Imaging Glycosylation In Vivo by Metabolic Labeling and Magnetic Resonance Imaging

Abstract Glycosylation is a ubiquitous post‐translational modification, present in over 50 % of the proteins in the human genome,1 with important roles in cell–cell communication and migration. Interest in glycome profiling has increased with the realization that glycans can be used as biomarkers of many diseases,2 including cancer.3 We report here the first tomographic imaging of glycosylated tissues in live mice by using metabolic labeling and a gadolinium‐based bioorthogonal MRI probe. Significant N‐azidoacetylgalactosamine dependent T 1 contrast was observed in vivo two hours after probe administration. Tumor, kidney, and liver showed significant contrast, and several other tissues, including the pancreas, spleen, heart, and intestines, showed a very high contrast (>10‐fold). This approach has the potential to enable the rapid and non‐invasive magnetic resonance imaging of glycosylated tissues in vivo in preclinical models of disease.

Abstract: Glycosylation is au biquitous post-translational modification, present in over 50 %o ft he proteins in the human genome, [1] with important roles in cell-cell communication and migration. Interest in glycome profiling has increased with the realization that glycans can be used as biomarkers of many diseases, [2] including cancer. [3] We report here the first tomographic imaging of glycosylated tissues in live mice by using metabolic labeling and agadolinium-based bioorthogonal MRI probe.S ignificant N-azidoacetylgalactosamine dependent T 1 contrast was observed in vivo two hours after probe administration. Tumor,k idney,a nd liver showed significant contrast, and several other tissues,i ncluding the pancreas,s pleen, heart, and intestines,s howed av ery high contrast (> 10-fold). This approach has the potential to enable the rapid and non-invasive magnetic resonance imaging of glycosylated tissues in vivo in preclinical models of disease.
Variousprobeshavebeenreportedforimagingglycosylation in vivo,including antibodies, [4] peptides, [5] boronic acid derivatives, [6] and lectins. [7] Most approaches give as tatic view of cell-surface glycosylation. We,a nd others,h ave used metabolic glycan labeling [8] in combination with bioorthogonal reactions [9] to image the dynamics of glycan biosynthesis in vivo.W ed emonstrated non-invasive imaging of tumor glycosylation in live mice [10] by metabolic labeling of tumor glycans with azido sugars followed by Staudinger ligation with ab iotinylated phosphine and subsequent imaging using fluorescent or radionuclide-labeled avidins.M ore recently, we described "double-click" reagents, [11] where azido-modified cell-surface glycoproteins were detected using abivalent double-click bioorthogonal probe.T he latter consisted of as trained tetramethoxydibenzocyclooctyne (TMDIBO), [12] which reacted specifically with azido sugar labeled glycans, [9] and a trans-cyclooctene( TCO), which reacted very rapidly with af luorescently labeled tetrazine for fluorescence imaging (FLI). FLI gives high sensitivity and throughput; [13] however, al imitation is light absorption and scattering, which prevents deep imaging in opaque organisms such as mice.
Recently,c ells metabolically labeled with N-azidoacetylmannosamine were imaged by magnetic resonance imaging (MRI) and axenon ( 129 Xe) biosensor.The azido group in cellsurface sialic acid residues was detected using ab ifunctional reagent incorporating bicyclo[6.1.0]nonyne,w hich reacted with the azido group,a nd ac ryptophane,w hich bound hyperpolarized 129 Xe.Bound xenon was detected by magnetization transfer measurements between free and bound xenon (hyper-CEST). [14] Although hyperpolarized 129 Xe is very sensitive to MR detection, the change in signal intensity was relatively small (ca. 30-50 %), and only demonstrated for encapsulated cells in ab ioreactor.ACEST-based label-free method for imaging underglycosylated mucin-1 expression in vivo has also been described recently. [15] Our aim was to develop aprobe for the tomographic,noninvasive MR imaging of metabolically labeled glycans in mice. Apreviously reported MRI probe,consisting of aphosphine conjugated to ag adolinium chelate, [11] gave no detectable azido sugar dependent contrast in vivo owing to high levels of non-specific binding,which we attributed to its hydrophobicity.Wedescribe here TMDIBO-Lys-Gd (2; Figure 1), anovel water-soluble probe that combines as trained cyclooctyne TMDIBO linked, via ahydrophilic lysine linker,t oagadolinium DOTAc helate,aclinically approved MRI contrast agent. [16] This probe was used to image metabolically labeled cell-surface glycans on tumor cells in vitro and in vivo.T he probe also showed significant labeling of other mouse tissues, including the pancreas,spleen, kidney,l iver,and gut.
The T 1 relaxivity of 2 ( Figure 1) in buffer at 7Twas 6.3 AE 0.1 mm À1 s À1 ,s imilar to published values for Gd DOTA complexes. [16] Ther eactivity of 2 with cell-surface azido sugar labeled glycans was determined by culturing Lewis lung (LL2) adenocarcinoma cells with N-azidoacetylgalactosamine (Ac 4 GalNAz, 1; Figure 1) for 24 h, and then incubating them with 2 for 45 min, after which the cells were washed. The R 1 (1/T 1 )relaxation rates were measured in pelleted cells that had been incubated with (+ / + /À)orwithout (+ /À/À) 1 and/ or 2 (+ / + / + , + /À/ + ;F igure 2). There was as ignificant increase (P < 0.005) in R 1 in azido sugar (1)treated LL2 cells that had been incubated with 2 (+ / + / + ,1 .38 AE 0.10 s À1 ), when compared with cells not cultured with the azido sugar (+ /À/ + ,0 .79 AE 0.02 s À1 )o rw ith cells not incubated with either 1 or 2 (+ /À/À,0 .64 AE 0.03 s À1 , P < 0.001). Thes mall increase in R 1 (by af actor of 1.23 AE 0.01) between cells that had not been treated with either 1 or 2 (+ /À/À)a nd cells incubated with 2 alone (+ /À/ +)showed that there were only low levels of non-specific binding of 2.T his was considerably less than observed previously with af luorescently labeled version of TMDIBO,w here this ratio was 3.3 AE 0.1. [11] Mice with flank tumors,o btained by subcutaneous injection of LL2 cells,w ere injected daily,f or three days, with Ac 4 GalNAz (1;300 mg kg À1 ,i.p.) or with solvent vehicle, and then injected with 2 (0.25 mmol kg À1 ,i .v.; gadoliniumbased contrast media are used clinically at 0.1-0.3 mmol kg À1 ) on day 4. [17] Them etabolic labeling of glycans with 1 and subsequent bioorthogonal detection with 2 was confirmed by ICP-MS measurements of the gadolinium content in excised tissues ( Figure 3) obtained after the imaging experiments (Figures 4a nd 5), and 24 ha fter the injection of 2.M ost tissues showed N-Ac 4 GalNAz dependent labeling.T he gadolinium content was highest in the kidney (29.9 AE 5.8 nmol Gd per gram of tissue). However,about one third of this was due to the non-specific retention of 2 and thus not N-Ac 4 GalNAz dependent (10.9 AE 2.2 nmol Gd per gram of tissue), and metabolic labeling was relatively modest (2.7 AE 0.5-fold increase in gadolinium content in animals injected with 1 and 2 (+ / +)c ompared with those injected with 2 alone (À/ +)). Thehigh background signal in the kidney is likely due to this organ being the preferred clearance route for molecules < 1kDa. Thel iver also showed relatively high levels of non-specific retention of 2 (4.5 AE 0.8 nmol Gd per gram of tissue) and similar levels of N-Ac 4 GalNAz dependent labeling as the kidney (2.3 AE 0.3-fold). Thel evels of kidney and liver labeling are in agreement with our previous work on FLI. [11] Other tissues showed very low levels of non-specific The R 1 rates were also measured in the buffer in which the cells had been suspended (À/À/ À)a nd in ab uffer to which 2 had been added (1.0 mm; À/À/ +). Data represent the mean AE standard error of the mean (SEM) (n = 3). **P < 0.01, ***P < 0.005, ****P < 0.001. Two-tailed unpaired T-test with Mann-Whitney correction.  Figure 1). The Gd content was normalizedt othe wet tissue weight. Data represent the mean AE SEM (n = 5). *P < 0.05, **P < 0.01, ***P < 0.005. Two-tailed unpaired T-test with Mann-Whitney correction.

Angewandte Chemie
Communications background retention of 2 and consequently high levels of N-Ac 4 GalNAz dependent labeling. Theg adolinium concentration ratios for tissues from animals injected with N-Ac 4 GalNAz (+ / +)relative to those injected with the solvent vehicle (À/ +)w ere 57 AE 8f or the heart, 39 AE 4f or spleen, 14 AE 2f or the pancreas and the small and large intestine,and 5 AE 1for the lungs.T umors showed much higher levels of nonspecific retention of 2,and therefore,the gadolinium concentration ratio was much lower (3.3 AE 1.1). Glycan-labeling methods based on the delivery of labeled sugars using targeted liposomes [18] could be used to improve tumor selectivity.
Serum showed significant metabolic labeling (the serum gadolinium concentration ratio for animals injected with N-Ac 4 GalNAz and those injected with the solvent vehicle was 12 AE 1, P < 0.005;F igure 3). Metabolic labeling of mouse serum [19] is thought to result from the incorporation of azido sugars into the major glycosylated proteins present. [20] We estimated the contribution of labeled serum glycoproteins to labeling of the small intestine,s pleen, kidney,a nd liver from the serum contents of these tissues,w hich have been estimated to be 5.0, 9.2, 19.1, and 20.2 %o ft he tissue volume,r espectively. [21] Thec ontribution of labeled serum proteins was estimated to be only 1.0, 2.4, 1.0, and 3.4 %ofthe total tissue glycan labeling,respectively.T umors have amuch larger interstitial volume fraction (20-40 %), [22] and their leaky neovasculature results in the retention of macromolecules. [23] However, the contribution to tumor labeling,due to retention of labeled serum proteins,was estimated to be only 15-30% of the total.
T 2 -weighted images and T 1 relaxation rate (R 1 )m aps ( Figure 4) were acquired in vivo before and at 2and 24 hpost injection of 2.T he relaxation rates (R 1 )f or water protons in the tumor,k idney,a nd liver (Figure 4) were significantly higher in animals injected with 1 and 2 (+ / +)t han for the controls injected with the vehicle and 2 (À/ +), demonstrating Thetumor contrast observed 2hafter probe injection was similar to that reported recently at 24 ha fter administration of aboronic acid based MRI contrast agent for detecting sialic acid in am urine model of melanoma. [6] Other organs also showed appreciable labeling (Figure 4e, + / +); however, analysis of the T 1 maps was difficult owing to organ motion. A semi-quantitative estimate of labeling based on coronal T 1weighted images acquired in vivo (Figure 5and Movie S1) 2h post administration of 2 indicated significant azido sugar dependent labeling of tumor (P < 0.05), kidney (P < 0.05), gut (P < 0.005), liver (P < 0.05), and spleen (P < 0.005).
Thegadolinium concentrations in tumor, kidney,and liver were estimated using the T 1 relaxivity of 2,and the relaxation rates of these tissues measured in vivo (Figure 4). This gave gadolinium concentrations of 42 AE 3, 80 AE 6a nd 93 AE 12 mm, respectively,at2hpost administration of 2 and 15 AE 1, 38 AE 3, and 17 AE 1 mm at 24 h. Theg adolinium concentration in the kidney measured ex vivo after 24 hw as 30 AE 6 mm (Figure 3), which is comparable with that estimated by the in vivo experiment. There was poorer agreement between the estimated in vivo concentrations in the tumor and liver at 24 ha nd those measured ex vivo,w hich were 1.8 AE 0.4 and 11 AE 1 mm,r espectively ( Figure 3). However,t his can be explained by the fact that the estimated minimum MRIdetectable tissue concentration of gadolinium is approximately 10 mm. [24] In summary,the gadolinium-labeled bioorthogonal probe described here can be used for the non-invasive in vivo imaging of tissue glycosylation by magnetic resonance imaging. Most tissues showed only low levels of non-specific retention of TMDIBO-Lys-Gd (2), and as ignificant Nazidoacetylgalactosamine dependent contrast was observed within two hours of probe administration. As altered cellsurface glycosylation is ah allmark of disease,p articularly cancer, and MRI is aw idely used imaging technique,t his novel method may enable the rapid assessment of diseaserelated changes in glycosylation in vivo. Figure 5. Imaging tissue glycosylation in vivo using T 1 -weighted MRI. Coronal T 1 -weighted images, before (a) and 2hafter (b) the injection of 2. Maximum intensity projection of T 1 -weighted signals (c) from representative mice injected with solvent vehicle and 2 (À/ +)or1 and 2 (+ / +). Coronal T 1 -weighted images (b, left to right) are displayed from the ventral towards the dorsal side. Maximum intensity projections (c) onto the sagittal (left) and coronal (right) planes. b) Metabolic labelingw as observed in the tumor (white arrows), kidney (cyan), liver (orange), gut (purple), and spleen (green) 2hafter the injection of 2.Bladder and gallbladder are indicated by red and yellow arrows, respectively.d )The N-Ac 4 GalNAz dependent contrast was analyzed semi-quantitatively,u sing regions of interest defined in the T 1 -weighted images for the tumor, kidney,gut, liver,a nd spleen;the mean signal intensity (MSI) for these tissues was divided by the MSI of muscle. Data represent mean AE SEM (n = 4). *P < 0.05, **P < 0.005. Two-tailedu npaired T-test with Mann-Whitney correction.