Exploiting Optical Imaging for the Spatiotemporal Monitoring of Response to Cancer Therapies

Abstract

Recurrence of residual cancer cells that evade therapy is a leading cause of death in many cancer types. This clinical challenge is particularly pronounced in Triple Negative Breast Cancer (TNBC) due to the lack of well-defined molecular targets and the extensive heterogeneity found across patients, leading to up to 80% of patients having high risks of recurrence. TNBC is a subtype of breast cancer that accounts for about 20% of all breast cancer cases. TNBC tumors are devoid of detectable estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) gene amplification, rendering chemotherapy as the current main therapeutic option. Despite the improvement in clinical outcomes in TNBC tumors after standard-of-care therapy, recurrence rates remain high as patients develop resistance. The highly heterogeneous nature of TNBC tumors contributes to variability in treatment responses. Indeed, TNBC patients treated with cytotoxic therapies such as paclitaxel or Gemcitabine record a durable response rate of less than 20%. Given that early detection of locally recurrent tumors significantly improves survival rates, and that Pathological Complete Response (PCR) is strongly associated with relapse-free survival in TNBC, it is critical to assess early responses as tumors evolve following therapy to inform on long-term survival and optimize treatment regimens. In an attempt to identify biomarkers of cancer, research groups have widely studied the hallmarks of cancer such as genomic mutations, sustained proliferation, resistance to cell death, replicative immortality, and altered metabolism. Recently, altered metabolism has gained a lot of traction as a promising metric of disease progression given that metabolic changes have been observed to occur in parallel to therapy resistance and precede anatomical changes after treatment. Indeed, ample evidence in the literature shows the relevance of myriad metabolic pathways in TNBC. Aerobic glycolysis allows TNBC tumors to sustain a proliferative state in the absence of normal growth signals, though cells can also adopt a high mitochondrial respiration phenotype to resist treatment. In addition, TNBC cells can rewire their metabolism to utilize alternative fuel sources such as lactate, amino acids, and lipids to improve their survivability. Monitoring dysregulated tumor metabolism is desirable since it reflects both cellular and tissue-level dynamics. Indeed, in addition to the evolutionary propensity of cancer cells to metabolically adapt to variable conditions, tumor metabolism is intricately linked to phenotypical conditions of the tumor microenvironment such as the surrounding local vasculature or neighboring “normal” cells. Taken together, such adaptations lead to spatiotemporal heterogeneities in tumor subpopulations, phenomena that have been recognized as leading causes of tumor recurrence. For instance, the chaotic vasculature in the tumor microenvironment can result in an unbalanced blood supply and varying degrees of perfusion leading to the emergence of specialized niches of metabolically distinct drug-resistant tumor subpopulations. Visualizing spatiotemporal adaptations in metabolic heterogeneity within the tumor microenvironment during therapy will, therefore, be critical in tailoring adaptive therapies aimed at eradicating residual cells or prolonging tumor dormancy. Current prospective studies investigating metabolism as a biomarker of chemotherapy response in TNBC utilize 18F-FDG-PET to quantify differences in standardized FDG uptake between a baseline scan and after a few cycles of chemotherapy. Decreases in tracer uptake between time points are typical of predicted responders and are associated with pCR. 18F-FCH-PET, a choline analog, is also being studied to trace lipid metabolism for monitoring treatment response in a 4T1 murine model of TNBC. Further, 13C-hyperpolarized Magnetic Resonance Spectroscopy Imaging is also promising in detecting treatment response in TNBC xenografts with higher sensitivity than PET. However, each of these studies was limited to one metabolic endpoint, required completion of at least one treatment cycle before evaluation, and was limited to evaluating one treatment regimen at a time. While metabolite profiling of blood serum, transporter expression, and over-expression of enzymes involved in lipid metabolism have also emerged as potential biomarkers of recurrence, these are static sample analyses. Additionally, the above methods do not account for intra-tumor heterogeneity which could allow drug-resistant sub-populations to evade therapy and eventually resurge. Optical imaging offers a solution by quantifying multiple metabolic endpoints longitudinally at cellular level resolution. Promising results using optical imaging show accurate long-term predictions of patient recurrences in pancreatic cancer by tracking early metabolic changes following therapy. However, the majority of these studies relied on the autofluorescence of endogenous fluorophores NADH and FAD. Therefore, direct measures of metabolic pathways to inform on druggable targets could not be obtained. The Center for Global Women’s Health Technologies has extensively validated the use of exogenous fluorophores 2-[N-(7-nitrobenz-2-oxa-1, 3-diazol-4-yl) amino]-2-deoxy-D-glucose (2-NBDG), Tetramethylrhodamine ethyl ester (TMRE), and Difluoro-5,7-Dimethyl-4-Bora-3a,4a-Diaza-s-Indacene-3-Hexadecanoic Acid (Bodipy FL C16) to directly report on key metabolic pathways involved in promoting therapy resistance in in vitro and in vivo models. 2-NBDG is a glucose analog and enables the quantification of glucose uptake. TMRE is a cation that is attracted to the proton gradient of the mitochondria during ATP synthesis and thus measures mitochondrial membrane potential. Bodipy FL C16 is a fluorescently tagged long-chain saturated fatty acid molecule that is taken up by the cell, and reports on fatty acid uptake. Such metabolic imaging using multiple exogenous fluorophores is advantageous as it provides holistic insights into specific targetable metabolic vulnerabilities associated with poor clinical outcomes. The work presented here demonstrates a quantitative technique to visualize spatiotemporal metabolic heterogeneity within the tumor microenvironment. This platform enables longitudinal measurements across three major metabolic axes in the context of the tumor microenvironment in a variety of clinically relevant model systems such as xenografts or patient-derived organoids. This approach has been applied to study behaviors of resistant tumors during treatment, thus informing on emergent, druggable targets. Optical imaging techniques employed pinpointed the time at which significant metabolic changes emerged following treatment, a temporal phenomenon that cannot be captured with static gold-standard techniques such as metabolomics. Additionally, the work leverages multi-scale imaging technologies to quantify heterogeneity both at the bulk tumor level in the context of its microenvironment and at the cellular level. Taken together, this work is well poised to study the evolutionary dynamics of residual tumor subpopulations, thus revealing potential targetable vulnerabilities of residual tumor subpopulations across cancer types. Toward this goal of quantifying spatiotemporal changes for monitoring response to cancer therapies in preclinical models, two specific aims were proposed. Aim 1 characterizes differences in metabolic reprogramming across TNBC xenograft models with differential responses to standard-of-care, maximum dose density chemotherapy. 2-NBDG, TMRE, and Bodipy FL C16 were used to quantify glucose uptake, mitochondrial membrane potential, and fatty acid uptake as resistant and sensitive tumors responded to treatment. This technique enabled metabolic shifts to be assessed as resistant and sensitive tumors transitioned from their primary (untreated) state to early regression (active treatment) to late regression (after drug withdrawal) to residual disease (stable tumor volume) to recurrence (if applicable). Image-based clustering analyses were performed on multiplexed metabolic endpoints to visualize subtle changes in the metabolic preferences of distinct tumoral subpopulations. At key time points corresponding to each disease stage, tumors from paired cohorts of mice were harvested for comprehensive gold-standard assays such as metabolomics and immunohistochemistry (IHC). Optical metabolic imaging of chemotherapy-treated resistant and sensitive tumors demonstrated that recurring tumors exhibited a shift from glycolysis to non-glucose-driven mitochondrial respiration during tumor regression and residual disease compared to untreated counterparts. Conversely, sensitive tumors showed a similar metabolic reliance during the treatment naïve primary state and during residual disease, indicating a lack of metabolic reprogramming. Further, through sub-population analyses, increased heterogeneity was observed over time in tumors that ultimately recurred compared to sensitive tumors. Gold-standard metabolomic analyses were consistent and complementary to imaging data, corroborating the shift away from glycolysis in resistant tumors while pointing to amino acid metabolism as a relevant pathway during residual disease. Given the limitations of current in vitro, ex vivo, or in vivo methods of metabolic phenotyping, the imaging platform employed in this work enables an unprecedented in vivo view of a tumor’s metabolic landscape. In vivo, longitudinal optical imaging allowed for transient metabolic phenotypes to be visualized, a feature that will be critical to inform on adaptive treatment approaches. Further, the methodology utilized enabled the capture of multiple metabolic endpoints repeatedly in the same animal, thus enabling multiplexed analyses aimed at assessing metabolic dependencies of tumoral subpopulations. Visualizing the emergence of metabolically distinct populations associated with therapy resistance has the potential of informing on both the risk of tumor recurrence and the approach to tailor therapies to target specific vulnerabilities of residual subpopulations to reduce tumor recurrence. In Aim 2, a fluorescence microscope, called the Capcell, capable of multi-scale quantitative fluorescent imaging was developed to analyze changes both across a bulk tumor microenvironment and across localized hotspots. A Gabor filter/Dijkstra segmentation-based analysis pipeline was established to assess the unique association between multi-pathway intra-tumoral metabolic heterogeneity and local vasculature within the tumor microenvironment. Through the utilization of a vascular disrupting agent Combretastatin A-1, both metabolic and vascular parameters were quantified in vivo over the course of treatment. Combining the metabolic and vascular analyses workflows enabled robust intra-image analysis while providing insights into the intricate link between metabolic activity and vascular features. We demonstrated that the Capcell microscope allows for the visualization of cellular- and capillary-level features in preclinical window chamber models. Widefield imaging was used to guide the selection of high-resolution images to enable a closer look into metabolic heterogeneity at specific local hotspots. Further, through small molecule perturbation studies, the ability to quantify spatiotemporal metabolic and vascular changes over the course of treatment was validated. A decrease in mitochondrial metabolism and vessel density was seen following treatment with the vascular disrupting agent. Intra-tumoral heterogeneity analyses showed a decrease in tumoral subpopulations dependent on either mitochondrial metabolism only or both mitochondrial metabolism and metabolic substrate in the absence of local vasculature. These observations prime the Capcell for future studies on quantifying intra-image metabolic heterogeneity in the context of vasculature in clinically-relevant biological models. Visualizing the unique relationship between metabolism and vasculature longitudinally and in vivo suggests the potential of using the Capcell as a monitoring tool for understanding response to therapeutics and for determining optimal dosing schedules. While targeting a tumor’s metabolic vulnerabilities is a promising clinical strategy, the challenge is to identify biomarkers that indicate the likelihood of long-term response. Monitoring early responses to cancer therapies will enable clinicians to tailor the timing and type of targeted therapies administered such that residual cells can be specifically eradicated, thus reducing or delaying the onset of tumor recurrence. The work presented herein was aimed towards enabling the multiplexed quantification of features associated with therapy resistance longitudinally, in vivo, and at spatial resolutions capable of visualizing small, residual tumor subpopulations. This platform fills a critical gap in available technologies for metabolic/vascular phenotyping and can be leveraged to holistically monitor behaviors of residual tumors, thus enabling the development of highly effective metabolically targeted therapies for currently uncurable cancers. Future directions will focus on implementing this technology in patient-derived samples to identify druggable targets and assessing the impact of oncogene-targeted therapy on spatial metabolic and vascular heterogeneity. Further, this platform can be extended to distinguish between distinct microenvironmental cell types such as tumor cells, adipocytes, and immune cells, thus discerning the contribution of different cellular neighborhoods in promoting therapy resistance.