A novel fluorescein-bisphosphonate based diagnostic tool for the detection of hydroxyapatite in both cell and tissue models

A rapid and efficient method for the detection of hydroxyapatite (HAP) has been developed which shows superiority to existing well-established methods. This fluorescein-bisphosphonate probe is highly selective for HAP over other calcium minerals and is capable of detecting lower levels of calcification in cellular models than either hydrochloric acid-based calcium leaching assays or the Alizarin S stain. The probe has been shown to be effective in both in vitro vascular calcification models and in vitro bone calcification models. Moreover we have demonstrated binding of this probe to vascular calcification in rat aorta and to areas of microcalcification, in human vascular tissue, beyond the resolution of computed tomography in human atherosclerotic plaques. Fluorescein-BP is therefore a highly sensitive and specific imaging probe for the detection of vascular calcification, with the potential to improve not only ex vivo assessments of HAP deposition but also the detection of vascular microcalcification in humans.

A rapid and efficient method for the detection of hydroxyapatite (HAP) has been developed which shows superiority to existing well-established methods. This fluorescein-bisphosphonate probe is highly selective for HAP over other calcium minerals and is capable of detecting lower levels of calcification in cellular models than either hydrochloric acid-based calcium leaching assays or the Alizarin S stain. The probe has been shown to be effective in both in vitro vascular calcification models and in vitro bone calcification models. Moreover we have demonstrated binding of this probe to vascular calcification in rat aorta and to areas of microcalcification, in human vascular tissue, beyond the resolution of computed tomography in human atherosclerotic plaques. Fluorescein-BP is therefore a highly sensitive and specific imaging probe for the detection of vascular calcification, with the potential to improve not only ex vivo assessments of HAP deposition but also the detection of vascular microcalcification in humans.
Calcium is a crucial intracellular element that is responsible for regulating many cellular processes across every cell type in biological organisms 1 . Calcium is found in either the free ion form or as a mineral phase, for example in the form of hydroxyapatite (HAP) that makes up bone and teeth. However mineralisation is also a critical component of a wide range of diseases such cancer, arthritis and cardiovascular disease (CVD) [2][3][4] . In atherosclerosis the presence of microscopic deposits of HAP can weaken the fibrous cap of an atherosclerotic plaque, leading to rupture of the plaque and vessel occlusion [5][6][7][8][9] . In aortic stenosis the progressive accumulation of HAP in the valve leads to increasing obstruction to the flow of blood out of the heart, whilst in abdominal aortic aneurysm disease microcalcification is associated with aneurysms that expand more quickly and are at increased risk of rupture or requiring repair. Thus the development of a calcium assay which is able to selectively detect HAP with high sensitivity and specificity could both improve understanding of disease pathophysiology and aid the diagnosis of a range of clinical disorders including vascular calcification.
The most common assays for calcium phosphate mineralisation are calcium leaching by hydrochloric acid (HCl); and staining with Alizarin S and von Kossa stains 10,11 . Each comes with their own limitations. The calcium leaching assay is used in laboratory settings to determine the presence of calcification in vascular and bone cell models. The assay involves incubating cell monolayers in the presence of HCl which allows the extraction of free calcium ions from the cells. The calcium content is then detected using a colourimetric assay and normalised against the total number of cells per sample 12 . Given the timeframe of the calcium leaching assay (24-48 h) it is not suitable for high-throughput applications. The colourimetric stain Alizarin S detects calcium, and not phosphate, thus it can also bind to calcium-binding proteins and proteoglycans without discriminating for the presence of HAP 13 . Finally, the colourimetric von Kossa stain detects phosphates but only in acidic environments, even though the presence of phosphate does not necessarily imply the presence of calcium or even HAP 10,13 . Both Alizarin S and von Kossa stains are used to indicate calcium depositions but at best give only semi-quantitative calcium readings 14 . In this study, the fluorescein-bisphosphonate conjugate 1 (Fluorescein-BP, Fig. 1

) 15 was
Use of Fluorescein-BP in cell-based models of vascular calcification. A series of experiments were designed to probe the ability of the Fluorescein-BP probe 1 to define the location of HAP in cellular calcification assays, exploiting both the selectivity and sensitivity which was demonstrated in the mineral-based assays. Vascular smooth muscle cells (VSMCs), the principal cell type involved in vascular calcification, can undergo a phenotypic transition under specific environmental conditions to calcifying osteoblastic cells 24 . The mouse VSMC line MOVAS-1 exhibits a smooth muscle cell phenotype and has been previously employed to investigate the VSMC cell cycle 25 . Validation of the Fluorescein-BP probe 1 was carried out using MOVAS-1 and primary mouse VSMCs.  (2). Calcium minerals were incubated with Fluorescein-BP (5 μM) (A) or 1% w/w Alizarin S (B) in water for 2 hours and the un-bound probe was removed and quantified. The histograms show the relative amounts of bound fluorophore for each of the calcium minerals. Data shown are representative of at least 3 independent experiments yielding comparable results. *P < 0.05, **P < 0.01, ***P < 0.001 compared to HAP, n = 6. Monolayers of MOVAS-1 cells were incubated with 2.4 mM Ca and 1.4 mM Pi for 7 days; then treated with Fluorescein-BP 1 for 2 h prior to imaging. Confocal microscopy ( Fig. 2A-C) shows extracellular localisation of probe 1 (in green); indicative of the presence of HAP in the extracellular matrix in accord with previous reports [26][27][28][29] . In contrast, no signal was observed when the same cells were incubated with a fluorescein-amino-1-pentanol conjugate (SI Fig. 1); giving a clear indication that the bisphosphonate motif is responsible for Fluorescein-BP binding to HAP in a cellular environment. Raman spectroscopic analysis of the observed mineral deposits confirmed the presence of HAP, with its distinctive peak at 960 cm −1 (Fig. 2D) 30 .
To probe the range of conditions under which the Fluorescein-BP probe 1 can be used, and compare its efficacy with the hydrochloric acid-based calcium leaching assay and Alizarin S stain, a further series of calcification assays were conducted using the MOVAS-1 cell line and primary mouse VSMCs. Monolayers of MOVAS-1 cells were incubated in both phosphate-enriched (3.0 mM Pi; 5.0 mM Pi), and phosphate and calcium-enriched (2.4 mM Ca and 1.4 mM Pi; 2.7 mM Ca and 2.5 mM Pi), growth media. Both the hydrochloric acid-based calcium leaching assay (4-fold, p < 0.001) (Fig. 3A) and the Fluorescein-BP 1 fluorescence assay (4-fold, p < 0.001) (Fig. 3B) show significant levels of calcium deposition in the presence of calcium-enriched media. The fluorescence assay (LOD 0.0065 µmol/L, SI Fig. 2) is able to detect the presence of HAP when using media enriched with only phosphate in comparison to the calcium leaching assay (LOD 225 µmol/L) 31 demonstrating the greater sensitivity of Fluorescein-BP 1. A comparison of the fluorescence images obtained for well plates with images of Alizarin S staining of equivalent plates (Fig. 3C), shows that both fail to detect calcification in media only enriched with phosphate. However, when cells are incubated in the presence of both calcium and phosphate, the fluorescent probe is able to detect the deposition of microcrystalline HAP with greater sensitivity than the Alizarin S stain. This is particularly notable when the incubation media have been enriched with 2.4 mM Ca and 1.4 mM Pi, as commonly used in vascular calcification models 32 .
When comparable studies were performed using primary mouse VSMCs ( Fig. 4A-C), incubating the monolayers with increasing levels of phosphate to induce calcification (1.8 mM Pi, 2.6 mM Pi and 3.0 mM Pi), the fluorescent probe (5-fold, p < 0.001) was also shown to be more sensitive than either the Alizarin S or calcium leaching assays (1.2-fold, p < 0.001).
In further experiments, MOVAS-1 cells (SI Fig. 3) and primary mouse VSMCs (SI Fig. 4) were incubated under standard calcification conditions (2.4 mM Ca and 1.4 mM Pi) and the calcification on days 3, 5 and 7 was detected using Fluorescein-BP probe 1 (2.5-fold, p < 0.001 for MOVAS-1; 2.5-fold, p < 0.001 for VSMCs) and compared with calcium leaching using the hydrochloric acid-based assay (5-fold, p < 0.001 for MOVAS-1; 2.5-fold, p < 0.001 for VSMCs) and staining by Alizarin S. The fluorescent probe was shown to be more sensitive than both the other assays. Similarly, when MC3T3 cells (SI Fig. 5), an osteoblast derived cell line which is commonly used for studies in the field of bone and skeletal mineralisation 33,34 , were incubated with Fluorescein-BP, calcium mineralisation was readily detected (2-fold, p < 0.001).
Ex vivo rat aorta and human tissue study. VSMC monolayer cultures are of only limited use in the study of calcification in CVD: they lack the architecture and matrix of normal vessels; they reach confluence over a short timeframe; they rapidly convert to a proliferative, secretory phenotype 35 ; morphological variations have been seen at high passage numbers; and variations in the gene phenotype can also occur. Cultured aortae can provide complementary information that bridges the gap between traditional cell culture and animal models, under almost in vivo conditions 35,36 . Previous studies have cultured sectioned thoracic aortas dissected from mice and estimated the status of calcification under Pi stimulation at 2.6-3.0 mM Pi 36 ; these phosphate concentrations correlate with myocardial infarction in human studies 36 . Ex vivo aortic rings from rats were incubated under different calcification conditions (1.8 mM Pi, 2.6 mM Pi and 3.0 mM Pi) for 7 days and subsequently incubated with Fluorescein-BP 1 followed by Alizarin S (Fig. 5, SI Fig. 6). From both staining methods, it is apparent that with increasing calcification conditions there is an increase in signal in both the Alizarin S and Fluorescein channel. When the two channels are merged, it is clear that the Alizarin S and Fluorescein-BP stains colocalize in most areas. No fluorescence signal is observed when the rings are cultured with control media confirming that the Fluorescein-BP probe is selective for HAP. Next, a more detailed analysis of calcification in aortic rings was conducted using 5 µm sections (Fig. 6). Calcified rat aortic rings were fixed and cryosectioned on day 7; the sections were then incubated with Fluorescein-BP probe 1 followed by Alizarin S. Fluorescent microscopy of these sections confirms differences in calcification between the control and 3.0 mM Pi specimens. This also allows the location of the calcification to be determined; which in this rat model occurs in the medial layer of the aorta 35 .
PET/CT imaging using Na 18 F has been reported as a novel tool for vascular disease diagnosis 9,37 . Na 18 F is able to identify areas of micro-calcification that are associated with unstable plaque phenotype and which are beyond the resolution of CT. Indeed Na 18 F PET is currently being explored as a method for detecting high risk coronary plaques and improving cardiovascular risk prediction (clinicaltrials.gov NCT02278211) 9,37,38 . We next compared Fluorescein-BP binding to the detection of vascular calcification by computed tomography and Na 18 F PET in 8 samples of human arterial tissues (7 affected by calcific atherosclerotic disease and 1 control tissue which was free of atherosclerosis). All the tissue samples were assessed initially using PET/CT scans followed by sectioning of the tissue as required to allow appropriate fluorescence images to be obtained (as depicted for a representative sample in Fig. 7, additional examples shown in SI Fig. 7). PET/CT imaging ( Fig. 7A-C) showed both areas of macrocalcification, as detected by CT, and areas of microcalcification, as detected by PET using a Na 18 F tracer. Samples from both areas were sectioned and stained (Fig. 7D-K). The images show that Fluorescein-BP probe 1 binds to both macro-and micro-calcification and colocalizes with Alizarin S staining (Fig. 7D,E). The control tissue sample (Fig. 7L,M), where there was no clinical history of coronary disease, showed neither PET/CT signal nor fluorescence staining.

Discussion
Here, we demonstrate that Fluorescein-BP is a highly sensitivity and specific marker of HAP in vascular tissue. Fluorescein-BP has been compared to the current gold standard techniques for identifying calcification, the Alizarin S and HCl decalcification assays, and in both instances shows greater sensitivity and selectivity for HAP. Moreover we have confirmed its binding to HAP in both in vitro cell models of vascular and bone calcification, ex vivo rat aorta tissue and human coronary atherosclerotic plaques. Given that its component parts are already FDA approved, Fluorescein-BP holds great translational potential for the detection of crystalline HAP in humans. Vascular mineralisation in atherosclerotic plaques contains high concentrations of crystalline HAP 39 . Within the arterial walls, extracellular vesicles containing HAP are extruded from lipid-rich macrophages and apoptotic smooth muscle cells which increase the plaque structural stress and risk of fibroathermatous cap fracture 40 . The inflammatory milieu driving the production of HAP deposition is associated with a greater propensity to destabilise plaques and results in clinical cardiovascular events 41,42 . Recent clinical trials have focused attention on the vital role that early calcification plays in carotid, aortic aneurysmal and coronary artery disease and hence there is a clinical imperative to better understand the mechanisms governing early calcification in the vascular wall [41][42][43] . The ability to differentiate calcium derivatives in regions of calcification, specifically by detecting HAP, will facilitate the dichotomisation of unstable and stable cardiovascular phenotypes in future clinical trials.
The detection of HAP deposition has been a field that has gained a lot of attention in the past 20 years 1,44-46 . Histological stains such as Alizarin S and von Kossa are widely used to study calcium deposits but their effectiveness has been disputed since their binding modes are somewhat indiscriminate 45 . Additionally, calcein is a routinely used tool for imaging calcium, however, not only does it bind across a range of calcium species, but it is also non-specific for calcium and interacts with a range of other metals including aluminium and zinc 47,48 . Another recently reported probe used in animal based models of both bone and vascular disease is Osteosense (680 and 750), a commercially available, near infrared (NIR) probe conjugated to a bisphosphonate [49][50][51] . However, the benefits offered by a NIR probe for in vivo studies, are off-set by the need for non-standard filters for plate readers and microscopes in laboratory based experiments which limited our capacity to directly compare the two probes. Fluorescein-BP probe 1, in comparison to Calcein and Osteosense, presents a more selective, sensitive and widely-applicable method of detecting calcification in vascular and bone models. In addition, the two low-cost reagents required for the synthesis of the probe are both commercially available, ensuring a wider user base for the new probe. Future in vivo studies might be enabled by intravascular administration of probe 1 using catheters fitted with a confocal endomicroscope probe, thus negating the need for conjugation to a costly NIR dye.

Conclusions
Fluorescein-BP probe 1 gives a dramatically improved signal output over more conventional imaging methods such as Alizarin S and von Kossa stains for the detection of calcium phosphate in a range of in vitro studies. It has been shown to be mineral-specific giving a markedly increased signal in the presence of HAP over other calcium species, and can be used in both condition-specific and temporal studies of calcification in cell-based models of CVD and bone disease. Staining with probe 1 is rapid, allowing high-throughput assays in multi-well format which are not possible with comparable assays such as the hydrochloric acid-based quantification method typically used in the field. When applied to ex vivo aortic sections and to human tissue presenting both macro-and micro-calcification, Fluorescein-BP 1 shows binding to both types of vascular calcification which are not readily detectable using one stand-alone technique. The employment of this probe may enable the elucidation of key mechanisms underpinning these pathological processes.  Semi-preparative RP-HPLC was performed using a Waters 600E pump, a Waters 486 tuneable absorbance detector controlled by Water Empower software which was equipped with a Phenomenex Luna C18(2), 5 µm particle size, 250 × 21.2 mm column at a flow rate of 18 mL min −1 . 1 H and 13 C NMR spectra were obtained on Bruker AVA600 instrument at the stated frequency using TMS as a reference and residual solvent as an internal standard. Infra-red spectra were recorded neat on a Shimadzu IRAffinity-1. Electrospray (ESI) mass spectra were obtained on a Bruker micrOTOF II instrument. UV/VIS absorption spectra were measured using a Shimadzu UV-1800 spectrometer. Melting points were determined on a Gallenkamp Electrothermal Melting Point apparatus and are uncorrected. Fluorescence data was obtained using SPEX Fluoromax-3.

Fluorescein-BP and Alizarin S incorporation in Calcium minerals.
Aqueous suspensions of calcium species (HAP, CaOx, CaPi, CaPyr, Sigma and Fisher Scientific) (50 mg) were incubated with Fluorescein-BP 1 (5 µM) for 2 hours. Unbound probe was removed by centrifugation of the suspension at 670 g for 5 minutes, the supernatant was removed and the precipitated solid was resuspended in water with continuous agitation for 15 minutes. This process of centrifugation and resuspension was repeated three times, until there was no further fluorescence in the supernatant after centrifugation. Fluorescence analysis of the combined aqueous extracts was performed using a Synergy HT Multi Mode Microplate reader at 488 nm.
VSMC isolation and culture. Primary aortic VSMCs were isolated from five week old C57BL/6 mice (Charles River Laboratories) as previously described 1,52,53 . Mice were euthanized by cervical dislocation, the aorta was then dissected and the adventitia removed. After washing with Hanks' balanced salt solution (HBSS; Life Technologies), the aorta was cut longitudinally to expose the endothelial layer. Eight aortae were pooled together and incubated for 10 minutes at 37 °C in 1 mg/mL trypsin (Life Technologies) to remove any remaining adventitia and endothelium. Aortae were washed and incubated overnight at 37 °C in VSMC growth medium containing Minimum Essential Medium Eagle alpha modification (α-MEM; Life Technologies), 10% foetal bovine serum (FBS; Life Technologies) and 1% gentamicin (Life Technologies) in a humidified 5% CO 2 incubator. Tissues were then washed and incubated in 425 UI/mL collagenase type II (Worthington Biochemical Corporation) for 4 hours at 37 °C. The resulting cell suspension was centrifuged at 320 g for 5 minutes. VSMC pellets were resuspended in culture medium and cultured for two passages in T25 tissue culture flasks (Corning) coated with 0.25 μg/cm 2 laminin (Sigma) to promote maintenance of the contractile differentiation state 54 .

Induction of calcification in MOVAS-1 and VSMCs.
Calcification was introduced as previously described 1,32,53 . In brief, for MOVAS-1, cells were grown to confluence (day 0) and switched to calcification medium, which was prepared by adding 1 M inorganic phosphate (mixture of NaH 2 PO 4 and Na 2 HPO 4 , pH = 7.4) and 1 M CaCl 2 to reach a final concentrations of 3.0 mM Pi, 5.0 mM Pi, 2.4 mM Ca and 1.4 mM Pi or 2.7 mM Ca and 2.5 mM Pi. For VSMCs, cells were grown to confluence (day 0) and switched to calcification medium, which was prepared by adding 1 M inorganic phosphate to reach a final concentrations of 1.8 mM Pi, 2.6 mM Pi and 3.0 mM Pi. Cells were incubated for 7 days in 95% air/5% CO 2 , changing media on alternate days.

Fluorescence staining of fixed cells using Fluorescein-BP and CellMask Orange Plasma Membrane
Stain. MOVAS-1 cells were grown on glass coverslips and calcified as described above. After incubation with Fluorescein-BP probe 1 (1 μM) for 2 hours, the media was changed, the monolayer was washed twice with HBSS and fresh HBSS containing 500 nM CellMask Orange Plasma Membrane Stain (Thermo Fisher) was added for 10 minutes. The monolayer was washed with phosphate buffered saline (PBS, 2×) and water (1×), fixed with 10% neutral buffered formalin (NBF) for 15 minutes and washed with PBS (2×) and water (1×). The cell monolayers were permeabilised with Triton X-100 (0.05%, 3×) for 5 minutes and incubated with DAPI (300 nM; Life Technologies) for 5 minutes. Excess DAPI was removed by washing with water (1× Calcium leaching assay. Calcium deposition was quantified as previously described 32,52 . Briefly, cells were rinsed with phosphate buffered saline (PBS) and decalcified with 0.6 N HCl at room temperature for 24 hours. Free calcium was determined calorimetrically by a stable interaction with o-cresolphthalein complexone using a commercially available kit (Randox Laboratories Ltd.) and corrected for total protein concentration (Bio-Rad Laboratories Ltd).
Alizarin S staining of cell monolayers. Cells were grown in plates and calcified as described above. Media was removed and the cell monolayer was subsequently washed with HBSS. Cells were fixed in 10% NBF for 15 minutes before washing twice with PBS. The cells were then stained with Alizarin S (2%, pH 4.2; 500 µL) for 10 minutes at rt. The supernatant was discarded and the cell monolayer was washed with water (3×) and then imaged at 530 nm (Alizarin S).
Ex vivo aorta isolation and culture. Aortic rings were dissected from eight week old Fischer male rats.
Rats were euthanized by cervical dislocation, the aorta was then dissected and the adventitia removed. After washing with HBSS the aortae were cut into ~4 mm thick rings and cultured in α-MEM supplemented with 10% FBS and 1% gentamicin at 37 °C in 95% air/5% CO 2 . After 1 day, the media was changed to calcifying media (1.8 mM Pi, 2.6 mM Pi, 3.0 mM Pi) for 7 days, changing media on alternate days. On day 7, aortas were incubated with Fluorescein-BP (1 µM) for 2 hours. Aortas were washed with PBS (2×) and fixed with 10% NBF for 15 minutes. Residual NBF was removed by washing with PBS (2×) and water (1×). For Alizarin S staining, the fixed aortas were permeabilised with 1% KOH for 1 hour followed by overnight staining with 0.00005% Alizarin S in 1% KOH. Residual Alizarin S was removed by washing with water (3×). Rings were then imaged under a Zeiss Axiovert 25 inverted microscope at 488 nm (Fluorescein-BP) and 530 nm (Alizarin S).
Aortic sections. Aortic rings were dissected and calcified following the standard protocol. Once calcified, the rings were fixed in 10% NBF for 15 minutes and then cryosectioned into 5 μm thick sections. Sections were then imaged under a Leica DMBL-2 upright fluorescent microscope at 488 nm (Fluorescein-BP), 530 nm (Alizarin S) and 350 nm (DAPI).
Human tissue. Atherosclerotic layers of left main coronary arteries were analysed from a range of patients, both male and female between the age of 48-71, for which main cause of death was either haemopericardium or ischaemic and hypertensive heart disease. Upon approval, the tissue was frozen at −80 °C.
PET/CT analysis of human tissue. Thawed non-decalcified coronary artery specimens were incubated for 20 minutes in 18 F-sodium fluoride 100 kBq/mL solution (10.5 MBq 18F-NaF in 99.5 mL 0.9% NaCl). Samples were twice washed in 10 mL 0.9% NaCl for 5 minutes to remove unbound 18 F-fluoride. Coronary artery specimens were scanned using high-resolution micro-positron emission tomography and non-contrast computed tomography [50 kV p tube voltage, 300 ms exposure time] (Mediso nanoScan PET/CT). After whole specimen imaging, the coronary arteries were fixed in 10% buffered formaldehyde before being dehydrated, embedded in paraffin wax and sectioned (5 μm thickness).
Alizarin S Staining of human tissue. Sections were de-waxed in xylene and stained with 2% Alizarin S for 5 minutes, washed three times with water and dehydrated to visualize calcium deposition. Images were obtained using a Nikon Ni1 Brightfield microscope.
Fluorescein-BP and Alizarin S Staining of human tissue. Sections were de-waxed in xylene and incubated with Fluorescein-BP probe 1(1 µM) for 2 hours, washed in water (2×) followed by incubation with 2% Alizarin S (250 µL) for 5 minutes. Sections were washed in water (3×) and subsequently incubated with DAPI (500 nM) for 5 minutes. Sections were washed with water (1×) and then mounted using ProLong Gold Antifade. Fluorescence signal was detected using a Leica DMRB fluorescence microscope at 488 nm (Fluorescein-BP), 530 nm (Alizarin S) and 350 nm (DAPI).