Molecular MR Contrast Agents

Abstract

Molecular magnetic resonance (MR) imaging utilizes molecular probes to provide added biochemical or cellular information to what can already be achieved with anatomical and functional MR imaging. This review provides an overview of molecular MR and focuses specifically on molecular MR contrast agents that provide contrast by shortening the T₁ time. We describe the requirements for a successful molecular MR contrast agent and the challenges for clinical translation. The review highlights work from the last five years and places an emphasis on new contrast agents that have been validated in multiple preclinical models. Applications of molecular MR include imaging of inflammation, fibrosis, fibrogenesis, thromboembolic disease, and cancers. Molecular MR is positioned to move beyond detection of disease to the quantitative staging of disease and measurement of treatment response.

I. Introduction

MRI is an integral part of diagnostic medicine. It provides high resolution, 3D images that are not constrained by tissue depth and does not require ionizing radiation. The utility of MRI is further enhanced by the ability to obtain different types of image contrast depending on the pulse sequence used to acquire the image, e.g. T₁-weighted, diffusion weighted, etc. Additional contrast can be achieved by administration of a paramagnetic or superparamagnetic contrast agent. Besides providing high resolution images of anatomical structures and pathology, contrast agents can also assess organ perfusion and tissue permeability. It is important to note that these different types of image acquisition are typically performed in a single session and the information provided by the various anatomical and functional scans are additive. The dominant class of contrast agents is extracellular fluid gadolinium-based contrast agents, ECF GBCAs,¹ which distribute freely in the extracellular space and are used for a wide range of indications including assessment of blood brain barrier permeability,^{2, 3} detection of breast lesions,^{4, 5} detection of abnormal vascular anatomy,^{6, 7} and myocardial perfusion and viability.⁸ There are also liver-specific contrast agents which are used to detect liver lesions and report on liver function.^{9, 10} A serum albumin binding contrast agent, gadofosveset, was developed for angiographic imaging and other applications that relied on albumin binding.^{11, 12} There have also been iron oxide nanoparticles approved for human use that were used to provide angiographic information or provide contrast due to their uptake in the reticuloendothelial system, i.e. liver and lymph nodes.^{13, 14}

Additional information can be provided by the use of a molecular MR contrast agent which targets a specific protein, cell-type, or biological process. Molecular contrast agents are often referred to as molecular probes, and we will use the terms interchangeably here. A molecular MR probe could improve disease detection sensitivity or specificity, could be used to quantitatively stage disease, could measure disease activity, or monitor disease progression and response to treatment. The molecular information obtained with such a probe would be in addition to the anatomical and functional information that one can already achieve in clinical MRI, Figure 1. Gadoxetic acid, taken up by hepatocytes through the adenosine triphosphate–dependent organic anion transporting polypeptide 1, and albumin-binding gadofosveset can be thought of as molecularly targeted agents. At the preclinical stage, there are several exciting new molecular probes to specifically image inflammation, fibrosis, fibrogenesis, thromboembolic disease, and cancers that have been validated in multiple preclinical models and these form the basis of this review. However, in order to be successful, a molecular MR contrast agent must fulfill a number of criteria. These are outlined below.

Figure 1.

MRI provides detailed anatomical information owing to different contrast mechanisms with and without exogenous contrast agents. Functional information can be added in the same imaging exam. Molecular MR imaging offers potential to add further quantitative biological characterization of pathology without sacrificing anatomical and functional information.

Detection sensitivity.

For molecular imaging, the probe should target a specific protein or cell type and/or have its relaxivity modulated by a biological process. However regardless of target specificity, the probe must be present in high enough concentration to be detected by MRI. Contrast agents are not detected directly but rather by their effect on tissue relaxation times. Tissue T₁ values depend on the organ and field strength but range from ~300 ms for fat up to several seconds for cerebrospinal fluid.¹⁵ The ability of a contrast agent to change T₁ (or T₂) is termed relaxivity r₁ where r₁ = Δ(1/T₁)/[M] where Δ(1/T₁) refers to the change in the reciprocal relaxation time measured in the presence and absence of the contrast agent and [M] refers to the concentration of the agent in mM. r₁ has units mM⁻¹s⁻¹ and r₁ depends on field strength, temperature, and the medium in which the measurement is made (e.g. plasma, whole blood, water). Taking a hypothetical tissue with a T₁ of 1 s and a contrast agent with r₁ = 4 mM⁻¹s⁻¹ (typical of GBCAs), one would require 25 μM of contrast agent to reduce T₁ by 10% and 500 μM to reduce the T₁ by 200%, at which point the signal intensity would appears as bright as fat on a T1-weighted image. The detection sensitivity would be higher if the baseline tissue T₁ is longer and/or the relaxivity is higher, but these are still values in the micromolar range.

For molecular MR imaging, therefore, the molecular target should be very abundant (μM), or there needs to be some sort of signal amplification mechanism such as high molecular relaxivity (multiple metal ions per probe) or catalytic enzymatic activation in order to detect sub-μM concentration targets.

Pharmacokinetics.

Since contrast agents are detected indirectly it is usually necessary to compare an image obtained prior to probe injection with one obtained after injection in order to see the signal change caused by the probe. The pharmacokinetics of the molecular probe should show fast localization at the target and fast clearance of non-specific background signal such that the pre- and post-injection images can be obtained in a single scan session. If it takes a long time (> 1 hour) for the probe to localize and for nonspecific enhancement to subside, then this will require two separate scan sessions which will increase costs, decrease patient compliance, and make it more challenging to register the pre-probe injection image to the post-probe injection image. Contrast agents with molecular weights under 10 kDa are typically small enough to rapidly extravasate into the extracellular, extravascular space where they may bind to their target, and also small enough to be rapidly filtered into the urine by the kidneys. Nanoparticles and macromolecules on the other hand typically take longer times to accumulate at their target and also often long times to clear from the circulation.

Safety.

Since contrast agents confer no therapeutic benefit, the safety margin must be very high. Molecular MR probes, like any contrast agent, should be well tolerated, have a low adverse event rate, and have a very low serious adverse event rate.¹⁶ For instance, one study found that only 0.36% of patients who underwent cardiac MRI with GBCAs showed symptoms of acute side effects, while only 0.033% were classified as severe adverse events.¹⁷ There are safety concerns regarding GBCAs that are directly related to the presence of gadolinium in the molecule. Nephrogenic systemic fibrosis (NSF) is a gadolinium-associated toxicity that can occur in patients with poor renal function who are exposed to GBCAs.^{18, 19} It has been also well established that all GBCA administrations result in trace accumulation of gadolinium, in some chemical form (elemental or chelated), in the body and brain.^20-22 The long-term consequences of this retained gadolinium are unknown but given the devastating effects of NSF and its delayed onset, Gd retention is worrisome. GBCAs used clinically are excreted efficiently in subjects with normal renal function. For molecular MR probes, targeting may also result in a longer residence time in the body.

There are different strategies to minimize the safety risk due to retained probe. First, the probe should be designed to be excreted as efficiently as possible.²³ Second, increasing the relaxivity will allow for the dose to be reduced and would also help with detection sensitivity.²⁴ Finally, it is also possible to use an endogenous metal ion like iron (Fe) or manganese (Mn) instead of Gd as the signal generating moiety.²⁵ It is believed that small residual amounts of Fe or Mn remaining would pose less of a risk since the metal ion could be eliminated through endogenous mechanisms that maintain metal homeostasis or could be incorporated into the body’s pool of Fe or Mn. Figure 2 highlights how two of these strategies might be employed. Figure 2A shows contrast enhanced MRI in a rat glioma model for a new high relaxivity ECF agent gadopiclinol, compared to the clinical agent gadobutrol. Gadopiclenol is a new macrocyclic contrast agent currently undergoing Phase III clinical trials.^26-28 In contradistinction to other GBCAs, gadopiclenol has two exchangeable water molecules bound to gadolinium ion and an increased molecular weight, resulting in more than a 2-fold increase in relaxivity (12.8 mM⁻¹s⁻¹ at 1.5T) compared to currently used GBCAs. Figure 2A shows the benefit of the high relaxivity which results in effective contrast enhancement at a lower dose than gadobutrol. Figure 2B,C shows a comparison of contrast enhanced MR angiography images in a baboon obtained with the Mn-based contrast agent Mn-PyC3A (2B) and the GBCA Gd-DTPA (2C) in which the animal was administered each agent at the same dose.²⁹ Mn-PyC3A has similar relaxivity to Gd-DTPA and demonstrates that the chelated Mn²⁺ ion can provide equally effective MR contrast enhancement as chelated Gd³⁺.

Figure 2.

A) Intra-animal comparison of gadopiclenol enhanced (left) and gadobutrol-enhanced MR images of a rat glioma model showing the high relaxivity of gadopiclenol, which provides similar tumor enhancement at half the dose of gadobutrol (top) and much stronger enhancement at the same dose (bottom). B) Mn-PyC3A enhanced MR angiography in a baboon showing abdominal aorta and renal arteries compared to C) an equivalent dose of Gd-DTPA in the same animal. Adapted with permission from ²⁸ and ²⁵.

Unmet need.

Is there a need for a novel molecular probe or can the diagnostic information safely be obtained by other means? For instance, a molecular MR probe could be used to detect pathology at an earlier stage than can be seen with conventional approaches. Or there could be an existing test that is imperfect, e.g. high false positive (or negative) rate; is semi-invasive and requires sedation; is time consuming; results in radiation exposure.

Market size.

For regulatory approval and clinical registration, a new contrast agent must undergo a multiphase, multiyear clinical development program similar to that for therapeutic drugs. Although there are some efficiencies relative to a novel therapeutic, the estimated development costs are in excess of $100 million.³⁰ Contrast agents are typically only administered once, and because they are diagnostics, there is a limit on what these can be sold for. In order to recoup the development costs and be profitable, a new contrast agent must have a high sales volume. This could involve a single indication or having an agent that is effective for multiple high-volume indications.

With all these factors in mind, we describe below a number of remarkable examples of molecular MR probes that address all of these requirements to some extent. All are focused on large disease areas where there is unmet need. Most of these agents are small(ish) molecules that have rapid uptake at the site of pathology and rapid elimination. Most are gadolinium-based but we are beginning to see Fe and Mn alternatives. Detection sensitivity reflects a range of strategies from high relaxivity, amplification, and judicious choice of a target. We focus on agents that have been evaluated in preclinical trials within the last 5 years, and especially on those that have been used in more than one animal model which represents some level of validation. As will be seen, many of these agents can detect pathological changes in different disease state models. Despite pathological differences between different diseases, there can often be a common biomarker, that is overexpressed in concentrations sufficient to be targeted by MRI agents. For instance, many extracellular matrix (ECM) proteins are upregulated in cancer and fibrosis, and therefore the majority of the probes discussed below produce a contrast enhancement upon binding to one of those proteins. At the same time, different ECM proteins can be overexpressed at different stages of the disease, allowing for non-invasive disease staging using a specific targeted probe. The same probe can be also used to verify whether the expression levels of these proteins are responsive to the treatment plan, which can be amended accordingly. Besides direct protein binding, another strategy involves exploiting an enzyme present in pathology to amplify the contrast agent signal. Myeloperoxidase (MPO) is highly expressed in a number of inflammatory states. Below we describe two classes of molecular MR agents that rely on oxidation by MPO to boost relaxivity and aid detection.

II. Enzyme activatable probes

Enzymatic reactions lie at the heart of many essential biological processes, where substrates and (or) products of these reactions often serve as promising molecular biomarkers. Certain enzymes may be upregulated upon onset of a pathological condition, making these potentially useful imaging targets. But most enzyme concentrations are in the nanomolar range and not detectable by direct targeting with a molecular MR probe. However molecular probes can be designed that are acted on by the enzyme resulting in a modulation of MR signal. For instance, a contrast agent can be converted into a product with a substantially different relaxivity value resulting in a detectable signal change. Or the enzyme can cause a chemical reaction that causes the molecular probe to be retained in tissue at the site of the enzymatic activity.

The enzyme myeloperoxidase (MPO) is a heme-containing peroxidase released by neutrophils and macrophages and represents a ubiquitous extracellular biomarker of inflammation. Because MPO is produced by immune cells, it is found throughout the body as part of the inflammatory response and therefore potentially can be used to image inflammation and its resolution across a range of diseases.³¹

Bogdanov, Chen, and Weissleder pioneered the use of hydroxytryptamine derivatized GBCAs that are oxidized to forms radicals in the presence of MPO. The radical can then react with other hydroxytryptamine groups which results in oligomerization of the contrast agent (with molecular weights exceeding 10 kDa),³³ or the radical can add to tyrosine residues on matrix proteins. Either process results in the Gd-chelate associated with a larger molecule resulting in increased relaxivity and retention at the site of inflammation (Figure 3A).^{34, 35} The first iteration of the probe, MPO-Gd, used the Gd-DTPA-bis(amide) chelate, but macrocyclic Gd-chelates were also intriduced.^35-38. The specificity of the probe towards MPO was demonstrated by imaging disease state models in the absence and presence of a MPO inhibitor (4-aminobenzoic acid hydrazide, ABAH),³⁹ and also by using MPO knockout mice. In a mouse stroke model no signs of MPO-Gd activation were observed in MPO knockout mice, whereas a strong signal enhancement was detected in wild type animals.⁴⁰ This specificity was also later confirmed in other preclinical models.^{41, 42} The utility of this and structurally-related probes was evaluated in multiple preclinical disease models ranging from autoimmune diseases⁴³ to lung fibrosis.³⁸ In particular, cardiovascular^{41, 44, 45} and neurovascular^{40, 46, 47} inflammatory processes were among the most studied. More recently, MPO-Gd was evaluated in preclinical models of multiple sclerosis^48-50 and spinal cord injury.⁵¹

Figure 3.

MPO oxidizes the hydroxytryptamine moiety on the molecular probe resulting in oligomerization and/or covalent addition to matrix proteins, both of which result in increased relaxivity and retention at the site of inflammation. (A). Comparison between axial T_1w MR images of mice with a diet-induced NASH phenotype (top row) and steatosis only phenotype (bottom row) before and 60 min post injection with MPO-Gd showing strong enhancement of the liver in the NASH animals that is further reflected in the CNR (contrast-to-noise ratio) maps overlaid on the right (B). Adapted with permission from ³².

In some cases, the expression of MPO strongly correlates with the severity of the pathological condition, and therefore MPO-Gd can be used for the disease staging. Experimental autoimmune encephalomyelitis (EAE) is the most widely used animal model of multiple sclerosis (MS), where a significant correlation between a clinical stage of the disease and MPO levels was observed.⁴⁷ The areas of demyelination were found to be co-localized with the regions of upregulated MPO activity, and therefore their volumes, which are strongly correlated with the severity of the disease, were quantified using Gd-MPO. A significantly better correlation between enhancing area and percent demyelination was observed for Gd-MPO (R² = 0.96, P = 0.002) than conventional imaging (R² = 0.65, P = 0.098), allowing for a reliable non-invasive staging of the disease.

The severity of EAE can be attenuated by ABAH, found in demyelinated plaques of patients with MS, which irreversibly inhibits the activity of MPO and therefore reduces demyelination and tissue damage caused by MPO-generated ROS. At the same time, the first-line immunomodulatory drug, glatiramer acetate (GA), is believed to recruit protective anti-inflammatory T helper type 2 lymphocytes, which inhibit inflammatory processes. Simultaneous administration of these two drugs, targeting different branches of the inflammatory cascade was hypothesized to have an increased therapeutic effect compared to either of the drugs alone. To test this hypothesis, MPO-Gd was used to follow a response to a synergistic combination therapy targeting both myeloid and lymphoid inflammation by co-administering sub-optimal concentrations of ABAH and GA in a mouse model of autoimmune neuroinflammation.⁵⁰ In spite of the fact that GA does not directly target MPO, unlike ABAH, a clear signal decrease using Gd-MPO was observed in the animals administered with the combination therapy compared to either ABAH or GA alone. This observation illustrates the intricate structure of inflammatory cascades, where the activity of MPO senses the changes from upstream modulation, affected by GA. Therefore, MPO-Gd can be successfully utilized in tracking therapeutic effect of anti-inflammatory drugs, regardless of their molecular target.

Another potential application of MPO imaging is in transplant rejection. For instance, the current gold standard for monitoring graft rejections in heart transplant recipients is serial endomyocardial biopsy. Graft rejection is accompanied by the accumulation of specific myeloid cells, including Ly-6C^hi, which accumulate progressively in allografts, and were shown to secret 10-fold higher levels of MPO.⁴² In a mouse model, MPO-Gd was used to selectively detect allograft rejection, while no MPO-Gd signal enhancement was observed in isograft recipients and allograft recipients in MPO-deficient mice. In this study, the therapeutic effect of some immunosuppressive drugs on allograft rejection was monitored by MPO-Gd MRI. A rapid drop in the population of myeloid cells upon administration of cyclosporin, prednisolone and azathioprine, as revealed by flow cytometry, resulted in a complete loss of myocardial enhancement. After administration of sub-optimal doses of prednisolone, MPO-Gd MRI could detect low-grade rejection was demonstrated, while higher doses of prednisolone resulted in lower contrast enhancement.

In gliomas an inflammatory response can often interfere with brain tumor imaging and can impede with treatment planning, as active inflammation can mimic tumor regrowth. In particular, oncolytic viral (OV) therapy has been successfully used to treat glioma in inoperable patients, but the immunogenic nature of these viruses causes their rapid clearance from the tumor by inflammatory cells. By using MPO-Gd MRI to detect intra- and peritumoral inflammation in rodent gliomas, higher contrast enhancement was observed in animals with a higher OV dose, in line with a stronger inflammatory response at a higher viral loading. At the same time, smaller tumors were reported for animals with a higher OV dose. However, a gradual clearance of OV by inflammatory cells could be observed by MPO-Gd 1-7 days after administration, and promoted continuous tumor growth.⁵²

Nonalcoholic fatty liver disease (NAFLD) is the most common liver disease in the world, which is projected to take even a heavier toll on the healthcare system in the upcoming years. Due to lower levels of MPO expression at early stages of NAFLD, MPO-Gd MRI allowed for differentiation between a relatively benign early-stage steatosis with no significant signal enhancement and a considerably more dangerous late-stage steatohepatitis (NASH), which often progresses into cirrhosis (Figure 3B).³² The potency of the probe was further validated by imaging human tissue samples extracted from patients with and without NASH.

It was shown that MPO-Gd is not taken up by cells, such as activated macrophages, and does not exhibit significant cytotoxicity at concentrations up to 5 mM.^{40, 53} Extensive preclinical evaluation of MPO-Gd in various animal models revealed no obvious signs of toxicity, although detailed high dose safety and toxicology studies have not yet been reported.

The original MPO-Gd was synthesized using a linear DTPA-bisamide chelator, which is unstable to Gd release in vivo. To enhance the stability, a DOTAGA-based complex, Gd-5-HT-DOTAGA, bearing just one hydroxytryptamide moiety was synthesized and was successfully used to detect pulmonary inflammation in a bleomycin induced lung injury mouse model, as well as to detect inflammation in human aneurysm tissue samples.^{37, 38}

MPO-Gd sensed MPO by having its relaxivity increase and by being retained in tissue. An alternative approach is to design a probe that has extremely low relaxivity rendering it undetectable in the absence of MPO but converted to a high relaxivity species by MPO. Gadolinium only has one accessible oxidation state but transition metals like iron and manganese can stably exist in different oxidation states and different spin states. For instance, high spin Mn²⁺ is a potent relaxation agent but Mn³⁺ is typically a poor relaxation agent. In a series of papers it was shown that Mn-based probes could be prepared that were chemically stable to Mn release in both the Mn²⁺ and Mn³⁺ forms, and that the redox potential for the probe could be tuned to such that the Mn³⁺ form was produced in the presence of oxidants like hydrogen peroxide while the Mn²⁺ form could be produced by reduction with glutathione.^{54, 55} The difference in the relaxivity between the “off” Mn³⁺ form and the “on” Mn²⁺ form was more than 10-fold and independent of field strength.⁵⁶

For Mn-based probes, the signal is turned off (relaxivity decreased) upon oxidation. For sensing, it would be preferred to have the signal increase in the presence of pathology. For Fe, the 2+ form is low relaxivity while the 3+ oxidation state is the high relaxivity form. This approach was taken recently with Fe-PyC3A, which can cycle between two stable oxidation states Fe²⁺ and Fe³⁺, and the probe was used to image oxidative stress associated with pancreatitis in a murine model.⁵⁷ By rationally designing the metal chelator, an optimal Fe²⁺/Fe³⁺ redox potential was attained for Fe-PyC3A, allowing for reversible switching between an effectively MRI silent Fe²⁺ state (r₁ = 0.18 mM⁻¹s⁻¹ at 4.7 T) and an MRI-active Fe³⁺ state (r₁ = 2.4 mM⁻¹s⁻¹ at 4.7 T). The authors showed that the complex is stable to Fe release in both oxidation states, and that the Fe²⁺ is extremely rapidly oxidized to the Fe³⁺ form by MPO. They showed that the Fe²⁺ probe administered to normal mice showed little to no enhancement, even in the excretory organs because of the low relaxivity. However, in a model of pancreatitis, i.v. injection of the Fe²⁺ probe showed selective enhancement of the pancreas (Figure 4), and that the signal change observed in the pancreas correlated linearly with the activity of MPO measured ex vivo.

Figure 4.

Axial T_1w MR images of saline-control (A) and caerulein/LPS (LPS - lipopolysaccharides) injured mice (B) prior to- and 6 minutes after IV injection of Fe-PyC3A before (C, saline; D, LPS). Organs are labelled as follows: P - pancreas, Sp – spleen, St – stomach, K – kidney, B – bowel, M – muscle. Reproduced with permission from ⁵⁷. Schematic representation of the molecular mechanism behind the signal enhancement in the presence of ROS (E).

III. Protein-based GBCAs

While most contrast agents are based on discrete metal chelates, there have also been biological approaches to molecular MR.⁵⁸ Half a century ago, Gd³⁺ and Mn²⁺ were used as paramagnetic probes to study calcium binding proteins by NMR. Ca²⁺ and Gd³⁺ are similar in size and both have preference for oxygen donor atoms, although Gd³⁺ binds with higher affinity because of its higher charge. Using bovine serum albumin as a model macromolecule, a remarkable increase of Gd³⁺ relaxivity was observed upon its binding to a protein (r₁ = 72 mM⁻¹s⁻¹ at 20 MHz).⁵⁹ Even higher relaxivity values were reported for immunoglobulin G (r₁ = 112 mM⁻¹s⁻¹ at 20 MHz) and glutamine synthetase (r₁ = 148 mM⁻¹s⁻¹ at 22.5 MHz).^{58, 60, 61} Until very recently,⁶² no endogenous metalloproteins containing lanthanide ions had been reported, and therefore calcium binding sites of proteins such as parvalbumin were repurposed to bind lanthanides.⁶³ To improve selectivity towards lanthanides, subtle chemical modifications were introduced by the Yang group at Georgia State University.⁶⁴ For instance a single aspartate mutation in the Ca²⁺ binding parvalbumin protein resulted in a higher binding affinity and selectivity towards gadolinium over competing transition metal ions and effectively precludes transmetallation of gadolinium.⁶⁵ There are two Gd ions in this small 12 kDa protein and the rigid binding site results in very high relaxivity with r₁ = 33.2 mM⁻¹s⁻¹ and r₂ = 44.6 mM⁻¹s⁻¹ per Gd³⁺ at 1.4 T and r₁ = 18.9 mM⁻¹s⁻¹ and r₂ = 48.6 mM⁻¹s⁻¹ per Gd³⁺ at 7 T. The mutant protein was PEGylated to increase solubility, blood circulation time and liver uptake, and the resulting probe, ProCA32-P40, was successfully used to visualize small liver tumors in animal models using both T_1w and T_2w images for detection.

The ProCA32 high relaxivity protein is readily derivatized for targeting. In one study, a vector that targets the chemokine receptor 4 (CXCR4) was conjugated to the protein and used to detect liver tumors in a mouse model (Figure 5).⁶⁶ The same ProCA32 protein was also attached to a type I-collagen targeting peptide (ProCA32.collagen1) to image liver fibrosis (discussed below) and a prostate specific membrane antigen (PSMA)-targeting peptide (ProCA32.PSMA) to visualize implanted PSMA-positive tumors in mice.^{67, 68} In the latter case, high r₂ relaxivity of the probe resulted in a significantly lower signal from tumors in T_2w images, while a more moderate signal change was reported for T_1w images, reflecting a decrease in the r₁ relaxivity at the higher field strength used in that study (7T). Although the ProCA32 construct is relatively small it does have high retention in the liver, and optimal imaging in terms of target to background contrast was often only achieved a day after administration.

Figure 5.

Schematic representation of ProCA32.CXCR4 bound to a targeted protein CXCR4 (A). Axial T_1w MR images of mice bearing subcutaneous tumors before (B) and 24h after (C) administration of ProCA32.CXCR4. For comparison, axial T_1w images show lower signal enhancement in animals administered with a CXCR4-blocking agent 24h after injection with ProCA32.CXCR4 (E) relative to the baseline (D). Adapted with permission from ⁶⁶.

IV. Probes targeting the fibrin-fibronectin complex and extradomain-B fibronectin

Epithelial-Mesenchymal Transition (EMT) is an essential part of tissue patterning and organization processes during embryonic development, which involves transdifferentiation of epithelial cells into mesenchymal cells.⁶⁹ At the same time, this process can be implicated in various pathologies, including many cancers, where epithelial cells of the primary tumor lose their intracellular adhesion and acquire migratory properties after EMT activation.⁷⁰ Since EMT causes pronounced changes in the content and structure of extracellular matrix (ECM), some ECM proteins were identified as promising biomarkers, which can be used as selective targets in molecular imaging. One of them, an extracellular glycoprotein fibronectin, is known to form complexes with other proteins in the matrix. The Lu group at Case Western has been developing probes to fibronectin. Fibrin is present in the tumor ECM as a result of its precursor fibrinogen accumulating through extravasation from leaky tumor blood vessels. The fibrin-fibronectin (FB-FN) complex was investigated as a promising target for molecular MRI by using a Gd-DTPA-based probe CLT1-(Gd-DTPA) where CLT1 is a FB-FN-selective cyclic peptide CLT1.⁷¹ The CLT1 peptide was identified by an in vivo phage display screen.⁷² Despite the high specificity of the molecular agent towards tumor cells, as confirmed by immunohistochemical and blocking studies, a relatively low contrast enhancement was observed. A higher signal intensity was achieved by increasing the metal loading per probe in two nanoglobular Gd-DOTA-monoamide conjugates bearing the same CLK1 peptide.⁷³ With approximately 25-43 gadolinium ions and 2-3 peptides per nanoparticle, a significant increase in relaxivity was reported (r₁ ~ 8 mM⁻¹s⁻¹ vs r₁ = 4.2 mM⁻¹s⁻¹ per Gd at 3T), which eventually translated in a larger signal enhancement in an animal model of breast and orthotopic prostate cancer and enabled administration of lower doses of the contrast agent (0.03 mmol/kg vs 0.10 mmol/kg) .^{73, 74} In an alternative approach to boost contrast enhancement, a low molecular weight probe CLT1-dL-(Gd-DOTA)₄, comprised of four DOTA-monoamide moieties attached to a single CLT1 peptide (r₁ = 10.1 mM⁻¹s⁻¹ at 1.5T), was tested in an animal model of orthotropic prostate cancer.⁷⁵ Another reported FB-FN-specific pentapeptide CREKA (Cys-Arg-Glu-Lys-Ala), identified from an in vivo phage display screen,⁷⁶ was conjugated to four DOTA-monoamide moieties to give the probe CREKA-dL-(Gd-DOTA)₄ (r₁ = 8.3 mM⁻¹s⁻¹ at 1.5 T).⁷⁷ Compared to the CLT1 probe, the CREKA probe showed much lower residual uptake in non-targeted organs in the same animal model of orthotropic prostate cancer 48 h post-injection.

A second fibronectin related target, the oncofetal isomorph of fibronectin, extradomain-B fibronectin (EDB-FN), is selectively overexpressed in many tumors. For instance aggressive prostate tumors with high metastatic potential express EDB-FN and this marker could be used to distinguish them from benign prostatic hyperplasia.⁷⁸ With this in mind, a HP-DO3A-based gadolinium probe ZD2-Gd(HP-DO3A) with an attached EDB-FN-targeting peptide ZD2 was used to distinguish between aggressive, androgen insensitive PC3 derived mouse tumor models, which exhibit high levels of EDB-FN from slower growing, androgen-sensitive LNCaP derived tumors.⁷⁹ Indeed, a selective uptake of ZD2-Gd(HP-DO3A) was observed only in animals bearing PC3 cancer cells, but not LNCaP cells. On the other hand, no significant difference between two groups was revealed when using the CREKA-Gd(HP-DO3A) probe, targeting the FB-FN complex, or the GBCA gadoteridol. ZD2-Gd(HP-DO3A) was also evaluated in the preclinical model of metastatic breast cancer using both murine and human cell lines,⁸⁰ since high expression of EDB-FN is associated with poor survival rate in patients with breast cancer.⁸¹ In animals bearing either of the cell lines, a specific uptake of ZD2-Gd(HP-DO3A) was detected both in primary tumors and in metastases (adrenal glands and lymph nodes), while no specific uptake was evident after injecting gadoteridol.

V. MR probes targeting fibrin: EP-2104R

Fibrin, formed by the action of the protease thrombin on soluble fibrinogen, is a principal component of blood clots. The insoluble fibrin strands then help in platelet aggregation to form a fibrin-based provisional matrix which can entrap platelets forming a plug to stem blood loss.^{82, 83} Fibrin is an attractive target for molecular imaging as it is only present at the site of injury and its limited solubility precludes its circulation in the blood. Additionally, fibrin deposition is specific to injury but can result from a myriad of unrelated diseases and conditions. Targeting fibrin with molecular imaging MR probes has the potential for a breadth of applications.

EP-2104R is a fibrin-targeted MR probe comprised of an 11-amino acid fibrin-binding peptide coupled to four Gd-DOTA chelates and originally developed by EPIX Pharmaceuticals (Figure 6A). The parent peptide was found in a phage display screen against fibrin in which the phage library was first depleted of fibrinogen- and serum protein binding phage.⁸⁴ This negative preselection engineered in selectivity for fibrin over fibrinogen and other plasma proteins. Further medicinal chemistry efforts optimized the probe with respect to fibrin affinity, metabolic stability and relaxivity.^85-87 The relaxivity of EP-2104R bound to fibrin is 18 mM⁻¹s⁻¹ at 1.4 T per Gd and 72 mM⁻¹s⁻¹ per molecule allowing it to be used at low doses (Figure 6B).⁸⁸

Figure 6.

Chemical structure of EP-2104R (A). T₁ relaxivity of EP-2104 R in PBS (filled triangle), in the presence of 30 μM fibrinogen (hollow circle) or in the presence of 30 μM fibrin (filled circle) (B, figure and caption adapted from ⁸⁸ ). Visualization of early (2 weeks, C) and advanced/remodeled (4 weeks, D) AAA-associated thrombi by EP-2104R enhanced MRI. On pre-contrast agent scans and control scans with Gd-DTPA (C and D top left and center columns), no significant enhancement was measured. On the EP-2104R enhanced MRI, a strong signal enhancement was measured at the location of the fibrin-rich thrombus (C, top left, red arrows). Corresponding histology (C, bottom left and center) and immunohistochemistry (C, bottom right) show the formation of a fresh fibrin-rich thrombus with a strong signal using a fibrin-specific antibody. After 4 weeks of angiotensin II infusion, the EP-2104R enhanced MRI scan shows a weak-to-moderate signal enhancement at the location of the fibrin-rich thrombus (D, top right, red arrows). Corresponding histology and immunohistochemistry (D, bottom row) show the formation of an advanced remodeled thrombus, with low signal of fibrin-specific antibody (D, bottom right) with significant expression of elastin and collagen fibers in areas adjacent to the thrombus (D, bottom left and center), indicating that remodeling took place. C and D, figures and caption adapted from ¹⁰². EP-2104R enhanced MRI of injured rat liver (E, Figure and caption adapted from reference ¹⁰⁸). T₁-weighted axial images of rats dosed with PBS (left), with DEN and imaged 7 days after fourth DEN dose (center), and imaged 1 day after fourth DEN dose (right). False color overlay represents subtraction image of pre-contrast image from the 1-minute post EP-2104R image. Peak liver signal enhancement (%ΔSI) post injection of EP-2104R is significantly higher for DEN animals compared to PBS controls

EP-2104R enhanced MRI was successfully used to detect thrombi in animal models of stroke,⁸⁹ pulmonary embolism,^90-92 deep vein thrombosis,^{93, 94} cardiac chamber clot,^{89, 90} coronary artery thrombus, and atherosclerotic plaque rupture.⁹⁵ EP-2104R advanced to Phase 1 and Phase 2 proof of concept studies in humans at a dose of 0.004 mmol/kg and was well tolerated.⁹⁶ The probe was able to detect thrombus in a range of vascular territories: heart chambers, lungs, aorta, deep veins.⁹⁷ Recently a version of this probe was synthesized replacing the Gd with stable Mn-chelates and used to image carotid artery thrombus in a rat model.⁹⁸

It has become increasingly recognized that fibrin represents a useful imaging target outside the blood vessels as well. In many pathologies the endothelium is disrupted resulting in leakage of plasma proteins like fibrinogen into the interstitial space. When fibrinogen extravasates it becomes converted to fibrin in the extracellular matrix and plays a role in inflammation, angiogenesis, wound healing, and cancer. In the context of cancer, EP-2104R was used to detect fibrin in a rat model of glioma and a mouse model of breast cancer.^{99, 100} Fibrin imaging in cancer may provide more specificity for malignant versus benign lesions.

More recently, EP-2104R was used to image fibrin associated with the vascular leak that occurs in the lung during pulmonary fibrosis.¹⁰¹ In this study it was shown that EP-2104R enhanced MRI could quantify fibrin deposition in mouse lungs using an ultrashort echo time sequence when the mice were injured with bleomycin compared to uninjured mice. The same study showed that the direct thrombin inhibitor dabigatran, an oral anticoagulant, acted as an effective antifibrotic therapy in this model. EP-2104R enhanced MRI was used to monitor the treatment effect of the dabigatran treatment. This study demonstrates the potential utility of MR imaging of fibrin in drug development and translation.

The utility of EP-2104R was also demonstrated in quantifying fibrin in abdominal aortic aneurysms (AAAs) using an angiotensin-II (Ang-II) infusion mouse mode.¹⁰² Injury to the aortic wall leads to release of prothrombotic factors and the formation of a fibrin-rich hematoma, initializing formation of an aneurysm.^{103, 104} As the size of the aneurysm increases, a fibrin-rich thrombus is formed which is stabilized as it matures by increasing levels of extracellular matrix proteins like collagen and elastin.^{103, 104} Clinically, aneurysms are characterized only by their diameter and no methods can assess the age or the risk for rupture of an aneurysm.¹⁰⁵ Fibrin-specific MRI with EP-2104R in vivo quantified changes in fibrin content in AAAs and enabled differentiation between early, fibrin rich thrombi and advanced stable aneurysms. The MR signal changes increased gradually from 1, 2 and 3 weeks after Ang-II infusion and decreased at week 4 post Agn(II) infusion (Figure 6C and 6D). The changes in MR signal at different stages of aneurysm development matched ex vivo immunohistomorphometry measures of fibrin and correlated with decrease of fibrin content and increase of collagen and elastin content in thrombi. This study suggests that fibrin-specific MRI with EP-2104R could provide a novel in vivo biomarker for aortic aneurysms which can ultimately improve the risk stratification of patients.¹⁰²

Coagulation and repair mechanisms involving fibrin play a role in inflammatory and fibroproliferative components of liver injury and abnormal extravascular deposition of fibrin has been observed in rodent models of liver injury.^{106, 107} Fibrin-targeted MRI with EP-2104R was used to detect hepatic tissue injury and inflammation in diethylnitrosamine (DEN)-injured rats. EP-2104R signal enhancement was significantly higher in the liver of animals imaged at 1 day post-DEN dosing compared with 7 days post-DEN or in control rats (Figure 6E). This enhancement was consistent with the histological activity index score for inflammation, indicating that EP-2104R enhancement is indicative of inflammation. Using a modified version of EP-2104R, which could not bind fibrin, as a control it was shown that the signal enhancement observed in the acutely injured liver was specific to fibrin. The study also showed that fibrin associated inflammation could be selectively imaged on a background of liver fibrosis, i.e. that the presence of fibrosis did not affect the ability of EP-2104R enhanced MRI to detect. These results demonstrate that EP-2104R enhanced MRI can assess hepatic inflammation and serve as a biomarker for tissue injury and may be used to monitor modulation of disease in response to therapeutic intervention.¹⁰⁸

VI. MR probes targeting collagen: EP-3533, CM-101 and ProCA32.collagen1

Formation of collagen-rich scar is the final step in the wound healing cascade replacing injured tissue. However, many chronic diseases or acute injuries can result in a buildup of scar tissue or fibrosis that can lead to organ dysfunction, organ failure, or death. Fibrosis is essentially an accumulation of extracellular matrix, with chief components being fibronectin, elastin and type I collagen, where elastin content increases in older injury.¹⁰⁹ EP-3533 is a type I collagen-specific MR probe comprised of four gadolinium chelates and a peptide that binds type I collagen with moderate affinity (K_d=1.8 μM) and high specificity (Figure 7A).¹¹⁰ The peptide in EP-3533 was identified by a phage display screen and its affinity improved via a medicinal chemistry campaign.^{87, 111} EP-3533 and its next generation analog CM-101 have high potential for clinical translation because 1) both probes are relatively small and clear rapidly from the blood allowing for delayed phase imaging shortly after probe administration, 2) they have high T₁ relaxivity at clinically relevant magnetic field strengths, and 3) excessive deposition of type I collagen is specific to fibrosis across different organs systems and pathologies, 4) diseases with a fibroproliferative component collectively account for at least one third of deaths worldwide,¹¹² and 5) there is a critical but yet unmet need for non-invasive methods to detect, stage, and monitor changes in tissue fibrosis.

Figure 7.

Chemical structure of the Type I collagen-binding probes EP-3533, CM-101 and ProCA32.collagen1 (A). Collagen imaging with EP-3533 can assess reduction of liver fibrosis with EDP-305 treatment in BDL rats (B-C). B: Representative axial R₁ map images of the liver for each treatment group before and after injection of the EP-3533 probe. C: Change in relaxivity (ΔR₁) in liver tissue as a surrogate measure of probe binding (B-C, figures and caption adapted from ¹¹⁸). D: T₁-weighted MR images at baseline (left), 5 min post injection (center) and 55 min post injection (right) of a tumor imaged with EP-3533 (top row), and for healthy pancreas imaged with EP-3533 (bottom row). Tumor and healthy pancreas are outlined. D: MR signal enhancement maps for the tissue in C showing sustained enhancement at 55 min post injection with EP-3533 in tumor (top row) but not in healthy pancreas (bottom row). Figure and caption adapted from ¹²⁹. E-F: ProCA32.collagen1 can detect early stages of liver fibrosis in TAA/Alcohol model. R₁ maps (F) and ΔR₁ values (G) of normal (Ishak stage 0 of 6, bottom image), early-stage (Ishak stage 3 of 6, middle image), and late-stage (Ishak stage 5 of 6, top image) liver fibrosis before and 24 h after injection of ProCA32.collagen1. ProCA32.collagen1 can distinguish early-stage liver fibrosis from normal and late-stage fibrotic liver with ΔR₁ derived from a R₁ map at 24 h time point (F). Figure and caption adapted from reference ⁶⁷.

EP-3533 enhanced MR imaging has been applied extensively to detect and stage liver fibrosis in preclinical models. Early work showed that EP-3533 could detect disease in rat and mouse models and that the liver signal increased with increasing fibrosis in a mouse carbon tetrachloride model of liver fibrosis.^{113, 114} More recently, EP-3533 has been used to accurately stage hepatic fibrosis in a rat of bile duct ligation (BDL) model mimicking fibrosis caused by biliary stasis.¹¹⁵ EP-3533 enhanced MRI was performed 4, 10, or 18 days following BDL. Quantification of the change in liver longitudinal relaxation rate (ΔR₁) following i.v. EP-3533 administration revealed significant increases in ΔR₁ in the livers of BDL rats compared to rats that underwent a sham procedure, and that the ΔR1 values increased with stage of disease as assessed by ex vivo analyses including blinded scoring of the tissue by a pathologist, digital pathology, and biochemical quantification of hydroxyproline (a proxy for total collagen).¹¹⁵ Receiver operating characteristic (ROC) curve analysis showed that EP-3533 MRI could accurately discriminate different stages of fibrosis. In the same study it was shown that EP-3533 MR could prospectively and accurately determine treatment response compared to the gold standard of histology.¹¹⁵ In another study EP-3533 was used to assess the antifibrotic effect an FXR agonist in clinical trials for treatment for non-alcoholic steaohepatitis (NASH).^{116, 117} Collagen imaging with EP-3533 accurately and non-invasively detected treatment response of EDP-305 in BDL rat models of hepatic injury (Figure 7B), and the reduction of fibrosis was consistent with biochemical (hydroxyproline), genetic (Col1a1 mRNA) and histological (collagen proportional area, CPA) ex vivo measures of fibrosis.¹¹⁸ In a diethylnitrosamine (DEN)-induced liver injury rat model of hepatic fibrosis, collagen imaging with EP-3533 was compared with MR elastography (MRE), a technique that measures stiffness and mechanoelastic properties of tissue and has been applied to assess advanced stages of liver fibrosis and cirrhosis.¹¹⁹ EP-3533 enhanced MRI was shown to be more sensitive to early stages of fibrosis, complementary to MRE which is more sensitive to advanced fibrosis.¹²⁰ In a recent study, EP-3533 was used in a rat model of NASH to detect and stage fibrosis disease progression and to monitor the effect of either dietary or pharmacological intervention.¹²¹ This study again showed the utility of the probe in detecting fibrosis, this time on a complex background of steatosis and inflammation. Together, these studies demonstrate that collagen imaging with EP-3533 can sensitively characterize hepatic fibrosis in multiple animal models, degrees of disease severities, and can sensitively and non-invasively detect and assess treatment response in fibrotic livers, which historically has relied on histologic analyses.^{122, 123}

EP-3533 has been used to image other disease models with a fibrotic component such as myocardial infarction and pulmonary fibrosis.^{110, 124, 125} Targeting collagen in the heart has also been used to assess myocardial perfusion in a steady state protocol analogous to nuclear cardiology.¹²⁶ The utility of EP-3533 is further demonstrated in the noninvasive quantification of muscle fibrosis a mouse mdx model of Duchenne muscular dystrophy (DMD). Although muscular dystrophies, including DMD, are generally considered to be conditions of muscle degeneration, muscle fibrosis is an important yet underrecognized component of pathogenesis of muscular dystrophies.¹²⁷ Administration of the type I collagen-specific probe EP-3533 to mdx mice resulted in significant changes in MRI relaxation rate (R₁) within the gastrocnemius and tibialis anterior muscle tissue, while no changes were observed in the same tissues of normal mice. These changes in R₁ correlated with ex vivo hydroxyproline content, and support the utility of EP-3533 in the noninvasive quantification of fibrosis within skeletal muscles.¹²⁸

Fibrosis is a hallmark of some solid tumors, such as pancreatic cancers where fibrosis is often referred to as the desmoplastic reaction. In an orthotopic syngeneic mouse model of pancreatic ductal adenocarcinoma (PDAC, Figure 7D, 7E), it was shown that EP-3533 showed collagen-specific enhancement of the tumor and could be distinguished from non-specific tumor enhancement.¹²⁹ EP-3533 MR enhancement of PDAC tumors correlated with the spatial distribution of collagen in the tumors as determined by second harmonic generation imaging.¹²⁹

CM-101 is an improved version of EP-3533 and bears the same type 1 collagen-binding peptide. In CM-101 the gadolinium chelate Gd-DTPA is replaced with the highly stable macrocyclic Gd-DOTA chelate.¹³⁰ In rats, CM-101 showed faster elimination than EP-3533 and was more completely eliminated from the body with no accumulation of Gd in bone which was observed with EP-3533. Similar to EP-3533, CM-101 detects fibrosis in rat BDL and mouse carbon tetrachloride (CCl₄) models of liver fibrosis.¹³⁰ CM-101 was also used in a mouse model of PDAC to measure changes in fibrosis in response to FOLFIRINOX chemotherapy.¹³¹ MR signal enhancement with CM-101 was greater and more prolonged in tumor regions than in the surrounding pancreas, which was not observed with the non-targeting Gd-DOTA. In FOLFIRINOX-treated tumor mice, there was increased histological fibrosis which associated with increased uptake of CM-101 in tumor regions as compared to untreated tumors, demonstrating that CM-101 can monitor the desmoplastic response of PDAC to neoadjuvant chemotherapy.¹³¹

The same type I collagen targeting peptide used in EP-3533 and CM-101 was used in a protein-based contrast agent described above, ProCA32. Linking the collagen specific peptide to ProCA32 resulted in collagen-specific ProCA32.collagen1, which was used for the early diagnosis and noninvasive detection of liver fibrosis in alcohol/chemical-induced mouse models of liver fibrosis and a diet-induced mouse model of NASH (Figure 7F and 7G).⁶⁷ Because of its much larger size compared to EP-3533 and CM-101, ProCA32.collagen1 has a longer blood residence time, requiring delayed imaging hours or a day after probe injection.⁶⁷ Nonetheless, the very high relaxivity of this protein-based MR probe increases its potential for clinical translation.

VII. Collagen cross-linking: Gd-Hyd, Gd-CHyd and Gd-OA

The cross-linking of collagen and elastin is crucial to stabilizing the extracellular matrix during wound healing and formation, and is catalyzed by the action of lysyl oxidase (LOX) enzymes.¹³² LOX enzymes oxidize lysine amino groups to aldehyde containing allysine residue that further reacts to form stable cross-links.^{133, 134} Because of the abundance of collagen in fibrosis and the large number of amino groups per collagen monomer, allysine is an attractive and suitable molecular MR target and marker of fibrogenesis, and several probes have been developed to quantify fibrogenesis in this manner.^135-137

Gd-Hyd (Figure 8A), is a small molecule Gd-DOTA-based probe containing a reactive hydrazine that readily undergoes a reversible covalent bond reaction with allysine. Gd-Hyd was first used to image lung fibrogenesis in a bleomycin induced pulmonary injury in mice.¹³⁵ Gd-Hyd, but not a negative control probe Gd-DiMe, was able to specifically detect and quantify fibrogenesis in the lung, distinguish fibrogenesis (active disease) from stable scar, and could monitor response to treatment with an inhibitor of lysyl oxidase.¹³⁵ Gd-Hyd resulted in 5-fold higher lung to muscle contrast to noise ratio 2 weeks after injury compared to the lungs of mice that underwent a sham procedure (Fig. 8D). Furthermore, Gd-Hyd lung enhancement decreased at 4 weeks after bleomycin injury, consistent with formation of stable scar as fibrogenic activity decreases. Gd-Hyd lung enhancement also decreased in the lungs of bleomycin injured mice that received the pan-LOX inhibitor, β-aminopropionitrile (BAPN) in response to decreased fibrogenesis (Figure 8D). Additionally, in vivo lung enhancement with Gd-Hyd correlated with ex vivo biochemical quantification of LOX activity and expression, and allysine content, the molecular target of Gd-Hyd (Figures 8B and 8C).¹³⁵.

Figure 8.

Chemical structures of allysine-binding MR probes Gd-Hyd, Gd-CHyd and Gd-OA (A). Gd-Hyd quantifies lung fibrogenesis and response to treatment in bleomycin-injured mouse lungs. Elevated levels of hydroxyproline (B, HYP) a biochemical proxy for collagen, and allysine (C) the molecular target of Gd-Hyd, correlate with the contrast to noise ratio change, ΔCNR, from Gd-Hyd injection (D). Gd-Hyd enhanced coronal MR images of mouse lungs (E). (B-E, Figures and caption adapted from ¹³⁵). Quantification of ΔCNR in the lungs of bleomycin-injured mice after injection of Gd-Hyd or Gd-CHyd reveals Gd-CHyd as superior to Gd-Hyd in MR enhancement of bleomycin-injured mouse lungs. Gd-CHyd lung enhancement is always greater than Gd-Hyd enhancement at all three time points measured (F). Pairwise analysis of the ΔCNR in the lungs of BM animals observed arising from Gd-Hyd or Gd-CHyd injection (G). (F-G, Figure and caption adapted from ¹³⁷).

Gd-Hyd was also applied in the detection of hepatic fibrogenesis caused by CCl₄ in a mouse model of liver fibrosis.¹³⁵ Gd-Hyd enhanced MR of the liver was sensitive to progression of liver injury with time and to reduction in fibrogenesis in mice after withdrawal of CCl₄ as assessed by histology and lysyl oxidase gene expression.¹³⁵ In a second model of NASH liver fibrosis, Gd-Hyd sensitively detected reduced fibrogenesis after treatment with the farnesoid X receptor agonist EDP-305. Response to treatment was verified by standard ex vivo measures including pathological scoring, digital pathology, quantification of hydroxyproline, and quantification of lysyl oxidase gene expression.¹³⁵ A recent multi-parametric MRI study compared the utility of molecular MRI with the type I collagen specific probe EP-3533, the fibrogenesis probe Gd-Hyd, MRE, native T₁, and Dixon MRI in assessment of disease progression and treatment response in rat dietary model of NASH.¹²¹ Gd-Hyd enhanced MRI was able to stage fibrosis and had the highest accuracy in detecting treatment response in NASH.¹²¹

Gd-CHyd (Figure 8A), is another allysine-binding MR probe that has improved reactivity compared to Gd-Hyd.¹³⁷ Gd-CHyd contains a slightly modified hydrazine binding unit that results in faster reactivity with and higher affinity towards aldehydes while maintaining general structure, pharmacokinetics and clearance.¹³⁷ Gd-CHyd was compared directly to Gd-Hyd in a crossover study in a bleomycin injury lung fibrosis mouse model. Gd-CHyd i.v. administration resulted in higher and more prolonged signal enhancement in the lungs of injured mice compared to Gd-Hyd, resulting in improved sensitivity (Figure 8C).¹³⁷

Gd-OA (Figure 8A), is another improved version of the allysine binding probe Gd-Hyd. The allysine-reactive hydrazide featured in Gd-Hyd is replaced with an oxyamine group in Gd-OA, which forms a more stable bond with allysine with respect to hydrolysis. Gd-OA also binds allysine with a greater affinity than Gd-Hyd and displayed minimal non-specific accumulation and rapid renal excretion, and improved imaging of pulmonary fibrosis.¹³⁶ The stability of the Gd-OA allysine bond made it a suitable candidate for non-invasive imaging of renal fibrosis where it remains bound to its target after the unbound Gd-OA has cleared from the kidneys. In a nephrotoxic nephritis (NTN) model of kidney fibrosis MR imaging with Gd-OA resulted in a 7-fold increased uptake in the renal cortex of fibrotic kidney compared to kidneys in control mice. This increased uptake correlates with increased fibrotic burden was confirmed by ex vivo standard measures of fibrosis.¹³⁸ Significantly, no difference was found in the medulla or renal pelvis which is consistent with the distribution of fibrosis in this model.

Collectively, allysine binding MR probes have demonstrated great utility in assessment of pulmonary, renal and hepatic fibrogenesis and can sensitively monitor response to treatment and can distinguish active disease from stable scar. The ability to determine disease activity would also be useful in the advancement of novel treatments through clinical trials since the stratification of patients based on disease activity would enable enrolment of patients that are the most at risk for disease progression and would allow for cohort enrichment. This strategy would enable reduction of sample size and would overall improve trial feasibility.

VIII. Molecular imaging of elastin: Gd-ESMA

Elastin is an abundant protein in arterial vessel wall and its expression is increased during plaque formation. The gadolinium-based elastin-specific MR probe Gd-ESMA (originally named BMS753951, Figure 9A) was developed to selectively image arterial walls¹³⁹ and was then used to non-invasively quantify elastin in atherosclerotic plaque burden in mice and to assess coronary artery remodeling following coronary injury in swine model.^{140, 141} Gd-ESMA was also used to detect plaques that are at risk of rupture in a rabbit model of atherosclerosis,¹⁴² and was able to characterize atherosclerotic plaques in large animal models in clinically relevant setting. Additionally, Gd-ESMA was used to assess plaque burden in animal models of atherosclerosis,^{142, 143} and to monitor elastin content response to therapeutic interventions.¹⁴⁴

Figure 9.

Chemical structure of elastin-binding MR probe Gd-ESMA (A). Images of nondisrupted (B) and disrupted (C) artherosclerotic plaques in a rabbit model. Cross-sectional unenhanced T1-weighted black blood images obtained before pharmacologic triggering show vessel wall thickening (B, center image). Corresponding images obtained after triggering (C, center) for plaque disruption show presence of thrombus (arrow) overlying disrupted plaque. Photomicrographs (Masson trichrome stain; original magnification, ×2.5) show atherosclerosis and thrombus (B and C, right). Figure and caption adapted from ¹⁴². D: Coronal T1-weighted MR images before and 24 hours after intravenous injection of Gd-ESMA and Gd-DTPA in healthy and adenine-induced fibrotic kidneys. Quantification of normalized MRI signal intensities in renal cortex and medulla of Gd-ESMA- and Gd-DTPA–injected mice with healthy and fibrotic kidneys shows higher difference in the change in kidney to muscle contrast to noise ratio between healthy and fibrotic kidney when Gd-ESMA is used as the contrast agent. Figure and caption adapted from ¹⁵².

In more recent studies, Gd-ESMA was used to track myocardial remodeling post myocardial infarction (MI) in a mouse model and was compared with the non-specific agent Gd-DTPA.¹⁴⁵ In infarcted tissue, Gd-ESMA uptake (as measured by R₁) increased gradually from day 3 to day 21 after MI injury and this increase correlated with histological findings which demonstrate increased synthesis of elastin. Additionally, Gd-DTPA showed no increase in R1 in the same time frame post MI.¹⁴⁵ Furthermore, Gd-ESMA can also be used to perform standard clinical reference techniques for visualization of arterial territories such as contrast enhanced magnetic resonance angiography (CE-MRA) and produce images of high quality similar to the clinically used gadobutrol.¹⁴⁶ Elastin-specific MRI with Gd-ESMA was also used in combination with imaging agents specific to inflammation. In a murine model of permanent coronary occlusion, combined ¹H MR using Gd-ESMA and ¹⁹F MR with perfluorocarbon nanoparticles allowed for simultaneous assessment of inflammatory response and ECM remodeling. The authors demonstrated that both an acute and prolonged inflammatory response or a lack of immune response had adverse effects on left ventricle remodeling and cardiac function. The authors demonstrated the prognostic value of simultaneously quantifying inflammation and elastin turnover and concluded that optimal healing post MI is achieved with a moderate well-balanced inflammatory response.¹⁴⁷ Clinical translation of this multichannel technique, however, is limited by the low sensitivity of ¹⁹F MR, which requires injection of high amounts of contrast agent. This limitation was addressed in another study by using macrophage avid iron oxide nanoparticles as T₂* agents. Elastin-specific imaging with Gd-ESMA was utilized along with macrophage specific imaging with iron-oxide probe ferumoxytol to predict aneurysm rupture in a mouse model of abdominal aortic aneurysm (AAA). ^{147, 148} Previous studies have demonstrated Gd-ESMA assessment of vascular remodeling in the prediction of vulnerable plaque (Figure 9B and C).¹⁴² The combined MR imaging of inflammatory activity with ferumoxytol and elastin remodeling with Gd-ESMA in response to injury in one imaging session proved to be more effective at predicting aneurysm rupture than either technique alone.¹⁴⁸ This strategy provides an efficient method for characterization of AAA in vivo and has high potential for clinical translation. Ferumoxytol is approved for treatment of iron deficiency in the USA and is used off-label as an MR contrast agent. Additionally, the authors in this study demonstrated that Gd-ESMA did not interfere with the T₂*-weighted images of ferumoxytol and ferumoxytol did not interfere with T₁-weighted images of Gd-ESMA when both these agents were administered at their anticipated clinical dose.¹⁴⁸

Gd-ESMA has also been used in the context of fibrotic disease. Excessive deposition of elastin and other extracellular matrix (ECM) proteins is also a common endpoint of virtually all progressive fibrotic disease. Elastin accumulates in advanced stages of liver fibrosis due to both its increased synthesis and decreased degradation,^{149, 150} and Gd-ESMA has been successfully used to detect ECM-remodeling during liver fibrosis.¹⁵¹ More recently, Gd-ESMA was used to assess ECM deposition following renal injury in three murine kidney fibrosis models of different etiology, including unilateral ureteral obstruction (UUO) and unilateral ischemia/reperfusion injury (IRI) as well as in adenine-induced nephropathy.¹⁵² MR signal due to Gd-ESMA accumulation in fibrotic kidneys reported as ΔCNR was significantly higher than in healthy kidneys, and in fibrotic kidneys the accumulation of Gd-ESMA was much higher than that with the non-targeted Gd-DTPA (Figure 9D and E). The Gd-ESMA enhanced MR signal correlated with histological and protein expression measures of elastin as well as ex vivo quantification of gadolinium concentration.¹⁵² The strongly up-regulated expression of elastin was also verified in renal tissue obtained from 6 patients with progressive CKD and compared to renal tissue obtained from healthy human kidneys.¹⁵² Ex vivo MR of human renal tissue with Gd-ESMA also demonstrated significantly elevated uptake of the probe in fibrotic tissue. Furthermore, Gd-ESMA-based molecular imaging was able to stage fibrosis in an adenine diet-induced nephropathy mouse model over 21 days and detect treatment response in the same model to the inflammasome inhibitor CRID3.¹⁵²

Conclusion

Molecular MRI remains a dynamically evolving field which can significantly benefit the quality and accessibility of clinical diagnostics. Among routinely used diagnostic procedures, invasive biopsy remains a first-line choice for many pathologies, although it suffers from sampling error and potentially harmful complications. On the other hand, non-invasive imaging modalities, such as molecular MRI, would address these shortcomings in current diagnostic methods. Molecular imaging can non-invasively quantify disease activity and monitor early response to therapy as has been demonstrated for the probes discussed above. Additionally, molecular imaging can identify and quantify targets for therapy and can better inform on effect of drugs on their targets.

The selected examples of recently designed molecular probes, which we discussed in this article, have been evaluated in different preclinical models and demonstrate the enormous potential within this field. However, a good amount of work still has to be done to capitalize on these promising preclinical results and address the needs of the clinical community. Probes that can meet the requirements of addressing unmet need(s), have a large addressable market, show strong imaging efficacy, and a clean safety profile have the highest chances of successful clinical translation. We hope to see new molecular MR contrast agents moving into clinical development in the near future.