Expanded LUXendin Color Palette for GLP1R Detection and Visualization In Vitro and In Vivo

The glucagon-like peptide-1 receptor (GLP1R) is expressed in peripheral tissues and the brain, where it exerts pleiotropic actions on metabolic and inflammatory processes. Detection and visualization of GLP1R remains challenging, partly due to a lack of validated reagents. Previously, we generated LUXendins, antagonistic red and far-red fluorescent probes for specific labeling of GLP1R in live and fixed cells/tissues. We now extend this concept to the green and near-infrared color ranges by synthesizing and testing LUXendin492, LUXendin551, LUXendin615, and LUXendin762. All four probes brightly and specifically label GLP1R in cells and pancreatic islets. Further, LUXendin551 acts as a chemical beta cell reporter in preclinical rodent models, while LUXendin762 allows noninvasive imaging, highlighting differentially accessible GLP1R populations. We thus expand the color palette of LUXendins to seven different spectra, opening up a range of experiments using wide-field microscopy available in most labs through super-resolution imaging and whole animal imaging. With this, we expect that LUXendins will continue to generate novel and specific insights into GLP1R biology.


■ INTRODUCTION
The glucagon-like peptide-1 receptor (GLP1R) is a class B G protein-coupled receptor involved in the regulation of glucose homeostasis, food intake, and inflammation. 1,2 As such, GLP1R agonist (GLP1RA) therapy has become a mainstay of type-2 diabetes treatment during the past decade, with a number of drugs on the market based upon stabilized analogues of glucagon-like peptide-1. 3 Most recently, phase III trials of the third-generation semaglutide have shown a ∼15% reduction in body weight when combined with lifestyle interventions, 4 leading to the approval of GLP1RAs as the first nonsurgical treatment for complex obesity. Despite this, information concerning the localization of GLP1R is lacking, primarily due to the paucity of reliable and specific reagents for its detection and visualization. 5 Without this knowledge, it is difficult to elucidate the exact cellular mechanisms underlying GLP1R actions, many of which could be key to developing even more specific or effective treatments for metabolic/ inflammatory disease states, for instance, by tissue-targeted delivery. 6 For example, GLP1RAs have been shown to reduce the progression from nonalcoholic fatty liver disease/nonalcoholic steatohepatitis to fulminant fibrosis, 7,8 yet where and how the GLP1R acts is currently uncertain. Along similar lines, GLP1RAs exert inhibitory (and beneficial) effects on glucagon secretion, yet pancreatic GLP1R distribution and signaling remain debated. 5 Lastly, the neural circuits that GLP1RAs are able to access to exert effects on food intake remain to be fully delineated. 9−11 Reagents to detect GLP1R in tissues include antibodies, reporter mice, and fluorescent ligands. 5 Historically, studies with antibodies have been confounded by the use of nonspecific antisera, which detect non-GLP1R targets. 12, 13 Four specific antibodies now exist and have been extensively validated, including in the GLP1R knockout tissue, or cells heterologously expressing human GLP1R. 14 However, the available antibodies do not perform well for immunofluorescence staining in the brain and cannot be used for live visualization of the GLP1R using microscopy. Reporter mice, where cells that express(ed) GLP1R are selectively labeled with high fidelity, have been used to address this limitation, demonstrating excellent concurrence with other approaches. 15,16 However, reporter alleles neither visualize the receptor itself nor differentiate cells that once expressed GLP1R, but no longer do so (the cell will be indelibly marked). Fluorescent agonists bind the GLP1R orthosteric site in live tissues and can also be fixed to allow further immunohistochemical analysis. 10,17,18 However, this approach is confounded by activation of GLP1R, and as such the unstimulated fraction cannot be studied in live cells.
Recently, we have developed fluorescent antagonists, which are capable of detecting GLP1R in its unstimulated/ antagonized state in the membrane. 19 Advantageously, these probes, termed LUXendins, are equipotent to native antagonists, work well in the periphery and brain, display excellent brightness, and can be formalin-fixed. 19 To date, LUXendins have been freely and widely distributed to dozens of other labs for academic use, 20−22 opening up new GLP1R biology. The LUXendins were necessarily furnished with red and far-red fluorophores, not only allowing conventional microscopy but also for the aims of our study, total internal reflection (TIRF) microscopy and stimulated emission depletion (STED) nanoscopy. 19 Aiming for more experimental modalities and taking on board comments from end users, we now expand the color palette of the LUXendins, further increasing their utility for wide-field, confocal, intravital, and near-infrared microscopy, allowing imaging from the single cell to the whole animal.

■ RESULTS
Design and Synthesis of LUXendin492, LUXendin551, LUXendin615, and LUXendin762 Exendin4(9−39) was employed as a scaffold for modification with fluorophores. Using solid-phase peptide synthesis (SPSS), exendin4(9−39)-S39C (S39C-Ex4) was generated, bearing a C-terminal serine to cysteine substitution for functionalization via the introduced thiol handle. CF488A-, Cy3-, CPY-, and Cy7-conjugated versions were produced using cysteine− maleimide reactions and termed LUXendin492, LUXen-din551, LUXendin615, and LUXendin762, respectively ( Figure 1A), according to their maximal absorption values. Spectral properties were determined using UV/vis and fluorescence spectroscopy ( Figure 1B,C) ( Table 1) and were in line with known properties of the fluorophores used, for which extinction coefficients and quantum yields are reported. Full compound characterization and purity assessment are provided in the Supporting Information.
LUXendin492, LUXendin551, and LUXendin615 Specifically Label GLP1R To establish the labeling efficacy and specificity of the novel LUXendins, SNAP-GLP1R:CHO-K1 cells were incubated with each probe, before washing and orthogonal SNAP labeling with cell impermeable SBG-TMR or SBG-SiR. 24 High-resolution confocal images showed predominantly membrane-localized LUXendin staining in SNAP-GLP1R:CHO-K1 cells, which overlapped with labeling of the SNAP-tag located on the GLP1R N-terminus ( Figure 2A). Labeling efficiency was close to 100% for all probes investigated ( Figure 2B). No signal was detected in mock (nontransfected) CHO-K1 cell controls (Supporting Information, Figure S1). LUXendins were also able to label stably transfected SNAP-GLP1R:INS1 832/3 rat beta cells ( Figure  2C), as well as native INS1 832/3, which endogenously express GLP1R ( Figure 2D). Demonstrating high specificity, the signal was absent in INS1 832/3 GL1PR −/− cells, CRISPR deleted for the GL1PR ( Figure 2C). Of note, LUXendin492 and LUXendin615 staining was less "clean" than LUXen-din551, with some fluorescent signals present in the cytoplasm. We have previously reported a similar staining distribution for LUXendin555 (TMR) versus LUXendin645 (Cy5), 19 demonstrating a general preference toward cyaninebased dyes over their xanthene-based counterparts for cell labeling. To gain further insight into this observation, we applied LUXendin492 and LUXendin615 to SNAP-GLP1R:CHO-K1 cells, in parallel with cell-permeable SNAP labels 24 (Supporting Information, Figure S2). A similar experiment was performed but using cell impermeable SNAP labels, 25 before chasing with LUXendin492 and LUXen-din615 (Supporting Information, Figure S3). In both cases, no overlap with SNAP label was noticed, suggesting that intracellular LUXendin492 and LUXendin615 staining patterns are unlikely to stem from bound GLP1R. Nonetheless, all the LUXendins tested clearly label membrane GLP1R.
We next validated LUXendins for use in wide-field microscopy, which is widely available in most labs, serves to illustrate the robustness of labeling, and has the added advantage of allowing detection of near-infrared probes using cost efficient and fast switchable LED excitation and sensitive sCMOS detectors. As for confocal imaging, a similar pattern of LUXendin492, LUXendin551, and LUXendin615 staining was seen, with the cyanine-based dye (LUX551) performing superiorly (Supporting Information, Figure S4).

LUXendin492, LUXendin551, and LUXendin615 Specifically Label Endogenous GLP1R
One of the major advantages of LUXendins is that they can be used to visualize GLP1R in both live and fixed complex tissues. Pancreatic islets of Langerhans served as the testbed for the novel LUXendins because they express GLP1R, which is predominantly localized to the beta cell compartment. 5,19 Following 1 h of incubation with LUXendin492, LUXen-din551, and LUXendin615, intense labeling was observed throughout the islet, with large gaps apparent (presumably representing the GLP1R-negative alpha cell compartment, which comprises ∼20% of the rodent islet, as reported 19 ) ( Figure 3A). In all cases, labeling with the novel LUXendins could still be observed following formalin-fixation ( Figure 3B), further expanding the utility of the novel LUXendins for protein identification together with immunohistochemistry. Confirming specificity, LUXendin492, LUXendin551, and LUXendin615 signals co-localized with specific GLP1R monoclonal antibody staining (Novo Nordisk 7F38, fully validated in GLP1R −/− tissue 19 ) ( Figure 3B).

LUXendin551 Allows In Vivo Fluorescent Labeling of Islets in NOD Mice
The NOD mouse is a type-1 diabetes model that develops insulitis at 4−8 weeks of age, with frank diabetes occurring from 30 weeks of age. 26 However, identifying beta cells during disease trajectory is challenging because the polygenic NOD genetic background cannot be easily recombined with common inbred beta cell reporter strains (e.g., Ins1Cre; R26YFP). We and others have previously shown that GLP1R expression is beta cell specific 16,19 and we thus hypothesized that LUXendins might open up the possibility to identify beta cells in NOD (and other polygenic) mice.
To investigate this, the pancreas was exposed in 8-week-old anesthetized NOD mice through a small abdominal incision before being subjected to two-photon microscopy ( Figure 4A). Baseline images were acquired following retro-orbital injection of Hoechst33342 and albumin-AF647 to label the nuclei and vasculature, respectively. Prior to LUXendin551 injection there was no detectable signal ( Figure 4B). Rapid labeling occurred following the administration of LUXendin551 and was detected for at least 30 min post-injection ( Figure 4B). These studies also demonstrated that LUXendin551 is highly specific to islets and provides the ability to distinguish islets and beta cells from exocrine tissue ( Figure 4C). Due to its near-infrared excitation, we surmised that a Cy7linked GLP1R antagonist, LUXendin762, might allow intravital labeling of GLP1R, using the widely available and noninvasive IVIS in vivo imaging systems. We first tested LUXendin762 in cellulo in SNAP-GLP1R:CHO-K1 cells and in keeping with its pharmacology were able to detect strong membrane labeling, with little evidence of intracellular accumulation, again pointing to the high performance of cyanine-based dyes ( Figure 5A). Quantifying staining against SNAP-positive cells labeled with SBG-SiR, we found maximal efficiency ( Figure 5B), that is, all cells were positive for both stains (also see Supporting Information, Figure S5). LUXendin762 was next used to label primary islets, again showing cell membrane localization (Supporting Information, Figure S6A), shown to be GLP1R-positive using validated monoclonal antibodies (Supporting Information, Figure S6B). No spectral overlap could be detected between Cy5 (LUXendin645) and Cy7 (LUXendin762) channels (Supporting Information, Figure S6A,B). Freeing the far-red channel allowed us to perform multicolor experiments with commercially available far-red SiR-tubulin ( Figure 5C) and SPY650-DNA probes ( Figure 5D) that mark microtubule and JACS Au pubs.acs.org/jacsau Article DNA structures, respectively, providing further possibilities for cellular imaging. Confident that LUXendin762 was able to specifically label GLP1R, we next injected Nude mice with the probe before imaging. A strong fluorescent signal could be detected in the abdomen and brain at 30 and 60 min, following intraperitoneal or subcutaneous injection, with fluorescence levels ∼2−5-fold higher than in animals receiving saline vehicle ( Figure 5E). Because the signal intensity from the injection site is far brighter than in the brain, various organs were harvested. The pancreas of mice receiving intraperitoneal LUXendin762 showed the highest fluorescent signal, while the brain, lung, heart, and liver were similar to saline-treated controls ( Figure  5F, Supporting Information, Figure S7). By contrast, mice receiving subcutaneous LUXendin762 displayed the highest probe levels in the brain, whereas no signal was detected in the pancreas, lung, and heart versus saline-treated control ( Figure  5G, Supporting Information, Figure S7). Notably, the brain and pancreas are known to be GLP1R-positive, 5 whereas the GLP1R is only expressed in small cell populations (or absent) in the lung, kidney, liver, and heart (e.g., smooth muscle of arterioles). 27,28 Together, these studies show that LUXen-din762 can be detected in vivo in the whole organism and reveal a novel role for the injection route in determining GLP1R access.

■ DISCUSSION
In the present study, we synthesize and validate LUXen-din492, LUXendin551, LUXendin615, and LUXendin762, antagonist probes spanning green to near infrared for the visualization of GLP1R in cells, tissues, and animals. Together with our previous LUXendin555, LUXendin645, and LUXendin651 probes, 19 we now extend the LUXendin color palette to seven different spectra. These probes contain a range of different fluorophores suitable for wide-field, confocal, super-resolution, intravital microscopy, and small animal optical imaging, as well as FACS.
Pharmacologically, the novel LUXendins behave as full antagonists at the GLP1R, with similar potency to benchmark Exendin4(9−39). These studies further validate the robustness of the synthetic approach used and highlight the advantages of the S39C C-terminally substituted backbone used previously for Exendin4(9−39) 19 and Exendin4(1−39). 29 We envisage that in the future a similar backbone might be amenable to functionalization with biotin, complexed lanthanides, singlet oxygen generators or even nanoparticles, for example to allow nonfluorescent labeling for mass spectrometry, magnetic resonance imaging, or electron microscopy. With our observation that cyanine fluorophores behave more "cleanly" for microscopy, we are eager to find out how other molecular markers and tracers behave, and these endeavors are ongoing in our laboratories.
Of note, labeling with the novel LUXendins was co-localized with both SNAP-GLP1R and specific monoclonal antibody staining, as expected given the previous thorough validation of LUXendin555, LUXendin645, and LUXendin651 stablemates. 19 Moreover, no LUXendin signal could be detected in INS1 832/3 cells CRISPR-deleted for the GLP1R. These  Graphs show mean ± SEM. #p = 0.08, *p < 0.05 (unpaired t-test for each tissue).

JACS Au
pubs.acs.org/jacsau Article data also confirm that the Exendin4-S39C scaffold tolerates a most fluorophores without significant effects on labeling or pharmacology. While some punctate staining was seen with non-cyanine dyes, this does not reflect GLP1R activation, since: (1) all LUXendins were potent antagonists; (2) no colocalization from intracellular signals were seen in SNAP-GLP1R cell systems; and (3) we showed that punctate LUXendin signal was not co-localized with GLP1R monoclonal antibody. 19 By performing pulse-chase experiments using permeable and impermeable labels against SNAP-GLP1R, we further confirmed that punctate staining for LUXendin492 and LUXendin615 does not reflect activated GLP1R. One explanation for this observation could be preferred cellular uptake of xanthene-based LUXendin492 and LUXendin615 by macropinocytosis, a pathway for cells to uptake extracellular material caused by membrane ruffles. The presence of GLP1R is likely needed to increase local concentration of LUXendin492/LUXendin615 at the cell surface because we did not see dye uptake in cells without GLP1R (mock-transfected). Indeed, recent studies have shown increased uptake of rhodamines when conjugated to peptidic, alpha-helical backbones. 30 This is further supported by studies on fluorophore-labeled cell-penetrating peptides, in which rhodamines were found to exhibit a high hydrophobicity, leading to increased membrane penetration depth in liposomes. 31 As such, we observed pronounced increases in performance of cyanine dyes (Cy3, Cy5, and Cy7) when compared to CF488, TMR, and CPY, most probably due to their molecular nature.
Using novel LUXendins, we were able to perform unprecedented experiments and reveal new biology regarding GLP1R. As the best performing dye, LUXendin551 allowed GLP1R and thus beta cells to be reported in intravital experiments of a type-1 diabetes preclinical mouse model, which is not readily amenable to further genetic manipulation. Such experiments are important because we are still lacking information on the changes that occur in beta cell mass (and GLP1R expression) during insulitis and autoimmune destruction. 32 To allow noninvasive imaging, Cy7 was installed on the LUXendin backbone to produce LUXendin762, a nearinfrared probe. We were able to demonstrate that the LUXendin762 signal can be recorded in vivo (compared to saline-treated controls) and sequesters in organs known to express the GLP1R such as the pancreas and brain. 5 Of interest, LUXendin762 highlighted differential access routes to peripheral and brain GLP1R sites, with subcutaneous and not intraperitoneal injection labeling the latter. While the mechanisms are currently unknown, we speculate that ligand injected subcutaneously is less prone to the first pass effect and as such is able to abundantly enter the carotid arteries for entry into the brain. LUXendin762 thus opens up for the first time noninvasive longitudinal studies of GLP1R in mice using readily accessible platforms available in most academic/ industrial animal facilities. In addition, increasing the LUXendin762 dose, covering the injection sites, or using a more direct injection route (e.g., intracerebroventricular injection) might allow imaging of probe arrival in the pancreas and uptake in the brain. Such studies are particularly pertinent because GLP1R is also a readout for beta cell mass in preclinical models of type-2 diabetes and other metabolic syndromes. 33 Furthermore, longitudinal measures in the same animal are statistically more powerful and refined compared to assessment of various timepoints in multiple cohorts.
In summary, a total of seven LUXendins now allow detection and labeling of GLP1R in five different colors, with fluorophores tailored for various imaging modalities. We anticipate that these specific and validated probes will provide further insights into GLP1R biology in the periphery and brain, with implications for treatment with GLP1RAs.

Animals
All studies with harvested tissue used 7−10 week old male C57BL6/J mice and were regulated by the Animals (Scientific Procedures) Act 1986 of the U.K (Personal Project Licenses P2ABC3A83 and PP1778740). Approval was granted by the University of Birmingham's Animal Welfare and Ethical Review Body. All in vivo imaging experiments were performed with approval and oversight from the Indiana University Institutional Animal Care and Use Committee (IACUC).

Islet Isolation
Animals were humanely euthanized using cervical dislocation, before injection of collagenase 1 mg/mL (Serva NB8) into the bile duct. Inflated pancreases were digested for 12 min at 37°C and islets separated using a Ficoll (Sigma-Aldrich) gradient. Islets were cultured in RPMI medium containing 10% FCS, 100 units/mL penicillin, and 100 μg/mL streptomycin.

Two-Photon In Vivo Imaging
Female NOD/ShiLtJ mice 8 weeks of age were anesthetized with isoflurane. A small, vertical incision was made to expose the intact pancreas. Then, the exposed pancreas was placed on a 50 mm glassbottom dish for imaging on an inverted microscope. The body temperature was maintained using heating pads and heating elements on the objective. The mouse received, via retro-orbital injection, Hoechst 33342 (1 mg/kg in PBS) to label nuclei, albumin-AF647 (1 mg/kg in PBS) to label vasculature, and 75 μL of 30 μM LUXendin551. Images were collected using a Leica SP8 microscope, equipped with a 25×/0.95 NA objective and Spectra Physics MaiTai DeepSee multiphoton laser. Excitation was delivered at λ = 800 nm for Hoechst and Albumin-AF647, with signals collected at λ = 410− 500 nm and λ = 550−590 nm, respectively. LUXendin551 was excited at λ = 1050, with the signal collected at 650−700 nm. A conventional PMT was used for Hoechst, with a HyD detector used for Albumin-AF647 and LUXendin551. Blood was collected from the tail vein prior to and 30 min after LUXendin555 injection, and glucose was measured using an AlphaTrak2 glucometer. After imaging, unconscious mice were euthanized by cervical dislocation.

Noninvasive In Vivo Imaging
Whole body fluorescence accumulation and distribution was assessed in male athymic nude mice 8 weeks of age using an IVIS Spectral CT (Perkin Elmer). Mice were anesthetized with inhaled isoflurane and baseline images were acquired. Then, mice were intraperitoneally or subcutaneously injected with 100 μL of saline or 5 μM LUXendin762. Images were collected using a broad excitation and emission series combination ranging from 640 to 675 nm and 680 to 760 nm, respectively, at 30 min and 1 h post-injection. At the end point, animals were sacrificed, and tissues (pancreas, heart, brain, lung, and liver) were harvested for ex vivo fluorescence analysis. Spectral unmixing and quantification were analyzed using Living Image software.
Chemical synthesis, characterization, labeling of live CHO-K1 cells and fixed islets, wide-field imaging, and images of individual tissues (PDF)