functionalflowpatternsandstaticbloodpooling in tumors … · potential to facilitate the monitoring...

Integrated Systems and Technologies

Functional FlowPatterns and Static Blood Poolingin Tumors Revealed by Combined Contrast-Enhanced Ultrasound and Photoacoustic ImagingAvinoam Bar-Zion1, Melissa Yin2, Dan Adam1, and F. Stuart Foster2,3

Abstract

Alterations in tumor perfusion and microenvironment havebeen shown to be associated with aggressive cancer phenotypes,raising the need for noninvasive methods of tracking thesechanges. Dynamic contrast–enhanced ultrasound (DCEUS) andphotoacoustic (PA) imaging serve as promising candidates—onehas the ability to measure tissue perfusion, whereas the other canbe used to monitor tissue oxygenation and hemoglobin concen-tration. In this study,we investigated the relationship between thedifferent functional parameters measured with DCEUS and PAimaging, using two morphologically different hind-limb tumormodels and drug-induced alterations in an orthotopic breast

tumor model. Imaging results showed some correlationbetween perfusion and oxygen saturation maps and the abilityto sensitively monitor antivascular treatment. In addition,DCEUS measurements revealed different vascular densities inthe core of specific tumors compared with their rims. Non-correlated perfusion and hemoglobin concentration measure-ments facilitated discrimination between blood lakes andnecrotic areas. Taken together, our results illustrate the utilityof a combined contrast-enhancedultrasoundmethodwithphoto-acoustic imaging to visualize blood flow patterns in tumors.Cancer Res; 76(15); 4320–31. �2016 AACR.

IntroductionTo ensure survival and progression, tumor cells have acquired a

number of characteristics or "hallmarks" that differentiate themfrom normal cells. Theses specific traits, including the ability toinduce abnormal angiogenesis, have been amajor area of focus incancer research (1). In addition to genetically altered cellularcharacteristics, tumor microenvironmental factors, such as hyp-oxia, also play an important part in cancer progression andaggressive phenotypes (2). Knowledge of changes in tumor per-fusion and microenvironmental patterns is of great importancefor detecting aggressive cancer phenotypes and determining theeffects of different treatments. Therefore, there is a need fornoninvasive imaging methods capable of tracking both changesin tumor perfusion and itsmicroenvironment.Dynamic contrast–enhanced ultrasound (DCEUS) and photoacoustic (PA) imagingare two noninvasive functional imaging modalities, having thepotential to facilitate the monitoring of tumor development andtreatment efficiency.

Contrast-enhanced ultrasound involves the use of gas-filledmicrobubbles. Using special pulse sequences and processingtechniques the signal originating from the microbubbles can beseparated from the soft tissue background (3). Furthermore,microbubbles are similar in size to red blood cells making themideal for the imaging of blood (4). DCEUS is an establishedimaging mode for the measurement of functional tissue perfu-sion, providing us with insights on blood flow and blood volumeinside localized tissue regions (5).

PA imaging combines optic and acoustic imaging into a singlemodality. Short laser pulses directed to the tissue generate region-al thermo-elastic expansion and create acoustic waves that aredetected by an ultrasound transducer at the surface of the tissue(6, 7). By encoding optical contrast onto acoustic waves, thepenetration depth is greatly increased, while retaining the highcontrast and spectral specificity of optical imaging. PA imaging isan excellent tool for the visualization of blood, or more specif-ically hemoglobin in tissue (8, 9). Furthermore, the differingabsorption spectra of deoxygenated and oxygenated hemoglo-bin make it possible to monitor oxygen saturation (SO2) in vivothrough the use of multiwavelength imaging, allowing for hyp-oxia detection in tissue.

There are two commonly accepted forms of hypoxia: diffusion-limited and perfusion-limited (2). Tumor cells located aroundfunctional blood vessels are supplied with oxygen and nutrients.When the distance from these vessels increases, the result isdecreased pO2 levels, characterizing diffusion-limited hypoxia.Perfusion-limited hypoxia on the other hand is transient andcaused by irregularities in blood flow within the tumor vascula-ture. As both the structure and functional status of tumor vascu-lature are important factors that affect the oxygenation of tumortissue, it is important to have imaging tools that can be used toinvestigate changes in perfusion, such as DCEUS, and monitorchanges the oxygenation, such as PA imaging.

1Department of Biomedical Engineering, Technion - Israel Institute ofTechnology, Haifa, Israel. 2Physical Sciences Platform, SunnybrookResearch Institute,Toronto,Canada. 3DepartmentofMedical Biophys-ics, University of Toronto, Toronto, Canada.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

A. Bar-Zion and M. Yin contributed equally to this article.

Corresponding Authors: Avinoam Bar-Zion, Department of Biomedical Engi-neering, Technion - Israel Institute of Technology, Haifa 32000, Israel. Phone:972-4829-4136; Fax: 972-4829-4599; E-mail: [email protected]; and F.Stuart Foster, University of Toronto, S-658, 2075 Bayview Avenue, Toronto,Ontario M4N 3M5, Canada. Phone: 416-480-5716; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-16-0376

�2016 American Association for Cancer Research.

CancerResearch

Cancer Res; 76(15) August 1, 20164320

on April 8, 2021. © 2016 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst June 20, 2016; DOI: 10.1158/0008-5472.CAN-16-0376

http://cancerres.aacrjournals.org/

In addition to SO2, hemoglobin concentration can also beestimated from PA measurements. Hemoglobin is usuallypurely intravascular. However, the leakiness of tumor vascula-ture can lead to extravasation of erythrocytes outside the vesselwalls to form static hemorrhagic regions (blood lakes) in sometumors. Defects in the endothelial monolayer are known toexplain the leakiness of tumor vessels (10). Both blood lakesand necrotic regions do not incorporate functional vasculature,thus have low perfusion values and potentially high hemoglo-bin content.

In this study, we examined the advantages of using DCEUScombined with PA imaging. Previous work performed usingscanners combining DCEUS and PA imaging tend to analyzephysiologic properties estimated using each modality separately(e.g., ref. 11). This sort of analysis could miss functional char-acteristics that may be derived from comparing the two differentbut complimenting datasets.

This study assessed the relationship between perfusion para-meters measured using DCEUS, and SO2 and hemoglobin con-centrationmeasured using PA imaging in two experiments. In thefirst experiment, the spatial relationship between DCEUS and PAparameters was studied at high resolution using two differenthind-limb tumor models with different vascular morphology. Inthe second experiment, the ability of DCEUS and PA imaging todetect global drug–induced functional perfusion andoxygenationchanges in the tumor was studied using an orthotopic model ofhuman breast cancer. Histopathology was used to validate theimaging results.

Materials and MethodsTumor models

Xenograft tumors were induced in mice using either LS174Thuman colorectal cancer cells, or PC3 human prostate cancercells. Both cell lines were cultured in Gibco DMEM (LifeTechnologies Inc.) supplemented with 10% FBS (Hyclone).A total of 1 � 106 cells were injected subcutaneously into theleft hind limb of female 8-week-old SHO mice (Charles RiverLaboratories). The tumors were scanned after they reached adepth of 4–6 mm.

Orthotopic primary breast tumors were induced using 231/LM2-4 cells, a metastatic variant of the human MDA-MB-231breast cancer cell line (12). A small incision in the lower abdom-inal skin was made above the right inguinal mammary fat pad(MFP) of female 8-week-old SHO mice (Charles River Laborato-ries), followed by the injection of 2 � 106 cells into the exposedMFP. The skinwas then closedwith surgical staples and allowed toheal. Tumors were allowed to grow to an approximate volume of200 mm3 before being imaged and treated. All procedures werecompleted with the animal anesthetized under isoflurane and inaccordance with Sunnybrook Health Science Centre's approvedprotocol for Animal Care and Use.

Drug induction of tumor microenvironmental changes usingOxi-4503

Tumor ischemia was induced in the orthotopic breast tumorsmodel using a tubulin-binding vascular disrupting agent: Oxi-4503 (OXiGENE Inc.). Each mouse was given either an intraper-itoneal injection of 50mg/kg of Oxi-4503 or injectable saline andallowed 4 hours for the drug to take effect before being imagedagain for post-treatment effects.

Noninvasive imaging of tumorsAll in vivo imaging was performed using a laser integrated

high-frequency ultrasound system (Vevo LAZR, VisualSonicsInc.). In the first experiment, 6 hind-limb tumors were imagedfor each tumor type. In the second experiment involving druginduction of tumor microenvironmental changes, two groupsof mice were imaged (n ¼ 6 per group) at two separate timepoints, once before the injections were given, then a secondtime 4 hours after injection. All animals were anesthetized with2% isoflurane (Abbott Laboratories Limited) delivered in com-bination with 1 L/minute oxygen. A 27G butterfly needle wasinserted into a lateral tail vein for intravenous injection ofthe ultrasound contrast agent (UCA). A linear array transducerwith fiber optical bundles integrated to each side (LZ-250,fc ¼ 21 MHz) was used to deliver light from a tunable laser(680–970 nm). The tumor was first imaged employing B-modeimaging for region of interest (ROI) selection, followed by PAand contrast mode. PA images were acquired using two alter-nating wavelengths (750 nm and 850 nm). Contrast-enhancedimages were collected after a 50-mL bolus injection of Micro-Marker UCA (VisualSonics Inc.) using nonlinear contrast imag-ing. All three-dimensional (3D) PA data were acquired bymoving the transducer with a stepper motor, at a step size of50 mm.

Image analysisIn the hind-limb experiments, the recorded PA images and

DCEUS cines were loaded into a personal computer and analyzedoffline using in-house MATLAB programs. To estimate the localtime–intensity curves (TIC) in the DCEUS movies, temporalwavelet denoising was applied to each pixel in the scans (13).From the denoised scans, three different perfusion parameterswere estimated, producing high-resolution parametric maps ofthe tumors: peak enhancement, the difference betweenmaximumamplitude, and the baseline intensity of thewash-in curve, used asan indication of the tumor blood volume; wash-in rate, themaximum slope of the curve, used as an indication of bloodflow rate; and area under the curve, the integrated differencebetween the intensity value and its baseline. To estimate the localcorrelation between different perfusion parameters and oxygen-ation level, or hemoglobin concentration estimations, the nor-malized correlation was assessed between each pair of parametersfor each pixel.

In the experiment of tumor microenvironmental changescaused by the injected drug, all image analysis was completedoffline on theVevo Lab. PAdatawere collected as a 3D image stackcontaining 150 to 200 SO2 2D images. A ROI encompassingthe tumor was drawn for each image to determine the averagedSO2 value for each tumor. The ROI selection was done withreference to the anatomic B-mode image. After averaging thecontrast intensity in each ROI, global TICs were generated andquantified using Vevo CQ. Two parameters were taken from theTICs: peak enhancement, and wash-in rate.

Histology of tumorsAll animals were sacrificed after the last contrast-enhanced

ultrasound imaging, for tumor tissue collection. Tumors wereexcised and cut in half in the same plane as the imaging plane andfixed in formalin for 48 hours before processing. All tissues werecut into 5-mm sections, and stained with either hematoxylinand eosin (H&E), or immunostained with carbonic anhydrase

Functional Flow Patterns and Static Blood Pooling in Tumors

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 4321




9 (CA9) or cluster of differentiation 31 (CD31) for hypoxia andangiogenesis, respectively. For the first experiment, a single setof sections was collected from each tumor. As part of the secondexperiment, four sections spaced 350 mm apart were collectedfrom each tumor. The Leica SCN400 (Leica Microsystems Inc.)slide scanner was used for tumor visualization, under brightfield at 40� magnification.

CD31 staining was analyzed with Sedeen Viewer, version4.0.x., counting the number of vessel structures per mm2 area oftumor slice. Necrotic regions, blood vessels, hypoxic areas, andblood lakes were separated from viable tissue regions usingthresholding performed in MATLAB following the use ofthe color deconvolution plugin in the ImageJ software. Subse-quently, necrotic regions and blood lakes were classified using aMATLAB GUI. Regions were segmented and marked as bloodlakes if they included red blood cells or plasma and were notsurrounded by endothelial cells in the matching CD31 staining.CA9 expression was quantified in MATLAB as percentage areaof positive stain per tumor slice. To produce composite histo-logic images, the masks of hypoxic regions produced from CA9stains were aligned with the masks of necrotic regions, bloodlakes, and red blood cells derived from H&E stains (Supple-mentary Fig. S1), using the control point registration MATLABfunction.

Statistical analysisStatistical functions in MATLAB were used to assess the

statistical significance of differences between the various mea-surements performed on PC3 and LS174T tumors. In additionto the correlation between DCEUS and PA parameters, differenthistologic parameters were compared between the two tumors.As Gaussian distributions were not assumed, the Wilcoxonrank-sum test was used and the results were considered signi-ficant for values of 0.05 > P.

The program PASW Statistics 18 was used to assess thestatistical significance of differences in mean values of DCEUSand PA parameters and the histologic markers in response tothe Oxi-4503 treatment. A two-tailed independent samples ttest or paired samples t test was used to assess the significanceof the mean difference.

ResultsLS174T and PC3 tumor xenografts displayed significantlydifferent perfusion and oxygenation patterns

Comparison of DCEUS parametric maps and processed PAscans revealed detailed information about the spatial rela-tionship between perfusion, oxygenation, and hemoglobindistribution in the tumors. Sample images of these functionalmaps taken from LS174T and PC3 tumors are shown in Fig. 1.

DCEUS scans of LS174T tumors presented highly perfused rimwith resolvable blood vessels in their cores. These vessels weresurrounded by many small nonperfused regions (Fig. 1C). Oxy-gen saturation maps showed high SO2 in the rim of the tumorsand low SO2 in their cores, corresponding well to the DCEUSperfusion maps (Fig. 1E). In contrast to perfusion and SO2

patterns, the distribution of hemoglobin was almost uniform inthe tumor (Fig. 1G).

PC3 tumors showed distinctively different anatomic and func-tional maps. While the rims of PC3 tumors were highly perfused,their core contained large poorly perfused and nonperfused

regions (Fig. 1D). Large resolvable vessels were not detected inany of the perfusion maps. High SO2 levels were measuredthroughout the tumors' rim (Fig. 1F), with low SO2 detected inthe core—measured only in regions where the concentration ofhemoglobin was high enough to enable oxygen saturation esti-mations (Fig. 1H). However, vast core regions appeared black,with no detectable hemoglobin concentration.

Overall tumor perfusion correlated with oxygen saturation,but showed discrepancies when compared with hemoglobindistribution

Different perfusion parameters were estimated from high res-olution TICs produced from DCEUS scans. The parameters eval-uated included peak enhancement, wash-in rate, and area underthe curve (AUC, see Fig 1C andD). Similar patterns were observedin the tumor perfusion and oxygen saturation maps, with areasof high SO2 reflective of areas of high blood perfusion. To quanti-fy and evaluate the relationship between the different functionalparameters, the correlation between corresponding pixels in theperfusion parametric maps and oxygen saturation maps wereassessed. A similar analysis was also performed to evaluate thecorrelation between different perfusion parameters and hemo-globin concentration. On the basis of this local spatial analysis,reasonable correlation between the perfusionAUCparameter andSO2 was observed for both LS174T and PC3 tumors (Fig. 2A, r ¼0.63 � 0.06 and r ¼ 0.49 � 0.10 respectively, n ¼ 6). In contrast,the correlation betweenperfusion andhemoglobin concentrationvaried significantly between the two tumor models (Fig. 2B, P ¼0.0022, n ¼ 6) and was negligible for LS174T tumors.

Vessel distribution analysis using histology showedmarkedly different vascular morphology between LS174Tand PC3 tumors, confirming DCEUS observations

We further examined the differences in perfusion patternbetween LS174T and PC3 tumors using immunohistochemistry.Tumors were stained with CD31, a widely used endothelialmarker for tumor neovascularization (14, 15), to compare thevascular morphology of the two tumor models. In LS174Ttumors, the core was characterized by large deformed bloodvessels (Fig. 3A, top right, arrows) that were separated from oneanother by more than 100 mm—the resolution of DCEUS scanstaken in this study. This was reflective of what we observed in theperfusionmaps, where large resolvable vessels were present at thecore of the tumors. In contrast, PC3 tumors were characterized bysmall vessels distributed across its core (Fig. 3A, bottom right,arrows). In comparison with the LS174T tumors, vessels found inPC3 tumors were located much closer to each other. CD31þ

microvessel counts showed significantly lower density of largevessels in PC3 tumors comparedwith LS174T tumors (Fig. 3B, n¼6,P¼0.0087). In addition, LS174T tumors exhibited significantlyhigher density of open lumen vessels compared with PC3 tumors(Fig. 3C, n ¼ 6, P ¼ 0.0022).

Discrepancies between tumor perfusion and hemoglobindistribution were explained with further histologic analysis,revealing hemorrhagic regions in LS174T tumors

We conducted detailed examination of CD31 and H&E stainsto understand the disagreement observed between perfusion andhemoglobin distributions. Histology revealed distinctive mor-phology differences between the two tumor types, with variabledistributions of necrotic regions and blood lakes. CD31 staining

Bar-Zion et al.

Cancer Res; 76(15) August 1, 2016 Cancer Research4322




were used to differentiate functional vessels from blood lakesbased on the absence of endothelial cells (CD31þ stains) fromblood lakes' edges.

On the basis of H&E staining, LS174T tumors were char-acterized by a highly viable periphery (Fig. 4A, left, darkpurple), with a core that contained a mosaic of blood lakes(light pink, dashed arrows), large deformed blood vessels, andnecrotic regions (light purple, arrowheads). The presence ofblood lakes gave insight to the discrepancy that was observedin areas of high hemoglobin concentration where both per-fusion and SO2 were low. Similar to LS174T tumors, the rims

of PC3 tumors were also composed of highly viable tissue(Fig. 4A, right). However, the cores of PC3 tumors werecharacterized by large necrotic regions, with no signs of hem-orrhagic blood lakes. Necrotic regions in both tumors had lowerythrocytes content with few sparse extravasated erythrocytesdetectable (Fig. 4D, yellow arrows), pointing to the remainingof dead vessels.

Thefindings regarding thedifferent compositionof LS174T andPC3 tumors were confirmed with statistical analysis. The propor-tion of necrotic regions in PC3 tumors were significantly higherthan in LS174T tumors (Fig. 4B, n ¼ 6, P ¼ 0.004). On the other

Figure 1.

Spatial patterns of perfusion, oxygen saturation, and hemoglobin concentration in sample LS174T (left column) and PC3 (right column) tumors. A andB, anatomic B-mode images show the location of the hypoechogenic tumors. Mean DCEUS intensity (C and D; AUC normalized by time), oxygensaturation (E and F), and hemoglobin concentration (G and H) are compared. Scale bars, 1 mm.






hand, blood lake areas (10.94 � 6.61 %) were detected only inLS174T tumors (Fig. 4C).

Histology confirmation of tissue necrosis, hypoxia, andblood lakes suggest that DCEUS and PA imaging shouldbe used in conjunction to separate regions of necrosisfrom hypoxia

In addition to H&E and CD31, both LS147T and PC3 tumorswere also stained for CA9 to validate and compare patterns ofhypoxia in the two tumor types. To present a complete histologicimage displaying necrotic regions, blood lakes, and hypoxicregions, CA9 and H&E images were segmented and overlaid.A representative composite histologic image of each tumor isshown in Fig. 5A, along with the matching oxygen saturation andperfusion maps for comparison. Red blood cells are shown aswhite dots, but are not observable in PC3 tumors in the displayedresolution, due to the small sizes of the vessels.

Amidst the areas of necrosis and blood lakes prevalent inthe cores of LS174T tumors, there are also vast areas ofhypoxia (Fig. 5A, top left). This is correlative of the matchingoxygen saturation and perfusion maps (Fig. 5A, left, middle,and bottom plots), where regions of low SO2 were found toconcentrate at the core of the tumors, where only sparse largevessels were found. Areas of necrosis and hypoxia could notbe differentiated using oxygen saturation maps alone, as lowSO2 was observed in the core of the tumors even in the presenceof functional vessels. In contrast to LS174T, only small andscarce areas of hypoxia were found in PC3 tumors, surroundingthe borders of necrotic tumor regions (Fig. 5A, top right).Histology results were in agreement with matched oxygensaturation maps, showing areas of low to no SO2 at the tumorcore (Fig. 5A, right middle). However, in this case, low PAsignal was a result of tissue necrosis and sparse vessels, whichlimited the estimation of SO2. This was confirmed by contrast-enhanced perfusion maps showing large nonperfused areasat the center of PC3 tumors. To quantitatively evaluate thedifference between the SO2 patterns in the two tumor types,histograms of SO2 values normalized to the total tumor areawere calculated (Fig. 5B). LS174T tumor histogram showedhigher frequencies of low SO2 and much lower frequencies ofhigh SO2. In contrast, PC3 tumors displayed relatively simi-lar frequencies of SO2 across the board, but had much loweroverall detected SO2, suggesting that markedly larger regions

in PC3 tumors did not produce sufficient PA signal. Quanti-fication of CA9 stains showed significantly larger hypoxic re-gions in LS174T tumors compared with PC3 tumors (Fig. 5C,n ¼ 6, P ¼ 0.0022).

Functional imaging reveals reduced perfusion and oxygensaturation in response to Oxi-4503 treatment

Contrast assessment of tumor perfusion is shown in Fig. 6A,which displays sample images of an Oxi-4503–treated mousebefore and after treatment. In the pretreatment image, contrastsignal enhancement can be observed in the surrounding tissueand throughout the periphery of the tumor. This, however, ischanged 4 hours after the injection of Oxi-4503, showing avery dark tumor center with much less signs of enhancement.As illustrated in Fig. 6B, both groups of mice showed similarlevels of peak enhancement and wash-in rate before treatment.However, when tumor perfusion was measured again 4 hoursafter, a significant change was observed in the Oxi-4503–treated mice, with an 82.1% and 80.5% decrease in tumorblood volume and flow rate, respectively, P < 0.01. Thisstatistically significant change was not observed with thecontrol mice.

PA assessment of tumor oxygenation is shown in the twolower panels of Fig. 6A, which display sample images of an Oxi-4503–treated mouse before and after treatment. Areas of highoxygenation are shown in bright red, and dark red for lowoxygenation regions, while black represents no oxygenation orlow PA signal. From the images, it can be observed that thetumor is well oxygenated around the periphery, with very lowoxygenation toward the center. This, however, is changed afterthe injection of Oxi-4503, with very low to no SO2 signal seenthroughout the tumor 4 hours after treatment. This wasexpected, as Oxi-4503 is known to induce localized hypoxiainside the tumors as a result of vessel collapse. The averagedpercentage of SO2 for Oxi-4503–treated mice showed a 37.2%decrease in tumor oxygenation after treatment, with P < 0.01(Fig. 6C). This statistically significant change was not observedwith the control.

Histology validation of contrast and PA assessments ofresponse to Oxi-4503 treatment

All tumor slices taken from control and Oxi-4503–treatedmice were stained for CA9 and CD31 to validate contrast and PA

Figure 2.

Correlation between perfusion, oxygen saturation, and hemoglobin concentration in LS174T and PC3 tumors. A, correlation between matching pixels inthe perfusion (DCEUS-AUC) and oxygen saturation parametric maps is fair for both tumors. B, correlation between matching pixels in the perfusion(DCEUS-AUC) and hemoglobin concentration parametric maps is significantly lower for LS174T tumors (� , P < 0.05, n ¼ 6). Data are expressed asmean � SEM.

Bar-Zion et al.





findings. Shown in Fig. 7A are representative images from H&Eand CA9 stains for each group. The control H&E stain displays alarge nonviable region in the inner core of the tumor (lightpurple), and the adjacent image shows brown positive stainingfor CA9 localized as a rim lining this inner core. However, theOxi-4503–treated sample presented a different pattern, withmultiple foci of CA9-positive regions found throughout thetumor and in viable regions as shown by its H&E staining.Figure 7C demonstrates consistency in n¼ 6 with approximately40% more CA9 staining than the control group. Figure 7Bpresents sample sections of tumor slices from each groupstained for CD31, with positive brown staining for vessels. Asshown in the figure, the control group appears to have more

vessels per area than the drug-treated group. This finding wasconfirmed with all of the animals, with Fig. 7D showing that thecontrol group had an approximately 2-fold higher vessel densitythan the Oxi-4503–treated group.

DiscussionIn recent years, scanners combining several noninvasive

imaging modalities are becoming increasingly popular in clini-cal and preclinical imaging due to their ability to provide addi-tional and complimentary information on the imaged tumor(e.g., refs. 16 and 17). In this study, we examined the advan-tages of using DCEUS combined with PA as a diagnostic tool,

Figure 3.

Differences in blood vessel morphology between hind-limb tumor models. A, H&E and CD31 stains reveal noticeable differences in blood vessel morphologyin the core of LS174T and PC3 tumors. LS174T tumors are characterized by scattered large luminal vessels (arrows) and huge blood lakes (dashedarrows). In contrast, PC3 tumors' cores contain mainly small vessels clustered with higher density. Large necrotic regions (arrowheads) can be foundin location lacking functional blood vessels. B, the number of large vessels (surrounded by more than 5 endothelial cells) per tumor area was significantlyhigher in LS174 tumors. C, significantly higher density of luminal vessels was measured in LS174T tumors. �, P < 0.01.






and as a tool for monitoring antivascular treatment. It wasshown that a multimodality imaging scheme combiningDCEUS and PA produce additional information about thepresence of viable tissue, hemorrhagic regions, and necroticareas. To the best of our knowledge, this level of discriminationhas not previously been published with the individualmodalities.

Correlation and discrepancies between perfusion maps andoxygen saturation patterns

Previous studies have shown using histology, the spatialrelationship between hypoxic and perfused tissue regions inspecific xenograft tumors (18). Hypoxia was shown to developin tissue separated from functional vessels by a distance ofapproximately 100 mm. On the basis of this we expected to finda correlative relationship between perfusion and SO2 in ourimaging maps. The pixel-wise correlation between parametricperfusion and oxygen saturation maps in both LS174T and PC3tumors were moderately good (r ¼ 0.63 � 0.06 and r ¼ 0.49 �0.10, respectively, n¼ 6). However, it seems that a global linearregression does not fully capture the relation between thesefunctional parameters. Moreover, in the LS174T tumors, lowSO2 was detected throughout the core even though large

functional vessels were apparent in the DCEUS scans. Histo-pathologic analysis combining information from H&E, CA9,and CD31 stains showed the presence of hypoxia but notnecrosis in the vicinity of functional vessels, in accordancewith previous studies. This inconsistency between PA SO2

estimations and histology results may be due to the fact thattumor cells only show hypoxic stress under extremely lowoxygenation levels (2). The resolution needed for detectionof extremely low SO2 levels could be challenging for nonto-mographic PA scanners. These challenges are especially dom-inant in higher depths, due to the wavelength-dependentabsorption (19).

DCEUS perfusion maps reveal information on vascularmorphology

As single vessels could not be resolved in PA scans, DCEUSperfusion maps were crucial in detecting difference in vascularmorphology between different regions of each tumor, andbetween the two tumors types. Resolvable large vessels weredetected in the core of LS174T tumors, while the rim of thesetumors had relatively uniform perfusion. These results werevalidated using histology, which showed large deformed func-tional vessels in the center of LS174T tumors, surround by areas of

Figure 4.

Differences in blood lakes andnecrotic region distributionbetween hind-limb tumor models.A, H&E histologic staining of sampleLS174T and PC3 tumors illustratethe differences between theprevalence of necrotic regions(arrowheads) and blood lakes(dashed arrows) in these two celllines. Static blood lakes weredifferentiated from functionalvessels based on the presence ofendothelial cells in the lining of thevessels in matching CD31 stains(not presented). B, histologicanalysis of the proportion ofnecrotic region area in LS174T andPC3 tumors. PC3 tumors havesignificantly higher proportionof necrotic area. � , P < 0.01. C,histologic analysis of the proportionof tumor area containing bloodlakes in LS174T and PC3 tumors.LS174T tumors are characterized byhuge blood lakes, while nosignificant blood lakes weredetected in PC3 tumors. D,extravagated red blood cells weredetected in small amounts insidenecrotic regions of PC3 tumors(yellow arrows).

Bar-Zion et al.





Figure 5.

Hypoxia and oxygen saturation patterns in hind-limb tumor models. A, examples of composite histologic image of LS174T and PC3 tumors combininginformation from H&E, CA9, and CD31 stains (top) are presented against SO2 images (middle panels) and perfusion maps (bottom panels). Thecores of LS174T (top left) are embedded with blood lakes (red), small necrotic regions (blue), and large hypoxic regions (yellow). Viable tumortissue is presented in black and red blood cells appear as white dots. In agreement, the cores of LS174T tumors are characterized by low SO2 (middle left).The perfusion map of the LS174T tumor shows large vessels in the tumor core separated by nonperfused areas. In contrast, only small patches ofhypoxic cells are found in PC3 tumors concentrated around the rims of the vast necrotic regions (top right). The absence of hemoglobin innecrotic regions inside PC3 tumors and low signal originating from small vessels can explain the high percentage of tumor area with low PA signalin PC3 tumors (bottom middle). The perfusion map in the right bottom panel resembles the histologic composition of the tumor. Scale bars, 1 mm.B, histograms of SO2 values in LS174T and PC3 tumors normalized to the total tumor area. C, histologic analysis of hypoxia in LS174T and PC3tumors show significant difference in percentage of hypoxic area. � , P < 0.01.






hypoxia (Figs. 3A and 5A, respectively). Previous studies haveshown that larger and leakier vessels are generally associated withregions of hypoxia, affirming our observations (20). In contrast toLS174T tumors, large resolvable blood vessels could not be foundin any of the PC3 DCEUS scans, suggesting that the tumors werecomposed of relatively smaller, compact vessels. This was sup-ported by histology showing small and densely clustered vesselsin the viable regions of PC3 tumors, with minimal signs ofhypoxia (Figs. 3A and 5A, respectively). A certain (global) indi-cation of the hypoxic nature of the LS174T tumors can be derivedfrom the histograms of SO2 measurements from the two tumors

(Fig. 5B). These histograms show markedly higher frequencies oflow SO2 values in LS174T tumors. This is contrasted by PC3tumors, showing extensive areas that did not produce sufficientPA signal, indicative of necrotic regions or dysfunctional vascu-lature. The histologic results were consistent among all tumors,with quantitative analysis showing significantly higher density oflarge and open lumen vessels in LS174T tumors. Previous studieshave shown a positive correlation between the number of openlumen and perfusion (21, 22). A proposed reason for increasednumber of open lumen vessels was that there may be less com-pression from surrounding tumor cells.

Figure 6.

Functional imaging of reaction toOxi-4503 treatment. A, contrastenhanced ultrasound images of Oxi-4503 treated mice before injectionand after injection (top) comparedagainst parametric images of SO2 inOxi-4503–treated mice beforeinjection and after injection (bottom).B, analysis of peak enhancement andwash-in rate in mice injected withsaline or Oxi-4503, before and4 hours after treatment. � , P < 0.01;��, P < 0.001 using paired- andindependent-sample t test. C, analysisof oxygen saturation in mice injectedwith saline or Oxi-4503 before and4 hours after treatment. � , P < 0.01;��, P < 0.001 using paired- andindependent-sample t test. The scalebars are 1 mm long.

Bar-Zion et al.





Combined PA and DCEUS imaging enable distinction betweennecrotic regions and blood lakes

The last aspect of the spatial correlation between DCEUS andPA imaging presented in this study is the ability to distinguishbetween necrotic regions and blood lakes. Both nonviabletissue regions and blood lakes appear in DCEUS scans asnonperfused regions. Comparison between perfusion maps andPA-derived hemoglobin concentration estimations facilitate theclassification of nonperfused areas into two categories: necrotic

regions and blood lakes. While necrotic regions include at mostsparse clusters of extravasated erythrocytes, blood lakes arecomposed of large populations of red blood cells that haveescaped from leaky vessels. By using DCEUS in combinationwith PA imaging the two tissue characteristics can be separated.Nonperfused regions with high concentration of hemoglobinwould likely signify areas with blood lakes, while nonperfusedregions with low hemoglobin concentration would more likelysignify regions of necrosis. Furthermore, correlation between

Figure 7.

Histopathologic signs of Oxi-4503treatment. A, histologic staining (H&Eand CA9) of tumors from control andOxi-4503–treated mice. Histologicstaining of H&E and CA9 revealsregions of nonviable tissue (H&Estaining in light purple) and hypoxia(CA9 staining in brown). Note that thecontrol group displays a largenonviable region in the inner core ofthe tumor surrounded by localizedbrown positive staining for CA9 whilethe Oxi-4503 presented a differentpattern, with multiple foci of CA9-positive regions found throughout thetumor. B, histologic CD31 stainingshows microvessel densitydistribution in regions of tumors fromOxi-4503–treated and control mice.The treated tumors show lowermicrovessel density compared withthe control tumors. Images aretaken at �20 magnification.C, histologic analysis of hypoxia inOxi-4503–treated and control tumors.� , P < 0.01. D, histologic analysis ofmicrovessel density in tumors ofOxi-4503–treated and control mice.�� , P < 0.001.






perfusion and hemoglobin concentration can also help indicatethe presence of blood lakes verses necrotic tissue. In tissueswhere substantial amount of blood lakes are present, correla-tion would likely be low, such as observed in LS174T tumors(Fig. 2B). On the other hand, higher correlation betweenperfusion and hemoglobin concentration would likely appearin necrosis dominated tissue, such as PC3 tumors. Our obser-vations were validated using histology, where large blood lakesdominated LS174T tumors (Fig. 4C), while necrosis was sig-nificantly higher in PC3 tumors (Fig. 4B).

Multimodality imaging facilitates functional assessment ofresponse to Oxi-4503 treatment

The ability of PA and DCEUS to facilitate functional assess-ment of response to Oxi-4503 treatment was examined;Oxi-4503 was chosen for its well-researched effects of causingtumor vessel collapse, leading to localized hypoxia, and even-tually hypoxia-induced necrosis (23, 24).

DCEUS imaging was able to detect the perfusion changes inthe drug-treated group showing almost no microbubble signalenhancement 4 hours after Oxi-4503 injection. The 80% de-crease in both tumor blood volume and flow rate indicatorswas comparable with previous studies that demonstrated onlya small viable rim remains in the tumor after treatment withOxi-4503 (25). There were also, however, slight peak enhance-ment and wash-in rate changes observed in the control group.These changes may have been a result of slight plane shifts, ortumor orientation shifts between the two imaging time points.We subsequently looked at CD31 staining of the tumor tissueto validate our contrast findings. The histologic analysis show-ed a 50% decrease in the number of vessel per area in the Oxi-4503 mice compared with the control. This level of microvesseldensity change was slightly lower than our quantified contrastchange, which may be due to the fact that CD31 stains for allexisting endothelial cells, whereas DCEUS address functionalvasculature in real time.

Relative quantification of SO2 was performed using recon-structed parametric maps of dual-wavelength PA imaging. Wewere able to observe a noticeable decrease of around 40% intumor oxygenationwithOxi-4503–treatedmice,whereas changeswere negligible in the control mice. Subsequent validation byhistology also showed 45% less hypoxia expression in the controlmice compared to Oxi-4503 treated. The presence of large viablehypoxic regions 4 hours after the Oxi-4503 treatment is reason-able as some tissues can survive for more than 5 hours withoutblood supply before undergoing necrosis (26).The histologicfindings in this study support PA findings of drug-inducedhypoxia.

Limitations and future research directionsThere are two main limitations to the imaging methods

studied in this work. The first is the limited resolution ofnontomographic PA imaging, both spatially and in separatinglow SO2 values. New semitomographic methods that enable

the combination of DCEUS and PA are being developed andthe effect of improved PA resolution on the capabilities of thesemultimodality scanners is a subject of future research. Thesecond limitation is that currently, DCEUS imaging using bolusinjections is limited to a single imaging plane. 3D ultrasoundprobes have gained popularity in the past few years and thecombination of 3D DCEUS and 3D PA could decrease thevariability of the scans and enable the study of the 3D mor-phology of the vasculature.

ConclusionsIn summary, DCEUS and PA produced high resolution maps

of perfusion and SO2 and hemoglobin concentration. Thesephysiologic parameters provide important information aboutmorphology and functionality of the vasculature feeding thetumor and its microenvironment. In addition, the combinationof these two modalities enables the classification of tumor areato viable regions, necrotic tissue regions, and blood lakes. Thisdetection of blood lakes could aid in determining the leakinessof the tumor. Furthermore, DCEUS and PA imaging were ableto monitor, with high sensitivity, known drug-induced changesin the tumors, with close correlations to the current goldstandard histology measurements. These results were validatedquantitatively using histology. Taken together, our findingsstrongly support that DCEUS and PA imaging could be effectivetools for the quantitative functional assessments of the micro-environment of various tumor models.

Disclosure of Potential Conflicts of InterestF.S. Foster reports receiving a commercial research grant from and is a

consultant/advisory board member for VisualSonics. No potential conflicts ofinterest were disclosed by the other authors.

Authors' ContributionsConception and design: A. Bar-Zion, M. Yin, D. Adam, F.S. FosterDevelopment of methodology: A. Bar-Zion, M. Yin, D. Adam, F.S. FosterAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): A. Bar-Zion, M. Yin, F.S. FosterAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): A. Bar-Zion, M. YinWriting, review, and/or revision of the manuscript: A. Bar-Zion, M. Yin,D. Adam, F.S. FosterAdministrative, technical, or material support (i.e., reporting or organiz-ing data, constructing databases): A. Bar-Zion, F.S. FosterStudy supervision: D. Adam, F.S. Foster

AcknowledgmentsThe authors thank the Canadian Institutes of Health Research, the Terry Fox

Foundation, and VisualSonics Inc. for financial support.The costs of publication of this article were defrayed in part by the

payment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received February 4, 2016; revised April 5, 2016; accepted May 26, 2016;published OnlineFirst June 20, 2016.

References1. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell

2011;144:646–74.2. H€ockel M, Vaupel P. Tumor hypoxia: definitions and current clinical,

biologic, and molecular aspects. J Natl Cancer Inst 2001;93:266–76.

3. Eckersley RJ, Chin CT, Burns PN. Optimising phase and ampli-tude modulation schemes for imaging microbubble contrastagents at low acoustic power. Ultrasound Med Biol 2005;31:213–9.

Bar-Zion et al.





4. Wilson SR, Burns PN. Microbubble-enhanced US in body imaging: whatrole? Radiology 2010;257:24–39.

5. Hudson JM, Williams R, Tremblay-Darveau C, Sheeran PS, Milot L, Bjar-nason GA, et al. Dynamic contrast enhanced ultrasound for therapymonitoring. Eur J Radiol 2015;84:1650–7.

6. Xu M, Wang LV. Photoacoustic imaging in biomedicine. Rev Sci Instrum2006;77:041101.

7. Beard P. Biomedical photoacoustic imaging. Interface Focus 2011;1:602–31.8. Zhang HF, Maslov K, Sivaramakrishnan M, Stoica G, Wang LV. Imaging

of hemoglobin oxygen saturation variations in single vessels in vivousing photoacoustic microscopy. Appl Phys Lett 2007;90:053901.

9. Hu S, Wang LV. Photoacoustic imaging and characterization of the micro-vasculature. J Biomed Opt 2010;15:011101.

10. Hashizume H, Baluk P, Morikawa S, McLean JW, Thurston G, Roberge S,et al. Openings between defective endothelial cells explain tumor vesselleakiness. Am J Pathol 2000;156:1363–80.

11. Gerling M, Zhao Y, Nania S, Norberg KJ, Verbeke CS, Englert B, et al. Real-time assessment of tissue hypoxia in vivo with combined photoacousticsand high-frequency ultrasound. Theranostics 2014;4:604–13.

12. Munoz R, Man S, Shaked Y, Lee CR, Wong J, Francia G, et al. Highlyefficacious nontoxic preclinical treatment for advanced metastatic breastcancer using combination oral UFT-cyclophosphamide metronomicchemotherapy. Cancer Res 2006;66:3386–91.

13. Bar Zion A, Tremblay-Darveau C, Yin M, Dan A, Foster F. Denoising ofcontrast enhanced ultrasound cine sequences based on a multiplicativemodel. IEEE Trans Biomed Eng 2015;99:1–1.

14. Wang D, Stockard CR, Harkins L, Lott P, Salih C, Yuan K, et al. Immuno-histochemistry in the evaluation of neovascularization in tumor xeno-grafts. Biotech Histochem 2008;83:179–89.

15. Vanzulli S, Gazzaniga S, Braidot MF, Vecchi A, Mantovani A, WainstokDCR. Detection of endothelial cells by MEC 13.3 monoclonal antibodyin mice mammary tumors. Biocell 1997;21:39–46.

16. Bar-Shalom R, YefremovN, Guralnik L, Gaitini D, Frenkel A, Kuten A, et al.Clinical performance of PET/CT in evaluation of cancer: additional value

for diagnostic imaging and patient management. J Nucl Med 2003;44:1200–9.

17. Deroose CM, De A, Loening AM, Chow PL, Ray P, Chatziioannou AF, et al.Multimodality imaging of tumor xenografts and metastases in mice withcombined small-animal PET, small-animal CT, and bioluminescenceimaging. J Nucl Med 2007;48:295–303.

18. Rijken PFJW, Bernsen HJJA, Peters JPW, Hodgkiss RJ, Raleigh JA, van derKogel AJ. Spatial relationship between hypoxia and the (perfused) vascularnetwork in a human glioma xenograft: a quantitative multi-parameteranalysis. Int J Radiat Oncol 2000;48:571–82.

19. Needles A,Heinmiller A, Sun J, TheodoropoulosC, BatesD,HirsonD, et al.Development and initial application of a fully integrated photoacousticmicro-ultrasound system. IEEE Trans Ultrason Ferroelectr Freq Control2013;60:888–97.

20. Jain RK. Normalization of tumor vasculature: an emerging concept inantiangiogenic therapy. Science 2005;307:58–62.

21. Chauhan VP, Martin JD, Liu H, Lacorre DA, Jain SR, Kozin SV, et al.Angiotensin inhibition enhances drug delivery and potentiates chemo-therapy by decompressing tumour blood vessels. Nat Commun 2013;4:2516.

22. Stylianopoulos T, Jain RK. Combining two strategies to improve perfusionand drug delivery in solid tumors. Proc Natl Acad Sci U S A 2013;110:18632–7.

23. Salmon HW, Siemann DW. Effect of the second-generation vascular disrupt-ing agent OXi4503 on tumor vascularity. Clin Cancer Res 2006;12:4090–4.

24. Siemann DW, Horsman MR. Vascular targeted therapies in oncology. CellTissue Res 2009;335:241–8.

25. Daenen LG, Shaked Y, Man S, Xu P, Voest EE, Hoffman RM, et al.Low-dose metronomic cyclophosphamide combined with vascu-lar disrupting therapy induces potent anti-tumor activity in pre-clinical human tumor xenograft models. Mol Cancer Ther 2009;8:2872–81.

26. Petrasek PF, Homer-Vanniasinkam S, Walker PM. Determinants of ische-mic injury to skeletal muscle. J Vasc Surg 1994;19:623–31.






2016;76:4320-4331. Published OnlineFirst June 20, 2016.Cancer Res Avinoam Bar-Zion, Melissa Yin, Dan Adam, et al. Photoacoustic ImagingRevealed by Combined Contrast-Enhanced Ultrasound and Functional Flow Patterns and Static Blood Pooling in Tumors

Updated version

10.1158/0008-5472.CAN-16-0376doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/06/18/0008-5472.CAN-16-0376.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/76/15/4320.full#ref-list-1

This article cites 26 articles, 7 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/76/15/4320.full#related-urls

This article has been cited by 6 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/76/15/4320To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-16-0376

http://cancerres.aacrjournals.org/content/suppl/2016/06/18/0008-5472.CAN-16-0376.DC1

http://cancerres.aacrjournals.org/content/76/15/4320.full#ref-list-1

http://cancerres.aacrjournals.org/content/76/15/4320.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/76/15/4320


functionalflowpatternsandstaticbloodpooling in tumors … · potential to facilitate the monitoring...

Documents