Imaging Extracellular Acidification and Immune Activation in Cancer

Abstract

Metabolism reveals pathways by which cells, in healthy and disease tissues, use nutrients to fuel their function and (re)growth. However, gene-centric views have dominated cancer hallmarks, relegating metabolic reprogramming that all cells in the tumor niche undergo as an incidental phenomenon. Aerobic glycolysis in cancer is well known, but recent evidence suggests that diverse symbolic traits of cancer cells are derived from oncogene-directed metabolism required for their sustenance and evolution. Cells in the tumor milieu actively metabolize different nutrients, but proficiently secrete acidic by-products using diverse mechanisms to create a hostile ecosystem for host cells, and where local immune cells suffer collateral damage. Since metabolic interactions between cancer and immune cells hold promise for future cancer therapies, here we focus on translational magnetic resonance methods enabling in vivo and simultaneous detection of tumor habitat acidification and immune activation – innovations for monitoring personalized treatments.

1. INTRODUCTION

Measuring metabolism is central to understanding how anatomy and physiology change with aging, injury, and/or disease, because this route of scientific enquiry answers basic questions of how much metabolic resources are used for (re)building cellular infrastructure and their functional activities ¹. These questions are extremely relevant to tumor imaging in vivo because cancer is a disease of uncontrolled growth, which requires fuel and that is fundamentally a metabolic issue ².

Normal cells advancing to a neoplastic state acquire specific abilities, commonly referred to as cancer hallmarks ^3,4. These include evading apoptosis, uncontrolled growth, replicative immortality, limitless proliferation, sustained angiogenesis, and invasion and metastasis (Figure 1A). In other words, cancer cells hijack normal cellular mechanisms that regulate the usual timetables of survival, growth, and proliferation. These mutations then lead to uncontrolled cell division to promote tumor formation and progression. Furthermore tumor-promoting potentials are enhanced by genetic diversity arising from genome instability of cancer cells and even the pro-inflammatory potential of necrotic cells in the tumor milieu (Figure 1A). While these provide insights into genetic complexities of tumorigenesis, a paradigm shift in cancer research is realizing that oncogenes are key drivers of tumor growth due to reprogrammed metabolism ^5–8, thereby enabling cancer cells to thrive in their unusual habitat, by suppressing their own immune system (Figure 1A). The metabolic interplay between cancer and immune cells within the tumor microenvironment is a novel target that could potentially provide personalized therapeutic opportunities for cancer patients ⁹.

Figure 1. Metabolic features of the tumor microenvironment.

(A) Tumorigenesis is often depicted by gene-centric cancer hallmarks, which describe processes of how normal cells advance to a neoplastic state by acquiring mutations that commandeer normal mechanisms to promote their own survival. These include evading apoptosis, uncontrolled growth, replicative immortality, limitless proliferation, sustained angiogenesis, and invasion and metastasis, but also tumor-promoting potentials from cancer cells’ genome instability and necrotic cells’ pro-inflammatory potential. Key drivers of tumor growth reprogram metabolism in the tumor milieu, suppressing their own immune system. (B) Diversity of cells in the tumor niche. Cancer cells selfishly promote their own survival at the cost of their immune system counterparts. Abbreviations: natural killer (NK) cells, tumor-associated macrophages (TAMs), myeloid-derives suppressor cells (MDSCs), and dendritic cells (DCs). (C) Metabolic pathways of energy production and biosynthesis. Glucose is transported across membrane by glucose transporter (GLUT1). Uncoupling between glucose metabolism (V_glc) and oxidative metabolism (V_O2) generate lactate and H⁺. Both oxidative and non-oxidative processes produce acidic by-products, which are transported out of the cell by monocarboxylic acid transporters (e.g., MCT) and ion exchangers (e.g., Na⁺-H⁺ exchanger and vacuolar H⁺-ATPase). The MCTs also allow transport of other substrates (i.e., acetate, BHB) into cells. The CO₂ generated by oxidative processes can also become acidic by-products using anionic exchangers (e.g., carbonic anhydrase 9 or CA IX), which converts generated CO₂ and H₂O into HCO₃⁻ (bicarbonate) and H⁺. To maintain intracellular pH homeostasis both lactate and H⁺ are extruded into the extracellular space to cause acidification of the tumor milieu. (D) Metabolic phenotype of cells in the tumor milieu, which span from glycolytic (Glyc) to oxidative phosphorylation (OxPhos) to biosynthesis (Synth). Both aerobic glycolysis and biosynthetic pathways are used by the diversity of cells to meet their functional and structural needs.

It is vital to probe the tumor microenvironment in vivo because monolayer cultures of cancer cells in vitro do not reflect the three-dimensional (3D) cellular growth of a tumor ¹⁰, and inaccuracies in estimates of tumor cell percentages on in situ tissue staining can be misinterpreted ¹¹. The anatomical and physiological complexities of solid tumors can now be studied in vivo with an impressive arsenal of biomedical imaging tools ¹², spanning from translational methods like magnetic resonance imaging (MRI) and spectroscopy (MRS) as well as positron emission tomography (PET) or single photon emission computed tomography (SPECT) used in both preclinical and clinical research, to optical imaging methods which are used primarily in basic science research. A distinctive property of MRI and MRS methods is ability to safely and reproducibly map the tumor niche in vivo at suitably high spatial resolution. Here we discuss recent progress made for in vivo detection of metabolic dysregulation of tumor microenvironment and tracking immune cell activation using MRI and MRS methods.

2. FROM TUMOR MILIEU TO CELL METABOLISM TO MAGNETIC RESONANCE

Below we discuss briefly some background concepts on metabolic nuances of tumor microenvironment ^2,13–16 and magnetic resonance ^12,17,18.

2.1 -. A PRIMER ON TUMOR MILIEU

Genetic diversity of cancer cells contributes to their superior evolution ^15,16, and as a consequence the tumor habitat is complex ¹³ where cancer cells stimulate severe changes to their own ecosystem to support their progression at the peril of host cells ¹⁴ (Figure 1B). Immune cells, key constituents of the tumor microenvironment, are supposed to protect the host cells but instead turn on them. The tumor niche also consists of a thin monolayer of endothelial cells that deliver nutrients and dispose waste. Because these blood vessels are immature, they are often quite leaky thereby possessing abnormal perfusion rates. Additionally, stromal cells are recruited from endogenous connective tissue to (re)shape the extracellular matrix for continuous growth of cancer cells. Stellate cells are quiescent stromal cells of mesenchymal origin specific to liver and pancreas. The tumor habitat also contains adipocytes, which are responsible for storing excess energy as fat. Next to cancer cells, the dominant cellular populations in the tumor niche comprise of cells that help form the connective tissue (i.e., fibroblasts and pericytes) to modify the extracellular matrix and diverse cells that are involved in pro- and anti-tumorigenic responses of the immune system. These innate and adaptive immune cells provide first and second lines of defense against pathogens.

The innate (or non-specific) immune response provides resistance against microorganisms to suppress initial spread of foreign pathogens. Examples of innate immune cells include the natural killer (NK) cells and tumor-associated macrophages (TAMs). NK cells are cytotoxic lymphocytes patrolling the circulation, and these in addition to neutrophils which make up majority of circulating leukocytes, provide the first line of defense in the bloodstream. Both NK cells and leukocytes can secrete inflammatory cytokines. TAMs play a major role in facilitating tumor growth and metastasis as they create both immunosuppressive and inflammatory states by producing signaling molecules (e.g., cytokines, chemokines, growth factors) and triggering the release of inhibitory immune checkpoint proteins in T-cells (see below). Also part of the innate immune response system is the recently classified immature myeloid-derived suppressor cells (MDSCs), which also suppress immune responses and expand inflammation.

Adaptive (or acquired) immunity is particular to a specific pathogen, but in a tumor microenvironment these cells (e.g., T-cells and B-cells) also make errors and attack themselves. There are three types of T-cells: cytotoxic T-cells (CD8+) produce molecules that kill the infected cell, helper T-cells (CD4+) produce cytokines that signal to other immune cells, and regulatory T-cells (CD4+) shut off the immune response when no longer needed. B-cells are usually found at the tumor margin, and are responsible for antibody production, antigen presentation, and secretion of cytokines. The B-cells can inhibit tumor growth by using the tumor-reactive antibodies to promote cancer cell death by NK cells and priming of T-cells (CD4+ and CD8+). Dendritic cells (DCs) process the antigen material from B-cells and present it on the surface of T-cells, and thus act as immune cell messengers.

2.2 -. A PRIMER ON CELL METABOLISM

While the cell’s structural and functional demands are largely met by glucose metabolism, breakdown of other substrates also contribute ^2,13–16 (Figure 1C). Transporters located at plasma and/or cell membranes mediate substrate delivery, e.g., glucose transporters (GLUT1) for glucose and monocarboxylate transporters (MCT) for acetate, lactate, and ketone bodies like β-hydroxybutyrate (BHB). The ATP yield is ~18 times higher with full glucose oxidation in the tricarboxylic acid (TCA) cycle compared to glycolysis. However acetate and BHB can bypass the glycolytic steps to directly enter acetyl-CoA. The cell’s synthetic needs are largely met by glycolytic intermediates, but TCA cycle metabolites also contribute.

Normal breakdown of these substrates maintain a homeostatic ATP level (2–4 mM), while endogenous energy reserves are low: glucose (1–3 mM), oxygen (50–100 μM), glycogen (2–4 mM), and creatine (8–10 mM) ¹⁹. Thus normal function needs an efficient blood circulation to provide the nutrients, but also to remove the waste ²⁰. Glucose can be stored as glycogen in astrocytes for the brain, and it may provide a mechanism for constant ATP supply ²¹. Total creatine, which represents both phosphocreatine (PCr) and creatine (Cr), can undergo a phosphorylation-dephosphorylation reaction catalyzed by creatine kinase. These alternate energy reserves (glycogen, PCr) together can provide energy support for short epochs, e.g., a few minutes under ischemia ¹. These non-oxidative cytosolic pathways can provide faster ATP (in ms range) compared to mitochondrial respiration ²¹.

Glycolytic breakdown of glucose to pyruvate generates 2 ATP and in the presence of oxygen pyruvate is fully oxidized to carbon dioxide and water in the TCA cycle (i.e., glucose + 6O₂ → 6CO₂ + 6H₂O), while generating 34 additional ATP per glucose (Figure 1C). However other substrates can bypass the glycolytic steps to enter the TCA cycle via acetyl-CoA pool (Figure 1C). While most ATP support functional needs, phosphorylation is also needed for some biosynthetic processes ²². Glycolytic intermediates are involved in making of nucleotides, amino acids, and lipids, but some TCA cycle products are also needed for lipid biosynthesis ²². For example, mutations in isocitrate dehydrogenase (IDH1 and IDH2), which frequently occur in glioma and other cancers, produce D-2-hydroxyglutarate (D-2HG) from α-ketoglutarate, an example of reprogrammed metabolism preventing ATP production in the TCA cycle. When oxygen becomes limiting (i.e., hypoxia), pyruvate is reduced to lactate. Both oxidative and non-oxidative substrate breakdown produce lactate and hydrogen ions (H⁺), although no net H⁺ production occurs unless ATP is hydrolyzed. In cancer, uncoupling between cytosolic glucose metabolism (V_glc) and mitochondrial oxidative metabolism (V_O2) generates extra lactate and H⁺(Figure 1C) ².

Several exchangers and transporters of acidic by-products help maintain near neutral cytosolic pH ² (Figure 1C). Upregulation of these mechanisms in cancer ^23,24 leads to dramatic extracellular acidification, which is directly related to augmented tumor invasion, increased chromosomal rearrangements, enhanced proliferation rate, increased angiogenesis, and decreased immune function ^2,13–16. For example, the Na⁺-H⁺ exchanger extrudes H⁺ out of the cell for Na⁺ entry into the cell, whereas the vacuolar H⁺-ATPase extrudes H⁺ out of the cell for ATP hydrolysis, both in 1:1 stoichiometry. Since uncontrolled tumor growth results in hypoxia within select areas of the tumor niche, even CO₂ generated from oxidative activities can contribute to extracellular acidification with anionic exchangers like carbonic anhydrase 9 (or CA IX), which converts CO₂ and H₂O (both oxidatively generated) into bicarbonate (HCO₃^–) and H⁺ within the extracellular space. Importantly, upregulated GLUT1 and MCTs help transport glucose and lactate, respectively, into and out of cells.

Recent studies of the tumor habitat show that while cancer and immune cells have some similarities, there are metabolic distinctions, and understanding these differences may provide important vulnerabilities that cancer immunotherapies can target ^13,25 (Figure 1D). The nutrient-depleted hostile environment of the tumor habitat (e.g., acidic and hypoxic) creates a scenario of metabolic competition between cancer and immune cells. In addition to oxidizing the conventional substrates, the immune system can also breakdown amino acids and fatty acids through the TCA cycle producing energy for quiescent, differentiated cells. Similar to cancer cell metabolism ²², immune cells can also upregulate aerobic glycolysis pathways ^13,25. Thus the metabolic phenotype of cells in the tumor niche is quite diverse, spanning from glycolytic (Glyc) to oxidative phosphorylation (OxPhos) to biosynthesis (Synth) phenotypes (Figure 1D). While no cells show only Glyc phenotype, T-cells and B-cells only possess OxPhos phenotype whereas fibroblasts, adipocytes, and necrotic cells only feature Synth phenotype. Immunosuppressive TAMs, MDSCs, cancer, stem, and NK cells feature both Glyc and OxPhos phenotypes, whereas inflammatory TAMs and cytotoxic T-cells possess both Glyc and Synth phenotypes. Thus heterogeneity of metabolic phenotypes within the cell population of the tumor niche is very important for designing immunotherapy targets ⁹.

2.3 -. A PRIMER ON MAGNETIC RESONANCE

Nuclei containing odd numbers of protons and neutrons possess intrinsic magnetic moments that align in a parallel fashion in the presence of a strong static magnetic field (B_o) ^12,17,18. Biologically relevant and non-radioactive isotopes detectable by magnetic resonance are ¹H, ²H, ¹³C, ¹⁵N, ¹⁷O, ¹⁹F, ²³Na, and ³¹P. Higher B_o leads to greater signal-to-noise ratio (SNR) because of greater polarization, but ultimately SNR depends on the gyromagnetic ratio and the natural abundance of the isotope. High spatial resolution MRI, in 2D or 3D, arises from strong ¹H signal from high water/fat content in soft tissues. Molecular specificity in MRS comes from dilute ¹H signals from biomolecules other than water/fat, where the data can also be viewed in 2D or 3D with chemical shift imaging (CSI), but the images do not appear as crisp as for MRI because the voxels are much larger, needed for sufficiently high enough SNR of dilute ¹H signals from biomolecules.

Image contrast in clinical ¹H-MRI methods primarily rely on either the transverse (R₂* by gradient-echo or R₂ by spin-echo) or longitudinal (R₁) relaxation rates of tissue water to map both structure and function ¹². Intrinsic ¹H-MRI contrasts provide anatomical separation between healthy tissue and lesion, while diffusion-weighted ¹H-MRI, which maps the extent of Brownian motion of tissue water, provides information on degree of cellularity. Both R₂* (or R₂) and R₁ contrasts between healthy tissue and lesion can be further enhanced with exogenous agents injected into the bloodstream. Paramagnetic metal ions, specifically gadolinium III ion (Gd³⁺), conjugated with a chelating molecule consisting of electron donors, provide dynamic ¹H-MRI contrast as the agent extravasates into the lesion through leaky blood vessels to enhance the lesion’s appearance vs. healthy tissue ¹². Two types of ¹H-MRI experiments can be conducted with Gd³⁺-agents to measure vascular permeability and blood volume, respectively: dynamic contrast enhanced (DCE) measures the R₁ effect (positive contrast), whereas dynamic susceptibility contrast (DSC) reflects the R₂* effect (negative contrast). The R₂* effect (negative contrast) can also be viewed with SuperParamagnetic Iron Oxide (SPIO) nanoparticles or MicroParticles of Iron Oxide (MPIO), which are materials that are designed to be intravascular but they can also extravasate like Gd³⁺-agents. Additionally, either the chelating molecules or the particles can be attached to antibodies for targeting specific cell types (e.g., membrane antigen) ²⁶.

In general, MRS uses differences in resonance frequency between dissimilar chemical groups, either intramolecular or intermolecular, to measure regional concentrations of endogenous molecules, e.g., lactate, glucose, glutamate, and glutamine ¹⁸. But MRS can also detect exogenous molecules. High levels of lactate (observed in all tumors) and D-2HG (observed in IDH1-mutant gliomas) are detectable in vivo with ¹H-MRS ^27,28. ¹³C-MRS, in combination with infusion of ¹³C-labeled substrates like glucose, acetate, etc, allows rates of ¹³C-label incorporation into cell-specific pools. In the brain, for example, turnover of ¹³C-label from ¹³C-labeled substrates into glutamate and GABA (predominantly neuronal), or glutamine (predominantly glial) - can be detected in vivo ²⁹. ¹³C-MRS applied to brain tumors show that the tumor actively takes up glucose, acetate, and BHB ^30–32. Given advances in ¹⁷O-MRS ^33,34 and ²H-MRS ^35,36, metabolism of oxygen and glucose can potentially be detected with increased sensitivity and specificity.

3. NOVEL METHODS FOR MOLECULAR IMAGING OF CANCER

Rapidly growing cancer cells possess elevated rates of cytosolic V_glc compared to mitochondrial V_O2 ². For example, gliomas that progress to grade IV glioblastoma, are extremely glycolytic. Uncoupling between V_glc and V_O2 generates excess lactate and H⁺ (Figure 1C), which if not extruded out can hinder intracellular functions. Thus in vivo mapping of intracellular pH (pH_i) and/or extracellular pH (pH_e) is important for cancer research. ³¹P-MRS has played an important role in assessing the pH_i-pH_e gradient ^37–39, where pH_i can be measured by the protonation level of endogenous phosphate compounds,

$${\text{pH}}_{\text{i}}={\text{pK}}_{\text{a}}+\log \left[\left(\Delta \delta -{\delta }_1\right)/\left({\delta }_2-\Delta \delta \right)\right]$$ eq. 1

where Δδ is the shift difference between P_i and PCr resonances, δ₁ = 3.23 ppm and δ₂ = 5.70 ppm are corresponding shift differences at low and high pH respectively, and pK_a = 6.77 is the logarithm of the equilibrium constant for the protonation-deprotonation reaction ^37–39. However, pH_e can also be measured by ³¹P-MRS from the shift of exogenous agents like 3-aminopropyl phosphonate (3-APP) where pK_a = 7.11 and δ₁ = 20.34 ppm and δ₂ = 23.84 ppm in eq. 1 (Figure 2A) ³⁷. Previous ³¹P-MRS studies with 3-APP show that tumors possess lower pH_e (6.2–6.9) compared to normal tissue, whereas pH_i is near neutral (7.1–7.6) for both tissues (Figure 2B) ³⁸.

Figure 2. ³¹P-MRS and Chemical Exchange Saturation Transfer (CEST) in cancer imaging.

(A) In vivo ³¹P-MRS spectrum (no ¹H-decoupling) of a tumor xenograft in a severe combined immunodeficient (SCID) mouse, revealing the exogenous pH marker 3-aminopropylphosphonate (3-APP) in relation to inorganic phosphate (Pi), phosphomonoesters (PME), phosphocreatine (PCr), and nucleoside triphosphates (α-ATP, β-ATP, γ-ATP), where the 3-APP chemical shift is ~20 parts per million (ppm) downfield of Pi. Due to lack of mature B-cells and T-cells, the SCID mouse is ideal for xenoengraftment of human cells and tissue. (B) ³¹P-MRS detection of 3-APP and Pi for extracellular pH (pH_e; blue) and intracellular pH (pH_i; red) in different tumor types, where black and gray arrows, respectively, reflect the pH_i- pH_e gradients in tumor and normal tissues. (C) Schematic diagram showing relevant solute exchangeable proton pools (amide, amine, hydroxyl, etc.), each with a specific rate constant (k_ex), which resonate at frequencies shifted by few ppm away from that of bulk water protons. Confounding processes that might affect accurate quantification of CEST effects include macromolecular exchange-relayed nuclear Overhauser effects (NOE) and macromolecular magnetization transfer (MT). (D) Amine and amide concentration-independent detection (AACID) is a diaCEST contrast that includes CEST effects from both amine and amide protons in a ratiometric manner. The pH calibration of AACID_CEST in vivo (dashed line) was achieved using standard ³¹P-MRS, in vivo and postmortem. (E) pH mapping with AACID_CEST (shown in radiologic orientation) in a mouse brain 5 hours after permanent middle cerebral artery occlusion to induce cerebral ischemia (i) demonstrate ischemic (left) and contralateral (right) regions. The ischemic region on the left hemisphere was confirmed histologically using 2,3,5-triphenyltetrazolium chloride (TTC) staining (ii) for cell damage. (F) AACID_CEST data from a mouse brain with an U87 glioblastoma multiforme treated with 100 mg/kg Lonidamine. The AACID_CEST images were acquired immediately before (left) and 50 min after (right) intraperitoneal injection of 100 mg/kg Lonidamine, and show AACID_CEST values consistently higher in tumor after Lonidamine injection relative to baseline. Panel A was modified from Raghunand ³⁹. Panel B was modified from Hashim et al ³⁸. Panels D and E were modified from McVicar et al. ⁴⁹ with permission. Panel F was modified from McVicar et al. ⁵⁰ with permission.

In addition to demand for pH imaging, there is an emerging need for in vivo imaging of immune cell activation in the tumor environment because of therapeutic opportunities ⁹. While PET and SPECT methods have contributed much to advanced strategies for in vivo tracking of immune cells, recent developments in ¹H-MRI with Gd³⁺- as well as MPIO- and/or SPIO-based agents show some promise as well ⁴⁰. Specifically, tracking of T-cells, NK cells, DCs, and TAMs can now be tailored for cell-based immunotherapy approaches ⁹.

3.1 -. INTRATUMORAL pH_i-WEIGHTED IMAGING

Diffusion- and relaxation-based ¹H-MRI is widely used in cancer diagnosis ^12,41. But a new promising ¹H-MRI contrast for cancer imaging is called Chemical Exchange Saturation Transfer (CEST). The CEST contrast is generated when a radio-frequency (RF) pulse (B₁) saturates a pool of exchangeable protons (e.g., amide/amine (-NH_x) or hydroxyl (-OH) protons) to attenuate the bulk water proton signal ⁴². The CEST contrasts from diamagnetic and paramagnetic molecules are called diaCEST and paraCEST, respectively. The B₁ saturation of a pool of exchangeable protons is transferred through chemical exchange and attenuates the bulk water proton signal by

$$\text{S}/{\text{S}}_{\circ }\approx 1/\left[1+{\text{k}}_1{\text{T}}_1\right]$$ eq. 2

where S and S_o are the bulk water proton signals with and without RF saturation, respectively, k₁ is the pseudo first-order exchange rate constant, and T₁ = 1/R₁ is the longitudinal relaxation time of bulk (tissue) water. k₁ depends on the rate constant of the agent (k_A), agent concentration ([agent]), and the number of exchange sites (n). Since k_A and T₁ vary with pH, the CEST effect can provide information about the acidity of agent’s environment, assuming that the k_A of different pools is distinct enough (Figure 2C). In CEST imaging a Z-spectrum, which is a plot of the bulk water signal attenuation versus off-resonance B₁ saturation frequency Δω (in units of parts per million, ppm) is commonly used. The Δω of exchangeable proton pools of diaCEST and paraCEST are, respectively, ~5 ppm and ~50 ppm from water resonance. Both endogenous and exogenous diaCEST agents have Z-spectrum peaks that are downfield of water, but there are also aliphatic peaks in the Z-spectrum that are upfield of water (Figure 2C).

While various diaCEST contrasts can be obtained in theory ⁴³, there are outstanding issues that complicate in vivo diaCEST imaging in practice. Chemical shifts of diamagnetic exchangeable protons are within 1–4 ppm of bulk water (Figure 2C), and thus the contrast is affected by off-resonance direct saturation of bulk water and/or macromolecular magnetization transfer (MT) effects from indirect saturation, raising concerns about absolute and accurate pH quantification using CEST. In addition, there might be unknown differences between normal and diseased tissues in R₁, temperature, water content, protein content, macromolecular exchange-relayed nuclear Overhauser (NOE) effects, B_o inhomogeneity, B₁ inhomogeneity. Moreover, the presence of multiple pools that contribute to the total CEST effect simultaneously represents another critical issue when accounting for the CEST effects arising from different agents, because they do not follow a general linear model (i.e., CEST _A+B ≠ CEST _A + CEST _B) ⁴⁴. Nevertheless, this method is very appealing for clinical applications because most diaCEST agents are endogenous ⁴⁵.

Amide Proton Transfer (APT) is a diaCEST contrast generated from proton exchange between bulk water and amide groups of endogenous proteins and peptides, enabled by saturation of amide signals at 3.5 ppm downfield of water. Because these exchangeable protons (assumed to be arising from endogenous mobile proteins and peptides in the cytoplasm) are more abundant in tumors compared to healthy tissues ⁴⁵, generally a 3–4% APT_CEST contrast increase is observed in the intratumoral region compared to the peritumoral region.

While the APT_CEST contrast increase suggests a pH_i increase, a quantitative and accurate verification has been lacking because the CEST contrast remains qualitative unless the exchangeable pool’s concentration is known. Recently, a study by Ray et al. ⁴⁶ suggested that majority of the APT_CEST contrast in brain metastasis is from protein concentration. While APT_CEST contrast has a positive relation to pH_i and amide content, a study by Krikken et al. ⁴⁷ in breast cancer showed a negative relationship between APT_CEST contrast and pH_i by ³¹P-MRS.

Regardless, a rat study by Sagiyama et al. ⁴⁸ used the APT_CEST contrast to monitor the pH_i response of glioblastoma to Temozolomide, which is an oral alkylating chemotherapeutic agent that disturbs DNA replication and induces apoptosis of cancer cells. The results indicate that 1-week after a single course of Temozolomide treatment, the APT_CEST contrast decreased in the intratumoral region in glioblastoma tumors compared to untreated tumors. No changes in tumor volume, cell density, or apoptosis were observed in the treated tumors. Thus, it was concluded that the change in the APT_CEST contrast was due to a pH_i decrease induced by Temozolomide.

Amine and Amide Concentration-Independent Detection (AACID) represents another diaCEST contrast based on effects from both amine and amide proton pools in a ratiometric manner, thus avoiding the measurement of the exchangeable pool’s concentrations ⁴⁹. The AACID_CEST contrast is based on independent saturations at 2.75 ppm (amine) and 3.5 ppm (amide). While the pH_i derived from AACID_CEST contrast can be validated at a given B_o by ³¹P-MRS (Figure 2D), it generates an opposite contrast to APT_CEST. Regardless, the efficacy of pH_i mapping with AACID_CEST at 9.4T was elegantly demonstrated in rodent brain with ischemia by McVicar et al. ⁴⁹, which was confirmed with histological data (Figure 2E). Another later rodent study at 9.4T by McVicar et al. ⁵⁰ used the AACID_CEST contrast to monitor the pH_i response to Lonidamine in glioblastoma (Figure 2F). Since Lonidamine inhibits lactate transport, accumulation of more lactate in the intratumoral region should decrease pH_i compared to peritumoral region. Indeed, the APT_CEST contrast in the intratumoral region was lower than in the peritumoral region. In contrast to APT_CEST, the concentration-independent AACID_CEST contrast can provide a more quantitative estimation of pH_i.

Both APT_CEST and AACID_CEST contrasts encourage clinical applications ^45,51 because they can be applied in conjunction with other MRI techniques ^12,41. However NOE effects in tumor diagnosis cannot be overlooked with CEST imaging ⁵². While endogenous diaCEST agents (amide or amine pools) allow for intratumoral pH_i-weighted imaging, some exogenous diaCEST agents (e.g., iobitridol, iopamidol) are being explored for pH_e imaging as well ^53–55. While these exogenous diaCEST agents are clinically approved for other purposes (e.g., X-ray agents, doses of 0.3–1 g iodine/kg body weight), these CEST methods (doses of 1 – 4 g iodine/kg body weight) may have potential for clinical imaging if the agent dose can be lowered. In general, for all diaCEST imaging the main challenge is the degree of overlap of these exchangeable pools (Figure 2C), as well as contributions to the Z-spectrum from multiple components, such as water direct saturation, macromolecular MT effects, and aliphatic NOE effects.

3.2 -. INTRATUMORAL AND PERITUMORAL pH_e IMAGING

In CEST imaging, to increase the chemical shift separation between the signals arising from exchangeable protons of an inner sphere of bound water and the bulk water, a special class of cyclen-derivatives containing paramagnetic lanthanide III ions (Ln³⁺) called paraCEST agents ⁵⁶ was developed. The paraCEST agents improve the CEST contrast by allowing the tuning of water exchange kinetics (i.e., higher k₁ in eq. 2), chelate structure, and/or choice of Ln³⁺ ion. Clinically approved MRI contrast agents for DCE or DSC methods (see above) are based on DOTA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid) and contain Gd³⁺ ions that affect bulk (tissue) water R₁ relaxation (positive contrast). Most paraCEST agents are also DOTA derivatives, but use other Ln³⁺ ions like thulium (Tm³⁺), europium (Eu³⁺), etc. However in vivo paraCEST imaging has been challenging primarily due to B_o inhomogeneity issues and the need for much higher B₁ power ⁵⁷.

Recently a new ¹H-MRS method called Biosensor Imaging of Redundant Deviation in Shifts (BIRDS), which uses probes that are similar to paraCEST agents, was developed for pH and/or temperature mapping ^58,59. Although these agents contain exchangeable protons (e.g., -OH or -NH_x), the main physiological readout for BIRDS arises from the non-exchangeable protons (i.e., -CH_y) on the chelate. BIRDS, therefore, combines high spatial resolution of MRI with high molecular specificity of MRS, but into a 3D-CSI platform. Moreover, since the BIRDS agents possess both non-exchangeable and exchangeable protons, both BIRDS and CEST methods are possible with the same probe ⁶⁰. A special feature of BIRDS is that signals from different agents follow a general linear model (i.e., BIRDS_A+B = BIRDS_A + BIRDS_B) ⁶¹, unlike the CEST contrast involving multiple pools of exchangeable protons present together ⁴⁴. Although BIRDS agents use the DOTA framework, clinical use will depend on using non-Gd³⁺ ions. Additionally, these same BIRDS agents also can provide a decent R₂ effect (negative contrast) due to their extravasation ⁶².

The BIRDS method is based on detecting the signals from the agent itself. A ¹H-spectrum of the chelate (i.e., without the Ln³⁺ ion) shows conventional diamagnetic shifts spanning ~5 ppm. However a ¹H-spectrum of the complexed agent (i.e., with the Ln³⁺ ion) shows paramagnetically-shifted resonances more than 100 ppm apart (Figure 3A(i)). These massively shifted resonances have unusual properties because the chelate’s non-exchangeable protons are juxtaposed near the paramagnetic field of the metal ion (i.e., pseudo-contact), and thus the chemical shifts of BIRDS agents are affected by proximity of the nuclear spin to unpaired electrons of the paramagnetic metal ion ¹⁷. As a consequence the longitudinal (T₁ = 1/R₁) and transverse (T₂ = 1/R₂) relaxation times of non-exchangeable protons are very short (in ms range), allowing high-resolution and high-speed CSI that is minimally affected by B_o inhomogeneity. In addition, BIRDS agents could potentially be used across a wide B_o range because the T₂/T₁ ratio remains high and the molecular readout is largely unaffected by B_o. Since the effect on chemical shift depends on vector L between the spin and unpaired electron(s), factors affecting L (like temperature) will influence the shift. Similarly, protonation of the complex can alter a molecule’s geometry and change the relative shift. Variation of the total shift term, Δδ_O, when both temperature (T) and pH change simultaneously, can be modeled as

$$\Delta {\delta }_{\text{O}}={\text{C}}_{\text{T}}\Delta \text{T}+{\text{C}}_{\text{pH}}\Delta \text{pH}+{\text{C}}_{\text{X}}\Delta [\text{X}]$$ eq. 3

where C_T = (Δδ_O/ΔT)_pH is the temperature dependence at a given pH, C_pH = (Δδ_O/ΔpH)_T is the pH dependence at a given temperature, and the much weaker C_X term is for dependence on concentration of cation X. BIRDS characterization depends on C_T and C_pH in relation to Δδ_O. In addition, the molecular readout does not depend on diffusion or blood flow. Although the distribution of BIRDS agents may depend on vessel permeability (e.g. normal versus tumor tissue), the readout is independent of agent dose. Validation of pH_e measurements using a pH-sensitive Tm³⁺-probe, TmDOTP⁵⁻ based on the DOTA-framework (see Figure 3A(i) inset for TmDOTP⁵⁻) with BIRDS has been made with ³¹P-MRS using 3-APP ⁶³ (Figure 3A(ii)), which had also been used for pH calibrations of other ³¹P-MRS measurements in vivo (Figure 2B).

Figure 3. Biosensor Imaging of Redundant Deviation in Shifts (BIRDS) in cancer imaging.

(A) Validation of pH_e measurements with ¹H BIRDS using ³¹P MRS. The pH can be calculated with ¹H BIRDS using the chemical shifts of H2, H3 and H6 protons of TmDOTP⁵⁻ (i), or the chemical shift difference between the ³¹P MRS signals from 3-APP or TmDOTP⁵⁻ and inorganic phosphate (Pi) (ii). The ³¹P signal from 3-APP has a sensitivity of 2.9ppm/pH unit (iii), while the ³¹P signal from TmDOTP⁵⁻ has a sensitivity of 14.6ppm/pH unit (iv). The MRS-based pH measurements were also compared with those obtained using the pH meter (v). The slopes of the pH with MRS versus pH electrode measurements were very close to unity, 0.9999 for ¹H with TmDOTP^5–, 0.9998 for ³¹P with TmDOTP⁵⁻ and 0.9976 for ³¹P with 3-APP. The identity line (gray dashed line) was used to better visualize the similarity of the pH measurements. (B) Representative extracellular pH (pH_e) imaging with BIRDS in rat brains bearing 9L (top row) and RG2 (bottom row) tumors. T₂-weighted images (i, v) show tumor localization (tumor, blue; brain, orange). Chemical shift imaging data (ii, vi) and examples of spectra from intratumoral and peritumoral voxels (iii, vii) for the slices shown in panels (i, v) indicate TmDOTP⁵⁻ signals throughout the brain. Quantitative pH_e maps from BIRDS with TmDOTP⁵⁻ (iv, viii) indicate lower pH_e in tumors compared to normal brain tissue. (C) Representative pH_e imaging with BIRDS in rabbit livers bearing VX2 tumors. T₂-weighted images (i, v) show tumor localization (tumor, red; liver, blue). Chemical shift imaging data (ii, vi) and examples of spectra from tumor (iii) and normal liver (vii) for the slices shown in panels (i, v) indicate TmDOTP⁵⁻ signals throughout the liver. Quantitative pH_e maps from BIRDS with TmDOTP⁵⁻ (iv, viii) indicate lower pH_e inside the tumor and in adjacent areas compared to normal liver. Panel (A) was modified from Savic et al. ⁶³ with permission. Panel (B) was modified from Coman et al. ⁶⁴ with permission. Panel (C) was modified from Coman et al. ⁷¹ with permission.

pH_e in both intratumoral and peritumoral regions of rat brain containing various types of gliomas was measured with BIRDS at B_o of 11.7T using TmDOTP^{5− 64}. Upon agent infusion, the tumor boundary was identified with MRI by enhanced R₂ effect (negative contrast) because the Tm³⁺-probe is paramagnetic, while BIRDS provided imaging of intratumoral-peritumoral pH_e gradients for two different gliomas, 9L and RG2 ⁶⁴. The pH_e measured by BIRDS was also validated by ³¹P-MRS with inorganic phosphate (Pi) and PCr. While the intratumoral pH_e was acidic for both tumor types, peritumoral pH_e varied with tumor type. The intratumoral-peritumoral pH_e gradient was larger for 9L tumors than for RG2 tumors (Figure 3B), because acidic pH_e regions were observed in distal peritumoral regions in RG2 tumors. Increased presence of Ki-67 positive cells beyond the RG2 tumors border determined by T₂-weighted MRI suggested that RG2 tumors were more invasive than 9L tumors. In contrast, these in vivo and in vitro parameters in glioblastoma were reversed by Temozolomide treatment ⁶⁵.

Because the surgical intervention used to increase the plasma concentration of Tm³⁺ probe ⁶⁶ prohibits longitudinal BIRDS scanning on the same subject, Huang et al. used probenecid (i.e., an organic anion transporter inhibitor) which, co-infused with the BIRDS agent, temporarily restricts its renal clearance, thereby facilitating BIRDS without surgical intervention. In vivo BIRDS in rat brains bearing RG2, 9L, and U87 brain tumors using the probenecid and Tm³⁺ probe co-infusion method showed intratumoral-peritumoral pH_e gradients that were unaffected by the probe doses used ⁶⁷. This co-infusion method allows repeated scans with BIRDS (e.g., over days and even weeks) in the same subject, for pH_e mapping in preclinical models of tumor progression and therapeutic monitoring. Furthermore, compatibility of BIRDS with various nanoensembles such as SPIO nanoparticles, liposomes, and even dendrimers ^68,69 may provide exciting opportunities for assessing therapeutic response with BIRDS ⁶⁵ and allowing multi-modal imaging of drug delivery with MRI ⁷⁰. Recently, BIRDS has also been implemented on a clinical scanner operating at a B_o of 3.0T ^71,72. A translational demonstration of this work was the application of pH_e mapping with BIRDS in rabbit livers bearing VX2 tumors at B_o of 3.0T to show clear acidification of tumor areas in relation to healthy tissue (Figure 3C), similar to the findings in rodent brain tumors (Figure 3B).

3.3 -. IMAGING INTRATUMORAL pH_e AND IMMUNE ACTIVATION

Although Gd³⁺-based MRI agents are commonly used in clinical MRI of tumor diagnosis ^12,41, there have been major efforts to develop novel Gd³⁺ agents with low toxicity but high relaxivity and specific binding to either cancer or immune cells ⁷³. A crucial mechanism of reducing Gd³⁺ toxicity is achieved by enhancing the agents’ thermodynamic and kinetic stability without altering the pharmacokinetic properties. The multiple pendant arms available on the DOTA-framework (e.g., see Figure 3A inset for TmDOTP⁵⁻) enables simple chemistry to attach antibodies for targeting specific cell types ²⁶. Furthermore Gd³⁺-based MRI can be scaled up for super-high relaxivity with dendrimers, e.g., poly(amidoamine) (PAMAM), to allow super-high R₁ effect (positive contrast). Compared to commonly used MRI monomeric agents like GdDTPA²⁻ (<1 nm) which are quickly excreted by the kidneys, a GdDTPA²⁻ conjugated to higher generation and larger sized PAMAM-G9 dendrimeric agent (14 nm) show much slower renal excretion, thereby enhancing tissue contrast (Figure 4A) ⁷⁴. Similarly ligand-conjugated SPIO or MPIO have the potential to provide high MRI contrast with R₂* effect (negative contrast) and selectivity by appropriate chemistry applied to their surfaces ²⁶. Ligand-targeted SPIO or MPIO can be coupled covalently through peptide linkers that can be selectively cleaved by specific events without compromising on relaxivity, toxicity, and rapid clearance. Antibody-targeting of MPIOs or SPIOs can also be made to contain multiple fluorescent labels (e.g., AF488, green), and based on their size the excretion can be controlled. A recent study by Perez-Balderas et al. ⁷⁵ shows that MPIOs can be synthesized bearing either an anti-VCAM-1 antibody (αVCAM-AF488-MPIO) or a corresponding IgG control antibody (IgG-AF488-MPIO) for in vivo tracking of cerebral inflammation (Figure 4B). The mice were injected intracerebrally with interleukin-1β (IL-1β) in the left striatum to induce endothelial activation and VCAM-1 expression, where mice which were injected with αVCAM-AF488-MPIO a notable T₂ contrast was noted compared to mice injected with IgG-AF488-MPIO.

Figure 4. Molecular MRI tools to reveal resistance mechanisms caused by the immune-metabolic interplay.

(A) Whole body MRI of mice injected with 0.03 mmol-Gd/kg of PAMAM-G9 and 0.1 mmol-Gd/kg of Gd-DTPA (d) at 3 min post-injection. (B) Select T₂-weigthed MRI from mice injected intrastriatally with IL-1β several hours before intravenous injection of either αVCAM-AF488-MPIO (to) and control non-targeted IgG-AF488-MPIO (bottom). (C) Extracellular pH (pH_e) imaging with BIRDS shows that pH_e normalization in tumors restores immune permissiveness after conventional transarterial chemoembolization (cTACE). The BIRDS peaks shown in red were overlaid on the corresponding T₁-weighted MR images (top row), and were used to calculate the tumor pH_e (illustrated with color map overlays). Staining for hematoxylin-eosin (H&E, middle row) and human leukocyte antigen–DR isotope (HLA-DR, bottom row) receptors reveals peritumoral immune cell infiltration in acidic untreated tumors. Tumor pH_e changes after conventional TACE remained insignificant and immune cell infiltrates were similar or decreased compared to untreated tumors. However, conventional TACE with bicarbonate (cTACE w/bicarb) significantly increased pH_e of tumor and tumor edge, also boosting peritumoral immune cell infiltration. (D) Molecular imaging of antigen-presenting immune cells with gadolinium (¹⁶⁰Gd)–labeled anti–human leukocyte antigen–DR isotope (HLA-DR) antibodies. T₁-weighted MR imaging of VX2 liver tumors (*) before (i) and 24 hours after (ii) intra-arterial administration of ¹⁶⁰Gd-labeled anti–HLA-DR antibody shows peritumoral rim enhancement (white arrows), which indicates peritumoral immune cell infiltrate. Bright field image (iii) and colored images from ex vivo imaging mass cytometry (iv, v) of tissue harvested from the same rabbit confirm deposition of ¹⁶⁰Gd-labeled anti–HLA-DR antibody (green) in the peritumoral rim. The box in (iv) shows the area of magnification for (v). (E) Macrophage infiltration in the peritumoral rim demonstrated with molecular imaging of superparamagnetic iron oxide nanoparticles (SPIONs). T₂-weighted spin-echo MR imaging of VX2 liver tumors (*) before ((i), in axial orientation) and 24 hours after SPION administration in axial (ii) and coronal (iii) orientations show hypointense peritumoral rim (white arrows) indicating peritumoral iron retention. Photomicrographs of iron (Prussian blue stain) reveal deposition of SPIONs primarily in the peritumoral rim at ×1 (iv) and ×5 (v) magnification after phagocytosis by macrophages as seen at ×20 magnification (vi). The boxes in (iv) and (v) indicate areas of magnification for (v) and (vi), respectively. The yellow lines in (v) outline the peritumoral rim. L = liver, R = peritumoral rim, T = tumor. Panel A was modified from Kobayashi and Brechbiel ⁷⁴ with permission. Panel B was modified from Perez-Balderas et al. ⁷⁵ with permission. Panels C-E were modified from Savic et al. ⁷⁷ with permission.

However it is also possible to combine molecular MRI techniques with pH imaging. Since the pH_e readout with BIRDS remains unaffected by presence of paramagnetic or superparamagnetic fields created, respectively, by Gd³⁺- and SPIO-based MRI agents ^68,76, Savic et al. have combined BIRDS with specifically designed Gd³⁺- and SPIO-based MRI agents to study the immuno-metabolic interplay in a translational liver cancer model of VX2 tumors in rabbits ⁷⁷. Since this tumor type shows immune activation and pH_e normalization with treatments like conventional oil-based transarterial chemoembolization (TACE) or when adding bicarbonate to TACE to reduce the extracellular acidification (Figure 4C), they also sought to image the immune responses (Figures 4D and 4E). Thus, they labeled anti-human leukocyte antigen-DR isotope (HLA-DR) antibodies (¹⁶⁰Gd-labeled anti-HLA-DR) with gadolinium (¹⁶⁰Gd) to track antigen-presenting immune cells, which could be imaged in vivo with T₁-weighted MRI (positive contrast) and ex vivo with imaging mass cytometry (Figure 4D). Similarly, rhodamine-conjugated SPIO nanoparticles were used to detect macrophage infiltration upon phagocytosis, where these could be imaged in vivo with T₂*-weighted MRI (negative contrast) and ex vivo with staining (Figure 4E). After successful in vivo administrations of these agents with intra-arterial and intra-venous injections, the immune activation was readily observable at the tumor rim areas (Figures 4B and 4C).

4. CONCLUDING REMARKS AND FUTURE PERSPECTIVES

Almost everything the tumor niche is enabled to achieve given the cancer hallmarks, the development and maintenance of a reversed pH gradient across the cell membrane is directly related to secretion of acidic by-products into the extracellular space, which in turn enhances tumor aggressiveness but also resistance to chemotherapy. While dominance of genetic perspectives on cancer biology has created a view where tailored therapy for a given patient is based on the tumors’ gene expression pattern ⁷⁸, metabolic interventions may enable patient’s immune response to treat the cancer itself ⁹. If one is interested in how cancer and immune cells work, what are their main nutrients, how they get these nutrients, how they use the nutrients to support growth and function at various stages of treatments, then specific in vivo MRI and MRS techniques are needed to quantitatively measure metabolism and track immune response.

Advanced MRI and MRS methods can study metabolic changes of the tumor milieu, and how acidification and immune response are altered with treatment. Since acidic pH_e favors tumor growth and metastasis by activating matrix metalloproteases and cathepsins ³⁸ and even drive local tumor cell invasion beyond the tumor core ⁷⁹, CEST and BIRDS can help assess the efficacy of different tumor treatments using pH_i and pH_e mapping, whereas MRI with Gd³⁺- and SPIO-based agents can track various immune responses. Moreover, all of these methods can be performed together with more conventional MRI and MRS methods. Our focus on magnetic resonance is not intended to dismiss the importance of other biomedical imaging tools, as the methods utilized should be driven by the questions being asked, since different methods can provide complementary information ²⁹. While these magnetic resonance methods for imaging metabolism and immune responses demonstrate translational potential for better diagnosis and tracking of cancer, given the potential of combining optical imaging with these MRI/MRS techniques in vivo ⁸⁰, there is greater possibility to expand higher spatial resolution inspection of the interesting ongoing changes at the tumor rim.