Near-simultaneous quantification of glucose uptake, mitochondrial membrane potential, and vascular parameters in murine flank tumors using quantitative diffuse reflectance and fluorescence spectroscopy.

The shifting metabolic landscape of aggressive tumors, with fluctuating oxygenation conditions and temporal changes in glycolysis and mitochondrial metabolism, is a critical phenomenon to study in order to understand negative treatment outcomes. Recently, we have demonstrated near-simultaneous optical imaging of mitochondrial membrane potential (MMP) and glucose uptake in non-tumor window chambers, using the fluorescent probes tetramethylrhodamine ethyl ester (TMRE) and 2-N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose (2-NBDG). Here, we demonstrate a complementary technique to perform near-simultaneous in vivo optical spectroscopy of tissue vascular parameters, glucose uptake, and MMP in a solid tumor model that is most often used for therapeutic studies. Our study demonstrates the potential of optical spectroscopy as an effective tool to quantify the vascular and metabolic characteristics of a tumor, which is an important step towards understanding the mechanisms underlying cancer progression, metastasis, and resistance to therapies.


Introduction
Interest in tumor metabolism and vasculature continues to grow in the field of cancer research. Beyond the extensively studied "Warburg effect" [1], which reflects a tumor's propensity for aerobic glycolysis, mitochondria have recently gained recognition for their distinct contribution to tumor oxidative metabolism. In fact, some tumors can switch their primary metabolic mode between glycolysis and oxidative phosphorylation to meet increased energy demands required for proliferation and metastasis [2][3][4], as well as adapt to stressors including chemo and molecular therapies [5]. The vascular microenvironment is another critical enabler of tumor survival and recurrence [6] since it will influence tumor metabolism. Specifically, vascular oxygen saturation (SO 2 ) and hemoglobin concentration ([Hb]) within the tumor microenvironment influences metabolism by affecting metabolic substrate availability; conversely, metabolic needs affect vascular parameters by dictating substrate demand [7]. Taken together, glycolysis, mitochondrial oxidative phosphorylation, and vascular parameters all play a key role in understanding how the metabolic characteristics of tumors impact therapeutic outcome, and the ability to monitor all of them simultaneously can play an important role in cancer pharmacology research. A common tool for assessing glycolysis and oxidative phosphorylation is the ubiquitous Seahorse Assay, which treats cells in vitro with chemical perturbations to measure two functional endpoints: oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) [8][9][10][11][12][13]. The Seahorse instrument is widely used for biomedical research as evidenced by over 500 journal articles in 2017 alone [14]. In the cancer research field, the Seahorse Assay has been frequently used to investigate metabolic characteristics associated with cell proliferation and apoptosis [15], response to therapeutic stress [16], and metabolic reprogramming in metastatic breast cancer cells [17], to name just a few examples. However, a major limitation of the Seahorse Assay is its utility is restricted to in vitro cell samples, preventing both the investigation of metabolism in vivo and the effect of the vasculature within the tumor microenvironment on tumor metabolism. Metabolomics, the gold standard for measurement of metabolites, can quickly screen a large number of molecules and map metabolic networks [18,19]. Metabolomics has been frequently used to evaluate the effect of chemotherapy on metabolite levels [20], understand how the "Warburg effect" promotes tumor survival [21], and identify unique metabolites in metastatic tumors [22]. However, metabolomics is mainly used for ex vivo tissue samples and only provides a snapshot of the tissue's metabolic state. By using 13 C labeled glucose or other metabolites, it is possible to acquire metabolic fluxes [23]. However, this requires highly sophisticated technology and software for such analyses. There are several techniques currently available for in vivo metabolic imaging including Positron Emission Tomography (PET) and magnetic resonance imaging (MRI). PET is able to quantify glucose uptake using 18 F-FDG [24][25][26][27][28][29]. PET can also image hypoxia using other radio-labeled probes (e.g. 18 FFMISO) [30], however, both glucose uptake and hypoxia cannot be measured simultaneously. Magnetic resonance spectral imaging (MR(S)I) can report on mitochondrial metabolism and glycolysis using special tracers such as 31 P or hyperpolarized 13  . While each of these tools enable organ-level imaging, the endpoints provide relative measurements and the endpoints cannot be imaged simultaneously. Ultimately, however, there are tradeoffs between the MRI, PET and optical spectroscopy and we believe that the latter will fit well within the suite of technologies available for metabolic and vascular assessment of tumors.
Optical spectroscopy and imaging can leverage endogenous contrast or be coupled with appropriate indicators to provide quantitative endpoints related tumor metabolism and its associated vasculature in vivo [38][39][40][41][42][43][44]. Two endogenous fluorophores in tissue, reduced nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD) [43,45] provide insights into the reduction-oxidation state in the electron transport chain, as NADH fluorescence is increased in tumors reliant on glycolysis, whereas increased FAD fluorescence corresponds to more oxidative tumors [46]. It should be noted that the ratio of NADH and FAD does not directly report on glucose uptake nor mitochondrial activity given that cytosolic NAD(P)H also contributes to endogenous fluorescence that is similar to that of NADH, and FAD fluorescence is relatively weak [47]. Our group has developed a novel technique to quantify glucose uptake and mitochondrial metabolism using the fluorescent In this study, we demonstrate near-simultaneous optical spectroscopy and MC modeling [44, 55, 56] to quantify vascular SO 2 and [Hb] along with 2-NBDG and TMRE uptake of a flank tumor model in vivo. We were able to adopt the sequential injection strategy previously developed for intravital microscopy to perform near-simultaneous quantification of glucose uptake and mitochondrial membrane potential in vivo. We observed that 4T1 tumors had a significant increase in 2-NBDG uptake, TMRE uptake and [Hb], but decreased SO 2 and scattering, compared to normal flank tissues. We also demonstrate that glucose uptake and oxidative phosphorylation quantified from spectroscopy were concordant with our formerly reported microcopy results [50]. Although 4T1 tumors have been considered as primarily glycolytic in the past [16, 22], our results demonstrate that it also relies on mitochondrial metabolism. A strong correlation was also observed between the metabolic (2-NBDG and TMRE) and vascular endpoints (SO 2 and [Hb]) in 4T1 tumors. None of these correlations were observed in normal tissue. The correlation studies suggest a strong coupling between substrate availability and demand in 4T1 tumors, but not in quiescent normal tissue. Quantitative optical spectroscopy is an effective tool for near-simultaneous in vivo quantification of vascularity and the major axes of metabolism in cancer, which is important for studying the mechanisms underlying cancer progression, metastasis, and resistance to therapies. Our technology will be able to seamlessly connect insights from in vitro studies to those from whole animal or patient imaging through local tissue measurements in vivo.

Flank tumor model
All in vivo experiments described here were performed according to a protocol approved by Duke University Institutional Animal Care and Use Committee (IACUC). Female athymic nude mice (nu/nu, NCI, Frederick, Maryland) aged 8 to 10 weeks were used for these studies. All animals were housed in an on-site housing facility with ad libitum access to food and water and standard 12-hour light/dark cycles. A non-tumor animal study was used to further evaluate the sequential injection protocol for near-simultaneous optical spectroscopy of 2-NBDG and TMRE in a flank tumor model. A total of 20 non-tumor animals were randomly assigned to (1) TMRE only group (N = 6, 100 µL dose of 75 µM); (2) 2-NBDG only group (N = 7, 100 µL dose of 6 mM); and (3) TMRE→2-NBDG with 20-minute delay group (N = 7), the amounts and concentrations of TMRE and 2-NBDG used in the sequential injection group were exactly same as those used in the 2-NBDG only or TMRE only groups. Mice were fasted for 6 hours before optical spectroscopy to minimize variance in metabolic demand [48]. A separate group of tumor animal study (N = 9) was performed to demonstrate the usability of our protocol in solid tumors. These mice received a subcutaneous injection of 4T1 cells (0.1 mL of cell solution with a concentration of 10 6 cells/mL) in the right flank under isoflurane anesthesia. The mice were returned to the cage and monitored continuously for two weeks after the tumor cell injection. On day 10 after the tumor injection (tumor size is ~6 mm in diameter), mice were fasted for 6 hours and anesthetized with isoflurane for the spectroscopy study. All tumor animals received injection protocol (3) TMRE→2-NBDG with 20-minutes delay (100 µL dose of 75 µM TMRE→100 µL dose of 6 mM 2-NBDG).

Optical measurements
The optical spectroscopy system and fiber-probe were described in detail previously [44]. The optical measurement system is briefly illustrated in Fig. 1(A). The fiber-optics probe ( Fig.  1(B)) consisted of 19 fibers for illumination and 18 surrounding fibers for detection. The numerical apertures of illumination fibers and detection fibers were all 0.22. The core diameters of the illumination and detection fibers were 200 μm. The sensing depth of the probe was estimated to be 1.5 mm from tissue-mimicking phantom studies. To count the lamp throughput change with time, all diffuse reflectance and fluorescence spectra on each set of experiments were calibrated using a diffuse reflectance standard (20%, Spectralon, Labsphere) and a fluorescence standard (USF 210-010, LabSphere), respectively. To further correct fluorescence spectra for wavelength response, the fluorescence spectra were further calibrated by a NIST-approved tungsten calibration lamp (Optronic Laboratories Inc.). All the optical measurements were performed on animals under anesthesia using a mixture of isoflurane and room air (1.5% v/v), and a heating pad was used to help mice maintain body temperature. Optical measurements on normal and tumor-bearing mice were obtained by placing the fiber probe gently on the flank ( Fig. 1(C)) with the help of a custom designed probe holder ( Fig. 1(B)). Diffuse reflectance spectra were acquired from 400 nm to 650 nm (integration time: 3.8 ms). Fluorescence emission spectra were acquired from 520 nm to 600 nm (integration time: 2 s) using excitation at 488 nm and from 565 nm to 650 nm (integration time: 5 s) using excitation at 555 nm. The 488 nm light was typically used for 2-NBDG excitation, while the 555 nm light was used for TMRE excitation. As reported previously, the use of these two light sources will result in negligible optical cross-talk between the fluorescence channels when biologically relevant concentrated TMRE (75 µM of 100 µL) and 2-NBDG (6 mM of 100 µL) were used [50]. All measurements were acquired in a dark room to minimize background noise. Prior to any injection, baseline diffuse reflectance and background fluorescence spectra (excited by 488 nm and 555 nm respectively) were measured from the tissue region of interest. After the injection, optical measurements on each mouse were acquired continuously for a period of 80 minutes (TMRE only group or TMRE→2-NBDG with 20-minute delay group) or 60 minutes (2-NBDG only group). Each animal was euthanized after the completion of all optical measurements based on the IACUC protocol. After optical measurements on each animal were complete, reference spectra on a diffuse reflectance standard (20%, Spectralon, Labsphere) and a fluorescence standard (USF 210-010, LabSphere) were measured for future calibration.

Calculation of vascular and metabolic parameters
In order to compare fluorescence intensities between tumor and normal tissue, it is essential to correct the tissue absorption and scattering induced distortions of the measured fluorescence signal. Both diffuse reflectance and fluorescence spectra across the wavelength range 500 nm to 650 nm are needed to remove absorption and scattering distortions on fluorescence using an inverse MC model. Our previously developed scalable inverse MC model [44] was used to extract tissue scattering, absorption, native fluorescence of 2-NBDG, and native fluorescence of TMRE from in vivo optically measured spectra. The reflectance and fluorescence-based inversion MC model has been decried in detail previously [44,55,56]. Generally, the MC model assumes oxygenated hemoglobin, deoxygenated hemoglobin, and overlying mice skin as absorbers while cells and cellular components as scatterers. The tissue absorption coefficient is calculated by utilizing the widely used extinction coefficients reported by Scott Prahl [57], and the tissue scattering is calculated using Mie theory for spherical particles [58]. The MC inverse model adaptively fits the modeled diffuse reflectance to the measured tissue reflectance until the sum of squares error between the modeled and measured diffuse reflectance is minimized. Since the MC model works on an absolute scale while the tissue measurements are relative to a reflectance standard, a reference phantom with known optical properties is created to scale the tissue optical properties accurately [44]. The MC fluorescence model assumes that the measured fluorescence is a function of fluorophore concentration, absorbed energy probability and fluorescence escape probability [55]. The absorbed energy probability and fluorescence escape probability rely on optical properties at excitation wavelength and emission wavelength respectively [55], thus they can be easily simulated once the absorption and scattering information is extracted to quantify intrinsic fluorescence. The extracted absorption spectra between 520 nm and 600 nm were used to estimate SO 2 and total hemoglobin concentration. The extracted intrinsic 2-NBDG and TMRE fluorescence spectra were used to estimate the glucose uptake and MMP. Specifically, the mean of the peak-band (emission peak wavelength ± 10 nm) fluorescence intensity of intrinsic 2-NBDG and TMRE spectra were used to represent the 2-NBDG and TMRE signal. The 2-NBDG and TMRE signals taken at different time points were used to create the kinetic uptake curves. Comparison of mean kinetic curves across animal groups was performed using a two-way analysis of variance (ANOVA) test followed by Tukey-Kramer post-hoc tests. The fluorescence intensities, [Hb], SO 2 , or average scattering among different groups were compared with a two-sample t-test. A p-value < 0.05 was considered to be statistically different among the two groups under comparison.

Sequential injection of TMRE and 2-NBDG enables near-simultaneous measurement of glucose uptake and MMP in a flank model
Our former study [50] has demonstrated that there is neither significant optical cross-talk in tissue mimicking phantoms nor chemical crosstalk by mass spectrometry analysis between 2-NBDG and TMRE, suggesting that they are suitable for combined fluorescence imaging. However our in vivo non-tumor window chamber study showed that there was strong biological cross-talk when we injected the two probes at the same time. Specifically, TMRE signal was significantly attenuated by 2-NBDG when the two probes were injected simultaneously. We established that injecting TMRE first and then 2-NBDG after a 10-20 minute delay results in negligible crosstalk during simultaneous imaging of TMRE and 2-NBDG in a non-tumor window chamber model. Here we wanted to further establish that our previously validated sequential injection strategy [50] is applicable for quantitative optical spectroscopy of turbidity corrected TMRE and 2-NBDG fluorescence in a flank tumor model. Towards this goal, we compared turbidity-corrected fluorescence spectra measured from mice receiving injections of either 2-NBDG alone, TMRE alone, or TMRE followed by 2-NBDG with a 20-min NBDG alone) fluorescent in Fig cal enabler of t ting substrate a so quantified t cient (µ s ‫,)׳‬ and after the admin s was measure [59]. Figure 4 tissue and a 4T resent MC mod on coefficient 4(C) shows th in normal tissu SO 2 was signi ( Fig. 4(D)), w ith normal tiss significantly l ng levels are sign mors compared with les) and MC mod d We have previously demonstrated near-simultaneous imaging of glucose uptake, mitochondrial membrane potential of tumors in dorsal window chamber models using intravital microscopy [50]. Our former study confirmed that there is neither significant optical cross-talk in phantoms nor chemical cross-talk in mass spectrometry samples between 2-NBDG and TMRE [50], suggesting that they are suitable for combined fluorescence imaging when injected in a staggered manner (TMRE followed by 2-NBDG after a 10-20 minute delay) into the animal model. In this study, we further confirmed the effectiveness of the sequential injection strategy for near-simultaneous optical spectroscopy of 2-NBDG and TMRE uptake in a flank tumor model. The consistency in the data obtained with the spectroscopy system and our formerly reported microscopy systems demonstrates the robustness of our technique for near-simultaneous measurement of glucose uptake and mitochondrial membrane potential. Dorsal window chambers are by design optically thin and therefore provide an excellent model system to image small tumors with an intact tissue microenvironment via microscopy, however the dorsal window chamber cannot be imaged longitudinally beyond a few weeks, which limits its application to short-term in vivo studies [60]. In contrast, solid tumor models are particularly well-suited for long-term in vivo studies using optical spectroscopy such as monitoring tumor growth and assessing response to treatment [61], thus it is most often used for therapeutic studies [52].
Using our method, we were able to recapitulate known metabolic phenotypes in normal tissue and tumors. We demonstrated that 4T1 tumors have high mitochondrial activity in addition to glycolysis. The increased mitochondrial metabolism and glycolysis in 4T1 tumors compared to the normal flank recapitulated the results from our previous intravital microscopy studies [50,51]. The increased 2-NBDG uptake in 4T1 tumors also matched well with published reports including a Seahorse Assay study [2], magnetic resonance spectroscopy studies [2,62], metabolomics study [63], and FDG-PET study [62], all suggesting that 4T1 tumors have increased glucose uptake compared to normal tissue. It is interesting to note that the 4T1 tumors also have high mitochondrial activity in addition to glycolysis [5], which is consistent with the results published previously [16]. Specifically, Simões et al [16] utilized both magnetic resonance spectroscopy and the Seahorse Assay to characterize the tumor metabolism in which they captured both increased glycolysis and increased oxidative phosphorylation activity in 4T1 tumors compared to its sibling nonmetastatic tumor lines. Several studies showed that 4T1 tumors, which are highly metastatic and aggressive, are likely more "adaptable" to micro-environmental changes [16,22,64]. These "adaptable" tumors have the capacity to rely on both glycolytic and mitochondrial metabolism under a range of oxygen conditions allowing them to survive therapeutic stress, promoting negative outcomes such as increased recurrence [65], migration [66] and metastatic propensity [67].
The absorption and scattering properties for both normal flanks and small solid 4T1 flank tumors measured by our quantitative optical spectroscopy system are consistent with published reports [68,69]. We observed that 4T1 tumors have higher absorption and lower scattering coefficients compared to normal tissues. The increased absorption in 4T1 tumors is likely caused by angiogenesis [44, 70,71] as evidenced by the increased hemoglobin concentration in tumors. In fact, angiogenesis induced by 4T1 tumor cells begins at a very early stage [72], i.e., when the tumor mass contains roughly 100-300 tumor cells. Moreover, 4T1 tumors were found to have robust HIF-1a and VEGF levels [73] that could promote angiogenesis and hypoxia. Our spectroscopy study showed that solid flank 4T1 tumors had significantly lower baseline SO 2 values compared to normal tissues, suggesting that these solid flank tumors likely have regions of hypoxia [74,75]. It is worth mentioning that our former microscopy study found that the baseline SO 2 values were comparable between normal tissue and 4T1 tumors in window chamber models [51]. This difference in the baseline SO 2 values of 4T1 tumors in window chambers and flanks suggests that the baseline SO 2 levels in tumors are related to tumor size; small 4T1 tumors in window chambers (~2 mm in diameter) have higher oxygenation values while the larger 4T1 tumors in flank tumors (~6 mm in diameter) are more likely to be hypoxic. Hypoxia plays an important role in tumor cells that evade traditional therapies [76][77][78]. Our non-invasive optical quantification of vascular endpoints in animals offers an opportunity to study the tumor microenvironment and its effect on tumor metabolism.
We also demonstrated that the metabolic endpoints are positively correlated with their corresponding vascular endpoints for 4T1 tumors but not for normal tissue. Moreover, optical spectroscopy of key metabolic endpoints reveals a strong positive correlation between 2-NBDG and TMRE uptake in 4T1 tumors but not in normal tissue. We did not observe correlations between metabolic endpoints and baseline SO 2 values for normal animals which is likely attributed to the high baseline SO 2 values seen in normal tissue. In contrast, there was considerable variance in vascular endpoints for the solid 4T1 tumors which is likely due to the fact that tumor hypoxia is unstable both spatially and temporally [79]. The positive correlations between metabolic and vascular endpoints suggest that 4T1 tumors highly rely on substrate availability [16].

Conclusion
Our preclinical study demonstrates that optical spectroscopy is an effective tool for simultaneously quantifying vascularity and metabolism of cancer, which is critical to understand mechanisms underlying cancer progression, metastasis and resistance to therapies. Quantitation of tumor mitochondrial membrane potential and glucose uptake could provide insight into how metabolism modulates therapeutic outcomes and tumor cell survival following therapy. The associated tumor vasculature within the tumor microenvironment also influences tumor metabolism and being able to quantify both energy supply and demand will provide a holistic view of tumor bioenergetics. Quantitative optical spectroscopy enables longitudinal in vivo studies, at a length scale that complements existing methods, and thus has the potential to facilitate novel inter-disciplinary studies in cancer pharmacology.