2.1.

Chemicals

All chemicals, reagents and solvents were purchased from indicated suppliers and used without further purification: SNARF-4F 5-(and-6)-carboxylic acid (Invitrogen; Carlsbad, California), American Chemical Society (ACS)-grade hydrochloric acid (HCl) (EMD; Gibbstown, New Jersey), ACS-grade sodium hydroxide (NaOH) (Fisher Scientific; Pittsburgh, Pennsylvania), laboratory-grade gelatin from porcine skin, type A (Sigma Aldrich; Milwaukee, Wisconsin), bovine hemoglobin (Sigma Aldrich; Milwaukee, Wisconsin), and intralipid 20% (Fresenius Kabi AB).

2.2.

Aqueous pH Solutions

Solutions ranging from pH 3–10 in 0.5-pH increments were prepared from a 1.0-μM aqueous stock solution of SNARF-4F separated into 10-mL aliquots. The pH of each aliquot was adjusted via microadditions of concentrated HCl and/or NaOH. Solution pH adjustments were monitored with a calibrated pH meter (MP220 pH Meter, Mettler Toledo, Columbus, OH) under positive nitrogen pressure with vigorous stirring. Over the course of the experiment, the solutions were wrapped in aluminum foil to minimize light exposure. Spectroscopic analysis and/or MSFI of pH-adjusted samples were performed immediately following pH adjustment and stabilization.

2.3.

Preparation of Biological Phantoms

To validate the ability of MSFI to measure pH in biological environments, tissuelike phantoms with the approximate photon scattering and absorption properties of biological tissues were prepared.²¹ First, a 10% gelatin solution was prepared by adding gelatin to heated deionized water (40–50°C) with constant stirring. The solution was then cooled to 30–40°C, at which the desired amounts of bovine hemoglobin and intralipid were added, resulting in a 42.5-μM hemoglobin and 1% intralipid phantom mixture. Three parts buffered dye solution [acetate buffer (pH 4.3, 5.6), phosphate buffer (pH 6.2, 7.3), tris-HCl buffer (pH 8.4), and carbonate buffer (pH 9.5)] and one part phantom mixture were combined and dispensed into a chilled well plate and refrigerated at 4°C until solid.

2.4.

Fluorimetry

The spectroscopic characteristics of pH-adjusted aqueous SNARF-4F solutions were investigated using a Photon Technology International QuantaMaster^TM 50 fluorimeter (Birmingham, New Jersey) and a standard, 1-cm path-length quartz cuvette (NSG Precision Cells 517BES10). An excitation wavelength (λ_ex) of 523 nm was used for single-wavelength studies, whereas an excitation range of 425–650 nm in increments of 25nm was used for multiple-wavelength studies. Emission scans were collected every nanometer over the emission wavelength (λ_em) range of 400–800 nm with an integration time of 0.1 s and a shutter width of 1.5 nm. Three acquisitions were averaged for each λ_ex. Excitation emission matrices of fluorescence emission as a function of excitation were generated in MATLAB (The MathWorks, Inc., Natick, Massachusetts). Fluorimetry data were used in conjunction with

2.5.

MSFI: Aqueous Solutions

Aqueous SNARF-4F solutions ranging in pH from four to eight in increments of ~1.0 pH unit were imaged using the Maestro^TM Q FLEX In Vivo Imaging System from CRi, Inc. (Woburn, Massachusetts). Solutions were imaged in 1.5-mL microcentrifuge tubes that were prerinsed with pH-adjusted solution prior to addition of dye solution. All images were obtained using Maestro Q Filter Set B, which features a bandpass excitation filter with transmission centered at 525 nm (full width at half maximum = 47 nm) and a long-pass emission filter with 560 nm cut in. A spectral library was generated that included emission spectra for fully protonated/deprotonated dye species as well as autofluorescence from the microcentrifuge tube plastic. This library was used throughout the solution phase experiments for spectral unmixing. Quantitative analysis of the autofluorescence signal arising from the microcentrifuge tubes demonstrated that >95% of the autofluorescence could be removed by spectral unmixing (data not shown). Unmixed dye fluorescence intensity data were obtained from a manually drawn region of interest (ROI) shaped and sized to include the central region of the microcentrifuge tube. The data were fitted to a nonlinear sigmoidal model and plotted against the analytically measured solution pH. Uncorrected and corrected SNARF-4F pK_a values were calculated.

2.6.

MSFI—Biological Phantoms

Using a protocol similar to that used for the solution phase experiments, dye containing tissuelike phantoms ranging in a pH of 4–9.5 were imaged with the Maestro Q system. Spectral unmixing was achieved by applying the previously derived solution phase spectral library with the inclusion of autofluorescence from tissuelike phantom controls. Unmixed fluorescence intensity data were obtained from an ellipselike manually drawn ROI sized to include the entirety of a single well. The data were fit to a nonlinear sigmoidal model and plotted against the analytically measured solution pH. Uncorrected and corrected SNARF-4F pK_a values were calculated.

2.7.

Ex Vivo pH Mapping of Human Xenograft Tumor Tissue

All studies involving the use of animal models were conducted in accordance with Vanderbilt University Institutional Animal Care and Use Committee and applicable federal guidelines. To generate the human colorectal cancer (CRC) model used for tumor imaging, 200,000 DiFi human CRC cells were subcutaneously injected on the left flank of athymic nude mice, as previously described.¹⁷ Experiments were conducted two to three weeks postinoculation, when tumors reached ~250 mm³. For pH_e mapping, 1.0 nmol of SNARF-4F in 200 μL of saline was administered to xenograft-bearing mice via intravenous injection under inhalation of anesthesia (2% isofluorane). Following dye administration, mice were allowed to briefly recover from anesthesia and given access to food and water ad libitum during a 15-min uptake period. Mice were then sacrificed and tumor tissue harvested. Immediately following collection (<2 min), tumors were macrodissected into two equivalent hemispheres, positioned with the intratumoral facets facing toward the camera, and imaged using the Maestro Q system. Fluorescence images were collected from 560 to 850 nm and unmixed using the previously described spectral library (excluding autofluorescence from plastic). Unmixed fluorescence intensity data were obtained from a manually drawn ROI (~2% of the total area) that was applied to multiple areas of the tumor in order to ascertain pH heterogeneity. A gradient pH_e map was generated in MATLAB from each MSFI unmixed image using

3.1.

Spectroscopy

The fluorescence spectroscopy of the SNARF family of organic dyes is dependent on the protonation state of the compound. Typically, the protonated and deprotonated states of the compound yield distinct, yet overlapping, emission peaks that may be separated by as much as 60 nm.²² For these studies, aimed at measuring pH in tissue, an important determinant was selection of a dye possessing a pK_a near the anticipated pH of the specimens being assayed. These studies utilized SNARF-4F, with a known pK_a of 6.4. Thus, at pH values well below its pK_a, the dye exists predominately in a protonated state and exhibits fluorescence emission at 580 nm when excited at 523 nm (Fig. 1). SNARF-4F exists in a predominately deprotonated state at pH values well above the pK_a and exhibits fluorescence emission at 640 nm when excited at 523 nm (Fig. 1). Importantly, across the physiologically relevant pH range of 5–7, both HA and A⁻ species exist in equilibrium. Because of the unique emission spectroscopy of the HA and A⁻ species, spectral unmixing and quantitative measurement of each species facilitate determination of pH in accordance with a modified Henderson–Hasselbalch equation.

Fig. 1

Combined fluorescence emission spectra (excitation 523 nm) of aqueous SNARF-4F samples ranging in pH from ~4 to 7. The emission spectrum of pH 3.9 is representative of subsequently decreasing pH values (data not shown) while the emission spectrum of pH 7.2 is representative of subsequently increasing pH values (data not shown).

Conceptually, the simplest pH imaging study coupling SNARF-4F and the Maestro system would feature utilization of a single excitation band and simultaneous collection and spectral unmixing of both the HA and A⁻ emissions. However, an important assumption of this approach is the requirement for equivalent excitation and collection of both the HA and A⁻ species, which may be thought of as two separate fluorophores in this experiment. Because the emission of the A⁻ species is significantly redshifted (ca. 60 nm) from the HA species, it follows that equivalent excitation of each species using a common excitation wavelength is unlikely. Therefore, we explored the relationships between excitation wavelength, pH, and the resulting emission spectroscopy of both the HA and A⁻ SNARF-4F species using standard fluorimetry. We found that the HA species exhibited maximum fluorescence emission at an excitation wavelength of ~525 nm (emission = 590 nm), while the A⁻ species exhibited maximum fluorescence emission at an excitation wavelength of ~575 nm (emission = 650 nm) (Fig. 2). Optimization of the imaging assay to utilize only a single filter set could be accomplished using a single filter set included with the Maestro system [denoted as “B”, bandpass excitation 525/25 nm, long-pass emission 560 nm, (Fig. 3)]. This filter set was ideally suited for exciting the HA species but was suboptimal for exciting the A⁻ species. We found the long-pass emission filter contained within the B filter set was suitable for simultaneous collection of both the HA and A⁻ emission, with minor attenuation (~5%) of the HA emission from 530 to 560 nm.

Fig. 2

Excitation emission matrices of SNARF-4F fluorescence emission as a function of excitation wavelength for HA [(a) pH 4.5] and A⁻ [(b) pH 10.0]. Red marks located along the excitation axis show the approximate bandpass transmission of the Maestro Q B excitation filter, while the red mark on the emission axis shows the cut-on point for the long-pass emission filter. The excitation-emission coordinates for the point at which maximum fluorescence emission intensity is achieved are (525, 587) for (a) and (575, 653) for (b).

Fig. 3

Transmission spectrum for (a) bandpass excitation filter and (b) long-pass emission filter supplied as Filter Set B with the Maestro Q imaging system.

To compensate for the suboptimal excitation of the A⁻ species, a correction factor (C _F = 2.1 ± 0.03) was derived from Eq. 1. Equation 1 defines the ratio of the observed emission intensity at 650 nm for the A⁻ species when excited at 525nm (λ₁), the excitation wavelength in the Maestro, compared to the emission intensity at 650 nm when optimally excited at 575 nm (λ₂).

3.2.

MSFI

To validate C _F for use in MSFI assays, aqueous SNARF-4F solutions with pH values between 4 and 8 and tissuelike phantoms with pH values between 4 and 9.5 were imaged using filter set B and spectrally unmixed using the previously established spectral library. Following spectral unmixing, manual ROIs were drawn on unmixed images corresponding to the HA and A⁻ species [Figs. 4a and 4b]. Accordingly, the measured fluorescence intensities for each species were used to calculate the observed, uncorrected pK_a for the dye using a modified Henderson–Hasselbalch equation, substituting the fluorescence emission intensity for each species in place of HA and A⁻ concentration. Prior to correction of the A⁻ intensity, fitting the unmixed HA and A⁻ intensities to a sigmoidal curve and plotting versus pH [Figs. 4c and 4d] resulted in intersection between the two curves (by definition, pK_a) at 6.7 ± 0.3 for both the solution phase and biological phantom data (Table 1). Though these observed values are slightly higher than the known pK_a of 6.4,²² application of the fluorimetry determined C _F to the A⁻ intensities of all data resulted in indistinguishable calculated pK_a values of ~6.4 (Table 1) using Eq. 2, therefore validating C _F derived from the fluorimetry experiments in multiple mediums.

Fig. 4

Multispectral fluorescence imaging of (a) aqueous SNARF-4F solutions ranging in pH from approximately 4 to 8 and (b) dye-containing tissuelike phantoms ranging in pH from approximately 4 to 9.5. Spectrally mixed, pseudocolor composite images are shown in row i of both (a) and (b). Spectrally mixed composite images were unmixed to visualize emission from the HA [row ii of both (a) and (b)] and A⁻ [row iii of both (a) and (b)] species. Pseudocolor, unmixed composite images are shown in row iv of both (a) and (b), where the HA fluorescence emission is pseudocolorized in green and the A⁻ fluorescence emission is pseudocolorized in red. Quantified average photon intensities from (a) are given in (c) while intensities from (b) are given in (d). HA (red) and A⁻ (blue) species are displayed across pH as assessed by MSFI. Nonlinear fits of MSFI data (solid colored lines) result in goodness of fits of 7 (c) and 33 (d) degrees of freedom. Observed pK_a values (Table 1), the pH at which [HA] = [A⁻], are noted by solid black lines. Correction of the A⁻ intensity using C _f at each point (dotted blue line) results in calculated pK_a values (noted by dashed black lines) that are equivalent within one standard deviation of one another (Table 1).

Fig. 5

MSFI of dye-perfused human CRC cell line xenograft. (a) Spectrally mixed pseudocolor fluorescence image. Spectrally unmixed images show (b) HA and (c) A⁻ fluorescence emission. (d) Map illustrating the selected ROI placement for pH measurements performed in tumor regions corresponding to the values shown in Table 2. Pixel-by-pixel pH_e map generated using the HA fluorescence emission and the corrected A⁻ emission (e). Color bar corresponds to the calculated pH in each pixel.

Table 2

MSFI-measured photon intensities and calculated pH values for DiFi CRC xenograft tumor. Location for each ROI is shown in Fig. 5d.

Table 1

Observed (uncorrected) and calculated (corrected) pKa values for different SNARF-4F containing mediums.

3.3.

Ex Vivo pH Mapping of Xenograft Tumors

We next explored utilizing the MSFI approach for spatially resolved determination of pH_e in freshly resected xenograft tumor tissue. Using the Maestro Q system, we observed perfusion of SNARF-4F throughout the entire tumor, though accumulation tended to be heterogeneous [Fig. 5a]. Increased levels of accumulation were noted in tissues exhibiting elevated necrosis. Nonuniform accumulation of dye in disorganized tissue such as tumors is not unexpected. It is therefore important to emphasize that the ratiometric nature of the proposed MSFI approach renders pH measurement inherently concentration independent with respect to the dye, provided that detectable quantities are present within tissues of interest. Applying the previously established spectral library for the HA and A⁻ species to the unmixed data, fluorescence emission of the HA [Fig. 5b] and A⁻ [Fig. 5c] species could be visualized within tumor tissue. To quantify intratumoral pH_e, the fluorescence intensity for the HA and A⁻ species was measured by manual ROI analysis [Fig. 5d]. The measured intensity of the A⁻ species was corrected using C _F, and tumor pH_e for each ROI was calculated using Eq. 2 (Table 2). We found that pH_e within the tumor was heterogeneous with a mean pH of 6.9 ± 0.1. These observations are in agreement with those reported by previous studies where pH_e was assessed across multiple tumor types by alternative techniques. ^{23, 24, 25, 26} Expanding on this, to generate a high-resolution pH map of tumor tissue, the HA and corrected A⁻ pixel intensities were imported into MATLAB, where the pH was calculated, fit to a color bar, and displayed across each image pixel [Fig. 5e]. Although much of the tumor could be considered acidic with respect to normal tissue, we observed that tumor regions exhibiting the lowest pH tended to coincide with the highest degrees of necrosis in this model. These results suggest the feasibility of MFSI for ex vivo measurement of pH_e in preclinical tissue specimens.