would equally affect all animals. The location of orthotopic tumors was identified based upon the presence of deoxy-hemoglobin, upon deoxy-hemoglobin spectra, and upon relative location of the spleen and kidney. In addition, a region of interest (ROI) method was applied to determine signal strength in the tumors, liver, and kidney of both 3D V7-CW-MSN and 3A V7-CW-MSN treated mice using ViewMSOT software and reported as MSOT a.u. The ROI was manually created with an ellipse drawing tool with location of tumor based upon deoxy-hemoglobin component. The deoxy-hemoglobin component was used to provide a basis for tumor segmentation in animals receiving 3A V7-CW-MSN as very little 3A signal was detected (Fig. 5). The ROI area was kept constant at 3.5 mm 2 for all image slices, thus creating a nonuniform elliptical prism volume of interest (VOI). The mean pixel intensity per cross-section in the VOI for the spectrally unmixed injected agent (compounds 3D and 3A control) was plotted as MSOT signal vs position to assess the signal strength in the tumor. This analysis produced a consistent parabolic shape of signal over distance in the tumor. The maximal "mean signal per cross-section" in the volume was used as a quantitative indicator of probe binding in the tumor. Because optoacoustic signals using the detection geometry of this system are subject to out-of-plane contributions, this method was used to find the center of signal intensity and minimize variability from out-of-plane artefacts. The capacity of this optoacoustic system to deliver semiquantitative data reflective of relative probe accumulation in vivo in murine models using the aforementioned reconstruction and multispectral unmixing methods was previously established 20,25,49 . The MSOT a.u. values for 3D V7-CW-MSN and 3A V7-CW-MSN control were compared using SAS 9.3 (SAS, Cary, NC, USA).

Statistics.
In vitro, comparison of signal intensity of 3D dye and 3A dye alone and within the nanoparticle was performed with T-test using SAS 9.3 (SAS, Cary, NC, USA). In vivo, the MSOT a.u. values for 3D V7-CW-MSN and 3A V7-CW-MSN control were compared using Wilcoxon sum-rank test and ANOVA followed by Tukey post hoc test using SAS 9.3 (SAS, Cary, NC, USA). Significance was observed where p < 0.05.