Precision detection of liver metastasis by collagen-targeted protein MRI contrast agent

Biomaterials 224 (2019) 119478

Contents lists available at ScienceDirect

Biomaterials journal homepage: www.elsevier.com/locate/biomaterials

Precision detection of liver metastasis by collagen-targeted protein MRI contrast agent

T

Mani Salariana, Hua Yangb, Ravi Chakra Turagac,1, Shanshan Tana,1, Jingjuan Qiaoa, Shenghui Xuea, Zongxiang Guia, Guangda Pengc, Hongwei Hanc, Pardeep Mittald, Hans E. Grossniklausb, Jenny J. Yanga,e,∗ a

Department of Chemistry, Georgia State University, Atlanta, GA, 30303, USA Department of Ophthalmology, Emory University School of Medicine, Atlanta, GA, 30322, USA c Department of Biology, Georgia State University, Atlanta, GA, 30303, USA d Medical College of Georgia, Augusta University, Augusta, GA, 30912, USA e Center for Diagnostics and Therapeutics, Georgia State University, Atlanta, GA, 30303, USA b

A R T I C LE I N FO

A B S T R A C T

Keywords: Protein contrast agent Magnetic resonance imaging Liver metastasis Collagen I

The Liver is the most common organ for metastasis for various cancers, including uveal melanoma, the most common primary intraocular tumor. Uveal melanoma metastasizes to the liver in ~90% of patients, and results in death in almost all cases due to late detection and lack of effective treatment. There is a pressing unmet medical need to develop MRI contrast agents and imaging methodologies with desired sensitivity and specificity to overcome the high heterogeneous background and in vivo properties as well as reduced toxicity. Herein, we report the development of a collagen targeting protein contrast agent (ProCA32.collagen1), since collagen is a diagnostic biomarker and therapeutic target for many types of primary and metastatic cancers and the tumor microenvironment. In addition to a strong affinity to collagen I, ProCA32.collagen1 possesses high relaxivities (r1 and r2 are 68.0 ± 0.25 and 100.0 ± 0.32 mM−1 s−1 at 1.4 T, respectively, and 42.6 ± 1.0 and 217 ± 2.4 mM−1s−1 at 7.0 T per particle). ProCA32.collagen1 also has strong serum stability against degradation, resistance to transmetallation, and 102 and 1013-fold higher metal selectivity for Gd3+ over Ca2+ and Zn2+, respectively, compared to clinical contrast agents. ProCA32.collagen1 does not exhibit any cell toxicity for various cell lines. Sensitive detection of liver lesions in animal models can be achieved using multiple imaging methodologies, taking advantage of the dual relaxation property of ProCA32.collagen1. ProCA32.collagen1 enables sensitive and early stage detection of hepatic micrometastasis as small as 0.144 mm2 and two different tumor growth patterns. Further development of ProCA32.collagen1 has the potential to greatly facilitate noninvasive, early detection and staging of primary and metastatic liver cancers, and devising effective treatments.

1. Introduction Liver, as the largest organ of the body, has a unique architecture for its diverse functions. It is one of the sites of primary hepatocellular carcinoma and metastatic tumors such as uveal melanoma (UM) [1,2]. Uveal melanoma, as the most common intraocular malignancy, typically metastasizes through blood [3–5]. Hepatic metastases, which occurs in 95% of patients with metastatic uveal melanoma, results in death in almost all cases [6–12]. This high death rate is related to the recognition of liver metastasis at a very late stage (stage III > 1.8 cm). In some cases, early detection may benefit from current treatments. The

prognosis for patients with uveal melanoma liver metastases is poor, with a median survival of less than 6 months [13–16]. Improvement of early detection by non-invasive MR imaging of liver metastasis has a strong translational potential to allow the use of precision therapy earlier in disease which will likely lead to improved outcomes for patients. Pathologically, human UM liver metastases exhibit nodular, infiltrative or combined patterns that have different responses to treatment [16,17]. We recently reported limitations of clinical MRI contrast agents such as Eovist and MultiHance to define growth patterns by radiology, versus pathology, due to tumor heterogeneity [17]. Currently, there is no suitable non-invasive method for detection of

∗

Corresponding author. Department of Chemistry, Georgia State University, Atlanta, GA, 30303, USA. E-mail address: [email protected] (J.J. Yang). 1 These authors contributed equally to this work. https://doi.org/10.1016/j.biomaterials.2019.119478 Received 17 February 2019; Received in revised form 21 August 2019; Accepted 5 September 2019 Available online 06 September 2019 0142-9612/ © 2019 Elsevier Ltd. All rights reserved.

Biomaterials 224 (2019) 119478

M. Salarian, et al.

ProCA32 with a flexible linker (Fig. 1). The protein was further modified by PEGylation with Methoxy Succinimidyl Carboxymethyl Ester (M-SCM-2000, 2 kDa, JenKem Technology). ProCA32.collagen1 was bacterially expressed in E. coli and purified using previously described procedures [31,35]. Purification and surface modification with PEG were further confirmed by SDS-PAGE Coomassie Brilliant Blue staining, Electrospray ionization mass spectrometry (ESI-MS), and chromatography [35]. ProCA32.collagen1 was loaded with Gd3+ in a 2:1 ratio. The concentration of the protein was measured with UV–Vis spectrometry by monitoring the Tryptophan (Trp) absorbance signal.

early small liver metastatic lesions with desired sensitivity and selectivity. Standard liver function chemistry tests are not sensitive and unable to reveal the location of metastasis [18,19]. Although ultrasound can detect metastasis with a size of 2 mm, it is limited to the surface of the liver and has low resolution [20]. Compared with other imaging techniques such as PET and CT, MRI with contrast agents provides the most sensitive detection of soft tissue including lesions in the liver [21–26]. Differentiation of micrometastasis from heterogeneous tissue background requires contrast agents and imaging methodology with significantly improved sensitivity and specificity as well as improved relaxation properties. Contrast agents based on iron oxide produce negative (dark) T2/T2* signal resulting in images with signal voids with limited accuracy [27,28]. Two Gd3+-based contrast agents approved for liver imaging, Gd-EOB-DTPA (Eovist, US; Primovist, Europe) and Gd-BOPTA (MultiHance) exhibit bright imaging [29]. Along with all other approved Gd3+ MRI contrast agents, they have per Gd3+ relaxivity (r1) values of ~5 mM−1s−1, which is ~20-fold lower than a theoretical value for molecular imaging [30]. Such low relaxivity limits their sensitivity for detecting small liver lesions. In addition, these agents have short half-lives with rapid washout that largely limits their accumulation at small lesions which affects sensitivity and specificity. Furthermore, a high injection dose of 0.1–0.3 mmol/kg is required to generate detectable contrast. These high injection doses contribute to metal toxicity and have resulted in a black box warning related to use of these agents [31,32]. While targeting disease biomarkers has been shown to significantly improve specificity [33], the development of sensitive biomarker-targeted MRI contrast agents for sensitive detection of liver metastasis is extremely challenging. To achieve the long and standing goal of developing a robust, sensitive MRI technique called precision molecular MR Imaging (“pMRI”) capable of monitoring the status of biomarker expression of diseases, it will require significant improvement of relaxation properties and strong binding to liver metastasis biomarkers as well as the capability to reduce metal toxicity. In this paper, we report the development of a collagen I targeted protein-based MRI contrast agent, ProCA32.collagen1 for sensitive and early detection of liver metastasis combined with MRI methodology (Fig. 1) [34].

2.2. Liver tumor mouse models All animal procedures performed in this study complied with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and complied with an approved animal protocol from Institutional Animal Care and Use Committee (IACUC) at Georgia State University and Emory University. To generate hepatic melanoma implantation model, C57BL/ 6 female mice, 10–12-week-old, were anesthetized with i.p. ketamine hydrochloride, 90 mg/kg, and xylazine, 10 mg/kg. After anesthesia, a small horizontal abdominal incision (1 cm) was made in the left upper quadrant such that the left lobe of the liver would be exposed. A total of 2 × 106 of mouse B16LS9 melanoma cells were drawn into a Hamilton syringe with a 30 1/2-gauge needle creating a 5-μL slurry that was injected into the parenchyma of the exposed liver under an operating microscope. Gentle pressure was applied to the hepatic injection site for 1 min with a cotton-tipped applicator after tumor cell injection. The left lobe of the liver was repositioned into the peritoneal cavity, and the abdominal wall was then closed with 5.0 sutures. Human uveal melanoma M20-09-196 cells were inoculated into the supra choroid space of the right eye using a transcorneal technique. For each inoculation, a million cells were delivered in a volume of 2.5 μL. The mice were anesthetized with intraperitoneal injection of ketamine and xylazine mixture. A tunnel was prepared from the limbus to the choroid with a 30 1/2-gauge needle under a surgical microscope. The tip of a 10 μL glass syringe with a 31-gauge/45-degree point metal needle (Hamilton, Reno, NV) was used to introduce cell suspension into the supra choroid space through the needle track. The right eye was enucleated at 2 weeks after tumor cell inoculation [32,33].

2. Materials and methods 2.1. Molecular cloning, expression, purification and lysine PEGylation

2.3. Cell lines ProCA32.collagen1 was designed by engineering a collagen type I targeting peptide (GGGKKWHCYTYFPHHYCVYG) to the C-terminal of

LX-2 (Human Hepatic stellate cell line). LX-2 cells were thawed

Fig. 1. Development of collagen-targeted protein-based MRI contrast agent, ProCA32.collagen1 by linking collagen type I targeting moiety (violet) at C-terminal of ProCA32 with two Gd3+ binding sites using a flexible hinge (orange) and PEGylation. Combination of ProCA32.collagen1 with dual relaxation property and short T1 inversion recovery imaging methodology results in detection of liver metastasis with high expression of collagen as small as 0.144 mm2. 2

Biomaterials 224 (2019) 119478

M. Salarian, et al.

microscope was used to image the cells, and images were analyzed by Zen 2.6.

in DMEM High Glucose (Millipore Cat. No. SLM-021-B), 10% FBS (Millipore Cat. No. ES009-B), 1X Pen/Strep (Millipore Cat. No. TMSAB2-C) and 1X Glutamine (Millipore Cat. No. TMS-002-C) media. Cells were expanded in 2% FBS media using the same components listed above. Cells were pelleted by centrifugation at 300×g for 2–3 min. Human Umbilical Vein Endothelial Cells (HUVEC) (Life Technologies Cat. No. C0035C) were cultured in Medium 200 (Life Technologies Cat no. M200500), with Low Serum Growth Supplement (LSGS) (Cat No. S00310), and for all experiments, 4–6 passage cells were used. Human Hepatic Sinusoidal Endothelial Cells (ScienCell Cat. No. 5000) were cultured in Endothelial Cell Medium (ScienCell Cat. No.1001), and for all experiments, 3–4 passage cells were used. HepG2 cells, purchased from ATCC, were cultured in Eagle's Minimum Essential Medium with 10% FBS. Human uveal melanoma cell lines (Mel 270, Mel290 and M2009-196) and mouse melanoma cell line (B16LS9) were obtained from Dr. Grossniklaus' laboratory in Emory University, and cultured at 37°C in 5% CO2 in complete culture medium (RPMI-1640 with HEPES and L-glutamine (Cellgro, Cat. No. 10-040-CV), 10% fetal bovine serum, 1% nonessential amino acids (Cellgro, Cat. No. 25-025-Cl), 1% sodium pyruvate solution (Cellgro, Cat. No. 25-000-Cl), 1% MEM vitamin solution, 1% antibiotic-antimycotic solution containing 100 U/mL penicillin G, 250 ng/mL amphotericin B, and 100 μg/mL streptomycin solution (Gibco, Grand Island, NY).

2.6. Stability study of targeting moiety of ProCA32.collagen1 using ELISA ELISA assay was used to test the stability of ProCA32.collagen1 targeting moiety in human serum (MilliporeSigma, Cat. No. H4522) following the protocol described below. 15 μL of 2.6 mM ProCA32.collagen1 was incubated with 15 μL of human serum at 4 °C and 37 °C for up to 5 days. 30 μg of collagen type I solution from rat tail (C3867 Sigma-Aldrich) was used to coat the ELISA plate (ELISA MAX™, Biolegend) in 100 μL of coating buffer (0.05 M Carbonate-Bicarbonate, pH 9.6) at 4 °C, overnight. After thoroughly washing with wash solution (50 mM Tris, 0.14 M NaCl, 0.1% Tween 20, pH 8.0) 3 times, at 5 min per wash, 5% BSA (bovine serum albumin, HyClone™) solution was used for blocking at ambient temperature for 2 h. ProCA32.collagen1 samples were diluted to different concentrations (from 0 μM to 65 μM) with 5% BSA and incubated at ambient temperature for 2 h. Following another thorough washing, 0.1% of ProCA32.collagen1 primary antibody (Polyclonal rabbit anti mouse, self-generated) was incubated at 4 °C overnight. Next, 0.1% stabilized peroxidase conjugated goat antirabbit (Cat. No. 31460, ThermoFisher) secondary antibody was used for incubation at room temperature for 1 h. After washing three times with washing buffer, 100 μL of 1-StepTM Ultra TMB-ELISA substrate was added to each well to develop color for approximately 5–10 min 100 μL of 2 M H2SO4 was then added into each well to stop the reaction. Absorbance was recorded by Victor3 plate reader.

2.4. ProCA32.collagen1 cell viability studies Cells were incubated with varying concentrations of Gd3+-loaded ProCA32.collagen1 for 24 h, then the media was discarded from cell cultures by careful aspiration. 200 μL of complete media for respective cells with MTT (Sigma-Aldrich Cat. No. M5655) solution was added into each well and incubated at 37 °C for 3 h. After 3 h the media was carefully aspirated and 100 μL of DMSO was added to solubilize MTT formazan, and absorbance was measured at 590 nm. Cell viability was also tested with flow cytometry. Cells lines including LX-2, raw, and HepG2 were tested, respectively. 1 × 105 of cells were seeded in each of 24 wells; after the cells attached, 200 μL of 500 μM Gd3+-loaded ProCA32.collagen1 were added into the culture medium and incubated for 24 h. Negative control cells were incubated with 500 μM BSA buffer. Up to 1 × 105 were harvested with Accutase™ cell detachment solution (BD biosciences) and washed by PBS 2 times, followed by centrifuging at 300×g for 5 min. Cells were then washed 2 times by flow cytometry staining buffer (FCS buffer, Invitrogen™). 100 ng Propidium Iodide (PI, Invitrogen) was added in 200 μL FCS buffer and incubated with cells for 15 min, while isolating the samples from sources of light. A FACScan flow cytometer (BD LSRFortessa™) was used to determine the PI fluorescence of each sample. 1 × 104 cells were computed in list mode and analyzed by FlowJo.

2.7. Determination of r1 and r2 relaxivity values ProCA32.collagen1 relaxation rates, also known as relaxivities, r1 (R1 = 1/T1) or r2 (R2 = 1/T2), were calculated using the equation below:

ri = (

1 1 − )/[Gd3 +] Tis T1ib

i= 1,2

(1)

where 1/Tis and 1/T1ib reflect the change in relaxation rate before and after generation of the contrast and [Gd3+] is the concentration of Gd3+. ProCA32.collagen1 r1 and r2 values were calculated at four different concentrations of GdCl3 and protein with 2:1 ratio and measurement of T1 and T2 at 37 °C with 1.4 T Bruker Minispec using saturation recovery and CPMG sequence. Using the equation above, the relaxation rates for both T1 and T2 were measured with the slope of the curve being the longitudinal (r1) and transverse (r2) relaxivities.

2.8. Metal binding affinity studies The Gd3+ binding affinity of PEGylated ProCA32.collagen1 was calculated using a competition assay described in previous publications with Tb3+ luminescence resonance energy transfer (LRET) experiment using the equations below [31,36].

2.5. Cell uptake assessment of ProCA32.collagen1 Gd3+-loaded ProCA32.collagen1 was labeled with NHS-Fluorescein (5/6-carboxyfluorescein succinimidyl ester, Thermo Scientific™), and used for testing cell uptake. Fluorescein labeled ProCA32 was used as control. HepG2 and M20-09-196 cells were seeded in cover slides and cultured till confluence reached 50–60%. 5 μM of ProCA32.collagen1 was used to incubate with HepG2 and M20-09-196 for 4 h at 37 °C. Cells were washed with PBS buffer and fixed with 4% paraformaldehyde (in PBS) for 15 min at ambient temperature. After fixation, cells were washed twice again and 0.2% Triton X-100 was used to permeabilize cell for 15 min at room temperature. Following permeabilization, cells were blocked with 1% BSA for 1 h at room temperature and 1:40 dilution of phalloidin (Alexa Fluor™ 633 phalloidin, Thermo Fisher) was used to stain actin, and ProlongTM gold antifade reagent (with DAPI) was used to stain the nucleus. A Zeiss LSM 510 laser scanning confocal

[Tb]Free =

f=

K d Tb, DTPA × [Tb − DTPA] [DTPA] Free

(2)

[Tb]nFree K dn Tb, ProCA + [Tb]nFree

(3) 3+

3+

concentration calIn these equations, [Tb ]free is the free Tb culated from the buffer system, Kd Tb,DTPA is the dissociation constant of Tb3+and DTPA. The dissociation constant of Gd3+ to ProCA32.collagen1 (KdGd, ProCA) was calculated by Equations (4) and (5): 3

Biomaterials 224 (2019) 119478

M. Salarian, et al.

estimating the mean intensity within the ROI, and dividing the mean intensity by the standard deviation (SD) of background. Furthermore, Relative Contrast of tumor was calculated by choosing an ROI within the tumor area and estimating the mean intensity within that ROI divided by the mean intensity of an ROI within the liver area. CNR of tumor was calculated by dividing SNR of tumor by SNR of liver.

([Tb]T + [Gd]T + K d app) f=

−

([Tb]T + [Gd]T + K d app

K d Gd, ProCA = K d app ×

)2

− 4 × [Tb]T × [Gd] T

2 × [Tb]T K d Tb, ProCA K d Tb, ProCA + [Tb]T

(4)

(5)

Calcium and zinc binding affinities of ProCA32.collagen1 were also calculated using previously described methods [31,36].

2.13. Organ distribution by ICP-OES Tissue samples (~0.2 g) from different organs were obtained after euthanasia and digested in 70% (wt/vol) ICP-grade HNO3 at 40 °C overnight. The solution was collected the next day, filtered, and diluted with 2% (wt/vol) HNO3 to 8 mL. The Gd3+ concentration (wavelength of 342.246 nm) was measured in each tissue sample by ICP-OES. As internal standard, 2 ppm YCl3 in 2% HNO3 (Cat. No. 100067-1, High Purity Standards) was used and different Gd3+ standard solutions with different concentrations ranging from 5 ppb to 1000 ppb were used to calculate the final Gd3+ concentration in tissue samples.

2.9. Immunofluorescence staining Immunofluorescence staining was performed on frozen liver tissue sections (5 μm thickness) collected from mice post-injection of ProCA32.collagen1. Tissue fixation was carried out with 4% (vol/vol) formaldehyde for 10 min at room temperature followed by wash step with Tris-buffered saline with Tween 20 (TBST). Bovine serum albumin (BSA) 1% in TBST was used as the blocking reagent, followed by incubation with anti-ProCA32.collagen1 primary antibody (rabbit, selfgenerated) for 1.5 h with 1:100 dilution. Alexa Fluor 594-labeled goat anti-rabbit IgG (Invitrogen; Cat. No. A-11012) secondary antibody was used and then the nuclei of the cells were stained by DAPI (Cat. No. 62248, 1:1000 dilution).

2.14. Transmetallation studies The relaxation rate changes of Gd3+-loaded ProCA32.collagen1 (1:1 ratio, 110 μM) and clinical MRI contrast agents were measured in the presence of 100 μM Zn2+ and 1.2 mM PO43− (phosphate buffer) over time. Then the ratios of R1 at 4620 min and R1 at 0 min (R1 (t=4620)/ R1(t=o) were calculated for difference contrast agents [37].

2.10. Histology analysis For detecting the expression of collagen I and verifying the character of melanoma in the mouse liver, paraffin-embedded liver sections were incubated with rabbit anti-Collagen I antibody (abcam, Cat. No. ab34710) and anti-S100 antibody (abcam, Cat. No. ab14849), and counterstained with hematoxylin. Sirius red and H&E staining were performed on paraffin-embedded liver sections cut at 4 μm. First, the sections were dried, deparaffinized, and hydrated with xylene and different percentages of ethanol. Sirius Red staining was performed using NovaUltra™ Sirius Red Stain Kit (IHC WORLD, IW-3012). For H& E Staining, liver sections were fixed in 10% (vol/vol) formalin and dehydrated with increasing concentrations of alcohol. In the last step, they were washed with xylene, and stained with H&E chemicals.

3. Results and discussion 3.1. Design and characterization of ProCA32.collagen1 The design of this contrast agent is based on several considerations. First, we have shown that type I collagen is an excellent biomarker for imaging liver metastasis. Collagen is upregulated in tissue remodeling but is absent in the normal choroid or present at very low levels. At the early avascular stage, activated hepatic stellate cells (HSCs) play a major role in invasion and growth by deposition of collagen (mainly type I) and extracellular matrix proteins (ECM) [38,39]. The presence of type I collagen in the melanoma tumor stroma reflects active remodeling of extracellular matrix microenvironment by the melanoma cells themselves [40]. Fig. 2A and B shows that collagen type I exhibits different expression levels and structural features depending on the liver metastasis growth patterns. In the infiltrative pattern, collagens surround the tumor, whereas in the nodular pattern, collagens are present both inside and outside of the tumor nodule (Fig. 2A) [16]. Collagen as the important part of tumor microenvironment can also affect regulation of ECM remodeled by collagen degradation and redeposition, as well as promotion of tumor infiltration, angiogenesis, invasion and migration [27]. Second, we have shown that protein contrast agent ProCA32 (PEGylated, non-targeted agent) with two engineered Gd3+ binding sites in Parvalbumin exhibits excellent Gd3+ binding capability and high r1 and r2 relaxivity [31,36]. ProCA32.collagen1 is created by engineering a collagen type I targeting peptide at the C-terminal of ProCA32 with lysine residues in the targeting moiety positioned toward collagen type I in the modeled structure (Fig. 1). A flexible hinge was used to maximize targeting capacity and relaxivity, while maintaining metal binding capability [31]. PEGylation was used to improve biocompatibilities including dose efficiency, and improved relaxivity by tuning correlation time as well as increasing the stability and blood retention time [41]. The designed ProCA32.collagen1 was bacterially-expressed, purified, modified, and formulated. ICP-OES analysis of the complex indicated the formation a 2:1 Gd3+- ProCA32.collagen1 complex [31,36,41]. Fig. 3A summarizes the relaxivity values of ProCA32.collagen1. At 1.4 T, r1 and r2 values of ProCA32.collagen1 were 34 ± 0.12 mM−1s−1 and 50 ± 0.16 mM−1s−1 per Gd3+, respectively at 37 °C. Since

2.11. MRI scan Mice were imaged on a 7-T Agilent MRI scanner at University of Georgia. Anesthetization was performed with isoflurane, and respiration rates were monitored with small animal physiological monitoring system. Respiration rate was maintained and controlled at 65 ± 5 breaths per minute. Fast spin echo-based inversion recovery sequence was acquired with inversion recovery times of 10, 222, 435, 648, 861, 1074, 1287, and 1500 ms, and the best inversion time was selected. Other acquisition parameters included: repetition time of TR = 5000 ms, Effective TE = 32.67 ms, field of view FOV = 35 × 35 mm, matrix = 256 × 256, slice thickness = 1.0 mm, and 12 image slices with no gap. T2-weighted images were collected with acquisition parameters as follows: TR = 2000 ms, Effective TE = 40 ms, field of view, FOV = 35 × 35 mm, matrix = 256 × 256, slice thickness = 1.0 mm, and 12 image slices with no gap. T1-weighted images were collected with acquisition parameters as follows: TR = 500 ms, TE = 14.89 ms, field of view, FOV = 35 × 35 mm, matrix = 256 × 256, slice thickness = 1.0 mm, and 10 image slices with no gap. 2.12. Statistical and image analysis All statistical analyses were performed using one-way ANOVA in GraphPad Prism 5 (GraphPad Software). The signal-to-noise ratio (SNR) of T1 inversion recovery in the liver and tumor was calculated by choosing a region of interest (ROI) within the liver and tumor, 4

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Fig. 2. A. ProCA32.collagen1 has two distribution phases: liver specific and tumor specific phases. In the liver specific phase, liver signal decreases, and in tumor phase, liver signal will be further suppressed and UM micrometastasis growth patterns (Nodular vs Infiltrative) and implanted tumors are detected. B. Histochemistry of two adjacent human liver tissues with UM stained with Sirius red (red, collagen) and ProCA32.collagen1 (light brown) shows the strong binding of the contrast agent to human collagens in the tumor.

ProCA32.collagen1 has two Gd3+ binding sites, the per particle r1 and r2 values were 68 ± 0.25 mM−1s−1 and 100 ± 0.32 mM−1s−1, respectively. These values at 1.4 T were 5–10 times higher for r1 and 12–17 times higher for r2 than clinical contrast agents such as Eovist, MultiHance, and Magnevist (Fig. 3A). Furthermore, at high magnetic field strength of 7.0 T and 37 °C, ProCA32.collagen1 had r1 (21 ± 0.5 mM−1s−1) and r2 (108 ± 1.2 mM−1s−1) values per Gd3+, and 42 ± 1 mM−1s−1 and 216 ± 2.4 mM−1s−1, per particle values, respectively. To our knowledge, these values are greater than simulation based on Solomon–Bloembergen–Morgan theory [30,31,36]. Such high relaxation properties at low and high magnetic field strengths are extremely advantageous for both clinical studies at medically relevant low field strength and preclinical mouse studies at high field due to 1000fold smaller size of mice than human. The addition of the collagen targeting moiety increased the relaxation property compared to ProCA32 without the biomarker targeting moiety (Fig. 3A). Another important criterion for an ideal Gd3+-based contrast agent is its high Gd3+ binding affinity as well as strong metal selectivity over endogenous physiological metals to prevent Gd3+ release that is attributed to nephrogenic systemic fibrosis (NSF) [29,30]. Using a competition assay as previously described [31,35,36], the Gd3+ binding affinity of ProCA32.collagen1 was calculated to be 2.0 ± 0.25 × 10−22 M which is comparable to clinical contrast agents.

Importantly, ProCA32.collagen1 exhibits strong metal selectivity for Gd3+ over Zn2+ and Ca2+ (1.33 ± 0.03 × 10−6 M for Zn2+ and 4.93 ± 0.03 × 10−8 M for Ca2+). This is unprecedented metal selectivity of ProCA32.collagen1 that is 108-1016 times higher towards Ca2+ and Zn2+, respectively, than clinically approved contrast agents including Magnevist [31]. To the best of our knowledge, ProCA32.collagen1 has the highest metal selectivity among all other Gd3+-based contrast agents. We further determined the kinetic stability and metal selectivity of ProCA32.collagen1 to Gd3+ over Zn2+ compared to other clinical contrast agents using a transmetallation assay in phosphate buffer supplemented with ZnCl2 as a function of time [31,37]. Fig. 3B shows that the relaxivity of ProCA32.collgen1 remains unchanged over 4 days, which is similar to results for “safe” clinical contrast agents, ProHance, and Dotarem. This “kinetic stability” is superior to two clinically approved liver contrast agents, Eovist and MultiHance, as well as other approved agents, Magnevist and Omniscan. These approved contrast agents have low metal selectivity for Gd3+ over other endogenous physiological metal ions such as Zn2+. The addition of free Zn2+ is able to compete Gd3+ in the Gd-contrast agent complex to release Gd3+. The released free Gd3+ is precipitated in the phosphate buffer, resulting in the decrease in relaxation rate. The strong metal selectivity of ProCA32.collagen1 is likely due to the increase in the charge number at the coordination shell for trivalent metal ions. 5

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Fig. 3. A. Relaxivity (r1 and r2) of PEGylated ProCA32.collagen at 1.4 T, 37°C compared to clinical contrast agents and ProCA32 (PEGylated, non-targeted agent). B. Relaxation rates changes of ProCA32.collagen1 compared with clinical contrast agents all loaded with Gd3+ in phosphate buffer in the presence of ZnCl2 after approximately 4 days. C, D, E. ELISA experiment showing that the ProCA32.collagen1 targeting moiety remained stable when incubated with human serum in a collagen I coated plate. ProCA32.collagen1 incubated with serum for up to 5 days at both 4° and 37 °C showed comparable collagen I targeting with positive control (PC) (ProCA32.collagen1 incubated with serum for 0 h and ProCA32.collagen1 by itself) due to absorbance caused by binding to collagen I. However, due to lack of collagen I targeting moiety, ProCA32 incubation group as the negative control (NC) did not show collagen I targeting and had significantly lower absorbance when compared with ProCA32.collagen1 (one-way ANOVA, *** P < 0.0001).

significant differences in cell viability were observed between normal (without treatment), BSA and protein treated groups. To conclude, both ProCA32.collagen1 and PEGylated ProCA32.collagen1 did not affect the viability of any of the three cell lines under the theoretical maximum dosage. More cell lines, including Mel 270, M20-09-196, Mel 290, HUVEC, and LSEC, were also tested using MTT assay showing similar results (Fig. S6), supporting the conclusion that ProCA32.collagen1 does not exhibit cellular toxicity. We have also examined the uptake of ProCA32.collagen1 by LX-2, HepG2 and M20-09-196 cell lines using confocal imaging with NHSFluorescein labeled amino group of ProCA32.collagen1. Non-targeted contrast agent, ProCA32 was also used as a control. Similar to ProCA32, no significant uptake was observed for ProCA32.collagen1 for either cell lines (Fig. 5A). Fig. 5B shows that incubation of 50 μM of ProCA32.collagen1 with HepG2 and LX-2 cells in reduced serum conditions at various time points did not result in any reduction of ProCA32.collagen1 stained by Ponceau red, which is consistent with the absorbance measurement shown in Fig. 5C. On the other hand, no ProCA32.collagen1 was detected in cell lysate using both Ponceau red staining or by western blotting. These studies indicated that ProCA32.collagean1 is located extracellularly without endocytosis.

In addition, ProCA32.collagen1 shows relatively high affinity to collagen type I with dissociation constant of Kd = 1.42 ± 0.2 μM [35] in a 1:1 binding model [42,43]. In contrast, no collagen binding was detected for ProCA32 without a collagen targeting moiety. In addition, ProCA32.collagen1 targeting moiety is stable up to 5 days in human serum at 37 °C without detachment as determined using ELISA assay degradation (Fig. 3). 3.2. Stability of ProCA32.collagen1 targeting moiety The stability of the protein and collagen type I targeting moiety on ProCA32.collagen1 was evaluated in the presence of human serum with protease activity. ELISA was used to detect the binding of collagen I precoated to the plate by ProCA32.collagen1 pre-incubated in serum for 5 days at either 4 °C or 37 °C. ProCA32.collagen 1 without incubation was used as a positive control while ProCA32 without targeting moiety was used as a negative control. The ProCA32.collagen1 protein incubated with serum for 5 days at either 4 °C or 37 °C showed comparable collagen I targeting without incubation (Fig. 3C, D, E). However, due to lack of collagen I targeting moiety, ProCA32 incubation in serum did not show collagen binding and had significantly lower absorbance compared to ProCA32.collagen1. These results indicated that ProCA32.collagen1 maintained its targeting capability against protease degradation in the presence of human serum.

3.4. MR imaging of liver implanted tumor Next, we demonstrate the development and application of multiple imaging methodologies with improved in vivo sensitivity and specificity taking advantage of the unique high dual relaxation properties of ProCA32.collagen1 using liver implanted model by injection of mouse melanoma (B16LS9) tumor cells in liver lobes. Fig. 6A illustrates that an MRI enhancement of tumor at 7T can be readily detected using T1-weighted, T2-weighted, and short T1 inversion recovery with long echo time (TE) following intravenous injection of the contrast agent, ProCA32.collagen1 (100 μL, 5 mM). This implanted tumor was verified by histological analysis of H&E staining (Fig. 6C). Injection of ProCA32.collagen1 resulted in approximately 2fold enhancement in the relative contrast of the tumor using T1-and T2wighted images, respectively (Fig. 6B). A short T1 inversion recovery with long TE pulse sequence was developed to suppress liver signal using both T1 and T2 properties of ProCA32.collagen1 to achieve a high contrast between liver and tumor. Indeed, short T1 inversion recovery

3.3. ProCA32.collagen1 cell viability and uptake To evaluate the effect of ProCA32.collagen1 on cell growth, propidium iodide and flow cytometry as well as MTT assays were used to analyze the cell viability of three major cell contents in the liver: hepatic stellate cells, macrophages and hepatocytes after incubation with ProCA32.collagen1 for 24 h. Considering the concentration of ProCA32.collagen1 after injection in mouse blood (5 mM of 100 μL through tail vein in 1 mL of mouse blood), 500 μM of the protein (ProCA32.collagen1 and PEGylated ProCA32.collagen1) was used in the study and tested on LX-2, RAW 264.7 and HepG2 cells lines. In addition, 500 μM BSA was used as negative control, and H2O2 was used as positive control. At 24 h post treatment, both ProCA32.collagen1 and PEGylated ProCA32.collagen1 did not affect the viability of any of the three cell lines (Fig. 4). Based on one-way ANOVA analysis, no 6

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Fig. 4. A. Cell viability of three different cell lines including LX-2, and RAW 264.7 and HepG2 were tested using propidium iodide and flow cytometry after incubating with 500 μM of both ProCA32.collagen1 and PEGylated ProCA32.collagen1. Based on propidium iodide histogram of LX-2, RAW 264.7, and HepG2 after different treatment for 24 h, no significant difference of PI staining was observed after treated with BSA, ProCA32.collagen1 and PEGylated ProCA32.collagen1. Both ProCA32.collagen1 and PEGylated ProCA32.collagen1 had no effect on cell viability. B. Statistical analysis of cell viability between different treatment groups. There was no significant difference in cell viability between normal cells without treatment, BSA, ProCA32.collagen1 and PEGylated ProCA32.collagen1 treated cells. They all showed significant differences when compared with H2O2 treated groups (*** P < 0.0001).

before. The observed heterogeneity of detected liver metastasis (Fig. 7A) and differences in their SNR (Fig. 7E) are likely due to their differential tumor microenvironment and growth patterns. Consistent with MRI and histological analysis, ICP-OES analysis of Gd3+ content in different organs revealed that the Gd3+ injection dosage of ProCA32.collagen1 in liver with uveal melanoma tumors (25%) was higher than that of ProCA32 (< 15%) (Fig. S5A). Immunofluorescence imaging further confirmed the accumulation of the contrast agent (red) in liver (Fig. S1). In addition, further analysis of liver resulted in detection of several UM micrometastasis ranging from 0.144 mm2 to 0.420 mm2 (Fig. 8A, red circle). Fig. 8A shows the detection of tumors in T1-weighted and T2-weighted images post-injection of ProCA32.collagen1 compared to tumors detected in T1 inversion recovery with long TE images. Analysis of CNR and SNR exhibited the higher sensitivity of T1 inversion recovery with long TE sequence compared to T1-and T2-weighted images (Fig. 8B and C). Histology correlation of tumors detected under MRI are shown in Fig. S4. ProCA32.collagen1 also had a high r2 value that enabled the application of several imaging methodologies which reduce the possibility of detecting artifacts (Fig. 8). Because of the high r1 and r2 values, a short T1 inversion recovery with long TE of 32.67 ms was used to increase the sensitivity and achieve a high contrast between liver and tumor. In this pulse sequence, the first step is liver signal suppression using appropriate inversion times with short liver T1 because of high r1 of contrast agent. In the second step, further liver signal suppression happens with long TE due to short T2 of liver becuase of high r2 value of

with long TE showed the highest increase in relative contrast (more than 6-fold) post-injection of ProCA32.collagen1 (Fig. 6B). In addition to significantly improved sensitivity, these imaging sequences provide multiple ways to eliminate MRI artifacts associated with heterogeneous liver background (i.e. false positives) and increase accuracy. 3.5. MR imaging of liver hepatic metastasis We then examined the in vivo non-invasive detection capability of liver micrometastasis of ProCA32.collagen1 using our developed mice liver metastasis model by inoculation of aggressive human uveal melanoma M20-09-196 cells in the eye [44]. As seen in Fig. 7A, six small liver micrometastasis were detected by T1 inversion recovery. In contrast, both Eovist and ProCA32 (5 mM, 100 μL) failed to detect any tumors. Histology analysis using S100 (red) and collagen I (brown) staining confirmed their identity of liver UM metastases and overexpression of collagen I (Fig. 7C). Both H&E and its corresponding Sirius Red staining of these UM tumors reveal the overexpression of collagen (red) in all detected liver lesions (Fig. 7B). The tumors detected by MRI were strongly correlated with H&E staining in terms of sizes and shapes (Fig. 7D). Detailed histological data, verified by a pathologist (HEG) revealed two liver growth patterns resembling those reported in human patient samples [16] and ranging from 0.250 mm2 to 595 mm2 (Fig. 7D). Due to double suppression of liver signal using T1 and T2 properties and specific collagen binding of ProCA32.collagen1 to the tumor, we were able to detect liver micrometastases with SNR of L1-L6 3-6-fold higher than liver SNR which has not been reported 7

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Fig. 5. A. Immunofluorescence imaging of HepG2 and M20-09-196 cells to study the uptake of ProCA32.collagen1. HepG2 and M20-09-196 cells were used to study the cellular location and uptake of ProCA32.collagen1. ProCA32.collagen1 (upper row) and ProCA32 (lower row) (5 μM) were labeled by NHS-fluorescein and incubated with HepG2 and M20-09-196 for 4 h under 37 °C. HepG2 cells incubated with fluorescein labeled ProCA32.collagen1 showed sparse green fluorescence in the cell, while other cells did not exhibit protein uptake. Magenta color represents actin staining, and blue color is nucleus staining. B. Incubation of 50 μM of ProCA32.collagen1 with HepG2 and LX-2 cells for monitoring the uptake in reduced serum conditions at various time points. Ponceau and western blotting was used to measure uptake of ProCA32.collagen1. C. The absorbance of ProCA32.collagen1 was measured in culture medium for uptake upon incubation with HepG2 cells.

ProCA32.collagen1 (Fig. 1). The effect of short T1 inversion recovery with long TE methodology could also be observed in detecting tumor and collagen heterogeneity as it shows higher sensitivity compared to other imaging pulse sequences based on higher CNR of tumor over liver (Figs. 7A and 8C).

days post-injection, including liver, spleen, heart, and brain, which is likely due to its relatively large size (Fig. S5B) [35]. Further analysis of MRI results suggested that the majority of ProCA32.collagen1 was cleared through kidney 48 h post-injection, as the SNR value was the highest among different time points (Fig. S5C).

3.6. Biodistribution and clinical chemistry tests of ProCA32.collagen1

4. Conclusion

The long-term (14 day post-injection) distribution of Gd3+ after injection of ProCA32.collagen1 in mice, at a dosage of 0.02 mmol/kg, showed the contrast agent has low accumulations in major organs other than liver, and specifically no accumulation in the brain compared to Eovist. In addition, ProCA32.collagen1 did not show any toxicity related to clinical chemistry tests as the values were within a normal range post-injection (Table S1) [35]. Furthermore, ProCA32.collagen1 did not show any significant Gd3+ accumulation in different organs 14

In summary, we have developed a collagen targeted MRI contrast agent, ProCA32.collagen1 with high dual relaxivity values for r1 and r2 at both 1.4 and 7.0 T as well as strong collagen I targeting capability. We further demonstrated that multiple MRI sequences especially short T1 inversion recovery methodology with long TE, taking advantage of dual relaxivity and collagen binding, are able to significantly improve sensitivity, limit of detection and SNR for in vivo detection of tumor overcoming major limitation of heterogeneous tissue backgrounds. Fig. 6. A. T1-weighted, T2-weighted and T1 inversion recovery images of B16LS9 mouse melanoma tumor (shown in red circle) before and 24 h postinjection of ProCA32.collagen1. B. Quantitative analysis of relative contrast demonstrates an enhancement post-injection of ProCA32.collagen1. C. H&E staining and image of implanted tumor in the liver.

8

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Fig. 7. A. MR imaging of liver before and post-injection of ProCA32.collagen1, ProCA32, and Eovist. Only ProCA32.collagen1 was capable of detecting small UM lesions in the liver 24 h post-injection. B. H&E staining of six UM metastasis tumors (1–6) and its corresponding Sirius red staining which shows the expression of collagen (red) and their corresponding locations in MRI. Tumor 2 exhibited an infiltrative pattern (green circle), while the other tumors exhibited a nodular pattern. C. Immunohistochemistry of tumor in the liver shows collagen type I surrounding the tumor (brown). S100 antibody staining (red) confirms the presence of UM metastasis. D. Sizes of the six tumors detected in MRI with histology demonstrated a linear correlation. E. Quantitative analysis of the SNR of six tumors detected in MRI T1 inversion recovery compared to liver at 24 h (L1: Lesion 1).

ProCA32.collgen1 enables the first robust early detection of UM micrometastasis to liver with both nodular and infiltrative UM tumor patterns at stage III as small as 0.144 mm2, which has not been achieved before. Such information is critical for devising treatment strategy since different growth patterns of liver metastasis were reported to have different responses to the treatment [16,17]. We have shown that ProCA32.collagen1 has reduced Gd3+ toxicity and no brain deposition due to its unprecedented metal selectivity and kinetic stability and low injection dose because of higher relaxivity. Furthermore, ProCA32.collagen1 has been shown to have strong human translational potential and application because of its high binding affinity and specificity to collagen in human UM liver metastases. The development of collagentargeted protein MRI contrast agent is expected to overcome the major

clinical barriers in early diagnosis, noninvasive detection and staging of UM micrometastasis to liver, and have strong application in facilitating effective treatment toward major clinical consequences. Although further work must define the diagnostic sensitivity and specificity and safety with additional animal and patient samples, these data set the stage for the new avenues for non-invasive detection, tumor stage, and define tumor subtypes of liver metastasis for precision medicine.

Data availability The raw and processed data required to reproduce these findings will be available upon request. Fig. 8. A. MR imaging of liver before and post-injection of ProCA32.collagen1 using multiple imaging pulse sequences (T1-weighted, T2-weighted, and T1 inversion recovery). B. Comparison of tumor and liver signal to noise ratio (SNR) post-injection of ProCA32.collagen1, in different pulse sequences. C. Contrast to noise ratio (CNR) of tumor over liver demonstrates the higher sensitivity of T1 inversion recovery for detection of tumor metastasis compared to T1-weighted and T2-weighted images.

9

Biomaterials 224 (2019) 119478

M. Salarian, et al.

Acknowledgements

[21] K.V. Sharma, J.E. Gould, J.W. Harbour, G.P. Linette, T.K. Pilgram, P.N. Dayani, D.B. Brown, Hepatic arterial chemoembolization for management of metastatic melanoma, AJR Am. J. Roentgenol. 190 (1) (2008) 99–104. [22] L. Pasquini, A. Napolitano, E. Visconti, D. Longo, A. Romano, P. Toma, M.C.R. Espagnet, Gadolinium-based contrast agent-related toxicities, CNS Drugs 32 (3) (2018) 229–240. [23] P. Caravan, J.J. Ellison, T.J. McMurry, R.B. Lauffer, Gadolinium(III) chelates as MRI contrast agents: structure, dynamics, and applications, Chem. Rev. 99 (9) (1999) 2293–2352. [24] Z. Zhou, M. Qutaish, Z. Han, R.M. Schur, Y. Liu, D.L. Wilson, Z.R. Lu, MRI detection of breast cancer micrometastases with a fibronectin-targeting contrast agent, Nat. Commun. 6 (2015) 7984. [25] D.F. Quail, J.A. Joyce, Microenvironmental regulation of tumor progression and metastasis, Nat. Med. 19 (11) (2013) 1423–1437. [26] N.E. Reticker-Flynn, D.F. Malta, M.M. Winslow, J.M. Lamar, M.J. Xu, G.H. Underhill, R.O. Hynes, T.E. Jacks, S.N. Bhatia, A combinatorial extracellular matrix platform identifies cell-extracellular matrix interactions that correlate with metastasis, Nat. Commun. 3 (2012) 1122. [27] M. Fang, J. Yuan, C. Peng, Y. Li, Collagen as a double-edged sword in tumor progression, Tumour. Biol. 35 (4) (2014) 2871–2882. [28] J. Choi, H. Park, T. Kim, Y. Jeong, M.H. Oh, T. Hyeon, A.A. Gilad, K.H. Lee, Engineered collagen hydrogels for the sustained release of biomolecules and imaging agents: promoting the growth of human gingival cells, Int. J. Nanomed. 9 (2014) 5189–5201. [29] S. Aime, P. Caravan, Biodistribution of gadolinium-based contrast agents, including gadolinium deposition, J. Magn. Reson. Imaging 30 (6) (2009) 1259–1267. [30] P. Caravan, Strategies for increasing the sensitivity of gadolinium based MRI contrast agents, Chem. Soc. Rev. 35 (6) (2006) 512–523. [31] S. Xue, H. Yang, J. Qiao, F. Pu, J. Jiang, K. Hubbard, K. Hekmatyar, J. Langley, M. Salarian, R.C. Long, R.G. Bryant, X.P. Hu, H.E. Grossniklaus, Z.R. Liu, J.J. Yang, Protein MRI contrast agent with unprecedented metal selectivity and sensitivity for liver cancer imaging, Proc. Natl. Acad. Sci. U. S. A. 112 (21) (2015) 6607–6612. [32] S. Dithmar, D. Rusciano, H.E. Grossniklaus, A new technique for implantation of tissue culture melanoma cells in a murine model of metastatic ocular melanoma, Melanoma Res. 10 (1) (2000) 2–8. [33] H. Yang, S. Dithmar, H.E. Grossniklaus, Interferon alpha 2b decreases hepatic micrometastasis in a murine model of ocular melanoma by activation of intrinsic hepatic natural killer cells, Investig. Ophthalmol. Vis. Sci. 45 (7) (2004) 2056–2064. [34] B. Sana, C.L. Poh, S. Lim, A manganese-ferritin nanocomposite as an ultrasensitive T2 contrast agent, Chem. Commun. 48 (6) (2012) 862–864. [35] M. Salarian, R.C. Turaga, S. Xue, M. Nezafati, K. Hekmatyar, J. Qiao, Y. Zhang, S. Tan, O.Y. Ibhagui, Y. Hai, J. Li, R. Mukkavilli, M. Sharma, P. Mittal, X. Min, S. Keilholz, L. Yu, G. Qin, A.B. Farris, Z. Liu, J.J. Yang, Early detection and staging of chronic liver diseases with a protein MRI contrast agent, Nat. Commun. (2019). [36] F. Pu, M. Salarian, S. Xue, J. Qiao, J. Feng, S. Tan, A. Patel, X. Li, K. Mamouni, K. Hekmatyar, J. Zou, D. Wu, J.J. Yang, Prostate-specific membrane antigen targeted protein contrast agents for molecular imaging of prostate cancer by MRI, Nanoscale 8 (25) (2016) 12668–12682. [37] S. Laurent, L. Vander Elst, C. Henoumont, R.N. Muller, How to measure the transmetallation of a gadolinium complex, Contrast Media Mol. Imaging 5 (6) (2010) 305–308. [38] H.E. Grossniklaus, Progression of ocular melanoma metastasis to the liver: the 2012 Zimmerman lecture, JAMA Ophthalmol. 131 (4) (2013) 462–469. [39] H. Yang, H.E. Grossniklaus, Combined immunologic and anti-angiogenic therapy reduces hepatic micrometastases in a murine ocular melanoma model, Curr. Eye Res. 31 (6) (2006) 557–562. [40] K.J. Daniels, H.C. Boldt, J.A. Martin, L.M. Gardner, M. Meyer, R. Folberg, Expression of type VI collagen in uveal melanoma: its role in pattern formation and tumor progression, Lab. Investig. 75 (1) (1996) 55–66. [41] S. Li, J. Jiang, J. Zou, J. Qiao, S. Xue, L. Wei, R. Long, L. Wang, A. Castiblanco, N. White, J. Ngo, H. Mao, Z.R. Liu, J.J. Yang, PEGylation of protein-based MRI contrast agents improves relaxivities and biocompatibilities, J. Inorg. Biochem. 107 (1) (2012) 111–118. [42] S. Federico, B.F. Pierce, S. Piluso, C. Wischke, A. Lendlein, A.T. Neffe, Design of decorin-based peptides that bind to collagen I and their potential as adhesion moieties in biomaterials, Angew Chem. Int. Ed. Engl. 54 (37) (2015) 10980–10984. [43] R. Tenni, M. Viola, F. Welser, P. Sini, C. Giudici, A. Rossi, M.E. Tira, Interaction of decorin with CNBr peptides from collagens I and II. Evidence for multiple binding sites and essential lysyl residues in collagen, Eur. J. Biochem. 269 (5) (2002) 1428–1437. [44] H. Yang, J. Cao, H.E. Grossniklaus, Uveal melanoma metastasis models, Ocul. Oncol. Pathol. 1 (3) (2015) 151–160.

We thank Dr. David H. Lawson, Zhi-Ren Liu, and Alton Brad Farris for their discussion and advice. Dr. Michael Kirberger for carefully editing the manuscript. Dr. Khan Hekmatyar for his help in MRI data acquisition. This work was supported in part by a Molecular Basis of Disease fellowship (to M. Salarian) and National Institute of Health (NIH) Research Grants CA183376 and AA025863 (to J.J.Y.). Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.biomaterials.2019.119478. References [1] B.L. Eckhardt, P.A. Francis, B.S. Parker, R.L. Anderson, Strategies for the discovery and development of therapies for metastatic breast cancer, Nature reviews, Drug Discover. 11 (6) (2012) 479–497. [2] G.G. Van den Eynden, A.W. Majeed, M. Illemann, P.B. Vermeulen, N.C. Bird, G. Hoyer-Hansen, R.L. Eefsen, A.R. Reynolds, P. Brodt, The multifaceted role of the microenvironment in liver metastasis: biology and clinical implications, Cancer Res. 73 (7) (2013) 2031–2043. [3] A.D. Singh, M.E. Turell, A.K. Topham, Uveal melanoma: trends in incidence, treatment, and survival, Ophthalmology 118 (9) (2011) 1881–1885. [4] C.C. McLaughlin, X.C. Wu, A. Jemal, H.J. Martin, L.M. Roche, V.W. Chen, Incidence of noncutaneous melanomas in the, U.S, Cancer 103 (5) (2005) 1000–1007. [5] F. Spagnolo, G. Caltabiano, P. Queirolo, Uveal melanoma, Cancer Treat Rev. 38 (5) (2012) 549–553. [6] Y. Chen, Y. Zhou, X. Lin, H.C. Wong, Q. Xu, J. Jiang, S. Wang, M.M. Lurtz, C.F. Louis, R.D. Veenstra, J.J. Yang, Molecular interaction and functional regulation of connexin50 gap junctions by calmodulin, Biochem. J. 435 (3) (2011) 711–722. [7] Y. Huang, Y. Zhou, H.C. Wong, A. Castiblanco, Y. Chen, E.M. Brown, J.J. Yang, Calmodulin regulates Ca2+-sensing receptor-mediated Ca2+ signaling and its cell surface expression, J. Biol. Chem. 285 (46) (2010) 35919–35931. [8] J. Jiang, Y. Zhou, J. Zou, Y. Chen, P. Patel, J.J. Yang, E.M. Balog, Site-specific modification of calmodulin Ca(2)(+) affinity tunes the skeletal muscle ryanodine receptor activation profile, Biochem. J. 432 (1) (2010) 89–99. [9] B. Mukherji, Immunology of melanoma, Clin. Dermatol. 31 (2) (2013) 156–165. [10] U.P. Hegde, N. Chakraborty, B. Mukherji, J.M. Grant Kels, Metastatic melanoma in the older patient: immunologic insights and treatment outcomes, Expert Rev. Pharmacoecon. Outcomes Res. 11 (2) (2011) 185–193. [11] T.H. Beckham, S. Elojeimy, J.C. Cheng, L.S. Turner, S.R. Hoffman, J.S. Norris, X. Liu, Targeting sphingolipid metabolism in head and neck cancer: rational therapeutic potentials, Expert Opin. Ther. Targets 14 (5) (2010) 529–539. [12] J.L. Isaac-Renton, Laboratory diagnosis of giardiasis, Clin. Lab. Med. 11 (4) (1991) 811–827. [13] A.D. Singh, E.C. Borden, Metastatic uveal melanoma, Ophthalmol Clin North Am 18 (1) (2005) 143–150 (ix). [14] K.M. Halenda, R.R. Kudchadkar, D.H. Lawson, D.D. Kies, K.E. Zhelnin, A.M. Krasinskas, H.E. Grossniklaus, Reduction of nodular growth pattern of metastatic uveal melanoma after radioembolization of hepatic metastases, Ocul. Oncol. Pathol. 2 (3) (2016) 160–165. [15] J.J. Augsburger, Z.M. Correa, A.H. Shaikh, Effectiveness of treatments for metastatic uveal melanoma, Am. J. Ophthalmol. 148 (1) (2009) 119–127. [16] H.E. Grossniklaus, Q. Zhang, S. You, C. McCarthy, S. Heegaard, S.E. Coupland, Metastatic ocular melanoma to the liver exhibits infiltrative and nodular growth patterns, Hum. Pathol. 57 (2016) 165–175. [17] A. Liao, P. Mittal, D.H. Lawson, J.J. Yang, E. Szalai, H.E. Grossniklaus, Radiologic and histopathologic correlation of different growth patterns of metastatic uveal melanoma to the liver, Ophthalmology 125 (4) (2018) 597–605. [18] T.J. Vogl, S. Kummel, R. Hammerstingl, M. Schellenbeck, G. Schumacher, T. Balzer, W. Schwarz, P.K. Muller, W.O. Bechstein, M.G. Mack, O. Sollner, R. Felix, Liver tumors: comparison of MR imaging with Gd-EOB-DTPA and Gd-DTPA, Radiology 200 (1) (1996) 59–67. [19] K. Hanaoka, A.J. Lubag, A. Castillo-Muzquiz, T. Kodadek, A.D. Sherry, The detection limit of a Gd3+-based T1 agent is substantially reduced when targeted to a protein microdomain, Magn. Reson. Imaging 26 (5) (2008) 608–617. [20] J.F. Gerstenmaier, R.N. Gibson, Ultrasound in chronic liver disease, Insight. Imag. 5 (4) (2014) 441–455.

10